perf bpf_counter: Move header declarations into C code
Subsystems (2)
Ian Rogers
reviewerPATCH 1/2 perf bpf_counter: Move header declarations into C code
Reduce the API surface that is in bpf_counter.h, this helps compiler analysis like unused static function, makes it easier to set a breakpoint and just makes it easier to see the code is self contained. When code is shared between BPF C code, put it inside HAVE_BPF_SKEL. Move transitively found #includes into appropriate C files.
No functional change.
--- tools/perf/util/bpf_counter.c | 62 ++++++++++++++++++++++- tools/perf/util/bpf_counter.h | 74 +++------------------------- tools/perf/util/bpf_counter_cgroup.c | 1 + tools/perf/util/bpf_ftrace.c | 1 + tools/perf/util/bpf_off_cpu.c | 1 + 5 files changed, 69 insertions(+), 70 deletions(-)
tools/perf/util/bpf_counter.c +60 -2
@@ -6,10 +6,14 @@#include <limits.h> #include <unistd.h> #include <sys/file.h> +#include <sys/resource.h> #include <sys/time.h> #include <linux/err.h> +#include <linux/list.h> #include <linux/zalloc.h> #include <api/fs/fs.h> +#include <bpf/bpf.h> +#include <bpf/btf.h> #include <perf/bpf_perf.h> #include "bpf_counter.h"
@@ -28,13 +32,67 @@#include "bpf_skel/bperf_leader.skel.h" #include "bpf_skel/bperf_follower.skel.h" +struct bpf_counter { +void *skel; +struct list_head list; +}; +#define ATTR_MAP_SIZE 16 -static inline void *u64_to_ptr (__u64 ptr) +static void *u64_to_ptr (__u64 ptr) {return (void *)(unsigned long )ptr; } + +void set_max_rlimit (void ) +{ +struct rlimit rinf = {RLIM_INFINITY ,RLIM_INFINITY }; + +setrlimit (RLIMIT_MEMLOCK , &rinf); +} + +static __u32 bpf_link_get_id (int fd) +{ +struct bpf_link_info link_info = { .id =0 , }; +__u32 link_info_len =sizeof (link_info); + +bpf_obj_get_info_by_fd (fd, &link_info, &link_info_len); +return link_info.id; +} + +static __u32 bpf_link_get_prog_id (int fd) +{ +struct bpf_link_info link_info = { .id =0 , }; +__u32 link_info_len =sizeof (link_info); + +bpf_obj_get_info_by_fd (fd, &link_info, &link_info_len); +return link_info.prog_id; +} + +static __u32 bpf_map_get_id (int fd) +{ +struct bpf_map_info map_info = { .id =0 , }; +__u32 map_info_len =sizeof (map_info); + +bpf_obj_get_info_by_fd (fd, &map_info, &map_info_len); +return map_info.id; +} + +/* trigger the leader program on a cpu */ +int bperf_trigger_reading (int prog_fd,int cpu) +{ +DECLARE_LIBBPF_OPTS (bpf_test_run_opts, opts, + .ctx_in = NULL, + .ctx_size_in = 0, + .flags = BPF_F_TEST_RUN_ON_CPU, + .cpu = cpu, + .retval = 0, + ); + +return bpf_prog_test_run_opts (prog_fd, &opts); +} +static struct bpf_counter *bpf_counter_alloc (void ) {struct bpf_counter *counter;
@@ -785,7 +843,7 @@ struct bpf_counter_ops bperf_ops = {extern struct bpf_counter_ops bperf_cgrp_ops; -static inline bool bpf_counter_skip (struct evsel *evsel) +static bool bpf_counter_skip (struct evsel *evsel) {return evsel->bpf_counter_ops ==NULL ; }
tools/perf/util/bpf_counter.h +6 -68
104 lines not on this page · raw message
tools/perf/util/bpf_counter_cgroup.c +1 -0
@@ -13,6 +13,7 @@#include <linux/zalloc.h> #include <linux/perf_event.h> #include <api/fs/fs.h> +#include <bpf/bpf.h> #include <perf/bpf_perf.h> #include "affinity.h"
tools/perf/util/bpf_ftrace.c +1 -0
@@ -3,6 +3,7 @@#include <stdint.h> #include <stdlib.h> +#include <bpf/bpf.h> #include <linux/err.h> #include "util/ftrace.h"
tools/perf/util/bpf_off_cpu.c +1 -0
@@ -13,6 +13,7 @@#include "util/cgroup.h" #include "util/strlist.h" #include <bpf/bpf.h> +#include <bpf/btf.h> #include <internal/xyarray.h> #include <linux/time64.h>
Signature · 1 line
2.51.0.618.g983fd99d29-goog
Ian Rogers
reviewerPATCH 2/2 perf bpf_counter: Fix handling of cpumap fixing hybrid
Don't open evsels on all CPUs, open them just on the CPUs they support. This avoids opening say an e-core event on a p-core and getting a failure - achieve this by getting rid of the "all_cpu_map".
