linux-kernel - Re: [PATCH] xfs: mark xfs_dir2_sf_entry_t as _

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20190308152924.GB20294@lst.de>
Date:   Fri, 8 Mar 2019 16:29:24 +0100
From:   Christoph Hellwig <hch@....de>
To:     Arnd Bergmann <arnd@...db.de>
Cc:     "Darrick J. Wong" <darrick.wong@...cle.com>,
        linux-xfs@...r.kernel.org, Christoph Hellwig <hch@....de>,
        Dave Chinner <dchinner@...hat.com>,
        Brian Foster <bfoster@...hat.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] xfs: mark xfs_dir2_sf_entry_t as __packed again

On Mon, Mar 04, 2019 at 08:36:47PM +0100, Arnd Bergmann wrote:
> For ARM OABI builds, we run into a compile time assertion:
> 
> inlined from 'init_xfs_fs' at /git/arm-soc/fs/xfs/xfs_super.c:1991:2:
> fs/xfs/xfs_ondisk.h:119:208: error: call to '__compiletime_assert_119' declared with attribute error: XFS: sizeof(xfs_dir2_sf_entry_t) is wrong, expected 3
> 
> While ARM OABI is pretty much dead and fails to build for typical
> configurations on modern architectures (ARMv6 or higher), and has
> been declared deprecated in user space since gcc-4.6, the kernel
> still allows it to used for building the kernel.
> 
> In commit 8353a649f577 ("xfs: kill xfs_dir2_sf_off_t"), Christoph
> removed the old __arch_pack annotation that made it possible to
> build xfs with oddball ABIs. However, OABI not only requrires
> padding around short structure but still adds padding after this
> change. There is no harm to unconditionally mark the structure as
> __packed now, and that will do the right thing here.
> 
> As of commit aa2dd0ad4d6d ("xfs: remove __arch_pack"), we need to
> use __packed here as well, instead of the old __arch_pack.

I don't think we want more __packed attributes than required.  Given
how dead OABI is can we just have XFS depend on !OABI?