linux-kernel - Re: [syzbot] [hfs] general protection fault in tomoyo_check

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

Date: Thu, 14 Mar 2024 18:27:31 +0100
From: Jan Kara <jack@...e.cz>
To: Aleksandr Nogikh <nogikh@...gle.com>
Cc: Jan Kara <jack@...e.cz>,
	syzbot <syzbot+28aaddd5a3221d7fd709@...kaller.appspotmail.com>,
	axboe@...nel.dk, brauner@...nel.org, jmorris@...ei.org,
	linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-security-module@...r.kernel.org, paul@...l-moore.com,
	serge@...lyn.com, syzkaller-bugs@...glegroups.com,
	Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>,
	Dmitry Vyukov <dvyukov@...gle.com>
Subject: Re: [syzbot] [hfs] general protection fault in tomoyo_check_acl (3)

Hi Aleksandr,

On Thu 14-03-24 17:21:30, Aleksandr Nogikh wrote:
> Yes, the CONFIG_BLK_DEV_WRITE_MOUNTED=n change did indeed break our C
> executor code (and therefore our C reproducers). I posted a fix[1]
> soon afterwards, but the problem is that syzbot will keep on using old
> reproducers for old bugs. Syzkaller descriptions change over time, so
> during bisection and patch testing we have to use the exact syzkaller
> revision that detected the original bug. All older syzkaller revisions
> now neither find nor reproduce fs bugs on newer Linux kernel revisions
> with CONFIG_BLK_DEV_WRITE_MOUNTED=n.

I see, thanks for explanation!

> If the stream of such bisection results is already bothering you and
> other fs people, a very quick fix could be to ban this commit from the
> possible bisection results (it's just a one line change in the syzbot
> config). Then such bugs would just get gradually obsoleted by syzbot
> without any noise.

It isn't bothering me as such but it results in
CONFIG_BLK_DEV_WRITE_MOUNTED=n breaking all fs-related reproducers and thus
making it difficult to evaluate whether the reproducer was somehow
corrupting the fs image or not. Practically it means closing most
fs-related syzbot bugs and (somewhat needlessly) starting over from scratch
with search for reproducers. I'm OK with that although it is a bit
unfortunate... But I'm pretty sure within a few months syzbot will deliver
a healthy portion of new issues :)

								Honza
-- 
Jan Kara <jack@...e.com>
SUSE Labs, CR