From e50c54189f7c6211a99539156e3978474f0b1a0b Mon Sep 17 00:00:00 2001 From: Marc Zyngier Date: Thu, 13 Sep 2012 16:40:46 +0100 Subject: ARM: perf: add guest vs host discrimination Add minimal guest support to perf, so it can distinguish whether the PMU interrupt was in the host or the guest, as well as collecting some very basic information (guest PC, user vs kernel mode). This is not feature complete though, as it doesn't support backtracing in the guest. Based on the x86 implementation, tested with KVM/ARM. Signed-off-by: Marc Zyngier Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index 53c0304b734a..f8406af03279 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c @@ -576,6 +576,10 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs) { struct frame_tail __user *tail; + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + /* We don't support guest os callchain now */ + return; + } tail = (struct frame_tail __user *)regs->ARM_fp - 1; @@ -603,9 +607,41 @@ perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs) { struct stackframe fr; + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + /* We don't support guest os callchain now */ + return; + } + fr.fp = regs->ARM_fp; fr.sp = regs->ARM_sp; fr.lr = regs->ARM_lr; fr.pc = regs->ARM_pc; walk_stackframe(&fr, callchain_trace, entry); } + +unsigned long perf_instruction_pointer(struct pt_regs *regs) +{ + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) + return perf_guest_cbs->get_guest_ip(); + + return instruction_pointer(regs); +} + +unsigned long perf_misc_flags(struct pt_regs *regs) +{ + int misc = 0; + + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + if (perf_guest_cbs->is_user_mode()) + misc |= PERF_RECORD_MISC_GUEST_USER; + else + misc |= PERF_RECORD_MISC_GUEST_KERNEL; + } else { + if (user_mode(regs)) + misc |= PERF_RECORD_MISC_USER; + else + misc |= PERF_RECORD_MISC_KERNEL; + } + + return misc; +} -- cgit From 513c99ce4e64245be1f83f56039ec4891b451955 Mon Sep 17 00:00:00 2001 From: Sudeep KarkadaNagesha Date: Tue, 31 Jul 2012 10:11:23 +0100 Subject: ARM: perf: allocate CPU PMU dynamically at probe time Supporting multiple, heterogeneous CPU PMUs requires us to allocate the arm_pmu structures dynamically as the devices are probed. This patch removes the static structure definitions for each CPU PMU type and instead passes pointers to the PMU-specific init functions. Signed-off-by: Sudeep KarkadaNagesha Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event_cpu.c | 47 ++++++++++------ arch/arm/kernel/perf_event_v6.c | 72 ++++++++++++------------ arch/arm/kernel/perf_event_v7.c | 106 +++++++++++++++++++----------------- arch/arm/kernel/perf_event_xscale.c | 72 ++++++++++++------------ 4 files changed, 153 insertions(+), 144 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event_cpu.c b/arch/arm/kernel/perf_event_cpu.c index 8d7d8d4de9d6..3863fd405fa1 100644 --- a/arch/arm/kernel/perf_event_cpu.c +++ b/arch/arm/kernel/perf_event_cpu.c @@ -23,6 +23,7 @@ #include #include #include +#include #include #include @@ -195,13 +196,13 @@ static struct platform_device_id __devinitdata cpu_pmu_plat_device_ids[] = { /* * CPU PMU identification and probing. */ -static struct arm_pmu *__devinit probe_current_pmu(void) +static int __devinit probe_current_pmu(struct arm_pmu *pmu) { - struct arm_pmu *pmu = NULL; int cpu = get_cpu(); unsigned long cpuid = read_cpuid_id(); unsigned long implementor = (cpuid & 0xFF000000) >> 24; unsigned long part_number = (cpuid & 0xFFF0); + int ret = -ENODEV; pr_info("probing PMU on CPU %d\n", cpu); @@ -211,25 +212,25 @@ static struct arm_pmu *__devinit probe_current_pmu(void) case 0xB360: /* ARM1136 */ case 0xB560: /* ARM1156 */ case 0xB760: /* ARM1176 */ - pmu = armv6pmu_init(); + ret = armv6pmu_init(pmu); break; case 0xB020: /* ARM11mpcore */ - pmu = armv6mpcore_pmu_init(); + ret = armv6mpcore_pmu_init(pmu); break; case 0xC080: /* Cortex-A8 */ - pmu = armv7_a8_pmu_init(); + ret = armv7_a8_pmu_init(pmu); break; case 0xC090: /* Cortex-A9 */ - pmu = armv7_a9_pmu_init(); + ret = armv7_a9_pmu_init(pmu); break; case 0xC050: /* Cortex-A5 */ - pmu = armv7_a5_pmu_init(); + ret = armv7_a5_pmu_init(pmu); break; case 0xC0F0: /* Cortex-A15 */ - pmu = armv7_a15_pmu_init(); + ret = armv7_a15_pmu_init(pmu); break; case 0xC070: /* Cortex-A7 */ - pmu = armv7_a7_pmu_init(); + ret = armv7_a7_pmu_init(pmu); break; } /* Intel CPUs [xscale]. */ @@ -237,39 +238,51 @@ static struct arm_pmu *__devinit probe_current_pmu(void) part_number = (cpuid >> 13) & 0x7; switch (part_number) { case 1: - pmu = xscale1pmu_init(); + ret = xscale1pmu_init(pmu); break; case 2: - pmu = xscale2pmu_init(); + ret = xscale2pmu_init(pmu); break; } } put_cpu(); - return pmu; + return ret; } static int __devinit cpu_pmu_device_probe(struct platform_device *pdev) { const struct of_device_id *of_id; - struct arm_pmu *(*init_fn)(void); + int (*init_fn)(struct arm_pmu *); struct device_node *node = pdev->dev.of_node; + struct arm_pmu *pmu; + int ret = -ENODEV; if (cpu_pmu) { pr_info("attempt to register multiple PMU devices!"); return -ENOSPC; } + pmu = kzalloc(sizeof(struct arm_pmu), GFP_KERNEL); + if (!pmu) { + pr_info("failed to allocate PMU device!"); + return -ENOMEM; + } + if (node && (of_id = of_match_node(cpu_pmu_of_device_ids, pdev->dev.of_node))) { init_fn = of_id->data; - cpu_pmu = init_fn(); + ret = init_fn(pmu); } else { - cpu_pmu = probe_current_pmu(); + ret = probe_current_pmu(pmu); } - if (!cpu_pmu) - return -ENODEV; + if (ret) { + pr_info("failed to register PMU devices!"); + kfree(pmu); + return ret; + } + cpu_pmu = pmu; cpu_pmu->plat_device = pdev; cpu_pmu_init(cpu_pmu); register_cpu_notifier(&cpu_pmu_hotplug_notifier); diff --git a/arch/arm/kernel/perf_event_v6.c b/arch/arm/kernel/perf_event_v6.c index 6ccc07971745..3908cb4e5566 100644 --- a/arch/arm/kernel/perf_event_v6.c +++ b/arch/arm/kernel/perf_event_v6.c @@ -649,24 +649,22 @@ static int armv6_map_event(struct perf_event *event) &armv6_perf_cache_map, 0xFF); } -static struct arm_pmu armv6pmu = { - .name = "v6", - .handle_irq = armv6pmu_handle_irq, - .enable = armv6pmu_enable_event, - .disable = armv6pmu_disable_event, - .read_counter = armv6pmu_read_counter, - .write_counter = armv6pmu_write_counter, - .get_event_idx = armv6pmu_get_event_idx, - .start = armv6pmu_start, - .stop = armv6pmu_stop, - .map_event = armv6_map_event, - .num_events = 3, - .max_period = (1LLU << 32) - 1, -}; - -static struct arm_pmu *__devinit armv6pmu_init(void) +static int __devinit armv6pmu_init(struct arm_pmu *cpu_pmu) { - return &armv6pmu; + cpu_pmu->name = "v6"; + cpu_pmu->handle_irq = armv6pmu_handle_irq; + cpu_pmu->enable = armv6pmu_enable_event; + cpu_pmu->disable = armv6pmu_disable_event; + cpu_pmu->read_counter = armv6pmu_read_counter; + cpu_pmu->write_counter = armv6pmu_write_counter; + cpu_pmu->get_event_idx = armv6pmu_get_event_idx; + cpu_pmu->start = armv6pmu_start; + cpu_pmu->stop = armv6pmu_stop; + cpu_pmu->map_event = armv6_map_event; + cpu_pmu->num_events = 3; + cpu_pmu->max_period = (1LLU << 32) - 1; + + return 0; } /* @@ -683,33 +681,31 @@ static int armv6mpcore_map_event(struct perf_event *event) &armv6mpcore_perf_cache_map, 0xFF); } -static struct arm_pmu armv6mpcore_pmu = { - .name = "v6mpcore", - .handle_irq = armv6pmu_handle_irq, - .enable = armv6pmu_enable_event, - .disable = armv6mpcore_pmu_disable_event, - .read_counter = armv6pmu_read_counter, - .write_counter = armv6pmu_write_counter, - .get_event_idx = armv6pmu_get_event_idx, - .start = armv6pmu_start, - .stop = armv6pmu_stop, - .map_event = armv6mpcore_map_event, - .num_events = 3, - .max_period = (1LLU << 32) - 1, -}; - -static struct arm_pmu *__devinit armv6mpcore_pmu_init(void) +static int __devinit armv6mpcore_pmu_init(struct arm_pmu *cpu_pmu) { - return &armv6mpcore_pmu; + cpu_pmu->name = "v6mpcore"; + cpu_pmu->handle_irq = armv6pmu_handle_irq; + cpu_pmu->enable = armv6pmu_enable_event; + cpu_pmu->disable = armv6mpcore_pmu_disable_event; + cpu_pmu->read_counter = armv6pmu_read_counter; + cpu_pmu->write_counter = armv6pmu_write_counter; + cpu_pmu->get_event_idx = armv6pmu_get_event_idx; + cpu_pmu->start = armv6pmu_start; + cpu_pmu->stop = armv6pmu_stop; + cpu_pmu->map_event = armv6mpcore_map_event; + cpu_pmu->num_events = 3; + cpu_pmu->max_period = (1LLU << 32) - 1; + + return 0; } #else -static struct arm_pmu *__devinit armv6pmu_init(void) +static int armv6pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } -static struct arm_pmu *__devinit armv6mpcore_pmu_init(void) +static int armv6mpcore_pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } #endif /* CONFIG_CPU_V6 || CONFIG_CPU_V6K */ diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c index bd4b090ebcfd..b189403f30e4 100644 --- a/arch/arm/kernel/perf_event_v7.c +++ b/arch/arm/kernel/perf_event_v7.c @@ -18,8 +18,6 @@ #ifdef CONFIG_CPU_V7 -static struct arm_pmu armv7pmu; - /* * Common ARMv7 event types * @@ -1014,7 +1012,7 @@ static void armv7pmu_enable_event(struct hw_perf_event *hwc, int idx) * We only need to set the event for the cycle counter if we * have the ability to perform event filtering. */ - if (armv7pmu.set_event_filter || idx != ARMV7_IDX_CYCLE_COUNTER) + if (cpu_pmu->set_event_filter || idx != ARMV7_IDX_CYCLE_COUNTER) armv7_pmnc_write_evtsel(idx, hwc->config_base); /* @@ -1232,17 +1230,18 @@ static int armv7_a7_map_event(struct perf_event *event) &armv7_a7_perf_cache_map, 0xFF); } -static struct arm_pmu armv7pmu = { - .handle_irq = armv7pmu_handle_irq, - .enable = armv7pmu_enable_event, - .disable = armv7pmu_disable_event, - .read_counter = armv7pmu_read_counter, - .write_counter = armv7pmu_write_counter, - .get_event_idx = armv7pmu_get_event_idx, - .start = armv7pmu_start, - .stop = armv7pmu_stop, - .reset = armv7pmu_reset, - .max_period = (1LLU << 32) - 1, +static void armv7pmu_init(struct arm_pmu *cpu_pmu) +{ + cpu_pmu->handle_irq = armv7pmu_handle_irq; + cpu_pmu->enable = armv7pmu_enable_event; + cpu_pmu->disable = armv7pmu_disable_event; + cpu_pmu->read_counter = armv7pmu_read_counter; + cpu_pmu->write_counter = armv7pmu_write_counter; + cpu_pmu->get_event_idx = armv7pmu_get_event_idx; + cpu_pmu->start = armv7pmu_start; + cpu_pmu->stop = armv7pmu_stop; + cpu_pmu->reset = armv7pmu_reset; + cpu_pmu->max_period = (1LLU << 32) - 1; }; static u32 __devinit armv7_read_num_pmnc_events(void) @@ -1256,70 +1255,75 @@ static u32 __devinit armv7_read_num_pmnc_events(void) return nb_cnt + 1; } -static struct arm_pmu *__devinit armv7_a8_pmu_init(void) +static int __devinit armv7_a8_pmu_init(struct arm_pmu *cpu_pmu) { - armv7pmu.name = "ARMv7 Cortex-A8"; - armv7pmu.map_event = armv7_a8_map_event; - armv7pmu.num_events = armv7_read_num_pmnc_events(); - return &armv7pmu; + armv7pmu_init(cpu_pmu); + cpu_pmu->name = "ARMv7 Cortex-A8"; + cpu_pmu->map_event = armv7_a8_map_event; + cpu_pmu->num_events = armv7_read_num_pmnc_events(); + return 0; } -static struct arm_pmu *__devinit armv7_a9_pmu_init(void) +static int __devinit armv7_a9_pmu_init(struct arm_pmu *cpu_pmu) { - armv7pmu.name = "ARMv7 Cortex-A9"; - armv7pmu.map_event = armv7_a9_map_event; - armv7pmu.num_events = armv7_read_num_pmnc_events(); - return &armv7pmu; + armv7pmu_init(cpu_pmu); + cpu_pmu->name = "ARMv7 Cortex-A9"; + cpu_pmu->map_event = armv7_a9_map_event; + cpu_pmu->num_events = armv7_read_num_pmnc_events(); + return 0; } -static struct arm_pmu *__devinit armv7_a5_pmu_init(void) +static int __devinit armv7_a5_pmu_init(struct arm_pmu *cpu_pmu) { - armv7pmu.name = "ARMv7 Cortex-A5"; - armv7pmu.map_event = armv7_a5_map_event; - armv7pmu.num_events = armv7_read_num_pmnc_events(); - return &armv7pmu; + armv7pmu_init(cpu_pmu); + cpu_pmu->name = "ARMv7 Cortex-A5"; + cpu_pmu->map_event = armv7_a5_map_event; + cpu_pmu->num_events = armv7_read_num_pmnc_events(); + return 0; } -static struct arm_pmu *__devinit armv7_a15_pmu_init(void) +static int __devinit armv7_a15_pmu_init(struct arm_pmu *cpu_pmu) { - armv7pmu.name = "ARMv7 Cortex-A15"; - armv7pmu.map_event = armv7_a15_map_event; - armv7pmu.num_events = armv7_read_num_pmnc_events(); - armv7pmu.set_event_filter = armv7pmu_set_event_filter; - return &armv7pmu; + armv7pmu_init(cpu_pmu); + cpu_pmu->name = "ARMv7 Cortex-A15"; + cpu_pmu->map_event = armv7_a15_map_event; + cpu_pmu->num_events = armv7_read_num_pmnc_events(); + cpu_pmu->set_event_filter = armv7pmu_set_event_filter; + return 0; } -static struct arm_pmu *__devinit armv7_a7_pmu_init(void) +static int __devinit armv7_a7_pmu_init(struct arm_pmu *cpu_pmu) { - armv7pmu.name = "ARMv7 Cortex-A7"; - armv7pmu.map_event = armv7_a7_map_event; - armv7pmu.num_events = armv7_read_num_pmnc_events(); - armv7pmu.set_event_filter = armv7pmu_set_event_filter; - return &armv7pmu; + armv7pmu_init(cpu_pmu); + cpu_pmu->name = "ARMv7 Cortex-A7"; + cpu_pmu->map_event = armv7_a7_map_event; + cpu_pmu->num_events = armv7_read_num_pmnc_events(); + cpu_pmu->set_event_filter = armv7pmu_set_event_filter; + return 0; } #else -static struct arm_pmu *__devinit armv7_a8_pmu_init(void) +static inline int armv7_a8_pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } -static struct arm_pmu *__devinit armv7_a9_pmu_init(void) +static inline int armv7_a9_pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } -static struct arm_pmu *__devinit armv7_a5_pmu_init(void) +static inline int armv7_a5_pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } -static struct arm_pmu *__devinit armv7_a15_pmu_init(void) +static inline int armv7_a15_pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } -static struct arm_pmu *__devinit armv7_a7_pmu_init(void) +static inline int armv7_a7_pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } #endif /* CONFIG_CPU_V7 */ diff --git a/arch/arm/kernel/perf_event_xscale.c b/arch/arm/kernel/perf_event_xscale.c index 426e19f380a2..131ede6c2fdf 100644 --- a/arch/arm/kernel/perf_event_xscale.c +++ b/arch/arm/kernel/perf_event_xscale.c @@ -434,24 +434,22 @@ static int xscale_map_event(struct perf_event *event) &xscale_perf_cache_map, 0xFF); } -static struct arm_pmu xscale1pmu = { - .name = "xscale1", - .handle_irq = xscale1pmu_handle_irq, - .enable = xscale1pmu_enable_event, - .disable = xscale1pmu_disable_event, - .read_counter = xscale1pmu_read_counter, - .write_counter = xscale1pmu_write_counter, - .get_event_idx = xscale1pmu_get_event_idx, - .start = xscale1pmu_start, - .stop = xscale1pmu_stop, - .map_event = xscale_map_event, - .num_events = 3, - .max_period = (1LLU << 32) - 1, -}; - -static struct arm_pmu *__devinit xscale1pmu_init(void) +static int __devinit xscale1pmu_init(struct arm_pmu *cpu_pmu) { - return &xscale1pmu; + cpu_pmu->name = "xscale1"; + cpu_pmu->handle_irq = xscale1pmu_handle_irq; + cpu_pmu->enable = xscale1pmu_enable_event; + cpu_pmu->disable = xscale1pmu_disable_event; + cpu_pmu->read_counter = xscale1pmu_read_counter; + cpu_pmu->write_counter = xscale1pmu_write_counter; + cpu_pmu->get_event_idx = xscale1pmu_get_event_idx; + cpu_pmu->start = xscale1pmu_start; + cpu_pmu->stop = xscale1pmu_stop; + cpu_pmu->map_event = xscale_map_event; + cpu_pmu->num_events = 3; + cpu_pmu->max_period = (1LLU << 32) - 1; + + return 0; } #define XSCALE2_OVERFLOWED_MASK 0x01f @@ -801,33 +799,31 @@ xscale2pmu_write_counter(int counter, u32 val) } } -static struct arm_pmu xscale2pmu = { - .name = "xscale2", - .handle_irq = xscale2pmu_handle_irq, - .enable = xscale2pmu_enable_event, - .disable = xscale2pmu_disable_event, - .read_counter = xscale2pmu_read_counter, - .write_counter = xscale2pmu_write_counter, - .get_event_idx = xscale2pmu_get_event_idx, - .start = xscale2pmu_start, - .stop = xscale2pmu_stop, - .map_event = xscale_map_event, - .num_events = 5, - .max_period = (1LLU << 32) - 1, -}; - -static struct arm_pmu *__devinit xscale2pmu_init(void) +static int __devinit xscale2pmu_init(struct arm_pmu *cpu_pmu) { - return &xscale2pmu; + cpu_pmu->name = "xscale2"; + cpu_pmu->handle_irq = xscale2pmu_handle_irq; + cpu_pmu->enable = xscale2pmu_enable_event; + cpu_pmu->disable = xscale2pmu_disable_event; + cpu_pmu->read_counter = xscale2pmu_read_counter; + cpu_pmu->write_counter = xscale2pmu_write_counter; + cpu_pmu->get_event_idx = xscale2pmu_get_event_idx; + cpu_pmu->start = xscale2pmu_start; + cpu_pmu->stop = xscale2pmu_stop; + cpu_pmu->map_event = xscale_map_event; + cpu_pmu->num_events = 5; + cpu_pmu->max_period = (1LLU << 32) - 1; + + return 0; } #else -static struct arm_pmu *__devinit xscale1pmu_init(void) +static inline int xscale1pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } -static struct arm_pmu *__devinit xscale2pmu_init(void) +static inline int xscale2pmu_init(struct arm_pmu *cpu_pmu) { - return NULL; + return -ENODEV; } #endif /* CONFIG_CPU_XSCALE */ -- cgit From ed6f2a522398c26559f4da23a80aa6195e6284c7 Mon Sep 17 00:00:00 2001 From: Sudeep KarkadaNagesha Date: Mon, 30 Jul 2012 12:00:02 +0100 Subject: ARM: perf: consistently use struct perf_event in arm_pmu functions The arm_pmu functions have wildly varied parameters which can often be derived from struct perf_event. This patch changes the arm_pmu function prototypes so that struct perf_event pointers are passed in preference to fields that can be derived from the event. Signed-off-by: Sudeep KarkadaNagesha Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event.c | 44 +++++++++---------- arch/arm/kernel/perf_event_cpu.c | 8 ++-- arch/arm/kernel/perf_event_v6.c | 54 ++++++++++++----------- arch/arm/kernel/perf_event_v7.c | 46 +++++++++++++------- arch/arm/kernel/perf_event_xscale.c | 85 +++++++++++++++++++++---------------- 5 files changed, 131 insertions(+), 106 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index f8406af03279..1cfa3f35713e 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c @@ -86,12 +86,10 @@ armpmu_map_event(struct perf_event *event, return -ENOENT; } -int -armpmu_event_set_period(struct perf_event *event, - struct hw_perf_event *hwc, - int idx) +int armpmu_event_set_period(struct perf_event *event) { struct arm_pmu *armpmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; s64 left = local64_read(&hwc->period_left); s64 period = hwc->sample_period; int ret = 0; @@ -119,24 +117,22 @@ armpmu_event_set_period(struct perf_event *event, local64_set(&hwc->prev_count, (u64)-left); - armpmu->write_counter(idx, (u64)(-left) & 0xffffffff); + armpmu->write_counter(event, (u64)(-left) & 0xffffffff); perf_event_update_userpage(event); return ret; } -u64 -armpmu_event_update(struct perf_event *event, - struct hw_perf_event *hwc, - int idx) +u64 armpmu_event_update(struct perf_event *event) { struct arm_pmu *armpmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; u64 delta, prev_raw_count, new_raw_count; again: prev_raw_count = local64_read(&hwc->prev_count); - new_raw_count = armpmu->read_counter(idx); + new_raw_count = armpmu->read_counter(event); if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, new_raw_count) != prev_raw_count) @@ -159,7 +155,7 @@ armpmu_read(struct perf_event *event) if (hwc->idx < 0) return; - armpmu_event_update(event, hwc, hwc->idx); + armpmu_event_update(event); } static void @@ -173,14 +169,13 @@ armpmu_stop(struct perf_event *event, int flags) * PERF_EF_UPDATE, see comments in armpmu_start(). */ if (!(hwc->state & PERF_HES_STOPPED)) { - armpmu->disable(hwc, hwc->idx); - armpmu_event_update(event, hwc, hwc->idx); + armpmu->disable(event); + armpmu_event_update(event); hwc->state |= PERF_HES_STOPPED | PERF_HES_UPTODATE; } } -static void -armpmu_start(struct perf_event *event, int flags) +static void armpmu_start(struct perf_event *event, int flags) { struct arm_pmu *armpmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; @@ -200,8 +195,8 @@ armpmu_start(struct perf_event *event, int flags) * get an interrupt too soon or *way* too late if the overflow has * happened since disabling. */ - armpmu_event_set_period(event, hwc, hwc->idx); - armpmu->enable(hwc, hwc->idx); + armpmu_event_set_period(event); + armpmu->enable(event); } static void @@ -233,7 +228,7 @@ armpmu_add(struct perf_event *event, int flags) perf_pmu_disable(event->pmu); /* If we don't have a space for the counter then finish early. */ - idx = armpmu->get_event_idx(hw_events, hwc); + idx = armpmu->get_event_idx(hw_events, event); if (idx < 0) { err = idx; goto out; @@ -244,7 +239,7 @@ armpmu_add(struct perf_event *event, int flags) * sure it is disabled. */ event->hw.idx = idx; - armpmu->disable(hwc, idx); + armpmu->disable(event); hw_events->events[idx] = event; hwc->state = PERF_HES_STOPPED | PERF_HES_UPTODATE; @@ -264,13 +259,12 @@ validate_event(struct pmu_hw_events *hw_events, struct perf_event *event) { struct arm_pmu *armpmu = to_arm_pmu(event->pmu); - struct hw_perf_event fake_event = event->hw; struct pmu *leader_pmu = event->group_leader->pmu; if (event->pmu != leader_pmu || event->state <= PERF_EVENT_STATE_OFF) return 1; - return armpmu->get_event_idx(hw_events, &fake_event) >= 0; + return armpmu->get_event_idx(hw_events, event) >= 0; } static int @@ -316,7 +310,7 @@ static irqreturn_t armpmu_dispatch_irq(int irq, void *dev) static void armpmu_release_hardware(struct arm_pmu *armpmu) { - armpmu->free_irq(); + armpmu->free_irq(armpmu); pm_runtime_put_sync(&armpmu->plat_device->dev); } @@ -330,7 +324,7 @@ armpmu_reserve_hardware(struct arm_pmu *armpmu) return -ENODEV; pm_runtime_get_sync(&pmu_device->dev); - err = armpmu->request_irq(armpmu_dispatch_irq); + err = armpmu->request_irq(armpmu, armpmu_dispatch_irq); if (err) { armpmu_release_hardware(armpmu); return err; @@ -465,13 +459,13 @@ static void armpmu_enable(struct pmu *pmu) int enabled = bitmap_weight(hw_events->used_mask, armpmu->num_events); if (enabled) - armpmu->start(); + armpmu->start(armpmu); } static void armpmu_disable(struct pmu *pmu) { struct arm_pmu *armpmu = to_arm_pmu(pmu); - armpmu->stop(); + armpmu->stop(armpmu); } #ifdef CONFIG_PM_RUNTIME diff --git a/arch/arm/kernel/perf_event_cpu.c b/arch/arm/kernel/perf_event_cpu.c index 3863fd405fa1..02244faa5396 100644 --- a/arch/arm/kernel/perf_event_cpu.c +++ b/arch/arm/kernel/perf_event_cpu.c @@ -71,7 +71,7 @@ static struct pmu_hw_events *cpu_pmu_get_cpu_events(void) return &__get_cpu_var(cpu_hw_events); } -static void cpu_pmu_free_irq(void) +static void cpu_pmu_free_irq(struct arm_pmu *cpu_pmu) { int i, irq, irqs; struct platform_device *pmu_device = cpu_pmu->plat_device; @@ -87,7 +87,7 @@ static void cpu_pmu_free_irq(void) } } -static int cpu_pmu_request_irq(irq_handler_t handler) +static int cpu_pmu_request_irq(struct arm_pmu *cpu_pmu, irq_handler_t handler) { int i, err, irq, irqs; struct platform_device *pmu_device = cpu_pmu->plat_device; @@ -148,7 +148,7 @@ static void __devinit cpu_pmu_init(struct arm_pmu *cpu_pmu) /* Ensure the PMU has sane values out of reset. */ if (cpu_pmu && cpu_pmu->reset) - on_each_cpu(cpu_pmu->reset, NULL, 1); + on_each_cpu(cpu_pmu->reset, cpu_pmu, 1); } /* @@ -164,7 +164,7 @@ static int __cpuinit cpu_pmu_notify(struct notifier_block *b, return NOTIFY_DONE; if (cpu_pmu && cpu_pmu->reset) - cpu_pmu->reset(NULL); + cpu_pmu->reset(cpu_pmu); return NOTIFY_OK; } diff --git a/arch/arm/kernel/perf_event_v6.c b/arch/arm/kernel/perf_event_v6.c index 3908cb4e5566..f3e22ff8b6a2 100644 --- a/arch/arm/kernel/perf_event_v6.c +++ b/arch/arm/kernel/perf_event_v6.c @@ -401,9 +401,10 @@ armv6_pmcr_counter_has_overflowed(unsigned long pmcr, return ret; } -static inline u32 -armv6pmu_read_counter(int counter) +static inline u32 armv6pmu_read_counter(struct perf_event *event) { + struct hw_perf_event *hwc = &event->hw; + int counter = hwc->idx; unsigned long value = 0; if (ARMV6_CYCLE_COUNTER == counter) @@ -418,10 +419,11 @@ armv6pmu_read_counter(int counter) return value; } -static inline void -armv6pmu_write_counter(int counter, - u32 value) +static inline void armv6pmu_write_counter(struct perf_event *event, u32 value) { + struct hw_perf_event *hwc = &event->hw; + int counter = hwc->idx; + if (ARMV6_CYCLE_COUNTER == counter) asm volatile("mcr p15, 0, %0, c15, c12, 1" : : "r"(value)); else if (ARMV6_COUNTER0 == counter) @@ -432,12 +434,13 @@ armv6pmu_write_counter(int counter, WARN_ONCE(1, "invalid counter number (%d)\n", counter); } -static void -armv6pmu_enable_event(struct hw_perf_event *hwc, - int idx) +static void armv6pmu_enable_event(struct perf_event *event) { unsigned long val, mask, evt, flags; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; if (ARMV6_CYCLE_COUNTER == idx) { mask = 0; @@ -473,7 +476,8 @@ armv6pmu_handle_irq(int irq_num, { unsigned long pmcr = armv6_pmcr_read(); struct perf_sample_data data; - struct pmu_hw_events *cpuc; + struct arm_pmu *cpu_pmu = (struct arm_pmu *)dev; + struct pmu_hw_events *cpuc = cpu_pmu->get_hw_events(); struct pt_regs *regs; int idx; @@ -489,7 +493,6 @@ armv6pmu_handle_irq(int irq_num, */ armv6_pmcr_write(pmcr); - cpuc = &__get_cpu_var(cpu_hw_events); for (idx = 0; idx < cpu_pmu->num_events; ++idx) { struct perf_event *event = cpuc->events[idx]; struct hw_perf_event *hwc; @@ -506,13 +509,13 @@ armv6pmu_handle_irq(int irq_num, continue; hwc = &event->hw; - armpmu_event_update(event, hwc, idx); + armpmu_event_update(event); perf_sample_data_init(&data, 0, hwc->last_period); - if (!armpmu_event_set_period(event, hwc, idx)) + if (!armpmu_event_set_period(event)) continue; if (perf_event_overflow(event, &data, regs)) - cpu_pmu->disable(hwc, idx); + cpu_pmu->disable(event); } /* @@ -527,8 +530,7 @@ armv6pmu_handle_irq(int irq_num, return IRQ_HANDLED; } -static void -armv6pmu_start(void) +static void armv6pmu_start(struct arm_pmu *cpu_pmu) { unsigned long flags, val; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -540,8 +542,7 @@ armv6pmu_start(void) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void -armv6pmu_stop(void) +static void armv6pmu_stop(struct arm_pmu *cpu_pmu) { unsigned long flags, val; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -555,10 +556,11 @@ armv6pmu_stop(void) static int armv6pmu_get_event_idx(struct pmu_hw_events *cpuc, - struct hw_perf_event *event) + struct perf_event *event) { + struct hw_perf_event *hwc = &event->hw; /* Always place a cycle counter into the cycle counter. */ - if (ARMV6_PERFCTR_CPU_CYCLES == event->config_base) { + if (ARMV6_PERFCTR_CPU_CYCLES == hwc->config_base) { if (test_and_set_bit(ARMV6_CYCLE_COUNTER, cpuc->used_mask)) return -EAGAIN; @@ -579,12 +581,13 @@ armv6pmu_get_event_idx(struct pmu_hw_events *cpuc, } } -static void -armv6pmu_disable_event(struct hw_perf_event *hwc, - int idx) +static void armv6pmu_disable_event(struct perf_event *event) { unsigned long val, mask, evt, flags; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; if (ARMV6_CYCLE_COUNTER == idx) { mask = ARMV6_PMCR_CCOUNT_IEN; @@ -613,12 +616,13 @@ armv6pmu_disable_event(struct hw_perf_event *hwc, raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void -armv6mpcore_pmu_disable_event(struct hw_perf_event *hwc, - int idx) +static void armv6mpcore_pmu_disable_event(struct perf_event *event) { unsigned long val, mask, flags, evt = 0; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; if (ARMV6_CYCLE_COUNTER == idx) { mask = ARMV6_PMCR_CCOUNT_IEN; diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c index b189403f30e4..1183c81087b2 100644 --- a/arch/arm/kernel/perf_event_v7.c +++ b/arch/arm/kernel/perf_event_v7.c @@ -840,8 +840,10 @@ static inline int armv7_pmnc_select_counter(int idx) return idx; } -static inline u32 armv7pmu_read_counter(int idx) +static inline u32 armv7pmu_read_counter(struct perf_event *event) { + struct hw_perf_event *hwc = &event->hw; + int idx = hwc->idx; u32 value = 0; if (!armv7_pmnc_counter_valid(idx)) @@ -855,8 +857,11 @@ static inline u32 armv7pmu_read_counter(int idx) return value; } -static inline void armv7pmu_write_counter(int idx, u32 value) +static inline void armv7pmu_write_counter(struct perf_event *event, u32 value) { + struct hw_perf_event *hwc = &event->hw; + int idx = hwc->idx; + if (!armv7_pmnc_counter_valid(idx)) pr_err("CPU%u writing wrong counter %d\n", smp_processor_id(), idx); @@ -991,10 +996,13 @@ static void armv7_pmnc_dump_regs(void) } #endif -static void armv7pmu_enable_event(struct hw_perf_event *hwc, int idx) +static void armv7pmu_enable_event(struct perf_event *event) { unsigned long flags; + struct hw_perf_event *hwc = &event->hw; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; /* * Enable counter and interrupt, and set the counter to count @@ -1028,10 +1036,13 @@ static void armv7pmu_enable_event(struct hw_perf_event *hwc, int idx) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void armv7pmu_disable_event(struct hw_perf_event *hwc, int idx) +static void armv7pmu_disable_event(struct perf_event *event) { unsigned long flags; + struct hw_perf_event *hwc = &event->hw; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; /* * Disable counter and interrupt @@ -1055,7 +1066,8 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev) { u32 pmnc; struct perf_sample_data data; - struct pmu_hw_events *cpuc; + struct arm_pmu *cpu_pmu = (struct arm_pmu *)dev; + struct pmu_hw_events *cpuc = cpu_pmu->get_hw_events(); struct pt_regs *regs; int idx; @@ -1075,7 +1087,6 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev) */ regs = get_irq_regs(); - cpuc = &__get_cpu_var(cpu_hw_events); for (idx = 0; idx < cpu_pmu->num_events; ++idx) { struct perf_event *event = cpuc->events[idx]; struct hw_perf_event *hwc; @@ -1092,13 +1103,13 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev) continue; hwc = &event->hw; - armpmu_event_update(event, hwc, idx); + armpmu_event_update(event); perf_sample_data_init(&data, 0, hwc->last_period); - if (!armpmu_event_set_period(event, hwc, idx)) + if (!armpmu_event_set_period(event)) continue; if (perf_event_overflow(event, &data, regs)) - cpu_pmu->disable(hwc, idx); + cpu_pmu->disable(event); } /* @@ -1113,7 +1124,7 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev) return IRQ_HANDLED; } -static void armv7pmu_start(void) +static void armv7pmu_start(struct arm_pmu *cpu_pmu) { unsigned long flags; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -1124,7 +1135,7 @@ static void armv7pmu_start(void) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void armv7pmu_stop(void) +static void armv7pmu_stop(struct arm_pmu *cpu_pmu) { unsigned long flags; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -1136,10 +1147,12 @@ static void armv7pmu_stop(void) } static int armv7pmu_get_event_idx(struct pmu_hw_events *cpuc, - struct hw_perf_event *event) + struct perf_event *event) { int idx; - unsigned long evtype = event->config_base & ARMV7_EVTYPE_EVENT; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; + unsigned long evtype = hwc->config_base & ARMV7_EVTYPE_EVENT; /* Always place a cycle counter into the cycle counter. */ if (evtype == ARMV7_PERFCTR_CPU_CYCLES) { @@ -1190,11 +1203,14 @@ static int armv7pmu_set_event_filter(struct hw_perf_event *event, static void armv7pmu_reset(void *info) { + struct arm_pmu *cpu_pmu = (struct arm_pmu *)info; u32 idx, nb_cnt = cpu_pmu->num_events; /* The counter and interrupt enable registers are unknown at reset. */ - for (idx = ARMV7_IDX_CYCLE_COUNTER; idx < nb_cnt; ++idx) - armv7pmu_disable_event(NULL, idx); + for (idx = ARMV7_IDX_CYCLE_COUNTER; idx < nb_cnt; ++idx) { + armv7_pmnc_disable_counter(idx); + armv7_pmnc_disable_intens(idx); + } /* Initialize & Reset PMNC: C and P bits */ armv7_pmnc_write(ARMV7_PMNC_P | ARMV7_PMNC_C); diff --git a/arch/arm/kernel/perf_event_xscale.c b/arch/arm/kernel/perf_event_xscale.c index 131ede6c2fdf..0c8265e53d5f 100644 --- a/arch/arm/kernel/perf_event_xscale.c +++ b/arch/arm/kernel/perf_event_xscale.c @@ -224,7 +224,8 @@ xscale1pmu_handle_irq(int irq_num, void *dev) { unsigned long pmnc; struct perf_sample_data data; - struct pmu_hw_events *cpuc; + struct arm_pmu *cpu_pmu = (struct arm_pmu *)dev; + struct pmu_hw_events *cpuc = cpu_pmu->get_hw_events(); struct pt_regs *regs; int idx; @@ -248,7 +249,6 @@ xscale1pmu_handle_irq(int irq_num, void *dev) regs = get_irq_regs(); - cpuc = &__get_cpu_var(cpu_hw_events); for (idx = 0; idx < cpu_pmu->num_events; ++idx) { struct perf_event *event = cpuc->events[idx]; struct hw_perf_event *hwc; @@ -260,13 +260,13 @@ xscale1pmu_handle_irq(int irq_num, void *dev) continue; hwc = &event->hw; - armpmu_event_update(event, hwc, idx); + armpmu_event_update(event); perf_sample_data_init(&data, 0, hwc->last_period); - if (!armpmu_event_set_period(event, hwc, idx)) + if (!armpmu_event_set_period(event)) continue; if (perf_event_overflow(event, &data, regs)) - cpu_pmu->disable(hwc, idx); + cpu_pmu->disable(event); } irq_work_run(); @@ -280,11 +280,13 @@ xscale1pmu_handle_irq(int irq_num, void *dev) return IRQ_HANDLED; } -static void -xscale1pmu_enable_event(struct hw_perf_event *hwc, int idx) +static void xscale1pmu_enable_event(struct perf_event *event) { unsigned long val, mask, evt, flags; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; switch (idx) { case XSCALE_CYCLE_COUNTER: @@ -314,11 +316,13 @@ xscale1pmu_enable_event(struct hw_perf_event *hwc, int idx) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void -xscale1pmu_disable_event(struct hw_perf_event *hwc, int