diff options
author | Dave Chinner <david@fromorbit.com> | 2010-05-20 22:07:08 -0400 |
---|---|---|
committer | Alex Elder <aelder@sgi.com> | 2010-05-24 11:34:00 -0400 |
commit | ed3b4d6cdc81e8feefdbfa3c584614be301b6d39 (patch) | |
tree | 5b8cd5735dfbc5eb834f96d25a8eb587186715be /fs/xfs | |
parent | 955833cf2ad0aa39b336e853cad212d867199984 (diff) |
xfs: Improve scalability of busy extent tracking
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Diffstat (limited to 'fs/xfs')
-rw-r--r-- | fs/xfs/linux-2.6/xfs_buf.c | 9 | ||||
-rw-r--r-- | fs/xfs/linux-2.6/xfs_quotaops.c | 1 | ||||
-rw-r--r-- | fs/xfs/linux-2.6/xfs_trace.h | 83 | ||||
-rw-r--r-- | fs/xfs/xfs_ag.h | 24 | ||||
-rw-r--r-- | fs/xfs/xfs_alloc.c | 357 | ||||
-rw-r--r-- | fs/xfs/xfs_alloc.h | 7 | ||||
-rw-r--r-- | fs/xfs/xfs_alloc_btree.c | 2 | ||||
-rw-r--r-- | fs/xfs/xfs_trans.c | 41 | ||||
-rw-r--r-- | fs/xfs/xfs_trans.h | 35 | ||||
-rw-r--r-- | fs/xfs/xfs_trans_item.c | 109 | ||||
-rw-r--r-- | fs/xfs/xfs_trans_priv.h | 4 |
11 files changed, 350 insertions, 322 deletions
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index f01de3c55c43..649ade8ef598 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c | |||
@@ -37,6 +37,7 @@ | |||
37 | 37 | ||
38 | #include "xfs_sb.h" | 38 | #include "xfs_sb.h" |
39 | #include "xfs_inum.h" | 39 | #include "xfs_inum.h" |
40 | #include "xfs_log.h" | ||
40 | #include "xfs_ag.h" | 41 | #include "xfs_ag.h" |
41 | #include "xfs_dmapi.h" | 42 | #include "xfs_dmapi.h" |
42 | #include "xfs_mount.h" | 43 | #include "xfs_mount.h" |
@@ -850,6 +851,12 @@ xfs_buf_lock_value( | |||
850 | * Note that this in no way locks the underlying pages, so it is only | 851 | * Note that this in no way locks the underlying pages, so it is only |
851 | * useful for synchronizing concurrent use of buffer objects, not for | 852 | * useful for synchronizing concurrent use of buffer objects, not for |
852 | * synchronizing independent access to the underlying pages. | 853 | * synchronizing independent access to the underlying pages. |
854 | * | ||
855 | * If we come across a stale, pinned, locked buffer, we know that we | ||
856 | * are being asked to lock a buffer that has been reallocated. Because | ||
857 | * it is pinned, we know that the log has not been pushed to disk and | ||
858 | * hence it will still be locked. Rather than sleeping until someone | ||
859 | * else pushes the log, push it ourselves before trying to get the lock. | ||
853 | */ | 860 | */ |
854 | void | 861 | void |
855 | xfs_buf_lock( | 862 | xfs_buf_lock( |
@@ -857,6 +864,8 @@ xfs_buf_lock( | |||
857 | { | 864 | { |
858 | trace_xfs_buf_lock(bp, _RET_IP_); | 865 | trace_xfs_buf_lock(bp, _RET_IP_); |
859 | 866 | ||
867 | if (atomic_read(&bp->b_pin_count) && (bp->b_flags & XBF_STALE)) | ||
868 | xfs_log_force(bp->b_mount, 0); | ||
860 | if (atomic_read(&bp->b_io_remaining)) | 869 | if (atomic_read(&bp->b_io_remaining)) |
861 | blk_run_address_space(bp->b_target->bt_mapping); | 870 | blk_run_address_space(bp->b_target->bt_mapping); |
862 | down(&bp->b_sema); | 871 | down(&bp->b_sema); |
diff --git a/fs/xfs/linux-2.6/xfs_quotaops.c b/fs/xfs/linux-2.6/xfs_quotaops.c index 1947514ce1ad..2e73688dae9c 100644 --- a/fs/xfs/linux-2.6/xfs_quotaops.c +++ b/fs/xfs/linux-2.6/xfs_quotaops.c | |||
@@ -19,6 +19,7 @@ | |||
19 | #include "xfs_dmapi.h" | 19 | #include "xfs_dmapi.h" |
20 | #include "xfs_sb.h" | 20 | #include "xfs_sb.h" |
21 | #include "xfs_inum.h" | 21 | #include "xfs_inum.h" |
22 | #include "xfs_log.h" | ||
22 | #include "xfs_ag.h" | 23 | #include "xfs_ag.h" |
23 | #include "xfs_mount.h" | 24 | #include "xfs_mount.h" |
24 | #include "xfs_quota.h" | 25 | #include "xfs_quota.h" |
diff --git a/fs/xfs/linux-2.6/xfs_trace.h b/fs/xfs/linux-2.6/xfs_trace.h index 8a319cfd2901..ff6bc797baf2 100644 --- a/fs/xfs/linux-2.6/xfs_trace.h +++ b/fs/xfs/linux-2.6/xfs_trace.h | |||
@@ -1059,83 +1059,112 @@ TRACE_EVENT(xfs_bunmap, | |||
1059 | 1059 | ||
1060 | ); | 1060 | ); |
1061 | 1061 | ||
1062 | #define XFS_BUSY_SYNC \ | ||
1063 | { 0, "async" }, \ | ||
1064 | { 1, "sync" } | ||
1065 | |||
1062 | TRACE_EVENT(xfs_alloc_busy, | 1066 | TRACE_EVENT(xfs_alloc_busy, |
1063 | TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t agbno, | 1067 | TP_PROTO(struct xfs_trans *trans, xfs_agnumber_t agno, |
1064 | xfs_extlen_t len, int slot), | 1068 | xfs_agblock_t agbno, xfs_extlen_t len, int sync), |
1065 | TP_ARGS(mp, agno, agbno, len, slot), | 1069 | TP_ARGS(trans, agno, agbno, len, sync), |
1066 | TP_STRUCT__entry( | 1070 | TP_STRUCT__entry( |
1067 | __field(dev_t, dev) | 1071 | __field(dev_t, dev) |
1072 | __field(struct xfs_trans *, tp) | ||
1073 | __field(int, tid) | ||
1068 | __field(xfs_agnumber_t, agno) | 1074 | __field(xfs_agnumber_t, agno) |
1069 | __field(xfs_agblock_t, agbno) | 1075 | __field(xfs_agblock_t, agbno) |
1070 | __field(xfs_extlen_t, len) | 1076 | __field(xfs_extlen_t, len) |
1071 | __field(int, slot) | 1077 | __field(int, sync) |
1072 | ), | 1078 | ), |
1073 | TP_fast_assign( | 1079 | TP_fast_assign( |
1074 | __entry->dev = mp->m_super->s_dev; | 1080 | __entry->dev = trans->t_mountp->m_super->s_dev; |
1081 | __entry->tp = trans; | ||
1082 | __entry->tid = trans->t_ticket->t_tid; | ||
1075 | __entry->agno = agno; | 1083 | __entry->agno = agno; |
1076 | __entry->agbno = agbno; | 1084 | __entry->agbno = agbno; |
1077 | __entry->len = len; | 1085 | __entry->len = len; |
1078 | __entry->slot = slot; | 1086 | __entry->sync = sync; |
1079 | ), | 1087 | ), |
1080 | TP_printk("dev %d:%d agno %u agbno %u len %u slot %d", | 1088 | TP_printk("dev %d:%d trans 0x%p tid 0x%x agno %u agbno %u len %u %s", |
1081 | MAJOR(__entry->dev), MINOR(__entry->dev), | 1089 | MAJOR(__entry->dev), MINOR(__entry->dev), |
1090 | __entry->tp, | ||
1091 | __entry->tid, | ||
1082 | __entry->agno, | 1092 | __entry->agno, |
1083 | __entry->agbno, | 1093 | __entry->agbno, |
1084 | __entry->len, | 1094 | __entry->len, |
1085 | __entry->slot) | 1095 | __print_symbolic(__entry->sync, XFS_BUSY_SYNC)) |
1086 | 1096 | ||
1087 | ); | 1097 | ); |
1088 | 1098 | ||
1089 | #define XFS_BUSY_STATES \ | ||
1090 | { 0, "found" }, \ | ||
1091 | { 1, "missing" } | ||
