linux-ext4 - Re: [PATCH 2/2] ext4: Use slab allocator for sub-page sized allocations

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: <20091202201234.GG6278@thunk.org>
Date:	Wed, 2 Dec 2009 15:12:34 -0500
From:	tytso@....edu
To:	Eric Sandeen <sandeen@...hat.com>
Cc:	Ext4 Developers List <linux-ext4@...r.kernel.org>
Subject: Re: [PATCH 2/2] ext4: Use slab allocator for sub-page sized
 allocations

On Wed, Dec 02, 2009 at 10:53:26AM -0600, Eric Sandeen wrote:
> 
> So, this undoes commit c089d490dfbf53bc0893dc9ef57cf3ee6448314d
> more or less, right:
> 
> JBD: Replace slab allocations with page allocations

Close, but not quite.  We still use getfreepage() in the case where
blocksize==pagesize.  So in the most common case of 4k blocks on
x86/x86_64, we won't be sufferring the overhead of using the
SLAB/SLUB/SLOB/SLQB machinery.

> Was alignment the only reason that commit went in?

Alignment and if debugging was enabled, it meant that you needed two
contiguous pages for every 4k shadow buffer_head used by the JBD
layer, not to mention the overhead of tracking each page in the slab
cache.

> Also, there is a race issue here I think, that we've had bugs
> on internally, but w/ the above commit it didn't matter anymore:
> 
> If you simultaneously mount several jbd(2)-based filesystems
> at once, before the slabs get created, nothing protects jbd2_slab[]
> and we can race on creation/initialization of that array.
> 
> See a proposed patch:
> 
> https://bugzilla.lustre.org/attachment.cgi?id=22906
> 
> but I think that patch is a little heavy-handed, the down_reads
> of the slab lock seem to be extraneous because if the fs is mounted,
> nobody will be tearing down that slab anyway, since that only
> happens on module unload, right?

Yeah, we only need the lock at mount time, when we are filling in the
jbd_slab[] array.  Thanks for pointing that out; I'll fix it and
resend the patch.

					- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html