aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ocfs2
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ocfs2')
-rw-r--r--fs/ocfs2/alloc.c101
-rw-r--r--fs/ocfs2/alloc.h2
-rw-r--r--fs/ocfs2/aops.c50
-rw-r--r--fs/ocfs2/aops.h2
-rw-r--r--fs/ocfs2/cluster/heartbeat.c31
-rw-r--r--fs/ocfs2/cluster/nodemanager.c198
-rw-r--r--fs/ocfs2/cluster/nodemanager.h17
-rw-r--r--fs/ocfs2/cluster/quorum.c4
-rw-r--r--fs/ocfs2/cluster/tcp.c240
-rw-r--r--fs/ocfs2/cluster/tcp.h8
-rw-r--r--fs/ocfs2/cluster/tcp_internal.h23
-rw-r--r--fs/ocfs2/dir.c35
-rw-r--r--fs/ocfs2/dir.h2
-rw-r--r--fs/ocfs2/dlm/dlmcommon.h2
-rw-r--r--fs/ocfs2/dlm/dlmdomain.c9
-rw-r--r--fs/ocfs2/dlm/dlmfs.c10
-rw-r--r--fs/ocfs2/dlm/dlmlock.c4
-rw-r--r--fs/ocfs2/dlm/dlmmaster.c4
-rw-r--r--fs/ocfs2/dlm/dlmrecovery.c11
-rw-r--r--fs/ocfs2/dlm/userdlm.c10
-rw-r--r--fs/ocfs2/dlmglue.c149
-rw-r--r--fs/ocfs2/dlmglue.h9
-rw-r--r--fs/ocfs2/export.c2
-rw-r--r--fs/ocfs2/extent_map.c2
-rw-r--r--fs/ocfs2/file.c372
-rw-r--r--fs/ocfs2/file.h11
-rw-r--r--fs/ocfs2/heartbeat.c9
-rw-r--r--fs/ocfs2/inode.c36
-rw-r--r--fs/ocfs2/inode.h11
-rw-r--r--fs/ocfs2/ioctl.c10
-rw-r--r--fs/ocfs2/journal.c318
-rw-r--r--fs/ocfs2/journal.h85
-rw-r--r--fs/ocfs2/localalloc.c128
-rw-r--r--fs/ocfs2/localalloc.h3
-rw-r--r--fs/ocfs2/mmap.c17
-rw-r--r--fs/ocfs2/namei.c304
-rw-r--r--fs/ocfs2/namei.h2
-rw-r--r--fs/ocfs2/ocfs2.h11
-rw-r--r--fs/ocfs2/ocfs2_fs.h14
-rw-r--r--fs/ocfs2/slot_map.c2
-rw-r--r--fs/ocfs2/suballoc.c180
-rw-r--r--fs/ocfs2/suballoc.h16
-rw-r--r--fs/ocfs2/super.c140
-rw-r--r--fs/ocfs2/symlink.c4
-rw-r--r--fs/ocfs2/uptodate.c2
-rw-r--r--fs/ocfs2/vote.c7
46 files changed, 1488 insertions, 1119 deletions
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index f43bc5f18a35..f27e5378caf2 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -52,14 +52,14 @@ static int ocfs2_extent_contig(struct inode *inode,
52 u64 blkno); 52 u64 blkno);
53 53
54static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, 54static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
55 struct ocfs2_journal_handle *handle, 55 handle_t *handle,
56 struct inode *inode, 56 struct inode *inode,
57 int wanted, 57 int wanted,
58 struct ocfs2_alloc_context *meta_ac, 58 struct ocfs2_alloc_context *meta_ac,
59 struct buffer_head *bhs[]); 59 struct buffer_head *bhs[]);
60 60
61static int ocfs2_add_branch(struct ocfs2_super *osb, 61static int ocfs2_add_branch(struct ocfs2_super *osb,
62 struct ocfs2_journal_handle *handle, 62 handle_t *handle,
63 struct inode *inode, 63 struct inode *inode,
64 struct buffer_head *fe_bh, 64 struct buffer_head *fe_bh,
65 struct buffer_head *eb_bh, 65 struct buffer_head *eb_bh,
@@ -67,14 +67,14 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
67 struct ocfs2_alloc_context *meta_ac); 67 struct ocfs2_alloc_context *meta_ac);
68 68
69static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, 69static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
70 struct ocfs2_journal_handle *handle, 70 handle_t *handle,
71 struct inode *inode, 71 struct inode *inode,
72 struct buffer_head *fe_bh, 72 struct buffer_head *fe_bh,
73 struct ocfs2_alloc_context *meta_ac, 73 struct ocfs2_alloc_context *meta_ac,
74 struct buffer_head **ret_new_eb_bh); 74 struct buffer_head **ret_new_eb_bh);
75 75
76static int ocfs2_do_insert_extent(struct ocfs2_super *osb, 76static int ocfs2_do_insert_extent(struct ocfs2_super *osb,
77 struct ocfs2_journal_handle *handle, 77 handle_t *handle,
78 struct inode *inode, 78 struct inode *inode,
79 struct buffer_head *fe_bh, 79 struct buffer_head *fe_bh,
80 u64 blkno, 80 u64 blkno,
@@ -152,7 +152,7 @@ bail:
152 * l_count for you 152 * l_count for you
153 */ 153 */
154static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, 154static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
155 struct ocfs2_journal_handle *handle, 155 handle_t *handle,
156 struct inode *inode, 156 struct inode *inode,
157 int wanted, 157 int wanted,
158 struct ocfs2_alloc_context *meta_ac, 158 struct ocfs2_alloc_context *meta_ac,
@@ -253,7 +253,7 @@ bail:
253 * contain a single record with e_clusters == 0. 253 * contain a single record with e_clusters == 0.
254 */ 254 */
255static int ocfs2_add_branch(struct ocfs2_super *osb, 255static int ocfs2_add_branch(struct ocfs2_super *osb,
256 struct ocfs2_journal_handle *handle, 256 handle_t *handle,
257 struct inode *inode, 257 struct inode *inode,
258 struct buffer_head *fe_bh, 258 struct buffer_head *fe_bh,
259 struct buffer_head *eb_bh, 259 struct buffer_head *eb_bh,
@@ -418,7 +418,7 @@ bail:
418 * after this call. 418 * after this call.
419 */ 419 */
420static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, 420static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
421 struct ocfs2_journal_handle *handle, 421 handle_t *handle,
422 struct inode *inode, 422 struct inode *inode,
423 struct buffer_head *fe_bh, 423 struct buffer_head *fe_bh,
424 struct ocfs2_alloc_context *meta_ac, 424 struct ocfs2_alloc_context *meta_ac,
@@ -520,7 +520,7 @@ bail:
520 * down. 520 * down.
521 */ 521 */
522static int ocfs2_do_insert_extent(struct ocfs2_super *osb, 522static int ocfs2_do_insert_extent(struct ocfs2_super *osb,
523 struct ocfs2_journal_handle *handle, 523 handle_t *handle,
524 struct inode *inode, 524 struct inode *inode,
525 struct buffer_head *fe_bh, 525 struct buffer_head *fe_bh,
526 u64 start_blk, 526 u64 start_blk,
@@ -809,7 +809,7 @@ bail:
809 809
810/* the caller needs to update fe->i_clusters */ 810/* the caller needs to update fe->i_clusters */
811int ocfs2_insert_extent(struct ocfs2_super *osb, 811int ocfs2_insert_extent(struct ocfs2_super *osb,
812 struct ocfs2_journal_handle *handle, 812 handle_t *handle,
813 struct inode *inode, 813 struct inode *inode,
814 struct buffer_head *fe_bh, 814 struct buffer_head *fe_bh,
815 u64 start_blk, 815 u64 start_blk,
@@ -951,7 +951,7 @@ static int ocfs2_truncate_log_can_coalesce(struct ocfs2_truncate_log *tl,
951} 951}
952 952
953static int ocfs2_truncate_log_append(struct ocfs2_super *osb, 953static int ocfs2_truncate_log_append(struct ocfs2_super *osb,
954 struct ocfs2_journal_handle *handle, 954 handle_t *handle,
955 u64 start_blk, 955 u64 start_blk,
956 unsigned int num_clusters) 956 unsigned int num_clusters)
957{ 957{
@@ -1034,7 +1034,7 @@ bail:
1034} 1034}
1035 1035
1036static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, 1036static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
1037 struct ocfs2_journal_handle *handle, 1037 handle_t *handle,
1038 struct inode *data_alloc_inode, 1038 struct inode *data_alloc_inode,
1039 struct buffer_head *data_alloc_bh) 1039 struct buffer_head *data_alloc_bh)
1040{ 1040{
@@ -1113,7 +1113,7 @@ static int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
1113{ 1113{
1114 int status; 1114 int status;
1115 unsigned int num_to_flush; 1115 unsigned int num_to_flush;
1116 struct ocfs2_journal_handle *handle = NULL; 1116 handle_t *handle;
1117 struct inode *tl_inode = osb->osb_tl_inode; 1117 struct inode *tl_inode = osb->osb_tl_inode;
1118 struct inode *data_alloc_inode = NULL; 1118 struct inode *data_alloc_inode = NULL;
1119 struct buffer_head *tl_bh = osb->osb_tl_bh; 1119 struct buffer_head *tl_bh = osb->osb_tl_bh;
@@ -1130,7 +1130,7 @@ static int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
1130 if (!OCFS2_IS_VALID_DINODE(di)) { 1130 if (!OCFS2_IS_VALID_DINODE(di)) {
1131 OCFS2_RO_ON_INVALID_DINODE(osb->sb, di); 1131 OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
1132 status = -EIO; 1132 status = -EIO;
1133 goto bail; 1133 goto out;
1134 } 1134 }
1135 1135
1136 num_to_flush = le16_to_cpu(tl->tl_used); 1136 num_to_flush = le16_to_cpu(tl->tl_used);
@@ -1138,14 +1138,7 @@ static int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
1138 num_to_flush, (unsigned long long)OCFS2_I(tl_inode)->ip_blkno); 1138 num_to_flush, (unsigned long long)OCFS2_I(tl_inode)->ip_blkno);
1139 if (!num_to_flush) { 1139 if (!num_to_flush) {
1140 status = 0; 1140 status = 0;
1141 goto bail; 1141 goto out;
1142 }
1143
1144 handle = ocfs2_alloc_handle(osb);
1145 if (!handle) {
1146 status = -ENOMEM;
1147 mlog_errno(status);
1148 goto bail;
1149 } 1142 }
1150 1143
1151 data_alloc_inode = ocfs2_get_system_file_inode(osb, 1144 data_alloc_inode = ocfs2_get_system_file_inode(osb,
@@ -1154,41 +1147,40 @@ static int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
1154 if (!data_alloc_inode) { 1147 if (!data_alloc_inode) {
1155 status = -EINVAL; 1148 status = -EINVAL;
1156 mlog(ML_ERROR, "Could not get bitmap inode!\n"); 1149 mlog(ML_ERROR, "Could not get bitmap inode!\n");
1157 goto bail; 1150 goto out;
1158 } 1151 }
1159 1152
1160 ocfs2_handle_add_inode(handle, data_alloc_inode); 1153 mutex_lock(&data_alloc_inode->i_mutex);
1161 status = ocfs2_meta_lock(data_alloc_inode, handle, &data_alloc_bh, 1); 1154
1155 status = ocfs2_meta_lock(data_alloc_inode, &data_alloc_bh, 1);
1162 if (status < 0) { 1156 if (status < 0) {
1163 mlog_errno(status); 1157 mlog_errno(status);
1164 goto bail; 1158 goto out_mutex;
1165 } 1159 }
1166 1160
1167 handle = ocfs2_start_trans(osb, handle, OCFS2_TRUNCATE_LOG_UPDATE); 1161 handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
1168 if (IS_ERR(handle)) { 1162 if (IS_ERR(handle)) {
1169 status = PTR_ERR(handle); 1163 status = PTR_ERR(handle);
1170 handle = NULL;
1171 mlog_errno(status); 1164 mlog_errno(status);
1172 goto bail; 1165 goto out_unlock;
1173 } 1166 }
1174 1167
1175 status = ocfs2_replay_truncate_records(osb, handle, data_alloc_inode, 1168 status = ocfs2_replay_truncate_records(osb, handle, data_alloc_inode,
1176 data_alloc_bh); 1169 data_alloc_bh);
1177 if (status < 0) { 1170 if (status < 0)
1178 mlog_errno(status); 1171 mlog_errno(status);
1179 goto bail;
1180 }
1181 1172
1182bail: 1173 ocfs2_commit_trans(osb, handle);
1183 if (handle)
1184 ocfs2_commit_trans(handle);
1185 1174
1186 if (data_alloc_inode) 1175out_unlock:
1187 iput(data_alloc_inode); 1176 brelse(data_alloc_bh);
1177 ocfs2_meta_unlock(data_alloc_inode, 1);
1188 1178
1189 if (data_alloc_bh) 1179out_mutex:
1190 brelse(data_alloc_bh); 1180 mutex_unlock(&data_alloc_inode->i_mutex);
1181 iput(data_alloc_inode);
1191 1182
1183out:
1192 mlog_exit(status); 1184 mlog_exit(status);
1193 return status; 1185 return status;
1194} 1186}
@@ -1205,10 +1197,12 @@ int ocfs2_flush_truncate_log(struct ocfs2_super *osb)
1205 return status; 1197 return status;
1206} 1198}
1207 1199
1208static void ocfs2_truncate_log_worker(void *data) 1200static void ocfs2_truncate_log_worker(struct work_struct *work)
1209{ 1201{
1210 int status; 1202 int status;
1211 struct ocfs2_super *osb = data; 1203 struct ocfs2_super *osb =
1204 container_of(work, struct ocfs2_super,
1205 osb_truncate_log_wq.work);
1212 1206
1213 mlog_entry_void(); 1207 mlog_entry_void();
1214 1208
@@ -1347,7 +1341,7 @@ int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
1347 int i; 1341 int i;
1348 unsigned int clusters, num_recs, start_cluster; 1342 unsigned int clusters, num_recs, start_cluster;
1349 u64 start_blk; 1343 u64 start_blk;
1350 struct ocfs2_journal_handle *handle; 1344 handle_t *handle;
1351 struct inode *tl_inode = osb->osb_tl_inode; 1345 struct inode *tl_inode = osb->osb_tl_inode;
1352 struct ocfs2_truncate_log *tl; 1346 struct ocfs2_truncate_log *tl;
1353 1347
@@ -1373,8 +1367,7 @@ int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
1373 } 1367 }
1374 } 1368 }
1375 1369
1376 handle = ocfs2_start_trans(osb, NULL, 1370 handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
1377 OCFS2_TRUNCATE_LOG_UPDATE);
1378 if (IS_ERR(handle)) { 1371 if (IS_ERR(handle)) {
1379 status = PTR_ERR(handle); 1372 status = PTR_ERR(handle);
1380 mlog_errno(status); 1373 mlog_errno(status);
@@ -1387,7 +1380,7 @@ int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
1387 1380
1388 status = ocfs2_truncate_log_append(osb, handle, 1381 status = ocfs2_truncate_log_append(osb, handle,
1389 start_blk, clusters); 1382 start_blk, clusters);
1390 ocfs2_commit_trans(handle); 1383 ocfs2_commit_trans(osb, handle);
1391 if (status < 0) { 1384 if (status < 0) {
1392 mlog_errno(status); 1385 mlog_errno(status);
1393 goto bail_up; 1386 goto bail_up;
@@ -1441,7 +1434,8 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb)
1441 /* ocfs2_truncate_log_shutdown keys on the existence of 1434 /* ocfs2_truncate_log_shutdown keys on the existence of
1442 * osb->osb_tl_inode so we don't set any of the osb variables 1435 * osb->osb_tl_inode so we don't set any of the osb variables
1443 * until we're sure all is well. */ 1436 * until we're sure all is well. */
1444 INIT_WORK(&osb->osb_truncate_log_wq, ocfs2_truncate_log_worker, osb); 1437 INIT_DELAYED_WORK(&osb->osb_truncate_log_wq,
1438 ocfs2_truncate_log_worker);
1445 osb->osb_tl_bh = tl_bh; 1439 osb->osb_tl_bh = tl_bh;
1446 osb->osb_tl_inode = tl_inode; 1440 osb->osb_tl_inode = tl_inode;
1447 1441
@@ -1543,7 +1537,7 @@ static int ocfs2_do_truncate(struct ocfs2_super *osb,
1543 struct inode *inode, 1537 struct inode *inode,
1544 struct buffer_head *fe_bh, 1538 struct buffer_head *fe_bh,
1545 struct buffer_head *old_last_eb_bh, 1539 struct buffer_head *old_last_eb_bh,
1546 struct ocfs2_journal_handle *handle, 1540 handle_t *handle,
1547 struct ocfs2_truncate_context *tc) 1541 struct ocfs2_truncate_context *tc)
1548{ 1542{
1549 int status, i, depth; 1543 int status, i, depth;
@@ -1782,7 +1776,7 @@ int ocfs2_commit_truncate(struct ocfs2_super *osb,
1782 struct ocfs2_extent_block *eb; 1776 struct ocfs2_extent_block *eb;
1783 struct ocfs2_extent_list *el; 1777 struct ocfs2_extent_list *el;
1784 struct buffer_head *last_eb_bh; 1778 struct buffer_head *last_eb_bh;
1785 struct ocfs2_journal_handle *handle = NULL; 1779 handle_t *handle = NULL;
1786 struct inode *tl_inode = osb->osb_tl_inode; 1780 struct inode *tl_inode = osb->osb_tl_inode;
1787 1781
1788 mlog_entry_void(); 1782 mlog_entry_void();
@@ -1868,7 +1862,7 @@ start:
1868 1862
1869 credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del, 1863 credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
1870 fe, el); 1864 fe, el);
1871 handle = ocfs2_start_trans(osb, NULL, credits); 1865 handle = ocfs2_start_trans(osb, credits);
1872 if (IS_ERR(handle)) { 1866 if (IS_ERR(handle)) {
1873 status = PTR_ERR(handle); 1867 status = PTR_ERR(handle);
1874 handle = NULL; 1868 handle = NULL;
@@ -1891,7 +1885,7 @@ start:
1891 mutex_unlock(&tl_inode->i_mutex); 1885 mutex_unlock(&tl_inode->i_mutex);
1892 tl_sem = 0; 1886 tl_sem = 0;
1893 1887
1894 ocfs2_commit_trans(handle); 1888 ocfs2_commit_trans(osb, handle);
1895 handle = NULL; 1889 handle = NULL;
1896 1890
1897 BUG_ON(le32_to_cpu(fe->i_clusters) < target_i_clusters); 1891 BUG_ON(le32_to_cpu(fe->i_clusters) < target_i_clusters);
@@ -1906,7 +1900,7 @@ bail:
1906 mutex_unlock(&tl_inode->i_mutex); 1900 mutex_unlock(&tl_inode->i_mutex);
1907 1901
1908 if (handle) 1902 if (handle)
1909 ocfs2_commit_trans(handle); 1903 ocfs2_commit_trans(osb, handle);
1910 1904
1911 if (last_eb_bh) 1905 if (last_eb_bh)
1912 brelse(last_eb_bh); 1906 brelse(last_eb_bh);
@@ -1965,7 +1959,7 @@ int ocfs2_prepare_truncate(struct ocfs2_super *osb,
1965 goto bail; 1959 goto bail;
1966 } 1960 }
1967 1961
1968 *tc = kcalloc(1, sizeof(struct ocfs2_truncate_context), GFP_KERNEL); 1962 *tc = kzalloc(sizeof(struct ocfs2_truncate_context), GFP_KERNEL);
1969 if (!(*tc)) { 1963 if (!(*tc)) {
1970 status = -ENOMEM; 1964 status = -ENOMEM;
1971 mlog_errno(status); 1965 mlog_errno(status);
@@ -2011,10 +2005,7 @@ int ocfs2_prepare_truncate(struct ocfs2_super *osb,
2011 mutex_lock(&ext_alloc_inode->i_mutex); 2005 mutex_lock(&ext_alloc_inode->i_mutex);
2012 (*tc)->tc_ext_alloc_inode = ext_alloc_inode; 2006 (*tc)->tc_ext_alloc_inode = ext_alloc_inode;
2013 2007
2014 status = ocfs2_meta_lock(ext_alloc_inode, 2008 status = ocfs2_meta_lock(ext_alloc_inode, &ext_alloc_bh, 1);
2015 NULL,
2016 &ext_alloc_bh,
2017 1);
2018 if (status < 0) { 2009 if (status < 0) {
2019 mlog_errno(status); 2010 mlog_errno(status);
2020 goto bail; 2011 goto bail;
diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h
index 12ba897743f4..0b82e8044325 100644
--- a/fs/ocfs2/alloc.h
+++ b/fs/ocfs2/alloc.h
@@ -28,7 +28,7 @@
28 28
29struct ocfs2_alloc_context; 29struct ocfs2_alloc_context;
30int ocfs2_insert_extent(struct ocfs2_super *osb, 30int ocfs2_insert_extent(struct ocfs2_super *osb,
31 struct ocfs2_journal_handle *handle, 31 handle_t *handle,
32 struct inode *inode, 32 struct inode *inode,
33 struct buffer_head *fe_bh, 33 struct buffer_head *fe_bh,
34 u64 blkno, 34 u64 blkno,
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 3d7c082a8f58..93628b02ef5d 100644
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -200,7 +200,7 @@ static int ocfs2_readpage(struct file *file, struct page *page)
200 200
201 mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0)); 201 mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0));
202 202
203 ret = ocfs2_meta_lock_with_page(inode, NULL, NULL, 0, page); 203 ret = ocfs2_meta_lock_with_page(inode, NULL, 0, page);
204 if (ret != 0) { 204 if (ret != 0) {
205 if (ret == AOP_TRUNCATED_PAGE) 205 if (ret == AOP_TRUNCATED_PAGE)
206 unlock = 0; 206 unlock = 0;
@@ -305,7 +305,7 @@ static int ocfs2_prepare_write(struct file *file, struct page *page,
305 305
306 mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to); 306 mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to);
307 307
308 ret = ocfs2_meta_lock_with_page(inode, NULL, NULL, 0, page); 308 ret = ocfs2_meta_lock_with_page(inode, NULL, 0, page);
309 if (ret != 0) { 309 if (ret != 0) {
310 mlog_errno(ret); 310 mlog_errno(ret);
311 goto out; 311 goto out;
@@ -355,16 +355,16 @@ static int walk_page_buffers( handle_t *handle,
355 return ret; 355 return ret;
356} 356}
357 357
358struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode, 358handle_t *ocfs2_start_walk_page_trans(struct inode *inode,
359 struct page *page, 359 struct page *page,
360 unsigned from, 360 unsigned from,
361 unsigned to) 361 unsigned to)
362{ 362{
363 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 363 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
364 struct ocfs2_journal_handle *handle = NULL; 364 handle_t *handle = NULL;
365 int ret = 0; 365 int ret = 0;
366 366
367 handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS); 367 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
368 if (!handle) { 368 if (!handle) {
369 ret = -ENOMEM; 369 ret = -ENOMEM;
370 mlog_errno(ret); 370 mlog_errno(ret);
@@ -372,7 +372,7 @@ struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode,
372 } 372 }
373 373
374 if (ocfs2_should_order_data(inode)) { 374 if (ocfs2_should_order_data(inode)) {
375 ret = walk_page_buffers(handle->k_handle, 375 ret = walk_page_buffers(handle,
376 page_buffers(page), 376 page_buffers(page),
377 from, to, NULL, 377 from, to, NULL,
378 ocfs2_journal_dirty_data); 378 ocfs2_journal_dirty_data);
@@ -382,7 +382,7 @@ struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode,
382out: 382out:
383 if (ret) { 383 if (ret) {
384 if (handle) 384 if (handle)
385 ocfs2_commit_trans(handle); 385 ocfs2_commit_trans(osb, handle);
386 handle = ERR_PTR(ret); 386 handle = ERR_PTR(ret);
387 } 387 }
388 return handle; 388 return handle;
@@ -394,7 +394,7 @@ static int ocfs2_commit_write(struct file *file, struct page *page,
394 int ret; 394 int ret;
395 struct buffer_head *di_bh = NULL; 395 struct buffer_head *di_bh = NULL;
396 struct inode *inode = page->mapping->host; 396 struct inode *inode = page->mapping->host;
397 struct ocfs2_journal_handle *handle = NULL; 397 handle_t *handle = NULL;
398 struct ocfs2_dinode *di; 398 struct ocfs2_dinode *di;
399 399
400 mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to); 400 mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to);
@@ -412,7 +412,7 @@ static int ocfs2_commit_write(struct file *file, struct page *page,
412 * stale inode allocation image (i_size, i_clusters, etc). 412 * stale inode allocation image (i_size, i_clusters, etc).
413 */ 413 */
414 414
415 ret = ocfs2_meta_lock_with_page(inode, NULL, &di_bh, 1, page); 415 ret = ocfs2_meta_lock_with_page(inode, &di_bh, 1, page);
416 if (ret != 0) { 416 if (ret != 0) {
417 mlog_errno(ret); 417 mlog_errno(ret);
418 goto out; 418 goto out;
@@ -464,7 +464,7 @@ static int ocfs2_commit_write(struct file *file, struct page *page,
464 } 464 }
465 465
466out_commit: 466out_commit:
467 ocfs2_commit_trans(handle); 467 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
468out_unlock_data: 468out_unlock_data:
469 ocfs2_data_unlock(inode, 1); 469 ocfs2_data_unlock(inode, 1);
470out_unlock_meta: 470out_unlock_meta:
@@ -490,7 +490,7 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block)
490 * accessed concurrently from multiple nodes. 490 * accessed concurrently from multiple nodes.
491 */ 491 */
492 if (!INODE_JOURNAL(inode)) { 492 if (!INODE_JOURNAL(inode)) {
493 err = ocfs2_meta_lock(inode, NULL, NULL, 0); 493 err = ocfs2_meta_lock(inode, NULL, 0);
494 if (err) { 494 if (err) {
495 if (err != -ENOENT) 495 if (err != -ENOENT)
496 mlog_errno(err); 496 mlog_errno(err);
@@ -540,8 +540,7 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock,
540 struct buffer_head *bh_result, int create) 540 struct buffer_head *bh_result, int create)
541{ 541{
542 int ret; 542 int ret;
543 u64 vbo_max; /* file offset, max_blocks from iblock */ 543 u64 p_blkno, inode_blocks;
544 u64 p_blkno;
545 int contig_blocks; 544 int contig_blocks;
546 unsigned char blocksize_bits = inode->i_sb->s_blocksize_bits; 545 unsigned char blocksize_bits = inode->i_sb->s_blocksize_bits;
547 unsigned long max_blocks = bh_result->b_size >> inode->i_blkbits; 546 unsigned long max_blocks = bh_result->b_size >> inode->i_blkbits;
@@ -550,12 +549,23 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock,
550 * nicely aligned and of the right size, so there's no need 549 * nicely aligned and of the right size, so there's no need
551 * for us to check any of that. */ 550 * for us to check any of that. */
552 551
553 vbo_max = ((u64)iblock + max_blocks) << blocksize_bits;
554
555 spin_lock(&OCFS2_I(inode)->ip_lock); 552 spin_lock(&OCFS2_I(inode)->ip_lock);
556 if ((iblock + max_blocks) > 553 inode_blocks = ocfs2_clusters_to_blocks(inode->i_sb,
557 ocfs2_clusters_to_blocks(inode->i_sb, 554 OCFS2_I(inode)->ip_clusters);
558 OCFS2_I(inode)->ip_clusters)) { 555
556 /*
557 * For a read which begins past the end of file, we return a hole.
558 */
559 if (!create && (iblock >= inode_blocks)) {
560 spin_unlock(&OCFS2_I(inode)->ip_lock);
561 ret = 0;
562 goto bail;
563 }
564
565 /*
566 * Any write past EOF is not allowed because we'd be extending.
