perf/core improvements and fixes
Arnaldo Carvalho de Melo
maintainerHi Ingo,
Please consider pulling,
- Arnaldo
The following changes since commit 9059b284caecb628fac826c2c5cc8ee85708eec1:
Merge tag 'perf-core-for-mingo-2' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core (2015-09-15 08:50:59 +0200)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-core-for-mingo
for you to fetch changes up to bbbe6bf6037d77816c4a19aaf35f4cecf662b49a:
perf tools: Introduce regs_query_register_offset() for x86 (2015-09-15 09:48:33 -0300)
----------------------------------------------------------------
perf/core improvements and fixes:
User visible:
- Enhance the error reporting of tracepoint event parsing, e.g.:
$ oldperf record -e sched:sched_switc usleep 1
event syntax error: 'sched:sched_switc'
\___ unknown tracepoint
Run 'perf list' for a list of valid eventsNow we get the much nicer:
$ perf record -e sched:sched_switc ls
event syntax error: 'sched:sched_switc'
\___ can't access trace events Error: No permissions to read /sys/kernel/debug/tracing/events/sched/sched_switc
Hint: Try 'sudo mount -o remount,mode=755 /sys/kernel/debug'
And after we have those mount point permissions fixed:
$ perf record -e sched:sched_switc ls
event syntax error: 'sched:sched_switc'
\___ unknown tracepoint Error: File /sys/kernel/debug/tracing/events/sched/sched_switc not found.
Hint: Perhaps this kernel misses some CONFIG_ setting to enable this feature?.
Now its just a matter of using what git uses to suggest alternatives when we make a typo, i.e. that it is just an 'h' missing :-)
I.e. basically now the event parsing routing uses the strerror_open() routines introduced by and used in 'perf trace' work. (Jiri Olsa)
Infrastructure:
- Export init/exit_probe_symbol_maps() from 'perf probe' for use in eBPF (Namhyung Kim)
- Free perf_probe_event in cleanup_perf_probe_events() (Namhyung Kim)
- regs_query_register_offset() infrastructure + implementation for x86. First user will be the perf/eBPF code (Wang Nan)
----------------------------------------------------------------
Jiri Olsa (4):
tools: Add err.h with ERR_PTR PTR_ERR interface perf tools: Propagate error info for the tracepoint parsing perf evsel: Propagate error info from tp_format perf tools: Enhance parsing events tracepoint error output
Namhyung Kim (2):
perf probe: Free perf_probe_event in cleanup_perf_probe_events() perf probe: Export init/exit_probe_symbol_maps()
Wang Nan (2):
perf tools: regs_query_register_offset() infrastructure perf tools: Introduce regs_query_register_offset() for x86
tools/include/linux/err.h | 49 +++++++++++ tools/perf/arch/x86/Makefile | 1 + tools/perf/arch/x86/util/dwarf-regs.c | 122 ++++++++++++++++++++-------- tools/perf/builtin-probe.c | 5 ++ tools/perf/builtin-trace.c | 19 +++-- tools/perf/config/Makefile | 4 + tools/perf/tests/evsel-tp-sched.c | 10 ++- tools/perf/tests/mmap-basic.c | 3 +- tools/perf/tests/openat-syscall-all-cpus.c | 3 +- tools/perf/tests/openat-syscall-tp-fields.c | 3 +- tools/perf/tests/openat-syscall.c | 3 +- tools/perf/util/evlist.c | 3 +- tools/perf/util/evsel.c | 16 +++- tools/perf/util/evsel.h | 3 + tools/perf/util/include/dwarf-regs.h | 8 ++ tools/perf/util/parse-events.c | 66 +++++++++++---- tools/perf/util/parse-events.h | 3 +- tools/perf/util/parse-events.y | 16 ++-- tools/perf/util/probe-event.c | 32 ++++---- tools/perf/util/probe-event.h | 2 + tools/perf/util/trace-event.c | 15 +++- 21 files changed, 291 insertions(+), 95 deletions(-) create mode 100644 tools/include/linux/err.h
Arnaldo Carvalho de Melo
maintainerPATCH 1/8 perf probe: Free perf_probe_event in cleanup_perf_probe_events()
From: Namhyung Kim <[email protected]>
The cleanup_perf_probe_events() frees all resources related to a perf probe event. However it only freed resources in trace probe events, not perf probe events. So call clear_perf_probe_event() too.
--- tools/perf/util/probe-event.c | 1 + 1 file changed, 1 insertion(+)
tools/perf/util/probe-event.c +1 -0
@@ -2792,6 +2792,7 @@ void cleanup_perf_probe_events(struct perf_probe_event *pevs, int npevs)clear_probe_trace_event (&pevs[i].tevs[j]);zfree (&pevs[i].tevs); pevs[i].ntevs =0 ; +clear_perf_probe_event (&pevs[i]); }exit_symbol_maps ();
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 2/8 perf probe: Export init/exit_probe_symbol_maps()
In reply to Arnaldo Carvalho de MeloFrom: Namhyung Kim <[email protected]>
The init/exit_symbols_maps() functions are to setup and cleanup necessary info for probe events. But they need to be called from out of the probe code now, so this patch exports them.
