From 8381cdd0e32dd748bd34ca3ace476949948bd793 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 18 Jan 2017 14:14:56 +0900 Subject: [PATCH 1/4] perf diff: Fix segfault on 'perf diff -o N' option The -o/--order option is to select column number to sort a diff result. It does the job by adding a hpp field at the beginning of the sort list. But it should not be added to the output field list as it has no callbacks required by a output field. During the setup_sorting(), the perf_hpp__setup_output_field() appends the given sort keys to the output field if it's not there already. Originally it was checked by fmt->list being non-empty. But commit 3f931f2c4274 ("perf hists: Make hpp setup function generic") changed it to check the ->equal callback. Anyways, we don't need to add the pseudo hpp field to the output field list since it won't be used for output. So just skip fields if they have no ->color or ->entry callbacks. Signed-off-by: Namhyung Kim Acked-by: Jiri Olsa Cc: Peter Zijlstra Fixes: 3f931f2c4274 ("perf hists: Make hpp setup function generic") Link: http://lkml.kernel.org/r/20170118051457.30946-1-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/hist.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c index 37388397b5bc..4ec79b2f9416 100644 --- a/tools/perf/ui/hist.c +++ b/tools/perf/ui/hist.c @@ -560,6 +560,10 @@ void perf_hpp__setup_output_field(struct perf_hpp_list *list) perf_hpp_list__for_each_sort_list(list, fmt) { struct perf_hpp_fmt *pos; + /* skip sort-only fields ("sort_compute" in perf diff) */ + if (!fmt->entry && !fmt->color) + continue; + perf_hpp_list__for_each_format(list, pos) { if (fmt_equal(fmt, pos)) goto next; From a1c9f97f0b64e6337d9cfcc08c134450934fdd90 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 18 Jan 2017 14:14:57 +0900 Subject: [PATCH 2/4] perf diff: Fix -o/--order option behavior (again) Commit 21e6d8428664 ("perf diff: Use perf_hpp__register_sort_field interface") changed list_add() to perf_hpp__register_sort_field(). This resulted in a behavior change since the field was added to the tail instead of the head. So the -o option is mostly ignored due to its order in the list. This patch fixes it by adding perf_hpp__prepend_sort_field(). Signed-off-by: Namhyung Kim Acked-by: Jiri Olsa Cc: Peter Zijlstra Fixes: 21e6d8428664 ("perf diff: Use perf_hpp__register_sort_field interface") Link: http://lkml.kernel.org/r/20170118051457.30946-2-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-diff.c | 2 +- tools/perf/ui/hist.c | 6 ++++++ tools/perf/util/hist.h | 7 +++++++ 3 files changed, 14 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index 9ff0db4e2d0c..933aeec46f4a 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -1199,7 +1199,7 @@ static int ui_init(void) BUG_ON(1); } - perf_hpp__register_sort_field(fmt); + perf_hpp__prepend_sort_field(fmt); return 0; } diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c index 4ec79b2f9416..18cfcdc90356 100644 --- a/tools/perf/ui/hist.c +++ b/tools/perf/ui/hist.c @@ -521,6 +521,12 @@ void perf_hpp_list__register_sort_field(struct perf_hpp_list *list, list_add_tail(&format->sort_list, &list->sorts); } +void perf_hpp_list__prepend_sort_field(struct perf_hpp_list *list, + struct perf_hpp_fmt *format) +{ + list_add(&format->sort_list, &list->sorts); +} + void perf_hpp__column_unregister(struct perf_hpp_fmt *format) { list_del(&format->list); diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index d4b6514eeef5..28c216e3d5b7 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -283,6 +283,8 @@ void perf_hpp_list__column_register(struct perf_hpp_list *list, struct perf_hpp_fmt *format); void perf_hpp_list__register_sort_field(struct perf_hpp_list *list, struct perf_hpp_fmt *format); +void perf_hpp_list__prepend_sort_field(struct perf_hpp_list *list, + struct