From 67e9c74b8a873408c27ac9a8e4c1d1c8d72c93ff Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 16 Nov 2015 11:13:34 -0500 Subject: cgroup: replace __DEVEL__sane_behavior with cgroup2 fs type With major controllers - cpu, memory and io - shaping up for the unified hierarchy, cgroup2 is about ready to be, gradually, released into the wild. Replace __DEVEL__sane_behavior flag which was used to select the unified hierarchy with a separate filesystem type "cgroup2" so that unified hierarchy can be mounted as follows. mount -t cgroup2 none $MOUNT_POINT The cgroup2 fs has its own magic number - 0x63677270 ("cgrp"). v2: Assign a different magic number to cgroup2 fs. Signed-off-by: Tejun Heo Acked-by: Li Zefan Cc: Johannes Weiner --- kernel/cgroup.c | 47 +++++++++++++++++++++++------------------------ 1 file changed, 23 insertions(+), 24 deletions(-) (limited to 'kernel') diff --git a/kernel/cgroup.c b/kernel/cgroup.c index b316debadeb3..af0886262f58 100644 --- a/kernel/cgroup.c +++ b/kernel/cgroup.c @@ -211,6 +211,7 @@ static unsigned long have_free_callback __read_mostly; /* Ditto for the can_fork callback. */ static unsigned long have_canfork_callback __read_mostly; +static struct file_system_type cgroup2_fs_type; static struct cftype cgroup_dfl_base_files[]; static struct cftype cgroup_legacy_base_files[]; @@ -1641,10 +1642,6 @@ static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts) all_ss = true; continue; } - if (!strcmp(token, "__DEVEL__sane_behavior")) { - opts->flags |= CGRP_ROOT_SANE_BEHAVIOR; - continue; - } if (!strcmp(token, "noprefix")) { opts->flags |= CGRP_ROOT_NOPREFIX; continue; @@ -1711,15 +1708,6 @@ static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts) return -ENOENT; } - if (opts->flags & CGRP_ROOT_SANE_BEHAVIOR) { - pr_warn("sane_behavior: this is still under development and its behaviors will change, proceed at your own risk\n"); - if (nr_opts != 1) { - pr_err("sane_behavior: no other mount options allowed\n"); - return -EINVAL; - } - return 0; - } - /* * If the 'all' option was specified select all the subsystems, * otherwise if 'none', 'name=' and a subsystem name options were @@ -1998,6 +1986,7 @@ static struct dentry *cgroup_mount(struct file_system_type *fs_type, int flags, const char *unused_dev_name, void *data) { + bool is_v2 = fs_type == &cgroup2_fs_type; struct super_block *pinned_sb = NULL; struct cgroup_subsys *ss; struct cgroup_root *root; @@ -2014,6 +2003,17 @@ static struct dentry *cgroup_mount(struct file_system_type *fs_type, if (!use_task_css_set_links) cgroup_enable_task_cg_lists(); + if (is_v2) { + if (data) { + pr_err("cgroup2: unknown option \"%s\"\n", (char *)data); + return ERR_PTR(-EINVAL); + } + cgrp_dfl_root_visible = true; + root = &cgrp_dfl_root; + cgroup_get(&root->cgrp); + goto out_mount; + } + mutex_lock(&cgroup_mutex); /* First find the desired set of subsystems */ @@ -2021,15 +2021,6 @@ static struct dentry *cgroup_mount(struct file_system_type *fs_type, if (ret) goto out_unlock; - /* look for a matching existing root */ - if (opts.flags & CGRP_ROOT_SANE_BEHAVIOR) { - cgrp_dfl_root_visible = true; - root = &cgrp_dfl_root; - cgroup_get(&root->cgrp); - ret = 0; - goto out_unlock; - } - /* * Destruction of cgroup root is asynchronous, so subsystems may * still be dying after the previous unmount. Let's drain the @@ -2140,9 +2131,10 @@ out_free: if (ret) return ERR_PTR(ret); - +out_mount: dentry = kernfs_mount(fs_type, flags, root->kf_root, - CGROUP_SUPER_MAGIC, &new_sb); + is_v2 ? CGROUP2_SUPER_MAGIC : CGROUP_SUPER_MAGIC, + &new_sb); if (IS_ERR(dentry) || !new_sb) cgroup_put(&root->cgrp); @@ -2185,6 +2177,12 @@ static struct file_system_type cgroup_fs_type = { .kill_sb = cgroup_kill_sb, }; +static struct file_system_type cgroup2_fs_type = { + .name = "cgroup2", + .mount = cgroup_mount, + .kill_sb = cgroup_kill_sb, +}; + /** * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy * @task: target task @@ -5315,6 +5313,7 @@ int __init cgroup_init(void) WARN_ON(sysfs_create_mount_point(fs_kobj, "cgroup")); WARN_ON(register_filesystem(&cgroup_fs_type)); + WARN_ON(register_filesystem(&cgroup2_fs_type)); WARN_ON(!proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations)); return 0; -- cgit From a946e8c717f9355d1abd5408ed0adc0002d1aed1 Mon Sep 17 00:00:00 2001 From: Sudeep Holla Date: Wed, 4 Nov 2015 18:32:37 +0000 Subject: genirq: Delay incrementing interrupt count if it's disabled/pending In case of a wakeup interrupt, irq_pm_check_wakeup disables the interrupt and marks it pending and suspended, disables it and notifies the pm core about the wake event. The interrupt gets handled later once the system is resumed. However the irq stats is updated twice: once when it's disabled waiting for the system to resume and later when it's handled, resulting in wrong counting of the wakeup interrupt when waking up the system. This patch updates the interrupt count so that it's updated only when the interrupt gets handled. It's already handled correctly in handle_edge_irq and handle_edge_eoi_irq. Reported-by: Manoil Claudiu Signed-off-by: Sudeep Holla Cc: Marc Zyngier Link: http://lkml.kernel.org/r/1446661957-1019-1-git-send-email-sudeep.holla@arm.com Signed-off-by: Thomas Gleixner --- kernel/irq/chip.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index 15206453b12a..05e29de57933 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -338,7 +338,6 @@ void handle_nested_irq(unsigned int irq) raw_spin_lock_irq(&desc->lock); desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING); - kstat_incr_irqs_this_cpu(desc); action = desc->action; if (unlikely(!action || irqd_irq_disabled(&desc->irq_data))) { @@ -346,6 +345,7 @@ void handle_nested_irq(unsigned int irq) goto out_unlock; } + kstat_incr_irqs_this_cpu(desc); irqd_set(&desc->irq_data, IRQD_IRQ_INPROGRESS); raw_spin_unlock_irq(&desc->lock); @@ -412,13 +412,13 @@ void handle_simple_irq(struct irq_desc *desc) goto out_unlock; desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING); - kstat_incr_irqs_this_cpu(desc); if (unlikely(!desc->action || irqd_irq_disabled(&desc->irq_data))) { desc->istate |= IRQS_PENDING; goto out_unlock; } + kstat_incr_irqs_this_cpu(desc); handle_irq_event(desc); out_unlock: @@ -462,7 +462,6 @@ void handle_level_irq(struct irq_desc *desc) goto out_unlock; desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING); - kstat_incr_irqs_this_cpu(desc); /* * If its disabled or no action available @@ -473,6 +472,7 @@ void handle_level_irq(struct irq_desc *desc) goto out_unlock; } + kstat_incr_irqs_this_cpu(desc); handle_irq_event(desc); cond_unmask_irq(desc); @@ -532,7 +532,6 @@ void handle_fasteoi_irq(struct irq_desc *desc) goto out; desc->istate &= ~(IRQS_REPLAY | IRQS_WAITING); - kstat_incr_irqs_this_cpu(desc); /* * If its disabled or no action available @@ -544,6 +543,7 @@ void handle_fasteoi_irq(struct irq_desc *desc) goto out; } + kstat_incr_irqs_this_cpu(desc); if (desc->istate & IRQS_ONESHOT) mask_irq(desc); -- cgit From 38c6ade2dd4dcc3bca06c981e2a1b91289046177 Mon Sep 17 00:00:00 2001 From: Dietmar Eggemann Date: Tue, 20 Oct 2015 13:04:41 +0100 Subject: sched/fair: Remove empty idle enter and exit functions Commit cd126afe838d ("sched/fair: Remove rq's runnable avg") got rid of rq->avg and so there is no need to update it any more when entering or exiting idle. Remove the now empty functions idle_{enter|exit}_fair(). Signed-off-by: Dietmar Eggemann Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Yuyang Du Link: http://lkml.kernel.org/r/1445342681-17171-1-git-send-email-dietmar.eggemann@arm.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 24 +----------------------- kernel/sched/idle_task.c | 1 - kernel/sched/sched.h | 8 -------- 3 files changed, 1 insertion(+), 32 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index f04fda8f669c..2779dece43b2 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2835,24 +2835,6 @@ void remove_entity_load_avg(struct sched_entity *se) atomic_long_add(se->avg.util_avg, &cfs_rq->removed_util_avg); } -/* - * Update the rq's load with the elapsed running time before entering - * idle. if the last scheduled task is not a CFS task, idle_enter will - * be the only way to update the runnable statistic. - */ -void idle_enter_fair(struct rq *this_rq) -{ -} - -/* - * Update the rq's load with the elapsed idle time before a task is - * scheduled. if the newly scheduled task is not a CFS task, idle_exit will - * be the only way to update the runnable statistic. - */ -void idle_exit_fair(struct rq *this_rq) -{ -} - static inline unsigned long cfs_rq_runnable_load_avg(struct cfs_rq *cfs_rq) { return cfs_rq->runnable_load_avg; @@ -7248,8 +7230,6 @@ static int idle_balance(struct rq *this_rq) int pulled_task = 0; u64 curr_cost = 0; - idle_enter_fair(this_rq); - /* * We must set idle_stamp _before_ calling idle_balance(), such that we * measure the duration of idle_balance() as idle time. @@ -7330,10 +7310,8 @@ out: if (this_rq->nr_running != this_rq->cfs.h_nr_running) pulled_task = -1; - if (pulled_task) { - idle_exit_fair(this_rq); + if (pulled_task) this_rq->idle_stamp = 0; - } return pulled_task; } diff --git a/kernel/sched/idle_task.c b/kernel/sched/idle_task.c index c4ae0f1fdf9b..47ce94931f1b 100644 --- a/kernel/sched/idle_task.c +++ b/kernel/sched/idle_task.c @@ -47,7 +47,6 @@ dequeue_task_idle(struct rq *rq, struct task_struct *p, int flags) static void put_prev_task_idle(struct rq *rq, struct task_struct *prev) { - idle_exit_fair(rq); rq_last_tick_reset(rq); } diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index efd3bfc7e347..2eb2002aa336 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1249,16 +1249,8 @@ extern void update_group_capacity(struct sched_domain *sd, int cpu); extern void trigger_load_balance(struct rq *rq); -extern void idle_enter_fair(struct rq *this_rq); -extern void idle_exit_fair(struct rq *this_rq); - extern void set_cpus_allowed_common(struct task_struct *p, const struct cpumask *new_mask); -#else - -static inline void idle_enter_fair(struct rq *rq) { } -static inline void idle_exit_fair(struct rq *rq) { } - #endif #ifdef CONFIG_CPU_IDLE -- cgit From d937cdc59e363baf8d5c757d944b13ebfa33e729 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 19 Oct 2015 13:49:30 +0200 Subject: sched/fair: Clean up the explanation around decaying load update misses Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 53 ++++++++++++++++++++++++----------------------------- 1 file changed, 24 insertions(+), 29 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 2779dece43b2..8f3905e3b986 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4222,42 +4222,37 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) */ /* - * The exact cpuload at various idx values, calculated at every tick would be - * load = (2^idx - 1) / 2^idx * load + 1 / 2^idx * cur_load + * The exact cpuload calculated at every tick would be: * - * If a cpu misses updates for n-1 ticks (as it was idle) and update gets called - * on nth tick when cpu may be busy, then we have: - * load = ((2^idx - 1) / 2^idx)^(n-1) * load - * load = (2^idx - 1) / 2^idx) * load + 1 / 2^idx * cur_load + * load' = (1 - 1/2^i) * load + (1/2^i) * cur_load + * + * If a cpu misses updates for n ticks (as it was idle) and update gets + * called on the n+1-th tick when cpu may be busy, then we have: + * + * load_n = (1 - 1/2^i)^n * load_0 + * load_n+1 = (1 - 1/2^i) * load_n + (1/2^i) * cur_load * * decay_load_missed() below does efficient calculation of - * load = ((2^idx - 1) / 2^idx)^(n-1) * load - * avoiding 0..n-1 loop doing load = ((2^idx - 1) / 2^idx) * load * - * The calculation is approximated on a 128 point scale. - * degrade_zero_ticks is the number of ticks after which load at any - * particular idx is approximated to be zero. - * degrade_factor is a precomputed table, a row for each load idx. - * Each column corresponds to degradation factor for a power of two ticks, - * based on 128 point scale. - * Example: - * row 2, col 3 (=12) says that the degradation at load idx 2 after - * 8 ticks is 12/128 (which is an approximation of exact factor 3^8/4^8). + * load' = (1 - 1/2^i)^n * load + * + * Because x^(n+m) := x^n * x^m we can decompose any x^n in power-of-2 factors. + * This allows us to precompute the above in said factors, thereby allowing the + * reduction of an arbitrary n in O(log_2 n) steps. (See also + * fixed_power_int()) * - * With this power of 2 load factors, we can degrade the load n times - * by looking at 1 bits in n and doing as many mult/shift instead of - * n mult/shifts needed by the exact degradation. + * The calculation is approximated on a 128 point scale. */ #define DEGRADE_SHIFT 7 -static const unsigned char - degrade_zero_ticks[CPU_LOAD_IDX_MAX] = {0, 8, 32, 64, 128}; -static const unsigned char - degrade_factor[CPU_LOAD_IDX_MAX][DEGRADE_SHIFT + 1] = { - {0, 0, 0, 0, 0, 0, 0, 0}, - {64, 32, 8, 0, 0, 0, 0, 0}, - {96, 72, 40, 12, 1, 0, 0}, - {112, 98, 75, 43, 15, 1, 0}, - {120, 112, 98, 76, 45, 16, 2} }; + +static const u8 degrade_zero_ticks[CPU_LOAD_IDX_MAX] = {0, 8, 32, 64, 128}; +static const u8 degrade_factor[CPU_LOAD_IDX_MAX][DEGRADE_SHIFT + 1] = { + { 0, 0, 0, 0, 0, 0, 0, 0 }, + { 64, 32, 8, 0, 0, 0, 0, 0 }, + { 96, 72, 40, 12, 1, 0, 0, 0 }, + { 112, 98, 75, 43, 15, 1, 0, 0 }, + { 120, 112, 98, 76, 45, 16, 2, 0 } +}; /* * Update cpu_load for any missed ticks, due to tickless idle. The backlog -- cgit From 59543275488d18d878cd2ab2b1072efc1e9ac1c4 Mon Sep 17 00:00:00 2001 From: Byungchul Park Date: Wed, 14 Oct 2015 18:47:35 +0900 Subject: sched/fair: Prepare __update_cpu_load() to handle active tickless There are some cases where distance between ticks is more than one tick while the CPU is not idle, e.g. full NOHZ. However __update_cpu_load() assumes it is the idle tickless case if the distance between ticks is more than 1, even though it can be the active tickless case as well. Thus in the active tickless case, updating the CPU load will not be performed correctly. Where the current code assumes the load for each tick is zero, this is (obviously) not true in non-idle tickless case. We can approximately consider the load ~= this_rq->cpu_load[0] during tickless in non-idle tickless case. Signed-off-by: Byungchul Park Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1444816056-11886-2-git-send-email-byungchul.park@lge.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 49 +++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 41 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 8f3905e3b986..404006ae1ac9 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4283,14 +4283,46 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) return load; } -/* +/** + * __update_cpu_load - update the rq->cpu_load[] statistics + * @this_rq: The rq to update statistics for + * @this_load: The current load + * @pending_updates: The number of missed updates + * @active: !0 for NOHZ_FULL + * * Update rq->cpu_load[] statistics. This function is usually called every - * scheduler tick (TICK_NSEC). With tickless idle this will not be called - * every tick. We fix it up based on jiffies. + * scheduler tick (TICK_NSEC). + * + * This function computes a decaying average: + * + * load[i]' = (1 - 1/2^i) * load[i] + (1/2^i) * load + * + * Because of NOHZ it might not get called on every tick which gives need for + * the @pending_updates argument. + * + * load[i]_n = (1 - 1/2^i) * load[i]_n-1 + (1/2^i) * load_n-1 + * = A * load[i]_n-1 + B ; A := (1 - 1/2^i), B := (1/2^i) * load + * = A * (A * load[i]_n-2 + B) + B + * = A * (A * (A * load[i]_n-3 + B) + B) + B + * = A^3 * load[i]_n-3 + (A^2 + A + 1) * B + * = A^n * load[i]_0 + (A^(n-1) + A^(n-2) + ... + 1) * B + * = A^n * load[i]_0 + ((1 - A^n) / (1 - A)) * B + * = (1 - 1/2^i)^n * (load[i]_0 - load) + load + * + * In the above we've assumed load_n := load, which is true for NOHZ_FULL as + * any change in load would have resulted in the tick being turned back on. + * + * For regular NOHZ, this reduces to: + * + * load[i]_n = (1 - 1/2^i)^n * load[i]_0 + * + * see decay_load_misses(). For NOHZ_FULL we get to subtract and add the extra + * term. See the @active paramter. */ static void __update_cpu_load(struct rq *this_rq, unsigned long this_load, - unsigned long pending_updates) + unsigned long pending_updates, int active) { + unsigned long tickless_load = active ? this_rq->cpu_load[0] : 0; int i, scale; this_rq->nr_load_updates++; @@ -4302,8 +4334,9 @@ static void __update_cpu_load(struct rq *this_rq, unsigned long this_load, /* scale is effectively 1 << i now, and >> i divides by scale */ - old_load = this_rq->cpu_load[i]; + old_load = this_rq->cpu_load[i] - tickless_load; old_load = decay_load_missed(old_load, pending_updates - 1, i); + old_load += tickless_load; new_load = this_load; /* * Round up the averaging division if load is increasing. This @@ -4358,7 +4391,7 @@ static void update_idle_cpu_load(struct rq *this_rq) pending_updates = curr_jiffies - this_rq->last_load_update_tick; this_rq->last_load_update_tick = curr_jiffies; - __update_cpu_load(this_rq, load, pending_updates); + __update_cpu_load(this_rq, load, pending_updates, 0); } /* @@ -4381,7 +4414,7 @@ void update_cpu_load_nohz(void) * We were idle, this means load 0, the current load might be * !0 due to remote wakeups and the sort. */ - __update_cpu_load(this_rq, 0, pending_updates); + __update_cpu_load(this_rq, 0, pending_updates, 0); } raw_spin_unlock(&this_rq->lock); } @@ -4397,7 +4430,7 @@ void update_cpu_load_active(struct rq *this_rq) * See the mess around update_idle_cpu_load() / update_cpu_load_nohz(). */ this_rq->last_load_update_tick = jiffies; - __update_cpu_load(this_rq, load, 1); + __update_cpu_load(this_rq, load, 1, 1); } /* -- cgit From 525705d15e63b7455977408e4601e76e6bc41524 Mon Sep 17 00:00:00 2001 From: Byungchul Park Date: Tue, 10 Nov 2015 09:36:02 +0900 Subject: sched/fair: Consider missed ticks in NOHZ_FULL in update_cpu_load_nohz() Usually the tick can be stopped for an idle CPU in NOHZ. However in NOHZ_FULL mode, a non-idle CPU's tick can also be stopped. However, update_cpu_load_nohz() does not consider the case a non-idle CPU's tick has been stopped at all. This patch makes the update_cpu_load_nohz() know if the calling path comes from NOHZ_FULL or idle NOHZ. Signed-off-by: Byungchul Park Signed-off-by: Peter Zijlstra (Intel) Acked-by: Frederic Weisbecker Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447115762-19734-3-git-send-email-byungchul.park@lge.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 10 ++++++---- kernel/time/tick-sched.c | 8 ++++---- 2 files changed, 10 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 404006ae1ac9..309b1d551f25 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4397,10 +4397,11 @@ static void update_idle_cpu_load(struct rq *this_rq) /* * Called from tick_nohz_idle_exit() -- try and fix up the ticks we missed. */ -void update_cpu_load_nohz(void) +void update_cpu_load_nohz(int active) { struct rq *this_rq = this_rq(); unsigned long curr_jiffies = READ_ONCE(jiffies); + unsigned long load = active ? weighted_cpuload(cpu_of(this_rq)) : 0; unsigned long pending_updates; if (curr_jiffies == this_rq->last_load_update_tick) @@ -4411,10 +4412,11 @@ void update_cpu_load_nohz(void) if (pending_updates) { this_rq->last_load_update_tick = curr_jiffies; /* - * We were idle, this means load 0, the current load might be - * !0 due to remote wakeups and the sort. + * In the regular NOHZ case, we were idle, this means load 0. + * In the NOHZ_FULL case, we were non-idle, we should consider + * its weighted load. */ - __update_cpu_load(this_rq, 0, pending_updates, 0); + __update_cpu_load(this_rq, load, pending_updates, active); } raw_spin_unlock(&this_rq->lock); } diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 7c7ec4515983..515edf3eb62b 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -694,11 +694,11 @@ out: return tick; } -static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now) +static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now, int active) { /* Update jiffies first */ tick_do_update_jiffies64(now); - update_cpu_load_nohz(); + update_cpu_load_nohz(active); calc_load_exit_idle(); touch_softlockup_watchdog(); @@ -725,7 +725,7 @@ static void tick_nohz_full_update_tick(struct tick_sched *ts) if (can_stop_full_tick()) tick_nohz_stop_sched_tick(ts, ktime_get(), cpu); else if (ts->tick_stopped) - tick_nohz_restart_sched_tick(ts, ktime_get()); + tick_nohz_restart_sched_tick(ts, ktime_get(), 1); #endif } @@ -916,7 +916,7 @@ void tick_nohz_idle_exit(void) tick_nohz_stop_idle(ts, now); if (ts->tick_stopped) { - tick_nohz_restart_sched_tick(ts, now); + tick_nohz_restart_sched_tick(ts, now, 0); tick_nohz_account_idle_ticks(ts); } -- cgit From 51170840fe91dfca10fd533b303ea39b2524782a Mon Sep 17 00:00:00 2001 From: Rik van Riel Date: Thu, 5 Nov 2015 15:56:23 -0500 Subject: sched/numa: Cap PTE scanning overhead to 3% of run time There is a fundamental mismatch between the runtime based NUMA scanning at the task level, and the wall clock time NUMA scanning at the mm level. On a severely overloaded system, with very large processes, this mismatch can cause the system to spend all of its time in change_prot_numa(). This can happen if the task spends at least two ticks in change_prot_numa(), and only gets two ticks of CPU time in the real time between two scan intervals of the mm. This patch ensures that a task never spends more than 3% of run time scanning PTEs. It does that by ensuring that in-between task_numa_work() runs, the task spends at least 32x as much time on other things than it did on task_numa_work(). This is done stochastically: if a timer tick happens, or the task gets rescheduled during task_numa_work(), we delay a future run of task_numa_work() until the task has spent at least 32x the amount of CPU time doing something else, as it spent inside task_numa_work(). The longer task_numa_work() takes, the more likely it is this happens. If task_numa_work() takes very little time, chances are low that that code will do anything, but we will not care. Reported-and-tested-by: Jan Stancek Signed-off-by: Rik van Riel Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: mgorman@suse.de Link: http://lkml.kernel.org/r/1446756983-28173-3-git-send-email-riel@redhat.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 309b1d551f25..95b944ecf7e4 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2155,6 +2155,7 @@ void task_numa_work(struct callback_head *work) unsigned long migrate, next_scan, now = jiffies; struct task_struct *p = current; struct mm_struct *mm = p->mm; + u64 runtime = p->se.sum_exec_runtime; struct vm_area_struct *vma; unsigned long start, end; unsigned long nr_pte_updates = 0; @@ -2277,6 +2278,17 @@ out: else reset_ptenuma_scan(p); up_read(&mm->mmap_sem); + + /* + * Make sure tasks use at least 32x as much time to run other code + * than they used here, to limit NUMA PTE scanning overhead to 3% max. + * Usually update_task_scan_period slows down scanning enough; on an + * overloaded system we need to limit overhead on a per task basis. + */ + if (unlikely(p->se.sum_exec_runtime != runtime)) { + u64 diff = p->se.sum_exec_runtime - runtime; + p->node_stamp += 32 * diff; + } } /* -- cgit From 3ea94de15ce9f3a217f6d0a7e9e0f48388902bb7 Mon Sep 17 00:00:00 2001 From: Joonwoo Park Date: Thu, 12 Nov 2015 19:38:54 -0800 Subject: sched/core: Fix incorrect wait time and wait count statistics At present scheduler resets task's wait start timestamp when the task migrates to another rq. This misleads scheduler itself into reporting less wait time than actual by omitting time spent for waiting prior to migration and also more wait count than actual by counting migration as wait end event which can be seen by trace or /proc//sched with CONFIG_SCHEDSTATS=y. Carry forward migrating task's wait time prior to migration and don't count migration as a wait end event to fix such statistics error. In order to determine whether task is migrating mark task->on_rq with TASK_ON_RQ_MIGRATING while dequeuing and enqueuing due to migration. Signed-off-by: Joonwoo Park Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: ohaugan@codeaurora.org Link: http://lkml.kernel.org/r/20151113033854.GA4247@codeaurora.org Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 15 ++++++++++-- kernel/sched/fair.c | 67 +++++++++++++++++++++++++++++++++++++---------------- 2 files changed, 60 insertions(+), 22 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 4d568ac9319e..1b7cb5e95816 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1071,8 +1071,8 @@ static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new { lockdep_assert_held(&rq->lock); - dequeue_task(rq, p, 0); p->on_rq = TASK_ON_RQ_MIGRATING; + dequeue_task(rq, p, 0); set_task_cpu(p, new_cpu); raw_spin_unlock(&rq->lock); @@ -1080,8 +1080,8 @@ static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new raw_spin_lock(&rq->lock); BUG_ON(task_cpu(p) != new_cpu); - p->on_rq = TASK_ON_RQ_QUEUED; enqueue_task(rq, p, 0); + p->on_rq = TASK_ON_RQ_QUEUED; check_preempt_curr(rq, p, 0); return rq; @@ -1274,6 +1274,15 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu) WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING && !p->on_rq); + /* + * Migrating fair class task must have p->on_rq = TASK_ON_RQ_MIGRATING, + * because schedstat_wait_{start,end} rebase migrating task's wait_start + * time relying on p->on_rq. + */ + WARN_ON_ONCE(p->state == TASK_RUNNING && + p->sched_class == &fair_sched_class && + (p->on_rq && !task_on_rq_migrating(p))); + #ifdef CONFIG_LOCKDEP /* * The caller should hold either p->pi_lock or rq->lock, when changing @@ -1310,9 +1319,11 @@ static void __migrate_swap_task(struct task_struct *p, int cpu) src_rq = task_rq(p); dst_rq = cpu_rq(cpu); + p->on_rq = TASK_ON_RQ_MIGRATING; deactivate_task(src_rq, p, 0); set_task_cpu(p, cpu); activate_task(dst_rq, p, 0); + p->on_rq = TASK_ON_RQ_QUEUED; check_preempt_curr(dst_rq, p, 0); } else { /* diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 95b944ecf7e4..f7017ad91400 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -738,12 +738,56 @@ static void update_curr_fair(struct rq *rq) update_curr(cfs_rq_of(&rq->curr->se)); } +#ifdef CONFIG_SCHEDSTATS static inline void update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se) { - schedstat_set(se->statistics.wait_start, rq_clock(rq_of(cfs_rq))); + u64 wait_start = rq_clock(rq_of(cfs_rq)); + + if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) && + likely(wait_start > se->statistics.wait_start)) + wait_start -= se->statistics.wait_start; + + se->statistics.wait_start = wait_start; } +static void +update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se) +{ + struct task_struct *p; + u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start; + + if (entity_is_task(se)) { + p = task_of(se); + if (task_on_rq_migrating(p)) { + /* + * Preserve migrating task's wait time so wait_start + * time stamp can be adjusted to accumulate wait time + * prior to migration. + */ + se->statistics.wait_start = delta; + return; + } + trace_sched_stat_wait(p, delta); + } + + se->statistics.wait_max = max(se->statistics.wait_max, delta); + se->statistics.wait_count++; + se->statistics.wait_sum += delta; + se->statistics.wait_start = 0; +} +#else +static inline void +update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se) +{ +} + +static inline void +update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se) +{ +} +#endif + /* * Task is being enqueued - update stats: */ @@ -757,23 +801,6 @@ static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) update_stats_wait_start(cfs_rq, se); } -static void -update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se) -{ - schedstat_set(se->statistics.wait_max, max(se->statistics.wait_max, - rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start)); - schedstat_set(se->statistics.wait_count, se->statistics.wait_count + 1); - schedstat_set(se->statistics.wait_sum, se->statistics.wait_sum + - rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start); -#ifdef CONFIG_SCHEDSTATS - if (entity_is_task(se)) { - trace_sched_stat_wait(task_of(se), - rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start); - } -#endif - schedstat_set(se->statistics.wait_start, 0); -} - static inline void update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) { @@ -5745,8 +5772,8 @@ static void detach_task(struct task_struct *p, struct lb_env *env) { lockdep_assert_held(&env->src_rq->lock); - deactivate_task(env->src_rq, p, 0); p->on_rq = TASK_ON_RQ_MIGRATING; + deactivate_task(env->src_rq, p, 0); set_task_cpu(p, env->dst_cpu); } @@ -5879,8 +5906,8 @@ static void attach_task(struct rq *rq, struct task_struct *p) lockdep_assert_held(&rq->lock); BUG_ON(task_rq(p) != rq); - p->on_rq = TASK_ON_RQ_QUEUED; activate_task(rq, p, 0); + p->on_rq = TASK_ON_RQ_QUEUED; check_preempt_curr(rq, p, 0); } -- cgit From 01783e0d452736d7deff1b920c5eccad67adc428 Mon Sep 17 00:00:00 2001 From: Geliang Tang Date: Sun, 15 Nov 2015 18:18:40 +0800 Subject: sched/core: Use list_is_singular() in sched_can_stop_tick() Use list_is_singular() to check if run_list has only one entry. Signed-off-by: Geliang Tang Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/a5453fafd735affcf28e53a1d0a3d6965cb5dbb5.1447582547.git.geliangtang@163.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 1b7cb5e95816..5b420d29bce3 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -731,7 +731,7 @@ bool sched_can_stop_tick(void) if (current->policy == SCHED_RR) { struct sched_rt_entity *rt_se = ¤t->rt; - return rt_se->run_list.prev == rt_se->run_list.next; + return list_is_singular(&rt_se->run_list); } /* -- cgit From 64038f292a1b33c7d46bd11f62f7798101152c00 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:11 +0100 Subject: stop_machine: Fix possible cpu_stopper_thread() crash stop_one_cpu_nowait(fn) will crash the kernel if the callback returns nonzero, work->done == NULL in this case. This needs more cleanups, cpu_stop_signal_done() is called right after we check done != NULL and it does the same check. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193311.GA8242@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 867bc20e1ef1..1a66a9569915 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -454,7 +454,7 @@ repeat: preempt_disable(); ret = fn(arg); - if (ret) + if (ret && done) done->ret = ret; /* restore preemption and check it's still balanced */ -- cgit From 6a19005157c464b47b2082f2617d12bc11198a0d Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:14 +0100 Subject: stop_machine: Don't disable preemption in stop_two_cpus() Now that stop_two_cpus() path does not check cpu_active() we can remove preempt_disable(), it was only needed to ensure that stop_machine() can not be called after we observe cpu_active() == T and before we queue the new work. Also, turn the pointless and confusing ->executed check into WARN_ON(). We know that both works must be executed, otherwise we have a bug. And in fact I think that done->executed should die, see the next changes. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193314.GA8249@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 1a66a9569915..17f01a9dc3df 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -258,7 +258,6 @@ int stop_two_cpus(unsigned int cpu1, unsigned int cpu2, cpu_stop_fn_t fn, void * struct cpu_stop_work work1, work2; struct multi_stop_data msdata; - preempt_disable(); msdata = (struct multi_stop_data){ .fn = fn, .data = arg, @@ -277,16 +276,12 @@ int stop_two_cpus(unsigned int cpu1, unsigned int cpu2, cpu_stop_fn_t fn, void * if (cpu1 > cpu2) swap(cpu1, cpu2); - if (cpu_stop_queue_two_works(cpu1, &work1, cpu2, &work2)) { - preempt_enable(); + if (cpu_stop_queue_two_works(cpu1, &work1, cpu2, &work2)) return -ENOENT; - } - - preempt_enable(); wait_for_completion(&done.completion); - - return done.executed ? done.ret : -ENOENT; + WARN_ON(!done.executed); + return done.ret; } /** -- cgit From 1b034bd989aa4a396c13d305759c376c52595a97 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Tue, 17 Nov 2015 18:05:23 +0100 Subject: stop_machine: Make cpu_stop_queue_work() and stop_one_cpu_nowait() return bool Change cpu_stop_queue_work() to return true if the work was queued and change stop_one_cpu_nowait() to return the result of cpu_stop_queue_work(). This makes it more useful, for example now you can alloc cpu_stop_work for stop_one_cpu_nowait() and free it in the callback or if stop_one_cpu_nowait() fails, currently this is impossible because you can't know if @fn will be called or not. Also, this allows to kill cpu_stop_done->executed, see the next changes. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151117170523.GA13955@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 17f01a9dc3df..0ec1f16da379 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -81,17 +81,21 @@ static void __cpu_stop_queue_work(struct cpu_stopper *stopper, } /* queue @work to @stopper. if offline, @work is completed immediately */ -static void cpu_stop_queue_work(unsigned int cpu, struct cpu_stop_work *work) +static bool cpu_stop_queue_work(unsigned int cpu, struct cpu_stop_work *work) { struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu); unsigned long flags; + bool enabled; spin_lock_irqsave(&stopper->lock, flags); - if (stopper->enabled) + enabled = stopper->enabled; + if (enabled) __cpu_stop_queue_work(stopper, work); else cpu_stop_signal_done(work->done, false); spin_unlock_irqrestore(&stopper->lock, flags); + + return enabled; } /** @@ -297,12 +301,16 @@ int stop_two_cpus(unsigned int cpu1, unsigned int cpu2, cpu_stop_fn_t fn, void * * * CONTEXT: * Don't care. + * + * RETURNS: + * true if cpu_stop_work was queued successfully and @fn will be called, + * false otherwise. */ -void stop_one_cpu_nowait(unsigned int cpu, cpu_stop_fn_t fn, void *arg, +bool stop_one_cpu_nowait(unsigned int cpu, cpu_stop_fn_t fn, void *arg, struct cpu_stop_work *work_buf) { *work_buf = (struct cpu_stop_work){ .fn = fn, .arg = arg, }; - cpu_stop_queue_work(cpu, work_buf); + return cpu_stop_queue_work(cpu, work_buf); } /* static data for stop_cpus */ -- cgit From 958c5f848e17e216df138cc2161b07b7120e2d15 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:20 +0100 Subject: stop_machine: Change stop_one_cpu() to rely on cpu_stop_queue_work() Change stop_one_cpu() to return -ENOENT if cpu_stop_queue_work() fails. Otherwise we know that ->executed must be true after wait_for_completion() so we can just return done.ret. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193320.GA8259@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 0ec1f16da379..68b73c400408 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -128,9 +128,11 @@ int stop_one_cpu(unsigned int cpu, cpu_stop_fn_t fn, void *arg) struct cpu_stop_work work = { .fn = fn, .arg = arg, .done = &done }; cpu_stop_init_done(&done, 1); - cpu_stop_queue_work(cpu, &work); + if (!cpu_stop_queue_work(cpu, &work)) + return -ENOENT; wait_for_completion(&done.completion); - return done.executed ? done.ret : -ENOENT; + WARN_ON(!done.executed); + return done.ret; } /* This controls the threads on each CPU. */ -- cgit From 4aff1ca6970afbf9cd916c34a9c442c8ccba905e Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:23 +0100 Subject: stop_machine: Change __stop_cpus() to rely on cpu_stop_queue_work() Change queue_stop_cpus_work() to return true if it queues at least one work, this means that the caller should wait. __stop_cpus() can check the value returned by queue_stop_cpus_work() and avoid done.executed, just like stop_one_cpu() does. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193323.GA8262@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 68b73c400408..ed2019ac192d 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -318,12 +318,13 @@ bool stop_one_cpu_nowait(unsigned int cpu, cpu_stop_fn_t fn, void *arg, /* static data for stop_cpus */ static DEFINE_MUTEX(stop_cpus_mutex); -static void queue_stop_cpus_work(const struct cpumask *cpumask, +static bool queue_stop_cpus_work(const struct cpumask *cpumask, cpu_stop_fn_t fn, void *arg, struct cpu_stop_done *done) { struct cpu_stop_work *work; unsigned int cpu; + bool queued = false; /* * Disable preemption while queueing to avoid getting @@ -336,9 +337,12 @@ static void queue_stop_cpus_work(const struct cpumask *cpumask, work->fn = fn; work->arg = arg; work->done = done; - cpu_stop_queue_work(cpu, work); + if (cpu_stop_queue_work(cpu, work)) + queued = true; } lg_global_unlock(&stop_cpus_lock); + + return queued; } static int __stop_cpus(const struct cpumask *cpumask, @@ -347,9 +351,11 @@ static int __stop_cpus(const struct cpumask *cpumask, struct cpu_stop_done done; cpu_stop_init_done(&done, cpumask_weight(cpumask)); - queue_stop_cpus_work(cpumask, fn, arg, &done); + if (!queue_stop_cpus_work(cpumask, fn, arg, &done)) + return -ENOENT; wait_for_completion(&done.completion); - return done.executed ? done.ret : -ENOENT; + WARN_ON(!done.executed); + return done.ret; } /** -- cgit From 6fa3b826bcb3309157166e6e523a4be236fe267a Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:26 +0100 Subject: stop_machine: Kill cpu_stop_done->executed Now that cpu_stop_done->executed becomes write-only (ignoring WARN_ON() checks) we can remove it. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193326.GA8269@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 12 +++--------- 1 file changed, 3 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index ed2019ac192d..09eb83fc9f0e 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -28,7 +28,6 @@ */ struct cpu_stop_done { atomic_t nr_todo; /* nr left to execute */ - bool executed; /* actually executed? */ int ret; /* collected return value */ struct completion completion; /* fired if nr_todo reaches 0 */ }; @@ -63,11 +62,9 @@ static void cpu_stop_init_done(struct cpu_stop_done *done, unsigned int nr_todo) } /* signal completion unless @done is NULL */ -static void cpu_stop_signal_done(struct cpu_stop_done *done, bool executed) +static void cpu_stop_signal_done(struct cpu_stop_done *done) { if (done) { - if (executed) - done->executed = true; if (atomic_dec_and_test(&done->nr_todo)) complete(&done->completion); } @@ -92,7 +89,7 @@ static bool cpu_stop_queue_work(unsigned int cpu, struct cpu_stop_work *work) if (enabled) __cpu_stop_queue_work(stopper, work); else - cpu_stop_signal_done(work->done, false); + cpu_stop_signal_done(work->done); spin_unlock_irqrestore(&stopper->lock, flags); return enabled; @@ -131,7 +128,6 @@ int stop_one_cpu(unsigned int cpu, cpu_stop_fn_t fn, void *arg) if (!cpu_stop_queue_work(cpu, &work)) return -ENOENT; wait_for_completion(&done.completion); - WARN_ON(!done.executed); return done.ret; } @@ -286,7 +282,6 @@ int stop_two_cpus(unsigned int cpu1, unsigned int cpu2, cpu_stop_fn_t fn, void * return -ENOENT; wait_for_completion(&done.completion); - WARN_ON(!done.executed); return done.ret; } @@ -354,7 +349,6 @@ static int __stop_cpus(const struct cpumask *cpumask, if (!queue_stop_cpus_work(cpumask, fn, arg, &done)) return -ENOENT; wait_for_completion(&done.completion); - WARN_ON(!done.executed); return done.ret; } @@ -467,6 +461,7 @@ repeat: ret = fn(arg); if (ret && done) done->ret = ret; + cpu_stop_signal_done(done); /* restore preemption and check it's still balanced */ preempt_enable(); @@ -475,7 +470,6 @@ repeat: kallsyms_lookup((unsigned long)fn, NULL, NULL, NULL, ksym_buf), arg); - cpu_stop_signal_done(done, true); goto repeat; } } -- cgit From dd2e3121e3cb16d03a6e3f2db48f260f046f39c2 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:29 +0100 Subject: stop_machine: Shift the 'done != NULL' check from cpu_stop_signal_done() to callers Change cpu_stop_queue_work() and cpu_stopper_thread() to check done != NULL before cpu_stop_signal_done(done). This makes the code more clean imo, note that cpu_stopper_thread() has to do this check anyway. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193329.GA8274@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 09eb83fc9f0e..7ff7acee2c76 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -64,10 +64,8 @@ static void cpu_stop_init_done(struct cpu_stop_done *done, unsigned int nr_todo) /* signal completion unless @done is NULL */ static void cpu_stop_signal_done(struct cpu_stop_done *done) { - if (done) { - if (atomic_dec_and_test(&done->nr_todo)) - complete(&done->completion); - } + if (atomic_dec_and_test(&done->nr_todo)) + complete(&done->completion); } static void __cpu_stop_queue_work(struct cpu_stopper *stopper, @@ -88,7 +86,7 @@ static bool cpu_stop_queue_work(unsigned int cpu, struct cpu_stop_work *work) enabled = stopper->enabled; if (enabled) __cpu_stop_queue_work(stopper, work); - else + else if (work->done) cpu_stop_signal_done(work->done); spin_unlock_irqrestore(&stopper->lock, flags); @@ -457,12 +455,12 @@ repeat: /* cpu stop callbacks are not allowed to sleep */ preempt_disable(); - ret = fn(arg); - if (ret && done) - done->ret = ret; - cpu_stop_signal_done(done); - + if (done) { + if (ret) + done->ret = ret; + cpu_stop_signal_done(done); + } /* restore preemption and check it's still balanced */ preempt_enable(); WARN_ONCE(preempt_count(), -- cgit From accaf6ea3db6f5fb997f096b6eefd5431d03f7e5 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Sun, 15 Nov 2015 20:33:32 +0100 Subject: stop_machine: Clean up the usage of the preemption counter in cpu_stopper_thread() 1. Change this code to use preempt_count_inc/preempt_count_dec; this way it works even if CONFIG_PREEMPT_COUNT=n, and we avoid the unnecessary __preempt_schedule() check (stop_sched_class is not preemptible). And this makes clear that we only want to make preempt_count() != 0 for __might_sleep() / schedule_debug(). 2. Change WARN_ONCE() to use %pf to print the function name and remove kallsyms_lookup/ksym_buf. 3. Move "int ret" into the "if (work)" block, this looks more consistent. Signed-off-by: Oleg Nesterov Signed-off-by: Peter Zijlstra (Intel) Acked-by: Tejun Heo Cc: Linus Torvalds Cc: Mike Galbraith Cc: Milos Vyletel Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20151115193332.GA8281@redhat.com Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 7ff7acee2c76..61101193967e 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -435,7 +435,6 @@ static void cpu_stopper_thread(unsigned int cpu) { struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu); struct cpu_stop_work *work; - int ret; repeat: work = NULL; @@ -451,23 +450,19 @@ repeat: cpu_stop_fn_t fn = work->fn; void *arg = work->arg; struct cpu_stop_done *done = work->done; - char ksym_buf[KSYM_NAME_LEN] __maybe_unused; + int ret; - /* cpu stop callbacks are not allowed to sleep */ - preempt_disable(); + /* cpu stop callbacks must not sleep, make in_atomic() == T */ + preempt_count_inc(); ret = fn(arg); if (done) { if (ret) done->ret = ret; cpu_stop_signal_done(done); } - /* restore preemption and check it's still balanced */ - preempt_enable(); + preempt_count_dec(); WARN_ONCE(preempt_count(), - "cpu_stop: %s(%p) leaked preempt count\n", - kallsyms_lookup((unsigned long)fn, NULL, NULL, NULL, - ksym_buf), arg); - + "cpu_stop: %pf(%p) leaked preempt count\n", fn, arg); goto repeat; } } -- cgit From 525628c73bd6af65f27d927e699e7460d7d55ed3 Mon Sep 17 00:00:00 2001 From: Byungchul Park Date: Wed, 18 Nov 2015 09:34:59 +0900 Subject: sched/fair: Modify the comment about lock assumptions in migrate_task_rq_fair() The comment describing migrate_task_rq_fair() says that the caller should hold p->pi_lock. But in some cases the caller can hold task_rq(p)->lock instead of p->pi_lock. So the comment is broken and this patch fixes it. Signed-off-by: Byungchul Park Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447806899-20303-1-git-send-email-byungchul.park@lge.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index f7017ad91400..ff8ec8695957 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -5058,8 +5058,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f /* * Called immediately before a task is migrated to a new cpu; task_cpu(p) and * cfs_rq_of(p) references at time of call are still valid and identify the - * previous cpu. However, the caller only guarantees p->pi_lock is held; no - * other assumptions, including the state of rq->lock, should be made. + * previous cpu. The caller guarantees p->pi_lock or task_rq(p)->lock is held. */ static void migrate_task_rq_fair(struct task_struct *p) { -- cgit From 64d816cba06c67eeee455b8c78ebcda349d49c24 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Mon, 9 Nov 2015 19:09:21 -0500 Subject: locking/qspinlock: Use _acquire/_release() versions of cmpxchg() & xchg() This patch replaces the cmpxchg() and xchg() calls in the native qspinlock code with the more relaxed _acquire or _release versions of those calls to enable other architectures to adopt queued spinlocks with less memory barrier performance overhead. Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447114167-47185-2-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock.c | 29 ++++++++++++++++++++++++----- 1 file changed, 24 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c index 87e9ce6a63c5..7868418ea586 100644 --- a/kernel/locking/qspinlock.c +++ b/kernel/locking/qspinlock.c @@ -14,8 +14,9 @@ * (C) Copyright 2013-2015 Hewlett-Packard Development Company, L.P. * (C) Copyright 2013-2014 Red Hat, Inc. * (C) Copyright 2015 Intel Corp. + * (C) Copyright 2015 Hewlett-Packard Enterprise Development LP * - * Authors: Waiman Long + * Authors: Waiman Long * Peter Zijlstra */ @@ -176,7 +177,12 @@ static __always_inline u32 xchg_tail(struct qspinlock *lock, u32 tail) { struct __qspinlock *l = (void *)lock; - return (u32)xchg(&l->tail, tail >> _Q_TAIL_OFFSET) << _Q_TAIL_OFFSET; + /* + * Use release semantics to make sure that the MCS node is properly + * initialized before changing the tail code. + */ + return (u32)xchg_release(&l->tail, + tail >> _Q_TAIL_OFFSET) << _Q_TAIL_OFFSET; } #else /* _Q_PENDING_BITS == 8 */ @@ -208,7 +214,11 @@ static __always_inline u32 xchg_tail(struct qspinlock *lock, u32 tail) for (;;) { new = (val & _Q_LOCKED_PENDING_MASK) | tail; - old = atomic_cmpxchg(&lock->val, val, new); + /* + * Use release semantics to make sure that the MCS node is + * properly initialized before changing the tail code. + */ + old = atomic_cmpxchg_release(&lock->val, val, new); if (old == val) break; @@ -319,7 +329,11 @@ void queued_spin_lock_slowpath(struct qspinlock *lock, u32 val) if (val == new) new |= _Q_PENDING_VAL; - old = atomic_cmpxchg(&lock->val, val, new); + /* + * Acquire semantic is required here as the function may + * return immediately if the lock was free. + */ + old = atomic_cmpxchg_acquire(&lock->val, val, new); if (old == val) break; @@ -426,7 +440,12 @@ queue: set_locked(lock); break; } - old = atomic_cmpxchg(&lock->val, val, _Q_LOCKED_VAL); + /* + * The smp_load_acquire() call above has provided the necessary + * acquire semantics required for locking. At most two + * iterations of this loop may be ran. + */ + old = atomic_cmpxchg_relaxed(&lock->val, val, _Q_LOCKED_VAL); if (old == val) goto release; /* No contention */ -- cgit From 81b5598665a24083dd889fbd8cb08b0d8de4b8ad Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Mon, 9 Nov 2015 19:09:22 -0500 Subject: locking/qspinlock: Prefetch the next node cacheline A queue head CPU, after acquiring the lock, will have to notify the next CPU in the wait queue that it has became the new queue head. This involves loading a new cacheline from the MCS node of the next CPU. That operation can be expensive and add to the latency of locking operation. This patch addes code to optmistically prefetch the next MCS node cacheline if the next pointer is defined and it has been spinning for the MCS lock for a while. This reduces the locking latency and improves the system throughput. The performance change will depend on whether the prefetch overhead can be hidden within the latency of the lock spin loop. On really short critical section, there may not be performance gain at all. With longer critical section, however, it was found to have a performance boost of 5-10% over a range of different queue depths with a spinlock loop microbenchmark. Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447114167-47185-3-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'kernel') diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c index 7868418ea586..365b2033f55e 100644 --- a/kernel/locking/qspinlock.c +++ b/kernel/locking/qspinlock.c @@ -407,6 +407,16 @@ queue: pv_wait_node(node); arch_mcs_spin_lock_contended(&node->locked); + + /* + * While waiting for the MCS lock, the next pointer may have + * been set by another lock waiter. We optimistically load + * the next pointer & prefetch the cacheline for writing + * to reduce latency in the upcoming MCS unlock operation. + */ + next = READ_ONCE(node->next); + if (next) + prefetchw(next); } /* -- cgit From aa68744f80bfb6f26fbe7f10e42876066f7dac1b Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Mon, 9 Nov 2015 19:09:23 -0500 Subject: locking/qspinlock: Avoid redundant read of next pointer With optimistic prefetch of the next node cacheline, the next pointer may have been properly inititalized. As a result, the reading of node->next in the contended path may be redundant. This patch eliminates the redundant read if the next pointer value is not NULL. Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447114167-47185-4-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c index 365b2033f55e..986207887def 100644 --- a/kernel/locking/qspinlock.c +++ b/kernel/locking/qspinlock.c @@ -396,6 +396,7 @@ queue: * p,*,* -> n,*,* */ old = xchg_tail(lock, tail); + next = NULL; /* * if there was a previous node; link it and wait until reaching the @@ -463,10 +464,12 @@ queue: } /* - * contended path; wait for next, release. + * contended path; wait for next if not observed yet, release. */ - while (!(next = READ_ONCE(node->next))) - cpu_relax(); + if (!next) { + while (!(next = READ_ONCE(node->next))) + cpu_relax(); + } arch_mcs_spin_unlock_contended(&next->locked); pv_kick_node(lock, next); -- cgit From d78045306c41bd9334b956e4e7fa77cc72f06a40 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Mon, 9 Nov 2015 19:09:24 -0500 Subject: locking/pvqspinlock, x86: Optimize the PV unlock code path The unlock function in queued spinlocks was optimized for better performance on bare metal systems at the expense of virtualized guests. For x86-64 systems, the unlock call needs to go through a PV_CALLEE_SAVE_REGS_THUNK() which saves and restores 8 64-bit registers before calling the real __pv_queued_spin_unlock() function. The thunk code may also be in a separate cacheline from __pv_queued_spin_unlock(). This patch optimizes the PV unlock code path by: 1) Moving the unlock slowpath code from the fastpath into a separate __pv_queued_spin_unlock_slowpath() function to make the fastpath as simple as possible.. 2) For x86-64, hand-coded an assembly function to combine the register saving thunk code with the fastpath code. Only registers that are used in the fastpath will be saved and restored. If the fastpath fails, the slowpath function will be called via another PV_CALLEE_SAVE_REGS_THUNK(). For 32-bit, it falls back to the C __pv_queued_spin_unlock() code as the thunk saves and restores only one 32-bit register. With a microbenchmark of 5M lock-unlock loop, the table below shows the execution times before and after the patch with different number of threads in a VM running on a 32-core Westmere-EX box with x86-64 4.2-rc1 based kernels: Threads Before patch After patch % Change ------- ------------ ----------- -------- 1 134.1 ms 119.3 ms -11% 2 1286 ms 953 ms -26% 3 3715 ms 3480 ms -6.3% 4 4092 ms 3764 ms -8.0% Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447114167-47185-5-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock_paravirt.h | 43 +++++++++++++++++++++++-------------- 1 file changed, 27 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h index f0450ff4829b..4bd323d38c60 100644 --- a/kernel/locking/qspinlock_paravirt.h +++ b/kernel/locking/qspinlock_paravirt.h @@ -308,23 +308,14 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) } /* - * PV version of the unlock function to be used in stead of - * queued_spin_unlock(). + * PV versions of the unlock fastpath and slowpath functions to be used + * instead of queued_spin_unlock(). */ -__visible void __pv_queued_spin_unlock(struct qspinlock *lock) +__visible void +__pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked) { struct __qspinlock *l = (void *)lock; struct pv_node *node; - u8 locked; - - /* - * We must not unlock if SLOW, because in that case we must first - * unhash. Otherwise it would be possible to have multiple @lock - * entries, which would be BAD. - */ - locked = cmpxchg(&l->locked, _Q_LOCKED_VAL, 0); - if (likely(locked == _Q_LOCKED_VAL)) - return; if (unlikely(locked != _Q_SLOW_VAL)) { WARN(!debug_locks_silent, @@ -363,12 +354,32 @@ __visible void __pv_queued_spin_unlock(struct qspinlock *lock) */ pv_kick(node->cpu); } + /* * Include the architecture specific callee-save thunk of the * __pv_queued_spin_unlock(). This thunk is put together with - * __pv_queued_spin_unlock() near the top of the file to make sure - * that the callee-save thunk and the real unlock function are close - * to each other sharing consecutive instruction cachelines. + * __pv_queued_spin_unlock() to make the callee-save thunk and the real unlock + * function close to each other sharing consecutive instruction cachelines. + * Alternatively, architecture specific version of __pv_queued_spin_unlock() + * can be defined. */ #include +#ifndef __pv_queued_spin_unlock +__visible void __pv_queued_spin_unlock(struct qspinlock *lock) +{ + struct __qspinlock *l = (void *)lock; + u8 locked; + + /* + * We must not unlock if SLOW, because in that case we must first + * unhash. Otherwise it would be possible to have multiple @lock + * entries, which would be BAD. + */ + locked = cmpxchg(&l->locked, _Q_LOCKED_VAL, 0); + if (likely(locked == _Q_LOCKED_VAL)) + return; + + __pv_queued_spin_unlock_slowpath(lock, locked); +} +#endif /* __pv_queued_spin_unlock */ -- cgit From 2a67e741bbbc022e0fadf8c6dbc3a76019ecd0cf Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 8 Oct 2015 12:24:23 +0200 Subject: rcu: Create transitive rnp->lock acquisition functions Providing RCU's memory-ordering guarantees requires that the rcu_node tree's locking provide transitive memory ordering, which the Linux kernel's spinlocks currently do not provide unless smp_mb__after_unlock_lock() is used. Having a separate smp_mb__after_unlock_lock() after each and every lock acquisition is error-prone, hard to read, and a bit annoying, so this commit provides wrapper functions that pull in the smp_mb__after_unlock_lock() invocations. Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 78 ++++++++++++++++-------------------------------- kernel/rcu/tree.h | 39 ++++++++++++++++++++++++ kernel/rcu/tree_plugin.h | 18 ++++------- 3 files changed, 71 insertions(+), 64 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index f07343b54fe5..daf17e248757 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -1534,10 +1534,8 @@ rcu_start_future_gp(struct rcu_node *rnp, struct rcu_data *rdp, * hold it, acquire the root rcu_node structure's lock in order to * start one (if needed). */ - if (rnp != rnp_root) { - raw_spin_lock(&rnp_root->lock); - smp_mb__after_unlock_lock(); - } + if (rnp != rnp_root) + raw_spin_lock_rcu_node(rnp_root); /* * Get a new grace-period number. If there really is no grace @@ -1786,11 +1784,10 @@ static void note_gp_changes(struct rcu_state *rsp, struct rcu_data *rdp) if ((rdp->gpnum == READ_ONCE(rnp->gpnum) && rdp->completed == READ_ONCE(rnp->completed) && !unlikely(READ_ONCE(rdp->gpwrap))) || /* w/out lock. */ - !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */ + !raw_spin_trylock_rcu_node(rnp)) { /* irqs already off, so later. */ local_irq_restore(flags); return; } - smp_mb__after_unlock_lock(); needwake = __note_gp_changes(rsp, rnp, rdp); raw_spin_unlock_irqrestore(&rnp->lock, flags); if (needwake) @@ -1814,8 +1811,7 @@ static int rcu_gp_init(struct rcu_state *rsp) struct rcu_node *rnp = rcu_get_root(rsp); WRITE_ONCE(rsp->gp_activity, jiffies); - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_irq_rcu_node(rnp); if (!READ_ONCE(rsp->gp_flags)) { /* Spurious wakeup, tell caller to go back to sleep. */ raw_spin_unlock_irq(&rnp->lock); @@ -1847,8 +1843,7 @@ static int rcu_gp_init(struct rcu_state *rsp) */ rcu_for_each_leaf_node(rsp, rnp) { rcu_gp_slow(rsp, gp_preinit_delay); - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_irq_rcu_node(rnp); if (rnp->qsmaskinit == rnp->qsmaskinitnext && !rnp->wait_blkd_tasks) { /* Nothing to do on this leaf rcu_node structure. */ @@ -1904,8 +1899,7 @@ static int rcu_gp_init(struct rcu_state *rsp) */ rcu_for_each_node_breadth_first(rsp, rnp) { rcu_gp_slow(rsp, gp_init_delay); - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_irq_rcu_node(rnp); rdp = this_cpu_ptr(rsp->rda); rcu_preempt_check_blocked_tasks(rnp); rnp->qsmask = rnp->qsmaskinit; @@ -1973,8 +1967,7 @@ static void rcu_gp_fqs(struct rcu_state *rsp, bool first_time) } /* Clear flag to prevent immediate re-entry. */ if (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) { - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_irq_rcu_node(rnp); WRITE_ONCE(rsp->gp_flags, READ_ONCE(rsp->gp_flags) & ~RCU_GP_FLAG_FQS); raw_spin_unlock_irq(&rnp->lock); @@ -1993,8 +1986,7 @@ static void rcu_gp_cleanup(struct rcu_state *rsp) struct rcu_node *rnp = rcu_get_root(rsp); WRITE_ONCE(rsp->gp_activity, jiffies); - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_irq_rcu_node(rnp); gp_duration = jiffies - rsp->gp_start; if (gp_duration > rsp->gp_max) rsp->gp_max = gp_duration; @@ -2019,8 +2011,7 @@ static void rcu_gp_cleanup(struct rcu_state *rsp) * grace period is recorded in any of the rcu_node structures. */ rcu_for_each_node_breadth_first(rsp, rnp) { - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_irq_rcu_node(rnp); WARN_ON_ONCE(rcu_preempt_blocked_readers_cgp(rnp)); WARN_ON_ONCE(rnp->qsmask); WRITE_ONCE(rnp->completed, rsp->gpnum); @@ -2035,8 +2026,7 @@ static void rcu_gp_cleanup(struct rcu_state *rsp) rcu_gp_slow(rsp, gp_cleanup_delay); } rnp = rcu_get_root(rsp); - raw_spin_lock_irq(&rnp->lock); - smp_mb__after_unlock_lock(); /* Order GP before ->completed update. */ + raw_spin_lock_irq_rcu_node(rnp); /* Order GP before ->completed update. */ rcu_nocb_gp_set(rnp, nocb); /* Declare grace period done. */ @@ -2284,8 +2274,7 @@ rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp, raw_spin_unlock_irqrestore(&rnp->lock, flags); rnp_c = rnp; rnp = rnp->parent; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); oldmask = rnp_c->qsmask; } @@ -2332,8 +2321,7 @@ static void rcu_report_unblock_qs_rnp(struct rcu_state *rsp, gps = rnp->gpnum; mask = rnp->grpmask; raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */ - raw_spin_lock(&rnp_p->lock); /* irqs already disabled. */ - smp_mb__after_unlock_lock(); + raw_spin_lock_rcu_node(rnp_p); /* irqs already disabled. */ rcu_report_qs_rnp(mask, rsp, rnp_p, gps, flags); } @@ -2355,8 +2343,7 @@ rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp) struct rcu_node *rnp; rnp = rdp->mynode; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if ((rdp->cpu_no_qs.b.norm && rdp->rcu_qs_ctr_snap == __this_cpu_read(rcu_qs_ctr)) || rdp->gpnum != rnp->gpnum || rnp->completed == rnp->gpnum || @@ -2582,8 +2569,7 @@ static void rcu_cleanup_dead_rnp(struct rcu_node *rnp_leaf) rnp = rnp->parent; if (!rnp) break; - raw_spin_lock(&rnp->lock); /* irqs already disabled. */ - smp_mb__after_unlock_lock(); /* GP memory ordering. */ + raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ rnp->qsmaskinit &= ~mask; rnp->qsmask &= ~mask; if (rnp->qsmaskinit) { @@ -2611,8 +2597,7 @@ static void rcu_cleanup_dying_idle_cpu(int cpu, struct rcu_state *rsp) /* Remove outgoing CPU from mask in the leaf rcu_node structure. */ mask = rdp->grpmask; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); /* Enforce GP memory-order guarantee. */ + raw_spin_lock_irqsave_rcu_node(rnp, flags); /* Enforce GP memory-order guarantee. */ rnp->qsmaskinitnext &= ~mask; raw_spin_unlock_irqrestore(&rnp->lock, flags); } @@ -2809,8 +2794,7 @@ static void force_qs_rnp(struct rcu_state *rsp, rcu_for_each_leaf_node(rsp, rnp) { cond_resched_rcu_qs(); mask = 0; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (rnp->qsmask == 0) { if (rcu_state_p == &rcu_sched_state || rsp != rcu_state_p || @@ -2881,8 +2865,7 @@ static void force_quiescent_state(struct rcu_state *rsp) /* rnp_old == rcu_get_root(rsp), rnp == NULL. */ /* Reached the root of the rcu_node tree, acquire lock. */ - raw_spin_lock_irqsave(&rnp_old->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp_old, flags); raw_spin_unlock(&rnp_old->fqslock); if (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) { rsp->n_force_qs_lh++; @@ -3005,8 +2988,7 @@ static void __call_rcu_core(struct rcu_state *rsp, struct rcu_data *rdp, if (!rcu_gp_in_progress(rsp)) { struct rcu_node *rnp_root = rcu_get_root(rsp); - raw_spin_lock(&rnp_root->lock); - smp_mb__after_unlock_lock(); + raw_spin_lock_rcu_node(rnp_root); needwake = rcu_start_gp(rsp); raw_spin_unlock(&rnp_root->lock); if (needwake) @@ -3426,8 +3408,7 @@ static void sync_exp_reset_tree_hotplug(struct rcu_state *rsp) * CPUs for the current rcu_node structure up the rcu_node tree. */ rcu_for_each_leaf_node(rsp, rnp) { - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (rnp->expmaskinit == rnp->expmaskinitnext) { raw_spin_unlock_irqrestore(&rnp->lock, flags); continue; /* No new CPUs, nothing to do. */ @@ -3447,8 +3428,7 @@ static void sync_exp_reset_tree_hotplug(struct rcu_state *rsp) rnp_up = rnp->parent; done = false; while (rnp_up) { - raw_spin_lock_irqsave(&rnp_up->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp_up, flags); if (rnp_up->expmaskinit) done = true; rnp_up->expmaskinit |= mask; @@ -3472,8 +3452,7 @@ static void __maybe_unused sync_exp_reset_tree(struct rcu_state *rsp) sync_exp_reset_tree_hotplug(rsp); rcu_for_each_node_breadth_first(rsp, rnp) { - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); WARN_ON_ONCE(rnp->expmask); rnp->expmask = rnp->expmaskinit; raw_spin_unlock_irqrestore(&rnp->lock, flags); @@ -3531,8 +3510,7 @@ static void __rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp, mask = rnp->grpmask; raw_spin_unlock(&rnp->lock); /* irqs remain disabled */ rnp = rnp->parent; - raw_spin_lock(&rnp->lock); /* irqs already disabled */ - smp_mb__after_unlock_lock(); + raw_spin_lock_rcu_node(rnp); /* irqs already disabled */ WARN_ON_ONCE(!(rnp->expmask & mask)); rnp->expmask &= ~mask; } @@ -3549,8 +3527,7 @@ static void __maybe_unused rcu_report_exp_rnp(struct rcu_state *rsp, { unsigned long flags; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); __rcu_report_exp_rnp(rsp, rnp, wake, flags); } @@ -3564,8 +3541,7 @@ static void rcu_report_exp_cpu_mult(struct rcu_state *rsp, struct rcu_node *rnp, { unsigned long flags; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (!(rnp->expmask & mask)) { raw_spin_unlock_irqrestore(&rnp->lock, flags); return; @@ -3708,8 +3684,7 @@ static void sync_rcu_exp_select_cpus(struct rcu_state *rsp, sync_exp_reset_tree(rsp); rcu_for_each_leaf_node(rsp, rnp) { - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); /* Each pass checks a CPU for identity, offline, and idle. */ mask_ofl_test = 0; @@ -4198,8 +4173,7 @@ rcu_init_percpu_data(int cpu, struct rcu_state *rsp) */ rnp = rdp->mynode; mask = rdp->grpmask; - raw_spin_lock(&rnp->lock); /* irqs already disabled. */ - smp_mb__after_unlock_lock(); + raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ rnp->qsmaskinitnext |= mask; rnp->expmaskinitnext |= mask; if (!rdp->beenonline) diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index 9fb4e238d4dc..f32bebb6bc90 100644 --- a/kernel/rcu/tree.h +++ b/kernel/rcu/tree.h @@ -664,3 +664,42 @@ static inline void rcu_nocb_q_lengths(struct rcu_data *rdp, long *ql, long *qll) #else /* #ifdef CONFIG_PPC */ #define smp_mb__after_unlock_lock() do { } while (0) #endif /* #else #ifdef CONFIG_PPC */ + +/* + * Wrappers for the rcu_node::lock acquire. + * + * Because the rcu_nodes form a tree, the tree traversal locking will observe + * different lock values, this in turn means that an UNLOCK of one level + * followed by a LOCK of another level does not imply a full memory barrier; + * and most importantly transitivity is lost. + * + * In order to restore full ordering between tree levels, augment the regular + * lock acquire functions with smp_mb__after_unlock_lock(). + */ +static inline void raw_spin_lock_rcu_node(struct rcu_node *rnp) +{ + raw_spin_lock(&rnp->lock); + smp_mb__after_unlock_lock(); +} + +static inline void raw_spin_lock_irq_rcu_node(struct rcu_node *rnp) +{ + raw_spin_lock_irq(&rnp->lock); + smp_mb__after_unlock_lock(); +} + +#define raw_spin_lock_irqsave_rcu_node(rnp, flags) \ +do { \ + typecheck(unsigned long, flags); \ + raw_spin_lock_irqsave(&(rnp)->lock, flags); \ + smp_mb__after_unlock_lock(); \ +} while (0) + +static inline bool raw_spin_trylock_rcu_node(struct rcu_node *rnp) +{ + bool locked = raw_spin_trylock(&rnp->lock); + + if (locked) + smp_mb__after_unlock_lock(); + return locked; +} diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index 630c19772630..fa0e3b96a9ed 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -301,8 +301,7 @@ static void rcu_preempt_note_context_switch(void) /* Possibly blocking in an RCU read-side critical section. */ rdp = this_cpu_ptr(rcu_state_p->rda); rnp = rdp->mynode; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); t->rcu_read_unlock_special.b.blocked = true; t->rcu_blocked_node = rnp; @@ -457,8 +456,7 @@ void rcu_read_unlock_special(struct task_struct *t) */ for (;;) { rnp = t->rcu_blocked_node; - raw_spin_lock(&rnp->lock); /* irqs already disabled. */ - smp_mb__after_unlock_lock(); + raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ if (rnp == t->rcu_blocked_node) break; WARN_ON_ONCE(1); @@ -989,8 +987,7 @@ static int rcu_boost(struct rcu_node *rnp) READ_ONCE(rnp->boost_tasks) == NULL) return 0; /* Nothing left to boost. */ - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); /* * Recheck under the lock: all tasks in need of boosting @@ -1176,8 +1173,7 @@ static int rcu_spawn_one_boost_kthread(struct rcu_state *rsp, "rcub/%d", rnp_index); if (IS_ERR(t)) return PTR_ERR(t); - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); rnp->boost_kthread_task = t; raw_spin_unlock_irqrestore(&rnp->lock, flags); sp.sched_priority = kthread_prio; @@ -1567,8 +1563,7 @@ static void rcu_prepare_for_idle(void) if (!*rdp->nxttail[RCU_DONE_TAIL]) continue; rnp = rdp->mynode; - raw_spin_lock(&rnp->lock); /* irqs already disabled. */ - smp_mb__after_unlock_lock(); + raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ needwake = rcu_accelerate_cbs(rsp, rnp, rdp); raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */ if (needwake) @@ -2068,8 +2063,7 @@ static void rcu_nocb_wait_gp(struct rcu_data *rdp) bool needwake; struct rcu_node *rnp = rdp->mynode; - raw_spin_lock_irqsave(&rnp->lock, flags); - smp_mb__after_unlock_lock(); + raw_spin_lock_irqsave_rcu_node(rnp, flags); needwake = rcu_start_future_gp(rnp, rdp, &c); raw_spin_unlock_irqrestore(&rnp->lock, flags); if (needwake) -- cgit From 6cf10081220ae21175a867d446b3167bcbcb937b Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Thu, 8 Oct 2015 15:36:54 -0700 Subject: rcu: Add transitivity to remaining rcu_node ->lock acquisitions The rule is that all acquisitions of the rcu_node structure's ->lock must provide transitivity: The lock is not acquired that frequently, and sorting out exactly which required it and which did not would be a maintenance nightmare. This commit therefore supplies the needed transitivity to the remaining ->lock acquisitions. Reported-by: Peter Zijlstra Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 24 ++++++++++++------------ kernel/rcu/tree_plugin.h | 2 +- kernel/rcu/tree_trace.c | 2 +- 3 files changed, 14 insertions(+), 14 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index daf17e248757..81aa1cdc6bc9 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -1214,7 +1214,7 @@ static void rcu_dump_cpu_stacks(struct rcu_state *rsp) struct rcu_node *rnp; rcu_for_each_leaf_node(rsp, rnp) { - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (rnp->qsmask != 0) { for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++) if (rnp->qsmask & (1UL << cpu)) @@ -1237,7 +1237,7 @@ static void print_other_cpu_stall(struct rcu_state *rsp, unsigned long gpnum) /* Only let one CPU complain about others per time interval. */ - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); delta = jiffies - READ_ONCE(rsp->jiffies_stall); if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) { raw_spin_unlock_irqrestore(&rnp->lock, flags); @@ -1256,7 +1256,7 @@ static void print_other_cpu_stall(struct rcu_state *rsp, unsigned long gpnum) rsp->name); print_cpu_stall_info_begin(); rcu_for_each_leaf_node(rsp, rnp) { - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); ndetected += rcu_print_task_stall(rnp); if (rnp->qsmask != 0) { for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++) @@ -1327,7 +1327,7 @@ static void print_cpu_stall(struct rcu_state *rsp) rcu_dump_cpu_stacks(rsp); - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (ULONG_CMP_GE(jiffies, READ_ONCE(rsp->jiffies_stall))) WRITE_ONCE(rsp->jiffies_stall, jiffies + 3 * rcu_jiffies_till_stall_check() + 3); @@ -2897,7 +2897,7 @@ __rcu_process_callbacks(struct rcu_state *rsp) /* Does this CPU require a not-yet-started grace period? */ local_irq_save(flags); if (cpu_needs_another_gp(rsp, rdp)) { - raw_spin_lock(&rcu_get_root(rsp)->lock); /* irqs disabled. */ + raw_spin_lock_rcu_node(rcu_get_root(rsp)); /* irqs disabled. */ needwake = rcu_start_gp(rsp); raw_spin_unlock_irqrestore(&rcu_get_root(rsp)->lock, flags); if (needwake) @@ -3718,7 +3718,7 @@ retry_ipi: mask_ofl_ipi &= ~mask; } else { /* Failed, raced with offline. */ - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (cpu_online(cpu) && (rnp->expmask & mask)) { raw_spin_unlock_irqrestore(&rnp->lock, @@ -3727,8 +3727,8 @@ retry_ipi: if (cpu_online(cpu) && (rnp->expmask & mask)) goto retry_ipi; - raw_spin_lock_irqsave(&rnp->lock, - flags); + raw_spin_lock_irqsave_rcu_node(rnp, + flags); } if (!(rnp->expmask & mask)) mask_ofl_ipi &= ~mask; @@ -4110,7 +4110,7 @@ static void rcu_init_new_rnp(struct rcu_node *rnp_leaf) rnp = rnp->parent; if (rnp == NULL) return; - raw_spin_lock(&rnp->lock); /* Interrupts already disabled. */ + raw_spin_lock_rcu_node(rnp); /* Interrupts already disabled. */ rnp->qsmaskinit |= mask; raw_spin_unlock(&rnp->lock); /* Interrupts remain disabled. */ } @@ -4127,7 +4127,7 @@ rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp) struct rcu_node *rnp = rcu_get_root(rsp); /* Set up local state, ensuring consistent view of global state. */ - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo); rdp->dynticks = &per_cpu(rcu_dynticks, cpu); WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE); @@ -4154,7 +4154,7 @@ rcu_init_percpu_data(int cpu, struct rcu_state *rsp) struct rcu_node *rnp = rcu_get_root(rsp); /* Set up local state, ensuring consistent view of global state. */ - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); rdp->qlen_last_fqs_check = 0; rdp->n_force_qs_snap = rsp->n_force_qs; rdp->blimit = blimit; @@ -4301,7 +4301,7 @@ static int __init rcu_spawn_gp_kthread(void) t = kthread_create(rcu_gp_kthread, rsp, "%s", rsp->name); BUG_ON(IS_ERR(t)); rnp = rcu_get_root(rsp); - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); rsp->gp_kthread = t; if (kthread_prio) { sp.sched_priority = kthread_prio; diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index fa0e3b96a9ed..57ba873d2f18 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -525,7 +525,7 @@ static void rcu_print_detail_task_stall_rnp(struct rcu_node *rnp) unsigned long flags; struct task_struct *t; - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); if (!rcu_preempt_blocked_readers_cgp(rnp)) { raw_spin_unlock_irqrestore(&rnp->lock, flags); return; diff --git a/kernel/rcu/tree_trace.c b/kernel/rcu/tree_trace.c index ef7093cc9b5c..8efaba870d96 100644 --- a/kernel/rcu/tree_trace.c +++ b/kernel/rcu/tree_trace.c @@ -319,7 +319,7 @@ static void show_one_rcugp(struct seq_file *m, struct rcu_state *rsp) unsigned long gpmax; struct rcu_node *rnp = &rsp->node[0]; - raw_spin_lock_irqsave(&rnp->lock, flags); + raw_spin_lock_irqsave_rcu_node(rnp, flags); completed = READ_ONCE(rsp->completed); gpnum = READ_ONCE(rsp->gpnum); if (completed == gpnum) -- cgit From ed11a7f1b3bd482bd7d6ef7bc2859c41fb43b9ee Mon Sep 17 00:00:00 2001 From: Andy Lutomirski Date: Thu, 12 Nov 2015 12:59:01 -0800 Subject: context_tracking: Switch to new static_branch API This is much less error-prone than the old code. Signed-off-by: Andy Lutomirski Cc: Andy Lutomirski Cc: Borislav Petkov Cc: Brian Gerst Cc: Denys Vlasenko Cc: Frederic Weisbecker Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/812df7e64f120c5c7c08481f36a8caa9f53b2199.1447361906.git.luto@kernel.org Signed-off-by: Ingo Molnar --- kernel/context_tracking.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c index d8560ee3bab7..9ad37b9e44a7 100644 --- a/kernel/context_tracking.c +++ b/kernel/context_tracking.c @@ -24,7 +24,7 @@ #define CREATE_TRACE_POINTS #include -struct static_key context_tracking_enabled = STATIC_KEY_INIT_FALSE; +DEFINE_STATIC_KEY_FALSE(context_tracking_enabled); EXPORT_SYMBOL_GPL(context_tracking_enabled); DEFINE_PER_CPU(struct context_tracking, context_tracking); @@ -191,7 +191,7 @@ void __init context_tracking_cpu_set(int cpu) if (!per_cpu(context_tracking.active, cpu)) { per_cpu(context_tracking.active, cpu) = true; - static_key_slow_inc(&context_tracking_enabled); + static_branch_inc(&context_tracking_enabled); } if (initialized) -- cgit From 8573636ea794fa088f459429e65e47d7776532cf Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Tue, 17 Nov 2015 14:03:11 -0500 Subject: ring-buffer: Use READ_ONCE() for most tail_page access As cpu_buffer->tail_page may be modified by interrupts at almost any time, the flow of logic is very important. Do not let gcc get smart with re-reading cpu_buffer->tail_page by adding READ_ONCE() around most of its accesses. Signed-off-by: Steven Rostedt --- kernel/trace/ring_buffer.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 9c6045a27ba3..ab102e6259bc 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -1036,7 +1036,7 @@ static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer, * it is, then it is up to us to update the tail * pointer. */ - if (tail_page == cpu_buffer->tail_page) { + if (tail_page == READ_ONCE(cpu_buffer->tail_page)) { /* Zero the write counter */ unsigned long val = old_write & ~RB_WRITE_MASK; unsigned long eval = old_entries & ~RB_WRITE_MASK; @@ -2036,12 +2036,15 @@ rb_handle_head_page(struct ring_buffer_per_cpu *cpu_buffer, * the tail page would have moved. */ if (ret == RB_PAGE_NORMAL) { + struct buffer_page *buffer_tail_page; + + buffer_tail_page = READ_ONCE(cpu_buffer->tail_page); /* * If the tail had moved passed next, then we need * to reset the pointer. */ - if (cpu_buffer->tail_page != tail_page && - cpu_buffer->tail_page != next_page) + if (buffer_tail_page != tail_page && + buffer_tail_page != next_page) rb_head_page_set_normal(cpu_buffer, new_head, next_page, RB_PAGE_HEAD); @@ -2362,7 +2365,7 @@ rb_try_to_discard(struct ring_buffer_per_cpu *cpu_buffer, addr = (unsigned long)event; addr &= PAGE_MASK; - bpage = cpu_buffer->tail_page; + bpage = READ_ONCE(cpu_buffer->tail_page); if (bpage->page == (void *)addr && rb_page_write(bpage) == old_index) { unsigned long write_mask = @@ -2410,7 +2413,7 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer) again: max_count = cpu_buffer->nr_pages * 100; - while (cpu_buffer->commit_page != cpu_buffer->tail_page) { + while (cpu_buffer->commit_page != READ_ONCE(cpu_buffer->tail_page)) { if (RB_WARN_ON(cpu_buffer, !(--max_count))) return; if (RB_WARN_ON(cpu_buffer, @@ -2443,7 +2446,7 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer) * and pushed the tail page forward, we will be left with * a dangling commit that will never go forward. */ - if (unlikely(cpu_buffer->commit_page != cpu_buffer->tail_page)) + if (unlikely(cpu_buffer->commit_page != READ_ONCE(cpu_buffer->tail_page))) goto again; } @@ -2699,7 +2702,8 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer, if (unlikely(info->add_timestamp)) info->length += RB_LEN_TIME_EXTEND; - tail_page = info->tail_page = cpu_buffer->tail_page; + /* Don't let the compiler play games with cpu_buffer->tail_page */ + tail_page = info->tail_page = READ_ONCE(cpu_buffer->tail_page); write = local_add_return(info->length, &tail_page->write); /* set write to only the index of the write */ -- cgit From 70004986ffdf36d8bc787403af2571aeeef96595 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Tue, 17 Nov 2015 15:15:19 -0500 Subject: ring-buffer: Remove redundant update of page timestamp The first commit of a buffer page updates the timestamp of that page. No need to have the update to the next page add the timestamp too. It will only be replaced by the first commit on that page anyway. Only update to a page if it contains an event. Signed-off-by: Steven Rostedt --- kernel/trace/ring_buffer.c | 32 ++++++++------------------------ 1 file changed, 8 insertions(+), 24 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index ab102e6259bc..631541a53baf 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -1001,17 +1001,13 @@ static int rb_head_page_replace(struct buffer_page *old, /* * rb_tail_page_update - move the tail page forward - * - * Returns 1 if moved tail page, 0 if someone else did. */ -static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer, +static void rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer, struct buffer_page *tail_page, struct buffer_page *next_page) { - struct buffer_page *old_tail; unsigned long old_entries; unsigned long old_write; - int ret = 0; /* * The tail page now needs to be moved forward. @@ -1061,14 +1057,9 @@ static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer, */ local_set(&next_page->page->commit, 0); - old_tail = cmpxchg(&cpu_buffer->tail_page, - tail_page, next_page); - - if (old_tail == tail_page) - ret = 1; + /* Again, either we update tail_page or an interrupt does */ + (void)cmpxchg(&cpu_buffer->tail_page, tail_page, next_page); } - - return ret; } static int rb_check_bpage(struct ring_buffer_per_cpu *cpu_buffer, @@ -2150,7 +2141,6 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer, struct ring_buffer *buffer = cpu_buffer->buffer; struct buffer_page *next_page; int ret; - u64 ts; next_page = tail_page; @@ -2224,15 +2214,7 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer, } } - ret = rb_tail_page_update(cpu_buffer, tail_page, next_page); - if (ret) { - /* - * Nested commits always have zero deltas, so - * just reread the time stamp - */ - ts = rb_time_stamp(buffer); - next_page->page->time_stamp = ts; - } + rb_tail_page_update(cpu_buffer, tail_page, next_page); out_again: @@ -2422,8 +2404,10 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer) local_set(&cpu_buffer->commit_page->page->commit, rb_page_write(cpu_buffer->commit_page)); rb_inc_page(cpu_buffer, &cpu_buffer->commit_page); - cpu_buffer->write_stamp = - cpu_buffer->commit_page->page->time_stamp; + /* Only update the write stamp if the page has an event */ + if (rb_page_write(cpu_buffer->commit_page)) + cpu_buffer->write_stamp = + cpu_buffer->commit_page->page->time_stamp; /* add barrier to keep gcc from optimizing too much */ barrier(); } -- cgit From d2b4365809060b256330a99289de9797a5dd6967 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Wed, 25 Nov 2015 16:16:55 +0100 Subject: cpuset: Replace all instances of time_t with time64_t The following patch replaces all instances of time_t with time64_t i.e. change the type used for representing time from 32-bit to 64-bit. All 32-bit kernels to date use a signed 32-bit time_t type, which can only represent time until January 2038. Since embedded systems running 32-bit Linux are going to survive beyond that date, we have to change all current uses, in a backwards compatible way. The patch also changes the function get_seconds() that returns a 32-bit integer to ktime_get_seconds() that returns seconds as 64-bit integer. The patch changes the type of ticks from time_t to u32. We keep ticks as 32-bits as the function uses 32-bit arithmetic which would prove less expensive than 64-bit arithmetic and the function is expected to be called atleast once every 32 seconds. Signed-off-by: Heena Sirwani Reviewed-by: Arnd Bergmann Signed-off-by: Arnd Bergmann Signed-off-by: Tejun Heo --- kernel/cpuset.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/cpuset.c b/kernel/cpuset.c index 10ae73611d80..c9ea63ff70a7 100644 --- a/kernel/cpuset.c +++ b/kernel/cpuset.c @@ -51,6 +51,7 @@ #include #include #include +#include #include #include @@ -68,7 +69,7 @@ struct static_key cpusets_enabled_key __read_mostly = STATIC_KEY_INIT_FALSE; struct fmeter { int cnt; /* unprocessed events count */ int val; /* most recent output value */ - time_t time; /* clock (secs) when val computed */ + time64_t time; /* clock (secs) when val computed */ spinlock_t lock; /* guards read or write of above */ }; @@ -1374,7 +1375,7 @@ out: */ #define FM_COEF 933 /* coefficient for half-life of 10 secs */ -#define FM_MAXTICKS ((time_t)99) /* useless computing more ticks than this */ +#define FM_MAXTICKS ((u32)99) /* useless computing more ticks than this */ #define FM_MAXCNT 1000000 /* limit cnt to avoid overflow */ #define FM_SCALE 1000 /* faux fixed point scale */ @@ -1390,8 +1391,11 @@ static void fmeter_init(struct fmeter *fmp) /* Internal meter update - process cnt events and update value */ static void fmeter_update(struct fmeter *fmp) { - time_t now = get_seconds(); - time_t ticks = now - fmp->time; + time64_t now; + u32 ticks; + + now = ktime_get_seconds(); + ticks = now - fmp->time; if (ticks == 0) return; -- cgit From 4239c38fe0b3847e1e6d962c74b41b08ba0e2990 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Tue, 17 Nov 2015 16:36:06 -0500 Subject: ring-buffer: Process commits whenever moving to a new page. When crossing over to a new page, commit the current work. This will allow readers to get data with less latency, and also simplifies the work to get timestamps working for interrupted events. Signed-off-by: Steven Rostedt --- kernel/trace/ring_buffer.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 631541a53baf..95181e36891a 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -2129,6 +2129,8 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer, local_sub(length, &tail_page->write); } +static inline void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer); + /* * This is the slow path, force gcc not to inline it. */ @@ -2220,6 +2222,11 @@ rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer, rb_reset_tail(cpu_buffer, tail, info); + /* Commit what we have for now. */ + rb_end_commit(cpu_buffer); + /* rb_end_commit() decs committing */ + local_inc(&cpu_buffer->committing); + /* fail and let the caller try again */ return ERR_PTR(-EAGAIN); -- cgit From 4e4a4d75700da67c1209c151d5f11a309c3b8367 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Mon, 23 Nov 2015 13:51:16 -0600 Subject: tracing: Update cond flag when enabling or disabling a trigger When a trigger is enabled, the cond flag should be set beforehand, otherwise a trigger that's expecting to process a trace record (e.g. one with post_trigger set) could be invoked without one. Likewise a trigger's cond flag should be reset after it's disabled, not before. Link: http://lkml.kernel.org/r/a420b52a67b1c2d3cab017914362d153255acb99.1448303214.git.tom.zanussi@linux.intel.com Signed-off-by: Tom Zanussi Signed-off-by: Daniel Wagner Reviewed-by: Masami Hiramatsu Reviewed-by: Namhyung Kim Tested-by: Masami Hiramatsu Signed-off-by: Steven Rostedt --- kernel/trace/trace_events_trigger.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events_trigger.c b/kernel/trace/trace_events_trigger.c index 42a4009fd75a..4d2f3ccc56f6 100644 --- a/kernel/trace/trace_events_trigger.c +++ b/kernel/trace/trace_events_trigger.c @@ -543,11 +543,12 @@ static int register_trigger(char *glob, struct event_trigger_ops *ops, list_add_rcu(&data->list, &file->triggers); ret++; + update_cond_flag(file); if (trace_event_trigger_enable_disable(file, 1) < 0) { list_del_rcu(&data->list); + update_cond_flag(file); ret--; } - update_cond_flag(file); out: return ret; } @@ -575,8 +576,8 @@ static void unregister_trigger(char *glob, struct event_trigger_ops *ops, if (data->cmd_ops->trigger_type == test->cmd_ops->trigger_type) { unregistered = true; list_del_rcu(&data->list); - update_cond_flag(file); trace_event_trigger_enable_disable(file, 0); + update_cond_flag(file); break; } } @@ -1319,11 +1320,12 @@ static int event_enable_register_trigger(char *glob, list_add_rcu(&data->list, &file->triggers); ret++; + update_cond_flag(file); if (trace_event_trigger_enable_disable(file, 1) < 0) { list_del_rcu(&data->list); + update_cond_flag(file); ret--; } - update_cond_flag(file); out: return ret; } @@ -1344,8 +1346,8 @@ static void event_enable_unregister_trigger(char *glob, (enable_data->file == test_enable_data->file)) { unregistered = true; list_del_rcu(&data->list); - update_cond_flag(file); trace_event_trigger_enable_disable(file, 0); + update_cond_flag(file); break; } } -- cgit From 02a392a0439ffdc62b4d8f17bd18d68736b166a9 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Wed, 25 Nov 2015 12:50:47 -0500 Subject: ftrace: Add new type to distinguish what kind of ftrace_bug() The ftrace function hook utility has several internal checks to make sure that whatever it modifies is exactly what it expects to be modifying. This is essential as modifying running code can be extremely dangerous to the system. When an anomaly is detected, ftrace_bug() is called which sends a splat to the console and disables function tracing. There's some extra information that is printed to help diagnose the issue. One thing that is missing though is output of what ftrace was doing at the time of the crash. Was it updating a call site or perhaps converting a call site to a nop? A new global enum variable is created to state what ftrace was doing at the time of the anomaly, and this is reported in ftrace_bug(). Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 37 ++++++++++++++++++++++++++++++++++++- 1 file changed, 36 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 3f743b147247..b95efcac9dfe 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1953,6 +1953,28 @@ static void print_ip_ins(const char *fmt, unsigned char *p) static struct ftrace_ops * ftrace_find_tramp_ops_any(struct dyn_ftrace *rec); +enum ftrace_bug_type ftrace_bug_type; + +static void print_bug_type(void) +{ + switch (ftrace_bug_type) { + case FTRACE_BUG_UNKNOWN: + break; + case FTRACE_BUG_INIT: + pr_info("Initializing ftrace call sites\n"); + break; + case FTRACE_BUG_NOP: + pr_info("Setting ftrace call site to NOP\n"); + break; + case FTRACE_BUG_CALL: + pr_info("Setting ftrace call site to call ftrace function\n"); + break; + case FTRACE_BUG_UPDATE: + pr_info("Updating ftrace call site to call a different ftrace function\n"); + break; + } +} + /** * ftrace_bug - report and shutdown function tracer * @failed: The failed type (EFAULT, EINVAL, EPERM) @@ -1992,6 +2014,7 @@ void ftrace_bug(int failed, struct dyn_ftrace *rec) pr_info("ftrace faulted on unknown error "); print_ip_sym(ip); } + print_bug_type(); if (rec) { struct ftrace_ops *ops = NULL; @@ -2016,6 +2039,8 @@ static int ftrace_check_record(struct dyn_ftrace *rec, int enable, int update) { unsigned long flag = 0UL; + ftrace_bug_type = FTRACE_BUG_UNKNOWN; + /* * If we are updating calls: * @@ -2077,9 +2102,12 @@ static int ftrace_check_record(struct dyn_ftrace *rec, int enable, int update) * from the save regs, to a non-save regs function or * vice versa, or from a trampoline call. */ - if (flag & FTRACE_FL_ENABLED) + if (flag & FTRACE_FL_ENABLED) { + ftrace_bug_type = FTRACE_BUG_CALL; return FTRACE_UPDATE_MAKE_CALL; + } + ftrace_bug_type = FTRACE_BUG_UPDATE; return FTRACE_UPDATE_MODIFY_CALL; } @@ -2096,6 +2124,7 @@ static int ftrace_check_record(struct dyn_ftrace *rec, int enable, int update) FTRACE_FL_REGS_EN); } + ftrace_bug_type = FTRACE_BUG_NOP; return FTRACE_UPDATE_MAKE_NOP; } @@ -2307,17 +2336,22 @@ __ftrace_replace_code(struct dyn_ftrace *rec, int enable) ret = ftrace_update_record(rec, enable); + ftrace_bug_type = FTRACE_BUG_UNKNOWN; + switch (ret) { case FTRACE_UPDATE_IGNORE: return 0; case FTRACE_UPDATE_MAKE_CALL: + ftrace_bug_type = FTRACE_BUG_CALL; return ftrace_make_call(rec, ftrace_addr); case FTRACE_UPDATE_MAKE_NOP: + ftrace_bug_type = FTRACE_BUG_NOP; return ftrace_make_nop(NULL, rec, ftrace_old_addr); case FTRACE_UPDATE_MODIFY_CALL: + ftrace_bug_type = FTRACE_BUG_UPDATE; return ftrace_modify_call(rec, ftrace_old_addr, ftrace_addr); } @@ -2425,6 +2459,7 @@ ftrace_code_disable(struct module *mod, struct dyn_ftrace *rec) ret = ftrace_make_nop(mod, rec, MCOUNT_ADDR); if (ret) { + ftrace_bug_type = FTRACE_BUG_INIT; ftrace_bug(ret, rec); return 0; } -- cgit From b05086c77a162dd8ef79606cb4723f1fc1448bb1 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Wed, 25 Nov 2015 14:13:11 -0500 Subject: ftrace: Add variable ftrace_expected for archs to show expected code When an anomaly is found while modifying function code, ftrace_bug() is called which disables the function tracing infrastructure and reports information about what failed. If the code that is to be replaced does not match what is expected, then actual code is shown. Currently there is no arch generic way to show what was expected. Add a new variable pointer calld ftrace_expected that the arch code can set to point to what it expected so that ftrace_bug() can report the actual text as well as the text that was expected to be there. Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index b95efcac9dfe..7870c03b4c4d 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1940,7 +1940,7 @@ static int ftrace_hash_ipmodify_update(struct ftrace_ops *ops, return __ftrace_hash_update_ipmodify(ops, old_hash, new_hash); } -static void print_ip_ins(const char *fmt, unsigned char *p) +static void print_ip_ins(const char *fmt, const unsigned char *p) { int i; @@ -1954,6 +1954,7 @@ static struct ftrace_ops * ftrace_find_tramp_ops_any(struct dyn_ftrace *rec); enum ftrace_bug_type ftrace_bug_type; +const void *ftrace_expected; static void print_bug_type(void) { @@ -2001,8 +2002,12 @@ void ftrace_bug(int failed, struct dyn_ftrace *rec) FTRACE_WARN_ON_ONCE(1); pr_info("ftrace failed to modify "); print_ip_sym(ip); - print_ip_ins(" actual: ", (unsigned char *)ip); + print_ip_ins(" actual: ", (unsigned char *)ip); pr_cont("\n"); + if (ftrace_expected) { + print_ip_ins(" expected: ", ftrace_expected); + pr_cont("\n"); + } break; case -EPERM: FTRACE_WARN_ON_ONCE(1); -- cgit From 39daa7b9e89512f234b7fb5d55812a78318251fc Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Wed, 25 Nov 2015 15:12:38 -0500 Subject: ftrace: Show all tramps registered to a record on ftrace_bug() When an anomaly is detected in the function call modification code, ftrace_bug() is called to disable function tracing as well as give any information that may help debug the problem. Currently, only the first found trampoline that is attached to the failed record is reported. Instead, show all trampolines that are hooked to it. Also, not only show the ops pointer but also report the function it calls. While at it, add this info to the enabled_functions debug file too. Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 46 +++++++++++++++++++++++++++++++++++++--------- 1 file changed, 37 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 7870c03b4c4d..e371aed51fcf 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1952,6 +1952,8 @@ static void print_ip_ins(const char *fmt, const unsigned char *p) static struct ftrace_ops * ftrace_find_tramp_ops_any(struct dyn_ftrace *rec); +static struct ftrace_ops * +ftrace_find_tramp_ops_next(struct dyn_ftrace *rec, struct ftrace_ops *ops); enum ftrace_bug_type ftrace_bug_type; const void *ftrace_expected; @@ -2028,15 +2030,19 @@ void ftrace_bug(int failed, struct dyn_ftrace *rec) rec->flags & FTRACE_FL_REGS ? " R" : " "); if (rec->flags & FTRACE_FL_TRAMP_EN) { ops = ftrace_find_tramp_ops_any(rec); - if (ops) - pr_cont("\ttramp: %pS", - (void *)ops->trampoline); - else + if (ops) { + do { + pr_cont("\ttramp: %pS (%pS)", + (void *)ops->trampoline, + (void *)ops->func); + ops = ftrace_find_tramp_ops_next(rec, ops); + } while (ops); + } else pr_cont("\ttramp: ERROR!"); } ip = ftrace_get_addr_curr(rec); - pr_cont(" expected tramp: %lx\n", ip); + pr_cont("\n expected tramp: %lx\n", ip); } } @@ -2178,6 +2184,24 @@ ftrace_find_tramp_ops_any(struct dyn_ftrace *rec) return NULL; } +static struct ftrace_ops * +ftrace_find_tramp_ops_next(struct dyn_ftrace *rec, + struct ftrace_ops *op) +{ + unsigned long ip = rec->ip; + + while_for_each_ftrace_op(op) { + + if (!op->trampoline) + continue; + + if (hash_contains_ip(ip, op->func_hash)) + return op; + } + + return NULL; +} + static struct ftrace_ops * ftrace_find_tramp_ops_curr(struct dyn_ftrace *rec) { @@ -3306,10 +3330,14 @@ static int t_show(struct seq_file *m, void *v) rec->flags & FTRACE_FL_IPMODIFY ? " I" : " "); if (rec->flags & FTRACE_FL_TRAMP_EN) { ops = ftrace_find_tramp_ops_any(rec); - if (ops) - seq_printf(m, "\ttramp: %pS", - (void *)ops->trampoline); - else + if (ops) { + do { + seq_printf(m, "\ttramp: %pS (%pS)", + (void *)ops->trampoline, + (void *)ops->func); + ops = ftrace_find_tramp_ops_next(rec, ops); + } while (ops); + } else seq_puts(m, "\ttramp: ERROR!"); } -- cgit From 82bbe34b3d895fb026b2fc0e7da2e641797bfaed Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 19 Nov 2015 17:21:06 +0100 Subject: nohz: Clarify magic in tick_nohz_stop_sched_tick() While going through the nohz code I got stumped by some of it. This patch adds a few comments clarifying the code; based on discussion with Thomas. Signed-off-by: Peter Zijlstra (Intel) Cc: Frederic Weisbecker Link: http://lkml.kernel.org/r/20151119162106.GO3816@twins.programming.kicks-ass.net Signed-off-by: Thomas Gleixner --- kernel/time/tick-sched.