summaryrefslogtreecommitdiff
path: root/kernel
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2017-09-04 08:39:02 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2017-09-04 08:39:02 -0700
commit9657752cb5039c7498d4b27c4a75530f93b87d9b (patch)
treeef4198ba427da0ef5e1cb8fb4ec62843b645aed9 /kernel
parent0081a0ce809b611c1f37da5d6ae5bc8027ffd1c4 (diff)
parent1b2f76d77a277bb70d38ad0991ed7f16bbc115a9 (diff)
Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull perf updates from Ingo Molnar: "Kernel side changes: - Add branch type profiling/tracing support. (Jin Yao) - Add the PERF_SAMPLE_PHYS_ADDR ABI to allow the tracing/profiling of physical memory addresses, where the PMU supports it. (Kan Liang) - Export some PMU capability details in the new /sys/bus/event_source/devices/cpu/caps/ sysfs directory. (Andi Kleen) - Aux data fixes and updates (Will Deacon) - kprobes fixes and updates (Masami Hiramatsu) - AMD uncore PMU driver fixes and updates (Janakarajan Natarajan) On the tooling side, here's a (limited!) list of highlights - there were many other changes that I could not list, see the shortlog and git history for details: UI improvements: - Implement a visual marker for fused x86 instructions in the annotate TUI browser, available now in 'perf report', more work needed to have it available as well in 'perf top' (Jin Yao) Further explanation from one of Jin's patches: │ ┌──cmpl $0x0,argp_program_version_hook 81.93 │ ├──je 20 │ │ lock cmpxchg %esi,0x38a9a4(%rip) │ │↓ jne 29 │ │↓ jmp 43 11.47 │20:└─→cmpxch %esi,0x38a999(%rip) That means the cmpl+je is a fused instruction pair and they should be considered together. - Record the branch type and then show statistics and info about in callchain entries (Jin Yao) Example from one of Jin's patches: # perf record -g -j any,save_type # perf report --branch-history --stdio --no-children 38.50% div.c:45 [.] main div | ---main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) namespaces support: - Add initial support for namespaces, using setns to access files in namespaces, grabbing their build-ids, etc. (Krister Johansen) perf trace enhancements: - Beautify pkey_{alloc,free,mprotect} arguments in 'perf trace' (Arnaldo Carvalho de Melo) - Add initial 'clone' syscall args beautifier in 'perf trace' (Arnaldo Carvalho de Melo) - Ignore 'fd' and 'offset' args for MAP_ANONYMOUS in 'perf trace' (Arnaldo Carvalho de Melo) - Beautifiers for the 'cmd' arg of several ioctl types, including: sound, DRM, KVM, vhost virtio and perf_events. (Arnaldo Carvalho de Melo) - Add PERF_SAMPLE_CALLCHAIN and PERF_RECORD_MMAP[2] to 'perf data' CTF conversion, allowing CTF trace visualization tools to show callchains and to resolve symbols (Geneviève Bastien) - Beautify the fcntl syscall, which is an interesting one in the sense that infrastructure had to be put in place to change the formatters of some arguments according to the value in a previous one, i.e. cmd dictates how arg and the syscall return will be formatted. (Arnaldo Carvalho de Melo perf stat enhancements: - Use group read for event groups in 'perf stat', reducing overhead when groups are defined in the event specification, i.e. when using {} to enclose a list of events, asking them to be read at the same time, e.g.: "perf stat -e '{cycles,instructions}'" (Jiri Olsa) pipe mode improvements: - Process tracing data in 'perf annotate' pipe mode (David Carrillo-Cisneros) - Add header record types to pipe-mode, now this command: $ perf record -o - -e cycles sleep 1 | perf report --stdio --header Will show the same as in non-pipe mode, i.e. involving a perf.data file (David Carrillo-Cisneros) Vendor specific hardware event support updates/enhancements: - Update POWER9 vendor events tables (Sukadev Bhattiprolu) - Add POWER9 PMU events Sukadev (Bhattiprolu) - Support additional POWER8+ PVR in PMU mapfile (Shriya) - Add Skylake server uncore JSON vendor events (Andi Kleen) - Support exporting Intel PT data to sqlite3 with python perf scripts, this is in addition to the postgresql support that was already there (Adrian Hunter)" * 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (253 commits) perf symbols: Fix plt entry calculation for ARM and AARCH64 perf probe: Fix kprobe blacklist checking condition perf/x86: Fix caps/ for !Intel perf/core, x86: Add PERF_SAMPLE_PHYS_ADDR perf/core, pt, bts: Get rid of itrace_started perf trace beauty: Beautify pkey_{alloc,free,mprotect} arguments tools headers: Sync cpu features kernel ABI headers with tooling headers perf tools: Pass full path of FEATURES_DUMP perf tools: Robustify detection of clang binary tools lib: Allow external definition of CC, AR and LD perf tools: Allow external definition of flex and bison binary names tools build tests: Don't hardcode gcc name perf report: Group stat values on global event id perf values: Zero value buffers perf values: Fix allocation check perf values: Fix thread index bug perf report: Add dump_read function perf record: Set read_format for inherit_stat perf c2c: Fix remote HITM detection for Skylake perf tools: Fix static build with newer toolchains ...
