From 075e0b00857e166dcc3e39037a1fc5a90acac709 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:40 +0200 Subject: perf: Optimize ctx_sched_out() Oleg noted that ctx_sched_out() disables the PMU even though it might not actually do something, avoid needless PMU-disabling. Reported-by: Oleg Nesterov Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.665385503@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index d863b3c057bb..4d9a1f014286 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1760,7 +1760,6 @@ static void ctx_sched_out(struct perf_event_context *ctx, struct perf_event *event; raw_spin_lock(&ctx->lock); - perf_pmu_disable(ctx->pmu); ctx->is_active = 0; if (likely(!ctx->nr_events)) goto out; @@ -1770,6 +1769,7 @@ static void ctx_sched_out(struct perf_event_context *ctx, if (!ctx->nr_active) goto out; + perf_pmu_disable(ctx->pmu); if (event_type & EVENT_PINNED) { list_for_each_entry(event, &ctx->pinned_groups, group_entry) group_sched_out(event, cpuctx, ctx); @@ -1779,8 +1779,8 @@ static void ctx_sched_out(struct perf_event_context *ctx, list_for_each_entry(event, &ctx->flexible_groups, group_entry) group_sched_out(event, cpuctx, ctx); } -out: perf_pmu_enable(ctx->pmu); +out: raw_spin_unlock(&ctx->lock); } -- cgit v1.2.3 From 9137fb28ac74d05eb66d1d8e6778eaa14e6fed43 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:41 +0200 Subject: perf: Clean up 'ctx' reference counting Small cleanup to how we refcount in find_get_context(), this also allows us to use put_ctx() to free things instead of using kfree(). Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.719340481@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 4d9a1f014286..d665ac4242f2 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -2835,16 +2835,12 @@ retry: unclone_ctx(ctx); ++ctx->pin_count; raw_spin_unlock_irqrestore(&ctx->lock, flags); - } - - if (!ctx) { + } else { ctx = alloc_perf_context(pmu, task); err = -ENOMEM; if (!ctx) goto errout; - get_ctx(ctx); - err = 0; mutex_lock(&task->perf_event_mutex); /* @@ -2856,14 +2852,14 @@ retry: else if (task->perf_event_ctxp[ctxn]) err = -EAGAIN; else { + get_ctx(ctx); ++ctx->pin_count; rcu_assign_pointer(task->perf_event_ctxp[ctxn], ctx); } mutex_unlock(&task->perf_event_mutex); if (unlikely(err)) { - put_task_struct(task); - kfree(ctx); + put_ctx(ctx); if (err == -EAGAIN) goto retry; -- cgit v1.2.3 From facc43071cc0d4821c176d7d34570714eb348df9 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:42 +0200 Subject: perf: Optimize event scheduling locking Currently we only hold one ctx->lock at a time, which results in us flipping back and forth between cpuctx->ctx.lock and task_ctx->lock. Avoid this and gain large atomic regions by holding both locks. We nest the task lock inside the cpu lock, since with task scheduling we might have to change task ctx while holding the cpu ctx lock. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.769881865@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 61 +++++++++++++++++++++++++++++++--------------------- 1 file changed, 36 insertions(+), 25 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index d665ac4242f2..d243af954dcc 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -200,6 +200,22 @@ __get_cpu_context(struct perf_event_context *ctx) return this_cpu_ptr(ctx->pmu->pmu_cpu_context); } +static void perf_ctx_lock(struct perf_cpu_context *cpuctx, + struct perf_event_context *ctx) +{ + raw_spin_lock(&cpuctx->ctx.lock); + if (ctx) + raw_spin_lock(&ctx->lock); +} + +static void perf_ctx_unlock(struct perf_cpu_context *cpuctx, + struct perf_event_context *ctx) +{ + if (ctx) + raw_spin_unlock(&ctx->lock); + raw_spin_unlock(&cpuctx->ctx.lock); +} + #ifdef CONFIG_CGROUP_PERF /* @@ -340,11 +356,8 @@ void perf_cgroup_switch(struct task_struct *task, int mode) rcu_read_lock(); list_for_each_entry_rcu(pmu, &pmus, entry) { - cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); - perf_pmu_disable(cpuctx->ctx.pmu); - /* * perf_cgroup_events says at least one * context on this CPU has cgroup events. @@ -353,6 +366,8 @@ void perf_cgroup_switch(struct task_struct *task, int mode) * events for a context. */ if (cpuctx->ctx.nr_cgroups > 0) { + perf_ctx_lock(cpuctx, cpuctx->task_ctx); + perf_pmu_disable(cpuctx->ctx.pmu); if (mode & PERF_CGROUP_SWOUT) { cpu_ctx_sched_out(cpuctx, EVENT_ALL); @@ -372,9 +387,9 @@ void perf_cgroup_switch(struct task_struct *task, int mode) cpuctx->cgrp = perf_cgroup_from_task(task); cpu_ctx_sched_in(cpuctx, EVENT_ALL, task); } + perf_pmu_enable(cpuctx->ctx.pmu); + perf_ctx_unlock(cpuctx, cpuctx->task_ctx); } - - perf_pmu_enable(cpuctx->ctx.pmu); } rcu_read_unlock(); @@ -1759,15 +1774,14 @@ static void ctx_sched_out(struct perf_event_context *ctx, { struct perf_event *event; - raw_spin_lock(&ctx->lock); ctx->is_active = 0; if (likely(!ctx->nr_events)) - goto out; + return; + update_context_time(ctx); update_cgrp_time_from_cpuctx(cpuctx); - if (!ctx->nr_active) - goto out; + return; perf_pmu_disable(ctx->pmu); if (event_type & EVENT_PINNED) { @@ -1780,8 +1794,6 @@ static void ctx_sched_out(struct perf_event_context *ctx, group_sched_out(event, cpuctx, ctx); } perf_pmu_enable(ctx->pmu); -out: - raw_spin_unlock(&ctx->lock); } /* @@ -1929,8 +1941,10 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, rcu_read_unlock(); if (do_switch) { + raw_spin_lock(&ctx->lock); ctx_sched_out(ctx, cpuctx, EVENT_ALL); cpuctx->task_ctx = NULL; + raw_spin_unlock(&ctx->lock); } } @@ -2056,10 +2070,9 @@ ctx_sched_in(struct perf_event_context *ctx, { u64 now; - raw_spin_lock(&ctx->lock); ctx->is_active = 1; if (likely(!ctx->nr_events)) - goto out; + return; now = perf_clock(); ctx->timestamp = now; @@ -2074,9 +2087,6 @@ ctx_sched_in(struct perf_event_context *ctx, /* Then walk through the lower prio flexible groups */ if (event_type & EVENT_FLEXIBLE) ctx_flexible_sched_in(ctx, cpuctx); - -out: - raw_spin_unlock(&ctx->lock); } static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx, @@ -2110,6 +2120,7 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx, if (cpuctx->task_ctx == ctx) return; + perf_ctx_lock(cpuctx, ctx); perf_pmu_disable(ctx->pmu); /* * We want to keep the following priority order: @@ -2124,12 +2135,14 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx, cpuctx->task_ctx = ctx; + perf_pmu_enable(ctx->pmu); + perf_ctx_unlock(cpuctx, ctx); + /* * Since these rotations are per-cpu, we need to ensure the * cpu-context we got scheduled on is actually rotating. */ perf_pmu_rotate_start(ctx->pmu); - perf_pmu_enable(ctx->pmu); } /* @@ -2269,7 +2282,6 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period) u64 interrupts, now; s64 delta; - raw_spin_lock(&ctx->lock); list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { if (event->state != PERF_EVENT_STATE_ACTIVE) continue; @@ -2301,7 +2313,6 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period) if (delta > 0) perf_adjust_period(event, period, delta); } - raw_spin_unlock(&ctx->lock); } /* @@ -2309,16 +2320,12 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period) */ static void rotate_ctx(struct perf_event_context *ctx) { - raw_spin_lock(&ctx->lock); - /* * Rotate the first entry last of non-pinned groups. Rotation might be * disabled by the inheritance code. */ if (!ctx->rotate_disable) list_rotate_left(&ctx->flexible_groups); - - raw_spin_unlock(&ctx->lock); } /* @@ -2345,6 +2352,7 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) rotate = 1; } + perf_ctx_lock(cpuctx, cpuctx->task_ctx); perf_pmu_disable(cpuctx->ctx.pmu); perf_ctx_adjust_freq(&cpuctx->ctx, interval); if (ctx) @@ -2370,6 +2378,7 @@ done: list_del_init(&cpuctx->rotation_list); perf_pmu_enable(cpuctx->ctx.pmu); + perf_ctx_unlock(cpuctx, cpuctx->task_ctx); } void perf_event_task_tick(void) @@ -2424,9 +2433,9 @@ static void perf_event_enable_on_exec(struct perf_event_context *ctx) * in. */ perf_cgroup_sched_out(current); - task_ctx_sched_out(ctx, EVENT_ALL); raw_spin_lock(&ctx->lock); + task_ctx_sched_out(ctx, EVENT_ALL); list_for_each_entry(event, &ctx->pinned_groups, group_entry) { ret = event_enable_on_exec(event, ctx); @@ -5982,6 +5991,7 @@ free_dev: } static struct lock_class_key cpuctx_mutex; +static struct lock_class_key cpuctx_lock; int perf_pmu_register(struct pmu *pmu, char *name, int type) { @@ -6032,6 +6042,7 @@ skip_type: cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); __perf_event_init_context(&cpuctx->ctx); lockdep_set_class(&cpuctx->ctx.mutex, &cpuctx_mutex); + lockdep_set_class(&cpuctx->ctx.lock, &cpuctx_lock); cpuctx->ctx.type = cpu_context; cpuctx->ctx.pmu = pmu; cpuctx->jiffies_interval = 1; @@ -6776,7 +6787,6 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) * our context. */ child_ctx = rcu_dereference_raw(child->perf_event_ctxp[ctxn]); - task_ctx_sched_out(child_ctx, EVENT_ALL); /* * Take the context lock here so that if find_get_context is @@ -6784,6 +6794,7 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) * incremented the context's refcount before we do put_ctx below. */ raw_spin_lock(&child_ctx->lock); + task_ctx_sched_out(child_ctx, EVENT_ALL); child->perf_event_ctxp[ctxn] = NULL; /* * If this context is a clone; unclone it so it can't get -- cgit v1.2.3 From 04dc2dbbfe1c6f81b996d4dab255da75f9efbb4a Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:43 +0200 Subject: perf: Remove task_ctx_sched_in() Make task_ctx_sched_*() imply EVENT_ALL, since anything less will not actually have scheduled the task in/out at all. Since there's no site that schedules all of a task in (due to the interleave with flexible cpuctx) we can remove this function. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.817893268@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 26 ++++++-------------------- 1 file changed, 6 insertions(+), 20 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index d243af954dcc..66b3dd809409 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1979,8 +1979,7 @@ void __perf_event_task_sched_out(struct task_struct *task, perf_cgroup_sched_out(task); } -static void task_ctx_sched_out(struct perf_event_context *ctx, - enum event_type_t event_type) +static void task_ctx_sched_out(struct perf_event_context *ctx) { struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); @@ -1990,7 +1989,7 @@ static void task_ctx_sched_out(struct perf_event_context *ctx, if (WARN_ON_ONCE(ctx != cpuctx->task_ctx)) return; - ctx_sched_out(ctx, cpuctx, event_type); + ctx_sched_out(ctx, cpuctx, EVENT_ALL); cpuctx->task_ctx = NULL; } @@ -2098,19 +2097,6 @@ static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx, ctx_sched_in(ctx, cpuctx, event_type, task); } -static void task_ctx_sched_in(struct perf_event_context *ctx, - enum event_type_t event_type) -{ - struct perf_cpu_context *cpuctx; - - cpuctx = __get_cpu_context(ctx); - if (cpuctx->task_ctx == ctx) - return; - - ctx_sched_in(ctx, cpuctx, event_type, NULL); - cpuctx->task_ctx = ctx; -} - static void perf_event_context_sched_in(struct perf_event_context *ctx, struct task_struct *task) { @@ -2363,7 +2349,7 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); if (ctx) - task_ctx_sched_out(ctx, EVENT_FLEXIBLE); + ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE); rotate_ctx(&cpuctx->ctx); if (ctx) @@ -2371,7 +2357,7 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, current); if (ctx) - task_ctx_sched_in(ctx, EVENT_FLEXIBLE); + ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, current); done: if (remove) @@ -2435,7 +2421,7 @@ static void perf_event_enable_on_exec(struct perf_event_context *ctx) perf_cgroup_sched_out(current); raw_spin_lock(&ctx->lock); - task_ctx_sched_out(ctx, EVENT_ALL); + task_ctx_sched_out(ctx); list_for_each_entry(event, &ctx->pinned_groups, group_entry) { ret = event_enable_on_exec(event, ctx); @@ -6794,7 +6780,7 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) * incremented the context's refcount before we do put_ctx below. */ raw_spin_lock(&child_ctx->lock); - task_ctx_sched_out(child_ctx, EVENT_ALL); + task_ctx_sched_out(child_ctx); child->perf_event_ctxp[ctxn] = NULL; /* * If this context is a clone; unclone it so it can't get -- cgit v1.2.3 From 2c29ef0fef8aaff1f91263fc75c749d659da6972 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:44 +0200 Subject: perf: Simplify and fix __perf_install_in_context() Currently __perf_install_in_context() will try and schedule in the event irrespective of our event scheduling rules, that is, we try to schedule CPU-pinned, TASK-pinned, CPU-flexible, TASK-flexible, but when creating a new event we simply try and schedule it on top of whatever is already on the PMU, this can lead to errors for pinned events. Therefore, simplify things and simply schedule everything out, add the event to the corresponding context and schedule everything back in. This also nicely handles the case where with __ARCH_WANT_INTERRUPTS_ON_CTXSW the IPI can come right in the middle of schedule, before we managed to call perf_event_task_sched_in(). Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.870894224@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 80 +++++++++++++++++++++++----------------------------- 1 file changed, 35 insertions(+), 45 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 66b3dd809409..60b333ae0bcf 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1469,8 +1469,12 @@ static void add_event_to_ctx(struct perf_event *event, event->tstamp_stopped = tstamp; } -static void perf_event_context_sched_in(struct perf_event_context *ctx, - struct task_struct *tsk); +static void task_ctx_sched_out(struct perf_event_context *ctx); +static void +ctx_sched_in(struct perf_event_context *ctx, + struct perf_cpu_context *cpuctx, + enum event_type_t event_type, + struct task_struct *task); /* * Cross CPU call to install and enable a performance event @@ -1481,20 +1485,31 @@ static int __perf_install_in_context(void *info) { struct perf_event *event = info; struct perf_event_context *ctx = event->ctx; - struct perf_event *leader = event->group_leader; struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); - int err; + struct perf_event_context *task_ctx = cpuctx->task_ctx; + struct task_struct *task = current; + + perf_ctx_lock(cpuctx, cpuctx->task_ctx); + perf_pmu_disable(cpuctx->ctx.pmu); /* - * In case we're installing a new context to an already running task, - * could also happen before perf_event_task_sched_in() on architectures - * which do context switches with IRQs enabled. + * If there was an active task_ctx schedule it out. */ - if (ctx->task && !cpuctx->task_ctx) - perf_event_context_sched_in(ctx, ctx->task); + if (task_ctx) { + task_ctx_sched_out(task_ctx); + /* + * If the context we're installing events in is not the + * active task_ctx, flip them. + */ + if (ctx->task && task_ctx != ctx) { + raw_spin_unlock(&cpuctx->ctx.lock); + raw_spin_lock(&ctx->lock); + cpuctx->task_ctx = task_ctx = ctx; + } + task = task_ctx->task; + } + cpu_ctx_sched_out(cpuctx, EVENT_ALL); - raw_spin_lock(&ctx->lock); - ctx->is_active = 1; update_context_time(ctx); /* * update cgrp time only if current cgrp @@ -1505,43 +1520,18 @@ static int __perf_install_in_context(void *info) add_event_to_ctx(event, ctx); - if (!event_filter_match(event)) - goto unlock; - - /* - * Don't put the event on if it is disabled or if - * it is in a group and the group isn't on. - */ - if (event->state != PERF_EVENT_STATE_INACTIVE || - (leader != event && leader->state != PERF_EVENT_STATE_ACTIVE)) - goto unlock; - /* - * An exclusive event can't go on if there are already active - * hardware events, and no hardware event can go on if there - * is already an exclusive event on. + * Schedule everything back in */ - if (!group_can_go_on(event, cpuctx, 1)) - err = -EEXIST; - else - err = event_sched_in(event, cpuctx, ctx); - - if (err) { - /* - * This event couldn't go on. If it is in a group - * then we have to pull the whole group off. - * If the event group is pinned then put it in error state. - */ - if (leader != event) - group_sched_out(leader, cpuctx, ctx); - if (leader->attr.pinned) { - update_group_times(leader); - leader->state = PERF_EVENT_STATE_ERROR; - } - } + cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task); + if (task_ctx) + ctx_sched_in(task_ctx, cpuctx, EVENT_PINNED, task); + cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task); + if (task_ctx) + ctx_sched_in(task_ctx, cpuctx, EVENT_FLEXIBLE, task); -unlock: - raw_spin_unlock(&ctx->lock); + perf_pmu_enable(cpuctx->ctx.pmu); + perf_ctx_unlock(cpuctx, task_ctx); return 0; } -- cgit v1.2.3 From db24d33e08b88e990991760a44d72006a5dc6102 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:45 +0200 Subject: perf: Change and simplify ctx::is_active semantics Instead of tracking if a context is active or not, track which events of the context are active. By making it a bitmask of EVENT_PINNED|EVENT_FLEXIBLE we can simplify some of the scheduling routines since it can avoid adding events that are already active. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.930282378@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 60b333ae0bcf..71c2d44ff95d 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1763,8 +1763,9 @@ static void ctx_sched_out(struct perf_event_context *ctx, enum event_type_t event_type) { struct perf_event *event; + int is_active = ctx->is_active; - ctx->is_active = 0; + ctx->is_active &= ~event_type; if (likely(!ctx->nr_events)) return; @@ -1774,12 +1775,12 @@ static void ctx_sched_out(struct perf_event_context *ctx, return; perf_pmu_disable(ctx->pmu); - if (event_type & EVENT_PINNED) { + if ((is_active & EVENT_PINNED) && (event_type & EVENT_PINNED)) { list_for_each_entry(event, &ctx->pinned_groups, group_entry) group_sched_out(event, cpuctx, ctx); } - if (event_type & EVENT_FLEXIBLE) { + if ((is_active & EVENT_FLEXIBLE) && (event_type & EVENT_FLEXIBLE)) { list_for_each_entry(event, &ctx->flexible_groups, group_entry) group_sched_out(event, cpuctx, ctx); } @@ -2058,8 +2059,9 @@ ctx_sched_in(struct perf_event_context *ctx, struct task_struct *task) { u64 now; + int is_active = ctx->is_active; - ctx->is_active = 1; + ctx->is_active |= event_type; if (likely(!ctx->nr_events)) return; @@ -2070,11 +2072,11 @@ ctx_sched_in(struct perf_event_context *ctx, * First go through the list and put on any pinned groups * in order to give them the best chance of going on. */ - if (event_type & EVENT_PINNED) + if (!(is_active & EVENT_PINNED) && (event_type & EVENT_PINNED)) ctx_pinned_sched_in(ctx, cpuctx); /* Then walk through the lower prio flexible groups */ - if (event_type & EVENT_FLEXIBLE) + if (!(is_active & EVENT_FLEXIBLE) && (event_type & EVENT_FLEXIBLE)) ctx_flexible_sched_in(ctx, cpuctx); } -- cgit v1.2.3 From dce5855bba5df9e87bb04584d505c1f1b103c652 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:46 +0200 Subject: perf: Collect the schedule-in rules in one function This was scattered out - refactor it into a single function. No change in functionality. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192141.979862055@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 71c2d44ff95d..802f3b24eeef 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1476,6 +1476,18 @@ ctx_sched_in(struct perf_event_context *ctx, enum event_type_t event_type, struct task_struct *task); +static void perf_event_sched_in(struct perf_cpu_context *cpuctx, + struct perf_event_context *ctx, + struct task_struct *task) +{ + cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task); + if (ctx) + ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task); + cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task); + if (ctx) + ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task); +} + /* * Cross CPU call to install and enable a performance event * @@ -1523,12 +1535,7 @@ static int __perf_install_in_context(void *info) /* * Schedule everything back in */ - cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task); - if (task_ctx) - ctx_sched_in(task_ctx, cpuctx, EVENT_PINNED, task); - cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task); - if (task_ctx) - ctx_sched_in(task_ctx, cpuctx, EVENT_FLEXIBLE, task); + perf_event_sched_in(cpuctx, task_ctx, task); perf_pmu_enable(cpuctx->ctx.pmu); perf_ctx_unlock(cpuctx, task_ctx); @@ -2107,9 +2114,7 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx, */ cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); - ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task); - cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task); - ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task); + perf_event_sched_in(cpuctx, ctx, task); cpuctx->task_ctx = ctx; @@ -2347,9 +2352,7 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) if (ctx) rotate_ctx(ctx); - cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, current); - if (ctx) - ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, current); + perf_event_sched_in(cpuctx, ctx, current); done: if (remove) -- cgit v1.2.3 From e03a9a55b4e45377af9ca3d464135f9ea280b8f8 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:47 +0200 Subject: perf: Change close() semantics for group events In order to always call list_del_event() on the correct cpu if the event is part of an active context and avoid having to do two IPIs, change the close() semantics slightly. The current perf_event_disable() call would disable a whole group if the event that's being closed is the group leader, whereas the new code keeps the group siblings enabled. People should not rely on this behaviour and I don't think they do, but in case we find they do, the fix is easy and we have to take the double IPI cost. Signed-off-by: Peter Zijlstra Cc: Vince Weaver Link: http://lkml.kernel.org/r/20110409192142.038377551@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 802f3b24eeef..c378062da277 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -2920,12 +2920,6 @@ int perf_event_release_kernel(struct perf_event *event) { struct perf_event_context *ctx = event->ctx; - /* - * Remove from the PMU, can't get re-enabled since we got - * here because the last ref went. - */ - perf_event_disable(event); - WARN_ON_ONCE(ctx->parent_ctx); /* * There are two ways this annotation is useful: @@ -2942,8 +2936,8 @@ int perf_event_release_kernel(struct perf_event *event) mutex_lock_nested(&ctx->mutex, SINGLE_DEPTH_NESTING); raw_spin_lock_irq(&ctx->lock); perf_group_detach(event); - list_del_event(event, ctx); raw_spin_unlock_irq(&ctx->lock); + perf_remove_from_context(event); mutex_unlock(&ctx->mutex); free_event(event); -- cgit v1.2.3 From 64ce312618ef0e11d88def80effcefd1b59fdb1e Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 9 Apr 2011 21:17:48 +0200 Subject: perf: De-schedule a task context when removing the last event Since perf_install_in_context() will now install a context when we add the first event, we can de-schedule the context when the last event is removed. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110409192142.090431763@chello.nl Signed-off-by: Ingo Molnar --- kernel/events/core.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index c378062da277..cc5d57d1d0b6 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1120,6 +1120,10 @@ static int __perf_remove_from_context(void *info) raw_spin_lock(&ctx->lock); event_sched_out(event, cpuctx, ctx); list_del_event(event, ctx); + if (!ctx->nr_events && cpuctx->task_ctx == ctx) { + ctx->is_active = 0; + cpuctx->task_ctx = NULL; + } raw_spin_unlock(&ctx->lock); return 0; -- cgit v1.2.3 From b58f6b0dd3d677338b9065388cc2cc942b86338e Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 7 Jun 2011 00:23:28 +0200 Subject: perf, core: Fix initial task_ctx/event installation A lost Quilt refresh of 2c29ef0fef8 (perf: Simplify and fix __perf_install_in_context()) is causing grief and lockups, reported by Jiri Olsa. When installing an event in a task context, there's a number of issues: - there might not be an existing task context, in which case we should install the now current context; - there might already be a context, not the current one, in which case we should de-schedule the old and install the new; these cases were dealt with in the lost refresh, however there is one further case that was found in testing: - there might already be a context, the current one, in which case we should still de-schedule, and should take care to re-install it (note that task_ctx_sched_out() clears cpuctx->task_ctx). Reported-by: Jiri Olsa Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1307399008.2497.971.camel@laptop Signed-off-by: Ingo Molnar --- kernel/events/core.c | 28 +++++++++++++++++----------- 1 file changed, 17 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index ba89f40abe6a..5e8c7b1389bc 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1505,25 +1505,31 @@ static int __perf_install_in_context(void *info) struct perf_event_context *task_ctx = cpuctx->task_ctx; struct task_struct *task = current; - perf_ctx_lock(cpuctx, cpuctx->task_ctx); + perf_ctx_lock(cpuctx, task_ctx); perf_pmu_disable(cpuctx->ctx.pmu); /* * If there was an active task_ctx schedule it out. */ - if (task_ctx) { + if (task_ctx) task_ctx_sched_out(task_ctx); - /* - * If the context we're installing events in is not the - * active task_ctx, flip them. - */ - if (ctx->task && task_ctx != ctx) { - raw_spin_unlock(&cpuctx->ctx.lock); - raw_spin_lock(&ctx->lock); - cpuctx->task_ctx = task_ctx = ctx; - } + + /* + * If the context we're installing events in is not the + * active task_ctx, flip them. + */ + if (ctx->task && task_ctx != ctx) { + if (task_ctx) + raw_spin_unlock(&task_ctx->lock); + raw_spin_lock(&ctx->lock); + task_ctx = ctx; + } + + if (task_ctx) { + cpuctx->task_ctx = task_ctx; task = task_ctx->task; } + cpu_ctx_sched_out(cpuctx, EVENT_ALL); update_context_time(ctx); -- cgit v1.2.3 From 76369139ceb955deefc509e6e12ce9d6ce50ccab Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 19 May 2011 19:55:04 +0200 Subject: perf: Split up buffer handling from core code And create the internal perf events header. v2: Keep an internal inlined perf_output_copy() Signed-off-by: Frederic Weisbecker Acked-by: Peter Zijlstra Cc: Borislav Petkov Cc: Stephane Eranian Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Link: http://lkml.kernel.org/r/1305827704-5607-1-git-send-email-fweisbec@gmail.com [ v3: use clearer 'ring_buffer' and 'rb' naming ] Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 33 +-- kernel/events/Makefile | 2 +- kernel/events/core.c | 568 ++++++-------------------------------------- kernel/events/internal.h | 97 ++++++++ kernel/events/ring_buffer.c | 399 +++++++++++++++++++++++++++++++ 5 files changed, 572 insertions(+), 527 deletions(-) create mode 100644 kernel/events/internal.h create mode 100644 kernel/events/ring_buffer.c (limited to 'kernel') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 3412684ce5d5..779f6ed54d52 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -680,33 +680,6 @@ enum perf_event_active_state { }; struct file; - -#define PERF_BUFFER_WRITABLE 0x01 - -struct perf_buffer { - atomic_t refcount; - struct rcu_head rcu_head; -#ifdef CONFIG_PERF_USE_VMALLOC - struct work_struct work; - int page_order; /* allocation order */ -#endif - int nr_pages; /* nr of data pages */ - int writable; /* are we writable */ - - atomic_t poll; /* POLL_ for wakeups */ - - local_t head; /* write position */ - local_t nest; /* nested writers */ - local_t events; /* event limit */ - local_t wakeup; /* wakeup stamp */ - local_t lost; /* nr records lost */ - - long watermark; /* wakeup watermark */ - - struct perf_event_mmap_page *user_page; - void *data_pages[0]; -}; - struct perf_sample_data; typedef void (*perf_overflow_handler_t)(struct perf_event *, int, @@ -745,6 +718,8 @@ struct perf_cgroup { }; #endif +struct ring_buffer; + /** * struct perf_event - performance event kernel representation: */ @@ -834,7 +809,7 @@ struct perf_event { atomic_t mmap_count; int mmap_locked; struct user_struct *mmap_user; - struct perf_buffer *buffer; + struct ring_buffer *rb; /* poll related */ wait_queue_head_t waitq; @@ -945,7 +920,7 @@ struct perf_cpu_context { struct perf_output_handle { struct perf_event *event; - struct perf_buffer *buffer; + struct ring_buffer *rb; unsigned long wakeup; unsigned long size; void *addr; diff --git a/kernel/events/Makefile b/kernel/events/Makefile index 1ce23d3d8394..89e5e8aa4c36 100644 --- a/kernel/events/Makefile +++ b/kernel/events/Makefile @@ -2,5 +2,5 @@ ifdef CONFIG_FUNCTION_TRACER CFLAGS_REMOVE_core.o = -pg endif -obj-y := core.o +obj-y := core.o ring_buffer.o obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o diff --git a/kernel/events/core.c b/kernel/events/core.c index 5e8c7b1389bc..5e70f62752a2 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -36,6 +36,8 @@ #include #include +#include "internal.h" + #include struct remote_function_call { @@ -2886,7 +2888,7 @@ static void free_event_rcu(struct rcu_head *head) kfree(event); } -static void perf_buffer_put(struct perf_buffer *buffer); +static void ring_buffer_put(struct ring_buffer *rb); static void free_event(struct perf_event *event) { @@ -2909,9 +2911,9 @@ static void free_event(struct perf_event *event) } } - if (event->buffer) { - perf_buffer_put(event->buffer); - event->buffer = NULL; + if (event->rb) { + ring_buffer_put(event->rb); + event->rb = NULL; } if (is_cgroup_event(event)) @@ -3139,13 +3141,13 @@ perf_read(struct file *file, char __user *buf, size_t count, loff_t *ppos) static unsigned int perf_poll(struct file *file, poll_table *wait) { struct perf_event *event = file->private_data; - struct perf_buffer *buffer; + struct ring_buffer *rb; unsigned int events = POLL_HUP; rcu_read_lock(); - buffer = rcu_dereference(event->buffer); - if (buffer) - events = atomic_xchg(&buffer->poll, 0); + rb = rcu_dereference(event->rb); + if (rb) + events = atomic_xchg(&rb->poll, 0); rcu_read_unlock(); poll_wait(file, &event->waitq, wait); @@ -3356,14 +3358,14 @@ static int perf_event_index(struct perf_event *event) void perf_event_update_userpage(struct perf_event *event) { struct perf_event_mmap_page *userpg; - struct perf_buffer *buffer; + struct ring_buffer *rb; rcu_read_lock(); - buffer = rcu_dereference(event->buffer); - if (!buffer) + rb = rcu_dereference(event->rb); + if (!rb) goto unlock; - userpg = buffer->user_page; + userpg = rb->user_page; /* * Disable preemption so as to not let the corresponding user-space @@ -3390,220 +3392,10 @@ unlock: rcu_read_unlock(); } -static unsigned long perf_data_size(struct perf_buffer *buffer); - -static void -perf_buffer_init(struct perf_buffer *buffer, long watermark, int flags) -{ - long max_size = perf_data_size(buffer); - - if (watermark) - buffer->watermark = min(max_size, watermark); - - if (!buffer->watermark) - buffer->watermark = max_size / 2; - - if (flags & PERF_BUFFER_WRITABLE) - buffer->writable = 1; - - atomic_set(&buffer->refcount, 1); -} - -#ifndef CONFIG_PERF_USE_VMALLOC - -/* - * Back perf_mmap() with regular GFP_KERNEL-0 pages. - */ - -static struct page * -perf_mmap_to_page(struct perf_buffer *buffer, unsigned long pgoff) -{ - if (pgoff > buffer->nr_pages) - return NULL; - - if (pgoff == 0) - return virt_to_page(buffer->user_page); - - return virt_to_page(buffer->data_pages[pgoff - 1]); -} - -static void *perf_mmap_alloc_page(int cpu) -{ - struct page *page; - int node; - - node = (cpu == -1) ? cpu : cpu_to_node(cpu); - page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0); - if (!page) - return NULL; - - return page_address(page); -} - -static struct perf_buffer * -perf_buffer_alloc(int nr_pages, long watermark, int cpu, int flags) -{ - struct perf_buffer *buffer; - unsigned long size; - int i; - - size = sizeof(struct perf_buffer); - size += nr_pages * sizeof(void *); - - buffer = kzalloc(size, GFP_KERNEL); - if (!buffer) - goto fail; - - buffer->user_page = perf_mmap_alloc_page(cpu); - if (!buffer->user_page) - goto fail_user_page; - - for (i = 0; i < nr_pages; i++) { - buffer->data_pages[i] = perf_mmap_alloc_page(cpu); - if (!buffer->data_pages[i]) - goto fail_data_pages; - } - - buffer->nr_pages = nr_pages; - - perf_buffer_init(buffer, watermark, flags); - - return buffer; - -fail_data_pages: - for (i--; i >= 0; i--) - free_page((unsigned long)buffer->data_pages[i]); - - free_page((unsigned long)buffer->user_page); - -fail_user_page: - kfree(buffer); - -fail: - return NULL; -} - -static void perf_mmap_free_page(unsigned long addr) -{ - struct page *page = virt_to_page((void *)addr); - - page->mapping = NULL; - __free_page(page); -} - -static void perf_buffer_free(struct perf_buffer *buffer) -{ - int i; - - perf_mmap_free_page((unsigned long)buffer->user_page); - for (i = 0; i < buffer->nr_pages; i++) - perf_mmap_free_page((unsigned long)buffer->data_pages[i]); - kfree(buffer); -} - -static inline int page_order(struct perf_buffer *buffer) -{ - return 0; -} - -#else - -/* - * Back perf_mmap() with vmalloc memory. - * - * Required for architectures that have d-cache aliasing issues. - */ - -static inline int page_order(struct perf_buffer *buffer) -{ - return buffer->page_order; -} - -static struct page * -perf_mmap_to_page(struct perf_buffer *buffer, unsigned long pgoff) -{ - if (pgoff > (1UL << page_order(buffer))) - return NULL; - - return vmalloc_to_page((void *)buffer->user_page + pgoff * PAGE_SIZE); -} - -static void perf_mmap_unmark_page(void *addr) -{ - struct page *page = vmalloc_to_page(addr); - - page->mapping = NULL; -} - -static void perf_buffer_free_work(struct work_struct *work) -{ - struct perf_buffer *buffer; - void *base; - int i, nr; - - buffer = container_of(work, struct perf_buffer, work); - nr = 1 << page_order(buffer); - - base = buffer->user_page; - for (i = 0; i < nr + 1; i++) - perf_mmap_unmark_page(base + (i * PAGE_SIZE)); - - vfree(base); - kfree(buffer); -} - -static void perf_buffer_free(struct perf_buffer *buffer) -{ - schedule_work(&buffer->work); -} - -static struct perf_buffer * -perf_buffer_alloc(int nr_pages, long watermark, int cpu, int flags) -{ - struct perf_buffer *buffer; - unsigned long size; - void *all_buf; - - size = sizeof(struct perf_buffer); - size += sizeof(void *); - - buffer = kzalloc(size, GFP_KERNEL); - if (!buffer) - goto fail; - - INIT_WORK(&buffer->work, perf_buffer_free_work); - - all_buf = vmalloc_user((nr_pages + 1) * PAGE_SIZE); - if (!all_buf) - goto fail_all_buf; - - buffer->user_page = all_buf; - buffer->data_pages[0] = all_buf + PAGE_SIZE; - buffer->page_order = ilog2(nr_pages); - buffer->nr_pages = 1; - - perf_buffer_init(buffer, watermark, flags); - - return buffer; - -fail_all_buf: - kfree(buffer); - -fail: - return NULL; -} - -#endif - -static unsigned long perf_data_size(struct perf_buffer *buffer) -{ - return buffer->nr_pages << (PAGE_SHIFT + page_order(buffer)); -} - static int perf_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf) { struct perf_event *event = vma->vm_file->private_data; - struct perf_buffer *buffer; + struct ring_buffer *rb; int ret = VM_FAULT_SIGBUS; if (vmf->flags & FAULT_FLAG_MKWRITE) { @@ -3613,14 +3405,14 @@ static int perf_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf) } rcu_read_lock(); - buffer = rcu_dereference(event->buffer); - if (!buffer) + rb = rcu_dereference(event->rb); + if (!rb) goto unlock; if (vmf->pgoff && (vmf->flags & FAULT_FLAG_WRITE)) goto unlock; - vmf->page = perf_mmap_to_page(buffer, vmf->pgoff); + vmf->page = perf_mmap_to_page(rb, vmf->pgoff); if (!vmf->page) goto unlock; @@ -3635,35 +3427,35 @@ unlock: return ret; } -static void perf_buffer_free_rcu(struct rcu_head *rcu_head) +static void rb_free_rcu(struct rcu_head *rcu_head) { - struct perf_buffer *buffer; + struct ring_buffer *rb; - buffer = container_of(rcu_head, struct perf_buffer, rcu_head); - perf_buffer_free(buffer); + rb = container_of(rcu_head, struct ring_buffer, rcu_head); + rb_free(rb); } -static struct perf_buffer *perf_buffer_get(struct perf_event *event) +static struct ring_buffer *ring_buffer_get(struct perf_event *event) { - struct perf_buffer *buffer; + struct ring_buffer *rb; rcu_read_lock(); - buffer = rcu_dereference(event->buffer); - if (buffer) { - if (!atomic_inc_not_zero(&buffer->refcount)) - buffer = NULL; + rb = rcu_dereference(event->rb); + if (rb) { + if (!atomic_inc_not_zero(&rb->refcount)) + rb = NULL; } rcu_read_unlock(); - return buffer; + return rb; } -static void perf_buffer_put(struct perf_buffer *buffer) +static void ring_buffer_put(struct ring_buffer *rb) { - if (!atomic_dec_and_test(&buffer->refcount)) + if (!atomic_dec_and_test(&rb->refcount)) return; - call_rcu(&buffer->rcu_head, perf_buffer_free_rcu); + call_rcu(&rb->rcu_head, rb_free_rcu); } static void perf_mmap_open(struct vm_area_struct *vma) @@ -3678,16 +3470,16 @@ static void perf_mmap_close(struct vm_area_struct *vma) struct perf_event *event = vma->vm_file->private_data; if (atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex)) { - unsigned long size = perf_data_size(event->buffer); + unsigned long size = perf_data_size(event->rb); struct user_struct *user = event->mmap_user; - struct perf_buffer *buffer = event->buffer; + struct ring_buffer *rb = event->rb; atomic_long_sub((size >> PAGE_SHIFT) + 1, &user->locked_vm); vma->vm_mm->locked_vm -= event->mmap_locked; - rcu_assign_pointer(event->buffer, NULL); + rcu_assign_pointer(event->rb, NULL); mutex_unlock(&event->mmap_mutex); - perf_buffer_put(buffer); + ring_buffer_put(rb); free_uid(user); } } @@ -3705,7 +3497,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) unsigned long user_locked, user_lock_limit; struct user_struct *user = current_user(); unsigned long locked, lock_limit; - struct perf_buffer *buffer; + struct ring_buffer *rb; unsigned long vma_size; unsigned long nr_pages; long user_extra, extra; @@ -3714,7 +3506,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) /* * Don't allow mmap() of inherited per-task counters. This would * create a performance issue due to all children writing to the - * same buffer. + * same rb. */ if (event->cpu == -1 && event->attr.inherit) return -EINVAL; @@ -3726,7 +3518,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) nr_pages = (vma_size / PAGE_SIZE) - 1; /* - * If we have buffer pages ensure they're a power-of-two number, so we + * If we have rb pages ensure they're a power-of-two number, so we * can do bitmasks instead of modulo. */ if (nr_pages != 0 && !is_power_of_2(nr_pages)) @@ -3740,9 +3532,9 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) WARN_ON_ONCE(event->ctx->parent_ctx); mutex_lock(&event->mmap_mutex); - if (event->buffer) { - if (event->buffer->nr_pages == nr_pages) - atomic_inc(&event->buffer->refcount); + if (event->rb) { + if (event->rb->nr_pages == nr_pages) + atomic_inc(&event->rb->refcount); else ret = -EINVAL; goto unlock; @@ -3772,18 +3564,18 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) goto unlock; } - WARN_ON(event->buffer); + WARN_ON(event->rb); if (vma->vm_flags & VM_WRITE) - flags |= PERF_BUFFER_WRITABLE; + flags |= RING_BUFFER_WRITABLE; - buffer = perf_buffer_alloc(nr_pages, event->attr.wakeup_watermark, + rb = rb_alloc(nr_pages, event->attr.wakeup_watermark, event->cpu, flags); - if (!buffer) { + if (!rb) { ret = -ENOMEM; goto unlock; } - rcu_assign_pointer(event->buffer, buffer); + rcu_assign_pointer(event->rb, rb); atomic_long_add(user_extra, &user->locked_vm); event->mmap_locked = extra; @@ -3882,117 +3674,6 @@ int perf_unregister_guest_info_callbacks(struct perf_guest_info_callbacks *cbs) } EXPORT_SYMBOL_GPL(perf_unregister_guest_info_callbacks); -/* - * Output - */ -static bool perf_output_space(struct perf_buffer *buffer, unsigned long tail, - unsigned long offset, unsigned long head) -{ - unsigned long mask; - - if (!buffer->writable) - return true; - - mask = perf_data_size(buffer) - 1; - - offset = (offset - tail) & mask; - head = (head - tail) & mask; - - if ((int)(head - offset) < 0) - return false; - - return true; -} - -static void perf_output_wakeup(struct perf_output_handle *handle) -{ - atomic_set(&handle->buffer->poll, POLL_IN); - - if (handle->nmi) { - handle->event->pending_wakeup = 1; - irq_work_queue(&handle->event->pending); - } else - perf_event_wakeup(handle->event); -} - -/* - * We need to ensure a later event_id doesn't publish a head when a former - * event isn't done writing. However since we need to deal with NMIs we - * cannot fully serialize things. - * - * We only publish the head (and generate a wakeup) when the outer-most - * event completes. - */ -static void perf_output_get_handle(struct perf_output_handle *handle) -{ - struct perf_buffer *buffer = handle->buffer; - - preempt_disable(); - local_inc(&buffer->nest); - handle->wakeup = local_read(&buffer->wakeup); -} - -static void perf_output_put_handle(struct perf_output_handle *handle) -{ - struct perf_buffer *buffer = handle->buffer; - unsigned long head; - -again: - head = local_read(&buffer->head); - - /* - * IRQ/NMI can happen here, which means we can miss a head update. - */ - - if (!local_dec_and_test(&buffer->nest)) - goto out; - - /* - * Publish the known good head. Rely on the full barrier implied - * by atomic_dec_and_test() order the buffer->head read and this - * write. - */ - buffer->user_page->data_head = head; - - /* - * Now check if we missed an update, rely on the (compiler) - * barrier in atomic_dec_and_test() to re-read buffer->head. - */ - if (unlikely(head != local_read(&buffer->head))) { - local_inc(&buffer->nest); - goto again; - } - - if (handle->wakeup != local_read(&buffer->wakeup)) - perf_output_wakeup(handle); - -out: - preempt_enable(); -} - -__always_inline void perf_output_copy(struct perf_output_handle *handle, - const void *buf, unsigned int len) -{ - do { - unsigned long size = min_t(unsigned long, handle->size, len); - - memcpy(handle->addr, buf, size); - - len -= size; - handle->addr += size; - buf += size; - handle->size -= size; - if (!handle->size) { - struct perf_buffer *buffer = handle->buffer; - - handle->page++; - handle->page &= buffer->nr_pages - 1; - handle->addr = buffer->data_pages[handle->page]; - handle->size = PAGE_SIZE << page_order(buffer); - } - } while (len); -} - static void __perf_event_header__init_id(struct perf_event_header *header, struct perf_sample_data *data, struct perf_event *event) @@ -4023,9 +3704,9 @@ static void __perf_event_header__init_id(struct perf_event_header *header, } } -static void perf_event_header__init_id(struct perf_event_header *header, - struct perf_sample_data *data, - struct perf_event *event) +void perf_event_header__init_id(struct perf_event_header *header, + struct perf_sample_data *data, + struct perf_event *event) { if (event->attr.sample_id_all) __perf_event_header__init_id(header, data, event); @@ -4052,121 +3733,14 @@ static void __perf_event__output_id_sample(struct perf_output_handle *handle, perf_output_put(handle, data->cpu_entry); } -static void perf_event__output_id_sample(struct perf_event *event, - struct perf_output_handle *handle, - struct perf_sample_data *sample) +void perf_event__output_id_sample(struct perf_event *event, + struct perf_output_handle *handle, + struct perf_sample_data *sample) { if (event->attr.sample_id_all) __perf_event__output_id_sample(handle, sample); } -int perf_output_begin(struct perf_output_handle *handle, - struct perf_event *event, unsigned int size, - int nmi, int sample) -{ - struct perf_buffer *buffer; - unsigned long tail, offset, head; - int have_lost; - struct perf_sample_data sample_data; - struct { - struct perf_event_header header; - u64 id; - u64 lost; - } lost_event; - - rcu_read_lock(); - /* - * For inherited events we send all the output towards the parent. - */ - if (event->parent) - event = event->parent; - - buffer = rcu_dereference(event->buffer); - if (!buffer) - goto out; - - handle->buffer = buffer; - handle->event = event; - handle->nmi = nmi; - handle->sample = sample; - - if (!buffer->nr_pages) - goto out; - - have_lost = local_read(&buffer->lost); - if (have_lost) { - lost_event.header.size = sizeof(lost_event); - perf_event_header__init_id(&lost_event.header, &sample_data, - event); - size += lost_event.header.size; - } - - perf_output_get_handle(handle); - - do { - /* - * Userspace could choose to issue a mb() before updating the - * tail pointer. So that all reads will be completed before the - * write is issued. - */ - tail = ACCESS_ONCE(buffer->user_page->data_tail); - smp_rmb(); - offset = head = local_read(&buffer->head); - head += size; - if (unlikely(!perf_output_space(buffer, tail, offset, head))) - goto fail; - } while (local_cmpxchg(&buffer->head, offset, head) != offset); - - if (head - local_read(&buffer->wakeup) > buffer->watermark) - local_add(buffer->watermark, &buffer->wakeup); - - handle->page = offset >> (PAGE_SHIFT + page_order(buffer)); - handle->page &= buffer->nr_pages - 1; - handle->size = offset & ((PAGE_SIZE << page_order(buffer)) - 1); - handle->addr = buffer->data_pages[handle->page]; - handle->addr += handle->size; - handle->size = (PAGE_SIZE << page_order(buffer)) - handle->size; - - if (have_lost) { - lost_event.header.type = PERF_RECORD_LOST; - lost_event.header.misc = 0; - lost_event.id = event->id; - lost_event.lost = local_xchg(&buffer->lost, 0); - - perf_output_put(handle, lost_event); - perf_event__output_id_sample(event, handle, &sample_data); - } - - return 0; - -fail: - local_inc(&buffer->lost); - perf_output_put_handle(handle); -out: - rcu_read_unlock(); - - return -ENOSPC; -} - -void perf_output_end(struct perf_output_handle *handle) -{ - struct perf_event *event = handle->event; - struct perf_buffer *buffer = handle->buffer; - - int wakeup_events = event->attr.wakeup_events; - - if (handle->sample && wakeup_events) { - int events = local_inc_return(&buffer->events); - if (events >= wakeup_events) { - local_sub(wakeup_events, &buffer->events); - local_inc(&buffer->wakeup); - } - } - - perf_output_put_handle(handle); - rcu_read_unlock(); -} - static void perf_output_read_one(struct perf_output_handle *handle, struct perf_event *event, u64 enabled, u64 running) @@ -4187,7 +3761,7 @@ static void perf_output_read_one(struct perf_output_handle *handle, if (read_format & PERF_FORMAT_ID) values[n++] = primary_event_id(event); - perf_output_copy(handle, values, n * sizeof(u64)); + __output_copy(handle, values, n * sizeof(u64)); } /* @@ -4217,7 +3791,7 @@ static void perf_output_read_group(struct perf_output_handle *handle, if (read_format & PERF_FORMAT_ID) values[n++] = primary_event_id(leader); - perf_output_copy(handle, values, n * sizeof(u64)); + __output_copy(handle, values, n * sizeof(u64)); list_for_each_entry(sub, &leader->sibling_list, group_entry) { n = 0; @@ -4229,7 +3803,7 @@ static void perf_output_read_group(struct perf_output_handle *handle, if (read_format & PERF_FORMAT_ID) values[n++] = primary_event_id(sub); - perf_output_copy(handle, values, n * sizeof(u64)); + __output_copy(handle, values, n * sizeof(u64)); } } @@ -4309,7 +3883,7 @@ void perf_output_sample(struct perf_output_handle *handle, size *= sizeof(u64); - perf_output_copy(handle, data->callchain, size); + __output_copy(handle, data->callchain, size); } else { u64 nr = 0; perf_output_put(handle, nr); @@ -4319,8 +3893,8 @@ void perf_output_sample(struct perf_output_handle *handle, if (sample_type & PERF_SAMPLE_RAW) { if (data->raw) { perf_output_put(handle, data->raw->size); - perf_output_copy(handle, data->raw->data, - data->raw->size); + __output_copy(handle, data->raw->data, + data->raw->size); } else { struct { u32 size; @@ -4617,7 +4191,7 @@ static void perf_event_comm_output(struct perf_event *event, comm_event->event_id.tid = perf_event_tid(event, comm_event->task); perf_output_put(&handle, comm_event->event_id); - perf_output_copy(&handle, comm_event->comm, + __output_copy(&handle, comm_event->comm, comm_event->comm_size); perf_event__output_id_sample(event, &handle, &sample); @@ -4763,7 +4337,7 @@ static void perf_event_mmap_output(struct perf_event *event, mmap_event->event_id.tid = perf_event_tid(event, current); perf_output_put(&handle, mmap_event->event_id); - perf_output_copy(&handle, mmap_event->file_name, + __output_copy(&handle, mmap_event->file_name, mmap_event->file_size); perf_event__output_id_sample(event, &handle, &sample); @@ -4819,7 +4393,7 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event) if (file) { /* - * d_path works from the end of the buffer backwards, so we + * d_path works from the end of the rb backwards, so we * need to add enough zero bytes after the string to handle * the 64bit alignment we do later. */ @@ -6346,7 +5920,7 @@ err_size: static int perf_event_set_output(struct perf_event *event, struct perf_event *output_event) { - struct perf_buffer *buffer = NULL, *old_buffer = NULL; + struct ring_buffer *rb = NULL, *old_rb = NULL; int ret = -EINVAL; if (!output_event) @@ -6363,7 +5937,7 @@ perf_event_set_output(struct perf_event *event, struct perf_event *output_event) goto out; /* - * If its not a per-cpu buffer, it must be the same task. + * If its not a per-cpu rb, it must be the same task. */ if (output_event->cpu == -1 && output_event->ctx != event->ctx) goto out; @@ -6375,20 +5949,20 @@ set: goto unlock; if (output_event) { - /* get the buffer we want to redirect to */ - buffer = perf_buffer_get(output_event); - if (!buffer) + /* get the rb we want to redirect to */ + rb = ring_buffer_get(output_event); + if (!rb) goto unlock; } - old_buffer = event->buffer; - rcu_assign_pointer(event->buffer, buffer); + old_rb = event->rb; + rcu_assign_pointer(event->rb, rb); ret = 0; unlock: mutex_unlock(&event->mmap_mutex); - if (old_buffer) - perf_buffer_put(old_buffer); + if (old_rb) + ring_buffer_put(old_rb); out: return ret; } diff --git a/kernel/events/internal.h b/kernel/events/internal.h new file mode 100644 index 000000000000..114f27f3a624 --- /dev/null +++ b/kernel/events/internal.h @@ -0,0 +1,97 @@ +#ifndef _KERNEL_EVENTS_INTERNAL_H +#define _KERNEL_EVENTS_INTERNAL_H + +#define RING_BUFFER_WRITABLE 0x01 + +struct ring_buffer { + atomic_t refcount; + struct rcu_head rcu_head; +#ifdef CONFIG_PERF_USE_VMALLOC + struct work_struct work; + int page_order; /* allocation order */ +#endif + int nr_pages; /* nr of data pages */ + int writable; /* are we writable */ + + atomic_t poll; /* POLL_ for wakeups */ + + local_t head; /* write position */ + local_t nest; /* nested writers */ + local_t events; /* event limit */ + local_t wakeup; /* wakeup stamp */ + local_t lost; /* nr records lost */ + + long watermark; /* wakeup watermark */ + + struct perf_event_mmap_page *user_page; + void *data_pages[0]; +}; + + +extern void rb_free(struct ring_buffer *rb); +extern struct ring_buffer * +rb_alloc(int nr_pages, long watermark, int cpu, int flags); +extern void perf_event_wakeup(struct perf_event *event); + +extern void +perf_event_header__init_id(struct perf_event_header *header, + struct perf_sample_data *data, + struct perf_event *event); +extern void +perf_event__output_id_sample(struct perf_event *event, + struct perf_output_handle *handle, + struct perf_sample_data *sample); + +extern struct page * +perf_mmap_to_page(struct ring_buffer *rb, unsigned long pgoff); + +#ifdef CONFIG_PERF_USE_VMALLOC +/* + * Back perf_mmap() with vmalloc memory. + * + * Required for architectures that have d-cache aliasing issues. + */ + +static inline int page_order(struct ring_buffer *rb) +{ + return rb->page_order; +} + +#else + +static inline int page_order(struct ring_buffer *rb) +{ + return 0; +} +#endif + +static unsigned long perf_data_size(struct ring_buffer *rb) +{ + return rb->nr_pages << (PAGE_SHIFT + page_order(rb)); +} + +static inline void +__output_copy(struct perf_output_handle *handle, + const void *buf, unsigned int len) +{ + do { + unsigned long size = min_t(unsigned long, handle->size, len); + + memcpy(handle->addr, buf, size); + + len -= size; + handle->addr += size; + buf += size; + handle->size -= size; + if (!handle->size) { + struct ring_buffer *rb = handle->rb; + + handle->page++; + handle->page &= rb->nr_pages - 1; + handle->addr = rb->data_pages[handle->page]; + handle->size = PAGE_SIZE << page_order(rb); + } + } while (len); +} + +#endif /* _KERNEL_EVENTS_INTERNAL_H */ diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c new file mode 100644 index 000000000000..fde52595d8f7 --- /dev/null +++ b/kernel/events/ring_buffer.c @@ -0,0 +1,399 @@ +/* + * Performance events ring-buffer code: + * + * Copyright (C) 2008 Thomas Gleixner + * Copyright (C) 2008-2011 Red Hat, Inc., Ingo Molnar + * Copyright (C) 2008-2011 Red Hat, Inc., Peter Zijlstra + * Copyright © 2009 Paul Mackerras, IBM Corp. + * + * For licensing details see kernel-base/COPYING + */ + +#include +#include +#include + +#include "internal.h" + +static bool perf_output_space(struct ring_buffer *rb, unsigned long tail, + unsigned long offset, unsigned long head) +{ + unsigned long mask; + + if (!rb->writable) + return true; + + mask = perf_data_size(rb) - 1; + + offset = (offset - tail) & mask; + head = (head - tail) & mask; + + if ((int)(head - offset) < 0) + return false; + + return true; +} + +static void perf_output_wakeup(struct perf_output_handle *handle) +{ + atomic_set(&handle->rb->poll, POLL_IN); + + if (handle->nmi) { + handle->event->pending_wakeup = 1; + irq_work_queue(&handle->event->pending); + } else + perf_event_wakeup(handle->event); +} + +/* + * We need to ensure a later event_id doesn't publish a head when a former + * event isn't done writing. However since we need to deal with NMIs we + * cannot fully serialize things. + * + * We only publish the head (and generate a wakeup) when the outer-most + * event completes. + */ +static void perf_output_get_handle(struct perf_output_handle *handle) +{ + struct ring_buffer *rb = handle->rb; + + preempt_disable(); + local_inc(&rb->nest); + handle->wakeup = local_read(&rb->wakeup); +} + +static void perf_output_put_handle(struct perf_output_handle *handle) +{ + struct ring_buffer *rb = handle->rb; + unsigned long head; + +again: + head = local_read(&rb->head); + + /* + * IRQ/NMI can happen here, which means we can miss a head update. + */ + + if (!local_dec_and_test(&rb->nest)) + goto out; + + /* + * Publish the known good head. Rely on the full barrier implied + * by atomic_dec_and_test() order the rb->head read and this + * write. + */ + rb->user_page->data_head = head; + + /* + * Now check if we missed an update, rely on the (compiler) + * barrier in atomic_dec_and_test() to re-read rb->head. + */ + if (unlikely(head != local_read(&rb->head))) { + local_inc(&rb->nest); + goto again; + } + + if (handle->wakeup != local_read(&rb->wakeup)) + perf_output_wakeup(handle); + +out: + preempt_enable(); +} + +int perf_output_begin(struct perf_output_handle *handle, + struct perf_event *event, unsigned int size, + int nmi, int sample) +{ + struct ring_buffer *rb; + unsigned long tail, offset, head; + int have_lost; + struct perf_sample_data sample_data; + struct { + struct perf_event_header header; + u64 id; + u64 lost; + } lost_event; + + rcu_read_lock(); + /* + * For inherited events we send all the output towards the parent. + */ + if (event->parent) + event = event->parent; + + rb = rcu_dereference(event->rb); + if (!rb) + goto out; + + handle->rb = rb; + handle->event = event; + handle->nmi = nmi; + handle->sample = sample; + + if (!rb->nr_pages) + goto out; + + have_lost = local_read(&rb->lost); + if (have_lost) { + lost_event.header.size = sizeof(lost_event); + perf_event_header__init_id(&lost_event.header, &sample_data, + event); + size += lost_event.header.size; + } + + perf_output_get_handle(handle); + + do { + /* + * Userspace could choose to issue a mb() before updating the + * tail pointer. So that all reads will be completed before the + * write is issued. + */ + tail = ACCESS_ONCE(rb->user_page->data_tail); + smp_rmb(); + offset = head = local_read(&rb->head); + head += size; + if (unlikely(!perf_output_space(rb, tail, offset, head))) + goto fail; + } while (local_cmpxchg(&rb->head, offset, head) != offset); + + if (head - local_read(&rb->wakeup) > rb->watermark) + local_add(rb->watermark, &rb->wakeup); + + handle->page = offset >> (PAGE_SHIFT + page_order(rb)); + handle->page &= rb->nr_pages - 1; + handle->size = offset & ((PAGE_SIZE << page_order(rb)) - 1); + handle->addr = rb->data_pages[handle->page]; + handle->addr += handle->size; + handle->size = (PAGE_SIZE << page_order(rb)) - handle->size; + + if (have_lost) { + lost_event.header.type = PERF_RECORD_LOST; + lost_event.header.misc = 0; + lost_event.id = event->id; + lost_event.lost = local_xchg(&rb->lost, 0); + + perf_output_put(handle, lost_event); + perf_event__output_id_sample(event, handle, &sample_data); + } + + return 0; + +fail: + local_inc(&rb->lost); + perf_output_put_handle(handle); +out: + rcu_read_unlock(); + + return -ENOSPC; +} + +void perf_output_copy(struct perf_output_handle *handle, + const void *buf, unsigned int len) +{ + __output_copy(handle, buf, len); +} + +void perf_output_end(struct perf_output_handle *handle) +{ + struct perf_event *event = handle->event; + struct ring_buffer *rb = handle->rb; + + int wakeup_events = event->attr.wakeup_events; + + if (handle->sample && wakeup_events) { + int events = local_inc_return(&rb->events); + if (events >= wakeup_events) { + local_sub(wakeup_events, &rb->events); + local_inc(&rb->wakeup); + } + } + + perf_output_put_handle(handle); + rcu_read_unlock(); +} + +static void +ring_buffer_init(struct ring_buffer *rb, long watermark, int flags) +{ + long max_size = perf_data_size(rb); + + if (watermark) + rb->watermark = min(max_size, watermark); + + if (!rb->watermark) + rb->watermark = max_size / 2; + + if (flags & RING_BUFFER_WRITABLE) + rb->writable = 1; + + atomic_set(&rb->refcount, 1); +} + +#ifndef CONFIG_PERF_USE_VMALLOC + +/* + * Back perf_mmap() with regular GFP_KERNEL-0 pages. + */ + +struct page * +perf_mmap_to_page(struct ring_buffer *rb, unsigned long pgoff) +{ + if (pgoff > rb->nr_pages) + return NULL; + + if (pgoff == 0) + return virt_to_page(rb->user_page); + + return virt_to_page(rb->data_pages[pgoff - 1]); +} + +static void *perf_mmap_alloc_page(int cpu) +{ + struct page *page; + int node; + + node = (cpu == -1) ? cpu : cpu_to_node(cpu); + page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0); + if (!page) + return NULL; + + return page_address(page); +} + +struct ring_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags) +{ + struct ring_buffer *rb; + unsigned long size; + int i; + + size = sizeof(struct ring_buffer); + size += nr_pages * sizeof(void *); + + rb = kzalloc(size, GFP_KERNEL); + if (!rb) + goto fail; + + rb->user_page = perf_mmap_alloc_page(cpu); + if (!rb->user_page) + goto fail_user_page; + + for (i = 0; i < nr_pages; i++) { + rb->data_pages[i] = perf_mmap_alloc_page(cpu); + if (!rb->data_pages[i]) + goto fail_data_pages; + } + + rb->nr_pages = nr_pages; + + ring_buffer_init(rb, watermark, flags); + + return rb; + +fail_data_pages: + for (i--; i >= 0; i--) + free_page((unsigned long)rb->data_pages[i]); + + free_page((unsigned long)rb->user_page); + +fail_user_page: + kfree(rb); + +fail: + return NULL; +} + +static void perf_mmap_free_page(unsigned long addr) +{ + struct page *page = virt_to_page((void *)addr); + + page->mapping = NULL; + __free_page(page); +} + +void rb_free(struct ring_buffer *rb) +{ + int i; + + perf_mmap_free_page((unsigned long)rb->user_page); + for (i = 0; i < rb->nr_pages; i++) + perf_mmap_free_page((unsigned long)rb->data_pages[i]); + kfree(rb); +} + +#else + +struct page * +perf_mmap_to_page(struct ring_buffer *rb, unsigned long pgoff) +{ + if (pgoff > (1UL << page_order(rb))) + return NULL; + + return vmalloc_to_page((void *)rb->user_page + pgoff * PAGE_SIZE); +} + +static void perf_mmap_unmark_page(void *addr) +{ + struct page *page = vmalloc_to_page(addr); + + page->mapping = NULL; +} + +static void rb_free_work(struct work_struct *work) +{ + struct ring_buffer *rb; + void *base; + int i, nr; + + rb = container_of(work, struct ring_buffer, work); + nr = 1 << page_order(rb); + + base = rb->user_page; + for (i = 0; i < nr + 1; i++) + perf_mmap_unmark_page(base + (i * PAGE_SIZE)); + + vfree(base); + kfree(rb); +} + +void rb_free(struct ring_buffer *rb) +{ + schedule_work(&rb->work); +} + +struct ring_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags) +{ + struct ring_buffer *rb; + unsigned long size; + void *all_buf; + + size = sizeof(struct ring_buffer); + size += sizeof(void *); + + rb = kzalloc(size, GFP_KERNEL); + if (!rb) + goto fail; + + INIT_WORK(&rb->work, rb_free_work); + + all_buf = vmalloc_user((nr_pages + 1) * PAGE_SIZE); + if (!all_buf) + goto fail_all_buf; + + rb->user_page = all_buf; + rb->data_pages[0] = all_buf + PAGE_SIZE; + rb->page_order = ilog2(nr_pages); + rb->nr_pages = 1; + + ring_buffer_init(rb, watermark, flags); + + return rb; + +fail_all_buf: + kfree(rb); + +fail: + return NULL; +} + +#endif -- cgit v1.2.3 From e7e2ee89a9dbf48d70a922d5625cd7320a27cbff Mon Sep 17 00:00:00 2001 From: Vaibhav Nagarnaik Date: Tue, 10 May 2011 13:27:21 -0700 Subject: tracing: Schedule a delayed work to call wakeup() In using syscall tracing by concurrent processes, the wakeup() that is called in the event commit function causes contention on the spin lock of the waitqueue. I enabled sys_enter_getuid and sys_exit_getuid tracepoints, and by running getuid_microbench from autotest in parallel I found that the contention causes exponential latency increase in the tracing path. The autotest binary getuid_microbench calls getuid() in a tight loop for the given number of iterations and measures the average time required to complete a single invocation of syscall. The patch schedules a delayed work after 2 ms once an event commit calls to wake up the trace wait_queue. This removes the delay caused by contention on spin lock in wakeup() and amortizes the wakeup() calls scheduled over the 2 ms period. In the following example, the script enables the sys_enter_getuid and sys_exit_getuid tracepoints and runs the getuid_microbench in parallel with the given number of processes. The output clearly shows the latency increase caused by contentions. $ ~/getuid.sh 1 1000000 calls in 0.720974253 s (720.974253 ns/call) $ ~/getuid.sh 2 1000000 calls in 1.166457554 s (1166.457554 ns/call) 1000000 calls in 1.168933765 s (1168.933765 ns/call) $ ~/getuid.sh 3 1000000 calls in 1.783827516 s (1783.827516 ns/call) 1000000 calls in 1.795553270 s (1795.553270 ns/call) 1000000 calls in 1.796493376 s (1796.493376 ns/call) $ ~/getuid.sh 4 1000000 calls in 4.483041796 s (4483.041796 ns/call) 1000000 calls in 4.484165388 s (4484.165388 ns/call) 1000000 calls in 4.484850762 s (4484.850762 ns/call) 1000000 calls in 4.485643576 s (4485.643576 ns/call) $ ~/getuid.sh 5 1000000 calls in 6.497521653 s (6497.521653 ns/call) 1000000 calls in 6.502000236 s (6502.000236 ns/call) 1000000 calls in 6.501709115 s (6501.709115 ns/call) 1000000 calls in 6.502124100 s (6502.124100 ns/call) 1000000 calls in 6.502936358 s (6502.936358 ns/call) After the patch, the latencies scale better. 1000000 calls in 0.728720455 s (728.720455 ns/call) 1000000 calls in 0.842782857 s (842.782857 ns/call) 1000000 calls in 0.883803135 s (883.803135 ns/call) 1000000 calls in 0.902077764 s (902.077764 ns/call) 1000000 calls in 0.902838202 s (902.838202 ns/call) 1000000 calls in 0.908896885 s (908.896885 ns/call) 1000000 calls in 0.932523515 s (932.523515 ns/call) 1000000 calls in 0.958009672 s (958.009672 ns/call) 1000000 calls in 0.986188020 s (986.188020 ns/call) 1000000 calls in 0.989771102 s (989.771102 ns/call) 1000000 calls in 0.933518391 s (933.518391 ns/call) 1000000 calls in 0.958897947 s (958.897947 ns/call) 1000000 calls in 1.031038897 s (1031.038897 ns/call) 1000000 calls in 1.089516025 s (1089.516025 ns/call) 1000000 calls in 1.141998347 s (1141.998347 ns/call) Signed-off-by: Vaibhav Nagarnaik Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Michael Rubin Cc: David Sharp Cc: Linus Torvalds Link: http://lkml.kernel.org/r/1305059241-7629-1-git-send-email-vnagarnaik@google.com Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index ee9c921d7f21..71777c8fe36b 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -343,26 +343,27 @@ unsigned long trace_flags = TRACE_ITER_PRINT_PARENT | TRACE_ITER_PRINTK | static int trace_stop_count; static DEFINE_SPINLOCK(tracing_start_lock); +static void wakeup_work_handler(struct work_struct *work) +{ + wake_up(&trace_wait); +} + +static DECLARE_DELAYED_WORK(wakeup_work, wakeup_work_handler); + /** * trace_wake_up - wake up tasks waiting for trace input * - * Simply wakes up any task that is blocked on the trace_wait - * queue. These is used with trace_poll for tasks polling the trace. + * Schedules a delayed work to wake up any task that is blocked on the + * trace_wait queue. These is used with trace_poll for tasks polling the + * trace. */ void trace_wake_up(void) { - int cpu; + const unsigned long delay = msecs_to_jiffies(2); if (trace_flags & TRACE_ITER_BLOCK) return; - /* - * The runqueue_is_locked() can fail, but this is the best we - * have for now: - */ - cpu = get_cpu(); - if (!runqueue_is_locked(cpu)) - wake_up(&trace_wait); - put_cpu(); + schedule_delayed_work(&wakeup_work, delay); } static int __init set_buf_size(char *str) -- cgit v1.2.3 From 7ea5906405a1f3fc1c0033dfd7e02f2cfd1de5e5 Mon Sep 17 00:00:00 2001 From: Vaibhav Nagarnaik Date: Tue, 3 May 2011 17:56:42 -0700 Subject: tracing: Use NUMA allocation for per-cpu ring buffer pages The tracing ring buffer is a group of per-cpu ring buffers where allocation and logging is done on a per-cpu basis. The events that are generated on a particular CPU are logged in the corresponding buffer. This is to provide wait-free writes between CPUs and good NUMA node locality while accessing the ring buffer. However, the allocation routines consider NUMA locality only for buffer page metadata and not for the actual buffer page. This causes the pages to be allocated on the NUMA node local to the CPU where the allocation routine is running at the time. This patch fixes the problem by using a NUMA node specific allocation routine so that the pages are allocated from a NUMA node local to the logging CPU. I tested with the getuid_microbench from autotest. It is a simple binary that calls getuid() in a loop and measures the average time for the syscall to complete. The following command was used to test: $ getuid_microbench 1000000 Compared the numbers found on kernel with and without this patch and found that logging latency decreases by 30-50 ns/call. tracing with non-NUMA allocation - 569 ns/call tracing with NUMA allocation - 512 ns/call Signed-off-by: Vaibhav Nagarnaik Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Michael Rubin Cc: David Sharp Link: http://lkml.kernel.org/r/1304470602-20366-1-git-send-email-vnagarnaik@google.com Signed-off-by: Steven Rostedt --- include/linux/ring_buffer.h | 2 +- kernel/trace/ring_buffer.c | 36 +++++++++++++++++++----------------- kernel/trace/ring_buffer_benchmark.c | 2 +- kernel/trace/trace.c | 7 +++---- 4 files changed, 24 insertions(+), 23 deletions(-) (limited to 'kernel') diff --git a/include/linux/ring_buffer.h b/include/linux/ring_buffer.h index ab38ac80b0f9..b891de96000f 100644 --- a/include/linux/ring_buffer.h +++ b/include/linux/ring_buffer.h @@ -169,7 +169,7 @@ void ring_buffer_set_clock(struct ring_buffer *buffer, size_t ring_buffer_page_len(void *page); -void *ring_buffer_alloc_read_page(struct ring_buffer *buffer); +void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu); void ring_buffer_free_read_page(struct ring_buffer *buffer, void *data); int ring_buffer_read_page(struct ring_buffer *buffer, void **data_page, size_t len, int cpu, int full); diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index b0c7aa407943..725153d6cf73 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -997,13 +997,14 @@ static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned nr_pages) { struct buffer_page *bpage, *tmp; - unsigned long addr; LIST_HEAD(pages); unsigned i; WARN_ON(!nr_pages); for (i = 0; i < nr_pages; i++) { + struct page *page; + bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), GFP_KERNEL, cpu_to_node(cpu_buffer->cpu)); if (!bpage) @@ -1013,10 +1014,11 @@ static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, list_add(&bpage->list, &pages); - addr = __get_free_page(GFP_KERNEL); - if (!addr) + page = alloc_pages_node(cpu_to_node(cpu_buffer->cpu), + GFP_KERNEL, 0); + if (!page) goto free_pages; - bpage->page = (void *)addr; + bpage->page = page_address(page); rb_init_page(bpage->page); } @@ -1045,7 +1047,7 @@ rb_allocate_cpu_buffer(struct ring_buffer *buffer, int cpu) { struct ring_buffer_per_cpu *cpu_buffer; struct buffer_page *bpage; - unsigned long addr; + struct page *page; int ret; cpu_buffer = kzalloc_node(ALIGN(sizeof(*cpu_buffer), cache_line_size()), @@ -1067,10 +1069,10 @@ rb_allocate_cpu_buffer(struct ring_buffer *buffer, int cpu) rb_check_bpage(cpu_buffer, bpage); cpu_buffer->reader_page = bpage; - addr = __get_free_page(GFP_KERNEL); - if (!addr) + page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0); + if (!page) goto fail_free_reader; - bpage->page = (void *)addr; + bpage->page = page_address(page); rb_init_page(bpage->page); INIT_LIST_HEAD(&cpu_buffer->reader_page->list); @@ -1314,7 +1316,6 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size) unsigned nr_pages, rm_pages, new_pages; struct buffer_page *bpage, *tmp; unsigned long buffer_size; - unsigned long addr; LIST_HEAD(pages); int i, cpu; @@ -1375,16 +1376,17 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size) for_each_buffer_cpu(buffer, cpu) { for (i = 0; i < new_pages; i++) { + struct page *page; bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), GFP_KERNEL, cpu_to_node(cpu)); if (!bpage) goto free_pages; list_add(&bpage->list, &pages); - addr = __get_free_page(GFP_KERNEL); - if (!addr) + page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0); + if (!page) goto free_pages; - bpage->page = (void *)addr; + bpage->page = page_address(page); rb_init_page(bpage->page); } } @@ -3730,16 +3732,16 @@ EXPORT_SYMBOL_GPL(ring_buffer_swap_cpu); * Returns: * The page allocated, or NULL on error. */ -void *ring_buffer_alloc_read_page(struct ring_buffer *buffer) +void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu) { struct buffer_data_page *bpage; - unsigned long addr; + struct page *page; - addr = __get_free_page(GFP_KERNEL); - if (!addr) + page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0); + if (!page) return NULL; - bpage = (void *)addr; + bpage = page_address(page); rb_init_page(bpage); diff --git a/kernel/trace/ring_buffer_benchmark.c b/kernel/trace/ring_buffer_benchmark.c index 302f8a614635..a5457d577b98 100644 --- a/kernel/trace/ring_buffer_benchmark.c +++ b/kernel/trace/ring_buffer_benchmark.c @@ -106,7 +106,7 @@ static enum event_status read_page(int cpu) int inc; int i; - bpage = ring_buffer_alloc_read_page(buffer); + bpage = ring_buffer_alloc_read_page(buffer, cpu); if (!bpage) return EVENT_DROPPED; diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 71777c8fe36b..61fda6b6f1ab 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -3697,7 +3697,7 @@ tracing_buffers_read(struct file *filp, char __user *ubuf, return 0; if (!info->spare) - info->spare = ring_buffer_alloc_read_page(info->tr->buffer); + info->spare = ring_buffer_alloc_read_page(info->tr->buffer, info->cpu); if (!info->spare) return -ENOMEM; @@ -3854,7 +3854,7 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, ref->ref = 1; ref->buffer = info->tr->buffer; - ref->page = ring_buffer_alloc_read_page(ref->buffer); + ref->page = ring_buffer_alloc_read_page(ref->buffer, info->cpu); if (!ref->page) { kfree(ref); break; @@ -3863,8 +3863,7 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, r = ring_buffer_read_page(ref->buffer, &ref->page, len, info->cpu, 1); if (r < 0) { - ring_buffer_free_read_page(ref->buffer, - ref->page); + ring_buffer_free_read_page(ref->buffer, ref->page); kfree(ref); break; } -- cgit v1.2.3 From 4f271a2a60c748599b30bb4dafff30d770439b96 Mon Sep 17 00:00:00 2001 From: Vaibhav Nagarnaik Date: Mon, 13 Jun 2011 17:51:57 -0700 Subject: tracing: Add a proc file to stop tracing and free buffer The proc file entry buffer_size_kb is used to set the size of tracing buffer. The memory to expand the buffer size is kernel memory. Consider a use case where tracing is handled by a user space utility, which acts as a gate keeper for tracing requests. In an OOM condition, tracing is considered a low priority task and if the utility gets killed the ring buffer memory cannot be released back to the kernel. This patch adds a proc file called "free_buffer" whose purpose is to stop tracing and free up the ring buffer when it is closed. The user space process can then set the desired size in buffer_size_kb file and open the fd to the "free_buffer" file. Under OOM condition, if the process gets killed, the kernel closes the file descriptor. The release handler stops the tracing and releases the kernel memory automatically. Cc: Ingo Molnar Cc: Frederic Weisbecker Cc: Michael Rubin Cc: David Sharp Signed-off-by: Vaibhav Nagarnaik Link: http://lkml.kernel.org/r/1308012717-11148-1-git-send-email-vnagarnaik@google.com Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 108 ++++++++++++++++++++++++++++++++++----------------- 1 file changed, 73 insertions(+), 35 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 61fda6b6f1ab..9c557ae6a21e 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -2768,7 +2768,7 @@ int tracer_init(struct tracer *t, struct trace_array *tr) return t->init(tr); } -static int tracing_resize_ring_buffer(unsigned long size) +static int __tracing_resize_ring_buffer(unsigned long size) { int ret; @@ -2820,6 +2820,41 @@ static int tracing_resize_ring_buffer(unsigned long size) return ret; } +static ssize_t tracing_resize_ring_buffer(unsigned long size) +{ + int cpu, ret = size; + + mutex_lock(&trace_types_lock); + + tracing_stop(); + + /* disable all cpu buffers */ + for_each_tracing_cpu(cpu) { + if (global_trace.data[cpu]) + atomic_inc(&global_trace.data[cpu]->disabled); + if (max_tr.data[cpu]) + atomic_inc(&max_tr.data[cpu]->disabled); + } + + if (size != global_trace.entries) + ret = __tracing_resize_ring_buffer(size); + + if (ret < 0) + ret = -ENOMEM; + + for_each_tracing_cpu(cpu) { + if (global_trace.data[cpu]) + atomic_dec(&global_trace.data[cpu]->disabled); + if (max_tr.data[cpu]) + atomic_dec(&max_tr.data[cpu]->disabled); + } + + tracing_start(); + mutex_unlock(&trace_types_lock); + + return ret; +} + /** * tracing_update_buffers - used by tracing facility to expand ring buffers @@ -2837,7 +2872,7 @@ int tracing_update_buffers(void) mutex_lock(&trace_types_lock); if (!ring_buffer_expanded) - ret = tracing_resize_ring_buffer(trace_buf_size); + ret = __tracing_resize_ring_buffer(trace_buf_size); mutex_unlock(&trace_types_lock); return ret; @@ -2861,7 +2896,7 @@ static int tracing_set_tracer(const char *buf) mutex_lock(&trace_types_lock); if (!ring_buffer_expanded) { - ret = tracing_resize_ring_buffer(trace_buf_size); + ret = __tracing_resize_ring_buffer(trace_buf_size); if (ret < 0) goto out; ret = 0; @@ -3436,7 +3471,7 @@ tracing_entries_write(struct file *filp, const char __user *ubuf, { unsigned long val; char buf[64]; - int ret, cpu; + int ret; if (cnt >= sizeof(buf)) return -EINVAL; @@ -3454,48 +3489,43 @@ tracing_entries_write(struct file *filp, const char __user *ubuf, if (!val) return -EINVAL; - mutex_lock(&trace_types_lock); - - tracing_stop(); - - /* disable all cpu buffers */ - for_each_tracing_cpu(cpu) { - if (global_trace.data[cpu]) - atomic_inc(&global_trace.data[cpu]->disabled); - if (max_tr.data[cpu]) - atomic_inc(&max_tr.data[cpu]->disabled); - } - /* value is in KB */ val <<= 10; - if (val != global_trace.entries) { - ret = tracing_resize_ring_buffer(val); - if (ret < 0) { - cnt = ret; - goto out; - } - } + ret = tracing_resize_ring_buffer(val); + if (ret < 0) + return ret; *ppos += cnt; - /* If check pages failed, return ENOMEM */ - if (tracing_disabled) - cnt = -ENOMEM; - out: - for_each_tracing_cpu(cpu) { - if (global_trace.data[cpu]) - atomic_dec(&global_trace.data[cpu]->disabled); - if (max_tr.data[cpu]) - atomic_dec(&max_tr.data[cpu]->disabled); - } + return cnt; +} - tracing_start(); - mutex_unlock(&trace_types_lock); +static ssize_t +tracing_free_buffer_write(struct file *filp, const char __user *ubuf, + size_t cnt, loff_t *ppos) +{ + /* + * There is no need to read what the user has written, this function + * is just to make sure that there is no error when "echo" is used + */ + + *ppos += cnt; return cnt; } +static int +tracing_free_buffer_release(struct inode *inode, struct file *filp) +{ + /* disable tracing */ + tracing_off(); + /* resize the ring buffer to 0 */ + tracing_resize_ring_buffer(0); + + return 0; +} + static int mark_printk(const char *fmt, ...) { int ret; @@ -3641,6 +3671,11 @@ static const struct file_operations tracing_entries_fops = { .llseek = generic_file_llseek, }; +static const struct file_operations tracing_free_buffer_fops = { + .write = tracing_free_buffer_write, + .release = tracing_free_buffer_release, +}; + static const struct file_operations tracing_mark_fops = { .open = tracing_open_generic, .write = tracing_mark_write, @@ -4365,6 +4400,9 @@ static __init int tracer_init_debugfs(void) trace_create_file("buffer_size_kb", 0644, d_tracer, &global_trace, &tracing_entries_fops); + trace_create_file("free_buffer", 0644, d_tracer, + &global_trace, &tracing_free_buffer_fops); + trace_create_file("trace_marker", 0220, d_tracer, NULL, &tracing_mark_fops); -- cgit v1.2.3 From cf30cf67d6c7592c670ec946d89fc15ee0deb0eb Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 14 Jun 2011 22:44:07 -0400 Subject: tracing: Add disable_on_free option Add a trace option to disable tracing on free. When this option is set, a write into the free_buffer file will not only shrink the ring buffer down to zero, but it will also disable tracing. Cc: Vaibhav Nagarnaik Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 6 ++++-- kernel/trace/trace.h | 1 + 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 9c557ae6a21e..42fdf3adff31 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -425,6 +425,7 @@ static const char *trace_options[] = { "graph-time", "record-cmd", "overwrite", + "disable_on_free", NULL }; @@ -3518,8 +3519,9 @@ tracing_free_buffer_write(struct file *filp, const char __user *ubuf, static int tracing_free_buffer_release(struct inode *inode, struct file *filp) { - /* disable tracing */ - tracing_off(); + /* disable tracing ? */ + if (trace_flags & TRACE_ITER_STOP_ON_FREE) + tracing_off(); /* resize the ring buffer to 0 */ tracing_resize_ring_buffer(0); diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 229f8591f61d..742f545ae185 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -609,6 +609,7 @@ enum trace_iterator_flags { TRACE_ITER_GRAPH_TIME = 0x80000, TRACE_ITER_RECORD_CMD = 0x100000, TRACE_ITER_OVERWRITE = 0x200000, + TRACE_ITER_STOP_ON_FREE = 0x400000, }; /* -- cgit v1.2.3 From bd38c0e6f98326132a691d73b2056b426423c638 Mon Sep 17 00:00:00 2001 From: Paul McQuade Date: Tue, 31 May 2011 20:51:55 +0100 Subject: ftrace: Fixed an include coding style issue Removed because was already declared. Braces of struct's coding style fixed. Cc: Frederic Weisbecker Signed-off-by: Paul McQuade Link: http://lkml.kernel.org/r/4DE59711.3090900@gmail.com Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 1ee417fcbfa5..e1538071660d 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -32,7 +32,6 @@ #include -#include #include #include "trace_output.h" @@ -82,8 +81,7 @@ static int ftrace_disabled __read_mostly; static DEFINE_MUTEX(ftrace_lock); -static struct ftrace_ops ftrace_list_end __read_mostly = -{ +static struct ftrace_ops ftrace_list_end __read_mostly = { .func = ftrace_stub, }; @@ -785,8 +783,7 @@ static void unregister_ftrace_profiler(void) unregister_ftrace_graph(); } #else -static struct ftrace_ops ftrace_profile_ops __read_mostly = -{ +static struct ftrace_ops ftrace_profile_ops __read_mostly = { .func = function_profile_call, }; -- cgit v1.2.3 From 84c15027a7f2fbd7f1180d7cbb60e41abbbaedd2 Mon Sep 17 00:00:00 2001 From: Paul McQuade Date: Tue, 31 May 2011 20:51:55 +0100 Subject: async: Fixed an include coding style issue Added , and Removed . Added KERN_DEBUG to printk() functions. Acked-by: Arjan van de Ven Signed-off-by: Paul McQuade Link: http://lkml.kernel.org/r/4DE596B4.7030904@gmail.com Signed-off-by: Steven Rostedt --- kernel/async.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/async.c b/kernel/async.c index cd9dbb913c77..d5fe7af0de2e 100644 --- a/kernel/async.c +++ b/kernel/async.c @@ -49,12 +49,13 @@ asynchronous and synchronous parts of the kernel. */ #include +#include +#include #include #include #include #include #include -#include static async_cookie_t next_cookie = 1; @@ -128,7 +129,8 @@ static void async_run_entry_fn(struct work_struct *work) /* 2) run (and print duration) */ if (initcall_debug && system_state == SYSTEM_BOOTING) { - printk("calling %lli_%pF @ %i\n", (long long)entry->cookie, + printk(KERN_DEBUG "calling %lli_%pF @ %i\n", + (long long)entry->cookie, entry->func, task_pid_nr(current)); calltime = ktime_get(); } @@ -136,7 +138,7 @@ static void async_run_entry_fn(struct work_struct *work) if (initcall_debug && system_state == SYSTEM_BOOTING) { rettime = ktime_get(); delta = ktime_sub(rettime, calltime); - printk("initcall %lli_%pF returned 0 after %lld usecs\n", + printk(KERN_DEBUG "initcall %lli_%pF returned 0 after %lld usecs\n", (long long)entry->cookie, entry->func, (long long)ktime_to_ns(delta) >> 10); @@ -270,7 +272,7 @@ void async_synchronize_cookie_domain(async_cookie_t cookie, ktime_t starttime, delta, endtime; if (initcall_debug && system_state == SYSTEM_BOOTING) { - printk("async_waiting @ %i\n", task_pid_nr(current)); + printk(KERN_DEBUG "async_waiting @ %i\n", task_pid_nr(current)); starttime = ktime_get(); } @@ -280,7 +282,7 @@ void async_synchronize_cookie_domain(async_cookie_t cookie, endtime = ktime_get(); delta = ktime_sub(endtime, starttime); - printk("async_continuing @ %i after %lli usec\n", + printk(KERN_DEBUG "async_continuing @ %i after %lli usec\n", task_pid_nr(current), (long long)ktime_to_ns(delta) >> 10); } -- cgit v1.2.3 From 321e68b095addc473ca52ced9acfa59be88f76e6 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Fri, 3 Jun 2011 16:58:47 +0200 Subject: tracing, function_graph: Remove dependency of abstime and duration fields on latency The display of absolute time and duration fields is based on the latency field. This was added during the irqsoff/wakeup tracers graph support changes. It's causing confusion in what fields will be displayed for the function_graph tracer itself. So I'm removing this depency, and adding absolute time and duration fields to the preemptirqsoff preemptoff irqsoff wakeup tracers. With following commands: # echo function_graph > ./current_tracer # cat trace This is what it looked like before: # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 0) 0.068 us | } /* page_add_file_rmap */ 0) | _raw_spin_unlock() { ... This is what it looks like now: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 0) 0.068 us | } /* add_preempt_count */ 0) 0.993 us | } /* vfsmount_lock_local_lock */ ... For preemptirqsoff preemptoff irqsoff wakeup tracers, this is what it looked like before: SNIP # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / _-=> lock-depth # |||| / # CPU TASK/PID ||||| DURATION FUNCTION CALLS # | | | ||||| | | | | | | 1) -0 | d..1 0.000 us | acpi_idle_enter_simple(); ... This is what it looks like now: SNIP # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / # TIME CPU TASK/PID |||| DURATION FUNCTION CALLS # | | | | |||| | | | | | | 19.847735 | 1) -0 | d..1 0.000 us | acpi_idle_enter_simple(); ... Signed-off-by: Jiri Olsa Link: http://lkml.kernel.org/r/1307113131-10045-2-git-send-email-jolsa@redhat.com Signed-off-by: Steven Rostedt --- kernel/trace/trace_functions_graph.c | 19 +++---------------- kernel/trace/trace_irqsoff.c | 4 +++- kernel/trace/trace_sched_wakeup.c | 4 +++- 3 files changed, 9 insertions(+), 18 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c index 962cdb24ed81..352652eb9aae 100644 --- a/kernel/trace/trace_functions_graph.c +++ b/kernel/trace/trace_functions_graph.c @@ -1207,7 +1207,7 @@ print_graph_comment(struct trace_seq *s, struct trace_entry *ent, enum print_line_t -__print_graph_function_flags(struct trace_iterator *iter, u32 flags) +print_graph_function_flags(struct trace_iterator *iter, u32 flags) { struct ftrace_graph_ent_entry *field; struct fgraph_data *data = iter->private; @@ -1270,18 +1270,7 @@ __print_graph_function_flags(struct trace_iterator *iter, u32 flags) static enum print_line_t print_graph_function(struct trace_iterator *iter) { - return __print_graph_function_flags(iter, tracer_flags.val); -} - -enum print_line_t print_graph_function_flags(struct trace_iterator *iter, - u32 flags) -{ - if (trace_flags & TRACE_ITER_LATENCY_FMT) - flags |= TRACE_GRAPH_PRINT_DURATION; - else - flags |= TRACE_GRAPH_PRINT_ABS_TIME; - - return __print_graph_function_flags(iter, flags); + return print_graph_function_flags(iter, tracer_flags.val); } static enum print_line_t @@ -1364,9 +1353,7 @@ void print_graph_headers_flags(struct seq_file *s, u32 flags) return; print_trace_header(s, iter); - flags |= TRACE_GRAPH_PRINT_DURATION; - } else - flags |= TRACE_GRAPH_PRINT_ABS_TIME; + } __print_graph_headers_flags(s, flags); } diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c index c77424be284d..667aa8cc0cfc 100644 --- a/kernel/trace/trace_irqsoff.c +++ b/kernel/trace/trace_irqsoff.c @@ -226,7 +226,9 @@ static void irqsoff_trace_close(struct trace_iterator *iter) } #define GRAPH_TRACER_FLAGS (TRACE_GRAPH_PRINT_CPU | \ - TRACE_GRAPH_PRINT_PROC) + TRACE_GRAPH_PRINT_PROC | \ + TRACE_GRAPH_PRINT_ABS_TIME | \ + TRACE_GRAPH_PRINT_DURATION) static enum print_line_t irqsoff_print_line(struct trace_iterator *iter) { diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c index f029dd4fd2ca..e4a70c0c71b6 100644 --- a/kernel/trace/trace_sched_wakeup.c +++ b/kernel/trace/trace_sched_wakeup.c @@ -227,7 +227,9 @@ static void wakeup_trace_close(struct trace_iterator *iter) graph_trace_close(iter); } -#define GRAPH_TRACER_FLAGS (TRACE_GRAPH_PRINT_PROC) +#define GRAPH_TRACER_FLAGS (TRACE_GRAPH_PRINT_PROC | \ + TRACE_GRAPH_PRINT_ABS_TIME | \ + TRACE_GRAPH_PRINT_DURATION) static enum print_line_t wakeup_print_line(struct trace_iterator *iter) { -- cgit v1.2.3 From ffeb80fc30acbf6bd51cb47a1815f621a9d017dc Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Fri, 3 Jun 2011 16:58:48 +0200 Subject: tracing, function_graph: Merge overhead and duration display functions Functions print_graph_overhead() and print_graph_duration() displays data for one field - DURATION. I merged them into single function print_graph_duration(), and added a way to display the empty parts of the field. This way the print_graph_irq() function can use this column to display the IRQ signs if needed and the DURATION field details stays inside the print_graph_duration() function. Signed-off-by: Jiri Olsa Link: http://lkml.kernel.org/r/1307113131-10045-3-git-send-email-jolsa@redhat.com Signed-off-by: Steven Rostedt --- kernel/trace/trace_functions_graph.c | 148 +++++++++++++++++------------------ 1 file changed, 74 insertions(+), 74 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c index 352652eb9aae..44b955fd87bf 100644 --- a/kernel/trace/trace_functions_graph.c +++ b/kernel/trace/trace_functions_graph.c @@ -74,6 +74,20 @@ static struct tracer_flags tracer_flags = { static struct trace_array *graph_array; +/* + * DURATION column is being also used to display IRQ signs, + * following values are used by print_graph_irq and others + * to fill in space into DURATION column. + */ +enum { + DURATION_FILL_FULL = -1, + DURATION_FILL_START = -2, + DURATION_FILL_END = -3, +}; + +static enum print_line_t +print_graph_duration(unsigned long long duration, struct trace_seq *s, + u32 flags); /* Add a function return address to the trace stack on thread info.*/ int @@ -577,32 +591,6 @@ get_return_for_leaf(struct trace_iterator *iter, return next; } -/* Signal a overhead of time execution to the output */ -static int -print_graph_overhead(unsigned long long duration, struct trace_seq *s, - u32 flags) -{ - /* If duration disappear, we don't need anything */ - if (!(flags & TRACE_GRAPH_PRINT_DURATION)) - return 1; - - /* Non nested entry or return */ - if (duration == -1) - return trace_seq_printf(s, " "); - - if (flags & TRACE_GRAPH_PRINT_OVERHEAD) { - /* Duration exceeded 100 msecs */ - if (duration > 100000ULL) - return trace_seq_printf(s, "! "); - - /* Duration exceeded 10 msecs */ - if (duration > 10000ULL) - return trace_seq_printf(s, "+ "); - } - - return trace_seq_printf(s, " "); -} - static int print_graph_abs_time(u64 t, struct trace_seq *s) { unsigned long usecs_rem; @@ -650,9 +638,9 @@ print_graph_irq(struct trace_iterator *iter, unsigned long addr, } /* No overhead */ - ret = print_graph_overhead(-1, s, flags); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; + ret = print_graph_duration(DURATION_FILL_START, s, flags); + if (ret != TRACE_TYPE_HANDLED) + return ret; if (type == TRACE_GRAPH_ENT) ret = trace_seq_printf(s, "==========>"); @@ -662,9 +650,10 @@ print_graph_irq(struct trace_iterator *iter, unsigned long addr, if (!ret) return TRACE_TYPE_PARTIAL_LINE; - /* Don't close the duration column if haven't one */ - if (flags & TRACE_GRAPH_PRINT_DURATION) - trace_seq_printf(s, " |"); + ret = print_graph_duration(DURATION_FILL_END, s, flags); + if (ret != TRACE_TYPE_HANDLED) + return ret; + ret = trace_seq_printf(s, "\n"); if (!ret) @@ -716,9 +705,48 @@ trace_print_graph_duration(unsigned long long duration, struct trace_seq *s) } static enum print_line_t -print_graph_duration(unsigned long long duration, struct trace_seq *s) +print_graph_duration(unsigned long long duration, struct trace_seq *s, + u32 flags) { - int ret; + int ret = -1; + + if (!(flags & TRACE_GRAPH_PRINT_DURATION)) + return TRACE_TYPE_HANDLED; + + /* No real adata, just filling the column with spaces */ + switch (duration) { + case DURATION_FILL_FULL: + ret = trace_seq_printf(s, " | "); + return ret ? TRACE_TYPE_HANDLED : TRACE_TYPE_PARTIAL_LINE; + case DURATION_FILL_START: + ret = trace_seq_printf(s, " "); + return ret ? TRACE_TYPE_HANDLED : TRACE_TYPE_PARTIAL_LINE; + case DURATION_FILL_END: + ret = trace_seq_printf(s, " |"); + return ret ? TRACE_TYPE_HANDLED : TRACE_TYPE_PARTIAL_LINE; + } + + /* Signal a overhead of time execution to the output */ + if (flags & TRACE_GRAPH_PRINT_OVERHEAD) { + /* Duration exceeded 100 msecs */ + if (duration > 100000ULL) + ret = trace_seq_printf(s, "! "); + /* Duration exceeded 10 msecs */ + else if (duration > 10000ULL) + ret = trace_seq_printf(s, "+ "); + } + + /* + * The -1 means we either did not exceed the duration tresholds + * or we dont want to print out the overhead. Either way we need + * to fill out the space. + */ + if (ret == -1) + ret = trace_seq_printf(s, " "); + + /* Catching here any failure happenned above */ + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; ret = trace_print_graph_duration(duration, s); if (ret != TRACE_TYPE_HANDLED) @@ -767,18 +795,11 @@ print_graph_entry_leaf(struct trace_iterator *iter, cpu_data->enter_funcs[call->depth] = 0; } - /* Overhead */ - ret = print_graph_overhead(duration, s, flags); - if (!ret) + /* Overhead and duration */ + ret = print_graph_duration(duration, s, flags); + if (ret == TRACE_TYPE_PARTIAL_LINE) return TRACE_TYPE_PARTIAL_LINE; - /* Duration */ - if (flags & TRACE_GRAPH_PRINT_DURATION) { - ret = print_graph_duration(duration, s); - if (ret == TRACE_TYPE_PARTIAL_LINE) - return TRACE_TYPE_PARTIAL_LINE; - } - /* Function */ for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) { ret = trace_seq_printf(s, " "); @@ -815,17 +836,10 @@ print_graph_entry_nested(struct trace_iterator *iter, cpu_data->enter_funcs[call->depth] = call->func; } - /* No overhead */ - ret = print_graph_overhead(-1, s, flags); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; - /* No time */ - if (flags & TRACE_GRAPH_PRINT_DURATION) { - ret = trace_seq_printf(s, " | "); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; - } + ret = print_graph_duration(DURATION_FILL_FULL, s, flags); + if (ret != TRACE_TYPE_HANDLED) + return ret; /* Function */ for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) { @@ -1078,18 +1092,11 @@ print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s, if (print_graph_prologue(iter, s, 0, 0, flags)) return TRACE_TYPE_PARTIAL_LINE; - /* Overhead */ - ret = print_graph_overhead(duration, s, flags); - if (!ret) + /* Overhead and duration */ + ret = print_graph_duration(duration, s, flags); + if (ret == TRACE_TYPE_PARTIAL_LINE) return TRACE_TYPE_PARTIAL_LINE; - /* Duration */ - if (flags & TRACE_GRAPH_PRINT_DURATION) { - ret = print_graph_duration(duration, s); - if (ret == TRACE_TYPE_PARTIAL_LINE) - return TRACE_TYPE_PARTIAL_LINE; - } - /* Closing brace */ for (i = 0; i < trace->depth * TRACE_GRAPH_INDENT; i++) { ret = trace_seq_printf(s, " "); @@ -1146,17 +1153,10 @@ print_graph_comment(struct trace_seq *s, struct trace_entry *ent, if (print_graph_prologue(iter, s, 0, 0, flags)) return TRACE_TYPE_PARTIAL_LINE; - /* No overhead */ - ret = print_graph_overhead(-1, s, flags); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; - /* No time */ - if (flags & TRACE_GRAPH_PRINT_DURATION) { - ret = trace_seq_printf(s, " | "); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; - } + ret = print_graph_duration(DURATION_FILL_FULL, s, flags); + if (ret != TRACE_TYPE_HANDLED) + return ret; /* Indentation */ if (depth > 0) -- cgit v1.2.3 From f56e7f8efb4ec200364f690a9902713410e24d47 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Fri, 3 Jun 2011 16:58:49 +0200 Subject: tracing, function: Fix trace header to follow context-info option The header display of function tracer does not follow the context-info option, so field names are displayed even if this option is off. Added check for TRACE_ITER_CONTEXT_INFO trace_flags. With following commands: # echo function > ./current_tracer # echo 0 > options/context-info # cat trace This is what it looked like before: # tracer: function # # TASK-PID CPU# TIMESTAMP FUNCTION # | | | | | add_preempt_count <-schedule rcu_note_context_switch <-schedule ... This is what it looks like now: # tracer: function # _raw_spin_unlock_irqrestore <-hrtimer_try_to_cancel ... Signed-off-by: Jiri Olsa Link: http://lkml.kernel.org/r/1307113131-10045-4-git-send-email-jolsa@redhat.com Signed-off-by: Steven Rostedt --- kernel/trace/trace.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 42fdf3adff31..cf22b4bf9896 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -2053,6 +2053,9 @@ void trace_default_header(struct seq_file *m) { struct trace_iterator *iter = m->private; + if (!(trace_flags & TRACE_ITER_CONTEXT_INFO)) + return; + if (iter->iter_flags & TRACE_FILE_LAT_FMT) { /* print nothing if the buffers are empty */ if (trace_empty(iter)) -- cgit v1.2.3 From 199abfab40389963b397c2982222e68ea782b2cf Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Fri, 3 Jun 2011 16:58:50 +0200 Subject: tracing, function_graph: Remove lock-depth from latency trace The lock_depth was removed in commit e6e1e25 tracing: Remove lock_depth from event entry Removing the lock_depth info from function_graph latency header. With following commands: # echo function_graph > ./current_tracer # echo 1 > options/latency-format # cat trace This is what it looked like before: # tracer: function_graph # # function_graph latency trace v1.1.5 on 3.0.0-rc1-tip+ # -------------------------------------------------------------------- # latency: 0 us, #59756/311298, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) # ----------------- # | task: -0 (uid:0 nice:0 policy:0 rt_prio:0) # ----------------- # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / _-=> lock-depth # |||| / # CPU||||| DURATION FUNCTION CALLS # | ||||| | | | | | | 0) .... 0.068 us | } /* __rcu_read_unlock */ ... This is what it looks like now: # tracer: function_graph # # function_graph latency trace v1.1.5 on 3.0.0-rc1-tip+ # -------------------------------------------------------------------- # latency: 0 us, #59747/1744610, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) # ----------------- # | task: -0 (uid:0 nice:0 policy:0 rt_prio:0) # ----------------- # # _-----=> irqs-off # / _----=> need-resched # | / _---=> hardirq/softirq # || / _--=> preempt-depth # ||| / # CPU|||| DURATION FUNCTION CALLS # | |||| | | | | | | 0) ..s. 1.641 us | } /* __rcu_process_callbacks */ ... Signed-off-by: Jiri Olsa Link: http://lkml.kernel.org/r/1307113131-10045-5-git-send-email-jolsa@redhat.com Signed-off-by: Steven Rostedt --- kernel/trace/trace_functions_graph.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c index 44b955fd87bf..1da5b97d2288 100644 --- a/kernel/trace/trace_functions_graph.c +++ b/kernel/trace/trace_functions_graph.c @@ -1298,8 +1298,7 @@ static void print_lat_header(struct seq_file *s, u32 flags) seq_printf(s, "#%.*s / _----=> need-resched \n", size, spaces); seq_printf(s, "#%.*s| / _---=> hardirq/softirq \n", size, spaces); seq_printf(s, "#%.*s|| / _--=> preempt-depth \n", size, spaces); - seq_printf(s, "#%.*s||| / _-=> lock-depth \n", size, spaces); - seq_printf(s, "#%.*s|||| / \n", size, spaces); + seq_printf(s, "#%.*s||| / \n", size, spaces); } static void __print_graph_headers_flags(struct seq_file *s, u32 flags) @@ -1318,7 +1317,7 @@ static void __print_graph_headers_flags(struct seq_file *s, u32 flags) if (flags & TRACE_GRAPH_PRINT_PROC) seq_printf(s, " TASK/PID "); if (lat) - seq_printf(s, "|||||"); + seq_printf(s, "||||"); if (flags & TRACE_GRAPH_PRINT_DURATION) seq_printf(s, " DURATION "); seq_printf(s, " FUNCTION CALLS\n"); @@ -1332,7 +1331,7 @@ static void __print_graph_headers_flags(struct seq_file *s, u32 flags) if (flags & TRACE_GRAPH_PRINT_PROC) seq_printf(s, " | | "); if (lat) - seq_printf(s, "|||||"); + seq_printf(s, "||||"); if (flags & TRACE_GRAPH_PRINT_DURATION) seq_printf(s, " | | "); seq_printf(s, " | | | |\n"); -- cgit v1.2.3 From 749230b06a753a22f6ed96e5dd60815d6ab12865 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Fri, 3 Jun 2011 16:58:51 +0200 Subject: tracing, function_graph: Add context-info support for function_graph tracer The function_graph tracer does not follow global context-info option. Adding TRACE_ITER_CONTEXT_INFO trace_flags check to enable it. With following commands: # echo function_graph > ./current_tracer # echo 0 > options/context-info # cat trace This is what it looked like before: # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 1) 0.079 us | } /* __vma_link_rb */ 1) 0.056 us | copy_page_range(); 1) | security_vm_enough_memory() { ... This is what it looks like now: # tracer: function_graph # } /* update_ts_time_stats */ timekeeping_max_deferment(); ... Signed-off-by: Jiri Olsa Link: http://lkml.kernel.org/r/1307113131-10045-6-git-send-email-jolsa@redhat.com Signed-off-by: Steven Rostedt --- kernel/trace/trace_functions_graph.c | 53 +++++++++++++++++++++--------------- 1 file changed, 31 insertions(+), 22 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c index 1da5b97d2288..e8d6bb55d719 100644 --- a/kernel/trace/trace_functions_graph.c +++ b/kernel/trace/trace_functions_graph.c @@ -613,28 +613,30 @@ print_graph_irq(struct trace_iterator *iter, unsigned long addr, addr >= (unsigned long)__irqentry_text_end) return TRACE_TYPE_UNHANDLED; - /* Absolute time */ - if (flags & TRACE_GRAPH_PRINT_ABS_TIME) { - ret = print_graph_abs_time(iter->ts, s); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; - } + if (trace_flags & TRACE_ITER_CONTEXT_INFO) { + /* Absolute time */ + if (flags & TRACE_GRAPH_PRINT_ABS_TIME) { + ret = print_graph_abs_time(iter->ts, s); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + } - /* Cpu */ - if (flags & TRACE_GRAPH_PRINT_CPU) { - ret = print_graph_cpu(s, cpu); - if (ret == TRACE_TYPE_PARTIAL_LINE) - return TRACE_TYPE_PARTIAL_LINE; - } + /* Cpu */ + if (flags & TRACE_GRAPH_PRINT_CPU) { + ret = print_graph_cpu(s, cpu); + if (ret == TRACE_TYPE_PARTIAL_LINE) + return TRACE_TYPE_PARTIAL_LINE; + } - /* Proc */ - if (flags & TRACE_GRAPH_PRINT_PROC) { - ret = print_graph_proc(s, pid); - if (ret == TRACE_TYPE_PARTIAL_LINE) - return TRACE_TYPE_PARTIAL_LINE; - ret = trace_seq_printf(s, " | "); - if (!ret) - return TRACE_TYPE_PARTIAL_LINE; + /* Proc */ + if (flags & TRACE_GRAPH_PRINT_PROC) { + ret = print_graph_proc(s, pid); + if (ret == TRACE_TYPE_PARTIAL_LINE) + return TRACE_TYPE_PARTIAL_LINE; + ret = trace_seq_printf(s, " | "); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + } } /* No overhead */ @@ -710,8 +712,9 @@ print_graph_duration(unsigned long long duration, struct trace_seq *s, { int ret = -1; - if (!