aboutsummaryrefslogtreecommitdiffstats
path: root/fs/xfs/xfs_inode.c
diff options
context:
space:
mode:
authorLachlan McIlroy <lachlan@sgi.com>2008-09-25 22:17:57 -0400
committerLachlan McIlroy <lachlan@redback.melbourne.sgi.com>2008-09-25 22:17:57 -0400
commit71a8c87fb300b601eacf7a86cc6c6322fe827bfd (patch)
tree9f0e12933f6b2d74a452b69438693527853486c8 /fs/xfs/xfs_inode.c
parentf1ccd2955157e1aff992f6aaaba0944209076220 (diff)
[XFS] Remove xfs_iext_irec_compact_full()
Yet another bug was found in xfs_iext_irec_compact_full() and while the source of the bug was found it wasn't an easy task to track it down because the conditions are very difficult to reproduce. A HUGE thank-you goes to Russell Cattelan and Eric Sandeen for their significant effort in tracking down the source of this corruption. xfs_iext_irec_compact_full() and xfs_iext_irec_compact_pages() are almost identical - they both compact indirect extent lists by moving extents from subsequent buffers into earlier ones. xfs_iext_irec_compact_pages() only moves extents if all of the extents in the next buffer will fit into the empty space in the buffer before it. xfs_iext_irec_compact_full() will go a step further and move part of the next buffer if all the extents wont fit. It will then shift the remaining extents in the next buffer up to the start of the buffer. The bug here was that we did not update er_extoff and this caused extent list corruption. It does not appear that this extra functionality gains us much. Calling xfs_iext_irec_compact_pages() instead will do a good enough job at compacting the indirect list and will be quicker too. For the case in xfs_iext_indirect_to_direct() the total number of extents in the indirect list will fit into one buffer so we will never need the extra functionality of xfs_iext_irec_compact_full() there. Also xfs_iext_irec_compact_pages() doesn't need to do a memmove() (the buffers will never overlap) so we don't want the performance hit that can incur. SGI-PV: 987159 SGI-Modid: xfs-linux-melb:xfs-kern:32166a Signed-off-by: Lachlan McIlroy <lachlan@sgi.com> Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Diffstat (limited to 'fs/xfs/xfs_inode.c')
-rw-r--r--fs/xfs/xfs_inode.c95
1 files changed, 3 insertions, 92 deletions
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 419cfc2eacb3..dbd9cef852ec 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -4118,7 +4118,7 @@ xfs_iext_indirect_to_direct(
4118 ASSERT(nextents <= XFS_LINEAR_EXTS); 4118 ASSERT(nextents <= XFS_LINEAR_EXTS);
4119 size = nextents * sizeof(xfs_bmbt_rec_t); 4119 size = nextents * sizeof(xfs_bmbt_rec_t);
4120 4120
4121 xfs_iext_irec_compact_full(ifp); 4121 xfs_iext_irec_compact_pages(ifp);
4122 ASSERT(ifp->if_real_bytes == XFS_IEXT_BUFSZ); 4122 ASSERT(ifp->if_real_bytes == XFS_IEXT_BUFSZ);
4123 4123
4124 ep = ifp->if_u1.if_ext_irec->er_extbuf; 4124 ep = ifp->if_u1.if_ext_irec->er_extbuf;
@@ -4449,8 +4449,7 @@ xfs_iext_irec_remove(
4449 * compaction policy is as follows: 4449 * compaction policy is as follows:
4450 * 4450 *
4451 * Full Compaction: Extents fit into a single page (or inline buffer) 4451 * Full Compaction: Extents fit into a single page (or inline buffer)
4452 * Full Compaction: Extents occupy less than 10% of allocated space 4452 * Partial Compaction: Extents occupy less than 50% of allocated space
4453 * Partial Compaction: Extents occupy > 10% and < 50% of allocated space
4454 * No Compaction: Extents occupy at least 50% of allocated space 4453 * No Compaction: Extents occupy at least 50% of allocated space
4455 */ 4454 */
4456void 4455void
@@ -4471,8 +4470,6 @@ xfs_iext_irec_compact(
4471 xfs_iext_direct_to_inline(ifp, nextents); 4470 xfs_iext_direct_to_inline(ifp, nextents);
4472 } else if (nextents <= XFS_LINEAR_EXTS) { 4471 } else if (nextents <= XFS_LINEAR_EXTS) {
4473 xfs_iext_indirect_to_direct(ifp); 4472 xfs_iext_indirect_to_direct(ifp);
4474 } else if (nextents < (nlists * XFS_LINEAR_EXTS) >> 3) {
4475 xfs_iext_irec_compact_full(ifp);
4476 } else if (nextents < (nlists * XFS_LINEAR_EXTS) >> 1) { 4473 } else if (nextents < (nlists * XFS_LINEAR_EXTS) >> 1) {
4477 xfs_iext_irec_compact_pages(ifp); 4474 xfs_iext_irec_compact_pages(ifp);
4478 } 4475 }
@@ -4496,7 +4493,7 @@ xfs_iext_irec_compact_pages(
4496 erp_next = erp + 1; 4493 erp_next = erp + 1;
4497 if (erp_next->er_extcount <= 4494 if (erp_next->er_extcount <=
4498 (XFS_LINEAR_EXTS - erp->er_extcount)) { 4495 (XFS_LINEAR_EXTS - erp->er_extcount)) {
4499 memmove(&erp->er_extbuf[erp->er_extcount], 4496 memcpy(&erp->er_extbuf[erp->er_extcount],
4500 erp_next->er_extbuf, erp_next->er_extcount * 4497 erp_next->er_extbuf, erp_next->er_extcount *
4501 sizeof(xfs_bmbt_rec_t)); 4498 sizeof(xfs_bmbt_rec_t));
4502 erp->er_extcount += erp_next->er_extcount; 4499 erp->er_extcount += erp_next->er_extcount;
@@ -4516,92 +4513,6 @@ xfs_iext_irec_compact_pages(
4516} 4513}
4517 4514
4518/* 4515/*
4519 * Fully compact the extent records managed by the indirection array.