However the names are too generic, so change them to have 'probe'. :)
--- tools/perf/builtin-probe.c | 5 +++++ tools/perf/util/probe-event.c | 31 +++++++++++++++---------------- tools/perf/util/probe-event.h | 2 ++ 3 files changed, 22 insertions(+), 16 deletions(-)
tools/perf/builtin-probe.c +5 -0
@@ -317,6 +317,10 @@ static int perf_add_probe_events(struct perf_probe_event *pevs, int npevs)int i, k;const char *event =NULL , *group =NULL ; + ret =init_probe_symbol_maps (pevs->uprobes); +if (ret <0 ) +return ret; + ret =convert_perf_probe_events (pevs, npevs);if (ret <0 ) goto out_cleanup;
@@ -354,6 +358,7 @@ static int perf_add_probe_events(struct perf_probe_event *pevs, int npevs) out_cleanup:cleanup_perf_probe_events (pevs, npevs); +exit_probe_symbol_maps ();return ret; }
tools/perf/util/probe-event.c +15 -16
@@ -71,7 +71,7 @@ static char *synthesize_perf_probe_point(struct perf_probe_point *pp);static struct machine *host_machine;/* Initialize symbol maps and path of vmlinux/modules */ -static int init_symbol_maps (bool user_only) +int init_probe_symbol_maps (bool user_only) {int ret;
@@ -101,7 +101,7 @@ out:return ret; } -static void exit_symbol_maps (void ) +void exit_probe_symbol_maps (void ) {if (host_machine) {machine__delete (host_machine);
@@ -859,11 +859,11 @@ int show_line_range(struct line_range *lr, const char *module, bool user) {int ret; - ret =init_symbol_maps (user); + ret =init_probe_symbol_maps (user);if (ret <0 )return ret; ret =__show_line_range (lr, module, user); -exit_symbol_maps (); +exit_probe_symbol_maps ();return ret; }
@@ -941,7 +941,7 @@ int show_available_vars(struct perf_probe_event *pevs, int npevs,int i, ret =0 ;struct debuginfo *dinfo; - ret =init_symbol_maps (pevs->uprobes); + ret =init_probe_symbol_maps (pevs->uprobes);if (ret <0 )return ret;
@@ -958,7 +958,7 @@ int show_available_vars(struct perf_probe_event *pevs, int npevs,debuginfo__delete (dinfo); out: -exit_symbol_maps (); +exit_probe_symbol_maps ();return ret; }
@@ -2262,7 +2262,7 @@ int show_perf_probe_events(struct strfilter *filter)setup_pager (); - ret =init_symbol_maps (false); + ret =init_probe_symbol_maps (false);if (ret <0 )return ret;
@@ -2278,7 +2278,7 @@ int show_perf_probe_events(struct strfilter *filter)close (kp_fd);if (up_fd >0 )close (up_fd); -exit_symbol_maps (); +exit_probe_symbol_maps ();return ret; }
@@ -2746,10 +2746,6 @@ int convert_perf_probe_events(struct perf_probe_event *pevs, int npevs) {int i, ret; - ret =init_symbol_maps (pevs->uprobes); -if (ret <0 ) -return ret; -/* Loop 1: convert all events */ for (i =0 ; i < npevs; i++) {/* Init kprobe blacklist if needed */
@@ -2794,20 +2790,23 @@ void cleanup_perf_probe_events(struct perf_probe_event *pevs, int npevs) pevs[i].ntevs =0 ;clear_perf_probe_event (&pevs[i]); } - -exit_symbol_maps (); }int add_perf_probe_events (struct perf_probe_event *pevs,int npevs) {int ret; + ret =init_probe_symbol_maps (pevs->uprobes); +if (ret <0 ) +return ret; + ret =convert_perf_probe_events (pevs, npevs);if (ret ==0 ) ret =apply_perf_probe_events (pevs, npevs);cleanup_perf_probe_events (pevs, npevs); +exit_probe_symbol_maps ();return ret; }
@@ -2867,7 +2866,7 @@ int show_available_funcs(const char *target, struct strfilter *_filter,struct map *map;int ret; - ret =init_symbol_maps (user); + ret =init_probe_symbol_maps (user);if (ret <0 )return ret;
@@ -2897,7 +2896,7 @@ end:if (user) {map__put (map); } -exit_symbol_maps (); +exit_probe_symbol_maps ();return ret; }
tools/perf/util/probe-event.h +2 -0
@@ -110,6 +110,8 @@ struct variable_list { };struct map ; +int init_probe_symbol_maps (bool user_only); +void exit_probe_symbol_maps (void );/* Command string to events */ extern int parse_perf_probe_command(const char *cmd,
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 3/8 tools: Add err.h with ERR_PTR PTR_ERR interface
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Adding part of the kernel's <linux/err.h> interface:
inline void * __must_check ERR_PTR(long error); inline long __must_check PTR_ERR(__force const void *ptr); inline bool __must_check IS_ERR(__force const void *ptr);
It will be used to propagate error through pointers in following patches.
