linux-mips
[Top] [All Lists]

Re: missing flush_dcache_page call in 2.4 kernel

To: pdh@colonel-panic.org
Subject: Re: missing flush_dcache_page call in 2.4 kernel
From: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Date: Sat, 27 Mar 2004 22:49:52 +0900 (JST)
Cc: phorton@bitbox.co.uk, linux-mips@linux-mips.org
In-reply-to: <20040326184317.GA3661@skeleton-jack>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <4062F1A1.9070005@bitbox.co.uk> <20040326.122258.41628012.nemoto@toshiba-tops.co.jp> <20040326184317.GA3661@skeleton-jack>
Sender: linux-mips-bounce@linux-mips.org
>>>>> On Fri, 26 Mar 2004 18:43:17 +0000, Peter Horton <pdh@colonel-panic.org> 
>>>>> said:

>> I'm quite sure that it's a kernel bug and may cause problems if any
>> PIO block device (PIO ide, ide-cs, mtdblock, etc.) are used on CPUs
>> which have d-cache aliases (not only MIPS).  We need a correct fix
>> ...

pdh> True. A proper fix would flush the relevant page after PIO
pdh> transfers into the page cache / swap pages. Unfortunately this
pdh> would require a hook in the generic kernel.

I found somewhat long discussions in linux-kernel ML.

Subject: [patch] cache flush bug in mm/filemap.c (all kernels >= 2.5.30(at 
least))
http://www.ussg.iu.edu/hypermail/linux/kernel/0305.2/1205.html
http://www.ussg.iu.edu/hypermail/linux/kernel/0305.3/0151.html

Still I do not understand whole story on the thread, David S. Miller
said that architecture defined IDE insw/outsw macro should do the
flushing in this case, if I understand correctly.  Definitely sparc64
__ide_insw do it.  Hmm ...

---
Atsushi Nemoto

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