Messages in this thread | | | Date | Fri, 24 Oct 2014 11:47:37 +0300 | From | Adrian Hunter <> | Subject | Re: [PATCH 04/16] perf tools: Add a thread stack for synthesizing call chains |
| |
On 23/10/14 23:51, Arnaldo Carvalho de Melo wrote: > Em Thu, Oct 23, 2014 at 01:45:12PM +0300, Adrian Hunter escreveu: >> Add a thread stack for synthesizing call chains from call >> and return events. >> >> Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> >> --- >> tools/perf/Makefile.perf | 2 + >> tools/perf/util/event.h | 26 +++++++ >> tools/perf/util/thread-stack.c | 151 +++++++++++++++++++++++++++++++++++++++++ >> tools/perf/util/thread-stack.h | 32 +++++++++ >> tools/perf/util/thread.c | 3 + >> tools/perf/util/thread.h | 3 + >> 6 files changed, 217 insertions(+) >> create mode 100644 tools/perf/util/thread-stack.c >> create mode 100644 tools/perf/util/thread-stack.h >> >> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf >> index 262916f..5bbe1ff 100644 >> --- a/tools/perf/Makefile.perf >> +++ b/tools/perf/Makefile.perf >> @@ -304,6 +304,7 @@ LIB_H += ui/util.h >> LIB_H += ui/ui.h >> LIB_H += util/data.h >> LIB_H += util/kvm-stat.h >> +LIB_H += util/thread-stack.h >> >> LIB_OBJS += $(OUTPUT)util/abspath.o >> LIB_OBJS += $(OUTPUT)util/alias.o >> @@ -380,6 +381,7 @@ LIB_OBJS += $(OUTPUT)util/srcline.o >> LIB_OBJS += $(OUTPUT)util/data.o >> LIB_OBJS += $(OUTPUT)util/tsc.o >> LIB_OBJS += $(OUTPUT)util/cloexec.o >> +LIB_OBJS += $(OUTPUT)util/thread-stack.o >> >> LIB_OBJS += $(OUTPUT)ui/setup.o >> LIB_OBJS += $(OUTPUT)ui/helpline.o >> diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h >> index c89518e..e84f929 100644 >> --- a/tools/perf/util/event.h >> +++ b/tools/perf/util/event.h >> @@ -143,6 +143,32 @@ struct branch_stack { >> struct branch_entry entries[0]; >> }; >> >> +enum { >> + PERF_FLAG_BRANCH = 1ULL << 0, >> + PERF_FLAG_CALL = 1ULL << 1, >> + PERF_FLAG_RETURN = 1ULL << 2, >> + PERF_FLAG_CONDITIONAL = 1ULL << 3, >> + PERF_FLAG_SYSCALLRET = 1ULL << 4, >> + PERF_FLAG_ASYNC = 1ULL << 5, >> + PERF_FLAG_INTERRUPT = 1ULL << 6, >> + PERF_FLAG_TX_ABORT = 1ULL << 7, >> + PERF_FLAG_TRACE_BEGIN = 1ULL << 8, >> + PERF_FLAG_TRACE_END = 1ULL << 9, >> + PERF_FLAG_IN_TX = 1ULL << 10, >> +}; >> + >> +#define PERF_BRANCH_MASK (\ >> + PERF_FLAG_BRANCH |\ >> + PERF_FLAG_CALL |\ >> + PERF_FLAG_RETURN |\ >> + PERF_FLAG_CONDITIONAL |\ >> + PERF_FLAG_SYSCALLRET |\ >> + PERF_FLAG_ASYNC |\ >> + PERF_FLAG_INTERRUPT |\ >> + PERF_FLAG_TX_ABORT |\ >> + PERF_FLAG_TRACE_BEGIN |\ >> + PERF_FLAG_TRACE_END) >> + >> struct perf_sample { >> u64 ip; >> u32 pid, tid; >> diff --git a/tools/perf/util/thread-stack.c b/tools/perf/util/thread-stack.c >> new file mode 100644 >> index 0000000..c1ca2a9 >> --- /dev/null >> +++ b/tools/perf/util/thread-stack.c >> @@ -0,0 +1,151 @@ >> +/* >> + * thread-stack.c: Synthesize a thread's stack using call / return events >> + * Copyright (c) 2014, Intel Corporation. >> + * >> + * This program is free software; you can redistribute it and/or modify it >> + * under the terms and conditions of the GNU General Public License, >> + * version 2, as published by the Free Software Foundation. >> + * >> + * This program is distributed in the hope it will be useful, but WITHOUT >> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or >> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for >> + * more details. >> + * >> + */ >> + >> +#include "thread.h" >> +#include "event.h" >> +#include "util.h" >> +#include "thread-stack.h" >> + >> +#define STACK_GROWTH 4096 >> + >> +struct thread_stack_entry { >> + u64 ret_addr; >> +}; >> + >> +struct thread_stack { >> + struct thread_stack_entry *stack; >> + size_t cnt; >> + size_t sz; >> + u64 trace_nr; >> +}; >> + >> +static void thread_stack__grow(struct thread_stack *ts) >> +{ >> + struct thread_stack_entry *new_stack; >> + size_t sz, new_sz; >> + >> + new_sz = ts->sz + STACK_GROWTH; >> + sz = new_sz * sizeof(struct thread_stack_entry); >> + new_stack = realloc(ts->stack, sz); >> + if (new_stack) { >> + ts->stack = new_stack; >> + ts->sz = new_sz; >> + } >> +} >> + >> +static struct thread_stack *thread_stack__new(void) >> +{ >> + struct thread_stack *ts; >> + >> + ts = zalloc(sizeof(struct thread_stack)); >> + if (!ts) >> + return NULL; >> + >> + thread_stack__grow(ts); >> + if (!ts->stack) { >> + free(ts); >> + return NULL; >> + } >> + >> + return ts; >> +} >> + >> +static void thread_stack__push(struct thread_stack *ts, u64 ret_addr) >> +{ >> + if (ts->cnt == ts->sz) { >> + thread_stack__grow(ts); >> + if (ts->cnt == ts->sz) >> + ts->cnt = 0; >> + } >> + >> + ts->stack[ts->cnt++].ret_addr = ret_addr; >> +} > > So can you elaborate on the use case, i.e. this silently trows the > existing stack contents away if it doesn't grow, looks strange :-\ > > Merits some explanation about why this is OK, to say the least.