In install_pe functions don't use the cpu_map_idx as a CPU number, translate the cpu_map_idx, which is a dense index into the cpu_map skipping holes at the beginning, to a proper CPU number.
Before:
```
$ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1
Performance counter stats for 'system wide':
<not supported> cpu_atom/cycles/
566,270,672 cpu_core/cycles/
<not supported> cpu_atom/instructions/
572,792,836 cpu_core/instructions/ # 1.01 insn per cycle
1.001595384 seconds time elapsed
```After:
```
$ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1
Performance counter stats for 'system wide':
443,299,201 cpu_atom/cycles/
1,233,919,737 cpu_core/cycles/
213,634,112 cpu_atom/instructions/ # 0.48 insn per cycle
2,758,965,527 cpu_core/instructions/ # 2.24 insn per cycle
1.001699485 seconds time elapsed
```--- tools/perf/util/bpf_counter.c | 26 ++++++++++---------------- tools/perf/util/bpf_counter_cgroup.c | 3 ++- 2 files changed, 12 insertions(+), 17 deletions(-)
tools/perf/util/bpf_counter.c +10 -16
@@ -336,6 +336,7 @@ static int bpf_program_profiler__install_pe(struct evsel *evsel, int cpu_map_idx {struct bpf_prog_profiler_bpf *skel;struct bpf_counter *counter; +int cpu =perf_cpu_map__cpu (evsel->core.cpus, cpu_map_idx).cpu;int ret;list_for_each_entry (counter, &evsel->bpf_counter_list, list) {
@@ -343,7 +344,7 @@ static int bpf_program_profiler__install_pe(struct evsel *evsel, int cpu_map_idxassert (skel !=NULL ); ret =bpf_map_update_elem (bpf_map__fd (skel->maps.events), - &cpu_map_idx, &fd,BPF_ANY ); + &cpu, &fd,BPF_ANY );if (ret)return ret; }
@@ -451,7 +452,6 @@ static int bperf_check_target(struct evsel *evsel,return 0 ; } -static struct perf_cpu_map *all_cpu_map;static __u32 filter_entry_cnt; static int bperf_reload_leader_program(struct evsel *evsel, int attr_map_fd,
@@ -495,7 +495,7 @@ static int bperf_reload_leader_program(struct evsel *evsel, int attr_map_fd,* following evsel__open_per_cpu call */ evsel->leader_skel = skel; -evsel__open_per_cpu (evsel, all_cpu_map,-1 ); +evsel__open (evsel, evsel->core.cpus, evsel->core.threads); out:bperf_leader_bpf__destroy (skel);
@@ -533,12 +533,6 @@ static int bperf__load(struct evsel *evsel, struct target *target)if (bperf_check_target (evsel, target, &filter_type, &filter_entry_cnt))return -1 ; -if (!all_cpu_map) { - all_cpu_map =perf_cpu_map__new_online_cpus (); -if (!all_cpu_map) -return -1 ; - } - evsel->bperf_leader_prog_fd =-1 ; evsel->bperf_leader_link_fd =-1 ;
@@ -656,9 +650,10 @@ static int bperf__load(struct evsel *evsel, struct target *target)static int bperf__install_pe (struct evsel *evsel,int cpu_map_idx,int fd) {struct bperf_leader_bpf *skel = evsel->leader_skel; +int cpu =perf_cpu_map__cpu (evsel->core.cpus, cpu_map_idx).cpu;return bpf_map_update_elem (bpf_map__fd (skel->maps.events), - &cpu_map_idx, &fd,BPF_ANY ); + &cpu, &fd,BPF_ANY ); }/*
@@ -667,13 +662,12 @@ static int bperf__install_pe(struct evsel *evsel, int cpu_map_idx, int fd)*/ static int bperf_sync_counters (struct evsel *evsel) { -int num_cpu, i, cpu; +struct perf_cpu cpu; +int idx; + +perf_cpu_map__for_each_cpu (cpu, idx, evsel->core.cpus) +bperf_trigger_reading (evsel->bperf_leader_prog_fd, cpu.cpu); - num_cpu =perf_cpu_map__nr (all_cpu_map); -for (i =0 ; i < num_cpu; i++) { - cpu =perf_cpu_map__cpu (all_cpu_map, i).cpu; -bperf_trigger_reading (evsel->bperf_leader_prog_fd, cpu); - }return 0 ; }
