Received-SPF: pass (google.com: domain of linux-kernel+bounces-172583-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240507183545.1236093-1-irogers@google.com> <20240507183545.1236093-2-irogers@google.com>
 <ZjqNKPgWR7mBFaV4@x1> <ZjqNjrJ2ElrT11iB@x1> <CAP-5=fUsQwKsCi3us+dp-Tj+PayNPrYTqTQeo-YLbvSuOt1=9w@mail.gmail.com>
 <ZjqXa25BnFncJmw-@x1> <ZjqYD-q7EqQW6hVE@x1> <CAP-5=fVRBXv0H3vRNYO_uwWsNGvBxWVXd78UOu0QvGK04+bERw@mail.gmail.com>
 <CA+JHD90W7PLBx=SEL9+7-_=LkjaMu4YM1S3kJ2oSkAYoHE7hPw@mail.gmail.com>
In-Reply-To: <CA+JHD90W7PLBx=SEL9+7-_=LkjaMu4YM1S3kJ2oSkAYoHE7hPw@mail.gmail.com>
From: Ian Rogers <irogers@google.com>
Date: Tue, 7 May 2024 18:17:47 -0700
Message-ID: <CAP-5=fU1E4xY-DEpSh0ThGj+y4hmGHD9sMF_FPFBpAUuK6Pt3Q@mail.gmail.com>
Subject: Re: [PATCH v1 1/8] perf ui browser: Don't save pointer to stack memory
To: Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>, Peter Zijlstra <peterz@infradead.org>, 
	Ingo Molnar <mingo@redhat.com>, Namhyung Kim <namhyung@kernel.org>, 
	Mark Rutland <mark.rutland@arm.com>, 
	Alexander Shishkin <alexander.shishkin@linux.intel.com>, Jiri Olsa <jolsa@kernel.org>, 
	Adrian Hunter <adrian.hunter@intel.com>, Kan Liang <kan.liang@linux.intel.com>, 
	Oliver Upton <oliver.upton@linux.dev>, James Clark <james.clark@arm.com>, 
	Tim Chen <tim.c.chen@linux.intel.com>, Yicong Yang <yangyicong@hisilicon.com>, 
	K Prateek Nayak <kprateek.nayak@amd.com>, Yanteng Si <siyanteng@loongson.cn>, 
	Sun Haiyong <sunhaiyong@loongson.cn>, Kajol Jain <kjain@linux.ibm.com>, 
	Ravi Bangoria <ravi.bangoria@amd.com>, Li Dong <lidong@vivo.com>, Paran Lee <p4ranlee@gmail.com>, 
	Ben Gainey <ben.gainey@arm.com>, Andi Kleen <ak@linux.intel.com>, 
	Athira Rajeev <atrajeev@linux.vnet.ibm.com>, 
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>, 
	linux-perf-users <linux-perf-users@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, May 7, 2024 at 6:07=E2=80=AFPM Arnaldo Carvalho de Melo
<arnaldo.melo@gmail.com> wrote:
>
>
>
> On Tue, May 7, 2024, 9:51 PM Ian Rogers <irogers@google.com> wrote:
>>
>> On Tue, May 7, 2024 at 2:07=E2=80=AFPM Arnaldo Carvalho de Melo <acme@ke=
rnel.org> wrote:
>> >
>> > On Tue, May 07, 2024 at 06:04:43PM -0300, Arnaldo Carvalho de Melo wro=
te:
>> > > On Tue, May 07, 2024 at 01:48:28PM -0700, Ian Rogers wrote:
>> > > > On Tue, May 7, 2024 at 1:22=E2=80=AFPM Arnaldo Carvalho de Melo <a=
cme@kernel.org> wrote:
>> > > > >
>> > > > > On Tue, May 07, 2024 at 05:20:59PM -0300, Arnaldo Carvalho de Me=
lo wrote:
>> > > > > > On Tue, May 07, 2024 at 11:35:38AM -0700, Ian Rogers wrote:
>> > > > > > > ui_browser__show is capturing the input title that is stack =
allocated
>> > > > > > > memory in hist_browser__run. Avoid a use after return by str=
dup-ing
>> > > > > > > the string.
>> > > > > >
>> > > > > > But everything happens in that context, i.e. hist_brower__run(=
) will
>> > > > > > call ui_browser__ methods and then exit.
>> > > > > >
>> > > > > > We end up having browser->title pointing to returned stack mem=
ory
>> > > > > > (invalid) but there will be no references to it, no?
>> > > > > >
>> > > > > > If we return to hist_browser__run() we then call ui_browser__s=
how
>> > > > > > passing a new title, for "live" stack memory, rinse repeat. Or=
 have you
>> > > > > > noticed an actual use-after-"free"?
>> > > > >
>> > > > > And I'll take the patch, I'm just trying to figure it out if it =
fixed a
>> > > > > real bug or if it just makes the code more future proof, i.e. to=
 avoid
>> > > > > us adding code that actually uses invalid stack memory.
>> > > >
>> > > > My command line using tui is:
>> > > > $ sudo bash -c 'rm /tmp/asan.log*; export
>> > > > ASAN_OPTIONS=3D"log_path=3D/tmp/asan.log"; /tmp/perf/perf mem reco=
rd -a
>> > > > sleep 1; /tmp/perf/perf mem report'
>> > > > I then go to the perf annotate view and quit. This triggers the as=
an
>> > > > error (from the log file):
>> > > > ```
>> > >
>> > > Thanks, it is indeed a bug, I'll keep that Fixes tag, people interes=
ted
>> > > in the full details can hopefully find this message going from the L=
ink:
>> > > tag.
>> >
>> > Nah, I added your explanation to the cset log message.
>>
>>
>> Okay, I found I needed this to avoid a segv introduced by this patch:
>> ```
>> diff --git a/tools/perf/ui/browser.c b/tools/perf/ui/browser.c
>> index c4cdf2ea69b7..19503e838738 100644
>> --- a/tools/perf/ui/browser.c
>> +++ b/tools/perf/ui/browser.c
>> @@ -203,7 +203,7 @@ void ui_browser__refresh_dimensions(struct
>> ui_browser *browser)
>> void ui_browser__handle_resize(struct ui_browser *browser)
>> {
>>        ui__refresh_dimensions(false);
>> -       ui_browser__show(browser, browser->title, ui_helpline__current);
>> +       ui_browser__show(browser, browser->title ?: "", ui_helpline__cur=
rent);
>>        ui_browser__refresh(browser);
>> }
>> ```
>> I also found a use-after-free issue with patch 5. I'll send a v2.
>>
>> Please send a fix, it's already in perf-tools-next.
>

Okay. It looks like you accidentally pushed tmp.perf-tools.next, that
is a .next rather than a -next (dot not dash). I'll work from
perf-tools-next.

Thanks,
Ian

> Thanks,
>
> - Arnaldo
>
>> Thanks,
>> Ian
>>
>> > Thanks,
>> >
>> > - Arnaldo