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 7c7ec4515983..7ea28ed3109d 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -603,15 +603,31 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts, /* * If the tick is due in the next period, keep it ticking or - * restart it proper. + * force prod the timer. */ delta = next_tick - basemono; if (delta <= (u64)TICK_NSEC) { tick.tv64 = 0; + /* + * We've not stopped the tick yet, and there's a timer in the + * next period, so no point in stopping it either, bail. + */ if (!ts->tick_stopped) goto out; + + /* + * If, OTOH, we did stop it, but there's a pending (expired) + * timer reprogram the timer hardware to fire now. + * + * We will not restart the tick proper, just prod the timer + * hardware into firing an interrupt to process the pending + * timers. Just like tick_irq_exit() will not restart the tick + * for 'normal' interrupts. + * + * Only once we exit the idle loop will we re-enable the tick, + * see tick_nohz_idle_exit(). + */ if (delta == 0) { - /* Tick is stopped, but required now. Enforce it */ tick_nohz_restart(ts, now); goto out; } -- cgit From 29732938a6289a15e907da234d6692a2ead71855 Mon Sep 17 00:00:00 2001 From: Zach Brown Date: Tue, 10 Nov 2015 16:53:30 -0500 Subject: vfs: add copy_file_range syscall and vfs helper Add a copy_file_range() system call for offloading copies between regular files. This gives an interface to underlying layers of the storage stack which can copy without reading and writing all the data. There are a few candidates that should support copy offloading in the nearer term: - btrfs shares extent references with its clone ioctl - NFS has patches to add a COPY command which copies on the server - SCSI has a family of XCOPY commands which copy in the device This system call avoids the complexity of also accelerating the creation of the destination file by operating on an existing destination file descriptor, not a path. Currently the high level vfs entry point limits copy offloading to files on the same mount and super (and not in the same file). This can be relaxed if we get implementations which can copy between file systems safely. Signed-off-by: Zach Brown [Anna Schumaker: Change -EINVAL to -EBADF during file verification, Change flags parameter from int to unsigned int, Add function to include/linux/syscalls.h, Check copy len after file open mode, Don't forbid ranges inside the same file, Use rw_verify_area() to veriy ranges, Use file_out rather than file_in, Add COPY_FR_REFLINK flag] Signed-off-by: Anna Schumaker Reviewed-by: Christoph Hellwig Signed-off-by: Al Viro --- kernel/sys_ni.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c index 0623787ec67a..2c5e3a8e00d7 100644 --- a/kernel/sys_ni.c +++ b/kernel/sys_ni.c @@ -174,6 +174,7 @@ cond_syscall(sys_setfsuid); cond_syscall(sys_setfsgid); cond_syscall(sys_capget); cond_syscall(sys_capset); +cond_syscall(sys_copy_file_range); /* arch-specific weak syscall entries */ cond_syscall(sys_pciconfig_read); -- cgit From b53202e6308939d33ba0c78712e850f891b4e76f Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Thu, 3 Dec 2015 10:24:08 -0500 Subject: cgroup: kill cgrp_ss_priv[CGROUP_CANFORK_COUNT] and friends Now that nobody use the "priv" arg passed to can_fork/cancel_fork/fork we can kill CGROUP_CANFORK_COUNT/SUBSYS_TAG/etc and cgrp_ss_priv[] in copy_process(). Signed-off-by: Oleg Nesterov Signed-off-by: Tejun Heo --- kernel/cgroup.c | 30 +++++++----------------------- kernel/cgroup_freezer.c | 2 +- kernel/cgroup_pids.c | 4 ++-- kernel/fork.c | 7 +++---- kernel/sched/core.c | 2 +- 5 files changed, 14 insertions(+), 31 deletions(-) (limited to 'kernel') diff --git a/kernel/cgroup.c b/kernel/cgroup.c index ad35ac033d9b..7f2f007397fe 100644 --- a/kernel/cgroup.c +++ b/kernel/cgroup.c @@ -5488,19 +5488,6 @@ static const struct file_operations proc_cgroupstats_operations = { .release = single_release, }; -static void **subsys_canfork_priv_p(void *ss_priv[CGROUP_CANFORK_COUNT], int i) -{ - if (CGROUP_CANFORK_START <= i && i < CGROUP_CANFORK_END) - return &ss_priv[i - CGROUP_CANFORK_START]; - return NULL; -} - -static void *subsys_canfork_priv(void *ss_priv[CGROUP_CANFORK_COUNT], int i) -{ - void **private = subsys_canfork_priv_p(ss_priv, i); - return private ? *private : NULL; -} - /** * cgroup_fork - initialize cgroup related fields during copy_process() * @child: pointer to task_struct of forking parent process. @@ -5523,14 +5510,13 @@ void cgroup_fork(struct task_struct *child) * returns an error, the fork aborts with that error code. This allows for * a cgroup subsystem to conditionally allow or deny new forks. */ -int cgroup_can_fork(struct task_struct *child, - void *ss_priv[CGROUP_CANFORK_COUNT]) +int cgroup_can_fork(struct task_struct *child) { struct cgroup_subsys *ss; int i, j, ret; for_each_subsys_which(ss, i, &have_canfork_callback) { - ret = ss->can_fork(child, subsys_canfork_priv_p(ss_priv, i)); + ret = ss->can_fork(child); if (ret) goto out_revert; } @@ -5542,7 +5528,7 @@ out_revert: if (j >= i) break; if (ss->cancel_fork) - ss->cancel_fork(child, subsys_canfork_priv(ss_priv, j)); + ss->cancel_fork(child); } return ret; @@ -5555,15 +5541,14 @@ out_revert: * This calls the cancel_fork() callbacks if a fork failed *after* * cgroup_can_fork() succeded. */ -void cgroup_cancel_fork(struct task_struct *child, - void *ss_priv[CGROUP_CANFORK_COUNT]) +void cgroup_cancel_fork(struct task_struct *child) { struct cgroup_subsys *ss; int i; for_each_subsys(ss, i) if (ss->cancel_fork) - ss->cancel_fork(child, subsys_canfork_priv(ss_priv, i)); + ss->cancel_fork(child); } /** @@ -5576,8 +5561,7 @@ void cgroup_cancel_fork(struct task_struct *child, * cgroup_task_iter_start() - to guarantee that the new task ends up on its * list. */ -void cgroup_post_fork(struct task_struct *child, - void *old_ss_priv[CGROUP_CANFORK_COUNT]) +void cgroup_post_fork(struct task_struct *child) { struct cgroup_subsys *ss; int i; @@ -5621,7 +5605,7 @@ void cgroup_post_fork(struct task_struct *child, * and addition to css_set. */ for_each_subsys_which(ss, i, &have_fork_callback) - ss->fork(child, subsys_canfork_priv(old_ss_priv, i)); + ss->fork(child); } /** diff --git a/kernel/cgroup_freezer.c b/kernel/cgroup_freezer.c index 2d3df82c54f2..1b72d56edce5 100644 --- a/kernel/cgroup_freezer.c +++ b/kernel/cgroup_freezer.c @@ -200,7 +200,7 @@ static void freezer_attach(struct cgroup_taskset *tset) * to do anything as freezer_attach() will put @task into the appropriate * state. */ -static void freezer_fork(struct task_struct *task, void *private) +static void freezer_fork(struct task_struct *task) { struct freezer *freezer; diff --git a/kernel/cgroup_pids.c b/kernel/cgroup_pids.c index b50d5a167fda..18107aea2895 100644 --- a/kernel/cgroup_pids.c +++ b/kernel/cgroup_pids.c @@ -209,7 +209,7 @@ static void pids_cancel_attach(struct cgroup_taskset *tset) * task_css_check(true) in pids_can_fork() and pids_cancel_fork() relies * on threadgroup_change_begin() held by the copy_process(). */ -static int pids_can_fork(struct task_struct *task, void **priv_p) +static int pids_can_fork(struct task_struct *task) { struct cgroup_subsys_state *css; struct pids_cgroup *pids; @@ -219,7 +219,7 @@ static int pids_can_fork(struct task_struct *task, void **priv_p) return pids_try_charge(pids, 1); } -static void pids_cancel_fork(struct task_struct *task, void *priv) +static void pids_cancel_fork(struct task_struct *task) { struct cgroup_subsys_state *css; struct pids_cgroup *pids; diff --git a/kernel/fork.c b/kernel/fork.c index fce002ee3ddf..ba7d1c037490 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1249,7 +1249,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, { int retval; struct task_struct *p; - void *cgrp_ss_priv[CGROUP_CANFORK_COUNT] = {}; if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) return ERR_PTR(-EINVAL); @@ -1526,7 +1525,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, * between here and cgroup_post_fork() if an organisation operation is in * progress. */ - retval = cgroup_can_fork(p, cgrp_ss_priv); + retval = cgroup_can_fork(p); if (retval) goto bad_fork_free_pid; @@ -1608,7 +1607,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, write_unlock_irq(&tasklist_lock); proc_fork_connector(p); - cgroup_post_fork(p, cgrp_ss_priv); + cgroup_post_fork(p); threadgroup_change_end(current); perf_event_fork(p); @@ -1618,7 +1617,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, return p; bad_fork_cancel_cgroup: - cgroup_cancel_fork(p, cgrp_ss_priv); + cgroup_cancel_fork(p); bad_fork_free_pid: if (pid != &init_struct_pid) free_pid(pid); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index a9db4819e586..b7d2271cd948 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -8212,7 +8212,7 @@ static void cpu_cgroup_css_offline(struct cgroup_subsys_state *css) sched_offline_group(tg); } -static void cpu_cgroup_fork(struct task_struct *task, void *private) +static void cpu_cgroup_fork(struct task_struct *task) { sched_move_task(task); } -- cgit From a0e3213f836640065e2a23ad55fa6f72e812a4f1 Mon Sep 17 00:00:00 2001 From: zhuo-hao Date: Tue, 17 Nov 2015 20:08:07 +0800 Subject: alarmtimer: Avoid unexpected rtc interrupt when system resume from S3 Before the system go to suspend (S3), if user create a timer with clockid CLOCK_REALTIME_ALARM/CLOCK_BOOTTIME_ALARM and set a "large" timeout value to this timer. The function alarmtimer_suspend will be called to setup a timeout value to RTC timer to avoid the system sleep over time. However, if the system wakeup early than RTC timeout, the RTC timer will not be cleared. And this will cause the hpet_rtc_interrupt come unexpectedly until the RTC timeout. To fix this problem, just adding alarmtimer_resume to cancel the RTC timer. This was noticed because the HPET RTC emulation fires an interrupt every 16ms(=1/2^DEFAULT_RTC_SHIFT) up to the point where the alarm time is reached. This program always hits this situation (https://lkml.org/lkml/2015/11/8/326), if system wake up earlier than alarm time. Cc: Thomas Gleixner Cc: John Stultz Signed-off-by: Zhuo-hao Lee [jstultz: Tweak commit subject & formatting slightly] Signed-off-by: John Stultz --- kernel/time/alarmtimer.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) (limited to 'kernel') diff --git a/kernel/time/alarmtimer.c b/kernel/time/alarmtimer.c index 7fbba635a549..e840ed867a5d 100644 --- a/kernel/time/alarmtimer.c +++ b/kernel/time/alarmtimer.c @@ -271,11 +271,27 @@ static int alarmtimer_suspend(struct device *dev) __pm_wakeup_event(ws, MSEC_PER_SEC); return ret; } + +static int alarmtimer_resume(struct device *dev) +{ + struct rtc_device *rtc; + + rtc = alarmtimer_get_rtcdev(); + if (rtc) + rtc_timer_cancel(rtc, &rtctimer); + return 0; +} + #else static int alarmtimer_suspend(struct device *dev) { return 0; } + +static int alarmtimer_resume(struct device *dev) +{ + return 0; +} #endif static void alarmtimer_freezerset(ktime_t absexp, enum alarmtimer_type type) @@ -800,6 +816,7 @@ out: /* Suspend hook structures */ static const struct dev_pm_ops alarmtimer_pm_ops = { .suspend = alarmtimer_suspend, + .resume = alarmtimer_resume, }; static struct platform_driver alarmtimer_driver = { -- cgit From b3e0b1b6d841a4b2f64fc09ea728913da8218424 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 16 Oct 2015 14:39:38 +0200 Subject: locking, sched: Introduce smp_cond_acquire() and use it Introduce smp_cond_acquire() which combines a control dependency and a read barrier to form acquire semantics. This primitive has two benefits: - it documents control dependencies, - its typically cheaper than using smp_load_acquire() in a loop. Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Linus Torvalds Cc: Mike Galbraith Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Thomas Gleixner Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock.c | 3 +-- kernel/sched/core.c | 8 +------- kernel/sched/sched.h | 2 +- 3 files changed, 3 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c index 986207887def..ed9d96708f93 100644 --- a/kernel/locking/qspinlock.c +++ b/kernel/locking/qspinlock.c @@ -433,8 +433,7 @@ queue: * */ pv_wait_head(lock, node); - while ((val = smp_load_acquire(&lock->val.counter)) & _Q_LOCKED_PENDING_MASK) - cpu_relax(); + smp_cond_acquire(!((val = atomic_read(&lock->val)) & _Q_LOCKED_PENDING_MASK)); /* * claim the lock: diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 7063c6a07440..9f7862da2cd1 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1968,19 +1968,13 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) /* * If the owning (remote) cpu is still in the middle of schedule() with * this task as prev, wait until its done referencing the task. - */ - while (p->on_cpu) - cpu_relax(); - /* - * Combined with the control dependency above, we have an effective - * smp_load_acquire() without the need for full barriers. * * Pairs with the smp_store_release() in finish_lock_switch(). * * This ensures that tasks getting woken will be fully ordered against * their previous state and preserve Program Order. */ - smp_rmb(); + smp_cond_acquire(!p->on_cpu); p->sched_contributes_to_load = !!task_contributes_to_load(p); p->state = TASK_WAKING; diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index b242775bf670..1e0bb4afe3fd 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1076,7 +1076,7 @@ static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev) * In particular, the load of prev->state in finish_task_switch() must * happen before this. * - * Pairs with the control dependency and rmb in try_to_wake_up(). + * Pairs with the smp_cond_acquire() in try_to_wake_up(). */ smp_store_release(&prev->on_cpu, 0); #endif -- cgit From 8643cda549ca49a403160892db68504569ac9052 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 17 Nov 2015 19:01:11 +0100 Subject: sched/core, locking: Document Program-Order guarantees These are some notes on the scheduler locking and how it provides program order guarantees on SMP systems. ( This commit is in the locking tree, because the new documentation refers to a newly introduced locking primitive. ) Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Boqun Feng Cc: David Howells Cc: Jonathan Corbet Cc: Linus Torvalds Cc: Michal Hocko Cc: Mike Galbraith Cc: Oleg Nesterov Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Will Deacon Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 91 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 91 insertions(+) (limited to 'kernel') diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 9f7862da2cd1..91db75018652 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1905,6 +1905,97 @@ static void ttwu_queue(struct task_struct *p, int cpu) raw_spin_unlock(&rq->lock); } +/* + * Notes on Program-Order guarantees on SMP systems. + * + * MIGRATION + * + * The basic program-order guarantee on SMP systems is that when a task [t] + * migrates, all its activity on its old cpu [c0] happens-before any subsequent + * execution on its new cpu [c1]. + * + * For migration (of runnable tasks) this is provided by the following means: + * + * A) UNLOCK of the rq(c0)->lock scheduling out task t + * B) migration for t is required to synchronize *both* rq(c0)->lock and + * rq(c1)->lock (if not at the same time, then in that order). + * C) LOCK of the rq(c1)->lock scheduling in task + * + * Transitivity guarantees that B happens after A and C after B. + * Note: we only require RCpc transitivity. + * Note: the cpu doing B need not be c0 or c1 + * + * Example: + * + * CPU0 CPU1 CPU2 + * + * LOCK rq(0)->lock + * sched-out X + * sched-in Y + * UNLOCK rq(0)->lock + * + * LOCK rq(0)->lock // orders against CPU0 + * dequeue X + * UNLOCK rq(0)->lock + * + * LOCK rq(1)->lock + * enqueue X + * UNLOCK rq(1)->lock + * + * LOCK rq(1)->lock // orders against CPU2 + * sched-out Z + * sched-in X + * UNLOCK rq(1)->lock + * + * + * BLOCKING -- aka. SLEEP + WAKEUP + * + * For blocking we (obviously) need to provide the same guarantee as for + * migration. However the means are completely different as there is no lock + * chain to provide order. Instead we do: + * + * 1) smp_store_release(X->on_cpu, 0) + * 2) smp_cond_acquire(!X->on_cpu) + * + * Example: + * + * CPU0 (schedule) CPU1 (try_to_wake_up) CPU2 (schedule) + * + * LOCK rq(0)->lock LOCK X->pi_lock + * dequeue X + * sched-out X + * smp_store_release(X->on_cpu, 0); + * + * smp_cond_acquire(!X->on_cpu); + * X->state = WAKING + * set_task_cpu(X,2) + * + * LOCK rq(2)->lock + * enqueue X + * X->state = RUNNING + * UNLOCK rq(2)->lock + * + * LOCK rq(2)->lock // orders against CPU1 + * sched-out Z + * sched-in X + * UNLOCK rq(2)->lock + * + * UNLOCK X->pi_lock + * UNLOCK rq(0)->lock + * + * + * However; for wakeups there is a second guarantee we must provide, namely we + * must observe the state that lead to our wakeup. That is, not only must our + * task observe its own prior state, it must also observe the stores prior to + * its wakeup. + * + * This means that any means of doing remote wakeups must order the CPU doing + * the wakeup against the CPU the task is going to end up running on. This, + * however, is already required for the regular Program-Order guarantee above, + * since the waking CPU is the one issueing the ACQUIRE (smp_cond_acquire). + * + */ + /** * try_to_wake_up - wake up a thread * @p: the thread to be awakened -- cgit From ad936d8658fd348338cb7d42c577dac77892b074 Mon Sep 17 00:00:00 2001 From: Byungchul Park Date: Sat, 24 Oct 2015 01:16:19 +0900 Subject: sched/fair: Make it possible to account fair load avg consistently The current code accounts for the time a task was absent from the fair class (per ATTACH_AGE_LOAD). However it does not work correctly when a task got migrated or moved to another cgroup while outside of the fair class. This patch tries to address that by aging on migration. We locklessly read the 'last_update_time' stamp from both the old and new cfs_rq, ages the load upto the old time, and sets it to the new time. These timestamps should in general not be more than 1 tick apart from one another, so there is a definite bound on things. Signed-off-by: Byungchul Park [ Changelog, a few edits and !SMP build fix ] Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1445616981-29904-2-git-send-email-byungchul.park@lge.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 4 ++++ kernel/sched/fair.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ kernel/sched/sched.h | 11 ++++++++++- 3 files changed, 60 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8969a9a9dab5..32d83e49cfd7 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2120,6 +2120,10 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) p->se.vruntime = 0; INIT_LIST_HEAD(&p->se.group_node); +#ifdef CONFIG_FAIR_GROUP_SCHED + p->se.cfs_rq = NULL; +#endif + #ifdef CONFIG_SCHEDSTATS memset(&p->se.statistics, 0, sizeof(p->se.statistics)); #endif diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index ff8ec8695957..efd664c4926e 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2715,6 +2715,52 @@ static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force) } } +/* + * Called within set_task_rq() right before setting a task's cpu. The + * caller only guarantees p->pi_lock is held; no other assumptions, + * including the state of rq->lock, should be made. + */ +void set_task_rq_fair(struct sched_entity *se, + struct cfs_rq *prev, struct cfs_rq *next) +{ + if (!sched_feat(ATTACH_AGE_LOAD)) + return; + + /* + * We are supposed to update the task to "current" time, then its up to + * date and ready to go to new CPU/cfs_rq. But we have difficulty in + * getting what current time is, so simply throw away the out-of-date + * time. This will result in the wakee task is less decayed, but giving + * the wakee more load sounds not bad. + */ + if (se->avg.last_update_time && prev) { + u64 p_last_update_time; + u64 n_last_update_time; + +#ifndef CONFIG_64BIT + u64 p_last_update_time_copy; + u64 n_last_update_time_copy; + + do { + p_last_update_time_copy = prev->load_last_update_time_copy; + n_last_update_time_copy = next->load_last_update_time_copy; + + smp_rmb(); + + p_last_update_time = prev->avg.last_update_time; + n_last_update_time = next->avg.last_update_time; + + } while (p_last_update_time != p_last_update_time_copy || + n_last_update_time != n_last_update_time_copy); +#else + p_last_update_time = prev->avg.last_update_time; + n_last_update_time = next->avg.last_update_time; +#endif + __update_load_avg(p_last_update_time, cpu_of(rq_of(prev)), + &se->avg, 0, 0, NULL); + se->avg.last_update_time = n_last_update_time; + } +} #else /* CONFIG_FAIR_GROUP_SCHED */ static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force) {} #endif /* CONFIG_FAIR_GROUP_SCHED */ diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index cdae23dabfdc..9a029fadcfbe 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -335,7 +335,15 @@ extern void sched_move_task(struct task_struct *tsk); #ifdef CONFIG_FAIR_GROUP_SCHED extern int sched_group_set_shares(struct task_group *tg, unsigned long shares); -#endif + +#ifdef CONFIG_SMP +extern void set_task_rq_fair(struct sched_entity *se, + struct cfs_rq *prev, struct cfs_rq *next); +#else /* !CONFIG_SMP */ +static inline void set_task_rq_fair(struct sched_entity *se, + struct cfs_rq *prev, struct cfs_rq *next) { } +#endif /* CONFIG_SMP */ +#endif /* CONFIG_FAIR_GROUP_SCHED */ #else /* CONFIG_CGROUP_SCHED */ @@ -933,6 +941,7 @@ static inline void set_task_rq(struct task_struct *p, unsigned int cpu) #endif #ifdef CONFIG_FAIR_GROUP_SCHED + set_task_rq_fair(&p->se, p->se.cfs_rq, tg->cfs_rq[cpu]); p->se.cfs_rq = tg->cfs_rq[cpu]; p->se.parent = tg->se[cpu]; #endif -- cgit From 7877a0ba5ec63c7b0111b06c773f1696fa17b35a Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Thu, 19 Nov 2015 16:47:29 +0100 Subject: sched/cputime: Remove extra cost in task_cputime() There is an extra cost in task_cputime() and task_cputime_scaled() when nohz_full is not activated. When vtime accounting is not enabled, we don't need to get deltas of utime and stime under vtime seqlock. This patch removes that cost with adding a shortcut route if vtime accounting is not enabled. Use context_tracking_is_enabled() to check if vtime is accounting on some cpu, in which case only we need to check the tickless cputime delta. Signed-off-by: Hiroshi Shimamoto Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Chris Metcalf Cc: Christoph Lameter Cc: Linus Torvalds Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447948054-28668-3-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/sched/cputime.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'kernel') diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 05de80b48586..1128d4ba6c55 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -853,6 +853,14 @@ void task_cputime(struct task_struct *t, cputime_t *utime, cputime_t *stime) { cputime_t udelta, sdelta; + if (!context_tracking_is_enabled()) { + if (utime) + *utime = t->utime; + if (stime) + *stime = t->stime; + return; + } + fetch_task_cputime(t, utime, stime, &t->utime, &t->stime, &udelta, &sdelta); if (utime) @@ -866,6 +874,14 @@ void task_cputime_scaled(struct task_struct *t, { cputime_t udelta, sdelta; + if (!context_tracking_is_enabled()) { + if (utimescaled) + *utimescaled = t->utimescaled; + if (stimescaled) + *stimescaled = t->stimescaled; + return; + } + fetch_task_cputime(t, utimescaled, stimescaled, &t->utimescaled, &t->stimescaled, &udelta, &sdelta); if (utimescaled) -- cgit From 7098c1eac75dc03fdbb7249171a6e68ce6044a5a Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 19 Nov 2015 16:47:30 +0100 Subject: sched/cputime: Clarify vtime symbols and document them VTIME_SLEEPING state happens either when: 1) The task is sleeping and no tickless delta is to be added on the task cputime stats. 2) The CPU isn't running vtime at all, so the same properties of 1) applies. Lets rename the vtime symbol to reflect both states. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Chris Metcalf Cc: Christoph Lameter Cc: Hiroshi Shimamoto Cc: Linus Torvalds Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447948054-28668-4-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/fork.c | 2 +- kernel/sched/cputime.c | 6 +++--- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/fork.c b/kernel/fork.c index f97f2c449f5c..c0a13706b1a7 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1350,7 +1350,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN seqlock_init(&p->vtime_seqlock); p->vtime_snap = 0; - p->vtime_snap_whence = VTIME_SLEEPING; + p->vtime_snap_whence = VTIME_INACTIVE; #endif #if defined(SPLIT_RSS_COUNTING) diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 1128d4ba6c55..4a18a6ed7723 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -680,7 +680,7 @@ static cputime_t get_vtime_delta(struct task_struct *tsk) { unsigned long long delta = vtime_delta(tsk); - WARN_ON_ONCE(tsk->vtime_snap_whence == VTIME_SLEEPING); + WARN_ON_ONCE(tsk->vtime_snap_whence == VTIME_INACTIVE); tsk->vtime_snap += delta; /* CHECKME: always safe to convert nsecs to cputime? */ @@ -764,7 +764,7 @@ void vtime_account_idle(struct task_struct *tsk) void arch_vtime_task_switch(struct task_struct *prev) { write_seqlock(&prev->vtime_seqlock); - prev->vtime_snap_whence = VTIME_SLEEPING; + prev->vtime_snap_whence = VTIME_INACTIVE; write_sequnlock(&prev->vtime_seqlock); write_seqlock(¤t->vtime_seqlock); @@ -829,7 +829,7 @@ fetch_task_cputime(struct task_struct *t, *s_dst = *s_src; /* Task is sleeping, nothing to add */ - if (t->vtime_snap_whence == VTIME_SLEEPING || + if (t->vtime_snap_whence == VTIME_INACTIVE || is_idle_task(t)) continue; -- cgit From cab245d68c38afff1a4c4d018ab7e1d316982f5d Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 19 Nov 2015 16:47:31 +0100 Subject: sched/cputime: Correctly handle task guest time on housekeepers When a task runs on a housekeeper (a CPU running with the periodic tick with neighbours running tickless), it doesn't account cputime using vtime but relies on the tick. Such a task has its vtime_snap_whence value set to VTIME_INACTIVE. Readers won't handle that correctly though. As long as vtime is running on some CPU, readers incorretly assume that vtime runs on all CPUs and always compute the tickless cputime delta, which is only junk on housekeepers. So lets fix this with checking that the target runs on a vtime CPU through the appropriate state check before computing the tickless delta. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Chris Metcalf Cc: Christoph Lameter Cc: Hiroshi Shimamoto Cc: Linus Torvalds Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447948054-28668-5-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/sched/cputime.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 4a18a6ed7723..5cf24e743c6e 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -795,7 +795,7 @@ cputime_t task_gtime(struct task_struct *t) seq = read_seqbegin(&t->vtime_seqlock); gtime = t->gtime; - if (t->flags & PF_VCPU) + if (t->vtime_snap_whence == VTIME_SYS && t->flags & PF_VCPU) gtime += vtime_delta(t); } while (read_seqretry(&t->vtime_seqlock, seq)); -- cgit From 55dbdcfa05533f44c9416070b8a9f6432b22314a Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 19 Nov 2015 16:47:32 +0100 Subject: sched/cputime: Rename vtime_accounting_enabled() to vtime_accounting_cpu_enabled() vtime_accounting_enabled() checks if vtime is running on the current CPU and is as such a misnomer. Lets rename it to a function that reflect its locality. We are going to need the current name for a function that tells if vtime runs at all on some CPU. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Chris Metcalf Cc: Christoph Lameter Cc: Hiroshi Shimamoto Cc: Linus Torvalds Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447948054-28668-6-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/sched/cputime.c | 2 +- kernel/time/tick-sched.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 5cf24e743c6e..572721756359 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -466,7 +466,7 @@ void account_process_tick(struct task_struct *p, int user_tick) cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy); struct rq *rq = this_rq(); - if (vtime_accounting_enabled()) + if (vtime_accounting_cpu_enabled()) return; if (sched_clock_irqtime) { diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 515edf3eb62b..11ce59916c1a 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -875,7 +875,7 @@ static void tick_nohz_account_idle_ticks(struct tick_sched *ts) #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE unsigned long ticks; - if (vtime_accounting_enabled()) + if (vtime_accounting_cpu_enabled()) return; /* * We stopped the tick in idle. Update process times would miss the -- cgit From e592539466380279a9e6e6fdfe4545aa54f22593 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 19 Nov 2015 16:47:33 +0100 Subject: sched/cputime: Introduce vtime accounting check for readers Readers need to know if vtime runs at all on some CPU somewhere, this is a fast-path check to determine if we need to check further the need to add up any tickless cputime delta. This fast path check uses context tracking state because vtime is tied to context tracking as of now. This check appears to be confusing though so lets use a vtime function that deals with context tracking details in vtime implementation instead. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Chris Metcalf Cc: Christoph Lameter Cc: Hiroshi Shimamoto Cc: Linus Torvalds Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447948054-28668-7-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/sched/cputime.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 572721756359..9989c3f61723 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -788,7 +788,7 @@ cputime_t task_gtime(struct task_struct *t) unsigned int seq; cputime_t gtime; - if (!context_tracking_is_enabled()) + if (!vtime_accounting_enabled()) return t->gtime; do { @@ -853,7 +853,7 @@ void task_cputime(struct task_struct *t, cputime_t *utime, cputime_t *stime) { cputime_t udelta, sdelta; - if (!context_tracking_is_enabled()) { + if (!vtime_accounting_enabled()) { if (utime) *utime = t->utime; if (stime) @@ -874,7 +874,7 @@ void task_cputime_scaled(struct task_struct *t, { cputime_t udelta, sdelta; - if (!context_tracking_is_enabled()) { + if (!vtime_accounting_enabled()) { if (utimescaled) *utimescaled = t->utimescaled; if (stimescaled) -- cgit From b7ce2277f087fd052e7e1bbf432f7fecbee82bb6 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 19 Nov 2015 16:47:34 +0100 Subject: sched/cputime: Convert vtime_seqlock to seqcount The cputime can only be updated by the current task itself, even in vtime case. So we can safely use seqcount instead of seqlock as there is no writer concurrency involved. Signed-off-by: Frederic Weisbecker Signed-off-by: Peter Zijlstra (Intel) Cc: Chris Metcalf Cc: Christoph Lameter Cc: Hiroshi Shimamoto Cc: Linus Torvalds Cc: Luiz Capitulino Cc: Mike Galbraith Cc: Paul E . McKenney Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Rik van Riel Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447948054-28668-8-git-send-email-fweisbec@gmail.com Signed-off-by: Ingo Molnar --- kernel/fork.c | 2 +- kernel/sched/cputime.c | 46 ++++++++++++++++++++++++---------------------- 2 files changed, 25 insertions(+), 23 deletions(-) (limited to 'kernel') diff --git a/kernel/fork.c b/kernel/fork.c index c0a13706b1a7..eea32b55432a 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1348,7 +1348,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, prev_cputime_init(&p->prev_cputime); #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN - seqlock_init(&p->vtime_seqlock); + seqcount_init(&p->vtime_seqcount); p->vtime_snap = 0; p->vtime_snap_whence = VTIME_INACTIVE; #endif diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 9989c3f61723..d5ff5c6bf829 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -696,37 +696,37 @@ static void __vtime_account_system(struct task_struct *tsk) void vtime_account_system(struct task_struct *tsk) { - write_seqlock(&tsk->vtime_seqlock); + write_seqcount_begin(&tsk->vtime_seqcount); __vtime_account_system(tsk); - write_sequnlock(&tsk->vtime_seqlock); + write_seqcount_end(&tsk->vtime_seqcount); } void vtime_gen_account_irq_exit(struct task_struct *tsk) { - write_seqlock(&tsk->vtime_seqlock); + write_seqcount_begin(&tsk->vtime_seqcount); __vtime_account_system(tsk); if (context_tracking_in_user()) tsk->vtime_snap_whence = VTIME_USER; - write_sequnlock(&tsk->vtime_seqlock); + write_seqcount_end(&tsk->vtime_seqcount); } void vtime_account_user(struct task_struct *tsk) { cputime_t delta_cpu; - write_seqlock(&tsk->vtime_seqlock); + write_seqcount_begin(&tsk->vtime_seqcount); delta_cpu = get_vtime_delta(tsk); tsk->vtime_snap_whence = VTIME_SYS; account_user_time(tsk, delta_cpu, cputime_to_scaled(delta_cpu)); - write_sequnlock(&tsk->vtime_seqlock); + write_seqcount_end(&tsk->vtime_seqcount); } void vtime_user_enter(struct task_struct *tsk) { - write_seqlock(&tsk->vtime_seqlock); + write_seqcount_begin(&tsk->vtime_seqcount); __vtime_account_system(tsk); tsk->vtime_snap_whence = VTIME_USER; - write_sequnlock(&tsk->vtime_seqlock); + write_seqcount_end(&tsk->vtime_seqcount); } void vtime_guest_enter(struct task_struct *tsk) @@ -738,19 +738,19 @@ void vtime_guest_enter(struct task_struct *tsk) * synchronization against the reader (task_gtime()) * that can thus safely catch up with a tickless delta. */ - write_seqlock(&tsk->vtime_seqlock); + write_seqcount_begin(&tsk->vtime_seqcount); __vtime_account_system(tsk); current->flags |= PF_VCPU; - write_sequnlock(&tsk->vtime_seqlock); + write_seqcount_end(&tsk->vtime_seqcount); } EXPORT_SYMBOL_GPL(vtime_guest_enter); void vtime_guest_exit(struct task_struct *tsk) { - write_seqlock(&tsk->vtime_seqlock); + write_seqcount_begin(&tsk->vtime_seqcount); __vtime_account_system(tsk); current->flags &= ~PF_VCPU; - write_sequnlock(&tsk->vtime_seqlock); + write_seqcount_end(&tsk->vtime_seqcount); } EXPORT_SYMBOL_GPL(vtime_guest_exit); @@ -763,24 +763,26 @@ void vtime_account_idle(struct task_struct *tsk) void arch_vtime_task_switch(struct task_struct *prev) { - write_seqlock(&prev->vtime_seqlock); + write_seqcount_begin(&prev->vtime_seqcount); prev->vtime_snap_whence = VTIME_INACTIVE; - write_sequnlock(&prev->vtime_seqlock); + write_seqcount_end(&prev->vtime_seqcount); - write_seqlock(¤t->vtime_seqlock); + write_seqcount_begin(¤t->vtime_seqcount); current->vtime_snap_whence = VTIME_SYS; current->vtime_snap = sched_clock_cpu(smp_processor_id()); - write_sequnlock(¤t->vtime_seqlock); + write_seqcount_end(¤t->vtime_seqcount); } void vtime_init_idle(struct task_struct *t, int cpu) { unsigned long flags; - write_seqlock_irqsave(&t->vtime_seqlock, flags); + local_irq_save(flags); + write_seqcount_begin(&t->vtime_seqcount); t->vtime_snap_whence = VTIME_SYS; t->vtime_snap = sched_clock_cpu(cpu); - write_sequnlock_irqrestore(&t->vtime_seqlock, flags); + write_seqcount_end(&t->vtime_seqcount); + local_irq_restore(flags); } cputime_t task_gtime(struct task_struct *t) @@ -792,13 +794,13 @@ cputime_t task_gtime(struct task_struct *t) return t->gtime; do { - seq = read_seqbegin(&t->vtime_seqlock); + seq = read_seqcount_begin(&t->vtime_seqcount); gtime = t->gtime; if (t->vtime_snap_whence == VTIME_SYS && t->flags & PF_VCPU) gtime += vtime_delta(t); - } while (read_seqretry(&t->vtime_seqlock, seq)); + } while (read_seqcount_retry(&t->vtime_seqcount, seq)); return gtime; } @@ -821,7 +823,7 @@ fetch_task_cputime(struct task_struct *t, *udelta = 0; *sdelta = 0; - seq = read_seqbegin(&t->vtime_seqlock); + seq = read_seqcount_begin(&t->vtime_seqcount); if (u_dst) *u_dst = *u_src; @@ -845,7 +847,7 @@ fetch_task_cputime(struct task_struct *t, if (t->vtime_snap_whence == VTIME_SYS) *sdelta = delta; } - } while (read_seqretry(&t->vtime_seqlock, seq)); + } while (read_seqcount_retry(&t->vtime_seqcount, seq)); } -- cgit From ed82b8a1ff76ed7b2709e36ed361ddd022fe2407 Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Sun, 29 Nov 2015 20:59:43 -0800 Subject: sched/core: Move the sched_to_prio[] arrays out of line When building a kernel with a gcc 6 snapshot the compiler complains about unused const static variables for prio_to_weight and prio_to_mult for multiple scheduler files (all but core.c and autogroup.c) The way the array is currently declared it will be duplicated in every scheduler file that includes sched.h, which seems rather wasteful. Move the array out of line into core.c. I also added a sched_ prefix to avoid any potential name space collisions. Signed-off-by: Andi Kleen Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1448859583-3252-1-git-send-email-andi@firstfloor.org Signed-off-by: Ingo Molnar --- kernel/sched/auto_group.c | 2 +- kernel/sched/core.c | 45 +++++++++++++++++++++++++++++++++++++++++++-- kernel/sched/sched.h | 42 ++---------------------------------------- 3 files changed, 46 insertions(+), 43 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/auto_group.c b/kernel/sched/auto_group.c index 750ed601ddf7..a5d966cb8891 100644 --- a/kernel/sched/auto_group.c +++ b/kernel/sched/auto_group.c @@ -212,7 +212,7 @@ int proc_sched_autogroup_set_nice(struct task_struct *p, int nice) ag = autogroup_task_get(p); down_write(&ag->lock); - err = sched_group_set_shares(ag->tg, prio_to_weight[nice + 20]); + err = sched_group_set_shares(ag->tg, sched_prio_to_weight[nice + 20]); if (!err) ag->nice = nice; up_write(&ag->lock); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 32d83e49cfd7..d591db1f2d6d 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -823,8 +823,8 @@ static void set_load_weight(struct task_struct *p) return; } - load->weight = scale_load(prio_to_weight[prio]); - load->inv_weight = prio_to_wmult[prio]; + load->weight = scale_load(sched_prio_to_weight[prio]); + load->inv_weight = sched_prio_to_wmult[prio]; } static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags) @@ -8625,3 +8625,44 @@ void dump_cpu_task(int cpu) pr_info("Task dump for CPU %d:\n", cpu); sched_show_task(cpu_curr(cpu)); } + +/* + * Nice levels are multiplicative, with a gentle 10% change for every + * nice level changed. I.e. when a CPU-bound task goes from nice 0 to + * nice 1, it will get ~10% less CPU time than another CPU-bound task + * that remained on nice 0. + * + * The "10% effect" is relative and cumulative: from _any_ nice level, + * if you go up 1 level, it's -10% CPU usage, if you go down 1 level + * it's +10% CPU usage. (to achieve that we use a multiplier of 1.25. + * If a task goes up by ~10% and another task goes down by ~10% then + * the relative distance between them is ~25%.) + */ +const int sched_prio_to_weight[40] = { + /* -20 */ 88761, 71755, 56483, 46273, 36291, + /* -15 */ 29154, 23254, 18705, 14949, 11916, + /* -10 */ 9548, 7620, 6100, 4904, 3906, + /* -5 */ 3121, 2501, 1991, 1586, 1277, + /* 0 */ 1024, 820, 655, 526, 423, + /* 5 */ 335, 272, 215, 172, 137, + /* 10 */ 110, 87, 70, 56, 45, + /* 15 */ 36, 29, 23, 18, 15, +}; + +/* + * Inverse (2^32/x) values of the sched_prio_to_weight[] array, precalculated. + * + * In cases where the weight does not change often, we can use the + * precalculated inverse to speed up arithmetics by turning divisions + * into multiplications: + */ +const u32 sched_prio_to_wmult[40] = { + /* -20 */ 48388, 59856, 76040, 92818, 118348, + /* -15 */ 147320, 184698, 229616, 287308, 360437, + /* -10 */ 449829, 563644, 704093, 875809, 1099582, + /* -5 */ 1376151, 1717300, 2157191, 2708050, 3363326, + /* 0 */ 4194304, 5237765, 6557202, 8165337, 10153587, + /* 5 */ 12820798, 15790321, 19976592, 24970740, 31350126, + /* 10 */ 39045157, 49367440, 61356676, 76695844, 95443717, + /* 15 */ 119304647, 148102320, 186737708, 238609294, 286331153, +}; diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 9a029fadcfbe..472cd142e4f4 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1122,46 +1122,8 @@ static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev) #define WEIGHT_IDLEPRIO 3 #define WMULT_IDLEPRIO 1431655765 -/* - * Nice levels are multiplicative, with a gentle 10% change for every - * nice level changed. I.e. when a CPU-bound task goes from nice 0 to - * nice 1, it will get ~10% less CPU time than another CPU-bound task - * that remained on nice 0. - * - * The "10% effect" is relative and cumulative: from _any_ nice level, - * if you go up 1 level, it's -10% CPU usage, if you go down 1 level - * it's +10% CPU usage. (to achieve that we use a multiplier of 1.25. - * If a task goes up by ~10% and another task goes down by ~10% then - * the relative distance between them is ~25%.) - */ -static const int prio_to_weight[40] = { - /* -20 */ 88761, 71755, 56483, 46273, 36291, - /* -15 */ 29154, 23254, 18705, 14949, 11916, - /* -10 */ 9548, 7620, 6100, 4904, 3906, - /* -5 */ 3121, 2501, 1991, 1586, 1277, - /* 0 */ 1024, 820, 655, 526, 423, - /* 5 */ 335, 272, 215, 172, 137, - /* 10 */ 110, 87, 70, 56, 45, - /* 15 */ 36, 29, 23, 18, 15, -}; - -/* - * Inverse (2^32/x) values of the prio_to_weight[] array, precalculated. - * - * In cases where the weight does not change often, we can use the - * precalculated inverse to speed up arithmetics by turning divisions - * into multiplications: - */ -static const u32 prio_to_wmult[40] = { - /* -20 */ 48388, 59856, 76040, 92818, 118348, - /* -15 */ 147320, 184698, 229616, 287308, 360437, - /* -10 */ 449829, 563644, 704093, 875809, 1099582, - /* -5 */ 1376151, 1717300, 2157191, 2708050, 3363326, - /* 0 */ 4194304, 5237765, 6557202, 8165337, 10153587, - /* 5 */ 12820798, 15790321, 19976592, 24970740, 31350126, - /* 10 */ 39045157, 49367440, 61356676, 76695844, 95443717, - /* 15 */ 119304647, 148102320, 186737708, 238609294, 286331153, -}; +extern const int sched_prio_to_weight[40]; +extern const u32 sched_prio_to_wmult[40]; #define ENQUEUE_WAKEUP 0x01 #define ENQUEUE_HEAD 0x02 -- cgit From a426f99c91d1036767a7819aaaba6bd3191b7f06 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Wed, 25 Nov 2015 14:09:38 -0500 Subject: sched/fair: Avoid redundant idle_cpu() call in update_sg_lb_stats() Part of the responsibility of the update_sg_lb_stats() function is to update the idle_cpus statistical counter in struct sg_lb_stats. This check is done by calling idle_cpu(). The idle_cpu() function, in turn, checks a number of fields within the run queue structure such as rq->curr and rq->nr_running. With the current layout of the run queue structure, rq->curr and rq->nr_running are in separate cachelines. The rq->curr variable is checked first followed by nr_running. As nr_running is also accessed by update_sg_lb_stats() earlier, it makes no sense to load another cacheline when nr_running is not 0 as idle_cpu() will always return false in this case. This patch eliminates this redundant cacheline load by checking the cached nr_running before calling idle_cpu(). Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Douglas Hatch Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1448478580-26467-2-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index efd664c4926e..4b0e8b8700fd 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -6398,7 +6398,7 @@ static inline void update_sg_lb_stats(struct lb_env *env, bool *overload) { unsigned long load; - int i; + int i, nr_running; memset(sgs, 0, sizeof(*sgs)); @@ -6415,7 +6415,8 @@ static inline void update_sg_lb_stats(struct lb_env *env, sgs->group_util += cpu_util(i); sgs->sum_nr_running += rq->cfs.h_nr_running; - if (rq->nr_running > 1) + nr_running = rq->nr_running; + if (nr_running > 1) *overload = true; #ifdef CONFIG_NUMA_BALANCING @@ -6423,7 +6424,10 @@ static inline void update_sg_lb_stats(struct lb_env *env, sgs->nr_preferred_running += rq->nr_preferred_running; #endif sgs->sum_weighted_load += weighted_cpuload(i); - if (idle_cpu(i)) + /* + * No need to call idle_cpu() if nr_running is not 0 + */ + if (!nr_running && idle_cpu(i)) sgs->idle_cpus++; } -- cgit From b0367629acf62a78404c467cd09df447c2fea804 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Wed, 2 Dec 2015 13:41:49 -0500 Subject: sched/fair: Move the cache-hot 'load_avg' variable into its own cacheline If a system with large number of sockets was driven to full utilization, it was found that the clock tick handling occupied a rather significant proportion of CPU time when fair group scheduling and autogroup were enabled. Running a java benchmark on a 16-socket IvyBridge-EX system, the perf profile looked like: 10.52% 0.00% java [kernel.vmlinux] [k] smp_apic_timer_interrupt 9.66% 0.05% java [kernel.vmlinux] [k] hrtimer_interrupt 8.65% 0.03% java [kernel.vmlinux] [k] tick_sched_timer 8.56% 0.00% java [kernel.vmlinux] [k] update_process_times 8.07% 0.03% java [kernel.vmlinux] [k] scheduler_tick 6.91% 1.78% java [kernel.vmlinux] [k] task_tick_fair 5.24% 5.04% java [kernel.vmlinux] [k] update_cfs_shares In particular, the high CPU time consumed by update_cfs_shares() was mostly due to contention on the cacheline that contained the task_group's load_avg statistical counter. This cacheline may also contains variables like shares, cfs_rq & se which are accessed rather frequently during clock tick processing. This patch moves the load_avg variable into another cacheline separated from the other frequently accessed variables. It also creates a cacheline aligned kmemcache for task_group to make sure that all the allocated task_group's are cacheline aligned. By doing so, the perf profile became: 9.44% 0.00% java [kernel.vmlinux] [k] smp_apic_timer_interrupt 8.74% 0.01% java [kernel.vmlinux] [k] hrtimer_interrupt 7.83% 0.03% java [kernel.vmlinux] [k] tick_sched_timer 7.74% 0.00% java [kernel.vmlinux] [k] update_process_times 7.27% 0.03% java [kernel.vmlinux] [k] scheduler_tick 5.94% 1.74% java [kernel.vmlinux] [k] task_tick_fair 4.15% 3.92% java [kernel.vmlinux] [k] update_cfs_shares The %cpu time is still pretty high, but it is better than before. The benchmark results before and after the patch was as follows: Before patch - Max-jOPs: 907533 Critical-jOps: 134877 After patch - Max-jOPs: 916011 Critical-jOps: 142366 Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Ben Segall Cc: Douglas Hatch Cc: Linus Torvalds Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Paul Turner Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Cc: Yuyang Du Link: http://lkml.kernel.org/r/1449081710-20185-3-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/sched/core.c | 10 +++++++--- kernel/sched/sched.h | 7 ++++++- 2 files changed, 13 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/core.c b/kernel/sched/core.c index d591db1f2d6d..aa3f97869217 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -7370,6 +7370,9 @@ int in_sched_functions(unsigned long addr) */ struct task_group root_task_group; LIST_HEAD(task_groups); + +/* Cacheline aligned slab cache for task_group */ +static struct kmem_cache *task_group_cache __read_mostly; #endif DECLARE_PER_CPU(cpumask_var_t, load_balance_mask); @@ -7427,11 +7430,12 @@ void __init sched_init(void) #endif /* CONFIG_RT_GROUP_SCHED */ #ifdef CONFIG_CGROUP_SCHED + task_group_cache = KMEM_CACHE(task_group, 0); + list_add(&root_task_group.list, &task_groups); INIT_LIST_HEAD(&root_task_group.children); INIT_LIST_HEAD(&root_task_group.siblings); autogroup_init(&init_task); - #endif /* CONFIG_CGROUP_SCHED */ for_each_possible_cpu(i) { @@ -7712,7 +7716,7 @@ static void free_sched_group(struct task_group *tg) free_fair_sched_group(tg); free_rt_sched_group(tg); autogroup_free(tg); - kfree(tg); + kmem_cache_free(task_group_cache, tg); } /* allocate runqueue etc for a new task group */ @@ -7720,7 +7724,7 @@ struct task_group *sched_create_group(struct task_group *parent) { struct task_group *tg; - tg = kzalloc(sizeof(*tg), GFP_KERNEL); + tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO); if (!tg) return ERR_PTR(-ENOMEM); diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 472cd142e4f4..a5a6b3e60868 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -248,7 +248,12 @@ struct task_group { unsigned long shares; #ifdef CONFIG_SMP - atomic_long_t load_avg; + /* + * load_avg can be heavily contended at clock tick time, so put + * it in its own cacheline separated from the fields above which + * will also be accessed at each tick. + */ + atomic_long_t load_avg ____cacheline_aligned; #endif #endif -- cgit From aa0b7ae06387d40a988ce16a189082dee6e570bc Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Wed, 2 Dec 2015 13:41:50 -0500 Subject: sched/fair: Disable the task group load_avg update for the root_task_group Currently, the update_tg_load_avg() function attempts to update the tg's load_avg value whenever the load changes even for root_task_group where the load_avg value will never be used. This patch will disable the load_avg update when the given task group is the root_task_group. Running a Java benchmark with noautogroup and a 4.3 kernel on a 16-socket IvyBridge-EX system, the amount of CPU time (as reported by perf) consumed by task_tick_fair() which includes update_tg_load_avg() decreased from 0.71% to 0.22%, a more than 3X reduction. The Max-jOPs results also increased slightly from 983015 to 986449. Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Reviewed-by: Ben Segall Cc: Douglas Hatch Cc: Linus Torvalds Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Paul Turner Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Cc: Yuyang Du Link: http://lkml.kernel.org/r/1449081710-20185-4-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 4b0e8b8700fd..1093873dcd0f 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2709,6 +2709,12 @@ static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force) { long delta = cfs_rq->avg.load_avg - cfs_rq->tg_load_avg_contrib; + /* + * No need to update load_avg for root_task_group as it is not used. + */ + if (cfs_rq->tg == &root_task_group) + return; + if (force || abs(delta) > cfs_rq->tg_load_avg_contrib / 64) { atomic_long_add(delta, &cfs_rq->tg->load_avg); cfs_rq->tg_load_avg_contrib = cfs_rq->avg.load_avg; -- cgit From 45e898b735620f426eddf105fc886d2966593a58 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Mon, 9 Nov 2015 19:09:25 -0500 Subject: locking/pvqspinlock: Collect slowpath lock statistics This patch enables the accumulation of kicking and waiting related PV qspinlock statistics when the new QUEUED_LOCK_STAT configuration option is selected. It also enables the collection of data which enable us to calculate the kicking and wakeup latencies which have a heavy dependency on the CPUs being used. The statistical counters are per-cpu variables to minimize the performance overhead in their updates. These counters are exported via the debugfs filesystem under the qlockstat directory. When the corresponding debugfs files are read, summation and computing of the required data are then performed. The measured latencies for different CPUs are: CPU Wakeup Kicking --- ------ ------- Haswell-EX 63.6us 7.4us Westmere-EX 67.6us 9.3us The measured latencies varied a bit from run-to-run. The wakeup latency is much higher than the kicking latency. A sample of statistical counters after system bootup (with vCPU overcommit) was: pv_hash_hops=1.00 pv_kick_unlock=1148 pv_kick_wake=1146 pv_latency_kick=11040 pv_latency_wake=194840 pv_spurious_wakeup=7 pv_wait_again=4 pv_wait_head=23 pv_wait_node=1129 Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447114167-47185-6-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock_paravirt.h | 32 +++- kernel/locking/qspinlock_stat.h | 281 ++++++++++++++++++++++++++++++++++++ 2 files changed, 308 insertions(+), 5 deletions(-) create mode 100644 kernel/locking/qspinlock_stat.h (limited to 'kernel') diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h index 4bd323d38c60..aaeeefb791f8 100644 --- a/kernel/locking/qspinlock_paravirt.h +++ b/kernel/locking/qspinlock_paravirt.h @@ -40,6 +40,11 @@ struct pv_node { u8 state; }; +/* + * Include queued spinlock statistics code + */ +#include "qspinlock_stat.h" + /* * Lock and MCS node addresses hash table for fast lookup * @@ -100,10 +105,13 @@ static struct qspinlock **pv_hash(struct qspinlock *lock, struct pv_node *node) { unsigned long offset, hash = hash_ptr(lock, pv_lock_hash_bits); struct pv_hash_entry *he; + int hopcnt = 0; for_each_hash_entry(he, offset, hash) { + hopcnt++; if (!cmpxchg(&he->lock, NULL, lock)) { WRITE_ONCE(he->node, node); + qstat_hop(hopcnt); return &he->lock; } } @@ -164,9 +172,11 @@ static void pv_init_node(struct mcs_spinlock *node) static void pv_wait_node(struct mcs_spinlock *node) { struct pv_node *pn = (struct pv_node *)node; + int waitcnt = 0; int loop; - for (;;) { + /* waitcnt processing will be compiled out if !QUEUED_LOCK_STAT */ + for (;; waitcnt++) { for (loop = SPIN_THRESHOLD; loop; loop--) { if (READ_ONCE(node->locked)) return; @@ -184,12 +194,16 @@ static void pv_wait_node(struct mcs_spinlock *node) */ smp_store_mb(pn->state, vcpu_halted); - if (!READ_ONCE(node->locked)) + if (!READ_ONCE(node->locked)) { + qstat_inc(qstat_pv_wait_node, true); + qstat_inc(qstat_pv_wait_again, waitcnt); pv_wait(&pn->state, vcpu_halted); + } /* - * If pv_kick_node() changed us to vcpu_hashed, retain that value - * so that pv_wait_head() knows to not also try to hash this lock. + * If pv_kick_node() changed us to vcpu_hashed, retain that + * value so that pv_wait_head() knows to not also try to hash + * this lock. */ cmpxchg(&pn->state, vcpu_halted, vcpu_running); @@ -200,6 +214,7 @@ static void pv_wait_node(struct mcs_spinlock *node) * So it is better to spin for a while in the hope that the * MCS lock will be released soon. */ + qstat_inc(qstat_pv_spurious_wakeup, !READ_ONCE(node->locked)); } /* @@ -250,6 +265,7 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) struct pv_node *pn = (struct pv_node *)node; struct __qspinlock *l = (void *)lock; struct qspinlock **lp = NULL; + int waitcnt = 0; int loop; /* @@ -259,7 +275,7 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) if (READ_ONCE(pn->state) == vcpu_hashed) lp = (struct qspinlock **)1; - for (;;) { + for (;; waitcnt++) { for (loop = SPIN_THRESHOLD; loop; loop--) { if (!READ_ONCE(l->locked)) return; @@ -290,14 +306,19 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) return; } } + qstat_inc(qstat_pv_wait_head, true); + qstat_inc(qstat_pv_wait_again, waitcnt); pv_wait(&l->locked, _Q_SLOW_VAL); + if (!READ_ONCE(l->locked)) + return; /* * The unlocker should have freed the lock before kicking the * CPU. So if the lock is still not free, it is a spurious * wakeup and so the vCPU should wait again after spinning for * a while. */ + qstat_inc(qstat_pv_spurious_wakeup, true); } /* @@ -352,6 +373,7 @@ __pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked) * vCPU is harmless other than the additional latency in completing * the unlock. */ + qstat_inc(qstat_pv_kick_unlock, true); pv_kick(node->cpu); } diff --git a/kernel/locking/qspinlock_stat.h b/kernel/locking/qspinlock_stat.h new file mode 100644 index 000000000000..b1553adec2e7 --- /dev/null +++ b/kernel/locking/qspinlock_stat.h @@ -0,0 +1,281 @@ +/* + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * Authors: Waiman Long + */ + +/* + * When queued spinlock statistical counters are enabled, the following + * debugfs files will be created for reporting the counter values: + * + * /qlockstat/ + * pv_hash_hops - average # of hops per hashing operation + * pv_kick_unlock - # of vCPU kicks issued at unlock time + * pv_kick_wake - # of vCPU kicks used for computing pv_latency_wake + * pv_latency_kick - average latency (ns) of vCPU kick operation + * pv_latency_wake - average latency (ns) from vCPU kick to wakeup + * pv_spurious_wakeup - # of spurious wakeups + * pv_wait_again - # of vCPU wait's that happened after a vCPU kick + * pv_wait_head - # of vCPU wait's at the queue head + * pv_wait_node - # of vCPU wait's at a non-head queue node + * + * Writing to the "reset_counters" file will reset all the above counter + * values. + * + * These statistical counters are implemented as per-cpu variables which are + * summed and computed whenever the corresponding debugfs files are read. This + * minimizes added overhead making the counters usable even in a production + * environment. + * + * There may be slight difference between pv_kick_wake and pv_kick_unlock. + */ +enum qlock_stats { + qstat_pv_hash_hops, + qstat_pv_kick_unlock, + qstat_pv_kick_wake, + qstat_pv_latency_kick, + qstat_pv_latency_wake, + qstat_pv_spurious_wakeup, + qstat_pv_wait_again, + qstat_pv_wait_head, + qstat_pv_wait_node, + qstat_num, /* Total number of statistical counters */ + qstat_reset_cnts = qstat_num, +}; + +#ifdef CONFIG_QUEUED_LOCK_STAT +/* + * Collect pvqspinlock statistics + */ +#include +#include +#include + +static const char * const qstat_names[qstat_num + 1] = { + [qstat_pv_hash_hops] = "pv_hash_hops", + [qstat_pv_kick_unlock] = "pv_kick_unlock", + [qstat_pv_kick_wake] = "pv_kick_wake", + [qstat_pv_spurious_wakeup] = "pv_spurious_wakeup", + [qstat_pv_latency_kick] = "pv_latency_kick", + [qstat_pv_latency_wake] = "pv_latency_wake", + [qstat_pv_wait_again] = "pv_wait_again", + [qstat_pv_wait_head] = "pv_wait_head", + [qstat_pv_wait_node] = "pv_wait_node", + [qstat_reset_cnts] = "reset_counters", +}; + +/* + * Per-cpu counters + */ +static DEFINE_PER_CPU(unsigned long, qstats[qstat_num]); +static DEFINE_PER_CPU(u64, pv_kick_time); + +/* + * Function to read and return the qlock statistical counter values + * + * The following counters are handled specially: + * 1. qstat_pv_latency_kick + * Average kick latency (ns) = pv_latency_kick/pv_kick_unlock + * 2. qstat_pv_latency_wake + * Average wake latency (ns) = pv_latency_wake/pv_kick_wake + * 3. qstat_pv_hash_hops + * Average hops/hash = pv_hash_hops/pv_kick_unlock + */ +static ssize_t qstat_read(struct file *file, char __user *user_buf, + size_t count, loff_t *ppos) +{ + char buf[64]; + int cpu, counter, len; + u64 stat = 0, kicks = 0; + + /* + * Get the counter ID stored in file->f_inode->i_private + */ + if (!file->f_inode) { + WARN_ON_ONCE(1); + return -EBADF; + } + counter = (long)(file->f_inode->i_private); + + if (counter >= qstat_num) + return -EBADF; + + for_each_possible_cpu(cpu) { + stat += per_cpu(qstats[counter], cpu); + /* + * Need to sum additional counter for some of them + */ + switch (counter) { + + case qstat_pv_latency_kick: + case qstat_pv_hash_hops: + kicks += per_cpu(qstats[qstat_pv_kick_unlock], cpu); + break; + + case qstat_pv_latency_wake: + kicks += per_cpu(qstats[qstat_pv_kick_wake], cpu); + break; + } + } + + if (counter == qstat_pv_hash_hops) { + u64 frac; + + frac = 100ULL * do_div(stat, kicks); + frac = DIV_ROUND_CLOSEST_ULL(frac, kicks); + + /* + * Return a X.XX decimal number + */ + len = snprintf(buf, sizeof(buf) - 1, "%llu.%02llu\n", stat, frac); + } else { + /* + * Round to the nearest ns + */ + if ((counter == qstat_pv_latency_kick) || + (counter == qstat_pv_latency_wake)) { + stat = 0; + if (kicks) + stat = DIV_ROUND_CLOSEST_ULL(stat, kicks); + } + len = snprintf(buf, sizeof(buf) - 1, "%llu\n", stat); + } + + return simple_read_from_buffer(user_buf, count, ppos, buf, len); +} + +/* + * Function to handle write request + * + * When counter = reset_cnts, reset all the counter values. + * Since the counter updates aren't atomic, the resetting is done twice + * to make sure that the counters are very likely to be all cleared. + */ +static ssize_t qstat_write(struct file *file, const char __user *user_buf, + size_t count, loff_t *ppos) +{ + int cpu; + + /* + * Get the counter ID stored in file->f_inode->i_private + */ + if (!file->f_inode) { + WARN_ON_ONCE(1); + return -EBADF; + } + if ((long)(file->f_inode->i_private) != qstat_reset_cnts) + return count; + + for_each_possible_cpu(cpu) { + int i; + unsigned long *ptr = per_cpu_ptr(qstats, cpu); + + for (i = 0 ; i < qstat_num; i++) + WRITE_ONCE(ptr[i], 0); + for (i = 0 ; i < qstat_num; i++) + WRITE_ONCE(ptr[i], 0); + } + return count; +} + +/* + * Debugfs data structures + */ +static const struct file_operations fops_qstat = { + .read = qstat_read, + .write = qstat_write, + .llseek = default_llseek, +}; + +/* + * Initialize debugfs for the qspinlock statistical counters + */ +static int __init init_qspinlock_stat(void) +{ + struct dentry *d_qstat = debugfs_create_dir("qlockstat", NULL); + int i; + + if (!d_qstat) { + pr_warn("Could not create 'qlockstat' debugfs directory\n"); + return 0; + } + + /* + * Create the debugfs files + * + * As reading from and writing to the stat files can be slow, only + * root is allowed to do the read/write to limit impact to system + * performance. + */ + for (i = 0; i < qstat_num; i++) + debugfs_create_file(qstat_names[i], 0400, d_qstat, + (void *)(long)i, &fops_qstat); + + debugfs_create_file(qstat_names[qstat_reset_cnts], 0200, d_qstat, + (void *)(long)qstat_reset_cnts, &fops_qstat); + return 0; +} +fs_initcall(init_qspinlock_stat); + +/* + * Increment the PV qspinlock statistical counters + */ +static inline void qstat_inc(enum qlock_stats stat, bool cond) +{ + if (cond) + this_cpu_inc(qstats[stat]); +} + +/* + * PV hash hop count + */ +static inline void qstat_hop(int hopcnt) +{ + this_cpu_add(qstats[qstat_pv_hash_hops], hopcnt); +} + +/* + * Replacement function for pv_kick() + */ +static inline void __pv_kick(int cpu) +{ + u64 start = sched_clock(); + + per_cpu(pv_kick_time, cpu) = start; + pv_kick(cpu); + this_cpu_add(qstats[qstat_pv_latency_kick], sched_clock() - start); +} + +/* + * Replacement function for pv_wait() + */ +static inline void __pv_wait(u8 *ptr, u8 val) +{ + u64 *pkick_time = this_cpu_ptr(&pv_kick_time); + + *pkick_time = 0; + pv_wait(ptr, val); + if (*pkick_time) { + this_cpu_add(qstats[qstat_pv_latency_wake], + sched_clock() - *pkick_time); + qstat_inc(qstat_pv_kick_wake, true); + } +} + +#define pv_kick(c) __pv_kick(c) +#define pv_wait(p, v) __pv_wait(p, v) + +#else /* CONFIG_QUEUED_LOCK_STAT */ + +static inline void qstat_inc(enum qlock_stats stat, bool cond) { } +static inline void qstat_hop(int hopcnt) { } + +#endif /* CONFIG_QUEUED_LOCK_STAT */ -- cgit From 1c4941fd53afb46ab15826628e4819866d008a28 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Tue, 10 Nov 2015 16:18:56 -0500 Subject: locking/pvqspinlock: Allow limited lock stealing This patch allows one attempt for the lock waiter to steal the lock when entering the PV slowpath. To prevent lock starvation, the pending bit will be set by the queue head vCPU when it is in the active lock spinning loop to disable any lock stealing attempt. This helps to reduce the performance penalty caused by lock waiter preemption while not having much of the downsides of a real unfair lock. The pv_wait_head() function was renamed as pv_wait_head_or_lock() as it was modified to acquire the lock before returning. This is necessary because of possible lock stealing attempts from other tasks. Linux kernel builds were run in KVM guest on an 8-socket, 4 cores/socket Westmere-EX system and a 4-socket, 8 cores/socket Haswell-EX system. Both systems are configured to have 32 physical CPUs. The kernel build times before and after the patch were: Westmere Haswell Patch 32 vCPUs 48 vCPUs 32 vCPUs 48 vCPUs ----- -------- -------- -------- -------- Before patch 3m15.6s 10m56.1s 1m44.1s 5m29.1s After patch 3m02.3s 5m00.2s 1m43.7s 3m03.5s For the overcommited case (48 vCPUs), this patch is able to reduce kernel build time by more than 54% for Westmere and 44% for Haswell. Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447190336-53317-1-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock.c | 26 +++++-- kernel/locking/qspinlock_paravirt.h | 141 ++++++++++++++++++++++++++++++------ kernel/locking/qspinlock_stat.h | 16 ++++ 3 files changed, 155 insertions(+), 28 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c index ed9d96708f93..2ea42999d2d8 100644 --- a/kernel/locking/qspinlock.c +++ b/kernel/locking/qspinlock.c @@ -251,15 +251,16 @@ static __always_inline void __pv_init_node(struct mcs_spinlock *node) { } static __always_inline void __pv_wait_node(struct mcs_spinlock *node) { } static __always_inline void __pv_kick_node(struct qspinlock *lock, struct mcs_spinlock *node) { } -static __always_inline void __pv_wait_head(struct qspinlock *lock, - struct mcs_spinlock *node) { } +static __always_inline u32 __pv_wait_head_or_lock(struct qspinlock *lock, + struct mcs_spinlock *node) + { return 0; } #define pv_enabled() false #define pv_init_node __pv_init_node #define pv_wait_node __pv_wait_node #define pv_kick_node __pv_kick_node -#define pv_wait_head __pv_wait_head +#define pv_wait_head_or_lock __pv_wait_head_or_lock #ifdef CONFIG_PARAVIRT_SPINLOCKS #define queued_spin_lock_slowpath native_queued_spin_lock_slowpath @@ -431,10 +432,22 @@ queue: * sequentiality; this is because the set_locked() function below * does not imply a full barrier. * + * The PV pv_wait_head_or_lock function, if active, will acquire + * the lock and return a non-zero value. So we have to skip the + * smp_load_acquire() call. As the next PV queue head hasn't been + * designated yet, there is no way for the locked value to become + * _Q_SLOW_VAL. So both the set_locked() and the + * atomic_cmpxchg_relaxed() calls will be safe. + * + * If PV isn't active, 0 will be returned instead. + * */ - pv_wait_head(lock, node); + if ((val = pv_wait_head_or_lock(lock, node))) + goto locked; + smp_cond_acquire(!((val = atomic_read(&lock->val)) & _Q_LOCKED_PENDING_MASK)); +locked: /* * claim the lock: * @@ -446,7 +459,8 @@ queue: * to grab the lock. */ for (;;) { - if (val != tail) { + /* In the PV case we might already have _Q_LOCKED_VAL set */ + if ((val & _Q_TAIL_MASK) != tail) { set_locked(lock); break; } @@ -493,7 +507,7 @@ EXPORT_SYMBOL(queued_spin_lock_slowpath); #undef pv_init_node #undef pv_wait_node #undef pv_kick_node -#undef pv_wait_head +#undef pv_wait_head_or_lock #undef queued_spin_lock_slowpath #define queued_spin_lock_slowpath __pv_queued_spin_lock_slowpath diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h index aaeeefb791f8..ace60a451b4f 100644 --- a/kernel/locking/qspinlock_paravirt.h +++ b/kernel/locking/qspinlock_paravirt.h @@ -40,6 +40,89 @@ struct pv_node { u8 state; }; +/* + * By replacing the regular queued_spin_trylock() with the function below, + * it will be called once when a lock waiter enter the PV slowpath before + * being queued. By allowing one lock stealing attempt here when the pending + * bit is off, it helps to reduce the performance impact of lock waiter + * preemption without the drawback of lock starvation. + */ +#define queued_spin_trylock(l) pv_queued_spin_steal_lock(l) +static inline bool pv_queued_spin_steal_lock(struct qspinlock *lock) +{ + struct __qspinlock *l = (void *)lock; + + return !(atomic_read(&lock->val) & _Q_LOCKED_PENDING_MASK) && + (cmpxchg(&l->locked, 0, _Q_LOCKED_VAL) == 0); +} + +/* + * The pending bit is used by the queue head vCPU to indicate that it + * is actively spinning on the lock and no lock stealing is allowed. + */ +#if _Q_PENDING_BITS == 8 +static __always_inline void set_pending(struct qspinlock *lock) +{ + struct __qspinlock *l = (void *)lock; + + WRITE_ONCE(l->pending, 1); +} + +static __always_inline void clear_pending(struct qspinlock *lock) +{ + struct __qspinlock *l = (void *)lock; + + WRITE_ONCE(l->pending, 0); +} + +/* + * The pending bit check in pv_queued_spin_steal_lock() isn't a memory + * barrier. Therefore, an atomic cmpxchg() is used to acquire the lock + * just to be sure that it will get it. + */ +static __always_inline int trylock_clear_pending(struct qspinlock *lock) +{ + struct __qspinlock *l = (void *)lock; + + return !READ_ONCE(l->locked) && + (cmpxchg(&l->locked_pending, _Q_PENDING_VAL, _Q_LOCKED_VAL) + == _Q_PENDING_VAL); +} +#else /* _Q_PENDING_BITS == 8 */ +static __always_inline void set_pending(struct qspinlock *lock) +{ + atomic_set_mask(_Q_PENDING_VAL, &lock->val); +} + +static __always_inline void clear_pending(struct qspinlock *lock) +{ + atomic_clear_mask(_Q_PENDING_VAL, &lock->val); +} + +static __always_inline int trylock_clear_pending(struct qspinlock *lock) +{ + int val = atomic_read(&lock->val); + + for (;;) { + int old, new; + + if (val & _Q_LOCKED_MASK) + break; + + /* + * Try to clear pending bit & set locked bit + */ + old = val; + new = (val & ~_Q_PENDING_MASK) | _Q_LOCKED_VAL; + val = atomic_cmpxchg(&lock->val, old, new); + + if (val == old) + return 1; + } + return 0; +} +#endif /* _Q_PENDING_BITS == 8 */ + /* * Include queued spinlock statistics code */ @@ -202,8 +285,8 @@ static void pv_wait_node(struct mcs_spinlock *node) /* * If pv_kick_node() changed us to vcpu_hashed, retain that - * value so that pv_wait_head() knows to not also try to hash - * this lock. + * value so that pv_wait_head_or_lock() knows to not also try + * to hash this lock. */ cmpxchg(&pn->state, vcpu_halted, vcpu_running); @@ -227,8 +310,9 @@ static void pv_wait_node(struct mcs_spinlock *node) /* * Called after setting next->locked = 1 when we're the lock owner. * - * Instead of waking the waiters stuck in pv_wait_node() advance their state such - * that they're waiting in pv_wait_head(), this avoids a wake/sleep cycle. + * Instead of waking the waiters stuck in pv_wait_node() advance their state + * such that they're waiting in pv_wait_head_or_lock(), this avoids a + * wake/sleep cycle. */ static void pv_kick_node(struct qspinlock *lock, struct mcs_spinlock *node) { @@ -257,10 +341,14 @@ static void pv_kick_node(struct qspinlock *lock, struct mcs_spinlock *node) } /* - * Wait for l->locked to become clear; halt the vcpu after a short spin. + * Wait for l->locked to become clear and acquire the lock; + * halt the vcpu after a short spin. * __pv_queued_spin_unlock() will wake us. + * + * The current value of the lock will be returned for additional processing. */ -static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) +static u32 +pv_wait_head_or_lock(struct qspinlock *lock, struct mcs_spinlock *node) { struct pv_node *pn = (struct pv_node *)node; struct __qspinlock *l = (void *)lock; @@ -276,11 +364,18 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) lp = (struct qspinlock **)1; for (;; waitcnt++) { + /* + * Set the pending bit in the active lock spinning loop to + * disable lock stealing before attempting to acquire the lock. + */ + set_pending(lock); for (loop = SPIN_THRESHOLD; loop; loop--) { - if (!READ_ONCE(l->locked)) - return; + if (trylock_clear_pending(lock)) + goto gotlock; cpu_relax(); } + clear_pending(lock); + if (!lp) { /* ONCE */ lp = pv_hash(lock, pn); @@ -296,36 +391,38 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node) * * Matches the smp_rmb() in __pv_queued_spin_unlock(). */ - if (!cmpxchg(&l->locked, _Q_LOCKED_VAL, _Q_SLOW_VAL)) { + if (xchg(&l->locked, _Q_SLOW_VAL) == 0) { /* - * The lock is free and _Q_SLOW_VAL has never - * been set. Therefore we need to unhash before - * getting the lock. + * The lock was free and now we own the lock. + * Change the lock value back to _Q_LOCKED_VAL + * and unhash the table. */ + WRITE_ONCE(l->locked, _Q_LOCKED_VAL); WRITE_ONCE(*lp, NULL); - return; + goto gotlock; } } qstat_inc(qstat_pv_wait_head, true); qstat_inc(qstat_pv_wait_again, waitcnt); pv_wait(&l->locked, _Q_SLOW_VAL); - if (!READ_ONCE(l->locked)) - return; /* * The unlocker should have freed the lock before kicking the * CPU. So if the lock is still not free, it is a spurious - * wakeup and so the vCPU should wait again after spinning for - * a while. + * wakeup or another vCPU has stolen the lock. The current + * vCPU should spin again. */ - qstat_inc(qstat_pv_spurious_wakeup, true); + qstat_inc(qstat_pv_spurious_wakeup, READ_ONCE(l->locked)); } /* - * Lock is unlocked now; the caller will acquire it without waiting. - * As with pv_wait_node() we rely on the caller to do a load-acquire - * for us. + * The cmpxchg() or xchg() call before coming here provides the + * acquire semantics for locking. The dummy ORing of _Q_LOCKED_VAL + * here is to indicate to the compiler that the value will always + * be nozero to enable better code optimization. */ +gotlock: + return (u32)(atomic_read(&lock->val) | _Q_LOCKED_VAL); } /* @@ -350,7 +447,7 @@ __pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked) * so we need a barrier to order the read of the node data in * pv_unhash *after* we've read the lock being _Q_SLOW_VAL. * - * Matches the cmpxchg() in pv_wait_head() setting _Q_SLOW_VAL. + * Matches the cmpxchg() in pv_wait_head_or_lock() setting _Q_SLOW_VAL. */ smp_rmb(); diff --git a/kernel/locking/qspinlock_stat.h b/kernel/locking/qspinlock_stat.h index b1553adec2e7..94d4533fe984 100644 --- a/kernel/locking/qspinlock_stat.h +++ b/kernel/locking/qspinlock_stat.h @@ -22,6 +22,7 @@ * pv_kick_wake - # of vCPU kicks used for computing pv_latency_wake * pv_latency_kick - average latency (ns) of vCPU kick operation * pv_latency_wake - average latency (ns) from vCPU kick to wakeup + * pv_lock_stealing - # of lock stealing operations * pv_spurious_wakeup - # of spurious wakeups * pv_wait_again - # of vCPU wait's that happened after a vCPU kick * pv_wait_head - # of vCPU wait's at the queue head @@ -43,6 +44,7 @@ enum qlock_stats { qstat_pv_kick_wake, qstat_pv_latency_kick, qstat_pv_latency_wake, + qstat_pv_lock_stealing, qstat_pv_spurious_wakeup, qstat_pv_wait_again, qstat_pv_wait_head, @@ -66,6 +68,7 @@ static const char * const qstat_names[qstat_num + 1] = { [qstat_pv_spurious_wakeup] = "pv_spurious_wakeup", [qstat_pv_latency_kick] = "pv_latency_kick", [qstat_pv_latency_wake] = "pv_latency_wake", + [qstat_pv_lock_stealing] = "pv_lock_stealing", [qstat_pv_wait_again] = "pv_wait_again", [qstat_pv_wait_head] = "pv_wait_head", [qstat_pv_wait_node] = "pv_wait_node", @@ -273,6 +276,19 @@ static inline void __pv_wait(u8 *ptr, u8 val) #define pv_kick(c) __pv_kick(c) #define pv_wait(p, v) __pv_wait(p, v) +/* + * PV unfair trylock count tracking function + */ +static inline int qstat_spin_steal_lock(struct qspinlock *lock) +{ + int ret = pv_queued_spin_steal_lock(lock); + + qstat_inc(qstat_pv_lock_stealing, ret); + return ret; +} +#undef queued_spin_trylock +#define queued_spin_trylock(l) qstat_spin_steal_lock(l) + #else /* CONFIG_QUEUED_LOCK_STAT */ static inline void qstat_inc(enum qlock_stats stat, bool cond) { } -- cgit From cd0272fab785077c121aa91ec2401090965bbc37 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Mon, 9 Nov 2015 19:09:27 -0500 Subject: locking/pvqspinlock: Queue node adaptive spinning In an overcommitted guest where some vCPUs have to be halted to make forward progress in other areas, it is highly likely that a vCPU later in the spinlock queue will be spinning while the ones earlier in the queue would have been halted. The spinning in the later vCPUs is then just a waste of precious CPU cycles because they are not going to get the lock soon as the earlier ones have to be woken up and take their turn to get the lock. This patch implements an adaptive spinning mechanism where the vCPU will call pv_wait() if the previous vCPU is not running. Linux kernel builds were run in KVM guest on an 8-socket, 4 cores/socket Westmere-EX system and a 4-socket, 8 cores/socket Haswell-EX system. Both systems are configured to have 32 physical CPUs. The kernel build times before and after the patch were: Westmere Haswell Patch 32 vCPUs 48 vCPUs 32 vCPUs 48 vCPUs ----- -------- -------- -------- -------- Before patch 3m02.3s 5m00.2s 1m43.7s 3m03.5s After patch 3m03.0s 4m37.5s 1m43.0s 2m47.2s For 32 vCPUs, this patch doesn't cause any noticeable change in performance. For 48 vCPUs (over-committed), there is about 8% performance improvement. Signed-off-by: Waiman Long Signed-off-by: Peter Zijlstra (Intel) Cc: Andrew Morton Cc: Davidlohr Bueso Cc: Douglas Hatch Cc: H. Peter Anvin Cc: Linus Torvalds Cc: Paul E. McKenney Cc: Peter Zijlstra Cc: Scott J Norton Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1447114167-47185-8-git-send-email-Waiman.Long@hpe.com Signed-off-by: Ingo Molnar --- kernel/locking/qspinlock.c | 5 ++-- kernel/locking/qspinlock_paravirt.h | 46 +++++++++++++++++++++++++++++++++++-- kernel/locking/qspinlock_stat.h | 3 +++ 3 files changed, 50 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c index 2ea42999d2d8..393d1874b9e0 100644 --- a/kernel/locking/qspinlock.c +++ b/kernel/locking/qspinlock.c @@ -248,7 +248,8 @@ static __always_inline void set_locked(struct qspinlock *lock) */ static __always_inline void __pv_init_node(struct mcs_spinlock *node) { } -static __always_inline void __pv_wait_node(struct mcs_spinlock *node) { } +static __always_inline void __pv_wait_node(struct mcs_spinlock *node, + struct mcs_spinlock *prev) { } static __always_inline void __pv_kick_node(struct qspinlock *lock, struct mcs_spinlock *node) { } static __always_inline u32 __pv_wait_head_or_lock(struct qspinlock *lock, @@ -407,7 +408,7 @@ queue: prev = decode_tail(old); WRITE_ONCE(prev->next, node); - pv_wait_node(node); + pv_wait_node(node, prev); arch_mcs_spin_lock_contended(&node->locked); /* diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h index ace60a451b4f..87bb235c3448 100644 --- a/kernel/locking/qspinlock_paravirt.h +++ b/kernel/locking/qspinlock_paravirt.h @@ -22,6 +22,20 @@ #define _Q_SLOW_VAL (3U << _Q_LOCKED_OFFSET) +/* + * Queue Node Adaptive Spinning + * + * A queue node vCPU will stop spinning if the vCPU in the previous node is + * not running. The one lock stealing attempt allowed at slowpath entry + * mitigates the slight slowdown for non-overcommitted guest with this + * aggressive wait-early mechanism. + * + * The status of the previous node will be checked at fixed interval + * controlled by PV_PREV_CHECK_MASK. This is to ensure that we won't + * pound on the cacheline of the previous node too heavily. + */ +#define PV_PREV_CHECK_MASK 0xff + /* * Queue node uses: vcpu_running & vcpu_halted. * Queue head uses: vcpu_running & vcpu_hashed. @@ -234,6 +248,20 @@ static struct pv_node *pv_unhash(struct qspinlock *lock) BUG(); } +/* + * Return true if when it is time to check the previous node which is not + * in a running state. + */ +static inline bool +pv_wait_early(struct pv_node *prev, int loop) +{ + + if ((loop & PV_PREV_CHECK_MASK) != 0) + return false; + + return READ_ONCE(prev->state) != vcpu_running; +} + /* * Initialize the PV part of the mcs_spinlock node. */ @@ -252,17 +280,23 @@ static void pv_init_node(struct mcs_spinlock *node) * pv_kick_node() is used to set _Q_SLOW_VAL and fill in hash table on its * behalf. */ -static void pv_wait_node(struct mcs_spinlock *node) +static void pv_wait_node(struct mcs_spinlock *node, struct mcs_spinlock *prev) { struct pv_node *pn = (struct pv_node *)node; + struct pv_node *pp = (struct pv_node *)prev; int waitcnt = 0; int loop; + bool wait_early; /* waitcnt processing will be compiled out if !QUEUED_LOCK_STAT */ for (;; waitcnt++) { - for (loop = SPIN_THRESHOLD; loop; loop--) { + for (wait_early = false, loop = SPIN_THRESHOLD; loop; loop--) { if (READ_ONCE(node->locked)) return; + if (pv_wait_early(pp, loop)) { + wait_early = true; + break; + } cpu_relax(); } @@ -280,6 +314,7 @@ static void pv_wait_node(struct mcs_spinlock *node) if (!READ_ONCE(node->locked)) { qstat_inc(qstat_pv_wait_node, true); qstat_inc(qstat_pv_wait_again, waitcnt); + qstat_inc(qstat_pv_wait_early, wait_early); pv_wait(&pn->state, vcpu_halted); } @@ -364,6 +399,12 @@ pv_wait_head_or_lock(struct qspinlock *lock, struct mcs_spinlock *node) lp = (struct qspinlock **)1; for (;; waitcnt++) { + /* + * Set correct vCPU state to be used by queue node wait-early + * mechanism. + */ + WRITE_ONCE(pn->state, vcpu_running); + /* * Set the pending bit in the active lock spinning loop to * disable lock stealing before attempting to acquire the lock. @@ -402,6 +443,7 @@ pv_wait_head_or_lock(struct qspinlock *lock, struct mcs_spinlock *node) goto gotlock; } } + WRITE_ONCE(pn->state, vcpu_halted); qstat_inc(qstat_pv_wait_head, true); qstat_inc(qstat_pv_wait_again, waitcnt); pv_wait(&l->locked, _Q_SLOW_VAL); diff --git a/kernel/locking/qspinlock_stat.h b/kernel/locking/qspinlock_stat.h index 94d4533fe984..640dcecdd1df 100644 --- a/kernel/locking/qspinlock_stat.h +++ b/kernel/locking/qspinlock_stat.h @@ -25,6 +25,7 @@ * pv_lock_stealing - # of lock stealing operations * pv_spurious_wakeup - # of spurious wakeups * pv_wait_again - # of vCPU wait's that happened after a vCPU kick + * pv_wait_early - # of early vCPU wait's * pv_wait_head - # of vCPU wait's at the queue head * pv_wait_node - # of vCPU wait's at a non-head queue node * @@ -47,6 +48,7 @@ enum qlock_stats { qstat_pv_lock_stealing, qstat_pv_spurious_wakeup, qstat_pv_wait_again, + qstat_pv_wait_early, qstat_pv_wait_head, qstat_pv_wait_node, qstat_num, /* Total number of statistical counters */ @@ -70,6 +72,7 @@ static const char * const qstat_names[qstat_num + 1] = { [qstat_pv_latency_wake] = "pv_latency_wake", [qstat_pv_lock_stealing] = "pv_lock_stealing", [qstat_pv_wait_again] = "pv_wait_again", + [qstat_pv_wait_early] = "pv_wait_early", [qstat_pv_wait_head] = "pv_wait_head", [qstat_pv_wait_node] = "pv_wait_node", [qstat_reset_cnts] = "reset_counters", -- cgit From 06f60de19d3141f07d954c9275fe7ccca8e96b42 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 08:15:52 -0700 Subject: rcu: Short-circuit synchronize_sched_expedited() if only one CPU If there is only one CPU, then invoking synchronize_sched_expedited() is by definition a grace period. This commit checks for this condition and does a short-circuit return in that case. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 81aa1cdc6bc9..bd2605c144cc 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3822,6 +3822,10 @@ void synchronize_sched_expedited(void) struct rcu_node *rnp; struct rcu_state *rsp = &rcu_sched_state; + /* If only one CPU, this is automatically a grace period. */ + if (rcu_blocking_is_gp()) + return; + /* Take a snapshot of the sequence number. */ s = rcu_exp_gp_seq_snap(rsp); -- cgit From 1de6e56ddc043437d335ee0455a1b34b73510c91 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 09:45:00 -0700 Subject: rcu: Clarify role of ->expmaskinitnext Analogy with the ->qsmaskinitnext field might lead one to believe that ->expmaskinitnext tracks online CPUs. This belief is incorrect: Any CPU that has ever been online will have its bit set in the ->expmaskinitnext field. This commit therefore adds a comment to make this clear, at least to people who read comments. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel') diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index f32bebb6bc90..8151971a8978 100644 --- a/kernel/rcu/tree.h +++ b/kernel/rcu/tree.h @@ -178,6 +178,8 @@ struct rcu_node { /* beginning of each expedited GP. */ unsigned long expmaskinitnext; /* Online CPUs for next expedited GP. */ + /* Any CPU that has ever been online will */ + /* have its bit set. */ unsigned long grpmask; /* Mask to apply to parent qsmask. */ /* Only one bit will be set in this mask. */ int grplo; /* lowest-numbered CPU or group here. */ -- cgit From 886ef5a18a4a771d5fdc0e23ae9373bb35d529e7 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 12:34:40 -0700 Subject: rcu: Move smp_mb() from rcu_seq_snap() to rcu_exp_gp_seq_snap() The memory barrier in rcu_seq_snap() is needed only for grace periods, so this commit moves it to the grace-period-oriented wrapper rcu_exp_gp_seq_snap(). Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index bd2605c144cc..a4a0475aede9 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3347,7 +3347,6 @@ static unsigned long rcu_seq_snap(unsigned long *sp) { unsigned long s; - smp_mb(); /* Caller's modifications seen first by other CPUs. */ s = (READ_ONCE(*sp) + 3) & ~0x1; smp_mb(); /* Above access must not bleed into critical section. */ return s; @@ -3374,6 +3373,7 @@ static void rcu_exp_gp_seq_end(struct rcu_state *rsp) } static unsigned long rcu_exp_gp_seq_snap(struct rcu_state *rsp) { + smp_mb(); /* Caller's modifications seen first by other CPUs. */ return rcu_seq_snap(&rsp->expedited_sequence); } static bool rcu_exp_gp_seq_done(struct rcu_state *rsp, unsigned long s) -- cgit From 1307f2148719cc9e9d12f5fa7d5b3b61ec5aef72 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 15:29:21 -0700 Subject: rcu: Invert sync_rcu_exp_select_cpus() "if" statement This commit saves a couple lines of code and reduces indentation by inverting the sense of an "if" statement in the function sync_rcu_exp_select_cpus(). Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 30 ++++++++++++++---------------- 1 file changed, 14 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index a4a0475aede9..00f07d6436ce 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3716,24 +3716,22 @@ retry_ipi: ret = smp_call_function_single(cpu, func, rsp, 0); if (!ret) { mask_ofl_ipi &= ~mask; - } else { - /* Failed, raced with offline. */ - raw_spin_lock_irqsave_rcu_node(rnp, flags); - if (cpu_online(cpu) && - (rnp->expmask & mask)) { - raw_spin_unlock_irqrestore(&rnp->lock, - flags); - schedule_timeout_uninterruptible(1); - if (cpu_online(cpu) && - (rnp->expmask & mask)) - goto retry_ipi; - raw_spin_lock_irqsave_rcu_node(rnp, - flags); - } - if (!(rnp->expmask & mask)) - mask_ofl_ipi &= ~mask; + continue; + } + /* Failed, raced with offline. */ + raw_spin_lock_irqsave_rcu_node(rnp, flags); + if (cpu_online(cpu) && + (rnp->expmask & mask)) { raw_spin_unlock_irqrestore(&rnp->lock, flags); + schedule_timeout_uninterruptible(1); + if (cpu_online(cpu) && + (rnp->expmask & mask)) + goto retry_ipi; + raw_spin_lock_irqsave_rcu_node(rnp, flags); } + if (!(rnp->expmask & mask)) + mask_ofl_ipi &= ~mask; + raw_spin_unlock_irqrestore(&rnp->lock, flags); } /* Report quiescent states for those that went offline. */ mask_ofl_test |= mask_ofl_ipi; -- cgit From df5bd5144a80a9f6c3807383b11f735dae9caf9d Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Thu, 1 Oct 2015 10:26:24 -0700 Subject: rcu: Reduce expedited GP memory contention via per-CPU variables Currently, the piggybacked-work checks carried out by sync_exp_work_done() atomically increment a small set of variables (the ->expedited_workdone0, ->expedited_workdone1, ->expedited_workdone2, ->expedited_workdone3 fields in the rcu_state structure), which will form a memory-contention bottleneck given a sufficiently large number of CPUs concurrently invoking either synchronize_rcu_expedited() or synchronize_sched_expedited(). This commit therefore moves these for fields to the per-CPU rcu_data structure, eliminating the memory contention. The show_rcuexp() function also changes to sum up each field in the rcu_data structures. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 11 +++++------ kernel/rcu/tree.h | 8 ++++---- kernel/rcu/tree_trace.c | 18 ++++++++++++------ 3 files changed, 21 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 00f07d6436ce..33d7e2551165 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3585,7 +3585,7 @@ static bool sync_exp_work_done(struct rcu_state *rsp, struct rcu_node *rnp, */ static struct rcu_node *exp_funnel_lock(struct rcu_state *rsp, unsigned long s) { - struct rcu_data *rdp; + struct rcu_data *rdp = per_cpu_ptr(rsp->rda, raw_smp_processor_id()); struct rcu_node *rnp0; struct rcu_node *rnp1 = NULL; @@ -3599,7 +3599,7 @@ static struct rcu_node *exp_funnel_lock(struct rcu_state *rsp, unsigned long s) if (!mutex_is_locked(&rnp0->exp_funnel_mutex)) { if (mutex_trylock(&rnp0->exp_funnel_mutex)) { if (sync_exp_work_done(rsp, rnp0, NULL, - &rsp->expedited_workdone0, s)) + &rdp->expedited_workdone0, s)) return NULL; return rnp0; } @@ -3613,14 +3613,13 @@ static struct rcu_node *exp_funnel_lock(struct rcu_state *rsp, unsigned long s) * can be inexact, as it is just promoting locality and is not * strictly needed for correctness. */ - rdp = per_cpu_ptr(rsp->rda, raw_smp_processor_id()); - if (sync_exp_work_done(rsp, NULL, NULL, &rsp->expedited_workdone1, s)) + if (sync_exp_work_done(rsp, NULL, NULL, &rdp->expedited_workdone1, s)) return NULL; mutex_lock(&rdp->exp_funnel_mutex); rnp0 = rdp->mynode; for (; rnp0 != NULL; rnp0 = rnp0->parent) { if (sync_exp_work_done(rsp, rnp1, rdp, - &rsp->expedited_workdone2, s)) + &rdp->expedited_workdone2, s)) return NULL; mutex_lock(&rnp0->exp_funnel_mutex); if (rnp1) @@ -3630,7 +3629,7 @@ static struct rcu_node *exp_funnel_lock(struct rcu_state *rsp, unsigned long s) rnp1 = rnp0; } if (sync_exp_work_done(rsp, rnp1, rdp, - &rsp->expedited_workdone3, s)) + &rdp->expedited_workdone3, s)) return NULL; return rnp1; } diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index 8151971a8978..6cbec31f99d6 100644 --- a/kernel/rcu/tree.h +++ b/kernel/rcu/tree.h @@ -386,6 +386,10 @@ struct rcu_data { struct rcu_head oom_head; #endif /* #ifdef CONFIG_RCU_FAST_NO_HZ */ struct mutex exp_funnel_mutex; + atomic_long_t expedited_workdone0; /* # done by others #0. */ + atomic_long_t expedited_workdone1; /* # done by others #1. */ + atomic_long_t expedited_workdone2; /* # done by others #2. */ + atomic_long_t expedited_workdone3; /* # done by others #3. */ /* 7) Callback offloading. */ #ifdef CONFIG_RCU_NOCB_CPU @@ -500,10 +504,6 @@ struct rcu_state { /* End of fields guarded by barrier_mutex. */ unsigned long expedited_sequence; /* Take a ticket. */ - atomic_long_t expedited_workdone0; /* # done by others #0. */ - atomic_long_t expedited_workdone1; /* # done by others #1. */ - atomic_long_t expedited_workdone2; /* # done by others #2. */ - atomic_long_t expedited_workdone3; /* # done by others #3. */ atomic_long_t expedited_normal; /* # fallbacks to normal. */ atomic_t expedited_need_qs; /* # CPUs left to check in. */ wait_queue_head_t expedited_wq; /* Wait for check-ins. */ diff --git a/kernel/rcu/tree_trace.c b/kernel/rcu/tree_trace.c index 8efaba870d96..d43649450ea4 100644 --- a/kernel/rcu/tree_trace.c +++ b/kernel/rcu/tree_trace.c @@ -183,14 +183,20 @@ static const struct file_operations rcudata_fops = { static int show_rcuexp(struct seq_file *m, void *v) { + int cpu; struct rcu_state *rsp = (struct rcu_state *)m->private; - + struct rcu_data *rdp; + unsigned long s0 = 0, s1 = 0, s2 = 0, s3 = 0; + + for_each_possible_cpu(cpu) { + rdp = per_cpu_ptr(rsp->rda, cpu); + s0 += atomic_long_read(&rdp->expedited_workdone0); + s1 += atomic_long_read(&rdp->expedited_workdone1); + s2 += atomic_long_read(&rdp->expedited_workdone2); + s3 += atomic_long_read(&rdp->expedited_workdone3); + } seq_printf(m, "s=%lu wd0=%lu wd1=%lu wd2=%lu wd3=%lu n=%lu enq=%d sc=%lu\n", - rsp->expedited_sequence, - atomic_long_read(&rsp->expedited_workdone0), - atomic_long_read(&rsp->expedited_workdone1), - atomic_long_read(&rsp->expedited_workdone2), - atomic_long_read(&rsp->expedited_workdone3), + rsp->expedited_sequence, s0, s1, s2, s3, atomic_long_read(&rsp->expedited_normal), atomic_read(&rsp->expedited_need_qs), rsp->expedited_sequence / 2); -- cgit From 73f36f9de8bed78bcda2704a348594c20518b455 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 17 Nov 2015 10:56:55 -0800 Subject: rcu: Make expedited grace periods resolve stall-warning ties Currently, if a grace period ends just as the stall-warning timeout fires, an empty stall warning will be printed. This is not helpful, so this commit avoids these useless warnings by rechecking completion after awakening in synchronize_sched_expedited_wait(). Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 33d7e2551165..bc6b79716a86 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3757,7 +3757,7 @@ static void synchronize_sched_expedited_wait(struct rcu_state *rsp) rsp->expedited_wq, sync_rcu_preempt_exp_done(rnp_root), jiffies_stall); - if (ret > 0) + if (ret > 0 || sync_rcu_preempt_exp_done(rnp_root)) return; if (ret < 0) { /* Hit a signal, disable CPU stall warnings. */ -- cgit From 72611ab9f5d2d384a04e72d560c9c82463115cbf Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 17 Nov 2015 13:25:21 -0800 Subject: rcu: Add more diagnostics to expedited stall warning messages. This commit adds print statements that check the rcu_node structure to find which ->expmask bits and which ->exp_tasks structures are blocking the current expedited grace period. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 24 +++++++++++++++++++++--- 1 file changed, 21 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index bc6b79716a86..6a652d1f3d7f 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3745,6 +3745,7 @@ static void synchronize_sched_expedited_wait(struct rcu_state *rsp) unsigned long jiffies_stall; unsigned long jiffies_start; unsigned long mask; + int ndetected; struct rcu_node *rnp; struct rcu_node *rnp_root = rcu_get_root(rsp); int ret; @@ -3767,14 +3768,16 @@ static void synchronize_sched_expedited_wait(struct rcu_state *rsp) } pr_err("INFO: %s detected expedited stalls on CPUs/tasks: {", rsp->name); + ndetected = 0; rcu_for_each_leaf_node(rsp, rnp) { - (void)rcu_print_task_exp_stall(rnp); + ndetected = rcu_print_task_exp_stall(rnp); mask = 1; for (cpu = rnp->grplo; cpu <= rnp->grphi; cpu++, mask <<= 1) { struct rcu_data *rdp; if (!(rnp->expmask & mask)) continue; + ndetected++; rdp = per_cpu_ptr(rsp->rda, cpu); pr_cont(" %d-%c%c%c", cpu, "O."