tools/perf/util/bpf_counter_cgroup.c +2 -1
@@ -186,7 +186,8 @@ static int bperf_cgrp__load(struct evsel *evsel, }static int bperf_cgrp__install_pe (struct evsel *evsel__maybe_unused , -int cpu__maybe_unused ,int fd__maybe_unused ) +int cpu_map_idx__maybe_unused , +int fd__maybe_unused ) {/* nothing to do */ return 0 ;
Signature · 1 line
2.51.0.618.g983fd99d29-goog
Ian Rogers
reviewerRe: [PATCH v1 2/2] perf bpf_counter: Fix handling of cpumap fixing hybrid
Quoted 37 lines · Ian Rogers
Don't open evsels on all CPUs, open them just on the CPUs they support. This avoids opening say an e-core event on a p-core and getting a failure - achieve this by getting rid of the "
all_cpu_map".In
install_pefunctions don't use thecpu_map_idxas a CPU number, translate thecpu_map_idx, which is a dense index into thecpu_mapskipping holes at the beginning, to a proper CPU number.Before:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': <not supported> cpu_atom/cycles/ 566,270,672 cpu_core/cycles/ <not supported> cpu_atom/instructions/ 572,792,836 cpu_core/instructions/ # 1.01 insn per cycle 1.001595384 seconds time elapsed ```After:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': 443,299,201 cpu_atom/cycles/ 1,233,919,737 cpu_core/cycles/ 213,634,112 cpu_atom/instructions/ # 0.48 insn per cycle 2,758,965,527 cpu_core/instructions/ # 2.24 insn per cycle 1.001699485 seconds time elapsed ```
Quoted 2 lines
Fixes 7fac83aaf2ee ("perf stat: Introduce 'bperf' to share hardware PMCs with BPF")Signed-off-by Ian Rogers <[email protected]>
+Thomas Falcon
I think it'd be nice to get this quite major fix for --bpf-counters/bperf for hybrid architectures into v6.18 and stable builds. Thomas would it be possible for you to give a Tested-by tag using the reproduction in the commit message?
Thanks,
Ian
Quoted 104 lines
--- tools/perf/util/bpf_counter.c | 26 ++++++++++---------------- tools/perf/util/bpf_counter_cgroup.c | 3 ++- 2 files changed, 12 insertions(+), 17 deletions(-)tools/perf/util/bpf_counter.c +10
-16@@ -336,6 +336,7 @@ static int bpf_program_profiler__install_pe(struct evsel *evsel, int cpu_map_idx {struct bpf_prog_profiler_bpf *skel;struct bpf_counter *counter; +int cpu =perf_cpu_map__cpu (evsel->core.cpus, cpu_map_idx).cpu;int ret;list_for_each_entry (counter, &evsel->bpf_counter_list, list) {@@ -343,7 +344,7 @@ static int bpf_program_profiler__install_pe(struct evsel *evsel, int cpu_map_idxassert (skel !=NULL ); ret =bpf_map_update_elem (bpf_map__fd (skel->maps.events), - &cpu_map_idx, &fd,BPF_ANY ); + &cpu, &fd,BPF_ANY );if (ret)return ret; }@@ -451,7 +452,6 @@ static int bperf_check_target(struct evsel *evsel,return 0 ; } -static struct perf_cpu_map *all_cpu_map;static __u32 filter_entry_cnt; static int bperf_reload_leader_program(struct evsel *evsel, int attr_map_fd,@@ -495,7 +495,7 @@ static int bperf_reload_leader_program(struct evsel *evsel, int attr_map_fd,* following evsel__open_per_cpu call */ evsel->leader_skel = skel; -evsel__open_per_cpu (evsel, all_cpu_map,-1 ); +evsel__open (evsel, evsel->core.cpus, evsel->core.threads); out:bperf_leader_bpf__destroy (skel);@@ -533,12 +533,6 @@ static int bperf__load(struct evsel *evsel, struct target *target)if (bperf_check_target (evsel, target, &filter_type, &filter_entry_cnt))return -1 ; -if (!all_cpu_map) { - all_cpu_map =perf_cpu_map__new_online_cpus (); -if (!all_cpu_map) -return -1 ; - } - evsel->bperf_leader_prog_fd =-1 ; evsel->bperf_leader_link_fd =-1 ;@@ -656,9 +650,10 @@ static int bperf__load(struct evsel *evsel, struct target *target)static int bperf__install_pe (struct evsel *evsel,int cpu_map_idx,int fd) {struct bperf_leader_bpf *skel = evsel->leader_skel; +int cpu =perf_cpu_map__cpu (evsel->core.cpus, cpu_map_idx).cpu;return bpf_map_update_elem (bpf_map__fd (skel->maps.events), - &cpu_map_idx, &fd,BPF_ANY ); + &cpu, &fd,BPF_ANY ); }/* @@ -667,13 +662,12 @@ static int bperf__install_pe(struct evsel *evsel, int cpu_map_idx, int fd)*/ static int bperf_sync_counters (struct evsel *evsel) { -int num_cpu, i, cpu; +struct perf_cpu cpu; +int idx; + +perf_cpu_map__for_each_cpu (cpu, idx, evsel->core.cpus) +bperf_trigger_reading (evsel->bperf_leader_prog_fd, cpu.cpu); - num_cpu =perf_cpu_map__nr (all_cpu_map); -for (i =0 ; i < num_cpu; i++) { - cpu =perf_cpu_map__cpu (all_cpu_map, i).cpu; -bperf_trigger_reading (evsel->bperf_leader_prog_fd, cpu); - }return 0 ; }tools/perf/util/bpf_counter_cgroup.c +2
-2@@ -186,7 +186,8 @@ static int bperf_cgrp__load(struct evsel *evsel, }static int bperf_cgrp__install_pe (struct evsel *evsel__maybe_unused , -int cpu__maybe_unused ,int fd__maybe_unused ) +int cpu_map_idx__maybe_unused , +int fd__maybe_unused ) {/* nothing to do */ return 0 ; --2.51.0.618.g983fd99d29-goog
Arnaldo Carvalho de Melo
maintainerRe: [PATCH v1 2/2] perf bpf_counter: Fix handling of cpumap fixing hybrid
Quoted 41 lines · Ian Rogers
Quoted 37 lines · Ian Rogers
Don't open evsels on all CPUs, open them just on the CPUs they support. This avoids opening say an e-core event on a p-core and getting a failure - achieve this by getting rid of the "
all_cpu_map".In
install_pefunctions don't use thecpu_map_idxas a CPU number, translate thecpu_map_idx, which is a dense index into thecpu_mapskipping holes at the beginning, to a proper CPU number.Before:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': <not supported> cpu_atom/cycles/ 566,270,672 cpu_core/cycles/ <not supported> cpu_atom/instructions/ 572,792,836 cpu_core/instructions/ # 1.01 insn per cycle 1.001595384 seconds time elapsed ```After:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': 443,299,201 cpu_atom/cycles/ 1,233,919,737 cpu_core/cycles/ 213,634,112 cpu_atom/instructions/ # 0.48 insn per cycle 2,758,965,527 cpu_core/instructions/ # 2.24 insn per cycle 1.001699485 seconds time elapsed ```Quoted 2 lines
Fixes 7fac83aaf2ee ("perf stat: Introduce 'bperf' to share hardware PMCs with BPF")Signed-off-by Ian Rogers <[email protected]>
Quoted 6 lines
+Thomas Falcon
I think it'd be nice to get this quite major fix for --bpf-counters/bperf for hybrid architectures into v6.18 and stable builds. Thomas would it be possible for you to give a Tested-by tag using the reproduction in the commit message?