567 */
568 if (create && (iblock + max_blocks) > inode_blocks) {
559 spin_unlock(&OCFS2_I(inode)->ip_lock); 569 spin_unlock(&OCFS2_I(inode)->ip_lock);
560 ret = -EIO; 570 ret = -EIO;
561 goto bail; 571 goto bail;
@@ -595,7 +605,7 @@ static void ocfs2_dio_end_io(struct kiocb *iocb,
595 ssize_t bytes, 605 ssize_t bytes,
596 void *private) 606 void *private)
597{ 607{
598 struct inode *inode = iocb->ki_filp->f_dentry->d_inode; 608 struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;
599 609
600 /* this io's submitter should not have unlocked this before we could */ 610 /* this io's submitter should not have unlocked this before we could */
601 BUG_ON(!ocfs2_iocb_is_rw_locked(iocb)); 611 BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
@@ -611,7 +621,7 @@ static ssize_t ocfs2_direct_IO(int rw,
611 unsigned long nr_segs) 621 unsigned long nr_segs)
612{ 622{
613 struct file *file = iocb->ki_filp; 623 struct file *file = iocb->ki_filp;
614 struct inode *inode = file->f_dentry->d_inode->i_mapping->host; 624 struct inode *inode = file->f_path.dentry->d_inode->i_mapping->host;
615 int ret; 625 int ret;
616 626
617 mlog_entry_void(); 627 mlog_entry_void();
diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h
index e88c3f0b8fa9..f446a15eab88 100644
--- a/fs/ocfs2/aops.h
+++ b/fs/ocfs2/aops.h
@@ -25,7 +25,7 @@
25int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page, 25int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page,
26 unsigned from, unsigned to); 26 unsigned from, unsigned to);
27 27
28struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode, 28handle_t *ocfs2_start_walk_page_trans(struct inode *inode,
29 struct page *page, 29 struct page *page,
30 unsigned from, 30 unsigned from,
31 unsigned to); 31 unsigned to);
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index 305cba3681fe..277ca67a2ad6 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -141,7 +141,7 @@ struct o2hb_region {
141 * recognizes a node going up and down in one iteration */ 141 * recognizes a node going up and down in one iteration */
142 u64 hr_generation; 142 u64 hr_generation;
143 143
144 struct work_struct hr_write_timeout_work; 144 struct delayed_work hr_write_timeout_work;
145 unsigned long hr_last_timeout_start; 145 unsigned long hr_last_timeout_start;
146 146
147 /* Used during o2hb_check_slot to hold a copy of the block 147 /* Used during o2hb_check_slot to hold a copy of the block
@@ -156,9 +156,11 @@ struct o2hb_bio_wait_ctxt {
156 int wc_error; 156 int wc_error;
157}; 157};
158 158
159static void o2hb_write_timeout(void *arg) 159static void o2hb_write_timeout(struct work_struct *work)
160{ 160{
161 struct o2hb_region *reg = arg; 161 struct o2hb_region *reg =
162 container_of(work, struct o2hb_region,
163 hr_write_timeout_work.work);
162 164
163 mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u " 165 mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
164 "milliseconds\n", reg->hr_dev_name, 166 "milliseconds\n", reg->hr_dev_name,
@@ -1404,7 +1406,7 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1404 goto out; 1406 goto out;
1405 } 1407 }
1406 1408
1407 INIT_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout, reg); 1409 INIT_DELAYED_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout);
1408 1410
1409 /* 1411 /*
1410 * A node is considered live after it has beat LIVE_THRESHOLD 1412 * A node is considered live after it has beat LIVE_THRESHOLD
@@ -1445,6 +1447,15 @@ out:
1445 return ret; 1447 return ret;
1446} 1448}
1447 1449
1450static ssize_t o2hb_region_pid_read(struct o2hb_region *reg,
1451 char *page)
1452{
1453 if (!reg->hr_task)
1454 return 0;
1455
1456 return sprintf(page, "%u\n", reg->hr_task->pid);
1457}
1458
1448struct o2hb_region_attribute { 1459struct o2hb_region_attribute {
1449 struct configfs_attribute attr; 1460 struct configfs_attribute attr;
1450 ssize_t (*show)(struct o2hb_region *, char *); 1461 ssize_t (*show)(struct o2hb_region *, char *);
@@ -1483,11 +1494,19 @@ static struct o2hb_region_attribute o2hb_region_attr_dev = {
1483 .store = o2hb_region_dev_write, 1494 .store = o2hb_region_dev_write,
1484}; 1495};
1485 1496
1497static struct o2hb_region_attribute o2hb_region_attr_pid = {
1498 .attr = { .ca_owner = THIS_MODULE,
1499 .ca_name = "pid",
1500 .ca_mode = S_IRUGO | S_IRUSR },
1501 .show = o2hb_region_pid_read,
1502};
1503
1486static struct configfs_attribute *o2hb_region_attrs[] = { 1504static struct configfs_attribute *o2hb_region_attrs[] = {
1487 &o2hb_region_attr_block_bytes.attr, 1505 &o2hb_region_attr_block_bytes.attr,
1488 &o2hb_region_attr_start_block.attr, 1506 &o2hb_region_attr_start_block.attr,
1489 &o2hb_region_attr_blocks.attr, 1507 &o2hb_region_attr_blocks.attr,
1490 &o2hb_region_attr_dev.attr, 1508 &o2hb_region_attr_dev.attr,
1509 &o2hb_region_attr_pid.attr,
1491 NULL, 1510 NULL,
1492}; 1511};
1493 1512
@@ -1551,7 +1570,7 @@ static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *g
1551 struct o2hb_region *reg = NULL; 1570 struct o2hb_region *reg = NULL;
1552 struct config_item *ret = NULL; 1571 struct config_item *ret = NULL;
1553 1572
1554 reg = kcalloc(1, sizeof(struct o2hb_region), GFP_KERNEL); 1573 reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
1555 if (reg == NULL) 1574 if (reg == NULL)
1556 goto out; /* ENOMEM */ 1575 goto out; /* ENOMEM */
1557 1576
@@ -1677,7 +1696,7 @@ struct config_group *o2hb_alloc_hb_set(void)
1677 struct o2hb_heartbeat_group *hs = NULL; 1696 struct o2hb_heartbeat_group *hs = NULL;
1678 struct config_group *ret = NULL; 1697 struct config_group *ret = NULL;
1679 1698
1680 hs = kcalloc(1, sizeof(struct o2hb_heartbeat_group), GFP_KERNEL); 1699 hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
1681 if (hs == NULL) 1700 if (hs == NULL)
1682 goto out; 1701 goto out;
1683 1702
diff --git a/fs/ocfs2/cluster/nodemanager.c b/fs/ocfs2/cluster/nodemanager.c
index d11753c50bc1..b17333a0606b 100644
--- a/fs/ocfs2/cluster/nodemanager.c
+++ b/fs/ocfs2/cluster/nodemanager.c
@@ -35,7 +35,7 @@
35/* for now we operate under the assertion that there can be only one 35/* for now we operate under the assertion that there can be only one
36 * cluster active at a time. Changing this will require trickling 36 * cluster active at a time. Changing this will require trickling
37 * cluster references throughout where nodes are looked up */ 37 * cluster references throughout where nodes are looked up */
38static struct o2nm_cluster *o2nm_single_cluster = NULL; 38struct o2nm_cluster *o2nm_single_cluster = NULL;
39 39
40#define OCFS2_MAX_HB_CTL_PATH 256 40#define OCFS2_MAX_HB_CTL_PATH 256
41static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl"; 41static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl";
@@ -97,17 +97,6 @@ const char *o2nm_get_hb_ctl_path(void)
97} 97}
98EXPORT_SYMBOL_GPL(o2nm_get_hb_ctl_path); 98EXPORT_SYMBOL_GPL(o2nm_get_hb_ctl_path);
99 99
100struct o2nm_cluster {
101 struct config_group cl_group;
102 unsigned cl_has_local:1;
103 u8 cl_local_node;
104 rwlock_t cl_nodes_lock;
105 struct o2nm_node *cl_nodes[O2NM_MAX_NODES];
106 struct rb_root cl_node_ip_tree;
107 /* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */
108 unsigned long cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
109};
110
111struct o2nm_node *o2nm_get_node_by_num(u8 node_num) 100struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
112{ 101{
113 struct o2nm_node *node = NULL; 102 struct o2nm_node *node = NULL;
@@ -543,6 +532,179 @@ static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
543} 532}
544#endif 533#endif
545 534
535struct o2nm_cluster_attribute {
536 struct configfs_attribute attr;
537 ssize_t (*show)(struct o2nm_cluster *, char *);
538 ssize_t (*store)(struct o2nm_cluster *, const char *, size_t);
539};
540
541static ssize_t o2nm_cluster_attr_write(const char *page, ssize_t count,
542 unsigned int *val)
543{
544 unsigned long tmp;
545 char *p = (char *)page;
546
547 tmp = simple_strtoul(p, &p, 0);
548 if (!p || (*p && (*p != '\n')))
549 return -EINVAL;
550
551 if (tmp == 0)
552 return -EINVAL;
553 if (tmp >= (u32)-1)
554 return -ERANGE;
555
556 *val = tmp;
557
558 return count;
559}
560
561static ssize_t o2nm_cluster_attr_idle_timeout_ms_read(
562 struct o2nm_cluster *cluster, char *page)
563{
564 return sprintf(page, "%u\n", cluster->cl_idle_timeout_ms);
565}
566
567static ssize_t o2nm_cluster_attr_idle_timeout_ms_write(
568 struct o2nm_cluster *cluster, const char *page, size_t count)
569{
570 ssize_t ret;
571 unsigned int val;
572
573 ret = o2nm_cluster_attr_write(page, count, &val);
574
575 if (ret > 0) {
576 if (cluster->cl_idle_timeout_ms != val
577 && o2net_num_connected_peers()) {
578 mlog(ML_NOTICE,
579 "o2net: cannot change idle timeout after "
580 "the first peer has agreed to it."
581 " %d connected peers\n",
582 o2net_num_connected_peers());
583 ret = -EINVAL;
584 } else if (val <= cluster->cl_keepalive_delay_ms) {
585 mlog(ML_NOTICE, "o2net: idle timeout must be larger "
586 "than keepalive delay\n");
587 ret = -EINVAL;
588 } else {
589 cluster->cl_idle_timeout_ms = val;
590 }
591 }
592
593 return ret;
594}
595
596static ssize_t o2nm_cluster_attr_keepalive_delay_ms_read(
597 struct o2nm_cluster *cluster, char *page)
598{
599 return sprintf(page, "%u\n", cluster->cl_keepalive_delay_ms);
600}
601
602static ssize_t o2nm_cluster_attr_keepalive_delay_ms_write(
603 struct o2nm_cluster *cluster, const char *page, size_t count)
604{
605 ssize_t ret;
606 unsigned int val;
607
608 ret = o2nm_cluster_attr_write(page, count, &val);
609
610 if (ret > 0) {
611 if (cluster->cl_keepalive_delay_ms != val
612 && o2net_num_connected_peers()) {
613 mlog(ML_NOTICE,
614 "o2net: cannot change keepalive delay after"
615 " the first peer has agreed to it."
616 " %d connected peers\n",
617 o2net_num_connected_peers());
618 ret = -EINVAL;
619 } else if (val >= cluster->cl_idle_timeout_ms) {
620 mlog(ML_NOTICE, "o2net: keepalive delay must be "
621 "smaller than idle timeout\n");
622 ret = -EINVAL;
623 } else {
624 cluster->cl_keepalive_delay_ms = val;
625 }
626 }
627
628 return ret;
629}
630
631static ssize_t o2nm_cluster_attr_reconnect_delay_ms_read(
632 struct o2nm_cluster *cluster, char *page)
633{
634 return sprintf(page, "%u\n", cluster->cl_reconnect_delay_ms);
635}
636
637static ssize_t o2nm_cluster_attr_reconnect_delay_ms_write(
638 struct o2nm_cluster *cluster, const char *page, size_t count)
639{
640 return o2nm_cluster_attr_write(page, count,
641 &cluster->cl_reconnect_delay_ms);
642}
643static struct o2nm_cluster_attribute o2nm_cluster_attr_idle_timeout_ms = {
644 .attr = { .ca_owner = THIS_MODULE,
645 .ca_name = "idle_timeout_ms",
646 .ca_mode = S_IRUGO | S_IWUSR },
647 .show = o2nm_cluster_attr_idle_timeout_ms_read,
648 .store = o2nm_cluster_attr_idle_timeout_ms_write,
649};
650
651static struct o2nm_cluster_attribute o2nm_cluster_attr_keepalive_delay_ms = {
652 .attr = { .ca_owner = THIS_MODULE,
653 .ca_name = "keepalive_delay_ms",
654 .ca_mode = S_IRUGO | S_IWUSR },
655 .show = o2nm_cluster_attr_keepalive_delay_ms_read,
656 .store = o2nm_cluster_attr_keepalive_delay_ms_write,
657};
658
659static struct o2nm_cluster_attribute o2nm_cluster_attr_reconnect_delay_ms = {
660 .attr = { .ca_owner = THIS_MODULE,
661 .ca_name = "reconnect_delay_ms",
662 .ca_mode = S_IRUGO | S_IWUSR },
663 .show = o2nm_cluster_attr_reconnect_delay_ms_read,
664 .store = o2nm_cluster_attr_reconnect_delay_ms_write,
665};
666
667static struct configfs_attribute *o2nm_cluster_attrs[] = {
668 &o2nm_cluster_attr_idle_timeout_ms.attr,
669 &o2nm_cluster_attr_keepalive_delay_ms.attr,
670 &o2nm_cluster_attr_reconnect_delay_ms.attr,
671 NULL,
672};
673static ssize_t o2nm_cluster_show(struct config_item *item,
674 struct configfs_attribute *attr,
675 char *page)
676{
677 struct o2nm_cluster *cluster = to_o2nm_cluster(item);
678 struct o2nm_cluster_attribute *o2nm_cluster_attr =
679 container_of(attr, struct o2nm_cluster_attribute, attr);
680 ssize_t ret = 0;
681
682 if (o2nm_cluster_attr->show)
683 ret = o2nm_cluster_attr->show(cluster, page);
684 return ret;
685}
686
687static ssize_t o2nm_cluster_store(struct config_item *item,
688 struct configfs_attribute *attr,
689 const char *page, size_t count)
690{
691 struct o2nm_cluster *cluster = to_o2nm_cluster(item);
692 struct o2nm_cluster_attribute *o2nm_cluster_attr =
693 container_of(attr, struct o2nm_cluster_attribute, attr);
694 ssize_t ret;
695
696 if (o2nm_cluster_attr->store == NULL) {
697 ret = -EINVAL;
698 goto out;
699 }
700
701 ret = o2nm_cluster_attr->store(cluster, page, count);
702 if (ret < count)
703 goto out;
704out:
705 return ret;
706}
707
546static struct config_item *o2nm_node_group_make_item(struct config_group *group, 708static struct config_item *o2nm_node_group_make_item(struct config_group *group,
547 const char *name) 709 const char *name)
548{ 710{
@@ -552,7 +714,7 @@ static struct config_item *o2nm_node_group_make_item(struct config_group *group,
552 if (strlen(name) > O2NM_MAX_NAME_LEN) 714 if (strlen(name) > O2NM_MAX_NAME_LEN)
553 goto out; /* ENAMETOOLONG */ 715 goto out; /* ENAMETOOLONG */
554 716
555 node = kcalloc(1, sizeof(struct o2nm_node), GFP_KERNEL); 717 node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL);
556 if (node == NULL) 718 if (node == NULL)
557 goto out; /* ENOMEM */ 719 goto out; /* ENOMEM */
558 720
@@ -624,10 +786,13 @@ static void o2nm_cluster_release(struct config_item *item)
624 786
625static struct configfs_item_operations o2nm_cluster_item_ops = { 787static struct configfs_item_operations o2nm_cluster_item_ops = {
626 .release = o2nm_cluster_release, 788 .release = o2nm_cluster_release,
789 .show_attribute = o2nm_cluster_show,
790 .store_attribute = o2nm_cluster_store,
627}; 791};
628 792
629static struct config_item_type o2nm_cluster_type = { 793static struct config_item_type o2nm_cluster_type = {
630 .ct_item_ops = &o2nm_cluster_item_ops, 794 .ct_item_ops = &o2nm_cluster_item_ops,
795 .ct_attrs = o2nm_cluster_attrs,
631 .ct_owner = THIS_MODULE, 796 .ct_owner = THIS_MODULE,
632}; 797};
633 798
@@ -660,8 +825,8 @@ static struct config_group *o2nm_cluster_group_make_group(struct config_group *g
660 if (o2nm_single_cluster) 825 if (o2nm_single_cluster)
661 goto out; /* ENOSPC */ 826 goto out; /* ENOSPC */
662 827
663 cluster = kcalloc(1, sizeof(struct o2nm_cluster), GFP_KERNEL); 828 cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL);
664 ns = kcalloc(1, sizeof(struct o2nm_node_group), GFP_KERNEL); 829 ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL);
665 defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL); 830 defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
666 o2hb_group = o2hb_alloc_hb_set(); 831 o2hb_group = o2hb_alloc_hb_set();
667 if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL) 832 if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL)
@@ -678,6 +843,9 @@ static struct config_group *o2nm_cluster_group_make_group(struct config_group *g
678 cluster->cl_group.default_groups[2] = NULL; 843 cluster->cl_group.default_groups[2] = NULL;
679 rwlock_init(&cluster->cl_nodes_lock); 844 rwlock_init(&cluster->cl_nodes_lock);
680 cluster->cl_node_ip_tree = RB_ROOT; 845 cluster->cl_node_ip_tree = RB_ROOT;
846 cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT;
847 cluster->cl_idle_timeout_ms = O2NET_IDLE_TIMEOUT_MS_DEFAULT;
848 cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT;
681 849
682 ret = &cluster->cl_group; 850 ret = &cluster->cl_group;
683 o2nm_single_cluster = cluster; 851 o2nm_single_cluster = cluster;
diff --git a/fs/ocfs2/cluster/nodemanager.h b/fs/ocfs2/cluster/nodemanager.h
index fce8033c310f..8fb23cacc2f5 100644
--- a/fs/ocfs2/cluster/nodemanager.h
+++ b/fs/ocfs2/cluster/nodemanager.h
@@ -53,6 +53,23 @@ struct o2nm_node {
53 unsigned long nd_set_attributes; 53 unsigned long nd_set_attributes;
54}; 54};
55 55
56struct o2nm_cluster {
57 struct config_group cl_group;
58 unsigned cl_has_local:1;
59 u8 cl_local_node;
60 rwlock_t cl_nodes_lock;
61 struct o2nm_node *cl_nodes[O2NM_MAX_NODES];
62 struct rb_root cl_node_ip_tree;
63 unsigned int cl_idle_timeout_ms;
64 unsigned int cl_keepalive_delay_ms;
65 unsigned int cl_reconnect_delay_ms;
66
67 /* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */
68 unsigned long cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
69};
70
71extern struct o2nm_cluster *o2nm_single_cluster;
72
56u8 o2nm_this_node(void); 73u8 o2nm_this_node(void);
57 74
58int o2nm_configured_node_map(unsigned long *map, unsigned bytes); 75int o2nm_configured_node_map(unsigned long *map, unsigned bytes);
diff --git a/fs/ocfs2/cluster/quorum.c b/fs/ocfs2/cluster/quorum.c
index 7bba98fbfc15..4705d659fe57 100644
--- a/fs/ocfs2/cluster/quorum.c
+++ b/fs/ocfs2/cluster/quorum.c
@@ -88,7 +88,7 @@ void o2quo_disk_timeout(void)
88 o2quo_fence_self(); 88 o2quo_fence_self();
89} 89}
90 90
91static void o2quo_make_decision(void *arg) 91static void o2quo_make_decision(struct work_struct *work)
92{ 92{
93 int quorum; 93 int quorum;
94 int lowest_hb, lowest_reachable = 0, fence = 0; 94 int lowest_hb, lowest_reachable = 0, fence = 0;
@@ -306,7 +306,7 @@ void o2quo_init(void)
306 struct o2quo_state *qs = &o2quo_state; 306 struct o2quo_state *qs = &o2quo_state;
307 307
308 spin_lock_init(&qs->qs_lock); 308 spin_lock_init(&qs->qs_lock);
309 INIT_WORK(&qs->qs_work, o2quo_make_decision, NULL); 309 INIT_WORK(&qs->qs_work, o2quo_make_decision);
310} 310}
311 311
312void o2quo_exit(void) 312void o2quo_exit(void)
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index b650efa8c8be..ae4ff4a6636b 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -140,13 +140,35 @@ static int o2net_sys_err_translations[O2NET_ERR_MAX] =
140 [O2NET_ERR_DIED] = -EHOSTDOWN,}; 140 [O2NET_ERR_DIED] = -EHOSTDOWN,};
141 141
142/* can't quite avoid *all* internal declarations :/ */ 142/* can't quite avoid *all* internal declarations :/ */
143static void o2net_sc_connect_completed(void *arg); 143static void o2net_sc_connect_completed(struct work_struct *work);
144static void o2net_rx_until_empty(void *arg); 144static void o2net_rx_until_empty(struct work_struct *work);
145static void o2net_shutdown_sc(void *arg); 145static void o2net_shutdown_sc(struct work_struct *work);
146static void o2net_listen_data_ready(struct sock *sk, int bytes); 146static void o2net_listen_data_ready(struct sock *sk, int bytes);
147static void o2net_sc_send_keep_req(void *arg); 147static void o2net_sc_send_keep_req(struct work_struct *work);
148static void o2net_idle_timer(unsigned long data); 148static void o2net_idle_timer(unsigned long data);
149static void o2net_sc_postpone_idle(struct o2net_sock_container *sc); 149static void o2net_sc_postpone_idle(struct o2net_sock_container *sc);
150static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc);
151
152/*
153 * FIXME: These should use to_o2nm_cluster_from_node(), but we end up
154 * losing our parent link to the cluster during shutdown. This can be
155 * solved by adding a pre-removal callback to configfs, or passing
156 * around the cluster with the node. -jeffm
157 */
158static inline int o2net_reconnect_delay(struct o2nm_node *node)
159{
160 return o2nm_single_cluster->cl_reconnect_delay_ms;
161}
162
163static inline int o2net_keepalive_delay(struct o2nm_node *node)
164{
165 return o2nm_single_cluster->cl_keepalive_delay_ms;
166}
167
168static inline int o2net_idle_timeout(struct o2nm_node *node)
169{
170 return o2nm_single_cluster->cl_idle_timeout_ms;
171}
150 172
151static inline int o2net_sys_err_to_errno(enum o2net_system_error err) 173static inline int o2net_sys_err_to_errno(enum o2net_system_error err)
152{ 174{
@@ -271,6 +293,8 @@ static void sc_kref_release(struct kref *kref)
271{ 293{
272 struct o2net_sock_container *sc = container_of(kref, 294 struct o2net_sock_container *sc = container_of(kref,
273 struct o2net_sock_container, sc_kref); 295 struct o2net_sock_container, sc_kref);
296 BUG_ON(timer_pending(&sc->sc_idle_timeout));
297
274 sclog(sc, "releasing\n"); 298 sclog(sc, "releasing\n");
275 299
276 if (sc->sc_sock) { 300 if (sc->sc_sock) {
@@ -300,7 +324,7 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node)
300 struct page *page = NULL; 324 struct page *page = NULL;
301 325
302 page = alloc_page(GFP_NOFS); 326 page = alloc_page(GFP_NOFS);
303 sc = kcalloc(1, sizeof(*sc), GFP_NOFS); 327 sc = kzalloc(sizeof(*sc), GFP_NOFS);
304 if (sc == NULL || page == NULL) 328 if (sc == NULL || page == NULL)
305 goto out; 329 goto out;
306 330
@@ -308,10 +332,10 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node)
308 o2nm_node_get(node); 332 o2nm_node_get(node);
309 sc->sc_node = node; 333 sc->sc_node = node;
310 334
311 INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed, sc); 335 INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed);
312 INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty, sc); 336 INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty);
313 INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc, sc); 337 INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc);
314 INIT_WORK(&sc->sc_keepalive_work, o2net_sc_send_keep_req, sc); 338 INIT_DELAYED_WORK(&sc->sc_keepalive_work, o2net_sc_send_keep_req);
315 339
316 init_timer(&sc->sc_idle_timeout); 340 init_timer(&sc->sc_idle_timeout);
317 sc->sc_idle_timeout.function = o2net_idle_timer; 341 sc->sc_idle_timeout.function = o2net_idle_timer;
@@ -342,7 +366,7 @@ static void o2net_sc_queue_work(struct o2net_sock_container *sc,
342 sc_put(sc); 366 sc_put(sc);
343} 367}
344static void o2net_sc_queue_delayed_work(struct o2net_sock_container *sc, 368static void o2net_sc_queue_delayed_work(struct o2net_sock_container *sc,
345 struct work_struct *work, 369 struct delayed_work *work,
346 int delay) 370 int delay)
347{ 371{
348 sc_get(sc); 372 sc_get(sc);
@@ -350,12 +374,19 @@ static void o2net_sc_queue_delayed_work(struct o2net_sock_container *sc,
350 sc_put(sc); 374 sc_put(sc);
351} 375}
352static void o2net_sc_cancel_delayed_work(struct o2net_sock_container *sc, 376static void o2net_sc_cancel_delayed_work(struct o2net_sock_container *sc,
353 struct work_struct *work) 377 struct delayed_work *work)
354{ 378{
355 if (cancel_delayed_work(work)) 379 if (cancel_delayed_work(work))
356 sc_put(sc); 380 sc_put(sc);
357} 381}
358 382
383static atomic_t o2net_connected_peers = ATOMIC_INIT(0);
384
385int o2net_num_connected_peers(void)
386{
387 return atomic_read(&o2net_connected_peers);
388}
389
359static void o2net_set_nn_state(struct o2net_node *nn, 390static void o2net_set_nn_state(struct o2net_node *nn,
360 struct o2net_sock_container *sc, 391 struct o2net_sock_container *sc,
361 unsigned valid, int err) 392 unsigned valid, int err)
@@ -366,6 +397,11 @@ static void o2net_set_nn_state(struct o2net_node *nn,
366 397
367 assert_spin_locked(&nn->nn_lock); 398 assert_spin_locked(&nn->nn_lock);
368 399
400 if (old_sc && !sc)
401 atomic_dec(&o2net_connected_peers);
402 else if (!old_sc && sc)
403 atomic_inc(&o2net_connected_peers);
404
369 /* the node num comparison and single connect/accept path should stop 405 /* the node num comparison and single connect/accept path should stop
370 * an non-null sc from being overwritten with another */ 406 * an non-null sc from being overwritten with another */
371 BUG_ON(sc && nn->nn_sc && nn->nn_sc != sc); 407 BUG_ON(sc && nn->nn_sc && nn->nn_sc != sc);
@@ -424,9 +460,9 @@ static void o2net_set_nn_state(struct o2net_node *nn,
424 /* delay if we're withing a RECONNECT_DELAY of the 460 /* delay if we're withing a RECONNECT_DELAY of the
425 * last attempt */ 461 * last attempt */
426 delay = (nn->nn_last_connect_attempt + 462 delay = (nn->nn_last_connect_attempt +
427 msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS)) 463 msecs_to_jiffies(o2net_reconnect_delay(sc->sc_node)))
428 - jiffies; 464 - jiffies;
429 if (delay > msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS)) 465 if (delay > msecs_to_jiffies(o2net_reconnect_delay(sc->sc_node)))
430 delay = 0; 466 delay = 0;
431 mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay); 467 mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay);
432 queue_delayed_work(o2net_wq, &nn->nn_connect_work, delay); 468 queue_delayed_work(o2net_wq, &nn->nn_connect_work, delay);
@@ -564,9 +600,11 @@ static void o2net_ensure_shutdown(struct o2net_node *nn,
564 * ourselves as state_change couldn't get the nn_lock and call set_nn_state 600 * ourselves as state_change couldn't get the nn_lock and call set_nn_state
565 * itself. 601 * itself.
566 */ 602 */
567static void o2net_shutdown_sc(void *arg) 603static void o2net_shutdown_sc(struct work_struct *work)
568{ 604{
569 struct o2net_sock_container *sc = arg; 605 struct o2net_sock_container *sc =
606 container_of(work, struct o2net_sock_container,
607 sc_shutdown_work);
570 struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); 608 struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
571 609
572 sclog(sc, "shutting down\n"); 610 sclog(sc, "shutting down\n");
@@ -676,7 +714,7 @@ int o2net_register_handler(u32 msg_type, u32 key, u32 max_len,
676 goto out; 714 goto out;
677 } 715 }
678 716
679 nmh = kcalloc(1, sizeof(struct o2net_msg_handler), GFP_NOFS); 717 nmh = kzalloc(sizeof(struct o2net_msg_handler), GFP_NOFS);
680 if (nmh == NULL) { 718 if (nmh == NULL) {
681 ret = -ENOMEM; 719 ret = -ENOMEM;
682 goto out; 720 goto out;
@@ -1097,13 +1135,51 @@ static int o2net_check_handshake(struct o2net_sock_container *sc)
1097 return -1; 1135 return -1;
1098 } 1136 }
1099 1137
1138 /*
1139 * Ensure timeouts are consistent with other nodes, otherwise
1140 * we can end up with one node thinking that the other must be down,
1141 * but isn't. This can ultimately cause corruption.
