linux/kernel/trace
Peter Zijlstra (Intel) 86038c5ea8 perf: Avoid horrible stack usage
Both Linus (most recent) and Steve (a while ago) reported that perf
related callbacks have massive stack bloat.

The problem is that software events need a pt_regs in order to
properly report the event location and unwind stack. And because we
could not assume one was present we allocated one on stack and filled
it with minimal bits required for operation.

Now, pt_regs is quite large, so this is undesirable. Furthermore it
turns out that most sites actually have a pt_regs pointer available,
making this even more onerous, as the stack space is pointless waste.

This patch addresses the problem by observing that software events
have well defined nesting semantics, therefore we can use static
per-cpu storage instead of on-stack.

Linus made the further observation that all but the scheduler callers
of perf_sw_event() have a pt_regs available, so we change the regular
perf_sw_event() to require a valid pt_regs (where it used to be
optional) and add perf_sw_event_sched() for the scheduler.

We have a scheduler specific call instead of a more generic _noregs()
like construct because we can assume non-recursion from the scheduler
and thereby simplify the code further (_noregs would have to put the
recursion context call inline in order to assertain which __perf_regs
element to use).

One last note on the implementation of perf_trace_buf_prepare(); we
allow .regs = NULL for those cases where we already have a pt_regs
pointer available and do not need another.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Javi Merino <javi.merino@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Petr Mladek <pmladek@suse.cz>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Tom Zanussi <tom.zanussi@linux.intel.com>
Cc: Vaibhav Nagarnaik <vnagarnaik@google.com>
Link: http://lkml.kernel.org/r/20141216115041.GW3337@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-01-14 15:11:45 +01:00
..
blktrace.c Merge branch 'for-3.19/core' of git://git.kernel.dk/linux-block 2014-12-13 14:14:23 -08:00
ftrace.c ftrace, kprobes: Support IPMODIFY flag to find IP modify conflict 2014-11-21 14:42:10 -05:00
Kconfig tracing: Remove function_trace_stop and HAVE_FUNCTION_TRACE_MCOUNT_TEST 2014-07-18 13:58:12 -04:00
Makefile tracing / PM: Replace CONFIG_PM_RUNTIME with CONFIG_PM 2014-12-13 02:23:30 +01:00
power-traces.c PM / tracing: remove deprecated power trace API 2013-01-26 00:39:12 +01:00
ring_buffer_benchmark.c sched, cleanup, treewide: Remove set_current_state(TASK_RUNNING) after schedule() 2014-09-19 12:35:17 +02:00
ring_buffer.c There was a lot of clean ups and minor fixes. One of those clean ups was 2014-12-10 19:58:13 -08:00
rpm-traces.c PM / Runtime: Introduce trace points for tracing rpm_* functions 2011-09-27 22:53:27 +02:00
trace_benchmark.c tracing: Only calculate stats of tracepoint benchmarks for 2^32 times 2014-06-06 00:41:38 -04:00
trace_benchmark.h tracing: Add tracepoint benchmark tracepoint 2014-05-29 22:49:54 -04:00
trace_branch.c tracing: Have branch tracer use trace_handle_return() helper function 2014-11-19 15:25:41 -05:00
trace_clock.c tracing: Fix wraparound problems in "uptime" trace clock 2014-07-21 09:56:12 -04:00
trace_entries.h tracing: Add trace_puts() for even faster trace_printk() tracing 2013-03-15 00:35:55 -04:00
trace_event_perf.c perf: Avoid horrible stack usage 2015-01-14 15:11:45 +01:00
trace_events_filter_test.h
trace_events_filter.c tracing: Allow NOT to filter AND and OR clauses 2014-12-03 10:00:27 -05:00
trace_events_trigger.c trace: Replace single-character seq_puts with seq_putc 2014-11-14 07:55:55 -05:00
trace_events.c As the merge window is still open, and this code was not as complex 2014-12-16 12:53:59 -08:00
trace_export.c tracing: Fix anonymous unions in struct ftrace_event_call 2014-04-09 20:02:55 -04:00
trace_functions_graph.c This code is a fork from the trace-3.19 pull as it needed the trace_seq 2014-12-10 20:35:41 -08:00
trace_functions.c tracing/trivial: Fix typos and make an int into a bool 2014-11-20 10:05:36 -05:00
trace_irqsoff.c tracing: Allow irq/preempt tracers to be used by instances 2014-04-21 13:59:29 -04:00
trace_kdb.c KGDB/KDB fixes and cleanups 2015-01-09 20:51:10 -08:00
trace_kprobe.c perf: Avoid horrible stack usage 2015-01-14 15:11:45 +01:00
trace_mmiotrace.c tracing: Do not check return values of trace_seq_p*() for mmio tracer 2014-11-19 15:25:44 -05:00
trace_nop.c tracing: Remove mock up poll wait function 2014-04-30 08:40:05 -04:00
trace_output.c tracing: Add additional marks to signal very large time deltas 2014-12-03 17:10:13 -05:00
trace_output.h tracing: Add trace_seq_has_overflowed() and trace_handle_return() 2014-11-19 15:25:39 -05:00
trace_printk.c trace: Replace single-character seq_puts with seq_putc 2014-11-14 07:55:55 -05:00
trace_probe.c tracing/probes: Do not use return value of trace_seq_printf() 2014-11-19 15:25:44 -05:00
trace_probe.h kprobes, ftrace: Use NOKPROBE_SYMBOL macro in ftrace 2014-04-24 10:26:39 +02:00
trace_sched_switch.c tracing: Move tracing_sched_{switch,wakeup}() into wakeup tracer 2014-11-11 12:43:15 -05:00
trace_sched_wakeup.c tracing: Move tracing_sched_{switch,wakeup}() into wakeup tracer 2014-11-11 12:43:15 -05:00
trace_selftest_dynamic.c
trace_selftest.c Seems that Peter Zijlstra added a new check that is making old 2014-10-12 07:28:55 -04:00
trace_seq.c tracing: Use trace_seq_used() and seq_buf_used() instead of len 2014-11-19 22:01:15 -05:00
trace_stack.c sched: Add helper for task stack page overrun checking 2014-09-19 12:35:23 +02:00
trace_stat.c trace/trace_stat: use rbtree postorder iteration helper instead of opencoding 2013-11-05 16:01:47 -05:00
trace_stat.h
trace_syscalls.c perf: Avoid horrible stack usage 2015-01-14 15:11:45 +01:00
trace_uprobe.c perf: Avoid horrible stack usage 2015-01-14 15:11:45 +01:00
trace.c As the merge window is still open, and this code was not as complex 2014-12-16 12:53:59 -08:00
trace.h tracing: Add tp_printk cmdline to have tracepoints go to printk() 2014-12-15 10:17:38 -05:00