Em Tue, Jul 18, 2017 at 06:18:36PM -0700, David Carrillo-Cisneros escreveu:
> Atomic reference counters were replaced by refcount_t in
> commit 79c5fe6db8c7 ("perf/core: Fix error handling in perf_event_alloc()")
>
> In util/cgroup.c atomic_inc was replaced by refcount_inc, but the latter
> is not mean to initiliaze refcounts with zero value. Add a path
> to initialize cgrp->refcnt == 0 using refcount_set.
>
> Before this patch:
>
> $ perf stat -e cycles -C 0 -G /
> perf_before: /usr/local/.../tools/include/linux/refcount.h:108:
> refcount_inc: Assertion `!(!refcount_inc_not_zero(r))' failed.
> Aborted (core dumped)
>
> After this patch:
>
> $ perf stat -e cycles -C 0 -G /
> Performance counter stats for 'CPU(s) 0':
> 17,516,664 cycles /
Ok, so this one was also reported by Brendan and I came up with an
alternative patch, that follows the usual sequence of steps, see below.
- Arnaldo
commit b13ca9843c4ea07c5a1dbf0295986b5dfeb6ef6f
Author: Arnaldo Carvalho de Melo <arnaldo.m...@gmail.com>
Date: Tue Jul 18 20:20:19 2017 -0300
perf cgroup: Fix refcount usage
When converting from atomic_t to refcount_t we didn't follow the usual
step of initializing it to one before taking any new reference, which
trips over checking if taking a reference for a freed refcount_t, fix
it.
Brendan's report:
---
It's 4.12-rc7, with node v4.4.1. I'm building 4.13-rc1 now, as I hit
what I think is another unrelated perf bug and I'm starting to wonder
what else is broken on that version:
(root) /mnt/src/linux-4.12-rc7/tools/perf # ./perf record -F 99 -a -e
cpu-clock
--cgroup=docker/f9e9d5df065b14646e8a11edc837a13877fd90c171137b2ba3feb67a0201cb65
-g
perf: /mnt/src/linux-4.12-rc7/tools/include/linux/refcount.h:108:
refcount_inc: Assertion `!(!refcount_inc_not_zero(r))' failed.
Aborted
that used to work...
---
Testing it:
Before:
# perf stat -e cycles -C 0 --cgroup /
perf: /home/acme/git/linux/tools/include/linux/refcount.h:108:
refcount_inc: Assertion `!(!refcount_inc_not_zero(r))' failed.
Aborted (core dumped)
#
After:
# perf stat -e cycles -C 0 --cgroup /
^C
Performance counter stats for 'CPU(s) 0':
132,081,393 cycles /
2.492942763 seconds time elapsed
#
Reported-by: Brendan Gregg <brendan.d.gr...@gmail.com>
Acked-by: Elena Reshetova <elena.reshet...@intel.com>
Cc: Alexander Shishkin <alexander.shish...@linux.intel.com>
Cc: David Ahern <dsah...@gmail.com>
Cc: David Carrillo-Cisneros <davi...@google.com>
Cc: Kees Kook <keesc...@chromium.org>
Cc: Krister Johansen <k...@templeofstupid.com>
Cc: Paul Turner <p...@google.com>
Cc: Peter Zijlstra <pet...@infradead.org>
Cc: Stephane Eranian <eran...@google.com>
Cc: Sudeep Holla <sudeep.ho...@arm.com>
Cc: Thomas-Mich Richter <tmri...@linux.vnet.ibm.com>
Cc: Wang Nan <wangn...@huawei.com>
Fixes: 79c5fe6db8c7 ("perf cgroup: Convert cgroup_sel.refcnt from atomic_t
to refcount_t")
Link: http://lkml.kernel.org/n/tip-l7ovfblq14ip2i08m1g0f...@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <a...@redhat.com>
diff --git a/tools/perf/util/cgroup.c b/tools/perf/util/cgroup.c
index 03347748f3fa..0e77bc9e5f3c 100644
--- a/tools/perf/util/cgroup.c
+++ b/tools/perf/util/cgroup.c
@@ -98,8 +98,10 @@ static int add_cgroup(struct perf_evlist *evlist, char *str)
cgrp = counter->cgrp;
if (!cgrp)
continue;
- if (!strcmp(cgrp->name, str))
+ if (!strcmp(cgrp->name, str)) {
+ refcount_inc(&cgrp->refcnt);
break;
+ }
cgrp = NULL;
}
@@ -110,6 +112,7 @@ static int add_cgroup(struct perf_evlist *evlist, char *str)
return -1;
cgrp->name = str;
+ refcount_set(&cgrp->refcnt, 1);
cgrp->fd = open_cgroup(str);
if (cgrp->fd == -1) {
@@ -128,12 +131,11 @@ static int add_cgroup(struct perf_evlist *evlist, char
*str)
goto found;
n++;
}
- if (refcount_read(&cgrp->refcnt) == 0)
+ if (refcount_dec_and_test(&cgrp->refcnt))
free(cgrp);
return -1;
found:
- refcount_inc(&cgrp->refcnt);
counter->cgrp = cgrp;
return 0;
}