1092 | |||
1093 | TRACE_EVENT(xfs_alloc_unbusy, | 1099 | TRACE_EVENT(xfs_alloc_unbusy, |
1094 | TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, | 1100 | TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, |
1095 | int slot, int found), | 1101 | xfs_agblock_t agbno, xfs_extlen_t len), |
1096 | TP_ARGS(mp, agno, slot, found), | 1102 | TP_ARGS(mp, agno, agbno, len), |
1097 | TP_STRUCT__entry( | 1103 | TP_STRUCT__entry( |
1098 | __field(dev_t, dev) | 1104 | __field(dev_t, dev) |
1099 | __field(xfs_agnumber_t, agno) | 1105 | __field(xfs_agnumber_t, agno) |
1100 | __field(int, slot) | 1106 | __field(xfs_agblock_t, agbno) |
1101 | __field(int, found) | 1107 | __field(xfs_extlen_t, len) |
1102 | ), | 1108 | ), |
1103 | TP_fast_assign( | 1109 | TP_fast_assign( |
1104 | __entry->dev = mp->m_super->s_dev; | 1110 | __entry->dev = mp->m_super->s_dev; |
1105 | __entry->agno = agno; | 1111 | __entry->agno = agno; |
1106 | __entry->slot = slot; | 1112 | __entry->agbno = agbno; |
1107 | __entry->found = found; | 1113 | __entry->len = len; |
1108 | ), | 1114 | ), |
1109 | TP_printk("dev %d:%d agno %u slot %d %s", | 1115 | TP_printk("dev %d:%d agno %u agbno %u len %u", |
1110 | MAJOR(__entry->dev), MINOR(__entry->dev), | 1116 | MAJOR(__entry->dev), MINOR(__entry->dev), |
1111 | __entry->agno, | 1117 | __entry->agno, |
1112 | __entry->slot, | 1118 | __entry->agbno, |
1113 | __print_symbolic(__entry->found, XFS_BUSY_STATES)) | 1119 | __entry->len) |
1114 | ); | 1120 | ); |
1115 | 1121 | ||
1122 | #define XFS_BUSY_STATES \ | ||
1123 | { 0, "missing" }, \ | ||
1124 | { 1, "found" } | ||
1125 | |||
1116 | TRACE_EVENT(xfs_alloc_busysearch, | 1126 | TRACE_EVENT(xfs_alloc_busysearch, |
1117 | TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t agbno, | 1127 | TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, |
1118 | xfs_extlen_t len, xfs_lsn_t lsn), | 1128 | xfs_agblock_t agbno, xfs_extlen_t len, int found), |
1119 | TP_ARGS(mp, agno, agbno, len, lsn), | 1129 | TP_ARGS(mp, agno, agbno, len, found), |
1120 | TP_STRUCT__entry( | 1130 | TP_STRUCT__entry( |
1121 | __field(dev_t, dev) | 1131 | __field(dev_t, dev) |
1122 | __field(xfs_agnumber_t, agno) | 1132 | __field(xfs_agnumber_t, agno) |
1123 | __field(xfs_agblock_t, agbno) | 1133 | __field(xfs_agblock_t, agbno) |
1124 | __field(xfs_extlen_t, len) | 1134 | __field(xfs_extlen_t, len) |
1125 | __field(xfs_lsn_t, lsn) | 1135 | __field(int, found) |
1126 | ), | 1136 | ), |
1127 | TP_fast_assign( | 1137 | TP_fast_assign( |
1128 | __entry->dev = mp->m_super->s_dev; | 1138 | __entry->dev = mp->m_super->s_dev; |
1129 | __entry->agno = agno; | 1139 | __entry->agno = agno; |
1130 | __entry->agbno = agbno; | 1140 | __entry->agbno = agbno; |
1131 | __entry->len = len; | 1141 | __entry->len = len; |
1132 | __entry->lsn = lsn; | 1142 | __entry->found = found; |
1133 | ), | 1143 | ), |
1134 | TP_printk("dev %d:%d agno %u agbno %u len %u force lsn 0x%llx", | 1144 | TP_printk("dev %d:%d agno %u agbno %u len %u %s", |
1135 | MAJOR(__entry->dev), MINOR(__entry->dev), | 1145 | MAJOR(__entry->dev), MINOR(__entry->dev), |
1136 | __entry->agno, | 1146 | __entry->agno, |
1137 | __entry->agbno, | 1147 | __entry->agbno, |
1138 | __entry->len, | 1148 | __entry->len, |
1149 | __print_symbolic(__entry->found, XFS_BUSY_STATES)) | ||
1150 | ); | ||
1151 | |||
1152 | TRACE_EVENT(xfs_trans_commit_lsn, | ||
1153 | TP_PROTO(struct xfs_trans *trans), | ||
1154 | TP_ARGS(trans), | ||
1155 | TP_STRUCT__entry( | ||
1156 | __field(dev_t, dev) | ||
1157 | __field(struct xfs_trans *, tp) | ||
1158 | __field(xfs_lsn_t, lsn) | ||
1159 | ), | ||
1160 | TP_fast_assign( | ||
1161 | __entry->dev = trans->t_mountp->m_super->s_dev; | ||
1162 | __entry->tp = trans; | ||
1163 | __entry->lsn = trans->t_commit_lsn; | ||
1164 | ), | ||
1165 | TP_printk("dev %d:%d trans 0x%p commit_lsn 0x%llx", | ||
1166 | MAJOR(__entry->dev), MINOR(__entry->dev), | ||
1167 | __entry->tp, | ||
1139 | __entry->lsn) | 1168 | __entry->lsn) |
1140 | ); | 1169 | ); |
1141 | 1170 | ||
diff --git a/fs/xfs/xfs_ag.h b/fs/xfs/xfs_ag.h index abb8222b88c9..401f364ad36c 100644 --- a/fs/xfs/xfs_ag.h +++ b/fs/xfs/xfs_ag.h | |||
@@ -175,14 +175,20 @@ typedef struct xfs_agfl { | |||
175 | } xfs_agfl_t; | 175 | } xfs_agfl_t; |
176 | 176 | ||
177 | /* | 177 | /* |
178 | * Busy block/extent entry. Used in perag to mark blocks that have been freed | 178 | * Busy block/extent entry. Indexed by a rbtree in perag to mark blocks that |
179 | * but whose transactions aren't committed to disk yet. | 179 | * have been freed but whose transactions aren't committed to disk yet. |
180 | * | ||
181 | * Note that we use the transaction ID to record the transaction, not the | ||
182 | * transaction structure itself. See xfs_alloc_busy_insert() for details. | ||
180 | */ | 183 | */ |
181 | typedef struct xfs_perag_busy { | 184 | struct xfs_busy_extent { |
182 | xfs_agblock_t busy_start; | 185 | struct rb_node rb_node; /* ag by-bno indexed search tree */ |
183 | xfs_extlen_t busy_length; | 186 | struct list_head list; /* transaction busy extent list */ |
184 | struct xfs_trans *busy_tp; /* transaction that did the free */ | 187 | xfs_agnumber_t agno; |
185 | } xfs_perag_busy_t; | 188 | xfs_agblock_t bno; |
189 | xfs_extlen_t length; | ||
190 | xlog_tid_t tid; /* transaction that created this */ | ||
191 | }; | ||
186 | 192 | ||
187 | /* | 193 | /* |
188 | * Per-ag incore structure, copies of information in agf and agi, | 194 | * Per-ag incore structure, copies of information in agf and agi, |
@@ -216,7 +222,8 @@ typedef struct xfs_perag { | |||
216 | xfs_agino_t pagl_leftrec; | 222 | xfs_agino_t pagl_leftrec; |
217 | xfs_agino_t pagl_rightrec; | 223 | xfs_agino_t pagl_rightrec; |
218 | #ifdef __KERNEL__ | 224 | #ifdef __KERNEL__ |
219 | spinlock_t pagb_lock; /* lock for pagb_list */ | 225 | spinlock_t pagb_lock; /* lock for pagb_tree */ |
226 | struct rb_root pagb_tree; /* ordered tree of busy extents */ | ||
220 | 227 | ||
221 | atomic_t pagf_fstrms; /* # of filestreams active in this AG */ | 228 | atomic_t pagf_fstrms; /* # of filestreams active in this AG */ |
222 | 229 | ||
