From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Sat, 20 Jul 2002 18:31:33 -0700 (PDT) Message-Id: <20020720.183133.67807986.davem@redhat.com> Subject: Re: [PATCH] generalized spin_lock_bit From: "David S. Miller" In-Reply-To: <1027211185.17234.48.camel@irongate.swansea.linux.org.uk> References: <1027196511.1555.767.camel@sinai> <20020720.152703.102669295.davem@redhat.com> <1027211185.17234.48.camel@irongate.swansea.linux.org.uk> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-linux-mm@kvack.org Return-Path: To: alan@lxorguk.ukuu.org.uk Cc: rml@tech9.net, torvalds@transmeta.com, linux-kernel@vger.kernel.org, linux-mm@kvack.org, riel@conectiva.com.br, wli@holomorphy.com List-ID: Secondly many platforms want to implement their locks in other ways. Atomic bitops are an x86 luxury so your proposal simply generates hideously inefficient code compared to arch specific sanity For an asm-generic/bitlock.h implementation it is more than fine. That way we get asm-i386/bitlock.h that does whatever it wants to do and the rest of asm-*/bitlock.h includes the generic version until the arch maintainer sees fit to do otherwise. See the difference between what we have here now? It means all ports will at least sort-of work when the change gets installed. A lot of testing gets lost because ports break on a daily basis due to changes when done like this. Look at the asm/rmap.h stuff, that was done right and platforms kept at least compiling when that change went in. I don't mind architecture breakage when truly necessary to change stuff, but when it can be avoided reasonably it should. -- 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/