1142 */
1143 if (be32_to_cpu(hand->o2net_idle_timeout_ms) !=
1144 o2net_idle_timeout(sc->sc_node)) {
1145 mlog(ML_NOTICE, SC_NODEF_FMT " uses a network idle timeout of "
1146 "%u ms, but we use %u ms locally. disconnecting\n",
1147 SC_NODEF_ARGS(sc),
1148 be32_to_cpu(hand->o2net_idle_timeout_ms),
1149 o2net_idle_timeout(sc->sc_node));
1150 o2net_ensure_shutdown(nn, sc, -ENOTCONN);
1151 return -1;
1152 }
1153
1154 if (be32_to_cpu(hand->o2net_keepalive_delay_ms) !=
1155 o2net_keepalive_delay(sc->sc_node)) {
1156 mlog(ML_NOTICE, SC_NODEF_FMT " uses a keepalive delay of "
1157 "%u ms, but we use %u ms locally. disconnecting\n",
1158 SC_NODEF_ARGS(sc),
1159 be32_to_cpu(hand->o2net_keepalive_delay_ms),
1160 o2net_keepalive_delay(sc->sc_node));
1161 o2net_ensure_shutdown(nn, sc, -ENOTCONN);
1162 return -1;
1163 }
1164
1165 if (be32_to_cpu(hand->o2hb_heartbeat_timeout_ms) !=
1166 O2HB_MAX_WRITE_TIMEOUT_MS) {
1167 mlog(ML_NOTICE, SC_NODEF_FMT " uses a heartbeat timeout of "
1168 "%u ms, but we use %u ms locally. disconnecting\n",
1169 SC_NODEF_ARGS(sc),
1170 be32_to_cpu(hand->o2hb_heartbeat_timeout_ms),
1171 O2HB_MAX_WRITE_TIMEOUT_MS);
1172 o2net_ensure_shutdown(nn, sc, -ENOTCONN);
1173 return -1;
1174 }
1175
1100 sc->sc_handshake_ok = 1; 1176 sc->sc_handshake_ok = 1;
1101 1177
1102 spin_lock(&nn->nn_lock); 1178 spin_lock(&nn->nn_lock);
1103 /* set valid and queue the idle timers only if it hasn't been 1179 /* set valid and queue the idle timers only if it hasn't been
1104 * shut down already */ 1180 * shut down already */
1105 if (nn->nn_sc == sc) { 1181 if (nn->nn_sc == sc) {
1106 o2net_sc_postpone_idle(sc); 1182 o2net_sc_reset_idle_timer(sc);
1107 o2net_set_nn_state(nn, sc, 1, 0); 1183 o2net_set_nn_state(nn, sc, 1, 0);
1108 } 1184 }
1109 spin_unlock(&nn->nn_lock); 1185 spin_unlock(&nn->nn_lock);
@@ -1129,6 +1205,23 @@ static int o2net_advance_rx(struct o2net_sock_container *sc)
1129 sclog(sc, "receiving\n"); 1205 sclog(sc, "receiving\n");
1130 do_gettimeofday(&sc->sc_tv_advance_start); 1206 do_gettimeofday(&sc->sc_tv_advance_start);
1131 1207
1208 if (unlikely(sc->sc_handshake_ok == 0)) {
1209 if(sc->sc_page_off < sizeof(struct o2net_handshake)) {
1210 data = page_address(sc->sc_page) + sc->sc_page_off;
1211 datalen = sizeof(struct o2net_handshake) - sc->sc_page_off;
1212 ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen);
1213 if (ret > 0)
1214 sc->sc_page_off += ret;
1215 }
1216
1217 if (sc->sc_page_off == sizeof(struct o2net_handshake)) {
1218 o2net_check_handshake(sc);
1219 if (unlikely(sc->sc_handshake_ok == 0))
1220 ret = -EPROTO;
1221 }
1222 goto out;
1223 }
1224
1132 /* do we need more header? */ 1225 /* do we need more header? */
1133 if (sc->sc_page_off < sizeof(struct o2net_msg)) { 1226 if (sc->sc_page_off < sizeof(struct o2net_msg)) {
1134 data = page_address(sc->sc_page) + sc->sc_page_off; 1227 data = page_address(sc->sc_page) + sc->sc_page_off;
@@ -1136,15 +1229,6 @@ static int o2net_advance_rx(struct o2net_sock_container *sc)
1136 ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen); 1229 ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen);
1137 if (ret > 0) { 1230 if (ret > 0) {
1138 sc->sc_page_off += ret; 1231 sc->sc_page_off += ret;
1139
1140 /* this working relies on the handshake being
1141 * smaller than the normal message header */
1142 if (sc->sc_page_off >= sizeof(struct o2net_handshake)&&
1143 !sc->sc_handshake_ok && o2net_check_handshake(sc)) {
1144 ret = -EPROTO;
1145 goto out;
1146 }
1147
1148 /* only swab incoming here.. we can 1232 /* only swab incoming here.. we can
1149 * only get here once as we cross from 1233 * only get here once as we cross from
1150 * being under to over */ 1234 * being under to over */
@@ -1201,9 +1285,10 @@ out:
1201/* this work func is triggerd by data ready. it reads until it can read no 1285/* this work func is triggerd by data ready. it reads until it can read no
1202 * more. it interprets 0, eof, as fatal. if data_ready hits while we're doing 1286 * more. it interprets 0, eof, as fatal. if data_ready hits while we're doing
1203 * our work the work struct will be marked and we'll be called again. */ 1287 * our work the work struct will be marked and we'll be called again. */
1204static void o2net_rx_until_empty(void *arg) 1288static void o2net_rx_until_empty(struct work_struct *work)
1205{ 1289{
1206 struct o2net_sock_container *sc = arg; 1290 struct o2net_sock_container *sc =
1291 container_of(work, struct o2net_sock_container, sc_rx_work);
1207 int ret; 1292 int ret;
1208 1293
1209 do { 1294 do {
@@ -1245,26 +1330,43 @@ static int o2net_set_nodelay(struct socket *sock)
1245 return ret; 1330 return ret;
1246} 1331}
1247 1332
1333static void o2net_initialize_handshake(void)
1334{
1335 o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
1336 O2HB_MAX_WRITE_TIMEOUT_MS);
1337 o2net_hand->o2net_idle_timeout_ms = cpu_to_be32(
1338 o2net_idle_timeout(NULL));
1339 o2net_hand->o2net_keepalive_delay_ms = cpu_to_be32(
1340 o2net_keepalive_delay(NULL));
1341 o2net_hand->o2net_reconnect_delay_ms = cpu_to_be32(
1342 o2net_reconnect_delay(NULL));
1343}
1344
1248/* ------------------------------------------------------------ */ 1345/* ------------------------------------------------------------ */
1249 1346
1250/* called when a connect completes and after a sock is accepted. the 1347/* called when a connect completes and after a sock is accepted. the
1251 * rx path will see the response and mark the sc valid */ 1348 * rx path will see the response and mark the sc valid */
1252static void o2net_sc_connect_completed(void *arg) 1349static void o2net_sc_connect_completed(struct work_struct *work)
1253{ 1350{
1254 struct o2net_sock_container *sc = arg; 1351 struct o2net_sock_container *sc =
1352 container_of(work, struct o2net_sock_container,
1353 sc_connect_work);
1255 1354
1256 mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n", 1355 mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n",
1257 (unsigned long long)O2NET_PROTOCOL_VERSION, 1356 (unsigned long long)O2NET_PROTOCOL_VERSION,
1258 (unsigned long long)be64_to_cpu(o2net_hand->connector_id)); 1357 (unsigned long long)be64_to_cpu(o2net_hand->connector_id));
1259 1358
1359 o2net_initialize_handshake();
1260 o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand)); 1360 o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand));
1261 sc_put(sc); 1361 sc_put(sc);
1262} 1362}
1263 1363
1264/* this is called as a work_struct func. */ 1364/* this is called as a work_struct func. */
1265static void o2net_sc_send_keep_req(void *arg) 1365static void o2net_sc_send_keep_req(struct work_struct *work)
1266{ 1366{
1267 struct o2net_sock_container *sc = arg; 1367 struct o2net_sock_container *sc =
1368 container_of(work, struct o2net_sock_container,
1369 sc_keepalive_work.work);
1268 1370
1269 o2net_sendpage(sc, o2net_keep_req, sizeof(*o2net_keep_req)); 1371 o2net_sendpage(sc, o2net_keep_req, sizeof(*o2net_keep_req));
1270 sc_put(sc); 1372 sc_put(sc);
@@ -1280,8 +1382,10 @@ static void o2net_idle_timer(unsigned long data)
1280 1382
1281 do_gettimeofday(&now); 1383 do_gettimeofday(&now);
1282 1384
1283 printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT " has been idle for 10 " 1385 printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u "
1284 "seconds, shutting it down.\n", SC_NODEF_ARGS(sc)); 1386 "seconds, shutting it down.\n", SC_NODEF_ARGS(sc),
1387 o2net_idle_timeout(sc->sc_node) / 1000,
1388 o2net_idle_timeout(sc->sc_node) % 1000);
1285 mlog(ML_NOTICE, "here are some times that might help debug the " 1389 mlog(ML_NOTICE, "here are some times that might help debug the "
1286 "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv " 1390 "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv "
1287 "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n", 1391 "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n",
@@ -1299,14 +1403,21 @@ static void o2net_idle_timer(unsigned long data)
1299 o2net_sc_queue_work(sc, &sc->sc_shutdown_work); 1403 o2net_sc_queue_work(sc, &sc->sc_shutdown_work);
1300} 1404}
1301 1405
1302static void o2net_sc_postpone_idle(struct o2net_sock_container *sc) 1406static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc)
1303{ 1407{
1304 o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work); 1408 o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
1305 o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work, 1409 o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work,
1306 O2NET_KEEPALIVE_DELAY_SECS * HZ); 1410 msecs_to_jiffies(o2net_keepalive_delay(sc->sc_node)));
1307 do_gettimeofday(&sc->sc_tv_timer); 1411 do_gettimeofday(&sc->sc_tv_timer);
1308 mod_timer(&sc->sc_idle_timeout, 1412 mod_timer(&sc->sc_idle_timeout,
1309 jiffies + (O2NET_IDLE_TIMEOUT_SECS * HZ)); 1413 jiffies + msecs_to_jiffies(o2net_idle_timeout(sc->sc_node)));
1414}
1415
1416static void o2net_sc_postpone_idle(struct o2net_sock_container *sc)
1417{
1418 /* Only push out an existing timer */
1419 if (timer_pending(&sc->sc_idle_timeout))
1420 o2net_sc_reset_idle_timer(sc);
1310} 1421}
1311 1422
1312/* this work func is kicked whenever a path sets the nn state which doesn't 1423/* this work func is kicked whenever a path sets the nn state which doesn't
@@ -1314,14 +1425,15 @@ static void o2net_sc_postpone_idle(struct o2net_sock_container *sc)
1314 * having a connect attempt fail, etc. This centralizes the logic which decides 1425 * having a connect attempt fail, etc. This centralizes the logic which decides
1315 * if a connect attempt should be made or if we should give up and all future 1426 * if a connect attempt should be made or if we should give up and all future
1316 * transmit attempts should fail */ 1427 * transmit attempts should fail */
1317static void o2net_start_connect(void *arg) 1428static void o2net_start_connect(struct work_struct *work)
1318{ 1429{
1319 struct o2net_node *nn = arg; 1430 struct o2net_node *nn =
1431 container_of(work, struct o2net_node, nn_connect_work.work);
1320 struct o2net_sock_container *sc = NULL; 1432 struct o2net_sock_container *sc = NULL;
1321 struct o2nm_node *node = NULL, *mynode = NULL; 1433 struct o2nm_node *node = NULL, *mynode = NULL;
1322 struct socket *sock = NULL; 1434 struct socket *sock = NULL;
1323 struct sockaddr_in myaddr = {0, }, remoteaddr = {0, }; 1435 struct sockaddr_in myaddr = {0, }, remoteaddr = {0, };
1324 int ret = 0; 1436 int ret = 0, stop;
1325 1437
1326 /* if we're greater we initiate tx, otherwise we accept */ 1438 /* if we're greater we initiate tx, otherwise we accept */
1327 if (o2nm_this_node() <= o2net_num_from_nn(nn)) 1439 if (o2nm_this_node() <= o2net_num_from_nn(nn))
@@ -1342,10 +1454,9 @@ static void o2net_start_connect(void *arg)
1342 1454
1343 spin_lock(&nn->nn_lock); 1455 spin_lock(&nn->nn_lock);
1344 /* see if we already have one pending or have given up */ 1456 /* see if we already have one pending or have given up */
1345 if (nn->nn_sc || nn->nn_persistent_error) 1457 stop = (nn->nn_sc || nn->nn_persistent_error);
1346 arg = NULL;
1347 spin_unlock(&nn->nn_lock); 1458 spin_unlock(&nn->nn_lock);
1348 if (arg == NULL) /* *shrug*, needed some indicator */ 1459 if (stop)
1349 goto out; 1460 goto out;
1350 1461
1351 nn->nn_last_connect_attempt = jiffies; 1462 nn->nn_last_connect_attempt = jiffies;
@@ -1421,24 +1532,29 @@ out:
1421 return; 1532 return;
1422} 1533}
1423 1534
1424static void o2net_connect_expired(void *arg) 1535static void o2net_connect_expired(struct work_struct *work)
1425{ 1536{
1426 struct o2net_node *nn = arg; 1537 struct o2net_node *nn =
1538 container_of(work, struct o2net_node, nn_connect_expired.work);
1427 1539
1428 spin_lock(&nn->nn_lock); 1540 spin_lock(&nn->nn_lock);
1429 if (!nn->nn_sc_valid) { 1541 if (!nn->nn_sc_valid) {
1542 struct o2nm_node *node = nn->nn_sc->sc_node;
1430 mlog(ML_ERROR, "no connection established with node %u after " 1543 mlog(ML_ERROR, "no connection established with node %u after "
1431 "%u seconds, giving up and returning errors.\n", 1544 "%u.%u seconds, giving up and returning errors.\n",
1432 o2net_num_from_nn(nn), O2NET_IDLE_TIMEOUT_SECS); 1545 o2net_num_from_nn(nn),
1546 o2net_idle_timeout(node) / 1000,
1547 o2net_idle_timeout(node) % 1000);
1433 1548
1434 o2net_set_nn_state(nn, NULL, 0, -ENOTCONN); 1549 o2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
1435 } 1550 }
1436 spin_unlock(&nn->nn_lock); 1551 spin_unlock(&nn->nn_lock);
1437} 1552}
1438 1553
1439static void o2net_still_up(void *arg) 1554static void o2net_still_up(struct work_struct *work)
1440{ 1555{
1441 struct o2net_node *nn = arg; 1556 struct o2net_node *nn =
1557 container_of(work, struct o2net_node, nn_still_up.work);
1442 1558
1443 o2quo_hb_still_up(o2net_num_from_nn(nn)); 1559 o2quo_hb_still_up(o2net_num_from_nn(nn));
1444} 1560}
@@ -1469,6 +1585,8 @@ static void o2net_hb_node_down_cb(struct o2nm_node *node, int node_num,
1469 1585
1470 if (node_num != o2nm_this_node()) 1586 if (node_num != o2nm_this_node())
1471 o2net_disconnect_node(node); 1587 o2net_disconnect_node(node);
1588
1589 BUG_ON(atomic_read(&o2net_connected_peers) < 0);
1472} 1590}
1473 1591
1474static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num, 1592static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num,
@@ -1480,14 +1598,14 @@ static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num,
1480 1598
1481 /* ensure an immediate connect attempt */ 1599 /* ensure an immediate connect attempt */
1482 nn->nn_last_connect_attempt = jiffies - 1600 nn->nn_last_connect_attempt = jiffies -
1483 (msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS) + 1); 1601 (msecs_to_jiffies(o2net_reconnect_delay(node)) + 1);
1484 1602
1485 if (node_num != o2nm_this_node()) { 1603 if (node_num != o2nm_this_node()) {
1486 /* heartbeat doesn't work unless a local node number is 1604 /* heartbeat doesn't work unless a local node number is
1487 * configured and doing so brings up the o2net_wq, so we can 1605 * configured and doing so brings up the o2net_wq, so we can
1488 * use it.. */ 1606 * use it.. */
1489 queue_delayed_work(o2net_wq, &nn->nn_connect_expired, 1607 queue_delayed_work(o2net_wq, &nn->nn_connect_expired,
1490 O2NET_IDLE_TIMEOUT_SECS * HZ); 1608 msecs_to_jiffies(o2net_idle_timeout(node)));
1491 1609
1492 /* believe it or not, accept and node hearbeating testing 1610 /* believe it or not, accept and node hearbeating testing
1493 * can succeed for this node before we got here.. so 1611 * can succeed for this node before we got here.. so
@@ -1632,6 +1750,7 @@ static int o2net_accept_one(struct socket *sock)
1632 o2net_register_callbacks(sc->sc_sock->sk, sc); 1750 o2net_register_callbacks(sc->sc_sock->sk, sc);
1633 o2net_sc_queue_work(sc, &sc->sc_rx_work); 1751 o2net_sc_queue_work(sc, &sc->sc_rx_work);
1634 1752
1753 o2net_initialize_handshake();
1635 o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand)); 1754 o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand));
1636 1755
1637out: 1756out:
@@ -1644,9 +1763,9 @@ out:
1644 return ret; 1763 return ret;
1645} 1764}
1646 1765
1647static void o2net_accept_many(void *arg) 1766static void o2net_accept_many(struct work_struct *work)
1648{ 1767{
1649 struct socket *sock = arg; 1768 struct socket *sock = o2net_listen_sock;
1650 while (o2net_accept_one(sock) == 0) 1769 while (o2net_accept_one(sock) == 0)
1651 cond_resched(); 1770 cond_resched();
1652} 1771}
@@ -1700,7 +1819,7 @@ static int o2net_open_listening_sock(__be16 port)
1700 write_unlock_bh(&sock->sk->sk_callback_lock); 1819 write_unlock_bh(&sock->sk->sk_callback_lock);
1701 1820
1702 o2net_listen_sock = sock; 1821 o2net_listen_sock = sock;
1703 INIT_WORK(&o2net_listen_work, o2net_accept_many, sock); 1822 INIT_WORK(&o2net_listen_work, o2net_accept_many);
1704 1823
1705 sock->sk->sk_reuse = 1; 1824 sock->sk->sk_reuse = 1;
1706 ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin)); 1825 ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
@@ -1799,9 +1918,9 @@ int o2net_init(void)
1799 1918
1800 o2quo_init(); 1919 o2quo_init();
1801 1920
1802 o2net_hand = kcalloc(1, sizeof(struct o2net_handshake), GFP_KERNEL); 1921 o2net_hand = kzalloc(sizeof(struct o2net_handshake), GFP_KERNEL);
1803 o2net_keep_req = kcalloc(1, sizeof(struct o2net_msg), GFP_KERNEL); 1922 o2net_keep_req = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL);
1804 o2net_keep_resp = kcalloc(1, sizeof(struct o2net_msg), GFP_KERNEL); 1923 o2net_keep_resp = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL);
1805 if (!o2net_hand || !o2net_keep_req || !o2net_keep_resp) { 1924 if (!o2net_hand || !o2net_keep_req || !o2net_keep_resp) {
1806 kfree(o2net_hand); 1925 kfree(o2net_hand);
1807 kfree(o2net_keep_req); 1926 kfree(o2net_keep_req);
@@ -1819,9 +1938,10 @@ int o2net_init(void)
1819 struct o2net_node *nn = o2net_nn_from_num(i); 1938 struct o2net_node *nn = o2net_nn_from_num(i);
1820 1939
1821 spin_lock_init(&nn->nn_lock); 1940 spin_lock_init(&nn->nn_lock);
1822 INIT_WORK(&nn->nn_connect_work, o2net_start_connect, nn); 1941 INIT_DELAYED_WORK(&nn->nn_connect_work, o2net_start_connect);
1823 INIT_WORK(&nn->nn_connect_expired, o2net_connect_expired, nn); 1942 INIT_DELAYED_WORK(&nn->nn_connect_expired,
1824 INIT_WORK(&nn->nn_still_up, o2net_still_up, nn); 1943 o2net_connect_expired);
1944 INIT_DELAYED_WORK(&nn->nn_still_up, o2net_still_up);
1825 /* until we see hb from a node we'll return einval */ 1945 /* until we see hb from a node we'll return einval */
1826 nn->nn_persistent_error = -ENOTCONN; 1946 nn->nn_persistent_error = -ENOTCONN;
1827 init_waitqueue_head(&nn->nn_sc_wq); 1947 init_waitqueue_head(&nn->nn_sc_wq);
diff --git a/fs/ocfs2/cluster/tcp.h b/fs/ocfs2/cluster/tcp.h
index 616ff2b8434a..21a4e43df836 100644
--- a/fs/ocfs2/cluster/tcp.h
+++ b/fs/ocfs2/cluster/tcp.h
@@ -54,6 +54,13 @@ typedef int (o2net_msg_handler_func)(struct o2net_msg *msg, u32 len, void *data)
54 54
55#define O2NET_MAX_PAYLOAD_BYTES (4096 - sizeof(struct o2net_msg)) 55#define O2NET_MAX_PAYLOAD_BYTES (4096 - sizeof(struct o2net_msg))
56 56
57/* same as hb delay, we're waiting for another node to recognize our hb */
58#define O2NET_RECONNECT_DELAY_MS_DEFAULT 2000
59
60#define O2NET_KEEPALIVE_DELAY_MS_DEFAULT 5000
61#define O2NET_IDLE_TIMEOUT_MS_DEFAULT 10000
62
63
57/* TODO: figure this out.... */ 64/* TODO: figure this out.... */
58static inline int o2net_link_down(int err, struct socket *sock) 65static inline int o2net_link_down(int err, struct socket *sock)
59{ 66{
@@ -101,6 +108,7 @@ void o2net_unregister_hb_callbacks(void);
101int o2net_start_listening(struct o2nm_node *node); 108int o2net_start_listening(struct o2nm_node *node);
102void o2net_stop_listening(struct o2nm_node *node); 109void o2net_stop_listening(struct o2nm_node *node);
103void o2net_disconnect_node(struct o2nm_node *node); 110void o2net_disconnect_node(struct o2nm_node *node);
111int o2net_num_connected_peers(void);
104 112
105int o2net_init(void); 113int o2net_init(void);
106void o2net_exit(void); 114void o2net_exit(void);
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index 4b46aac7d243..b700dc9624d1 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -27,23 +27,20 @@
27#define O2NET_MSG_KEEP_REQ_MAGIC ((u16)0xfa57) 27#define O2NET_MSG_KEEP_REQ_MAGIC ((u16)0xfa57)
28#define O2NET_MSG_KEEP_RESP_MAGIC ((u16)0xfa58) 28#define O2NET_MSG_KEEP_RESP_MAGIC ((u16)0xfa58)
29 29
30/* same as hb delay, we're waiting for another node to recognize our hb */
31#define O2NET_RECONNECT_DELAY_MS O2HB_REGION_TIMEOUT_MS
32
33/* we're delaying our quorum decision so that heartbeat will have timed 30/* we're delaying our quorum decision so that heartbeat will have timed
34 * out truly dead nodes by the time we come around to making decisions 31 * out truly dead nodes by the time we come around to making decisions
35 * on their number */ 32 * on their number */
36#define O2NET_QUORUM_DELAY_MS ((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS) 33#define O2NET_QUORUM_DELAY_MS ((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS)
37 34
38#define O2NET_KEEPALIVE_DELAY_SECS 5
39#define O2NET_IDLE_TIMEOUT_SECS 10
40
41/* 35/*
42 * This version number represents quite a lot, unfortunately. It not 36 * This version number represents quite a lot, unfortunately. It not
43 * only represents the raw network message protocol on the wire but also 37 * only represents the raw network message protocol on the wire but also
44 * locking semantics of the file system using the protocol. It should 38 * locking semantics of the file system using the protocol. It should
45 * be somewhere else, I'm sure, but right now it isn't. 39 * be somewhere else, I'm sure, but right now it isn't.
46 * 40 *
41 * New in version 5:
42 * - Network timeout checking protocol
43 *
47 * New in version 4: 44 * New in version 4:
48 * - Remove i_generation from lock names for better stat performance. 45 * - Remove i_generation from lock names for better stat performance.