perf_hpp_fmt *format); static inline void perf_hpp__column_register(struct perf_hpp_fmt *format) { @@ -294,6 +296,11 @@ static inline void perf_hpp__register_sort_field(struct perf_hpp_fmt *format) perf_hpp_list__register_sort_field(&perf_hpp_list, format); } +static inline void perf_hpp__prepend_sort_field(struct perf_hpp_fmt *format) +{ + perf_hpp_list__prepend_sort_field(&perf_hpp_list, format); +} + #define perf_hpp_list__for_each_format(_list, format) \ list_for_each_entry(format, &(_list)->fields, list) From aa33b9b9a2ebb00d33c83a5312d4fbf2d5aeba36 Mon Sep 17 00:00:00 2001 From: Krister Johansen Date: Thu, 5 Jan 2017 22:23:31 -0800 Subject: [PATCH 3/4] perf callchain: Reference count maps If dso__load_kcore frees all of the existing maps, but one has already been attached to a callchain cursor node, then we can get a SIGSEGV in any function that happens to try to use this invalid cursor. Use the existing map refcount mechanism to forestall cleanup of a map until the cursor iterates past the node. Signed-off-by: Krister Johansen Tested-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Masami Hiramatsu Cc: Namhyung Kim Cc: stable@kernel.org Fixes: 84c2cafa2889 ("perf tools: Reference count struct map") Link: http://lkml.kernel.org/r/20170106062331.GB2707@templeofstupid.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/callchain.c | 11 +++++++++-- tools/perf/util/callchain.h | 6 ++++++ tools/perf/util/hist.c | 7 +++++++ 3 files changed, 22 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/callchain.c b/tools/perf/util/callchain.c index 42922512c1c6..8b610dd9e2f6 100644 --- a/tools/perf/util/callchain.c +++ b/tools/perf/util/callchain.c @@ -437,7 +437,7 @@ fill_node(struct callchain_node *node, struct callchain_cursor *cursor) } call->ip = cursor_node->ip; call->ms.sym = cursor_node->sym; - call->ms.map = cursor_node->map; + call->ms.map = map__get(cursor_node->map); if (cursor_node->branch) { call->branch_count = 1; @@ -477,6 +477,7 @@ add_child(struct callchain_node *parent, list_for_each_entry_safe(call, tmp, &new->val, list) { list_del(&call->list); + map__zput(call->ms.map); free(call); } free(new); @@ -761,6 +762,7 @@ merge_chain_branch(struct callchain_cursor *cursor, list->ms.map, list->ms.sym, false, NULL, 0, 0); list_del(&list->list); + map__zput(list->ms.map); free(list); } @@ -811,7 +813,8 @@ int callchain_cursor_append(struct callchain_cursor *cursor, } node->ip = ip; - node->map = map; + map__zput(node->map); + node->map = map__get(map); node->sym = sym; node->branch = branch; node->nr_loop_iter = nr_loop_iter; @@ -1142,11 +1145,13 @@ static void free_callchain_node(struct callchain_node *node) list_for_each_entry_safe(list, tmp, &node->parent_val, list) { list_del(&list->list); + map__zput(list->ms.map); free(list); } list_for_each_entry_safe(list, tmp, &node->val, list) { list_del(&list->list); + map__zput(list->ms.map); free(list); } @@ -1210,6 +1215,7 @@ int callchain_node__make_parent_list(struct callchain_node *node) goto out; *new = *chain; new->has_children = false; + map__get(new->ms.map); list_add_tail(&new->list, &head); } parent = parent->parent; @@ -1230,6 +1236,7 @@ int callchain_node__make_parent_list(struct callchain_node *node) out: list_for_each_entry_safe(chain, new, &head, list) { list_del(&chain->list); + map__zput(chain->ms.map); free(chain); } return -ENOMEM; diff --git a/tools/perf/util/callchain.h b/tools/perf/util/callchain.h index 35c8e379530f..4f4b60f1558a 100644 --- a/tools/perf/util/callchain.h +++ b/tools/perf/util/callchain.h @@ -5,6 +5,7 @@ #include #include #include "event.h" +#include "map.h" #include "symbol.h" #define HELP_PAD "\t\t\t\t" @@ -184,8 +185,13 @@ int callchain_merge(struct callchain_cursor *cursor, */ static inline void callchain_cursor_reset(struct callchain_cursor *cursor) { + struct