@@ -226,7 +233,6 @@ typedef struct xfs_perag { | |||
226 | int pag_ici_reclaimable; /* reclaimable inodes */ | 233 | int pag_ici_reclaimable; /* reclaimable inodes */ |
227 | #endif | 234 | #endif |
228 | int pagb_count; /* pagb slots in use */ | 235 | int pagb_count; /* pagb slots in use */ |
229 | xfs_perag_busy_t pagb_list[XFS_PAGB_NUM_SLOTS]; /* unstable blocks */ | ||
230 | } xfs_perag_t; | 236 | } xfs_perag_t; |
231 | 237 | ||
232 | /* | 238 | /* |
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c index 94cddbfb2560..a7fbe8a99b12 100644 --- a/fs/xfs/xfs_alloc.c +++ b/fs/xfs/xfs_alloc.c | |||
@@ -46,11 +46,9 @@ | |||
46 | #define XFSA_FIXUP_BNO_OK 1 | 46 | #define XFSA_FIXUP_BNO_OK 1 |
47 | #define XFSA_FIXUP_CNT_OK 2 | 47 | #define XFSA_FIXUP_CNT_OK 2 |
48 | 48 | ||
49 | STATIC void | 49 | static int |
50 | xfs_alloc_search_busy(xfs_trans_t *tp, | 50 | xfs_alloc_busy_search(struct xfs_mount *mp, xfs_agnumber_t agno, |
51 | xfs_agnumber_t agno, | 51 | xfs_agblock_t bno, xfs_extlen_t len); |
52 | xfs_agblock_t bno, | ||
53 | xfs_extlen_t len); | ||
54 | 52 | ||
55 | /* | 53 | /* |
56 | * Prototypes for per-ag allocation routines | 54 | * Prototypes for per-ag allocation routines |
@@ -540,9 +538,16 @@ xfs_alloc_ag_vextent( | |||
540 | be32_to_cpu(agf->agf_length)); | 538 | be32_to_cpu(agf->agf_length)); |
541 | xfs_alloc_log_agf(args->tp, args->agbp, | 539 | xfs_alloc_log_agf(args->tp, args->agbp, |
542 | XFS_AGF_FREEBLKS); | 540 | XFS_AGF_FREEBLKS); |
543 | /* search the busylist for these blocks */ | 541 | /* |
544 | xfs_alloc_search_busy(args->tp, args->agno, | 542 | * Search the busylist for these blocks and mark the |
545 | args->agbno, args->len); | 543 | * transaction as synchronous if blocks are found. This |
544 | * avoids the need to block due to a synchronous log | ||
545 | * force to ensure correct ordering as the synchronous | ||
546 | * transaction will guarantee that for us. | ||
547 | */ | ||
548 | if (xfs_alloc_busy_search(args->mp, args->agno, | ||
549 | args->agbno, args->len)) | ||
550 | xfs_trans_set_sync(args->tp); | ||
546 | } | 551 | } |
547 | if (!args->isfl) | 552 | if (!args->isfl) |
548 | xfs_trans_mod_sb(args->tp, | 553 | xfs_trans_mod_sb(args->tp, |
@@ -1693,7 +1698,7 @@ xfs_free_ag_extent( | |||
1693 | * when the iclog commits to disk. If a busy block is allocated, | 1698 | * when the iclog commits to disk. If a busy block is allocated, |
1694 | * the iclog is pushed up to the LSN that freed the block. | 1699 | * the iclog is pushed up to the LSN that freed the block. |
1695 | */ | 1700 | */ |
1696 | xfs_alloc_mark_busy(tp, agno, bno, len); | 1701 | xfs_alloc_busy_insert(tp, agno, bno, len); |
1697 | return 0; | 1702 | return 0; |
1698 | 1703 | ||
1699 | error0: | 1704 | error0: |
@@ -1989,14 +1994,20 @@ xfs_alloc_get_freelist( | |||
1989 | *bnop = bno; | 1994 | *bnop = bno; |
1990 | 1995 | ||
1991 | /* | 1996 | /* |
1992 | * As blocks are freed, they are added to the per-ag busy list | 1997 | * As blocks are freed, they are added to the per-ag busy list and |
1993 | * and remain there until the freeing transaction is committed to | 1998 | * remain there until the freeing transaction is committed to disk. |
1994 | * disk. Now that we have allocated blocks, this list must be | 1999 | * Now that we have allocated blocks, this list must be searched to see |
1995 | * searched to see if a block is being reused. If one is, then | 2000 | * if a block is being reused. If one is, then the freeing transaction |
1996 | * the freeing transaction must be pushed to disk NOW by forcing | 2001 | * must be pushed to disk before this transaction. |
1997 | * to disk all iclogs up that transaction's LSN. | 2002 | * |
2003 | * We do this by setting the current transaction to a sync transaction | ||
2004 | * which guarantees that the freeing transaction is on disk before this | ||
2005 | * transaction. This is done instead of a synchronous log force here so | ||
2006 | * that we don't sit and wait with the AGF locked in the transaction | ||
2007 | * during the log force. | ||
1998 | */ | 2008 | */ |
1999 | xfs_alloc_search_busy(tp, be32_to_cpu(agf->agf_seqno), bno, 1); | 2009 | if (xfs_alloc_busy_search(mp, be32_to_cpu(agf->agf_seqno), bno, 1)) |
2010 | xfs_trans_set_sync(tp); | ||
2000 | return 0; | 2011 | return 0; |
2001 | } | 2012 | } |
2002 | 2013 | ||
@@ -2201,7 +2212,7 @@ xfs_alloc_read_agf( | |||
2201 | be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]); | 2212 | be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]); |
2202 | spin_lock_init(&pag->pagb_lock); | 2213 | spin_lock_init(&pag->pagb_lock); |
2203 | pag->pagb_count = 0; | 2214 | pag->pagb_count = 0; |
2204 | memset(pag->pagb_list, 0, sizeof(pag->pagb_list)); | 2215 | pag->pagb_tree = RB_ROOT; |
2205 | pag->pagf_init = 1; | 2216 | pag->pagf_init = 1; |
2206 | } | 2217 | } |
2207 | #ifdef DEBUG | 2218 | #ifdef DEBUG |
@@ -2479,127 +2490,263 @@ error0: | |||
2479 | * list is reused, the transaction that freed it must be forced to disk | 2490 | * list is reused, the transaction that freed it must be forced to disk |
2480 | * before continuing to use the block. | 2491 | * before continuing to use the block. |
2481 | * | 2492 | * |
2482 | * xfs_alloc_mark_busy - add to the per-ag busy list | 2493 | * xfs_alloc_busy_insert - add to the per-ag busy list |
2483 | * xfs_alloc_clear_busy - remove an item from the per-ag busy list | 2494 | * xfs_alloc_busy_clear - remove an item from the per-ag busy list |
2495 | * xfs_alloc_busy_search - search for a busy extent | ||
2496 | */ | ||
2497 | |||
2498 | /* | ||
2499 | * Insert a new extent into the busy tree. | ||
2500 | * | ||
2501 | * The busy extent tree is indexed by the start block of the busy extent. | ||
2502 | * there can be multiple overlapping ranges in the busy extent tree but only | ||
2503 | * ever one entry at a given start block. The reason for this is that | ||
2504 | * multi-block extents can be freed, then smaller chunks of that extent | ||
2505 | * allocated and freed again before the first transaction commit is on disk. | ||
2506 | * If the exact same start block is freed a second time, we have to wait for | ||
