linux-ext4 - Re: [PATCH 0/5] ext4/overlayfs: fiemap related fixes

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]

Message-Id: <20200424232024.A39974C046@d06av22.portsmouth.uk.ibm.com>
Date:   Sat, 25 Apr 2020 04:50:23 +0530
From:   Ritesh Harjani <riteshh@...ux.ibm.com>
To:     Christoph Hellwig <hch@...radead.org>
Cc:     linux-ext4@...r.kernel.org, jack@...e.cz, tytso@....edu,
        adilger@...ger.ca, darrick.wong@...cle.com,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Dan Carpenter <dan.carpenter@...cle.com>,
        "Aneesh Kumar K . V" <aneesh.kumar@...ux.ibm.com>,
        Murphy Zhou <jencce.kernel@...il.com>,
        Miklos Szeredi <miklos@...redi.hu>,
        Amir Goldstein <amir73il@...il.com>,
        linux-fsdevel@...r.kernel.org, linux-unionfs@...r.kernel.org
Subject: Re: [PATCH 0/5] ext4/overlayfs: fiemap related fixes

Hello Christoph,

Thanks for your review comments.

On 4/24/20 3:41 PM, Christoph Hellwig wrote:
> I think the right fix is to move fiemap_check_ranges into all the ->fiemap

I do welcome your suggestion here. But I am not sure of what you are
suggesting should be done as a 1st round of changes for the immediate
reported problem.
So currently these patches take the same approach on overlayfs on how 
VFS does it. So as a fix to the overlayfs over ext4 reported problems in
thread [1] & [2]. I think these patches are doing the right thing.

Also maybe I am biased in some way because as I see these are the right
fixes with minimal changes only at places which does have a direct
problem.

But I do agree that in the second round (as a right approach for the
long term), we could just get rid of fiemap_check_ranges() from
ioctl_fiemap() & ovl_fiemap(), and better add those in all filesystem
specific implementations of ->fiemap() call.
(e.g. ext4_fiemap(), f2fs_fiemap() etc.).

> instances (we only have a few actual implementation minus the wrappers
> around iomap/generic).  
 >
Ok, got it. So for filesystem specific ->fiemap implementations,
we should add fiemap_check_ranges() in there implementations.
And for those FS which are calling iomap_fiemap() or
generic_block_fiemap(), what you are suggesting it to add
fiemap_check_ranges() in iomap_fiemap() & generic_block_fiemap().
Is this understanding correct?

> Then add a version if iomap_fiemap that can pass
> in maxbytes explicitly for ext4, similar to what we've done with various
> other generic helpers.

Sorry I am not sure if I followed it correctly. Help me understand pls.
Also some e.g about "what we've done with various other generic helpers"

iomap_fiemap(), will already get a FS specific inode from which we can
calculate inode->i_sb->s_maxbytes. So why pass maxbytes explicitly?

> 
> The idea of validating input against file systems specific paramaters
> before we call into the fs is just bound to cause problems.
> 
Sure, but as I was saying. The changes you are suggesting will have
changes in all filesystem's individual ->fiemap() implementations.
But as a fix for the reported problem of [1] & [2], I think these
patches could be taken. Once those are merged, I can work on the changes
that you are suggesting.

Does that sound ok to you?

[1]: https://lkml.org/lkml/2020/4/11/46
[2]: 
https://patchwork.ozlabs.org/project/linux-ext4/patch/20200418233231.z767yvfiupy7hwgp@xzhoux.usersys.redhat.com/ 

-ritesh