From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from digeo-nav01.digeo.com (digeo-nav01.digeo.com [192.168.1.233]) by packet.digeo.com (8.9.3+Sun/8.9.3) with SMTP id LAA12540 for ; Fri, 6 Sep 2002 11:58:38 -0700 (PDT) Message-ID: <3D78FAD6.269EF2FB@digeo.com> Date: Fri, 06 Sep 2002 11:58:30 -0700 From: Andrew Morton MIME-Version: 1.0 Subject: Re: 0-order allocation failures in LTP run of Last nights bk tree References: <1031322426.30394.4.camel@plars.austin.ibm.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-linux-mm@kvack.org Return-Path: To: Paul Larson Cc: lkml , linux-mm List-ID: Paul Larson wrote: > > In the nightly ltp run against the bk 2.5 tree last night I saw this > show up in the logs. > > It happened on the 2-way PIII-550, 2gb physical ram, but not on the > smaller UP box I test on. > > mtest01: page allocation failure. order:0, mode:0x50 scsi, I assume? This will be failed bounce buffer allocation attempts. That's fine, normal. block will fall back to the mempool and will wait. Of course, your shouldn't be bounce buffering at all. This is happening because of the block-highmem problem. There's a workaround at http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.33/2.5.33-mm4/broken-out/scsi_hack.patch But please bear in mind, this "page allocation failure" message is purely a developer diagnostic thing. The reason it is there is so that if some random toaster driver oopses over a failure to handle an allocation failure, the person who reports the bug can say "I saw an allocation failure and then your driver crashed". Which tells the driver developer where to look. Under heavy load, page allocation attempts _will_ fail, and that's OK. The mempool-backed memory will become available. It's a bit CPU-inefficient, and I have code under test which changes GFP_NOFS mempool allocators to not even bother trying to enter page reclaim if the nonblocking allocation attempt failed. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/