linux-kernel - Re: [patch] convert aio event reap to use atomic-op instead of spin

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: <20070411200348.GN13621@kvack.org>
Date:	Wed, 11 Apr 2007 16:03:48 -0400
From:	Benjamin LaHaise <bcrl@...ck.org>
To:	Zach Brown <zach.brown@...cle.com>
Cc:	Ken Chen <kenchen@...gle.com>, akpm@...ux-foundation.org,
	linux-aio@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [patch] convert aio event reap to use atomic-op instead of spin_lock

On Wed, Apr 11, 2007 at 12:52:56PM -0700, Zach Brown wrote:
> I'm worried that virtual aliasing spells doom for the current
> home-brewed serialization that fs/aio.c is doing with the shared ring
> head/tail accesses.  Am I worrying about nothing here?

Adding a flush_dcache_page() should fix that, but it might also need a 
change in the layout of the ring buffer header to get things completely 
correct.  What I'm thinking of is that the head and tail bits might need 
to be on different cachelines to ensure any aliasing that does occur will 
not result in updates colliding.  I'm not that much of an expert on virtually 
aliased caches, though.

> > > I will look into this aside from this patch.
> > 
> > That's probably the case.  Also, any changes in this area *must* correctly 
> > update the compat/incompat feature flags in the ring buffer header.  That 
> > has been missed in the past...
> 
> Do you know of anyone using the current ring info ABI?
> 
> The *only* user I know of is the check of ctx->magic in libaio.

But... any libaio implementing the new sematics must be able to run on old 
kernels by falling back to the syscall when it notices that required bits 
aren't set in the header.  It's easy enough to implement the checks, they 
just need to be carefully checked before being shipped.

		-ben
-- 
"Time is of no importance, Mr. President, only life is important."
Don't Email: <zyntrop@...ck.org>.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/