linux-ext4 - Re: [PATCH 12/31] fuse: implement direct IO with iomap

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: <acd28342-0674-482f-922c-e9af077d8ab0@meta.com>
Date: Fri, 6 Feb 2026 09:27:36 -0500
From: Chris Mason <clm@...a.com>
To: "Darrick J. Wong" <djwong@...nel.org>
Cc: miklos@...redi.hu, joannelkoong@...il.com, bernd@...ernd.com,
        neal@...pa.dev, linux-ext4@...r.kernel.org,
        linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH 12/31] fuse: implement direct IO with iomap

On 2/6/26 12:08 AM, Darrick J. Wong wrote:
> On Thu, Feb 05, 2026 at 09:52:29PM -0500, Chris Mason wrote:
>> On 2/5/26 9:08 PM, Darrick J. Wong wrote:
>>> On Thu, Feb 05, 2026 at 11:19:11AM -0800, Chris Mason wrote:
>>>> "Darrick J. Wong" <djwong@...nel.org> wrote:
>>>>> From: Darrick J. Wong <djwong@...nel.org>
>>>>>
>>>>> Start implementing the fuse-iomap file I/O paths by adding direct I/O
>>>>> support and all the signalling flags that come with it.  Buffered I/O
>>>>> is much more complicated, so we leave that to a subsequent patch.
>>>>>
>>>>
>>>> Hi everyone,
>>>>
>>>> I'm trying out my AI review prompts on a few more trees, and I ran it
>>>> on the fuse-iomap-cache branch:
>>>>
>>>> https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/log/?h=fuse-iomap-cache  
>>>
>>> I should mention: I appreciate you rolling out the AI reviews by slowly
>>> expanding the number of trees you scan for bugs.
>>
>> Thanks, I'm trying not to send unsolicited AI unless it feels like it's
>> mostly right.  I did discard one false positive, which looked accurate
>> but also looked like intended behavior.
>>
>> Now that the false positive rate is pretty reasonable, I'll try to
>> collect some fs/* Fixes: tagged patches and see if I can teach claude
>> how to spot the bugs.  In past kernel-wide scans, it gets ~35%, which is
>> better than 0, but not as good as I was hoping for.
> 
> <nod> You've found some very good bugs, especially in the fuse-iomap
> branch!  At some point I'm going to have to figure out how to run these
> tools myself, but until then you're quite welcome to keep scanning my
> dev trees. :)
> > I wonder, have you tried it on non-kernel repos like e2fsprogs (ha!) or
> fstests?

The prompts are here:

https://github.com/masoncl/review-prompts

And thanks to Christian they now have both kernel and systemd specific
directories.  The original versions of the prompts had a lot of details
about exactly how to review code, but recent models don't seem to need
(or follow) that level of detail.

Instead it's really just forcing larger chunks of the call graph into
the AI context window, and adding some kernel specific knowledge about
locking, rcu, gfp masks, sleepable vs irq context etc.  Basically the
weird stuff that we've forgotten is weird.

So non-kernel projects would mostly work but would need a few fixups
depending on how far they stray from kernel semantics.  It's easy enough
to add branches into the prompts, with the asterisk that from time to
time AI ignores all the instructions and does what it wants.

-chris