From mboxrd@z Thu Jan 1 00:00:00 1970 From: Xiao Ni Subject: One patch just review the code Date: Tue, 5 Aug 2014 03:12:07 -0400 (EDT) Message-ID: <1122994629.17915693.1407222727992.JavaMail.zimbra@redhat.com> References: <705044700.17914512.1407222497904.JavaMail.zimbra@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <705044700.17914512.1407222497904.JavaMail.zimbra@redhat.com> Sender: linux-raid-owner@vger.kernel.org To: linux-raid@vger.kernel.org Cc: Jes Sorensen List-Id: linux-raid.ids Hi all I'm reading the code of md. I find there is a problem. I know now there are arrays mark[SYNC_MARKS] mark_cnt[SYNC_MARKS] store the information about how many sectors finish recovery and the moment. 7825 currspeed = ((unsigned long)(io_sectors-mddev->resync_mark_cnt))/2 7826 /((jiffies-mddev->resync_mark)/HZ +1) +1; But when calculate the speed of recovery, the sectors used to calculate contains the sectors which are not finished recovery. When assign value to mark_cnt[next], it subtract the sectors which don't finish recovery. So I think when calculate the recovery speed we should subtract the sectors not finishing recovery too. 7638 mark_cnt[next] = io_sectors - atomic_read(&mddev->recovery_active); So I try to modify and the patch is: --- linux-stable/drivers/md/md.c 2014-07-30 14:36:37.327535805 +0800 +++ fix/md.c 2014-07-31 16:40:57.151493177 +0800 @@ -7652,7 +7652,7 @@ */ cond_resched(); - currspeed = ((unsigned long)(io_sectors-mddev->resync_mark_cnt))/2 + currspeed = ((unsigned long)(io_sectors-atomic_read(&mddev->recovery_active)-mddev->resync_mark_cnt))/2 /((jiffies-mddev->resync_mark)/HZ +1) +1; if (currspeed > speed_min(mddev)) { Am I right? Best Regards Xiao