--- tools/include/linux/err.h | 49 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 49 insertions(+) create mode 100644 tools/include/linux/err.h
tools/include/linux/err.h new file +49 -0
@@ -0,0 +1,49 @@ +#ifndef __TOOLS_LINUX_ERR_H +#define __TOOLS_LINUX_ERR_H + +#include <linux/compiler.h> +#include <linux/types.h> + +#include <asm/errno.h> + +/* +* Original kernel header comment: +* +* Kernel pointers have redundant information, so we can use a +* scheme where we can return either an error code or a normal +* pointer with the same return value. +* +* This should be a per-architecture thing, to allow different +* error and pointer decisions. +* +* Userspace note: +* The same principle works for userspace, because 'error' pointers +* fall down to the unused hole far from user space, as described +* in Documentation/x86/x86_64/mm.txt for x86_64 arch: +* +* 0000000000000000 - 00007fffffffffff (=47 bits) user space, different per mm hole caused by [48:63] sign extension +* ffffffffffe00000 - ffffffffffffffff (=2 MB) unused hole +* +* It should be the same case for other architectures, because +* this code is used in generic kernel code. +*/ +#define MAX_ERRNO 4095 + +#define IS_ERR_VALUE (x) unlikely((x) >= (unsigned long)-MAX_ERRNO) + +static inline void *__must_check ERR_PTR (long error) +{ +return (void *) error; +} + +static inline long __must_check PTR_ERR (__force const void *ptr) +{ +return (long ) ptr; +} + +static inline bool __must_check IS_ERR (__force const void *ptr) +{ +return IS_ERR_VALUE ((unsigned long )ptr); +} + +#endif /* _LINUX_ERR_H */
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 4/8 perf tools: Propagate error info for the tracepoint parsing
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Pass 'struct parse_events_error *error' to the parse-event.c tracepoint adding path. It will be filled with error data in following patches.
--- tools/perf/util/parse-events.c | 27 ++++++++++++++++----------- tools/perf/util/parse-events.h | 3 ++- tools/perf/util/parse-events.y | 4 ++-- 3 files changed, 20 insertions(+), 14 deletions(-)
tools/perf/util/parse-events.c +16 -11
@@ -387,7 +387,8 @@ int parse_events_add_cache(struct list_head *list, int *idx, }static int add_tracepoint (struct list_head *list,int *idx, -char *sys_name,char *evt_name) +char *sys_name,char *evt_name, +struct parse_events_error *error__maybe_unused ) {struct perf_evsel *evsel;
@@ -401,7 +402,8 @@ static int add_tracepoint(struct list_head *list, int *idx, }static int add_tracepoint_multi_event (struct list_head *list,int *idx, -char *sys_name,char *evt_name) +char *sys_name,char *evt_name, +struct parse_events_error *error) {char evt_path[MAXPATHLEN ];struct dirent *evt_ent;
@@ -425,7 +427,7 @@ static int add_tracepoint_multi_event(struct list_head *list, int *idx,if (!strglobmatch (evt_ent->d_name, evt_name))continue ; - ret =add_tracepoint (list, idx, sys_name, evt_ent->d_name); + ret =add_tracepoint (list, idx, sys_name, evt_ent->d_name, error); }closedir (evt_dir);
@@ -433,15 +435,17 @@ static int add_tracepoint_multi_event(struct list_head *list, int *idx, }static int add_tracepoint_event (struct list_head *list,int *idx, -char *sys_name,char *evt_name) +char *sys_name,char *evt_name, +struct parse_events_error *error) {return strpbrk (evt_name,"*?" ) ? -add_tracepoint_multi_event (list, idx, sys_name, evt_name) : -add_tracepoint (list, idx, sys_name, evt_name); +add_tracepoint_multi_event (list, idx, sys_name, evt_name, error) : +add_tracepoint (list, idx, sys_name, evt_name, error); }static int add_tracepoint_multi_sys (struct list_head *list,int *idx, -char *sys_name,char *evt_name) +char *sys_name,char *evt_name, +struct parse_events_error *error) {struct dirent *events_ent;DIR *events_dir;
@@ -465,7 +469,7 @@ static int add_tracepoint_multi_sys(struct list_head *list, int *idx,continue ; ret =add_tracepoint_event (list, idx, events_ent->d_name, - evt_name); + evt_name, error); }closedir (events_dir);
@@ -473,12 +477,13 @@ static int add_tracepoint_multi_sys(struct list_head *list, int *idx, }int parse_events_add_tracepoint (struct list_head *list,int *idx, -char *sys,char *event) +char *sys,char *event, +struct parse_events_error *error) {if (strpbrk (sys,"*?" )) -return add_tracepoint_multi_sys (list, idx, sys, event); +return add_tracepoint_multi_sys (list, idx, sys, event, error);else -return add_tracepoint_event (list, idx, sys, event); +return add_tracepoint_event (list, idx, sys, event, error); } static int
tools/perf/util/parse-events.h +2 -1
@@ -118,7 +118,8 @@ int parse_events__modifier_event(struct list_head *list, char *str, bool add);int parse_events__modifier_group (struct list_head *list,char *event_mod);int parse_events_name (struct list_head *list,char *name);int parse_events_add_tracepoint (struct list_head *list,int *idx, -char *sys,char *event); +char *sys,char *event, +struct parse_events_error *error); int parse_events_add_numeric(struct parse_events_evlist *data, struct list_head *list, u32 type, u64 config,
tools/perf/util/parse-events.y +2 -2
@@ -376,7 +376,7 @@ PE_NAME '-' PE_NAME ':' PE_NAME snprintf(&sys_name, 128, "%s-%s", $1, $3); ALLOC_LIST(list); - ABORT_ON(parse_events_add_tracepoint(list, &data->idx, &sys_name, $5)); + ABORT_ON(parse_events_add_tracepoint(list, &data->idx, &sys_name, $5, data->error)); $$ = list; } |
@@ -386,7 +386,7 @@ PE_NAME ':' PE_NAME struct list_head *list; ALLOC_LIST(list); - if (parse_events_add_tracepoint(list, &data->idx, $1, $3)) { + if (parse_events_add_tracepoint(list, &data->idx, $1, $3, data->error)) { struct parse_events_error *error = data->error; if (error) {
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 5/8 perf evsel: Propagate error info from tp_format
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Propagate error info from tp_format via ERR_PTR to get it all the way down to the parse-event.c tracepoint adding routines. Following functions now return pointer with encoded error:
- tp_format
- trace_event__tp_format
- perf_evsel__newtp_idx
- perf_evsel__newtp
This affects several other places in perf, that cannot use pointer check anymore, but must utilize the err.h interface, when getting error information from above functions list.