[cpu_online(cpu)], @@ -3783,8 +3786,23 @@ static void synchronize_sched_expedited_wait(struct rcu_state *rsp) } mask <<= 1; } - pr_cont(" } %lu jiffies s: %lu\n", - jiffies - jiffies_start, rsp->expedited_sequence); + pr_cont(" } %lu jiffies s: %lu root: %#lx/%c\n", + jiffies - jiffies_start, rsp->expedited_sequence, + rnp_root->expmask, ".T"[!!rnp_root->exp_tasks]); + if (!ndetected) { + pr_err("blocking rcu_node structures:"); + rcu_for_each_node_breadth_first(rsp, rnp) { + if (rnp == rnp_root) + continue; /* printed unconditionally */ + if (sync_rcu_preempt_exp_done(rnp)) + continue; + pr_cont(" l=%u:%d-%d:%#lx/%c", + rnp->level, rnp->grplo, rnp->grphi, + rnp->expmask, + ".T"[!!rnp->exp_tasks]); + } + pr_cont("\n"); + } rcu_for_each_leaf_node(rsp, rnp) { mask = 1; for (cpu = rnp->grplo; cpu <= rnp->grphi; cpu++, mask <<= 1) { -- cgit From 5a9be7c628c5273f84abacebf7faf2488376e0f0 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 24 Nov 2015 15:44:06 -0800 Subject: rcu: Add rcu_normal kernel parameter to suppress expediting Although expedited grace periods can be quite useful, and although their OS jitter has been greatly reduced, they can still pose problems for extreme real-time workloads. This commit therefore adds a rcu_normal kernel boot parameter (which can also be manipulated via sysfs) to suppress expedited grace periods, that is, to treat requests for expedited grace periods as if they were requests for normal grace periods. If both rcu_expedited and rcu_normal are specified, rcu_normal wins. This means that if you are relying on expedited grace periods to speed up boot, you will want to specify rcu_expedited on the kernel command line, and then specify rcu_normal via sysfs once boot completes. Signed-off-by: Paul E. McKenney --- kernel/ksysfs.c | 22 ++++++++++++++++++++-- kernel/rcu/srcu.c | 2 +- kernel/rcu/tree.c | 6 ++++++ kernel/rcu/tree_plugin.h | 6 ++++++ kernel/rcu/update.c | 12 ++++++++++++ 5 files changed, 45 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/ksysfs.c b/kernel/ksysfs.c index e83b26464061..b4e2fa52d8bc 100644 --- a/kernel/ksysfs.c +++ b/kernel/ksysfs.c @@ -20,7 +20,7 @@ #include #include -#include /* rcu_expedited */ +#include /* rcu_expedited and rcu_normal */ #define KERNEL_ATTR_RO(_name) \ static struct kobj_attribute _name##_attr = __ATTR_RO(_name) @@ -148,7 +148,7 @@ int rcu_expedited; static ssize_t rcu_expedited_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { - return sprintf(buf, "%d\n", rcu_expedited); + return sprintf(buf, "%d\n", READ_ONCE(rcu_expedited)); } static ssize_t rcu_expedited_store(struct kobject *kobj, struct kobj_attribute *attr, @@ -161,6 +161,23 @@ static ssize_t rcu_expedited_store(struct kobject *kobj, } KERNEL_ATTR_RW(rcu_expedited); +int rcu_normal; +static ssize_t rcu_normal_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + return sprintf(buf, "%d\n", READ_ONCE(rcu_normal)); +} +static ssize_t rcu_normal_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + if (kstrtoint(buf, 0, &rcu_normal)) + return -EINVAL; + + return count; +} +KERNEL_ATTR_RW(rcu_normal); + /* * Make /sys/kernel/notes give the raw contents of our kernel .notes section. */ @@ -203,6 +220,7 @@ static struct attribute * kernel_attrs[] = { &vmcoreinfo_attr.attr, #endif &rcu_expedited_attr.attr, + &rcu_normal_attr.attr, NULL }; diff --git a/kernel/rcu/srcu.c b/kernel/rcu/srcu.c index a63a1ea5a41b..9b9cdd549caa 100644 --- a/kernel/rcu/srcu.c +++ b/kernel/rcu/srcu.c @@ -489,7 +489,7 @@ static void __synchronize_srcu(struct srcu_struct *sp, int trycount) */ void synchronize_srcu(struct srcu_struct *sp) { - __synchronize_srcu(sp, rcu_gp_is_expedited() + __synchronize_srcu(sp, (rcu_gp_is_expedited() && !rcu_gp_is_normal()) ? SYNCHRONIZE_SRCU_EXP_TRYCOUNT : SYNCHRONIZE_SRCU_TRYCOUNT); } diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 6a652d1f3d7f..489992997c06 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -3841,6 +3841,12 @@ void synchronize_sched_expedited(void) if (rcu_blocking_is_gp()) return; + /* If expedited grace periods are prohibited, fall back to normal. */ + if (rcu_gp_is_normal()) { + wait_rcu_gp(call_rcu_sched); + return; + } + /* Take a snapshot of the sequence number. */ s = rcu_exp_gp_seq_snap(rsp); diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index 57ba873d2f18..d45df3781551 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -746,6 +746,12 @@ void synchronize_rcu_expedited(void) struct rcu_state *rsp = rcu_state_p; unsigned long s; + /* If expedited grace periods are prohibited, fall back to normal. */ + if (rcu_gp_is_normal()) { + wait_rcu_gp(call_rcu); + return; + } + s = rcu_exp_gp_seq_snap(rsp); rnp_unlock = exp_funnel_lock(rsp, s); diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c index 5f748c5a40f0..8fccda3a794d 100644 --- a/kernel/rcu/update.c +++ b/kernel/rcu/update.c @@ -61,6 +61,7 @@ MODULE_ALIAS("rcupdate"); #define MODULE_PARAM_PREFIX "rcupdate." module_param(rcu_expedited, int, 0); +module_param(rcu_normal, int, 0); #if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_PREEMPT_COUNT) /** @@ -113,6 +114,17 @@ EXPORT_SYMBOL(rcu_read_lock_sched_held); #ifndef CONFIG_TINY_RCU +/* + * Should expedited grace-period primitives always fall back to their + * non-expedited counterparts? Intended for use within RCU. Note + * that if the user specifies both rcu_expedited and rcu_normal, then + * rcu_normal wins. + */ +bool rcu_gp_is_normal(void) +{ + return READ_ONCE(rcu_normal); +} + static atomic_t rcu_expedited_nesting = ATOMIC_INIT(IS_ENABLED(CONFIG_RCU_EXPEDITE_BOOT) ? 1 : 0); -- cgit From 3e42ec1aa716f10c68294b8492ae3ea684528699 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Wed, 25 Nov 2015 18:56:00 -0800 Subject: rcu: Allow expedited grace periods to be disabled at init Expedited grace periods can speed up boot, but are undesirable in aggressive real-time systems. This commit therefore introduces a kernel parameter rcupdate.rcu_normal_after_boot that disables expedited grace periods just before init is spawned. Signed-off-by: Paul E. McKenney --- kernel/rcu/update.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'kernel') diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c index 8fccda3a794d..12b91f5a60a6 100644 --- a/kernel/rcu/update.c +++ b/kernel/rcu/update.c @@ -63,6 +63,9 @@ MODULE_ALIAS("rcupdate"); module_param(rcu_expedited, int, 0); module_param(rcu_normal, int, 0); +static int rcu_normal_after_boot; +module_param(rcu_normal_after_boot, int, 0); + #if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_PREEMPT_COUNT) /** * rcu_read_lock_sched_held() - might we be in RCU-sched read-side critical section? @@ -178,6 +181,8 @@ void rcu_end_inkernel_boot(void) { if (IS_ENABLED(CONFIG_RCU_EXPEDITE_BOOT)) rcu_unexpedite_gp(); + if (rcu_normal_after_boot) + WRITE_ONCE(rcu_normal, 1); } #ifdef CONFIG_PREEMPT_RCU -- cgit From 3dc5dbe9a1b815b659a6b04540fc6fd4b4e3831b Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Sat, 26 Sep 2015 14:51:24 -0700 Subject: rcu: Move lock_class_key to local scope Currently, the rcu_node_class[], rcu_fqs_class[], and rcu_exp_class[] arrays needlessly pollute the global namespace within tree.c. This commit therefore converts them to static local variables within rcu_init_one(). Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 81aa1cdc6bc9..23df2661c899 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -68,10 +68,6 @@ MODULE_ALIAS("rcutree"); /* Data structures. */ -static struct lock_class_key rcu_node_class[RCU_NUM_LVLS]; -static struct lock_class_key rcu_fqs_class[RCU_NUM_LVLS]; -static struct lock_class_key rcu_exp_class[RCU_NUM_LVLS]; - /* * In order to export the rcu_state name to the tracing tools, it * needs to be added in the __tracepoint_string section. @@ -4365,6 +4361,9 @@ static void __init rcu_init_one(struct rcu_state *rsp, static const char * const buf[] = RCU_NODE_NAME_INIT; static const char * const fqs[] = RCU_FQS_NAME_INIT; static const char * const exp[] = RCU_EXP_NAME_INIT; + static struct lock_class_key rcu_node_class[RCU_NUM_LVLS]; + static struct lock_class_key rcu_fqs_class[RCU_NUM_LVLS]; + static struct lock_class_key rcu_exp_class[RCU_NUM_LVLS]; static u8 fl_mask = 0x1; int levelcnt[RCU_NUM_LVLS]; /* # nodes in each level. */ -- cgit From 47dbc90663f697a4515a8dd5c99ae43dba108cb4 Mon Sep 17 00:00:00 2001 From: Paul Gortmaker Date: Sun, 27 Sep 2015 19:14:57 -0400 Subject: kernel: Make rcu/tree_trace.c explicitly non-modular The Kconfig currently controlling compilation of this code is: init/Kconfig:config TREE_RCU_TRACE init/Kconfig: def_bool RCU_TRACE && ( TREE_RCU || PREEMPT_RCU ) ...meaning that it currently is not being built as a module by anyone. Lets remove the modular code that is essentially orphaned, so that when reading the file there is no doubt it is builtin-only. Since module_init translates to device_initcall in the non-modular case, the init ordering remains unchanged with this commit. We could consider moving this to an earlier initcall if desired. We don't replace module.h with init.h since the file already has that. We also delete the moduleparam.h include that is left over from commit 64db4cfff99c04cd5f550357edcc8780f96b54a2 (""Tree RCU": scalable classic RCU implementation") since it is not needed here either. We morph some tags like MODULE_AUTHOR into the comments at the top of the file for documentation purposes. Cc: "Paul E. McKenney" Cc: Josh Triplett Reviewed-by: Josh Triplett Cc: Steven Rostedt Cc: Mathieu Desnoyers Cc: Lai Jiangshan Signed-off-by: Paul Gortmaker Signed-off-by: Paul E. McKenney --- kernel/rcu/tree_trace.c | 19 +++---------------- 1 file changed, 3 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree_trace.c b/kernel/rcu/tree_trace.c index 8efaba870d96..82aca98b18f8 100644 --- a/kernel/rcu/tree_trace.c +++ b/kernel/rcu/tree_trace.c @@ -1,5 +1,5 @@ /* - * Read-Copy Update tracing for classic implementation + * Read-Copy Update tracing for hierarchical implementation. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -16,6 +16,7 @@ * http://www.gnu.org/licenses/gpl-2.0.html. * * Copyright IBM Corporation, 2008 + * Author: Paul E. McKenney * * Papers: http://www.rdrop.com/users/paulmck/RCU * @@ -33,9 +34,7 @@ #include #include #include -#include #include -#include #include #include #include @@ -487,16 +486,4 @@ free_out: debugfs_remove_recursive(rcudir); return 1; } - -static void __exit rcutree_trace_cleanup(void) -{ - debugfs_remove_recursive(rcudir); -} - - -module_init(rcutree_trace_init); -module_exit(rcutree_trace_cleanup); - -MODULE_AUTHOR("Paul E. McKenney"); -MODULE_DESCRIPTION("Read-Copy Update tracing for hierarchical implementation"); -MODULE_LICENSE("GPL"); +device_initcall(rcutree_trace_init); -- cgit From fecbf6f01fbd83e6419ccb7f61d9a6eb987f1d92 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Mon, 28 Sep 2015 18:19:24 -0700 Subject: rcu: Simplify rcu_sched_qs() control flow This commit applies an early-exit approach to rcu_sched_qs(), reducing the nesting level and saving a line of code. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 29 ++++++++++++++--------------- 1 file changed, 14 insertions(+), 15 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 23df2661c899..ed3bc0578cc5 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -244,22 +244,21 @@ void rcu_sched_qs(void) { unsigned long flags; - if (__this_cpu_read(rcu_sched_data.cpu_no_qs.s)) { - trace_rcu_grace_period(TPS("rcu_sched"), - __this_cpu_read(rcu_sched_data.gpnum), - TPS("cpuqs")); - __this_cpu_write(rcu_sched_data.cpu_no_qs.b.norm, false); - if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)) - return; - local_irq_save(flags); - if (__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)) { - __this_cpu_write(rcu_sched_data.cpu_no_qs.b.exp, false); - rcu_report_exp_rdp(&rcu_sched_state, - this_cpu_ptr(&rcu_sched_data), - true); - } - local_irq_restore(flags); + if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.s)) + return; + trace_rcu_grace_period(TPS("rcu_sched"), + __this_cpu_read(rcu_sched_data.gpnum), + TPS("cpuqs")); + __this_cpu_write(rcu_sched_data.cpu_no_qs.b.norm, false); + if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)) + return; + local_irq_save(flags); + if (__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)) { + __this_cpu_write(rcu_sched_data.cpu_no_qs.b.exp, false); + rcu_report_exp_rdp(&rcu_sched_state, + this_cpu_ptr(&rcu_sched_data), true); } + local_irq_restore(flags); } void rcu_bh_qs(void) -- cgit From 8ba9153b2c3ab733d64e22adb57820ccb6afc496 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 07:55:41 -0700 Subject: rcu: Remove lock-acquisition loop from rcu_read_unlock_special() Several releases have come and gone without the warning triggering, so remove the lock-acquisition loop. Retain the WARN_ON_ONCE() out of sheer paranoia. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree_plugin.h | 18 ++++++------------ 1 file changed, 6 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index 57ba873d2f18..ae4ce2b665f8 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -449,19 +449,13 @@ void rcu_read_unlock_special(struct task_struct *t) /* * Remove this task from the list it blocked on. The task - * now remains queued on the rcu_node corresponding to - * the CPU it first blocked on, so the first attempt to - * acquire the task's rcu_node's ->lock will succeed. - * Keep the loop and add a WARN_ON() out of sheer paranoia. + * now remains queued on the rcu_node corresponding to the + * CPU it first blocked on, so there is no longer any need + * to loop. Retain a WARN_ON_ONCE() out of sheer paranoia. */ - for (;;) { - rnp = t->rcu_blocked_node; - raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ - if (rnp == t->rcu_blocked_node) - break; - WARN_ON_ONCE(1); - raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */ - } + rnp = t->rcu_blocked_node; + raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ + WARN_ON_ONCE(rnp != t->rcu_blocked_node); empty_norm = !rcu_preempt_blocked_readers_cgp(rnp); empty_exp = sync_rcu_preempt_exp_done(rnp); smp_mb(); /* ensure expedited fastpath sees end of RCU c-s. */ -- cgit From 699d40352059e64a4d993af170272585c41988d0 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 08:47:49 -0700 Subject: rcu: Fix obsolete rcu_bootup_announce_oddness() comment This function no longer has #ifdefs, so this commit removes the header comment calling them out. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree_plugin.h | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index ae4ce2b665f8..42df93721e6f 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -63,8 +63,7 @@ static bool __read_mostly rcu_nocb_poll; /* Offload kthread are to poll. */ /* * Check the RCU kernel configuration parameters and print informative - * messages about anything out of the ordinary. If you like #ifdef, you - * will love this function. + * messages about anything out of the ordinary. */ static void __init rcu_bootup_announce_oddness(void) { -- cgit From f0f2e7d307fff226e0c1df5a07101a1216a46d8a Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 29 Sep 2015 08:59:32 -0700 Subject: rcu: Avoid tick_nohz_active checks on NOCBs CPUs Currently, rcu_prepare_for_idle() checks for tick_nohz_active, even on individual NOCBs CPUs, unless all CPUs are marked as NOCBs CPUs at build time. This check is pointless on NOCBs CPUs because they never have any callbacks posted, given that all of their callbacks are handed off to the corresponding rcuo kthread. There is a check for individually designated NOCBs CPUs, but it pointelessly follows the check for tick_nohz_active. This commit therefore moves the check for individually designated NOCBs CPUs up with the check for CONFIG_RCU_NOCB_CPU_ALL. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree_plugin.h | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index 42df93721e6f..8e9d4a4d0326 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -1513,7 +1513,8 @@ static void rcu_prepare_for_idle(void) struct rcu_state *rsp; int tne; - if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_ALL)) + if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_ALL) || + rcu_is_nocb_cpu(smp_processor_id())) return; /* Handle nohz enablement switches conservatively. */ @@ -1527,10 +1528,6 @@ static void rcu_prepare_for_idle(void) if (!tne) return; - /* If this is a no-CBs CPU, no callbacks, just return. */ - if (rcu_is_nocb_cpu(smp_processor_id())) - return; - /* * If a non-lazy callback arrived at a CPU having only lazy * callbacks, invoke RCU core for the side-effect of recalculating -- cgit From 46a5d164db53ba6066b11889abb7fa6bddbe5cf7 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Wed, 7 Oct 2015 09:10:48 -0700 Subject: rcu: Stop disabling interrupts in scheduler fastpaths We need the scheduler's fastpaths to be, well, fast, and unnecessarily disabling and re-enabling interrupts is not necessarily consistent with this goal. Especially given that there are regions of the scheduler that already have interrupts disabled. This commit therefore moves the call to rcu_note_context_switch() to one of the interrupts-disabled regions of the scheduler, and removes the now-redundant disabling and re-enabling of interrupts from rcu_note_context_switch() and the functions it calls. Reported-by: Peter Zijlstra Signed-off-by: Paul E. McKenney [ paulmck: Shift rcu_note_context_switch() to avoid deadlock, as suggested by Peter Zijlstra. ] --- kernel/rcu/tree.c | 27 ++++++++++++--------------- kernel/rcu/tree_plugin.h | 14 ++++++-------- kernel/sched/core.c | 6 ++++-- 3 files changed, 22 insertions(+), 25 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index ed3bc0578cc5..93941d3434ad 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -242,8 +242,6 @@ static int rcu_gp_in_progress(struct rcu_state *rsp) */ void rcu_sched_qs(void) { - unsigned long flags; - if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.s)) return; trace_rcu_grace_period(TPS("rcu_sched"), @@ -252,13 +250,9 @@ void rcu_sched_qs(void) __this_cpu_write(rcu_sched_data.cpu_no_qs.b.norm, false); if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)) return; - local_irq_save(flags); - if (__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)) { - __this_cpu_write(rcu_sched_data.cpu_no_qs.b.exp, false); - rcu_report_exp_rdp(&rcu_sched_state, - this_cpu_ptr(&rcu_sched_data), true); - } - local_irq_restore(flags); + __this_cpu_write(rcu_sched_data.cpu_no_qs.b.exp, false); + rcu_report_exp_rdp(&rcu_sched_state, + this_cpu_ptr(&rcu_sched_data), true); } void rcu_bh_qs(void) @@ -295,17 +289,16 @@ EXPORT_PER_CPU_SYMBOL_GPL(rcu_qs_ctr); * We inform the RCU core by emulating a zero-duration dyntick-idle * period, which we in turn do by incrementing the ->dynticks counter * by two. + * + * The caller must have disabled interrupts. */ static void rcu_momentary_dyntick_idle(void) { - unsigned long flags; struct rcu_data *rdp; struct rcu_dynticks *rdtp; int resched_mask; struct rcu_state *rsp; - local_irq_save(flags); - /* * Yes, we can lose flag-setting operations. This is OK, because * the flag will be set again after some delay. @@ -335,13 +328,12 @@ static void rcu_momentary_dyntick_idle(void) smp_mb__after_atomic(); /* Later stuff after QS. */ break; } - local_irq_restore(flags); } /* * Note a context switch. This is a quiescent state for RCU-sched, * and requires special handling for preemptible RCU. - * The caller must have disabled preemption. + * The caller must have disabled interrupts. */ void rcu_note_context_switch(void) { @@ -371,9 +363,14 @@ EXPORT_SYMBOL_GPL(rcu_note_context_switch); */ void rcu_all_qs(void) { + unsigned long flags; + barrier(); /* Avoid RCU read-side critical sections leaking down. */ - if (unlikely(raw_cpu_read(rcu_sched_qs_mask))) + if (unlikely(raw_cpu_read(rcu_sched_qs_mask))) { + local_irq_save(flags); rcu_momentary_dyntick_idle(); + local_irq_restore(flags); + } this_cpu_inc(rcu_qs_ctr); barrier(); /* Avoid RCU read-side critical sections leaking up. */ } diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index 8e9d4a4d0326..e6da888cc908 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -146,8 +146,8 @@ static void __init rcu_bootup_announce(void) * the corresponding expedited grace period will also be the end of the * normal grace period. */ -static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp, - unsigned long flags) __releases(rnp->lock) +static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp) + __releases(rnp->lock) /* But leaves rrupts disabled. */ { int blkd_state = (rnp->gp_tasks ? RCU_GP_TASKS : 0) + (rnp->exp_tasks ? RCU_EXP_TASKS : 0) + @@ -235,7 +235,7 @@ static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp, rnp->gp_tasks = &t->rcu_node_entry; if (!rnp->exp_tasks && (blkd_state & RCU_EXP_BLKD)) rnp->exp_tasks = &t->rcu_node_entry; - raw_spin_unlock(&rnp->lock); + raw_spin_unlock(&rnp->lock); /* rrupts remain disabled. */ /* * Report the quiescent state for the expedited GP. This expedited @@ -250,7 +250,6 @@ static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp, } else { WARN_ON_ONCE(t->rcu_read_unlock_special.b.exp_need_qs); } - local_irq_restore(flags); } /* @@ -285,12 +284,11 @@ static void rcu_preempt_qs(void) * predating the current grace period drain, in other words, until * rnp->gp_tasks becomes NULL. * - * Caller must disable preemption. + * Caller must disable interrupts. */ static void rcu_preempt_note_context_switch(void) { struct task_struct *t = current; - unsigned long flags; struct rcu_data *rdp; struct rcu_node *rnp; @@ -300,7 +298,7 @@ static void rcu_preempt_note_context_switch(void) /* Possibly blocking in an RCU read-side critical section. */ rdp = this_cpu_ptr(rcu_state_p->rda); rnp = rdp->mynode; - raw_spin_lock_irqsave_rcu_node(rnp, flags); + raw_spin_lock_rcu_node(rnp); t->rcu_read_unlock_special.b.blocked = true; t->rcu_blocked_node = rnp; @@ -316,7 +314,7 @@ static void rcu_preempt_note_context_switch(void) (rnp->qsmask & rdp->grpmask) ? rnp->gpnum : rnp->gpnum + 1); - rcu_preempt_ctxt_queue(rnp, rdp, flags); + rcu_preempt_ctxt_queue(rnp, rdp); } else if (t->rcu_read_lock_nesting < 0 && t->rcu_read_unlock_special.s) { diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 4d568ac9319e..ec72de234feb 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -3085,7 +3085,6 @@ static void __sched notrace __schedule(bool preempt) cpu = smp_processor_id(); rq = cpu_rq(cpu); - rcu_note_context_switch(); prev = rq->curr; /* @@ -3104,13 +3103,16 @@ static void __sched notrace __schedule(bool preempt) if (sched_feat(HRTICK)) hrtick_clear(rq); + local_irq_disable(); + rcu_note_context_switch(); + /* * Make sure that signal_pending_state()->signal_pending() below * can't be reordered with __set_current_state(TASK_INTERRUPTIBLE) * done by the caller to avoid the race with signal_wake_up(). */ smp_mb__before_spinlock(); - raw_spin_lock_irq(&rq->lock); + raw_spin_lock(&rq->lock); lockdep_pin_lock(&rq->lock); rq->clock_skip_update <<= 1; /* promote REQ to ACT */ -- cgit From a0e3a3aa2841d5720a277de53b6882eb8b2ef698 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Sat, 5 Dec 2015 17:34:10 -0800 Subject: rcutorture: Flag nonexistent RCU GP kthread Currently, if the RCU grace-period kthread has not yet been created, in which case the starvation-check code will print zero for the state, which maps to TASK_RUNNING. This could clearly be quite confusing, so this commit prints ~0, which does not map to any legal ->state value. Signed-off-by: Paul E. McKenney Reviewed-by: Josh Triplett --- kernel/rcu/tree.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 81aa1cdc6bc9..e2315fb57b23 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -1201,7 +1201,7 @@ static void rcu_check_gp_kthread_starvation(struct rcu_state *rsp) rsp->name, j - gpa, rsp->gpnum, rsp->completed, rsp->gp_flags, rsp->gp_state, - rsp->gp_kthread ? rsp->gp_kthread->state : 0); + rsp->gp_kthread ? rsp->gp_kthread->state : ~0); } /* -- cgit From b1adb3e2736b695821badc715d2c7a5d873b8b94 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Thu, 1 Oct 2015 10:38:16 -0700 Subject: rcutorture: Dump stack when GP kthread stalls This commit increases debug information in the case where the grace-period kthread is being prevented from running by dumping that kthread's stack. Signed-off-by: Paul E. McKenney [ paulmck: Split into prior commit and this commit, as suggested by Josh Triplett. ] Reviewed-by: Josh Triplett --- kernel/rcu/tree.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index e2315fb57b23..7b78c88e19a3 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -1196,12 +1196,15 @@ static void rcu_check_gp_kthread_starvation(struct rcu_state *rsp) j = jiffies; gpa = READ_ONCE(rsp->gp_activity); - if (j - gpa > 2 * HZ) + if (j - gpa > 2 * HZ) { pr_err("%s kthread starved for %ld jiffies! g%lu c%lu f%#x s%d ->state=%#lx\n", rsp->name, j - gpa, rsp->gpnum, rsp->completed, rsp->gp_flags, rsp->gp_state, rsp->gp_kthread ? rsp->gp_kthread->state : ~0); + if (rsp->gp_kthread) + sched_show_task(rsp->gp_kthread); + } } /* -- cgit From 18aff33e7314253b9437234bd6d69ddc4827de70 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 17 Nov 2015 13:35:28 -0800 Subject: rcutorture: Print symbolic name for rcu_torture_writer_state Currently, rcu_torture_writer_state is printed as an integer, which slows debugging. This commit therefore prints a symbolic name in addition to the integer. Signed-off-by: Paul E. McKenney [ paulmck: More "const", as suggested by Josh Triplett. ] Reviewed-by: Josh Triplett --- kernel/rcu/rcutorture.c | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c index d89328e260df..d2988d047d66 100644 --- a/kernel/rcu/rcutorture.c +++ b/kernel/rcu/rcutorture.c @@ -162,6 +162,27 @@ static int rcu_torture_writer_state; #define RTWS_SYNC 7 #define RTWS_STUTTER 8 #define RTWS_STOPPING 9 +static const char * const rcu_torture_writer_state_names[] = { + "RTWS_FIXED_DELAY", + "RTWS_DELAY", + "RTWS_REPLACE", + "RTWS_DEF_FREE", + "RTWS_EXP_SYNC", + "RTWS_COND_GET", + "RTWS_COND_SYNC", + "RTWS_SYNC", + "RTWS_STUTTER", + "RTWS_STOPPING", +}; + +static const char *rcu_torture_writer_state_getname(void) +{ + unsigned int i = READ_ONCE(rcu_torture_writer_state); + + if (i >= ARRAY_SIZE(rcu_torture_writer_state_names)) + return "???"; + return rcu_torture_writer_state_names[i]; +} #if defined(MODULE) || defined(CONFIG_RCU_TORTURE_TEST_RUNNABLE) #define RCUTORTURE_RUNNABLE_INIT 1 @@ -1307,7 +1328,8 @@ rcu_torture_stats_print(void) rcutorture_get_gp_data(cur_ops->ttype, &flags, &gpnum, &completed); - pr_alert("??? Writer stall state %d g%lu c%lu f%#x\n", + pr_alert("??? Writer stall state %s(%d) g%lu c%lu f%#x\n", + rcu_torture_writer_state_getname(), rcu_torture_writer_state, gpnum, completed, flags); show_rcu_gp_kthreads(); -- cgit From 6b50e119c440b7532ed749b635a58b3839f62992 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 17 Nov 2015 14:39:26 -0800 Subject: rcutorture: Print symbolic name for ->gp_state Currently, ->gp_state is printed as an integer, which slows debugging. This commit therefore prints a symbolic name in addition to the integer. Signed-off-by: Paul E. McKenney [ paulmck: Updated to fix relational operator called out by Dan Carpenter. ] [ paulmck: More "const", as suggested by Josh Triplett. ] Reviewed-by: Josh Triplett --- kernel/rcu/tree.c | 15 +++++++++++++-- kernel/rcu/tree.h | 12 ++++++++++++ 2 files changed, 25 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 7b78c88e19a3..316354109734 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -1186,6 +1186,16 @@ static void record_gp_stall_check_time(struct rcu_state *rsp) rsp->n_force_qs_gpstart = READ_ONCE(rsp->n_force_qs); } +/* + * Convert a ->gp_state value to a character string. + */ +static const char *gp_state_getname(short gs) +{ + if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names)) + return "???"; + return gp_state_names[gs]; +} + /* * Complain about starvation of grace-period kthread. */ @@ -1197,10 +1207,11 @@ static void rcu_check_gp_kthread_starvation(struct rcu_state *rsp) j = jiffies; gpa = READ_ONCE(rsp->gp_activity); if (j - gpa > 2 * HZ) { - pr_err("%s kthread starved for %ld jiffies! g%lu c%lu f%#x s%d ->state=%#lx\n", + pr_err("%s kthread starved for %ld jiffies! g%lu c%lu f%#x %s(%d) ->state=%#lx\n", rsp->name, j - gpa, rsp->gpnum, rsp->completed, - rsp->gp_flags, rsp->gp_state, + rsp->gp_flags, + gp_state_getname(rsp->gp_state), rsp->gp_state, rsp->gp_kthread ? rsp->gp_kthread->state : ~0); if (rsp->gp_kthread) sched_show_task(rsp->gp_kthread); diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index f32bebb6bc90..a3fb6fe94127 100644 --- a/kernel/rcu/tree.h +++ b/kernel/rcu/tree.h @@ -545,6 +545,18 @@ struct rcu_state { #define RCU_GP_CLEANUP 5 /* Grace-period cleanup started. */ #define RCU_GP_CLEANED 6 /* Grace-period cleanup complete. */ +#ifndef RCU_TREE_NONCORE +static const char * const gp_state_names[] = { + "RCU_GP_IDLE", + "RCU_GP_WAIT_GPS", + "RCU_GP_DONE_GPS", + "RCU_GP_WAIT_FQS", + "RCU_GP_DOING_FQS", + "RCU_GP_CLEANUP", + "RCU_GP_CLEANED", +}; +#endif /* #ifndef RCU_TREE_NONCORE */ + extern struct list_head rcu_struct_flavors; /* Sequence through rcu_state structures for each RCU flavor. */ -- cgit From 0017960f38a2470e70d9f1991228e2b55b2abe0c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 30 Nov 2015 16:26:35 +0100 Subject: perf/core: Collapse common IPI pattern Various functions implement the same pattern to send IPIs to an event's CPU. Collapse the easy ones in a common helper function to reduce duplication. Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Signed-off-by: Ingo Molnar --- kernel/events/core.c | 180 ++++++++++++++++++++++----------------------------- 1 file changed, 76 insertions(+), 104 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 39cf4a40aa4c..c3d61b92d805 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -126,6 +126,37 @@ static int cpu_function_call(int cpu, remote_function_f func, void *info) return data.ret; } +static void event_function_call(struct perf_event *event, + int (*active)(void *), + void (*inactive)(void *), + void *data) +{ + struct perf_event_context *ctx = event->ctx; + struct task_struct *task = ctx->task; + + if (!task) { + cpu_function_call(event->cpu, active, data); + return; + } + +again: + if (!task_function_call(task, active, data)) + return; + + raw_spin_lock_irq(&ctx->lock); + if (ctx->is_active) { + /* + * Reload the task pointer, it might have been changed by + * a concurrent perf_event_context_sched_out(). + */ + task = ctx->task; + raw_spin_unlock_irq(&ctx->lock); + goto again; + } + inactive(data); + raw_spin_unlock_irq(&ctx->lock); +} + #define EVENT_OWNER_KERNEL ((void *) -1) static bool is_kernel_event(struct perf_event *event) @@ -1629,6 +1660,17 @@ struct remove_event { bool detach_group; }; +static void ___perf_remove_from_context(void *info) +{ + struct remove_event *re = info; + struct perf_event *event = re->event; + struct perf_event_context *ctx = event->ctx; + + if (re->detach_group) + perf_group_detach(event); + list_del_event(event, ctx); +} + /* * Cross CPU call to remove a performance event * @@ -1656,7 +1698,6 @@ static int __perf_remove_from_context(void *info) return 0; } - /* * Remove the event from a task's (or a CPU's) list of events. * @@ -1673,7 +1714,6 @@ static int __perf_remove_from_context(void *info) static void perf_remove_from_context(struct perf_event *event, bool detach_group) { struct perf_event_context *ctx = event->ctx; - struct task_struct *task = ctx->task; struct remove_event re = { .event = event, .detach_group = detach_group, @@ -1681,44 +1721,8 @@ static void perf_remove_from_context(struct perf_event *event, bool detach_group lockdep_assert_held(&ctx->mutex); - if (!task) { - /* - * Per cpu events are removed via an smp call. The removal can - * fail if the CPU is currently offline, but in that case we - * already called __perf_remove_from_context from - * perf_event_exit_cpu. - */ - cpu_function_call(event->cpu, __perf_remove_from_context, &re); - return; - } - -retry: - if (!task_function_call(task, __perf_remove_from_context, &re)) - return; - - raw_spin_lock_irq(&ctx->lock); - /* - * If we failed to find a running task, but find the context active now - * that we've acquired the ctx->lock, retry. - */ - if (ctx->is_active) { - raw_spin_unlock_irq(&ctx->lock); - /* - * Reload the task pointer, it might have been changed by - * a concurrent perf_event_context_sched_out(). - */ - task = ctx->task; - goto retry; - } - - /* - * Since the task isn't running, its safe to remove the event, us - * holding the ctx->lock ensures the task won't get scheduled in. - */ - if (detach_group) - perf_group_detach(event); - list_del_event(event, ctx); - raw_spin_unlock_irq(&ctx->lock); + event_function_call(event, __perf_remove_from_context, + ___perf_remove_from_context, &re); } /* @@ -2067,6 +2071,18 @@ static void perf_event_sched_in(struct perf_cpu_context *cpuctx, ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task); } +static void ___perf_install_in_context(void *info) +{ + struct perf_event *event = info; + struct perf_event_context *ctx = event->ctx; + + /* + * Since the task isn't running, its safe to add the event, us holding + * the ctx->lock ensures the task won't get scheduled in. + */ + add_event_to_ctx(event, ctx); +} + /* * Cross CPU call to install and enable a performance event * @@ -2143,48 +2159,14 @@ perf_install_in_context(struct perf_event_context *ctx, struct perf_event *event, int cpu) { - struct task_struct *task = ctx->task; - lockdep_assert_held(&ctx->mutex); event->ctx = ctx; if (event->cpu != -1) event->cpu = cpu; - if (!task) { - /* - * Per cpu events are installed via an smp call and - * the install is always successful. - */ - cpu_function_call(cpu, __perf_install_in_context, event); - return; - } - -retry: - if (!task_function_call(task, __perf_install_in_context, event)) - return; - - raw_spin_lock_irq(&ctx->lock); - /* - * If we failed to find a running task, but find the context active now - * that we've acquired the ctx->lock, retry. - */ - if (ctx->is_active) { - raw_spin_unlock_irq(&ctx->lock); - /* - * Reload the task pointer, it might have been changed by - * a concurrent perf_event_context_sched_out(). - */ - task = ctx->task; - goto retry; - } - - /* - * Since the task isn't running, its safe to add the event, us holding - * the ctx->lock ensures the task won't get scheduled in. - */ - add_event_to_ctx(event, ctx); - raw_spin_unlock_irq(&ctx->lock); + event_function_call(event, __perf_install_in_context, + ___perf_install_in_context, event); } /* @@ -4154,6 +4136,22 @@ struct period_event { u64 value; }; +static void ___perf_event_period(void *info) +{ + struct period_event *pe = info; + struct perf_event *event = pe->event; + u64 value = pe->value; + + if (event->attr.freq) { + event->attr.sample_freq = value; + } else { + event->attr.sample_period = value; + event->hw.sample_period = value; + } + + local64_set(&event->hw.period_left, 0); +} + static int __perf_event_period(void *info) { struct period_event *pe = info; @@ -4190,8 +4188,6 @@ static int __perf_event_period(void *info) static int perf_event_period(struct perf_event *event, u64 __user *arg) { struct period_event pe = { .event = event, }; - struct perf_event_context *ctx = event->ctx; - struct task_struct *task; u64 value; if (!is_sampling_event(event)) @@ -4206,34 +4202,10 @@ static int perf_event_period(struct perf_event *event, u64 __user *arg) if (event->attr.freq && value > sysctl_perf_event_sample_rate) return -EINVAL; - task = ctx->task; pe.value = value; - if (!task) { - cpu_function_call(event->cpu, __perf_event_period, &pe); - return 0; - } - -retry: - if (!task_function_call(task, __perf_event_period, &pe)) - return 0; - - raw_spin_lock_irq(&ctx->lock); - if (ctx->is_active) { - raw_spin_unlock_irq(&ctx->lock); - task = ctx->task; - goto retry; - } - - if (event->attr.freq) { - event->attr.sample_freq = value; - } else { - event->attr.sample_period = value; - event->hw.sample_period = value; - } - - local64_set(&event->hw.period_left, 0); - raw_spin_unlock_irq(&ctx->lock); + event_function_call(event, __perf_event_period, + ___perf_event_period, &pe); return 0; } -- cgit From 35a4933a895927990772ae96fdcfd2f806929ee2 Mon Sep 17 00:00:00 2001 From: David Gibson Date: Mon, 30 Nov 2015 12:30:30 +1100 Subject: time: Avoid signed overflow in timekeeping_get_ns() 1e75fa8 "time: Condense timekeeper.xtime into xtime_sec" replaced a call to clocksource_cyc2ns() from timekeeping_get_ns() with an open-coded version of the same logic to avoid keeping a semi-redundant struct timespec in struct timekeeper. However, the commit also introduced a subtle semantic change - where clocksource_cyc2ns() uses purely unsigned math, the new version introduces a signed temporary, meaning that if (delta * tk->mult) has a 63-bit overflow the following shift will still give a negative result. The choice of 'maxsec' in __clocksource_updatefreq_scale() means this will generally happen if there's a ~10 minute pause in examining the clocksource. This can be triggered on a powerpc KVM guest by stopping it from qemu for a bit over 10 minutes. After resuming time has jumped backwards several minutes causing numerous problems (jiffies does not advance, msleep()s can be extended by minutes..). It doesn't happen on x86 KVM guests, because the guest TSC is effectively frozen while the guest is stopped, which is not the case for the powerpc timebase. Obviously an unsigned (64 bit) overflow will only take twice as long as a signed, 63-bit overflow. I don't know the time code well enough to know if that will still cause incorrect calculations, or if a 64-bit overflow is avoided elsewhere. Still, an incorrect forwards clock adjustment will cause less trouble than time going backwards. So, this patch removes the potential for intermediate signed overflow. Cc: stable@vger.kernel.org (3.7+) Suggested-by: Laurent Vivier Tested-by: Laurent Vivier Signed-off-by: David Gibson Signed-off-by: John Stultz --- kernel/time/timekeeping.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index d563c1960302..99188ee5d9d0 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -305,8 +305,7 @@ static inline s64 timekeeping_get_ns(struct tk_read_base *tkr) delta = timekeeping_get_delta(tkr); - nsec = delta * tkr->mult + tkr->xtime_nsec; - nsec >>= tkr->shift; + nsec = (delta * tkr->mult + tkr->xtime_nsec) >> tkr->shift; /* If arch requires, add in get_arch_timeoffset() */ return nsec + arch_gettimeoffset(); -- cgit From 390dd67c471a43a8a0f36c7d5177de49e7749c59 Mon Sep 17 00:00:00 2001 From: Seiichi Ikarashi Date: Thu, 10 Sep 2015 18:01:56 +0900 Subject: clocksource: Add CPU info to clocksource watchdog reporting The clocksource watchdog reporting was improved by 0b046b217ad4c6. I want to add the info of CPU where the watchdog detects a deviation because it is necessary to identify the trouble spot if the clocksource is TSC. Signed-off-by: Seiichi Ikarashi [jstultz: Tweaked commit message] Signed-off-by: John Stultz --- kernel/time/clocksource.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index 1347882d131e..664de539299b 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -218,8 +218,8 @@ static void clocksource_watchdog(unsigned long data) /* Check the deviation from the watchdog clocksource. */ if (abs(cs_nsec - wd_nsec) > WATCHDOG_THRESHOLD) { - pr_warn("timekeeping watchdog: Marking clocksource '%s' as unstable because the skew is too large:\n", - cs->name); + pr_warn("timekeeping watchdog on CPU%d: Marking clocksource '%s' as unstable because the skew is too large:\n", + smp_processor_id(), cs->name); pr_warn(" '%s' wd_now: %llx wd_last: %llx mask: %llx\n", watchdog->name, wdnow, wdlast, watchdog->mask); pr_warn(" '%s' cs_now: %llx cs_last: %llx mask: %llx\n", -- cgit From 79cfea0273876d9c438f3227b8f68c8c7ae31583 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Mon, 7 Dec 2015 13:09:52 -0800 Subject: rcu: Remove TINY_RCU bloat from pointless boot parameters The rcu_expedited, rcu_normal, and rcu_normal_after_boot kernel boot parameters are pointless in the case of TINY_RCU because in that case synchronous grace periods, both expedited and normal, are no-ops. However, these three symbols contribute several hundred bytes of bloat. This commit therefore uses CPP directives to avoid compiling this code in TINY_RCU kernels. Reported-by: kbuild test robot Signed-off-by: Paul E. McKenney Reviewed-by: Josh Triplett --- kernel/ksysfs.c | 4 ++++ kernel/rcu/update.c | 7 ++++--- 2 files changed, 8 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/ksysfs.c b/kernel/ksysfs.c index b4e2fa52d8bc..152da4a48867 100644 --- a/kernel/ksysfs.c +++ b/kernel/ksysfs.c @@ -144,6 +144,7 @@ static ssize_t fscaps_show(struct kobject *kobj, } KERNEL_ATTR_RO(fscaps); +#ifndef CONFIG_TINY_RCU int rcu_expedited; static ssize_t rcu_expedited_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) @@ -177,6 +178,7 @@ static ssize_t rcu_normal_store(struct kobject *kobj, return count; } KERNEL_ATTR_RW(rcu_normal); +#endif /* #ifndef CONFIG_TINY_RCU */ /* * Make /sys/kernel/notes give the raw contents of our kernel .notes section. @@ -219,8 +221,10 @@ static struct attribute * kernel_attrs[] = { &kexec_crash_size_attr.attr, &vmcoreinfo_attr.attr, #endif +#ifndef CONFIG_TINY_RCU &rcu_expedited_attr.attr, &rcu_normal_attr.attr, +#endif NULL }; diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c index 12b91f5a60a6..76b94e19430b 100644 --- a/kernel/rcu/update.c +++ b/kernel/rcu/update.c @@ -60,11 +60,12 @@ MODULE_ALIAS("rcupdate"); #endif #define MODULE_PARAM_PREFIX "rcupdate." +#ifndef CONFIG_TINY_RCU module_param(rcu_expedited, int, 0); module_param(rcu_normal, int, 0); - static int rcu_normal_after_boot; module_param(rcu_normal_after_boot, int, 0); +#endif /* #ifndef CONFIG_TINY_RCU */ #if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_PREEMPT_COUNT) /** @@ -172,8 +173,6 @@ void rcu_unexpedite_gp(void) } EXPORT_SYMBOL_GPL(rcu_unexpedite_gp); -#endif /* #ifndef CONFIG_TINY_RCU */ - /* * Inform RCU of the end of the in-kernel boot sequence. */ @@ -185,6 +184,8 @@ void rcu_end_inkernel_boot(void) WRITE_ONCE(rcu_normal, 1); } +#endif /* #ifndef CONFIG_TINY_RCU */ + #ifdef CONFIG_PREEMPT_RCU /* -- cgit From a87f203e2731ab477386c678e59033ee103018c0 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Tue, 20 Oct 2015 12:38:49 -0700 Subject: rcu: Eliminate unused rcu_init_one() argument Now that the rcu_state structure's ->rda field is compile-time initialized, there is no need to pass the per-CPU rcu_data structure into rcu_init_one(). This commit therefore eliminates this now-unused parameter. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 7 +++---- kernel/rcu/tree_plugin.h | 2 +- 2 files changed, 4 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 93941d3434ad..9a4c8c0653ff 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -4351,8 +4351,7 @@ static void __init rcu_init_levelspread(int *levelspread, const int *levelcnt) /* * Helper function for rcu_init() that initializes one rcu_state structure. */ -static void __init rcu_init_one(struct rcu_state *rsp, - struct rcu_data __percpu *rda) +static void __init rcu_init_one(struct rcu_state *rsp) { static const char * const buf[] = RCU_NODE_NAME_INIT; static const char * const fqs[] = RCU_FQS_NAME_INIT; @@ -4545,8 +4544,8 @@ void __init rcu_init(void) rcu_bootup_announce(); rcu_init_geometry(); - rcu_init_one(&rcu_bh_state, &rcu_bh_data); - rcu_init_one(&rcu_sched_state, &rcu_sched_data); + rcu_init_one(&rcu_bh_state); + rcu_init_one(&rcu_sched_state); if (dump_tree) rcu_dump_rcu_node_tree(&rcu_sched_state); __rcu_init_preempt(); diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index e6da888cc908..fccef5d4b198 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -777,7 +777,7 @@ EXPORT_SYMBOL_GPL(rcu_barrier); */ static void __init __rcu_init_preempt(void) { - rcu_init_one(rcu_state_p, rcu_data_p); + rcu_init_one(rcu_state_p); } /* -- cgit From d117c8aa1d511f76401337620b9c4ffb4c886579 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Sat, 31 Oct 2015 00:01:18 -0700 Subject: rcu: Make cpu_needs_another_gp() be bool The cpu_needs_another_gp() function is currently of type int, but only returns zero or one. Bow to reality and make it be of type bool. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 9a4c8c0653ff..d6863bceeb45 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -597,25 +597,25 @@ static int rcu_future_needs_gp(struct rcu_state *rsp) * The caller must have disabled interrupts to prevent races with * normal callback registry. */ -static int +static bool cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp) { int i; if (rcu_gp_in_progress(rsp)) - return 0; /* No, a grace period is already in progress. */ + return false; /* No, a grace period is already in progress. */ if (rcu_future_needs_gp(rsp)) - return 1; /* Yes, a no-CBs CPU needs one. */ + return true; /* Yes, a no-CBs CPU needs one. */ if (!rdp->nxttail[RCU_NEXT_TAIL]) - return 0; /* No, this is a no-CBs (or offline) CPU. */ + return false; /* No, this is a no-CBs (or offline) CPU. */ if (*rdp->nxttail[RCU_NEXT_READY_TAIL]) - return 1; /* Yes, this CPU has newly registered callbacks. */ + return true; /* Yes, CPU has newly registered callbacks. */ for (i = RCU_WAIT_TAIL; i < RCU_NEXT_TAIL; i++) if (rdp->nxttail[i - 1] != rdp->nxttail[i] && ULONG_CMP_LT(READ_ONCE(rsp->completed), rdp->nxtcompleted[i])) - return 1; /* Yes, CBs for future grace period. */ - return 0; /* No grace period needed. */ + return true; /* Yes, CBs for future grace period. */ + return false; /* No grace period needed. */ } /* -- cgit From 7c9906ca5e582a773fff696975e312cef58a7386 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Sat, 31 Oct 2015 00:59:01 -0700 Subject: rcu: Don't redundantly disable irqs in rcu_irq_{enter,exit}() This commit replaces a local_irq_save()/local_irq_restore() pair with a lockdep assertion that interrupts are already disabled. This should remove the corresponding overhead from the interrupt entry/exit fastpaths. This change was inspired by the fact that Iftekhar Ahmed's mutation testing showed that removing rcu_irq_enter()'s call to local_ird_restore() had no effect, which might indicate that interrupts were always enabled anyway. Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 32 ++++++++++++++++++++++++++------ 1 file changed, 26 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index d6863bceeb45..40940b0d0310 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -732,7 +732,7 @@ void rcu_user_enter(void) * * Exit from an interrupt handler, which might possibly result in entering * idle mode, in other words, leaving the mode in which read-side critical - * sections can occur. + * sections can occur. The caller must have disabled interrupts. * * This code assumes that the idle loop never does anything that might * result in unbalanced calls to irq_enter() and irq_exit(). If your @@ -745,11 +745,10 @@ void rcu_user_enter(void) */ void rcu_irq_exit(void) { - unsigned long flags; long long oldval; struct rcu_dynticks *rdtp; - local_irq_save(flags); + RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_irq_exit() invoked with irqs enabled!!!"); rdtp = this_cpu_ptr(&rcu_dynticks); oldval = rdtp->dynticks_nesting; rdtp->dynticks_nesting--; @@ -760,6 +759,17 @@ void rcu_irq_exit(void) else rcu_eqs_enter_common(oldval, true); rcu_sysidle_enter(1); +} + +/* + * Wrapper for rcu_irq_exit() where interrupts are enabled. + */ +void rcu_irq_exit_irqson(void) +{ + unsigned long flags; + + local_irq_save(flags); + rcu_irq_exit(); local_irq_restore(flags); } @@ -857,7 +867,7 @@ void rcu_user_exit(void) * * Enter an interrupt handler, which might possibly result in exiting * idle mode, in other words, entering the mode in which read-side critical - * sections can occur. + * sections can occur. The caller must have disabled interrupts. * * Note that the Linux kernel is fully capable of entering an interrupt * handler that it never exits, for example when doing upcalls to @@ -873,11 +883,10 @@ void rcu_user_exit(void) */ void rcu_irq_enter(void) { - unsigned long flags; struct rcu_dynticks *rdtp; long long oldval; - local_irq_save(flags); + RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_irq_enter() invoked with irqs enabled!!!"); rdtp = this_cpu_ptr(&rcu_dynticks); oldval = rdtp->dynticks_nesting; rdtp->dynticks_nesting++; @@ -888,6 +897,17 @@ void rcu_irq_enter(void) else rcu_eqs_exit_common(oldval, true); rcu_sysidle_exit(1); +} + +/* + * Wrapper for rcu_irq_enter() where interrupts are enabled. + */ +void rcu_irq_enter_irqson(void) +{ + unsigned long flags; + + local_irq_save(flags); + rcu_irq_enter(); local_irq_restore(flags); } -- cgit From e11f13355b09df970495c45ed0eac1dc85dcf5c1 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 4 Nov 2015 08:22:05 -0800 Subject: rcu: Move wakeup out from under rnp->lock This patch removes a potential deadlock hazard by moving the wake_up_process() in rcu_spawn_gp_kthread() out from under rnp->lock. Signed-off-by: Peter Zijlstra Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 40940b0d0310..87b604d0b0d2 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -4319,8 +4319,8 @@ static int __init rcu_spawn_gp_kthread(void) sp.sched_priority = kthread_prio; sched_setscheduler_nocheck(t, SCHED_FIFO, &sp); } - wake_up_process(t); raw_spin_unlock_irqrestore(&rnp->lock, flags); + wake_up_process(t); } rcu_spawn_nocb_kthreads(); rcu_spawn_boost_kthreads(); -- cgit From 45fed3e7cfb4001c80cd4bd25249d194a52bfed3 Mon Sep 17 00:00:00 2001 From: "Paul E. McKenney" Date: Sat, 7 Nov 2015 23:35:00 -0800 Subject: rcu: Make rcu_gp_init() be bool rather than int The return value from rcu_gp_init() is always used as a bool, so this commit makes it be a bool. Reported-by: Iftekhar Ahmed Signed-off-by: Paul E. McKenney --- kernel/rcu/tree.