| To: | Alex Elder <aelder@xxxxxxx> |
|---|---|
| Subject: | Re: [PATCH 5/8] xfsprogs: Drop root inode refrerence in libxfs_umount() |
| From: | Christoph Hellwig <hch@xxxxxxxxxxxxx> |
| Date: | Sun, 13 Nov 2011 07:07:04 -0500 |
| Cc: | xfs@xxxxxxxxxxx, Kevan Rehm <kfr@xxxxxxx> |
| In-reply-to: | <8634b2514c7642f4cbcfdaf3b6852e129417c372.1320955676.git.aelder@xxxxxxx> |
| References: | <3ed89446f6dc1231ec1de712f242522052952b7e.1320955675.git.aelder@xxxxxxx> <8634b2514c7642f4cbcfdaf3b6852e129417c372.1320955676.git.aelder@xxxxxxx> |
| User-agent: | Mutt/1.5.21 (2010-09-15) |
On Thu, Nov 10, 2011 at 02:35:15PM -0600, Alex Elder wrote: > From: Kevan Rehm <kfr@xxxxxxx> > > Routine libxfs_umount() did not call libxfs_iput for the m_rootip > inode, so updates made to that inode could be lost. This adds the > missing call, and re-initializes the m_rootip pointer to be null. > > Since the root inode reference is now dropped by libxfs_umount(), it > should *not* be dropped in mkfs parseproto(). Looks fine in general, but as mentioned before live would be a lot easier if we simply made the root inode a local variable in libxfs_mount. In fact I wonder if we even want to do that, there seems very little reason to even do the root inode iget there - the only caller requesting it will fall back if it fails. |
| Previous by Date: | Re: [PATCH 4/8] xfsprogs: mkfs.xfs: let libxfs_umount() do its thing, Christoph Hellwig |
|---|---|
| Next by Date: | Re: [PATCH 6/8] xfsprogs: xfs_db: unmount fs before exiting, Christoph Hellwig |
| Previous by Thread: | [PATCH 5/8] xfsprogs: Drop root inode refrerence in libxfs_umount(), Alex Elder |
| Next by Thread: | [PATCH 8/8] xfsprogs: xfs_db: add new "resvsp" command, Alex Elder |
| Indexes: | [Date] [Thread] [Top] [All Lists] |