2507 | * that busy extent to pass out of the tree before the new extent is inserted. | ||
2508 | * There are two main cases we have to handle here. | ||
2509 | * | ||
2510 | * The first case is a transaction that triggers a "free - allocate - free" | ||
2511 | * cycle. This can occur during btree manipulations as a btree block is freed | ||
2512 | * to the freelist, then allocated from the free list, then freed again. In | ||
2513 | * this case, the second extxpnet free is what triggers the duplicate and as | ||
2514 | * such the transaction IDs should match. Because the extent was allocated in | ||
2515 | * this transaction, the transaction must be marked as synchronous. This is | ||
2516 | * true for all cases where the free/alloc/free occurs in the one transaction, | ||
2517 | * hence the addition of the ASSERT(tp->t_flags & XFS_TRANS_SYNC) to this case. | ||
2518 | * This serves to catch violations of the second case quite effectively. | ||
2519 | * | ||
2520 | * The second case is where the free/alloc/free occur in different | ||
2521 | * transactions. In this case, the thread freeing the extent the second time | ||
2522 | * can't mark the extent busy immediately because it is already tracked in a | ||
2523 | * transaction that may be committing. When the log commit for the existing | ||
2524 | * busy extent completes, the busy extent will be removed from the tree. If we | ||
2525 | * allow the second busy insert to continue using that busy extent structure, | ||
2526 | * it can be freed before this transaction is safely in the log. Hence our | ||
2527 | * only option in this case is to force the log to remove the existing busy | ||
2528 | * extent from the list before we insert the new one with the current | ||
2529 | * transaction ID. | ||
2530 | * | ||
2531 | * The problem we are trying to avoid in the free-alloc-free in separate | ||
2532 | * transactions is most easily described with a timeline: | ||
2533 | * | ||
2534 | * Thread 1 Thread 2 Thread 3 xfslogd | ||
2535 | * xact alloc | ||
2536 | * free X | ||
2537 | * mark busy | ||
2538 | * commit xact | ||
2539 | * free xact | ||
2540 | * xact alloc | ||
2541 | * alloc X | ||
2542 | * busy search | ||
2543 | * mark xact sync | ||
2544 | * commit xact | ||
2545 | * free xact | ||
2546 | * force log | ||
2547 | * checkpoint starts | ||
2548 | * .... | ||
2549 | * xact alloc | ||
2550 | * free X | ||
2551 | * mark busy | ||
2552 | * finds match | ||
2553 | * *** KABOOM! *** | ||
2554 | * .... | ||
2555 | * log IO completes | ||
2556 | * unbusy X | ||
2557 | * checkpoint completes | ||
2558 | * | ||
2559 | * By issuing a log force in thread 3 @ "KABOOM", the thread will block until | ||
2560 | * the checkpoint completes, and the busy extent it matched will have been | ||
2561 | * removed from the tree when it is woken. Hence it can then continue safely. | ||
2562 | * | ||
2563 | * However, to ensure this matching process is robust, we need to use the | ||
2564 | * transaction ID for identifying transaction, as delayed logging results in | ||
2565 | * the busy extent and transaction lifecycles being different. i.e. the busy | ||
2566 | * extent is active for a lot longer than the transaction. Hence the | ||
2567 | * transaction structure can be freed and reallocated, then mark the same | ||
2568 | * extent busy again in the new transaction. In this case the new transaction | ||
2569 | * will have a different tid but can have the same address, and hence we need | ||
2570 | * to check against the tid. | ||
2571 | * | ||
2572 | * Future: for delayed logging, we could avoid the log force if the extent was | ||
2573 | * first freed in the current checkpoint sequence. This, however, requires the | ||
2574 | * ability to pin the current checkpoint in memory until this transaction | ||
2575 | * commits to ensure that both the original free and the current one combine | ||
2576 | * logically into the one checkpoint. If the checkpoint sequences are | ||
2577 | * different, however, we still need to wait on a log force. | ||
2484 | */ | 2578 | */ |
2485 | void | 2579 | void |
2486 | xfs_alloc_mark_busy(xfs_trans_t *tp, | 2580 | xfs_alloc_busy_insert( |
2487 | xfs_agnumber_t agno, | 2581 | struct xfs_trans *tp, |
2488 | xfs_agblock_t bno, | 2582 | xfs_agnumber_t agno, |
2489 | xfs_extlen_t len) | 2583 | xfs_agblock_t bno, |
2584 | xfs_extlen_t len) | ||
2490 | { | 2585 | { |
2491 | xfs_perag_busy_t *bsy; | 2586 | struct xfs_busy_extent *new; |
2587 | struct xfs_busy_extent *busyp; | ||
2492 | struct xfs_perag *pag; | 2588 | struct xfs_perag *pag; |
2493 | int n; | 2589 | struct rb_node **rbp; |
2590 | struct rb_node *parent; | ||
2591 | int match; | ||
2494 | 2592 | ||
2495 | pag = xfs_perag_get(tp->t_mountp, agno); | ||
2496 | spin_lock(&pag->pagb_lock); | ||
2497 | 2593 | ||
2498 | /* search pagb_list for an open slot */ | 2594 | new = kmem_zalloc(sizeof(struct xfs_busy_extent), KM_MAYFAIL); |
2499 | for (bsy = pag->pagb_list, n = 0; | 2595 | if (!new) { |
2500 | n < XFS_PAGB_NUM_SLOTS; | 2596 | /* |
2501 | bsy++, n++) { | 2597 | * No Memory! Since it is now not possible to track the free |
2502 | if (bsy->busy_tp == NULL) { | 2598 | * block, make this a synchronous transaction to insure that |
2503 | break; | 2599 | * the block is not reused before this transaction commits. |
2504 | } | 2600 | */ |
2601 | trace_xfs_alloc_busy(tp, agno, bno, len, 1); | ||
2602 | xfs_trans_set_sync(tp); | ||
2603 | return; | ||
2505 | } | 2604 | } |
2506 | 2605 | ||
2507 | trace_xfs_alloc_busy(tp->t_mountp, agno, bno, len, n); | 2606 | new->agno = agno; |
2607 | new->bno = bno; | ||
2608 | new->length = len; | ||
2609 | new->tid = xfs_log_get_trans_ident(tp); | ||
2508 | 2610 | ||
2509 | if (n < XFS_PAGB_NUM_SLOTS) { | 2611 | INIT_LIST_HEAD(&new->list); |
2510 | bsy = &pag->pagb_list[n]; | 2612 | |
2511 | pag->pagb_count++; | 2613 | /* trace before insert to be able to see failed inserts */ |
2512 | bsy->busy_start = bno; | 2614 | trace_xfs_alloc_busy(tp, agno, bno, len, 0); |
2513 | bsy->busy_length = len; | 2615 | |
2514 | bsy->busy_tp = tp; | 2616 | pag = xfs_perag_get(tp->t_mountp, new->agno); |