--- tools/perf/builtin-trace.c | 19 +++++++++++-------- tools/perf/tests/evsel-tp-sched.c | 10 ++++++++-- tools/perf/tests/mmap-basic.c | 3 ++- tools/perf/tests/openat-syscall-all-cpus.c | 3 ++- tools/perf/tests/openat-syscall-tp-fields.c | 3 ++- tools/perf/tests/openat-syscall.c | 3 ++- tools/perf/util/evlist.c | 3 ++- tools/perf/util/evsel.c | 16 +++++++++++++--- tools/perf/util/evsel.h | 3 +++ tools/perf/util/parse-events.c | 6 +++--- tools/perf/util/trace-event.c | 15 ++++++++++++--- 11 files changed, 60 insertions(+), 24 deletions(-)
tools/perf/builtin-trace.c +11 -8
@@ -38,6 +38,7 @@#include <stdlib.h> #include <sys/mman.h> #include <linux/futex.h> +#include <linux/err.h> /* For older distros: */ #ifndef MAP_STACK
@@ -245,13 +246,14 @@ static struct perf_evsel *perf_evsel__syscall_newtp(const char *direction, voidstruct perf_evsel *evsel =perf_evsel__newtp ("raw_syscalls" , direction);/* older kernel (e.g., RHEL6) use syscalls:{enter,exit} */ -if (evsel ==NULL ) +if (IS_ERR (evsel)) evsel =perf_evsel__newtp ("syscalls" , direction); -if (evsel) { -if (perf_evsel__init_syscall_tp (evsel, handler)) - goto out_delete; - } +if (IS_ERR (evsel)) +return NULL ; + +if (perf_evsel__init_syscall_tp (evsel, handler)) + goto out_delete;return evsel;
@@ -1705,12 +1707,12 @@ static int trace__read_syscall_info(struct trace *trace, int id)snprintf (tp_name,sizeof (tp_name),"sys_enter_%s" , sc->name); sc->tp_format =trace_event__tp_format ("syscalls" , tp_name); -if (sc->tp_format ==NULL && sc->fmt && sc->fmt->alias) { +if (IS_ERR (sc->tp_format) && sc->fmt && sc->fmt->alias) {snprintf (tp_name,sizeof (tp_name),"sys_enter_%s" , sc->fmt->alias); sc->tp_format =trace_event__tp_format ("syscalls" , tp_name); } -if (sc->tp_format ==NULL ) +if (IS_ERR (sc->tp_format))return -1 ; sc->args = sc->tp_format->format.fields;
@@ -2390,7 +2392,8 @@ static size_t trace__fprintf_thread_summary(struct trace *trace, FILE *fp);static bool perf_evlist__add_vfs_getname (struct perf_evlist *evlist) {struct perf_evsel *evsel =perf_evsel__newtp ("probe" ,"vfs_getname" ); -if (evsel ==NULL ) + +if (IS_ERR (evsel))return false;if (perf_evsel__field (evsel,"pathname" ) ==NULL ) {
tools/perf/tests/evsel-tp-sched.c +8 -2
@@ -1,3 +1,4 @@ +#include <linux/err.h> #include <traceevent/event-parse.h> #include "evsel.h" #include "tests.h"
@@ -36,8 +37,8 @@ int test__perf_evsel__tp_sched_test(void)struct perf_evsel *evsel =perf_evsel__newtp ("sched" ,"sched_switch" );int ret =0 ; -if (evsel ==NULL ) { -pr_debug ("perf_evsel__new\n" ); +if (IS_ERR (evsel)) { +pr_debug ("perf_evsel__newtp failed with %ld\n" ,PTR_ERR (evsel));return -1 ; }
@@ -66,6 +67,11 @@ int test__perf_evsel__tp_sched_test(void) evsel =perf_evsel__newtp ("sched" ,"sched_wakeup" ); +if (IS_ERR (evsel)) { +pr_debug ("perf_evsel__newtp failed with %ld\n" ,PTR_ERR (evsel)); +return -1 ; + } +if (perf_evsel__test_field (evsel,"comm" ,16 , true)) ret =-1 ;
tools/perf/tests/mmap-basic.c +2 -1
@@ -3,6 +3,7 @@#include "thread_map.h" #include "cpumap.h" #include "tests.h" +#include <linux/err.h> /* * This test will generate random numbers of calls to some getpid syscalls,
@@ -65,7 +66,7 @@ int test__basic_mmap(void)snprintf (name,sizeof (name),"sys_enter_%s" , syscall_names[i]); evsels[i] =perf_evsel__newtp ("syscalls" , name); -if (evsels[i] ==NULL ) { +if (IS_ERR (evsels[i])) {pr_debug ("perf_evsel__new\n" ); goto out_delete_evlist; }
tools/perf/tests/openat-syscall-all-cpus.c +2 -1
@@ -1,4 +1,5 @@#include <api/fs/fs.h> +#include <linux/err.h> #include "evsel.h" #include "tests.h" #include "thread_map.h"
@@ -31,7 +32,7 @@ int test__openat_syscall_event_on_all_cpus(void)CPU_ZERO (&cpu_set); evsel =perf_evsel__newtp ("syscalls" ,"sys_enter_openat" ); -if (evsel ==NULL ) { +if (IS_ERR (evsel)) {tracing_path__strerror_open_tp (errno, errbuf,sizeof (errbuf),"syscalls" ,"sys_enter_openat" );pr_err ("%s\n" , errbuf); goto out_thread_map_delete;
tools/perf/tests/openat-syscall-tp-fields.c +2 -1
