linux-mips
[Top] [All Lists]

Re: [PATCH] [RFC] r4k_dma_cache_wback_inv function fails when size=0...

To: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Subject: Re: [PATCH] [RFC] r4k_dma_cache_wback_inv function fails when size=0...
From: Ralf Baechle <ralf@linux-mips.org>
Date: Tue, 16 Mar 2004 17:01:14 +0100
Cc: sjhill@realitydiluted.com, linux-mips@linux-mips.org
In-reply-to: <20040316.230928.74756852.anemo@mba.ocn.ne.jp>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <4055E320.8080808@realitydiluted.com> <20040316.230928.74756852.anemo@mba.ocn.ne.jp>
Sender: linux-mips-bounce@linux-mips.org
User-agent: Mutt/1.4.1i
On Tue, Mar 16, 2004 at 11:09:28PM +0900, Atsushi Nemoto wrote:

> sjhill> The 'r4k_dma_cache_wback_inv' function will fail when the
> sjhill> requested size equals 0 AND when the address is a multiple of
> sjhill> the line size. I discovered this bug while using the National
> sjhill> Semiconductor DP8381x series PCI ethernet driver. I have
> sjhill> attached a test program showing the bug as well as a patch for
> sjhill> comment. Okay to apply?
> 
> I think your patch is overkill.  It flushes many one line then needed.

So far I was simply considering any use with size 0 a bug.  In this case
a 0 argument was passed to pci_unmap_XXX, so I'll have to figure out with
the other architecture maintainers if that's just odd usage of the API
or an outright bug.

  Ralf

<Prev in Thread] Current Thread [Next in Thread>