[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87y5bee2qc.fsf@xmission.com>
Date: Thu, 16 May 2013 19:53:31 -0700
From: ebiederm@...ssion.com (Eric W. Biederman)
To: "H. Peter Anvin" <hpa@...or.com>
Cc: HATAYAMA Daisuke <d.hatayama@...fujitsu.com>, vgoyal@...hat.com,
akpm@...ux-foundation.org, cpw@....com,
kumagai-atsushi@....nes.nec.co.jp, lisa.mitchell@...com,
kexec@...ts.infradead.org, linux-kernel@...r.kernel.org,
zhangyanfei@...fujitsu.com, jingbai.ma@...com, linux-mm@...ck.org,
riel@...hat.com, walken@...gle.com, hughd@...gle.com,
kosaki.motohiro@...fujitsu.com
Subject: Re: [PATCH v6 0/8] kdump, vmcore: support mmap() on /proc/vmcore
"H. Peter Anvin" <hpa@...or.com> writes:
> On 05/15/2013 02:05 AM, HATAYAMA Daisuke wrote:
>> Currently, read to /proc/vmcore is done by read_oldmem() that uses
>> ioremap/iounmap per a single page. For example, if memory is 1GB,
>> ioremap/iounmap is called (1GB / 4KB)-times, that is, 262144
>> times. This causes big performance degradation.
>
> read_oldmem() is fundamentally broken and unsafe. It needs to be
> unified with the plain /dev/mem code and any missing functionality fixed
> instead of "let's just do a whole new driver".
That is completely and totally orthogonal to this change.
read_oldmem may have problems but in practice on a large systems those
problems are totally dwarfed by real life performance issues that come
from playing too much with the page tables.
I really don't find bringing up whatever foundational issues you have
with read_oldmem() appropriate or relevant here.
Eric
--
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/
Powered by blists - more mailing lists