2515 | xfs_trans_add_busy(tp, agno, n); | 2617 | restart: |
2516 | } else { | 2618 | spin_lock(&pag->pagb_lock); |
2619 | rbp = &pag->pagb_tree.rb_node; | ||
2620 | parent = NULL; | ||
2621 | busyp = NULL; | ||
2622 | match = 0; | ||
2623 | while (*rbp && match >= 0) { | ||
2624 | parent = *rbp; | ||
2625 | busyp = rb_entry(parent, struct xfs_busy_extent, rb_node); | ||
2626 | |||
2627 | if (new->bno < busyp->bno) { | ||
2628 | /* may overlap, but exact start block is lower */ | ||
2629 | rbp = &(*rbp)->rb_left; | ||
2630 | if (new->bno + new->length > busyp->bno) | ||
2631 | match = busyp->tid == new->tid ? 1 : -1; | ||
2632 | } else if (new->bno > busyp->bno) { | ||
2633 | /* may overlap, but exact start block is higher */ | ||
2634 | rbp = &(*rbp)->rb_right; | ||
2635 | if (bno < busyp->bno + busyp->length) | ||
2636 | match = busyp->tid == new->tid ? 1 : -1; | ||
2637 | } else { | ||
2638 | match = busyp->tid == new->tid ? 1 : -1; | ||
2639 | break; | ||
2640 | } | ||
2641 | } | ||
2642 | if (match < 0) { | ||
2643 | /* overlap marked busy in different transaction */ | ||
2644 | spin_unlock(&pag->pagb_lock); | ||
2645 | xfs_log_force(tp->t_mountp, XFS_LOG_SYNC); | ||
2646 | goto restart; | ||
2647 | } | ||
2648 | if (match > 0) { | ||
2517 | /* | 2649 | /* |
2518 | * The busy list is full! Since it is now not possible to | 2650 | * overlap marked busy in same transaction. Update if exact |
2519 | * track the free block, make this a synchronous transaction | 2651 | * start block match, otherwise combine the busy extents into |
2520 | * to insure that the block is not reused before this | 2652 | * a single range. |
2521 | * transaction commits. | ||
2522 | */ | 2653 | */ |
2523 | xfs_trans_set_sync(tp); | 2654 | if (busyp->bno == new->bno) { |
2524 | } | 2655 | busyp->length = max(busyp->length, new->length); |
2656 | spin_unlock(&pag->pagb_lock); | ||
2657 | ASSERT(tp->t_flags & XFS_TRANS_SYNC); | ||
2658 | xfs_perag_put(pag); | ||
2659 | kmem_free(new); | ||
2660 | return; | ||
2661 | } | ||
2662 | rb_erase(&busyp->rb_node, &pag->pagb_tree); | ||
2663 | new->length = max(busyp->bno + busyp->length, | ||
2664 | new->bno + new->length) - | ||
2665 | min(busyp->bno, new->bno); | ||
2666 | new->bno = min(busyp->bno, new->bno); | ||
2667 | } else | ||
2668 | busyp = NULL; | ||
2525 | 2669 | ||
2670 | rb_link_node(&new->rb_node, parent, rbp); | ||
2671 | rb_insert_color(&new->rb_node, &pag->pagb_tree); | ||
2672 | |||
2673 | list_add(&new->list, &tp->t_busy); | ||
2526 | spin_unlock(&pag->pagb_lock); | 2674 | spin_unlock(&pag->pagb_lock); |
2527 | xfs_perag_put(pag); | 2675 | xfs_perag_put(pag); |
2676 | kmem_free(busyp); | ||
2528 | } | 2677 | } |
2529 | 2678 | ||
2530 | void | 2679 | /* |
2531 | xfs_alloc_clear_busy(xfs_trans_t *tp, | 2680 | * Search for a busy extent within the range of the extent we are about to |
2532 | xfs_agnumber_t agno, | 2681 | * allocate. You need to be holding the busy extent tree lock when calling |
2533 | int idx) | 2682 | * xfs_alloc_busy_search(). This function returns 0 for no overlapping busy |
2683 | * extent, -1 for an overlapping but not exact busy extent, and 1 for an exact | ||
2684 | * match. This is done so that a non-zero return indicates an overlap that | ||
2685 | * will require a synchronous transaction, but it can still be | ||
2686 | * used to distinguish between a partial or exact match. | ||
2687 | */ | ||
2688 | static int | ||
2689 | xfs_alloc_busy_search( | ||
2690 | struct xfs_mount *mp, | ||
2691 | xfs_agnumber_t agno, | ||
2692 | xfs_agblock_t bno, | ||
2693 | xfs_extlen_t len) | ||
2534 | { | 2694 | { |
2535 | struct xfs_perag *pag; | 2695 | struct xfs_perag *pag; |
2536 | xfs_perag_busy_t *list; | 2696 | struct rb_node *rbp; |
2697 | struct xfs_busy_extent *busyp; | ||
2698 | int match = 0; | ||
2537 | 2699 | ||
2538 | ASSERT(idx < XFS_PAGB_NUM_SLOTS); | 2700 | pag = xfs_perag_get(mp, agno); |
2539 | pag = xfs_perag_get(tp->t_mountp, agno); | ||
2540 | spin_lock(&pag->pagb_lock); | 2701 | spin_lock(&pag->pagb_lock); |
2541 | list = pag->pagb_list; | ||
2542 | 2702 | ||
2543 | trace_xfs_alloc_unbusy(tp->t_mountp, agno, idx, list[idx].busy_tp == tp); | 2703 | rbp = pag->pagb_tree.rb_node; |
2544 | 2704 | ||
2545 | if (list[idx].busy_tp == tp) { | 2705 | /* find closest start bno overlap */ |
2546 | list[idx].busy_tp = NULL; | 2706 | while (rbp) { |
2547 | pag->pagb_count--; | 2707 | busyp = rb_entry(rbp, struct xfs_busy_extent, rb_node); |
2708 | if (bno < busyp->bno) { | ||
2709 | /* may overlap, but exact start block is lower */ | ||
2710 | if (bno + len > busyp->bno) | ||
2711 | match = -1; | ||
2712 | rbp = rbp->rb_left; | ||
2713 | } else if (bno > busyp->bno) { | ||
2714 | /* may overlap, but exact start block is higher */ | ||
2715 | if (bno < busyp->bno + busyp->length) | ||
2716 | match = -1; | ||
2717 | rbp = rbp->rb_right; | ||
2718 | } else { | ||
2719 | /* bno matches busyp, length determines exact match */ | ||
2720 | match = (busyp->length == len) ? 1 : -1; | ||
2721 | break; | ||
2722 | } | ||
2548 | } | 2723 | } |
2549 | |||
2550 | spin_unlock(&pag->pagb_lock); | 2724 | spin_unlock(&pag->pagb_lock); |
2725 | trace_xfs_alloc_busysearch(mp, agno, bno, len, !!match); | ||
2551 | xfs_perag_put(pag); | 2726 | xfs_perag_put(pag); |
2727 | return match; | ||
2552 | } | 2728 | } |
2553 | 2729 | ||
2554 | 2730 | void | |
2555 | /* | 2731 | xfs_alloc_busy_clear( |
2556 | * If we find the extent in the busy list, force the log out to get the | 2732 | struct xfs_mount *mp, |
2557 | * extent out of the busy list so the caller can use it straight away. | 2733 | struct xfs_busy_extent *busyp) |
2558 | */ | ||
2559 | STATIC void | ||
2560 | xfs_alloc_search_busy(xfs_trans_t *tp, | ||
2561 | xfs_agnumber_t agno, | ||
2562 | xfs_agblock_t bno, | ||
2563 | xfs_extlen_t len) | ||
2564 | { | 2734 | { |
2565 | struct xfs_perag *pag; | 2735 | struct xfs_perag *pag; |
2566 | xfs_perag_busy_t *bsy; | ||
2567 | xfs_agblock_t uend, bend; | ||
2568 | xfs_lsn_t lsn = 0; | ||
2569 | int cnt; | ||
2570 | 2736 | ||
2571 | pag = xfs_perag_get(tp->t_mountp, agno); | 2737 | trace_xfs_alloc_unbusy(mp, busyp->agno, busyp->bno, |
