diff options
author | Steven Rostedt <rostedt@goodmis.org> | 2008-10-01 13:14:09 -0400 |
---|---|---|
committer | Ingo Molnar <mingo@elte.hu> | 2008-10-14 10:39:09 +0200 |
commit | 38697053fa006411224a1790e2adb8216440ab0f (patch) | |
tree | 30daab3a6ba93f1c8c922397ffe8a0d6a220e8b1 | |
parent | e4c2ce82ca2710e17cb4df8eb2b249fa2eb5af30 (diff) | |
download | linux-38697053fa006411224a1790e2adb8216440ab0f.tar.bz2 |
ftrace: preempt disable over interrupt disable
With the new ring buffer infrastructure in ftrace, I'm trying to make
ftrace a little more light weight.
This patch converts a lot of the local_irq_save/restore into
preempt_disable/enable. The original preempt count in a lot of cases
has to be sent in as a parameter so that it can be recorded correctly.
Some places were recording it incorrectly before anyway.
This is also laying the ground work to make ftrace a little bit
more reentrant, and remove all locking. The function tracers must
still protect from reentrancy.
Note: All the function tracers must be careful when using preempt_disable.
It must do the following:
resched = need_resched();
preempt_disable_notrace();
[...]
if (resched)
preempt_enable_no_resched_notrace();
else
preempt_enable_notrace();
The reason is that if this function traces schedule() itself, the
preempt_enable_notrace() will cause a schedule, which will lead
us into a recursive failure.
If we needed to reschedule before calling preempt_disable, we
should have already scheduled. Since we did not, this is most
likely that we should not and are probably inside a schedule
function.
If resched was not set, we still need to catch the need resched
flag being set when preemption was off and the if case at the
end will catch that for us.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
-rw-r--r-- | kernel/trace/trace.c | 123 | ||||
-rw-r--r-- | kernel/trace/trace.h | 13 | ||||
-rw-r--r-- | kernel/trace/trace_boot.c | 2 | ||||
-rw-r--r-- | kernel/trace/trace_irqsoff.c | 13 | ||||
-rw-r--r-- | kernel/trace/trace_mmiotrace.c | 4 | ||||
-rw-r--r-- | kernel/trace/trace_sched_switch.c | 9 | ||||
-rw-r--r-- | kernel/trace/trace_sched_wakeup.c | 13 |
7 files changed, 97 insertions, 80 deletions
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 948f7d821c62..1cd2e8143bb4 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -652,12 +652,10 @@ void tracing_record_cmdline(struct task_struct *tsk) } void -tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags) +tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags, + int pc) { struct task_struct *tsk = current; - unsigned long pc; - - pc = preempt_count(); entry->preempt_count = pc & 0xff; entry->pid = (tsk) ? tsk->pid : 0; @@ -670,7 +668,8 @@ tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags) void trace_function(struct trace_array *tr, struct trace_array_cpu *data, - unsigned long ip, unsigned long parent_ip, unsigned long flags) + unsigned long ip, unsigned long parent_ip, unsigned long flags, + int pc) { struct ring_buffer_event *event; struct ftrace_entry *entry; @@ -685,7 +684,7 @@ trace_function(struct trace_array *tr, struct trace_array_cpu *data, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, flags); + tracing_generic_entry_update(&entry->ent, flags, pc); entry->ent.type = TRACE_FN; entry->ip = ip; entry->parent_ip = parent_ip; @@ -694,16 +693,17 @@ trace_function(struct trace_array *tr, struct trace_array_cpu *data, void ftrace(struct trace_array *tr, struct trace_array_cpu *data, - unsigned long ip, unsigned long parent_ip, unsigned long flags) + unsigned long ip, unsigned long parent_ip, unsigned long flags, + int pc) { if (likely(!atomic_read(&data->disabled))) - trace_function(tr, data, ip, parent_ip, flags); + trace_function(tr, data, ip, parent_ip, flags, pc); } -void __trace_stack(struct trace_array *tr, - struct trace_array_cpu *data, - unsigned long flags, - int skip) +static void ftrace_trace_stack(struct trace_array *tr, + struct trace_array_cpu *data, + unsigned long flags, + int skip, int pc) { struct ring_buffer_event *event; struct stack_entry *entry; @@ -718,7 +718,7 @@ void __trace_stack(struct trace_array *tr, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, flags); + tracing_generic_entry_update(&entry->ent, flags, pc); entry->ent.type = TRACE_STACK; memset(&entry->caller, 0, sizeof(entry->caller)); @@ -732,9 +732,18 @@ void __trace_stack(struct trace_array *tr, ring_buffer_unlock_commit(tr->buffer, event, irq_flags); } -void -__trace_special(void *__tr, void *__data, - unsigned long arg1, unsigned long arg2, unsigned long arg3) +void __trace_stack(struct trace_array *tr, + struct trace_array_cpu *data, + unsigned long flags, + int skip) +{ + ftrace_trace_stack(tr, data, flags, skip, preempt_count()); +} + +static void +ftrace_trace_special(void *__tr, void *__data, + unsigned long arg1, unsigned long arg2, unsigned long arg3, + int pc) { struct ring_buffer_event *event; struct trace_array_cpu *data = __data; @@ -747,23 +756,30 @@ __trace_special(void *__tr, void *__data, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, 0); + tracing_generic_entry_update(&entry->ent, 0, pc); entry->ent.type = TRACE_SPECIAL; entry->arg1 = arg1; entry->arg2 = arg2; entry->arg3 = arg3; ring_buffer_unlock_commit(tr->buffer, event, irq_flags); - __trace_stack(tr, data, irq_flags, 4); + ftrace_trace_stack(tr, data, irq_flags, 4, pc); trace_wake_up(); } void +__trace_special(void *__tr, void *__data, + unsigned long arg1, unsigned long arg2, unsigned long arg3) +{ + ftrace_trace_special(__tr, __data, arg1, arg2, arg3, preempt_count()); +} + +void tracing_sched_switch_trace(struct trace_array *tr, struct trace_array_cpu *data, struct task_struct *prev, struct task_struct *next, - unsigned long flags) + unsigned long flags, int pc) { struct ring_buffer_event *event; struct ctx_switch_entry *entry; @@ -774,7 +790,7 @@ tracing_sched_switch_trace(struct trace_array *tr, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, flags); + tracing_generic_entry_update(&entry->ent, flags, pc); entry->ent.type = TRACE_CTX; entry->prev_pid = prev->pid; entry->prev_prio = prev->prio; @@ -784,7 +800,7 @@ tracing_sched_switch_trace(struct trace_array *tr, entry->next_state = next->state; entry->next_cpu = task_cpu(next); ring_buffer_unlock_commit(tr->buffer, event, irq_flags); - __trace_stack(tr, data, flags, 5); + ftrace_trace_stack(tr, data, flags, 5, pc); } void @@ -792,7 +808,7 @@ tracing_sched_wakeup_trace(struct trace_array *tr, struct trace_array_cpu *data, struct task_struct *wakee, struct task_struct *curr, - unsigned long flags) + unsigned long flags, int pc) { struct ring_buffer_event *event; struct ctx_switch_entry *entry; @@ -803,7 +819,7 @@ tracing_sched_wakeup_trace(struct trace_array *tr, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, flags); + tracing_generic_entry_update(&entry->ent, flags, pc); entry->ent.type = TRACE_WAKE; entry->prev_pid = curr->pid; entry->prev_prio = curr->prio; @@ -813,7 +829,7 @@ tracing_sched_wakeup_trace(struct trace_array *tr, entry->next_state = wakee->state; entry->next_cpu = task_cpu(wakee); ring_buffer_unlock_commit(tr->buffer, event, irq_flags); - __trace_stack(tr, data, flags, 6); + ftrace_trace_stack(tr, data, flags, 6, pc); trace_wake_up(); } @@ -823,23 +839,24 @@ ftrace_special(unsigned long arg1, unsigned long arg2, unsigned long arg3) { struct trace_array *tr = &global_trace; struct trace_array_cpu *data; - unsigned long flags; long disabled; int cpu; + int pc; if (tracing_disabled || !tr->ctrl) return; - local_irq_save(flags); + pc = preempt_count(); + preempt_disable_notrace(); cpu = raw_smp_processor_id(); data = tr->data[cpu]; disabled = atomic_inc_return(&data->disabled); if (likely(disabled == 1)) - __trace_special(tr, data, arg1, arg2, arg3); + ftrace_trace_special(tr, data, arg1, arg2, arg3, pc); atomic_dec(&data->disabled); - local_irq_restore(flags); + preempt_enable_notrace(); } #ifdef