The ls-time test in coreutils-7.5 fails on btrfs filesystems. I've tracked it down to ctime not being updated when adding hardlinks. The following script (based on the coreutils test) shows the problem. #!/bin/sh touch a b # Make sure ctime is at least one second apart sleep 2 ln b c # Now, C should have ctime more recent than A. set `ls -ct a b` if test "$*" = 'b a'; then : ok else # In spite of documentation, (e.g., stat(2)), neither link nor chmod # update a file's st_ctime on SunOS4.1.4. cat >&2 << \EOF failed ls ctime test -- this failure is expected at least for SunOS4.1.4 and for tmpfs file systems on Solaris 5.5.1. In the output below, `b' should have had a ctime more recent than that of `a', but does not. EOF #' ls -ctl --full-time a b fi I'm not familiar with the btrfs code base at all but the following patch fixes the problem and shouldn't introduce any regressions as far as I can tell. >From d66f0bcf30785e821973885f693cb285857055c4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bryan=20=C3=98stergaard?= <kloeri@xxxxxxxxxxx> Date: Tue, 25 Aug 2009 14:04:42 +0200 Subject: [PATCH] Fix ctime issue in btrfs btrfs doesn't update ctime when creating new hardlinks as required by stat(2). --- fs/btrfs/inode.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 59cba18..b045d42 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3879,6 +3879,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, drop_inode = 1; btrfs_update_inode_block_group(trans, dir); + inode->i_ctime = CURRENT_TIME; err = btrfs_update_inode(trans, root, inode); if (err) -- 1.6.4.1 Regards, Bryan Østergaard -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
