perf stat: Add/fix bperf cgroup max events workarounds
Ian Rogers
reviewerPATCH 1/1 perf stat: Add/fix bperf cgroup max events workarounds
Commit b8308511f6e0 bumped the max events to 1024 but this results in BPF verifier issues if the number of command line events is too large. Workaround this by:
1) moving the constants to a header file to share between BPF and perf C code,
2) testing that the maximum number of events doesn't cause BPF verifier issues in debug builds,
3) lower the max events from 1024 to 128,
4) in perf stat, if there are more events than the BPF counters can support then disable BPF counter usage.
The rodata setup is factored into its own function to avoid duplicating it in the testing code.
Signed-off-by Ian Rogers <[email protected]>
Fixes b8308511f6e0 ("perf stat bperf cgroup: Increase MAX_EVENTS from 32 to 1024")
--- tools/perf/builtin-stat.c | 13 +++- tools/perf/util/bpf_counter_cgroup.c | 79 +++++++++++++++------ tools/perf/util/bpf_skel/bperf_cgroup.bpf.c | 18 +++-- 3 files changed, 76 insertions(+), 34 deletions(-)
tools/perf/builtin-stat.c +12 -1
@@ -96,6 +96,10 @@#include <perf/evlist.h> #include <internal/threadmap.h> +#ifdef HAVE_BPF_SKEL +#include "util/bpf_skel/bperf_cgroup.h" +#endif +#define DEFAULT_SEPARATOR " "#define FREEZE_ON_SMI_PATH "bus/event_source/devices/cpu/freeze_on_smi"
@@ -2852,7 +2856,14 @@ int cmd_stat(int argc, const char **argv) goto out; } } - +#ifdef HAVE_BPF_SKEL +if (target.use_bpf && + (evsel_list->core.nr_entries / nr_cgroups) >BPERF_CGROUP__MAX_EVENTS ) { +pr_warning ("Disabling BPF counters due to more events (%d) than the max (%d)\n" , + evsel_list->core.nr_entries / nr_cgroups,BPERF_CGROUP__MAX_EVENTS ); + target.use_bpf = false; + } +#endif // HAVE_BPF_SKEL evlist__warn_user_requested_cpus (evsel_list, target.cpu_list);evlist__for_each_entry (evsel_list, counter) {
tools/perf/util/bpf_counter_cgroup.c +56 -23
103 lines not on this page · raw message
tools/perf/util/bpf_skel/bperf_cgroup.bpf.c +8 -10
@@ -1,14 +1,12 @@// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) // Copyright (c) 2021 Facebook // Copyright (c) 2021 Google +#include "bperf_cgroup.h" #include "vmlinux.h" #include <bpf/bpf_helpers.h> #include <bpf/bpf_tracing.h> #include <bpf/bpf_core_read.h> -#define MAX_LEVELS 10 // max cgroup hierarchy level: arbitrary -#define MAX_EVENTS 1024 // max events per cgroup: arbitrary -// NOTE: many of map and global data will be modified before loading // from the userspace (perf tool) using the skeleton helpers.
@@ -97,7 +95,7 @@ static inline int get_cgroup_v1_idx(__u32 *cgrps, int size) cgrp =BPF_CORE_READ (p, cgroups, subsys[perf_subsys_id], cgroup); level =BPF_CORE_READ (cgrp, level); -for (cnt =0 ; i <MAX_LEVELS ; i++) { +for (cnt =0 ; i <BPERF_CGROUP__MAX_LEVELS ; i++) {__u64 cgrp_id;if (i > level)
@@ -123,7 +121,7 @@ static inline int get_cgroup_v2_idx(__u32 *cgrps, int size)__u32 *elem;int cnt; -for (cnt =0 ; i <MAX_LEVELS ; i++) { +for (cnt =0 ; i <BPERF_CGROUP__MAX_LEVELS ; i++) {__u64 cgrp_id =bpf_get_current_ancestor_cgroup_id (i);if (cgrp_id ==0 )
@@ -148,17 +146,17 @@ static int bperf_cgroup_count(void) registerint c =0 ;struct bpf_perf_event_value val, delta, *prev_val, *cgrp_val;__u32 cpu =bpf_get_smp_processor_id (); -__u32 cgrp_idx[MAX_LEVELS ]; +__u32 cgrp_idx[BPERF_CGROUP__MAX_LEVELS ];int cgrp_cnt;__u32 key, cgrp;long err;if (use_cgroup_v2) - cgrp_cnt =get_cgroup_v2_idx (cgrp_idx,MAX_LEVELS ); + cgrp_cnt =get_cgroup_v2_idx (cgrp_idx,BPERF_CGROUP__MAX_LEVELS );else - cgrp_cnt =get_cgroup_v1_idx (cgrp_idx,MAX_LEVELS ); + cgrp_cnt =get_cgroup_v1_idx (cgrp_idx,BPERF_CGROUP__MAX_LEVELS ); -for ( ; idx <MAX_EVENTS ; idx++) { +for ( ; idx <BPERF_CGROUP__MAX_EVENTS ; idx++) {if (idx == num_events)break ;
@@ -186,7 +184,7 @@ static int bperf_cgroup_count(void) delta.enabled = val.enabled - prev_val->enabled; delta.running = val.running - prev_val->running; -for (c =0 ; c <MAX_LEVELS ; c++) { +for (c =0 ; c <BPERF_CGROUP__MAX_LEVELS ; c++) {if (c == cgrp_cnt)break ;
Signature · 1 line
2.51.1.821.gb6fe4d2222-goog