[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240902200515.2103769-5-namhyung@kernel.org>
Date: Mon, 2 Sep 2024 13:05:14 -0700
From: Namhyung Kim <namhyung@...nel.org>
To: Arnaldo Carvalho de Melo <acme@...nel.org>,
Ian Rogers <irogers@...gle.com>,
Kan Liang <kan.liang@...ux.intel.com>
Cc: Jiri Olsa <jolsa@...nel.org>,
Adrian Hunter <adrian.hunter@...el.com>,
Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...nel.org>,
LKML <linux-kernel@...r.kernel.org>,
linux-perf-users@...r.kernel.org,
Song Liu <song@...nel.org>,
bpf@...r.kernel.org
Subject: [PATCH 4/5] perf lock contention: Constify control data for BPF
The control knobs set before loading BPF programs should be declared as
'const volatile' so that it can be optimized by the BPF core.
Signed-off-by: Namhyung Kim <namhyung@...nel.org>
---
tools/perf/util/bpf_lock_contention.c | 45 ++++++++++---------
.../perf/util/bpf_skel/lock_contention.bpf.c | 27 +++++------
2 files changed, 38 insertions(+), 34 deletions(-)
diff --git a/tools/perf/util/bpf_lock_contention.c b/tools/perf/util/bpf_lock_contention.c
index bc4e92c0c08b8b20..41a1ad08789511c3 100644
--- a/tools/perf/util/bpf_lock_contention.c
+++ b/tools/perf/util/bpf_lock_contention.c
@@ -46,14 +46,22 @@ int lock_contention_prepare(struct lock_contention *con)
else
bpf_map__set_max_entries(skel->maps.stacks, 1);
- if (target__has_cpu(target))
+ if (target__has_cpu(target)) {
+ skel->rodata->has_cpu = 1;
ncpus = perf_cpu_map__nr(evlist->core.user_requested_cpus);
- if (target__has_task(target))
+ }
+ if (target__has_task(target)) {
+ skel->rodata->has_task = 1;
ntasks = perf_thread_map__nr(evlist->core.threads);
- if (con->filters->nr_types)
+ }
+ if (con->filters->nr_types) {
+ skel->rodata->has_type = 1;
ntypes = con->filters->nr_types;
- if (con->filters->nr_cgrps)
+ }
+ if (con->filters->nr_cgrps) {
+ skel->rodata->has_cgroup = 1;
ncgrps = con->filters->nr_cgrps;
+ }
/* resolve lock name filters to addr */
if (con->filters->nr_syms) {
@@ -82,6 +90,7 @@ int lock_contention_prepare(struct lock_contention *con)
con->filters->addrs = addrs;
}
naddrs = con->filters->nr_addrs;
+ skel->rodata->has_addr = 1;
}
bpf_map__set_max_entries(skel->maps.cpu_filter, ncpus);
@@ -90,6 +99,16 @@ int lock_contention_prepare(struct lock_contention *con)
bpf_map__set_max_entries(skel->maps.addr_filter, naddrs);
bpf_map__set_max_entries(skel->maps.cgroup_filter, ncgrps);
+ skel->rodata->stack_skip = con->stack_skip;
+ skel->rodata->aggr_mode = con->aggr_mode;
+ skel->rodata->needs_callstack = con->save_callstack;
+ skel->rodata->lock_owner = con->owner;
+
+ if (con->aggr_mode == LOCK_AGGR_CGROUP || con->filters->nr_cgrps) {
+ if (cgroup_is_v2("perf_event"))
+ skel->rodata->use_cgroup_v2 = 1;
+ }
+
if (lock_contention_bpf__load(skel) < 0) {
pr_err("Failed to load lock-contention BPF skeleton\n");
return -1;
@@ -99,7 +118,6 @@ int lock_contention_prepare(struct lock_contention *con)
u32 cpu;
u8 val = 1;
- skel->bss->has_cpu = 1;
fd = bpf_map__fd(skel->maps.cpu_filter);
for (i = 0; i < ncpus; i++) {
@@ -112,7 +130,6 @@ int lock_contention_prepare(struct lock_contention *con)
u32 pid;
u8 val = 1;
- skel->bss->has_task = 1;
fd = bpf_map__fd(skel->maps.task_filter);
for (i = 0; i < ntasks; i++) {
@@ -125,7 +142,6 @@ int lock_contention_prepare(struct lock_contention *con)
u32 pid = evlist->workload.pid;
u8 val = 1;
- skel->bss->has_task = 1;
fd = bpf_map__fd(skel->maps.task_filter);
bpf_map_update_elem(fd, &pid, &val, BPF_ANY);
}
@@ -133,7 +149,6 @@ int lock_contention_prepare(struct lock_contention *con)
if (con->filters->nr_types) {
u8 val = 1;
- skel->bss->has_type = 1;
fd = bpf_map__fd(skel->maps.type_filter);
for (i = 0; i < con->filters->nr_types; i++)
@@ -143,7 +158,6 @@ int lock_contention_prepare(struct lock_contention *con)
if (con->filters->nr_addrs) {
u8 val = 1;
- skel->bss->has_addr = 1;
fd = bpf_map__fd(skel->maps.addr_filter);
for (i = 0; i < con->filters->nr_addrs; i++)
@@ -153,25 +167,14 @@ int lock_contention_prepare(struct lock_contention *con)
if (con->filters->nr_cgrps) {
u8 val = 1;
- skel->bss->has_cgroup = 1;
fd = bpf_map__fd(skel->maps.cgroup_filter);
for (i = 0; i < con->filters->nr_cgrps; i++)
bpf_map_update_elem(fd, &con->filters->cgrps[i], &val, BPF_ANY);
}
- /* these don't work well if in the rodata section */
- skel->bss->stack_skip = con->stack_skip;
- skel->bss->aggr_mode = con->aggr_mode;
- skel->bss->needs_callstack = con->save_callstack;
- skel->bss->lock_owner = con->owner;
-
- if (con->aggr_mode == LOCK_AGGR_CGROUP) {
- if (cgroup_is_v2("perf_event"))
- skel->bss->use_cgroup_v2 = 1;
-
+ if (con->aggr_mode == LOCK_AGGR_CGROUP)
read_all_cgroups(&con->cgroups);
- }
bpf_program__set_autoload(skel->progs.collect_lock_syms, false);
diff --git a/tools/perf/util/bpf_skel/lock_contention.bpf.c b/tools/perf/util/bpf_skel/lock_contention.bpf.c
index 52a876b4269917fe..1069bda5d733887f 100644
--- a/tools/perf/util/bpf_skel/lock_contention.bpf.c
+++ b/tools/perf/util/bpf_skel/lock_contention.bpf.c
@@ -117,21 +117,22 @@ struct mm_struct___new {
} __attribute__((preserve_access_index));
/* control flags */
-int enabled;
-int has_cpu;
-int has_task;
-int has_type;
-int has_addr;
-int has_cgroup;
-int needs_callstack;
-int stack_skip;
-int lock_owner;
-
-int use_cgroup_v2;
-int perf_subsys_id = -1;
+const volatile int has_cpu;
+const volatile int has_task;
+const volatile int has_type;
+const volatile int has_addr;
+const volatile int has_cgroup;
+const volatile int needs_callstack;
+const volatile int stack_skip;
+const volatile int lock_owner;
+const volatile int use_cgroup_v2;
/* determine the key of lock stat */
-int aggr_mode;
+const volatile int aggr_mode;
+
+int enabled;
+
+int perf_subsys_id = -1;
__u64 end_ts;
--
2.46.0.469.g59c65b2a67-goog
Powered by blists - more mailing lists