@@ -1,3 +1,4 @@ +#include <linux/err.h> #include "perf.h" #include "evlist.h" #include "evsel.h"
@@ -30,7 +31,7 @@ int test__syscall_openat_tp_fields(void) } evsel =perf_evsel__newtp ("syscalls" ,"sys_enter_openat" ); -if (evsel ==NULL ) { +if (IS_ERR (evsel)) {pr_debug ("%s: perf_evsel__newtp\n" , __func__); goto out_delete_evlist; }
tools/perf/tests/openat-syscall.c +2 -1
@@ -1,4 +1,5 @@#include <api/fs/tracing_path.h> +#include <linux/err.h> #include "thread_map.h" #include "evsel.h" #include "debug.h"
@@ -19,7 +20,7 @@ int test__openat_syscall_event(void) } evsel =perf_evsel__newtp ("syscalls" ,"sys_enter_openat" ); -if (evsel ==NULL ) { +if (IS_ERR (evsel)) {tracing_path__strerror_open_tp (errno, errbuf,sizeof (errbuf),"syscalls" ,"sys_enter_openat" );pr_err ("%s\n" , errbuf); goto out_thread_map_delete;
tools/perf/util/evlist.c +2 -1
@@ -25,6 +25,7 @@#include <linux/bitops.h> #include <linux/hash.h> #include <linux/log2.h> +#include <linux/err.h> static void perf_evlist__mmap_put (struct perf_evlist *evlist,int idx);static void __perf_evlist__munmap (struct perf_evlist *evlist,int idx);
@@ -265,7 +266,7 @@ int perf_evlist__add_newtp(struct perf_evlist *evlist, {struct perf_evsel *evsel =perf_evsel__newtp (sys, name); -if (evsel ==NULL ) +if (IS_ERR (evsel))return -1 ; evsel->handler = handler;
tools/perf/util/evsel.c +13 -3
@@ -13,6 +13,7 @@#include <traceevent/event-parse.h> #include <linux/hw_breakpoint.h> #include <linux/perf_event.h> +#include <linux/err.h> #include <sys/resource.h> #include "asm/bug.h" #include "callchain.h"
@@ -225,11 +226,17 @@ struct perf_evsel *perf_evsel__new_idx(struct perf_event_attr *attr, int idx)return evsel; } +/* +* Returns pointer with encoded error via <linux/err.h> interface. +*/ struct perf_evsel *perf_evsel__newtp_idx(const char *sys, const char *name, int idx) { struct perf_evsel *evsel = zalloc(perf_evsel__object.size); + int err = -ENOMEM; - if (evsel != NULL) { + if (evsel == NULL) { + goto out_err; + } else { struct perf_event_attr attr = { .type = PERF_TYPE_TRACEPOINT, .sample_type = (PERF_SAMPLE_RAW | PERF_SAMPLE_TIME |
@@ -240,8 +247,10 @@ struct perf_evsel *perf_evsel__newtp_idx(const char *sys, const char *name, int goto out_free; evsel->tp_format =trace_event__tp_format (sys, name); -if (evsel->tp_format ==NULL ) +if (IS_ERR (evsel->tp_format)) { + err =PTR_ERR (evsel->tp_format); goto out_free; + }event_attr_init (&attr); attr.config = evsel->tp_format->id;
@@ -254,7 +263,8 @@ struct perf_evsel *perf_evsel__newtp_idx(const char *sys, const char *name, int out_free:zfree (&evsel->name);free (evsel); -return NULL ; +out_err: +return ERR_PTR (err); }const char *perf_evsel__hw_names[PERF_COUNT_HW_MAX ] = {
tools/perf/util/evsel.h +3 -0
@@ -160,6 +160,9 @@ static inline struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr)struct perf_evsel *perf_evsel__newtp_idx (const char *sys,const char *name,int idx); +/* +* Returns pointer with encoded error via <linux/err.h> interface. +*/ static inline struct perf_evsel *perf_evsel__newtp (const char *sys,const char *name) {return perf_evsel__newtp_idx (sys, name,0 );
tools/perf/util/parse-events.c +3 -3
@@ -1,4 +1,5 @@#include <linux/hw_breakpoint.h> +#include <linux/err.h> #include "util.h" #include "../perf.h" #include "evlist.h"
@@ -393,11 +394,10 @@ static int add_tracepoint(struct list_head *list, int *idx,struct perf_evsel *evsel; evsel =perf_evsel__newtp_idx (sys_name, evt_name, (*idx)++); -if (!evsel) -return -ENOMEM ; +if (IS_ERR (evsel)) +return PTR_ERR (evsel);list_add_tail (&evsel->node, list); -return 0 ; }
tools/perf/util/trace-event.c +12 -3
@@ -7,6 +7,7 @@#include <sys/stat.h> #include <fcntl.h> #include <linux/kernel.h> +#include <linux/err.h> #include <traceevent/event-parse.h> #include <api/fs/tracing_path.h> #include "trace-event.h"
@@ -66,6 +67,9 @@ void trace_event__cleanup(struct trace_event *t)pevent_free (t->pevent); } +/* +* Returns pointer with encoded error via <linux/err.h> interface. +*/ static struct event_format *tp_format (const char *sys,const char *name) {