Diffstat (limited to 'kernel')
-rw-r--r--kernel/events/core.c85
-rw-r--r--kernel/events/internal.h6
-rw-r--r--kernel/events/ring_buffer.c31
3 files changed, 92 insertions, 30 deletions
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 03ac9c8b02fb..ce64f3fed5c6 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -1249,26 +1249,31 @@ unclone_ctx(struct perf_event_context *ctx)
return parent_ctx;
}
-static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
+static u32 perf_event_pid_type(struct perf_event *event, struct task_struct *p,
+ enum pid_type type)
{
+ u32 nr;
/*
* only top level events have the pid namespace they were created in
*/
if (event->parent)
event = event->parent;
- return task_tgid_nr_ns(p, event->ns);
+ nr = __task_pid_nr_ns(p, type, event->ns);
+ /* avoid -1 if it is idle thread or runs in another ns */
+ if (!nr && !pid_alive(p))
+ nr = -1;
+ return nr;
}
-static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
+static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
{
- /*
- * only top level events have the pid namespace they were created in
- */
- if (event->parent)
- event = event->parent;
+ return perf_event_pid_type(event, p, __PIDTYPE_TGID);
+}
- return task_pid_nr_ns(p, event->ns);
+static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
+{
+ return perf_event_pid_type(event, p, PIDTYPE_PID);
}
/*
@@ -1570,6 +1575,9 @@ static void __perf_event_header_size(struct perf_event *event, u64 sample_type)
if (sample_type & PERF_SAMPLE_TRANSACTION)
size += sizeof(data->txn);
+ if (sample_type & PERF_SAMPLE_PHYS_ADDR)
+ size += sizeof(data->phys_addr);
+
event->header_size = size;
}
@@ -3211,6 +3219,13 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
return;
perf_ctx_lock(cpuctx, ctx);
+ /*
+ * We must check ctx->nr_events while holding ctx->lock, such
+ * that we serialize against perf_install_in_context().
+ */
+ if (!ctx->nr_events)
+ goto unlock;
+
perf_pmu_disable(ctx->pmu);
/*
* We want to keep the following priority order:
@@ -3224,6 +3239,8 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
perf_event_sched_in(cpuctx, ctx, task);
perf_pmu_enable(ctx->pmu);
+
+unlock:
perf_ctx_unlock(cpuctx, ctx);
}
@@ -6003,6 +6020,9 @@ void perf_output_sample(struct perf_output_handle *handle,
}
}
+ if (sample_type & PERF_SAMPLE_PHYS_ADDR)
+ perf_output_put(handle, data->phys_addr);
+
if (!event->attr.watermark) {
int wakeup_events = event->attr.wakeup_events;
@@ -6018,6 +6038,38 @@ void perf_output_sample(struct perf_output_handle *handle,
}
}
+static u64 perf_virt_to_phys(u64 virt)
+{
+ u64 phys_addr = 0;
+ struct page *p = NULL;
+
+ if (!virt)
+ return 0;
+
+ if (virt >= TASK_SIZE) {
+ /* If it's vmalloc()d memory, leave phys_addr as 0 */
+ if (virt_addr_valid((void *)(uintptr_t)virt) &&
+ !(virt >= VMALLOC_START && virt < VMALLOC_END))
+ phys_addr = (u64)virt_to_phys((void *)(uintptr_t)virt);
+ } else {
+ /*
+ * Walking the pages tables for user address.
+ * Interrupts are disabled, so it prevents any tear down
+ * of the page tables.
+ * Try IRQ-safe __get_user_pages_fast first.
+ * If failed, leave phys_addr as 0.