49 * 46 *
@@ -54,10 +51,14 @@
54 * - full 64 bit i_size in the metadata lock lvbs 51 * - full 64 bit i_size in the metadata lock lvbs
55 * - introduction of "rw" lock and pushing meta/data locking down 52 * - introduction of "rw" lock and pushing meta/data locking down
56 */ 53 */
57#define O2NET_PROTOCOL_VERSION 4ULL 54#define O2NET_PROTOCOL_VERSION 5ULL
58struct o2net_handshake { 55struct o2net_handshake {
59 __be64 protocol_version; 56 __be64 protocol_version;
60 __be64 connector_id; 57 __be64 connector_id;
58 __be32 o2hb_heartbeat_timeout_ms;
59 __be32 o2net_idle_timeout_ms;
60 __be32 o2net_keepalive_delay_ms;
61 __be32 o2net_reconnect_delay_ms;
61}; 62};
62 63
63struct o2net_node { 64struct o2net_node {
@@ -86,18 +87,18 @@ struct o2net_node {
86 * connect attempt fails and so can be self-arming. shutdown is 87 * connect attempt fails and so can be self-arming. shutdown is
87 * careful to first mark the nn such that no connects will be attempted 88 * careful to first mark the nn such that no connects will be attempted
88 * before canceling delayed connect work and flushing the queue. */ 89 * before canceling delayed connect work and flushing the queue. */
89 struct work_struct nn_connect_work; 90 struct delayed_work nn_connect_work;
90 unsigned long nn_last_connect_attempt; 91 unsigned long nn_last_connect_attempt;
91 92
92 /* this is queued as nodes come up and is canceled when a connection is 93 /* this is queued as nodes come up and is canceled when a connection is
93 * established. this expiring gives up on the node and errors out 94 * established. this expiring gives up on the node and errors out
94 * transmits */ 95 * transmits */
95 struct work_struct nn_connect_expired; 96 struct delayed_work nn_connect_expired;
96 97
97 /* after we give up on a socket we wait a while before deciding 98 /* after we give up on a socket we wait a while before deciding
98 * that it is still heartbeating and that we should do some 99 * that it is still heartbeating and that we should do some
99 * quorum work */ 100 * quorum work */
100 struct work_struct nn_still_up; 101 struct delayed_work nn_still_up;
101}; 102};
102 103
103struct o2net_sock_container { 104struct o2net_sock_container {
@@ -129,7 +130,7 @@ struct o2net_sock_container {
129 struct work_struct sc_shutdown_work; 130 struct work_struct sc_shutdown_work;
130 131
131 struct timer_list sc_idle_timeout; 132 struct timer_list sc_idle_timeout;
132 struct work_struct sc_keepalive_work; 133 struct delayed_work sc_keepalive_work;
133 134
134 unsigned sc_handshake_ok:1; 135 unsigned sc_handshake_ok:1;
135 136
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index 04e01915b86e..66821e178167 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -79,9 +79,10 @@ int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir)
79 struct buffer_head * bh, * tmp; 79 struct buffer_head * bh, * tmp;
80 struct ocfs2_dir_entry * de; 80 struct ocfs2_dir_entry * de;
81 int err; 81 int err;
82 struct inode *inode = filp->f_dentry->d_inode; 82 struct inode *inode = filp->f_path.dentry->d_inode;
83 struct super_block * sb = inode->i_sb; 83 struct super_block * sb = inode->i_sb;
84 unsigned int ra_sectors = 16; 84 unsigned int ra_sectors = 16;
85 int lock_level = 0;
85 86
86 mlog_entry("dirino=%llu\n", 87 mlog_entry("dirino=%llu\n",
87 (unsigned long long)OCFS2_I(inode)->ip_blkno); 88 (unsigned long long)OCFS2_I(inode)->ip_blkno);
@@ -89,7 +90,15 @@ int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir)
89 stored = 0; 90 stored = 0;
90 bh = NULL; 91 bh = NULL;
91 92
92 error = ocfs2_meta_lock(inode, NULL, NULL, 0); 93 error = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level);
94 if (lock_level && error >= 0) {
95 /* We release EX lock which used to update atime
96 * and get PR lock again to reduce contention
97 * on commonly accessed directories. */
98 ocfs2_meta_unlock(inode, 1);
99 lock_level = 0;
100 error = ocfs2_meta_lock(inode, NULL, 0);
101 }
93 if (error < 0) { 102 if (error < 0) {
94 if (error != -ENOENT) 103 if (error != -ENOENT)
95 mlog_errno(error); 104 mlog_errno(error);
@@ -198,7 +207,7 @@ revalidate:
198 207
199 stored = 0; 208 stored = 0;
200bail: 209bail:
201 ocfs2_meta_unlock(inode, 0); 210 ocfs2_meta_unlock(inode, lock_level);
202 211
203bail_nolock: 212bail_nolock:
204 mlog_exit(stored); 213 mlog_exit(stored);
@@ -340,7 +349,7 @@ int ocfs2_empty_dir(struct inode *inode)
340 349
341/* returns a bh of the 1st new block in the allocation. */ 350/* returns a bh of the 1st new block in the allocation. */
342int ocfs2_do_extend_dir(struct super_block *sb, 351int ocfs2_do_extend_dir(struct super_block *sb,
343 struct ocfs2_journal_handle *handle, 352 handle_t *handle,
344 struct inode *dir, 353 struct inode *dir,
345 struct buffer_head *parent_fe_bh, 354 struct buffer_head *parent_fe_bh,
346 struct ocfs2_alloc_context *data_ac, 355 struct ocfs2_alloc_context *data_ac,
@@ -398,7 +407,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
398 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) parent_fe_bh->b_data; 407 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
399 struct ocfs2_alloc_context *data_ac = NULL; 408 struct ocfs2_alloc_context *data_ac = NULL;
400 struct ocfs2_alloc_context *meta_ac = NULL; 409 struct ocfs2_alloc_context *meta_ac = NULL;
401 struct ocfs2_journal_handle *handle = NULL; 410 handle_t *handle = NULL;
402 struct buffer_head *new_bh = NULL; 411 struct buffer_head *new_bh = NULL;
403 struct ocfs2_dir_entry * de; 412 struct ocfs2_dir_entry * de;
404 struct super_block *sb = osb->sb; 413 struct super_block *sb = osb->sb;
@@ -409,13 +418,6 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
409 mlog(0, "extending dir %llu (i_size = %lld)\n", 418 mlog(0, "extending dir %llu (i_size = %lld)\n",
410 (unsigned long long)OCFS2_I(dir)->ip_blkno, dir_i_size); 419 (unsigned long long)OCFS2_I(dir)->ip_blkno, dir_i_size);
411 420
412 handle = ocfs2_alloc_handle(osb);
413 if (handle == NULL) {
414 status = -ENOMEM;
415 mlog_errno(status);
416 goto bail;
417 }
418
419 /* dir->i_size is always block aligned. */ 421 /* dir->i_size is always block aligned. */
420 spin_lock(&OCFS2_I(dir)->ip_lock); 422 spin_lock(&OCFS2_I(dir)->ip_lock);
421 if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) { 423 if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) {
@@ -428,8 +430,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
428 } 430 }
429 431
430 if (!num_free_extents) { 432 if (!num_free_extents) {
431 status = ocfs2_reserve_new_metadata(osb, handle, 433 status = ocfs2_reserve_new_metadata(osb, fe, &meta_ac);
432 fe, &meta_ac);
433 if (status < 0) { 434 if (status < 0) {
434 if (status != -ENOSPC) 435 if (status != -ENOSPC)
435 mlog_errno(status); 436 mlog_errno(status);
@@ -437,7 +438,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
437 } 438 }
438 } 439 }
439 440
440 status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac); 441 status = ocfs2_reserve_clusters(osb, 1, &data_ac);
441 if (status < 0) { 442 if (status < 0) {
442 if (status != -ENOSPC) 443 if (status != -ENOSPC)
443 mlog_errno(status); 444 mlog_errno(status);
@@ -450,7 +451,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
450 credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS; 451 credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS;
451 } 452 }
452 453
453 handle = ocfs2_start_trans(osb, handle, credits); 454 handle = ocfs2_start_trans(osb, credits);
454 if (IS_ERR(handle)) { 455 if (IS_ERR(handle)) {
455 status = PTR_ERR(handle); 456 status = PTR_ERR(handle);
456 handle = NULL; 457 handle = NULL;
@@ -496,7 +497,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
496 get_bh(*new_de_bh); 497 get_bh(*new_de_bh);
497bail: 498bail:
498 if (handle) 499 if (handle)
499 ocfs2_commit_trans(handle); 500 ocfs2_commit_trans(osb, handle);
500 501
501 if (data_ac) 502 if (data_ac)
502 ocfs2_free_alloc_context(data_ac); 503 ocfs2_free_alloc_context(data_ac);
diff --git a/fs/ocfs2/dir.h b/fs/ocfs2/dir.h
index 5f614ec9649c..3f67e146864a 100644
--- a/fs/ocfs2/dir.h
+++ b/fs/ocfs2/dir.h
@@ -45,7 +45,7 @@ int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb,
45 struct buffer_head **ret_de_bh); 45 struct buffer_head **ret_de_bh);
46struct ocfs2_alloc_context; 46struct ocfs2_alloc_context;
47int ocfs2_do_extend_dir(struct super_block *sb, 47int ocfs2_do_extend_dir(struct super_block *sb,
48 struct ocfs2_journal_handle *handle, 48 handle_t *handle,
49 struct inode *dir, 49 struct inode *dir,
50 struct buffer_head *parent_fe_bh, 50 struct buffer_head *parent_fe_bh,
51 struct ocfs2_alloc_context *data_ac, 51 struct ocfs2_alloc_context *data_ac,
diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index fa968180b072..6b6ff76538c5 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -153,7 +153,7 @@ static inline struct hlist_head *dlm_lockres_hash(struct dlm_ctxt *dlm, unsigned
153 * called functions that cannot be directly called from the 153 * called functions that cannot be directly called from the
154 * net message handlers for some reason, usually because 154 * net message handlers for some reason, usually because
155 * they need to send net messages of their own. */ 155 * they need to send net messages of their own. */
156void dlm_dispatch_work(void *data); 156void dlm_dispatch_work(struct work_struct *work);
157 157
158struct dlm_lock_resource; 158struct dlm_lock_resource;
159struct dlm_work_item; 159struct dlm_work_item;
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 8d1065f8b3bd..f0b25f2dd205 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -68,7 +68,8 @@ static void **dlm_alloc_pagevec(int pages)
68 goto out_free; 68 goto out_free;
69 69
70 mlog(0, "Allocated DLM hash pagevec; %d pages (%lu expected), %lu buckets per page\n", 70 mlog(0, "Allocated DLM hash pagevec; %d pages (%lu expected), %lu buckets per page\n",
71 pages, DLM_HASH_PAGES, (unsigned long)DLM_BUCKETS_PER_PAGE); 71 pages, (unsigned long)DLM_HASH_PAGES,
72 (unsigned long)DLM_BUCKETS_PER_PAGE);
72 return vec; 73 return vec;
73out_free: 74out_free:
74 dlm_free_pagevec(vec, i); 75 dlm_free_pagevec(vec, i);
@@ -919,7 +920,7 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm)
919 920
920 mlog_entry("%p", dlm); 921 mlog_entry("%p", dlm);
921 922
922 ctxt = kcalloc(1, sizeof(*ctxt), GFP_KERNEL); 923 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
923 if (!ctxt) { 924 if (!ctxt) {
924 status = -ENOMEM; 925 status = -ENOMEM;
925 mlog_errno(status); 926 mlog_errno(status);
@@ -1222,7 +1223,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
1222 int i; 1223 int i;
1223 struct dlm_ctxt *dlm = NULL; 1224 struct dlm_ctxt *dlm = NULL;
1224 1225
1225 dlm = kcalloc(1, sizeof(*dlm), GFP_KERNEL); 1226 dlm = kzalloc(sizeof(*dlm), GFP_KERNEL);
1226 if (!dlm) { 1227 if (!dlm) {
1227 mlog_errno(-ENOMEM); 1228 mlog_errno(-ENOMEM);
1228 goto leave; 1229 goto leave;
@@ -1296,7 +1297,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
1296 1297
1297 spin_lock_init(&dlm->work_lock); 1298 spin_lock_init(&dlm->work_lock);
1298 INIT_LIST_HEAD(&dlm->work_list); 1299 INIT_LIST_HEAD(&dlm->work_list);
1299 INIT_WORK(&dlm->dispatched_work, dlm_dispatch_work, dlm); 1300 INIT_WORK(&dlm->dispatched_work, dlm_dispatch_work);
1300 1301
1301 kref_init(&dlm->dlm_refs); 1302 kref_init(&dlm->dlm_refs);
1302 dlm->dlm_state = DLM_CTXT_NEW; 1303 dlm->dlm_state = DLM_CTXT_NEW;
diff --git a/fs/ocfs2/dlm/dlmfs.c b/fs/ocfs2/dlm/dlmfs.c
index 16b8d1ba7066..b7f0ba97a1a2 100644
--- a/fs/ocfs2/dlm/dlmfs.c
+++ b/fs/ocfs2/dlm/dlmfs.c
@@ -66,7 +66,7 @@ static struct file_operations dlmfs_file_operations;
66static struct inode_operations dlmfs_dir_inode_operations; 66static struct inode_operations dlmfs_dir_inode_operations;
67static struct inode_operations dlmfs_root_inode_operations; 67static struct inode_operations dlmfs_root_inode_operations;
68static struct inode_operations dlmfs_file_inode_operations; 68static struct inode_operations dlmfs_file_inode_operations;
69static kmem_cache_t *dlmfs_inode_cache; 69static struct kmem_cache *dlmfs_inode_cache;
70 70
71struct workqueue_struct *user_dlm_worker; 71struct workqueue_struct *user_dlm_worker;
72 72
@@ -176,7 +176,7 @@ static ssize_t dlmfs_file_read(struct file *filp,
176 int bytes_left; 176 int bytes_left;
177 ssize_t readlen; 177 ssize_t readlen;
178 char *lvb_buf; 178 char *lvb_buf;
179 struct inode *inode = filp->f_dentry->d_inode; 179 struct inode *inode = filp->f_path.dentry->d_inode;
180 180
181 mlog(0, "inode %lu, count = %zu, *ppos = %llu\n", 181 mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
182 inode->i_ino, count, *ppos); 182 inode->i_ino, count, *ppos);
@@ -220,7 +220,7 @@ static ssize_t dlmfs_file_write(struct file *filp,
220 int bytes_left; 220 int bytes_left;
221 ssize_t writelen; 221 ssize_t writelen;
222 char *lvb_buf; 222 char *lvb_buf;
223 struct inode *inode = filp->f_dentry->d_inode; 223 struct inode *inode = filp->f_path.dentry->d_inode;
224 224
225 mlog(0, "inode %lu, count = %zu, *ppos = %llu\n", 225 mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
226 inode->i_ino, count, *ppos); 226 inode->i_ino, count, *ppos);
@@ -257,7 +257,7 @@ static ssize_t dlmfs_file_write(struct file *filp,
257} 257}
258 258
259static void dlmfs_init_once(void *foo, 259static void dlmfs_init_once(void *foo,
260 kmem_cache_t *cachep, 260 struct kmem_cache *cachep,
261 unsigned long flags) 261 unsigned long flags)
262{ 262{
263 struct dlmfs_inode_private *ip = 263 struct dlmfs_inode_private *ip =
@@ -276,7 +276,7 @@ static struct inode *dlmfs_alloc_inode(struct super_block *sb)
276{ 276{
277 struct dlmfs_inode_private *ip; 277 struct dlmfs_inode_private *ip;
278 278
279 ip = kmem_cache_alloc(dlmfs_inode_cache, SLAB_NOFS); 279 ip = kmem_cache_alloc(dlmfs_inode_cache, GFP_NOFS);
280 if (!ip) 280 if (!ip)
281 return NULL; 281 return NULL;
282 282
diff --git a/fs/ocfs2/dlm/dlmlock.c b/fs/ocfs2/dlm/dlmlock.c
index 42a1b91979b5..e5ca3db197f6 100644
--- a/fs/ocfs2/dlm/dlmlock.c
+++ b/fs/ocfs2/dlm/dlmlock.c
@@ -408,13 +408,13 @@ struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie,
408 struct dlm_lock *lock; 408 struct dlm_lock *lock;
409 int kernel_allocated = 0; 409 int kernel_allocated = 0;
410 410
411 lock = kcalloc(1, sizeof(*lock), GFP_NOFS); 411 lock = kzalloc(sizeof(*lock), GFP_NOFS);
412 if (!lock) 412 if (!lock)
413 return NULL; 413 return NULL;
414 414
415 if (!lksb) { 415 if (!lksb) {
416 /* zero memory only if kernel-allocated */ 416 /* zero memory only if kernel-allocated */
417 lksb = kcalloc(1, sizeof(*lksb), GFP_NOFS); 417 lksb = kzalloc(sizeof(*lksb), GFP_NOFS);
418 if (!lksb) { 418 if (!lksb) {
419 kfree(lock); 419 kfree(lock);
420 return NULL; 420 return NULL;
diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
index f784177b6241..0ad872055cb3 100644
--- a/fs/ocfs2/dlm/dlmmaster.c
+++ b/fs/ocfs2/dlm/dlmmaster.c
@@ -221,7 +221,7 @@ EXPORT_SYMBOL_GPL(dlm_dump_all_mles);
221#endif /* 0 */ 221#endif /* 0 */
222 222
223 223
224static kmem_cache_t *dlm_mle_cache = NULL; 224static struct kmem_cache *dlm_mle_cache = NULL;
225 225
226 226
227static void dlm_mle_release(struct kref *kref); 227static void dlm_mle_release(struct kref *kref);
@@ -1939,7 +1939,7 @@ int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
1939 int ignore_higher, u8 request_from, u32 flags) 1939 int ignore_higher, u8 request_from, u32 flags)
1940{ 1940{
1941 struct dlm_work_item *item; 1941 struct dlm_work_item *item;
1942 item = kcalloc(1, sizeof(*item), GFP_NOFS); 1942 item = kzalloc(sizeof(*item), GFP_NOFS);
1943 if (!item) 1943 if (!item)
1944 return -ENOMEM; 1944 return -ENOMEM;
1945 1945
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index 9d950d7cea38..367a11e9e2ed 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -153,9 +153,10 @@ static inline void dlm_reset_recovery(struct dlm_ctxt *dlm)
153} 153}
154 154
155/* Worker function used during recovery. */ 155/* Worker function used during recovery. */
156void dlm_dispatch_work(void *data) 156void dlm_dispatch_work(struct work_struct *work)
157{ 157{
158 struct dlm_ctxt *dlm = (struct dlm_ctxt *)data; 158 struct dlm_ctxt *dlm =
159 container_of(work, struct dlm_ctxt, dispatched_work);
159 LIST_HEAD(tmp_list); 160 LIST_HEAD(tmp_list);
160 struct list_head *iter, *iter2; 161 struct list_head *iter, *iter2;
161 struct dlm_work_item *item; 162 struct dlm_work_item *item;
@@ -756,7 +757,7 @@ static int dlm_init_recovery_area(struct dlm_ctxt *dlm, u8 dead_node)
756 } 757 }
757 BUG_ON(num == dead_node); 758 BUG_ON(num == dead_node);
758 759
759 ndata = kcalloc(1, sizeof(*ndata), GFP_NOFS); 760 ndata = kzalloc(sizeof(*ndata), GFP_NOFS);
760 if (!ndata) { 761 if (!ndata) {
761 dlm_destroy_recovery_area(dlm, dead_node); 762 dlm_destroy_recovery_area(dlm, dead_node);
762 return -ENOMEM; 763 return -ENOMEM;
@@ -841,7 +842,7 @@ int dlm_request_all_locks_handler(struct o2net_msg *msg, u32 len, void *data)
841 } 842 }
842 BUG_ON(lr->dead_node != dlm->reco.dead_node); 843 BUG_ON(lr->dead_node != dlm->reco.dead_node);
843 844
844 item = kcalloc(1, sizeof(*item), GFP_NOFS); 845 item = kzalloc(sizeof(*item), GFP_NOFS);
845 if (!item) { 846 if (!item) {
846 dlm_put(dlm); 847 dlm_put(dlm);
847 return -ENOMEM; 848 return -ENOMEM;
@@ -1322,7 +1323,7 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data)
1322 1323
1323 ret = -ENOMEM; 1324 ret = -ENOMEM;
1324 buf = kmalloc(be16_to_cpu(msg->data_len), GFP_NOFS); 1325 buf = kmalloc(be16_to_cpu(msg->data_len), GFP_NOFS);
1325 item = kcalloc(1, sizeof(*item), GFP_NOFS); 1326 item = kzalloc(sizeof(*item), GFP_NOFS);
1326 if (!buf || !item) 1327 if (!buf || !item)
1327 goto leave; 1328 goto leave;
1328 1329
diff --git a/fs/ocfs2/dlm/userdlm.c b/fs/ocfs2/dlm/userdlm.c
index eead48bbfac6..7d2f578b267d 100644
--- a/fs/ocfs2/dlm/userdlm.c
+++ b/fs/ocfs2/dlm/userdlm.c
@@ -171,15 +171,14 @@ static inline void user_dlm_grab_inode_ref(struct user_lock_res *lockres)
171 BUG(); 171 BUG();
172} 172}
173 173
174static void user_dlm_unblock_lock(void *opaque); 174static void user_dlm_unblock_lock(struct work_struct *work);
175 175
176static void __user_dlm_queue_lockres(struct user_lock_res *lockres) 176static void __user_dlm_queue_lockres(struct user_lock_res *lockres)
177{ 177{
178 if (!(lockres->l_flags & USER_LOCK_QUEUED)) { 178 if (!(lockres->l_flags & USER_LOCK_QUEUED)) {
179 user_dlm_grab_inode_ref(lockres); 179 user_dlm_grab_inode_ref(lockres);
180 180
181 INIT_WORK(&lockres->l_work, user_dlm_unblock_lock, 181 INIT_WORK(&lockres->l_work, user_dlm_unblock_lock);
182 lockres);
183 182
184 queue_work(user_dlm_worker, &lockres->l_work); 183 queue_work(user_dlm_worker, &lockres->l_work);
185 lockres->l_flags |= USER_LOCK_QUEUED; 184 lockres->l_flags |= USER_LOCK_QUEUED;
@@ -279,10 +278,11 @@ static inline void user_dlm_drop_inode_ref(struct user_lock_res *lockres)
279 iput(inode); 278 iput(inode);
280} 279}
281 280
282static void user_dlm_unblock_lock(void *opaque) 281static void user_dlm_unblock_lock(struct work_struct *work)
283{ 282{
284 int new_level, status; 283 int new_level, status;
285 struct user_lock_res *lockres = (struct user_lock_res *) opaque; 284 struct user_lock_res *lockres =
285 container_of(work, struct user_lock_res, l_work);
286 struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres); 286 struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres);
287 287
288 mlog(0, "processing lockres %.*s\n", lockres->l_namelen, 288 mlog(0, "processing lockres %.*s\n", lockres->l_namelen,
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index 8801e41afe80..e335541727f9 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -49,6 +49,7 @@
49#include "dcache.h" 49#include "dcache.h"
50#include "dlmglue.h" 50#include "dlmglue.h"
51#include "extent_map.h" 51#include "extent_map.h"
52#include "file.h"
52#include "heartbeat.h" 53#include "heartbeat.h"
53#include "inode.h" 54#include "inode.h"
54#include "journal.h" 55#include "journal.h"
@@ -769,7 +770,7 @@ static int ocfs2_lock_create(struct ocfs2_super *osb,
769 int dlm_flags) 770 int dlm_flags)
770{ 771{
771 int ret = 0; 772 int ret = 0;
772 enum dlm_status status; 773 enum dlm_status status = DLM_NORMAL;
773 unsigned long flags; 774 unsigned long flags;
774 775
775 mlog_entry_void(); 776 mlog_entry_void();
@@ -1063,10 +1064,10 @@ static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
1063 mlog_exit_void(); 1064 mlog_exit_void();
1064} 1065}
1065 1066
1066int ocfs2_create_new_lock(struct ocfs2_super *osb, 1067static int ocfs2_create_new_lock(struct ocfs2_super *osb,
1067 struct ocfs2_lock_res *lockres, 1068 struct ocfs2_lock_res *lockres,
1068 int ex, 1069 int ex,
1069 int local) 1070 int local)
1070{ 1071{
1071 int level = ex ? LKM_EXMODE : LKM_PRMODE; 1072 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1072 unsigned long flags; 1073 unsigned long flags;
@@ -1137,6 +1138,7 @@ int ocfs2_rw_lock(struct inode *inode, int write)
1137{ 1138{
1138 int status, level; 1139 int status, level;
1139 struct ocfs2_lock_res *lockres; 1140 struct ocfs2_lock_res *lockres;
1141 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1140 1142
1141 BUG_ON(!inode); 1143 BUG_ON(!inode);
1142 1144
@@ -1146,6 +1148,9 @@ int ocfs2_rw_lock(struct inode *inode, int write)
1146 (unsigned long long)OCFS2_I(inode)->ip_blkno, 1148 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1147 write ? "EXMODE" : "PRMODE"); 1149 write ? "EXMODE" : "PRMODE");
1148 1150
1151 if (ocfs2_mount_local(osb))
1152 return 0;
1153
1149 lockres = &OCFS2_I(inode)->ip_rw_lockres; 1154 lockres = &OCFS2_I(inode)->ip_rw_lockres;
1150 1155
1151 level = write ? LKM_EXMODE : LKM_PRMODE; 1156 level = write ? LKM_EXMODE : LKM_PRMODE;
@@ -1163,6 +1168,7 @@ void ocfs2_rw_unlock(struct inode *inode, int write)
1163{ 1168{
1164 int level = write ? LKM_EXMODE : LKM_PRMODE; 1169 int level = write ? LKM_EXMODE : LKM_PRMODE;
1165 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres; 1170 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres;
1171 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1166 1172
1167 mlog_entry_void(); 1173 mlog_entry_void();
1168 1174
@@ -1170,7 +1176,8 @@ void ocfs2_rw_unlock(struct inode *inode, int write)
1170 (unsigned long long)OCFS2_I(inode)->ip_blkno, 1176 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1171 write ? "EXMODE" : "PRMODE"); 1177 write ? "EXMODE" : "PRMODE");
1172 1178
1173 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); 1179 if (!ocfs2_mount_local(osb))
1180 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1174 1181
1175 mlog_exit_void(); 1182 mlog_exit_void();
1176} 1183}
@@ -1181,6 +1188,7 @@ int ocfs2_data_lock_full(struct inode *inode,
1181{ 1188{
1182 int status = 0, level; 1189 int status = 0, level;
1183 struct ocfs2_lock_res *lockres; 1190 struct ocfs2_lock_res *lockres;
1191 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1184 1192
1185 BUG_ON(!inode); 1193 BUG_ON(!inode);
1186 1194
@@ -1200,6 +1208,9 @@ int ocfs2_data_lock_full(struct inode *inode,
1200 goto out; 1208 goto out;
1201 } 1209 }
1202 1210
1211 if (ocfs2_mount_local(osb))
1212 goto out;
1213
1203 lockres = &OCFS2_I(inode)->ip_data_lockres; 1214 lockres = &OCFS2_I(inode)->ip_data_lockres;
1204 1215
1205 level = write ? LKM_EXMODE : LKM_PRMODE; 1216 level = write ? LKM_EXMODE : LKM_PRMODE;
@@ -1268,6 +1279,7 @@ void ocfs2_data_unlock(struct inode *inode,
1268{ 1279{
1269 int level = write ? LKM_EXMODE : LKM_PRMODE; 1280 int level = write ? LKM_EXMODE : LKM_PRMODE;
1270 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres; 1281 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres;
1282 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1271 1283
1272 mlog_entry_void(); 1284 mlog_entry_void();
1273 1285
@@ -1275,7 +1287,8 @@ void ocfs2_data_unlock(struct inode *inode,
1275 (unsigned long long)OCFS2_I(inode)->ip_blkno, 1287 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1276 write ? "EXMODE" : "PRMODE"); 1288 write ? "EXMODE" : "PRMODE");
1277 1289
1278 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) 1290 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)) &&
1291 !ocfs2_mount_local(osb))
1279 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); 1292 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1280 1293
1281 mlog_exit_void(); 1294 mlog_exit_void();
@@ -1466,8 +1479,9 @@ static int ocfs2_meta_lock_update(struct inode *inode,
1466{ 1479{
1467 int status = 0; 1480 int status = 0;
1468 struct ocfs2_inode_info *oi = OCFS2_I(inode); 1481 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1469 struct ocfs2_lock_res *lockres; 1482 struct ocfs2_lock_res *lockres = NULL;
1470 struct ocfs2_dinode *fe; 1483 struct ocfs2_dinode *fe;
1484 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1471 1485
1472 mlog_entry_void(); 1486 mlog_entry_void();
1473 1487
@@ -1482,10 +1496,12 @@ static int ocfs2_meta_lock_update(struct inode *inode,
1482 } 1496 }
1483 spin_unlock(&oi->ip_lock); 1497 spin_unlock(&oi->ip_lock);
1484 1498
1485 lockres = &oi->ip_meta_lockres; 1499 if (!ocfs2_mount_local(osb)) {
1500 lockres = &oi->ip_meta_lockres;
1486 1501
1487 if (!ocfs2_should_refresh_lock_res(lockres)) 1502 if (!ocfs2_should_refresh_lock_res(lockres))
1488 goto bail; 1503 goto bail;
1504 }
1489 1505
1490 /* This will discard any caching information we might have had 1506 /* This will discard any caching information we might have had
1491 * for the inode metadata. */ 1507 * for the inode metadata. */
@@ -1495,7 +1511,7 @@ static int ocfs2_meta_lock_update(struct inode *inode,
1495 * map (directories, bitmap files, etc) */ 1511 * map (directories, bitmap files, etc) */
1496 ocfs2_extent_map_trunc(inode, 0); 1512 ocfs2_extent_map_trunc(inode, 0);
1497 1513
1498 if (ocfs2_meta_lvb_is_trustable(inode, lockres)) { 1514 if (lockres && ocfs2_meta_lvb_is_trustable(inode, lockres)) {
1499 mlog(0, "Trusting LVB on inode %llu\n", 1515 mlog(0, "Trusting LVB on inode %llu\n",
1500 (unsigned long long)oi->ip_blkno); 1516 (unsigned long long)oi->ip_blkno);
1501 ocfs2_refresh_inode_from_lvb(inode); 1517 ocfs2_refresh_inode_from_lvb(inode);
@@ -1542,7 +1558,8 @@ static int ocfs2_meta_lock_update(struct inode *inode,
1542 1558
1543 status = 0; 1559 status = 0;
1544bail_refresh: 1560bail_refresh:
1545 ocfs2_complete_lock_res_refresh(lockres, status); 1561 if (lockres)
1562 ocfs2_complete_lock_res_refresh(lockres, status);
1546bail: 1563bail:
1547 mlog_exit(status); 1564 mlog_exit(status);
1548 return status; 1565 return status;
@@ -1579,13 +1596,12 @@ static int ocfs2_assign_bh(struct inode *inode,
1579 * the result of the lock will be communicated via the callback. 1596 * the result of the lock will be communicated via the callback.
1580 */ 1597 */
1581int ocfs2_meta_lock_full(struct inode *inode, 1598int ocfs2_meta_lock_full(struct inode *inode,
1582 struct ocfs2_journal_handle *handle,
1583 struct buffer_head **ret_bh, 1599 struct buffer_head **ret_bh,
1584 int ex, 1600 int ex,
1585 int arg_flags) 1601 int arg_flags)
1586{ 1602{
1587 int status, level, dlm_flags, acquired; 1603 int status, level, dlm_flags, acquired;
1588 struct ocfs2_lock_res *lockres; 1604 struct ocfs2_lock_res *lockres = NULL;
1589 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 1605 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1590 struct buffer_head *local_bh = NULL; 1606 struct buffer_head *local_bh = NULL;
1591 1607
@@ -1607,6 +1623,9 @@ int ocfs2_meta_lock_full(struct inode *inode,
1607 goto bail; 1623 goto bail;
1608 } 1624 }
1609 1625
1626 if (ocfs2_mount_local(osb))
1627 goto local;
1628
1610 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) 1629 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
1611 wait_event(osb->recovery_event, 1630 wait_event(osb->recovery_event,
1612 ocfs2_node_map_is_empty(osb, &osb->recovery_map)); 1631 ocfs2_node_map_is_empty(osb, &osb->recovery_map));
@@ -1636,6 +1655,7 @@ int ocfs2_meta_lock_full(struct inode *inode,
1636 wait_event(osb->recovery_event, 1655 wait_event(osb->recovery_event,
1637 ocfs2_node_map_is_empty(osb, &osb->recovery_map)); 1656 ocfs2_node_map_is_empty(osb, &osb->recovery_map));
1638 1657
1658local:
1639 /* 1659 /*
1640 * We only see this flag if we're being called from 1660 * We only see this flag if we're being called from
1641 * ocfs2_read_locked_inode(). It means we're locking an inode 1661 * ocfs2_read_locked_inode(). It means we're locking an inode
@@ -1644,7 +1664,8 @@ int ocfs2_meta_lock_full(struct inode *inode,
1644 */ 1664 */
1645 if (inode->i_state & I_NEW) { 1665 if (inode->i_state & I_NEW) {
1646 status = 0; 1666 status = 0;
1647 ocfs2_complete_lock_res_refresh(lockres, 0); 1667 if (lockres)
1668 ocfs2_complete_lock_res_refresh(lockres, 0);
1648 goto bail; 1669 goto bail;
1649 } 1670 }
1650 1671
@@ -1668,12 +1689,6 @@ int ocfs2_meta_lock_full(struct inode *inode,
1668 } 1689 }
1669 } 1690 }
1670 1691
1671 if (handle) {
1672 status = ocfs2_handle_add_lock(handle, inode);
1673 if (status < 0)
1674 mlog_errno(status);
1675 }
1676
1677bail: 1692bail:
1678 if (status < 0) { 1693 if (status < 0) {
1679 if (ret_bh && (*ret_bh)) { 1694 if (ret_bh && (*ret_bh)) {
@@ -1713,18 +1728,16 @@ bail:
1713 * the lock inversion simply. 1728 * the lock inversion simply.
