perf cpumap: Add reference count checking

Enabled when REFCNT_CHECKING is defined. The change adds a memory
allocated pointer that is interposed between the reference counted cpu
map at a get and freed by a put. The pointer replaces the original
perf_cpu_map struct, so use of the perf_cpu_map via APIs remains
unchanged. Any use of the cpu map without the API requires two versions,
handled via the RC_CHK_ACCESS macro.

This change is intended to catch:

 - use after put: using a cpumap after you have put it will cause a
   segv.
 - unbalanced puts: two puts for a get will result in a double free
   that can be captured and reported by tools like address sanitizer,
   including with the associated stack traces of allocation and frees.
 - missing puts: if a put is missing then the get turns into a memory
   leak that can be reported by leak sanitizer, including the stack
   trace at the point the get occurs.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Bayduraev <alexey.v.bayduraev@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Dmitriy Vyukov <dvyukov@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: German Gomez <german.gomez@arm.com>
Cc: Hao Luo <haoluo@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Miaoqian Lin <linmq006@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Riccardo Mancini <rickyman7@gmail.com>
Cc: Shunsuke Nakamura <nakamura.shun@fujitsu.com>
Cc: Song Liu <song@kernel.org>
Cc: Stephen Brennan <stephen.s.brennan@oracle.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Thomas Richter <tmricht@linux.ibm.com>,
Cc: Yury Norov <yury.norov@gmail.com>
Link: https://lore.kernel.org/lkml/20230407230405.2931830-3-irogers@google.com
[ Extracted from a larger patch ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

diff --git a/tools/lib/perf/Makefile b/tools/lib/perf/Makefile
index d8cad124e4c5de57592c1cf25cb03e22611d8e4b..3a9b2140aa048ea919c69ed2240bf0ea444dbf21 100644 (file)
--- a/tools/lib/perf/Makefile
+++ b/tools/lib/perf/Makefile
@@ -188,7 +188,7 @@ install_lib: libs
                cp -fpR $(LIBPERF_ALL) $(DESTDIR)$(libdir_SQ)
 
 HDRS := bpf_perf.h core.h cpumap.h threadmap.h evlist.h evsel.h event.h mmap.h
-INTERNAL_HDRS := cpumap.h evlist.h evsel.h lib.h mmap.h threadmap.h xyarray.h
+INTERNAL_HDRS := cpumap.h evlist.h evsel.h lib.h mmap.h rc_check.h threadmap.h xyarray.h
 
 INSTALL_HDRS_PFX := $(DESTDIR)$(prefix)/include/perf
 INSTALL_HDRS := $(addprefix $(INSTALL_HDRS_PFX)/, $(HDRS))

diff --git a/tools/lib/perf/cpumap.c b/tools/lib/perf/cpumap.c
index 27c3e73c6db29b57401df946836b73a1bc956395..1229b18bcdb168220d0db4db43a2b6b977d0a235 100644 (file)
--- a/tools/lib/perf/cpumap.c
+++ b/tools/lib/perf/cpumap.c
@@ -12,19 +12,19 @@
 
 void perf_cpu_map__set_nr(struct perf_cpu_map *map, int nr_cpus)
 {
-       map->nr = nr_cpus;
+       RC_CHK_ACCESS(map)->nr = nr_cpus;
 }
 
 struct perf_cpu_map *perf_cpu_map__alloc(int nr_cpus)
 {
-       struct perf_cpu_map *cpus = malloc(sizeof(*cpus) + sizeof(struct perf_cpu) * nr_cpus);
+       RC_STRUCT(perf_cpu_map) *cpus = malloc(sizeof(*cpus) + sizeof(struct perf_cpu) * nr_cpus);
+       struct perf_cpu_map *result;
 