(flags & TRACE_GRAPH_PRINT_DURATION)) - return TRACE_TYPE_HANDLED; + if (!(flags & TRACE_GRAPH_PRINT_DURATION) || + !(trace_flags & TRACE_ITER_CONTEXT_INFO)) + return TRACE_TYPE_HANDLED; /* No real adata, just filling the column with spaces */ switch (duration) { @@ -879,6 +882,9 @@ print_graph_prologue(struct trace_iterator *iter, struct trace_seq *s, return TRACE_TYPE_PARTIAL_LINE; } + if (!(trace_flags & TRACE_ITER_CONTEXT_INFO)) + return 0; + /* Absolute time */ if (flags & TRACE_GRAPH_PRINT_ABS_TIME) { ret = print_graph_abs_time(iter->ts, s); @@ -1346,6 +1352,9 @@ void print_graph_headers_flags(struct seq_file *s, u32 flags) { struct trace_iterator *iter = s->private; + if (!(trace_flags & TRACE_ITER_CONTEXT_INFO)) + return; + if (trace_flags & TRACE_ITER_LATENCY_FMT) { /* print nothing if the buffers are empty */ if (trace_empty(iter)) -- cgit v1.2.3 From 22fe9b54d859e53bfbbbdc1a0a77a82bc453927c Mon Sep 17 00:00:00 2001 From: Peter Huewe Date: Tue, 7 Jun 2011 21:58:27 +0200 Subject: tracing: Convert to kstrtoul_from_user This patch replaces the code for getting an unsigned long from a userspace buffer by a simple call to kstroul_from_user. This makes it easier to read and less error prone. Signed-off-by: Peter Huewe Link: http://lkml.kernel.org/r/1307476707-14762-1-git-send-email-peterhuewe@gmx.de Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 13 ++------- kernel/trace/ring_buffer.c | 13 ++------- kernel/trace/trace.c | 65 +++++++-------------------------------------- kernel/trace/trace_events.c | 26 +++--------------- kernel/trace/trace_stack.c | 13 ++------- 5 files changed, 20 insertions(+), 110 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index e1538071660d..458018a1ac9a 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -803,19 +803,10 @@ ftrace_profile_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { unsigned long val; - char buf[64]; /* big enough to hold a number */ int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; val = !!val; diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 725153d6cf73..f00ede314eb6 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -3980,20 +3980,11 @@ rb_simple_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { unsigned long *p = filp->private_data; - char buf[64]; unsigned long val; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; if (val) diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index cf22b4bf9896..c977018e87c2 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -2706,20 +2706,11 @@ tracing_ctrl_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { struct trace_array *tr = filp->private_data; - char buf[64]; unsigned long val; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; val = !!val; @@ -3006,20 +2997,11 @@ tracing_max_lat_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { unsigned long *ptr = filp->private_data; - char buf[64]; unsigned long val; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; *ptr = val * 1000; @@ -3474,19 +3456,10 @@ tracing_entries_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { unsigned long val; - char buf[64]; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; /* must have at least 1 entry */ @@ -4139,19 +4112,10 @@ trace_options_write(struct file *filp, const char __user *ubuf, size_t cnt, { struct trace_option_dentry *topt = filp->private_data; unsigned long val; - char buf[64]; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; if (val != 0 && val != 1) @@ -4199,20 +4163,11 @@ trace_options_core_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { long index = (long)filp->private_data; - char buf[64]; unsigned long val; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; if (val != 0 && val != 1) diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 686ec399f2a8..4d7e1498ae91 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -486,20 +486,11 @@ event_enable_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { struct ftrace_event_call *call = filp->private_data; - char buf[64]; unsigned long val; int ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; ret = tracing_update_buffers(); @@ -571,19 +562,10 @@ system_enable_write(struct file *filp, const char __user *ubuf, size_t cnt, { const char *system = filp->private_data; unsigned long val; - char buf[64]; ssize_t ret; - if (cnt >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, cnt)) - return -EFAULT; - - buf[cnt] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, cnt, 10, &val); + if (ret) return ret; ret = tracing_update_buffers(); diff --git a/kernel/trace/trace_stack.c b/kernel/trace/trace_stack.c index b0b53b8e4c25..77575b386d97 100644 --- a/kernel/trace/trace_stack.c +++ b/kernel/trace/trace_stack.c @@ -156,20 +156,11 @@ stack_max_size_write(struct file *filp, const char __user *ubuf, { long *ptr = filp->private_data; unsigned long val, flags; - char buf[64]; int ret; int cpu; - if (count >= sizeof(buf)) - return -EINVAL; - - if (copy_from_user(&buf, ubuf, count)) - return -EFAULT; - - buf[count] = 0; - - ret = strict_strtoul(buf, 10, &val); - if (ret < 0) + ret = kstrtoul_from_user(ubuf, count, 10, &val); + if (ret) return ret; local_irq_save(flags); -- cgit v1.2.3 From d7ec4bfed6c97405c6417970ba06c439e08ab8e3 Mon Sep 17 00:00:00 2001 From: Vaibhav Nagarnaik Date: Tue, 7 Jun 2011 17:01:42 -0700 Subject: ring-buffer: Set __GFP_NORETRY flag for ring buffer allocating process The tracing ring buffer is allocated from kernel memory. While allocating a large chunk of memory, OOM might happen which destabilizes the system. Thus random processes might get killed during the allocation. This patch adds __GFP_NORETRY flag to the ring buffer allocation calls to make it fail more gracefully if the system will not be able to complete the allocation request. Acked-by: David Rientjes Signed-off-by: Vaibhav Nagarnaik Cc: Ingo Molnar Cc: Frederic Weisbecker Cc: Michael Rubin Cc: David Sharp Link: http://lkml.kernel.org/r/1307491302-9236-1-git-send-email-vnagarnaik@google.com Signed-off-by: Steven Rostedt --- kernel/trace/ring_buffer.c | 25 +++++++++++++++++++------ 1 file changed, 19 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index f00ede314eb6..731201bf4acc 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -1004,9 +1004,14 @@ static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, for (i = 0; i < nr_pages; i++) { struct page *page; - + /* + * __GFP_NORETRY flag makes sure that the allocation fails + * gracefully without invoking oom-killer and the system is + * not destabilized. + */ bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), - GFP_KERNEL, cpu_to_node(cpu_buffer->cpu)); + GFP_KERNEL | __GFP_NORETRY, + cpu_to_node(cpu_buffer->cpu)); if (!bpage) goto free_pages; @@ -1015,7 +1020,7 @@ static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, list_add(&bpage->list, &pages); page = alloc_pages_node(cpu_to_node(cpu_buffer->cpu), - GFP_KERNEL, 0); + GFP_KERNEL | __GFP_NORETRY, 0); if (!page) goto free_pages; bpage->page = page_address(page); @@ -1377,13 +1382,20 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size) for_each_buffer_cpu(buffer, cpu) { for (i = 0; i < new_pages; i++) { struct page *page; + /* + * __GFP_NORETRY flag makes sure that the allocation + * fails gracefully without invoking oom-killer and + * the system is not destabilized. + */ bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), - GFP_KERNEL, cpu_to_node(cpu)); + GFP_KERNEL | __GFP_NORETRY, + cpu_to_node(cpu)); if (!bpage) goto free_pages; list_add(&bpage->list, &pages); - page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0); + page = alloc_pages_node(cpu_to_node(cpu), + GFP_KERNEL | __GFP_NORETRY, 0); if (!page) goto free_pages; bpage->page = page_address(page); @@ -3737,7 +3749,8 @@ void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu) struct buffer_data_page *bpage; struct page *page; - page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0); + page = alloc_pages_node(cpu_to_node(cpu), + GFP_KERNEL | __GFP_NORETRY, 0); if (!page) return NULL; -- cgit v1.2.3 From c624d33f61cd05241e85b906311f0b712fdb0f32 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 8 Jun 2011 16:09:27 +0900 Subject: stack_trace: Add weak save_stack_trace_regs() Add weak symbol of save_stack_trace_regs() as same as save_stack_trace_tsk() since that is not implemented except x86 yet. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: yrl.pp-manager.tt@hitachi.com Cc: Peter Zijlstra Cc: Namhyung Kim Link: http://lkml.kernel.org/r/20110608070927.17777.37895.stgit@fedora15 Signed-off-by: Steven Rostedt --- kernel/stacktrace.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'kernel') diff --git a/kernel/stacktrace.c b/kernel/stacktrace.c index eb212f8f8bc8..d20c6983aad9 100644 --- a/kernel/stacktrace.c +++ b/kernel/stacktrace.c @@ -26,12 +26,18 @@ void print_stack_trace(struct stack_trace *trace, int spaces) EXPORT_SYMBOL_GPL(print_stack_trace); /* - * Architectures that do not implement save_stack_trace_tsk get this - * weak alias and a once-per-bootup warning (whenever this facility - * is utilized - for example by procfs): + * Architectures that do not implement save_stack_trace_tsk or + * save_stack_trace_regs get this weak alias and a once-per-bootup warning + * (whenever this facility is utilized - for example by procfs): */ __weak void save_stack_trace_tsk(struct task_struct *tsk, struct stack_trace *trace) { WARN_ONCE(1, KERN_INFO "save_stack_trace_tsk() not implemented yet.\n"); } + +__weak void +save_stack_trace_regs(struct pt_regs *regs, struct stack_trace *trace) +{ + WARN_ONCE(1, KERN_INFO "save_stack_trace_regs() not implemented yet.\n"); +} -- cgit v1.2.3 From 1fd8df2c3970c9e7e4e262354154ee39e58bdd7c Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 8 Jun 2011 16:09:34 +0900 Subject: tracing/kprobes: Fix kprobe-tracer to support stack trace Fix to support kernel stack trace correctly on kprobe-tracer. Since the execution path of kprobe-based dynamic events is different from other tracepoint-based events, normal ftrace_trace_stack() doesn't work correctly. To fix that, this introduces ftrace_trace_stack_regs() which traces stack via pt_regs instead of current stack register. e.g. # echo p schedule+4 > /sys/kernel/debug/tracing/kprobe_events # echo 1 > /sys/kernel/debug/tracing/options/stacktrace # echo 1 > /sys/kernel/debug/tracing/events/kprobes/enable # head -n 20 /sys/kernel/debug/tracing/trace bash-2968 [000] 10297.050245: p_schedule_4: (schedule+0x4/0x4ca) bash-2968 [000] 10297.050247: => schedule_timeout => n_tty_read => tty_read => vfs_read => sys_read => system_call_fastpath kworker/0:1-2940 [000] 10297.050265: p_schedule_4: (schedule+0x4/0x4ca) kworker/0:1-2940 [000] 10297.050266: => worker_thread => kthread => kernel_thread_helper sshd-1132 [000] 10297.050365: p_schedule_4: (schedule+0x4/0x4ca) sshd-1132 [000] 10297.050365: => sysret_careful Note: Even with this fix, the first entry will be skipped if the probe is put on the function entry area before the frame pointer is set up (usually, that is 4 bytes (push %bp; mov %sp %bp) on x86), because stack unwinder depends on the frame pointer. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: yrl.pp-manager.tt@hitachi.com Cc: Peter Zijlstra Cc: Namhyung Kim Link: http://lkml.kernel.org/r/20110608070934.17777.17116.stgit@fedora15 Signed-off-by: Steven Rostedt --- include/linux/ftrace_event.h | 4 ++++ kernel/trace/trace.c | 34 +++++++++++++++++++++++++++++----- kernel/trace/trace.h | 9 +++++++++ kernel/trace/trace_kprobe.c | 6 ++++-- 4 files changed, 46 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h index 59d3ef100eb9..b1e69eefc203 100644 --- a/include/linux/ftrace_event.h +++ b/include/linux/ftrace_event.h @@ -129,6 +129,10 @@ void trace_current_buffer_unlock_commit(struct ring_buffer *buffer, void trace_nowake_buffer_unlock_commit(struct ring_buffer *buffer, struct ring_buffer_event *event, unsigned long flags, int pc); +void trace_nowake_buffer_unlock_commit_regs(struct ring_buffer *buffer, + struct ring_buffer_event *event, + unsigned long flags, int pc, + struct pt_regs *regs); void trace_current_buffer_discard_commit(struct ring_buffer *buffer, struct ring_buffer_event *event); diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index c977018e87c2..d9c16123f6e2 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -1193,6 +1193,18 @@ void trace_nowake_buffer_unlock_commit(struct ring_buffer *buffer, } EXPORT_SYMBOL_GPL(trace_nowake_buffer_unlock_commit); +void trace_nowake_buffer_unlock_commit_regs(struct ring_buffer *buffer, + struct ring_buffer_event *event, + unsigned long flags, int pc, + struct pt_regs *regs) +{ + ring_buffer_unlock_commit(buffer, event); + + ftrace_trace_stack_regs(buffer, flags, 0, pc, regs); + ftrace_trace_userstack(buffer, flags, pc); +} +EXPORT_SYMBOL_GPL(trace_nowake_buffer_unlock_commit_regs); + void trace_current_buffer_discard_commit(struct ring_buffer *buffer, struct ring_buffer_event *event) { @@ -1238,7 +1250,7 @@ ftrace(struct trace_array *tr, struct trace_array_cpu *data, #ifdef CONFIG_STACKTRACE static void __ftrace_trace_stack(struct ring_buffer *buffer, unsigned long flags, - int skip, int pc) + int skip, int pc, struct pt_regs *regs) { struct ftrace_event_call *call = &event_kernel_stack; struct ring_buffer_event *event; @@ -1257,24 +1269,36 @@ static void __ftrace_trace_stack(struct ring_buffer *buffer, trace.skip = skip; trace.entries = entry->caller; - save_stack_trace(&trace); + if (regs) + save_stack_trace_regs(regs, &trace); + else + save_stack_trace(&trace); if (!filter_check_discard(call, entry, buffer, event)) ring_buffer_unlock_commit(buffer, event); } +void ftrace_trace_stack_regs(struct ring_buffer *buffer, unsigned long flags, + int skip, int pc, struct pt_regs *regs) +{ + if (!(trace_flags & TRACE_ITER_STACKTRACE)) + return; + + __ftrace_trace_stack(buffer, flags, skip, pc, regs); +} + void ftrace_trace_stack(struct ring_buffer *buffer, unsigned long flags, int skip, int pc) { if (!(trace_flags & TRACE_ITER_STACKTRACE)) return; - __ftrace_trace_stack(buffer, flags, skip, pc); + __ftrace_trace_stack(buffer, flags, skip, pc, NULL); } void __trace_stack(struct trace_array *tr, unsigned long flags, int skip, int pc) { - __ftrace_trace_stack(tr->buffer, flags, skip, pc); + __ftrace_trace_stack(tr->buffer, flags, skip, pc, NULL); } /** @@ -1290,7 +1314,7 @@ void trace_dump_stack(void) local_save_flags(flags); /* skipping 3 traces, seems to get us at the caller of this function */ - __ftrace_trace_stack(global_trace.buffer, flags, 3, preempt_count()); + __ftrace_trace_stack(global_trace.buffer, flags, 3, preempt_count(), NULL); } static DEFINE_PER_CPU(int, user_stack_count); diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 742f545ae185..a3e2db708072 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -389,6 +389,9 @@ void update_max_tr_single(struct trace_array *tr, void ftrace_trace_stack(struct ring_buffer *buffer, unsigned long flags, int skip, int pc); +void ftrace_trace_stack_regs(struct ring_buffer *buffer, unsigned long flags, + int skip, int pc, struct pt_regs *regs); + void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags, int pc); @@ -400,6 +403,12 @@ static inline void ftrace_trace_stack(struct ring_buffer *buffer, { } +static inline void ftrace_trace_stack_regs(struct ring_buffer *buffer, + unsigned long flags, int skip, + int pc, struct pt_regs *regs) +{ +} + static inline void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags, int pc) { diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index f925c45f0afa..7053ef3d73d2 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -1397,7 +1397,8 @@ static __kprobes void kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs) store_trace_args(sizeof(*entry), tp, regs, (u8 *)&entry[1], dsize); if (!filter_current_check_discard(buffer, call, entry, event)) - trace_nowake_buffer_unlock_commit(buffer, event, irq_flags, pc); + trace_nowake_buffer_unlock_commit_regs(buffer, event, + irq_flags, pc, regs); } /* Kretprobe handler */ @@ -1429,7 +1430,8 @@ static __kprobes void kretprobe_trace_func(struct kretprobe_instance *ri, store_trace_args(sizeof(*entry), tp, regs, (u8 *)&entry[1], dsize); if (!filter_current_check_discard(buffer, call, entry, event)) - trace_nowake_buffer_unlock_commit(buffer, event, irq_flags, pc); + trace_nowake_buffer_unlock_commit_regs(buffer, event, + irq_flags, pc, regs); } /* Event entry printers */ -- cgit v1.2.3 From 4ec8363dfc1451f8c8f86825731fe712798ada02 Mon Sep 17 00:00:00 2001 From: Vince Weaver Date: Wed, 1 Jun 2011 15:15:36 -0400 Subject: perf_events: Fix perf buffer watermark setting Since 2.6.36 (specifically commit d57e34fdd60b ("perf: Simplify the ring-buffer logic: make perf_buffer_alloc() do everything needed"), the perf_buffer_init_code() has been mis-setting the buffer watermark if perf_event_attr.wakeup_events has a non-zero value. This is because perf_event_attr.wakeup_events is a union with perf_event_attr.wakeup_watermark. This commit re-enables the check for perf_event_attr.watermark being set before continuing with setting a non-default watermark. This bug is most noticable when you are trying to use PERF_IOC_REFRESH with a value larger than one and perf_event_attr.wakeup_events is set to one. In this case the buffer watermark will be set to 1 and you will get extraneous POLL_IN overflows rather than POLL_HUP as expected. [ avoid using attr.wakeup_events when attr.watermark is set ] Signed-off-by: Vince Weaver Signed-off-by: Peter Zijlstra Cc: Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1106011506390.5384@cl320.eecs.utk.edu Signed-off-by: Ingo Molnar --- kernel/events/core.c | 6 ++++-- kernel/events/ring_buffer.c | 16 +++++++++------- 2 files changed, 13 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 5e70f62752a2..e4aee519572d 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3569,8 +3569,10 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) if (vma->vm_flags & VM_WRITE) flags |= RING_BUFFER_WRITABLE; - rb = rb_alloc(nr_pages, event->attr.wakeup_watermark, - event->cpu, flags); + rb = rb_alloc(nr_pages, + event->attr.watermark ? event->attr.wakeup_watermark : 0, + event->cpu, flags); + if (!rb) { ret = -ENOMEM; goto unlock; diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c index fde52595d8f7..fc2701c99207 100644 --- a/kernel/events/ring_buffer.c +++ b/kernel/events/ring_buffer.c @@ -199,13 +199,15 @@ void perf_output_end(struct perf_output_handle *handle) struct perf_event *event = handle->event; struct ring_buffer *rb = handle->rb; - int wakeup_events = event->attr.wakeup_events; - - if (handle->sample && wakeup_events) { - int events = local_inc_return(&rb->events); - if (events >= wakeup_events) { - local_sub(wakeup_events, &rb->events); - local_inc(&rb->wakeup); + if (handle->sample && !event->attr.watermark) { + int wakeup_events = event->attr.wakeup_events; + + if (wakeup_events) { + int events = local_inc_return(&rb->events); + if (events >= wakeup_events) { + local_sub(wakeup_events, &rb->events); + local_inc(&rb->wakeup); + } } } -- cgit v1.2.3 From b7526f0ca6dc68f57ca467ce503151b1d476a3e4 Mon Sep 17 00:00:00 2001 From: Eric B Munson Date: Thu, 23 Jun 2011 16:34:37 -0400 Subject: events: Add note to update_event_times comment about holding ctx->lock Signed-off-by: Eric B Munson Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1308861279-15216-1-git-send-email-emunson@mgebm.net Signed-off-by: Ingo Molnar --- kernel/events/core.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index e4aee519572d..2293e0b084a9 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -748,6 +748,7 @@ static u64 perf_event_time(struct perf_event *event) /* * Update the total_time_enabled and total_time_running fields for a event. + * The caller of this function needs to hold the ctx->lock. */ static void update_event_times(struct perf_event *event) { -- cgit v1.2.3 From c4794295917ebeda8013b6cb9c8d71ab4f74a1fa Mon Sep 17 00:00:00 2001 From: Eric B Munson Date: Thu, 23 Jun 2011 16:34:38 -0400 Subject: events: Move lockless timer calculation into helper function Take the timer calculation from perf_output_read and move it to a helper function for any place that needs timer values but cannot take the ctx->lock. Signed-off-by: Eric B Munson Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1308861279-15216-2-git-send-email-emunson@mgebm.net Signed-off-by: Ingo Molnar --- kernel/events/core.c | 22 +++++++++++++++------- 1 file changed, 15 insertions(+), 7 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 2293e0b084a9..c851d707821f 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3351,6 +3351,18 @@ static int perf_event_index(struct perf_event *event) return event->hw.idx + 1 - PERF_EVENT_INDEX_OFFSET; } +static void calc_timer_values(struct perf_event *event, + u64 *running, + u64 *enabled) +{ + u64 now, ctx_time; + + now = perf_clock(); + ctx_time = event->shadow_ctx_time + now; + *enabled = ctx_time - event->tstamp_enabled; + *running = ctx_time - event->tstamp_running; +} + /* * Callers need to ensure there can be no nesting of this function, otherwise * the seqlock logic goes bad. We can not serialize this because the arch @@ -3816,7 +3828,7 @@ static void perf_output_read_group(struct perf_output_handle *handle, static void perf_output_read(struct perf_output_handle *handle, struct perf_event *event) { - u64 enabled = 0, running = 0, now, ctx_time; + u64 enabled = 0, running = 0; u64 read_format = event->attr.read_format; /* @@ -3828,12 +3840,8 @@ static void perf_output_read(struct perf_output_handle *handle, * because of locking issue as we are called in * NMI context */ - if (read_format & PERF_FORMAT_TOTAL_TIMES) { - now = perf_clock(); - ctx_time = event->shadow_ctx_time + now; - enabled = ctx_time - event->tstamp_enabled; - running = ctx_time - event->tstamp_running; - } + if (read_format & PERF_FORMAT_TOTAL_TIMES) + calc_timer_values(event, &enabled, &running); if (event->attr.read_format & PERF_FORMAT_GROUP) perf_output_read_group(handle, event, enabled, running); -- cgit v1.2.3 From 0d6412085b7ff58612af52e51ffa864f0df4b8fd Mon Sep 17 00:00:00 2001 From: Eric B Munson Date: Fri, 24 Jun 2011 12:26:26 -0400 Subject: events: Ensure that timers are updated without requiring read() call The event tracing infrastructure exposes two timers which should be updated each time the value of the counter is updated. Currently, these counters are only updated when userspace calls read() on the fd associated with an event. This means that counters which are read via the mmap'd page exclusively never have their timers updated. This patch adds ensures that the timers are updated each time the values in the mmap'd page are updated. Signed-off-by: Eric B Munson Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1308932786-5111-1-git-send-email-emunson@mgebm.net Signed-off-by: Ingo Molnar --- kernel/events/core.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index c851d707821f..270e32f9fc06 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3372,8 +3372,19 @@ void perf_event_update_userpage(struct perf_event *event) { struct perf_event_mmap_page *userpg; struct ring_buffer *rb; + u64 enabled, running; rcu_read_lock(); + /* + * compute total_time_enabled, total_time_running + * based on snapshot values taken when the event + * was last scheduled in. + * + * we cannot simply called update_context_time() + * because of locking issue as we can be called in + * NMI context + */ + calc_timer_values(event, &enabled, &running); rb = rcu_dereference(event->rb); if (!rb) goto unlock; @@ -3392,10 +3403,10 @@ void perf_event_update_userpage(struct perf_event *event) if (event->state == PERF_EVENT_STATE_ACTIVE) userpg->offset -= local64_read(&event->hw.prev_count); - userpg->time_enabled = event->total_time_enabled + + userpg->time_enabled = enabled + atomic64_read(&event->child_total_time_enabled); - userpg->time_running = event->total_time_running + + userpg->time_running = running + atomic64_read(&event->child_total_time_running); barrier(); -- cgit v1.2.3 From 1880c4ae182afb5650c5678949ecfe7ff66a724e Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Thu, 23 Jun 2011 16:49:18 +0400 Subject: perf, x86: Add hw_watchdog_set_attr() in a sake of nmi-watchdog on P4 Due to restriction and specifics of Netburst PMU we need a separated event for NMI watchdog. In particular every Netburst event consumes not just a counter and a config register, but also an additional ESCR register. Since ESCR registers are grouped upon counters (i.e. if ESCR is occupied for some event there is no room for another event to enter until its released) we need to pick up the "least" used ESCR (or the most available one) for nmi-watchdog purposes -- so MSR_P4_CRU_ESCR2/3 was chosen. With this patch nmi-watchdog and perf top should be able to run simultaneously. Signed-off-by: Cyrill Gorcunov CC: Lin Ming CC: Arnaldo Carvalho de Melo CC: Frederic Weisbecker Tested-and-reviewed-by: Don Zickus Tested-and-reviewed-by: Stephane Eranian Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20110623124918.GC13050@sun Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 7 +++++++ arch/x86/kernel/cpu/perf_event_p4.c | 26 ++++++++++++++++++++++++++ kernel/watchdog.c | 6 +++++- 3 files changed, 38 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 3a0338b4b179..8a57f9aa8e36 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -233,6 +233,7 @@ struct x86_pmu { void (*enable_all)(int added); void (*enable)(struct perf_event *); void (*disable)(struct perf_event *); + void (*hw_watchdog_set_attr)(struct perf_event_attr *attr); int (*hw_config)(struct perf_event *event); int (*schedule_events)(struct cpu_hw_events *cpuc, int n, int *assign); unsigned eventsel; @@ -315,6 +316,12 @@ static u64 __read_mostly hw_cache_extra_regs [PERF_COUNT_HW_CACHE_OP_MAX] [PERF_COUNT_HW_CACHE_RESULT_MAX]; +void hw_nmi_watchdog_set_attr(struct perf_event_attr *wd_attr) +{ + if (x86_pmu.hw_watchdog_set_attr) + x86_pmu.hw_watchdog_set_attr(wd_attr); +} + /* * Propagate event elapsed time into the generic event. * Can only be executed on the CPU where the event is active. diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c index ead584fb6a7d..f76fddf63381 100644 --- a/arch/x86/kernel/cpu/perf_event_p4.c +++ b/arch/x86/kernel/cpu/perf_event_p4.c @@ -705,6 +705,31 @@ static int p4_validate_raw_event(struct perf_event *event) return 0; } +static void p4_hw_watchdog_set_attr(struct perf_event_attr *wd_attr) +{ + /* + * Watchdog ticks are special on Netburst, we use + * that named "non-sleeping" ticks as recommended + * by Intel SDM Vol3b. + */ + WARN_ON_ONCE(wd_attr->type != PERF_TYPE_HARDWARE || + wd_attr->config != PERF_COUNT_HW_CPU_CYCLES); + + wd_attr->type = PERF_TYPE_RAW; + wd_attr->config = + p4_config_pack_escr(P4_ESCR_EVENT(P4_EVENT_EXECUTION_EVENT) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS0) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS1) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS2) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS3) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS0) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS1) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS2) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS3)) | + p4_config_pack_cccr(P4_CCCR_THRESHOLD(15) | P4_CCCR_COMPLEMENT | + P4_CCCR_COMPARE); +} + static int p4_hw_config(struct perf_event *event) { int cpu = get_cpu(); @@ -1179,6 +1204,7 @@ static __initconst const struct x86_pmu p4_pmu = { .cntval_bits = ARCH_P4_CNTRVAL_BITS, .cntval_mask = ARCH_P4_CNTRVAL_MASK, .max_period = (1ULL << (ARCH_P4_CNTRVAL_BITS - 1)) - 1, + .hw_watchdog_set_attr = p4_hw_watchdog_set_attr, .hw_config = p4_hw_config, .schedule_events = p4_pmu_schedule_events, /* diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 3d0c56ad4792..752b75ba662b 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -200,6 +200,8 @@ static int is_softlockup(unsigned long touch_ts) } #ifdef CONFIG_HARDLOCKUP_DETECTOR +void __weak hw_nmi_watchdog_set_attr(struct perf_event_attr *wd_attr) { } + static struct perf_event_attr wd_hw_attr = { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, @@ -368,9 +370,11 @@ static int watchdog_nmi_enable(int cpu) if (event != NULL) goto out_enable; - /* Try to register using hardware perf events */ wd_attr = &wd_hw_attr; wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh); + hw_nmi_watchdog_set_attr(wd_attr); + + /* Try to register using hardware perf events */ event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback); if (!IS_ERR(event)) { printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n"); -- cgit v1.2.3 From a8b0ca17b80e92faab46ee7179ba9e99ccb61233 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 27 Jun 2011 14:41:57 +0200 Subject: perf: Remove the nmi parameter from the swevent and overflow interface The nmi parameter indicated if we could do wakeups from the current context, if not, we would set some state and self-IPI and let the resulting interrupt do the wakeup. For the various event classes: - hardware: nmi=0; PMI is in fact an NMI or we run irq_work_run from the PMI-tail (ARM etc.) - tracepoint: nmi=0; since tracepoint could be from NMI context. - software: nmi=[0,1]; some, like the schedule thing cannot perform wakeups, and hence need 0. As one can see, there is very little nmi=1 usage, and the down-side of not using it is that on some platforms some software events can have a jiffy delay in wakeup (when arch_irq_work_raise isn't implemented). The up-side however is that we can remove the nmi parameter and save a bunch of conditionals in fast paths. Signed-off-by: Peter Zijlstra Cc: Michael Cree Cc: Will Deacon Cc: Deng-Cheng Zhu Cc: Anton Blanchard Cc: Eric B Munson Cc: Heiko Carstens Cc: Paul Mundt Cc: David S. Miller Cc: Frederic Weisbecker Cc: Jason Wessel Cc: Don Zickus Link: http://lkml.kernel.org/n/tip-agjev8eu666tvknpb3iaj0fg@git.kernel.org Signed-off-by: Ingo Molnar --- arch/alpha/kernel/perf_event.c | 2 +- arch/arm/kernel/perf_event_v6.c | 2 +- arch/arm/kernel/perf_event_v7.c | 2 +- arch/arm/kernel/perf_event_xscale.c | 4 +- arch/arm/kernel/ptrace.c | 2 +- arch/arm/kernel/swp_emulate.c | 2 +- arch/arm/mm/fault.c | 6 +-- arch/mips/kernel/perf_event.c | 2 +- arch/mips/kernel/traps.c | 8 ++-- arch/mips/kernel/unaligned.c | 5 +-- arch/mips/math-emu/cp1emu.c | 3 +- arch/mips/mm/fault.c | 8 ++-- arch/powerpc/include/asm/emulated_ops.h | 4 +- arch/powerpc/kernel/perf_event.c | 6 +-- arch/powerpc/kernel/perf_event_fsl_emb.c | 6 +-- arch/powerpc/kernel/ptrace.c | 2 +- arch/powerpc/mm/fault.c | 6 +-- arch/s390/mm/fault.c | 6 +-- arch/sh/kernel/ptrace_32.c | 2 +- arch/sh/kernel/traps_32.c | 2 +- arch/sh/kernel/traps_64.c | 8 ++-- arch/sh/math-emu/math.c | 2 +- arch/sh/mm/fault_32.c | 6 +-- arch/sh/mm/tlbflush_64.c | 6 +-- arch/sparc/kernel/perf_event.c | 2 +- arch/sparc/kernel/unaligned_32.c | 4 +- arch/sparc/kernel/unaligned_64.c | 12 +++--- arch/sparc/kernel/visemul.c | 2 +- arch/sparc/math-emu/math_32.c | 2 +- arch/sparc/math-emu/math_64.c | 2 +- arch/sparc/mm/fault_32.c | 8 ++-- arch/sparc/mm/fault_64.c | 8 ++-- arch/x86/kernel/cpu/perf_event.c | 2 +- arch/x86/kernel/cpu/perf_event_intel.c | 2 +- arch/x86/kernel/cpu/perf_event_intel_ds.c | 4 +- arch/x86/kernel/cpu/perf_event_p4.c | 2 +- arch/x86/kernel/kgdb.c | 2 +- arch/x86/kernel/ptrace.c | 2 +- arch/x86/mm/fault.c | 6 +-- include/linux/perf_event.h | 18 ++++----- kernel/events/core.c | 63 ++++++++++++++----------------- kernel/events/internal.h | 1 - kernel/events/ring_buffer.c | 10 ++--- kernel/sched.c | 2 +- kernel/watchdog.c | 2 +- samples/hw_breakpoint/data_breakpoint.c | 2 +- 46 files changed, 119 insertions(+), 141 deletions(-) (limited to 'kernel') diff --git a/arch/alpha/kernel/perf_event.c b/arch/alpha/kernel/perf_event.c index 90561c45e7d8..8e47709160f8 100644 --- a/arch/alpha/kernel/perf_event.c +++ b/arch/alpha/kernel/perf_event.c @@ -847,7 +847,7 @@ static void alpha_perf_event_irq_handler(unsigned long la_ptr, data.period = event->hw.last_period; if (alpha_perf_event_set_period(event, hwc, idx)) { - if (perf_event_overflow(event, 1, &data, regs)) { + if (perf_event_overflow(event, &data, regs)) { /* Interrupts coming too quickly; "throttle" the * counter, i.e., disable it for a little while. */ diff --git a/arch/arm/kernel/perf_event_v6.c b/arch/arm/kernel/perf_event_v6.c index f1e8dd94afe8..38dc4da1d612 100644 --- a/arch/arm/kernel/perf_event_v6.c +++ b/arch/arm/kernel/perf_event_v6.c @@ -479,7 +479,7 @@ armv6pmu_handle_irq(int irq_num, if (!armpmu_event_set_period(event, hwc, idx)) continue; - if (perf_event_overflow(event, 0, &data, regs)) + if (perf_event_overflow(event, &data, regs)) armpmu->disable(hwc, idx); } diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c index 4960686afb58..6e5f8752303b 100644 --- a/arch/arm/kernel/perf_event_v7.c +++ b/arch/arm/kernel/perf_event_v7.c @@ -787,7 +787,7 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev) if (!armpmu_event_set_period(event, hwc, idx)) continue; - if (perf_event_overflow(event, 0, &data, regs)) + if (perf_event_overflow(event, &data, regs)) armpmu->disable(hwc, idx); } diff --git a/arch/arm/kernel/perf_event_xscale.c b/arch/arm/kernel/perf_event_xscale.c index 39affbe4fdb2..99b6b85c7e49 100644 --- a/arch/arm/kernel/perf_event_xscale.c +++ b/arch/arm/kernel/perf_event_xscale.c @@ -251,7 +251,7 @@ xscale1pmu_handle_irq(int irq_num, void *dev) if (!armpmu_event_set_period(event, hwc, idx)) continue; - if (perf_event_overflow(event, 0, &data, regs)) + if (perf_event_overflow(event, &data, regs)) armpmu->disable(hwc, idx); } @@ -583,7 +583,7 @@ xscale2pmu_handle_irq(int irq_num, void *dev) if (!armpmu_event_set_period(event, hwc, idx)) continue; - if (perf_event_overflow(event, 0, &data, regs)) + if (perf_event_overflow(event, &data, regs)) armpmu->disable(hwc, idx); } diff --git a/arch/arm/kernel/ptrace.c b/arch/arm/kernel/ptrace.c index 97260060bf26..0c9b1054f790 100644 --- a/arch/arm/kernel/ptrace.c +++ b/arch/arm/kernel/ptrace.c @@ -396,7 +396,7 @@ static long ptrace_hbp_idx_to_num(int idx) /* * Handle hitting a HW-breakpoint. */ -static void ptrace_hbptriggered(struct perf_event *bp, int unused, +static void ptrace_hbptriggered(struct perf_event *bp, struct perf_sample_data *data, struct pt_regs *regs) { diff --git a/arch/arm/kernel/swp_emulate.c b/arch/arm/kernel/swp_emulate.c index 40ee7e5045e4..5f452f8fde05 100644 --- a/arch/arm/kernel/swp_emulate.c +++ b/arch/arm/kernel/swp_emulate.c @@ -183,7 +183,7 @@ static int swp_handler(struct pt_regs *regs, unsigned int instr) unsigned int address, destreg, data, type; unsigned int res = 0; - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, regs->ARM_pc); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, regs->ARM_pc); if (current->pid != previous_pid) { pr_debug("\"%s\" (%ld) uses deprecated SWP{B} instruction\n", diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c index bc0e1d88fd3b..9ea4f7ddd665 100644 --- a/arch/arm/mm/fault.c +++ b/arch/arm/mm/fault.c @@ -318,11 +318,11 @@ do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs) fault = __do_page_fault(mm, addr, fsr, tsk); up_read(&mm->mmap_sem); - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, addr); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); if (fault & VM_FAULT_MAJOR) - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, regs, addr); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, addr); else if (fault & VM_FAULT_MINOR) - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, regs, addr); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, addr); /* * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR diff --git a/arch/mips/kernel/perf_event.c b/arch/mips/kernel/perf_event.c index a8244854d3dc..d0deaab9ace2 100644 --- a/arch/mips/kernel/perf_event.c +++ b/arch/mips/kernel/perf_event.c @@ -527,7 +527,7 @@ handle_associated_event(struct cpu_hw_events *cpuc, if (!mipspmu_event_set_period(event, hwc, idx)) return; - if (perf_event_overflow(event, 0, data, regs)) + if (perf_event_overflow(event, data, regs)) mipspmu->disable_event(idx); } diff --git a/arch/mips/kernel/traps.c b/arch/mips/kernel/traps.c index e9b3af27d844..b7517e3abc85 100644 --- a/arch/mips/kernel/traps.c +++ b/arch/mips/kernel/traps.c @@ -578,12 +578,12 @@ static int simulate_llsc(struct pt_regs *regs, unsigned int opcode) { if ((opcode & OPCODE) == LL) { perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, - 1, 0, regs, 0); + 1, regs, 0); return simulate_ll(regs, opcode); } if ((opcode & OPCODE) == SC) { perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, - 1, 0, regs, 0); + 1, regs, 0); return simulate_sc(regs, opcode); } @@ -602,7 +602,7 @@ static int simulate_rdhwr(struct pt_regs *regs, unsigned int opcode) int rd = (opcode & RD) >> 11; int rt = (opcode & RT) >> 16; perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, - 1, 0, regs, 0); + 1, regs, 0); switch (rd) { case 0: /* CPU number */ regs->regs[rt] = smp_processor_id(); @@ -640,7 +640,7 @@ static int simulate_sync(struct pt_regs *regs, unsigned int opcode) { if ((opcode & OPCODE) == SPEC0 && (opcode & FUNC) == SYNC) { perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, - 1, 0, regs, 0); + 1, regs, 0); return 0; } diff --git a/arch/mips/kernel/unaligned.c b/arch/mips/kernel/unaligned.c index cfea1adfa153..eb319b580353 100644 --- a/arch/mips/kernel/unaligned.c +++ b/arch/mips/kernel/unaligned.c @@ -111,8 +111,7 @@ static void emulate_load_store_insn(struct pt_regs *regs, unsigned long value; unsigned int res; - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, - 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); /* * This load never faults. @@ -517,7 +516,7 @@ asmlinkage void do_ade(struct pt_regs *regs) mm_segment_t seg; perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, - 1, 0, regs, regs->cp0_badvaddr); + 1, regs, regs->cp0_badvaddr); /* * Did we catch a fault trying to load an instruction? * Or are we running in MIPS16 mode? diff --git a/arch/mips/math-emu/cp1emu.c b/arch/mips/math-emu/cp1emu.c index d32cb0503110..dbf2f93a5091 100644 --- a/arch/mips/math-emu/cp1emu.c +++ b/arch/mips/math-emu/cp1emu.c @@ -272,8 +272,7 @@ static int cop1Emulate(struct pt_regs *xcp, struct mips_fpu_struct *ctx, } emul: - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, - 1, 0, xcp, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, xcp, 0); MIPS_FPU_EMU_INC_STATS(emulated); switch (MIPSInst_OPCODE(ir)) { case ldc1_op:{ diff --git a/arch/mips/mm/fault.c b/arch/mips/mm/fault.c index 137ee76a0045..937cf3368164 100644 --- a/arch/mips/mm/fault.c +++ b/arch/mips/mm/fault.c @@ -145,7 +145,7 @@ good_area: * the fault. */ fault = handle_mm_fault(mm, vma, address, write ? FAULT_FLAG_WRITE : 0); - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); if (unlikely(fault & VM_FAULT_ERROR)) { if (fault & VM_FAULT_OOM) goto out_of_memory; @@ -154,12 +154,10 @@ good_area: BUG(); } if (fault & VM_FAULT_MAJOR) { - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, - 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); tsk->maj_flt++; } else { - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, - 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); tsk->min_flt++; } diff --git a/arch/powerpc/include/asm/emulated_ops.h b/arch/powerpc/include/asm/emulated_ops.h index 45921672b97a..2cc41c715d2b 100644 --- a/arch/powerpc/include/asm/emulated_ops.h +++ b/arch/powerpc/include/asm/emulated_ops.h @@ -78,14 +78,14 @@ extern void ppc_warn_emulated_print(const char *type); #define PPC_WARN_EMULATED(type, regs) \ do { \ perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, \ - 1, 0, regs, 0); \ + 1, regs, 0); \ __PPC_WARN_EMULATED(type); \ } while (0) #define PPC_WARN_ALIGNMENT(type, regs) \ do { \ perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, \ - 1, 0, regs, regs->dar); \ + 1, regs, regs->dar); \ __PPC_WARN_EMULATED(type); \ } while (0) diff --git a/arch/powerpc/kernel/perf_event.c b/arch/powerpc/kernel/perf_event.c index 822f63008ae1..14967de98876 100644 --- a/arch/powerpc/kernel/perf_event.c +++ b/arch/powerpc/kernel/perf_event.c @@ -1207,7 +1207,7 @@ struct pmu power_pmu = { * here so there is no possibility of being interrupted. */ static void record_and_restart(struct perf_event *event, unsigned long val, - struct pt_regs *regs, int nmi) + struct pt_regs *regs) { u64 period = event->hw.sample_period; s64 prev, delta, left; @@ -1258,7 +1258,7 @@ static void record_and_restart(struct perf_event *event, unsigned long val, if (event->attr.sample_type & PERF_SAMPLE_ADDR) perf_get_data_addr(regs, &data.addr); - if (perf_event_overflow(event, nmi, &data, regs)) + if (perf_event_overflow(event, &data, regs)) power_pmu_stop(event, 0); } } @@ -1346,7 +1346,7 @@ static void perf_event_interrupt(struct pt_regs *regs) if ((int)val < 0) { /* event has overflowed */ found = 1; - record_and_restart(event, val, regs, nmi); + record_and_restart(event, val, regs); } } diff --git a/arch/powerpc/kernel/perf_event_fsl_emb.c b/arch/powerpc/kernel/perf_event_fsl_emb.c index b0dc8f7069cd..0a6d2a9d569c 100644 --- a/arch/powerpc/kernel/perf_event_fsl_emb.c +++ b/arch/powerpc/kernel/perf_event_fsl_emb.c @@ -568,7 +568,7 @@ static struct pmu fsl_emb_pmu = { * here so there is no possibility of being interrupted. */ static void record_and_restart(struct perf_event *event, unsigned long val, - struct pt_regs *regs, int nmi) + struct pt_regs *regs) { u64 period = event->hw.sample_period; s64 prev, delta, left; @@ -616,7 +616,7 @@ static void record_and_restart(struct perf_event *event, unsigned long val, perf_sample_data_init(&data, 0); data.period = event->hw.last_period; - if (perf_event_overflow(event, nmi, &data, regs)) + if (perf_event_overflow(event, &data, regs)) fsl_emb_pmu_stop(event, 0); } } @@ -644,7 +644,7 @@ static void perf_event_interrupt(struct pt_regs *regs) if (event) { /* event has overflowed */ found = 1; - record_and_restart(event, val, regs, nmi); + record_and_restart(event, val, regs); } else { /* * Disabled counter is negative, diff --git a/arch/powerpc/kernel/ptrace.c b/arch/powerpc/kernel/ptrace.c index cb22024f2b42..3177617af2ef 100644 --- a/arch/powerpc/kernel/ptrace.c +++ b/arch/powerpc/kernel/ptrace.c @@ -882,7 +882,7 @@ void user_disable_single_step(struct task_struct *task) } #ifdef CONFIG_HAVE_HW_BREAKPOINT -void ptrace_triggered(struct perf_event *bp, int nmi, +void ptrace_triggered(struct perf_event *bp, struct perf_sample_data *data, struct pt_regs *regs) { struct perf_event_attr attr; diff --git a/arch/powerpc/mm/fault.c b/arch/powerpc/mm/fault.c index 54f4fb994e99..dbc48254c6cc 100644 --- a/arch/powerpc/mm/fault.c +++ b/arch/powerpc/mm/fault.c @@ -173,7 +173,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, die("Weird page fault", regs, SIGSEGV); } - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); /* When running in the kernel we expect faults to occur only to * addresses in user space. All other faults represent errors in the @@ -319,7 +319,7 @@ good_area: } if (ret & VM_FAULT_MAJOR) { current->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); #ifdef CONFIG_PPC_SMLPAR if (firmware_has_feature(FW_FEATURE_CMO)) { @@ -330,7 +330,7 @@ good_area: #endif } else { current->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } up_read(&mm->mmap_sem); diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c index fe103e891e7a..095f782a5512 100644 --- a/arch/s390/mm/fault.c +++ b/arch/s390/mm/fault.c @@ -299,7 +299,7 @@ static inline int do_exception(struct pt_regs *regs, int access, goto out; address = trans_exc_code & __FAIL_ADDR_MASK; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); flags = FAULT_FLAG_ALLOW_RETRY; if (access == VM_WRITE || (trans_exc_code & store_indication) == 0x400) flags |= FAULT_FLAG_WRITE; @@ -345,11 +345,11 @@ retry: if (flags & FAULT_FLAG_ALLOW_RETRY) { if (fault & VM_FAULT_MAJOR) { tsk->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); } else { tsk->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } if (fault & VM_FAULT_RETRY) { diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c index 3d7b209b2178..8051976100a6 100644 --- a/arch/sh/kernel/ptrace_32.c +++ b/arch/sh/kernel/ptrace_32.c @@ -63,7 +63,7 @@ static inline int put_stack_long(struct task_struct *task, int offset, return 0; } -void ptrace_triggered(struct perf_event *bp, int nmi, +void ptrace_triggered(struct perf_event *bp, struct perf_sample_data *data, struct pt_regs *regs) { struct perf_event_attr attr; diff --git a/arch/sh/kernel/traps_32.c b/arch/sh/kernel/traps_32.c index b51a17104b5f..d9006f8ffc14 100644 --- a/arch/sh/kernel/traps_32.c +++ b/arch/sh/kernel/traps_32.c @@ -393,7 +393,7 @@ int handle_unaligned_access(insn_size_t instruction, struct pt_regs *regs, */ if (!expected) { unaligned_fixups_notify(current, instruction, regs); - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, address); } diff --git a/arch/sh/kernel/traps_64.c b/arch/sh/kernel/traps_64.c index 6713ca97e553..67110be83fd7 100644 --- a/arch/sh/kernel/traps_64.c +++ b/arch/sh/kernel/traps_64.c @@ -434,7 +434,7 @@ static int misaligned_load(struct pt_regs *regs, return error; } - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, address); destreg = (opcode >> 4) & 0x3f; if (user_mode(regs)) { @@ -512,7 +512,7 @@ static int misaligned_store(struct pt_regs *regs, return error; } - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, address); srcreg = (opcode >> 4) & 0x3f; if (user_mode(regs)) { @@ -588,7 +588,7 @@ static int misaligned_fpu_load(struct pt_regs *regs, return error; } - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, address); destreg = (opcode >> 4) & 0x3f; if (user_mode(regs)) { @@ -665,7 +665,7 @@ static int misaligned_fpu_store(struct pt_regs *regs, return error; } - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, address); srcreg = (opcode >> 4) & 0x3f; if (user_mode(regs)) { diff --git a/arch/sh/math-emu/math.c b/arch/sh/math-emu/math.c index f76a5090d5d1..977195210653 100644 --- a/arch/sh/math-emu/math.c +++ b/arch/sh/math-emu/math.c @@ -620,7 +620,7 @@ int do_fpu_inst(unsigned short inst, struct pt_regs *regs) struct task_struct *tsk = current; struct sh_fpu_soft_struct *fpu = &(tsk->thread.xstate->softfpu); - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); if (!(task_thread_info(tsk)->status & TS_USEDFPU)) { /* initialize once. */ diff --git a/arch/sh/mm/fault_32.c b/arch/sh/mm/fault_32.c index d4c34d757f0d..7bebd044f2a1 100644 --- a/arch/sh/mm/fault_32.c +++ b/arch/sh/mm/fault_32.c @@ -160,7 +160,7 @@ asmlinkage void __kprobes do_page_fault(struct pt_regs *regs, if ((regs->sr & SR_IMASK) != SR_IMASK) local_irq_enable(); - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); /* * If we're in an interrupt, have no user context or are running @@ -210,11 +210,11 @@ good_area: } if (fault & VM_FAULT_MAJOR) { tsk->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); } else { tsk->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } diff --git a/arch/sh/mm/tlbflush_64.c b/arch/sh/mm/tlbflush_64.c index 7f5810f5dfdc..e3430e093d43 100644 --- a/arch/sh/mm/tlbflush_64.c +++ b/arch/sh/mm/tlbflush_64.c @@ -116,7 +116,7 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long writeaccess, /* Not an IO address, so reenable interrupts */ local_irq_enable(); - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); /* * If we're in an interrupt or have no user @@ -200,11 +200,11 @@ good_area: if (fault & VM_FAULT_MAJOR) { tsk->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); } else { tsk->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c index 2cb0e1c001e2..0b32f2e9e08d 100644 --- a/arch/sparc/kernel/perf_event.c +++ b/arch/sparc/kernel/perf_event.c @@ -1277,7 +1277,7 @@ static int __kprobes perf_event_nmi_handler(struct notifier_block *self, if (!sparc_perf_event_set_period(event, hwc, idx)) continue; - if (perf_event_overflow(event, 1, &data, regs)) + if (perf_event_overflow(event, &data, regs)) sparc_pmu_stop(event, 0); } diff --git a/arch/sparc/kernel/unaligned_32.c b/arch/sparc/kernel/unaligned_32.c index 4491f4cb2695..7efbb2f9e77f 100644 --- a/arch/sparc/kernel/unaligned_32.c +++ b/arch/sparc/kernel/unaligned_32.c @@ -247,7 +247,7 @@ asmlinkage void kernel_unaligned_trap(struct pt_regs *regs, unsigned int insn) unsigned long addr = compute_effective_address(regs, insn); int err; - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, addr); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, addr); switch (dir) { case load: err = do_int_load(fetch_reg_addr(((insn>>25)&0x1f), @@ -338,7 +338,7 @@ asmlinkage void user_unaligned_trap(struct pt_regs *regs, unsigned int insn) } addr = compute_effective_address(regs, insn); - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, addr); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, addr); switch(dir) { case load: err = do_int_load(fetch_reg_addr(((insn>>25)&0x1f), diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c index b2b019ea8caa..35cff1673aa4 100644 --- a/arch/sparc/kernel/unaligned_64.c +++ b/arch/sparc/kernel/unaligned_64.c @@ -317,7 +317,7 @@ asmlinkage void kernel_unaligned_trap(struct pt_regs *regs, unsigned int insn) addr = compute_effective_address(regs, insn, ((insn >> 25) & 0x1f)); - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, addr); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, addr); switch (asi) { case ASI_NL: case ASI_AIUPL: @@ -384,7 +384,7 @@ int handle_popc(u32 insn, struct pt_regs *regs) int ret, i, rd = ((insn >> 25) & 0x1f); int from_kernel = (regs->tstate & TSTATE_PRIV) != 0; - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); if (insn & 0x2000) { maybe_flush_windows(0, 0, rd, from_kernel); value = sign_extend_imm13(insn); @@ -431,7 +431,7 @@ int handle_ldf_stq(u32 insn, struct pt_regs *regs) int asi = decode_asi(insn, regs); int flag = (freg < 32) ? FPRS_DL : FPRS_DU; - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); save_and_clear_fpu(); current_thread_info()->xfsr[0] &= ~0x1c000; @@ -554,7 +554,7 @@ void handle_ld_nf(u32 insn, struct pt_regs *regs) int from_kernel = (regs->tstate & TSTATE_PRIV) != 0; unsigned long *reg; - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); maybe_flush_windows(0, 0, rd, from_kernel); reg = fetch_reg_addr(rd, regs); @@ -586,7 +586,7 @@ void handle_lddfmna(struct pt_regs *regs, unsigned long sfar, unsigned long sfsr if (tstate & TSTATE_PRIV) die_if_kernel("lddfmna from kernel", regs); - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, sfar); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, sfar); if (test_thread_flag(TIF_32BIT)) pc = (u32)pc; if (get_user(insn, (u32 __user *) pc) != -EFAULT) { @@ -647,7 +647,7 @@ void handle_stdfmna(struct pt_regs *regs, unsigned long sfar, unsigned long sfsr if (tstate & TSTATE_PRIV) die_if_kernel("stdfmna from kernel", regs); - perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, 0, regs, sfar); + perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, sfar); if (test_thread_flag(TIF_32BIT)) pc = (u32)pc; if (get_user(insn, (u32 __user *) pc) != -EFAULT) { diff --git a/arch/sparc/kernel/visemul.c b/arch/sparc/kernel/visemul.c index 36357717d691..32b626c9d815 100644 --- a/arch/sparc/kernel/visemul.c +++ b/arch/sparc/kernel/visemul.c @@ -802,7 +802,7 @@ int vis_emul(struct pt_regs *regs, unsigned int insn) BUG_ON(regs->tstate & TSTATE_PRIV); - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); if (test_thread_flag(TIF_32BIT)) pc = (u32)pc; diff --git a/arch/sparc/math-emu/math_32.c b/arch/sparc/math-emu/math_32.c index a3fccde894ec..aa4d55b0bdf0 100644 --- a/arch/sparc/math-emu/math_32.c +++ b/arch/sparc/math-emu/math_32.c @@ -164,7 +164,7 @@ int do_mathemu(struct pt_regs *regs, struct task_struct *fpt) int retcode = 0; /* assume all succeed */ unsigned long insn; - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); #ifdef DEBUG_MATHEMU printk("In do_mathemu()... pc is %08lx\n", regs->pc); diff --git a/arch/sparc/math-emu/math_64.c b/arch/sparc/math-emu/math_64.c index 56d2c44747b8..e575bd2fe381 100644 --- a/arch/sparc/math-emu/math_64.c +++ b/arch/sparc/math-emu/math_64.c @@ -184,7 +184,7 @@ int do_mathemu(struct pt_regs *regs, struct fpustate *f) if (tstate & TSTATE_PRIV) die_if_kernel("unfinished/unimplemented FPop from kernel", regs); - perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, 0, regs, 0); + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0); if (test_thread_flag(TIF_32BIT)) pc = (u32)pc; if (get_user(insn, (u32 __user *) pc) != -EFAULT) { diff --git a/arch/sparc/mm/fault_32.c b/arch/sparc/mm/fault_32.c index 7543ddbdadb2..aa1c1b1ce5cc 100644 --- a/arch/sparc/mm/fault_32.c +++ b/arch/sparc/mm/fault_32.c @@ -251,7 +251,7 @@ asmlinkage void do_sparc_fault(struct pt_regs *regs, int text_fault, int write, if (in_atomic() || !mm) goto no_context; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); down_read(&mm->mmap_sem); @@ -301,12 +301,10 @@ good_area: } if (fault & VM_FAULT_MAJOR) { current->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, - regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); } else { current->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, - regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } up_read(&mm->mmap_sem); return; diff --git a/arch/sparc/mm/fault_64.c b/arch/sparc/mm/fault_64.c index f92ce56a8b22..504c0622f729 100644 --- a/arch/sparc/mm/fault_64.c +++ b/arch/sparc/mm/fault_64.c @@ -325,7 +325,7 @@ asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs) if (in_atomic() || !mm) goto intr_or_no_mm; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); if (!down_read_trylock(&mm->mmap_sem)) { if ((regs->tstate & TSTATE_PRIV) && @@ -433,12 +433,10 @@ good_area: } if (fault & VM_FAULT_MAJOR) { current->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, - regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); } else { current->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, - regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } up_read(&mm->mmap_sem); diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 8a57f9aa8e36..5b86ec51534c 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1339,7 +1339,7 @@ static int x86_pmu_handle_irq(struct pt_regs *regs) if (!x86_perf_event_set_period(event)) continue; - if (perf_event_overflow(event, 1, &data, regs)) + if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); } diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 41178c826c48..d38b0020f775 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1003,7 +1003,7 @@ again: data.period = event->hw.last_period; - if (perf_event_overflow(event, 1, &data, regs)) + if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); } diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c index bab491b8ee25..0941f93f2940 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_ds.c +++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c @@ -340,7 +340,7 @@ static int intel_pmu_drain_bts_buffer(void) */ perf_prepare_sample(&header, &data, event, ®s); - if (perf_output_begin(&handle, event, header.size * (top - at), 1, 1)) + if (perf_output_begin(&handle, event, header.size * (top - at), 1)) return 1; for (; at < top; at++) { @@ -616,7 +616,7 @@ static void __intel_pmu_pebs_event(struct perf_event *event, else regs.flags &= ~PERF_EFLAGS_EXACT; - if (perf_event_overflow(event, 1, &data, ®s)) + if (perf_event_overflow(event, &data, ®s)) x86_pmu_stop(event, 0); } diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c index f76fddf63381..d6e6a67b9608 100644 --- a/arch/x86/kernel/cpu/perf_event_p4.c +++ b/arch/x86/kernel/cpu/perf_event_p4.c @@ -970,7 +970,7 @@ static int p4_pmu_handle_irq(struct pt_regs *regs) if (!x86_perf_event_set_period(event)) continue; - if (perf_event_overflow(event, 1, &data, regs)) + if (perf_event_overflow(event, &data, regs)) x86_pmu_stop(event, 0); } diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c index 5f9ecff328b5..98da6a7b5e82 100644 --- a/arch/x86/kernel/kgdb.c +++ b/arch/x86/kernel/kgdb.c @@ -608,7 +608,7 @@ int kgdb_arch_init(void) return register_die_notifier(&kgdb_notifier); } -static void kgdb_hw_overflow_handler(struct perf_event *event, int nmi, +static void kgdb_hw_overflow_handler(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { struct task_struct *tsk = current; diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index 807c2a2b80f1..11db2e9b860a 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -528,7 +528,7 @@ static int genregs_set(struct task_struct *target, return ret; } -static void ptrace_triggered(struct perf_event *bp, int nmi, +static void ptrace_triggered(struct perf_event *bp, struct perf_sample_data *data, struct pt_regs *regs) { diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c index 2dbf6bf4c7e5..4d09df054e39 100644 --- a/arch/x86/mm/fault.c +++ b/arch/x86/mm/fault.c @@ -1059,7 +1059,7 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code) if (unlikely(error_code & PF_RSVD)) pgtable_bad(regs, error_code, address); - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address); + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); /* * If we're in an interrupt, have no user context or are running @@ -1161,11 +1161,11 @@ good_area: if (flags & FAULT_FLAG_ALLOW_RETRY) { if (fault & VM_FAULT_MAJOR) { tsk->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address); } else { tsk->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0, + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address); } if (fault & VM_FAULT_RETRY) { diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 2f7b5d42ab41..0946a8bc098d 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -682,7 +682,7 @@ enum perf_event_active_state { struct file; struct perf_sample_data; -typedef void (*perf_overflow_handler_t)(struct perf_event *, int, +typedef void (*perf_overflow_handler_t)(struct perf_event *, struct perf_sample_data *, struct pt_regs *regs); @@ -925,7 +925,6 @@ struct perf_output_handle { unsigned long size; void *addr; int page; - int nmi; int sample; }; @@ -993,7 +992,7 @@ extern void perf_prepare_sample(struct perf_event_header *header, struct perf_event *event, struct pt_regs *regs); -extern int perf_event_overflow(struct perf_event *event, int nmi, +extern int perf_event_overflow(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs); @@ -1012,7 +1011,7 @@ static inline int is_software_event(struct perf_event *event) extern struct jump_label_key perf_swevent_enabled[PERF_COUNT_SW_MAX]; -extern void __perf_sw_event(u32, u64, int, struct pt_regs *, u64); +extern void __perf_sw_event(u32, u64, struct pt_regs *, u64); #ifndef perf_arch_fetch_caller_regs static inline void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip) { } @@ -1034,7 +1033,7 @@ static inline void perf_fetch_caller_regs(struct pt_regs *regs) } static __always_inline void -perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr) +perf_sw_event(u32 event_id, u64 nr, struct pt_regs *regs, u64 addr) { struct pt_regs hot_regs; @@ -1043,7 +1042,7 @@ perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr) perf_fetch_caller_regs(&hot_regs); regs = &hot_regs; } - __perf_sw_event(event_id, nr, nmi, regs, addr); + __perf_sw_event(event_id, nr, regs, addr); } } @@ -1057,7 +1056,7 @@ static inline void perf_event_task_sched_in(struct task_struct *task) static inline void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next) { - perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, NULL, 0); + perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, NULL, 0); __perf_event_task_sched_out(task, next); } @@ -1119,7 +1118,7 @@ extern void perf_bp_event(struct perf_event *event, void *data); extern int perf_output_begin(struct perf_output_handle *handle, struct perf_event *event, unsigned int size, - int nmi, int sample); + int sample); extern void perf_output_end(struct perf_output_handle *handle); extern void perf_output_copy(struct perf_output_handle *handle, const void *buf, unsigned int len); @@ -1143,8 +1142,7 @@ static inline int perf_event_task_disable(void) { return -EINVAL; } static inline int perf_event_task_enable(void) { return -EINVAL; } static inline void -perf_sw_event(u32 event_id, u64 nr, int nmi, - struct pt_regs *regs, u64 addr) { } +perf_sw_event(u32 event_id, u64 nr, struct pt_regs *regs, u64 addr) { } static inline void perf_bp_event(struct perf_event *event, void *data) { } diff --git a/kernel/events/core.c b/kernel/events/core.c index 270e32f9fc06..dbd1ca75bd3c 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3972,7 +3972,7 @@ void perf_prepare_sample(struct perf_event_header *header, } } -static void perf_event_output(struct perf_event *event, int nmi, +static void perf_event_output(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { @@ -3984,7 +3984,7 @@ static void perf_event_output(struct perf_event *event, int nmi, perf_prepare_sample(&header, data, event, regs); - if (perf_output_begin(&handle, event, header.size, nmi, 1)) + if (perf_output_begin(&handle, event, header.size, 1)) goto exit; perf_output_sample(&handle, &header, data, event); @@ -4024,7 +4024,7 @@ perf_event_read_event(struct perf_event *event, int ret; perf_event_header__init_id(&read_event.header, &sample, event); - ret = perf_output_begin(&handle, event, read_event.header.size, 0, 0); + ret = perf_output_begin(&handle, event, read_event.header.size, 0); if (ret) return; @@ -4067,7 +4067,7 @@ static void perf_event_task_output(struct perf_event *event, perf_event_header__init_id(&task_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, - task_event->event_id.header.size, 0, 0); + task_event->event_id.header.size, 0); if (ret) goto out; @@ -4204,7 +4204,7 @@ static void perf_event_comm_output(struct perf_event *event, perf_event_header__init_id(&comm_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, - comm_event->event_id.header.size, 0, 0); + comm_event->event_id.header.size, 0); if (ret) goto out; @@ -4351,7 +4351,7 @@ static void perf_event_mmap_output(struct perf_event *event, perf_event_header__init_id(&mmap_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, - mmap_event->event_id.header.size, 0, 0); + mmap_event->event_id.header.size, 0); if (ret) goto out; @@ -4546,7 +4546,7 @@ static void perf_log_throttle(struct perf_event *event, int enable) perf_event_header__init_id(&throttle_event.header, &sample, event); ret = perf_output_begin(&handle, event, - throttle_event.header.size, 1, 0); + throttle_event.header.size, 0); if (ret) return; @@ -4559,7 +4559,7 @@ static void perf_log_throttle(struct perf_event *event, int enable) * Generic event overflow handling, sampling. */ -static int __perf_event_overflow(struct perf_event *event, int nmi, +static int __perf_event_overflow(struct perf_event *event, int throttle, struct perf_sample_data *data, struct pt_regs *regs) { @@ -4602,34 +4602,28 @@ static int __perf_event_overflow(struct perf_event *event, int nmi, if (events && atomic_dec_and_test(&event->event_limit)) { ret = 1; event->pending_kill = POLL_HUP; - if (nmi) { - event->pending_disable = 1; - irq_work_queue(&event->pending); - } else - perf_event_disable(event); + event->pending_disable = 1; + irq_work_queue(&event->pending); } if (event->overflow_handler) - event->overflow_handler(event, nmi, data, regs); + event->overflow_handler(event, data, regs); else - perf_event_output(event, nmi, data, regs); + perf_event_output(event, data, regs); if (event->fasync && event->pending_kill) { - if (nmi) { - event->pending_wakeup = 1; - irq_work_queue(&event->pending); - } else - perf_event_wakeup(event); + event->pending_wakeup = 1; + irq_work_queue(&event->pending); } return ret; } -int perf_event_overflow(struct perf_event *event, int nmi, +int perf_event_overflow(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { - return __perf_event_overflow(event, nmi, 1, data, regs); + return __perf_event_overflow(event, 1, data, regs); } /* @@ -4678,7 +4672,7 @@ again: } static void perf_swevent_overflow(struct perf_event *event, u64 overflow, - int nmi, struct perf_sample_data *data, + struct perf_sample_data *data, struct pt_regs *regs) { struct hw_perf_event *hwc = &event->hw; @@ -4692,7 +4686,7 @@ static void perf_swevent_overflow(struct perf_event *event, u64 overflow, return; for (; overflow; overflow--) { - if (__perf_event_overflow(event, nmi, throttle, + if (__perf_event_overflow(event, throttle, data, regs)) { /* * We inhibit the overflow from happening when @@ -4705,7 +4699,7 @@ static void perf_swevent_overflow(struct perf_event *event, u64 overflow, } static void perf_swevent_event(struct perf_event *event, u64 nr, - int nmi, struct perf_sample_data *data, + struct perf_sample_data *data, struct pt_regs *regs) { struct hw_perf_event *hwc = &event->hw; @@ -4719,12 +4713,12 @@ static void perf_swevent_event(struct perf_event *event, u64 nr, return; if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq) - return perf_swevent_overflow(event, 1, nmi, data, regs); + return perf_swevent_overflow(event, 1, data, regs); if (local64_add_negative(nr, &hwc->period_left)) return; - perf_swevent_overflow(event, 0, nmi, data, regs); + perf_swevent_overflow(event, 0, data, regs); } static int perf_exclude_event(struct perf_event *event, @@ -4812,7 +4806,7 @@ find_swevent_head(struct swevent_htable *swhash, struct perf_event *event) } static void do_perf_sw_event(enum perf_type_id type, u32 event_id, - u64 nr, int nmi, + u64 nr, struct perf_sample_data *data, struct pt_regs *regs) { @@ -4828,7 +4822,7 @@ static void do_perf_sw_event(enum perf_type_id type, u32 event_id, hlist_for_each_entry_rcu(event, node, head, hlist_entry) { if (perf_swevent_match(event, type, event_id, data, regs)) - perf_swevent_event(event, nr, nmi, data, regs); + perf_swevent_event(event, nr, data, regs); } end: rcu_read_unlock(); @@ -4849,8 +4843,7 @@ inline void perf_swevent_put_recursion_context(int rctx) put_recursion_context(swhash->recursion, rctx); } -void __perf_sw_event(u32 event_id, u64 nr, int nmi, - struct pt_regs *regs, u64 addr) +void __perf_sw_event(u32 event_id, u64 nr, struct pt_regs *regs, u64 addr) { struct perf_sample_data data; int rctx; @@ -4862,7 +4855,7 @@ void __perf_sw_event(u32 event_id, u64 nr, int nmi, perf_sample_data_init(&data, addr); - do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, nmi, &data, regs); + do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, &data, regs); perf_swevent_put_recursion_context(rctx); preempt_enable_notrace(); @@ -5110,7 +5103,7 @@ void perf_tp_event(u64 addr, u64 count, void *record, int entry_size, hlist_for_each_entry_rcu(event, node, head, hlist_entry) { if (perf_tp_event_match(event, &data, regs)) - perf_swevent_event(event, count, 1, &data, regs); + perf_swevent_event(event, count, &data, regs); } perf_swevent_put_recursion_context(rctx); @@ -5203,7 +5196,7 @@ void perf_bp_event(struct perf_event *bp, void *data) perf_sample_data_init(&sample, bp->attr.bp_addr); if (!bp->hw.state && !perf_exclude_event(bp, regs)) - perf_swevent_event(bp, 1, 1, &sample, regs); + perf_swevent_event(bp, 1, &sample, regs); } #endif @@ -5232,7 +5225,7 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer) if (regs && !perf_exclude_event(event, regs)) { if (!