Re: [PATCH] kernel/hung_task.c: Monitor killed tasks.

[Tetsuo Handa]

Hello, Stephen.

I want to send debug printk() patches to linux-next.git. Petr Mladek
is suggesting me to have a git tree for debug printk() patches.
But it seems that there is "git quiltimport" command, and I prefer
"subversion + quilt", and I don't have trees for sending "git pull"
requests. Therefore, just ignoring "git quiltimport" failure is fine.
What do you think?

On 2019/05/16 21:38, Tetsuo Handa wrote:
> On 2019/05/16 20:57, Petr Mladek wrote:
>> CCed Stephen to discuss linux-next related question at the bottom
>> of the mail.
>>
>> On Thu 2019-05-16 17:19:12, Tetsuo Handa wrote:
>>> On 2019/05/15 19:55, Petr Mladek wrote:
>>> But in the context of syzbot's testing where there are only 2 CPUs
>>> in the target VM (which means that only small number of threads and
>>> not so much memory) and threads get SIGKILL after 5 seconds from fork(),
>>> being unable to reach do_exit() within 10 seconds is likely a sign of
>>> something went wrong. For example, 6 out of 7 trials of a reproducer for
>>> https://syzkaller.appspot.com/bug?id=835a0b9e75b14b55112661cbc61ca8b8f0edf767
>>> resulted in "no output from test machine" rather than "task hung".
>>> This patch is revealing that such killed threads are failing to reach
>>> do_exit() because they are trapped at unkillable retry loop due to a
>>> race bug.
>>>
>>> Therefore, I would like to try this patch in linux-next.git for feasibility
>>> testing whether this patch helps finding more bugs and reproducers for such
>>> bugs, by bringing "unable to terminate threads" reports out of "no output from
>>> test machine" reports. We can add sysctl settings before sending to linux.git.
>>
>> In this case, the watchdog should get enabled on with
>> CONFIG_DEBUG_AID_FOR_SYZBOT
> 
> Since "[PATCH] printk: Monitor change of console loglevel." is one time (only
> needed until we find the reason of silence), testing on only linux-next.git
> is sufficient and it gets enabled on with CONFIG_DEBUG_AID_FOR_SYZBOT.
> 
>>
>> Also we should ask/inform Stephen about this. I am not sure
>> if he is willing to resolve eventual conflicts for these
>> syzboot-specific patches that are not upstream candidates.
>>
>> A solution might be to create sysbot-specific for-next branch
>> that Stephen might simply ignore when there are conflicts.
>> And you would be responsible for updating it.
> 
> syzbot tests not only linux-next.git but also various trees, and tests
> attempted depends on target git tree. Therefore, apart from whether we
> can introduce a kernel config option for fuzzing testing,
> "[PATCH] kernel/hung_task.c: Monitor killed tasks." is expected to be
> in linux.git. This patch will eventually become upstream candidate.
>