Its even already in linux-next:
⬢ [acme@toolbx perf-tools-next]$ git log -5 --oneline linux-next/master tools/perf/util/bpf_counter.c b91917c0c6fa6df9 perf bpf_counter: Fix handling of cpumap fixing hybrid8c519a825b4add85 perf bpf_counter: Move header declarations into C code07dc3a6de33098b0 perf stat: Support inherit events during fork() for bperfeffe957c6bb70cac libperf cpumap: Replace usage of perf_cpu_map__new(NULL) with perf_cpu_map__new_online_cpus() b84b3f47921568a8 perf bpf_counter: Fix a few memory leaks
⬢ [acme@toolbx perf-tools-next]$
Ian Rogers
reviewerRe: [PATCH v1 2/2] perf bpf_counter: Fix handling of cpumap fixing hybrid
Quoted 52 lines · Arnaldo Carvalho de Melo
Quoted 41 lines · Ian Rogers
Quoted 37 lines · Ian Rogers
Don't open evsels on all CPUs, open them just on the CPUs they support. This avoids opening say an e-core event on a p-core and getting a failure - achieve this by getting rid of the "
all_cpu_map".In
install_pefunctions don't use thecpu_map_idxas a CPU number, translate thecpu_map_idx, which is a dense index into thecpu_mapskipping holes at the beginning, to a proper CPU number.Before:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': <not supported> cpu_atom/cycles/ 566,270,672 cpu_core/cycles/ <not supported> cpu_atom/instructions/ 572,792,836 cpu_core/instructions/ # 1.01 insn per cycle 1.001595384 seconds time elapsed ```After:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': 443,299,201 cpu_atom/cycles/ 1,233,919,737 cpu_core/cycles/ 213,634,112 cpu_atom/instructions/ # 0.48 insn per cycle 2,758,965,527 cpu_core/instructions/ # 2.24 insn per cycle 1.001699485 seconds time elapsed ```Quoted 2 lines
Fixes 7fac83aaf2ee ("perf stat: Introduce 'bperf' to share hardware PMCs with BPF")Signed-off-by Ian Rogers <[email protected]>Quoted 6 lines
+Thomas Falcon
I think it'd be nice to get this quite major fix for --bpf-counters/bperf for hybrid architectures into v6.18 and stable builds. Thomas would it be possible for you to give a Tested-by tag using the reproduction in the commit message?
Its even already in linux-next:
Quoted 7 lines
⬢ [acme@toolbx perf-tools-next]$ git log -5 --oneline linux-next/master
tools/perf/util/bpf_counter.cb91917c0c6fa6df9perfbpf_counter: Fix handling of cpumap fixing hybrid8c519a825b4add85perfbpf_counter: Move header declarations into C code07dc3a6de33098b0perf stat: Support inherit events duringfork()for bperfeffe957c6bb70caclibperf cpumap: Replace usage ofperf_cpu_map__new(NULL)withperf_cpu_map__new_online_cpus()b84b3f47921568a8perfbpf_counter: Fix a few memory leaks
⬢ [acme@toolbx perf-tools-next]$
Oh, thanks! Sorry for missing it. Thanks Thomas for the tag!
Ian
Falcon, Thomas
Tested-byRe: [PATCH v1 2/2] perf bpf_counter: Fix handling of cpumap fixing hybrid
In reply to Ian RogersQuoted 49 lines · Ian Rogers
Quoted 41 lines · Ian Rogers
Don't open evsels on all CPUs, open them just on the CPUs they support. This avoids opening say an e-core event on a p-core and getting a failure - achieve this by getting rid of the "
all_cpu_map".In
install_pefunctions don't use thecpu_map_idxas a CPU number, translate thecpu_map_idx, which is a dense index into thecpu_mapskipping holes at the beginning, to a proper CPU number.Before:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': <not supported> cpu_atom/cycles/ 566,270,672 cpu_core/cycles/ <not supported> cpu_atom/instructions/ 572,792,836 cpu_core/instructions/ # 1.01 insn per cycle 1.001595384 seconds time elapsed ```After:
``` $ perf stat --bpf-counters -a -e cycles,instructions -- sleep 1 Performance counter stats for 'system wide': 443,299,201 cpu_atom/cycles/ 1,233,919,737 cpu_core/cycles/ 213,634,112 cpu_atom/instructions/ # 0.48 insn per cycle 2,758,965,527 cpu_core/instructions/ # 2.24 insn per cycle 1.001699485 seconds time elapsed ```Quoted 3 lines
Fixes 7fac83aaf2ee ("perf stat: Introduce 'bperf' to sharehardware PMCs with BPF")
Signed-off-by Ian Rogers <[email protected]>+Thomas Falcon
Quoted 8 lines
I think it'd be nice to get this quite major fix for --bpf-counters/bperf for hybrid architectures into v6.18 and stable builds. Thomas would it be possible for you to give a Tested-by tag using the reproduction in the commit message?
Thanks,
Ian
Sorry for missing this. Here's my reproduction on an alder lake
sudo ./perf stat --bpf-counters -a -e cycles,instructions -- sleep 1
...
Performance counter stats for 'system wide':
364,715,896 cpu_atom/cycles/
946,331,957 cpu_core/cycles/
169,842,929 cpu_atom/instructions/ # 0.47
insn per cycle
1,338,720,324 cpu_core/instructions/ # 1.41 insn per cycle
1.001667769 seconds time elapsed
It looks like it's in perf-tools-next already but have my tested by for what it's worth.