(event->attr.exclude_idle && current->pid == 0)) - if (perf_event_overflow(event, 0, &data, regs)) + if (perf_event_overflow(event, &data, regs)) ret = HRTIMER_NORESTART; } diff --git a/kernel/events/internal.h b/kernel/events/internal.h index 114f27f3a624..09097dd8116c 100644 --- a/kernel/events/internal.h +++ b/kernel/events/internal.h @@ -27,7 +27,6 @@ struct ring_buffer { void *data_pages[0]; }; - extern void rb_free(struct ring_buffer *rb); extern struct ring_buffer * rb_alloc(int nr_pages, long watermark, int cpu, int flags); diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c index fc2701c99207..8b3b73630fa4 100644 --- a/kernel/events/ring_buffer.c +++ b/kernel/events/ring_buffer.c @@ -38,11 +38,8 @@ static void perf_output_wakeup(struct perf_output_handle *handle) { atomic_set(&handle->rb->poll, POLL_IN); - if (handle->nmi) { - handle->event->pending_wakeup = 1; - irq_work_queue(&handle->event->pending); - } else - perf_event_wakeup(handle->event); + handle->event->pending_wakeup = 1; + irq_work_queue(&handle->event->pending); } /* @@ -102,7 +99,7 @@ out: int perf_output_begin(struct perf_output_handle *handle, struct perf_event *event, unsigned int size, - int nmi, int sample) + int sample) { struct ring_buffer *rb; unsigned long tail, offset, head; @@ -127,7 +124,6 @@ int perf_output_begin(struct perf_output_handle *handle, handle->rb = rb; handle->event = event; - handle->nmi = nmi; handle->sample = sample; if (!rb->nr_pages) diff --git a/kernel/sched.c b/kernel/sched.c index 3f2e502d609b..d08d110b8976 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2220,7 +2220,7 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu) if (task_cpu(p) != new_cpu) { p->se.nr_migrations++; - perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, 1, NULL, 0); + perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, NULL, 0); } __set_task_cpu(p, new_cpu); diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 752b75ba662b..a6708e677a0a 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -211,7 +211,7 @@ static struct perf_event_attr wd_hw_attr = { }; /* Callback function for perf event subsystem */ -static void watchdog_overflow_callback(struct perf_event *event, int nmi, +static void watchdog_overflow_callback(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { diff --git a/samples/hw_breakpoint/data_breakpoint.c b/samples/hw_breakpoint/data_breakpoint.c index 063653955f9f..7b164d3200ff 100644 --- a/samples/hw_breakpoint/data_breakpoint.c +++ b/samples/hw_breakpoint/data_breakpoint.c @@ -41,7 +41,7 @@ module_param_string(ksym, ksym_name, KSYM_NAME_LEN, S_IRUGO); MODULE_PARM_DESC(ksym, "Kernel symbol to monitor; this module will report any" " write operations on the kernel symbol"); -static void sample_hbp_handler(struct perf_event *bp, int nmi, +static void sample_hbp_handler(struct perf_event *bp, struct perf_sample_data *data, struct pt_regs *regs) { -- cgit v1.2.3 From a7ac67ea021b4603095d2aa458bc41641238f22c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 27 Jun 2011 16:47:16 +0200 Subject: perf: Remove the perf_output_begin(.sample) argument Since only samples call perf_output_sample() its much saner (and more correct) to put the sample logic in there than in the perf_output_begin()/perf_output_end() pair. Saves a useless argument, reduces conditionals and shrinks struct perf_output_handle, win! Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-2crpvsx3cqu67q3zqjbnlpsc@git.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel_ds.c | 2 +- include/linux/perf_event.h | 4 +--- kernel/events/core.c | 26 ++++++++++++++++++++------ kernel/events/ring_buffer.c | 19 +------------------ 4 files changed, 23 insertions(+), 28 deletions(-) (limited to 'kernel') diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c index 0941f93f2940..1b1ef3addcfd 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_ds.c +++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c @@ -340,7 +340,7 @@ static int intel_pmu_drain_bts_buffer(void) */ perf_prepare_sample(&header, &data, event, ®s); - if (perf_output_begin(&handle, event, header.size * (top - at), 1)) + if (perf_output_begin(&handle, event, header.size * (top - at))) return 1; for (; at < top; at++) { diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 0946a8bc098d..771b0b2845e4 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -925,7 +925,6 @@ struct perf_output_handle { unsigned long size; void *addr; int page; - int sample; }; #ifdef CONFIG_PERF_EVENTS @@ -1117,8 +1116,7 @@ extern void perf_bp_event(struct perf_event *event, void *data); #endif extern int perf_output_begin(struct perf_output_handle *handle, - struct perf_event *event, unsigned int size, - int sample); + struct perf_event *event, unsigned int size); extern void perf_output_end(struct perf_output_handle *handle); extern void perf_output_copy(struct perf_output_handle *handle, const void *buf, unsigned int len); diff --git a/kernel/events/core.c b/kernel/events/core.c index dbd1ca75bd3c..81de28dcca8c 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3928,6 +3928,20 @@ void perf_output_sample(struct perf_output_handle *handle, perf_output_put(handle, raw); } } + + if (!event->attr.watermark) { + int wakeup_events = event->attr.wakeup_events; + + if (wakeup_events) { + struct ring_buffer *rb = handle->rb; + int events = local_inc_return(&rb->events); + + if (events >= wakeup_events) { + local_sub(wakeup_events, &rb->events); + local_inc(&rb->wakeup); + } + } + } } void perf_prepare_sample(struct perf_event_header *header, @@ -3984,7 +3998,7 @@ static void perf_event_output(struct perf_event *event, perf_prepare_sample(&header, data, event, regs); - if (perf_output_begin(&handle, event, header.size, 1)) + if (perf_output_begin(&handle, event, header.size)) goto exit; perf_output_sample(&handle, &header, data, event); @@ -4024,7 +4038,7 @@ perf_event_read_event(struct perf_event *event, int ret; perf_event_header__init_id(&read_event.header, &sample, event); - ret = perf_output_begin(&handle, event, read_event.header.size, 0); + ret = perf_output_begin(&handle, event, read_event.header.size); if (ret) return; @@ -4067,7 +4081,7 @@ static void perf_event_task_output(struct perf_event *event, perf_event_header__init_id(&task_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, - task_event->event_id.header.size, 0); + task_event->event_id.header.size); if (ret) goto out; @@ -4204,7 +4218,7 @@ static void perf_event_comm_output(struct perf_event *event, perf_event_header__init_id(&comm_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, - comm_event->event_id.header.size, 0); + comm_event->event_id.header.size); if (ret) goto out; @@ -4351,7 +4365,7 @@ static void perf_event_mmap_output(struct perf_event *event, perf_event_header__init_id(&mmap_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, - mmap_event->event_id.header.size, 0); + mmap_event->event_id.header.size); if (ret) goto out; @@ -4546,7 +4560,7 @@ static void perf_log_throttle(struct perf_event *event, int enable) perf_event_header__init_id(&throttle_event.header, &sample, event); ret = perf_output_begin(&handle, event, - throttle_event.header.size, 0); + throttle_event.header.size); if (ret) return; diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c index 8b3b73630fa4..a2a29205cc0f 100644 --- a/kernel/events/ring_buffer.c +++ b/kernel/events/ring_buffer.c @@ -98,8 +98,7 @@ out: } int perf_output_begin(struct perf_output_handle *handle, - struct perf_event *event, unsigned int size, - int sample) + struct perf_event *event, unsigned int size) { struct ring_buffer *rb; unsigned long tail, offset, head; @@ -124,7 +123,6 @@ int perf_output_begin(struct perf_output_handle *handle, handle->rb = rb; handle->event = event; - handle->sample = sample; if (!rb->nr_pages) goto out; @@ -192,21 +190,6 @@ void perf_output_copy(struct perf_output_handle *handle, void perf_output_end(struct perf_output_handle *handle) { - struct perf_event *event = handle->event; - struct ring_buffer *rb = handle->rb; - - if (handle->sample && !event->attr.watermark) { - int wakeup_events = event->attr.wakeup_events; - - if (wakeup_events) { - int events = local_inc_return(&rb->events); - if (events >= wakeup_events) { - local_sub(wakeup_events, &rb->events); - local_inc(&rb->wakeup); - } - } - } - perf_output_put_handle(handle); rcu_read_unlock(); } -- cgit v1.2.3 From 4dc0da86967d5463708631d02a70cfed5b104884 Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Wed, 29 Jun 2011 18:42:35 +0300 Subject: perf: Add context field to perf_event The perf_event overflow handler does not receive any caller-derived argument, so many callers need to resort to looking up the perf_event in their local data structure. This is ugly and doesn't scale if a single callback services many perf_events. Fix by adding a context parameter to perf_event_create_kernel_counter() (and derived hardware breakpoints APIs) and storing it in the perf_event. The field can be accessed from the callback as event->overflow_handler_context. All callers are updated. Signed-off-by: Avi Kivity Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1309362157-6596-2-git-send-email-avi@redhat.com Signed-off-by: Ingo Molnar --- arch/arm/kernel/ptrace.c | 3 ++- arch/powerpc/kernel/ptrace.c | 2 +- arch/sh/kernel/ptrace_32.c | 3 ++- arch/x86/kernel/kgdb.c | 2 +- arch/x86/kernel/ptrace.c | 3 ++- drivers/oprofile/oprofile_perf.c | 2 +- include/linux/hw_breakpoint.h | 10 ++++++++-- include/linux/perf_event.h | 4 +++- kernel/events/core.c | 21 +++++++++++++++------ kernel/events/hw_breakpoint.c | 10 +++++++--- kernel/watchdog.c | 2 +- samples/hw_breakpoint/data_breakpoint.c | 2 +- 12 files changed, 44 insertions(+), 20 deletions(-) (limited to 'kernel') diff --git a/arch/arm/kernel/ptrace.c b/arch/arm/kernel/ptrace.c index 0c9b1054f790..5c199610719f 100644 --- a/arch/arm/kernel/ptrace.c +++ b/arch/arm/kernel/ptrace.c @@ -479,7 +479,8 @@ static struct perf_event *ptrace_hbp_create(struct task_struct *tsk, int type) attr.bp_type = type; attr.disabled = 1; - return register_user_hw_breakpoint(&attr, ptrace_hbptriggered, tsk); + return register_user_hw_breakpoint(&attr, ptrace_hbptriggered, NULL, + tsk); } static int ptrace_gethbpregs(struct task_struct *tsk, long num, diff --git a/arch/powerpc/kernel/ptrace.c b/arch/powerpc/kernel/ptrace.c index 3177617af2ef..05b7dd217f60 100644 --- a/arch/powerpc/kernel/ptrace.c +++ b/arch/powerpc/kernel/ptrace.c @@ -973,7 +973,7 @@ int ptrace_set_debugreg(struct task_struct *task, unsigned long addr, &attr.bp_type); thread->ptrace_bps[0] = bp = register_user_hw_breakpoint(&attr, - ptrace_triggered, task); + ptrace_triggered, NULL, task); if (IS_ERR(bp)) { thread->ptrace_bps[0] = NULL; ptrace_put_breakpoints(task); diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c index 8051976100a6..92b3c276339a 100644 --- a/arch/sh/kernel/ptrace_32.c +++ b/arch/sh/kernel/ptrace_32.c @@ -91,7 +91,8 @@ static int set_single_step(struct task_struct *tsk, unsigned long addr) attr.bp_len = HW_BREAKPOINT_LEN_2; attr.bp_type = HW_BREAKPOINT_R; - bp = register_user_hw_breakpoint(&attr, ptrace_triggered, tsk); + bp = register_user_hw_breakpoint(&attr, ptrace_triggered, + NULL, tsk); if (IS_ERR(bp)) return PTR_ERR(bp); diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c index 98da6a7b5e82..00354d4919a9 100644 --- a/arch/x86/kernel/kgdb.c +++ b/arch/x86/kernel/kgdb.c @@ -638,7 +638,7 @@ void kgdb_arch_late(void) for (i = 0; i < HBP_NUM; i++) { if (breakinfo[i].pev) continue; - breakinfo[i].pev = register_wide_hw_breakpoint(&attr, NULL); + breakinfo[i].pev = register_wide_hw_breakpoint(&attr, NULL, NULL); if (IS_ERR((void * __force)breakinfo[i].pev)) { printk(KERN_ERR "kgdb: Could not allocate hw" "breakpoints\nDisabling the kernel debugger\n"); diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index 11db2e9b860a..82528799c5de 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -715,7 +715,8 @@ static int ptrace_set_breakpoint_addr(struct task_struct *tsk, int nr, attr.bp_type = HW_BREAKPOINT_W; attr.disabled = 1; - bp = register_user_hw_breakpoint(&attr, ptrace_triggered, tsk); + bp = register_user_hw_breakpoint(&attr, ptrace_triggered, + NULL, tsk); /* * CHECKME: the previous code returned -EIO if the addr wasn't diff --git a/drivers/oprofile/oprofile_perf.c b/drivers/oprofile/oprofile_perf.c index 9046f7b2ed79..59acf9ef78a4 100644 --- a/drivers/oprofile/oprofile_perf.c +++ b/drivers/oprofile/oprofile_perf.c @@ -79,7 +79,7 @@ static int op_create_counter(int cpu, int event) pevent = perf_event_create_kernel_counter(&counter_config[event].attr, cpu, NULL, - op_overflow_handler); + op_overflow_handler, NULL); if (IS_ERR(pevent)) return PTR_ERR(pevent); diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h index d1e55fed2c7d..6ae9c631a1be 100644 --- a/include/linux/hw_breakpoint.h +++ b/include/linux/hw_breakpoint.h @@ -73,6 +73,7 @@ static inline unsigned long hw_breakpoint_len(struct perf_event *bp) extern struct perf_event * register_user_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered, + void *context, struct task_struct *tsk); /* FIXME: only change from the attr, and don't unregister */ @@ -85,11 +86,13 @@ modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr); extern struct perf_event * register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr, perf_overflow_handler_t triggered, + void *context, int cpu); extern struct perf_event * __percpu * register_wide_hw_breakpoint(struct perf_event_attr *attr, - perf_overflow_handler_t triggered); + perf_overflow_handler_t triggered, + void *context); extern int register_perf_hw_breakpoint(struct perf_event *bp); extern int __register_perf_hw_breakpoint(struct perf_event *bp); @@ -115,6 +118,7 @@ static inline int __init init_hw_breakpoint(void) { return 0; } static inline struct perf_event * register_user_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered, + void *context, struct task_struct *tsk) { return NULL; } static inline int modify_user_hw_breakpoint(struct perf_event *bp, @@ -122,10 +126,12 @@ modify_user_hw_breakpoint(struct perf_event *bp, static inline struct perf_event * register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr, perf_overflow_handler_t triggered, + void *context, int cpu) { return NULL; } static inline struct perf_event * __percpu * register_wide_hw_breakpoint(struct perf_event_attr *attr, - perf_overflow_handler_t triggered) { return NULL; } + perf_overflow_handler_t triggered, + void *context) { return NULL; } static inline int register_perf_hw_breakpoint(struct perf_event *bp) { return -ENOSYS; } static inline int diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index a5f54b973bdb..2a08cacb1628 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -839,6 +839,7 @@ struct perf_event { u64 id; perf_overflow_handler_t overflow_handler; + void *overflow_handler_context; #ifdef CONFIG_EVENT_TRACING struct ftrace_event_call *tp_event; @@ -960,7 +961,8 @@ extern struct perf_event * perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, struct task_struct *task, - perf_overflow_handler_t callback); + perf_overflow_handler_t callback, + void *context); extern u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running); diff --git a/kernel/events/core.c b/kernel/events/core.c index 81de28dcca8c..ba8e0f4a20e6 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -5745,7 +5745,8 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, struct task_struct *task, struct perf_event *group_leader, struct perf_event *parent_event, - perf_overflow_handler_t overflow_handler) + perf_overflow_handler_t overflow_handler, + void *context) { struct pmu *pmu; struct perf_event *event; @@ -5803,10 +5804,13 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, #endif } - if (!overflow_handler && parent_event) + if (!overflow_handler && parent_event) { overflow_handler = parent_event->overflow_handler; + context = parent_event->overflow_handler_context; + } event->overflow_handler = overflow_handler; + event->overflow_handler_context = context; if (attr->disabled) event->state = PERF_EVENT_STATE_OFF; @@ -6073,7 +6077,8 @@ SYSCALL_DEFINE5(perf_event_open, } } - event = perf_event_alloc(&attr, cpu, task, group_leader, NULL, NULL); + event = perf_event_alloc(&attr, cpu, task, group_leader, NULL, + NULL, NULL); if (IS_ERR(event)) { err = PTR_ERR(event); goto err_task; @@ -6258,7 +6263,8 @@ err_fd: struct perf_event * perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, struct task_struct *task, - perf_overflow_handler_t overflow_handler) + perf_overflow_handler_t overflow_handler, + void *context) { struct perf_event_context *ctx; struct perf_event *event; @@ -6268,7 +6274,8 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, * Get the target context (task or percpu): */ - event = perf_event_alloc(attr, cpu, task, NULL, NULL, overflow_handler); + event = perf_event_alloc(attr, cpu, task, NULL, NULL, + overflow_handler, context); if (IS_ERR(event)) { err = PTR_ERR(event); goto err; @@ -6552,7 +6559,7 @@ inherit_event(struct perf_event *parent_event, parent_event->cpu, child, group_leader, parent_event, - NULL); + NULL, NULL); if (IS_ERR(child_event)) return child_event; get_ctx(child_ctx); @@ -6579,6 +6586,8 @@ inherit_event(struct perf_event *parent_event, child_event->ctx = child_ctx; child_event->overflow_handler = parent_event->overflow_handler; + child_event->overflow_handler_context + = parent_event->overflow_handler_context; /* * Precalculate sample_data sizes diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c index 086adf25a55e..b7971d6f38bf 100644 --- a/kernel/events/hw_breakpoint.c +++ b/kernel/events/hw_breakpoint.c @@ -431,9 +431,11 @@ int register_perf_hw_breakpoint(struct perf_event *bp) struct perf_event * register_user_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered, + void *context, struct task_struct *tsk) { - return perf_event_create_kernel_counter(attr, -1, tsk, triggered); + return perf_event_create_kernel_counter(attr, -1, tsk, triggered, + context); } EXPORT_SYMBOL_GPL(register_user_hw_breakpoint); @@ -502,7 +504,8 @@ EXPORT_SYMBOL_GPL(unregister_hw_breakpoint); */ struct perf_event * __percpu * register_wide_hw_breakpoint(struct perf_event_attr *attr, - perf_overflow_handler_t triggered) + perf_overflow_handler_t triggered, + void *context) { struct perf_event * __percpu *cpu_events, **pevent, *bp; long err; @@ -515,7 +518,8 @@ register_wide_hw_breakpoint(struct perf_event_attr *attr, get_online_cpus(); for_each_online_cpu(cpu) { pevent = per_cpu_ptr(cpu_events, cpu); - bp = perf_event_create_kernel_counter(attr, cpu, NULL, triggered); + bp = perf_event_create_kernel_counter(attr, cpu, NULL, + triggered, context); *pevent = bp; diff --git a/kernel/watchdog.c b/kernel/watchdog.c index a6708e677a0a..a933e3a0398b 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -375,7 +375,7 @@ static int watchdog_nmi_enable(int cpu) hw_nmi_watchdog_set_attr(wd_attr); /* Try to register using hardware perf events */ - event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback); + event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL); if (!IS_ERR(event)) { printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n"); goto out_save; diff --git a/samples/hw_breakpoint/data_breakpoint.c b/samples/hw_breakpoint/data_breakpoint.c index 7b164d3200ff..ef7f32291852 100644 --- a/samples/hw_breakpoint/data_breakpoint.c +++ b/samples/hw_breakpoint/data_breakpoint.c @@ -60,7 +60,7 @@ static int __init hw_break_module_init(void) attr.bp_len = HW_BREAKPOINT_LEN_4; attr.bp_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R; - sample_hbp = register_wide_hw_breakpoint(&attr, sample_hbp_handler); + sample_hbp = register_wide_hw_breakpoint(&attr, sample_hbp_handler, NULL); if (IS_ERR((void __force *)sample_hbp)) { ret = PTR_ERR((void __force *)sample_hbp); goto fail; -- cgit v1.2.3 From 26ca5c11fb45ae2b2ac7e3574b8db6b3a3c7d350 Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Wed, 29 Jun 2011 18:42:37 +0300 Subject: perf: export perf_event_refresh() to modules KVM needs one-shot samples, since a PMC programmed to -X will fire after X events and then again after 2^40 events (i.e. variable period). Signed-off-by: Avi Kivity Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1309362157-6596-4-git-send-email-avi@redhat.com Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 5 +++++ kernel/events/core.c | 3 ++- 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 2a08cacb1628..3f2711ccf910 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -955,6 +955,7 @@ extern void perf_pmu_disable(struct pmu *pmu); extern void perf_pmu_enable(struct pmu *pmu); extern int perf_event_task_disable(void); extern int perf_event_task_enable(void); +extern int perf_event_refresh(struct perf_event *event, int refresh); extern void perf_event_update_userpage(struct perf_event *event); extern int perf_event_release_kernel(struct perf_event *event); extern struct perf_event * @@ -1149,6 +1150,10 @@ static inline void perf_event_delayed_put(struct task_struct *task) { } static inline void perf_event_print_debug(void) { } static inline int perf_event_task_disable(void) { return -EINVAL; } static inline int perf_event_task_enable(void) { return -EINVAL; } +static inline int perf_event_refresh(struct perf_event *event, int refresh) +{ + return -EINVAL; +} static inline void perf_sw_event(u32 event_id, u64 nr, struct pt_regs *regs, u64 addr) { } diff --git a/kernel/events/core.c b/kernel/events/core.c index ba8e0f4a20e6..0567e32d71aa 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1764,7 +1764,7 @@ out: raw_spin_unlock_irq(&ctx->lock); } -static int perf_event_refresh(struct perf_event *event, int refresh) +int perf_event_refresh(struct perf_event *event, int refresh) { /* * not supported on inherited events @@ -1777,6 +1777,7 @@ static int perf_event_refresh(struct perf_event *event, int refresh) return 0; } +EXPORT_SYMBOL_GPL(perf_event_refresh); static void ctx_sched_out(struct perf_event_context *ctx, struct perf_cpu_context *cpuctx, -- cgit v1.2.3 From e9dbfae53eeb9fc3d4bb7da3df87fa9875f5da02 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 5 Jul 2011 11:36:06 -0400 Subject: tracing: Fix bug when reading system filters on module removal The event system is freed when its nr_events is set to zero. This happens when a module created an event system and then later the module is removed. Modules may share systems, so the system is allocated when it is created and freed when the modules are unloaded and all the events under the system are removed (nr_events set to zero). The problem arises when a task opened the "filter" file for the system. If the module is unloaded and it removed the last event for that system, the system structure is freed. If the task that opened the filter file accesses the "filter" file after the system has been freed, the system will access an invalid pointer. By adding a ref_count, and using it to keep track of what is using the event system, we can free it after all users are finished with the event system. Cc: Reported-by: Johannes Berg Signed-off-by: Steven Rostedt --- kernel/trace/trace.h | 1 + kernel/trace/trace_events.c | 86 +++++++++++++++++++++++++++++++++----- kernel/trace/trace_events_filter.c | 6 +++ 3 files changed, 82 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 229f8591f61d..f8074072d111 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -677,6 +677,7 @@ struct event_subsystem { struct dentry *entry; struct event_filter *filter; int nr_events; + int ref_count; }; #define FILTER_PRED_INVALID ((unsigned short)-1) diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 686ec399f2a8..ffc5b2884af1 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -244,6 +244,35 @@ static void ftrace_clear_events(void) mutex_unlock(&event_mutex); } +static void __put_system(struct event_subsystem *system) +{ + struct event_filter *filter = system->filter; + + WARN_ON_ONCE(system->ref_count == 0); + if (--system->ref_count) + return; + + if (filter) { + kfree(filter->filter_string); + kfree(filter); + } + kfree(system->name); + kfree(system); +} + +static void __get_system(struct event_subsystem *system) +{ + WARN_ON_ONCE(system->ref_count == 0); + system->ref_count++; +} + +static void put_system(struct event_subsystem *system) +{ + mutex_lock(&event_mutex); + __put_system(system); + mutex_unlock(&event_mutex); +} + /* * __ftrace_set_clr_event(NULL, NULL, NULL, set) will set/unset all events. */ @@ -826,6 +855,47 @@ event_filter_write(struct file *filp, const char __user *ubuf, size_t cnt, return cnt; } +static LIST_HEAD(event_subsystems); + +static int subsystem_open(struct inode *inode, struct file *filp) +{ + struct event_subsystem *system = NULL; + int ret; + + /* Make sure the system still exists */ + mutex_lock(&event_mutex); + list_for_each_entry(system, &event_subsystems, list) { + if (system == inode->i_private) { + /* Don't open systems with no events */ + if (!system->nr_events) { + system = NULL; + break; + } + __get_system(system); + break; + } + } + mutex_unlock(&event_mutex); + + if (system != inode->i_private) + return -ENODEV; + + ret = tracing_open_generic(inode, filp); + if (ret < 0) + put_system(system); + + return ret; +} + +static int subsystem_release(struct inode *inode, struct file *file) +{ + struct event_subsystem *system = inode->i_private; + + put_system(system); + + return 0; +} + static ssize_t subsystem_filter_read(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos) @@ -963,10 +1033,11 @@ static const struct file_operations ftrace_event_filter_fops = { }; static const struct file_operations ftrace_subsystem_filter_fops = { - .open = tracing_open_generic, + .open = subsystem_open, .read = subsystem_filter_read, .write = subsystem_filter_write, .llseek = default_llseek, + .release = subsystem_release, }; static const struct file_operations ftrace_system_enable_fops = { @@ -1002,8 +1073,6 @@ static struct dentry *event_trace_events_dir(void) return d_events; } -static LIST_HEAD(event_subsystems); - static struct dentry * event_subsystem_dir(const char *name, struct dentry *d_events) { @@ -1013,6 +1082,7 @@ event_subsystem_dir(const char *name, struct dentry *d_events) /* First see if we did not already create this dir */ list_for_each_entry(system, &event_subsystems, list) { if (strcmp(system->name, name) == 0) { + __get_system(system); system->nr_events++; return system->entry; } @@ -1035,6 +1105,7 @@ event_subsystem_dir(const char *name, struct dentry *d_events) } system->nr_events = 1; + system->ref_count = 1; system->name = kstrdup(name, GFP_KERNEL); if (!system->name) { debugfs_remove(system->entry); @@ -1184,16 +1255,9 @@ static void remove_subsystem_dir(const char *name) list_for_each_entry(system, &event_subsystems, list) { if (strcmp(system->name, name) == 0) { if (!--system->nr_events) { - struct event_filter *filter = system->filter; - debugfs_remove_recursive(system->entry); list_del(&system->list); - if (filter) { - kfree(filter->filter_string); - kfree(filter); - } - kfree(system->name); - kfree(system); + __put_system(system); } break; } diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c index 8008ddcfbf20..256764ecccd6 100644 --- a/kernel/trace/trace_events_filter.c +++ b/kernel/trace/trace_events_filter.c @@ -1886,6 +1886,12 @@ int apply_subsystem_event_filter(struct event_subsystem *system, mutex_lock(&event_mutex); + /* Make sure the system still has events */ + if (!system->nr_events) { + err = -ENODEV; + goto out_unlock; + } + if (!strcmp(strstrip(filter_string), "0")) { filter_free_subsystem_preds(system); remove_filter_string(system->filter); -- cgit v1.2.3 From 40ee4dffff061399eb9358e0c8fcfbaf8de4c8fe Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 5 Jul 2011 14:32:51 -0400 Subject: tracing: Have "enable" file use refcounts like the "filter" file The "enable" file for the event system can be removed when a module is unloaded and the event system only has events from that module. As the event system nr_events count goes to zero, it may be freed if its ref_count is also set to zero. Like the "filter" file, the "enable" file may be opened by a task and referenced later, after a module has been unloaded and the events for that event system have been removed. Although the "filter" file referenced the event system structure, the "enable" file only references a pointer to the event system name. Since the name is freed when the event system is removed, it is possible that an access to the "enable" file may reference a freed pointer. Update the "enable" file to use the subsystem_open() routine that the "filter" file uses, to keep a reference to the event system structure while the "enable" file is opened. Cc: Reported-by: Johannes Berg Signed-off-by: Steven Rostedt --- kernel/trace/trace_events.c | 31 ++++++++++++++++++++++--------- 1 file changed, 22 insertions(+), 9 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index ffc5b2884af1..3e2a7c91c548 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -557,7 +557,7 @@ system_enable_read(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos) { const char set_to_char[4] = { '?', '0', '1', 'X' }; - const char *system = filp->private_data; + struct event_subsystem *system = filp->private_data; struct ftrace_event_call *call; char buf[2]; int set = 0; @@ -568,7 +568,7 @@ system_enable_read(struct file *filp, char __user *ubuf, size_t cnt, if (!call->name || !call->class || !call->class->reg) continue; - if (system && strcmp(call->class->system, system) != 0) + if (system && strcmp(call->class->system, system->name) != 0) continue; /* @@ -598,7 +598,8 @@ static ssize_t system_enable_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos) { - const char *system = filp->private_data; + struct event_subsystem *system = filp->private_data; + const char *name = NULL; unsigned long val; char buf[64]; ssize_t ret; @@ -622,7 +623,14 @@ system_enable_write(struct file *filp, const char __user *ubuf, size_t cnt, if (val != 0 && val != 1) return -EINVAL; - ret = __ftrace_set_clr_event(NULL, system, NULL, val); + /* + * Opening of "enable" adds a ref count to system, + * so the name is safe to use. + */ + if (system) + name = system->name; + + ret = __ftrace_set_clr_event(NULL, name, NULL, val); if (ret) goto out; @@ -862,6 +870,9 @@ static int subsystem_open(struct inode *inode, struct file *filp) struct event_subsystem *system = NULL; int ret; + if (!inode->i_private) + goto skip_search; + /* Make sure the system still exists */ mutex_lock(&event_mutex); list_for_each_entry(system, &event_subsystems, list) { @@ -880,8 +891,9 @@ static int subsystem_open(struct inode *inode, struct file *filp) if (system != inode->i_private) return -ENODEV; + skip_search: ret = tracing_open_generic(inode, filp); - if (ret < 0) + if (ret < 0 && system) put_system(system); return ret; @@ -891,7 +903,8 @@ static int subsystem_release(struct inode *inode, struct file *file) { struct event_subsystem *system = inode->i_private; - put_system(system); + if (system) + put_system(system); return 0; } @@ -1041,10 +1054,11 @@ static const struct file_operations ftrace_subsystem_filter_fops = { }; static const struct file_operations ftrace_system_enable_fops = { - .open = tracing_open_generic, + .open = subsystem_open, .read = system_enable_read, .write = system_enable_write, .llseek = default_llseek, + .release = subsystem_release, }; static const struct file_operations ftrace_show_header_fops = { @@ -1133,8 +1147,7 @@ event_subsystem_dir(const char *name, struct dentry *d_events) "'%s/filter' entry\n", name); } - trace_create_file("enable", 0644, system->entry, - (void *)system->name, + trace_create_file("enable", 0644, system->entry, system, &ftrace_system_enable_fops); return system->entry; -- cgit v1.2.3 From 43dd61c9a09bd413e837df829e6bfb42159be52a Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 7 Jul 2011 11:09:22 -0400 Subject: ftrace: Fix regression of :mod:module function enabling The new code that allows different utilities to pick and choose what functions they trace broke the :mod: hook that allows users to trace only functions of a particular module. The reason is that the :mod: hook bypasses the hash that is setup to allow individual users to trace their own functions and uses the global hash directly. But if the global hash has not been set up, it will cause a bug: echo '*:mod:radeon' > /sys/kernel/debug/set_ftrace_filter produces: [drm:drm_mode_getfb] *ERROR* invalid framebuffer id [drm:radeon_crtc_page_flip] *ERROR* failed to reserve new rbo buffer before flip BUG: unable to handle kernel paging request at ffffffff8160ec90 IP: [] add_hash_entry+0x66/0xd0 PGD 1a05067 PUD 1a09063 PMD 80000000016001e1 Oops: 0003 [#1] SMP Jul 7 04:02:28 phyllis kernel: [55303.858604] CPU 1 Modules linked in: cryptd aes_x86_64 aes_generic binfmt_misc rfcomm bnep ip6table_filter hid radeon r8169 ahci libahci mii ttm drm_kms_helper drm video i2c_algo_bit intel_agp intel_gtt Pid: 10344, comm: bash Tainted: G WC 3.0.0-rc5 #1 Dell Inc. Inspiron N5010/0YXXJJ RIP: 0010:[] [] add_hash_entry+0x66/0xd0 RSP: 0018:ffff88003a96bda8 EFLAGS: 00010246 RAX: ffff8801301735c0 RBX: ffffffff8160ec80 RCX: 0000000000306ee0 RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff880137c92940 RBP: ffff88003a96bdb8 R08: ffff880137c95680 R09: 0000000000000000 R10: 0000000000000001 R11: 0000000000000000 R12: ffffffff81c9df78 R13: ffff8801153d1000 R14: 0000000000000000 R15: 0000000000000000 FS: 00007f329c18a700(0000) GS:ffff880137c80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffffffff8160ec90 CR3: 000000003002b000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process bash (pid: 10344, threadinfo ffff88003a96a000, task ffff88012fcfc470) Stack: 0000000000000fd0 00000000000000fc ffff88003a96be38 ffffffff810d92f5 ffff88011c4c4e00 ffff880000000000 000000000b69f4d0 ffffffff8160ec80 ffff8800300e6f06 0000000081130295 0000000000000282 ffff8800300e6f00 Call Trace: [] match_records+0x155/0x1b0 [] ftrace_mod_callback+0xbc/0x100 [] ftrace_regex_write+0x16f/0x210 [] ftrace_filter_write+0xf/0x20 [] vfs_write+0xc8/0x190 [] sys_write+0x51/0x90 [] system_call_fastpath+0x16/0x1b Code: 48 8b 33 31 d2 48 85 f6 75 33 49 89 d4 4c 03 63 08 49 8b 14 24 48 85 d2 48 89 10 74 04 48 89 42 08 49 89 04 24 4c 89 60 08 31 d2 RIP [] add_hash_entry+0x66/0xd0 RSP CR2: ffffffff8160ec90 ---[ end trace a5d031828efdd88e ]--- Reported-by: Brian Marete Signed-off-by: Steven Rostedt --- include/linux/ftrace.h | 3 ++- kernel/trace/ftrace.c | 12 +++--------- kernel/trace/trace_functions.c | 3 ++- 3 files changed, 7 insertions(+), 11 deletions(-) (limited to 'kernel') diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h index 9d88e1cb5dbb..ed0eb5254d1c 100644 --- a/include/linux/ftrace.h +++ b/include/linux/ftrace.h @@ -123,7 +123,8 @@ stack_trace_sysctl(struct ctl_table *table, int write, struct ftrace_func_command { struct list_head list; char *name; - int (*func)(char *func, char *cmd, + int (*func)(struct ftrace_hash *hash, + char *func, char *cmd, char *params, int enable); }; diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 908038f57440..1c4c0b087e1d 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -2407,10 +2407,9 @@ ftrace_match_module_records(struct ftrace_hash *hash, char *buff, char *mod) */ static int -ftrace_mod_callback(char *func, char *cmd, char *param, int enable) +ftrace_mod_callback(struct ftrace_hash *hash, + char *func, char *cmd, char *param, int enable) { - struct ftrace_ops *ops = &global_ops; - struct ftrace_hash *hash; char *mod; int ret = -EINVAL; @@ -2430,11 +2429,6 @@ ftrace_mod_callback(char *func, char *cmd, char *param, int enable) if (!strlen(mod)) return ret; - if (enable) - hash = ops->filter_hash; - else - hash = ops->notrace_hash; - ret = ftrace_match_module_records(hash, func, mod); if (!ret) ret = -EINVAL; @@ -2760,7 +2754,7 @@ static int ftrace_process_regex(struct ftrace_hash *hash, mutex_lock(&ftrace_cmd_mutex); list_for_each_entry(p, &ftrace_commands, list) { if (strcmp(p->name, command) == 0) { - ret = p->func(func, command, next, enable); + ret = p->func(hash, func, command, next, enable); goto out_unlock; } } diff --git a/kernel/trace/trace_functions.c b/kernel/trace/trace_functions.c index 8d0e1cc4e974..c7b0c6a7db09 100644 --- a/kernel/trace/trace_functions.c +++ b/kernel/trace/trace_functions.c @@ -324,7 +324,8 @@ ftrace_trace_onoff_unreg(char *glob, char *cmd, char *param) } static int -ftrace_trace_onoff_callback(char *glob, char *cmd, char *param, int enable) +ftrace_trace_onoff_callback(struct ftrace_hash *hash, + char *glob, char *cmd, char *param, int enable) { struct ftrace_probe_ops *ops; void *count = (void *)-1; -- cgit v1.2.3 From e4a3f541f0b67fdad98b326c851dfe7f4b6b6dad Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 14 Jun 2011 19:02:29 -0400 Subject: tracing: Still trace filtered irq functions when irq trace is disabled If a function is set to be traced by the set_graph_function, but the option funcgraph-irqs is zero, and the traced function happens to be called from a interrupt, it will not be traced. The point of funcgraph-irqs is to not trace interrupts when we are preempted by an irq, not to not trace functions we want to trace that happen to be *in* a irq. Luckily the current->trace_recursion element is perfect to add a flag to help us be able to trace functions within an interrupt even when we are not tracing interrupts that preempt the trace. Reported-by: Heiko Carstens Tested-by: Heiko Carstens Signed-off-by: Steven Rostedt --- kernel/trace/trace.h | 50 ++++++++++++++++++++++++------------ kernel/trace/trace_functions_graph.c | 2 +- 2 files changed, 35 insertions(+), 17 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index a3e2db708072..651f35be372a 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -278,6 +278,29 @@ struct tracer { }; +/* Only current can touch trace_recursion */ +#define trace_recursion_inc() do { (current)->trace_recursion++; } while (0) +#define trace_recursion_dec() do { (current)->trace_recursion--; } while (0) + +/* Ring buffer has the 10 LSB bits to count */ +#define trace_recursion_buffer() ((current)->trace_recursion & 0x3ff) + +/* for function tracing recursion */ +#define TRACE_INTERNAL_BIT (1<<11) +#define TRACE_GLOBAL_BIT (1<<12) +/* + * Abuse of the trace_recursion. + * As we need a way to maintain state if we are tracing the function + * graph in irq because we want to trace a particular function that + * was called in irq context but we have irq tracing off. Since this + * can only be modified by current, we can reuse trace_recursion. + */ +#define TRACE_IRQ_BIT (1<<13) + +#define trace_recursion_set(bit) do { (current)->trace_recursion |= (bit); } while (0) +#define trace_recursion_clear(bit) do { (current)->trace_recursion &= ~(bit); } while (0) +#define trace_recursion_test(bit) ((current)->trace_recursion & (bit)) + #define TRACE_PIPE_ALL_CPU -1 int tracer_init(struct tracer *t, struct trace_array *tr); @@ -516,8 +539,18 @@ static inline int ftrace_graph_addr(unsigned long addr) return 1; for (i = 0; i < ftrace_graph_count; i++) { - if (addr == ftrace_graph_funcs[i]) + if (addr == ftrace_graph_funcs[i]) { + /* + * If no irqs are to be traced, but a set_graph_function + * is set, and called by an interrupt handler, we still + * want to trace it. + */ + if (in_irq()) + trace_recursion_set(TRACE_IRQ_BIT); + else + trace_recursion_clear(TRACE_IRQ_BIT); return 1; + } } return 0; @@ -794,19 +827,4 @@ extern const char *__stop___trace_bprintk_fmt[]; FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print)) #include "trace_entries.h" -/* Only current can touch trace_recursion */ -#define trace_recursion_inc() do { (current)->trace_recursion++; } while (0) -#define trace_recursion_dec() do { (current)->trace_recursion--; } while (0) - -/* Ring buffer has the 10 LSB bits to count */ -#define trace_recursion_buffer() ((current)->trace_recursion & 0x3ff) - -/* for function tracing recursion */ -#define TRACE_INTERNAL_BIT (1<<11) -#define TRACE_GLOBAL_BIT (1<<12) - -#define trace_recursion_set(bit) do { (current)->trace_recursion |= (bit); } while (0) -#define trace_recursion_clear(bit) do { (current)->trace_recursion &= ~(bit); } while (0) -#define trace_recursion_test(bit) ((current)->trace_recursion & (bit)) - #endif /* _LINUX_KERNEL_TRACE_H */ diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c index e8d6bb55d719..a7d2a4c653d8 100644 --- a/kernel/trace/trace_functions_graph.c +++ b/kernel/trace/trace_functions_graph.c @@ -227,7 +227,7 @@ int __trace_graph_entry(struct trace_array *tr, static inline int ftrace_graph_ignore_irqs(void) { - if (!ftrace_graph_skip_irqs) + if (!ftrace_graph_skip_irqs || trace_recursion_test(TRACE_IRQ_BIT)) return 0; return in_irq(); -- cgit v1.2.3 From 4376cac66778b25e599be3f5d54f33f58ba8ead7 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Fri, 24 Jun 2011 23:28:13 -0400 Subject: ftrace: Do not disable interrupts for modules in mcount update When I mounted an NFS directory, it caused several modules to be loaded. At the time I was running the preemptirqsoff tracer, and it showed the following output: # tracer: preemptirqsoff # # preemptirqsoff latency trace v1.1.5 on 2.6.33.9-rt30-mrg-test # -------------------------------------------------------------------- # latency: 1177 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) # ----------------- # | task: modprobe-19370 (uid:0 nice:0 policy:0 rt_prio:0) # ----------------- # => started at: ftrace_module_notify # => ended at: ftrace_module_notify # # # _------=> CPU# # / _-----=> irqs-off # | / _----=> need-resched # || / _---=> hardirq/softirq # ||| / _--=> preempt-depth # |||| /_--=> lock-depth # |||||/ delay # cmd pid |||||| time | caller # \ / |||||| \ | / modprobe-19370 3d.... 0us!: ftrace_process_locs <-ftrace_module_notify modprobe-19370 3d.... 1176us : ftrace_process_locs <-ftrace_module_notify modprobe-19370 3d.... 1178us : trace_hardirqs_on <-ftrace_module_notify modprobe-19370 3d.... 1178us : => ftrace_process_locs => ftrace_module_notify => notifier_call_chain => __blocking_notifier_call_chain => blocking_notifier_call_chain => sys_init_module => system_call_fastpath That's over 1ms that interrupts are disabled on a Real-Time kernel! Looking at the cause (being the ftrace author helped), I found that the interrupts are disabled before the code modification of mcounts into nops. The interrupts only need to be disabled on start up around this code, not when modules are being loaded. Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index c997f7371c65..df93392aad89 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -3318,7 +3318,7 @@ static int ftrace_process_locs(struct module *mod, { unsigned long *p; unsigned long addr; - unsigned long flags; + unsigned long flags = 0; /* Shut up gcc */ mutex_lock(&ftrace_lock); p = start; @@ -3336,12 +3336,18 @@ static int ftrace_process_locs(struct module *mod, } /* - * Disable interrupts to prevent interrupts from executing - * code that is being modified. + * We only need to disable interrupts on start up + * because we are modifying code that an interrupt + * may execute, and the modification is not atomic. + * But for modules, nothing runs the code we modify + * until we are finished with it, and there's no + * reason to cause large interrupt latencies while we do it. */ - local_irq_save(flags); + if (!mod) + local_irq_save(flags); ftrace_update_code(mod); - local_irq_restore(flags); + if (!mod) + local_irq_restore(flags); mutex_unlock(&ftrace_lock); return 0; -- cgit v1.2.3 From 41fb61c2d08107ce96a5dcb3a6289b2afd3e135c Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 13 Jul 2011 15:03:44 -0400 Subject: ftrace: Balance records when updating the hash Whenever the hash of the ftrace_ops is updated, the record counts must be balance. This requires disabling the records that are set in the original hash, and then enabling the records that are set in the updated hash. Moving the update into ftrace_hash_move() removes the bug where the hash was updated but the records were not, which results in ftrace triggering a warning and disabling itself because the ftrace_ops filter is updated while the ftrace_ops was registered, and then the failure happens when the ftrace_ops is unregistered. The current code will not trigger this bug, but new code will. Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 49 +++++++++++++++++++++++++++++++++---------------- 1 file changed, 33 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index df93392aad89..853f6f0a4b4a 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1170,8 +1170,14 @@ alloc_and_copy_ftrace_hash(int size_bits, struct ftrace_hash *hash) return NULL; } +static void +ftrace_hash_rec_disable(struct ftrace_ops *ops, int filter_hash); +static void +ftrace_hash_rec_enable(struct ftrace_ops *ops, int filter_hash); + static int -ftrace_hash_move(struct ftrace_hash **dst, struct ftrace_hash *src) +ftrace_hash_move(struct ftrace_ops *ops, int enable, + struct ftrace_hash **dst, struct ftrace_hash *src) { struct ftrace_func_entry *entry; struct hlist_node *tp, *tn; @@ -1181,8 +1187,15 @@ ftrace_hash_move(struct ftrace_hash **dst, struct ftrace_hash *src) unsigned long key; int size = src->count; int bits = 0; + int ret; int i; + /* + * Remove the current set, update the hash and add + * them back. + */ + ftrace_hash_rec_disable(ops, enable); + /* * If the new source is empty, just free dst and assign it * the empty_hash. @@ -1203,9 +1216,10 @@ ftrace_hash_move(struct ftrace_hash **dst, struct ftrace_hash *src) if (bits > FTRACE_HASH_MAX_BITS) bits = FTRACE_HASH_MAX_BITS; + ret = -ENOMEM; new_hash = alloc_ftrace_hash(bits); if (!new_hash) - return -ENOMEM; + goto out; size = 1 << src->size_bits; for (i = 0; i < size; i++) { @@ -1224,7 +1238,16 @@ ftrace_hash_move(struct ftrace_hash **dst, struct ftrace_hash *src) rcu_assign_pointer(*dst, new_hash); free_ftrace_hash_rcu(old_hash); - return 0; + ret = 0; + out: + /* + * Enable regardless of ret: + * On success, we enable the new hash. + * On failure, we re-enable the original hash. + */ + ftrace_hash_rec_enable(ops, enable); + + return ret; } /* @@ -2845,7 +2868,7 @@ ftrace_set_regex(struct ftrace_ops *ops, unsigned char *buf, int len, ftrace_match_records(hash, buf, len); mutex_lock(&ftrace_lock); - ret = ftrace_hash_move(orig_hash, hash); + ret = ftrace_hash_move(ops, enable, orig_hash, hash); mutex_unlock(&ftrace_lock); mutex_unlock(&ftrace_regex_lock); @@ -3028,18 +3051,12 @@ ftrace_regex_release(struct inode *inode, struct file *file) orig_hash = &iter->ops->notrace_hash; mutex_lock(&ftrace_lock); - /* - * Remove the current set, update the hash and add - * them back. - */ - ftrace_hash_rec_disable(iter->ops, filter_hash); - ret = ftrace_hash_move(orig_hash, iter->hash); - if (!ret) { - ftrace_hash_rec_enable(iter->ops, filter_hash); - if (iter->ops->flags & FTRACE_OPS_FL_ENABLED - && ftrace_enabled) - ftrace_run_update_code(FTRACE_ENABLE_CALLS); - } + ret = ftrace_hash_move(iter->ops, filter_hash, + orig_hash, iter->hash); + if (!ret && (iter->ops->flags & FTRACE_OPS_FL_ENABLED) + && ftrace_enabled) + ftrace_run_update_code(FTRACE_ENABLE_CALLS); + mutex_unlock(&ftrace_lock); } free_ftrace_hash(iter->hash); -- cgit v1.2.3 From 072126f4529196f71a97960248bca54fd4554c2d Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 13 Jul 2011 15:08:31 -0400 Subject: ftrace: Update filter when tracing enabled in set_ftrace_filter() Currently, if set_ftrace_filter() is called when the ftrace_ops is active, the function filters will not be updated. They will only be updated when tracing is disabled and re-enabled. Update the functions immediately during set_ftrace_filter(). Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 853f6f0a4b4a..a0dc0de8d649 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -2869,6 +2869,10 @@ ftrace_set_regex(struct ftrace_ops *ops, unsigned char *buf, int len, mutex_lock(&ftrace_lock); ret = ftrace_hash_move(ops, enable, orig_hash, hash); + if (!ret && ops->flags & FTRACE_OPS_FL_ENABLED + && ftrace_enabled) + ftrace_run_update_code(FTRACE_ENABLE_CALLS); + mutex_unlock(&ftrace_lock); mutex_unlock(&ftrace_regex_lock); -- cgit v1.2.3 From 6331c28c962561aee59e5a493b7556a4bb585957 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 13 Jul 2011 15:11:02 -0400 Subject: ftrace: Fix dynamic selftest failure on some archs Archs that do not implement CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST, will fail the dynamic ftrace selftest. The function tracer has a quick 'off' variable that will prevent the call back functions from being called. This variable is called function_trace_stop. In x86, this is implemented directly in the mcount assembly, but for other archs, an intermediate function is used called ftrace_test_stop_func(). In dynamic ftrace, the function pointer variable ftrace_trace_function is used to update the caller code in the mcount caller. But for archs that do not have CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST set, it only calls ftrace_test_stop_func() instead, which in turn calls __ftrace_trace_function. When more than one ftrace_ops is registered, the function it calls is ftrace_ops_list_func(), which will iterate over all registered ftrace_ops and call the callbacks that have their hash matching. The issue happens when two ftrace_ops are registered for different functions and one is then unregistered. The __ftrace_trace_function is then pointed to the remaining ftrace_ops callback function directly. This mean it will be called for all functions that were registered to trace by both ftrace_ops that were registered. This is not an issue for archs with CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST, because the update of ftrace_trace_function doesn't happen until after all functions have been updated, and then the mcount caller is updated. But for those archs that do use the ftrace_test_stop_func(), the update is immediate. The dynamic selftest fails because it hits this situation, and the ftrace_ops that it registers fails to only trace what it was suppose to and instead traces all other functions. The solution is to delay the setting of __ftrace_trace_function until after all the functions have been updated according to the registered ftrace_ops. Also, function_trace_stop is set during the update to prevent function tracing from calling code that is caused by the function tracer itself. Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index a0dc0de8d649..62e26d930535 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -88,6 +88,7 @@ static struct ftrace_ops ftrace_list_end __read_mostly = { static struct ftrace_ops *ftrace_global_list __read_mostly = &ftrace_list_end; static struct ftrace_ops *ftrace_ops_list __read_mostly = &ftrace_list_end; ftrace_func_t ftrace_trace_function __read_mostly = ftrace_stub; +static ftrace_func_t __ftrace_trace_function_delay __read_mostly = ftrace_stub; ftrace_func_t __ftrace_trace_function __read_mostly = ftrace_stub; ftrace_func_t ftrace_pid_function __read_mostly = ftrace_stub; static struct ftrace_ops global_ops; @@ -146,9 +147,11 @@ void clear_ftrace_function(void) { ftrace_trace_function = ftrace_stub; __ftrace_trace_function = ftrace_stub; + __ftrace_trace_function_delay = ftrace_stub; ftrace_pid_function = ftrace_stub; } +#undef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST #ifndef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST /* * For those archs that do not test ftrace_trace_stop in their @@ -207,8 +210,13 @@ static void update_ftrace_function(void) #ifdef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST ftrace_trace_function = func; +#else +#ifdef CONFIG_DYNAMIC_FTRACE + /* do not update till all functions have been modified */ + __ftrace_trace_function_delay = func; #else __ftrace_trace_function = func; +#endif ftrace_trace_function = ftrace_test_stop_func; #endif } @@ -1607,6 +1615,12 @@ static int __ftrace_modify_code(void *data) { int *command = data; + /* + * Do not call function tracer while we update the code. + * We are in stop machine, no worrying about races. + */ + function_trace_stop++; + if (*command & FTRACE_ENABLE_CALLS) ftrace_replace_code(1); else if (*command & FTRACE_DISABLE_CALLS) @@ -1620,6 +1634,18 @@ static int __ftrace_modify_code(void *data) else if (*command & FTRACE_STOP_FUNC_RET) ftrace_disable_ftrace_graph_caller(); +#ifndef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST + /* + * For archs that call ftrace_test_stop_func(), we must + * wait till after we update all the function callers + * before we update the callback. This keeps different + * ops that record different functions from corrupting + * each other. + */ + __ftrace_trace_function = __ftrace_trace_function_delay; +#endif + function_trace_stop--; + return 0; } -- cgit v1.2.3 From 4a9bd3f134decd6d16ead8d288342d57aad486be Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 14 Jul 2011 16:36:53 -0400 Subject: tracing: Have dynamic size event stack traces Currently the stack trace per event in ftace is only 8 frames. This can be quite limiting and sometimes useless. Especially when the "ignore frames" is wrong and we also use up stack frames for the event processing itself. Change this to be dynamic by adding a percpu buffer that we can write a large stack frame into and then copy into the ring buffer. For interrupts and NMIs that come in while another event is being process, will only get to use the 8 frame stack. That should be enough as the task that it interrupted will have the full stack frame anyway. Requested-by: Thomas Gleixner Signed-off-by: Steven Rostedt --- include/linux/ftrace_event.h | 1 + kernel/trace/trace.c | 92 +++++++++++++++++++++++++++++++++++++------- kernel/trace/trace_entries.h | 3 +- kernel/trace/trace_output.c | 11 +++--- 4 files changed, 88 insertions(+), 19 deletions(-) (limited to 'kernel') diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h index b1e69eefc203..96efa6794ea5 100644 --- a/include/linux/ftrace_event.h +++ b/include/linux/ftrace_event.h @@ -76,6 +76,7 @@ struct trace_iterator { struct trace_entry *ent; unsigned long lost_events; int leftover; + int ent_size; int cpu; u64 ts; diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index d9c16123f6e2..e5df02c69b1d 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -1248,6 +1248,15 @@ ftrace(struct trace_array *tr, struct trace_array_cpu *data, } #ifdef CONFIG_STACKTRACE + +#define FTRACE_STACK_MAX_ENTRIES (PAGE_SIZE / sizeof(unsigned long)) +struct ftrace_stack { + unsigned long calls[FTRACE_STACK_MAX_ENTRIES]; +}; + +static DEFINE_PER_CPU(struct ftrace_stack, ftrace_stack); +static DEFINE_PER_CPU(int, ftrace_stack_reserve); + static void __ftrace_trace_stack(struct ring_buffer *buffer, unsigned long flags, int skip, int pc, struct pt_regs *regs) @@ -1256,25 +1265,77 @@ static void __ftrace_trace_stack(struct ring_buffer *buffer, struct ring_buffer_event *event; struct stack_entry *entry; struct stack_trace trace; + int use_stack; + int size = FTRACE_STACK_ENTRIES; + + trace.nr_entries = 0; + trace.skip = skip; + + /* + * Since events can happen in NMIs there's no safe way to + * use the per cpu ftrace_stacks. We reserve it and if an interrupt + * or NMI comes in, it will just have to use the default + * FTRACE_STACK_SIZE. + */ + preempt_disable_notrace(); + + use_stack = ++__get_cpu_var(ftrace_stack_reserve); + /* + * We don't need any atomic variables, just a barrier. + * If an interrupt comes in, we don't care, because it would + * have exited and put the counter back to what we want. + * We just need a barrier to keep gcc from moving things + * around. + */ + barrier(); + if (use_stack == 1) { + trace.entries = &__get_cpu_var(ftrace_stack).calls[0]; + trace.max_entries = FTRACE_STACK_MAX_ENTRIES; + + if (regs) + save_stack_trace_regs(regs, &trace); + else + save_stack_trace(&trace); + + if (trace.nr_entries > size) + size = trace.nr_entries; + } else + /* From now on, use_stack is a boolean */ + use_stack = 0; + + size *= sizeof(unsigned long); event = trace_buffer_lock_reserve(buffer, TRACE_STACK, - sizeof(*entry), flags, pc); + sizeof(*entry) + size, flags, pc); if (!event) - return; - entry = ring_buffer_event_data(event); - memset(&entry->caller, 0, sizeof(entry->caller)); + goto out; + entry = ring_buffer_event_data(event); - trace.nr_entries = 0; - trace.max_entries = FTRACE_STACK_ENTRIES; - trace.skip = skip; - trace.entries = entry->caller; + memset(&entry->caller, 0, size); + + if (use_stack) + memcpy(&entry->caller, trace.entries, + trace.nr_entries * sizeof(unsigned long)); + else { + trace.max_entries = FTRACE_STACK_ENTRIES; + trace.entries = entry->caller; + if (regs) + save_stack_trace_regs(regs, &trace); + else + save_stack_trace(&trace); + } + + entry->size = trace.nr_entries; - if (regs) - save_stack_trace_regs(regs, &trace); - else - save_stack_trace(&trace); if (!filter_check_discard(call, entry, buffer, event)) ring_buffer_unlock_commit(buffer, event); + + out: + /* Again, don't let gcc optimize things here */ + barrier(); + __get_cpu_var(ftrace_stack_reserve)--; + preempt_enable_notrace(); + } void ftrace_trace_stack_regs(struct ring_buffer *buffer, unsigned long flags, @@ -1562,7 +1623,12 @@ peek_next_entry(struct trace_iterator *iter, int cpu, u64 *ts, ftrace_enable_cpu(); - return event ? ring_buffer_event_data(event) : NULL; + if (event) { + iter->ent_size = ring_buffer_event_length(event); + return ring_buffer_event_data(event); + } + iter->ent_size = 0; + return NULL; } static struct trace_entry * diff --git a/kernel/trace/trace_entries.h b/kernel/trace/trace_entries.h index e32744c84d94..93365907f219 100644 --- a/kernel/trace/trace_entries.h +++ b/kernel/trace/trace_entries.h @@ -161,7 +161,8 @@ FTRACE_ENTRY(kernel_stack, stack_entry, TRACE_STACK, F_STRUCT( - __array( unsigned long, caller, FTRACE_STACK_ENTRIES ) + __field( int, size ) + __dynamic_array(unsigned long, caller ) ), F_printk("\t=> (%08lx)\n\t=> (%08lx)\n\t=> (%08lx)\n\t=> (%08lx)\n" diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c index e37de492a9e1..51999309a6cf 100644 --- a/kernel/trace/trace_output.c +++ b/kernel/trace/trace_output.c @@ -1107,19 +1107,20 @@ static enum print_line_t trace_stack_print(struct trace_iterator *iter, { struct stack_entry *field; struct trace_seq *s = &iter->seq; - int i; + unsigned long *p; + unsigned long *end; trace_assign_type(field, iter->ent); + end = (unsigned long *)((long)iter->ent + iter->ent_size); if (!trace_seq_puts(s, "\n")) goto partial; - for (i = 0; i < FTRACE_STACK_ENTRIES; i++) { - if (!field->caller[i] || (field->caller[i] == ULONG_MAX)) - break; + + for (p = field->caller; p && *p != ULONG_MAX && p < end; p++) { if (!trace_seq_puts(s, " => ")) goto partial; - if (!seq_print_ip_sym(s, field->caller[i], flags)) + if (!seq_print_ip_sym(s, *p, flags)) goto partial; if (!trace_seq_puts(s, "\n")) goto partial; -- cgit v1.2.3 From f91298709790b9a483752ca3c967845537df2af3 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 9 Jul 2011 00:17:12 +0400 Subject: perf, x86: P4 PMU - Introduce event alias feature Instead of hw_nmi_watchdog_set_attr() weak function and appropriate x86_pmu::hw_watchdog_set_attr() call we introduce even alias mechanism which allow us to drop this routines completely and isolate quirks of Netburst architecture inside P4 PMU code only. The main idea remains the same though -- to allow nmi-watchdog and perf top run simultaneously. Note the aliasing mechanism applies to generic PERF_COUNT_HW_CPU_CYCLES event only because arbitrary event (say passed as RAW initially) might have some additional bits set inside ESCR register changing the behaviour of event and we can't guarantee anymore that alias event will give the same result. P.S. Thanks a huge to Don and Steven for for testing and early review. Acked-by: Don Zickus Tested-by: Steven Rostedt Signed-off-by: Cyrill Gorcunov CC: Ingo Molnar CC: Peter Zijlstra CC: Stephane Eranian CC: Lin Ming CC: Arnaldo Carvalho de Melo CC: Frederic Weisbecker Link: http://lkml.kernel.org/r/20110708201712.GS23657@sun Signed-off-by: Steven Rostedt --- arch/x86/include/asm/perf_event_p4.h | 33 +++++++++ arch/x86/kernel/cpu/perf_event.c | 7 -- arch/x86/kernel/cpu/perf_event_p4.c | 135 +++++++++++++++++++++++++++-------- kernel/watchdog.c | 2 - 4 files changed, 139 insertions(+), 38 deletions(-) (limited to 'kernel') diff --git a/arch/x86/include/asm/perf_event_p4.h b/arch/x86/include/asm/perf_event_p4.h index 56fd9e3abbda..4d86c86178f2 100644 --- a/arch/x86/include/asm/perf_event_p4.h +++ b/arch/x86/include/asm/perf_event_p4.h @@ -101,6 +101,14 @@ #define P4_CONFIG_HT_SHIFT 63 #define P4_CONFIG_HT (1ULL << P4_CONFIG_HT_SHIFT) +/* + * If an event has alias it should be marked + * with a special bit. (Don't forget to check + * P4_PEBS_CONFIG_MASK and related bits on + * modification.) + */ +#define P4_CONFIG_ALIASABLE (1 << 9) + /* * The bits we allow to pass for RAW events */ @@ -123,6 +131,31 @@ (p4_config_pack_escr(P4_CONFIG_MASK_ESCR)) | \ (p4_config_pack_cccr(P4_CONFIG_MASK_CCCR)) +/* + * In case of event aliasing we need to preserve some + * caller bits otherwise the mapping won't be complete. + */ +#define P4_CONFIG_EVENT_ALIAS_MASK \ + (p4_config_pack_escr(P4_CONFIG_MASK_ESCR) | \ + p4_config_pack_cccr(P4_CCCR_EDGE | \ + P4_CCCR_THRESHOLD_MASK | \ + P4_CCCR_COMPLEMENT | \ + P4_CCCR_COMPARE)) + +#define P4_CONFIG_EVENT_ALIAS_IMMUTABLE_BITS \ + ((P4_CONFIG_HT) | \ + p4_config_pack_escr(P4_ESCR_T0_OS | \ + P4_ESCR_T0_USR | \ + P4_ESCR_T1_OS | \ + P4_ESCR_T1_USR) | \ + p4_config_pack_cccr(P4_CCCR_OVF | \ + P4_CCCR_CASCADE | \ + P4_CCCR_FORCE_OVF | \ + P4_CCCR_THREAD_ANY | \ + P4_CCCR_OVF_PMI_T0 | \ + P4_CCCR_OVF_PMI_T1 | \ + P4_CONFIG_ALIASABLE)) + static inline bool p4_is_event_cascaded(u64 config) { u32 cccr = p4_config_unpack_cccr(config); diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index c53d433c3dde..b7a010fce8c3 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -274,7 +274,6 @@ struct x86_pmu { void (*enable_all)(int added); void (*enable)(struct perf_event *); void (*disable)(struct perf_event *); - void (*hw_watchdog_set_attr)(struct perf_event_attr *attr); int (*hw_config)(struct perf_event *event); int (*schedule_events)(struct cpu_hw_events *cpuc, int n, int *assign); unsigned eventsel; @@ -360,12 +359,6 @@ static u64 __read_mostly hw_cache_extra_regs [PERF_COUNT_HW_CACHE_OP_MAX] [PERF_COUNT_HW_CACHE_RESULT_MAX]; -void hw_nmi_watchdog_set_attr(struct perf_event_attr *wd_attr) -{ - if (x86_pmu.hw_watchdog_set_attr) - x86_pmu.hw_watchdog_set_attr(wd_attr); -} - /* * Propagate event elapsed time into the generic event. * Can only be executed on the CPU where the event is active. diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c index fb901c5080f7..8b7a0c306784 100644 --- a/arch/x86/kernel/cpu/perf_event_p4.c +++ b/arch/x86/kernel/cpu/perf_event_p4.c @@ -570,11 +570,92 @@ static __initconst const u64 p4_hw_cache_event_ids }, }; +/* + * Because of Netburst being quite restricted in now + * many same events can run simultaneously, we use + * event aliases, ie different events which have the + * same functionallity but use non-intersected resources + * (ESCR/CCCR/couter registers). This allow us to run + * two or more semi-same events together. It is done + * transparently to a user space. + * + * Never set any cusom internal bits such as P4_CONFIG_HT, + * P4_CONFIG_ALIASABLE or bits for P4_PEBS_METRIC, they are + * either up-to-dated automatically either not appliable + * at all. + * + * And be really carefull choosing aliases! + */ +struct p4_event_alias { + u64 orig; + u64 alter; +} p4_event_aliases[] = { + { + /* + * Non-halted cycles can be substituted with + * non-sleeping cycles (see Intel SDM Vol3b for + * details). + */ + .orig = + p4_config_pack_escr(P4_ESCR_EVENT(P4_EVENT_GLOBAL_POWER_EVENTS) | + P4_ESCR_EMASK_BIT(P4_EVENT_GLOBAL_POWER_EVENTS, RUNNING)), + .alter = + p4_config_pack_escr(P4_ESCR_EVENT(P4_EVENT_EXECUTION_EVENT) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS0)| + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS1)| + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS2)| + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS3)| + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS0) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS1) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS2) | + P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS3))| + p4_config_pack_cccr(P4_CCCR_THRESHOLD(15) | P4_CCCR_COMPLEMENT | + P4_CCCR_COMPARE), + }, +}; + +static u64 p4_get_alias_event(u64 config) +{ + u64 config_match; + int i; + + /* + * Probably we're lucky and don't have to do + * matching over all config bits. + */ + if (!