2017-12-12 00:59:44

by Jia-Ju Bai

Subject: [BUG] fs/f2fs: a possible sleep-in-atomic bug in f2fs_trace_pid

According to fs/f2fs/trace.c, the kernel module may sleep under a spinlock.
The function call path is:
f2fs_trace_pid (acquire the spinlock)
f2fs_radix_tree_insert
cond_resched --> may sleep

I do not find a good way to fix it, so I only report.
This possible bug is found by my static analysis tool (DSAC) and my code
review.


Thanks,
Jia-Ju Bai

2017-12-12 02:01:13

by Chao Yu

Subject: Re: [BUG] fs/f2fs: a possible sleep-in-atomic bug in f2fs_trace_pid

Hi Jia-Ju,

Thanks for your report. :)

On 2017/12/12 8:59, Jia-Ju Bai wrote:
> According to fs/f2fs/trace.c, the kernel module may sleep under a spinlock.
> The function call path is:
> f2fs_trace_pid (acquire the spinlock)
> f2fs_radix_tree_insert
> cond_resched --> may sleep
>
> I do not find a good way to fix it, so I only report.
> This possible bug is found by my static analysis tool (DSAC) and my code
> review.
That's a known issue, and there are other problems in IO trace related codes,
so I want to refactor the code directly, you could check the code in below link:

https://git.kernel.org/pub/scm/linux/kernel/git/chao/linux.git/commit/?h=f2fs-dev&id=d895154be1733e2bd29c9e64b57e4e201f0e8a0c

But the patch still needs debugging so far. Since the IO trace feature is off
by default, I expect the impaction is limited, so I pend the debug job on this
patch.

Anyway, I think this issue you reported should be fixed first, let me make the
fixing part independent from original patch for fixing.

Thanks,

>
>
> Thanks,
> Jia-Ju Bai
>
>