idx) +static void xscale1pmu_disable_event(struct perf_event *event) { unsigned long val, mask, evt, flags; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; switch (idx) { case XSCALE_CYCLE_COUNTER: @@ -348,9 +352,10 @@ xscale1pmu_disable_event(struct hw_perf_event *hwc, int idx) static int xscale1pmu_get_event_idx(struct pmu_hw_events *cpuc, - struct hw_perf_event *event) + struct perf_event *event) { - if (XSCALE_PERFCTR_CCNT == event->config_base) { + struct hw_perf_event *hwc = &event->hw; + if (XSCALE_PERFCTR_CCNT == hwc->config_base) { if (test_and_set_bit(XSCALE_CYCLE_COUNTER, cpuc->used_mask)) return -EAGAIN; @@ -366,8 +371,7 @@ xscale1pmu_get_event_idx(struct pmu_hw_events *cpuc, } } -static void -xscale1pmu_start(void) +static void xscale1pmu_start(struct arm_pmu *cpu_pmu) { unsigned long flags, val; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -379,8 +383,7 @@ xscale1pmu_start(void) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void -xscale1pmu_stop(void) +static void xscale1pmu_stop(struct arm_pmu *cpu_pmu) { unsigned long flags, val; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -392,9 +395,10 @@ xscale1pmu_stop(void) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static inline u32 -xscale1pmu_read_counter(int counter) +static inline u32 xscale1pmu_read_counter(struct perf_event *event) { + struct hw_perf_event *hwc = &event->hw; + int counter = hwc->idx; u32 val = 0; switch (counter) { @@ -412,9 +416,11 @@ xscale1pmu_read_counter(int counter) return val; } -static inline void -xscale1pmu_write_counter(int counter, u32 val) +static inline void xscale1pmu_write_counter(struct perf_event *event, u32 val) { + struct hw_perf_event *hwc = &event->hw; + int counter = hwc->idx; + switch (counter) { case XSCALE_CYCLE_COUNTER: asm volatile("mcr p14, 0, %0, c1, c0, 0" : : "r" (val)); @@ -565,7 +571,8 @@ xscale2pmu_handle_irq(int irq_num, void *dev) { unsigned long pmnc, of_flags; struct perf_sample_data data; - struct pmu_hw_events *cpuc; + struct arm_pmu *cpu_pmu = (struct arm_pmu *)dev; + struct pmu_hw_events *cpuc = cpu_pmu->get_hw_events(); struct pt_regs *regs; int idx; @@ -583,7 +590,6 @@ xscale2pmu_handle_irq(int irq_num, void *dev) regs = get_irq_regs(); - cpuc = &__get_cpu_var(cpu_hw_events); for (idx = 0; idx < cpu_pmu->num_events; ++idx) { struct perf_event *event = cpuc->events[idx]; struct hw_perf_event *hwc; @@ -595,13 +601,13 @@ xscale2pmu_handle_irq(int irq_num, void *dev) continue; hwc = &event->hw; - armpmu_event_update(event, hwc, idx); + armpmu_event_update(event); perf_sample_data_init(&data, 0, hwc->last_period); - if (!armpmu_event_set_period(event, hwc, idx)) + if (!armpmu_event_set_period(event)) continue; if (perf_event_overflow(event, &data, regs)) - cpu_pmu->disable(hwc, idx); + cpu_pmu->disable(event); } irq_work_run(); @@ -615,11 +621,13 @@ xscale2pmu_handle_irq(int irq_num, void *dev) return IRQ_HANDLED; } -static void -xscale2pmu_enable_event(struct hw_perf_event *hwc, int idx) +static void xscale2pmu_enable_event(struct perf_event *event) { unsigned long flags, ien, evtsel; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; ien = xscale2pmu_read_int_enable(); evtsel = xscale2pmu_read_event_select(); @@ -659,11 +667,13 @@ xscale2pmu_enable_event(struct hw_perf_event *hwc, int idx) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void -xscale2pmu_disable_event(struct hw_perf_event *hwc, int idx) +static void xscale2pmu_disable_event(struct perf_event *event) { unsigned long flags, ien, evtsel, of_flags; + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); + struct hw_perf_event *hwc = &event->hw; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); + int idx = hwc->idx; ien = xscale2pmu_read_int_enable(); evtsel = xscale2pmu_read_event_select(); @@ -711,7 +721,7 @@ xscale2pmu_disable_event(struct hw_perf_event *hwc, int idx) static int xscale2pmu_get_event_idx(struct pmu_hw_events *cpuc, - struct hw_perf_event *event) + struct perf_event *event) { int idx = xscale1pmu_get_event_idx(cpuc, event); if (idx >= 0) @@ -725,8 +735,7 @@ out: return idx; } -static void -xscale2pmu_start(void) +static void xscale2pmu_start(struct arm_pmu *cpu_pmu) { unsigned long flags, val; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -738,8 +747,7 @@ xscale2pmu_start(void) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static void -xscale2pmu_stop(void) +static void xscale2pmu_stop(struct arm_pmu *cpu_pmu) { unsigned long flags, val; struct pmu_hw_events *events = cpu_pmu->get_hw_events(); @@ -751,9 +759,10 @@ xscale2pmu_stop(void) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static inline u32 -xscale2pmu_read_counter(int counter) +static inline u32 xscale2pmu_read_counter(struct perf_event *event) { + struct hw_perf_event *hwc = &event->hw; + int counter = hwc->idx; u32 val = 0; switch (counter) { @@ -777,9 +786,11 @@ xscale2pmu_read_counter(int counter) return val; } -static inline void -xscale2pmu_write_counter(int counter, u32 val) +static inline void xscale2pmu_write_counter(struct perf_event *event, u32 val) { + struct hw_perf_event *hwc = &event->hw; + int counter = hwc->idx; + switch (counter) { case XSCALE_CYCLE_COUNTER: asm volatile("mcr p14, 0, %0, c1, c1, 0" : : "r" (val)); -- cgit From 7279adbd9bb8ef8ff669da50f0e84c65a14022b5 Mon Sep 17 00:00:00 2001 From: Sudeep KarkadaNagesha Date: Fri, 20 Jul 2012 15:18:07 +0100 Subject: ARM: perf: check ARMv7 counter validity on a per-pmu basis Multi-cluster ARMv7 systems may have CPU PMUs with different number of counters. This patch updates armv7_pmnc_counter_valid so that it takes a pmu argument and checks the counter validity against that. We also remove a number of redundant counter checks whether the current PMU is not easily retrievable. Signed-off-by: Sudeep KarkadaNagesha Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event_v7.c | 94 +++++++++++++---------------------------- 1 file changed, 30 insertions(+), 64 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c index 1183c81087b2..7d0cce85d17e 100644 --- a/arch/arm/kernel/perf_event_v7.c +++ b/arch/arm/kernel/perf_event_v7.c @@ -736,7 +736,8 @@ static const unsigned armv7_a7_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] */ #define ARMV7_IDX_CYCLE_COUNTER 0 #define ARMV7_IDX_COUNTER0 1 -#define ARMV7_IDX_COUNTER_LAST (ARMV7_IDX_CYCLE_COUNTER + cpu_pmu->num_events - 1) +#define ARMV7_IDX_COUNTER_LAST(cpu_pmu) \ + (ARMV7_IDX_CYCLE_COUNTER + cpu_pmu->num_events - 1) #define ARMV7_MAX_COUNTERS 32 #define ARMV7_COUNTER_MASK (ARMV7_MAX_COUNTERS - 1) @@ -802,38 +803,20 @@ static inline int armv7_pmnc_has_overflowed(u32 pmnc) return pmnc & ARMV7_OVERFLOWED_MASK; } -static inline int armv7_pmnc_counter_valid(int idx) +static inline int armv7_pmnc_counter_valid(struct arm_pmu *cpu_pmu, int idx) { - return idx >= ARMV7_IDX_CYCLE_COUNTER && idx <= ARMV7_IDX_COUNTER_LAST; + return idx >= ARMV7_IDX_CYCLE_COUNTER && + idx <= ARMV7_IDX_COUNTER_LAST(cpu_pmu); } static inline int armv7_pmnc_counter_has_overflowed(u32 pmnc, int idx) { - int ret = 0; - u32 counter; - - if (!armv7_pmnc_counter_valid(idx)) { - pr_err("CPU%u checking wrong counter %d overflow status\n", - smp_processor_id(), idx); - } else { - counter = ARMV7_IDX_TO_COUNTER(idx); - ret = pmnc & BIT(counter); - } - - return ret; + return pmnc & BIT(ARMV7_IDX_TO_COUNTER(idx)); } static inline int armv7_pmnc_select_counter(int idx) { - u32 counter; - - if (!armv7_pmnc_counter_valid(idx)) { - pr_err("CPU%u selecting wrong PMNC counter %d\n", - smp_processor_id(), idx); - return -EINVAL; - } - - counter = ARMV7_IDX_TO_COUNTER(idx); + u32 counter = ARMV7_IDX_TO_COUNTER(idx); asm volatile("mcr p15, 0, %0, c9, c12, 5" : : "r" (counter)); isb(); @@ -842,11 +825,12 @@ static inline int armv7_pmnc_select_counter(int idx) static inline u32 armv7pmu_read_counter(struct perf_event *event) { + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; int idx = hwc->idx; u32 value = 0; - if (!armv7_pmnc_counter_valid(idx)) + if (!armv7_pmnc_counter_valid(cpu_pmu, idx)) pr_err("CPU%u reading wrong counter %d\n", smp_processor_id(), idx); else if (idx == ARMV7_IDX_CYCLE_COUNTER) @@ -859,10 +843,11 @@ static inline u32 armv7pmu_read_counter(struct perf_event *event) static inline void armv7pmu_write_counter(struct perf_event *event, u32 value) { + struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; int idx = hwc->idx; - if (!armv7_pmnc_counter_valid(idx)) + if (!armv7_pmnc_counter_valid(cpu_pmu, idx)) pr_err("CPU%u writing wrong counter %d\n", smp_processor_id(), idx); else if (idx == ARMV7_IDX_CYCLE_COUNTER) @@ -881,60 +866,28 @@ static inline void armv7_pmnc_write_evtsel(int idx, u32 val) static inline int armv7_pmnc_enable_counter(int idx) { - u32 counter; - - if (!armv7_pmnc_counter_valid(idx)) { - pr_err("CPU%u enabling wrong PMNC counter %d\n", - smp_processor_id(), idx); - return -EINVAL; - } - - counter = ARMV7_IDX_TO_COUNTER(idx); + u32 counter = ARMV7_IDX_TO_COUNTER(idx); asm volatile("mcr p15, 0, %0, c9, c12, 1" : : "r" (BIT(counter))); return idx; } static inline int armv7_pmnc_disable_counter(int idx) { - u32 counter; - - if (!armv7_pmnc_counter_valid(idx)) { - pr_err("CPU%u disabling wrong PMNC counter %d\n", - smp_processor_id(), idx); - return -EINVAL; - } - - counter = ARMV7_IDX_TO_COUNTER(idx); + u32 counter = ARMV7_IDX_TO_COUNTER(idx); asm volatile("mcr p15, 0, %0, c9, c12, 2" : : "r" (BIT(counter))); return idx; } static inline int armv7_pmnc_enable_intens(int idx) { - u32 counter; - - if (!armv7_pmnc_counter_valid(idx)) { - pr_err("CPU%u enabling wrong PMNC counter IRQ enable %d\n", - smp_processor_id(), idx); - return -EINVAL; - } - - counter = ARMV7_IDX_TO_COUNTER(idx); + u32 counter = ARMV7_IDX_TO_COUNTER(idx); asm volatile("mcr p15, 0, %0, c9, c14, 1" : : "r" (BIT(counter))); return idx; } static inline int armv7_pmnc_disable_intens(int idx) { - u32 counter; - - if (!armv7_pmnc_counter_valid(idx)) { - pr_err("CPU%u disabling wrong PMNC counter IRQ enable %d\n", - smp_processor_id(), idx); - return -EINVAL; - } - - counter = ARMV7_IDX_TO_COUNTER(idx); + u32 counter = ARMV7_IDX_TO_COUNTER(idx); asm volatile("mcr p15, 0, %0, c9, c14, 2" : : "r" (BIT(counter))); isb(); /* Clear the overflow flag in case an interrupt is pending. */ @@ -959,7 +912,7 @@ static inline u32 armv7_pmnc_getreset_flags(void) } #ifdef DEBUG -static void armv7_pmnc_dump_regs(void) +static void armv7_pmnc_dump_regs(struct arm_pmu *cpu_pmu) { u32 val; unsigned int cnt; @@ -984,7 +937,8 @@ static void armv7_pmnc_dump_regs(void) asm volatile("mrc p15, 0, %0, c9, c13, 0" : "=r" (val)); printk(KERN_INFO "CCNT =0x%08x\n", val); - for (cnt = ARMV7_IDX_COUNTER0; cnt <= ARMV7_IDX_COUNTER_LAST; cnt++) { + for (cnt = ARMV7_IDX_COUNTER0; + cnt <= ARMV7_IDX_COUNTER_LAST(cpu_pmu); cnt++) { armv7_pmnc_select_counter(cnt); asm volatile("mrc p15, 0, %0, c9, c13, 2" : "=r" (val)); printk(KERN_INFO "CNT[%d] count =0x%08x\n", @@ -1004,6 +958,12 @@ static void armv7pmu_enable_event(struct perf_event *event) struct pmu_hw_events *events = cpu_pmu->get_hw_events(); int idx = hwc->idx; + if (!armv7_pmnc_counter_valid(cpu_pmu, idx)) { + pr_err("CPU%u enabling wrong PMNC counter IRQ enable %d\n", + smp_processor_id(), idx); + return; + } + /* * Enable counter and interrupt, and set the counter to count * the event that we're interested in. @@ -1044,6 +1004,12 @@ static void armv7pmu_disable_event(struct perf_event *event) struct pmu_hw_events *events = cpu_pmu->get_hw_events(); int idx = hwc->idx; + if (!armv7_pmnc_counter_valid(cpu_pmu, idx)) { + pr_err("CPU%u disabling wrong PMNC counter IRQ enable %d\n", + smp_processor_id(), idx); + return; + } + /* * Disable counter and interrupt */ -- cgit From 2a4961ba89ffae388a553175db51dd58ce7c39a8 Mon Sep 17 00:00:00 2001 From: Mark Rutland Date: Fri, 21 Sep 2012 11:53:41 +0100 Subject: ARM: perf: register cpu_notifier at driver init The current practice of registering the cpu hotplug notifier at PMU registration time won't be safe with multiple PMUs, as we'll repeatedly attempt to register the notifier. This has the unfortunate effect of silently corrupting the notifier list, leading to boot stalling. Instead, register the notifier at init time. Its sanity checks will prevent anything bad from happening if the notifier is called before we have any PMUs registered. Signed-off-by: Mark Rutland Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event_cpu.