linux-ext4 - Re: [QUESTION] BUG_ON in ext4_mb_simple_scan

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: <20200430222558.805DEAE057@d06av26.portsmouth.uk.ibm.com>
Date:   Fri, 1 May 2020 03:55:57 +0530
From:   Ritesh Harjani <riteshh@...ux.ibm.com>
To:     linux-ext4@...r.kernel.org
Cc:     "zhangyi (F)" <yi.zhang@...wei.com>,
        yangerkun <yangerkun@...wei.com>, tytso@....edu, jack@...e.cz,
        dmonakhov@...il.com, adilger@...ger.ca, bob.liu@...cle.com,
        wshilong@....com
Subject: Re: [QUESTION] BUG_ON in ext4_mb_simple_scan_group

>>
>>> If above is true, then may be we should not call
>>> "SetPageUptodate(page)", in case of an error reading block bitmap?
>>> Thoughts?
>>>
>> Yeah, it's better to set page uptodate only if all block bitmap buffers
>> are uptodate represent by this page.
> 
> 
> So I guess the *easier* thing to do here is to abort the loop which
> calls ext4_wait_block_bitmap() in ext4_mb_init_cache, similar to how
> the loop above it does (which calls for ext4_read_block_bitmap_nowait())
> 
> Since if any of the block bitmap buffer (which belongs to that page)
> could not be read properly, then we should not set the PageUptodate.
> (including for blocksize < pagesize where groups_per_page > 1) and
> no need of even setting up the in memory buddy and bitmap information
> (since we are anyway not going to use it).
> 
> Others can comment, if something else needs to be done?
> But I think over optimizing this logic for blocksize < pagesize
> may become an overkill? (since this mostly happens during an I/O error).
> 

Ok, so as I see this bug was possibly introduced to handle blocksize < 
pagesize case itself [1]. So it will make no sense to just optimize it
for blocksize == pagesize again. So I guess we should look into this
more closely rather then simply implementing above logic.

[1]: https://www.spinics.net/lists/linux-ext4/msg48111.html

So, I can get back to this some time later maybe.

-ritesh