From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [RFC 4/5] MMC: Adjust unaligned write accesses. Date: Fri, 11 Mar 2011 11:23:20 +0100 Message-ID: <201103111123.21063.arnd@arndb.de> References: <1299718449-15172-1-git-send-email-andreiw@motorola.com> <201103102254.02041.arnd@arndb.de> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: Received: from moutng.kundenserver.de ([212.227.17.8]:51111 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750998Ab1CKKXg (ORCPT ); Fri, 11 Mar 2011 05:23:36 -0500 In-Reply-To: Sender: linux-mmc-owner@vger.kernel.org List-Id: linux-mmc@vger.kernel.org To: Andrei Warkentin Cc: linux-mmc@vger.kernel.org On Friday 11 March 2011, Andrei Warkentin wrote: > > I also don't know what the effect of setting physical_block_size > > and/or io_min is, possibly it no longer works if they are larger > > than the MMU page size. Need to try this. > > The other thing is figuring out the default size limit for when the > page align should be performed. I suppose it's safe enough to > set it to 1.5 size the super_page_size. But that number came from the > Toshiba card tests. Or it could be unbounded by default. I'd rather do > the later. You mean always splitting (multiples of) full super-pages from partial super-pages when the quirk flag is enabled? I think it depends on the performance numbers. Do you have any meaningful measurements without the quirk, with the current implementation and with the unbounded case? If the latter two are not much different on the toshiba card, that would be a simpler implementation, and more likely to be useful on other cards. I think we should also do measurements to see if the same quirk actually has any negative effects on other cards, or if there are even cases where it helps. Arnd