Re: [syzbot] [hfs?] general protection fault in hfs_find_init

["Arnd Bergmann" <arnd@arndb.de>]

On Sat, Sep 6, 2025, at 07:56, syzbot wrote:
> syzbot suspects this issue was fixed by commit:
>
> commit 42b0ef01e6b5e9c77b383d32c25a0ec2a735d08a
> Author: Arnd Bergmann <arnd@arndb.de>
> Date:   Fri Jul 11 08:46:51 2025 +0000
>
>     block: fix FS_IOC_GETLBMD_CAP parsing in blkdev_common_ioctl()
>
> bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=17c0d312580000
> start commit:   ee88bddf7f2f Merge tag 'bpf-fixes' of git://git.kernel.org..
> git tree:       upstream
> kernel config:  https://syzkaller.appspot.com/x/.config?x=28cc6f051378bb16
> dashboard link: https://syzkaller.appspot.com/bug?extid=7ca256d0da4af073b2e2
> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=1026b182580000
> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=159e0f0c580000

I took a look and concluded that my patch is unlikely to have
fixed the issue, because:

 - my patch was wrong and needed another fixup on top
 - the reproducer and kernel log show no reference to ioctl() calls,
   so they do not directly interact with the code I changed.

It is possible that my patch is hiding the root cause for the problem,
if part of the reproducer relies on a prior call to ioctl() on a block
device and this ioctl was broken by my patch. This still sounds like a
long shot though, and my first guess would be that the bisection
went wrong, possibly by running into more than one issue, or an
unreliable reproducer.

     Arnd