2572 | spin_lock(&pag->pagb_lock); | 2738 | busyp->length); |
2573 | cnt = pag->pagb_count; | ||
2574 | 2739 | ||
2575 | /* | 2740 | ASSERT(xfs_alloc_busy_search(mp, busyp->agno, busyp->bno, |
2576 | * search pagb_list for this slot, skipping open slots. We have to | 2741 | busyp->length) == 1); |
2577 | * search the entire array as there may be multiple overlaps and | ||
2578 | * we have to get the most recent LSN for the log force to push out | ||
2579 | * all the transactions that span the range. | ||
2580 | */ | ||
2581 | uend = bno + len - 1; | ||
2582 | for (cnt = 0; cnt < pag->pagb_count; cnt++) { | ||
2583 | bsy = &pag->pagb_list[cnt]; | ||
2584 | if (!bsy->busy_tp) | ||
2585 | continue; | ||
2586 | 2742 | ||
2587 | bend = bsy->busy_start + bsy->busy_length - 1; | 2743 | list_del_init(&busyp->list); |
2588 | if (bno > bend || uend < bsy->busy_start) | ||
2589 | continue; | ||
2590 | 2744 | ||
2591 | /* (start1,length1) within (start2, length2) */ | 2745 | pag = xfs_perag_get(mp, busyp->agno); |
2592 | if (XFS_LSN_CMP(bsy->busy_tp->t_commit_lsn, lsn) > 0) | 2746 | spin_lock(&pag->pagb_lock); |
2593 | lsn = bsy->busy_tp->t_commit_lsn; | 2747 | rb_erase(&busyp->rb_node, &pag->pagb_tree); |
2594 | } | ||
2595 | spin_unlock(&pag->pagb_lock); | 2748 | spin_unlock(&pag->pagb_lock); |
2596 | xfs_perag_put(pag); | 2749 | xfs_perag_put(pag); |
2597 | trace_xfs_alloc_busysearch(tp->t_mountp, agno, bno, len, lsn); | ||
2598 | 2750 | ||
2599 | /* | 2751 | kmem_free(busyp); |
2600 | * If a block was found, force the log through the LSN of the | ||
2601 | * transaction that freed the block | ||
2602 | */ | ||
2603 | if (lsn) | ||
2604 | xfs_log_force_lsn(tp->t_mountp, lsn, XFS_LOG_SYNC); | ||
2605 | } | 2752 | } |
diff --git a/fs/xfs/xfs_alloc.h b/fs/xfs/xfs_alloc.h index 599bffa39784..6d05199b667c 100644 --- a/fs/xfs/xfs_alloc.h +++ b/fs/xfs/xfs_alloc.h | |||
@@ -22,6 +22,7 @@ struct xfs_buf; | |||
22 | struct xfs_mount; | 22 | struct xfs_mount; |
23 | struct xfs_perag; | 23 | struct xfs_perag; |
24 | struct xfs_trans; | 24 | struct xfs_trans; |
25 | struct xfs_busy_extent; | ||
25 | 26 | ||
26 | /* | 27 | /* |
27 | * Freespace allocation types. Argument to xfs_alloc_[v]extent. | 28 | * Freespace allocation types. Argument to xfs_alloc_[v]extent. |
@@ -119,15 +120,13 @@ xfs_alloc_longest_free_extent(struct xfs_mount *mp, | |||
119 | #ifdef __KERNEL__ | 120 | #ifdef __KERNEL__ |
120 | 121 | ||
121 | void | 122 | void |
122 | xfs_alloc_mark_busy(xfs_trans_t *tp, | 123 | xfs_alloc_busy_insert(xfs_trans_t *tp, |
123 | xfs_agnumber_t agno, | 124 | xfs_agnumber_t agno, |
124 | xfs_agblock_t bno, | 125 | xfs_agblock_t bno, |
125 | xfs_extlen_t len); | 126 | xfs_extlen_t len); |
126 | 127 | ||
127 | void | 128 | void |
128 | xfs_alloc_clear_busy(xfs_trans_t *tp, | 129 | xfs_alloc_busy_clear(struct xfs_mount *mp, struct xfs_busy_extent *busyp); |
129 | xfs_agnumber_t ag, | ||
130 | int idx); | ||
131 | 130 | ||
132 | #endif /* __KERNEL__ */ | 131 | #endif /* __KERNEL__ */ |
133 | 132 | ||
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c index b726e10d2c1c..83f494218759 100644 --- a/fs/xfs/xfs_alloc_btree.c +++ b/fs/xfs/xfs_alloc_btree.c | |||
@@ -134,7 +134,7 @@ xfs_allocbt_free_block( | |||
134 | * disk. If a busy block is allocated, the iclog is pushed up to the | 134 | * disk. If a busy block is allocated, the iclog is pushed up to the |
135 | * LSN that freed the block. | 135 | * LSN that freed the block. |
136 | */ | 136 | */ |
137 | xfs_alloc_mark_busy(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1); | 137 | xfs_alloc_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1); |
138 | xfs_trans_agbtree_delta(cur->bc_tp, -1); | 138 | xfs_trans_agbtree_delta(cur->bc_tp, -1); |
139 | return 0; | 139 | return 0; |
140 | } | 140 | } |
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index be578ecb4af2..40d9595a8de2 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c | |||
@@ -44,6 +44,7 @@ | |||
44 | #include "xfs_trans_priv.h" | 44 | #include "xfs_trans_priv.h" |
45 | #include "xfs_trans_space.h" | 45 | #include "xfs_trans_space.h" |
46 | #include "xfs_inode_item.h" | 46 | #include "xfs_inode_item.h" |
47 | #include "xfs_trace.h" | ||
47 | 48 | ||
48 | kmem_zone_t *xfs_trans_zone; | 49 | kmem_zone_t *xfs_trans_zone; |
49 | 50 | ||
@@ -243,9 +244,8 @@ _xfs_trans_alloc( | |||
243 | tp->t_type = type; | 244 | tp->t_type = type; |
244 | tp->t_mountp = mp; | 245 | tp->t_mountp = mp; |
245 | tp->t_items_free = XFS_LIC_NUM_SLOTS; | 246 | tp->t_items_free = XFS_LIC_NUM_SLOTS; |
246 | tp->t_busy_free = XFS_LBC_NUM_SLOTS; | ||
247 | xfs_lic_init(&(tp->t_items)); | 247 | xfs_lic_init(&(tp->t_items)); |
248 | XFS_LBC_INIT(&(tp->t_busy)); | 248 | INIT_LIST_HEAD(&tp->t_busy); |
249 | return tp; | 249 | return tp; |
250 | } | 250 | } |
251 | 251 | ||
@@ -255,8 +255,13 @@ _xfs_trans_alloc( | |||
255 | */ | 255 | */ |
256 | STATIC void | 256 | STATIC void |
257 | xfs_trans_free( | 257 | xfs_trans_free( |
258 | xfs_trans_t *tp) | 258 | struct xfs_trans *tp) |
259 | { | 259 | { |
260 | struct xfs_busy_extent *busyp, *n; | ||
261 | |||
262 | list_for_each_entry_safe(busyp, n, &tp->t_busy, list) | ||
263 | xfs_alloc_busy_clear(tp->t_mountp, busyp); | ||
264 | |||
260 | atomic_dec(&tp->t_mountp->m_active_trans); | 265 | atomic_dec(&tp->t_mountp->m_active_trans); |
261 | xfs_trans_free_dqinfo(tp); | 266 | xfs_trans_free_dqinfo(tp); |
262 | kmem_zone_free(xfs_trans_zone, tp); | 267 | kmem_zone_free(xfs_trans_zone, tp); |
@@ -285,9 +290,8 @@ xfs_trans_dup( | |||
285 | ntp->t_type = tp->t_type; | 290 | ntp->t_type = tp->t_type; |
286 | ntp->t_mountp = tp->t_mountp; | 291 | ntp->t_mountp = tp->t_mountp; |
287 | ntp->t_items_free = XFS_LIC_NUM_SLOTS; | 292 | ntp->t_items_free = XFS_LIC_NUM_SLOTS; |
288 | ntp->t_busy_free = XFS_LBC_NUM_SLOTS; | ||
289 | xfs_lic_init(&(ntp->t_items)); | 293 | xfs_lic_init(&(ntp->t_items)); |
290 | XFS_LBC_INIT(&(ntp->t_busy)); | 294 | INIT_LIST_HEAD(&ntp->t_busy); |
291 | 295 | ||
292 | ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES); | 296 | ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES); |