CONFIG_FTRACE @@ -850,7 +867,8 @@ function_trace_call(unsigned long ip, unsigned long parent_ip) struct trace_array_cpu *data; unsigned long flags; long disabled; - int cpu; + int cpu, resched; + int pc; if (unlikely(!ftrace_function_enabled)) return; @@ -858,16 +876,22 @@ function_trace_call(unsigned long ip, unsigned long parent_ip) if (skip_trace(ip)) return; - local_irq_save(flags); + pc = preempt_count(); + resched = need_resched(); + preempt_disable_notrace(); + local_save_flags(flags); cpu = raw_smp_processor_id(); data = tr->data[cpu]; disabled = atomic_inc_return(&data->disabled); if (likely(disabled == 1)) - trace_function(tr, data, ip, parent_ip, flags); + trace_function(tr, data, ip, parent_ip, flags, pc); atomic_dec(&data->disabled); - local_irq_restore(flags); + if (resched) + preempt_enable_no_resched_notrace(); + else + preempt_enable_notrace(); } static struct ftrace_ops trace_ops __read_mostly = @@ -2508,9 +2532,6 @@ tracing_read_pipe(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos) { struct trace_iterator *iter = filp->private_data; -#ifdef CONFIG_FTRACE - int ftrace_save; -#endif ssize_t sret; /* return any leftover data */ @@ -2593,20 +2614,6 @@ waitagain: offsetof(struct trace_iterator, seq)); iter->pos = -1; - /* - * We need to stop all tracing on all CPUS to read the - * the next buffer. This is a bit expensive, but is - * not done often. We fill all what we can read, - * and then release the locks again. - */ - - local_irq_disable(); -#ifdef CONFIG_FTRACE - ftrace_save = ftrace_enabled; - ftrace_enabled = 0; -#endif - smp_wmb(); - while (find_next_entry_inc(iter) != NULL) { enum print_line_t ret; int len = iter->seq.len; @@ -2624,11 +2631,6 @@ waitagain: break; } -#ifdef CONFIG_FTRACE - ftrace_enabled = ftrace_save; -#endif - local_irq_enable(); - /* Now copy what we have to the user */ sret = trace_seq_to_user(&iter->seq, ubuf, cnt); if (iter->seq.readpos >= iter->seq.len) @@ -2960,12 +2962,13 @@ int trace_vprintk(unsigned long ip, const char *fmt, va_list args) struct print_entry *entry; unsigned long flags, irq_flags; long disabled; - int cpu, len = 0, size; + int cpu, len = 0, size, pc; if (!tr->ctrl || tracing_disabled) return 0; - local_irq_save(flags); + pc = preempt_count(); + preempt_disable_notrace(); cpu = raw_smp_processor_id(); data = tr->data[cpu]; disabled = atomic_inc_return(&data->disabled); @@ -2973,7 +2976,7 @@ int trace_vprintk(unsigned long ip, const char *fmt, va_list args) if (unlikely(disabled != 1)) goto out; - spin_lock(&trace_buf_lock); + spin_lock_irqsave(&trace_buf_lock, flags); len = vsnprintf(trace_buf, TRACE_BUF_SIZE, fmt, args); len = min(len, TRACE_BUF_SIZE-1); @@ -2984,7 +2987,7 @@ int trace_vprintk(unsigned long ip, const char *fmt, va_list args) if (!event) goto out_unlock; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, flags); + tracing_generic_entry_update(&entry->ent, flags, pc); entry->ent.type = TRACE_PRINT; entry->ip = ip; @@ -2993,11 +2996,11 @@ int trace_vprintk(unsigned long ip, const char *fmt, va_list args) ring_buffer_unlock_commit(tr->buffer, event, irq_flags); out_unlock: - spin_unlock(&trace_buf_lock); + spin_unlock_irqrestore(&trace_buf_lock, flags); out: atomic_dec(&data->disabled); - local_irq_restore(flags); + preempt_enable_notrace(); return len; } diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index f02042d0d828..f1f99572cde7 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -288,35 +288,36 @@ void init_tracer_sysprof_debugfs(struct dentry *d_tracer); struct trace_entry *tracing_get_trace_entry(struct trace_array *tr, struct trace_array_cpu *data); void tracing_generic_entry_update(struct trace_entry *entry, - unsigned long flags); + unsigned long flags, + int pc); void ftrace(struct trace_array *tr, struct trace_array_cpu *data, unsigned long