1714 */ 1729 */
1715int ocfs2_meta_lock_with_page(struct inode *inode, 1730int ocfs2_meta_lock_with_page(struct inode *inode,
1716 struct ocfs2_journal_handle *handle,
1717 struct buffer_head **ret_bh, 1731 struct buffer_head **ret_bh,
1718 int ex, 1732 int ex,
1719 struct page *page) 1733 struct page *page)
1720{ 1734{
1721 int ret; 1735 int ret;
1722 1736
1723 ret = ocfs2_meta_lock_full(inode, handle, ret_bh, ex, 1737 ret = ocfs2_meta_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK);
1724 OCFS2_LOCK_NONBLOCK);
1725 if (ret == -EAGAIN) { 1738 if (ret == -EAGAIN) {
1726 unlock_page(page); 1739 unlock_page(page);
1727 if (ocfs2_meta_lock(inode, handle, ret_bh, ex) == 0) 1740 if (ocfs2_meta_lock(inode, ret_bh, ex) == 0)
1728 ocfs2_meta_unlock(inode, ex); 1741 ocfs2_meta_unlock(inode, ex);
1729 ret = AOP_TRUNCATED_PAGE; 1742 ret = AOP_TRUNCATED_PAGE;
1730 } 1743 }
@@ -1732,11 +1745,50 @@ int ocfs2_meta_lock_with_page(struct inode *inode,
1732 return ret; 1745 return ret;
1733} 1746}
1734 1747
1748int ocfs2_meta_lock_atime(struct inode *inode,
1749 struct vfsmount *vfsmnt,
1750 int *level)
1751{
1752 int ret;
1753
1754 mlog_entry_void();
1755 ret = ocfs2_meta_lock(inode, NULL, 0);
1756 if (ret < 0) {
1757 mlog_errno(ret);
1758 return ret;
1759 }
1760
1761 /*
1762 * If we should update atime, we will get EX lock,
1763 * otherwise we just get PR lock.
1764 */
1765 if (ocfs2_should_update_atime(inode, vfsmnt)) {
1766 struct buffer_head *bh = NULL;
1767
1768 ocfs2_meta_unlock(inode, 0);
1769 ret = ocfs2_meta_lock(inode, &bh, 1);
1770 if (ret < 0) {
1771 mlog_errno(ret);
1772 return ret;
1773 }
1774 *level = 1;
1775 if (ocfs2_should_update_atime(inode, vfsmnt))
1776 ocfs2_update_inode_atime(inode, bh);
1777 if (bh)
1778 brelse(bh);
1779 } else
1780 *level = 0;
1781
1782 mlog_exit(ret);
1783 return ret;
1784}
1785
1735void ocfs2_meta_unlock(struct inode *inode, 1786void ocfs2_meta_unlock(struct inode *inode,
1736 int ex) 1787 int ex)
1737{ 1788{
1738 int level = ex ? LKM_EXMODE : LKM_PRMODE; 1789 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1739 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres; 1790 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres;
1791 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1740 1792
1741 mlog_entry_void(); 1793 mlog_entry_void();
1742 1794
@@ -1744,7 +1796,8 @@ void ocfs2_meta_unlock(struct inode *inode,
1744 (unsigned long long)OCFS2_I(inode)->ip_blkno, 1796 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1745 ex ? "EXMODE" : "PRMODE"); 1797 ex ? "EXMODE" : "PRMODE");
1746 1798
1747 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) 1799 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)) &&
1800 !ocfs2_mount_local(osb))
1748 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); 1801 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1749 1802
1750 mlog_exit_void(); 1803 mlog_exit_void();
@@ -1753,7 +1806,7 @@ void ocfs2_meta_unlock(struct inode *inode,
1753int ocfs2_super_lock(struct ocfs2_super *osb, 1806int ocfs2_super_lock(struct ocfs2_super *osb,
1754 int ex) 1807 int ex)
1755{ 1808{
1756 int status; 1809 int status = 0;
1757 int level = ex ? LKM_EXMODE : LKM_PRMODE; 1810 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1758 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; 1811 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
1759 struct buffer_head *bh; 1812 struct buffer_head *bh;
@@ -1764,6 +1817,9 @@ int ocfs2_super_lock(struct ocfs2_super *osb,
1764 if (ocfs2_is_hard_readonly(osb)) 1817 if (ocfs2_is_hard_readonly(osb))
1765 return -EROFS; 1818 return -EROFS;
1766 1819
1820 if (ocfs2_mount_local(osb))
1821 goto bail;
1822
1767 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0); 1823 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
1768 if (status < 0) { 1824 if (status < 0) {
1769 mlog_errno(status); 1825 mlog_errno(status);
@@ -1802,7 +1858,8 @@ void ocfs2_super_unlock(struct ocfs2_super *osb,
1802 int level = ex ? LKM_EXMODE : LKM_PRMODE; 1858 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1803 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; 1859 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
1804 1860
1805 ocfs2_cluster_unlock(osb, lockres, level); 1861 if (!ocfs2_mount_local(osb))
1862 ocfs2_cluster_unlock(osb, lockres, level);
1806} 1863}
1807 1864
1808int ocfs2_rename_lock(struct ocfs2_super *osb) 1865int ocfs2_rename_lock(struct ocfs2_super *osb)
@@ -1813,6 +1870,9 @@ int ocfs2_rename_lock(struct ocfs2_super *osb)
1813 if (ocfs2_is_hard_readonly(osb)) 1870 if (ocfs2_is_hard_readonly(osb))
1814 return -EROFS; 1871 return -EROFS;
1815 1872
1873 if (ocfs2_mount_local(osb))
1874 return 0;
1875
1816 status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0); 1876 status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0);
1817 if (status < 0) 1877 if (status < 0)
1818 mlog_errno(status); 1878 mlog_errno(status);
@@ -1824,7 +1884,8 @@ void ocfs2_rename_unlock(struct ocfs2_super *osb)
1824{ 1884{
1825 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; 1885 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
1826 1886
1827 ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE); 1887 if (!ocfs2_mount_local(osb))
1888 ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE);
1828} 1889}
1829 1890
1830int ocfs2_dentry_lock(struct dentry *dentry, int ex) 1891int ocfs2_dentry_lock(struct dentry *dentry, int ex)
@@ -1839,6 +1900,9 @@ int ocfs2_dentry_lock(struct dentry *dentry, int ex)
1839 if (ocfs2_is_hard_readonly(osb)) 1900 if (ocfs2_is_hard_readonly(osb))
1840 return -EROFS; 1901 return -EROFS;
1841 1902
1903 if (ocfs2_mount_local(osb))
1904 return 0;
1905
1842 ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0); 1906 ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0);
1843 if (ret < 0) 1907 if (ret < 0)
1844 mlog_errno(ret); 1908 mlog_errno(ret);
@@ -1852,7 +1916,8 @@ void ocfs2_dentry_unlock(struct dentry *dentry, int ex)
1852 struct ocfs2_dentry_lock *dl = dentry->d_fsdata; 1916 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
1853 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); 1917 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
1854 1918
1855 ocfs2_cluster_unlock(osb, &dl->dl_lockres, level); 1919 if (!ocfs2_mount_local(osb))
1920 ocfs2_cluster_unlock(osb, &dl->dl_lockres, level);
1856} 1921}
1857 1922
1858/* Reference counting of the dlm debug structure. We want this because 1923/* Reference counting of the dlm debug structure. We want this because
@@ -2115,12 +2180,15 @@ static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb)
2115 2180
2116int ocfs2_dlm_init(struct ocfs2_super *osb) 2181int ocfs2_dlm_init(struct ocfs2_super *osb)
2117{ 2182{
2118 int status; 2183 int status = 0;
2119 u32 dlm_key; 2184 u32 dlm_key;
2120 struct dlm_ctxt *dlm; 2185 struct dlm_ctxt *dlm = NULL;
2121 2186
2122 mlog_entry_void(); 2187 mlog_entry_void();
2123 2188
2189 if (ocfs2_mount_local(osb))
2190 goto local;
2191
2124 status = ocfs2_dlm_init_debug(osb); 2192 status = ocfs2_dlm_init_debug(osb);
2125 if (status < 0) { 2193 if (status < 0) {
2126 mlog_errno(status); 2194 mlog_errno(status);
@@ -2148,11 +2216,12 @@ int ocfs2_dlm_init(struct ocfs2_super *osb)
2148 goto bail; 2216 goto bail;
2149 } 2217 }
2150 2218
2219 dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb);
2220
2221local:
2151 ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb); 2222 ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb);
2152 ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb); 2223 ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb);
2153 2224
2154 dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb);
2155
2156 osb->dlm = dlm; 2225 osb->dlm = dlm;
2157 2226
2158 status = 0; 2227 status = 0;
@@ -2649,6 +2718,15 @@ static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
2649 inode = ocfs2_lock_res_inode(lockres); 2718 inode = ocfs2_lock_res_inode(lockres);
2650 mapping = inode->i_mapping; 2719 mapping = inode->i_mapping;
2651 2720
2721 /*
2722 * We need this before the filemap_fdatawrite() so that it can
2723 * transfer the dirty bit from the PTE to the
2724 * page. Unfortunately this means that even for EX->PR
2725 * downconverts, we'll lose our mappings and have to build
2726 * them up again.
2727 */
2728 unmap_mapping_range(mapping, 0, 0, 0);
2729
2652 if (filemap_fdatawrite(mapping)) { 2730 if (filemap_fdatawrite(mapping)) {
2653 mlog(ML_ERROR, "Could not sync inode %llu for downconvert!", 2731 mlog(ML_ERROR, "Could not sync inode %llu for downconvert!",
2654 (unsigned long long)OCFS2_I(inode)->ip_blkno); 2732 (unsigned long long)OCFS2_I(inode)->ip_blkno);
@@ -2656,7 +2734,6 @@ static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
2656 sync_mapping_buffers(mapping); 2734 sync_mapping_buffers(mapping);
2657 if (blocking == LKM_EXMODE) { 2735 if (blocking == LKM_EXMODE) {
2658 truncate_inode_pages(mapping, 0); 2736 truncate_inode_pages(mapping, 0);
2659 unmap_mapping_range(mapping, 0, 0, 0);
2660 } else { 2737 } else {
2661 /* We only need to wait on the I/O if we're not also 2738 /* We only need to wait on the I/O if we're not also
2662 * truncating pages because truncate_inode_pages waits 2739 * truncating pages because truncate_inode_pages waits
diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h
index 4a2769387229..c343fca68cf1 100644
--- a/fs/ocfs2/dlmglue.h
+++ b/fs/ocfs2/dlmglue.h
@@ -68,8 +68,6 @@ void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl,
68 u64 parent, struct inode *inode); 68 u64 parent, struct inode *inode);
69void ocfs2_lock_res_free(struct ocfs2_lock_res *res); 69void ocfs2_lock_res_free(struct ocfs2_lock_res *res);
70int ocfs2_create_new_inode_locks(struct inode *inode); 70int ocfs2_create_new_inode_locks(struct inode *inode);
71int ocfs2_create_new_lock(struct ocfs2_super *osb,
72 struct ocfs2_lock_res *lockres, int ex, int local);
73int ocfs2_drop_inode_locks(struct inode *inode); 71int ocfs2_drop_inode_locks(struct inode *inode);
74int ocfs2_data_lock_full(struct inode *inode, 72int ocfs2_data_lock_full(struct inode *inode,
75 int write, 73 int write,
@@ -82,19 +80,20 @@ void ocfs2_data_unlock(struct inode *inode,
82 int write); 80 int write);
83int ocfs2_rw_lock(struct inode *inode, int write); 81int ocfs2_rw_lock(struct inode *inode, int write);
84void ocfs2_rw_unlock(struct inode *inode, int write); 82void ocfs2_rw_unlock(struct inode *inode, int write);
83int ocfs2_meta_lock_atime(struct inode *inode,
84 struct vfsmount *vfsmnt,
85 int *level);
85int ocfs2_meta_lock_full(struct inode *inode, 86int ocfs2_meta_lock_full(struct inode *inode,
86 struct ocfs2_journal_handle *handle,
87 struct buffer_head **ret_bh, 87 struct buffer_head **ret_bh,
88 int ex, 88 int ex,
89 int arg_flags); 89 int arg_flags);
90int ocfs2_meta_lock_with_page(struct inode *inode, 90int ocfs2_meta_lock_with_page(struct inode *inode,
91 struct ocfs2_journal_handle *handle,
92 struct buffer_head **ret_bh, 91 struct buffer_head **ret_bh,
93 int ex, 92 int ex,
94 struct page *page); 93 struct page *page);
95/* 99% of the time we don't want to supply any additional flags -- 94/* 99% of the time we don't want to supply any additional flags --
96 * those are for very specific cases only. */ 95 * those are for very specific cases only. */
97#define ocfs2_meta_lock(i, h, b, e) ocfs2_meta_lock_full(i, h, b, e, 0) 96#define ocfs2_meta_lock(i, b, e) ocfs2_meta_lock_full(i, b, e, 0)
98void ocfs2_meta_unlock(struct inode *inode, 97void ocfs2_meta_unlock(struct inode *inode,
99 int ex); 98 int ex);
100int ocfs2_super_lock(struct ocfs2_super *osb, 99int ocfs2_super_lock(struct ocfs2_super *osb,
diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c
index fb91089a60a7..06be6e774cf9 100644
--- a/fs/ocfs2/export.c
+++ b/fs/ocfs2/export.c
@@ -100,7 +100,7 @@ static struct dentry *ocfs2_get_parent(struct dentry *child)
100 mlog(0, "find parent of directory %llu\n", 100 mlog(0, "find parent of directory %llu\n",
101 (unsigned long long)OCFS2_I(dir)->ip_blkno); 101 (unsigned long long)OCFS2_I(dir)->ip_blkno);
102 102
103 status = ocfs2_meta_lock(dir, NULL, NULL, 0); 103 status = ocfs2_meta_lock(dir, NULL, 0);
104 if (status < 0) { 104 if (status < 0) {
105 if (status != -ENOENT) 105 if (status != -ENOENT)
106 mlog_errno(status); 106 mlog_errno(status);
diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c
index fcd4475d1f89..80ac69f11d9f 100644
--- a/fs/ocfs2/extent_map.c
+++ b/fs/ocfs2/extent_map.c
@@ -61,7 +61,7 @@ struct ocfs2_em_insert_context {
61 struct ocfs2_extent_map_entry *right_ent; 61 struct ocfs2_extent_map_entry *right_ent;
62}; 62};
63 63
64static kmem_cache_t *ocfs2_em_ent_cachep = NULL; 64static struct kmem_cache *ocfs2_em_ent_cachep = NULL;
65 65
66 66
67static struct ocfs2_extent_map_entry * 67static struct ocfs2_extent_map_entry *
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 1be74c4e7814..10953a508f2f 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -31,6 +31,8 @@
31#include <linux/pagemap.h> 31#include <linux/pagemap.h>
32#include <linux/uio.h> 32#include <linux/uio.h>
33#include <linux/sched.h> 33#include <linux/sched.h>
34#include <linux/pipe_fs_i.h>
35#include <linux/mount.h>
34 36
35#define MLOG_MASK_PREFIX ML_INODE 37#define MLOG_MASK_PREFIX ML_INODE
36#include <cluster/masklog.h> 38#include <cluster/masklog.h>
@@ -66,7 +68,7 @@ static int ocfs2_file_open(struct inode *inode, struct file *file)
66 struct ocfs2_inode_info *oi = OCFS2_I(inode); 68 struct ocfs2_inode_info *oi = OCFS2_I(inode);
67 69
68 mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file, 70 mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file,
69 file->f_dentry->d_name.len, file->f_dentry->d_name.name); 71 file->f_path.dentry->d_name.len, file->f_path.dentry->d_name.name);
70 72
71 spin_lock(&oi->ip_lock); 73 spin_lock(&oi->ip_lock);
72 74
@@ -96,8 +98,8 @@ static int ocfs2_file_release(struct inode *inode, struct file *file)
96 struct ocfs2_inode_info *oi = OCFS2_I(inode); 98 struct ocfs2_inode_info *oi = OCFS2_I(inode);
97 99
98 mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file, 100 mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file,
99 file->f_dentry->d_name.len, 101 file->f_path.dentry->d_name.len,
100 file->f_dentry->d_name.name); 102 file->f_path.dentry->d_name.name);
101 103
102 spin_lock(&oi->ip_lock); 104 spin_lock(&oi->ip_lock);
103 if (!--oi->ip_open_count) 105 if (!--oi->ip_open_count)
@@ -134,7 +136,77 @@ bail:
134 return (err < 0) ? -EIO : 0; 136 return (err < 0) ? -EIO : 0;
135} 137}
136 138
137int ocfs2_set_inode_size(struct ocfs2_journal_handle *handle, 139int ocfs2_should_update_atime(struct inode *inode,
140 struct vfsmount *vfsmnt)
141{
142 struct timespec now;
143 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
144
145 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
146 return 0;
147
148 if ((inode->i_flags & S_NOATIME) ||
149 ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)))
150 return 0;
151
152 /*
153 * We can be called with no vfsmnt structure - NFSD will
154 * sometimes do this.
155 *
156 * Note that our action here is different than touch_atime() -
157 * if we can't tell whether this is a noatime mount, then we
158 * don't know whether to trust the value of s_atime_quantum.
159 */
160 if (vfsmnt == NULL)
161 return 0;
162
163 if ((vfsmnt->mnt_flags & MNT_NOATIME) ||
164 ((vfsmnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
165 return 0;
166
167 if (vfsmnt->mnt_flags & MNT_RELATIME) {
168 if ((timespec_compare(&inode->i_atime, &inode->i_mtime) <= 0) ||
169 (timespec_compare(&inode->i_atime, &inode->i_ctime) <= 0))
170 return 1;
171
172 return 0;
173 }
174
175 now = CURRENT_TIME;
176 if ((now.tv_sec - inode->i_atime.tv_sec <= osb->s_atime_quantum))
177 return 0;
178 else
179 return 1;
180}
181
182int ocfs2_update_inode_atime(struct inode *inode,
183 struct buffer_head *bh)
184{
185 int ret;
186 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
187 handle_t *handle;
188
189 mlog_entry_void();
190
191 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
192 if (handle == NULL) {
193 ret = -ENOMEM;
194 mlog_errno(ret);
195 goto out;
196 }
197
198 inode->i_atime = CURRENT_TIME;
199 ret = ocfs2_mark_inode_dirty(handle, inode, bh);
200 if (ret < 0)
201 mlog_errno(ret);
202
203 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
204out:
205 mlog_exit(ret);
206 return ret;
207}
208
209int ocfs2_set_inode_size(handle_t *handle,
138 struct inode *inode, 210 struct inode *inode,
139 struct buffer_head *fe_bh, 211 struct buffer_head *fe_bh,
140 u64 new_i_size) 212 u64 new_i_size)
@@ -163,10 +235,9 @@ static int ocfs2_simple_size_update(struct inode *inode,
163{ 235{
164 int ret; 236 int ret;
165 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 237 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
166 struct ocfs2_journal_handle *handle = NULL; 238 handle_t *handle = NULL;
167 239
168 handle = ocfs2_start_trans(osb, NULL, 240 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
169 OCFS2_INODE_UPDATE_CREDITS);
170 if (handle == NULL) { 241 if (handle == NULL) {
171 ret = -ENOMEM; 242 ret = -ENOMEM;
172 mlog_errno(ret); 243 mlog_errno(ret);
@@ -178,7 +249,7 @@ static int ocfs2_simple_size_update(struct inode *inode,
178 if (ret < 0) 249 if (ret < 0)
179 mlog_errno(ret); 250 mlog_errno(ret);
180 251
181 ocfs2_commit_trans(handle); 252 ocfs2_commit_trans(osb, handle);
182out: 253out:
183 return ret; 254 return ret;
184} 255}
@@ -189,14 +260,14 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
189 u64 new_i_size) 260 u64 new_i_size)
190{ 261{
191 int status; 262 int status;
192 struct ocfs2_journal_handle *handle; 263 handle_t *handle;
193 264
194 mlog_entry_void(); 265 mlog_entry_void();
195 266
196 /* TODO: This needs to actually orphan the inode in this 267 /* TODO: This needs to actually orphan the inode in this
197 * transaction. */ 268 * transaction. */
198 269
199 handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS); 270 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
200 if (IS_ERR(handle)) { 271 if (IS_ERR(handle)) {
201 status = PTR_ERR(handle); 272 status = PTR_ERR(handle);
202 mlog_errno(status); 273 mlog_errno(status);
@@ -207,7 +278,7 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
207 if (status < 0) 278 if (status < 0)
208 mlog_errno(status); 279 mlog_errno(status);
209 280
210 ocfs2_commit_trans(handle); 281 ocfs2_commit_trans(osb, handle);
211out: 282out:
212 mlog_exit(status); 283 mlog_exit(status);
213 return status; 284 return status;
@@ -328,7 +399,7 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
328 struct inode *inode, 399 struct inode *inode,
329 u32 clusters_to_add, 400 u32 clusters_to_add,
330 struct buffer_head *fe_bh, 401 struct buffer_head *fe_bh,
331 struct ocfs2_journal_handle *handle, 402 handle_t *handle,
332 struct ocfs2_alloc_context *data_ac, 403 struct ocfs2_alloc_context *data_ac,
333 struct ocfs2_alloc_context *meta_ac, 404 struct ocfs2_alloc_context *meta_ac,
334 enum ocfs2_alloc_restarted *reason_ret) 405 enum ocfs2_alloc_restarted *reason_ret)
@@ -433,7 +504,7 @@ static int ocfs2_extend_allocation(struct inode *inode,
433 u32 prev_clusters; 504 u32 prev_clusters;
434 struct buffer_head *bh = NULL; 505 struct buffer_head *bh = NULL;
435 struct ocfs2_dinode *fe = NULL; 506 struct ocfs2_dinode *fe = NULL;
436 struct ocfs2_journal_handle *handle = NULL; 507 handle_t *handle = NULL;
437 struct ocfs2_alloc_context *data_ac = NULL; 508 struct ocfs2_alloc_context *data_ac = NULL;
438 struct ocfs2_alloc_context *meta_ac = NULL; 509 struct ocfs2_alloc_context *meta_ac = NULL;
439 enum ocfs2_alloc_restarted why; 510 enum ocfs2_alloc_restarted why;
@@ -463,13 +534,6 @@ restart_all:
463 (unsigned long long)OCFS2_I(inode)->ip_blkno, i_size_read(inode), 534 (unsigned long long)OCFS2_I(inode)->ip_blkno, i_size_read(inode),
464 fe->i_clusters, clusters_to_add); 535 fe->i_clusters, clusters_to_add);
465 536
466 handle = ocfs2_alloc_handle(osb);
467 if (handle == NULL) {
468 status = -ENOMEM;
469 mlog_errno(status);
470 goto leave;
471 }
472
473 num_free_extents = ocfs2_num_free_extents(osb, 537 num_free_extents = ocfs2_num_free_extents(osb,
474 inode, 538 inode,
475 fe); 539 fe);
@@ -480,10 +544,7 @@ restart_all:
480 } 544 }
481 545
482 if (!num_free_extents) { 546 if (!num_free_extents) {
483 status = ocfs2_reserve_new_metadata(osb, 547 status = ocfs2_reserve_new_metadata(osb, fe, &meta_ac);
484 handle,
485 fe,
486 &meta_ac);
487 if (status < 0) { 548 if (status < 0) {
488 if (status != -ENOSPC) 549 if (status != -ENOSPC)
489 mlog_errno(status); 550 mlog_errno(status);
@@ -491,10 +552,7 @@ restart_all:
491 } 552 }
492 } 553 }
493 554
494 status = ocfs2_reserve_clusters(osb, 555 status = ocfs2_reserve_clusters(osb, clusters_to_add, &data_ac);
495 handle,
496 clusters_to_add,
497 &data_ac);
498 if (status < 0) { 556 if (status < 0) {
499 if (status != -ENOSPC) 557 if (status != -ENOSPC)
500 mlog_errno(status); 558 mlog_errno(status);
@@ -509,7 +567,7 @@ restart_all:
509 drop_alloc_sem = 1; 567 drop_alloc_sem = 1;
510 568
511 credits = ocfs2_calc_extend_credits(osb->sb, fe, clusters_to_add); 569 credits = ocfs2_calc_extend_credits(osb->sb, fe, clusters_to_add);
512 handle = ocfs2_start_trans(osb, handle, credits); 570 handle = ocfs2_start_trans(osb, credits);
513 if (IS_ERR(handle)) { 571 if (IS_ERR(handle)) {
514 status = PTR_ERR(handle); 572 status = PTR_ERR(handle);
515 handle = NULL; 573 handle = NULL;
@@ -589,7 +647,7 @@ leave:
589 drop_alloc_sem = 0; 647 drop_alloc_sem = 0;
590 } 648 }
591 if (handle) { 649 if (handle) {
592 ocfs2_commit_trans(handle); 650 ocfs2_commit_trans(osb, handle);
593 handle = NULL; 651 handle = NULL;
594 } 652 }
595 if (data_ac) { 653 if (data_ac) {
@@ -624,7 +682,7 @@ static int ocfs2_write_zero_page(struct inode *inode,
624 struct page *page; 682 struct page *page;
625 unsigned long index; 683 unsigned long index;
626 unsigned int offset; 684 unsigned int offset;
627 struct ocfs2_journal_handle *handle = NULL; 685 handle_t *handle = NULL;
628 int ret; 686 int ret;
629 687
630 offset = (size & (PAGE_CACHE_SIZE-1)); /* Within page */ 688 offset = (size & (PAGE_CACHE_SIZE-1)); /* Within page */
@@ -668,7 +726,7 @@ static int ocfs2_write_zero_page(struct inode *inode,
668 ret = 0; 726 ret = 0;
669 727
670 if (handle) 728 if (handle)
671 ocfs2_commit_trans(handle); 729 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
672out_unlock: 730out_unlock:
673 unlock_page(page); 731 unlock_page(page);
674 page_cache_release(page); 732 page_cache_release(page);
@@ -789,7 +847,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
789 struct super_block *sb = inode->i_sb; 847 struct super_block *sb = inode->i_sb;
790 struct ocfs2_super *osb = OCFS2_SB(sb); 848 struct ocfs2_super *osb = OCFS2_SB(sb);
791 struct buffer_head *bh = NULL; 849 struct buffer_head *bh = NULL;
792 struct ocfs2_journal_handle *handle = NULL; 850 handle_t *handle = NULL;
793 851
794 mlog_entry("(0x%p, '%.*s')\n", dentry, 852 mlog_entry("(0x%p, '%.*s')\n", dentry,
795 dentry->d_name.len, dentry->d_name.name); 853 dentry->d_name.len, dentry->d_name.name);
@@ -825,7 +883,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
825 } 883 }
826 } 884 }
827 885
828 status = ocfs2_meta_lock(inode, NULL, &bh, 1); 886 status = ocfs2_meta_lock(inode, &bh, 1);
829 if (status < 0) { 887 if (status < 0) {
830 if (status != -ENOENT) 888 if (status != -ENOENT)
831 mlog_errno(status); 889 mlog_errno(status);
@@ -845,7 +903,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
845 } 903 }
846 } 904 }
847 905
848 handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS); 906 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
849 if (IS_ERR(handle)) { 907 if (IS_ERR(handle)) {
850 status = PTR_ERR(handle); 908 status = PTR_ERR(handle);
851 mlog_errno(status); 909 mlog_errno(status);
@@ -863,7 +921,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
863 mlog_errno(status); 921 mlog_errno(status);
864 922
865bail_commit: 923bail_commit:
866 ocfs2_commit_trans(handle); 924 ocfs2_commit_trans(osb, handle);
867bail_unlock: 925bail_unlock:
868 ocfs2_meta_unlock(inode, 1); 926 ocfs2_meta_unlock(inode, 1);
869bail_unlock_rw: 927bail_unlock_rw:
@@ -906,19 +964,39 @@ bail:
906 return err; 964 return err;
907} 965}
908 966
967int ocfs2_permission(struct inode *inode, int mask, struct nameidata *nd)
968{
969 int ret;
970
971 mlog_entry_void();
972
973 ret = ocfs2_meta_lock(inode, NULL, 0);
974 if (ret) {
975 mlog_errno(ret);
976 goto out;
977 }
978
979 ret = generic_permission(inode, mask, NULL);
980
981 ocfs2_meta_unlock(inode, 0);
982out:
983 mlog_exit(ret);
984 return ret;
985}
986
909static int ocfs2_write_remove_suid(struct inode *inode) 987static int ocfs2_write_remove_suid(struct inode *inode)
910{ 988{
911 int ret; 989 int ret;
912 struct buffer_head *bh = NULL; 990 struct buffer_head *bh = NULL;
913 struct ocfs2_inode_info *oi = OCFS2_I(inode); 991 struct ocfs2_inode_info *oi = OCFS2_I(inode);
914 struct ocfs2_journal_handle *handle; 992 handle_t *handle;
915 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 993 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
916 struct ocfs2_dinode *di; 994 struct ocfs2_dinode *di;
917 995
918 mlog_entry("(Inode %llu, mode 0%o)\n", 996 mlog_entry("(Inode %llu, mode 0%o)\n",
919 (unsigned long long)oi->ip_blkno, inode->i_mode); 997 (unsigned long long)oi->ip_blkno, inode->i_mode);
920 998
921 handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS); 999 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
922 if (handle == NULL) { 1000 if (handle == NULL) {
923 ret = -ENOMEM; 1001 ret = -ENOMEM;
924 mlog_errno(ret); 1002 mlog_errno(ret);
@@ -951,75 +1029,29 @@ static int ocfs2_write_remove_suid(struct inode *inode)
951out_bh: 1029out_bh:
952 brelse(bh); 1030 brelse(bh);
953out_trans: 1031out_trans:
954 ocfs2_commit_trans(handle); 1032 ocfs2_commit_trans(osb, handle);
955out: 1033out:
956 mlog_exit(ret); 1034 mlog_exit(ret);
957 return ret; 1035 return ret;
958} 1036}
959 1037
960static inline int ocfs2_write_should_remove_suid(struct inode *inode) 1038static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
961{ 1039 loff_t *ppos,
962 mode_t mode = inode->i_mode; 1040 size_t count,
963 1041 int appending)
964 if (!capable(CAP_FSETID)) {
965 if (unlikely(mode & S_ISUID))
966 return 1;
967
968 if (unlikely((mode & S_ISGID) && (mode & S_IXGRP)))
969 return 1;
970 }
971 return 0;
972}
973
974static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
975 const struct iovec *iov,
976 unsigned long nr_segs,
977 loff_t pos)
978{ 1042{
979 int ret, rw_level = -1, meta_level = -1, have_alloc_sem = 0; 1043 int ret = 0, meta_level = appending;
1044 struct inode *inode = dentry->d_inode;
980 u32 clusters; 1045 u32 clusters;
981 struct file *filp = iocb->ki_filp;
982 struct inode *inode = filp->f_dentry->d_inode;
983 loff_t newsize, saved_pos; 1046 loff_t newsize, saved_pos;
984 1047
985 mlog_entry("(0x%p, %u, '%.*s')\n", filp,
986 (unsigned int)nr_segs,
987 filp->f_dentry->d_name.len,
988 filp->f_dentry->d_name.name);
989
990 /* happy write of zero bytes */
991 if (iocb->ki_left == 0)
992 return 0;
993
994 if (!inode) {
995 mlog(0, "bad inode\n");
996 return -EIO;
997 }
998
999 mutex_lock(&inode->i_mutex);
1000 /* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */
1001 if (filp->f_flags & O_DIRECT) {
1002 have_alloc_sem = 1;
1003 down_read(&inode->i_alloc_sem);
1004 }
1005
1006 /* concurrent O_DIRECT writes are allowed */
1007 rw_level = (filp->f_flags & O_DIRECT) ? 0 : 1;
1008 ret = ocfs2_rw_lock(inode, rw_level);
1009 if (ret < 0) {
1010 rw_level = -1;
1011 mlog_errno(ret);
1012 goto out;
1013 }
1014
1015 /* 1048 /*
1016 * We sample i_size under a read level meta lock to see if our write 1049 * We sample i_size under a read level meta lock to see if our write
1017 * is extending the file, if it is we back off and get a write level 1050 * is extending the file, if it is we back off and get a write level
1018 * meta lock. 1051 * meta lock.
