From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from neon.transmeta.com (neon-best.transmeta.com [206.184.214.10]) by kvack.org (8.8.7/8.8.7) with ESMTP id NAA00730 for ; Mon, 25 Jan 1999 13:53:17 -0500 Date: Mon, 25 Jan 1999 10:49:44 -0800 (PST) From: Linus Torvalds Subject: Re: MM deadlock [was: Re: arca-vm-8...] In-Reply-To: <199901251843.SAA08417@dax.scot.redhat.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org To: "Stephen C. Tweedie" Cc: Alan Cox , werner@suse.de, andrea@e-mind.com, riel@humbolt.geo.uu.nl, Zlatko.Calusic@CARNet.hr, ebiederm+eric@ccr.net, saw@msu.ru, steve@netplus.net, damonbrent@earthlink.net, reese@isn.net, kalle.andersson@mbox303.swipnet.se, bmccann@indusriver.com, bredelin@ucsd.edu, linux-kernel@vger.rutgers.edu, linux-mm@kvack.org List-ID: On Mon, 25 Jan 1999, Stephen C. Tweedie wrote: > > Correct: I haven't been testing any of the networking stuff myself so it > has been a non-issue for any of my workloads here. Obviously any check > for this case would have to be outside the GFP_WAIT conditional, but it > does make sense to set low_on_memory there anyway. In fact, I wonder if we shouldn't just get rid of the GFP_WAIT conditional in __get_free_pages(), and make all that unconditional, so that we track low memory situations correctly even for atomic network traffic - something that obviously is a GoodThing(tm) to do. Then we could just make sure that try_to_free_pages() returns immediately for anything that doesn't have GFP_WAIT set, and have all the kswapd logic there. That would even get rid of a test in the common path. Linus -- To unsubscribe, send a message with 'unsubscribe linux-mm my@address' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://humbolt.geo.uu.nl/Linux-MM/