From 3415d8b851307c75a1e8aa16030db9172306df78 Mon Sep 17 00:00:00 2001 From: Stanislav Fomichev Date: Mon, 20 Jan 2014 15:39:38 +0400 Subject: perf timechart: Fix wrong SVG height If we call perf timechart with -p 0 arguments, it means we don't want any tasks related data. It works, but space for tasks data is reserved in the generated SVG. Remove this unused empty space via passing 0 as count to the open_svg. Signed-off-by: Stanislav Fomichev Cc: Adrian Hunter Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1390217980-22424-2-git-send-email-stfomichev@yandex-team.ru Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 652af0b..25526d6 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -1045,6 +1045,9 @@ static void write_svg_file(struct timechart *tchart, const char *filename) thresh /= 10; } while (!process_filter && thresh && count < tchart->proc_num); + if (!tchart->proc_num) + count = 0; + open_svg(filename, tchart->numcpus, count, tchart->first_time, tchart->last_time); svg_time_grid(); -- cgit v0.10.2 From 8bac41cbfe2efe55e2b93673b84761ed7dd75f69 Mon Sep 17 00:00:00 2001 From: Stanislav Fomichev Date: Mon, 20 Jan 2014 15:39:39 +0400 Subject: perf session: Free cpu_map in perf_session__cpu_bitmap This method uses a temporary struct cpu_map to figure out the cpus present in the received cpu list in string form, but it failed to free it after returning. Fix it. Signed-off-by: Stanislav Fomichev Cc: Adrian Hunter Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1390217980-22424-3-git-send-email-stfomichev@yandex-team.ru [ Use goto + err = -1 to do the delete just once, in the normal exit path ] Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 7acc03e..0b39a48 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -1573,7 +1573,7 @@ next: int perf_session__cpu_bitmap(struct perf_session *session, const char *cpu_list, unsigned long *cpu_bitmap) { - int i; + int i, err = -1; struct cpu_map *map; for (i = 0; i < PERF_TYPE_MAX; ++i) { @@ -1602,13 +1602,17 @@ int perf_session__cpu_bitmap(struct perf_session *session, if (cpu >= MAX_NR_CPUS) { pr_err("Requested CPU %d too large. " "Consider raising MAX_NR_CPUS\n", cpu); - return -1; + goto out_delete_map; } set_bit(cpu, cpu_bitmap); } - return 0; + err = 0; + +out_delete_map: + cpu_map__delete(map); + return err; } void perf_session__fprintf_info(struct perf_session *session, FILE *fp, -- cgit v0.10.2 From 3a46817f674389d5fb2f4dce19fbc73ebd1976bb Mon Sep 17 00:00:00 2001 From: Baruch Siach Date: Mon, 13 Jan 2014 12:27:35 +0200 Subject: perf tools: Add support for the xtensa architecture Tested using kernel tracepoints on a QEMU simulated environment. Kernel support for perf depends on the patch "xtensa: enable HAVE_PERF_EVENTS", which is scheduled for v3.14. Hardware performance counters are not supported under xtensa yet. Acked-by: Ingo Molnar Acked-by: Max Filippov Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: linux-xtensa@linux-xtensa.org Link: http://lkml.kernel.org/r/aafcdb22f04e2d3188d2938528939481be56b649.1389608855.git.baruch@tkos.co.il Signed-off-by: Baruch Siach Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/perf.h b/tools/perf/perf.h index 3c2f213..7daa806 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -132,6 +132,13 @@ #define CPUINFO_PROC "CPU" #endif +#ifdef __xtensa__ +#define mb() asm volatile("memw" ::: "memory") +#define wmb() asm volatile("memw" ::: "memory") +#define rmb() asm volatile("" ::: "memory") +#define CPUINFO_PROC "core ID" +#endif + #define barrier() asm volatile ("" ::: "memory") #ifndef cpu_relax -- cgit v0.10.2 From 8a398897ff21f73cb8b15a19514660f032926882 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Fri, 17 Jan 2014 16:34:05 +0100 Subject: perf stat: fix NULL pointer reference bug with event unit