| To: | Lachlan McIlroy <lachlan@xxxxxxx> |
|---|---|
| Subject: | Re: [PATCH 3/4] simplify xfs_vn_getattr |
| From: | Christoph Hellwig <hch@xxxxxx> |
| Date: | Wed, 19 Sep 2007 13:04:38 +0200 |
| Cc: | Christoph Hellwig <hch@xxxxxx>, xfs@xxxxxxxxxxx |
| In-reply-to: | <46F08F98.9070102@xxxxxxx> |
| References: | <20070914162757.GD7110@xxxxxx> <46F08F98.9070102@xxxxxxx> |
| Sender: | xfs-bounce@xxxxxxxxxxx |
| User-agent: | Mutt/1.3.28i |
On Wed, Sep 19, 2007 at 12:55:20PM +1000, Lachlan McIlroy wrote: > Why do we copy the atime from the inode and not the xfs_inode? > > I think I see what's the deal here - the atime in the inode is > the authoritative atime. It gets updated from various places and > we synchronise it to the xfs inode before flushing the xfs inode > to disk. This means we shouldn't be using the atime in the > xfs_inode because it will be stale - is this correct? Yes. atime is updated by the vfs without callouts to the filesystem in various places unfortunately. |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: [PATCH 1/2] kill unnessecary ioops indirection, Christoph Hellwig |
|---|---|
| Next by Date: | Re: [PATCH 3/4] simplify xfs_vn_getattr, Lachlan McIlroy |
| Previous by Thread: | Re: [PATCH 3/4] simplify xfs_vn_getattr, Lachlan McIlroy |
| Next by Thread: | [PATCH 1/2] kill unnessecary ioops indirection, Christoph Hellwig |
| Indexes: | [Date] [Thread] [Top] [All Lists] |