293 | ASSERT(tp->t_ticket != NULL); | 297 | ASSERT(tp->t_ticket != NULL); |
@@ -423,7 +427,6 @@ undo_blocks: | |||
423 | return error; | 427 | return error; |
424 | } | 428 | } |
425 | 429 | ||
426 | |||
427 | /* | 430 | /* |
428 | * Record the indicated change to the given field for application | 431 | * Record the indicated change to the given field for application |
429 | * to the file system's superblock when the transaction commits. | 432 | * to the file system's superblock when the transaction commits. |
@@ -930,26 +933,6 @@ xfs_trans_item_committed( | |||
930 | IOP_UNPIN(lip); | 933 | IOP_UNPIN(lip); |
931 | } | 934 | } |
932 | 935 | ||
933 | /* Clear all the per-AG busy list items listed in this transaction */ | ||
934 | static void | ||
935 | xfs_trans_clear_busy_extents( | ||
936 | struct xfs_trans *tp) | ||
937 | { | ||
938 | xfs_log_busy_chunk_t *lbcp; | ||
939 | xfs_log_busy_slot_t *lbsp; | ||
940 | int i; | ||
941 | |||
942 | for (lbcp = &tp->t_busy; lbcp != NULL; lbcp = lbcp->lbc_next) { | ||
943 | i = 0; | ||
944 | for (lbsp = lbcp->lbc_busy; i < lbcp->lbc_unused; i++, lbsp++) { | ||
945 | if (XFS_LBC_ISFREE(lbcp, i)) | ||
946 | continue; | ||
947 | xfs_alloc_clear_busy(tp, lbsp->lbc_ag, lbsp->lbc_idx); | ||
948 | } | ||
949 | } | ||
950 | xfs_trans_free_busy(tp); | ||
951 | } | ||
952 | |||
953 | /* | 936 | /* |
954 | * This is typically called by the LM when a transaction has been fully | 937 | * This is typically called by the LM when a transaction has been fully |
955 | * committed to disk. It needs to unpin the items which have | 938 | * committed to disk. It needs to unpin the items which have |
@@ -984,7 +967,6 @@ xfs_trans_committed( | |||
984 | kmem_free(licp); | 967 | kmem_free(licp); |
985 | } | 968 | } |
986 | 969 | ||
987 | xfs_trans_clear_busy_extents(tp); | ||
988 | xfs_trans_free(tp); | 970 | xfs_trans_free(tp); |
989 | } | 971 | } |
990 | 972 | ||
@@ -1013,7 +995,6 @@ xfs_trans_uncommit( | |||
1013 | xfs_trans_unreserve_and_mod_dquots(tp); | 995 | xfs_trans_unreserve_and_mod_dquots(tp); |
1014 | 996 | ||
1015 | xfs_trans_free_items(tp, flags); | 997 | xfs_trans_free_items(tp, flags); |
1016 | xfs_trans_free_busy(tp); | ||
1017 | xfs_trans_free(tp); | 998 | xfs_trans_free(tp); |
1018 | } | 999 | } |
1019 | 1000 | ||
@@ -1075,6 +1056,8 @@ xfs_trans_commit_iclog( | |||
1075 | *commit_lsn = xfs_log_done(mp, tp->t_ticket, &commit_iclog, log_flags); | 1056 | *commit_lsn = xfs_log_done(mp, tp->t_ticket, &commit_iclog, log_flags); |
1076 | 1057 | ||
1077 | tp->t_commit_lsn = *commit_lsn; | 1058 | tp->t_commit_lsn = *commit_lsn; |
1059 | trace_xfs_trans_commit_lsn(tp); | ||
1060 | |||
1078 | if (nvec > XFS_TRANS_LOGVEC_COUNT) | 1061 | if (nvec > XFS_TRANS_LOGVEC_COUNT) |
1079 | kmem_free(log_vector); | 1062 | kmem_free(log_vector); |
1080 | 1063 | ||
@@ -1260,7 +1243,6 @@ out_unreserve: | |||
1260 | } | 1243 | } |
1261 | current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS); | 1244 | current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS); |
1262 | xfs_trans_free_items(tp, error ? XFS_TRANS_ABORT : 0); | 1245 | xfs_trans_free_items(tp, error ? XFS_TRANS_ABORT : 0); |
1263 | xfs_trans_free_busy(tp); | ||
1264 | xfs_trans_free(tp); | 1246 | xfs_trans_free(tp); |
1265 | 1247 | ||
1266 | XFS_STATS_INC(xs_trans_empty); | 1248 | XFS_STATS_INC(xs_trans_empty); |
@@ -1339,7 +1321,6 @@ xfs_trans_cancel( | |||
1339 | current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS); | 1321 | current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS); |
1340 | 1322 | ||
1341 | xfs_trans_free_items(tp, flags); | 1323 | xfs_trans_free_items(tp, flags); |
1342 | xfs_trans_free_busy(tp); | ||
1343 | xfs_trans_free(tp); | 1324 | xfs_trans_free(tp); |
1344 | } | 1325 | } |
1345 | 1326 | ||
diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h index c62beee0921e..ff7e9e6eee84 100644 --- a/fs/xfs/xfs_trans.h +++ b/fs/xfs/xfs_trans.h | |||
@@ -813,6 +813,7 @@ struct xfs_log_item_desc; | |||
813 | struct xfs_mount; | 813 | struct xfs_mount; |
814 | struct xfs_trans; | 814 | struct xfs_trans; |
815 | struct xfs_dquot_acct; | 815 | struct xfs_dquot_acct; |
816 | struct xfs_busy_extent; | ||
816 | 817 | ||
817 | typedef struct xfs_log_item { | 818 | typedef struct xfs_log_item { |
818 | struct list_head li_ail; /* AIL pointers */ | 819 | struct list_head li_ail; /* AIL pointers */ |
@@ -872,34 +873,6 @@ typedef struct xfs_item_ops { | |||
872 | #define XFS_ITEM_PUSHBUF 3 | 873 | #define XFS_ITEM_PUSHBUF 3 |
873 | 874 | ||
874 | /* | 875 | /* |
875 | * This structure is used to maintain a list of block ranges that have been | ||
876 | * freed in the transaction. The ranges are listed in the perag[] busy list | ||
877 | * between when they're freed and the transaction is committed to disk. | ||
878 | */ | ||
879 | |||
880 | typedef struct xfs_log_busy_slot { | ||
881 | xfs_agnumber_t lbc_ag; | ||
882 | ushort lbc_idx; /* index in perag.busy[] */ | ||
883 | } xfs_log_busy_slot_t; | ||
884 | |||
885 | #define XFS_LBC_NUM_SLOTS 31 | ||
886 | typedef struct xfs_log_busy_chunk { | ||
887 | struct xfs_log_busy_chunk *lbc_next; | ||
888 | uint lbc_free; /* free slots bitmask */ | ||
889 | ushort lbc_unused; /* first unused */ | ||
890 | xfs_log_busy_slot_t lbc_busy[XFS_LBC_NUM_SLOTS]; | ||
891 | } xfs_log_busy_chunk_t; | ||
892 | |||
893 | #define XFS_LBC_MAX_SLOT (XFS_LBC_NUM_SLOTS - 1) | ||
894 | #define XFS_LBC_FREEMASK ((1U << XFS_LBC_NUM_SLOTS) - 1) | ||
895 | |||
896 | #define XFS_LBC_INIT(cp) ((cp)->lbc_free = XFS_LBC_FREEMASK) | ||
897 | #define XFS_LBC_CLAIM(cp, slot) ((cp)->lbc_free &= ~(1 << (slot))) | ||
898 | #define XFS_LBC_SLOT(cp, slot) (&((cp)->lbc_busy[(slot)])) | ||
899 | #define XFS_LBC_VACANCY(cp) (((cp)->lbc_free) & XFS_LBC_FREEMASK) | ||
900 | #define XFS_LBC_ISFREE(cp, slot) ((cp)->lbc_free & (1 << (slot))) | ||
901 | |||
902 | /* | ||
903 | * This is the type of function which can be given to xfs_trans_callback() | 876 | * This is the type of function which can be given to xfs_trans_callback() |
904 | * to be called upon the transaction's commit to disk. | 877 | * to be called upon the transaction's commit to disk. |
