From 35b9d88ecd8c5fb720ba0dd325262f356d0b03e7 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 9 Nov 2011 08:47:15 -0200 Subject: [PATCH] perf evlist: Introduce {prepare,start}_workload refactored from 'perf record' So that we can easily start a workload in other tools. Cc: David Ahern Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-zdsksd4aphu0nltg2lpwsw3x@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 81 ++++++-------------------------------- tools/perf/perf.h | 1 + tools/perf/util/evlist.c | 96 +++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 10 +++++ 4 files changed, 120 insertions(+), 68 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index c3ac541..4799195 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -47,11 +47,9 @@ static struct perf_record_opts record_opts = { static unsigned int page_size; static unsigned int mmap_pages = UINT_MAX; static int output; -static int pipe_output = 0; static const char *output_name = NULL; static bool group = false; static int realtime_prio = 0; -static pid_t child_pid = -1; static enum write_mode_t write_mode = WRITE_FORCE; static bool no_buildid = false; static bool no_buildid_cache = false; @@ -144,9 +142,9 @@ static void sig_atexit(void) { int status; - if (child_pid > 0) { + if (evsel_list->workload.pid > 0) { if (!child_finished) - kill(child_pid, SIGTERM); + kill(evsel_list->workload.pid, SIGTERM); wait(&status); if (WIFSIGNALED(status)) @@ -304,7 +302,7 @@ static int process_buildids(void) static void atexit_header(void) { - if (!pipe_output) { + if (!record_opts.pipe_output) { session->header.data_size += bytes_written; if (!no_buildid) @@ -377,9 +375,7 @@ static int __cmd_record(int argc, const char **argv) int flags; int err; unsigned long waking = 0; - int child_ready_pipe[2], go_pipe[2]; const bool forks = argc > 0; - char buf; struct machine *machine; progname = argv[0]; @@ -391,20 +387,15 @@ static int __cmd_record(int argc, const char **argv) signal(SIGINT, sig_handler); signal(SIGUSR1, sig_handler); - if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) { - perror("failed to create pipes"); - exit(-1); - } - if (!output_name) { if (!fstat(STDOUT_FILENO, &st) && S_ISFIFO(st.st_mode)) - pipe_output = true; + record_opts.pipe_output = true; else output_name = "perf.data"; } if (output_name) { if (!strcmp(output_name, "-")) - pipe_output = true; + record_opts.pipe_output = true; else if (!stat(output_name, &st) && st.st_size) { if (write_mode == WRITE_FORCE) { char oldname[PATH_MAX]; @@ -424,7 +415,7 @@ static int __cmd_record(int argc, const char **argv) else flags |= O_TRUNC; - if (pipe_output) + if (record_opts.pipe_output) output = STDOUT_FILENO; else output = open(output_name, flags, S_IRUSR | S_IWUSR); @@ -470,57 +461,11 @@ static int __cmd_record(int argc, const char **argv) mmap_pages = (512 * 1024) / page_size; if (forks) { - child_pid = fork(); - if (child_pid < 0) { - perror("failed to fork"); - exit(-1); - } - - if (!child_pid) { - if (pipe_output) - dup2(2, 1); - close(child_ready_pipe[0]); - close(go_pipe[1]); - fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC); - - /* - * Do a dummy execvp to get the PLT entry resolved, - * so we avoid the resolver overhead on the real - * execvp call. - */ - execvp("", (char **)argv); - - /* - * Tell the parent we're ready to go - */ - close(child_ready_pipe[1]); - - /* - * Wait until the parent tells us to go. - */ - if (read(go_pipe[0], &buf, 1) == -1) - perror("unable to read pipe"); - - execvp(argv[0], (char **)argv); - - perror(argv[0]); - kill(getppid(), SIGUSR1); - exit(-1); - } - - if (!record_opts.system_wide && record_opts.target_tid == -1 && record_opts.target_pid == -1) - evsel_list->threads->map[0] = child_pid; - - close(child_ready_pipe[1]); - close(go_pipe[0]); - /* - * wait for child to settle - */ - if (read(child_ready_pipe[0], &buf, 1) == -1) { - perror("unable to read pipe"); - exit(-1); + err = perf_evlist__prepare_workload(evsel_list, &record_opts, argv); + if (err < 0) { + pr_err("Couldn't run the workload!