Re: [RFC] syzbot process

[Ozgur <ozgur@goosey.org>]

28.12.2017, 14:45, "Dmitry Vyukov" <dvyukov@google.com>:
> On Thu, Dec 28, 2017 at 11:51 AM, Ozgur <ozgur@goosey.org> wrote:
>>  28.12.2017, 13:41, "Dmitry Vyukov" <dvyukov@google.com>:
>>>  On Fri, Dec 22, 2017 at 4:32 AM, Eric Biggers <ebiggers3@gmail.com> wrote:
>>>>   On Thu, Dec 21, 2017 at 01:52:40PM +0100, Dmitry Vyukov wrote:
>>>>>   However, the cost is that it needs to understand statuses of bugs:
>>>>>   most importantly, what commit fixes what bug. It also has support for
>>>>>   marking a bug as "invalid", e.g. happened once but most likely was
>>>>>   caused by a previous silent memory corruption. And support for marking
>>>>>   bugs as duplicates of other bugs, i.e. the same root cause and will be
>>>>>   fixed when the target bug is fixed. These simple rules are outlined in
>>>>>   the footer of each report and also explained in more detail at the
>>>>>   referenced link:
>>>>>
>>>>>   ----------------------------------
>>>>>   This bug is generated by a dumb bot. It may contain errors.
>>>>>   See https://goo.gl/tpsmEJ for details.
>>>>>   Direct all questions to syzkaller@googlegroups.com.
>>>>>   Please credit me with: Reported-by: syzbot <syzkaller@googlegroups.com>
>>>>>   syzbot will keep track of this bug report.
>>>>>   Once a fix for this bug is merged into any tree, reply to this email with:
>>>>>   #syz fix: exact-commit-title
>>>>>   If you want to test a patch for this bug, please reply with:
>>>>>   #syz test: git://repo/address.git branch
>>>>>   and provide the patch inline or as an attachment.
>>>>>   To mark this as a duplicate of another syzbot report, please reply with:
>>>>>   #syz dup: exact-subject-of-another-report
>>>>>   If it's a one-off invalid bug report, please reply with:
>>>>>   #syz invalid
>>>>>   Note: if the crash happens again, it will cause creation of a new bug report.
>>>>>   Note: all commands must start from beginning of the line in the email body.
>>>>>   ----------------------------------
>>>>>
>>>>>   Status tracking allows syzbot to (1) keep track of still unfixed bugs
>>>>>   (more than half actually gets lost in LKML archives if nobody keeps
>>>>>   track of them), (2) be able to ever report similarly looking crashes
>>>>>   as new bugs in future, (3) be able to test fixes.
>>>>>
>>>>>   The problem is that these rules are mostly not followed.
>>>>
>>>>   As others mentioned, allowing a bug ID to be in the fix's commit message,
>>>>   perhaps in the Reported-by line which syzbot already suggests to include, would
>>>>   make things a bit easier.
>>>>
>>>>   But I think the larger problem is that people in the community don't have any
>>>>   visibility into the statuses of the bugs, so they don't have any motivation to
>>>>   manage the statuses.
>>>>
>>>>   Are you planning to make a dashboard app publicly available for upstream kernel
>>>>   bugs being tracked by syzbot? I think it would be very useful for the
>>>>   community, especially for finding more details about a bug, e.g. when was it
>>>>   last seen, how often was it seen, has it been seen in multiple trees. Also for
>>>>   finding duplicates which may not have been sent to the correct mailing list.
>>>
>>>  Hi Eric,
>>>
>>>  Good question. I would very much like to open the UI, and I hope to do
>>>  it in near future, but we need to do some additional work to make it
>>>  possible. The good news is that information is already accumulating
>>>  and we can do pings, etc.
>>
>>  Hello Dmitry,
>>
>>  I think not useful to be a GUI, for example it can be console based ui we can conenct and get information and fixed patches.
>
> Hi Ozgur,

Hello,

> We will do web UI first as it's something that's already partially
> there and syzbot itself is not a console process, it's a cloud
> service. It's also handy because there are lots of contextual
> information and in a web UI one can just just click links to navigate
> or download a blob. Later we could do an API for console clients, etc
> if there is an interest in developing these types of UIs. But
> generally UI is not the main business of syzbot, it's only a side
> thing that helps it achieve the main goal, so it's doesn't have a team
> of people assigned to it. But you are welcome to contribute, it's all
> open-source:
> https://github.com/google/syzkaller/tree/master/dashboard/app

I understand.

>>  So syzbot is perfectly, I founded a patc last time :)
>>
>>  https://09738734946362323617.googlegroups.com/attach/3c6ef7059f77c/patch.txt?part=0.2&view=1&vt=ANaJVrFm49WFVkkKiomlnsrdfnv4P-0znjiC4agFB72ibq9_6iqg1rmZtw9-DxS5VvoOoKx8Ikl88sYEQQ45X0vjrwFkKDRaZELV-oU9DVmmrRAMSfStn24
>>
>>  And, I have a my suggestions:
>>
>>  Please keep to short url addresses.
>
> Well, that's an URL generated by google groups, we don't have control
> over it. You also received the patch as an attachment in the syzbot
> email.

I know, understand. sure.

>>  and I think syzbot use to .txt file attached.
>>  .txt is not good.
>
> Why are not .txt attachments good? What do you propose to use?

I think I'm misunderstood that is good to have text output in a file but  not useful if the file extension is ".txt" 
Not comfortable use it for mutt / vim and diff.

I think needs to be an new extension, would be like this ".log" or ".syz" :)


> Thanks
>
>>>>   syzbot also should be sending out reminders for bugs that are still open if the
>>>>   crash is still occurring, and even moreso if there is a reproducer.
>>>
>>>  Agree. The reasons why this hasn't happen yet are:
>>>  1. syzbot is being built up as it's running, I am overwhelmed with
>>>  hundreds of bugs and also doing lots of work which may be not directly
>>>  visible but important (e.g. improving quality of generated
>>>  reproducers, increasing percent of cases when reproducers are created,
>>>  improving bug title extraction logic, implementing patch testing by
>>>  request, now this new Reported-by-based process, etc).
>>>  2. Just sending an email for each open bug every week is simple, but I
>>>  afraid it won't be warmly welcomed. The open questions are: how
>>>  frequently syzbot should ping? should repro/no repro affect this? what
>>>  to do if it stopped happening? stopped happenning for how long? and
>>>  what if it happened just few times, so we can't really conclude if it
>>>  still happens or not (but we've seen very bad races manifesting this
>>>  way)? how should it interact with the following point?
>>>
>>>>   However, if the crash isn't still occurring, then I expect it will become
>>>>   necessary to automatically invalidate the bug after some time, lest the list of
>>>>   bugs grow without bound due to bugs that have already been fixed that no one has
>>>>   time to debug to figure out exactly when/what the fix was, especially if there
>>>>   is no reproducer. Or perhaps the bug was only in linux-next and only existed
>>>>   due to a buggy patch which was dropped or modified before it reached mainline,
>>>>   so there is no "fix" commit.
>>>
>>>  Good point. I think we will need to do this in some form in future.
>>>  Again open questions:
>>>   - what is the precise formula behind "isn't still occurring"?
>>>   - should we only close "no repro" bugs?
>>>   - should we re-test bugs with repro? (re-testing is not 100% precise,
>>>  so we will lose some real subtle bugs this way)
>>>
>>>  Thanks