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event_cpu.c b/arch/arm/kernel/perf_event_cpu.c index 02244faa5396..71c824ce0203 100644 --- a/arch/arm/kernel/perf_event_cpu.c +++ b/arch/arm/kernel/perf_event_cpu.c @@ -285,7 +285,6 @@ static int __devinit cpu_pmu_device_probe(struct platform_device *pdev) cpu_pmu = pmu; cpu_pmu->plat_device = pdev; cpu_pmu_init(cpu_pmu); - register_cpu_notifier(&cpu_pmu_hotplug_notifier); armpmu_register(cpu_pmu, cpu_pmu->name, PERF_TYPE_RAW); return 0; @@ -303,6 +302,16 @@ static struct platform_driver cpu_pmu_driver = { static int __init register_pmu_driver(void) { - return platform_driver_register(&cpu_pmu_driver); + int err; + + err = register_cpu_notifier(&cpu_pmu_hotplug_notifier); + if (err) + return err; + + err = platform_driver_register(&cpu_pmu_driver); + if (err) + unregister_cpu_notifier(&cpu_pmu_hotplug_notifier); + + return err; } device_initcall(register_pmu_driver); -- cgit From 288700d16d4c1479aa00e3db13caed8bb7c83e9f Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 14:14:17 +0100 Subject: ARM: perf: return NOTIFY_DONE from cpu notifier when no available PMU When attempting to reset the PMU state for either a NULL PMU or a PMU implementation without a reset function, return NOTIFY_DONE from the CPU notifier as we don't care about the hotplug event. Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event_cpu.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event_cpu.c b/arch/arm/kernel/perf_event_cpu.c index 71c824ce0203..db9c6b530f3c 100644 --- a/arch/arm/kernel/perf_event_cpu.c +++ b/arch/arm/kernel/perf_event_cpu.c @@ -165,6 +165,8 @@ static int __cpuinit cpu_pmu_notify(struct notifier_block *b, if (cpu_pmu && cpu_pmu->reset) cpu_pmu->reset(cpu_pmu); + else + return NOTIFY_DONE; return NOTIFY_OK; } -- cgit From 0305230a3d92d6829db89c9e0c096d4d8733f317 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 14:23:47 +0100 Subject: ARM: perf: consistently use arm_pmu->name for PMU name Perf has three ways to name a PMU: either by passing an explicit char *, reading arm_pmu->name or accessing arm_pmu->pmu.name. Just use arm_pmu->name consistently in the ARM backend. Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event.c | 4 ++-- arch/arm/kernel/perf_event_cpu.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index 1cfa3f35713e..1243deda5bbb 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c @@ -511,12 +511,12 @@ static void __init armpmu_init(struct arm_pmu *armpmu) }; } -int armpmu_register(struct arm_pmu *armpmu, char *name, int type) +int armpmu_register(struct arm_pmu *armpmu, int type) { armpmu_init(armpmu); pr_info("enabled with %s PMU driver, %d counters available\n", armpmu->name, armpmu->num_events); - return perf_pmu_register(&armpmu->pmu, name, type); + return perf_pmu_register(&armpmu->pmu, armpmu->name, type); } /* diff --git a/arch/arm/kernel/perf_event_cpu.c b/arch/arm/kernel/perf_event_cpu.c index db9c6b530f3c..9a4f6307a016 100644 --- a/arch/arm/kernel/perf_event_cpu.c +++ b/arch/arm/kernel/perf_event_cpu.c @@ -46,7 +46,7 @@ const char *perf_pmu_name(void) if (!cpu_pmu) return NULL; - return cpu_pmu->pmu.name; + return cpu_pmu->name; } EXPORT_SYMBOL_GPL(perf_pmu_name); @@ -287,7 +287,7 @@ static int __devinit cpu_pmu_device_probe(struct platform_device *pdev) cpu_pmu = pmu; cpu_pmu->plat_device = pdev; cpu_pmu_init(cpu_pmu); - armpmu_register(cpu_pmu, cpu_pmu->name, PERF_TYPE_RAW); + armpmu_register(cpu_pmu, PERF_TYPE_RAW); return 0; } -- cgit From 2ac29a14a8b6b4a37c09c50db88dc893e6e7fc75 Mon Sep 17 00:00:00 2001 From: Jon Hunter Date: Thu, 25 Oct 2012 21:23:18 +0100 Subject: ARM: PMU: fix runtime PM enable Commit 7be2958 (ARM: PMU: Add runtime PM Support) updated the ARM PMU code to use runtime PM which was prototyped and validated on the OMAP devices. In this commit, there is no call pm_runtime_enable() and for OMAP devices pm_runtime_enable() is currently being called from the OMAP PMU code when the PMU device is created. However, there are two problems with this: 1. For any other ARM device wishing to use runtime PM for PMU they will need to call pm_runtime_enable() for runtime PM to work. 2. When booting with device-tree and using device-tree to create the PMU device, pm_runtime_enable() needs to be called from within the ARM PERF driver as we are no longer calling any device specific code to create the device. Hence, PMU does not work on OMAP devices that use the runtime PM callbacks when using device-tree to create the PMU device. Therefore, call pm_runtime_enable() directly from the ARM PMU driver when registering the device. For platforms that do not use runtime PM, pm_runtime_enable() does nothing and for platforms that do use runtime PM but may not require it specifically for PMU, this will just add a little overhead when initialising and uninitialising the PMU device. Tested with PERF on OMAP2420, OMAP3430 and OMAP4460. Acked-by: Kevin Hilman Acked-by: Tony Lindgren Signed-off-by: Jon Hunter Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index 1243deda5bbb..f9e8657dd241 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c @@ -514,6 +514,7 @@ static void __init armpmu_init(struct arm_pmu *armpmu) int armpmu_register(struct arm_pmu *armpmu, int type) { armpmu_init(armpmu); + pm_runtime_enable(&armpmu->plat_device->dev); pr_info("enabled with %s PMU driver, %d counters available\n", armpmu->name, armpmu->num_events); return perf_pmu_register(&armpmu->pmu, armpmu->name, type); -- cgit From e64877dcf5fd05d81fa195785a738f3a729587a3 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 14:53:13 +0100 Subject: ARM: hw_breakpoint: only clear OS lock when implemented on v7 The OS save and restore register are optional in debug architecture v7, so check the status register before attempting to clear the OS lock. Tested-by: Stephen Boyd Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index 281bf3301241..76a650a1a1d7 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -906,7 +906,7 @@ static struct undef_hook debug_reg_hook = { static void reset_ctrl_regs(void *unused) { int i, raw_num_brps, err = 0, cpu = smp_processor_id(); - u32 dbg_power; + u32 val; /* * v7 debug contains save and restore registers so that debug state @@ -926,16 +926,23 @@ static void reset_ctrl_regs(void *unused) * Ensure sticky power-down is clear (i.e. debug logic is * powered up). */ - asm volatile("mrc p14, 0, %0, c1, c5, 4" : "=r" (dbg_power)); - if ((dbg_power & 0x1) == 0) + asm volatile("mrc p14, 0, %0, c1, c5, 4" : "=r" (val)); + if ((val & 0x1) == 0) err = -EPERM; + + /* + * Check whether we implement OS save and restore. + */ + asm volatile("mrc p14, 0, %0, c1, c1, 4" : "=r" (val)); + if ((val & 0x9) == 0) + goto clear_vcr; break; case ARM_DEBUG_ARCH_V7_1: /* * Ensure the OS double lock is clear. */ - asm volatile("mrc p14, 0, %0, c1, c3, 4" : "=r" (dbg_power)); - if ((dbg_power & 0x1) == 1) + asm volatile("mrc p14, 0, %0, c1, c3, 4" : "=r" (val)); + if ((val & 0x1) == 1) err = -EPERM; break; } @@ -947,7 +954,7 @@ static void reset_ctrl_regs(void *unused) } /* - * Unconditionally clear the lock by writing a value + * Unconditionally clear the OS lock by writing a value * other than 0xC5ACCE55 to the access register. */ asm volatile("mcr p14, 0, %0, c1, c0, 4" : : "r" (0)); @@ -957,6 +964,7 @@ static void reset_ctrl_regs(void *unused) * Clear any configured vector-catch events before * enabling monitor mode. */ +clear_vcr: asm volatile("mcr p14, 0, %0, c0, c7, 0" : : "r" (0)); isb(); -- cgit From b59a540ca927ea84bb0590b9d8076f50c969abb4 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 15:08:17 +0100 Subject: ARM: hw_breakpoint: fix monitor mode detection with v7.1 Detecting whether halting debug is enabled is no longer possible via the DBGDSCR in v7.1, returning an UNKNOWN value for the HDBGen bit via CP14 when the OS lock is clear. This patch removes the halting mode check and ensures that accesses to the internal and external views of the DBGDSCR are serialised with an instruction barrier. Tested-by: Stephen Boyd Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 25 +++++-------------------- 1 file changed, 5 insertions(+), 20 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index 76a650a1a1d7..c87ea68d64ae 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -235,13 +235,6 @@ static int enable_monitor_mode(void) ARM_DBG_READ(c1, 0, dscr); - /* Ensure that halting mode is disabled. */ - if (WARN_ONCE(dscr & ARM_DSCR_HDBGEN, - "halting debug mode enabled. Unable to access hardware resources.\n")) { - ret = -EPERM; - goto out; - } - /* If monitor mode is already enabled, just return. */ if (dscr & ARM_DSCR_MDBGEN) goto out; @@ -255,6 +248,7 @@ static int enable_monitor_mode(void) case ARM_DEBUG_ARCH_V7_ECP14: case ARM_DEBUG_ARCH_V7_1: ARM_DBG_WRITE(c2, 2, (dscr | ARM_DSCR_MDBGEN)); + isb(); break; default: ret = -ENODEV; @@ -1000,8 +994,6 @@ static struct notifier_block __cpuinitdata dbg_reset_nb = { static int __init arch_hw_breakpoint_init(void) { - u32 dscr; - debug_arch = get_debug_arch(); if (!debug_arch_supported()) { @@ -1036,17 +1028,10 @@ static int __init arch_hw_breakpoint_init(void) core_num_brps, core_has_mismatch_brps() ? "(+1 reserved) " : "", core_num_wrps); - ARM_DBG_READ(c1, 0, dscr); - if (dscr & ARM_DSCR_HDBGEN) { - max_watchpoint_len = 4; - pr_warning("halting debug mode enabled. Assuming maximum watchpoint size of %u bytes.\n", - max_watchpoint_len); - } else { - /* Work out the maximum supported watchpoint length. */ - max_watchpoint_len = get_max_wp_len(); - pr_info("maximum watchpoint size is %u bytes.