905 | */ | 878 | */ |
@@ -950,8 +923,7 @@ typedef struct xfs_trans { | |||
950 | unsigned int t_items_free; /* log item descs free */ | 923 | unsigned int t_items_free; /* log item descs free */ |
951 | xfs_log_item_chunk_t t_items; /* first log item desc chunk */ | 924 | xfs_log_item_chunk_t t_items; /* first log item desc chunk */ |
952 | xfs_trans_header_t t_header; /* header for in-log trans */ | 925 | xfs_trans_header_t t_header; /* header for in-log trans */ |
953 | unsigned int t_busy_free; /* busy descs free */ | 926 | struct list_head t_busy; /* list of busy extents */ |
954 | xfs_log_busy_chunk_t t_busy; /* busy/async free blocks */ | ||
955 | unsigned long t_pflags; /* saved process flags state */ | 927 | unsigned long t_pflags; /* saved process flags state */ |
956 | } xfs_trans_t; | 928 | } xfs_trans_t; |
957 | 929 | ||
@@ -1025,9 +997,6 @@ int _xfs_trans_commit(xfs_trans_t *, | |||
1025 | void xfs_trans_cancel(xfs_trans_t *, int); | 997 | void xfs_trans_cancel(xfs_trans_t *, int); |
1026 | int xfs_trans_ail_init(struct xfs_mount *); | 998 | int xfs_trans_ail_init(struct xfs_mount *); |
1027 | void xfs_trans_ail_destroy(struct xfs_mount *); | 999 | void xfs_trans_ail_destroy(struct xfs_mount *); |
1028 | xfs_log_busy_slot_t *xfs_trans_add_busy(xfs_trans_t *tp, | ||
1029 | xfs_agnumber_t ag, | ||
1030 | xfs_extlen_t idx); | ||
1031 | 1000 | ||
1032 | extern kmem_zone_t *xfs_trans_zone; | 1001 | extern kmem_zone_t *xfs_trans_zone; |
1033 | 1002 | ||
diff --git a/fs/xfs/xfs_trans_item.c b/fs/xfs/xfs_trans_item.c index eb3fc57f9eef..2937a1e53318 100644 --- a/fs/xfs/xfs_trans_item.c +++ b/fs/xfs/xfs_trans_item.c | |||
@@ -438,112 +438,3 @@ xfs_trans_unlock_chunk( | |||
438 | 438 | ||
439 | return freed; | 439 | return freed; |
440 | } | 440 | } |
441 | |||
442 | |||
443 | /* | ||
444 | * This is called to add the given busy item to the transaction's | ||
445 | * list of busy items. It must find a free busy item descriptor | ||
446 | * or allocate a new one and add the item to that descriptor. | ||
447 | * The function returns a pointer to busy descriptor used to point | ||
448 | * to the new busy entry. The log busy entry will now point to its new | ||
449 | * descriptor with its ???? field. | ||
450 | */ | ||
451 | xfs_log_busy_slot_t * | ||
452 | xfs_trans_add_busy(xfs_trans_t *tp, xfs_agnumber_t ag, xfs_extlen_t idx) | ||
453 | { | ||
454 | xfs_log_busy_chunk_t *lbcp; | ||
455 | xfs_log_busy_slot_t *lbsp; | ||
456 | int i=0; | ||
457 | |||
458 | /* | ||
459 | * If there are no free descriptors, allocate a new chunk | ||
460 | * of them and put it at the front of the chunk list. | ||
461 | */ | ||
462 | if (tp->t_busy_free == 0) { | ||
463 | lbcp = (xfs_log_busy_chunk_t*) | ||
464 | kmem_alloc(sizeof(xfs_log_busy_chunk_t), KM_SLEEP); | ||
465 | ASSERT(lbcp != NULL); | ||
466 | /* | ||
467 | * Initialize the chunk, and then | ||
468 | * claim the first slot in the newly allocated chunk. | ||
469 | */ | ||
470 | XFS_LBC_INIT(lbcp); | ||
471 | XFS_LBC_CLAIM(lbcp, 0); | ||
472 | lbcp->lbc_unused = 1; | ||
473 | lbsp = XFS_LBC_SLOT(lbcp, 0); | ||
474 | |||
475 | /* | ||
476 | * Link in the new chunk and update the free count. | ||
477 | */ | ||
478 | lbcp->lbc_next = tp->t_busy.lbc_next; | ||
479 | tp->t_busy.lbc_next = lbcp; | ||
480 | tp->t_busy_free = XFS_LIC_NUM_SLOTS - 1; | ||
481 | |||
482 | /* | ||
483 | * Initialize the descriptor and the generic portion | ||
484 | * of the log item. | ||
485 | * | ||
486 | * Point the new slot at this item and return it. | ||
487 | * Also point the log item at its currently active | ||
488 | * descriptor and set the item's mount pointer. | ||
489 | */ | ||
490 | lbsp->lbc_ag = ag; | ||
491 | lbsp->lbc_idx = idx; | ||
492 | return lbsp; | ||
493 | } | ||
494 | |||
495 | /* | ||
496 | * Find the free descriptor. It is somewhere in the chunklist | ||
497 | * of descriptors. | ||
498 | */ | ||
499 | lbcp = &tp->t_busy; | ||
500 | while (lbcp != NULL) { | ||
501 | if (XFS_LBC_VACANCY(lbcp)) { | ||
502 | if (lbcp->lbc_unused <= XFS_LBC_MAX_SLOT) { | ||
503 | i = lbcp->lbc_unused; | ||
504 | break; | ||
505 | } else { | ||
506 | /* out-of-order vacancy */ | ||
507 | cmn_err(CE_DEBUG, "OOO vacancy lbcp 0x%p\n", lbcp); | ||
508 | ASSERT(0); | ||
509 | } | ||
510 | } | ||
511 | lbcp = lbcp->lbc_next; | ||
512 | } | ||
513 | ASSERT(lbcp != NULL); | ||
514 | /* | ||
515 | * If we find a free descriptor, claim it, | ||
516 | * initialize it, and return it. | ||
517 | */ | ||
518 | XFS_LBC_CLAIM(lbcp, i); | ||
519 | if (lbcp->lbc_unused <= i) { | ||
520 | lbcp->lbc_unused = i + 1; | ||
521 | } | ||
522 | lbsp = XFS_LBC_SLOT(lbcp, i); | ||
523 | tp->t_busy_free--; | ||
524 | lbsp->lbc_ag = ag; | ||
525 | lbsp->lbc_idx = idx; | ||
526 | return lbsp; | ||
527 | } | ||
528 | |||
529 | |||
530 | /* | ||
531 | * xfs_trans_free_busy | ||
532 | * Free all of the busy lists from a transaction | ||
533 | */ | ||
534 | void | ||
535 | xfs_trans_free_busy(xfs_trans_t *tp) | ||
536 | { | ||
537 | xfs_log_busy_chunk_t *lbcp; | ||
538 | xfs_log_busy_chunk_t *lbcq; | ||
539 | |||
540 | lbcp = tp->t_busy.lbc_next; | ||
541 | while (lbcp != NULL) { | ||
542 | lbcq = lbcp->lbc_next; | ||
543 | kmem_free(lbcp); | ||
544 | lbcp = lbcq; | ||
545 | } | ||
546 | |||
547 | XFS_LBC_INIT(&tp->t_busy); | ||
548 | tp->t_busy.lbc_unused = 0; | ||
549 | } | ||
diff --git a/fs/xfs/xfs_trans_priv.h b/fs/xfs/xfs_trans_priv.h index 73e2ad397432..901dc0f032da 100644 --- a/fs/xfs/xfs_trans_priv.h +++ b/fs/xfs/xfs_trans_priv.h | |||
@@ -38,10 +38,6 @@ struct xfs_log_item_desc *xfs_trans_next_item(struct xfs_trans *, | |||
38 | void xfs_trans_free_items(struct xfs_trans *, int); | 38 | void xfs_trans_free_items(struct xfs_trans *, int); |
39 | void xfs_trans_unlock_items(struct xfs_trans *, | 39 | void xfs_trans_unlock_items(struct xfs_trans *, |
40 | xfs_lsn_t); | 40 | xfs_lsn_t); |
41 | void xfs_trans_free_busy(xfs_trans_t *tp); | ||
42 | xfs_log_busy_slot_t *xfs_trans_add_busy(xfs_trans_t *tp, | ||
43 | xfs_agnumber_t ag, | ||
44 | xfs_extlen_t idx); | ||
45 | 41 | ||
46 | /* | 42 | /* |
47 | * AIL traversal cursor. | 43 | * AIL traversal cursor. |