-       if (cpus != NULL) {
+       if (ADD_RC_CHK(result, cpus)) {
                cpus->nr = nr_cpus;
                refcount_set(&cpus->refcnt, 1);
-
        }
-       return cpus;
+       return result;
 }
 
 struct perf_cpu_map *perf_cpu_map__dummy_new(void)
@@ -32,7 +32,7 @@ struct perf_cpu_map *perf_cpu_map__dummy_new(void)
        struct perf_cpu_map *cpus = perf_cpu_map__alloc(1);
 
        if (cpus)
-               cpus->map[0].cpu = -1;
+               RC_CHK_ACCESS(cpus)->map[0].cpu = -1;
 
        return cpus;
 }
@@ -42,21 +42,28 @@ static void cpu_map__delete(struct perf_cpu_map *map)
        if (map) {
                WARN_ONCE(refcount_read(perf_cpu_map__refcnt(map)) != 0,
                          "cpu_map refcnt unbalanced\n");
-               free(map);
+               RC_CHK_FREE(map);
        }
 }
 
 struct perf_cpu_map *perf_cpu_map__get(struct perf_cpu_map *map)
 {
-       if (map)
+       struct perf_cpu_map *result;
+
+       if (RC_CHK_GET(result, map))
                refcount_inc(perf_cpu_map__refcnt(map));
-       return map;
+
+       return result;
 }
 
 void perf_cpu_map__put(struct perf_cpu_map *map)
 {
-       if (map && refcount_dec_and_test(perf_cpu_map__refcnt(map)))
-               cpu_map__delete(map);
+       if (map) {
+               if (refcount_dec_and_test(perf_cpu_map__refcnt(map)))
+                       cpu_map__delete(map);
+               else
+                       RC_CHK_PUT(map);
+       }
 }
 
 static struct perf_cpu_map *cpu_map__default_new(void)
@@ -73,7 +80,7 @@ static struct perf_cpu_map *cpu_map__default_new(void)
                int i;
 
                for (i = 0; i < nr_cpus; ++i)
-                       cpus->map[i].cpu = i;
+                       RC_CHK_ACCESS(cpus)->map[i].cpu = i;
        }
 
        return cpus;
@@ -99,15 +106,15 @@ static struct perf_cpu_map *cpu_map__trim_new(int nr_cpus, const struct perf_cpu
        int i, j;
 
        if (cpus != NULL) {
-               memcpy(cpus->map, tmp_cpus, payload_size);
-               qsort(cpus->map, nr_cpus, sizeof(struct perf_cpu), cmp_cpu);
+               memcpy(RC_CHK_ACCESS(cpus)->map, tmp_cpus, payload_size);
+               qsort(RC_CHK_ACCESS(cpus)->map, nr_cpus, sizeof(struct perf_cpu), cmp_cpu);
                /* Remove dups */
                j = 0;
                for (i = 0; i < nr_cpus; i++) {
-                       if (i == 0 || cpus->map[i].cpu != cpus->map[i - 1].cpu)
-                               cpus->map[j++].cpu = cpus->map[i].cpu;
+                       if (i == 0 || RC_CHK_ACCESS(cpus)->map[i].cpu != RC_CHK_ACCESS(cpus)->map[i - 1].cpu)
+                               RC_CHK_ACCESS(cpus)->map[j++].cpu = RC_CHK_ACCESS(cpus)->map[i].cpu;
                }
-               cpus->nr = j;
+               perf_cpu_map__set_nr(cpus, j);
                assert(j <= nr_cpus);
        }
        return cpus;
@@ -268,20 +275,20 @@ struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx)
                .cpu = -1
        };
 
-       if (cpus && idx < cpus->nr)
-               return cpus->map[idx];
+       if (cpus && idx < RC_CHK_ACCESS(cpus)->nr)
+               return RC_CHK_ACCESS(cpus)->map[idx];
 
        return result;
 }
 
 int perf_cpu_map__nr(const struct perf_cpu_map *cpus)
 {
-       return cpus ? cpus->nr : 1;
+       return cpus ? RC_CHK_ACCESS(cpus)->nr : 1;
 }
 
 bool perf_cpu_map__empty(const struct perf_cpu_map *map)
 {
-       return map ? map->map[0].cpu == -1 : true;
+       return map ? RC_CHK_ACCESS(map)->map[0].cpu == -1 : true;
 }
 
 int perf_cpu_map__idx(const struct perf_cpu_map *cpus, struct perf_cpu cpu)
@@ -292,10 +299,10 @@ int perf_cpu_map__idx(const struct perf_cpu_map *cpus, struct perf_cpu cpu)
                return -1;
 
        low = 0;
-       high = cpus->nr;
+       high = RC_CHK_ACCESS(cpus)->nr;
        while (low < high) {
                int idx = (low + high) / 2;
-               struct perf_cpu cpu_at_idx = cpus->map[idx];
+               struct perf_cpu cpu_at_idx = RC_CHK_ACCESS(cpus)->map[idx];
 
                if (cpu_at_idx.cpu == cpu.cpu)
                        return idx;
@@ -321,7 +328,7 @@ struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map)
        };
 