1019 */ 1052 */
1020 meta_level = (filp->f_flags & O_APPEND) ? 1 : 0;
1021 for(;;) { 1053 for(;;) {
1022 ret = ocfs2_meta_lock(inode, NULL, NULL, meta_level); 1054 ret = ocfs2_meta_lock(inode, NULL, meta_level);
1023 if (ret < 0) { 1055 if (ret < 0) {
1024 meta_level = -1; 1056 meta_level = -1;
1025 mlog_errno(ret); 1057 mlog_errno(ret);
@@ -1035,7 +1067,7 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
1035 * inode. There's also the dinode i_size state which 1067 * inode. There's also the dinode i_size state which
1036 * can be lost via setattr during extending writes (we 1068 * can be lost via setattr during extending writes (we
1037 * set inode->i_size at the end of a write. */ 1069 * set inode->i_size at the end of a write. */
1038 if (ocfs2_write_should_remove_suid(inode)) { 1070 if (should_remove_suid(dentry)) {
1039 if (meta_level == 0) { 1071 if (meta_level == 0) {
1040 ocfs2_meta_unlock(inode, meta_level); 1072 ocfs2_meta_unlock(inode, meta_level);
1041 meta_level = 1; 1073 meta_level = 1;
@@ -1045,19 +1077,19 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
1045 ret = ocfs2_write_remove_suid(inode); 1077 ret = ocfs2_write_remove_suid(inode);
1046 if (ret < 0) { 1078 if (ret < 0) {
1047 mlog_errno(ret); 1079 mlog_errno(ret);
1048 goto out; 1080 goto out_unlock;
1049 } 1081 }
1050 } 1082 }
1051 1083
1052 /* work on a copy of ppos until we're sure that we won't have 1084 /* work on a copy of ppos until we're sure that we won't have
1053 * to recalculate it due to relocking. */ 1085 * to recalculate it due to relocking. */
1054 if (filp->f_flags & O_APPEND) { 1086 if (appending) {
1055 saved_pos = i_size_read(inode); 1087 saved_pos = i_size_read(inode);
1056 mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos); 1088 mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos);
1057 } else { 1089 } else {
1058 saved_pos = iocb->ki_pos; 1090 saved_pos = *ppos;
1059 } 1091 }
1060 newsize = iocb->ki_left + saved_pos; 1092 newsize = count + saved_pos;
1061 1093
1062 mlog(0, "pos=%lld newsize=%lld cursize=%lld\n", 1094 mlog(0, "pos=%lld newsize=%lld cursize=%lld\n",
1063 (long long) saved_pos, (long long) newsize, 1095 (long long) saved_pos, (long long) newsize,
@@ -1090,19 +1122,66 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
1090 if (!clusters) 1122 if (!clusters)
1091 break; 1123 break;
1092 1124
1093 ret = ocfs2_extend_file(inode, NULL, newsize, iocb->ki_left); 1125 ret = ocfs2_extend_file(inode, NULL, newsize, count);
1094 if (ret < 0) { 1126 if (ret < 0) {
1095 if (ret != -ENOSPC) 1127 if (ret != -ENOSPC)
1096 mlog_errno(ret); 1128 mlog_errno(ret);
1097 goto out; 1129 goto out_unlock;
1098 } 1130 }
1099 break; 1131 break;
1100 } 1132 }
1101 1133
1102 /* ok, we're done with i_size and alloc work */ 1134 if (appending)
1103 iocb->ki_pos = saved_pos; 1135 *ppos = saved_pos;
1136
1137out_unlock:
1104 ocfs2_meta_unlock(inode, meta_level); 1138 ocfs2_meta_unlock(inode, meta_level);
1105 meta_level = -1; 1139
1140out:
1141 return ret;
1142}
1143
1144static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
1145 const struct iovec *iov,
1146 unsigned long nr_segs,
1147 loff_t pos)
1148{
1149 int ret, rw_level, have_alloc_sem = 0;
1150 struct file *filp = iocb->ki_filp;
1151 struct inode *inode = filp->f_path.dentry->d_inode;
1152 int appending = filp->f_flags & O_APPEND ? 1 : 0;
1153
1154 mlog_entry("(0x%p, %u, '%.*s')\n", filp,
1155 (unsigned int)nr_segs,
1156 filp->f_path.dentry->d_name.len,
1157 filp->f_path.dentry->d_name.name);
1158
1159 /* happy write of zero bytes */
1160 if (iocb->ki_left == 0)
1161 return 0;
1162
1163 mutex_lock(&inode->i_mutex);
1164 /* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */
1165 if (filp->f_flags & O_DIRECT) {
1166 have_alloc_sem = 1;
1167 down_read(&inode->i_alloc_sem);
1168 }
1169
1170 /* concurrent O_DIRECT writes are allowed */
1171 rw_level = (filp->f_flags & O_DIRECT) ? 0 : 1;
1172 ret = ocfs2_rw_lock(inode, rw_level);
1173 if (ret < 0) {
1174 rw_level = -1;
1175 mlog_errno(ret);
1176 goto out;
1177 }
1178
1179 ret = ocfs2_prepare_inode_for_write(filp->f_path.dentry, &iocb->ki_pos,
1180 iocb->ki_left, appending);
1181 if (ret < 0) {
1182 mlog_errno(ret);
1183 goto out;
1184 }
1106 1185
1107 /* communicate with ocfs2_dio_end_io */ 1186 /* communicate with ocfs2_dio_end_io */
1108 ocfs2_iocb_set_rw_locked(iocb); 1187 ocfs2_iocb_set_rw_locked(iocb);
@@ -1128,8 +1207,6 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
1128 } 1207 }
1129 1208
1130out: 1209out:
1131 if (meta_level != -1)
1132 ocfs2_meta_unlock(inode, meta_level);
1133 if (have_alloc_sem) 1210 if (have_alloc_sem)
1134 up_read(&inode->i_alloc_sem); 1211 up_read(&inode->i_alloc_sem);
1135 if (rw_level != -1) 1212 if (rw_level != -1)
@@ -1140,19 +1217,90 @@ out:
1140 return ret; 1217 return ret;
1141} 1218}
1142 1219
1220static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
1221 struct file *out,
1222 loff_t *ppos,
1223 size_t len,
1224 unsigned int flags)
1225{
1226 int ret;
1227 struct inode *inode = out->f_path.dentry->d_inode;
1228
1229 mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", out, pipe,
1230 (unsigned int)len,
1231 out->f_path.dentry->d_name.len,
1232 out->f_path.dentry->d_name.name);
1233
1234 inode_double_lock(inode, pipe->inode);
1235
1236 ret = ocfs2_rw_lock(inode, 1);
1237 if (ret < 0) {
1238 mlog_errno(ret);
1239 goto out;
1240 }
1241
1242 ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, ppos, len, 0);
1243 if (ret < 0) {
1244 mlog_errno(ret);
1245 goto out_unlock;
1246 }
1247
1248 /* ok, we're done with i_size and alloc work */
1249 ret = generic_file_splice_write_nolock(pipe, out, ppos, len, flags);
1250
1251out_unlock:
1252 ocfs2_rw_unlock(inode, 1);
1253out:
1254 inode_double_unlock(inode, pipe->inode);
1255
1256 mlog_exit(ret);
1257 return ret;
1258}
1259
1260static ssize_t ocfs2_file_splice_read(struct file *in,
1261 loff_t *ppos,
1262 struct pipe_inode_info *pipe,
1263 size_t len,
1264 unsigned int flags)
1265{
1266 int ret = 0;
1267 struct inode *inode = in->f_path.dentry->d_inode;
1268
1269 mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", in, pipe,
1270 (unsigned int)len,
1271 in->f_path.dentry->d_name.len,
1272 in->f_path.dentry->d_name.name);
1273
1274 /*
1275 * See the comment in ocfs2_file_aio_read()
1276 */
1277 ret = ocfs2_meta_lock(inode, NULL, 0);
1278 if (ret < 0) {
1279 mlog_errno(ret);
1280 goto bail;
1281 }
1282 ocfs2_meta_unlock(inode, 0);
1283
1284 ret = generic_file_splice_read(in, ppos, pipe, len, flags);
1285
1286bail:
1287 mlog_exit(ret);
1288 return ret;
1289}
1290
1143static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, 1291static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
1144 const struct iovec *iov, 1292 const struct iovec *iov,
1145 unsigned long nr_segs, 1293 unsigned long nr_segs,
1146 loff_t pos) 1294 loff_t pos)
1147{ 1295{
1148 int ret = 0, rw_level = -1, have_alloc_sem = 0; 1296 int ret = 0, rw_level = -1, have_alloc_sem = 0, lock_level = 0;
1149 struct file *filp = iocb->ki_filp; 1297 struct file *filp = iocb->ki_filp;
1150 struct inode *inode = filp->f_dentry->d_inode; 1298 struct inode *inode = filp->f_path.dentry->d_inode;
1151 1299
1152 mlog_entry("(0x%p, %u, '%.*s')\n", filp, 1300 mlog_entry("(0x%p, %u, '%.*s')\n", filp,
1153 (unsigned int)nr_segs, 1301 (unsigned int)nr_segs,
1154 filp->f_dentry->d_name.len, 1302 filp->f_path.dentry->d_name.len,
1155 filp->f_dentry->d_name.name); 1303 filp->f_path.dentry->d_name.name);
1156 1304
1157 if (!inode) { 1305 if (!inode) {
1158 ret = -EINVAL; 1306 ret = -EINVAL;
@@ -1187,12 +1335,12 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
1187 * like i_size. This allows the checks down below 1335 * like i_size. This allows the checks down below
1188 * generic_file_aio_read() a chance of actually working. 1336 * generic_file_aio_read() a chance of actually working.
1189 */ 1337 */
1190 ret = ocfs2_meta_lock(inode, NULL, NULL, 0); 1338 ret = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level);
1191 if (ret < 0) { 1339 if (ret < 0) {
1192 mlog_errno(ret); 1340 mlog_errno(ret);
1193 goto bail; 1341 goto bail;
1194 } 1342 }
1195 ocfs2_meta_unlock(inode, 0); 1343 ocfs2_meta_unlock(inode, lock_level);
1196 1344
1197 ret = generic_file_aio_read(iocb, iov, nr_segs, iocb->ki_pos); 1345 ret = generic_file_aio_read(iocb, iov, nr_segs, iocb->ki_pos);
1198 if (ret == -EINVAL) 1346 if (ret == -EINVAL)
@@ -1220,11 +1368,13 @@ bail:
1220struct inode_operations ocfs2_file_iops = { 1368struct inode_operations ocfs2_file_iops = {
1221 .setattr = ocfs2_setattr, 1369 .setattr = ocfs2_setattr,
1222 .getattr = ocfs2_getattr, 1370 .getattr = ocfs2_getattr,
1371 .permission = ocfs2_permission,
1223}; 1372};
1224 1373
1225struct inode_operations ocfs2_special_file_iops = { 1374struct inode_operations ocfs2_special_file_iops = {
1226 .setattr = ocfs2_setattr, 1375 .setattr = ocfs2_setattr,
1227 .getattr = ocfs2_getattr, 1376 .getattr = ocfs2_getattr,
1377 .permission = ocfs2_permission,
1228}; 1378};
1229 1379
1230const struct file_operations ocfs2_fops = { 1380const struct file_operations ocfs2_fops = {
@@ -1238,6 +1388,8 @@ const struct file_operations ocfs2_fops = {
1238 .aio_read = ocfs2_file_aio_read, 1388 .aio_read = ocfs2_file_aio_read,
1239 .aio_write = ocfs2_file_aio_write, 1389 .aio_write = ocfs2_file_aio_write,
1240 .ioctl = ocfs2_ioctl, 1390 .ioctl = ocfs2_ioctl,
1391 .splice_read = ocfs2_file_splice_read,
1392 .splice_write = ocfs2_file_splice_write,
1241}; 1393};
1242 1394
1243const struct file_operations ocfs2_dops = { 1395const struct file_operations ocfs2_dops = {
diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h
index 740c9e7ca599..601a453f18a8 100644
--- a/fs/ocfs2/file.h
+++ b/fs/ocfs2/file.h
@@ -41,17 +41,24 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
41 struct inode *inode, 41 struct inode *inode,
42 u32 clusters_to_add, 42 u32 clusters_to_add,
43 struct buffer_head *fe_bh, 43 struct buffer_head *fe_bh,
44 struct ocfs2_journal_handle *handle, 44 handle_t *handle,
45 struct ocfs2_alloc_context *data_ac, 45 struct ocfs2_alloc_context *data_ac,
46 struct ocfs2_alloc_context *meta_ac, 46 struct ocfs2_alloc_context *meta_ac,
47 enum ocfs2_alloc_restarted *reason); 47 enum ocfs2_alloc_restarted *reason);
48int ocfs2_setattr(struct dentry *dentry, struct iattr *attr); 48int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
49int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry, 49int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
50 struct kstat *stat); 50 struct kstat *stat);
51int ocfs2_permission(struct inode *inode, int mask,
52 struct nameidata *nd);
51 53
52int ocfs2_set_inode_size(struct ocfs2_journal_handle *handle, 54int ocfs2_set_inode_size(handle_t *handle,
53 struct inode *inode, 55 struct inode *inode,
54 struct buffer_head *fe_bh, 56 struct buffer_head *fe_bh,
55 u64 new_i_size); 57 u64 new_i_size);
56 58
59int ocfs2_should_update_atime(struct inode *inode,
60 struct vfsmount *vfsmnt);
61int ocfs2_update_inode_atime(struct inode *inode,
62 struct buffer_head *bh);
63
57#endif /* OCFS2_FILE_H */ 64#endif /* OCFS2_FILE_H */
diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c
index cbfd45a97a63..8fc52d6d0ce7 100644
--- a/fs/ocfs2/heartbeat.c
+++ b/fs/ocfs2/heartbeat.c
@@ -154,6 +154,9 @@ int ocfs2_register_hb_callbacks(struct ocfs2_super *osb)
154{ 154{
155 int status; 155 int status;
156 156
157 if (ocfs2_mount_local(osb))
158 return 0;
159
157 status = o2hb_register_callback(&osb->osb_hb_down); 160 status = o2hb_register_callback(&osb->osb_hb_down);
158 if (status < 0) { 161 if (status < 0) {
159 mlog_errno(status); 162 mlog_errno(status);
@@ -172,6 +175,9 @@ void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb)
172{ 175{
173 int status; 176 int status;
174 177
178 if (ocfs2_mount_local(osb))
179 return;
180
175 status = o2hb_unregister_callback(&osb->osb_hb_down); 181 status = o2hb_unregister_callback(&osb->osb_hb_down);
176 if (status < 0) 182 if (status < 0)
177 mlog_errno(status); 183 mlog_errno(status);
@@ -186,6 +192,9 @@ void ocfs2_stop_heartbeat(struct ocfs2_super *osb)
186 int ret; 192 int ret;
187 char *argv[5], *envp[3]; 193 char *argv[5], *envp[3];
188 194
195 if (ocfs2_mount_local(osb))
196 return;
197
189 if (!osb->uuid_str) { 198 if (!osb->uuid_str) {
190 /* This can happen if we don't get far enough in mount... */ 199 /* This can happen if we don't get far enough in mount... */
191 mlog(0, "No UUID with which to stop heartbeat!\n\n"); 200 mlog(0, "No UUID with which to stop heartbeat!\n\n");
diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
index 16e8e74dc966..e4d91493d7d7 100644
--- a/fs/ocfs2/inode.c
+++ b/fs/ocfs2/inode.c
@@ -360,7 +360,6 @@ int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
360 inode); 360 inode);
361 361
362 ocfs2_set_inode_flags(inode); 362 ocfs2_set_inode_flags(inode);
363 inode->i_flags |= S_NOATIME;
364 363
365 status = 0; 364 status = 0;
366bail: 365bail:
@@ -424,7 +423,8 @@ static int ocfs2_read_locked_inode(struct inode *inode,
424 * cluster lock before trusting anything anyway. 423 * cluster lock before trusting anything anyway.
425 */ 424 */
426 can_lock = !(args->fi_flags & OCFS2_FI_FLAG_SYSFILE) 425 can_lock = !(args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
427 && !(args->fi_flags & OCFS2_FI_FLAG_NOLOCK); 426 && !(args->fi_flags & OCFS2_FI_FLAG_NOLOCK)
427 && !ocfs2_mount_local(osb);
428 428
429 /* 429 /*
430 * To maintain backwards compatibility with older versions of 430 * To maintain backwards compatibility with older versions of
@@ -441,7 +441,7 @@ static int ocfs2_read_locked_inode(struct inode *inode,
441 generation, inode); 441 generation, inode);
442 442
443 if (can_lock) { 443 if (can_lock) {
444 status = ocfs2_meta_lock(inode, NULL, NULL, 0); 444 status = ocfs2_meta_lock(inode, NULL, 0);
445 if (status) { 445 if (status) {
446 make_bad_inode(inode); 446 make_bad_inode(inode);
447 mlog_errno(status); 447 mlog_errno(status);
@@ -512,7 +512,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
512 struct buffer_head *fe_bh) 512 struct buffer_head *fe_bh)
513{ 513{
514 int status = 0; 514 int status = 0;
515 struct ocfs2_journal_handle *handle = NULL; 515 handle_t *handle = NULL;
516 struct ocfs2_truncate_context *tc = NULL; 516 struct ocfs2_truncate_context *tc = NULL;
517 struct ocfs2_dinode *fe; 517 struct ocfs2_dinode *fe;
518 518
@@ -524,7 +524,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
524 if (!fe->i_clusters) 524 if (!fe->i_clusters)
525 goto bail; 525 goto bail;
526 526
527 handle = ocfs2_start_trans(osb, handle, OCFS2_INODE_UPDATE_CREDITS); 527 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
528 if (IS_ERR(handle)) { 528 if (IS_ERR(handle)) {
529 status = PTR_ERR(handle); 529 status = PTR_ERR(handle);
530 handle = NULL; 530 handle = NULL;
@@ -538,7 +538,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
538 goto bail; 538 goto bail;
539 } 539 }
540 540
541 ocfs2_commit_trans(handle); 541 ocfs2_commit_trans(osb, handle);
542 handle = NULL; 542 handle = NULL;
543 543
544 status = ocfs2_prepare_truncate(osb, inode, fe_bh, &tc); 544 status = ocfs2_prepare_truncate(osb, inode, fe_bh, &tc);
@@ -554,7 +554,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
554 } 554 }
555bail: 555bail:
556 if (handle) 556 if (handle)
557 ocfs2_commit_trans(handle); 557 ocfs2_commit_trans(osb, handle);
558 558
559 mlog_exit(status); 559 mlog_exit(status);
560 return status; 560 return status;
@@ -568,7 +568,7 @@ static int ocfs2_remove_inode(struct inode *inode,
568 int status; 568 int status;
569 struct inode *inode_alloc_inode = NULL; 569 struct inode *inode_alloc_inode = NULL;
570 struct buffer_head *inode_alloc_bh = NULL; 570 struct buffer_head *inode_alloc_bh = NULL;
571 struct ocfs2_journal_handle *handle; 571 handle_t *handle;
572 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 572 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
573 struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; 573 struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
574 574
@@ -582,7 +582,7 @@ static int ocfs2_remove_inode(struct inode *inode,
582 } 582 }
583 583
584 mutex_lock(&inode_alloc_inode->i_mutex); 584 mutex_lock(&inode_alloc_inode->i_mutex);
585 status = ocfs2_meta_lock(inode_alloc_inode, NULL, &inode_alloc_bh, 1); 585 status = ocfs2_meta_lock(inode_alloc_inode, &inode_alloc_bh, 1);
586 if (status < 0) { 586 if (status < 0) {
587 mutex_unlock(&inode_alloc_inode->i_mutex); 587 mutex_unlock(&inode_alloc_inode->i_mutex);
588 588
@@ -590,7 +590,7 @@ static int ocfs2_remove_inode(struct inode *inode,
590 goto bail; 590 goto bail;
591 } 591 }
592 592
593 handle = ocfs2_start_trans(osb, NULL, OCFS2_DELETE_INODE_CREDITS); 593 handle = ocfs2_start_trans(osb, OCFS2_DELETE_INODE_CREDITS);
594 if (IS_ERR(handle)) { 594 if (IS_ERR(handle)) {
595 status = PTR_ERR(handle); 595 status = PTR_ERR(handle);
596 mlog_errno(status); 596 mlog_errno(status);
@@ -629,7 +629,7 @@ static int ocfs2_remove_inode(struct inode *inode,
629 mlog_errno(status); 629 mlog_errno(status);
630 630
631bail_commit: 631bail_commit:
632 ocfs2_commit_trans(handle); 632 ocfs2_commit_trans(osb, handle);
633bail_unlock: 633bail_unlock:
634 ocfs2_meta_unlock(inode_alloc_inode, 1); 634 ocfs2_meta_unlock(inode_alloc_inode, 1);
635 mutex_unlock(&inode_alloc_inode->i_mutex); 635 mutex_unlock(&inode_alloc_inode->i_mutex);
@@ -705,7 +705,7 @@ static int ocfs2_wipe_inode(struct inode *inode,
705 * delete_inode operation. We do this now to avoid races with 705 * delete_inode operation. We do this now to avoid races with
706 * recovery completion on other nodes. */ 706 * recovery completion on other nodes. */
707 mutex_lock(&orphan_dir_inode->i_mutex); 707 mutex_lock(&orphan_dir_inode->i_mutex);
708 status = ocfs2_meta_lock(orphan_dir_inode, NULL, &orphan_dir_bh, 1); 708 status = ocfs2_meta_lock(orphan_dir_inode, &orphan_dir_bh, 1);
709 if (status < 0) { 709 if (status < 0) {
710 mutex_unlock(&orphan_dir_inode->i_mutex); 710 mutex_unlock(&orphan_dir_inode->i_mutex);
711 711
@@ -933,7 +933,7 @@ void ocfs2_delete_inode(struct inode *inode)
933 * allocation lock here as it won't be needed - nobody will 933 * allocation lock here as it won't be needed - nobody will
934 * have the file open. 934 * have the file open.
935 */ 935 */
936 status = ocfs2_meta_lock(inode, NULL, &di_bh, 1); 936 status = ocfs2_meta_lock(inode, &di_bh, 1);
937 if (status < 0) { 937 if (status < 0) {
938 if (status != -ENOENT) 938 if (status != -ENOENT)
939 mlog_errno(status); 939 mlog_errno(status);
@@ -1067,12 +1067,6 @@ void ocfs2_clear_inode(struct inode *inode)
1067 mlog_bug_on_msg(oi->ip_open_count, 1067 mlog_bug_on_msg(oi->ip_open_count,
1068 "Clear inode of %llu has open count %d\n", 1068 "Clear inode of %llu has open count %d\n",
1069 (unsigned long long)oi->ip_blkno, oi->ip_open_count); 1069 (unsigned long long)oi->ip_blkno, oi->ip_open_count);
1070 mlog_bug_on_msg(!list_empty(&oi->ip_handle_list),
1071 "Clear inode of %llu has non empty handle list\n",
1072 (unsigned long long)oi->ip_blkno);
1073 mlog_bug_on_msg(oi->ip_handle,
1074 "Clear inode of %llu has non empty handle pointer\n",
1075 (unsigned long long)oi->ip_blkno);
1076 1070
1077 /* Clear all other flags. */ 1071 /* Clear all other flags. */
1078 oi->ip_flags = OCFS2_INODE_CACHE_INLINE; 1072 oi->ip_flags = OCFS2_INODE_CACHE_INLINE;
@@ -1186,7 +1180,7 @@ int ocfs2_inode_revalidate(struct dentry *dentry)
1186 1180
1187 /* Let ocfs2_meta_lock do the work of updating our struct 1181 /* Let ocfs2_meta_lock do the work of updating our struct
1188 * inode for us. */ 1182 * inode for us. */
1189 status = ocfs2_meta_lock(inode, NULL, NULL, 0); 1183 status = ocfs2_meta_lock(inode, NULL, 0);
1190 if (status < 0) { 1184 if (status < 0) {
1191 if (status != -ENOENT) 1185 if (status != -ENOENT)
1192 mlog_errno(status); 1186 mlog_errno(status);
@@ -1204,7 +1198,7 @@ bail:
1204 * struct inode. 1198 * struct inode.
1205 * Only takes ip_lock. 1199 * Only takes ip_lock.