This patch fixes a problem with the handling of the newly introduced optional event unit. The following cmdline caused a segfault: $ perf stat -e cpu/event-0x3c/ ls This patch fixes the problem with the default setting for alias->unit which was eventually causing the segfault. Signed-off-by: Stephane Eranian Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1389972846-6566-2-git-send-email-eranian@google.com Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index a7f1b6a..d248fca 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -635,7 +635,7 @@ int parse_events_add_pmu(struct list_head *list, int *idx, struct perf_event_attr attr; struct perf_pmu *pmu; struct perf_evsel *evsel; - char *unit; + const char *unit; double scale; pmu = perf_pmu__find(name); diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c index d9cab4d..b752ecb 100644 --- a/tools/perf/util/pmu.c +++ b/tools/perf/util/pmu.c @@ -105,7 +105,7 @@ static int perf_pmu__parse_scale(struct perf_pmu_alias *alias, char *dir, char * char scale[128]; int fd, ret = -1; char path[PATH_MAX]; - char *lc; + const char *lc; snprintf(path, PATH_MAX, "%s/%s.scale", dir, name); @@ -609,7 +609,7 @@ static struct perf_pmu_alias *pmu_find_alias(struct perf_pmu *pmu, static int check_unit_scale(struct perf_pmu_alias *alias, - char **unit, double *scale) + const char **unit, double *scale) { /* * Only one term in event definition can @@ -634,14 +634,18 @@ static int check_unit_scale(struct perf_pmu_alias *alias, * defined for the alias */ int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms, - char **unit, double *scale) + const char **unit, double *scale) { struct parse_events_term *term, *h; struct perf_pmu_alias *alias; int ret; + /* + * Mark unit and scale as not set + * (different from default values, see below) + */ *unit = NULL; - *scale = 0; + *scale = 0.0; list_for_each_entry_safe(term, h, head_terms, list) { alias = pmu_find_alias(pmu, term); @@ -658,6 +662,18 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms, list_del(&term->list); free(term); } + + /* + * if no unit or scale foundin aliases, then + * set defaults as for evsel + * unit cannot left to NULL + */ + if (*unit == NULL) + *unit = ""; + + if (*scale == 0.0) + *scale = 1.0; + return 0; } diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h index 9183380..8b64125 100644 --- a/tools/perf/util/pmu.h +++ b/tools/perf/util/pmu.h @@ -29,7 +29,7 @@ int perf_pmu__config_terms(struct list_head *formats, struct perf_event_attr *attr, struct list_head *head_terms); int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms, - char **unit, double *scale); + const char **unit, double *scale); struct list_head *perf_pmu__alias(struct perf_pmu *pmu, struct list_head *head_terms); int perf_pmu_wrap(void); -- cgit v0.10.2 From a761a2d8a7175b7b4e8525e0672e1a8d3c051001 Mon Sep 17 00:00:00 2001 From: Alan Cox Date: Mon, 20 Jan 2014 19:10:11 +0100 Subject: perf tools: Ensure sscanf does not overrun the "mem" field Make the parsing robust. (perf has some other assumptions that BUFSIZE <= MAX_PATH which are not touched here) Reported-by: Jackie Chang Signed-off-by: Alan Cox Cc: Alan Cox Cc: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-g2uoiwbrpiimb63rx32qv8ne@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index bb3e0ed..893f8e2 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -930,7 +930,7 @@ static int write_topo_node(int fd, int node) /* skip over invalid lines */ if (!strchr(buf, ':')) continue; - if (sscanf(buf, "%*s %*d %s %"PRIu64, field, &mem) != 2) + if (sscanf(buf, "%*s %*d %31s %"PRIu64, field, &mem) != 2) goto done; if (!strcmp(field, "MemTotal:")) mem_total = mem; -- cgit v0.10.2 From de256a4e6b9096070a5305950c7d693395150680 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Mon, 20 Jan 2014 16:15:13 -0300 Subject: perf evsel: Remove duplicate member zeroing after free No need to set evsel->fd to NULL after calling perf_evsel__free_fd(), as this method already does that. Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-wu6kul8fpapr8iyqm685ewtf@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 22e18a2..55407c5 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -1081,7 +1081,6 @@ void perf_evsel__close(struct perf_evsel *evsel, int ncpus, int nthreads) perf_evsel__close_fd(evsel, ncpus, nthreads); perf_evsel__free_fd(evsel); - evsel->fd = NULL; } static struct { -- cgit v0.10.2 From 8ad9219e08af12a5652892e273336dbd31b25b03 Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Fri, 17 Jan 2014 16:34:06 +0100 Subject: perf stat: Fix memory corruption of xyarray when cpumask is used This patch fixes a memory corruption problem with the xyarray when the evsel fds get closed at the end of the run_perf_stat() call. It could be triggered with: # perf stat -a -e power/energy-cores/ ls When cpumask are used by events (.e.g, RAPL or uncores) then the evsel fds are allocated based on the actual number of CPUs monitored. That number can be smaller than the total number of CPUs on the system. The problem arises at the end by perf stat closes the fds twice. When fds are closed, their entry in the xyarray are set to -1. The first close() on the evsel is made from __run_perf_stat() and it uses the actual number of CPUS for the event which is how the xyarray was allocated for. The second is from perf_evlist_close() but that one is on the total number of CPUs in the system, so it assume the xyarray was allocated to cover it. However it was not, and some writes corrupt memory. The fix is in perf_evlist_close() is to first try with the evsel->cpus if present, if not use the evlist->cpus. That fixes the problem. Signed-off-by: Stephane Eranian Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1389972846-6566-3-git-send-email-eranian@google.com Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 40bd2c0..59ef280 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -1003,9 +1003,12 @@ void perf_evlist__close(struct perf_evlist *evlist) struct perf_evsel *evsel; int ncpus = cpu_map__nr(evlist->cpus); int nthreads = thread_map__nr(evlist->threads); + int n; - evlist__for_each_reverse(evlist, evsel) - perf_evsel__close(evsel, ncpus, nthreads); + evlist__for_each_reverse(evlist, evsel) { + n = evsel->cpus ? evsel->cpus->nr : ncpus; + perf_evsel__close(evsel, n, nthreads); + } } int perf_evlist__open(struct perf_evlist *evlist) -- cgit v0.10.2 From 578c03c86fadcc6fd7319ddf41dd4d1d88aab77a Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Thu, 16 Jan 2014 10:49:31 +0900 Subject: perf symbols: Fix JIT symbol resolution on heap Gaurav reported that perf cannot profile JIT program if it executes the code on heap. This was because current map__new() only handle JIT on anon mappings - extends it to handle no_dso (heap, stack) case too. This patch assumes JIT profiling only provides dynamic function symbols so check the mapping type to distinguish the case. It'd provide no symbols for data mapping - if we need to support symbols on data mappings later it should be changed. Reported-by: Gaurav Jain Signed-off-by: Namhyung Kim Tested-by: Gaurav Jain Cc: Andi Kleen Cc: David Ahern Cc: Gaurav Jain Cc: Ingo Molnar Cc: Jiri Olsa Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1389836971-3549-1-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 9b9bd71..ee1dd68 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -69,7 +69,7 @@ struct map *map__new(struct list_head *dsos__list, u64 start, u64 len, map->ino = ino; map->ino_generation = ino_gen; - if (anon) { + if ((anon || no_dso) && type == MAP__FUNCTION) { snprintf(newfilename, sizeof(newfilename), "/tmp/perf-%d.map", pid); filename = newfilename; } @@ -93,7 +93,7 @@ struct map *map__new(struct list_head *dsos__list, u64 start, u64 len, * functions still return NULL, and we avoid the * unnecessary map__load warning. */ - if (no_dso) + if (type != MAP__FUNCTION) dso__set_loaded(dso, map->type); } } -- cgit v0.10.2 From b935a58dbff457c27fd63e1e1bb29db20b2ee6a8 Mon Sep 17 00:00:00 2001 From: Josh Boyer Date: Wed, 22 Jan 2014 10:01:48 -0500 Subject: perf tools: Fix traceevent plugin path definitions The plugindir_SQ definition contains $(prefix) which is not needed as the $(libdir) definition already contains prefix in it. This leads to the path including an extra prefix in it, e.g. /usr/usr/lib64. The -DPLUGIN_DIR defintion includes DESTDIR. This is incorrect, as it sets the plugin search path to include the value of DESTDIR. DESTDIR is a mechanism to install in a non-standard location such as a chroot or an RPM build root. In the RPM case, this leads to the search path being incorrect after the resulting RPM is installed (or in some cases an RPM build failure). Remove both of these unnecessary inclusions. Signed-off-by: Josh Boyer Acked-by: Jiri Olsa Cc: Ingo Molnar Cc: Jiri Olsa Cc: Steven Rostedt Link: http://lkml.kernel.org/r/20140122150147.GK16455@hansolo.jdub.homelinux.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/lib/traceevent/Makefile b/tools/lib/traceevent/Makefile index 56d52a3..005c9cc 100644 --- a/tools/lib/traceevent/Makefile +++ b/tools/lib/traceevent/Makefile @@ -63,7 +63,7 @@ endif endif ifeq ($(set_plugin_dir),1) -PLUGIN_DIR = -DPLUGIN_DIR="$(DESTDIR)/$(plugin_dir)" +PLUGIN_DIR = -DPLUGIN_DIR="$(plugin_dir)" PLUGIN_DIR_SQ = '$(subst ','\'',$(PLUGIN_DIR))' endif diff --git a/tools/perf/config/Makefile b/tools/perf/config/Makefile index d604e50..c48d449 100644 --- a/tools/perf/config/Makefile +++ b/tools/perf/config/Makefile @@ -600,5 +600,5 @@ perfexec_instdir_SQ = $(subst ','\'',$(perfexec_instdir)) # Otherwise we install plugins into the global $(libdir). ifdef DESTDIR plugindir=$(libdir)/traceevent/plugins -plugindir_SQ= $(subst ','\'',$(prefix)/$(plugindir)) +plugindir_SQ= $(subst ','\'',$(plugindir)) endif -- cgit v0.10.2 From 4afc81cd1caa93daa50c1c29a3ab747c978abc13 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 23 Jan 2014 02:29:50 +0000 Subject: perf symbols: Load map before using map->map_ip() In map_groups__find_symbol() map->map_ip is used without ensuring the map is loaded. Then the address passed to map->map_ip isn't mapped at the first time. E.g. below code always fails to get a symbol at the first call; addr = /* Somewhere in the kernel text */ symbol_conf.try_vmlinux_path = true; symbol__init(); host_machine = machine__new_host(); sym = machine__find_kernel_function(host_machine, addr, NULL, NULL); /* Note that machine__find_kernel_function calls map_groups__find_symbol */ This ensures it by calling map__load before using it in map_groups__find_symbol(). Signed-off-by: Masami Hiramatsu Cc: "David A. Long" Cc: David Ahern Cc: Ingo Molnar Cc: Namhyung Kim Cc: Oleg Nesterov Cc: Srikar Dronamraju Cc: "Steven Rostedt (Red Hat)" Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20140123022950.7206.17357.stgit@kbuild-fedora.yrl.intra.hitachi.co.jp Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index ee1dd68..3b97513 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -386,7 +386,8 @@ struct symbol *map_groups__find_symbol(struct map_groups *mg, { struct map *map = map_groups__find(mg, type, addr); - if (map != NULL) { + /* Ensure map is loaded before using map->map_ip */ + if (map != NULL && map__load(map, filter) >= 0) { if (mapp != NULL) *mapp = map; return map__find_symbol(map, map->map_ip(map, addr), filter); -- cgit v0.10.2