From mboxrd@z Thu Jan 1 00:00:00 1970 From: salyzyn@android.com (Mark Salyzyn) Date: Tue, 22 Sep 2015 07:15:00 -0700 Subject: ARM64 readahead: fault retry breaks mmap file read random detection In-Reply-To: <20150921235120.GI7356@arm.com> References: <1442868028-27055-1-git-send-email-salyzyn@android.com> <20150921210938.GC7356@arm.com> <56007868.8050605@android.com> <20150921235120.GI7356@arm.com> Message-ID: <56016264.50509@android.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 09/21/2015 04:51 PM, Will Deacon wrote: > On Mon, Sep 21, 2015 at 10:36:40PM +0100, Mark Salyzyn wrote: >> On 09/21/2015 02:09 PM, Will Deacon wrote: >>> On Mon, Sep 21, 2015 at 09:39:50PM +0100, Mark Salyzyn wrote: >>>> Description from commit 45cac65b0fcd >>>> ("readahead: fault retry breaks mmap file read random detection") >>>> . . . >>>> Yup, arm64 needs this too! Random read improves by 250%, sequential >>>> read improves by 40%, and random write by 400% to an eMMC device with >>>> dm crypto wrapped around it. >>> Thanks for this. This must've gone in whilst we were developing the initial >>> version of the arm64 port and has since gone unnoticed. >>> >>> I'll queue it on the arm64 fixes branch and send a pull request after >>> some testing. >>> >> As noted, this fix may need to be propagated to all the arch-specific >> code, I was not in a position to check this out on arm (32 bit) and the >> benchmarking code I used did not immediately port to 32-bit. > You lost me; which arch-specific code are you referring to? The original > patch (in mainline) touches a whole bunch of architectures. > > Will I see I am mistaken, arm64 was the _only_ one I see that is missing ... Sincerely -- Mark Salyzyn From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757187AbbIVOPF (ORCPT ); Tue, 22 Sep 2015 10:15:05 -0400 Received: from mail-pa0-f41.google.com ([209.85.220.41]:35770 "EHLO mail-pa0-f41.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751312AbbIVOPC (ORCPT ); Tue, 22 Sep 2015 10:15:02 -0400 Subject: Re: ARM64 readahead: fault retry breaks mmap file read random detection To: Will Deacon References: <1442868028-27055-1-git-send-email-salyzyn@android.com> <20150921210938.GC7356@arm.com> <56007868.8050605@android.com> <20150921235120.GI7356@arm.com> Cc: "linux-kernel@vger.kernel.org" , Riley Andrews , Shaohua Li , Shaohua Li , Rik van Riel , Wu Fengguang , Catalin Marinas , Vladimir Murzin , Dave P Martin , David Hildenbrand , James Morse , Mark Rutland , "linux-arm-kernel@lists.infradead.org" From: Mark Salyzyn Message-ID: <56016264.50509@android.com> Date: Tue, 22 Sep 2015 07:15:00 -0700 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.2.0 MIME-Version: 1.0 In-Reply-To: <20150921235120.GI7356@arm.com> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 09/21/2015 04:51 PM, Will Deacon wrote: > On Mon, Sep 21, 2015 at 10:36:40PM +0100, Mark Salyzyn wrote: >> On 09/21/2015 02:09 PM, Will Deacon wrote: >>> On Mon, Sep 21, 2015 at 09:39:50PM +0100, Mark Salyzyn wrote: >>>> Description from commit 45cac65b0fcd >>>> ("readahead: fault retry breaks mmap file read random detection") >>>> . . . >>>> Yup, arm64 needs this too! Random read improves by 250%, sequential >>>> read improves by 40%, and random write by 400% to an eMMC device with >>>> dm crypto wrapped around it. >>> Thanks for this. This must've gone in whilst we were developing the initial >>> version of the arm64 port and has since gone unnoticed. >>> >>> I'll queue it on the arm64 fixes branch and send a pull request after >>> some testing. >>> >> As noted, this fix may need to be propagated to all the arch-specific >> code, I was not in a position to check this out on arm (32 bit) and the >> benchmarking code I used did not immediately port to 32-bit. > You lost me; which arch-specific code are you referring to? The original > patch (in mainline) touches a whole bunch of architectures. > > Will I see I am mistaken, arm64 was the _only_ one I see that is missing ... Sincerely -- Mark Salyzyn