linux-kernel - Re: [GIT PULL||RFC 00/11] perf library and regression testing improvements

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <AANLkTik=kHNPPbAo7ynD4AY68Vd3t1bxW_2stZx2QAFH@mail.gmail.com>
Date:	Tue, 4 Jan 2011 16:10:17 +0100
From:	Stephane Eranian <eranian@...gle.com>
To:	Arnaldo Carvalho de Melo <acme@...radead.org>
Cc:	linux-kernel@...r.kernel.org,
	Frederic Weisbecker <fweisbec@...il.com>,
	Han Pingtian <phan@...hat.com>, Mike Galbraith <efault@....de>,
	Paul Mackerras <paulus@...ba.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Tom Zanussi <tzanussi@...il.com>,
	Thomas Gleixner <tglx@...utronix.de>, mingo@...e.hu
Subject: Re: [GIT PULL||RFC 00/11] perf library and regression testing improvements

Ok, The problem comes both files picking up a different verison for
perf_event.h.

In my case, util/evsel.c was using /usr/include/linux/perf_event.h whereas
builtin-stat.c was using ../../include/linux/perf_event.h. Both have a different
struct perf_event_attr.

When I remove the /usr/include/linux/perf_event.h file, then I cannot compile
perf anymore:
In file included from perf.c:15:
util/parse-events.h:7:30: error: linux/perf_event.h: No such file or directory

Looks like something changed in the Makefile and the util modules don't know
where to pickup perf_event.h


On Tue, Jan 4, 2011 at 3:59 PM, Stephane Eranian <eranian@...gle.com> wrote:
> The issue is related to a mismatch in the data structure layout
> between builtin-stat.c and util/evsel.c. The offset of the counts fields
> is different between the two files. Getting closer.....
>
> On Tue, Jan 4, 2011 at 3:46 PM, Stephane Eranian <eranian@...gle.com> wrote:
>> On Tue, Jan 4, 2011 at 3:36 PM, Arnaldo Carvalho de Melo
>> <acme@...radead.org> wrote:
>>> Em Tue, Jan 04, 2011 at 03:33:38PM +0100, Stephane Eranian escreveu:
>>>> No improvement with this patch.
>>>
>>> I was guessing you were using it somehow without calling
>>> perf_evsel__alloc_counts, now I guess you're just using what is in
>>> tip/perf/core, no local changes, right?
>>>
>> I am using:
>> git://git.eu.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip.git
>>
>>> Next step: there are anonymous structs inside anonymous unions, what
>>> compiler/distro?
>> Yes, I am looking at those right now.
>>>
>> gcc 4.4.3 or gcc 4.2.4 (ubuntu hardy or 10.04).
>>
>>> - Arnaldo
>>>
>>>> On Tue, Jan 4, 2011 at 3:27 PM, Arnaldo Carvalho de Melo
>>>> <acme@...radead.org> wrote:
>>>> > Em Tue, Jan 04, 2011 at 12:19:58PM -0200, Arnaldo Carvalho de Melo escreveu:
>>>> >> Em Tue, Jan 04, 2011 at 03:09:08PM +0100, Stephane Eranian escreveu:
>>>> >> > Arnaldo,
>>>> >> > Looks like what's wrong is not ps but count:
>>>> >> >
>>>> >> > Program received signal SIGSEGV, Segmentation fault.
>>>> >> > [Switching to Thread 0x7f96967fd6e0 (LWP 5156)]
>>>> >> > 0x0000000000412b58 in read_counter_aggr (counter=0x7d4820) at builtin-stat.c:206
>>>> >> > 206                 update_stats(&ps->res_stats[i], count[i]);
>>>> >> > (gdb) print ps
>>>> >> > $1 = (struct perf_stat *) 0x7d48a0
>>>> >> > (gdb) print *ps
>>>> >> > $2 = {res_stats = {{n = 0, mean = 0, M2 = 0}, {n = 0, mean = 0, M2 =
>>>> >> > 0}, {n = 0, mean = 0, M2 = 0}}}
>>>> >> > (gdb) print count
>>>> >> > $3 = (u64 *) 0x12
>>>> >> > (gdb) print *count
>>>> >> > Cannot access memory at address 0x12
>>>> >> > (gdb) print count
>>>> >> > $4 = (u64 *) 0x12
>>>> >>
>>>> >> Count is:
>>>> >>
>>>> >> u64 *count = counter->counts->aggr.values;
>>>> >
>>>> > Can you try with this patch?
>>>> >
>>>> > diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
>>>> > index 02b2d80..7876b11 100644
>>>> > --- a/tools/perf/builtin-stat.c
>>>> > +++ b/tools/perf/builtin-stat.c
>>>> > @@ -196,12 +196,14 @@ static inline int nsec_counter(struct perf_evsel *evsel)
>>>> >  static int read_counter_aggr(struct perf_evsel *counter)
>>>> >  {
>>>> >        struct perf_stat *ps = counter->priv;
>>>> > -       u64 *count = counter->counts->aggr.values;
>>>> > +       u64 *count;
>>>> >        int i;
>>>> >
>>>> >        if (__perf_evsel__read(counter, cpus->nr, threads->nr, scale) < 0)
>>>> >                return -1;
>>>> >
>>>> > +       count = counter->counts->aggr.values;
>>>> > +
>>>> >        for (i = 0; i < 3; i++)
>>>> >                update_stats(&ps->res_stats[i], count[i]);
>>>> >
>>>> > diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
>>>> > index d337761..26962ea 100644
>>>> > --- a/tools/perf/util/evsel.c
>>>> > +++ b/tools/perf/util/evsel.c
>>>> > @@ -89,8 +89,12 @@ int __perf_evsel__read(struct perf_evsel *evsel,
>>>> >  {
>>>> >        size_t nv = scale ? 3 : 1;
>>>> >        int cpu, thread;
>>>> > -       struct perf_counts_values *aggr = &evsel->counts->aggr, count;
>>>> > +       struct perf_counts_values *aggr, count;
>>>> >
>>>> > +       if (evsel->counts == NULL && perf_evsel__alloc_counts(evsel, ncpus) < 0)
>>>> > +               return -ENOMEM;
>>>> > +
>>>> > +       aggr = &evsel->counts->aggr;
>>>> >        aggr->val = 0;
>>>> >
>>>> >        for (cpu = 0; cpu < ncpus; cpu++) {
>>>> >
>>>
>>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/