callchain_cursor_node *node; + cursor->nr = 0; cursor->last = &cursor->first; + + for (node = cursor->first; node != NULL; node = node->next) + map__zput(node->map); } int callchain_cursor_append(struct callchain_cursor *cursor, u64 ip, diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index 6770a9645609..7d1b7d33e644 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c @@ -1,6 +1,7 @@ #include "util.h" #include "build-id.h" #include "hist.h" +#include "map.h" #include "session.h" #include "sort.h" #include "evlist.h" @@ -1019,6 +1020,10 @@ int hist_entry_iter__add(struct hist_entry_iter *iter, struct addr_location *al, int max_stack_depth, void *arg) { int err, err2; + struct map *alm = NULL; + + if (al && al->map) + alm = map__get(al->map); err = sample__resolve_callchain(iter->sample, &callchain_cursor, &iter->parent, iter->evsel, al, max_stack_depth); @@ -1058,6 +1063,8 @@ out: if (!err) err = err2; + map__put(alm); + return err; } From 451d24d1e5f40bad000fa9abe36ddb16fc9928cb Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 31 Jan 2017 11:27:10 +0100 Subject: [PATCH 4/4] perf/core: Fix crash in perf_event_read() Alexei had his box explode because doing read() on a package (rapl/uncore) event that isn't currently scheduled in ends up doing an out-of-bounds load. Rework the code to more explicitly deal with event->oncpu being -1. Reported-by: Alexei Starovoitov Tested-by: Alexei Starovoitov Tested-by: David Carrillo-Cisneros Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: eranian@google.com Fixes: d6a2f9035bfc ("perf/core: Introduce PMU_EV_CAP_READ_ACTIVE_PKG") Link: http://lkml.kernel.org/r/20170131102710.GL6515@twins.programming.kicks-ass.net Signed-off-by: Ingo Molnar --- kernel/events/core.c | 25 +++++++++++++++---------- 1 file changed, 15 insertions(+), 10 deletions(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index e5aaa806702d..e235bb991bdd 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3487,14 +3487,15 @@ struct perf_read_data { int ret; }; -static int find_cpu_to_read(struct perf_event *event, int local_cpu) +static int __perf_event_read_cpu(struct perf_event *event, int event_cpu) { - int event_cpu = event->oncpu; u16 local_pkg, event_pkg; if (event->group_caps & PERF_EV_CAP_READ_ACTIVE_PKG) { - event_pkg = topology_physical_package_id(event_cpu); - local_pkg = topology_physical_package_id(local_cpu); + int local_cpu = smp_processor_id(); + + event_pkg = topology_physical_package_id(event_cpu); + local_pkg = topology_physical_package_id(local_cpu); if (event_pkg == local_pkg) return local_cpu; @@ -3624,7 +3625,7 @@ u64 perf_event_read_local(struct perf_event *event) static int perf_event_read(struct perf_event *event, bool group) { - int ret = 0, cpu_to_read, local_cpu; + int event_cpu, ret = 0; /* * If event is enabled and currently active on a CPU, update the @@ -3637,21 +3638,25 @@ static int perf_event_read(struct perf_event *event, bool group) .ret = 0, }; - local_cpu = get_cpu(); - cpu_to_read = find_cpu_to_read(event, local_cpu); - put_cpu(); + event_cpu = READ_ONCE(event->oncpu); + if ((unsigned)event_cpu >= nr_cpu_ids) + return 0; + + preempt_disable(); + event_cpu = __perf_event_read_cpu(event, event_cpu); /* * Purposely ignore the smp_call_function_single() return * value. * - * If event->oncpu isn't a valid CPU it means the event got + * If event_cpu isn't a valid CPU it means the event got * scheduled out and that will have updated the event count. * * Therefore, either way, we'll have an up-to-date event count * after this. */ - (void)smp_call_function_single(cpu_to_read, __perf_event_read, &data, 1); + (void)smp_call_function_single(event_cpu, __perf_event_read, &data, 1); + preempt_enable(); ret = data.ret; } else if (event->state == PERF_EVENT_STATE_INACTIVE) { struct perf_event_context *ctx = event->ctx;