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 87b604d0b0d2..01a90a3bdf79 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -1814,9 +1814,9 @@ static void rcu_gp_slow(struct rcu_state *rsp, int delay) } /* - * Initialize a new grace period. Return 0 if no grace period required. + * Initialize a new grace period. Return false if no grace period required. */ -static int rcu_gp_init(struct rcu_state *rsp) +static bool rcu_gp_init(struct rcu_state *rsp) { unsigned long oldmask; struct rcu_data *rdp; @@ -1827,7 +1827,7 @@ static int rcu_gp_init(struct rcu_state *rsp) if (!READ_ONCE(rsp->gp_flags)) { /* Spurious wakeup, tell caller to go back to sleep. */ raw_spin_unlock_irq(&rnp->lock); - return 0; + return false; } WRITE_ONCE(rsp->gp_flags, 0); /* Clear all flags: New grace period. */ @@ -1837,7 +1837,7 @@ static int rcu_gp_init(struct rcu_state *rsp) * Not supposed to be able to happen. */ raw_spin_unlock_irq(&rnp->lock); - return 0; + return false; } /* Advance to a new grace period and initialize state. */ @@ -1929,7 +1929,7 @@ static int rcu_gp_init(struct rcu_state *rsp) WRITE_ONCE(rsp->gp_activity, jiffies); } - return 1; + return true; } /* -- cgit From f0cb32207307e9d7b3ee8117078b7a37f8d0166e Mon Sep 17 00:00:00 2001 From: Thomas Petazzoni Date: Tue, 20 Oct 2015 15:23:51 +0200 Subject: genirq: Implement irq_percpu_is_enabled() Certain interrupt controller drivers have a register set that does not make it easy to save/restore the mask of enabled/disabled interrupts at suspend/resume time. At resume time, such drivers rely on the core kernel irq subsystem to tell whether such or such interrupt is enabled or not, in order to restore the proper state in the interrupt controller register. While the irqd_irq_disabled() provides the relevant information for global interrupts, there is no similar function to query the enabled/disabled state of a per-CPU interrupt. Therefore, this commit complements the percpu_irq API with an irq_percpu_is_enabled() function. [ tglx: Simplified the implementation and added kerneldoc ] Signed-off-by: Thomas Petazzoni Cc: linux-arm-kernel@lists.infradead.org Cc: Tawfik Bayouk Cc: Nadav Haklai Cc: Lior Amsalem Cc: Andrew Lunn Cc: Sebastian Hesselbarth Cc: Gregory Clement Cc: Jason Cooper Cc: Marc Zyngier Link: http://lkml.kernel.org/r/1445347435-2333-2-git-send-email-thomas.petazzoni@free-electrons.com Signed-off-by: Thomas Gleixner --- kernel/irq/manage.c | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) (limited to 'kernel') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 0eebaeef317b..c84670c373f9 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -1743,6 +1743,31 @@ out: } EXPORT_SYMBOL_GPL(enable_percpu_irq); +/** + * irq_percpu_is_enabled - Check whether the per cpu irq is enabled + * @irq: Linux irq number to check for + * + * Must be called from a non migratable context. Returns the enable + * state of a per cpu interrupt on the current cpu. + */ +bool irq_percpu_is_enabled(unsigned int irq) +{ + unsigned int cpu = smp_processor_id(); + struct irq_desc *desc; + unsigned long flags; + bool is_enabled; + + desc = irq_get_desc_lock(irq, &flags, IRQ_GET_DESC_CHECK_PERCPU); + if (!desc) + return false; + + is_enabled = cpumask_test_cpu(cpu, desc->percpu_enabled); + irq_put_desc_unlock(desc, flags); + + return is_enabled; +} +EXPORT_SYMBOL_GPL(irq_percpu_is_enabled); + void disable_percpu_irq(unsigned int irq) { unsigned int cpu = smp_processor_id(); -- cgit From fca839c00a12d682cb59b3b620d109a1d850b262 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 7 Dec 2015 10:58:57 -0500 Subject: workqueue: warn if memory reclaim tries to flush !WQ_MEM_RECLAIM workqueue Task or work item involved in memory reclaim trying to flush a non-WQ_MEM_RECLAIM workqueue or one of its work items can lead to deadlock. Trigger WARN_ONCE() if such conditions are detected. Signed-off-by: Tejun Heo Cc: Peter Zijlstra --- kernel/workqueue.c | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) (limited to 'kernel') diff --git a/kernel/workqueue.c b/kernel/workqueue.c index c579dbab2e36..c7769c507bf5 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -2316,6 +2316,37 @@ repeat: goto repeat; } +/** + * check_flush_dependency - check for flush dependency sanity + * @target_wq: workqueue being flushed + * @target_work: work item being flushed (NULL for workqueue flushes) + * + * %current is trying to flush the whole @target_wq or @target_work on it. + * If @target_wq doesn't have %WQ_MEM_RECLAIM, verify that %current is not + * reclaiming memory or running on a workqueue which doesn't have + * %WQ_MEM_RECLAIM as that can break forward-progress guarantee leading to + * a deadlock. + */ +static void check_flush_dependency(struct workqueue_struct *target_wq, + struct work_struct *target_work) +{ + work_func_t target_func = target_work ? target_work->func : NULL; + struct worker *worker; + + if (target_wq->flags & WQ_MEM_RECLAIM) + return; + + worker = current_wq_worker(); + + WARN_ONCE(current->flags & PF_MEMALLOC, + "workqueue: PF_MEMALLOC task %d(%s) is flushing !WQ_MEM_RECLAIM %s:%pf", + current->pid, current->comm, target_wq->name, target_func); + WARN_ONCE(worker && (worker->current_pwq->wq->flags & WQ_MEM_RECLAIM), + "workqueue: WQ_MEM_RECLAIM %s:%pf is flushing !WQ_MEM_RECLAIM %s:%pf", + worker->current_pwq->wq->name, worker->current_func, + target_wq->name, target_func); +} + struct wq_barrier { struct work_struct work; struct completion done; @@ -2525,6 +2556,8 @@ void flush_workqueue(struct workqueue_struct *wq) list_add_tail(&this_flusher.list, &wq->flusher_overflow); } + check_flush_dependency(wq, NULL); + mutex_unlock(&wq->mutex); wait_for_completion(&this_flusher.done); @@ -2697,6 +2730,8 @@ static bool start_flush_work(struct work_struct *work, struct wq_barrier *barr) pwq = worker->current_pwq; } + check_flush_dependency(pwq->wq, work); + insert_wq_barrier(pwq, barr, work, worker); spin_unlock_irq(&pool->lock); -- cgit From 03e0d4610bf4d4a93bfa16b2474ed4fd5243aa71 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 8 Dec 2015 11:28:04 -0500 Subject: watchdog: introduce touch_softlockup_watchdog_sched() touch_softlockup_watchdog() is used to tell watchdog that scheduler stall is expected. One group of usage is from paths where the task may not be able to yield for a long time such as performing slow PIO to finicky device and coming out of suspend. The other is to account for scheduler and timer going idle. For scheduler softlockup detection, there's no reason to distinguish the two cases; however, workqueue lockup detector is planned and it can use the same signals from the former group while the latter would spuriously prevent detection. This patch introduces a new function touch_softlockup_watchdog_sched() and convert the latter group to call it instead. For now, it just calls touch_softlockup_watchdog() and there's no functional difference. Signed-off-by: Tejun Heo Cc: Ulrich Obergfell Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Andrew Morton --- kernel/sched/clock.c | 2 +- kernel/time/tick-sched.c | 6 +++--- kernel/watchdog.c | 15 ++++++++++++++- 3 files changed, 18 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/clock.c b/kernel/sched/clock.c index c0a205101c23..bf1f37507a49 100644 --- a/kernel/sched/clock.c +++ b/kernel/sched/clock.c @@ -354,7 +354,7 @@ void sched_clock_idle_wakeup_event(u64 delta_ns) return; sched_clock_tick(); - touch_softlockup_watchdog(); + touch_softlockup_watchdog_sched(); } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 7c7ec4515983..58219f6ff3c6 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -143,7 +143,7 @@ static void tick_sched_handle(struct tick_sched *ts, struct pt_regs *regs) * when we go busy again does not account too much ticks. */ if (ts->tick_stopped) { - touch_softlockup_watchdog(); + touch_softlockup_watchdog_sched(); if (is_idle_task(current)) ts->idle_jiffies++; } @@ -430,7 +430,7 @@ static void tick_nohz_update_jiffies(ktime_t now) tick_do_update_jiffies64(now); local_irq_restore(flags); - touch_softlockup_watchdog(); + touch_softlockup_watchdog_sched(); } /* @@ -701,7 +701,7 @@ static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now) update_cpu_load_nohz(); calc_load_exit_idle(); - touch_softlockup_watchdog(); + touch_softlockup_watchdog_sched(); /* * Cancel the scheduled timer and restore the tick */ diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 18f34cf75f74..9eaf3dbec7e8 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -225,7 +225,15 @@ static void __touch_watchdog(void) __this_cpu_write(watchdog_touch_ts, get_timestamp()); } -void touch_softlockup_watchdog(void) +/** + * touch_softlockup_watchdog_sched - touch watchdog on scheduler stalls + * + * Call when the scheduler may have stalled for legitimate reasons + * preventing the watchdog task from executing - e.g. the scheduler + * entering idle state. This should only be used for scheduler events. + * Use touch_softlockup_watchdog() for everything else. + */ +void touch_softlockup_watchdog_sched(void) { /* * Preemption can be enabled. It doesn't matter which CPU's timestamp @@ -233,6 +241,11 @@ void touch_softlockup_watchdog(void) */ raw_cpu_write(watchdog_touch_ts, 0); } + +void touch_softlockup_watchdog(void) +{ + touch_softlockup_watchdog_sched(); +} EXPORT_SYMBOL(touch_softlockup_watchdog); void touch_all_softlockup_watchdogs(void) -- cgit From 82607adcf9cdf40fb7b5331269780c8f70ec6e35 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 8 Dec 2015 11:28:04 -0500 Subject: workqueue: implement lockup detector Workqueue stalls can happen from a variety of usage bugs such as missing WQ_MEM_RECLAIM flag or concurrency managed work item indefinitely staying RUNNING. These stalls can be extremely difficult to hunt down because the usual warning mechanisms can't detect workqueue stalls and the internal state is pretty opaque. To alleviate the situation, this patch implements workqueue lockup detector. It periodically monitors all worker_pools periodically and, if any pool failed to make forward progress longer than the threshold duration, triggers warning and dumps workqueue state as follows. BUG: workqueue lockup - pool cpus=0 node=0 flags=0x0 nice=0 stuck for 31s! Showing busy workqueues and worker pools: workqueue events: flags=0x0 pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=17/256 pending: monkey_wrench_fn, e1000_watchdog, cache_reap, vmstat_shepherd, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, release_one_tty, cgroup_release_agent workqueue events_power_efficient: flags=0x80 pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=2/256 pending: check_lifetime, neigh_periodic_work workqueue cgroup_pidlist_destroy: flags=0x0 pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=1/1 pending: cgroup_pidlist_destroy_work_fn ... The detection mechanism is controller through kernel parameter workqueue.watchdog_thresh and can be updated at runtime through the sysfs module parameter file. v2: Decoupled from softlockup control knobs. Signed-off-by: Tejun Heo Acked-by: Don Zickus Cc: Ulrich Obergfell Cc: Michal Hocko Cc: Chris Mason Cc: Andrew Morton --- kernel/watchdog.c | 3 + kernel/workqueue.c | 174 ++++++++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 174 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 9eaf3dbec7e8..b04f680c4735 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -20,6 +20,7 @@ #include #include #include +#include #include #include @@ -245,6 +246,7 @@ void touch_softlockup_watchdog_sched(void) void touch_softlockup_watchdog(void) { touch_softlockup_watchdog_sched(); + wq_watchdog_touch(raw_smp_processor_id()); } EXPORT_SYMBOL(touch_softlockup_watchdog); @@ -259,6 +261,7 @@ void touch_all_softlockup_watchdogs(void) */ for_each_watchdog_cpu(cpu) per_cpu(watchdog_touch_ts, cpu) = 0; + wq_watchdog_touch(-1); } #ifdef CONFIG_HARDLOCKUP_DETECTOR diff --git a/kernel/workqueue.c b/kernel/workqueue.c index c7769c507bf5..1ecb588aae07 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -148,6 +148,8 @@ struct worker_pool { int id; /* I: pool ID */ unsigned int flags; /* X: flags */ + unsigned long watchdog_ts; /* L: watchdog timestamp */ + struct list_head worklist; /* L: list of pending works */ int nr_workers; /* L: total number of workers */ @@ -1083,6 +1085,8 @@ static void pwq_activate_delayed_work(struct work_struct *work) struct pool_workqueue *pwq = get_work_pwq(work); trace_workqueue_activate_work(work); + if (list_empty(&pwq->pool->worklist)) + pwq->pool->watchdog_ts = jiffies; move_linked_works(work, &pwq->pool->worklist, NULL); __clear_bit(WORK_STRUCT_DELAYED_BIT, work_data_bits(work)); pwq->nr_active++; @@ -1385,6 +1389,8 @@ retry: trace_workqueue_activate_work(work); pwq->nr_active++; worklist = &pwq->pool->worklist; + if (list_empty(worklist)) + pwq->pool->watchdog_ts = jiffies; } else { work_flags |= WORK_STRUCT_DELAYED; worklist = &pwq->delayed_works; @@ -2157,6 +2163,8 @@ recheck: list_first_entry(&pool->worklist, struct work_struct, entry); + pool->watchdog_ts = jiffies; + if (likely(!(*work_data_bits(work) & WORK_STRUCT_LINKED))) { /* optimization path, not strictly necessary */ process_one_work(worker, work); @@ -2240,6 +2248,7 @@ repeat: struct pool_workqueue, mayday_node); struct worker_pool *pool = pwq->pool; struct work_struct *work, *n; + bool first = true; __set_current_state(TASK_RUNNING); list_del_init(&pwq->mayday_node); @@ -2256,9 +2265,14 @@ repeat: * process'em. */ WARN_ON_ONCE(!list_empty(scheduled)); - list_for_each_entry_safe(work, n, &pool->worklist, entry) - if (get_work_pwq(work) == pwq) + list_for_each_entry_safe(work, n, &pool->worklist, entry) { + if (get_work_pwq(work) == pwq) { + if (first) + pool->watchdog_ts = jiffies; move_linked_works(work, scheduled, &n); + } + first = false; + } if (!list_empty(scheduled)) { process_scheduled_works(rescuer); @@ -3104,6 +3118,7 @@ static int init_worker_pool(struct worker_pool *pool) pool->cpu = -1; pool->node = NUMA_NO_NODE; pool->flags |= POOL_DISASSOCIATED; + pool->watchdog_ts = jiffies; INIT_LIST_HEAD(&pool->worklist); INIT_LIST_HEAD(&pool->idle_list); hash_init(pool->busy_hash); @@ -4343,7 +4358,9 @@ void show_workqueue_state(void) pr_info("pool %d:", pool->id); pr_cont_pool_info(pool); - pr_cont(" workers=%d", pool->nr_workers); + pr_cont(" hung=%us workers=%d", + jiffies_to_msecs(jiffies - pool->watchdog_ts) / 1000, + pool->nr_workers); if (pool->manager) pr_cont(" manager: %d", task_pid_nr(pool->manager->task)); @@ -5202,6 +5219,154 @@ static void workqueue_sysfs_unregister(struct workqueue_struct *wq) static void workqueue_sysfs_unregister(struct workqueue_struct *wq) { } #endif /* CONFIG_SYSFS */ +/* + * Workqueue watchdog. + * + * Stall may be caused by various bugs - missing WQ_MEM_RECLAIM, illegal + * flush dependency, a concurrency managed work item which stays RUNNING + * indefinitely. Workqueue stalls can be very difficult to debug as the + * usual warning mechanisms don't trigger and internal workqueue state is + * largely opaque. + * + * Workqueue watchdog monitors all worker pools periodically and dumps + * state if some pools failed to make forward progress for a while where + * forward progress is defined as the first item on ->worklist changing. + * + * This mechanism is controlled through the kernel parameter + * "workqueue.watchdog_thresh" which can be updated at runtime through the + * corresponding sysfs parameter file. + */ +#ifdef CONFIG_WQ_WATCHDOG + +static void wq_watchdog_timer_fn(unsigned long data); + +static unsigned long wq_watchdog_thresh = 30; +static struct timer_list wq_watchdog_timer = + TIMER_DEFERRED_INITIALIZER(wq_watchdog_timer_fn, 0, 0); + +static unsigned long wq_watchdog_touched = INITIAL_JIFFIES; +static DEFINE_PER_CPU(unsigned long, wq_watchdog_touched_cpu) = INITIAL_JIFFIES; + +static void wq_watchdog_reset_touched(void) +{ + int cpu; + + wq_watchdog_touched = jiffies; + for_each_possible_cpu(cpu) + per_cpu(wq_watchdog_touched_cpu, cpu) = jiffies; +} + +static void wq_watchdog_timer_fn(unsigned long data) +{ + unsigned long thresh = READ_ONCE(wq_watchdog_thresh) * HZ; + bool lockup_detected = false; + struct worker_pool *pool; + int pi; + + if (!thresh) + return; + + rcu_read_lock(); + + for_each_pool(pool, pi) { + unsigned long pool_ts, touched, ts; + + if (list_empty(&pool->worklist)) + continue; + + /* get the latest of pool and touched timestamps */ + pool_ts = READ_ONCE(pool->watchdog_ts); + touched = READ_ONCE(wq_watchdog_touched); + + if (time_after(pool_ts, touched)) + ts = pool_ts; + else + ts = touched; + + if (pool->cpu >= 0) { + unsigned long cpu_touched = + READ_ONCE(per_cpu(wq_watchdog_touched_cpu, + pool->cpu)); + if (time_after(cpu_touched, ts)) + ts = cpu_touched; + } + + /* did we stall? */ + if (time_after(jiffies, ts + thresh)) { + lockup_detected = true; + pr_emerg("BUG: workqueue lockup - pool"); + pr_cont_pool_info(pool); + pr_cont(" stuck for %us!\n", + jiffies_to_msecs(jiffies - pool_ts) / 1000); + } + } + + rcu_read_unlock(); + + if (lockup_detected) + show_workqueue_state(); + + wq_watchdog_reset_touched(); + mod_timer(&wq_watchdog_timer, jiffies + thresh); +} + +void wq_watchdog_touch(int cpu) +{ + if (cpu >= 0) + per_cpu(wq_watchdog_touched_cpu, cpu) = jiffies; + else + wq_watchdog_touched = jiffies; +} + +static void wq_watchdog_set_thresh(unsigned long thresh) +{ + wq_watchdog_thresh = 0; + del_timer_sync(&wq_watchdog_timer); + + if (thresh) { + wq_watchdog_thresh = thresh; + wq_watchdog_reset_touched(); + mod_timer(&wq_watchdog_timer, jiffies + thresh * HZ); + } +} + +static int wq_watchdog_param_set_thresh(const char *val, + const struct kernel_param *kp) +{ + unsigned long thresh; + int ret; + + ret = kstrtoul(val, 0, &thresh); + if (ret) + return ret; + + if (system_wq) + wq_watchdog_set_thresh(thresh); + else + wq_watchdog_thresh = thresh; + + return 0; +} + +static const struct kernel_param_ops wq_watchdog_thresh_ops = { + .set = wq_watchdog_param_set_thresh, + .get = param_get_ulong, +}; + +module_param_cb(watchdog_thresh, &wq_watchdog_thresh_ops, &wq_watchdog_thresh, + 0644); + +static void wq_watchdog_init(void) +{ + wq_watchdog_set_thresh(wq_watchdog_thresh); +} + +#else /* CONFIG_WQ_WATCHDOG */ + +static inline void wq_watchdog_init(void) { } + +#endif /* CONFIG_WQ_WATCHDOG */ + static void __init wq_numa_init(void) { cpumask_var_t *tbl; @@ -5325,6 +5490,9 @@ static int __init init_workqueues(void) !system_unbound_wq || !system_freezable_wq || !system_power_efficient_wq || !system_freezable_power_efficient_wq); + + wq_watchdog_init(); + return 0; } early_initcall(init_workqueues); -- cgit From 52d189f1b38810b1b483d5bac2e4fa90b9afd372 Mon Sep 17 00:00:00 2001 From: Sasha Levin Date: Thu, 3 Dec 2015 15:46:48 -0500 Subject: ntp: Verify offset doesn't overflow in ntp_update_offset We need to make sure that the offset is valid before manipulating it, otherwise it might overflow on the multiplication. Cc: Sasha Levin Cc: Richard Cochran Cc: Thomas Gleixner Signed-off-by: Sasha Levin [jstultz: Reworked one of the checks so it makes more sense] Signed-off-by: John Stultz --- kernel/time/ntp.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c index 149cc8086aea..125fc0342355 100644 --- a/kernel/time/ntp.c +++ b/kernel/time/ntp.c @@ -297,15 +297,17 @@ static void ntp_update_offset(long offset) if (!(time_status & STA_PLL)) return; - if (!(time_status & STA_NANO)) + if (!(time_status & STA_NANO)) { + /* Make sure the multiplication below won't overflow */ + offset = clamp(offset, -USEC_PER_SEC, USEC_PER_SEC); offset *= NSEC_PER_USEC; + } /* * Scale the phase adjustment and * clamp to the operating range. */ - offset = min(offset, MAXPHASE); - offset = max(offset, -MAXPHASE); + offset = clamp(offset, -MAXPHASE, MAXPHASE); /* * Select how the frequency is to be controlled -- cgit From 37cf4dc3370fbca0344e23bb96446eb2c3548ba7 Mon Sep 17 00:00:00 2001 From: John Stultz Date: Thu, 3 Dec 2015 22:09:31 -0500 Subject: time: Verify time values in adjtimex ADJ_SETOFFSET to avoid overflow For adjtimex()'s ADJ_SETOFFSET, make sure the tv_usec value is sane. We might multiply them later which can cause an overflow and undefined behavior. This patch introduces new helper functions to simplify the checking code and adds comments to clarify Orginally this patch was by Sasha Levin, but I've basically rewritten it, so he should get credit for finding the issue and I should get the blame for any mistakes made since. Also, credit to Richard Cochran for the phrasing used in the comment for what is considered valid here. Cc: Sasha Levin Cc: Richard Cochran Cc: Thomas Gleixner Reported-by: Sasha Levin Signed-off-by: John Stultz --- kernel/time/ntp.c | 10 ++++++++-- kernel/time/timekeeping.c | 2 +- 2 files changed, 9 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c index 125fc0342355..4073c9550af9 100644 --- a/kernel/time/ntp.c +++ b/kernel/time/ntp.c @@ -676,8 +676,14 @@ int ntp_validate_timex(struct timex *txc) return -EINVAL; } - if ((txc->modes & ADJ_SETOFFSET) && (!capable(CAP_SYS_TIME))) - return -EPERM; + if (txc->modes & ADJ_SETOFFSET) { + /* In order to inject time, you gotta be super-user! */ + if (!capable(CAP_SYS_TIME)) + return -EPERM; + + if (!timeval_inject_offset_valid(&txc->time)) + return -EINVAL; + } /* * Check for potential multiplication overflows that can diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 99188ee5d9d0..d9249daf14ba 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -958,7 +958,7 @@ int timekeeping_inject_offset(struct timespec *ts) struct timespec64 ts64, tmp; int ret = 0; - if ((unsigned long)ts->tv_nsec >= NSEC_PER_SEC) + if (!timespec_inject_offset_valid(ts)) return -EINVAL; ts64 = timespec_to_timespec64(*ts); -- cgit From abc7e40c81d113ef4bacb556f0a77ca63ac81d85 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sun, 13 Dec 2015 18:12:30 +0100 Subject: genirq: Prevent chip buslock deadlock MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If a interrupt chip utilizes chip->buslock then free_irq() can deadlock in the following way: CPU0 CPU1 interrupt(X) (Shared or spurious) free_irq(X) interrupt_thread(X) chip_bus_lock(X) irq_finalize_oneshot(X) chip_bus_lock(X) synchronize_irq(X) synchronize_irq() waits for the interrupt thread to complete, i.e. forever. Solution is simple: Drop chip_bus_lock() before calling synchronize_irq() as we do with the irq_desc lock. There is nothing to be protected after the point where irq_desc lock has been released. This adds chip_bus_lock/unlock() to the remove_irq() code path, but that's actually correct in the case where remove_irq() is called on such an interrupt. The current users of remove_irq() are not affected as none of those interrupts is on a chip which requires buslock. Reported-by: Fredrik Markström Signed-off-by: Thomas Gleixner Cc: stable@vger.kernel.org --- kernel/irq/manage.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c index 0eebaeef317b..6ead200370da 100644 --- a/kernel/irq/manage.c +++ b/kernel/irq/manage.c @@ -1434,6 +1434,7 @@ static struct irqaction *__free_irq(unsigned int irq, void *dev_id) if (!desc) return NULL; + chip_bus_lock(desc); raw_spin_lock_irqsave(&desc->lock, flags); /* @@ -1447,7 +1448,7 @@ static struct irqaction *__free_irq(unsigned int irq, void *dev_id) if (!action) { WARN(1, "Trying to free already-free IRQ %d\n", irq); raw_spin_unlock_irqrestore(&desc->lock, flags); - + chip_bus_sync_unlock(desc); return NULL; } @@ -1475,6 +1476,7 @@ static struct irqaction *__free_irq(unsigned int irq, void *dev_id) #endif raw_spin_unlock_irqrestore(&desc->lock, flags); + chip_bus_sync_unlock(desc); unregister_handler_proc(irq, action); @@ -1553,9 +1555,7 @@ void free_irq(unsigned int irq, void *dev_id) desc->affinity_notify = NULL; #endif - chip_bus_lock(desc); kfree(__free_irq(irq, dev_id)); - chip_bus_sync_unlock(desc); } EXPORT_SYMBOL(free_irq); -- cgit From 425a5072dcd1bd895eea90a6b495392b6358ebd0 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sun, 13 Dec 2015 18:02:22 +0100 Subject: genirq: Free irq_desc with rcu The new VMD device driver needs to iterate over a list of "demultiplexing" interrupts. Protecting that list with a lock is not possible because the list is also required in code pathes which hold irq descriptor lock. Therefor the demultiplexing interrupt handler would create a lock inversion scenario if it calls a demux handler with the list protection lock held. A solution for this is to free the irq descriptor via RCU, so the list can be walked with rcu read lock held. Signed-off-by: Thomas Gleixner Cc: Keith Busch --- kernel/irq/irqdesc.c | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c index 239e2ae2c947..0409da0bcc33 100644 --- a/kernel/irq/irqdesc.c +++ b/kernel/irq/irqdesc.c @@ -159,6 +159,7 @@ static struct irq_desc *alloc_desc(int irq, int node, struct module *owner) raw_spin_lock_init(&desc->lock); lockdep_set_class(&desc->lock, &irq_desc_lock_class); + init_rcu_head(&desc->rcu); desc_set_defaults(irq, desc, node, owner); @@ -171,6 +172,15 @@ err_desc: return NULL; } +static void delayed_free_desc(struct rcu_head *rhp) +{ + struct irq_desc *desc = container_of(rhp, struct irq_desc, rcu); + + free_masks(desc); + free_percpu(desc->kstat_irqs); + kfree(desc); +} + static void free_desc(unsigned int irq) { struct irq_desc *desc = irq_to_desc(irq); @@ -187,9 +197,12 @@ static void free_desc(unsigned int irq) delete_irq_desc(irq); mutex_unlock(&sparse_irq_lock); - free_masks(desc); - free_percpu(desc->kstat_irqs); - kfree(desc); + /* + * We free the descriptor, masks and stat fields via RCU. That + * allows demultiplex interrupts to do rcu based management of + * the child interrupts. + */ + call_rcu(&desc->rcu, delayed_free_desc); } static int alloc_descs(unsigned int start, unsigned int cnt, int node, -- cgit From fccd3af57100027e5330079819987aade07631ad Mon Sep 17 00:00:00 2001 From: Rami Rosen Date: Sun, 13 Dec 2015 22:13:08 +0200 Subject: cgroup_pids: fix a typo. This patch fixes a typo in pids_charge() method. Signed-off-by: Rami Rosen Signed-off-by: Tejun Heo --- kernel/cgroup_pids.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/cgroup_pids.c b/kernel/cgroup_pids.c index 18107aea2895..303097b37429 100644 --- a/kernel/cgroup_pids.c +++ b/kernel/cgroup_pids.c @@ -134,7 +134,7 @@ static void pids_charge(struct pids_cgroup *pids, int num) * * This function follows the set limit. It will fail if the charge would cause * the new value to exceed the hierarchical limit. Returns 0 if the charge - * succeded, otherwise -EAGAIN. + * succeeded, otherwise -EAGAIN. */ static int pids_try_charge(struct pids_cgroup *pids, int num) { -- cgit From c466595c416c04036e1ba36ecdc5fe9072c76228 Mon Sep 17 00:00:00 2001 From: Marc Zyngier Date: Mon, 23 Nov 2015 08:26:04 +0000 Subject: irqdomain: Make irq_domain_alloc_irqs_recursive available We are soon going to need the MSI layer to call into the domain allocators. Instead of open coding this, make the standard irq_domain_alloc_irqs_recursive function available to the MSI layer. Signed-off-by: Marc Zyngier --- kernel/irq/irqdomain.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c index 22aa9612ef7c..1c9973e1b316 100644 --- a/kernel/irq/irqdomain.c +++ b/kernel/irq/irqdomain.c @@ -1125,9 +1125,9 @@ static void irq_domain_free_irqs_recursive(struct irq_domain *domain, } } -static int irq_domain_alloc_irqs_recursive(struct irq_domain *domain, - unsigned int irq_base, - unsigned int nr_irqs, void *arg) +int irq_domain_alloc_irqs_recursive(struct irq_domain *domain, + unsigned int irq_base, + unsigned int nr_irqs, void *arg) { int ret = 0; struct irq_domain *parent = domain->parent; -- cgit From b2eba39bcab9d60a6c3b80c7fc2f3dacb77eeaae Mon Sep 17 00:00:00 2001 From: Marc Zyngier Date: Mon, 23 Nov 2015 08:26:05 +0000 Subject: genirq/msi: Make the .prepare callback reusable The .prepare callbacks are so far only called from msi_domain_alloc_irqs. In order to reuse that code, split that code and create a msi_domain_prepare_irqs function that the existing code can call into. Signed-off-by: Marc Zyngier --- kernel/irq/msi.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/irq/msi.c b/kernel/irq/msi.c index 6b0c0b74a2a1..9a85613d4227 100644 --- a/kernel/irq/msi.c +++ b/kernel/irq/msi.c @@ -252,6 +252,20 @@ struct irq_domain *msi_create_irq_domain(struct fwnode_handle *fwnode, &msi_domain_ops, info); } +int msi_domain_prepare_irqs(struct irq_domain *domain, struct device *dev, + int nvec, msi_alloc_info_t *arg) +{ + struct msi_domain_info *info = domain->host_data; + struct msi_domain_ops *ops = info->ops; + int ret; + + ret = ops->msi_check(domain, info, dev); + if (ret == 0) + ret = ops->msi_prepare(domain, dev, nvec, arg); + + return ret; +} + /** * msi_domain_alloc_irqs - Allocate interrupts from a MSI interrupt domain * @domain: The domain to allocate from @@ -270,9 +284,7 @@ int msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, struct msi_desc *desc; int i, ret, virq = -1; - ret = ops->msi_check(domain, info, dev); - if (ret == 0) - ret = ops->msi_prepare(domain, dev, nvec, &arg); + ret = msi_domain_prepare_irqs(domain, dev, nvec, &arg); if (ret) return ret; -- cgit From 2145ac9310b60c1c11294b7bea10fe154009be1d Mon Sep 17 00:00:00 2001 From: Marc Zyngier Date: Mon, 23 Nov 2015 08:26:06 +0000 Subject: genirq/msi: Add msi_domain_populate_irqs To be able to allocate interrupts from the MSI layer down, add a new msi_domain_populate_irqs entry point. Signed-off-by: Marc Zyngier --- kernel/irq/msi.c | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) (limited to 'kernel') diff --git a/kernel/irq/msi.c b/kernel/irq/msi.c index 9a85613d4227..15b249e7c673 100644 --- a/kernel/irq/msi.c +++ b/kernel/irq/msi.c @@ -266,6 +266,46 @@ int msi_domain_prepare_irqs(struct irq_domain *domain, struct device *dev, return ret; } +int msi_domain_populate_irqs(struct irq_domain *domain, struct device *dev, + int virq, int nvec, msi_alloc_info_t *arg) +{ + struct msi_domain_info *info = domain->host_data; + struct msi_domain_ops *ops = info->ops; + struct msi_desc *desc; + int ret = 0; + + for_each_msi_entry(desc, dev) { + /* Don't even try the multi-MSI brain damage. */ + if (WARN_ON(!desc->irq || desc->nvec_used != 1)) { + ret = -EINVAL; + break; + } + + if (!(desc->irq >= virq && desc->irq < (virq + nvec))) + continue; + + ops->set_desc(arg, desc); + /* Assumes the domain mutex is held! */ + ret = irq_domain_alloc_irqs_recursive(domain, virq, 1, arg); + if (ret) + break; + + irq_set_msi_desc_off(virq, 0, desc); + } + + if (ret) { + /* Mop up the damage */ + for_each_msi_entry(desc, dev) { + if (!(desc->irq >= virq && desc->irq < (virq + nvec))) + continue; + + irq_domain_free_irqs_common(domain, desc->irq, 1); + } + } + + return ret; +} + /** * msi_domain_alloc_irqs - Allocate interrupts from a MSI interrupt domain * @domain: The domain to allocate from -- cgit From dee3665416a8553279d10b62b5e62685cbe5daa8 Mon Sep 17 00:00:00 2001 From: DengChao Date: Sun, 13 Dec 2015 12:24:18 +0800 Subject: timekeeping: Provide internal function __ktime_get_real_seconds In order to fix Y2038 issues in the ntp code we will need replace get_seconds() with ktime_get_real_seconds() but as the ntp code uses the timekeeping lock which is also used by ktime_get_real_seconds(), we need a version without locking. Add a new function __ktime_get_real_seconds() in timekeeping to do this. Reviewed-by: John Stultz Signed-off-by: DengChao Signed-off-by: John Stultz --- kernel/time/timekeeping.c | 13 +++++++++++++ kernel/time/timekeeping_internal.h | 2 ++ 2 files changed, 15 insertions(+) (limited to 'kernel') diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index d9249daf14ba..21cc23918cbd 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -845,6 +845,19 @@ time64_t ktime_get_real_seconds(void) } EXPORT_SYMBOL_GPL(ktime_get_real_seconds); +/** + * __ktime_get_real_seconds - The same as ktime_get_real_seconds + * but without the sequence counter protect. This internal function + * is called just when timekeeping lock is already held. + */ +time64_t __ktime_get_real_seconds(void) +{ + struct timekeeper *tk = &tk_core.timekeeper; + + return tk->xtime_sec; +} + + #ifdef CONFIG_NTP_PPS /** diff --git a/kernel/time/timekeeping_internal.h b/kernel/time/timekeeping_internal.h index 4ea005a7f9da..e20466ffc208 100644 --- a/kernel/time/timekeeping_internal.h +++ b/kernel/time/timekeeping_internal.h @@ -26,4 +26,6 @@ static inline cycle_t clocksource_delta(cycle_t now, cycle_t last, cycle_t mask) } #endif +extern time64_t __ktime_get_real_seconds(void); + #endif /* _TIMEKEEPING_INTERNAL_H */ -- cgit From 0af864651b459afb0435ee8786a19cbe5a044cdb Mon Sep 17 00:00:00 2001 From: DengChao Date: Sun, 13 Dec 2015 12:24:19 +0800 Subject: ntp: Change time_reftime to time64_t and utilize 64bit __ktime_get_real_seconds The type of static variant "time_reftime" and the call of get_seconds in ntp are both not y2038 safe. So change the type of time_reftime to time64_t and replace get_seconds with __ktime_get_real_seconds. The local variant "secs" in ntp_update_offset represents seconds between now and last ntp adjustment, it seems impossible that this time will last more than 68 years, so keep its type as "long". Reviewed-by: John Stultz Signed-off-by: DengChao [jstultz: Tweaked commit message] Signed-off-by: John Stultz --- kernel/time/ntp.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c index 4073c9550af9..e947bfddd2c2 100644 --- a/kernel/time/ntp.c +++ b/kernel/time/ntp.c @@ -18,6 +18,8 @@ #include #include "ntp_internal.h" +#include "timekeeping_internal.h" + /* * NTP timekeeping variables: @@ -70,7 +72,7 @@ static long time_esterror = NTP_PHASE_LIMIT; static s64 time_freq; /* time at last adjustment (secs): */ -static long time_reftime; +static time64_t time_reftime; static long time_adjust; @@ -313,11 +315,11 @@ static void ntp_update_offset(long offset) * Select how the frequency is to be controlled * and in which mode (PLL or FLL). */ - secs = get_seconds() - time_reftime; + secs = (long)(__ktime_get_real_seconds() - time_reftime); if (unlikely(time_status & STA_FREQHOLD)) secs = 0; - time_reftime = get_seconds(); + time_reftime = __ktime_get_real_seconds(); offset64 = offset; freq_adj = ntp_update_offset_fll(offset64, secs); @@ -592,7 +594,7 @@ static inline void process_adj_status(struct timex *txc, struct timespec64 *ts) * reference time to current time. */ if (!(time_status & STA_PLL) && (txc->status & STA_PLL)) - time_reftime = get_seconds(); + time_reftime = __ktime_get_real_seconds(); /* only set allowed bits */ time_status &= STA_RONLY; -- cgit From c796348774f15c6e682834ed288bcae0f2c95707 Mon Sep 17 00:00:00 2001 From: DengChao Date: Sun, 13 Dec 2015 12:26:42 +0800 Subject: ntp: Fix second_overflow's input parameter type to be 64bits The function "second_overflow" uses "unsign long" as its input parameter type which will overflow after year 2106 on 32bit systems. Thus this patch replaces it with time64_t type. While the 64-bit division is expensive, "next_ntp_leap_sec" has been calculated already, so we can just re-use it in the TIME_INS/DEL cases, allowing one expensive division per leapsecond instead of re-doing the divsion once a second after the leap flag has been set. Signed-off-by: DengChao [jstultz: Tweaked commit message] Signed-off-by: John Stultz --- kernel/time/ntp.c | 16 +++++++++------- kernel/time/ntp_internal.h | 2 +- 2 files changed, 10 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c index e947bfddd2c2..36f2ca09aa5e 100644 --- a/kernel/time/ntp.c +++ b/kernel/time/ntp.c @@ -16,6 +16,7 @@ #include #include #include +#include #include "ntp_internal.h" #include "timekeeping_internal.h" @@ -394,10 +395,11 @@ ktime_t ntp_get_next_leap(void) * * Also handles leap second processing, and returns leap offset */ -int second_overflow(unsigned long secs) +int second_overflow(time64_t secs) { s64 delta; int leap = 0; + s32 rem; /* * Leap second processing. If in leap-insert state at the end of the @@ -408,19 +410,19 @@ int second_overflow(unsigned long secs) case TIME_OK: if (time_status & STA_INS) { time_state = TIME_INS; - ntp_next_leap_sec = secs + SECS_PER_DAY - - (secs % SECS_PER_DAY); + div_s64_rem(secs, SECS_PER_DAY, &rem); + ntp_next_leap_sec = secs + SECS_PER_DAY - rem; } else if (time_status & STA_DEL) { time_state = TIME_DEL; - ntp_next_leap_sec = secs + SECS_PER_DAY - - ((secs+1) % SECS_PER_DAY); + div_s64_rem(secs + 1, SECS_PER_DAY, &rem); + ntp_next_leap_sec = secs + SECS_PER_DAY - rem; } break; case TIME_INS: if (!(time_status & STA_INS)) { ntp_next_leap_sec = TIME64_MAX; time_state = TIME_OK; - } else if (secs % SECS_PER_DAY == 0) { + } else if (secs == ntp_next_leap_sec) { leap = -1; time_state = TIME_OOP; printk(KERN_NOTICE @@ -431,7 +433,7 @@ int second_overflow(unsigned long secs) if (!(time_status & STA_DEL)) { ntp_next_leap_sec = TIME64_MAX; time_state = TIME_OK; - } else if ((secs + 1) % SECS_PER_DAY == 0) { + } else if (secs == ntp_next_leap_sec) { leap = 1; ntp_next_leap_sec = TIME64_MAX; time_state = TIME_WAIT; diff --git a/kernel/time/ntp_internal.h b/kernel/time/ntp_internal.h index af924470eac0..d8a7c11fa71a 100644 --- a/kernel/time/ntp_internal.h +++ b/kernel/time/ntp_internal.h @@ -6,7 +6,7 @@ extern void ntp_clear(void); /* Returns how long ticks are at present, in ns / 2^NTP_SCALE_SHIFT. */ extern u64 ntp_tick_length(void); extern ktime_t ntp_get_next_leap(void); -extern int second_overflow(unsigned long secs); +extern int second_overflow(time64_t secs); extern int ntp_validate_timex(struct timex *); extern int __do_adjtimex(struct timex *, struct timespec64 *, s32 *); extern void __hardpps(const struct timespec64 *, const struct timespec64 *); -- cgit From ec02b076ceab63f99e5b3d80fd223d777266c236 Mon Sep 17 00:00:00 2001 From: John Stultz Date: Thu, 3 Dec 2015 10:23:30 -0800 Subject: timekeeping: Cap adjustments so they don't exceed the maxadj value Thus its been occasionally noted that users have seen confusing warnings like: Adjusting tsc more than 11% (5941981 vs 7759439) We try to limit the maximum total adjustment to 11% (10% tick adjustment + 0.5% frequency adjustment). But this is done by bounding the requested adjustment values, and the internal steering that is done by tracking the error from what was requested and what was applied, does not have any such limits. This is usually not problematic, but in some cases has a risk that an adjustment could cause the clocksource mult value to overflow, so its an indication things are outside of what is expected. It ends up most of the reports of this 11% warning are on systems using chrony, which utilizes the adjtimex() ADJ_TICK interface (which allows a +-10% adjustment). The original rational for ADJ_TICK unclear to me but my assumption it was originally added to allow broken systems to get a big constant correction at boot (see adjtimex userspace package for an example) which would allow the system to work w/ ntpd's 0.5% adjustment limit. Chrony uses ADJ_TICK to make very aggressive short term corrections (usually right at startup). Which push us close enough to the max bound that a few late ticks can cause the internal steering to push past the max adjust value (tripping the warning). Thus this patch adds some extra logic to enforce the max adjustment cap in the internal steering. Note: This has the potential to slow corrections when the ADJ_TICK value is furthest away from the default value. So it would be good to get some testing from folks using chrony, to make sure we don't cause any troubles there. Cc: Miroslav Lichvar Cc: Thomas Gleixner Cc: Richard Cochran Cc: Prarit Bhargava Cc: Andy Lutomirski Tested-by: Miroslav Lichvar Reported-by: Andy Lutomirski Signed-off-by: John Stultz --- kernel/time/timekeeping.c | 31 +++++++++++++++++++++++++++---- 1 file changed, 27 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 21cc23918cbd..34b4cedfa80d 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -1604,9 +1604,12 @@ static __always_inline void timekeeping_freqadjust(struct timekeeper *tk, { s64 interval = tk->cycle_interval; s64 xinterval = tk->xtime_interval; + u32 base = tk->tkr_mono.clock->mult; + u32 max = tk->tkr_mono.clock->maxadj; + u32 cur_adj = tk->tkr_mono.mult; s64 tick_error; bool negative; - u32 adj; + u32 adj_scale; /* Remove any current error adj from freq calculation */ if (tk->ntp_err_mult) @@ -1625,13 +1628,33 @@ static __always_inline void timekeeping_freqadjust(struct timekeeper *tk, /* preserve the direction of correction */ negative = (tick_error < 0); - /* Sort out the magnitude of the correction */ + /* If any adjustment would pass the max, just return */ + if (negative && (cur_adj - 1) <= (base - max)) + return; + if (!negative && (cur_adj + 1) >= (base + max)) + return; + /* + * Sort out the magnitude of the correction, but + * avoid making so large a correction that we go + * over the max adjustment. + */ + adj_scale = 0; tick_error = abs(tick_error); - for (adj = 0; tick_error > interval; adj++) + while (tick_error > interval) { + u32 adj = 1 << (adj_scale + 1); + + /* Check if adjustment gets us within 1 unit from the max */ + if (negative && (cur_adj - adj) <= (base - max)) + break; + if (!negative && (cur_adj + adj) >= (base + max)) + break; + + adj_scale++; tick_error >>= 1; + } /* scale the corrections */ - timekeeping_apply_adjustment(tk, offset, negative, adj); + timekeeping_apply_adjustment(tk, offset, negative, adj_scale); } /* -- cgit From 1717f2096b543cede7a380c858c765c41936bc35 Mon Sep 17 00:00:00 2001 From: Hidehiro Kawai Date: Mon, 14 Dec 2015 11:19:09 +0100 Subject: panic, x86: Fix re-entrance problem due to panic on NMI If panic on NMI happens just after panic() on the same CPU, panic() is recursively called. Kernel stalls, as a result, after failing to acquire panic_lock. To avoid this problem, don't call panic() in NMI context if we've already entered panic(). For that, introduce nmi_panic() macro to reduce code duplication. In the case of panic on NMI, don't return from NMI handlers if another CPU already panicked. Signed-off-by: Hidehiro Kawai Acked-by: Michal Hocko Cc: Aaron Tomlin Cc: Andrew Morton Cc: Andy Lutomirski Cc: Baoquan He Cc: Chris Metcalf Cc: David Hildenbrand Cc: Don Zickus Cc: "Eric W. Biederman" Cc: Frederic Weisbecker Cc: Gobinda Charan Maji Cc: HATAYAMA Daisuke Cc: "H. Peter Anvin" Cc: Ingo Molnar Cc: Javi Merino Cc: Jonathan Corbet Cc: kexec@lists.infradead.org Cc: linux-doc@vger.kernel.org Cc: lkml Cc: Masami Hiramatsu Cc: Michal Nazarewicz Cc: Nicolas Iooss Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Rasmus Villemoes Cc: Rusty Russell Cc: Seth Jennings Cc: Steven Rostedt Cc: Thomas Gleixner Cc: Ulrich Obergfell Cc: Vitaly Kuznetsov Cc: Vivek Goyal Link: http://lkml.kernel.org/r/20151210014626.25437.13302.stgit@softrs [ Cleanup comments, fixup formatting. ] Signed-off-by: Borislav Petkov Signed-off-by: Thomas Gleixner --- kernel/panic.c | 16 +++++++++++++--- kernel/watchdog.c | 2 +- 2 files changed, 14 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/panic.c b/kernel/panic.c index 4b150bc0c6c1..3344524cf6ff 100644 --- a/kernel/panic.c +++ b/kernel/panic.c @@ -61,6 +61,8 @@ void __weak panic_smp_self_stop(void) cpu_relax(); } +atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID); + /** * panic - halt the system * @fmt: The text string to print @@ -71,17 +73,17 @@ void __weak panic_smp_self_stop(void) */ void panic(const char *fmt, ...) { - static DEFINE_SPINLOCK(panic_lock); static char buf[1024]; va_list args; long i, i_next = 0; int state = 0; + int old_cpu, this_cpu; /* * Disable local interrupts. This will prevent panic_smp_self_stop * from deadlocking the first cpu that invokes the panic, since * there is nothing to prevent an interrupt handler (that runs - * after the panic_lock is acquired) from invoking panic again. + * after setting panic_cpu) from invoking panic() again. */ local_irq_disable(); @@ -94,8 +96,16 @@ void panic(const char *fmt, ...) * multiple parallel invocations of panic, all other CPUs either * stop themself or will wait until they are stopped by the 1st CPU * with smp_send_stop(). + * + * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which + * comes here, so go ahead. + * `old_cpu == this_cpu' means we came from nmi_panic() which sets + * panic_cpu to this CPU. In this case, this is also the 1st CPU. */ - if (!spin_trylock(&panic_lock)) + this_cpu = raw_smp_processor_id(); + old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu); + + if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu) panic_smp_self_stop(); console_verbose(); diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 18f34cf75f74..b9be18fae154 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -351,7 +351,7 @@ static void watchdog_overflow_callback(struct perf_event *event, trigger_allbutself_cpu_backtrace(); if (hardlockup_panic) - panic("Hard LOCKUP"); + nmi_panic("Hard LOCKUP"); __this_cpu_write(hard_watchdog_warn, true); return; -- cgit From 58c5661f2144c089bbc2e5d87c9ec1dc1d2964fe Mon Sep 17 00:00:00 2001 From: Hidehiro Kawai Date: Mon, 14 Dec 2015 11:19:10 +0100 Subject: panic, x86: Allow CPUs to save registers even if looping in NMI context Currently, kdump_nmi_shootdown_cpus(), a subroutine of crash_kexec(), sends an NMI IPI to CPUs which haven't called panic() to stop them, save their register information and do some cleanups for crash dumping. However, if such a CPU is infinitely looping in NMI context, we fail to save its register information into the crash dump. For example, this can happen when unknown NMIs are broadcast to all CPUs as follows: CPU 0 CPU 1 =========================== ========================== receive an unknown NMI unknown_nmi_error() panic() receive an unknown NMI spin_trylock(&panic_lock) unknown_nmi_error() crash_kexec() panic() spin_trylock(&panic_lock) panic_smp_self_stop() infinite loop kdump_nmi_shootdown_cpus() issue NMI IPI -----------> blocked until IRET infinite loop... Here, since CPU 1 is in NMI context, the second NMI from CPU 0 is blocked until CPU 1 executes IRET. However, CPU 1 never executes IRET, so the NMI is not handled and the callback function to save registers is never called. In practice, this can happen on some servers which broadcast NMIs to all CPUs when the NMI button is pushed. To save registers in this case, we need to: a) Return from NMI handler instead of looping infinitely or b) Call the callback function directly from the infinite loop Inherently, a) is risky because NMI is also used to prevent corrupted data from being propagated to devices. So, we chose b). This patch does the following: 1. Move the infinite looping of CPUs which haven't called panic() in NMI context (actually done by panic_smp_self_stop()) outside of panic() to enable us to refer pt_regs. Please note that panic_smp_self_stop() is still used for normal context. 