1206 */ 1200 */
1207int ocfs2_mark_inode_dirty(struct ocfs2_journal_handle *handle, 1201int ocfs2_mark_inode_dirty(handle_t *handle,
1208 struct inode *inode, 1202 struct inode *inode,
1209 struct buffer_head *bh) 1203 struct buffer_head *bh)
1210{ 1204{
diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h
index 9957810fdf85..1a7dd2945b34 100644
--- a/fs/ocfs2/inode.h
+++ b/fs/ocfs2/inode.h
@@ -48,13 +48,6 @@ struct ocfs2_inode_info
48 48
49 struct mutex ip_io_mutex; 49 struct mutex ip_io_mutex;
50 50
51 /* Used by the journalling code to attach an inode to a
52 * handle. These are protected by ip_io_mutex in order to lock
53 * out other I/O to the inode until we either commit or
54 * abort. */
55 struct list_head ip_handle_list;
56 struct ocfs2_journal_handle *ip_handle;
57
58 u32 ip_flags; /* see below */ 51 u32 ip_flags; /* see below */
59 u32 ip_attr; /* inode attributes */ 52 u32 ip_attr; /* inode attributes */
60 53
@@ -113,7 +106,7 @@ static inline struct ocfs2_inode_info *OCFS2_I(struct inode *inode)
113#define INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags & OCFS2_INODE_JOURNAL) 106#define INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags & OCFS2_INODE_JOURNAL)
114#define SET_INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags |= OCFS2_INODE_JOURNAL) 107#define SET_INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags |= OCFS2_INODE_JOURNAL)
115 108
116extern kmem_cache_t *ocfs2_inode_cache; 109extern struct kmem_cache *ocfs2_inode_cache;
117 110
118extern const struct address_space_operations ocfs2_aops; 111extern const struct address_space_operations ocfs2_aops;
119 112
@@ -143,7 +136,7 @@ ssize_t ocfs2_rw_direct(int rw, struct file *filp, char *buf,
143void ocfs2_sync_blockdev(struct super_block *sb); 136void ocfs2_sync_blockdev(struct super_block *sb);
144void ocfs2_refresh_inode(struct inode *inode, 137void ocfs2_refresh_inode(struct inode *inode,
145 struct ocfs2_dinode *fe); 138 struct ocfs2_dinode *fe);
146int ocfs2_mark_inode_dirty(struct ocfs2_journal_handle *handle, 139int ocfs2_mark_inode_dirty(handle_t *handle,
147 struct inode *inode, 140 struct inode *inode,
148 struct buffer_head *bh); 141 struct buffer_head *bh);
149int ocfs2_aio_read(struct file *file, struct kiocb *req, struct iocb *iocb); 142int ocfs2_aio_read(struct file *file, struct kiocb *req, struct iocb *iocb);
diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
index 3663cef80689..4768be5f3086 100644
--- a/fs/ocfs2/ioctl.c
+++ b/fs/ocfs2/ioctl.c
@@ -26,7 +26,7 @@ static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
26{ 26{
27 int status; 27 int status;
28 28
29 status = ocfs2_meta_lock(inode, NULL, NULL, 0); 29 status = ocfs2_meta_lock(inode, NULL, 0);
30 if (status < 0) { 30 if (status < 0) {
31 mlog_errno(status); 31 mlog_errno(status);
32 return status; 32 return status;
@@ -43,14 +43,14 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
43{ 43{
44 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode); 44 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
45 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 45 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
46 struct ocfs2_journal_handle *handle = NULL; 46 handle_t *handle = NULL;
47 struct buffer_head *bh = NULL; 47 struct buffer_head *bh = NULL;
48 unsigned oldflags; 48 unsigned oldflags;
49 int status; 49 int status;
50 50
51 mutex_lock(&inode->i_mutex); 51 mutex_lock(&inode->i_mutex);
52 52
53 status = ocfs2_meta_lock(inode, NULL, &bh, 1); 53 status = ocfs2_meta_lock(inode, &bh, 1);
54 if (status < 0) { 54 if (status < 0) {
55 mlog_errno(status); 55 mlog_errno(status);
56 goto bail; 56 goto bail;
@@ -67,7 +67,7 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
67 if (!S_ISDIR(inode->i_mode)) 67 if (!S_ISDIR(inode->i_mode))
68 flags &= ~OCFS2_DIRSYNC_FL; 68 flags &= ~OCFS2_DIRSYNC_FL;
69 69
70 handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS); 70 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
71 if (IS_ERR(handle)) { 71 if (IS_ERR(handle)) {
72 status = PTR_ERR(handle); 72 status = PTR_ERR(handle);
73 mlog_errno(status); 73 mlog_errno(status);
@@ -96,7 +96,7 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
96 if (status < 0) 96 if (status < 0)
97 mlog_errno(status); 97 mlog_errno(status);
98 98
99 ocfs2_commit_trans(handle); 99 ocfs2_commit_trans(osb, handle);
100bail_unlock: 100bail_unlock:
101 ocfs2_meta_unlock(inode, 1); 101 ocfs2_meta_unlock(inode, 1);
102bail: 102bail:
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index fd9734def551..825cb0ae1b4c 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -57,9 +57,6 @@ static int ocfs2_recover_node(struct ocfs2_super *osb,
57static int __ocfs2_recovery_thread(void *arg); 57static int __ocfs2_recovery_thread(void *arg);
58static int ocfs2_commit_cache(struct ocfs2_super *osb); 58static int ocfs2_commit_cache(struct ocfs2_super *osb);
59static int ocfs2_wait_on_mount(struct ocfs2_super *osb); 59static int ocfs2_wait_on_mount(struct ocfs2_super *osb);
60static void ocfs2_handle_cleanup_locks(struct ocfs2_journal *journal,
61 struct ocfs2_journal_handle *handle);
62static void ocfs2_commit_unstarted_handle(struct ocfs2_journal_handle *handle);
63static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb, 60static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
64 int dirty); 61 int dirty);
65static int ocfs2_trylock_journal(struct ocfs2_super *osb, 62static int ocfs2_trylock_journal(struct ocfs2_super *osb,
@@ -113,46 +110,18 @@ finally:
113 return status; 110 return status;
114} 111}
115 112
116struct ocfs2_journal_handle *ocfs2_alloc_handle(struct ocfs2_super *osb)
117{
118 struct ocfs2_journal_handle *retval = NULL;
119
120 retval = kcalloc(1, sizeof(*retval), GFP_NOFS);
121 if (!retval) {
122 mlog(ML_ERROR, "Failed to allocate memory for journal "
123 "handle!\n");
124 return NULL;
125 }
126
127 retval->max_buffs = 0;
128 retval->num_locks = 0;
129 retval->k_handle = NULL;
130
131 INIT_LIST_HEAD(&retval->locks);
132 INIT_LIST_HEAD(&retval->inode_list);
133 retval->journal = osb->journal;
134
135 return retval;
136}
137
138/* pass it NULL and it will allocate a new handle object for you. If 113/* pass it NULL and it will allocate a new handle object for you. If
139 * you pass it a handle however, it may still return error, in which 114 * you pass it a handle however, it may still return error, in which
140 * case it has free'd the passed handle for you. */ 115 * case it has free'd the passed handle for you. */
141struct ocfs2_journal_handle *ocfs2_start_trans(struct ocfs2_super *osb, 116handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int max_buffs)
142 struct ocfs2_journal_handle *handle,
143 int max_buffs)
144{ 117{
145 int ret;
146 journal_t *journal = osb->journal->j_journal; 118 journal_t *journal = osb->journal->j_journal;
147 119 handle_t *handle;
148 mlog_entry("(max_buffs = %d)\n", max_buffs);
149 120
150 BUG_ON(!osb || !osb->journal->j_journal); 121 BUG_ON(!osb || !osb->journal->j_journal);
151 122
152 if (ocfs2_is_hard_readonly(osb)) { 123 if (ocfs2_is_hard_readonly(osb))
153 ret = -EROFS; 124 return ERR_PTR(-EROFS);
154 goto done_free;
155 }
156 125
157 BUG_ON(osb->journal->j_state == OCFS2_JOURNAL_FREE); 126 BUG_ON(osb->journal->j_state == OCFS2_JOURNAL_FREE);
158 BUG_ON(max_buffs <= 0); 127 BUG_ON(max_buffs <= 0);
@@ -163,154 +132,41 @@ struct ocfs2_journal_handle *ocfs2_start_trans(struct ocfs2_super *osb,
163 BUG(); 132 BUG();
164 } 133 }
165 134
166 if (!handle)
167 handle = ocfs2_alloc_handle(osb);
168 if (!handle) {
169 ret = -ENOMEM;
170 mlog(ML_ERROR, "Failed to allocate memory for journal "
171 "handle!\n");
172 goto done_free;
173 }
174
175 handle->max_buffs = max_buffs;
176
177 down_read(&osb->journal->j_trans_barrier); 135 down_read(&osb->journal->j_trans_barrier);
178 136
179 /* actually start the transaction now */ 137 handle = journal_start(journal, max_buffs);
180 handle->k_handle = journal_start(journal, max_buffs); 138 if (IS_ERR(handle)) {
181 if (IS_ERR(handle->k_handle)) {
182 up_read(&osb->journal->j_trans_barrier); 139 up_read(&osb->journal->j_trans_barrier);
183 140
184 ret = PTR_ERR(handle->k_handle); 141 mlog_errno(PTR_ERR(handle));
185 handle->k_handle = NULL;
186 mlog_errno(ret);
187 142
188 if (is_journal_aborted(journal)) { 143 if (is_journal_aborted(journal)) {
189 ocfs2_abort(osb->sb, "Detected aborted journal"); 144 ocfs2_abort(osb->sb, "Detected aborted journal");
190 ret = -EROFS; 145 handle = ERR_PTR(-EROFS);
191 } 146 }
192 goto done_free; 147 } else {
148 if (!ocfs2_mount_local(osb))
149 atomic_inc(&(osb->journal->j_num_trans));
193 } 150 }
194 151
195 atomic_inc(&(osb->journal->j_num_trans));
196 handle->flags |= OCFS2_HANDLE_STARTED;
197
198 mlog_exit_ptr(handle);
199 return handle; 152 return handle;
200
201done_free:
202 if (handle)
203 ocfs2_commit_unstarted_handle(handle); /* will kfree handle */
204
205 mlog_exit(ret);
206 return ERR_PTR(ret);
207}
208
209void ocfs2_handle_add_inode(struct ocfs2_journal_handle *handle,
210 struct inode *inode)
211{
212 BUG_ON(!handle);
213 BUG_ON(!inode);
214
215 atomic_inc(&inode->i_count);
216
217 /* we're obviously changing it... */
218 mutex_lock(&inode->i_mutex);
219
220 /* sanity check */
221 BUG_ON(OCFS2_I(inode)->ip_handle);
222 BUG_ON(!list_empty(&OCFS2_I(inode)->ip_handle_list));
223
224 OCFS2_I(inode)->ip_handle = handle;
225 list_move_tail(&(OCFS2_I(inode)->ip_handle_list), &(handle->inode_list));
226}
227
228static void ocfs2_handle_unlock_inodes(struct ocfs2_journal_handle *handle)
229{
230 struct list_head *p, *n;
231 struct inode *inode;
232 struct ocfs2_inode_info *oi;
233
234 list_for_each_safe(p, n, &handle->inode_list) {
235 oi = list_entry(p, struct ocfs2_inode_info,
236 ip_handle_list);
237 inode = &oi->vfs_inode;
238
239 OCFS2_I(inode)->ip_handle = NULL;
240 list_del_init(&OCFS2_I(inode)->ip_handle_list);
241
242 mutex_unlock(&inode->i_mutex);
243 iput(inode);
244 }
245} 153}
246 154
247/* This is trivial so we do it out of the main commit 155int ocfs2_commit_trans(struct ocfs2_super *osb,
248 * paths. Beware, it can be called from start_trans too! */ 156 handle_t *handle)
249static void ocfs2_commit_unstarted_handle(struct ocfs2_journal_handle *handle)
250{ 157{
251 mlog_entry_void(); 158 int ret;
252 159 struct ocfs2_journal *journal = osb->journal;
253 BUG_ON(handle->flags & OCFS2_HANDLE_STARTED);
254
255 ocfs2_handle_unlock_inodes(handle);
256 /* You are allowed to add journal locks before the transaction
257 * has started. */
258 ocfs2_handle_cleanup_locks(handle->journal, handle);
259
260 kfree(handle);
261
262 mlog_exit_void();
263}
264
265void ocfs2_commit_trans(struct ocfs2_journal_handle *handle)
266{
267 handle_t *jbd_handle;
268 int retval;
269 struct ocfs2_journal *journal = handle->journal;
270
271 mlog_entry_void();
272 160
273 BUG_ON(!handle); 161 BUG_ON(!handle);
274 162
275 if (!(handle->flags & OCFS2_HANDLE_STARTED)) { 163 ret = journal_stop(handle);
276 ocfs2_commit_unstarted_handle(handle); 164 if (ret < 0)
277 mlog_exit_void(); 165 mlog_errno(ret);
278 return;
279 }
280
281 /* release inode semaphores we took during this transaction */
282 ocfs2_handle_unlock_inodes(handle);
283
284 /* ocfs2_extend_trans may have had to call journal_restart
285 * which will always commit the transaction, but may return
286 * error for any number of reasons. If this is the case, we
287 * clear k_handle as it's not valid any more. */
288 if (handle->k_handle) {
289 jbd_handle = handle->k_handle;
290
291 if (handle->flags & OCFS2_HANDLE_SYNC)
292 jbd_handle->h_sync = 1;
293 else
294 jbd_handle->h_sync = 0;
295
296 /* actually stop the transaction. if we've set h_sync,
297 * it'll have been committed when we return */
298 retval = journal_stop(jbd_handle);
299 if (retval < 0) {
300 mlog_errno(retval);
301 mlog(ML_ERROR, "Could not commit transaction\n");
302 BUG();
303 }
304
305 handle->k_handle = NULL; /* it's been free'd in journal_stop */
306 }
307
308 ocfs2_handle_cleanup_locks(journal, handle);
309 166
310 up_read(&journal->j_trans_barrier); 167 up_read(&journal->j_trans_barrier);
311 168
312 kfree(handle); 169 return ret;
313 mlog_exit_void();
314} 170}
315 171
316/* 172/*
@@ -326,20 +182,18 @@ void ocfs2_commit_trans(struct ocfs2_journal_handle *handle)
326 * good because transaction ids haven't yet been recorded on the 182 * good because transaction ids haven't yet been recorded on the
327 * cluster locks associated with this handle. 183 * cluster locks associated with this handle.
328 */ 184 */
329int ocfs2_extend_trans(struct ocfs2_journal_handle *handle, 185int ocfs2_extend_trans(handle_t *handle, int nblocks)
330 int nblocks)
331{ 186{
332 int status; 187 int status;
333 188
334 BUG_ON(!handle); 189 BUG_ON(!handle);
335 BUG_ON(!(handle->flags & OCFS2_HANDLE_STARTED));
336 BUG_ON(!nblocks); 190 BUG_ON(!nblocks);
337 191
338 mlog_entry_void(); 192 mlog_entry_void();
339 193
340 mlog(0, "Trying to extend transaction by %d blocks\n", nblocks); 194 mlog(0, "Trying to extend transaction by %d blocks\n", nblocks);
341 195
342 status = journal_extend(handle->k_handle, nblocks); 196 status = journal_extend(handle, nblocks);
343 if (status < 0) { 197 if (status < 0) {
344 mlog_errno(status); 198 mlog_errno(status);
345 goto bail; 199 goto bail;
@@ -347,15 +201,12 @@ int ocfs2_extend_trans(struct ocfs2_journal_handle *handle,
347 201
348 if (status > 0) { 202 if (status > 0) {
349 mlog(0, "journal_extend failed, trying journal_restart\n"); 203 mlog(0, "journal_extend failed, trying journal_restart\n");
350 status = journal_restart(handle->k_handle, nblocks); 204 status = journal_restart(handle, nblocks);
351 if (status < 0) { 205 if (status < 0) {
352 handle->k_handle = NULL;
353 mlog_errno(status); 206 mlog_errno(status);
354 goto bail; 207 goto bail;
355 } 208 }
356 handle->max_buffs = nblocks; 209 }
357 } else
358 handle->max_buffs += nblocks;
359 210
360 status = 0; 211 status = 0;
361bail: 212bail:
@@ -364,7 +215,7 @@ bail:
364 return status; 215 return status;
365} 216}
366 217
367int ocfs2_journal_access(struct ocfs2_journal_handle *handle, 218int ocfs2_journal_access(handle_t *handle,
368 struct inode *inode, 219 struct inode *inode,
369 struct buffer_head *bh, 220 struct buffer_head *bh,
370 int type) 221 int type)
@@ -374,7 +225,6 @@ int ocfs2_journal_access(struct ocfs2_journal_handle *handle,
374 BUG_ON(!inode); 225 BUG_ON(!inode);
375 BUG_ON(!handle); 226 BUG_ON(!handle);
376 BUG_ON(!bh); 227 BUG_ON(!bh);
377 BUG_ON(!(handle->flags & OCFS2_HANDLE_STARTED));
378 228
379 mlog_entry("bh->b_blocknr=%llu, type=%d (\"%s\"), bh->b_size = %zu\n", 229 mlog_entry("bh->b_blocknr=%llu, type=%d (\"%s\"), bh->b_size = %zu\n",
380 (unsigned long long)bh->b_blocknr, type, 230 (unsigned long long)bh->b_blocknr, type,
@@ -403,11 +253,11 @@ int ocfs2_journal_access(struct ocfs2_journal_handle *handle,
403 switch (type) { 253 switch (type) {
404 case OCFS2_JOURNAL_ACCESS_CREATE: 254 case OCFS2_JOURNAL_ACCESS_CREATE:
405 case OCFS2_JOURNAL_ACCESS_WRITE: 255 case OCFS2_JOURNAL_ACCESS_WRITE:
406 status = journal_get_write_access(handle->k_handle, bh); 256 status = journal_get_write_access(handle, bh);
407 break; 257 break;
408 258
409 case OCFS2_JOURNAL_ACCESS_UNDO: 259 case OCFS2_JOURNAL_ACCESS_UNDO:
410 status = journal_get_undo_access(handle->k_handle, bh); 260 status = journal_get_undo_access(handle, bh);
411 break; 261 break;
412 262
413 default: 263 default:
@@ -424,17 +274,15 @@ int ocfs2_journal_access(struct ocfs2_journal_handle *handle,
424 return status; 274 return status;
425} 275}
426 276
427int ocfs2_journal_dirty(struct ocfs2_journal_handle *handle, 277int ocfs2_journal_dirty(handle_t *handle,
428 struct buffer_head *bh) 278 struct buffer_head *bh)
429{ 279{
430 int status; 280 int status;
431 281
432 BUG_ON(!(handle->flags & OCFS2_HANDLE_STARTED));
433
434 mlog_entry("(bh->b_blocknr=%llu)\n", 282 mlog_entry("(bh->b_blocknr=%llu)\n",
435 (unsigned long long)bh->b_blocknr); 283 (unsigned long long)bh->b_blocknr);
436 284
437 status = journal_dirty_metadata(handle->k_handle, bh); 285 status = journal_dirty_metadata(handle, bh);
438 if (status < 0) 286 if (status < 0)
439 mlog(ML_ERROR, "Could not dirty metadata buffer. " 287 mlog(ML_ERROR, "Could not dirty metadata buffer. "
440 "(bh->b_blocknr=%llu)\n", 288 "(bh->b_blocknr=%llu)\n",
@@ -456,59 +304,6 @@ int ocfs2_journal_dirty_data(handle_t *handle,
456 return err; 304 return err;
457} 305}
458 306
459/* We always assume you're adding a metadata lock at level 'ex' */
460int ocfs2_handle_add_lock(struct ocfs2_journal_handle *handle,
461 struct inode *inode)
462{
463 int status;
464 struct ocfs2_journal_lock *lock;
465
466 BUG_ON(!inode);
467
468 lock = kmem_cache_alloc(ocfs2_lock_cache, GFP_NOFS);
469 if (!lock) {
470 status = -ENOMEM;
471 mlog_errno(-ENOMEM);
472 goto bail;
473 }
474
475 if (!igrab(inode))
476 BUG();
477 lock->jl_inode = inode;
478
479 list_add_tail(&(lock->jl_lock_list), &(handle->locks));
480 handle->num_locks++;
481
482 status = 0;
483bail:
484 mlog_exit(status);
485 return status;
486}
487
488static void ocfs2_handle_cleanup_locks(struct ocfs2_journal *journal,
489 struct ocfs2_journal_handle *handle)
490{
491 struct list_head *p, *n;
492 struct ocfs2_journal_lock *lock;
493 struct inode *inode;
494
495 list_for_each_safe(p, n, &(handle->locks)) {
496 lock = list_entry(p, struct ocfs2_journal_lock,
497 jl_lock_list);
498 list_del(&lock->jl_lock_list);
499 handle->num_locks--;
500
501 inode = lock->jl_inode;
502 ocfs2_meta_unlock(inode, 1);
503 if (atomic_read(&inode->i_count) == 1)
504 mlog(ML_ERROR,
505 "Inode %llu, I'm doing a last iput for!",
506 (unsigned long long)OCFS2_I(inode)->ip_blkno);
507 iput(inode);
508 kmem_cache_free(ocfs2_lock_cache, lock);
509 }
510}
511
512#define OCFS2_DEFAULT_COMMIT_INTERVAL (HZ * 5) 307#define OCFS2_DEFAULT_COMMIT_INTERVAL (HZ * 5)
513 308
514void ocfs2_set_journal_params(struct ocfs2_super *osb) 309void ocfs2_set_journal_params(struct ocfs2_super *osb)
@@ -562,8 +357,7 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty)
562 /* Skip recovery waits here - journal inode metadata never 357 /* Skip recovery waits here - journal inode metadata never
563 * changes in a live cluster so it can be considered an 358 * changes in a live cluster so it can be considered an
564 * exception to the rule. */ 359 * exception to the rule. */
565 status = ocfs2_meta_lock_full(inode, NULL, &bh, 1, 360 status = ocfs2_meta_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY);
566 OCFS2_META_LOCK_RECOVERY);
567 if (status < 0) { 361 if (status < 0) {
568 if (status != -ERESTARTSYS) 362 if (status != -ERESTARTSYS)
569 mlog(ML_ERROR, "Could not get lock on journal!\n"); 363 mlog(ML_ERROR, "Could not get lock on journal!\n");
@@ -715,9 +509,23 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
715 509
716 BUG_ON(atomic_read(&(osb->journal->j_num_trans)) != 0); 510 BUG_ON(atomic_read(&(osb->journal->j_num_trans)) != 0);
717 511
718 status = ocfs2_journal_toggle_dirty(osb, 0); 512 if (ocfs2_mount_local(osb)) {
719 if (status < 0) 513 journal_lock_updates(journal->j_journal);
720 mlog_errno(status); 514 status = journal_flush(journal->j_journal);
515 journal_unlock_updates(journal->j_journal);
516 if (status < 0)
517 mlog_errno(status);
518 }
519
520 if (status == 0) {
521 /*
522 * Do not toggle if flush was unsuccessful otherwise
523 * will leave dirty metadata in a "clean" journal
524 */
525 status = ocfs2_journal_toggle_dirty(osb, 0);
526 if (status < 0)
527 mlog_errno(status);
528 }
721 529
722 /* Shutdown the kernel journal system */ 530 /* Shutdown the kernel journal system */
723 journal_destroy(journal->j_journal); 531 journal_destroy(journal->j_journal);
@@ -757,7 +565,7 @@ static void ocfs2_clear_journal_error(struct super_block *sb,
757 } 565 }
758} 566}
759 567
760int ocfs2_journal_load(struct ocfs2_journal *journal) 568int ocfs2_journal_load(struct ocfs2_journal *journal, int local)
761{ 569{
762 int status = 0; 570 int status = 0;
763 struct ocfs2_super *osb; 571 struct ocfs2_super *osb;
@@ -784,14 +592,18 @@ int ocfs2_journal_load(struct ocfs2_journal *journal)
784 } 592 }
785 593
786 /* Launch the commit thread */ 594 /* Launch the commit thread */
787 osb->commit_task = kthread_run(ocfs2_commit_thread, osb, "ocfs2cmt"); 595 if (!local) {
788 if (IS_ERR(osb->commit_task)) { 596 osb->commit_task = kthread_run(ocfs2_commit_thread, osb,
789 status = PTR_ERR(osb->commit_task); 597 "ocfs2cmt");
598 if (IS_ERR(osb->commit_task)) {
599 status = PTR_ERR(osb->commit_task);
600 osb->commit_task = NULL;
601 mlog(ML_ERROR, "unable to launch ocfs2commit thread, "
602 "error=%d", status);
603 goto done;
604 }
605 } else
790 osb->commit_task = NULL; 606 osb->commit_task = NULL;
791 mlog(ML_ERROR, "unable to launch ocfs2commit thread, error=%d",
792 status);
793 goto done;
794 }
795 607
796done: 608done:
797 mlog_exit(status); 609 mlog_exit(status);
@@ -911,11 +723,12 @@ struct ocfs2_la_recovery_item {
911 * NOTE: This function can and will sleep on recovery of other nodes 723 * NOTE: This function can and will sleep on recovery of other nodes
912 * during cluster locking, just like any other ocfs2 process. 724 * during cluster locking, just like any other ocfs2 process.
913 */ 725 */
914void ocfs2_complete_recovery(void *data) 726void ocfs2_complete_recovery(struct work_struct *work)
915{ 727{
916 int ret; 728 int ret;
917 struct ocfs2_super *osb = data; 729 struct ocfs2_journal *journal =
918 struct ocfs2_journal *journal = osb->journal; 730 container_of(work, struct ocfs2_journal, j_recovery_work);
731 struct ocfs2_super *osb = journal->j_osb;
919 struct ocfs2_dinode *la_dinode, *tl_dinode; 732 struct ocfs2_dinode *la_dinode, *tl_dinode;
920 struct ocfs2_la_recovery_item *item; 733 struct ocfs2_la_recovery_item *item;
921 struct list_head *p, *n; 734 struct list_head *p, *n;
@@ -1160,8 +973,7 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb,
1160 } 973 }
1161 SET_INODE_JOURNAL(inode); 974 SET_INODE_JOURNAL(inode);
1162 975
1163 status = ocfs2_meta_lock_full(inode, NULL, &bh, 1, 976 status = ocfs2_meta_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY);
1164 OCFS2_META_LOCK_RECOVERY);
1165 if (status < 0) { 977 if (status < 0) {
1166 mlog(0, "status returned from ocfs2_meta_lock=%d\n", status); 978 mlog(0, "status returned from ocfs2_meta_lock=%d\n", status);
1167 if (status != -ERESTARTSYS) 979 if (status != -ERESTARTSYS)
@@ -1350,7 +1162,7 @@ static int ocfs2_trylock_journal(struct ocfs2_super *osb,
1350 SET_INODE_JOURNAL(inode); 1162 SET_INODE_JOURNAL(inode);
1351 1163
1352 flags = OCFS2_META_LOCK_RECOVERY | OCFS2_META_LOCK_NOQUEUE; 1164 flags = OCFS2_META_LOCK_RECOVERY | OCFS2_META_LOCK_NOQUEUE;
1353 status = ocfs2_meta_lock_full(inode, NULL, NULL, 1, flags); 1165 status = ocfs2_meta_lock_full(inode, NULL, 1, flags);
1354 if (status < 0) { 1166 if (status < 0) {
1355 if (status != -EAGAIN) 1167 if (status != -EAGAIN)
1356 mlog_errno(status); 1168 mlog_errno(status);
@@ -1433,7 +1245,7 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb,
1433 } 1245 }
1434 1246
1435 mutex_lock(&orphan_dir_inode->i_mutex); 1247 mutex_lock(&orphan_dir_inode->i_mutex);
1436 status = ocfs2_meta_lock(orphan_dir_inode, NULL, NULL, 0); 1248 status = ocfs2_meta_lock(orphan_dir_inode, NULL, 0);
1437 if (status < 0) { 1249 if (status < 0) {
1438 mlog_errno(status); 1250 mlog_errno(status);
1439 goto out; 1251 goto out;
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index 2f3a6acdac45..e1216364d191 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -37,7 +37,6 @@ enum ocfs2_journal_state {
37 37
38struct ocfs2_super; 38struct ocfs2_super;
39struct ocfs2_dinode; 39struct ocfs2_dinode;
40struct ocfs2_journal_handle;
41 40
42struct ocfs2_journal { 41struct ocfs2_journal {
43 enum ocfs2_journal_state j_state; /* Journals current state */ 42 enum ocfs2_journal_state j_state; /* Journals current state */
@@ -133,46 +132,8 @@ static inline void ocfs2_inode_set_new(struct ocfs2_super *osb,
133 spin_unlock(&trans_inc_lock); 132 spin_unlock(&trans_inc_lock);
134} 133}
135 134
136extern kmem_cache_t *ocfs2_lock_cache;
137
138struct ocfs2_journal_lock {
139 struct inode *jl_inode;
140 struct list_head jl_lock_list;
141};
142
143struct ocfs2_journal_handle {
144 handle_t *k_handle; /* kernel handle. */
145 struct ocfs2_journal *journal;
146 u32 flags; /* see flags below. */
147 int max_buffs; /* Buffs reserved by this handle */
148
149 /* The following two fields are for ocfs2_handle_add_lock */
150 int num_locks;
151 struct list_head locks; /* A bunch of locks to
152 * release on commit. This
153 * should be a list_head */
154
155 struct list_head inode_list;
156};
157
158#define OCFS2_HANDLE_STARTED 1
159/* should we sync-commit this handle? */
160#define OCFS2_HANDLE_SYNC 2
161static inline int ocfs2_handle_started(struct ocfs2_journal_handle *handle)
162{
163 return handle->flags & OCFS2_HANDLE_STARTED;
164}
165
166static inline void ocfs2_handle_set_sync(struct ocfs2_journal_handle *handle, int sync)
167{
168 if (sync)
169 handle->flags |= OCFS2_HANDLE_SYNC;
170 else
171 handle->flags &= ~OCFS2_HANDLE_SYNC;
172}
173
174/* Exported only for the journal struct init code in super.c. Do not call. */ 135/* Exported only for the journal struct init code in super.c. Do not call. */
175void ocfs2_complete_recovery(void *data); 136void ocfs2_complete_recovery(struct work_struct *work);
176 137
177/* 138/*
178 * Journal Control: 139 * Journal Control:
@@ -196,7 +157,7 @@ int ocfs2_journal_init(struct ocfs2_journal *journal,
196void ocfs2_journal_shutdown(struct ocfs2_super *osb); 157void ocfs2_journal_shutdown(struct ocfs2_super *osb);
197int ocfs2_journal_wipe(struct ocfs2_journal *journal, 158int ocfs2_journal_wipe(struct ocfs2_journal *journal,
198 int full); 159 int full);
199int ocfs2_journal_load(struct ocfs2_journal *journal); 160int ocfs2_journal_load(struct ocfs2_journal *journal, int local);
200int ocfs2_check_journals_nolocks(struct ocfs2_super *osb); 161int ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
201void ocfs2_recovery_thread(struct ocfs2_super *osb, 162void ocfs2_recovery_thread(struct ocfs2_super *osb,
202 int node_num); 163 int node_num);
@@ -213,6 +174,9 @@ static inline void ocfs2_checkpoint_inode(struct inode *inode)
213{ 174{
214 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 175 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
215 176
177 if (ocfs2_mount_local(osb))
178 return;
179
216 if (!ocfs2_inode_fully_checkpointed(inode)) { 180 if (!ocfs2_inode_fully_checkpointed(inode)) {
217 /* WARNING: This only kicks off a single 181 /* WARNING: This only kicks off a single
218 * checkpoint. If someone races you and adds more 182 * checkpoint. If someone races you and adds more
@@ -231,15 +195,14 @@ static inline void ocfs2_checkpoint_inode(struct inode *inode)
231 * Transaction Handling: 195 * Transaction Handling:
232 * Manage the lifetime of a transaction handle. 196 * Manage the lifetime of a transaction handle.