4520 */
4521void
4522xfs_iext_irec_compact_full(
4523 xfs_ifork_t *ifp) /* inode fork pointer */
4524{
4525 xfs_bmbt_rec_host_t *ep, *ep_next; /* extent record pointers */
4526 xfs_ext_irec_t *erp, *erp_next; /* extent irec pointers */
4527 int erp_idx = 0; /* extent irec index */
4528 int ext_avail; /* empty entries in ex list */
4529 int ext_diff; /* number of exts to add */
4530 int nlists; /* number of irec's (ex lists) */
4531
4532 ASSERT(ifp->if_flags & XFS_IFEXTIREC);
4533
4534 nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
4535 erp = ifp->if_u1.if_ext_irec;
4536 ep = &erp->er_extbuf[erp->er_extcount];
4537 erp_next = erp + 1;
4538 ep_next = erp_next->er_extbuf;
4539
4540 while (erp_idx < nlists - 1) {
4541 /*
4542 * Check how many extent records are available in this irec.
4543 * If there is none skip the whole exercise.
4544 */
4545 ext_avail = XFS_LINEAR_EXTS - erp->er_extcount;
4546 if (ext_avail) {
4547
4548 /*
4549 * Copy over as many as possible extent records into
4550 * the previous page.
4551 */
4552 ext_diff = MIN(ext_avail, erp_next->er_extcount);
4553 memcpy(ep, ep_next, ext_diff * sizeof(xfs_bmbt_rec_t));
4554 erp->er_extcount += ext_diff;
4555 erp_next->er_extcount -= ext_diff;
4556
4557 /*
4558 * If the next irec is empty now we can simply
4559 * remove it.
4560 */
4561 if (erp_next->er_extcount == 0) {
4562 /*
4563 * Free page before removing extent record
4564 * so er_extoffs don't get modified in
4565 * xfs_iext_irec_remove.
4566 */
4567 kmem_free(erp_next->er_extbuf);
4568 erp_next->er_extbuf = NULL;
4569 xfs_iext_irec_remove(ifp, erp_idx + 1);
4570 erp = &ifp->if_u1.if_ext_irec[erp_idx];
4571 nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ;
4572
4573 /*
4574 * If the next irec is not empty move up the content
4575 * that has not been copied to the previous page to
4576 * the beggining of this one.
4577 */
4578 } else {
4579 memmove(erp_next->er_extbuf, &ep_next[ext_diff],
4580 erp_next->er_extcount *
4581 sizeof(xfs_bmbt_rec_t));
4582 ep_next = erp_next->er_extbuf;
4583 memset(&ep_next[erp_next->er_extcount], 0,
4584 (XFS_LINEAR_EXTS -
4585 erp_next->er_extcount) *
4586 sizeof(xfs_bmbt_rec_t));
4587 erp_next->er_extoff += ext_diff;
4588 }
4589 }
4590
4591 if (erp->er_extcount == XFS_LINEAR_EXTS) {
4592 erp_idx++;
4593 if (erp_idx < nlists)
4594 erp = &ifp->if_u1.if_ext_irec[erp_idx];
4595 else
4596 break;
4597 }
4598 ep = &erp->er_extbuf[erp->er_extcount];
4599 erp_next = erp + 1;
4600 ep_next = erp_next->er_extbuf;
4601 }
4602}
4603
4604/*
4605 * This is called to update the er_extoff field in the indirection 4516 * This is called to update the er_extoff field in the indirection
4606 * array when extents have been added or removed from one of the 4517 * array when extents have been added or removed from one of the
4607 * extent lists. erp_idx contains the irec index to begin updating 4518 * extent lists. erp_idx contains the irec index to begin updating