\n", - max_watchpoint_len); - } + /* Work out the maximum supported watchpoint length. */ + max_watchpoint_len = get_max_wp_len(); + pr_info("maximum watchpoint size is %u bytes.\n", + max_watchpoint_len); /* Register debug fault handler. */ hook_fault_code(FAULT_CODE_DEBUG, hw_breakpoint_pending, SIGTRAP, -- cgit From 614bea500a88be2a841af0967469961470f2be83 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 15:38:26 +0100 Subject: ARM: hw_breakpoint: fix ordering of debug register reset sequence The debug register reset sequence for v7 and v7.1 is congruent with tap-dancing through a minefield. Rather than wait until we've blown ourselves to pieces, this patch instead checks the debug_err_mask after each potentially faulting operation. We also move the enabling of monitor_mode to the end of the sequence in order to prevent spurious debug events generated by UNKNOWN register values. Reported-by: Stephen Boyd Tested-by: Stephen Boyd Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 36 ++++++++++++++++++++++++++---------- 1 file changed, 26 insertions(+), 10 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index c87ea68d64ae..ae6bf80e3a5e 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -231,8 +231,6 @@ static int get_num_brps(void) static int enable_monitor_mode(void) { u32 dscr; - int ret = 0; - ARM_DBG_READ(c1, 0, dscr); /* If monitor mode is already enabled, just return. */ @@ -251,17 +249,18 @@ static int enable_monitor_mode(void) isb(); break; default: - ret = -ENODEV; - goto out; + return -ENODEV; } /* Check that the write made it through. */ ARM_DBG_READ(c1, 0, dscr); - if (!(dscr & ARM_DSCR_MDBGEN)) - ret = -EPERM; + if (WARN_ONCE(!(dscr & ARM_DSCR_MDBGEN), + "Failed to enable monitor mode on CPU %d.\n", + smp_processor_id())) + return -EPERM; out: - return ret; + return 0; } int hw_breakpoint_slots(int type) @@ -962,11 +961,16 @@ clear_vcr: asm volatile("mcr p14, 0, %0, c0, c7, 0" : : "r" (0)); isb(); -reset_regs: - if (enable_monitor_mode()) + if (cpumask_intersects(&debug_err_mask, cpumask_of(cpu))) { + pr_warning("CPU %d failed to disable vector catch\n", cpu); return; + } - /* We must also reset any reserved registers. */ +reset_regs: + /* + * The control/value register pairs are UNKNOWN out of reset so + * clear them to avoid spurious debug events. + */ raw_num_brps = get_num_brp_resources(); for (i = 0; i < raw_num_brps; ++i) { write_wb_reg(ARM_BASE_BCR + i, 0UL); @@ -977,6 +981,18 @@ reset_regs: write_wb_reg(ARM_BASE_WCR + i, 0UL); write_wb_reg(ARM_BASE_WVR + i, 0UL); } + + if (cpumask_intersects(&debug_err_mask, cpumask_of(cpu))) { + pr_warning("CPU %d failed to clear debug register pairs\n", cpu); + return; + } + + /* + * Have a crack at enabling monitor mode. We don't actually need + * it yet, but reporting an error early is useful if it fails. + */ + if (enable_monitor_mode()) + cpumask_or(&debug_err_mask, &debug_err_mask, cpumask_of(cpu)); } static int __cpuinit dbg_reset_notify(struct notifier_block *self, -- cgit From 7f4050a07be8ce5fad069722326ccd550577a93a Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 17:53:08 +0100 Subject: ARM: hw_breakpoint: don't try to clear v6 debug registers during boot v6 cores do not provide a way to clear the debug registers without first enabling monitor mode, meaning that we could take spurious debug exceptions. Instead, rely on the registers being in a sane state when we boot as they are defined to be disabled out of reset anyway. Tested-by: Stephen Boyd Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index ae6bf80e3a5e..8cd52faeb77e 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -912,8 +912,8 @@ static void reset_ctrl_regs(void *unused) switch (debug_arch) { case ARM_DEBUG_ARCH_V6: case ARM_DEBUG_ARCH_V6_1: - /* ARMv6 cores just need to reset the registers. */ - goto reset_regs; + /* ARMv6 cores clear the registers out of reset. */ + goto out_mdbgen; case ARM_DEBUG_ARCH_V7_ECP14: /* * Ensure sticky power-down is clear (i.e. debug logic is @@ -966,7 +966,6 @@ clear_vcr: return; } -reset_regs: /* * The control/value register pairs are UNKNOWN out of reset so * clear them to avoid spurious debug events. @@ -991,6 +990,7 @@ reset_regs: * Have a crack at enabling monitor mode. We don't actually need * it yet, but reporting an error early is useful if it fails. */ +out_mdbgen: if (enable_monitor_mode()) cpumask_or(&debug_err_mask, &debug_err_mask, cpumask_of(cpu)); } -- cgit From 5ad29ea24e58777aa1daaa2255670ffb40aefd99 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Fri, 21 Sep 2012 18:17:24 +0100 Subject: ARM: hw_breakpoint: make boot quieter without CPUID feature registers Booting on a v6 core without the CPUID feature registers (e.g. 1136) leads to a noisy dmesg complaining about their absence. This patch changes the pr_warning into a pr_warn_once to keep the log quieter. Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index 8cd52faeb77e..983558eb9023 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -136,8 +136,8 @@ static u8 get_debug_arch(void) /* Do we implement the extended CPUID interface? */ if (((read_cpuid_id() >> 16) & 0xf) != 0xf) { - pr_warning("CPUID feature registers not supported. " - "Assuming v6 debug is present.\n"); + pr_warn_once("CPUID feature registers not supported. " + "Assuming v6 debug is present.\n"); return ARM_DEBUG_ARCH_V6; } -- cgit From 0daa034e696ac601061cbf60fda41ad39678ae14 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Mon, 24 Sep 2012 18:01:13 +0100 Subject: ARM: hw_breakpoint: check if monitor mode is enabled during validation Rather than attempt to enable monitor mode explicitly when scheduling in a breakpoint event (which could raise an undefined exception trap when accessing DBGDSCRext), instead check that DBGDSCRint.MDBGen is set during event validation and report an error to the caller if not. Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 28 +++++++++++++++------------- 1 file changed, 15 insertions(+), 13 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index 983558eb9023..8e9532152bd9 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -228,6 +228,13 @@ static int get_num_brps(void) * be put into halting debug mode at any time by an external debugger * but there is nothing we can do to prevent that. */ +static int monitor_mode_enabled(void) +{ + u32 dscr; + ARM_DBG_READ(c1, 0, dscr); + return !!(dscr & ARM_DSCR_MDBGEN); +} + static int enable_monitor_mode(void) { u32 dscr; @@ -321,14 +328,9 @@ int arch_install_hw_breakpoint(struct perf_event *bp) { struct arch_hw_breakpoint *info = counter_arch_bp(bp); struct perf_event **slot, **slots; - int i, max_slots, ctrl_base, val_base, ret = 0; + int i, max_slots, ctrl_base, val_base; u32 addr, ctrl; - /* Ensure that we are in monitor mode and halting mode is disabled. */ - ret = enable_monitor_mode(); - if (ret) - goto out; - addr = info->address; ctrl = encode_ctrl_reg(info->ctrl) | 0x1; @@ -355,10 +357,8 @@ int arch_install_hw_breakpoint(struct perf_event *bp) } } - if (WARN_ONCE(i == max_slots, "Can't find any breakpoint slot\n")) { - ret = -EBUSY; - goto out; - } + if (WARN_ONCE(i == max_slots, "Can't find any breakpoint slot\n")) + return -EBUSY; /* Override the breakpoint data with the step data. */ if (info->step_ctrl.enabled) { @@ -376,9 +376,7 @@ int arch_install_hw_breakpoint(struct perf_event *bp) /* Setup the control register. */ write_wb_reg(ctrl_base + i, ctrl); - -out: - return ret; + return 0; } void arch_uninstall_hw_breakpoint(struct perf_event *bp) @@ -589,6 +587,10 @@ int arch_validate_hwbkpt_settings(struct perf_event *bp) int ret = 0; u32 offset, alignment_mask = 0x3; + /* Ensure that we are in monitor debug mode. */ + if (!monitor_mode_enabled()) + return -ENODEV; + /* Build the arch_hw_breakpoint. */ ret = arch_build_bp_info(bp); if (ret) -- cgit From 9e962f76602dbd293a57030f4ce5a4b57853e2ea Mon Sep 17 00:00:00 2001 From: Dietmar Eggemann Date: Wed, 26 Sep 2012 17:28:47 +0100 Subject: ARM: hw_breakpoint: use CRn as argument for debug reg accessor macros The coprocessor register CRn for accesses to the debug register can be a different one than C0. Take this into account for the ARM_DBG_READ and the ARM_DBG_WRITE macro. The inline assembler calls which used a coprocessor register CRn other than C0 are replaced by the ARM_DBG_READ or ARM_DBG_WRITE macro. Tested-by: Stephen Boyd Signed-off-by: Dietmar Eggemann Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 40 ++++++++++++++++++++-------------------- 1 file changed, 20 insertions(+), 20 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index 8e9532152bd9..05febbaecb44 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -52,14 +52,14 @@ static u8 debug_arch; /* Maximum supported watchpoint length. */ static u8 max_watchpoint_len; -#define READ_WB_REG_CASE(OP2, M, VAL) \ - case ((OP2 << 4) + M): \ - ARM_DBG_READ(c ## M, OP2, VAL); \ +#define READ_WB_REG_CASE(OP2, M, VAL) \ + case ((OP2 << 4) + M): \ + ARM_DBG_READ(c0, c ## M, OP2, VAL); \ break -#define WRITE_WB_REG_CASE(OP2, M, VAL) \ - case ((OP2 << 4) + M): \ - ARM_DBG_WRITE(c ## M, OP2, VAL);\ +#define WRITE_WB_REG_CASE(OP2, M, VAL) \ + case ((OP2 << 4) + M): \ + ARM_DBG_WRITE(c0, c ## M, OP2, VAL); \ break #define GEN_READ_WB_REG_CASES(OP2, VAL) \ @@ -141,7 +141,7 @@ static u8 get_debug_arch(void) return ARM_DEBUG_ARCH_V6; } - ARM_DBG_READ(c0, 0, didr); + ARM_DBG_READ(c0, c0, 0, didr); return (didr >> 16) & 0xf; } @@ -169,7 +169,7 @@ static int debug_exception_updates_fsr(void) static int get_num_wrp_resources(void) { u32 didr; - ARM_DBG_READ(c0, 0, didr); + ARM_DBG_READ(c0, c0, 0, didr); return ((didr >> 28) & 0xf) + 1; } @@ -177,7 +177,7 @@ static int get_num_wrp_resources(void) static int get_num_brp_resources(void) { u32 didr; - ARM_DBG_READ(c0, 0, didr); + ARM_DBG_READ(c0, c0, 0, didr); return ((didr >> 24) & 0xf) + 1; } @@ -231,14 +231,14 @@ static int get_num_brps(void) static int monitor_mode_enabled(void) { u32 dscr; - ARM_DBG_READ(c1, 0, dscr); + ARM_DBG_READ(c0, c1, 0, dscr); return !!(dscr & ARM_DSCR_MDBGEN); } static int enable_monitor_mode(void) { u32 dscr; - ARM_DBG_READ(c1, 0, dscr); + ARM_DBG_READ(c0, c1, 0, dscr); /* If monitor mode is already enabled, just return. */ if (dscr & ARM_DSCR_MDBGEN) @@ -248,11 +248,11 @@ static int enable_monitor_mode(void) switch (get_debug_arch()) { case ARM_DEBUG_ARCH_V6: case ARM_DEBUG_ARCH_V6_1: - ARM_DBG_WRITE(c1, 0, (dscr | ARM_DSCR_MDBGEN)); + ARM_DBG_WRITE(c0, c1, 0, (dscr | ARM_DSCR_MDBGEN)); break; case ARM_DEBUG_ARCH_V7_ECP14: case ARM_DEBUG_ARCH_V7_1: - ARM_DBG_WRITE(c2, 2, (dscr | ARM_DSCR_MDBGEN)); + ARM_DBG_WRITE(c0, c2, 2, (dscr | ARM_DSCR_MDBGEN)); isb(); break; default: @@ -260,7 +260,7 @@ static int enable_monitor_mode(void) } /* Check that the write made it through. */ - ARM_DBG_READ(c1, 0, dscr); + ARM_DBG_READ(c0, c1, 0, dscr); if (WARN_ONCE(!(dscr & ARM_DSCR_MDBGEN), "Failed to enable monitor mode on CPU %d.