@@ -74,12 +78,14 @@ tp_format(const char *sys, const char *name)char path[PATH_MAX ];size_t size;char *data; +int err;scnprintf (path,PATH_MAX ,"%s/%s/%s/format" , tracing_events_path, sys, name); -if (filename__read_str (path, &data, &size)) -return NULL ; + err =filename__read_str (path, &data, &size); +if (err) +return ERR_PTR (err);pevent_parse_format (pevent, &event, data, size, sys);
@@ -87,11 +93,14 @@ tp_format(const char *sys, const char *name)return event; } +/* +* Returns pointer with encoded error via <linux/err.h> interface. +*/ struct event_format *trace_event__tp_format (const char *sys,const char *name) {if (!tevent_initialized &&trace_event__init2 ()) -return NULL ; +return ERR_PTR (-ENOMEM );return tp_format (sys, name); }
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 6/8 perf tools: Enhance parsing events tracepoint error output
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Enhancing parsing events tracepoint error output. Adding more verbose output when the tracepoint is not found or the tracing event path cannot be access.
$ sudo perf record -e sched:sched_krava ls
event syntax error: 'sched:sched_krava'
\___ unknown tracepoint Error: File /sys/kernel/debug/tracing//tracing/events/sched/sched_krava not found.
Hint: Perhaps this kernel misses some CONFIG_ setting to enable this feature?.
Run 'perf list' for a list of valid events
...
$ perf record -e sched:sched_krava ls
event syntax error: 'sched:sched_krava'
\___ can't access trace events Error: No permissions to read /sys/kernel/debug/tracing//tracing/events/sched/sched_krava
Hint: Try 'sudo mount -o remount,mode=755 /sys/kernel/debug'
Run 'perf list' for a list of valid events
...
--- tools/perf/util/parse-events.c | 35 ++++++++++++++++++++++++++++++++--- tools/perf/util/parse-events.y | 16 +++++++++------- 2 files changed, 41 insertions(+), 10 deletions(-)
tools/perf/util/parse-events.c +32 -3
@@ -387,6 +387,33 @@ int parse_events_add_cache(struct list_head *list, int *idx,return add_event (list, idx, &attr, name,NULL ); } +static void tracepoint_error (struct parse_events_error *error,int err, +char *sys,char *name) +{ +char help[BUFSIZ ]; + +/* +* We get error directly from syscall errno ( > 0), +* or from encoded pointer's error ( < 0). +*/ + err =abs (err); + +switch (err) { +case EACCES : + error->str =strdup ("can't access trace events" ); +break ; +case ENOENT : + error->str =strdup ("unknown tracepoint" ); +break ; +default : + error->str =strdup ("failed to add tracepoint" ); +break ; + } + +tracing_path__strerror_open_tp (err, help,sizeof (help), sys, name); + error->help =strdup (help); +} +static int add_tracepoint (struct list_head *list,int *idx,char *sys_name,char *evt_name,struct parse_events_error *error__maybe_unused )
@@ -394,8 +421,10 @@ static int add_tracepoint(struct list_head *list, int *idx,struct perf_evsel *evsel; evsel =perf_evsel__newtp_idx (sys_name, evt_name, (*idx)++); -if (IS_ERR (evsel)) +if (IS_ERR (evsel)) { +tracepoint_error (error,PTR_ERR (evsel), sys_name, evt_name);return PTR_ERR (evsel); + }list_add_tail (&evsel->node, list);return 0 ;
@@ -413,7 +442,7 @@ static int add_tracepoint_multi_event(struct list_head *list, int *idx,snprintf (evt_path,MAXPATHLEN ,"%s/%s" , tracing_events_path, sys_name); evt_dir =opendir (evt_path);if (!evt_dir) { -perror ("Can't open event dir" ); +tracepoint_error (error, errno, sys_name, evt_name);return -1 ; }
@@ -453,7 +482,7 @@ static int add_tracepoint_multi_sys(struct list_head *list, int *idx, events_dir =opendir (tracing_events_path);if (!events_dir) { -perror ("Can't open event dir" ); +tracepoint_error (error, errno, sys_name, evt_name);return -1 ; }
tools/perf/util/parse-events.y +9 -7