\n"); + goto out_delete_session; } - close(child_ready_pipe[0]); } open_counters(evsel_list); @@ -530,7 +475,7 @@ static int __cmd_record(int argc, const char **argv) */ atexit(atexit_header); - if (pipe_output) { + if (record_opts.pipe_output) { err = perf_header__write_pipe(output); if (err < 0) return err; @@ -543,7 +488,7 @@ static int __cmd_record(int argc, const char **argv) post_processing_offset = lseek(output, 0, SEEK_CUR); - if (pipe_output) { + if (record_opts.pipe_output) { err = perf_session__synthesize_attrs(session, process_synthesized_event); if (err < 0) { @@ -629,7 +574,7 @@ static int __cmd_record(int argc, const char **argv) * Let the child rip */ if (forks) - close(go_pipe[1]); + perf_evlist__start_workload(evsel_list); for (;;) { int hits = samples; diff --git a/tools/perf/perf.h b/tools/perf/perf.h index eb6a138..32ee6ca 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -193,6 +193,7 @@ struct perf_record_opts { bool no_delay; bool no_inherit; bool no_samples; + bool pipe_output; bool raw_samples; bool sample_address; bool sample_time; diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index b774341..a472247 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -13,6 +13,7 @@ #include "thread_map.h" #include "evlist.h" #include "evsel.h" +#include #include "parse-events.h" @@ -33,6 +34,7 @@ void perf_evlist__init(struct perf_evlist *evlist, struct cpu_map *cpus, INIT_HLIST_HEAD(&evlist->heads[i]); INIT_LIST_HEAD(&evlist->entries); perf_evlist__set_maps(evlist, cpus, threads); + evlist->workload.pid = -1; } struct perf_evlist *perf_evlist__new(struct cpu_map *cpus, @@ -674,3 +676,97 @@ out_err: return err; } + +int perf_evlist__prepare_workload(struct perf_evlist *evlist, + struct perf_record_opts *opts, + const char *argv[]) +{ + int child_ready_pipe[2], go_pipe[2]; + char bf; + + if (pipe(child_ready_pipe) < 0) { + perror("failed to create 'ready' pipe"); + return -1; + } + + if (pipe(go_pipe) < 0) { + perror("failed to create 'go' pipe"); + goto out_close_ready_pipe; + } + + evlist->workload.pid = fork(); + if (evlist->workload.pid < 0) { + perror("failed to fork"); + goto out_close_pipes; + } + + if (!evlist->workload.pid) { + if (opts->pipe_output) + dup2(2, 1); + + close(child_ready_pipe[0]); + close(go_pipe[1]); + fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC); + + /* + * Do a dummy execvp to get the PLT entry resolved, + * so we avoid the resolver overhead on the real + * execvp call. + */ + execvp("", (char **)argv); + + /* + * Tell the parent we're ready to go + */ + close(child_ready_pipe[1]); + + /* + * Wait until the parent tells us to go. + */ + if (read(go_pipe[0], &bf, 1) == -1) + perror("unable to read pipe"); + + execvp(argv[0], (char **)argv); + + perror(argv[0]); + kill(getppid(), SIGUSR1); + exit(-1); + } + + if (!opts->system_wide && opts->target_tid == -1 && opts->target_pid == -1) + evlist->threads->map[0] = evlist->workload.pid; + + close(child_ready_pipe[1]); + close(go_pipe[0]); + /* + * wait for child to settle + */ + if (read(child_ready_pipe[0], &bf, 1) == -1) { + perror("unable to read pipe"); + goto out_close_pipes; + } + + evlist->workload.cork_fd = go_pipe[1]; + close(child_ready_pipe[0]); + return 0; + +out_close_pipes: + close(go_pipe[0]); + close(go_pipe[1]); +out_close_ready_pipe: + close(child_ready_pipe[0]); + close(child_ready_pipe[1]); + return -1; +} + +int perf_evlist__start_workload(struct perf_evlist *evlist) +{ + if (evlist->workload.cork_fd > 0) { + /* + * Remove the cork, let it rip! + */ + return close(evlist->workload.cork_fd); + } + + return 0; +} diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 231c06f..07d56b3 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -6,6 +6,7 @@ #include "../perf.h" #include "event.h" #include "util.h" +#include struct pollfd; struct thread_map; @@ -22,6 +23,10 @@ struct perf_evlist { int nr_fds; int nr_mmaps; int mmap_len; + struct { + int cork_fd; + pid_t pid; + } workload; bool overwrite; union perf_event event_copy; struct perf_mmap *mmap; @@ -68,6 +73,11 @@ int perf_evlist__open(struct perf_evlist *evlist, bool group); void perf_evlist__config_attrs(struct perf_evlist *evlist, struct perf_record_opts *opts); +int perf_evlist__prepare_workload(struct perf_evlist *evlist, + struct perf_record_opts *opts, + const char *argv[]); +int perf_evlist__start_workload(struct perf_evlist *evlist); + int perf_evlist__alloc_mmap(struct perf_evlist *evlist); int perf_evlist__mmap(struct perf_evlist *evlist, int pages, bool overwrite); void perf_evlist__munmap(struct perf_evlist *evlist); -- 2.7.4