        // cpu_map__trim_new() qsort()s it, cpu_map__default_new() sorts it as well.
-       return map->nr > 0 ? map->map[map->nr - 1] : result;
+       return RC_CHK_ACCESS(map)->nr > 0 ? RC_CHK_ACCESS(map)->map[RC_CHK_ACCESS(map)->nr - 1] : result;
 }
 
 /** Is 'b' a subset of 'a'. */
@@ -329,15 +336,15 @@ bool perf_cpu_map__is_subset(const struct perf_cpu_map *a, const struct perf_cpu
 {
        if (a == b || !b)
                return true;
-       if (!a || b->nr > a->nr)
+       if (!a || RC_CHK_ACCESS(b)->nr > RC_CHK_ACCESS(a)->nr)
                return false;
 
-       for (int i = 0, j = 0; i < a->nr; i++) {
-               if (a->map[i].cpu > b->map[j].cpu)
+       for (int i = 0, j = 0; i < RC_CHK_ACCESS(a)->nr; i++) {
+               if (RC_CHK_ACCESS(a)->map[i].cpu > RC_CHK_ACCESS(b)->map[j].cpu)
                        return false;
-               if (a->map[i].cpu == b->map[j].cpu) {
+               if (RC_CHK_ACCESS(a)->map[i].cpu == RC_CHK_ACCESS(b)->map[j].cpu) {
                        j++;
-                       if (j == b->nr)
+                       if (j == RC_CHK_ACCESS(b)->nr)
                                return true;
                }
        }
@@ -367,27 +374,27 @@ struct perf_cpu_map *perf_cpu_map__merge(struct perf_cpu_map *orig,
                return perf_cpu_map__get(other);
        }
 
-       tmp_len = orig->nr + other->nr;
+       tmp_len = RC_CHK_ACCESS(orig)->nr + RC_CHK_ACCESS(other)->nr;
        tmp_cpus = malloc(tmp_len * sizeof(struct perf_cpu));
        if (!tmp_cpus)
                return NULL;
 
        /* Standard merge algorithm from wikipedia */
        i = j = k = 0;
-       while (i < orig->nr && j < other->nr) {
-               if (orig->map[i].cpu <= other->map[j].cpu) {
-                       if (orig->map[i].cpu == other->map[j].cpu)
+       while (i < RC_CHK_ACCESS(orig)->nr && j < RC_CHK_ACCESS(other)->nr) {
+               if (RC_CHK_ACCESS(orig)->map[i].cpu <= RC_CHK_ACCESS(other)->map[j].cpu) {
+                       if (RC_CHK_ACCESS(orig)->map[i].cpu == RC_CHK_ACCESS(other)->map[j].cpu)
                                j++;
-                       tmp_cpus[k++] = orig->map[i++];
+                       tmp_cpus[k++] = RC_CHK_ACCESS(orig)->map[i++];
                } else
-                       tmp_cpus[k++] = other->map[j++];
+                       tmp_cpus[k++] = RC_CHK_ACCESS(other)->map[j++];
        }
 
-       while (i < orig->nr)
-               tmp_cpus[k++] = orig->map[i++];
+       while (i < RC_CHK_ACCESS(orig)->nr)
+               tmp_cpus[k++] = RC_CHK_ACCESS(orig)->map[i++];
 
-       while (j < other->nr)
-               tmp_cpus[k++] = other->map[j++];
+       while (j < RC_CHK_ACCESS(other)->nr)
+               tmp_cpus[k++] = RC_CHK_ACCESS(other)->map[j++];
        assert(k <= tmp_len);
 
        merged = cpu_map__trim_new(k, tmp_cpus);

diff --git a/tools/lib/perf/include/internal/cpumap.h b/tools/lib/perf/include/internal/cpumap.h
index 1e840dd53a11adc66b3934b71d37ceb3b0d6f8b0..49649eb51ce45e70db3bfd1870edb7fd21378efd 100644 (file)
--- a/tools/lib/perf/include/internal/cpumap.h
+++ b/tools/lib/perf/include/internal/cpumap.h
@@ -4,6 +4,7 @@
 