+ */
+ if ((current->mm != NULL) &&
+ (__get_user_pages_fast(virt, 1, 0, &p) == 1))
+ phys_addr = page_to_phys(p) + virt % PAGE_SIZE;
+
+ if (p)
+ put_page(p);
+ }
+
+ return phys_addr;
+}
+
void perf_prepare_sample(struct perf_event_header *header,
struct perf_sample_data *data,
struct perf_event *event,
@@ -6136,6 +6188,9 @@ void perf_prepare_sample(struct perf_event_header *header,
header->size += size;
}
+
+ if (sample_type & PERF_SAMPLE_PHYS_ADDR)
+ data->phys_addr = perf_virt_to_phys(data->addr);
}
static void __always_inline
@@ -7287,6 +7342,11 @@ static void perf_log_throttle(struct perf_event *event, int enable)
perf_output_end(&handle);
}
+void perf_event_itrace_started(struct perf_event *event)
+{
+ event->attach_state |= PERF_ATTACH_ITRACE;
+}
+
static void perf_log_itrace_start(struct perf_event *event)
{
struct perf_output_handle handle;
@@ -7302,7 +7362,7 @@ static void perf_log_itrace_start(struct perf_event *event)
event = event->parent;
if (!(event->pmu->capabilities & PERF_PMU_CAP_ITRACE) ||
- event->hw.itrace_started)
+ event->attach_state & PERF_ATTACH_ITRACE)
return;
rec.header.type = PERF_RECORD_ITRACE_START;
@@ -9890,6 +9950,11 @@ SYSCALL_DEFINE5(perf_event_open,
return -EINVAL;
}
+ /* Only privileged users can get physical addresses */
+ if ((attr.sample_type & PERF_SAMPLE_PHYS_ADDR) &&
+ perf_paranoid_kernel() && !capable(CAP_SYS_ADMIN))
+ return -EACCES;
+
if (!attr.sample_max_stack)
attr.sample_max_stack = sysctl_perf_event_max_stack;
diff --git a/kernel/events/internal.h b/kernel/events/internal.h
index 486fd78eb8d5..843e97047335 100644
--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -38,9 +38,9 @@ struct ring_buffer {
struct user_struct *mmap_user;
/* AUX area */
- local_t aux_head;
+ long aux_head;
local_t aux_nest;
- local_t aux_wakeup;
+ long aux_wakeup; /* last aux_watermark boundary crossed by aux_head */
unsigned long aux_pgoff;
int aux_nr_pages;
int aux_overwrite;
@@ -208,7 +208,7 @@ static inline int get_recursion_context(int *recursion)
{
int rctx;
- if (in_nmi())
+ if (unlikely(in_nmi()))
rctx = 3;
else if (in_irq())
rctx = 2;
diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c
index ee97196bb151..af71a84e12ee 100644
--- a/kernel/events/ring_buffer.c
+++ b/kernel/events/ring_buffer.c
@@ -367,7 +367,7 @@ void *perf_aux_output_begin(struct perf_output_handle *handle,
if (WARN_ON_ONCE(local_xchg(&rb->aux_nest, 1)))
goto err_put;
- aux_head = local_read(&rb->aux_head);
+ aux_head = rb->aux_head;
handle->rb = rb;
handle->event = event;
@@ -382,7 +382,7 @@ void *perf_aux_output_begin(struct perf_output_handle *handle,
*/
if (!rb->aux_overwrite) {
aux_tail = ACCESS_ONCE(rb->user_page->aux_tail);
- handle->wakeup = local_read(&rb->aux_wakeup) + rb->aux_watermark;
+ handle->wakeup = rb->aux_wakeup + rb->aux_watermark;
if (aux_head - aux_tail < perf_aux_size(rb))
handle->size = CIRC_SPACE(aux_head, aux_tail, perf_aux_size(rb));
@@ -433,12 +433,12 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size)
handle->aux_flags |= PERF_AUX_FLAG_OVERWRITE;
aux_head = handle->head;
- local_set(&rb->aux_head, aux_head);
+ rb->aux_head = aux_head;
} else {
handle->aux_flags &= ~PERF_AUX_FLAG_OVERWRITE;
- aux_head = local_read(&rb->aux_head);
- local_add(size, &rb->aux_head);
+ aux_head = rb->aux_head;
+ rb->aux_head += size;
}
if (size || handle->aux_flags) {
@@ -450,11 +450,10 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size)
handle->aux_flags);
}
- aux_head = rb->user_page->aux_head = local_read(&rb->aux_head);
-
- if (aux_head - local_read(&rb->aux_wakeup) >= rb->aux_watermark) {
+ rb->user_page->aux_head = rb->aux_head;
+ if (rb->aux_head - rb->aux_wakeup >= rb->aux_watermark) {
wakeup = true;
- local_add(rb->aux_watermark, &rb->aux_wakeup);
+ rb->aux_wakeup = rounddown(rb->aux_head, rb->aux_watermark);
}
if (wakeup) {
@@ -478,22 +477,20 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size)
int perf_aux_output_skip(struct perf_output_handle *handle, unsigned long size)
{
struct ring_buffer *rb = handle->rb;
- unsigned long aux_head;
if (size > handle->size)
return -ENOSPC;
- local_add(size, &rb->aux_head);
+ rb->aux_head += size;
- aux_head = rb->user_page->aux_head = local_read(&rb->aux_head);
- if (aux_head - local_read(&rb->aux_wakeup) >= rb->aux_watermark) {
+ rb->user_page->aux_head = rb->aux_head;
+ if (rb->aux_head - rb->aux_wakeup >= rb->aux_watermark) {
perf_output_wakeup(handle);
- local_add(rb->aux_watermark, &rb->aux_wakeup);
- handle->wakeup = local_read(&rb->aux_wakeup) +
- rb->aux_watermark;
+ rb->aux_wakeup = rounddown(rb->aux_head, rb->aux_watermark);
+ handle->wakeup = rb->aux_wakeup + rb->aux_watermark;
}
- handle->head = aux_head;
+ handle->head = rb->aux_head;
handle->size -= size;
return 0;