| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <pan.2014.12.16.22.19.18@googlemail.com> Date: Tue, 16 Dec 2014 22:19:18 +0000 (UTC) From: Holger Hoffstätte <holger.hoffstaette@...glemail.com> To: linux-kernel@...r.kernel.org Cc: linux-nfs@...r.kernel.org Subject: 3.18.1: broken directory with one file too many g (please CC: for followups) I just spent two hours trying to untangle a *weird* bug that I have not seen before. It might be new to 3.18.x but I don't know for sure. Apologies in advance for the long prelude but I figured I need to describe the problem scenario as precisely as possible. All this is on freshly baked 3.18.1 with Gentoo userland; the exported filesystem is ext4. On my NFS server I work with a git repo: holger>git clone ../work/kernel-patches.git Cloning into 'kernel-patches'... done. holger>cd kernel-patches holger>git status On branch master Your branch is up-to-date with 'origin/master'. nothing to commit, working directory clean holger>ll total 92K drwxr-xr-x 2 holger users 72K Dec 16 22:41 3.14/ drwxr-xr-x 2 holger users 16K Dec 16 22:41 3.18/ -rw-r--r-- 1 holger users 2.4K Dec 16 22:41 README.md holger> Looking fine! On my NFS client this directory is automounted via NFS: holger>mount | grep home tux:/home/holger on /mnt/tux/holger type nfs (rw,noatime,tcp,sloppy,vers=4,addr=192.168.100.222,clientaddr=192.168.100.128) This has worked for ages and never caused any problems. Let's see how my git repo is doing: holger>cd /mnt/tux/holger/Projects/kernel-patches holger>ll total 92K drwxr-xr-x 2 holger users 72K Dec 16 22:41 3.14/ drwxr-xr-x 2 holger users 16K Dec 16 22:41 3.18/ -rw-r--r-- 1 holger users 2.4K Dec 16 22:41 README.md holger>git status On branch master Your branch is up-to-date with 'origin/master'. Untracked files: (use "git add <file>..." to include in what will be committed) 3.14/btrfs-20 nothing added to commit but untracked files present (use "git add" to track) ..wait, what? There is no such file "btrfs-20" ! holger>ll 3.14 | head ls: cannot access 3.14/btrfs-20: No such file or directory total 4.5M -rw-r--r-- 1 holger users 3.3K Dec 16 22:41 bfq-v7r6-001-block-cgroups-kconfig-build-bits-for-BFQ-v7r6-3.14.patch -rw-r--r-- 1 holger users 219K Dec 16 22:41 bfq-v7r6-002-block-introduce-the-BFQ-v7r6-I-O-sched-for-3.14.patch -rw-r--r-- 1 holger users 41K Dec 16 22:41 bfq-v7r6-003-block-bfq-add-Early-Queue-Merge-EQM-to-BFQ-v7r6-for-3.14.0.patch -rw-r--r-- 1 holger users 237K Dec 16 22:41 bfs-454-001-sched-bfs.patch -rw-r--r-- 1 holger users 5.2K Dec 16 22:41 bfs-454-002-cpu-topology.patch -rw-r--r-- 1 holger users 13K Dec 16 22:41 bfs-454-003-smtnice-v6.patch -????????? ? ? ? ? ? btrfs-20 -rw-r--r-- 1 holger users 7.8K Dec 16 22:41 btrfs-20140114-don't-mix-the-ordered-extents-of-all-files-together-during-logging-the-inodes.patch -rw-r--r-- 1 holger users 1.2K Dec 16 22:41 btrfs-20140130-add-missing-error-check-in-incremental-send.patch holger> There is a "rogue" file messing up the directory?! This used to work until I added a specific file, so.. holger>ll 3.14/btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch -rw-r--r-- 1 holger users 2.3K Dec 16 22:41 3.14/btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch holger>stat 3.14/btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch File: ‘3.14/btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch’ Size: 2306 Blocks: 8 IO Block: 1048576 regular file Device: 18h/24d Inode: 22544856 Links: 1 Access: (0644/-rw-r--r--) Uid: ( 1000/ holger) Gid: ( 100/ users) Access: 2014-12-16 22:41:36.515665610 +0100 Modify: 2014-12-16 22:41:36.515665610 +0100 Change: 2014-12-16 22:41:36.515665610 +0100 Birth: - Looks fine..maybe try moving it to the parent? holger>mv 3.14/btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch . holger>ll total 96K drwxr-xr-x 2 holger users 72K Dec 16 22:44 3.14/ drwxr-xr-x 2 holger users 16K Dec 16 22:41 3.18/ -rw-r--r-- 1 holger users 2.4K Dec 16 22:41 README.md -rw-r--r-- 1 holger users 2.3K Dec 16 22:41 btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch holger>git status On branch master Your branch is up-to-date with 'origin/master'. Changes not staged for commit: (use "git add/rm <file>..." to update what will be committed) (use "git checkout -- <file>..." to discard changes in working directory) deleted: 3.14/btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch Untracked files: (use "git add <file>..." to include in what will be committed) btrfs-20141216-fix-a-warning-of-qgroup-account-on-shared-extents.patch no changes added to commit (use "git add" and/or "git commit -a") holger>ll 3.14 | head total 4.5M -rw-r--r-- 1 holger users 3.3K Dec 16 22:41 bfq-v7r6-001-block-cgroups-kconfig-build-bits-for-BFQ-v7r6-3.14.patch -rw-r--r-- 1 holger users 219K Dec 16 22:41 bfq-v7r6-002-block-introduce-the-BFQ-v7r6-I-O-sched-for-3.14.patch -rw-r--r-- 1 holger users 41K Dec 16 22:41 bfq-v7r6-003-block-bfq-add-Early-Queue-Merge-EQM-to-BFQ-v7r6-for-3.14.0.patch -rw-r--r-- 1 holger users 237K Dec 16 22:41 bfs-454-001-sched-bfs.patch -rw-r--r-- 1 holger users 5.2K Dec 16 22:41 bfs-454-002-cpu-topology.patch -rw-r--r-- 1 holger users 13K Dec 16 22:41 bfs-454-003-smtnice-v6.patch -rw-r--r-- 1 holger users 7.8K Dec 16 22:41 btrfs-20140114-don't-mix-the-ordered-extents-of-all-files-together-during-logging-the-inodes.patch -rw-r--r-- 1 holger users 1.2K Dec 16 22:41 btrfs-20140130-add-missing-error-check-in-incremental-send.patch -rw-r--r-- 1 holger users 5.2K Dec 16 22:41 btrfs-20140130-fix-32-64-bit-problem-with-BTRFS_SET_RECEIVED_SUBVOL-ioctl.patch I can move it back into 3.14/ and the directory is messed up again. All this is reproducible, in different export directories. Any ideas what this might be? A direntry hash collision maybe? There is a large number of files starting with btrfs-2014xxyy-.. but with the typical kernel patch names (some quite long), so that would be pretty bad. Also everything works locally on ext4 without problems, so I suspect it's an isolated NFS problem. Things that did not work - same results every time: - dropping caches on client & server - restarting NFS on client & server Any suggestions welcome; I'll gladly test patches. thanks Holger -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@...r.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists