[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <26394.37450.21740.720541@quad.stoffel.home>
Date: Thu, 24 Oct 2024 14:30:34 -0400
From: "John Stoffel" <john@...ffel.org>
To: Kent Overstreet <kent.overstreet@...ux.dev>
Cc: manas18244@...td.ac.in,
linux-bcachefs@...r.kernel.org,
linux-kernel@...r.kernel.org,
Anup Sharma <anupnewsmail@...il.com>,
Shuah Khan <skhan@...uxfoundation.org>,
syzbot+e8eff054face85d7ea41@...kaller.appspotmail.com
Subject: Re: [PATCH] Revert "bcachefs: Add asserts to
bch2_dev_btree_bitmap_marked_sectors()"
>>>>> "Kent" == Kent Overstreet <kent.overstreet@...ux.dev> writes:
> On Mon, Oct 21, 2024 at 10:18:57PM +0530, Manas via B4 Relay wrote:
>> From: Manas <manas18244@...td.ac.in>
>>
>> This reverts commit 60f2b1bcf519416dbffee219132aa949d0c39d0e.
>>
>> This syzbot bug[1] is triggered due to the BUG_ON assertions added in
>> __bch2_dev_btree_bitmap_mark. During runtime, m->btree_bitmap_shift is
>> 63 '?'. This triggers both the assertions.
> The BUG_ON() doesn't need to be deleted; we just need to fix the
> validation so it doesn't fire (it doesn't particularly matter if it's
> removed or not, ubsan would catch it without the BUG_ON()).
Shouldn't the BUG_ON() be replaced with making the filesystem readonly
instead if you're going to keep it? I'd rather have the filesystem
still be mounted and able to be read, but not writeable, instead of
having my system crash before I can do anything.
John
Powered by blists - more mailing lists