Return-Path: Received: from plane.gmane.org ([80.91.229.3]:46699 "EHLO plane.gmane.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752177AbbBXPlC (ORCPT ); Tue, 24 Feb 2015 10:41:02 -0500 Received: from list by plane.gmane.org with local (Exim 4.69) (envelope-from ) id 1YQHbV-0004a4-Rc for linux-btrfs@vger.kernel.org; Tue, 24 Feb 2015 16:40:58 +0100 Received: from netz.smurf.noris.de ([213.95.21.43]) by main.gmane.org with esmtp (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Tue, 24 Feb 2015 16:40:57 +0100 Received: from matthias by netz.smurf.noris.de with local (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Tue, 24 Feb 2015 16:40:57 +0100 To: linux-btrfs@vger.kernel.org From: Matthias Urlichs Subject: Read performance problem Date: Tue, 24 Feb 2015 15:40:46 +0000 (UTC) Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-btrfs-owner@vger.kernel.org List-ID: This is 3.19, scanning an otherwise-idle filesystem on otherwise-idle disks, mounted with noatime and nodatasum. 11:41:51.703418 llistxattr("cpool/0/0/5/00556827384267bee22ab9adfb891807", "", 1024) = 0 <0.000018> 11:41:51.703472 lstat("cpool/0/0/5/00524ed0f6edbadd2a92f9366bce00c4", {st_mode=S_IFREG|0640, st_size=1024, ...}) = 0 <0.002519> 11:41:51.706044 getxattr("cpool/0/0/5/00524ed0f6edbadd2a92f9366bce00c4", "system.posix_acl_access", 0x7fff32f63090, 132) = -1 ENODATA (No data available) <0.000028> 11:41:51.706125 stat("cpool/0/0/5/00524ed0f6edbadd2a92f9366bce00c4", {st_mode=S_IFREG|0640, st_size=1024, ...}) = 0 <0.000016> 11:41:51.706170 llistxattr("cpool/0/0/5/00524ed0f6edbadd2a92f9366bce00c4", "", 1024) = 0 <0.000016> 11:41:51.706218 lstat("cpool/0/0/5/0056f3e500258614f19569d7dc014c80", {st_mode=S_IFREG|0640, st_size=351, ...}) = 0 <0.214911> 11:41:51.921209 getxattr("cpool/0/0/5/0056f3e500258614f19569d7dc014c80", "system.posix_acl_access", 0x7fff32f63090, 132) = -1 ENODATA (No data available) <0.000023> 11:41:51.921294 stat("cpool/0/0/5/0056f3e500258614f19569d7dc014c80", {st_mode=S_IFREG|0640, st_size=351, ...}) = 0 <0.000008> 11:41:51.921342 llistxattr("cpool/0/0/5/0056f3e500258614f19569d7dc014c80", "", 1024) = 0 <0.000018> 11:41:51.921397 lstat("cpool/0/0/5/0058f2c720f7567989a0f4d33e5c8eb4", {st_mode=S_IFREG|0640, st_size=65, ...}) = 0 <0.007500> 11:41:51.928959 getxattr("cpool/0/0/5/0058f2c720f7567989a0f4d33e5c8eb4", "system.posix_acl_access", 0x7fff32f63090, 132) = -1 ENODATA (No data available) <0.000027> 0.2sec is not the longest delay I've seen here. iotop says: 7977 be/4 root 115.70 K/s 0.00 B/s 0.00 % 99.95 % rsync -aPHAX backup/backuppc/. /d/backup/backuppc/ 7259 be/4 root 160.07 K/s 586.40 K/s 0.00 % 99.14 % [btrfs-transacti] OK. So how can this even happen when there is no writing process here? And what's the hold-up anyway, given that the partition easily sustains >20 MByte/s write speed? The kernel thread is in 'D' state whenever I run "ps" and causes at most 2% of CPU load. https://bugzilla.kernel.org/show_bug.cgi?id=93801