\n", smp_processor_id())) @@ -853,7 +853,7 @@ static int hw_breakpoint_pending(unsigned long addr, unsigned int fsr, local_irq_enable(); /* We only handle watchpoints and hardware breakpoints. */ - ARM_DBG_READ(c1, 0, dscr); + ARM_DBG_READ(c0, c1, 0, dscr); /* Perform perf callbacks. */ switch (ARM_DSCR_MOE(dscr)) { @@ -921,14 +921,14 @@ static void reset_ctrl_regs(void *unused) * Ensure sticky power-down is clear (i.e. debug logic is * powered up). */ - asm volatile("mrc p14, 0, %0, c1, c5, 4" : "=r" (val)); + ARM_DBG_READ(c1, c5, 4, val); if ((val & 0x1) == 0) err = -EPERM; /* * Check whether we implement OS save and restore. */ - asm volatile("mrc p14, 0, %0, c1, c1, 4" : "=r" (val)); + ARM_DBG_READ(c1, c1, 4, val); if ((val & 0x9) == 0) goto clear_vcr; break; @@ -936,7 +936,7 @@ static void reset_ctrl_regs(void *unused) /* * Ensure the OS double lock is clear. */ - asm volatile("mrc p14, 0, %0, c1, c3, 4" : "=r" (val)); + ARM_DBG_READ(c1, c3, 4, val); if ((val & 0x1) == 1) err = -EPERM; break; @@ -952,7 +952,7 @@ static void reset_ctrl_regs(void *unused) * Unconditionally clear the OS lock by writing a value * other than 0xC5ACCE55 to the access register. */ - asm volatile("mcr p14, 0, %0, c1, c0, 4" : : "r" (0)); + ARM_DBG_WRITE(c1, c0, 4, 0); isb(); /* @@ -960,7 +960,7 @@ static void reset_ctrl_regs(void *unused) * enabling monitor mode. */ clear_vcr: - asm volatile("mcr p14, 0, %0, c0, c7, 0" : : "r" (0)); + ARM_DBG_WRITE(c0, c7, 0, 0); isb(); if (cpumask_intersects(&debug_err_mask, cpumask_of(cpu))) { -- cgit From f435ab79928e4d54082e2838c4562a165e37999c Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Thu, 25 Oct 2012 17:18:23 +0100 Subject: ARM: hw_breakpoint: kill WARN_ONCE usage WARN_ONCE is a bit OTT for some of the simple failure cases encountered in hw_breakpoint, so use either pr_warning or pr_warn_once instead. Reported-by: Stephen Boyd Signed-off-by: Will Deacon --- arch/arm/kernel/hw_breakpoint.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c index 05febbaecb44..5ff2e77782b1 100644 --- a/arch/arm/kernel/hw_breakpoint.c +++ b/arch/arm/kernel/hw_breakpoint.c @@ -261,10 +261,11 @@ static int enable_monitor_mode(void) /* Check that the write made it through. */ ARM_DBG_READ(c0, c1, 0, dscr); - if (WARN_ONCE(!(dscr & ARM_DSCR_MDBGEN), - "Failed to enable monitor mode on CPU %d.\n", - smp_processor_id())) + if (!(dscr & ARM_DSCR_MDBGEN)) { + pr_warn_once("Failed to enable monitor mode on CPU %d.\n", + smp_processor_id()); return -EPERM; + } out: return 0; @@ -357,8 +358,10 @@ int arch_install_hw_breakpoint(struct perf_event *bp) } } - if (WARN_ONCE(i == max_slots, "Can't find any breakpoint slot\n")) + if (i == max_slots) { + pr_warning("Can't find any breakpoint slot\n"); return -EBUSY; + } /* Override the breakpoint data with the step data. */ if (info->step_ctrl.enabled) { @@ -407,8 +410,10 @@ void arch_uninstall_hw_breakpoint(struct perf_event *bp) } } - if (WARN_ONCE(i == max_slots, "Can't find any breakpoint slot\n")) + if (i == max_slots) { + pr_warning("Can't find any breakpoint slot\n"); return; + } /* Ensure that we disable the mismatch breakpoint. */ if (info->ctrl.type != ARM_BREAKPOINT_EXECUTE && -- cgit From e8d432c9cf0a3d569b2d00877d12c9ffe9e55286 Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Tue, 6 Nov 2012 11:57:43 +0000 Subject: ARM: kernel: add MIDR to per-CPU information data The advent of big.LITTLE ARM platforms requires the kernel to be able to identify the MIDRs of all online CPUs upon request. MIDRs are stashed at boot time so that kernel subsystems can detect the MIDR of online CPUs by simply retrieving per-CPU data updated by all booted CPUs. Signed-off-by: Lorenzo Pieralisi Acked-by: Nicolas Pitre --- arch/arm/kernel/smp.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c index fbc8b2623d82..7eacd84cdc9c 100644 --- a/arch/arm/kernel/smp.c +++ b/arch/arm/kernel/smp.c @@ -281,6 +281,7 @@ static void __cpuinit smp_store_cpu_info(unsigned int cpuid) struct cpuinfo_arm *cpu_info = &per_cpu(cpu_data, cpuid); cpu_info->loops_per_jiffy = loops_per_jiffy; + cpu_info->cpuid = read_cpuid_id(); store_cpu_topology(cpuid); } -- cgit From b4b8f770eb10a1bccaf8aa0ec1956e2dd7ed1e0a Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Mon, 10 Sep 2012 18:55:21 +0100 Subject: ARM: kernel: update cpuinfo to print all online CPUs features Currently, reading /proc/cpuinfo provides userspace with CPU ID of the CPU carrying out the read from the file. This is fine as long as all CPUs in the system are the same. With the advent of big.LITTLE and heterogenous ARM systems this approach provides user space with incorrect bits of information since CPU ids in the system might differ from the one provided by the CPU reading the file. This patch updates the cpuinfo show function so that a read from /proc/cpuinfo prints HW information for all online CPUs at once, mirroring x86 behaviour. Signed-off-by: Lorenzo Pieralisi Acked-by: Nicolas Pitre --- arch/arm/kernel/setup.c | 70 ++++++++++++++++++++++++------------------------- 1 file changed, 35 insertions(+), 35 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/setup.c b/arch/arm/kernel/setup.c index da1d1aa20ad9..a15848f8b0ff 100644 --- a/arch/arm/kernel/setup.c +++ b/arch/arm/kernel/setup.c @@ -841,12 +841,9 @@ static const char *hwcap_str[] = { static int c_show(struct seq_file *m, void *v) { - int i; - - seq_printf(m, "Processor\t: %s rev %d (%s)\n", - cpu_name, read_cpuid_id() & 15, elf_platform); + int i, j; + u32 cpuid; -#if defined(CONFIG_SMP) for_each_online_cpu(i) { /* * glibc reads /proc/cpuinfo to determine the number of @@ -854,45 +851,48 @@ static int c_show(struct seq_file *m, void *v) * "processor". Give glibc what it expects. */ seq_printf(m, "processor\t: %d\n", i); - seq_printf(m, "BogoMIPS\t: %lu.%02lu\n\n", + cpuid = is_smp() ? per_cpu(cpu_data, i).cpuid : read_cpuid_id(); + seq_printf(m, "model name\t: %s rev %d (%s)\n", + cpu_name, cpuid & 15, elf_platform); + +#if defined(CONFIG_SMP) + seq_printf(m, "BogoMIPS\t: %lu.%02lu\n", per_cpu(cpu_data, i).loops_per_jiffy / (500000UL/HZ), (per_cpu(cpu_data, i).loops_per_jiffy / (5000UL/HZ)) % 100); - } -#else /* CONFIG_SMP */ - seq_printf(m, "BogoMIPS\t: %lu.%02lu\n", - loops_per_jiffy / (500000/HZ), - (loops_per_jiffy / (5000/HZ)) % 100); +#else + seq_printf(m, "BogoMIPS\t: %lu.%02lu\n", + loops_per_jiffy / (500000/HZ), + (loops_per_jiffy / (5000/HZ)) % 100); #endif + /* dump out the processor features */ + seq_puts(m, "Features\t: "); - /* dump out the processor features */ - seq_puts(m, "Features\t: "); - - for (i = 0; hwcap_str[i]; i++) - if (elf_hwcap & (1 << i)) - seq_printf(m, "%s ", hwcap_str[i]); + for (j = 0; hwcap_str[j]; j++) + if (elf_hwcap & (1 << j)) + seq_printf(m, "%s ", hwcap_str[j]); - seq_printf(m, "\nCPU implementer\t: 0x%02x\n", read_cpuid_id() >> 24); - seq_printf(m, "CPU architecture: %s\n", proc_arch[cpu_architecture()]); + seq_printf(m, "\nCPU implementer\t: 0x%02x\n", cpuid >> 24); + seq_printf(m, "CPU architecture: %s\n", + proc_arch[cpu_architecture()]); - if ((read_cpuid_id() & 0x0008f000) == 0x00000000) { - /* pre-ARM7 */ - seq_printf(m, "CPU part\t: %07x\n", read_cpuid_id() >> 4); - } else { - if ((read_cpuid_id() & 0x0008f000) == 0x00007000) { - /* ARM7 */ - seq_printf(m, "CPU variant\t: 0x%02x\n", - (read_cpuid_id() >> 16) & 127); + if ((cpuid & 0x0008f000) == 0x00000000) { + /* pre-ARM7 */ + seq_printf(m, "CPU part\t: %07x\n", cpuid >> 4); } else { - /* post-ARM7 */ - seq_printf(m, "CPU variant\t: 0x%x\n", - (read_cpuid_id() >> 20) & 15); + if ((cpuid & 0x0008f000) == 0x00007000) { + /* ARM7 */ + seq_printf(m, "CPU variant\t: 0x%02x\n", + (cpuid >> 16) & 127); + } else { + /* post-ARM7 */ + seq_printf(m, "CPU variant\t: 0x%x\n", + (cpuid >> 20) & 15); + } + seq_printf(m, "CPU part\t: 0x%03x\n", + (cpuid >> 4) & 0xfff); } - seq_printf(m, "CPU part\t: 0x%03x\n", - (read_cpuid_id() >> 4) & 0xfff); + seq_printf(m, "CPU revision\t: %d\n\n", cpuid & 15); } - seq_printf(m, "CPU revision\t: %d\n", read_cpuid_id() & 15); - - seq_puts(m, "\n"); seq_printf(m, "Hardware\t: %s\n", machine_name); seq_printf(m, "Revision\t: %04x\n", system_rev); -- cgit From dca463daa0151c5bbbd8ec8fd42882a3966d3c44 Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Thu, 15 Nov 2012 17:30:32 +0000 Subject: ARM: kernel: enhance MPIDR macro definitions Kernel subsystems other than the topology layer need the MPIDR mask definitions to access the MPIDR without relying on hardcoded masks. This patch moves the MPIDR register masks definition to a header file and defines a macro to simplify access to MPIDR bit fields representing affinity levels. Signed-off-by: Lorenzo Pieralisi Acked-by: Will Deacon Acked-by: Nicolas Pitre --- arch/arm/kernel/topology.c | 27 +-------------------------- 1 file changed, 1 insertion(+), 26 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/topology.c b/arch/arm/kernel/topology.c index 26c12c6440fc..cd68d1aa9c3d 100644 --- a/arch/arm/kernel/topology.c +++ b/arch/arm/kernel/topology.c @@ -196,32 +196,7 @@ static inline void parse_dt_topology(void) {} static inline void update_cpu_power(unsigned int cpuid, unsigned int mpidr) {} #endif - -/* - * cpu topology management - */ - -#define MPIDR_SMP_BITMASK (0x3 << 30) -#define MPIDR_SMP_VALUE (0x2 << 30) - -#define MPIDR_MT_BITMASK (0x1 << 24) - -/* - * These masks reflect the current use of the affinity levels. - * The affinity level can be up to 16 bits according to ARM ARM - */ -#define MPIDR_HWID_BITMASK 0xFFFFFF - -#define MPIDR_LEVEL0_MASK 0x3 -#define MPIDR_LEVEL0_SHIFT 0 - -#define MPIDR_LEVEL1_MASK 0xF -#define MPIDR_LEVEL1_SHIFT 8 - -#define MPIDR_LEVEL2_MASK 0xFF -#define MPIDR_LEVEL2_SHIFT 16 - -/* + /* * cpu topology table */ struct cputopo_arm cpu_topology[NR_CPUS]; -- cgit From 71db5bfec1349afcbfbd71268c01c658c357b4f3 Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Fri, 16 Nov 2012 15:24:06 +0000 Subject: ARM: kernel: update topology to use new MPIDR macros This patch updates the topology initialization code to use the newly defined accessors to retrieve the MPIDR affinity levels. Signed-off-by: Lorenzo Pieralisi Acked-by: Will Deacon Acked-by: Nicolas Pitre --- arch/arm/kernel/topology.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/topology.c b/arch/arm/kernel/topology.c index cd68d1aa9c3d..79282ebcd939 100644 --- a/arch/arm/kernel/topology.c +++ b/arch/arm/kernel/topology.c @@ -257,19 +257,14 @@ void store_cpu_topology(unsigned int cpuid) if (mpidr & MPIDR_MT_BITMASK) { /* core performance interdependency */ - cpuid_topo->thread_id = (mpidr >> MPIDR_LEVEL0_SHIFT) - & MPIDR_LEVEL0_MASK; - cpuid_topo->core_id = (mpidr >> MPIDR_LEVEL1_SHIFT) - & MPIDR_LEVEL1_MASK; - cpuid_topo->socket_id = (mpidr >> MPIDR_LEVEL2_SHIFT) - & MPIDR_LEVEL2_MASK; + cpuid_topo->thread_id = MPIDR_AFFINITY_LEVEL(mpidr, 0); + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL(mpidr, 1); + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL(mpidr, 2); } else { /* largely independent cores */ cpuid_topo->thread_id = -1; - cpuid_topo->core_id = (mpidr >> MPIDR_LEVEL0_SHIFT) - & MPIDR_LEVEL0_MASK; - cpuid_topo->socket_id = (mpidr >> MPIDR_LEVEL1_SHIFT) - & MPIDR_LEVEL1_MASK; + cpuid_topo->core_id = MPIDR_AFFINITY_LEVEL(mpidr, 0); + cpuid_topo->socket_id = MPIDR_AFFINITY_LEVEL(mpidr, 1); } } else { /* -- cgit From cb8cf4f821044f140ea5b9c8d4f816f0c05fab44 Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Thu, 8 Nov 2012 18:05:56 +0000 Subject: ARM: kernel: smp_setup_processor_id() updates This patch applies some basic changes to the smp_setup_processor_id() ARM implementation to make the code that builds cpu_logical_map more uniform across the kernel. The function now prints the full extent of the boot CPU MPIDR[23:0] and initializes the cpu_logical_map for CPUs up to nr_cpu_ids. Signed-off-by: Lorenzo Pieralisi Acked-by: Nicolas Pitre Acked-by: Will Deacon --- arch/arm/kernel/setup.