@@ -371,28 +371,30 @@ event_legacy_tracepoint: PE_NAME '-' PE_NAME ':' PE_NAME { struct parse_events_evlist *data = _data; + struct parse_events_error *error = data->error; struct list_head *list; char sys_name[128]; snprintf(&sys_name, 128, "%s-%s", $1, $3); ALLOC_LIST(list); - ABORT_ON(parse_events_add_tracepoint(list, &data->idx, &sys_name, $5, data->error)); + if (parse_events_add_tracepoint(list, &data->idx, &sys_name, $5, error)) { + if (error) + error->idx = @1.first_column; + return -1; + } $$ = list; } | PE_NAME ':' PE_NAME { struct parse_events_evlist *data = _data; + struct parse_events_error *error = data->error; struct list_head *list; ALLOC_LIST(list); - if (parse_events_add_tracepoint(list, &data->idx, $1, $3, data->error)) { - struct parse_events_error *error = data->error; - - if (error) { + if (parse_events_add_tracepoint(list, &data->idx, $1, $3, error)) { + if (error) error->idx = @1.first_column; - error->str = strdup("unknown tracepoint"); - } return -1; } $$ = list;
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 7/8 perf tools: regs_query_register_offset() infrastructure
In reply to Arnaldo Carvalho de MeloFrom: Wang Nan <[email protected]>
regs_query_register_offset() is a helper function which converts register name like "%rax" to offset of a register in 'struct pt_regs', which is required by BPF prologue generator.
PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET indicates an architecture supports converting name of a register to its offset in 'struct pt_regs'.
HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET is introduced as the corresponding CFLAGS of PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET.
--- tools/perf/config/Makefile | 4 ++++ tools/perf/util/include/dwarf-regs.h | 8 ++++++++ 2 files changed, 12 insertions(+)
tools/perf/config/Makefile +4 -0
@@ -109,6 +109,10 @@ endif # include ARCH specific config -include $(src-perf)/arch/$(ARCH)/Makefile +ifdef PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET + CFLAGS += -DHAVE_ARCH_REGS_QUERY_REGISTER_OFFSET +endif + include $(src-perf)/config/utilities.mak ifeq ($(call get-executable,$(FLEX)),)
tools/perf/util/include/dwarf-regs.h +8 -0
@@ -5,4 +5,12 @@const char *get_arch_regstr (unsigned int n);#endif +#ifdef HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET +/* +* Arch should support fetching the offset of a register in pt_regs +* by its name. See kernel's regs_query_register_offset in +* arch/xxx/kernel/ptrace.c. +*/ +int regs_query_register_offset (const char *name); +#endif #endif
Signature · 1 line
2.1.0
Arnaldo Carvalho de Melo
maintainerPATCH 8/8 perf tools: Introduce regs_query_register_offset() for x86
In reply to Arnaldo Carvalho de MeloFrom: Wang Nan <[email protected]>
regs_query_register_offset() is a helper function which converts register name like "%rax" to offset of a register in 'struct pt_regs', which is required by BPF prologue generator. Since the function is identical, try to reuse the code in arch/x86/kernel/ptrace.c.
Comment inside dwarf-regs.c list the differences between this implementation and kernel code.
get_arch_regstr() switches to regoffset_table and the old string table is dropped.
--- tools/perf/arch/x86/Makefile | 1 + tools/perf/arch/x86/util/dwarf-regs.c | 122 ++++++++++++++++++++++++---------- 2 files changed, 89 insertions(+), 34 deletions(-)
tools/perf/arch/x86/Makefile +1 -0
@@ -2,3 +2,4 @@ ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1 +PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1
tools/perf/arch/x86/util/dwarf-regs.c +88 -34
143 lines not on this page · raw message
Signature · 1 line
2.1.0
Quoted 6 lines · Arnaldo Carvalho de Melo
Hi Ingo,
Please consider pulling,
- Arnaldo
Quoted 7 lines
The following changes since commit
9059b284caecb628fac826c2c5cc8ee85708eec1:Merge tag 'perf-core-for-mingo-2' of
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linuxinto perf/core (2015-09-15 08:50:59 +0200)are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.gittags/perf-core-for-mingo
So your perf/urgent bits now conflict with the latest perf/core, in tools/perf/ui/browsers/hists.c. I have merged perf/urgent into perf/core - please double check my resolution (d71b0ad8d309).
