[Help Wanted] Syzkaller Corpus Download Link

[Hao Sun]

Hi
Dmitry, I'm Hao Sun, an MSE student from Tsinghua University.
Syzkaller is a really amazing great fuzzer. Appreciate your excellent
work. I'm doing some research work focusing on kernel fuzzing. I found
that Syzbot has already accumulated over 80,000 system sequences in
its corpus and I wonder if you can add a download link on the Syzbot
dashboard page. In this way, anyone who wants to deploy Syzkaller
locally can bootstrap with a rich initial corpus thereby speeding up
the fuzzing process. Besides, I also want to do some research work
about the existing corpus.
Looking forward to your response.

Best
Hao Sun

[逸风陈]

Hi

Here is a script used for crawl syz program https://drive.google.com/file/d/1uhQOQvayXY91QAeP1jtK5blxzwzwIHj-/view .

Hope it will help you.

[Hao Sun]

在2021年4月15日星期四 UTC+8 下午8:36:45<yifeng...@gmail.com> 写道：

Hi

Here is a script used for crawl syz program https://drive.google.com/file/d/1uhQOQvayXY91QAeP1jtK5blxzwzwIHj-/view .

Hope it will help you.

Hi

Thanks very much, yifeng.  This script seems like it will automatically crawl all the crashes and the corresponding reproduction programs, which is useful, thank you.

However, what I would like to see is that Syzkaller provides some kind of entry supporting people to download the corpus it has accumulated., as this would speed up the fuzz efficiency for everyone using Syzkaller.

I know syz-manager now regularly syncs the fuzz stats to the dashboard app, so a possible implementation would be to check if there are enough new progs while syncing the fuzz stats, and if so, then sync corpus to the dashboard app, and finally provide a download link in the dashboard page.

Hao Sun

[Dmitry Vyukov]

Hi Hao,

I've uploaded the corpus manually to:
https://storage.googleapis.com/syzkaller/temp/corpus.db

If you want to contribute automatic uploading that's welcome. But
please discuss the design first. Providing a live copy synced for
every program looks too expensive and unnecessary.

[Hao Sun]

Dmitry Vyukov <dvy...@google.com>于2021年4月16日 周五15:32写道：

On Thu, Apr 15, 2021 at 3:04 PM Hao Sun <sunh...@gmail.com> wrote:
>
> 在2021年4月15日星期四 UTC+8 下午8:36:45<yifeng...@gmail.com> 写道：
>>
>> Hi
>> Here is a script used for crawl syz program https://drive.google.com/file/d/1uhQOQvayXY91QAeP1jtK5blxzwzwIHj-/view .
>> Hope it will help you.
>>
> Hi
>
> Thanks very much, yifeng.  This script seems like it will automatically crawl all the crashes and the corresponding reproduction programs, which is useful, thank you.
> However, what I would like to see is that Syzkaller provides some kind of entry supporting people to download the corpus it has accumulated., as this would speed up the fuzz efficiency for everyone using Syzkaller.
> I know syz-manager now regularly syncs the fuzz stats to the dashboard app, so a possible implementation would be to check if there are enough new progs while syncing the fuzz stats, and if so, then sync corpus to the dashboard app, and finally provide a download link in the dashboard page.

Hi Hao,

I've uploaded the corpus manually to:
https://storage.googleapis.com/syzkaller/temp/corpus.db

Thank you very much!

If you want to contribute automatic uploading that's welcome. But
please discuss the design first. Providing a live copy synced for
every program looks too expensive and unnecessary.

That’s right. A more simple implementation would be sync the corpus regularly, e.g. every 1h, when the size of corpus is different from the size since last synchronization, in a background goroutine.