(config & P4_CONFIG_ALIASABLE)) + return 0; + + config_match = config & P4_CONFIG_EVENT_ALIAS_MASK; + + /* + * If an event was previously swapped to the alter config + * we should swap it back otherwise contnention on registers + * will return back. + */ + for (i = 0; i < ARRAY_SIZE(p4_event_aliases); i++) { + if (config_match == p4_event_aliases[i].orig) { + config_match = p4_event_aliases[i].alter; + break; + } else if (config_match == p4_event_aliases[i].alter) { + config_match = p4_event_aliases[i].orig; + break; + } + } + + if (i >= ARRAY_SIZE(p4_event_aliases)) + return 0; + + return config_match | + (config & P4_CONFIG_EVENT_ALIAS_IMMUTABLE_BITS); +} + static u64 p4_general_events[PERF_COUNT_HW_MAX] = { /* non-halted CPU clocks */ [PERF_COUNT_HW_CPU_CYCLES] = p4_config_pack_escr(P4_ESCR_EVENT(P4_EVENT_GLOBAL_POWER_EVENTS) | - P4_ESCR_EMASK_BIT(P4_EVENT_GLOBAL_POWER_EVENTS, RUNNING)), + P4_ESCR_EMASK_BIT(P4_EVENT_GLOBAL_POWER_EVENTS, RUNNING)) | + P4_CONFIG_ALIASABLE, /* * retired instructions @@ -719,31 +800,6 @@ static int p4_validate_raw_event(struct perf_event *event) return 0; } -static void p4_hw_watchdog_set_attr(struct perf_event_attr *wd_attr) -{ - /* - * Watchdog ticks are special on Netburst, we use - * that named "non-sleeping" ticks as recommended - * by Intel SDM Vol3b. - */ - WARN_ON_ONCE(wd_attr->type != PERF_TYPE_HARDWARE || - wd_attr->config != PERF_COUNT_HW_CPU_CYCLES); - - wd_attr->type = PERF_TYPE_RAW; - wd_attr->config = - p4_config_pack_escr(P4_ESCR_EVENT(P4_EVENT_EXECUTION_EVENT) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS0) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS1) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS2) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, NBOGUS3) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS0) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS1) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS2) | - P4_ESCR_EMASK_BIT(P4_EVENT_EXECUTION_EVENT, BOGUS3)) | - p4_config_pack_cccr(P4_CCCR_THRESHOLD(15) | P4_CCCR_COMPLEMENT | - P4_CCCR_COMPARE); -} - static int p4_hw_config(struct perf_event *event) { int cpu = get_cpu(); @@ -1159,6 +1215,8 @@ static int p4_pmu_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign struct p4_event_bind *bind; unsigned int i, thread, num; int cntr_idx, escr_idx; + u64 config_alias; + int pass; bitmap_zero(used_mask, X86_PMC_IDX_MAX); bitmap_zero(escr_mask, P4_ESCR_MSR_TABLE_SIZE); @@ -1167,6 +1225,17 @@ static int p4_pmu_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign hwc = &cpuc->event_list[i]->hw; thread = p4_ht_thread(cpu); + pass = 0; + +again: + /* + * Aliases are swappable so we may hit circular + * lock if both original config and alias need + * resources (MSR registers) which already busy. + */ + if (pass > 2) + goto done; + bind = p4_config_get_bind(hwc->config); escr_idx = p4_get_escr_idx(bind->escr_msr[thread]); if (unlikely(escr_idx == -1)) @@ -1180,8 +1249,17 @@ static int p4_pmu_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign } cntr_idx = p4_next_cntr(thread, used_mask, bind); - if (cntr_idx == -1 || test_bit(escr_idx, escr_mask)) - goto done; + if (cntr_idx == -1 || test_bit(escr_idx, escr_mask)) { + /* + * Probably an event alias is still available. + */ + config_alias = p4_get_alias_event(hwc->config); + if (!config_alias) + goto done; + hwc->config = config_alias; + pass++; + goto again; + } p4_pmu_swap_config_ts(hwc, cpu); if (assign) @@ -1218,7 +1296,6 @@ static __initconst const struct x86_pmu p4_pmu = { .cntval_bits = ARCH_P4_CNTRVAL_BITS, .cntval_mask = ARCH_P4_CNTRVAL_MASK, .max_period = (1ULL << (ARCH_P4_CNTRVAL_BITS - 1)) - 1, - .hw_watchdog_set_attr = p4_hw_watchdog_set_attr, .hw_config = p4_hw_config, .schedule_events = p4_pmu_schedule_events, /* diff --git a/kernel/watchdog.c b/kernel/watchdog.c index a933e3a0398b..36491cd5b7d4 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -200,7 +200,6 @@ static int is_softlockup(unsigned long touch_ts) } #ifdef CONFIG_HARDLOCKUP_DETECTOR -void __weak hw_nmi_watchdog_set_attr(struct perf_event_attr *wd_attr) { } static struct perf_event_attr wd_hw_attr = { .type = PERF_TYPE_HARDWARE, @@ -372,7 +371,6 @@ static int watchdog_nmi_enable(int cpu) wd_attr = &wd_hw_attr; wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh); - hw_nmi_watchdog_set_attr(wd_attr); /* Try to register using hardware perf events */ event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL); -- cgit v1.2.3 From 7143f168e2aa4bc7f8a8bcfe46d8dc52f7be869a Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 27 Jun 2011 16:26:36 +0900 Subject: tracing/kprobes: Rename probe_* to trace_probe_* Rename probe_* to trace_probe_* for avoiding namespace confliction. This also fixes improper names of find_probe_event() and cleanup_all_probes() to find_trace_probe() and release_all_trace_probes() respectively. Signed-off-by: Masami Hiramatsu Cc: Arnaldo Carvalho de Melo Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Frederic Weisbecker Link: http://lkml.kernel.org/r/20110627072636.6528.60374.stgit@fedora15 Signed-off-by: Steven Rostedt --- kernel/trace/trace_kprobe.c | 43 ++++++++++++++++++++++--------------------- 1 file changed, 22 insertions(+), 21 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 7db7b68c6c37..14b88ed65bb7 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -555,12 +555,12 @@ struct trace_probe { (sizeof(struct probe_arg) * (n))) -static __kprobes int probe_is_return(struct trace_probe *tp) +static __kprobes int trace_probe_is_return(struct trace_probe *tp) { return tp->rp.handler != NULL; } -static __kprobes const char *probe_symbol(struct trace_probe *tp) +static __kprobes const char *trace_probe_symbol(struct trace_probe *tp) { return tp->symbol ? tp->symbol : "unknown"; } @@ -671,7 +671,7 @@ static void free_trace_probe(struct trace_probe *tp) kfree(tp); } -static struct trace_probe *find_probe_event(const char *event, +static struct trace_probe *find_trace_probe(const char *event, const char *group) { struct trace_probe *tp; @@ -686,7 +686,7 @@ static struct trace_probe *find_probe_event(const char *event, /* Unregister a trace_probe and probe_event: call with locking probe_lock */ static void unregister_trace_probe(struct trace_probe *tp) { - if (probe_is_return(tp)) + if (trace_probe_is_return(tp)) unregister_kretprobe(&tp->rp); else unregister_kprobe(&tp->rp.kp); @@ -703,7 +703,7 @@ static int register_trace_probe(struct trace_probe *tp) mutex_lock(&probe_lock); /* register as an event */ - old_tp = find_probe_event(tp->call.name, tp->call.class->system); + old_tp = find_trace_probe(tp->call.name, tp->call.class->system); if (old_tp) { /* delete old event */ unregister_trace_probe(old_tp); @@ -716,7 +716,7 @@ static int register_trace_probe(struct trace_probe *tp) } tp->rp.kp.flags |= KPROBE_FLAG_DISABLED; - if (probe_is_return(tp)) + if (trace_probe_is_return(tp)) ret = register_kretprobe(&tp->rp); else ret = register_kprobe(&tp->rp.kp); @@ -1025,7 +1025,7 @@ static int create_trace_probe(int argc, char **argv) return -EINVAL; } mutex_lock(&probe_lock); - tp = find_probe_event(event, group); + tp = find_trace_probe(event, group); if (!tp) { mutex_unlock(&probe_lock); pr_info("Event %s/%s doesn't exist.\n", group, event); @@ -1144,7 +1144,7 @@ error: return ret; } -static void cleanup_all_probes(void) +static void release_all_trace_probes(void) { struct trace_probe *tp; @@ -1181,15 +1181,16 @@ static int probes_seq_show(struct seq_file *m, void *v) struct trace_probe *tp = v; int i; - seq_printf(m, "%c", probe_is_return(tp) ? 'r' : 'p'); + seq_printf(m, "%c", trace_probe_is_return(tp) ? 'r' : 'p'); seq_printf(m, ":%s/%s", tp->call.class->system, tp->call.name); if (!tp->symbol) seq_printf(m, " 0x%p", tp->rp.kp.addr); else if (tp->rp.kp.offset) - seq_printf(m, " %s+%u", probe_symbol(tp), tp->rp.kp.offset); + seq_printf(m, " %s+%u", trace_probe_symbol(tp), + tp->rp.kp.offset); else - seq_printf(m, " %s", probe_symbol(tp)); + seq_printf(m, " %s", trace_probe_symbol(tp)); for (i = 0; i < tp->nr_args; i++) seq_printf(m, " %s=%s", tp->args[i].name, tp->args[i].comm); @@ -1209,7 +1210,7 @@ static int probes_open(struct inode *inode, struct file *file) { if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) - cleanup_all_probes(); + release_all_trace_probes(); return seq_open(file, &probes_seq_op); } @@ -1518,7 +1519,7 @@ static int probe_event_enable(struct ftrace_event_call *call) struct trace_probe *tp = (struct trace_probe *)call->data; tp->flags |= TP_FLAG_TRACE; - if (probe_is_return(tp)) + if (trace_probe_is_return(tp)) return enable_kretprobe(&tp->rp); else return enable_kprobe(&tp->rp.kp); @@ -1530,7 +1531,7 @@ static void probe_event_disable(struct ftrace_event_call *call) tp->flags &= ~TP_FLAG_TRACE; if (!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE))) { - if (probe_is_return(tp)) + if (trace_probe_is_return(tp)) disable_kretprobe(&tp->rp); else disable_kprobe(&tp->rp.kp); @@ -1598,7 +1599,7 @@ static int __set_print_fmt(struct trace_probe *tp, char *buf, int len) const char *fmt, *arg; - if (!probe_is_return(tp)) { + if (!trace_probe_is_return(tp)) { fmt = "(%lx)"; arg = "REC->" FIELD_STRING_IP; } else { @@ -1722,7 +1723,7 @@ static int probe_perf_enable(struct ftrace_event_call *call) tp->flags |= TP_FLAG_PROFILE; - if (probe_is_return(tp)) + if (trace_probe_is_return(tp)) return enable_kretprobe(&tp->rp); else return enable_kprobe(&tp->rp.kp); @@ -1735,7 +1736,7 @@ static void probe_perf_disable(struct ftrace_event_call *call) tp->flags &= ~TP_FLAG_PROFILE; if (!(tp->flags & TP_FLAG_TRACE)) { - if (probe_is_return(tp)) + if (trace_probe_is_return(tp)) disable_kretprobe(&tp->rp); else disable_kprobe(&tp->rp.kp); @@ -1807,7 +1808,7 @@ static int register_probe_event(struct trace_probe *tp) /* Initialize ftrace_event_call */ INIT_LIST_HEAD(&call->class->fields); - if (probe_is_return(tp)) { + if (trace_probe_is_return(tp)) { call->event.funcs = &kretprobe_funcs; call->class->define_fields = kretprobe_event_define_fields; } else { @@ -1899,7 +1900,7 @@ static __init int kprobe_trace_self_tests_init(void) warn++; } else { /* Enable trace point */ - tp = find_probe_event("testprobe", KPROBE_EVENT_SYSTEM); + tp = find_trace_probe("testprobe", KPROBE_EVENT_SYSTEM); if (WARN_ON_ONCE(tp == NULL)) { pr_warning("error on getting new probe.\n"); warn++; @@ -1914,7 +1915,7 @@ static __init int kprobe_trace_self_tests_init(void) warn++; } else { /* Enable trace point */ - tp = find_probe_event("testprobe2", KPROBE_EVENT_SYSTEM); + tp = find_trace_probe("testprobe2", KPROBE_EVENT_SYSTEM); if (WARN_ON_ONCE(tp == NULL)) { pr_warning("error on getting new probe.\n"); warn++; @@ -1940,7 +1941,7 @@ static __init int kprobe_trace_self_tests_init(void) } end: - cleanup_all_probes(); + release_all_trace_probes(); if (warn) pr_cont("NG: Some tests are failed. Please check them.\n"); else -- cgit v1.2.3 From f7bc8b61f65726ff98f52e286b28e294499d7a08 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 14 Jul 2011 23:02:27 -0400 Subject: ftrace: Fix regression where ftrace breaks when modules are loaded Enabling function tracer to trace all functions, then load a module and then disable function tracing will cause ftrace to fail. This can also happen by enabling function tracing on the command line: ftrace=function and during boot up, modules are loaded, then you disable function tracing with 'echo nop > current_tracer' you will trigger a bug in ftrace that will shut itself down. The reason is, the new ftrace code keeps ref counts of all ftrace_ops that are registered for tracing. When one or more ftrace_ops are registered, all the records that represent the functions that the ftrace_ops will trace have a ref count incremented. If this ref count is not zero, when the code modification runs, that function will be enabled for tracing. If the ref count is zero, that function will be disabled from tracing. To make sure the accounting was working, FTRACE_WARN_ON()s were added to updating of the ref counts. If the ref count hits its max (> 2^30 ftrace_ops added), or if the ref count goes below zero, a FTRACE_WARN_ON() is triggered which disables all modification of code. Since it is common for ftrace_ops to trace all functions in the kernel, instead of creating > 20,000 hash items for the ftrace_ops, the hash count is just set to zero, and it represents that the ftrace_ops is to trace all functions. This is where the issues arrise. If you enable function tracing to trace all functions, and then add a module, the modules function records do not get the ref count updated. When the function tracer is disabled, all function records ref counts are subtracted. Since the modules never had their ref counts incremented, they go below zero and the FTRACE_WARN_ON() is triggered. The solution to this is rather simple. When modules are loaded, and their functions are added to the the ftrace pool, look to see if any ftrace_ops are registered that trace all functions. And for those, update the ref count for the module function records. Reported-by: Thomas Gleixner Signed-off-by: Steven Rostedt --- kernel/trace/ftrace.c | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 1c4c0b087e1d..ef9271b69b4f 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1744,10 +1744,36 @@ static cycle_t ftrace_update_time; static unsigned long ftrace_update_cnt; unsigned long ftrace_update_tot_cnt; +static int ops_traces_mod(struct ftrace_ops *ops) +{ + struct ftrace_hash *hash; + + hash = ops->filter_hash; + return !!(!hash || !hash->count); +} + static int ftrace_update_code(struct module *mod) { struct dyn_ftrace *p; cycle_t start, stop; + unsigned long ref = 0; + + /* + * When adding a module, we need to check if tracers are + * currently enabled and if they are set to trace all functions. + * If they are, we need to enable the module functions as well + * as update the reference counts for those function records. + */ + if (mod) { + struct ftrace_ops *ops; + + for (ops = ftrace_ops_list; + ops != &ftrace_list_end; ops = ops->next) { + if (ops->flags & FTRACE_OPS_FL_ENABLED && + ops_traces_mod(ops)) + ref++; + } + } start = ftrace_now(raw_smp_processor_id()); ftrace_update_cnt = 0; @@ -1760,7 +1786,7 @@ static int ftrace_update_code(struct module *mod) p = ftrace_new_addrs; ftrace_new_addrs = p->newlist; - p->flags = 0L; + p->flags = ref; /* * Do the initial record conversion from mcount jump @@ -1783,7 +1809,7 @@ static int ftrace_update_code(struct module *mod) * conversion puts the module to the correct state, thus * passing the ftrace_make_call check. */ - if (ftrace_start_up) { + if (ftrace_start_up && ref) { int failed = __ftrace_replace_code(p, 1); if (failed) { ftrace_bug(failed, p->ip); -- cgit v1.2.3 From 1538f888f1e793de04e0f90372352ac1b05833cf Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 27 Jun 2011 16:26:44 +0900 Subject: tracing/kprobes: Merge trace probe enable/disable functions Merge redundant enable/disable functions into enable_trace_probe() and disable_trace_probe(). Signed-off-by: Masami Hiramatsu Cc: Arnaldo Carvalho de Melo Cc: Ingo Molnar Cc: Peter Zijlstra Cc: yrl.pp-manager.tt@hitachi.com Cc: Frederic Weisbecker Cc: Ingo Molnar Link: http://lkml.kernel.org/r/20110627072644.6528.26910.stgit@fedora15 [ converted kprobe selftest to use enable_trace_probe ] Signed-off-by: Steven Rostedt --- kernel/trace/trace_kprobe.c | 92 ++++++++++++++++++--------------------------- 1 file changed, 36 insertions(+), 56 deletions(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 14b88ed65bb7..3be50c52d080 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -683,6 +683,34 @@ static struct trace_probe *find_trace_probe(const char *event, return NULL; } +/* Enable trace_probe - @flag must be TP_FLAG_TRACE or TP_FLAG_PROFILE */ +static int enable_trace_probe(struct trace_probe *tp, int flag) +{ + int ret = 0; + + tp->flags |= flag; + if (tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE)) { + if (trace_probe_is_return(tp)) + ret = enable_kretprobe(&tp->rp); + else + ret = enable_kprobe(&tp->rp.kp); + } + + return ret; +} + +/* Disable trace_probe - @flag must be TP_FLAG_TRACE or TP_FLAG_PROFILE */ +static void disable_trace_probe(struct trace_probe *tp, int flag) +{ + tp->flags &= ~flag; + if (!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE))) { + if (trace_probe_is_return(tp)) + disable_kretprobe(&tp->rp); + else + disable_kprobe(&tp->rp.kp); + } +} + /* Unregister a trace_probe and probe_event: call with locking probe_lock */ static void unregister_trace_probe(struct trace_probe *tp) { @@ -1514,30 +1542,6 @@ partial: return TRACE_TYPE_PARTIAL_LINE; } -static int probe_event_enable(struct ftrace_event_call *call) -{ - struct trace_probe *tp = (struct trace_probe *)call->data; - - tp->flags |= TP_FLAG_TRACE; - if (trace_probe_is_return(tp)) - return enable_kretprobe(&tp->rp); - else - return enable_kprobe(&tp->rp.kp); -} - -static void probe_event_disable(struct ftrace_event_call *call) -{ - struct trace_probe *tp = (struct trace_probe *)call->data; - - tp->flags &= ~TP_FLAG_TRACE; - if (!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE))) { - if (trace_probe_is_return(tp)) - disable_kretprobe(&tp->rp); - else - disable_kprobe(&tp->rp.kp); - } -} - #undef DEFINE_FIELD #define DEFINE_FIELD(type, item, name, is_signed) \ do { \ @@ -1716,49 +1720,25 @@ static __kprobes void kretprobe_perf_func(struct kretprobe_instance *ri, head = this_cpu_ptr(call->perf_events); perf_trace_buf_submit(entry, size, rctx, entry->ret_ip, 1, regs, head); } - -static int probe_perf_enable(struct ftrace_event_call *call) -{ - struct trace_probe *tp = (struct trace_probe *)call->data; - - tp->flags |= TP_FLAG_PROFILE; - - if (trace_probe_is_return(tp)) - return enable_kretprobe(&tp->rp); - else - return enable_kprobe(&tp->rp.kp); -} - -static void probe_perf_disable(struct ftrace_event_call *call) -{ - struct trace_probe *tp = (struct trace_probe *)call->data; - - tp->flags &= ~TP_FLAG_PROFILE; - - if (!(tp->flags & TP_FLAG_TRACE)) { - if (trace_probe_is_return(tp)) - disable_kretprobe(&tp->rp); - else - disable_kprobe(&tp->rp.kp); - } -} #endif /* CONFIG_PERF_EVENTS */ static __kprobes int kprobe_register(struct ftrace_event_call *event, enum trace_reg type) { + struct trace_probe *tp = (struct trace_probe *)event->data; + switch (type) { case TRACE_REG_REGISTER: - return probe_event_enable(event); + return enable_trace_probe(tp, TP_FLAG_TRACE); case TRACE_REG_UNREGISTER: - probe_event_disable(event); + disable_trace_probe(tp, TP_FLAG_TRACE); return 0; #ifdef CONFIG_PERF_EVENTS case TRACE_REG_PERF_REGISTER: - return probe_perf_enable(event); + return enable_trace_probe(tp, TP_FLAG_PROFILE); case TRACE_REG_PERF_UNREGISTER: - probe_perf_disable(event); + disable_trace_probe(tp, TP_FLAG_PROFILE); return 0; #endif } @@ -1905,7 +1885,7 @@ static __init int kprobe_trace_self_tests_init(void) pr_warning("error on getting new probe.\n"); warn++; } else - probe_event_enable(&tp->call); + enable_trace_probe(tp, TP_FLAG_TRACE); } ret = command_trace_probe("r:testprobe2 kprobe_trace_selftest_target " @@ -1920,7 +1900,7 @@ static __init int kprobe_trace_self_tests_init(void) pr_warning("error on getting new probe.\n"); warn++; } else - probe_event_enable(&tp->call); + enable_trace_probe(tp, TP_FLAG_TRACE); } if (warn) -- cgit v1.2.3 From bc81d48d13d8839fae6833c95794c403b2133f36 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 27 Jun 2011 16:26:50 +0900 Subject: kprobes: Return -ENOENT if probe point doesn't exist Return -ENOENT if probe point doesn't exist, but still returns -EINVAL if both of kprobe->addr and kprobe->symbol_name are specified or both are not specified. Acked-by: Ananth N Mavinakayanahalli Signed-off-by: Masami Hiramatsu Cc: Ananth N Mavinakayanahalli Cc: Arnaldo Carvalho de Melo Cc: Ingo Molnar Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Anil S Keshavamurthy Cc: "David S. Miller" Link: http://lkml.kernel.org/r/20110627072650.6528.67329.stgit@fedora15 Signed-off-by: Steven Rostedt --- kernel/kprobes.c | 33 +++++++++++++++++++++++---------- 1 file changed, 23 insertions(+), 10 deletions(-) (limited to 'kernel') diff --git a/kernel/kprobes.c b/kernel/kprobes.c index 77981813a1e7..b30fd54eb985 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c @@ -1255,19 +1255,29 @@ static int __kprobes in_kprobes_functions(unsigned long addr) /* * If we have a symbol_name argument, look it up and add the offset field * to it. This way, we can specify a relative address to a symbol. + * This returns encoded errors if it fails to look up symbol or invalid + * combination of parameters. */ static kprobe_opcode_t __kprobes *kprobe_addr(struct kprobe *p) { kprobe_opcode_t *addr = p->addr; + + if ((p->symbol_name && p->addr) || + (!p->symbol_name && !p->addr)) + goto invalid; + if (p->symbol_name) { - if (addr) - return NULL; kprobe_lookup_name(p->symbol_name, addr); + if (!addr) + return ERR_PTR(-ENOENT); } - if (!addr) - return NULL; - return (kprobe_opcode_t *)(((char *)addr) + p->offset); + addr = (kprobe_opcode_t *)(((char *)addr) + p->offset); + if (addr) + return addr; + +invalid: + return ERR_PTR(-EINVAL); } /* Check passed kprobe is valid and return kprobe in kprobe_table. */ @@ -1311,8 +1321,8 @@ int __kprobes register_kprobe(struct kprobe *p) kprobe_opcode_t *addr; addr = kprobe_addr(p); - if (!addr) - return -EINVAL; + if (IS_ERR(addr)) + return PTR_ERR(addr); p->addr = addr; ret = check_kprobe_rereg(p); @@ -1335,6 +1345,8 @@ int __kprobes register_kprobe(struct kprobe *p) */ probed_mod = __module_text_address((unsigned long) p->addr); if (probed_mod) { + /* Return -ENOENT if fail. */ + ret = -ENOENT; /* * We must hold a refcount of the probed module while updating * its code to prohibit unexpected unloading. @@ -1351,6 +1363,7 @@ int __kprobes register_kprobe(struct kprobe *p) module_put(probed_mod); goto fail_with_jump_label; } + /* ret will be updated by following code */ } preempt_enable(); jump_label_unlock(); @@ -1399,7 +1412,7 @@ out: fail_with_jump_label: preempt_enable(); jump_label_unlock(); - return -EINVAL; + return ret; } EXPORT_SYMBOL_GPL(register_kprobe); @@ -1686,8 +1699,8 @@ int __kprobes register_kretprobe(struct kretprobe *rp) if (kretprobe_blacklist_size) { addr = kprobe_addr(&rp->kp); - if (!addr) - return -EINVAL; + if (IS_ERR(addr)) + return PTR_ERR(addr); for (i = 0; kretprobe_blacklist[i].name != NULL; i++) { if (kretprobe_blacklist[i].addr == addr) -- cgit v1.2.3 From 614243181050436785f5a621749a7da2336a7916 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 27 Jun 2011 16:26:56 +0900 Subject: tracing/kprobes: Support module init function probing To support probing module init functions, kprobe-tracer allows user to define a probe on non-existed function when it is given with a module name. This also enables user to set a probe on a function on a specific module, even if a same name (but different) function is locally defined in another module. The module name must be in the front of function name and separated by a ':'. e.g. btrfs:btrfs_init_sysfs Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Link: http://lkml.kernel.org/r/20110627072656.6528.89970.stgit@fedora15 Signed-off-by: Steven Rostedt --- Documentation/trace/kprobetrace.txt | 9 +- kernel/trace/trace_kprobe.c | 164 ++++++++++++++++++++++++++++++------ 2 files changed, 143 insertions(+), 30 deletions(-) (limited to 'kernel') diff --git a/Documentation/trace/kprobetrace.txt b/Documentation/trace/kprobetrace.txt index c83bd6b4e6e8..d0d0bb9e3e25 100644 --- a/Documentation/trace/kprobetrace.txt +++ b/Documentation/trace/kprobetrace.txt @@ -22,14 +22,15 @@ current_tracer. Instead of that, add probe points via Synopsis of kprobe_events ------------------------- - p[:[GRP/]EVENT] SYMBOL[+offs]|MEMADDR [FETCHARGS] : Set a probe - r[:[GRP/]EVENT] SYMBOL[+0] [FETCHARGS] : Set a return probe + p[:[GRP/]EVENT] [MOD:]SYM[+offs]|MEMADDR [FETCHARGS] : Set a probe + r[:[GRP/]EVENT] [MOD:]SYM[+0] [FETCHARGS] : Set a return probe -:[GRP/]EVENT : Clear a probe GRP : Group name. If omitted, use "kprobes" for it. EVENT : Event name. If omitted, the event name is generated - based on SYMBOL+offs or MEMADDR. - SYMBOL[+offs] : Symbol+offset where the probe is inserted. + based on SYM+offs or MEMADDR. + MOD : Module name which has given SYM. + SYM[+offs] : Symbol+offset where the probe is inserted. MEMADDR : Address where the probe is inserted. FETCHARGS : Arguments. Each probe can have up to 128 args. diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 3be50c52d080..acc6664a6b22 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -536,6 +536,7 @@ struct probe_arg { /* Flags for trace_probe */ #define TP_FLAG_TRACE 1 #define TP_FLAG_PROFILE 2 +#define TP_FLAG_REGISTERED 4 struct trace_probe { struct list_head list; @@ -565,6 +566,39 @@ static __kprobes const char *trace_probe_symbol(struct trace_probe *tp) return tp->symbol ? tp->symbol : "unknown"; } +static __kprobes unsigned long trace_probe_offset(struct trace_probe *tp) +{ + return tp->rp.kp.offset; +} + +static __kprobes bool trace_probe_is_enabled(struct trace_probe *tp) +{ + return !!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE)); +} + +static __kprobes bool trace_probe_is_registered(struct trace_probe *tp) +{ + return !!(tp->flags & TP_FLAG_REGISTERED); +} + +static __kprobes bool trace_probe_has_gone(struct trace_probe *tp) +{ + return !!(kprobe_gone(&tp->rp.kp)); +} + +static __kprobes bool trace_probe_within_module(struct trace_probe *tp, + struct module *mod) +{ + int len = strlen(mod->name); + const char *name = trace_probe_symbol(tp); + return strncmp(mod->name, name, len) == 0 && name[len] == ':'; +} + +static __kprobes bool trace_probe_is_on_module(struct trace_probe *tp) +{ + return !!strchr(trace_probe_symbol(tp), ':'); +} + static int register_probe_event(struct trace_probe *tp); static void unregister_probe_event(struct trace_probe *tp); @@ -689,7 +723,8 @@ static int enable_trace_probe(struct trace_probe *tp, int flag) int ret = 0; tp->flags |= flag; - if (tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE)) { + if (trace_probe_is_enabled(tp) && trace_probe_is_registered(tp) && + !trace_probe_has_gone(tp)) { if (trace_probe_is_return(tp)) ret = enable_kretprobe(&tp->rp); else @@ -703,7 +738,7 @@ static int enable_trace_probe(struct trace_probe *tp, int flag) static void disable_trace_probe(struct trace_probe *tp, int flag) { tp->flags &= ~flag; - if (!(tp->flags & (TP_FLAG_TRACE | TP_FLAG_PROFILE))) { + if (!trace_probe_is_enabled(tp) && trace_probe_is_registered(tp)) { if (trace_probe_is_return(tp)) disable_kretprobe(&tp->rp); else @@ -711,13 +746,64 @@ static void disable_trace_probe(struct trace_probe *tp, int flag) } } -/* Unregister a trace_probe and probe_event: call with locking probe_lock */ -static void unregister_trace_probe(struct trace_probe *tp) +/* Internal register function - just handle k*probes and flags */ +static int __register_trace_probe(struct trace_probe *tp) { + int ret; + + if (trace_probe_is_registered(tp)) + return -EINVAL; + + /* Set/clear disabled flag according to tp->flag */ + if (trace_probe_is_enabled(tp)) + tp->rp.kp.flags &= ~KPROBE_FLAG_DISABLED; + else + tp->rp.kp.flags |= KPROBE_FLAG_DISABLED; + if (trace_probe_is_return(tp)) - unregister_kretprobe(&tp->rp); + ret = register_kretprobe(&tp->rp); else - unregister_kprobe(&tp->rp.kp); + ret = register_kprobe(&tp->rp.kp); + + if (ret == 0) + tp->flags |= TP_FLAG_REGISTERED; + else { + pr_warning("Could not insert probe at %s+%lu: %d\n", + trace_probe_symbol(tp), trace_probe_offset(tp), ret); + if (ret == -ENOENT && trace_probe_is_on_module(tp)) { + pr_warning("This probe might be able to register after" + "target module is loaded. Continue.\n"); + ret = 0; + } else if (ret == -EILSEQ) { + pr_warning("Probing address(0x%p) is not an " + "instruction boundary.\n", + tp->rp.kp.addr); + ret = -EINVAL; + } + } + + return ret; +} + +/* Internal unregister function - just handle k*probes and flags */ +static void __unregister_trace_probe(struct trace_probe *tp) +{ + if (trace_probe_is_registered(tp)) { + if (trace_probe_is_return(tp)) + unregister_kretprobe(&tp->rp); + else + unregister_kprobe(&tp->rp.kp); + tp->flags &= ~TP_FLAG_REGISTERED; + /* Cleanup kprobe for reuse */ + if (tp->rp.kp.symbol_name) + tp->rp.kp.addr = NULL; + } +} + +/* Unregister a trace_probe and probe_event: call with locking probe_lock */ +static void unregister_trace_probe(struct trace_probe *tp) +{ + __unregister_trace_probe(tp); list_del(&tp->list); unregister_probe_event(tp); } @@ -730,41 +816,65 @@ static int register_trace_probe(struct trace_probe *tp) mutex_lock(&probe_lock); - /* register as an event */ + /* Delete old (same name) event if exist */ old_tp = find_trace_probe(tp->call.name, tp->call.class->system); if (old_tp) { - /* delete old event */ unregister_trace_probe(old_tp); free_trace_probe(old_tp); } + + /* Register new event */ ret = register_probe_event(tp); if (ret) { pr_warning("Failed to register probe event(%d)\n", ret); goto end; } - tp->rp.kp.flags |= KPROBE_FLAG_DISABLED; - if (trace_probe_is_return(tp)) - ret = register_kretprobe(&tp->rp); - else - ret = register_kprobe(&tp->rp.kp); - - if (ret) { - pr_warning("Could not insert probe(%d)\n", ret); - if (ret == -EILSEQ) { - pr_warning("Probing address(0x%p) is not an " - "instruction boundary.\n", - tp->rp.kp.addr); - ret = -EINVAL; - } + /* Register k*probe */ + ret = __register_trace_probe(tp); + if (ret < 0) unregister_probe_event(tp); - } else + else list_add_tail(&tp->list, &probe_list); + end: mutex_unlock(&probe_lock); return ret; } +/* Module notifier call back, checking event on the module */ +static int trace_probe_module_callback(struct notifier_block *nb, + unsigned long val, void *data) +{ + struct module *mod = data; + struct trace_probe *tp; + int ret; + + if (val != MODULE_STATE_COMING) + return NOTIFY_DONE; + + /* Update probes on coming module */ + mutex_lock(&probe_lock); + list_for_each_entry(tp, &probe_list, list) { + if (trace_probe_within_module(tp, mod)) { + __unregister_trace_probe(tp); + ret = __register_trace_probe(tp); + if (ret) + pr_warning("Failed to re-register probe %s on" + "%s: %d\n", + tp->call.name, mod->name, ret); + } + } + mutex_unlock(&probe_lock); + + return NOTIFY_DONE; +} + +static struct notifier_block trace_probe_module_nb = { + .notifier_call = trace_probe_module_callback, + .priority = 1 /* Invoked after kprobe module callback */ +}; + /* Split symbol and offset. */ static int split_symbol_offset(char *symbol, unsigned long *offset) { @@ -990,8 +1100,8 @@ static int create_trace_probe(int argc, char **argv) { /* * Argument syntax: - * - Add kprobe: p[:[GRP/]EVENT] KSYM[+OFFS]|KADDR [FETCHARGS] - * - Add kretprobe: r[:[GRP/]EVENT] KSYM[+0] [FETCHARGS] + * - Add kprobe: p[:[GRP/]EVENT] [MOD:]KSYM[+OFFS]|KADDR [FETCHARGS] + * - Add kretprobe: r[:[GRP/]EVENT] [MOD:]KSYM[+0] [FETCHARGS] * Fetch args: * $retval : fetch return value * $stack : fetch stack address @@ -1186,7 +1296,6 @@ static void release_all_trace_probes(void) mutex_unlock(&probe_lock); } - /* Probes listing interfaces */ static void *probes_seq_start(struct seq_file *m, loff_t *pos) { @@ -1827,6 +1936,9 @@ static __init int init_kprobe_trace(void) struct dentry *d_tracer; struct dentry *entry; + if (register_module_notifier(&trace_probe_module_nb)) + return -EINVAL; + d_tracer = tracing_init_dentry(); if (!d_tracer) return 0; -- cgit v1.2.3 From 7f6878a3d707b947603e09d95df0c3a98987e3a4 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 27 Jun 2011 16:27:03 +0900 Subject: tracing/kprobe: Update symbol reference when loading module Since the address of a module-local variable can only be solved after the target module is loaded, the symbol fetch-argument should be updated when loading target module. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Link: http://lkml.kernel.org/r/20110627072703.6528.75042.stgit@fedora15 Signed-off-by: Steven Rostedt --- kernel/trace/trace_kprobe.c | 37 ++++++++++++++++++++++++++++++++++++- 1 file changed, 36 insertions(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index acc6664a6b22..5fb3697bf0e5 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -343,6 +343,14 @@ DEFINE_BASIC_FETCH_FUNCS(deref) DEFINE_FETCH_deref(string) DEFINE_FETCH_deref(string_size) +static __kprobes void update_deref_fetch_param(struct deref_fetch_param *data) +{ + if (CHECK_FETCH_FUNCS(deref, data->orig.fn)) + update_deref_fetch_param(data->orig.data); + else if (CHECK_FETCH_FUNCS(symbol, data->orig.fn)) + update_symbol_cache(data->orig.data); +} + static __kprobes void free_deref_fetch_param(struct deref_fetch_param *data) { if (CHECK_FETCH_FUNCS(deref, data->orig.fn)) @@ -376,6 +384,19 @@ DEFINE_BASIC_FETCH_FUNCS(bitfield) #define fetch_bitfield_string NULL #define fetch_bitfield_string_size NULL +static __kprobes void +update_bitfield_fetch_param(struct bitfield_fetch_param *data) +{ + /* + * Don't check the bitfield itself, because this must be the + * last fetch function. + */ + if (CHECK_FETCH_FUNCS(deref, data->orig.fn)) + update_deref_fetch_param(data->orig.data); + else if (CHECK_FETCH_FUNCS(symbol, data->orig.fn)) + update_symbol_cache(data->orig.data); +} + static __kprobes void free_bitfield_fetch_param(struct bitfield_fetch_param *data) { @@ -389,6 +410,7 @@ free_bitfield_fetch_param(struct bitfield_fetch_param *data) free_symbol_cache(data->orig.data); kfree(data); } + /* Default (unsigned long) fetch type */ #define __DEFAULT_FETCH_TYPE(t) u##t #define _DEFAULT_FETCH_TYPE(t) __DEFAULT_FETCH_TYPE(t) @@ -680,6 +702,16 @@ error: return ERR_PTR(ret); } +static void update_probe_arg(struct probe_arg *arg) +{ + if (CHECK_FETCH_FUNCS(bitfield, arg->fetch.fn)) + update_bitfield_fetch_param(arg->fetch.data); + else if (CHECK_FETCH_FUNCS(deref, arg->fetch.fn)) + update_deref_fetch_param(arg->fetch.data); + else if (CHECK_FETCH_FUNCS(symbol, arg->fetch.fn)) + update_symbol_cache(arg->fetch.data); +} + static void free_probe_arg(struct probe_arg *arg) { if (CHECK_FETCH_FUNCS(bitfield, arg->fetch.fn)) @@ -749,11 +781,14 @@ static void disable_trace_probe(struct trace_probe *tp, int flag) /* Internal register function - just handle k*probes and flags */ static int __register_trace_probe(struct trace_probe *tp) { - int ret; + int i, ret; if (trace_probe_is_registered(tp)) return -EINVAL; + for (i = 0; i < tp->nr_args; i++) + update_probe_arg(&tp->args[i]); + /* Set/clear disabled flag according to tp->flag */ if (trace_probe_is_enabled(tp)) tp->rp.kp.flags &= ~KPROBE_FLAG_DISABLED; -- cgit v1.2.3 From 9985c20f9e4aee6857c08246b273a3695a52b929 Mon Sep 17 00:00:00 2001 From: Lin Ming Date: Thu, 30 Jun 2011 08:09:55 +0000 Subject: perf: Remove perf_event_attr::type check PMU type id can be allocated dynamically, so perf_event_attr::type check when copying attribute from userspace to kernel is not valid. Signed-off-by: Lin Ming Cc: Robert Richter Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/1309421396-17438-4-git-send-email-ming.m.lin@intel.com Signed-off-by: Ingo Molnar --- kernel/events/core.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'kernel') diff --git a/kernel/events/core.c b/kernel/events/core.c index 0567e32d71aa..b8785e26ee1c 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -5926,13 +5926,6 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr, if (ret) return -EFAULT; - /* - * If the type exists, the corresponding creation will verify - * the attr->config. - */ - if (attr->type >= PERF_TYPE_MAX) - return -EINVAL; - if (attr->__reserved_1) return -EINVAL; -- cgit v1.2.3