Re: [PATCH v1 1/8] perf ui browser: Don't save pointer to stack memory

[Ian Rogers]

On Tue, May 7, 2024 at 6:07 PM Arnaldo Carvalho de Melo
<arnaldo.melo@xxxxxxxxx> wrote:
>
>
>
> On Tue, May 7, 2024, 9:51 PM Ian Rogers <irogers@xxxxxxxxxx> wrote:
>>
>> On Tue, May 7, 2024 at 2:07 PM Arnaldo Carvalho de Melo <acme@xxxxxxxxxx> wrote:
>> >
>> > On Tue, May 07, 2024 at 06:04:43PM -0300, Arnaldo Carvalho de Melo wrote:
>> > > On Tue, May 07, 2024 at 01:48:28PM -0700, Ian Rogers wrote:
>> > > > On Tue, May 7, 2024 at 1:22 PM Arnaldo Carvalho de Melo <acme@xxxxxxxxxx> wrote:
>> > > > >
>> > > > > On Tue, May 07, 2024 at 05:20:59PM -0300, Arnaldo Carvalho de Melo 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 strdup-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 memory
>> > > > > > (invalid) but there will be no references to it, no?
>> > > > > >
>> > > > > > If we return to hist_browser__run() we then call ui_browser__show
>> > > > > > 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="log_path=/tmp/asan.log"; /tmp/perf/perf mem record -a
>> > > > sleep 1; /tmp/perf/perf mem report'
>> > > > I then go to the perf annotate view and quit. This triggers the asan
>> > > > error (from the log file):
>> > > > ```
>> > >
>> > > Thanks, it is indeed a bug, I'll keep that Fixes tag, people interested
>> > > in the full details can hopefully find this message going from the Link:
>> > > 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__current);
>>        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