2. Call a callback of kdump_nmi_shootdown_cpus() directly to save registers and do some cleanups after setting waiting_for_crash_ipi which is used for counting down the number of CPUs which handled the callback Signed-off-by: Hidehiro Kawai Acked-by: Michal Hocko Cc: Aaron Tomlin Cc: Andrew Morton Cc: Andy Lutomirski Cc: Baoquan He Cc: Chris Metcalf Cc: Dave Young Cc: David Hildenbrand Cc: Don Zickus Cc: Eric Biederman Cc: Frederic Weisbecker Cc: Gobinda Charan Maji Cc: HATAYAMA Daisuke Cc: Hidehiro Kawai Cc: "H. Peter Anvin" Cc: Ingo Molnar Cc: Javi Merino Cc: Jiang Liu Cc: Jonathan Corbet Cc: kexec@lists.infradead.org Cc: linux-doc@vger.kernel.org Cc: lkml Cc: Masami Hiramatsu Cc: Michal Nazarewicz Cc: Nicolas Iooss Cc: Oleg Nesterov Cc: Peter Zijlstra Cc: Prarit Bhargava Cc: Rasmus Villemoes Cc: Seth Jennings Cc: Stefan Lippers-Hollmann Cc: Steven Rostedt Cc: Thomas Gleixner Cc: Ulrich Obergfell Cc: Vitaly Kuznetsov Cc: Vivek Goyal Cc: Yasuaki Ishimatsu Link: http://lkml.kernel.org/r/20151210014628.25437.75256.stgit@softrs [ Cleanup comments, fixup formatting. ] Signed-off-by: Borislav Petkov Signed-off-by: Thomas Gleixner --- kernel/panic.c | 9 +++++++++ kernel/watchdog.c | 2 +- 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/panic.c b/kernel/panic.c index 3344524cf6ff..06f31b49b3b4 100644 --- a/kernel/panic.c +++ b/kernel/panic.c @@ -61,6 +61,15 @@ void __weak panic_smp_self_stop(void) cpu_relax(); } +/* + * Stop ourselves in NMI context if another CPU has already panicked. Arch code + * may override this to prepare for crash dumping, e.g. save regs info. + */ +void __weak nmi_panic_self_stop(struct pt_regs *regs) +{ + panic_smp_self_stop(); +} + atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID); /** diff --git a/kernel/watchdog.c b/kernel/watchdog.c index b9be18fae154..84b5035cb6a5 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -351,7 +351,7 @@ static void watchdog_overflow_callback(struct perf_event *event, trigger_allbutself_cpu_backtrace(); if (hardlockup_panic) - nmi_panic("Hard LOCKUP"); + nmi_panic(regs, "Hard LOCKUP"); __this_cpu_write(hard_watchdog_warn, true); return; -- cgit From 7bbee5ca3896f69f09c68be549cb8997abe6bca6 Mon Sep 17 00:00:00 2001 From: Hidehiro Kawai Date: Mon, 14 Dec 2015 11:19:11 +0100 Subject: kexec: Fix race between panic() and crash_kexec() Currently, panic() and crash_kexec() can be called at the same time. For example (x86 case): CPU 0: oops_end() crash_kexec() mutex_trylock() // acquired nmi_shootdown_cpus() // stop other CPUs CPU 1: panic() crash_kexec() mutex_trylock() // failed to acquire smp_send_stop() // stop other CPUs infinite loop If CPU 1 calls smp_send_stop() before nmi_shootdown_cpus(), kdump fails. In another case: CPU 0: oops_end() crash_kexec() mutex_trylock() // acquired io_check_error() panic() crash_kexec() mutex_trylock() // failed to acquire infinite loop Clearly, this is an undesirable result. To fix this problem, this patch changes crash_kexec() to exclude others by using the panic_cpu atomic. Signed-off-by: Hidehiro Kawai Acked-by: Michal Hocko Cc: Andrew Morton Cc: Baoquan He Cc: Dave Young Cc: "Eric W. Biederman" Cc: HATAYAMA Daisuke Cc: "H. Peter Anvin" Cc: Ingo Molnar Cc: Jonathan Corbet Cc: kexec@lists.infradead.org Cc: linux-doc@vger.kernel.org Cc: Martin Schwidefsky Cc: Masami Hiramatsu Cc: Minfei Huang Cc: Peter Zijlstra Cc: Seth Jennings Cc: Steven Rostedt Cc: Thomas Gleixner Cc: Vitaly Kuznetsov Cc: Vivek Goyal Cc: x86-ml Link: http://lkml.kernel.org/r/20151210014630.25437.94161.stgit@softrs Signed-off-by: Borislav Petkov Signed-off-by: Thomas Gleixner --- kernel/kexec_core.c | 30 +++++++++++++++++++++++++++++- kernel/panic.c | 8 ++++++-- 2 files changed, 35 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c index 11b64a63c0f8..c823f3001e12 100644 --- a/kernel/kexec_core.c +++ b/kernel/kexec_core.c @@ -853,7 +853,12 @@ struct kimage *kexec_image; struct kimage *kexec_crash_image; int kexec_load_disabled; -void crash_kexec(struct pt_regs *regs) +/* + * No panic_cpu check version of crash_kexec(). This function is called + * only when panic_cpu holds the current CPU number; this is the only CPU + * which processes crash_kexec routines. + */ +void __crash_kexec(struct pt_regs *regs) { /* Take the kexec_mutex here to prevent sys_kexec_load * running on one cpu from replacing the crash kernel @@ -876,6 +881,29 @@ void crash_kexec(struct pt_regs *regs) } } +void crash_kexec(struct pt_regs *regs) +{ + int old_cpu, this_cpu; + + /* + * Only one CPU is allowed to execute the crash_kexec() code as with + * panic(). Otherwise parallel calls of panic() and crash_kexec() + * may stop each other. To exclude them, we use panic_cpu here too. + */ + this_cpu = raw_smp_processor_id(); + old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu); + if (old_cpu == PANIC_CPU_INVALID) { + /* This is the 1st CPU which comes here, so go ahead. */ + __crash_kexec(regs); + + /* + * Reset panic_cpu to allow another panic()/crash_kexec() + * call. + */ + atomic_set(&panic_cpu, PANIC_CPU_INVALID); + } +} + size_t crash_get_memory_size(void) { size_t size = 0; diff --git a/kernel/panic.c b/kernel/panic.c index 06f31b49b3b4..b333380c6bb2 100644 --- a/kernel/panic.c +++ b/kernel/panic.c @@ -136,9 +136,11 @@ void panic(const char *fmt, ...) * everything else. * If we want to run this after calling panic_notifiers, pass * the "crash_kexec_post_notifiers" option to the kernel. + * + * Bypass the panic_cpu check and call __crash_kexec directly. */ if (!crash_kexec_post_notifiers) - crash_kexec(NULL); + __crash_kexec(NULL); /* * Note smp_send_stop is the usual smp shutdown function, which @@ -161,9 +163,11 @@ void panic(const char *fmt, ...) * panic_notifiers and dumping kmsg before kdump. * Note: since some panic_notifiers can make crashed kernel * more unstable, it can increase risks of the kdump failure too. + * + * Bypass the panic_cpu check and call __crash_kexec directly. */ if (crash_kexec_post_notifiers) - crash_kexec(NULL); + __crash_kexec(NULL); bust_spinlocks(0); -- cgit From 1f45f1f33c8c8b96722dbc5e6b7acf74eaa721f7 Mon Sep 17 00:00:00 2001 From: Yang Yingliang Date: Sat, 31 Oct 2015 18:20:55 +0800 Subject: clocksource: Make clocksource validation work for all clocksources The clocksource validation which makes sure that the newly read value is not smaller than the last value only works if the clocksource mask is 64bit, i.e. the counter is 64bit wide. But we want to use that mechanism also for clocksources which are less than 64bit wide. So instead of checking whether bit 63 is set, we check whether the most significant bit of the clocksource mask is set in the delta result. If it is set, we return 0. [ tglx: Simplified the implementation, added a comment and massaged the commit message ] Suggested-by: Thomas Gleixner Signed-off-by: Yang Yingliang Cc: Link: http://lkml.kernel.org/r/56349607.6070708@huawei.com Signed-off-by: Thomas Gleixner --- kernel/time/timekeeping_internal.h | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/time/timekeeping_internal.h b/kernel/time/timekeeping_internal.h index e20466ffc208..5be76270ec4a 100644 --- a/kernel/time/timekeeping_internal.h +++ b/kernel/time/timekeeping_internal.h @@ -17,7 +17,11 @@ static inline cycle_t clocksource_delta(cycle_t now, cycle_t last, cycle_t mask) { cycle_t ret = (now - last) & mask; - return (s64) ret > 0 ? ret : 0; + /* + * Prevent time going backwards by checking the MSB of mask in + * the result. If set, return 0. + */ + return ret & ~(mask >> 1) ? 0 : ret; } #else static inline cycle_t clocksource_delta(cycle_t now, cycle_t last, cycle_t mask) -- cgit From a4289dc2ec3a5821076a78ee9678909b4eff297e Mon Sep 17 00:00:00 2001 From: Jake Oshins Date: Thu, 10 Dec 2015 17:52:59 +0000 Subject: genirq/msi: Export functions to allow MSI domains in modules The Linux kernel already has the concept of IRQ domain, wherein a component can expose a set of IRQs which are managed by a particular interrupt controller chip or other subsystem. The PCI driver exposes the notion of an IRQ domain for Message-Signaled Interrupts (MSI) from PCI Express devices. This patch exposes the functions which are necessary for creating a MSI IRQ domain within a module. [ tglx: Split it into x86 and core irq parts ] Signed-off-by: Jake Oshins Cc: gregkh@linuxfoundation.org Cc: kys@microsoft.com Cc: devel@linuxdriverproject.org Cc: olaf@aepfle.de Cc: apw@canonical.com Cc: vkuznets@redhat.com Cc: haiyangz@microsoft.com Cc: marc.zyngier@arm.com Cc: bhelgaas@google.com Link: http://lkml.kernel.org/r/1449769983-12948-4-git-send-email-jakeo@microsoft.com Signed-off-by: Thomas Gleixner --- kernel/irq/chip.c | 1 + kernel/irq/irqdomain.c | 4 ++++ 2 files changed, 5 insertions(+) (limited to 'kernel') diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c index 05e29de57933..5797909f4e5b 100644 --- a/kernel/irq/chip.c +++ b/kernel/irq/chip.c @@ -950,6 +950,7 @@ void irq_chip_ack_parent(struct irq_data *data) data = data->parent_data; data->chip->irq_ack(data); } +EXPORT_SYMBOL_GPL(irq_chip_ack_parent); /** * irq_chip_mask_parent - Mask the parent interrupt diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c index 1c9973e1b316..280a7fc43dd1 100644 --- a/kernel/irq/irqdomain.c +++ b/kernel/irq/irqdomain.c @@ -60,6 +60,7 @@ struct fwnode_handle *irq_domain_alloc_fwnode(void *data) fwid->fwnode.type = FWNODE_IRQCHIP; return &fwid->fwnode; } +EXPORT_SYMBOL_GPL(irq_domain_alloc_fwnode); /** * irq_domain_free_fwnode - Free a non-OF-backed fwnode_handle @@ -77,6 +78,7 @@ void irq_domain_free_fwnode(struct fwnode_handle *fwnode) kfree(fwid->name); kfree(fwid); } +EXPORT_SYMBOL_GPL(irq_domain_free_fwnode); /** * __irq_domain_add() - Allocate a new irq_domain data structure @@ -1013,6 +1015,7 @@ struct irq_data *irq_domain_get_irq_data(struct irq_domain *domain, return NULL; } +EXPORT_SYMBOL_GPL(irq_domain_get_irq_data); /** * irq_domain_set_hwirq_and_chip - Set hwirq and irqchip of @virq at @domain @@ -1343,6 +1346,7 @@ struct irq_data *irq_domain_get_irq_data(struct irq_domain *domain, return (irq_data && irq_data->domain == domain) ? irq_data : NULL; } +EXPORT_SYMBOL_GPL(irq_domain_get_irq_data); /** * irq_domain_set_info - Set the complete data for a @virq in @domain -- cgit From fb75a4282d0d9a3c7c44d940582c2d226cf3acfb Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 19 Dec 2015 20:07:38 +0000 Subject: futex: Drop refcount if requeue_pi() acquired the rtmutex If the proxy lock in the requeue loop acquires the rtmutex for a waiter then it acquired also refcount on the pi_state related to the futex, but the waiter side does not drop the reference count. Add the missing free_pi_state() call. Signed-off-by: Thomas Gleixner Cc: Peter Zijlstra Cc: Darren Hart Cc: Davidlohr Bueso Cc: Bhuvanesh_Surachari@mentor.com Cc: Andy Lowe Link: http://lkml.kernel.org/r/20151219200607.178132067@linutronix.de Signed-off-by: Thomas Gleixner Cc: stable@vger.kernel.org --- kernel/futex.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'kernel') diff --git a/kernel/futex.c b/kernel/futex.c index 684d7549825a..24fbc7765828 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -2755,6 +2755,11 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags, if (q.pi_state && (q.pi_state->owner != current)) { spin_lock(q.lock_ptr); ret = fixup_pi_state_owner(uaddr2, &q, current); + /* + * Drop the reference to the pi state which + * the requeue_pi() code acquired for us. + */ + free_pi_state(q.pi_state); spin_unlock(q.lock_ptr); } } else { -- cgit From 29e9ee5d48c35d6cf8afe09bdf03f77125c9ac11 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 19 Dec 2015 20:07:39 +0000 Subject: futex: Rename free_pi_state() to put_pi_state() free_pi_state() is confusing as it is in fact only freeing/caching the pi state when the last reference is gone. Rename it to put_pi_state() which reflects better what it is doing. Signed-off-by: Thomas Gleixner Cc: Peter Zijlstra Cc: Darren Hart Cc: Davidlohr Bueso Cc: Bhuvanesh_Surachari@mentor.com Cc: Andy Lowe Link: http://lkml.kernel.org/r/20151219200607.259636467@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/futex.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/futex.c b/kernel/futex.c index 24fbc7765828..f1581ff47122 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -725,9 +725,12 @@ static struct futex_pi_state * alloc_pi_state(void) } /* + * Drops a reference to the pi_state object and frees or caches it + * when the last reference is gone. + * * Must be called with the hb lock held. */ -static void free_pi_state(struct futex_pi_state *pi_state) +static void put_pi_state(struct futex_pi_state *pi_state) { if (!pi_state) return; @@ -1729,7 +1732,7 @@ retry_private: case 0: break; case -EFAULT: - free_pi_state(pi_state); + put_pi_state(pi_state); pi_state = NULL; double_unlock_hb(hb1, hb2); hb_waiters_dec(hb2); @@ -1746,7 +1749,7 @@ retry_private: * exit to complete. * - The user space value changed. */ - free_pi_state(pi_state); + put_pi_state(pi_state); pi_state = NULL; double_unlock_hb(hb1, hb2); hb_waiters_dec(hb2); @@ -1815,7 +1818,7 @@ retry_private: } else if (ret) { /* -EDEADLK */ this->pi_state = NULL; - free_pi_state(pi_state); + put_pi_state(pi_state); goto out_unlock; } } @@ -1824,7 +1827,7 @@ retry_private: } out_unlock: - free_pi_state(pi_state); + put_pi_state(pi_state); double_unlock_hb(hb1, hb2); wake_up_q(&wake_q); hb_waiters_dec(hb2); @@ -1973,7 +1976,7 @@ static void unqueue_me_pi(struct futex_q *q) __unqueue_futex(q); BUG_ON(!q->pi_state); - free_pi_state(q->pi_state); + put_pi_state(q->pi_state); q->pi_state = NULL; spin_unlock(q->lock_ptr); @@ -2759,7 +2762,7 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags, * Drop the reference to the pi state which * the requeue_pi() code acquired for us. */ - free_pi_state(q.pi_state); + put_pi_state(q.pi_state); spin_unlock(q.lock_ptr); } } else { -- cgit From ecb38b78f698a51988ec456751b20440e54702fb Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 19 Dec 2015 20:07:39 +0000 Subject: futex: Document pi_state refcounting in requeue code Documentation of the pi_state refcounting in the requeue code is non existent. Add it. Signed-off-by: Thomas Gleixner Cc: Peter Zijlstra Cc: Darren Hart Cc: Davidlohr Bueso Cc: Bhuvanesh_Surachari@mentor.com Cc: Andy Lowe Link: http://lkml.kernel.org/r/20151219200607.335938312@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/futex.c | 51 +++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 39 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/futex.c b/kernel/futex.c index f1581ff47122..20c468356b90 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -1709,27 +1709,31 @@ retry_private: * exist yet, look it up one more time to ensure we have a * reference to it. If the lock was taken, ret contains the * vpid of the top waiter task. + * If the lock was not taken, we have pi_state and an initial + * refcount on it. In case of an error we have nothing. */ if (ret > 0) { WARN_ON(pi_state); drop_count++; task_count++; /* - * If we acquired the lock, then the user - * space value of uaddr2 should be vpid. It - * cannot be changed by the top waiter as it - * is blocked on hb2 lock if it tries to do - * so. If something fiddled with it behind our - * back the pi state lookup might unearth - * it. So we rather use the known value than - * rereading and handing potential crap to - * lookup_pi_state. + * If we acquired the lock, then the user space value + * of uaddr2 should be vpid. It cannot be changed by + * the top waiter as it is blocked on hb2 lock if it + * tries to do so. If something fiddled with it behind + * our back the pi state lookup might unearth it. So + * we rather use the known value than rereading and + * handing potential crap to lookup_pi_state. + * + * If that call succeeds then we have pi_state and an + * initial refcount on it. */ ret = lookup_pi_state(ret, hb2, &key2, &pi_state); } switch (ret) { case 0: + /* We hold a reference on the pi state. */ break; case -EFAULT: put_pi_state(pi_state); @@ -1804,19 +1808,37 @@ retry_private: * of requeue_pi if we couldn't acquire the lock atomically. */ if (requeue_pi) { - /* Prepare the waiter to take the rt_mutex. */ + /* + * Prepare the waiter to take the rt_mutex. Take a + * refcount on the pi_state and store the pointer in + * the futex_q object of the waiter. + */ atomic_inc(&pi_state->refcount); this->pi_state = pi_state; ret = rt_mutex_start_proxy_lock(&pi_state->pi_mutex, this->rt_waiter, this->task); if (ret == 1) { - /* We got the lock. */ + /* + * We got the lock. We do neither drop the + * refcount on pi_state nor clear + * this->pi_state because the waiter needs the + * pi_state for cleaning up the user space + * value. It will drop the refcount after + * doing so. + */ requeue_pi_wake_futex(this, &key2, hb2); drop_count++; continue; } else if (ret) { - /* -EDEADLK */ + /* + * rt_mutex_start_proxy_lock() detected a + * potential deadlock when we tried to queue + * that waiter. Drop the pi_state reference + * which we took above and remove the pointer + * to the state from the waiters futex_q + * object. + */ this->pi_state = NULL; put_pi_state(pi_state); goto out_unlock; @@ -1827,6 +1849,11 @@ retry_private: } out_unlock: + /* + * We took an extra initial reference to the pi_state either + * in futex_proxy_trylock_atomic() or in lookup_pi_state(). We + * need to drop it here again. + */ put_pi_state(pi_state); double_unlock_hb(hb1, hb2); wake_up_q(&wake_q); -- cgit From 4959f2de11ca532a120a337429e5576fd283700f Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 19 Dec 2015 20:07:40 +0000 Subject: futex: Remove pointless put_pi_state calls in requeue() In the error handling cases we neither have pi_state nor a reference to it. Remove the pointless code. Signed-off-by: Thomas Gleixner Cc: Peter Zijlstra Cc: Darren Hart Cc: Davidlohr Bueso Cc: Bhuvanesh_Surachari@mentor.com Cc: Andy Lowe Link: http://lkml.kernel.org/r/20151219200607.432780944@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/futex.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/futex.c b/kernel/futex.c index 20c468356b90..dcec01856cf3 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -1735,9 +1735,9 @@ retry_private: case 0: /* We hold a reference on the pi state. */ break; + + /* If the above failed, then pi_state is NULL */ case -EFAULT: - put_pi_state(pi_state); - pi_state = NULL; double_unlock_hb(hb1, hb2); hb_waiters_dec(hb2); put_futex_key(&key2); @@ -1753,8 +1753,6 @@ retry_private: * exit to complete. * - The user space value changed. */ - put_pi_state(pi_state); - pi_state = NULL; double_unlock_hb(hb1, hb2); hb_waiters_dec(hb2); put_futex_key(&key2); -- cgit From 885c2cb770b5ac2507c41bc9f91a5d1c98337bee Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 19 Dec 2015 20:07:41 +0000 Subject: futex: Cleanup the goto confusion in requeue_pi() out_unlock: does not only drop the locks, it also drops the refcount on the pi_state. Really intuitive. Move the label after the put_pi_state() call and use 'break' in the error handling path of the requeue loop. Signed-off-by: Thomas Gleixner Cc: Peter Zijlstra Cc: Darren Hart Cc: Davidlohr Bueso Cc: Bhuvanesh_Surachari@mentor.com Cc: Andy Lowe Link: http://lkml.kernel.org/r/20151219200607.526665141@linutronix.de Signed-off-by: Thomas Gleixner --- kernel/futex.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/futex.c b/kernel/futex.c index dcec01856cf3..461d438f4816 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -1839,20 +1839,25 @@ retry_private: */ this->pi_state = NULL; put_pi_state(pi_state); - goto out_unlock; + /* + * We stop queueing more waiters and let user + * space deal with the mess. + */ + break; } } requeue_futex(this, hb1, hb2, &key2); drop_count++; } -out_unlock: /* * We took an extra initial reference to the pi_state either * in futex_proxy_trylock_atomic() or in lookup_pi_state(). We * need to drop it here again. */ put_pi_state(pi_state); + +out_unlock: double_unlock_hb(hb1, hb2); wake_up_q(&wake_q); hb_waiters_dec(hb2); -- cgit From 337f13046ff03717a9e99675284a817527440a49 Mon Sep 17 00:00:00 2001 From: Darren Hart Date: Fri, 18 Dec 2015 13:36:37 -0800 Subject: futex: Allow FUTEX_CLOCK_REALTIME with FUTEX_WAIT op While reviewing Michael Kerrisk's recent futex manpage update, I noticed that we allow the FUTEX_CLOCK_REALTIME flag for FUTEX_WAIT_BITSET but not for FUTEX_WAIT. FUTEX_WAIT is treated as a simple version for FUTEX_WAIT_BITSET internally (with a bitmask of FUTEX_BITSET_MATCH_ANY). As such, I cannot come up with a reason for this exclusion for FUTEX_WAIT. This change does modify the behavior of the futex syscall, changing a call with FUTEX_WAIT | FUTEX_CLOCK_REALTIME from returning -ENOSYS, to be equivalent to FUTEX_WAIT_BITSET | FUTEX_CLOCK_REALTIME with a bitset of FUTEX_BITSET_MATCH_ANY. Reported-by: Michael Kerrisk Signed-off-by: Darren Hart Cc: Peter Zijlstra Cc: Davidlohr Bueso Link: http://lkml.kernel.org/r/9f3bdc116d79d23f5ee72ceb9a2a857f5ff8fa29.1450474525.git.dvhart@linux.intel.com Signed-off-by: Thomas Gleixner --- kernel/futex.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/futex.c b/kernel/futex.c index 461d438f4816..8a310e240cda 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -3084,7 +3084,8 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout, if (op & FUTEX_CLOCK_REALTIME) { flags |= FLAGS_CLOCKRT; - if (cmd != FUTEX_WAIT_BITSET && cmd != FUTEX_WAIT_REQUEUE_PI) + if (cmd != FUTEX_WAIT && cmd != FUTEX_WAIT_BITSET && \ + cmd != FUTEX_WAIT_REQUEUE_PI) return -ENOSYS; } -- cgit From 75aba7b0e9ac416ca53c0c97680b8e9aedf09284 Mon Sep 17 00:00:00 2001 From: Suravee Suthikulpanit Date: Thu, 10 Dec 2015 08:55:28 -0800 Subject: irqdomain: Introduce is_fwnode_irqchip helper Since there will be several places checking if fwnode.type is equal FWNODE_IRQCHIP, this patch adds a convenient function for this purpose. Acked-by: Marc Zyngier Signed-off-by: Suravee Suthikulpanit Signed-off-by: Marc Zyngier --- kernel/irq/irqdomain.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c index 22aa9612ef7c..7f34d98ebfc4 100644 --- a/kernel/irq/irqdomain.c +++ b/kernel/irq/irqdomain.c @@ -70,7 +70,7 @@ void irq_domain_free_fwnode(struct fwnode_handle *fwnode) { struct irqchip_fwid *fwid; - if (WARN_ON(fwnode->type != FWNODE_IRQCHIP)) + if (WARN_ON(!is_fwnode_irqchip(fwnode))) return; fwid = container_of(fwnode, struct irqchip_fwid, fwnode); -- cgit From 1fe7c4ef88bd32e039f5f4126537c3f20c340414 Mon Sep 17 00:00:00 2001 From: Stefano Stabellini Date: Tue, 10 Nov 2015 12:36:46 +0000 Subject: missing include asm/paravirt.h in cputime.c Add include asm/paravirt.h to cputime.c, as steal_account_process_tick calls paravirt_steal_clock, which is defined in asm/paravirt.h. The ifdef CONFIG_PARAVIRT is necessary because not all archs have an asm/paravirt.h to include. The reason why currently cputime.c compiles, even though include is missing, is that on x86 asm/paravirt.h is included by one of the other headers included in kernel/sched/cputime.c: On arm and arm64, where I am about to introduce asm/paravirt.h and stolen time support, without #include in cputime.c, I would get an error. Signed-off-by: Stefano Stabellini Acked-by: Peter Zijlstra (Intel) --- kernel/sched/cputime.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'kernel') diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index 05de80b48586..851b00f344ae 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -5,6 +5,9 @@ #include #include #include "sched.h" +#ifdef CONFIG_PARAVIRT +#include +#endif #ifdef CONFIG_IRQ_TIME_ACCOUNTING -- cgit From b8ec330a63eb39127f5cfcae5f8524e969ef9f94 Mon Sep 17 00:00:00 2001 From: Li Bin Date: Mon, 30 Nov 2015 18:23:36 +0800 Subject: ftrace: Fix a typo in comment s/ARCH_SUPPORT_FTARCE_OPS/ARCH_SUPPORTS_FTRACE_OPS/ Link: http://lkml.kernel.org/r/1448879016-8659-1-git-send-email-huawei.libin@huawei.com Signed-off-by: Li Bin Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index e371aed51fcf..b8dfe8138aa2 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -5263,7 +5263,7 @@ out: * being NULL, or CONFIG_DYNAMIC_FTRACE_WITH_REGS. * Note, CONFIG_DYNAMIC_FTRACE_WITH_REGS expects a full regs to be saved. * An architecture can pass partial regs with ftrace_ops and still - * set the ARCH_SUPPORT_FTARCE_OPS. + * set the ARCH_SUPPORTS_FTRACE_OPS. */ #if ARCH_SUPPORTS_FTRACE_OPS static void ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip, -- cgit From 030f4e1cb86f059185572fd1678a55b5e8ff0d08 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Tue, 1 Dec 2015 12:24:45 -0500 Subject: ftrace: Fix output of enabled_functions for showing tramp When showing all tramps registered to a ftrace record in the file enabled_functions, it exits the loop with ops == NULL. But then it is suppose to show the function on the ops->trampoline and add_trampoline_func() is called with the given ops. But because ops is now NULL (to exit the loop), it always shows the static trampoline instead of the one that is really registered to the record. The call to add_trampoline_func() that shows the trampoline for the given ops needs to be called at every iteration. Fixes: 39daa7b9e895 "ftrace: Show all tramps registered to a record on ftrace_bug()" Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index b8dfe8138aa2..bf7bebcdad82 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -3322,7 +3322,7 @@ static int t_show(struct seq_file *m, void *v) seq_printf(m, "%ps", (void *)rec->ip); if (iter->flags & FTRACE_ITER_ENABLED) { - struct ftrace_ops *ops = NULL; + struct ftrace_ops *ops; seq_printf(m, " (%ld)%s%s", ftrace_rec_count(rec), @@ -3335,13 +3335,14 @@ static int t_show(struct seq_file *m, void *v) seq_printf(m, "\ttramp: %pS (%pS)", (void *)ops->trampoline, (void *)ops->func); + add_trampoline_func(m, ops, rec); ops = ftrace_find_tramp_ops_next(rec, ops); } while (ops); } else seq_puts(m, "\ttramp: ERROR!"); - + } else { + add_trampoline_func(m, NULL, rec); } - add_trampoline_func(m, ops, rec); } seq_putc(m, '\n'); -- cgit From ba27f2bc731135a0396f3968bdddb54f3bc72e64 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Mon, 30 Nov 2015 17:23:39 -0500 Subject: ftrace: Remove use of control list and ops Currently perf has its own list function within the ftrace infrastructure that seems to be used only to allow for it to have per-cpu disabling as well as a check to make sure that it's not called while RCU is not watching. It uses something called the "control_ops" which is used to iterate over ops under it with the control_list_func(). The problem is that this control_ops and control_list_func unnecessarily complicates the code. By replacing FTRACE_OPS_FL_CONTROL with two new flags (FTRACE_OPS_FL_RCU and FTRACE_OPS_FL_PER_CPU) we can remove all the code that is special with the control ops and add the needed checks within the generic ftrace_list_func(). Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 126 ++++++++++++---------------------------- kernel/trace/trace.h | 2 - kernel/trace/trace_event_perf.c | 2 +- 3 files changed, 39 insertions(+), 91 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index bf7bebcdad82..bc7f4eb6b4b0 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -62,8 +62,6 @@ #define FTRACE_HASH_DEFAULT_BITS 10 #define FTRACE_HASH_MAX_BITS 12 -#define FL_GLOBAL_CONTROL_MASK (FTRACE_OPS_FL_CONTROL) - #ifdef CONFIG_DYNAMIC_FTRACE #define INIT_OPS_HASH(opsname) \ .func_hash = &opsname.local_hash, \ @@ -113,11 +111,9 @@ static int ftrace_disabled __read_mostly; static DEFINE_MUTEX(ftrace_lock); -static struct ftrace_ops *ftrace_control_list __read_mostly = &ftrace_list_end; static struct ftrace_ops *ftrace_ops_list __read_mostly = &ftrace_list_end; ftrace_func_t ftrace_trace_function __read_mostly = ftrace_stub; static struct ftrace_ops global_ops; -static struct ftrace_ops control_ops; static void ftrace_ops_recurs_func(unsigned long ip, unsigned long parent_ip, struct ftrace_ops *op, struct pt_regs *regs); @@ -203,7 +199,7 @@ void clear_ftrace_function(void) ftrace_trace_function = ftrace_stub; } -static void control_ops_disable_all(struct ftrace_ops *ops) +static void per_cpu_ops_disable_all(struct ftrace_ops *ops) { int cpu; @@ -211,16 +207,19 @@ static void control_ops_disable_all(struct ftrace_ops *ops) *per_cpu_ptr(ops->disabled, cpu) = 1; } -static int control_ops_alloc(struct ftrace_ops *ops) +static int per_cpu_ops_alloc(struct ftrace_ops *ops) { int __percpu *disabled; + if (WARN_ON_ONCE(!(ops->flags & FTRACE_OPS_FL_PER_CPU))) + return -EINVAL; + disabled = alloc_percpu(int); if (!disabled) return -ENOMEM; ops->disabled = disabled; - control_ops_disable_all(ops); + per_cpu_ops_disable_all(ops); return 0; } @@ -256,10 +255,11 @@ static inline void update_function_graph_func(void) { } static ftrace_func_t ftrace_ops_get_list_func(struct ftrace_ops *ops) { /* - * If this is a dynamic ops or we force list func, + * If this is a dynamic, RCU, or per CPU ops, or we force list func, * then it needs to call the list anyway. */ - if (ops->flags & FTRACE_OPS_FL_DYNAMIC || FTRACE_FORCE_LIST_FUNC) + if (ops->flags & (FTRACE_OPS_FL_DYNAMIC | FTRACE_OPS_FL_PER_CPU | + FTRACE_OPS_FL_RCU) || FTRACE_FORCE_LIST_FUNC) return ftrace_ops_list_func; return ftrace_ops_get_func(ops); @@ -383,26 +383,6 @@ static int remove_ftrace_ops(struct ftrace_ops **list, struct ftrace_ops *ops) return 0; } -static void add_ftrace_list_ops(struct ftrace_ops **list, - struct ftrace_ops *main_ops, - struct ftrace_ops *ops) -{ - int first = *list == &ftrace_list_end; - add_ftrace_ops(list, ops); - if (first) - add_ftrace_ops(&ftrace_ops_list, main_ops); -} - -static int remove_ftrace_list_ops(struct ftrace_ops **list, - struct ftrace_ops *main_ops, - struct ftrace_ops *ops) -{ - int ret = remove_ftrace_ops(list, ops); - if (!ret && *list == &ftrace_list_end) - ret = remove_ftrace_ops(&ftrace_ops_list, main_ops); - return ret; -} - static void ftrace_update_trampoline(struct ftrace_ops *ops); static int __register_ftrace_function(struct ftrace_ops *ops) @@ -430,14 +410,12 @@ static int __register_ftrace_function(struct ftrace_ops *ops) if (!core_kernel_data((unsigned long)ops)) ops->flags |= FTRACE_OPS_FL_DYNAMIC; - if (ops->flags & FTRACE_OPS_FL_CONTROL) { - if (control_ops_alloc(ops)) + if (ops->flags & FTRACE_OPS_FL_PER_CPU) { + if (per_cpu_ops_alloc(ops)) return -ENOMEM; - add_ftrace_list_ops(&ftrace_control_list, &control_ops, ops); - /* The control_ops needs the trampoline update */ - ops = &control_ops; - } else - add_ftrace_ops(&ftrace_ops_list, ops); + } + + add_ftrace_ops(&ftrace_ops_list, ops); /* Always save the function, and reset at unregistering */ ops->saved_func = ops->func; @@ -460,11 +438,7 @@ static int __unregister_ftrace_function(struct ftrace_ops *ops) if (WARN_ON(!(ops->flags & FTRACE_OPS_FL_ENABLED))) return -EBUSY; - if (ops->flags & FTRACE_OPS_FL_CONTROL) { - ret = remove_ftrace_list_ops(&ftrace_control_list, - &control_ops, ops); - } else - ret = remove_ftrace_ops(&ftrace_ops_list, ops); + ret = remove_ftrace_ops(&ftrace_ops_list, ops); if (ret < 0) return ret; @@ -2630,7 +2604,7 @@ void __weak arch_ftrace_trampoline_free(struct ftrace_ops *ops) { } -static void control_ops_free(struct ftrace_ops *ops) +static void per_cpu_ops_free(struct ftrace_ops *ops) { free_percpu(ops->disabled); } @@ -2731,13 +2705,13 @@ static int ftrace_shutdown(struct ftrace_ops *ops, int command) if (!command || !ftrace_enabled) { /* - * If these are control ops, they still need their + * If these are per_cpu ops, they still need their * per_cpu field freed. Since, function tracing is * not currently active, we can just free them * without synchronizing all CPUs. */ - if (ops->flags & FTRACE_OPS_FL_CONTROL) - control_ops_free(ops); + if (ops->flags & FTRACE_OPS_FL_PER_CPU) + per_cpu_ops_free(ops); return 0; } @@ -2778,7 +2752,7 @@ static int ftrace_shutdown(struct ftrace_ops *ops, int command) /* * Dynamic ops may be freed, we must make sure that all * callers are done before leaving this function. - * The same goes for freeing the per_cpu data of the control + * The same goes for freeing the per_cpu data of the per_cpu * ops. * * Again, normal synchronize_sched() is not good enough. @@ -2789,13 +2763,13 @@ static int ftrace_shutdown(struct ftrace_ops *ops, int command) * infrastructure to do the synchronization, thus we must do it * ourselves. */ - if (ops->flags & (FTRACE_OPS_FL_DYNAMIC | FTRACE_OPS_FL_CONTROL)) { + if (ops->flags & (FTRACE_OPS_FL_DYNAMIC | FTRACE_OPS_FL_PER_CPU)) { schedule_on_each_cpu(ftrace_sync); arch_ftrace_trampoline_free(ops); - if (ops->flags & FTRACE_OPS_FL_CONTROL) - control_ops_free(ops); + if (ops->flags & FTRACE_OPS_FL_PER_CPU) + per_cpu_ops_free(ops); } return 0; @@ -5185,44 +5159,6 @@ void ftrace_reset_array_ops(struct trace_array *tr) tr->ops->func = ftrace_stub; } -static void -ftrace_ops_control_func(unsigned long ip, unsigned long parent_ip, - struct ftrace_ops *op, struct pt_regs *regs) -{ - if (unlikely(trace_recursion_test(TRACE_CONTROL_BIT))) - return; - - /* - * Some of the ops may be dynamically allocated, - * they must be freed after a synchronize_sched(). - */ - preempt_disable_notrace(); - trace_recursion_set(TRACE_CONTROL_BIT); - - /* - * Control funcs (perf) uses RCU. Only trace if - * RCU is currently active. - */ - if (!rcu_is_watching()) - goto out; - - do_for_each_ftrace_op(op, ftrace_control_list) { - if (!(op->flags & FTRACE_OPS_FL_STUB) && - !ftrace_function_local_disabled(op) && - ftrace_ops_test(op, ip, regs)) - op->func(ip, parent_ip, op, regs); - } while_for_each_ftrace_op(op); - out: - trace_recursion_clear(TRACE_CONTROL_BIT); - preempt_enable_notrace(); -} - -static struct ftrace_ops control_ops = { - .func = ftrace_ops_control_func, - .flags = FTRACE_OPS_FL_RECURSION_SAFE | FTRACE_OPS_FL_INITIALIZED, - INIT_OPS_HASH(control_ops) -}; - static inline void __ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip, struct ftrace_ops *ignored, struct pt_regs *regs) @@ -5239,8 +5175,22 @@ __ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip, * they must be freed after a synchronize_sched(). */ preempt_disable_notrace(); + do_for_each_ftrace_op(op, ftrace_ops_list) { - if (ftrace_ops_test(op, ip, regs)) { + /* + * Check the following for each ops before calling their func: + * if RCU flag is set, then rcu_is_watching() must be true + * if PER_CPU is set, then ftrace_function_local_disable() + * must be false + * Otherwise test if the ip matches the ops filter + * + * If any of the above fails then the op->func() is not executed. + */ + if ((!(op->flags & FTRACE_OPS_FL_RCU) || rcu_is_watching()) && + (!(op->flags & FTRACE_OPS_FL_PER_CPU) || + !ftrace_function_local_disabled(op)) && + ftrace_ops_test(op, ip, regs)) { + if (FTRACE_WARN_ON(!op->func)) { pr_warn("op=%p %pS\n", op, op); goto out; diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 919d9d07686f..d3980b87bf04 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -467,8 +467,6 @@ enum { TRACE_INTERNAL_IRQ_BIT, TRACE_INTERNAL_SIRQ_BIT, - TRACE_CONTROL_BIT, - TRACE_BRANCH_BIT, /* * Abuse of the trace_recursion. diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c index abfc903e741e..2649c85cd162 100644 --- a/kernel/trace/trace_event_perf.c +++ b/kernel/trace/trace_event_perf.c @@ -334,7 +334,7 @@ static int perf_ftrace_function_register(struct perf_event *event) { struct ftrace_ops *ops = &event->ftrace_ops; - ops->flags |= FTRACE_OPS_FL_CONTROL; + ops->flags |= FTRACE_OPS_FL_PER_CPU | FTRACE_OPS_FL_RCU; ops->func = perf_ftrace_function_call; return register_ftrace_function(ops); } -- cgit From c68c0fa29341754de86b6e5317b6074f1e334581 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Tue, 1 Dec 2015 13:28:16 -0500 Subject: ftrace: Have ftrace_ops_get_func() handle RCU and PER_CPU flags too Jiri Olsa noted that the change to replace the control_ops did not update the trampoline for when running perf on a single CPU and with CONFIG_PREEMPT disabled (where dynamic ops, like perf, can use trampolines directly). The result was that perf function could be called when RCU is not watching as well as not handle the ftrace_local_disable(). Modify the ftrace_ops_get_func() to also check the RCU and PER_CPU ops flags and use the recursive function if they are set. The recursive function is modified to check those flags and execute the appropriate checks if they are set. Link: http://lkml.kernel.org/r/20151201134213.GA14155@krava.brq.redhat.com Reported-by: Jiri Olsa Patch-fixed-up-by: Jiri Olsa Tested-by: Jiri Olsa Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 30 ++++++++++++++++++------------ 1 file changed, 18 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index bc7f4eb6b4b0..e290a30f2d0b 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -115,9 +115,6 @@ static struct ftrace_ops *ftrace_ops_list __read_mostly = &ftrace_list_end; ftrace_func_t ftrace_trace_function __read_mostly = ftrace_stub; static struct ftrace_ops global_ops; -static void ftrace_ops_recurs_func(unsigned long ip, unsigned long parent_ip, - struct ftrace_ops *op, struct pt_regs *regs); - #if ARCH_SUPPORTS_FTRACE_OPS static void ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip, struct ftrace_ops *op, struct pt_regs *regs); @@ -5231,20 +5228,29 @@ static void ftrace_ops_no_ops(unsigned long ip, unsigned long parent_ip) /* * If there's only one function registered but it does not support - * recursion, this function will be called by the mcount trampoline. - * This function will handle recursion protection. + * recursion, needs RCU protection and/or requires per cpu handling, then + * this function will be called by the mcount trampoline. */ -static void ftrace_ops_recurs_func(unsigned long ip, unsigned long parent_ip, +static void ftrace_ops_assist_func(unsigned long ip, unsigned long parent_ip, struct ftrace_ops *op, struct pt_regs *regs) { int bit; + if ((op->flags & FTRACE_OPS_FL_RCU) && !rcu_is_watching()) + return; + bit = trace_test_and_set_recursion(TRACE_LIST_START, TRACE_LIST_MAX); if (bit < 0) return; - op->func(ip, parent_ip, op, regs); + preempt_disable_notrace(); + if (!(op->flags & FTRACE_OPS_FL_PER_CPU) || + !ftrace_function_local_disabled(op)) { + op->func(ip, parent_ip, op, regs); + } + + preempt_enable_notrace(); trace_clear_recursion(bit); } @@ -5262,12 +5268,12 @@ static void ftrace_ops_recurs_func(unsigned long ip, unsigned long parent_ip, ftrace_func_t ftrace_ops_get_func(struct ftrace_ops *ops) { /* - * If the func handles its own recursion, call it directly. - * Otherwise call the recursion protected function that - * will call the ftrace ops function. + * If the function does not handle recursion, needs to be RCU safe, + * or does per cpu logic, then we need to call the assist handler. */ - if (!(ops->flags & FTRACE_OPS_FL_RECURSION_SAFE)) - return ftrace_ops_recurs_func; + if (!(ops->flags & FTRACE_OPS_FL_RECURSION_SAFE) || + ops->flags & (FTRACE_OPS_FL_RCU | FTRACE_OPS_FL_PER_CPU)) + return ftrace_ops_assist_func; return ops->func; } -- cgit From 27dff4e04199cf0ecf06239a26d0d225d3c046e9 Mon Sep 17 00:00:00 2001 From: Julia Lawall Date: Fri, 11 Dec 2015 18:35:59 +0100 Subject: bpf: Constify bpf_verifier_ops structure This bpf_verifier_ops structure is never modified, like the other bpf_verifier_ops structures, so declare it as const. Done with the help of Coccinelle. Link: http://lkml.kernel.org/r/1449855359-13724-1-git-send-email-Julia.Lawall@lip6.fr Signed-off-by: Julia Lawall Signed-off-by: Steven Rostedt --- kernel/trace/bpf_trace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c index 4228fd3682c3..45dd798bcd37 100644 --- a/kernel/trace/bpf_trace.c +++ b/kernel/trace/bpf_trace.c @@ -316,7 +316,7 @@ static bool kprobe_prog_is_valid_access(int off, int size, enum bpf_access_type return true; } -static struct bpf_verifier_ops kprobe_prog_ops = { +static const struct bpf_verifier_ops kprobe_prog_ops = { .get_func_proto = kprobe_prog_func_proto, .is_valid_access = kprobe_prog_is_valid_access, }; -- cgit From b6b71f66a16a9212b853c168f6ec1f303c5c7a7d Mon Sep 17 00:00:00 2001 From: Abel Vesa Date: Wed, 2 Dec 2015 15:39:57 +0100 Subject: ftrace: Join functions ftrace_module_init() and ftrace_init_module() Simple cleanup. No need for two functions here. The whole work can simply be done inside 'ftrace_module_init'. Link: http://lkml.kernel.org/r/1449067197-5718-1-git-send-email-abelvesa@linux.com Signed-off-by: Abel Vesa Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index e290a30f2d0b..29fcc2c76583 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -4980,19 +4980,16 @@ void ftrace_release_mod(struct module *mod) mutex_unlock(&ftrace_lock); } -static void ftrace_init_module(struct module *mod, - unsigned long *start, unsigned long *end) +void ftrace_module_init(struct module *mod) { + unsigned long *start = mod->ftrace_callsites; + unsigned long *end = mod->ftrace_callsites + + mod->num_ftrace_callsites; + if (ftrace_disabled || start == end) return; - ftrace_process_locs(mod, start, end); -} -void ftrace_module_init(struct module *mod) -{ - ftrace_init_module(mod, mod->ftrace_callsites, - mod->ftrace_callsites + - mod->num_ftrace_callsites); + ftrace_process_locs(mod, start, end); } static int ftrace_module_notify_exit(struct notifier_block *self, -- cgit From 97e9b4fca52bf4e2f7eed9463a0722f8e7afbe90 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Wed, 23 Dec 2015 12:12:22 -0500 Subject: ftrace: Clean up ftrace_module_init() code The start and end variables were only used when ftrace_module_init() was split up into multiple functions. No need to keep them around after the merger. Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 29fcc2c76583..0f7ee341f89f 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -4982,14 +4982,11 @@ void ftrace_release_mod(struct module *mod) void ftrace_module_init(struct module *mod) { - unsigned long *start = mod->ftrace_callsites; - unsigned long *end = mod->ftrace_callsites + - mod->num_ftrace_callsites; - - if (ftrace_disabled || start == end) + if (ftrace_disabled || !mod->num_ftrace_callsites) return; - ftrace_process_locs(mod, start, end); + ftrace_process_locs(mod, mod->ftrace_callsites, + mod->ftrace_callsites + mod->num_ftrace_callsites); } static int ftrace_module_notify_exit(struct notifier_block *self, -- cgit From 05a724bd44a68cd5c60bc4b949a946b57d4c9e55 Mon Sep 17 00:00:00 2001 From: Chuyu Hu Date: Tue, 22 Dec 2015 09:44:33 -0500 Subject: tracing: Fix comment to use tracing_on over tracing_enable The file tracing_enable is obsolete and does not exist anymore. Replace the comment that references it with the proper tracing_on file. Link: http://lkml.kernel.org/r/1450787141-45544-1-git-send-email-chuhu@redhat.com Signed-off-by: Chuyu Hu Signed-off-by: Steven Rostedt --- kernel/trace/trace.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index d3980b87bf04..8414fa40bf27 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -363,8 +363,8 @@ struct trace_option_dentry { * @name: the name chosen to select it on the available_tracers file * @init: called when one switches to this tracer (echo name > current_tracer) * @reset: called when one switches to another tracer - * @start: called when tracing is unpaused (echo 1 > tracing_enabled) - * @stop: called when tracing is paused (echo 0 > tracing_enabled) + * @start: called when tracing is unpaused (echo 1 > tracing_on) + * @stop: called when tracing is paused (echo 0 > tracing_on) * @update_thresh: called when tracing_thresh is updated * @open: called when the trace file is opened * @pipe_open: called when the trace_pipe file is opened -- cgit From 1b9f23727abb92c5e58f139e7d180befcaa06fe0 Mon Sep 17 00:00:00 2001 From: Richard Cochran Date: Tue, 22 Dec 2015 22:19:58 +0100 Subject: posix-clock: Fix return code on the poll method's error path The posix_clock_poll function is supposed to return a bit mask of POLLxxx values. However, in case the hardware has disappeared (due to hot plugging for example) this code returns -ENODEV in a futile attempt to throw an error at the file descriptor level. The kernel's file_operations interface does not accept such error codes from the poll method. Instead, this function aught to return POLLERR. The value -ENODEV does, in fact, contain the POLLERR bit (and almost all the other POLLxxx bits as well), but only by chance. This patch fixes code to return a proper bit mask. Credit goes to Markus Elfring for pointing out the suspicious signed/unsigned mismatch. Reported-by: Markus Elfring igned-off-by: Richard Cochran Cc: John Stultz Cc: Julia Lawall Link: http://lkml.kernel.org/r/1450819198-17420-1-git-send-email-richardcochran@gmail.com Cc: stable@vger.kernel.org Signed-off-by: Thomas Gleixner --- kernel/time/posix-clock.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/time/posix-clock.c b/kernel/time/posix-clock.c index ce033c7aa2e8..9cff0ab82b63 100644 --- a/kernel/time/posix-clock.c +++ b/kernel/time/posix-clock.c @@ -69,10 +69,10 @@ static ssize_t posix_clock_read(struct file *fp, char __user *buf, static unsigned int posix_clock_poll(struct file *fp, poll_table *wait) { struct posix_clock *clk = get_posix_clock(fp); - int result = 0; + unsigned int result = 0; if (!clk) - return -ENODEV; + return POLLERR; if (clk->ops.poll) result = clk->ops.poll(clk, fp, wait); -- cgit From a5ae989957cbc3f3b7bc40677bd9e459c0917528 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 29 Dec 2015 14:53:56 -0500 Subject: cgroup: demote subsystem init messages to KERN_DEBUG These are noisy during boot and not all that interesting. Signed-off-by: Tejun Heo --- kernel/cgroup.