| To: | linux-xfs@xxxxxxxxxxx |
|---|---|
| Subject: | reserving contiguous files |
| From: | "Rick Smith" <rgsmith72@xxxxxxxxxxx> |
| Date: | Thu, 13 Mar 2003 11:43:21 -0800 |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
|
Hello, I am trying to write sequences of contiguous files in XFS to minimize the drive head seeks that must occur during file reads. Using the XFS_RESVSP64 ioctl to reserve the set of files ahead of time, I am able to get individual files allocated with contiguous extents. However, the allocation of one file versus another (reserved immediately after each other) seems to be somewhat erratic. Is there a way to guarantee or further lock the allocation of one file with respect to another in a contiguous manner? Looking at xfs_vnodeops.c and xfs_bmap.c could this be possibly be achieved by initializing the firstfsb structure in xfs_alloc_file_space() only in the first reserve call for a group allocation? The value of this structure could be maintained for consecutive calls to allow the contiguous allocation of a group of files based on the preceeding reserve ioctl. This behavior could be added as an additional ioctl call (RESVSP64CONTIG?) to maintain the current working code. I am going to experiment with this, but I would appreciate suggestions/comments. Thanks. Rick Smith _________________________________________________________________ The new MSN 8: advanced junk mail protection and 2 months FREE* http://join.msn.com/?page=features/junkmail |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | [Bug 230] New: umount hangs after high disk load, bugzilla-daemon |
|---|---|
| Next by Date: | [NEWBIE] Why aren't directory mtimes updated like other filesystems?, Kevin P. Fleming |
| Previous by Thread: | [Bug 230] New: umount hangs after high disk load, bugzilla-daemon |
| Next by Thread: | [NEWBIE] Why aren't directory mtimes updated like other filesystems?, Kevin P. Fleming |
| Indexes: | [Date] [Thread] [Top] [All Lists] |