linux-kernel - Re: [syzbot] [ext4?] INFO: task hung in ext4_evict_ea

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CACT4Y+bmJAKw8J=oJgp5gh_-Tz1e-VkgtvKF4QFGcD-Ovffm8Q@mail.gmail.com>
Date: Mon, 16 Sep 2024 15:55:05 +0200
From: Dmitry Vyukov <dvyukov@...gle.com>
To: Bart Van Assche <bvanassche@....org>
Cc: syzbot <syzbot+38e6635a03c83c76297a@...kaller.appspotmail.com>, 
	adilger.kernel@...ger.ca, alim.akhtar@...sung.com, avri.altman@....com, 
	beanhuo@...ron.com, hdanton@...a.com, jejb@...ux.ibm.com, 
	linux-ext4@...r.kernel.org, linux-kernel@...r.kernel.org, 
	linux-scsi@...r.kernel.org, martin.petersen@...cle.com, 
	syzkaller-bugs@...glegroups.com, tytso@....edu, 
	wsa+renesas@...g-engineering.com, syzkaller <syzkaller@...glegroups.com>
Subject: Re: [syzbot] [ext4?] INFO: task hung in ext4_evict_ea_inode

On Mon, 16 Sept 2024 at 15:41, 'Bart Van Assche' via syzkaller
<syzkaller@...glegroups.com> wrote:
>
> On 9/16/24 6:28 AM, Dmitry Vyukov wrote:
> > On Fri, 3 Feb 2023 at 19:11, Bart Van Assche <bvanassche@....org> wrote:
> >>
> >> On 2/3/23 00:53, syzbot wrote:
> >>> syzbot has bisected this issue to:
> >>>
> >>> commit 82ede9c19839079e7953a47895729852a440080c
> >>> Author: Wolfram Sang <wsa+renesas@...g-engineering.com>
> >>> Date:   Tue Jun 21 14:46:53 2022 +0000
> >>>
> >>>       scsi: ufs: core: Fix typos in error messages
> >>
> >> To the syzbot maintainers: I think this is a good example of a bisection
> >> result that is wrong. It is unlikely that fixing typos in kernel
> >> messages would affect whether or not the kernel hangs. Additionally, as
> >> far as I know, the systems used by syzbot (Google Compute Engine virtual
> >> machines) do trigger any code in the UFS driver.
> >
> > Hi Bart,
> >
> > syzbot has logic to detect commits that don't affect builds.
> > It hashes SHF_ALLOC vmlinux sections to check if the commit actually
> > has any effect on the binary:
> > https://github.com/google/syzkaller/blob/c673ca06b23cea94091ab496ef62c3513e434585/pkg/build/linux.go#L253-L286
> >
> > Bug CONFIG_UFS_FS is enabled on syzbot, it has some coverage for it,
> > and strings affect the binary (can actually be the root cause for
> > bugs). So I don't see what else can be done here automatically.
>
> CONFIG_UFS_FS controls whether or not the UFS filesystem is enabled
> (fs/ufs/). The UFS driver is unrelated to the UFS filesystem and is
> controlled by CONFIG_SCSI_UFSHCD. Support for the UFS driver has been
> added recently in Qemu. Does that mean that it should be possible to
> test the UFS driver with syzbot? See also
> https://patchew.org/QEMU/20230616065816epcms2p82787f1aeb410ec4b8ab6ffedb6edf4d2@epcms2p8/

Oh, I see. Added this example to the issue:
https://github.com/google/syzkaller/issues/2297#issuecomment-2352986387

I've tried to follow all guidelines from reproducible builds:
https://docs.kernel.org/kbuild/reproducible-builds.html

https://github.com/google/syzkaller/blob/c673ca06b23cea94091ab496ef62c3513e434585/pkg/build/linux.go#L165-L176
https://github.com/google/syzkaller/blob/c673ca06b23cea94091ab496ef62c3513e434585/pkg/build/linux.go#L85-L102

Yet this commit still somehow affected the build:
culprit signature:
c1512384fd0e6d2bd48b3a1ce2034eb5c0de195562eca27cd2c0c6ac58cd4863
parent  signature:
1f86ef314029ed6e1f240aae874259f4bb0f7978f8b38b29d6c184b82965b331

I've debugged and fixed some of these non-determinism bugs, but it
seems something is still broken (not sure if it's the compiler, or
kbuild, or how we do deterministic build).