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/cgroup.c b/kernel/cgroup.c index 6b33631251b9..122ec5543335 100644 --- a/kernel/cgroup.c +++ b/kernel/cgroup.c @@ -5169,7 +5169,7 @@ static void __init cgroup_init_subsys(struct cgroup_subsys *ss, bool early) { struct cgroup_subsys_state *css; - printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name); + pr_debug("Initializing cgroup subsys %s\n", ss->name); mutex_lock(&cgroup_mutex); -- cgit From 16e5c1fc36040e592128a164499bc25eb138a80f Mon Sep 17 00:00:00 2001 From: Al Viro Date: Thu, 24 Dec 2015 00:06:05 -0500 Subject: convert a bunch of open-coded instances of memdup_user_nul() A _lot_ of ->write() instances were open-coding it; some are converted to memdup_user_nul(), a lot more remain... Signed-off-by: Al Viro --- kernel/trace/blktrace.c | 12 +++--------- 1 file changed, 3 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c index a990824c8604..2aeb6ffc0a1e 100644 --- a/kernel/trace/blktrace.c +++ b/kernel/trace/blktrace.c @@ -349,16 +349,10 @@ static ssize_t blk_msg_write(struct file *filp, const char __user *buffer, if (count >= BLK_TN_MAX_MSG) return -EINVAL; - msg = kmalloc(count + 1, GFP_KERNEL); - if (msg == NULL) - return -ENOMEM; - - if (copy_from_user(msg, buffer, count)) { - kfree(msg); - return -EFAULT; - } + msg = memdup_user_nul(buffer, count); + if (IS_ERR(msg)) + return PTR_ERR(msg); - msg[count] = '\0'; bt = filp->private_data; __trace_note_message(bt, "%s", msg); kfree(msg); -- cgit From 70f6cbb6f9c95535acd327d1ac1ce5fd078cff1e Mon Sep 17 00:00:00 2001 From: Al Viro Date: Thu, 24 Dec 2015 00:13:10 -0500 Subject: kernel/*: switch to memdup_user_nul() Signed-off-by: Al Viro --- kernel/sysctl.c | 79 +++++++++++++------------------------ kernel/trace/trace_events.c | 28 ++++--------- kernel/trace/trace_events_trigger.c | 15 +++---- kernel/user_namespace.c | 21 ++++------ 4 files changed, 48 insertions(+), 95 deletions(-) (limited to 'kernel') diff --git a/kernel/sysctl.c b/kernel/sysctl.c index dc6858d6639e..5faf89ac9ec0 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -2047,9 +2047,8 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table, void *data) { int *i, vleft, first = 1, err = 0; - unsigned long page = 0; size_t left; - char *kbuf; + char *kbuf = NULL, *p; if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) { *lenp = 0; @@ -2078,15 +2077,9 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table, if (left > PAGE_SIZE - 1) left = PAGE_SIZE - 1; - page = __get_free_page(GFP_TEMPORARY); - kbuf = (char *) page; - if (!kbuf) - return -ENOMEM; - if (copy_from_user(kbuf, buffer, left)) { - err = -EFAULT; - goto free; - } - kbuf[left] = 0; + p = kbuf = memdup_user_nul(buffer, left); + if (IS_ERR(kbuf)) + return PTR_ERR(kbuf); } for (; left && vleft--; i++, first=0) { @@ -2094,11 +2087,11 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table, bool neg; if (write) { - left -= proc_skip_spaces(&kbuf); + left -= proc_skip_spaces(&p); if (!left) break; - err = proc_get_long(&kbuf, &left, &lval, &neg, + err = proc_get_long(&p, &left, &lval, &neg, proc_wspace_sep, sizeof(proc_wspace_sep), NULL); if (err) @@ -2125,10 +2118,9 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table, if (!write && !first && left && !err) err = proc_put_char(&buffer, &left, '\n'); if (write && !err && left) - left -= proc_skip_spaces(&kbuf); -free: + left -= proc_skip_spaces(&p); if (write) { - free_page(page); + kfree(kbuf); if (first) return err ? : -EINVAL; } @@ -2310,9 +2302,8 @@ static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int { unsigned long *i, *min, *max; int vleft, first = 1, err = 0; - unsigned long page = 0; size_t left; - char *kbuf; + char *kbuf = NULL, *p; if (!data || !table->maxlen || !*lenp || (*ppos && !write)) { *lenp = 0; @@ -2340,15 +2331,9 @@ static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int if (left > PAGE_SIZE - 1) left = PAGE_SIZE - 1; - page = __get_free_page(GFP_TEMPORARY); - kbuf = (char *) page; - if (!kbuf) - return -ENOMEM; - if (copy_from_user(kbuf, buffer, left)) { - err = -EFAULT; - goto free; - } - kbuf[left] = 0; + p = kbuf = memdup_user_nul(buffer, left); + if (IS_ERR(kbuf)) + return PTR_ERR(kbuf); } for (; left && vleft--; i++, first = 0) { @@ -2357,9 +2342,9 @@ static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int if (write) { bool neg; - left -= proc_skip_spaces(&kbuf); + left -= proc_skip_spaces(&p); - err = proc_get_long(&kbuf, &left, &val, &neg, + err = proc_get_long(&p, &left, &val, &neg, proc_wspace_sep, sizeof(proc_wspace_sep), NULL); if (err) @@ -2385,10 +2370,9 @@ static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int if (!write && !first && left && !err) err = proc_put_char(&buffer, &left, '\n'); if (write && !err) - left -= proc_skip_spaces(&kbuf); -free: + left -= proc_skip_spaces(&p); if (write) { - free_page(page); + kfree(kbuf); if (first) return err ? : -EINVAL; } @@ -2650,34 +2634,27 @@ int proc_do_large_bitmap(struct ctl_table *table, int write, } if (write) { - unsigned long page = 0; - char *kbuf; + char *kbuf, *p; if (left > PAGE_SIZE - 1) left = PAGE_SIZE - 1; - page = __get_free_page(GFP_TEMPORARY); - kbuf = (char *) page; - if (!kbuf) - return -ENOMEM; - if (copy_from_user(kbuf, buffer, left)) { - free_page(page); - return -EFAULT; - } - kbuf[left] = 0; + p = kbuf = memdup_user_nul(buffer, left); + if (IS_ERR(kbuf)) + return PTR_ERR(kbuf); tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long), GFP_KERNEL); if (!tmp_bitmap) { - free_page(page); + kfree(kbuf); return -ENOMEM; } - proc_skip_char(&kbuf, &left, '\n'); + proc_skip_char(&p, &left, '\n'); while (!err && left) { unsigned long val_a, val_b; bool neg; - err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a, + err = proc_get_long(&p, &left, &val_a, &neg, tr_a, sizeof(tr_a), &c); if (err) break; @@ -2688,12 +2665,12 @@ int proc_do_large_bitmap(struct ctl_table *table, int write, val_b = val_a; if (left) { - kbuf++; + p++; left--; } if (c == '-') { - err = proc_get_long(&kbuf, &left, &val_b, + err = proc_get_long(&p, &left, &val_b, &neg, tr_b, sizeof(tr_b), &c); if (err) @@ -2704,16 +2681,16 @@ int proc_do_large_bitmap(struct ctl_table *table, int write, break; } if (left) { - kbuf++; + p++; left--; } } bitmap_set(tmp_bitmap, val_a, val_b - val_a + 1); first = 0; - proc_skip_char(&kbuf, &left, '\n'); + proc_skip_char(&p, &left, '\n'); } - free_page(page); + kfree(kbuf); } else { unsigned long bit_a, bit_b = 0; diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 4f6ef6912e00..f333e57c4614 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -1340,15 +1340,9 @@ event_filter_write(struct file *filp, const char __user *ubuf, size_t cnt, if (cnt >= PAGE_SIZE) return -EINVAL; - buf = (char *)__get_free_page(GFP_TEMPORARY); - if (!buf) - return -ENOMEM; - - if (copy_from_user(buf, ubuf, cnt)) { - free_page((unsigned long) buf); - return -EFAULT; - } - buf[cnt] = '\0'; + buf = memdup_user_nul(ubuf, cnt); + if (IS_ERR(buf)) + return PTR_ERR(buf); mutex_lock(&event_mutex); file = event_file_data(filp); @@ -1356,7 +1350,7 @@ event_filter_write(struct file *filp, const char __user *ubuf, size_t cnt, err = apply_event_filter(file, buf); mutex_unlock(&event_mutex); - free_page((unsigned long) buf); + kfree(buf); if (err < 0) return err; @@ -1507,18 +1501,12 @@ subsystem_filter_write(struct file *filp, const char __user *ubuf, size_t cnt, if (cnt >= PAGE_SIZE) return -EINVAL; - buf = (char *)__get_free_page(GFP_TEMPORARY); - if (!buf) - return -ENOMEM; - - if (copy_from_user(buf, ubuf, cnt)) { - free_page((unsigned long) buf); - return -EFAULT; - } - buf[cnt] = '\0'; + buf = memdup_user_nul(ubuf, cnt); + if (IS_ERR(buf)) + return PTR_ERR(buf); err = apply_subsystem_event_filter(dir, buf); - free_page((unsigned long) buf); + kfree(buf); if (err < 0) return err; diff --git a/kernel/trace/trace_events_trigger.c b/kernel/trace/trace_events_trigger.c index 42a4009fd75a..4b5e8ed68d77 100644 --- a/kernel/trace/trace_events_trigger.c +++ b/kernel/trace/trace_events_trigger.c @@ -237,28 +237,23 @@ static ssize_t event_trigger_regex_write(struct file *file, if (cnt >= PAGE_SIZE) return -EINVAL; - buf = (char *)__get_free_page(GFP_TEMPORARY); - if (!buf) - return -ENOMEM; + buf = memdup_user_nul(ubuf, cnt); + if (IS_ERR(buf)) + return PTR_ERR(buf); - if (copy_from_user(buf, ubuf, cnt)) { - free_page((unsigned long)buf); - return -EFAULT; - } - buf[cnt] = '\0'; strim(buf); mutex_lock(&event_mutex); event_file = event_file_data(file); if (unlikely(!event_file)) { mutex_unlock(&event_mutex); - free_page((unsigned long)buf); + kfree(buf); return -ENODEV; } ret = trigger_process_regex(event_file, buf); mutex_unlock(&event_mutex); - free_page((unsigned long)buf); + kfree(buf); if (ret < 0) goto out; diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c index 88fefa68c516..9bafc211930c 100644 --- a/kernel/user_namespace.c +++ b/kernel/user_namespace.c @@ -602,8 +602,7 @@ static ssize_t map_write(struct file *file, const char __user *buf, struct uid_gid_map new_map; unsigned idx; struct uid_gid_extent *extent = NULL; - unsigned long page = 0; - char *kbuf, *pos, *next_line; + char *kbuf = NULL, *pos, *next_line; ssize_t ret = -EINVAL; /* @@ -638,23 +637,18 @@ static ssize_t map_write(struct file *file, const char __user *buf, if (cap_valid(cap_setid) && !file_ns_capable(file, ns, CAP_SYS_ADMIN)) goto out; - /* Get a buffer */ - ret = -ENOMEM; - page = __get_free_page(GFP_TEMPORARY); - kbuf = (char *) page; - if (!page) - goto out; - /* Only allow < page size writes at the beginning of the file */ ret = -EINVAL; if ((*ppos != 0) || (count >= PAGE_SIZE)) goto out; /* Slurp in the user data */ - ret = -EFAULT; - if (copy_from_user(kbuf, buf, count)) + kbuf = memdup_user_nul(buf, count); + if (IS_ERR(kbuf)) { + ret = PTR_ERR(kbuf); + kbuf = NULL; goto out; - kbuf[count] = '\0'; + } /* Parse the user data */ ret = -EINVAL; @@ -756,8 +750,7 @@ static ssize_t map_write(struct file *file, const char __user *buf, ret = count; out: mutex_unlock(&userns_state_mutex); - if (page) - free_page(page); + kfree(kbuf); return ret; } -- cgit From a1e9ca6967d68209c70e616a224efa89a6b86ca6 Mon Sep 17 00:00:00 2001 From: "Rafael J. Wysocki" Date: Sat, 2 Jan 2016 03:09:16 +0100 Subject: PM / sleep: Add support for read-only sysfs attributes Some sysfs attributes in /sys/power/ should really be read-only, so add support for that, convert those attributes to read-only and drop the stub .show() routines from them. Original-by: Sergey Senozhatsky Signed-off-by: Rafael J. Wysocki --- kernel/power/main.c | 17 ++--------------- kernel/power/power.h | 9 +++++++++ 2 files changed, 11 insertions(+), 15 deletions(-) (limited to 'kernel') diff --git a/kernel/power/main.c b/kernel/power/main.c index b2dd4d999900..27946975eff0 100644 --- a/kernel/power/main.c +++ b/kernel/power/main.c @@ -280,13 +280,7 @@ static ssize_t pm_wakeup_irq_show(struct kobject *kobj, return pm_wakeup_irq ? sprintf(buf, "%u\n", pm_wakeup_irq) : -ENODATA; } -static ssize_t pm_wakeup_irq_store(struct kobject *kobj, - struct kobj_attribute *attr, - const char *buf, size_t n) -{ - return -EINVAL; -} -power_attr(pm_wakeup_irq); +power_attr_ro(pm_wakeup_irq); #else /* !CONFIG_PM_SLEEP_DEBUG */ static inline void pm_print_times_init(void) {} @@ -564,14 +558,7 @@ static ssize_t pm_trace_dev_match_show(struct kobject *kobj, return show_trace_dev_match(buf, PAGE_SIZE); } -static ssize_t -pm_trace_dev_match_store(struct kobject *kobj, struct kobj_attribute *attr, - const char *buf, size_t n) -{ - return -EINVAL; -} - -power_attr(pm_trace_dev_match); +power_attr_ro(pm_trace_dev_match); #endif /* CONFIG_PM_TRACE */ diff --git a/kernel/power/power.h b/kernel/power/power.h index caadb566e82b..efe1b3b17c88 100644 --- a/kernel/power/power.h +++ b/kernel/power/power.h @@ -77,6 +77,15 @@ static struct kobj_attribute _name##_attr = { \ .store = _name##_store, \ } +#define power_attr_ro(_name) \ +static struct kobj_attribute _name##_attr = { \ + .attr = { \ + .name = __stringify(_name), \ + .mode = S_IRUGO, \ + }, \ + .show = _name##_show, \ +} + /* Preferred image size in bytes (default 500 MB) */ extern unsigned long image_size; /* Size of memory reserved for drivers (default SPARE_PAGES x PAGE_SIZE) */ -- cgit From c127449944659543e5e2423002f08f0af98dba5c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 10 Dec 2015 20:57:40 +0100 Subject: perf: Fix race in perf_event_exec() I managed to tickle this warning: [ 2338.884942] ------------[ cut here ]------------ [ 2338.890112] WARNING: CPU: 13 PID: 35162 at ../kernel/events/core.c:2702 task_ctx_sched_out+0x6b/0x80() [ 2338.900504] Modules linked in: [ 2338.903933] CPU: 13 PID: 35162 Comm: bash Not tainted 4.4.0-rc4-dirty #244 [ 2338.911610] Hardware name: Intel Corporation S2600GZ/S2600GZ, BIOS SE5C600.86B.02.02.0002.122320131210 12/23/2013 [ 2338.923071] ffffffff81f1468e ffff8807c6457cb8 ffffffff815c680c 0000000000000000 [ 2338.931382] ffff8807c6457cf0 ffffffff810c8a56 ffffe8ffff8c1bd0 ffff8808132ed400 [ 2338.939678] 0000000000000286 ffff880813170380 ffff8808132ed400 ffff8807c6457d00 [ 2338.947987] Call Trace: [ 2338.950726] [] dump_stack+0x4e/0x82 [ 2338.956474] [] warn_slowpath_common+0x86/0xc0 [ 2338.963195] [] warn_slowpath_null+0x1a/0x20 [ 2338.969720] [] task_ctx_sched_out+0x6b/0x80 [ 2338.976244] [] perf_event_exec+0xe2/0x180 [ 2338.982575] [] setup_new_exec+0x6f/0x1b0 [ 2338.988810] [] load_elf_binary+0x393/0x1660 [ 2338.995339] [] ? get_user_pages+0x52/0x60 [ 2339.001669] [] search_binary_handler+0x97/0x200 [ 2339.008581] [] do_execveat_common.isra.33+0x543/0x6e0 [ 2339.016072] [] SyS_execve+0x3a/0x50 [ 2339.021819] [] stub_execve+0x5/0x5 [ 2339.027469] [] ? entry_SYSCALL_64_fastpath+0x12/0x71 [ 2339.034860] ---[ end trace ee1337c59a0ddeac ]--- Which is a WARN_ON_ONCE() indicating that cpuctx->task_ctx is not what we expected it to be. This is because context switches can swap the task_struct::perf_event_ctxp[] pointer around. Therefore you have to either disable preemption when looking at current, or hold ctx->lock. Fix perf_event_enable_on_exec(), it loads current->perf_event_ctxp[] before disabling interrupts, therefore a preemption in the right place can swap contexts around and we're using the wrong one. Signed-off-by: Peter Zijlstra (Intel) Cc: Alexander Potapenko Cc: Arnaldo Carvalho de Melo Cc: Arnaldo Carvalho de Melo Cc: Dmitry Vyukov Cc: Eric Dumazet Cc: Jiri Olsa Cc: Kostya Serebryany Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Sasha Levin Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: syzkaller Link: http://lkml.kernel.org/r/20151210195740.GG6357@twins.programming.kicks-ass.net Signed-off-by: Ingo Molnar --- kernel/events/core.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 39cf4a40aa4c..fd7de0418fbe 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3154,15 +3154,16 @@ static int event_enable_on_exec(struct perf_event *event, * Enable all of a task's events that have been marked enable-on-exec. * This expects task == current. */ -static void perf_event_enable_on_exec(struct perf_event_context *ctx) +static void perf_event_enable_on_exec(int ctxn) { - struct perf_event_context *clone_ctx = NULL; + struct perf_event_context *ctx, *clone_ctx = NULL; struct perf_event *event; unsigned long flags; int enabled = 0; int ret; local_irq_save(flags); + ctx = current->perf_event_ctxp[ctxn]; if (!ctx || !ctx->nr_events) goto out; @@ -3205,17 +3206,11 @@ out: void perf_event_exec(void) { - struct perf_event_context *ctx; int ctxn; rcu_read_lock(); - for_each_task_context_nr(ctxn) { - ctx = current->perf_event_ctxp[ctxn]; - if (!ctx) - continue; - - perf_event_enable_on_exec(ctx); - } + for_each_task_context_nr(ctxn) + perf_event_enable_on_exec(ctxn); rcu_read_unlock(); } -- cgit From 12ca6ad2e3a896256f086497a7c7406a547ee373 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 15 Dec 2015 13:49:05 +0100 Subject: perf: Fix race in swevent hash There's a race on CPU unplug where we free the swevent hash array while it can still have events on. This will result in a use-after-free which is BAD. Simply do not free the hash array on unplug. This leaves the thing around and no use-after-free takes place. When the last swevent dies, we do a for_each_possible_cpu() iteration anyway to clean these up, at which time we'll free it, so no leakage will occur. Reported-by: Sasha Levin Tested-by: Sasha Levin Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Signed-off-by: Ingo Molnar --- kernel/events/core.c | 20 +------------------- 1 file changed, 1 insertion(+), 19 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index fd7de0418fbe..0a791a2203dc 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -6488,9 +6488,6 @@ struct swevent_htable { /* Recursion avoidance in each contexts */ int recursion[PERF_NR_CONTEXTS]; - - /* Keeps track of cpu being initialized/exited */ - bool online; }; static DEFINE_PER_CPU(struct swevent_htable, swevent_htable); @@ -6748,14 +6745,8 @@ static int perf_swevent_add(struct perf_event *event, int flags) hwc->state = !(flags & PERF_EF_START); head = find_swevent_head(swhash, event); - if (!head) { - /* - * We can race with cpu hotplug code. Do not - * WARN if the cpu just got unplugged. - */ - WARN_ON_ONCE(swhash->online); + if (WARN_ON_ONCE(!head)) return -EINVAL; - } hlist_add_head_rcu(&event->hlist_entry, head); perf_event_update_userpage(event); @@ -6823,7 +6814,6 @@ static int swevent_hlist_get_cpu(struct perf_event *event, int cpu) int err = 0; mutex_lock(&swhash->hlist_mutex); - if (!swevent_hlist_deref(swhash) && cpu_online(cpu)) { struct swevent_hlist *hlist; @@ -9286,7 +9276,6 @@ static void perf_event_init_cpu(int cpu) struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu); mutex_lock(&swhash->hlist_mutex); - swhash->online = true; if (swhash->hlist_refcount > 0) { struct swevent_hlist *hlist; @@ -9328,14 +9317,7 @@ static void perf_event_exit_cpu_context(int cpu) static void perf_event_exit_cpu(int cpu) { - struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu); - perf_event_exit_cpu_context(cpu); - - mutex_lock(&swhash->hlist_mutex); - swhash->online = false; - swevent_hlist_release(swhash); - mutex_unlock(&swhash->hlist_mutex); } #else static inline void perf_event_exit_cpu(int cpu) { } -- cgit From 9e0e83a1eca66f8369e5a02973f85aad65c32416 Mon Sep 17 00:00:00 2001 From: Andrey Ryabinin Date: Mon, 14 Dec 2015 15:47:23 +0300 Subject: sched/fair: Fix multiplication overflow on 32-bit systems Make 'r' 64-bit type to avoid overflow in 'r * LOAD_AVG_MAX' on 32-bit systems: UBSAN: Undefined behaviour in kernel/sched/fair.c:2785:18 signed integer overflow: 87950 * 47742 cannot be represented in type 'int' The most likely effect of this bug are bad load average numbers resulting in weird scheduling. It's also likely that this can persist for a longer time - until the system goes idle for a long time so that all load avg numbers get reset. [ This is the CFS load average metric, not the procfs output, which is separate. ] Signed-off-by: Andrey Ryabinin Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Fixes: 9d89c257dfb9 ("sched/fair: Rewrite runnable load and utilization average tracking") Link: http://lkml.kernel.org/r/1450097243-30137-1-git-send-email-aryabinin@virtuozzo.com [ Improved the changelog. ] Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 90e26b11deaa..cfdc0e61066c 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2689,7 +2689,7 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) int decayed, removed = 0; if (atomic_long_read(&cfs_rq->removed_load_avg)) { - long r = atomic_long_xchg(&cfs_rq->removed_load_avg, 0); + s64 r = atomic_long_xchg(&cfs_rq->removed_load_avg, 0); sa->load_avg = max_t(long, sa->load_avg - r, 0); sa->load_sum = max_t(s64, sa->load_sum - r * LOAD_AVG_MAX, 0); removed = 1; -- cgit From 093e5840ae76f1082633503964d035f40ed0216d Mon Sep 17 00:00:00 2001 From: Sebastian Andrzej Siewior Date: Mon, 21 Dec 2015 18:17:10 +0100 Subject: sched/core: Reset task's lockless wake-queues on fork() In the following commit: 7675104990ed ("sched: Implement lockless wake-queues") we gained lockless wake-queues. The -RT kernel managed to lockup itself with those. There could be multiple attempts for task X to enqueue it for a wakeup _even_ if task X is already running. The reason is that task X could be runnable but not yet on CPU. The the task performing the wakeup did not leave the CPU it could performe multiple wakeups. With the proper timming task X could be running and enqueued for a wakeup. If this happens while X is performing a fork() then its its child will have a !NULL `wake_q` member copied. This is not a problem as long as the child task does not participate in lockless wakeups :) Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Peter Zijlstra (Intel) Cc: Davidlohr Bueso Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Steven Rostedt Cc: Thomas Gleixner Fixes: 7675104990ed ("sched: Implement lockless wake-queues") Link: http://lkml.kernel.org/r/20151221171710.GA5499@linutronix.de Signed-off-by: Ingo Molnar --- kernel/fork.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/fork.c b/kernel/fork.c index fce002ee3ddf..1155eac61687 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -380,6 +380,7 @@ static struct task_struct *dup_task_struct(struct task_struct *orig) #endif tsk->splice_pipe = NULL; tsk->task_frag.page = NULL; + tsk->wake_q.next = NULL; account_kernel_stack(ti, 1); -- cgit From 7d92de3a8285ab3dfd68aa3a99823acd5b190444 Mon Sep 17 00:00:00 2001 From: Wanpeng Li Date: Thu, 3 Dec 2015 17:42:10 +0800 Subject: sched/deadline: Fix the earliest_dl.next logic earliest_dl.next should cache deadline of the earliest ready task that is also enqueued in the pushable rbtree, as pull algorithm uses this information to find candidates for migration: if the earliest_dl.next deadline of source rq is earlier than the earliest_dl.curr deadline of destination rq, the task from the source rq can be pulled. However, current implementation only guarantees that earliest_dl.next is the deadline of the next ready task instead of the next pushable task; which will result in potentially holding both rqs' lock and find nothing to migrate because of affinity constraints. In addition, current logic doesn't update the next candidate for pushing in pick_next_task_dl(), even if the running task is never eligible. This patch fixes both problems by updating earliest_dl.next when pushable dl task is enqueued/dequeued, similar to what we already do for RT. Tested-by: Luca Abeni Signed-off-by: Wanpeng Li Signed-off-by: Peter Zijlstra (Intel) Acked-by: Juri Lelli Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/1449135730-27202-1-git-send-email-wanpeng.li@hotmail.com Signed-off-by: Ingo Molnar --- kernel/sched/deadline.c | 59 ++++++------------------------------------------- 1 file changed, 7 insertions(+), 52 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 8b0a15e285f9..cd64c979d0e1 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -176,8 +176,10 @@ static void enqueue_pushable_dl_task(struct rq *rq, struct task_struct *p) } } - if (leftmost) + if (leftmost) { dl_rq->pushable_dl_tasks_leftmost = &p->pushable_dl_tasks; + dl_rq->earliest_dl.next = p->dl.deadline; + } rb_link_node(&p->pushable_dl_tasks, parent, link); rb_insert_color(&p->pushable_dl_tasks, &dl_rq->pushable_dl_tasks_root); @@ -195,6 +197,10 @@ static void dequeue_pushable_dl_task(struct rq *rq, struct task_struct *p) next_node = rb_next(&p->pushable_dl_tasks); dl_rq->pushable_dl_tasks_leftmost = next_node; + if (next_node) { + dl_rq->earliest_dl.next = rb_entry(next_node, + struct task_struct, pushable_dl_tasks)->dl.deadline; + } } rb_erase(&p->pushable_dl_tasks, &dl_rq->pushable_dl_tasks_root); @@ -782,42 +788,14 @@ static void update_curr_dl(struct rq *rq) #ifdef CONFIG_SMP -static struct task_struct *pick_next_earliest_dl_task(struct rq *rq, int cpu); - -static inline u64 next_deadline(struct rq *rq) -{ - struct task_struct *next = pick_next_earliest_dl_task(rq, rq->cpu); - - if (next && dl_prio(next->prio)) - return next->dl.deadline; - else - return 0; -} - static void inc_dl_deadline(struct dl_rq *dl_rq, u64 deadline) { struct rq *rq = rq_of_dl_rq(dl_rq); if (dl_rq->earliest_dl.curr == 0 || dl_time_before(deadline, dl_rq->earliest_dl.curr)) { - /* - * If the dl_rq had no -deadline tasks, or if the new task - * has shorter deadline than the current one on dl_rq, we - * know that the previous earliest becomes our next earliest, - * as the new task becomes the earliest itself. - */ - dl_rq->earliest_dl.next = dl_rq->earliest_dl.curr; dl_rq->earliest_dl.curr = deadline; cpudl_set(&rq->rd->cpudl, rq->cpu, deadline, 1); - } else if (dl_rq->earliest_dl.next == 0 || - dl_time_before(deadline, dl_rq->earliest_dl.next)) { - /* - * On the other hand, if the new -deadline task has a - * a later deadline than the earliest one on dl_rq, but - * it is earlier than the next (if any), we must - * recompute the next-earliest. - */ - dl_rq->earliest_dl.next = next_deadline(rq); } } @@ -839,7 +817,6 @@ static void dec_dl_deadline(struct dl_rq *dl_rq, u64 deadline) entry = rb_entry(leftmost, struct sched_dl_entity, rb_node); dl_rq->earliest_dl.curr = entry->deadline; - dl_rq->earliest_dl.next = next_deadline(rq); cpudl_set(&rq->rd->cpudl, rq->cpu, entry->deadline, 1); } } @@ -1274,28 +1251,6 @@ static int pick_dl_task(struct rq *rq, struct task_struct *p, int cpu) return 0; } -/* Returns the second earliest -deadline task, NULL otherwise */ -static struct task_struct *pick_next_earliest_dl_task(struct rq *rq, int cpu) -{ - struct rb_node *next_node = rq->dl.rb_leftmost; - struct sched_dl_entity *dl_se; - struct task_struct *p = NULL; - -next_node: - next_node = rb_next(next_node); - if (next_node) { - dl_se = rb_entry(next_node, struct sched_dl_entity, rb_node); - p = dl_task_of(dl_se); - - if (pick_dl_task(rq, p, cpu)) - return p; - - goto next_node; - } - - return NULL; -} - /* * Return the earliest pushable rq's task, which is suitable to be executed * on the CPU, NULL otherwise: -- cgit From 0905f04eb21fc1c2e690bed5d0418a061d56c225 Mon Sep 17 00:00:00 2001 From: Yuyang Du Date: Thu, 17 Dec 2015 07:34:27 +0800 Subject: sched/fair: Fix new task's load avg removed from source CPU in wake_up_new_task() If a newly created task is selected to go to a different CPU in fork balance when it wakes up the first time, its load averages should not be removed from the source CPU since they are never added to it before. The same is also applicable to a never used group entity. Fix it in remove_entity_load_avg(): when entity's last_update_time is 0, simply return. This should precisely identify the case in question, because in other migrations, the last_update_time is set to 0 after remove_entity_load_avg(). Reported-by: Steve Muckle Signed-off-by: Yuyang Du [peterz: cfs_rq_last_update_time] Signed-off-by: Peter Zijlstra (Intel) Cc: Dietmar Eggemann Cc: Juri Lelli Cc: Linus Torvalds Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Patrick Bellasi Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Vincent Guittot Link: http://lkml.kernel.org/r/20151216233427.GJ28098@intel.com Signed-off-by: Ingo Molnar --- kernel/sched/fair.c | 38 ++++++++++++++++++++++++++++---------- 1 file changed, 28 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 93efb962c2e1..1926606ece80 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2900,27 +2900,45 @@ dequeue_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) max_t(s64, cfs_rq->runnable_load_sum - se->avg.load_sum, 0); } -/* - * Task first catches up with cfs_rq, and then subtract - * itself from the cfs_rq (task must be off the queue now). - */ -void remove_entity_load_avg(struct sched_entity *se) -{ - struct cfs_rq *cfs_rq = cfs_rq_of(se); - u64 last_update_time; - #ifndef CONFIG_64BIT +static inline u64 cfs_rq_last_update_time(struct cfs_rq *cfs_rq) +{ u64 last_update_time_copy; + u64 last_update_time; do { last_update_time_copy = cfs_rq->load_last_update_time_copy; smp_rmb(); last_update_time = cfs_rq->avg.last_update_time; } while (last_update_time != last_update_time_copy); + + return last_update_time; +} #else - last_update_time = cfs_rq->avg.last_update_time; +static inline u64 cfs_rq_last_update_time(struct cfs_rq *cfs_rq) +{ + return cfs_rq->avg.last_update_time; +} #endif +/* + * Task first catches up with cfs_rq, and then subtract + * itself from the cfs_rq (task must be off the queue now). + */ +void remove_entity_load_avg(struct sched_entity *se) +{ + struct cfs_rq *cfs_rq = cfs_rq_of(se); + u64 last_update_time; + + /* + * Newly created task or never used group entity should not be removed + * from its (source) cfs_rq + */ + if (se->avg.last_update_time == 0) + return; + + last_update_time = cfs_rq_last_update_time(cfs_rq); + __update_load_avg(last_update_time, cpu_of(rq_of(cfs_rq)), &se->avg, 0, 0, NULL); atomic_long_add(se->avg.load_avg, &cfs_rq->removed_load_avg); atomic_long_add(se->avg.util_avg, &cfs_rq->removed_util_avg); -- cgit From 7b648018f628eee73450b71dc68ebb3c3865465e Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 3 Dec 2015 18:35:21 +0100 Subject: perf/core: Collapse more IPI loops This patch collapses the two 'hard' cases, which are perf_event_{dis,en}able(). I cannot seem to convince myself the current code is correct. So starting with perf_event_disable(); we don't strictly need to test for event->state == ACTIVE, ctx->is_active is enough. If the event is not scheduled while the ctx is, __perf_event_disable() still does the right thing. Its a little less efficient to IPI in that case, over-all simpler. For perf_event_enable(); the same goes, but I think that's actually broken in its current form. The current condition is: ctx->is_active && event->state == OFF, that means it doesn't do anything when !ctx->active && event->state == OFF. This is wrong, it should still mark the event INACTIVE in that case, otherwise we'll still not try and schedule the event once the context becomes active again. This patch implements the two function using the new event_function_call() and does away with the tricky event->state tests. Signed-off-by: Peter Zijlstra (Intel) Reviewed-by: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Signed-off-by: Ingo Molnar --- kernel/events/core.c | 106 ++++++++++++++++----------------------------------- 1 file changed, 33 insertions(+), 73 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 2c7bb20afc43..bf8244190d0f 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1766,6 +1766,20 @@ int __perf_event_disable(void *info) return 0; } +void ___perf_event_disable(void *info) +{ + struct perf_event *event = info; + + /* + * Since we have the lock this context can't be scheduled + * in, so we can change the state safely. + */ + if (event->state == PERF_EVENT_STATE_INACTIVE) { + update_group_times(event); + event->state = PERF_EVENT_STATE_OFF; + } +} + /* * Disable a event. * @@ -1782,43 +1796,16 @@ int __perf_event_disable(void *info) static void _perf_event_disable(struct perf_event *event) { struct perf_event_context *ctx = event->ctx; - struct task_struct *task = ctx->task; - - if (!task) { - /* - * Disable the event on the cpu that it's on - */ - cpu_function_call(event->cpu, __perf_event_disable, event); - return; - } - -retry: - if (!task_function_call(task, __perf_event_disable, event)) - return; raw_spin_lock_irq(&ctx->lock); - /* - * If the event is still active, we need to retry the cross-call. - */ - if (event->state == PERF_EVENT_STATE_ACTIVE) { + if (event->state <= PERF_EVENT_STATE_OFF) { raw_spin_unlock_irq(&ctx->lock); - /* - * Reload the task pointer, it might have been changed by - * a concurrent perf_event_context_sched_out(). - */ - task = ctx->task; - goto retry; - } - - /* - * Since we have the lock this context can't be scheduled - * in, so we can change the state safely. - */ - if (event->state == PERF_EVENT_STATE_INACTIVE) { - update_group_times(event); - event->state = PERF_EVENT_STATE_OFF; + return; } raw_spin_unlock_irq(&ctx->lock); + + event_function_call(event, __perf_event_disable, + ___perf_event_disable, event); } /* @@ -2269,6 +2256,11 @@ unlock: return 0; } +void ___perf_event_enable(void *info) +{ + __perf_event_mark_enabled((struct perf_event *)info); +} + /* * Enable a event. * @@ -2281,58 +2273,26 @@ unlock: static void _perf_event_enable(struct perf_event *event) { struct perf_event_context *ctx = event->ctx; - struct task_struct *task = ctx->task; - if (!task) { - /* - * Enable the event on the cpu that it's on - */ - cpu_function_call(event->cpu, __perf_event_enable, event); + raw_spin_lock_irq(&ctx->lock); + if (event->state >= PERF_EVENT_STATE_INACTIVE) { + raw_spin_unlock_irq(&ctx->lock); return; } - raw_spin_lock_irq(&ctx->lock); - if (event->state >= PERF_EVENT_STATE_INACTIVE) - goto out; - /* * If the event is in error state, clear that first. - * That way, if we see the event in error state below, we - * know that it has gone back into error state, as distinct - * from the task having been scheduled away before the - * cross-call arrived. + * + * That way, if we see the event in error state below, we know that it + * has gone back into error state, as distinct from the task having + * been scheduled away before the cross-call arrived. */ if (event->state == PERF_EVENT_STATE_ERROR) event->state = PERF_EVENT_STATE_OFF; - -retry: - if (!ctx->is_active) { - __perf_event_mark_enabled(event); - goto out; - } - raw_spin_unlock_irq(&ctx->lock); - if (!task_function_call(task, __perf_event_enable, event)) - return; - - raw_spin_lock_irq(&ctx->lock); - - /* - * If the context is active and the event is still off, - * we need to retry the cross-call. - */ - if (ctx->is_active && event->state == PERF_EVENT_STATE_OFF) { - /* - * task could have been flipped by a concurrent - * perf_event_context_sched_out() - */ - task = ctx->task; - goto retry; - } - -out: - raw_spin_unlock_irq(&ctx->lock); + event_function_call(event, __perf_event_enable, + ___perf_event_enable, event); } /* -- cgit From 6201171e3b2c02992e62448636631a0dfe4e9d20 Mon Sep 17 00:00:00 2001 From: wanghaibin Date: Thu, 7 Jan 2016 20:38:59 +0800 Subject: workqueue: simplify the apply_workqueue_attrs_locked() If the apply_wqattrs_prepare() returns NULL, it has already cleaned up the related resources, so it can return directly and avoid calling the clean up function again. This doesn't introduce any functional changes. Signed-off-by: wanghaibin Signed-off-by: Tejun Heo --- kernel/workqueue.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/workqueue.c b/kernel/workqueue.c index 1ecb588aae07..61a0264e28f9 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -3651,7 +3651,6 @@ static int apply_workqueue_attrs_locked(struct workqueue_struct *wq, const struct workqueue_attrs *attrs) { struct apply_wqattrs_ctx *ctx; - int ret = -ENOMEM; /* only unbound workqueues can change attributes */ if (WARN_ON(!(wq->flags & WQ_UNBOUND))) @@ -3662,16 +3661,14 @@ static int apply_workqueue_attrs_locked(struct workqueue_struct *wq, return -EINVAL; ctx = apply_wqattrs_prepare(wq, attrs); + if (!ctx) + return -ENOMEM; /* the ctx has been prepared successfully, let's commit it */ - if (ctx) { - apply_wqattrs_commit(ctx); - ret = 0; - } - + apply_wqattrs_commit(ctx); apply_wqattrs_cleanup(ctx); - return ret; + return 0; } /** -- cgit From 049fb9bd416077b3622d317a45796be4f2431df3 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Tue, 5 Jan 2016 20:32:47 -0500 Subject: ftrace/module: Call clean up function when module init fails early If the module init code fails after calling ftrace_module_init() and before calling do_init_module(), we can suffer from a memory leak. This is because ftrace_module_init() allocates pages to store the locations that ftrace hooks are placed in the module text. If do_init_module() fails, it still calls the MODULE_GOING notifiers which will tell ftrace to do a clean up of the pages it allocated for the module. But if load_module() fails before then, the pages allocated by ftrace_module_init() will never be freed. Call ftrace_release_mod() on the module if load_module() fails before getting to do_init_module(). Link: http://lkml.kernel.org/r/567CEA31.1070507@intel.com Reported-by: "Qiu, PeiyangX" Fixes: a949ae560a511 "ftrace/module: Hardcode ftrace_module_init() call into load_module()" Cc: stable@vger.kernel.org # v2.6.38+ Acked-by: Rusty Russell Signed-off-by: Steven Rostedt --- kernel/module.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'kernel') diff --git a/kernel/module.c b/kernel/module.c index 8f051a106676..38c7bd5583ff 100644 --- a/kernel/module.c +++ b/kernel/module.c @@ -3571,6 +3571,12 @@ static int load_module(struct load_info *info, const char __user *uargs, synchronize_sched(); mutex_unlock(&module_mutex); free_module: + /* + * Ftrace needs to clean up what it initialized. + * This does nothing if ftrace_module_init() wasn't called, + * but it must be called outside of module_mutex. + */ + ftrace_release_mod(mod); /* Free lock-classes; relies on the preceding sync_rcu() */ lockdep_free_key_range(mod->module_core, mod->core_size); -- cgit From b7ffffbb46f205e7727a18bcc7a46c3c2b534f7c Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (Red Hat)" Date: Thu, 7 Jan 2016 15:40:01 -0500 Subject: ftrace: Add infrastructure for delayed enabling of module functions Qiu Peiyang pointed out that there's a race when enabling function tracing and loading a module. In order to make the modifications of converting nops in the prologue of functions into callbacks, the text needs to be converted from read-only to read-write. When enabling function tracing, the text permission is updated, the functions are modified, and then they are put back. When loading a module, the updates to convert function calls to mcount is done before the module text is set to read-only. But after it is done, the module text is visible by the function tracer. Thus we have the following race: CPU 0 CPU 1 ----- ----- start function tracing set text to read-write load_module add functions to ftrace set module text read-only update all functions to callbacks modify module functions too < Can't it's read-only > When this happens, ftrace detects the issue and disables itself till the next reboot. To fix this, a new DISABLED flag is added for ftrace records, which all module functions get when they are added. Then later, after the module code is all set, the records will have the DISABLED flag cleared, and they will be enabled if any callback wants all functions to be traced. Note, this doesn't add the delay to later. It simply changes the ftrace_module_init() to do both the setting of DISABLED records, and then immediately calls the enable code. This helps with testing this new code as it has the same behavior as previously. Another change will come after this to have the ftrace_module_enable() called after the text is set to read-only. Cc: Qiu Peiyang Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 161 +++++++++++++++++++++++++++++++++----------------- 1 file changed, 106 insertions(+), 55 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 0f7ee341f89f..23683b06b18c 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1658,6 +1658,9 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops, int in_hash = 0; int match = 0; + if (rec->flags & FTRACE_FL_DISABLED) + continue; + if (all) { /* * Only the filter_hash affects all records. @@ -2023,6 +2026,9 @@ static int ftrace_check_record(struct dyn_ftrace *rec, int enable, int update) ftrace_bug_type = FTRACE_BUG_UNKNOWN; + if (rec->flags & FTRACE_FL_DISABLED) + return FTRACE_UPDATE_IGNORE; + /* * If we are updating calls: * @@ -2833,9 +2839,9 @@ ops_references_rec(struct ftrace_ops *ops, struct dyn_ftrace *rec) if (!(ops->flags & FTRACE_OPS_FL_ENABLED)) return 0; - /* If ops traces all mods, we already accounted for it */ + /* If ops traces all then it includes this function */ if (ops_traces_mod(ops)) - return 0; + return 1; /* The function must be in the filter */ if (!ftrace_hash_empty(ops->func_hash->filter_hash) && @@ -2849,64 +2855,41 @@ ops_references_rec(struct ftrace_ops *ops, struct dyn_ftrace *rec) return 1; } -static int referenced_filters(struct dyn_ftrace *rec) -{ - struct ftrace_ops *ops; - int cnt = 0; - - for (ops = ftrace_ops_list; ops != &ftrace_list_end; ops = ops->next) { - if (ops_references_rec(ops, rec)) - cnt++; - } - - return cnt; -} - static int ftrace_update_code(struct module *mod, struct ftrace_page *new_pgs) { struct ftrace_page *pg; struct dyn_ftrace *p; cycle_t start, stop; unsigned long update_cnt = 0; - unsigned long ref = 0; - bool test = false; + unsigned long rec_flags = 0; int i; + start = ftrace_now(raw_smp_processor_id()); + /* - * When adding a module, we need to check if tracers are - * currently enabled and if they are set to trace all functions. - * If they are, we need to enable the module functions as well - * as update the reference counts for those function records. + * When a module is loaded, this function is called to convert + * the calls to mcount in its text to nops, and also to create + * an entry in the ftrace data. Now, if ftrace is activated + * after this call, but before the module sets its text to + * read-only, the modification of enabling ftrace can fail if + * the read-only is done while ftrace is converting the calls. + * To prevent this, the module's records are set as disabled + * and will be enabled after the call to set the module's text + * to read-only. */ - if (mod) { - struct ftrace_ops *ops; - - for (ops = ftrace_ops_list; - ops != &ftrace_list_end; ops = ops->next) { - if (ops->flags & FTRACE_OPS_FL_ENABLED) { - if (ops_traces_mod(ops)) - ref++; - else - test = true; - } - } - } - - start = ftrace_now(raw_smp_processor_id()); + if (mod) + rec_flags |= FTRACE_FL_DISABLED; for (pg = new_pgs; pg; pg = pg->next) { for (i = 0; i < pg->index; i++) { - int cnt = ref; /* If something went wrong, bail without enabling anything */ if (unlikely(ftrace_disabled)) return -1; p = &pg->records[i]; - if (test) - cnt += referenced_filters(p); - p->flags = cnt; + p->flags = rec_flags; /* * Do the initial record conversion from mcount jump @@ -2916,21 +2899,6 @@ static int ftrace_update_code(struct module *mod, struct ftrace_page *new_pgs) break; update_cnt++; - - /* - * If the tracing is enabled, go ahead and enable the record. - * - * The reason not to enable the record immediatelly is the - * inherent check of ftrace_make_nop/ftrace_make_call for - * correct previous instructions. Making first the NOP - * conversion puts the module to the correct state, thus - * passing the ftrace_make_call check. - */ - if (ftrace_start_up && cnt) { - int failed = __ftrace_replace_code(p, 1); - if (failed) - ftrace_bug(failed, p); - } } } @@ -4938,6 +4906,19 @@ static int ftrace_process_locs(struct module *mod, #define next_to_ftrace_page(p) container_of(p, struct ftrace_page, next) +static int referenced_filters(struct dyn_ftrace *rec) +{ + struct ftrace_ops *ops; + int cnt = 0; + + for (ops = ftrace_ops_list; ops != &ftrace_list_end; ops = ops->next) { + if (ops_references_rec(ops, rec)) + cnt++; + } + + return cnt; +} + void ftrace_release_mod(struct module *mod) { struct dyn_ftrace *rec; @@ -4980,6 +4961,75 @@ void ftrace_release_mod(struct module *mod) mutex_unlock(&ftrace_lock); } +static void ftrace_module_enable(struct module *mod) +{ + struct dyn_ftrace *rec; + struct ftrace_page *pg; + + mutex_lock(&ftrace_lock); + + if (ftrace_disabled) + goto out_unlock; + + /* + * If the tracing is enabled, go ahead and enable the record. + * + * The reason not to enable the record immediatelly is the + * inherent check of ftrace_make_nop/ftrace_make_call for + * correct previous instructions. Making first the NOP + * conversion puts the module to the correct state, thus + * passing the ftrace_make_call check. + * + * We also delay this to after the module code already set the + * text to read-only, as we now need to set it back to read-write + * so that we can modify the text. + */ + if (ftrace_start_up) + ftrace_arch_code_modify_prepare(); + + do_for_each_ftrace_rec(pg, rec) { + int cnt; + /* + * do_for_each_ftrace_rec() is a double loop. + * module text shares the pg. If a record is + * not part of this module, then skip this pg, + * which the "break" will do. + */ + if (!within_module_core(rec->ip, mod)) + break; + + cnt = 0; + + /* + * When adding a module, we need to check if tracers are + * currently enabled and if they are, and can trace this record, + * we need to enable the module functions as well as update the + * reference counts for those function records. + */ + if (ftrace_start_up) + cnt += referenced_filters(rec); + + /* This clears FTRACE_FL_DISABLED */ + rec->flags = cnt; + + if (ftrace_start_up && cnt) { + int failed = __ftrace_replace_code(rec, 1); + if (failed) { + ftrace_bug(failed, rec); + goto out_loop; + } + } + + } while_for_each_ftrace_rec(); + + out_loop: + if (ftrace_start_up) + ftrace_arch_code_modify_post_process(); + + out_unlock: + mutex_unlock(&ftrace_lock); +} + void ftrace_module_init(struct module *mod) { if (ftrace_disabled || !mod->num_ftrace_callsites) @@ -4987,6 +5037,7 @@ void ftrace_module_init(struct module *mod) ftrace_process_locs(mod, mod->ftrace_callsites, mod->ftrace_callsites + mod->num_ftrace_callsites); + ftrace_module_enable(mod); } static int ftrace_module_notify_exit(struct notifier_block *self, -- cgit From 5156dca34a3e1e1edac2d0dabf43d8632909b7aa Mon Sep 17 00:00:00 2001 From: Qiu Peiyang Date: Fri, 25 Dec 2015 14:46:00 +0800 Subject: ftrace: Fix the race between ftrace and insmod We hit ftrace_bug report when booting Android on a 64bit ATOM SOC chip. Basically, there is a race between insmod and ftrace_run_update_code. After load_module=>ftrace_module_init, another thread jumps in to call ftrace_run_update_code=>ftrace_arch_code_modify_prepare =>set_all_modules_text_rw, to change all modules as RW. Since the new module is at MODULE_STATE_UNFORMED, the text attribute is not changed. Then, the 2nd thread goes ahead to change codes. However, load_module continues to call complete_formation=>set_section_ro_nx, then 2nd thread would fail when probing the module's TEXT. The patch fixes it by using notifier to delay the enabling of ftrace records to the time when module is at state MODULE_STATE_COMING. Link: http://lkml.kernel.org/r/567CE628.3000609@intel.com Signed-off-by: Qiu Peiyang Signed-off-by: Zhang Yanmin Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 25 ++++++++++++++++--------- 1 file changed, 16 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 23683b06b18c..eca592f977b2 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -5037,29 +5037,36 @@ void ftrace_module_init(struct module *mod) ftrace_process_locs(mod, mod->ftrace_callsites, mod->ftrace_callsites + mod->num_ftrace_callsites); - ftrace_module_enable(mod); } -static int ftrace_module_notify_exit(struct notifier_block *self, - unsigned long val, void *data) +static int ftrace_module_notify(struct notifier_block *self, + unsigned long val, void *data) { struct module *mod = data; - if (val == MODULE_STATE_GOING) + switch (val) { + case MODULE_STATE_COMING: + ftrace_module_enable(mod); + break; + case MODULE_STATE_GOING: ftrace_release_mod(mod); + break; + default: + break; + } return 0; } #else -static int ftrace_module_notify_exit(struct notifier_block *self, - unsigned long val, void *data) +static int ftrace_module_notify(struct notifier_block *self, + unsigned long val, void *data) { return 0; } #endif /* CONFIG_MODULES */ -struct notifier_block ftrace_module_exit_nb = { - .notifier_call = ftrace_module_notify_exit, +struct notifier_block ftrace_module_nb = { + .notifier_call = ftrace_module_notify, .priority = INT_MIN, /* Run after anything that can remove kprobes */ }; @@ -5091,7 +5098,7 @@ void __init ftrace_init(void) __start_mcount_loc, __stop_mcount_loc); - ret = register_module_notifier(&ftrace_module_exit_nb); + ret = register_module_notifier(&ftrace_module_nb); if (ret) pr_warning("Failed to register trace ftrace module exit notifier\n"); -- cgit From 2cfa2b191e56fd4c5c6ffb72e1d1e82ed225676d Mon Sep 17 00:00:00 2001 From: Rami Rosen Date: Sat, 9 Jan 2016 23:33:06 +0200 Subject: cgroup: fix a typo. This patch fixes a typo in a comment in cgroup.c. Signed-off-by: Rami Rosen Signed-off-by: Tejun Heo --- kernel/cgroup.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/cgroup.c b/kernel/cgroup.c index 122ec5543335..effb6366fd08 100644 --- a/kernel/cgroup.c +++ b/kernel/cgroup.c @@ -4037,7 +4037,7 @@ int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from) goto out_err; /* - * Migrate tasks one-by-one until @form is empty. This fails iff + * Migrate tasks one-by-one until @from is empty. This fails iff * ->can_attach() fails. */ do { -- cgit