Quoted 32 lines
for you to fetch changes up to
bbbe6bf6037d77816c4a19aaf35f4cecf662b49a:perf tools: Introduce
regs_query_register_offset()for x86 (2015-09-15 09:48:33 -0300)----------------------------------------------------------------
perf/core improvements and fixes:User visible:
- Enhance the error reporting of tracepoint event parsing, e.g.:
$ oldperf record -e sched:sched_switc usleep 1 event syntax error: 'sched:sched_switc' \___ unknown tracepoint Run 'perf list' for a list of valid eventsNow we get the much nicer:
$ perf record -e sched:sched_switc ls event syntax error: 'sched:sched_switc' \___ can't access trace eventsError: No permissions to read
/sys/kernel/debug/tracing/events/sched/sched_switc
Hint: Try 'sudo mount -o remount,mode=755/sys/kernel/debug'And after we have those mount point permissions fixed:
$ perf record -e sched:sched_switc ls event syntax error: 'sched:sched_switc' \___ unknown tracepoint
Quoted 5 lines
Error: File
/sys/kernel/debug/tracing/events/sched/sched_switcnot found.
Hint: Perhaps this kernel misses someCONFIG_setting to enable this feature?.Now its just a matter of using what git uses to suggest alternatives when we make a typo, i.e. that it is just an 'h' missing :-)
Nice changes!
Btw., wouldn't it be even better to allow partial matches? Not allowing 'sched:sched_switc' is unnecessarily pedantic IMHO.
For example 'perf list' allows partial matches as well. As long as the resulting event is unique, we should allow partial matches. If it's not unique, we should print the first 3 matching entries or so.
There's a real UI advantage as well: I could abbreviate the command line with:
-e sched_sw
instead of always being forced to type out the full tracepoint name.
(Programmatic tracepoint usage and portable scripts should naturally always spell out the full event, to make sure new tracepoints don't cause overlaps - but ad-hoc usage can do abbreviations just fine.)
Quoted 45 lines
I.e. basically now the event parsing routing uses the
strerror_open()routines introduced by and used in 'perf trace' work. (Jiri Olsa)Infrastructure:
- Export init/exit_probe_symbol_maps() from 'perf probe' for use in eBPF (Namhyung Kim)
- Free
perf_probe_eventincleanup_perf_probe_events()(Namhyung Kim)-
regs_query_register_offset()infrastructure + implementation for x86. First user will be the perf/eBPF code (Wang Nan)Signed-off-by Arnaldo Carvalho de Melo <[email protected]>----------------------------------------------------------------
Jiri Olsa (4):
tools: Adderr.hwithERR_PTRPTR_ERRinterface perf tools: Propagate error info for the tracepoint parsing perf evsel: Propagate error info fromtp_formatperf tools: Enhance parsing events tracepoint error outputNamhyung Kim (2):
perf probe: Freeperf_probe_eventincleanup_perf_probe_events()perf probe: Export init/exit_probe_symbol_maps()Wang Nan (2):
perf tools:regs_query_register_offset()infrastructure perf tools: Introduceregs_query_register_offset()for x86tools/include/linux/err.h | 49 +++++++++++ tools/perf/arch/x86/Makefile | 1 + tools/perf/arch/x86/util/dwarf-regs.c | 122 ++++++++++++++++++++-------- tools/perf/builtin-probe.c | 5 ++ tools/perf/builtin-trace.c | 19 +++-- tools/perf/config/Makefile | 4 + tools/perf/tests/evsel-tp-sched.c | 10 ++- tools/perf/tests/mmap-basic.c | 3 +- tools/perf/tests/openat-syscall-all-cpus.c | 3 +- tools/perf/tests/openat-syscall-tp-fields.c | 3 +- tools/perf/tests/openat-syscall.c | 3 +- tools/perf/util/evlist.c | 3 +- tools/perf/util/evsel.c | 16 +++- tools/perf/util/evsel.h | 3 + tools/perf/util/include/dwarf-regs.h | 8 ++
Quoted 8 lines
tools/perf/util/parse-events.c | 66 +++++++++++---- tools/perf/util/parse-events.h | 3 +- tools/perf/util/parse-events.y | 16 ++-- tools/perf/util/probe-event.c | 32 ++++---- tools/perf/util/probe-event.h | 2 + tools/perf/util/trace-event.c | 15 +++- 21 files changed, 291 insertions(+), 95 deletions(-) create mode 100644 tools/include/linux/err.h
Pulled, thanks a lot Arnaldo!
Ingo
Arnaldo Carvalho de Melo
maintainerEm Wed, Sep 16, 2015 at 09:25:44AM +0200, Ingo Molnar escreveu:
Quoted 13 lines · Arnaldo Carvalho de Melo
Quoted 11 lines
Hi Ingo,
Please consider pulling,
- Arnaldo
The following changes since commit
9059b284caecb628fac826c2c5cc8ee85708eec1:Merge tag 'perf-core-for-mingo-2' of
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linuxinto perf/core (2015-09-15 08:50:59 +0200)
Quoted 7 lines
Quoted 3 lines
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.gittags/perf-core-for-mingoSo your perf/urgent bits now conflict with the latest perf/core, in
tools/perf/ui/browsers/hists.c. I have merged perf/urgent into perf/core - please double check my resolution (d71b0ad8d309).
Looks fine, tested it even, thanks!
- Arnaldo