 #include <linux/refcount.h>
 #include <perf/cpumap.h>
+#include <internal/rc_check.h>
 
 /**
  * A sized, reference counted, sorted array of integers representing CPU
@@ -12,7 +13,7 @@
  * gaps if CPU numbers were used. For events associated with a pid, rather than
  * a CPU, a single dummy map with an entry of -1 is used.
  */
-struct perf_cpu_map {
+DECLARE_RC_STRUCT(perf_cpu_map) {
        refcount_t      refcnt;
        /** Length of the map array. */
        int             nr;
@@ -32,6 +33,6 @@ void perf_cpu_map__set_nr(struct perf_cpu_map *map, int nr_cpus);
 
 static inline refcount_t *perf_cpu_map__refcnt(struct perf_cpu_map *map)
 {
-       return &map->refcnt;
+       return &RC_CHK_ACCESS(map)->refcnt;
 }
 #endif /* __LIBPERF_INTERNAL_CPUMAP_H */

diff --git a/tools/perf/util/cpumap.c b/tools/perf/util/cpumap.c
index 240557de129e7cd95e766f1007132326c670ad1e..75d9c73e018414226f34fa717b56838aa7da9226 100644 (file)
--- a/tools/perf/util/cpumap.c
+++ b/tools/perf/util/cpumap.c
@@ -77,9 +77,9 @@ static struct perf_cpu_map *cpu_map__from_entries(const struct perf_record_cpu_m
                         * otherwise it would become 65535.
                         */
                        if (data->cpus_data.cpu[i] == (u16) -1)
-                               map->map[i].cpu = -1;
+                               RC_CHK_ACCESS(map)->map[i].cpu = -1;
                        else
-                               map->map[i].cpu = (int) data->cpus_data.cpu[i];
+                               RC_CHK_ACCESS(map)->map[i].cpu = (int) data->cpus_data.cpu[i];
                }
        }
 
@@ -107,7 +107,7 @@ static struct perf_cpu_map *cpu_map__from_mask(const struct perf_record_cpu_map_
 
                perf_record_cpu_map_data__read_one_mask(data, i, local_copy);
                for_each_set_bit(cpu, local_copy, 64)
-                       map->map[j++].cpu = cpu + cpus_per_i;
+                       RC_CHK_ACCESS(map)->map[j++].cpu = cpu + cpus_per_i;
        }
        return map;
 
@@ -124,11 +124,11 @@ static struct perf_cpu_map *cpu_map__from_range(const struct perf_record_cpu_map
                return NULL;
 
        if (data->range_cpu_data.any_cpu)
-               map->map[i++].cpu = -1;
+               RC_CHK_ACCESS(map)->map[i++].cpu = -1;
 
        for (int cpu = data->range_cpu_data.start_cpu; cpu <= data->range_cpu_data.end_cpu;
             i++, cpu++)
-               map->map[i].cpu = cpu;
+               RC_CHK_ACCESS(map)->map[i].cpu = cpu;
 
        return map;
 }
@@ -164,7 +164,7 @@ struct perf_cpu_map *perf_cpu_map__empty_new(int nr)
 
        if (cpus != NULL) {
                for (int i = 0; i < nr; i++)
-                       cpus->map[i].cpu = -1;
+                       RC_CHK_ACCESS(cpus)->map[i].cpu = -1;
        }
 
        return cpus;

diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 760c848c9fa27728f8608c9c2a646efb042aa09e..65d0eb9a740a91066b513c050b15a24685430c9b 100644 (file)
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -2015,9 +2015,9 @@ int perf_pmu__cpus_match(struct perf_pmu *pmu, struct perf_cpu_map *cpus,
 
        perf_cpu_map__for_each_cpu(cpu, i, cpus) {
                if (!perf_cpu_map__has(pmu_cpus, cpu))
-                       unmatched_cpus->map[unmatched_nr++] = cpu;
+                       RC_CHK_ACCESS(unmatched_cpus)->map[unmatched_nr++] = cpu;
                else
-                       matched_cpus->map[matched_nr++] = cpu;
+                       RC_CHK_ACCESS(matched_cpus)->map[matched_nr++] = cpu;
        }
 
        perf_cpu_map__set_nr(unmatched_cpus, unmatched_nr);