Quoted 129 lines
Quoted 129 lines
---
tools/perf/util/bpf_counter.c | 26 ++++++++++-------------- --
tools/perf/util/bpf_counter_cgroup.c | 3 ++-2 files changed, 12 insertions(+), 17 deletions(-)tools/perf/util/bpf_counter.c +0
-0@@ -336,6 +336,7 @@ static intbpf_program_profiler__install_pe(struct evsel *evsel, int
cpu_map_idx
{
structbpf_prog_profiler_bpf*skel; structbpf_counter*counter;
+ int cpu =perf_cpu_map__cpu(evsel->core.cpus, cpu_map_idx).cpu;
int ret;
list_for_each_entry(counter, &evsel->bpf_counter_list, list){tools/perf/util/bpf_counter.c +0
-0@@ -343,7 +344,7 @@ static intbpf_program_profiler__install_pe(struct evsel *evsel, int
cpu_map_idx
assert(skel != NULL);ret = bpf_map_update_elem(bpf_map__fd(skel-
Quoted 1 line
maps.events),
-
&cpu_map_idx, &fd,BPF_ANY);
+ &cpu, &fd,BPF_ANY); if (ret)
return ret;
}tools/perf/util/bpf_counter.c +0
-1@@ -451,7 +452,6 @@ static int bperf_check_target(struct evsel *evsel, return 0; } -static struct perf_cpu_map *all_cpu_map; static __u32 filter_entry_cnt;static int bperf_reload_leader_program(struct evsel *evsel, int
attr_map_fd,tools/perf/util/bpf_counter.c +1
-1@@ -495,7 +495,7 @@ static int bperf_reload_leader_program(structevsel *evsel, int attr_map_fd, * following evsel__open_per_cpu call */ evsel->leader_skel = skel; - evsel__open_per_cpu(evsel, all_cpu_map, -1); + evsel__open(evsel, evsel->core.cpus, evsel->core.threads);out:
bperf_leader_bpf__destroy(skel);tools/perf/util/bpf_counter.c +0
-6@@ -533,12 +533,6 @@ static int bperf__load(struct evsel *evsel, struct target *target) if (bperf_check_target(evsel, target, &filter_type, &filter_entry_cnt)) return -1; - if (!all_cpu_map) { - all_cpu_map = perf_cpu_map__new_online_cpus(); - if (!all_cpu_map) - return -1; - } - evsel->bperf_leader_prog_fd = -1;
evsel->bperf_leader_link_fd= -1;tools/perf/util/bpf_counter.c +2
-1@@ -656,9 +650,10 @@ static int bperf__load(struct evsel *evsel, struct target *target) static int bperf__install_pe(struct evsel *evsel, int cpu_map_idx,int fd) { struct bperf_leader_bpf *skel = evsel->leader_skel; + int cpu = perf_cpu_map__cpu(evsel->core.cpus, cpu_map_idx).cpu; return bpf_map_update_elem(bpf_map__fd(skel->maps.events), - &cpu_map_idx, &fd, BPF_ANY); + &cpu, &fd, BPF_ANY);}
/*
tools/perf/util/bpf_counter.c +5
-6@@ -667,13 +662,12 @@ static int bperf__install_pe(struct evsel*evsel, int cpu_map_idx, int fd) */ static int bperf_sync_counters(struct evsel *evsel) { - int num_cpu, i, cpu; + struct perf_cpu cpu; + int idx; + + perf_cpu_map__for_each_cpu(cpu, idx, evsel->core.cpus) + bperf_trigger_reading(evsel->bperf_leader_prog_fd, cpu.cpu); - num_cpu = perf_cpu_map__nr(all_cpu_map); - for (i = 0; i < num_cpu; i++) { - cpu = perf_cpu_map__cpu(all_cpu_map, i).cpu; - bperf_trigger_reading(evsel->bperf_leader_prog_fd, cpu); - }return 0;
}tools/perf/util/bpf_counter_cgroup.c +0
-0@@ -186,7 +186,8 @@ static int bperf_cgrp__load(struct evsel *evsel, }static int bperf_cgrp__install_pe(struct evsel *evsel
__maybe_unused,
- int cpu__maybe_unused, int fd__maybe_unused)
+ intcpu_map_idx__maybe_unused, + int fd__maybe_unused) {
/* nothing to do */
return 0;
--
2.51.0.618.g983fd99d29-goog