I considered the call stack to be augmenting the information so not worth failing the whole session over. Equally memory allocation failure is likely to be fatal anyway due to oom-killers or other allocations failing.
But it could certainly print a warning, so I will add that.
> >> + >> +static void thread_stack__pop(struct thread_stack *ts, u64 ret_addr) >> +{ >> + if (!ts->cnt) >> + return; >> + >> + if (ts->stack[ts->cnt - 1].ret_addr == ret_addr) { >> + ts->cnt -= 1; >> + } else { >> + size_t i = ts->cnt - 1; >> + >> + while (i--) { >> + if (ts->stack[i].ret_addr == ret_addr) { >> + ts->cnt = i; >> + return; >> + } >> + } >> + } >> +} >> + >> +void thread_stack__event(struct thread *thread, u32 flags, u64 from_ip, >> + u64 to_ip, u16 insn_len, u64 trace_nr) >> +{ >> + if (!thread) >> + return; >> + >> + if (!thread->ts) { >> + thread->ts = thread_stack__new(); >> + if (!thread->ts) >> + return; >> + thread->ts->trace_nr = trace_nr; >> + } >> + >> + if (trace_nr != thread->ts->trace_nr) { >> + thread->ts->trace_nr = trace_nr; >> + thread->ts->cnt = 0; >> + } >> + >> + if (flags & PERF_FLAG_CALL) { >> + u64 ret_addr; >> + >> + if (!to_ip) >> + return; >> + ret_addr = from_ip + insn_len; >> + if (ret_addr == to_ip) >> + return; /* Zero-length calls are excluded */ >> + thread_stack__push(thread->ts, ret_addr); >> + } else if (flags & PERF_FLAG_RETURN) { >> + if (!from_ip) >> + return; >> + thread_stack__pop(thread->ts, to_ip); >> + } >> +} >> + >> +void thread_stack__free(struct thread *thread) >> +{ >> + if (thread->ts) { >> + zfree(&thread->ts->stack); >> + zfree(&thread->ts); >> + } >> +} >> + >> +void thread_stack__sample(struct thread *thread, struct ip_callchain *chain, >> + size_t sz, u64 ip) >> +{ >> + size_t i; >> + >> + if (!thread || !thread->ts) >> + chain->nr = 1; >> + else >> + chain->nr = min(sz, thread->ts->cnt + 1); >> + >> + chain->ips[0] = ip; >> + >> + for (i = 1; i < chain->nr; i++) >> + chain->ips[i] = thread->ts->stack[thread->ts->cnt - i].ret_addr; >> +} >> diff --git a/tools/perf/util/thread-stack.h b/tools/perf/util/thread-stack.h >> new file mode 100644 >> index 0000000..c0ba4cf >> --- /dev/null >> +++ b/tools/perf/util/thread-stack.h >> @@ -0,0 +1,32 @@ >> +/* >> + * thread-stack.h: Synthesize a thread's stack using call / return events >> + * Copyright (c) 2014, Intel Corporation. >> + * >> + * This program is free software; you can redistribute it and/or modify it >> + * under the terms and conditions of the GNU General Public License, >> + * version 2, as published by the Free Software Foundation. >> + * >> + * This program is distributed in the hope it will be useful, but WITHOUT >> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or >> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for >> + * more details. >> + * >> + */ >> + >> +#ifndef __PERF_THREAD_STACK_H >> +#define __PERF_THREAD_STACK_H >> + >> +#include <sys/types.h> >> + >> +#include <linux/types.h> >> + >> +struct thread; >> +struct ip_callchain; >> + >> +void thread_stack__event(struct thread *thread, u32 flags, u64 from_ip, >> + u64 to_ip, u16 insn_len, u64 trace_nr); >> +void thread_stack__sample(struct thread *thread, struct ip_callchain *chain, >> + size_t sz, u64 ip); >> +void thread_stack__free(struct thread *thread); >> + >> +#endif >> diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c >> index 2b7b2d9..4e3418d 100644 >> --- a/tools/perf/util/thread.c >> +++ b/tools/perf/util/thread.c >> @@ -4,6 +4,7 @@ >> #include <string.h> >> #include "session.h" >> #include "thread.h" >> +#include "thread-stack.h" >> #include "util.h" >> #include "debug.h" >> #include "comm.h" >> @@ -66,6 +67,8 @@ void thread__delete(struct thread *thread) >> { >> struct comm *comm, *tmp; >> >> + thread_stack__free(thread); >> + >> if (thread->mg) { >> map_groups__put(thread->mg); >> thread->mg = NULL; >> diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h >> index 8c75fa7..a057820 100644 >> --- a/tools/perf/util/thread.h >> +++ b/tools/perf/util/thread.h >> @@ -8,6 +8,8 @@ >> #include "symbol.h" >> #include <strlist.h> >> >> +struct thread_stack; >> + >> struct thread { >> union { >> struct rb_node rb_node; >> @@ -25,6 +27,7 @@ struct thread { >> int comm_len; >> >> void *priv; >> + struct thread_stack *ts; >> }; >> >> struct machine; >> -- >> 1.9.1 > >
| |