ip, unsigned long parent_ip, - unsigned long flags); + unsigned long flags, int pc); void tracing_sched_switch_trace(struct trace_array *tr, struct trace_array_cpu *data, struct task_struct *prev, struct task_struct *next, - unsigned long flags); + unsigned long flags, int pc); void tracing_record_cmdline(struct task_struct *tsk); void tracing_sched_wakeup_trace(struct trace_array *tr, struct trace_array_cpu *data, struct task_struct *wakee, struct task_struct *cur, - unsigned long flags); + unsigned long flags, int pc); void trace_special(struct trace_array *tr, struct trace_array_cpu *data, unsigned long arg1, unsigned long arg2, - unsigned long arg3); + unsigned long arg3, int pc); void trace_function(struct trace_array *tr, struct trace_array_cpu *data, unsigned long ip, unsigned long parent_ip, - unsigned long flags); + unsigned long flags, int pc); void tracing_start_cmdline_record(void); void tracing_stop_cmdline_record(void); diff --git a/kernel/trace/trace_boot.c b/kernel/trace/trace_boot.c index 43bde20b95bd..f2dac6f1cf06 100644 --- a/kernel/trace/trace_boot.c +++ b/kernel/trace/trace_boot.c @@ -95,7 +95,7 @@ void trace_boot(struct boot_trace *it) if (!event) goto out; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, 0); + tracing_generic_entry_update(&entry->ent, 0, 0); entry->ent.type = TRACE_BOOT; entry->initcall = *it; ring_buffer_unlock_commit(tr->buffer, event, irq_flags); diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c index 37ad49407f27..f925dbbff2a6 100644 --- a/kernel/trace/trace_irqsoff.c +++ b/kernel/trace/trace_irqsoff.c @@ -95,7 +95,7 @@ irqsoff_tracer_call(unsigned long ip, unsigned long parent_ip) disabled = atomic_inc_return(&data->disabled); if (likely(disabled == 1)) - trace_function(tr, data, ip, parent_ip, flags); + trace_function(tr, data, ip, parent_ip, flags, preempt_count()); atomic_dec(&data->disabled); } @@ -130,6 +130,7 @@ check_critical_timing(struct trace_array *tr, unsigned long latency, t0, t1; cycle_t T0, T1, delta; unsigned long flags; + int pc; /* * usecs conversion is slow so we try to delay the conversion @@ -144,13 +145,15 @@ check_critical_timing(struct trace_array *tr, if (!report_latency(delta)) goto out; + pc = preempt_count(); + spin_lock_irqsave(&max_trace_lock, flags); /* check if we are still the max latency */ if (!report_latency(delta)) goto out_unlock; - trace_function(tr, data, CALLER_ADDR0, parent_ip, flags); + trace_function(tr, data, CALLER_ADDR0, parent_ip, flags, pc); latency = nsecs_to_usecs(delta); @@ -174,7 +177,7 @@ out: data->critical_sequence = max_sequence; data->preempt_timestamp = ftrace_now(cpu); tracing_reset(tr, cpu); - trace_function(tr, data, CALLER_ADDR0, parent_ip, flags); + trace_function(tr, data, CALLER_ADDR0, parent_ip, flags, pc); } static inline void @@ -207,7 +210,7 @@ start_critical_timing(unsigned long ip, unsigned long parent_ip) local_save_flags(flags); - trace_function(tr, data, ip, parent_ip, flags); + trace_function(tr, data, ip, parent_ip, flags, preempt_count()); per_cpu(tracing_cpu, cpu) = 1; @@ -241,7 +244,7 @@ stop_critical_timing(unsigned long ip, unsigned long parent_ip) atomic_inc(&data->disabled); local_save_flags(flags); - trace_function(tr, data, ip, parent_ip, flags); + trace_function(tr, data, ip, parent_ip, flags, preempt_count()); check_critical_timing(tr, data, parent_ip ? : ip, cpu); data->critical_start = 0; atomic_dec(&data->disabled); diff --git a/kernel/trace/trace_mmiotrace.c b/kernel/trace/trace_mmiotrace.c index 0e819f47bb7a..f28484618ff0 100644 --- a/kernel/trace/trace_mmiotrace.c +++ b/kernel/trace/trace_mmiotrace.c @@ -324,7 +324,7 @@ static void __trace_mmiotrace_rw(struct trace_array *tr, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, 0); + tracing_generic_entry_update(&entry->ent, 0, preempt_count()); entry->ent.type = TRACE_MMIO_RW; entry->rw = *rw; ring_buffer_unlock_commit(tr->buffer, event, irq_flags); @@ -352,7 +352,7 @@ static void __trace_mmiotrace_map(struct trace_array *tr, if (!event) return; entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, 0); + tracing_generic_entry_update(&entry->ent, 0, preempt_count()); entry->ent.type = TRACE_MMIO_MAP; entry->map = *map; ring_buffer_unlock_commit(tr->buffer, event, irq_flags); diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c index e0b06db0f7af..c7fa08a5b7f4 100644 --- a/kernel/trace/trace_sched_switch.c +++ b/kernel/trace/trace_sched_switch.c @@ -26,6 +26,7 @@ probe_sched_switch(struct rq *__rq, struct task_struct *prev, unsigned long flags; long disabled; int cpu; + int pc; if (!atomic_read(&sched_ref)) return; @@ -36,13 +37,14 @@ probe_sched_switch(struct rq *__rq, struct task_struct *prev, if (!tracer_enabled) return; + pc = preempt_count(); local_irq_save(flags); cpu = raw_smp_processor_id(); data = ctx_trace->data[cpu]; disabled = atomic_inc_return(&data->disabled); if (likely(disabled == 1)) - tracing_sched_switch_trace(ctx_trace, data, prev, next, flags); + tracing_sched_switch_trace(ctx_trace, data, prev, next, flags, pc); atomic_dec(&data->disabled); local_irq_restore(flags); @@ -54,11 +56,12 @@ probe_sched_wakeup(struct rq *__rq, struct task_struct *wakee) struct trace_array_cpu *data; unsigned long flags; long disabled; - int cpu; + int cpu, pc; if (!likely(tracer_enabled)) return; + pc = preempt_count(); tracing_record_cmdline(current); local_irq_save(flags); @@ -68,7 +71,7 @@ probe_sched_wakeup(struct rq *__rq, struct task_struct *wakee) if (likely(disabled == 1)) tracing_sched_wakeup_trace(ctx_trace, data, wakee, current, - flags); + flags, pc); atomic_dec(&data->disabled); local_irq_restore(flags); diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c index 01e75e0639b7..fe4a252c2363 100644 --- a/kernel/trace/trace_sched_wakeup.c +++ b/kernel/trace/trace_sched_wakeup.c @@ -44,10 +44,12 @@ wakeup_tracer_call(unsigned long ip, unsigned long parent_ip) long disabled; int resched; int cpu; + int pc; if (likely(!wakeup_task)) return; + pc = preempt_count(); resched = need_resched(); preempt_disable_notrace(); @@ -70,7 +72,7 @@ wakeup_tracer_call(unsigned long ip, unsigned long parent_ip) if (task_cpu(wakeup_task) != cpu) goto unlock; - trace_function(tr, data, ip, parent_ip, flags); + trace_function(tr, data, ip, parent_ip, flags, pc); unlock: __raw_spin_unlock(&wakeup_lock); @@ -121,6 +123,7 @@ probe_wakeup_sched_switch(struct rq *rq, struct task_struct *prev, unsigned long flags; long disabled; int cpu; + int pc; tracing_record_cmdline(prev); @@ -139,6 +142,8 @@ probe_wakeup_sched_switch(struct rq *rq, struct task_struct *prev, if (next != wakeup_task) return; + pc = preempt_count(); + /* The task we are waiting for is waking up */ data = wakeup_trace->data[wakeup_cpu]; @@ -155,7 +160,7 @@ probe_wakeup_sched_switch(struct rq *rq, struct task_struct *prev, if (unlikely(!tracer_enabled || next != wakeup_task)) goto out_unlock; - trace_function(wakeup_trace, data, CALLER_ADDR1, CALLER_ADDR2, flags); + trace_function(wakeup_trace, data, CALLER_ADDR1, CALLER_ADDR2, flags, pc); /* * usecs conversion is slow so we try to delay the conversion @@ -220,6 +225,7 @@ probe_wakeup(struct rq *rq, struct task_struct *p) int cpu = smp_processor_id(); unsigned long flags; long disabled; + int pc; if (likely(!tracer_enabled)) return; @@ -232,6 +238,7 @@ probe_wakeup(struct rq *rq, struct task_struct *p) p->prio >= current->prio) return; + pc = preempt_count(); disabled = atomic_inc_return(&wakeup_trace->data[cpu]->disabled); if (unlikely(disabled != 1)) goto out; @@ -256,7 +263,7 @@ probe_wakeup(struct rq *rq, struct task_struct *p) wakeup_trace->data[wakeup_cpu]->preempt_timestamp = ftrace_now(cpu); trace_function(wakeup_trace, wakeup_trace->data[wakeup_cpu], - CALLER_ADDR1, CALLER_ADDR2, flags); + CALLER_ADDR1, CALLER_ADDR2, flags, pc); out_locked: __raw_spin_unlock(&wakeup_lock); |