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/setup.c b/arch/arm/kernel/setup.c index da1d1aa20ad9..4515bf6abee0 100644 --- a/arch/arm/kernel/setup.c +++ b/arch/arm/kernel/setup.c @@ -426,13 +426,14 @@ int __cpu_logical_map[NR_CPUS]; void __init smp_setup_processor_id(void) { int i; - u32 cpu = is_smp() ? read_cpuid_mpidr() & 0xff : 0; + u32 mpidr = is_smp() ? read_cpuid_mpidr() & MPIDR_HWID_BITMASK : 0; + u32 cpu = MPIDR_AFFINITY_LEVEL(mpidr, 0); cpu_logical_map(0) = cpu; - for (i = 1; i < NR_CPUS; ++i) + for (i = 1; i < nr_cpu_ids; ++i) cpu_logical_map(i) = i == cpu ? 0 : i; - printk(KERN_INFO "Booting Linux on physical CPU %d\n", cpu); + printk(KERN_INFO "Booting Linux on physical CPU 0x%x\n", mpidr); } static void __init setup_processor(void) -- cgit From a0ae02405076ac32bd17ece976e914b5b6075bb0 Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Thu, 17 Nov 2011 17:31:51 +0000 Subject: ARM: kernel: add device tree init map function When booting through a device tree, the kernel cpu logical id map can be initialized using device tree data passed by FW or through an embedded blob. This patch adds a function that parses device tree "cpu" nodes and retrieves the corresponding CPUs hardware identifiers (MPIDR). It sets the possible cpus and the cpu logical map values according to the number of CPUs defined in the device tree and respective properties. The device tree HW identifiers are considered valid if all CPU nodes contain a "reg" property, there are no duplicate "reg" entries and the DT defines a CPU node whose "reg" property matches the MPIDR[23:0] of the boot CPU. The primary CPU is assigned cpu logical number 0 to keep the current convention valid. Current bindings documentation is included in the patch: Documentation/devicetree/bindings/arm/cpus.txt Signed-off-by: Lorenzo Pieralisi Acked-by: Nicolas Pitre --- arch/arm/kernel/devtree.c | 100 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 100 insertions(+) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/devtree.c b/arch/arm/kernel/devtree.c index bee7f9d47f02..aaf9add497fe 100644 --- a/arch/arm/kernel/devtree.c +++ b/arch/arm/kernel/devtree.c @@ -19,8 +19,10 @@ #include #include +#include #include #include +#include #include #include @@ -61,6 +63,104 @@ void __init arm_dt_memblock_reserve(void) } } +/* + * arm_dt_init_cpu_maps - Function retrieves cpu nodes from the device tree + * and builds the cpu logical map array containing MPIDR values related to + * logical cpus + * + * Updates the cpu possible mask with the number of parsed cpu nodes + */ +void __init arm_dt_init_cpu_maps(void) +{ + /* + * Temp logical map is initialized with UINT_MAX values that are + * considered invalid logical map entries since the logical map must + * contain a list of MPIDR[23:0] values where MPIDR[31:24] must + * read as 0. + */ + struct device_node *cpu, *cpus; + u32 i, j, cpuidx = 1; + u32 mpidr = is_smp() ? read_cpuid_mpidr() & MPIDR_HWID_BITMASK : 0; + + u32 tmp_map[NR_CPUS] = { [0 ... NR_CPUS-1] = UINT_MAX }; + bool bootcpu_valid = false; + cpus = of_find_node_by_path("/cpus"); + + if (!cpus) + return; + + for_each_child_of_node(cpus, cpu) { + u32 hwid; + + pr_debug(" * %s...\n", cpu->full_name); + /* + * A device tree containing CPU nodes with missing "reg" + * properties is considered invalid to build the + * cpu_logical_map. + */ + if (of_property_read_u32(cpu, "reg", &hwid)) { + pr_debug(" * %s missing reg property\n", + cpu->full_name); + return; + } + + /* + * 8 MSBs must be set to 0 in the DT since the reg property + * defines the MPIDR[23:0]. + */ + if (hwid & ~MPIDR_HWID_BITMASK) + return; + + /* + * Duplicate MPIDRs are a recipe for disaster. + * Scan all initialized entries and check for + * duplicates. If any is found just bail out. + * temp values were initialized to UINT_MAX + * to avoid matching valid MPIDR[23:0] values. + */ + for (j = 0; j < cpuidx; j++) + if (WARN(tmp_map[j] == hwid, "Duplicate /cpu reg " + "properties in the DT\n")) + return; + + /* + * Build a stashed array of MPIDR values. Numbering scheme + * requires that if detected the boot CPU must be assigned + * logical id 0. Other CPUs get sequential indexes starting + * from 1. If a CPU node with a reg property matching the + * boot CPU MPIDR is detected, this is recorded so that the + * logical map built from DT is validated and can be used + * to override the map created in smp_setup_processor_id(). + */ + if (hwid == mpidr) { + i = 0; + bootcpu_valid = true; + } else { + i = cpuidx++; + } + + tmp_map[i] = hwid; + + if (cpuidx > nr_cpu_ids) + break; + } + + if (WARN(!bootcpu_valid, "DT missing boot CPU MPIDR[23:0], " + "fall back to default cpu_logical_map\n")) + return; + + /* + * Since the boot CPU node contains proper data, and all nodes have + * a reg property, the DT CPU list can be considered valid and the + * logical map created in smp_setup_processor_id() can be overridden + */ + for (i = 0; i < cpuidx; i++) { + set_cpu_possible(i, true); + cpu_logical_map(i) = tmp_map[i]; + pr_debug("cpu logical map 0x%x\n", cpu_logical_map(i)); + } +} + /** * setup_machine_fdt - Machine setup when an dtb was passed to the kernel * @dt_phys: physical address of dt blob -- cgit From 5587164eea4aad88fcb79d9b21dc8f14fea598cd Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Wed, 14 Dec 2011 16:01:24 +0000 Subject: ARM: kernel: add cpu logical map DT init in setup_arch As soon as the device tree is unflattened the cpu logical to physical mapping is carried out in setup_arch to build a proper array of MPIDR and corresponding logical indexes. The mapping could have been carried out using the flattened DT blob and related primitives, but since the mapping is not needed by early boot code it can safely be executed when the device tree has been uncompressed to its tree data structure. This patch adds the arm_dt_init_cpu maps() function call in setup_arch(). If the kernel is not compiled with DT support the function is empty and no logical mapping takes place through it; the mapping carried out in smp_setup_processor_id() is left unchanged. If DT is supported the mapping created in smp_setup_processor_id() is overriden. The DT mapping also sets the possible cpus mask, hence platform code need not set it again in the respective smp_init_cpus() functions. Signed-off-by: Lorenzo Pieralisi Acked-by: Will Deacon Acked-by: Nicolas Pitre --- arch/arm/kernel/setup.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/setup.c b/arch/arm/kernel/setup.c index 4515bf6abee0..d15f1c503f3d 100644 --- a/arch/arm/kernel/setup.c +++ b/arch/arm/kernel/setup.c @@ -759,6 +759,7 @@ void __init setup_arch(char **cmdline_p) unflatten_device_tree(); + arm_dt_init_cpu_maps(); #ifdef CONFIG_SMP if (is_smp()) { smp_set_ops(mdesc->smp); -- cgit From ce7b175656a1903605f0184bf33acebff70bfe7f Mon Sep 17 00:00:00 2001 From: Lorenzo Pieralisi Date: Thu, 22 Nov 2012 18:02:54 +0100 Subject: ARM: 7585/1: kernel: fix nr_cpu_ids check in DT logical map init If a kernel is configured with a DT containing more /cpu nodes than nr_cpu_ids, the number of cpus must be capped in the DT parsing code. Current code carries out the check, but fails to cap the value and the check is executed after the cpu logical index is used, which can lead to memory corruption due to index overflow. This patch refactors the check against nr_cpu_ids and move it before any computed index is used in the parsing code. Signed-off-by: Lorenzo Pieralisi Acked-by: Grant Likely Reported-by: Mark Rutland Signed-off-by: Russell King --- arch/arm/kernel/devtree.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/devtree.c b/arch/arm/kernel/devtree.c index aaf9add497fe..70f1bdeb241b 100644 --- a/arch/arm/kernel/devtree.c +++ b/arch/arm/kernel/devtree.c @@ -139,10 +139,14 @@ void __init arm_dt_init_cpu_maps(void) i = cpuidx++; } - tmp_map[i] = hwid; - - if (cpuidx > nr_cpu_ids) + if (WARN(cpuidx > nr_cpu_ids, "DT /cpu %u nodes greater than " + "max cores %u, capping them\n", + cpuidx, nr_cpu_ids)) { + cpuidx = nr_cpu_ids; break; + } + + tmp_map[i] = hwid; } if (WARN(!bootcpu_valid, "DT missing boot CPU MPIDR[23:0], " -- cgit From 14318efb322e2fe1a034c69463d725209eb9d548 Mon Sep 17 00:00:00 2001 From: Rob Herring Date: Thu, 29 Nov 2012 20:39:54 +0100 Subject: ARM: 7587/1: implement optimized percpu variable access Use the previously unused TPIDRPRW register to store percpu offsets. TPIDRPRW is only accessible in PL1, so it can only be used in the kernel. This replaces 2 loads with a mrc instruction for each percpu variable access. With hackbench, the performance improvement is 1.4% on Cortex-A9 (highbank). Taking an average of 30 runs of "hackbench -l 1000" yields: Before: 6.2191 After: 6.1348 Will Deacon reported similar delta on v6 with 11MPCore. The asm "memory clobber" are needed here to ensure the percpu offset gets reloaded. Testing by Will found that this would not happen in __schedule() which is a bit of a special case as preemption is disabled but the execution can move cores. Signed-off-by: Rob Herring Acked-by: Will Deacon Acked-by: Nicolas Pitre Signed-off-by: Russell King --- arch/arm/kernel/setup.c | 6 ++++++ arch/arm/kernel/smp.c | 4 +++- 2 files changed, 9 insertions(+), 1 deletion(-) (limited to 'arch/arm/kernel') diff --git a/arch/arm/kernel/setup.c b/arch/arm/kernel/setup.c index f739fb1d217a..9a89bf4aefe1 100644 --- a/arch/arm/kernel/setup.c +++ b/arch/arm/kernel/setup.c @@ -383,6 +383,12 @@ void cpu_init(void) BUG(); } + /* + * This only works on resume and secondary cores. For booting on the + * boot cpu, smp_prepare_boot_cpu is called after percpu area setup. + */ + set_my_cpu_offset(per_cpu_offset(cpu)); + cpu_proc_init(); /* diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c index 7eacd84cdc9c..f3a2be5837aa 100644 --- a/arch/arm/kernel/smp.c +++ b/arch/arm/kernel/smp.c @@ -314,9 +314,10 @@ asmlinkage void __cpuinit secondary_start_kernel(void) current->active_mm = mm; cpumask_set_cpu(cpu, mm_cpumask(mm)); + cpu_init(); + printk("CPU%u: Booted secondary processor\n", cpu); - cpu_init(); preempt_disable(); trace_hardirqs_off(); @@ -372,6 +373,7 @@ void __init smp_cpus_done(unsigned int max_cpus) void __init smp_prepare_boot_cpu(void) { + set_my_cpu_offset(per_cpu_offset(smp_processor_id())); } void __init smp_prepare_cpus(unsigned int max_cpus) -- cgit