233 * 197 *
234 * ocfs2_alloc_handle - Only allocate a handle so we can start putting
235 * cluster locks on it. To actually change blocks,
236 * call ocfs2_start_trans with the handle returned
237 * from this function. You may call ocfs2_commit_trans
238 * at any time in the lifetime of a handle.
239 * ocfs2_start_trans - Begin a transaction. Give it an upper estimate of 198 * ocfs2_start_trans - Begin a transaction. Give it an upper estimate of
240 * the number of blocks that will be changed during 199 * the number of blocks that will be changed during
241 * this handle. 200 * this handle.
242 * ocfs2_commit_trans - Complete a handle. 201 * ocfs2_commit_trans - Complete a handle. It might return -EIO if
202 * the journal was aborted. The majority of paths don't
203 * check the return value as an error there comes too
204 * late to do anything (and will be picked up in a
205 * later transaction).
243 * ocfs2_extend_trans - Extend a handle by nblocks credits. This may 206 * ocfs2_extend_trans - Extend a handle by nblocks credits. This may
244 * commit the handle to disk in the process, but will 207 * commit the handle to disk in the process, but will
245 * not release any locks taken during the transaction. 208 * not release any locks taken during the transaction.
@@ -249,24 +212,16 @@ static inline void ocfs2_checkpoint_inode(struct inode *inode)
249 * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data. 212 * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data.
250 * ocfs2_journal_dirty_data - Indicate that a data buffer should go out before 213 * ocfs2_journal_dirty_data - Indicate that a data buffer should go out before
251 * the current handle commits. 214 * the current handle commits.
252 * ocfs2_handle_add_lock - Sometimes we need to delay lock release
253 * until after a transaction has been completed. Use
254 * ocfs2_handle_add_lock to indicate that a lock needs
255 * to be released at the end of that handle. Locks
256 * will be released in the order that they are added.
257 * ocfs2_handle_add_inode - Add a locked inode to a transaction.
258 */ 215 */
259 216
260/* You must always start_trans with a number of buffs > 0, but it's 217/* You must always start_trans with a number of buffs > 0, but it's
261 * perfectly legal to go through an entire transaction without having 218 * perfectly legal to go through an entire transaction without having
262 * dirtied any buffers. */ 219 * dirtied any buffers. */
263struct ocfs2_journal_handle *ocfs2_alloc_handle(struct ocfs2_super *osb); 220handle_t *ocfs2_start_trans(struct ocfs2_super *osb,
264struct ocfs2_journal_handle *ocfs2_start_trans(struct ocfs2_super *osb,
265 struct ocfs2_journal_handle *handle,
266 int max_buffs); 221 int max_buffs);
267void ocfs2_commit_trans(struct ocfs2_journal_handle *handle); 222int ocfs2_commit_trans(struct ocfs2_super *osb,
268int ocfs2_extend_trans(struct ocfs2_journal_handle *handle, 223 handle_t *handle);
269 int nblocks); 224int ocfs2_extend_trans(handle_t *handle, int nblocks);
270 225
271/* 226/*
272 * Create access is for when we get a newly created buffer and we're 227 * Create access is for when we get a newly created buffer and we're
@@ -283,7 +238,7 @@ int ocfs2_extend_trans(struct ocfs2_journal_handle *handle,
283#define OCFS2_JOURNAL_ACCESS_WRITE 1 238#define OCFS2_JOURNAL_ACCESS_WRITE 1
284#define OCFS2_JOURNAL_ACCESS_UNDO 2 239#define OCFS2_JOURNAL_ACCESS_UNDO 2
285 240
286int ocfs2_journal_access(struct ocfs2_journal_handle *handle, 241int ocfs2_journal_access(handle_t *handle,
287 struct inode *inode, 242 struct inode *inode,
288 struct buffer_head *bh, 243 struct buffer_head *bh,
289 int type); 244 int type);
@@ -306,18 +261,10 @@ int ocfs2_journal_access(struct ocfs2_journal_handle *handle,
306 * <modify the bh> 261 * <modify the bh>
307 * ocfs2_journal_dirty(handle, bh); 262 * ocfs2_journal_dirty(handle, bh);
308 */ 263 */
309int ocfs2_journal_dirty(struct ocfs2_journal_handle *handle, 264int ocfs2_journal_dirty(handle_t *handle,
310 struct buffer_head *bh); 265 struct buffer_head *bh);
311int ocfs2_journal_dirty_data(handle_t *handle, 266int ocfs2_journal_dirty_data(handle_t *handle,
312 struct buffer_head *bh); 267 struct buffer_head *bh);
313int ocfs2_handle_add_lock(struct ocfs2_journal_handle *handle,
314 struct inode *inode);
315/*
316 * Use this to protect from other processes reading buffer state while
317 * it's in flight.
318 */
319void ocfs2_handle_add_inode(struct ocfs2_journal_handle *handle,
320 struct inode *inode);
321 268
322/* 269/*
323 * Credit Macros: 270 * Credit Macros:
diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c
index 1f17a4d08287..4dedd9789108 100644
--- a/fs/ocfs2/localalloc.c
+++ b/fs/ocfs2/localalloc.c
@@ -58,19 +58,18 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
58static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc); 58static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc);
59 59
60static int ocfs2_sync_local_to_main(struct ocfs2_super *osb, 60static int ocfs2_sync_local_to_main(struct ocfs2_super *osb,
61 struct ocfs2_journal_handle *handle, 61 handle_t *handle,
62 struct ocfs2_dinode *alloc, 62 struct ocfs2_dinode *alloc,
63 struct inode *main_bm_inode, 63 struct inode *main_bm_inode,
64 struct buffer_head *main_bm_bh); 64 struct buffer_head *main_bm_bh);
65 65
66static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb, 66static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb,
67 struct ocfs2_journal_handle *handle,
68 struct ocfs2_alloc_context **ac, 67 struct ocfs2_alloc_context **ac,
69 struct inode **bitmap_inode, 68 struct inode **bitmap_inode,
70 struct buffer_head **bitmap_bh); 69 struct buffer_head **bitmap_bh);
71 70
72static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb, 71static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
73 struct ocfs2_journal_handle *handle, 72 handle_t *handle,
74 struct ocfs2_alloc_context *ac); 73 struct ocfs2_alloc_context *ac);
75 74
76static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb, 75static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
@@ -196,7 +195,7 @@ bail:
196void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb) 195void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
197{ 196{
198 int status; 197 int status;
199 struct ocfs2_journal_handle *handle = NULL; 198 handle_t *handle;
200 struct inode *local_alloc_inode = NULL; 199 struct inode *local_alloc_inode = NULL;
201 struct buffer_head *bh = NULL; 200 struct buffer_head *bh = NULL;
202 struct buffer_head *main_bm_bh = NULL; 201 struct buffer_head *main_bm_bh = NULL;
@@ -207,7 +206,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
207 mlog_entry_void(); 206 mlog_entry_void();
208 207
209 if (osb->local_alloc_state == OCFS2_LA_UNUSED) 208 if (osb->local_alloc_state == OCFS2_LA_UNUSED)
210 goto bail; 209 goto out;
211 210
212 local_alloc_inode = 211 local_alloc_inode =
213 ocfs2_get_system_file_inode(osb, 212 ocfs2_get_system_file_inode(osb,
@@ -216,40 +215,34 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
216 if (!local_alloc_inode) { 215 if (!local_alloc_inode) {
217 status = -ENOENT; 216 status = -ENOENT;
218 mlog_errno(status); 217 mlog_errno(status);
219 goto bail; 218 goto out;
220 } 219 }
221 220
222 osb->local_alloc_state = OCFS2_LA_DISABLED; 221 osb->local_alloc_state = OCFS2_LA_DISABLED;
223 222
224 handle = ocfs2_alloc_handle(osb);
225 if (!handle) {
226 status = -ENOMEM;
227 mlog_errno(status);
228 goto bail;
229 }
230
231 main_bm_inode = ocfs2_get_system_file_inode(osb, 223 main_bm_inode = ocfs2_get_system_file_inode(osb,
232 GLOBAL_BITMAP_SYSTEM_INODE, 224 GLOBAL_BITMAP_SYSTEM_INODE,
233 OCFS2_INVALID_SLOT); 225 OCFS2_INVALID_SLOT);
234 if (!main_bm_inode) { 226 if (!main_bm_inode) {
235 status = -EINVAL; 227 status = -EINVAL;
236 mlog_errno(status); 228 mlog_errno(status);
237 goto bail; 229 goto out;
238 } 230 }
239 231
240 ocfs2_handle_add_inode(handle, main_bm_inode); 232 mutex_lock(&main_bm_inode->i_mutex);
241 status = ocfs2_meta_lock(main_bm_inode, handle, &main_bm_bh, 1); 233
234 status = ocfs2_meta_lock(main_bm_inode, &main_bm_bh, 1);
242 if (status < 0) { 235 if (status < 0) {
243 mlog_errno(status); 236 mlog_errno(status);
244 goto bail; 237 goto out_mutex;
245 } 238 }
246 239
247 /* WINDOW_MOVE_CREDITS is a bit heavy... */ 240 /* WINDOW_MOVE_CREDITS is a bit heavy... */
248 handle = ocfs2_start_trans(osb, handle, OCFS2_WINDOW_MOVE_CREDITS); 241 handle = ocfs2_start_trans(osb, OCFS2_WINDOW_MOVE_CREDITS);
249 if (IS_ERR(handle)) { 242 if (IS_ERR(handle)) {
250 mlog_errno(PTR_ERR(handle)); 243 mlog_errno(PTR_ERR(handle));
251 handle = NULL; 244 handle = NULL;
252 goto bail; 245 goto out_unlock;
253 } 246 }
254 247
255 bh = osb->local_alloc_bh; 248 bh = osb->local_alloc_bh;
@@ -258,7 +251,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
258 alloc_copy = kmalloc(bh->b_size, GFP_KERNEL); 251 alloc_copy = kmalloc(bh->b_size, GFP_KERNEL);
259 if (!alloc_copy) { 252 if (!alloc_copy) {
260 status = -ENOMEM; 253 status = -ENOMEM;
261 goto bail; 254 goto out_commit;
262 } 255 }
263 memcpy(alloc_copy, alloc, bh->b_size); 256 memcpy(alloc_copy, alloc, bh->b_size);
264 257
@@ -266,7 +259,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
266 OCFS2_JOURNAL_ACCESS_WRITE); 259 OCFS2_JOURNAL_ACCESS_WRITE);
267 if (status < 0) { 260 if (status < 0) {
268 mlog_errno(status); 261 mlog_errno(status);
269 goto bail; 262 goto out_commit;
270 } 263 }
271 264
272 ocfs2_clear_local_alloc(alloc); 265 ocfs2_clear_local_alloc(alloc);
@@ -274,7 +267,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
274 status = ocfs2_journal_dirty(handle, bh); 267 status = ocfs2_journal_dirty(handle, bh);
275 if (status < 0) { 268 if (status < 0) {
276 mlog_errno(status); 269 mlog_errno(status);
277 goto bail; 270 goto out_commit;
278 } 271 }
279 272
280 brelse(bh); 273 brelse(bh);
@@ -286,16 +279,20 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
286 if (status < 0) 279 if (status < 0)
287 mlog_errno(status); 280 mlog_errno(status);
288 281
289bail: 282out_commit:
290 if (handle) 283 ocfs2_commit_trans(osb, handle);
291 ocfs2_commit_trans(handle);
292 284
285out_unlock:
293 if (main_bm_bh) 286 if (main_bm_bh)
294 brelse(main_bm_bh); 287 brelse(main_bm_bh);
295 288
296 if (main_bm_inode) 289 ocfs2_meta_unlock(main_bm_inode, 1);
297 iput(main_bm_inode);
298 290
291out_mutex:
292 mutex_unlock(&main_bm_inode->i_mutex);
293 iput(main_bm_inode);
294
295out:
299 if (local_alloc_inode) 296 if (local_alloc_inode)
300 iput(local_alloc_inode); 297 iput(local_alloc_inode);
301 298
@@ -385,61 +382,59 @@ int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb,
385 struct ocfs2_dinode *alloc) 382 struct ocfs2_dinode *alloc)
386{ 383{
387 int status; 384 int status;
388 struct ocfs2_journal_handle *handle = NULL; 385 handle_t *handle;
389 struct buffer_head *main_bm_bh = NULL; 386 struct buffer_head *main_bm_bh = NULL;
390 struct inode *main_bm_inode = NULL; 387 struct inode *main_bm_inode;
391 388
392 mlog_entry_void(); 389 mlog_entry_void();
393 390
394 handle = ocfs2_alloc_handle(osb);
395 if (!handle) {
396 status = -ENOMEM;
397 mlog_errno(status);
398 goto bail;
399 }
400
401 main_bm_inode = ocfs2_get_system_file_inode(osb, 391 main_bm_inode = ocfs2_get_system_file_inode(osb,
402 GLOBAL_BITMAP_SYSTEM_INODE, 392 GLOBAL_BITMAP_SYSTEM_INODE,
403 OCFS2_INVALID_SLOT); 393 OCFS2_INVALID_SLOT);
404 if (!main_bm_inode) { 394 if (!main_bm_inode) {
405 status = -EINVAL; 395 status = -EINVAL;
406 mlog_errno(status); 396 mlog_errno(status);
407 goto bail; 397 goto out;
408 } 398 }
409 399
410 ocfs2_handle_add_inode(handle, main_bm_inode); 400 mutex_lock(&main_bm_inode->i_mutex);
411 status = ocfs2_meta_lock(main_bm_inode, handle, &main_bm_bh, 1); 401
402 status = ocfs2_meta_lock(main_bm_inode, &main_bm_bh, 1);
412 if (status < 0) { 403 if (status < 0) {
413 mlog_errno(status); 404 mlog_errno(status);
414 goto bail; 405 goto out_mutex;
415 } 406 }
416 407
417 handle = ocfs2_start_trans(osb, handle, OCFS2_WINDOW_MOVE_CREDITS); 408 handle = ocfs2_start_trans(osb, OCFS2_WINDOW_MOVE_CREDITS);
418 if (IS_ERR(handle)) { 409 if (IS_ERR(handle)) {
419 status = PTR_ERR(handle); 410 status = PTR_ERR(handle);
420 handle = NULL; 411 handle = NULL;
421 mlog_errno(status); 412 mlog_errno(status);
422 goto bail; 413 goto out_unlock;
423 } 414 }
424 415
425 /* we want the bitmap change to be recorded on disk asap */ 416 /* we want the bitmap change to be recorded on disk asap */
426 ocfs2_handle_set_sync(handle, 1); 417 handle->h_sync = 1;
427 418
428 status = ocfs2_sync_local_to_main(osb, handle, alloc, 419 status = ocfs2_sync_local_to_main(osb, handle, alloc,
429 main_bm_inode, main_bm_bh); 420 main_bm_inode, main_bm_bh);
430 if (status < 0) 421 if (status < 0)
431 mlog_errno(status); 422 mlog_errno(status);
432 423
433bail: 424 ocfs2_commit_trans(osb, handle);
434 if (handle) 425
435 ocfs2_commit_trans(handle); 426out_unlock:
427 ocfs2_meta_unlock(main_bm_inode, 1);
428
429out_mutex:
430 mutex_unlock(&main_bm_inode->i_mutex);
436 431
437 if (main_bm_bh) 432 if (main_bm_bh)
438 brelse(main_bm_bh); 433 brelse(main_bm_bh);
439 434
440 if (main_bm_inode) 435 iput(main_bm_inode);
441 iput(main_bm_inode);
442 436
437out:
443 mlog_exit(status); 438 mlog_exit(status);
444 return status; 439 return status;
445} 440}
@@ -452,7 +447,6 @@ bail:
452 * our own in order to shift windows. 447 * our own in order to shift windows.
453 */ 448 */
454int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb, 449int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
455 struct ocfs2_journal_handle *passed_handle,
456 u32 bits_wanted, 450 u32 bits_wanted,
457 struct ocfs2_alloc_context *ac) 451 struct ocfs2_alloc_context *ac)
458{ 452{
@@ -463,9 +457,7 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
463 457
464 mlog_entry_void(); 458 mlog_entry_void();
465 459
466 BUG_ON(!passed_handle);
467 BUG_ON(!ac); 460 BUG_ON(!ac);
468 BUG_ON(passed_handle->flags & OCFS2_HANDLE_STARTED);
469 461
470 local_alloc_inode = 462 local_alloc_inode =
471 ocfs2_get_system_file_inode(osb, 463 ocfs2_get_system_file_inode(osb,
@@ -476,7 +468,11 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
476 mlog_errno(status); 468 mlog_errno(status);
477 goto bail; 469 goto bail;
478 } 470 }
479 ocfs2_handle_add_inode(passed_handle, local_alloc_inode); 471
472 mutex_lock(&local_alloc_inode->i_mutex);
473
474 ac->ac_inode = local_alloc_inode;
475 ac->ac_which = OCFS2_AC_USE_LOCAL;
480 476
481 if (osb->local_alloc_state != OCFS2_LA_ENABLED) { 477 if (osb->local_alloc_state != OCFS2_LA_ENABLED) {
482 status = -ENOSPC; 478 status = -ENOSPC;
@@ -515,21 +511,17 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
515 } 511 }
516 } 512 }
517 513
518 ac->ac_inode = igrab(local_alloc_inode);
519 get_bh(osb->local_alloc_bh); 514 get_bh(osb->local_alloc_bh);
520 ac->ac_bh = osb->local_alloc_bh; 515 ac->ac_bh = osb->local_alloc_bh;
521 ac->ac_which = OCFS2_AC_USE_LOCAL;
522 status = 0; 516 status = 0;
523bail: 517bail:
524 if (local_alloc_inode)
525 iput(local_alloc_inode);
526 518
527 mlog_exit(status); 519 mlog_exit(status);
528 return status; 520 return status;
529} 521}
530 522
531int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, 523int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
532 struct ocfs2_journal_handle *handle, 524 handle_t *handle,
533 struct ocfs2_alloc_context *ac, 525 struct ocfs2_alloc_context *ac,
534 u32 min_bits, 526 u32 min_bits,
535 u32 *bit_off, 527 u32 *bit_off,
@@ -707,7 +699,7 @@ static void ocfs2_verify_zero_bits(unsigned long *bitmap,
707 * passed is used for caching. 699 * passed is used for caching.
708 */ 700 */
709static int ocfs2_sync_local_to_main(struct ocfs2_super *osb, 701static int ocfs2_sync_local_to_main(struct ocfs2_super *osb,
710 struct ocfs2_journal_handle *handle, 702 handle_t *handle,
711 struct ocfs2_dinode *alloc, 703 struct ocfs2_dinode *alloc,
712 struct inode *main_bm_inode, 704 struct inode *main_bm_inode,
713 struct buffer_head *main_bm_bh) 705 struct buffer_head *main_bm_bh)
@@ -778,21 +770,19 @@ bail:
778} 770}
779 771
780static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb, 772static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb,
781 struct ocfs2_journal_handle *handle,
782 struct ocfs2_alloc_context **ac, 773 struct ocfs2_alloc_context **ac,
783 struct inode **bitmap_inode, 774 struct inode **bitmap_inode,
784 struct buffer_head **bitmap_bh) 775 struct buffer_head **bitmap_bh)
785{ 776{
786 int status; 777 int status;
787 778
788 *ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL); 779 *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
789 if (!(*ac)) { 780 if (!(*ac)) {
790 status = -ENOMEM; 781 status = -ENOMEM;
791 mlog_errno(status); 782 mlog_errno(status);
792 goto bail; 783 goto bail;
793 } 784 }
794 785
795 (*ac)->ac_handle = handle;
796 (*ac)->ac_bits_wanted = ocfs2_local_alloc_window_bits(osb); 786 (*ac)->ac_bits_wanted = ocfs2_local_alloc_window_bits(osb);
797 787
798 status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac); 788 status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac);
@@ -821,7 +811,7 @@ bail:
821 * pass it the bitmap lock in lock_bh if you have it. 811 * pass it the bitmap lock in lock_bh if you have it.
822 */ 812 */
823static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb, 813static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
824 struct ocfs2_journal_handle *handle, 814 handle_t *handle,
825 struct ocfs2_alloc_context *ac) 815 struct ocfs2_alloc_context *ac)
826{ 816{
827 int status = 0; 817 int status = 0;
@@ -888,23 +878,15 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
888 int status = 0; 878 int status = 0;
889 struct buffer_head *main_bm_bh = NULL; 879 struct buffer_head *main_bm_bh = NULL;
890 struct inode *main_bm_inode = NULL; 880 struct inode *main_bm_inode = NULL;
891 struct ocfs2_journal_handle *handle = NULL; 881 handle_t *handle = NULL;
892 struct ocfs2_dinode *alloc; 882 struct ocfs2_dinode *alloc;
893 struct ocfs2_dinode *alloc_copy = NULL; 883 struct ocfs2_dinode *alloc_copy = NULL;
894 struct ocfs2_alloc_context *ac = NULL; 884 struct ocfs2_alloc_context *ac = NULL;
895 885
896 mlog_entry_void(); 886 mlog_entry_void();
897 887
898 handle = ocfs2_alloc_handle(osb);
899 if (!handle) {
900 status = -ENOMEM;
901 mlog_errno(status);
902 goto bail;
903 }
904
905 /* This will lock the main bitmap for us. */ 888 /* This will lock the main bitmap for us. */
906 status = ocfs2_local_alloc_reserve_for_window(osb, 889 status = ocfs2_local_alloc_reserve_for_window(osb,
907 handle,
908 &ac, 890 &ac,
909 &main_bm_inode, 891 &main_bm_inode,
910 &main_bm_bh); 892 &main_bm_bh);
@@ -914,7 +896,7 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
914 goto bail; 896 goto bail;
915 } 897 }
916 898
917 handle = ocfs2_start_trans(osb, handle, OCFS2_WINDOW_MOVE_CREDITS); 899 handle = ocfs2_start_trans(osb, OCFS2_WINDOW_MOVE_CREDITS);
918 if (IS_ERR(handle)) { 900 if (IS_ERR(handle)) {
919 status = PTR_ERR(handle); 901 status = PTR_ERR(handle);
920 handle = NULL; 902 handle = NULL;
@@ -972,7 +954,7 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
972 status = 0; 954 status = 0;
973bail: 955bail:
974 if (handle) 956 if (handle)
975 ocfs2_commit_trans(handle); 957 ocfs2_commit_trans(osb, handle);
976 958
977 if (main_bm_bh) 959 if (main_bm_bh)
978 brelse(main_bm_bh); 960 brelse(main_bm_bh);
diff --git a/fs/ocfs2/localalloc.h b/fs/ocfs2/localalloc.h
index 30f88ce14e46..385a10152f9c 100644
--- a/fs/ocfs2/localalloc.h
+++ b/fs/ocfs2/localalloc.h
@@ -42,12 +42,11 @@ int ocfs2_alloc_should_use_local(struct ocfs2_super *osb,
42 42
43struct ocfs2_alloc_context; 43struct ocfs2_alloc_context;
44int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb, 44int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
45 struct ocfs2_journal_handle *passed_handle,
46 u32 bits_wanted, 45 u32 bits_wanted,
47 struct ocfs2_alloc_context *ac); 46 struct ocfs2_alloc_context *ac);
48 47
49int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, 48int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
50 struct ocfs2_journal_handle *handle, 49 handle_t *handle,
51 struct ocfs2_alloc_context *ac, 50 struct ocfs2_alloc_context *ac,
52 u32 min_bits, 51 u32 min_bits,
53 u32 *bit_off, 52 u32 *bit_off,
diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c
index 83934e33e5b0..51b020447683 100644
--- a/fs/ocfs2/mmap.c
+++ b/fs/ocfs2/mmap.c
@@ -82,16 +82,27 @@ static struct vm_operations_struct ocfs2_file_vm_ops = {
82 82
83int ocfs2_mmap(struct file *file, struct vm_area_struct *vma) 83int ocfs2_mmap(struct file *file, struct vm_area_struct *vma)
84{ 84{
85 int ret = 0, lock_level = 0;
86 struct ocfs2_super *osb = OCFS2_SB(file->f_dentry->d_inode->i_sb);
87
85 /* We don't want to support shared writable mappings yet. */ 88 /* We don't want to support shared writable mappings yet. */
86 if (((vma->vm_flags & VM_SHARED) || (vma->vm_flags & VM_MAYSHARE)) 89 if (!ocfs2_mount_local(osb) &&
87 && ((vma->vm_flags & VM_WRITE) || (vma->vm_flags & VM_MAYWRITE))) { 90 ((vma->vm_flags & VM_SHARED) || (vma->vm_flags & VM_MAYSHARE)) &&
91 ((vma->vm_flags & VM_WRITE) || (vma->vm_flags & VM_MAYWRITE))) {
88 mlog(0, "disallow shared writable mmaps %lx\n", vma->vm_flags); 92 mlog(0, "disallow shared writable mmaps %lx\n", vma->vm_flags);
89 /* This is -EINVAL because generic_file_readonly_mmap 93 /* This is -EINVAL because generic_file_readonly_mmap
90 * returns it in a similar situation. */ 94 * returns it in a similar situation. */
91 return -EINVAL; 95 return -EINVAL;
92 } 96 }
93 97
94 file_accessed(file); 98 ret = ocfs2_meta_lock_atime(file->f_dentry->d_inode,
99 file->f_vfsmnt, &lock_level);
100 if (ret < 0) {
101 mlog_errno(ret);
102 goto out;
103 }
104 ocfs2_meta_unlock(file->f_dentry->d_inode, lock_level);
105out:
95 vma->vm_ops = &ocfs2_file_vm_ops; 106 vma->vm_ops = &ocfs2_file_vm_ops;
96 return 0; 107 return 0;
97} 108}
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index a57b751d4f40..9637039c2633 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -75,12 +75,12 @@ static int inline ocfs2_search_dirblock(struct buffer_head *bh,
75 unsigned long offset, 75 unsigned long offset,
76 struct ocfs2_dir_entry **res_dir); 76 struct ocfs2_dir_entry **res_dir);
77 77
78static int ocfs2_delete_entry(struct ocfs2_journal_handle *handle, 78static int ocfs2_delete_entry(handle_t *handle,
79 struct inode *dir, 79 struct inode *dir,
80 struct ocfs2_dir_entry *de_del, 80 struct ocfs2_dir_entry *de_del,
81 struct buffer_head *bh); 81 struct buffer_head *bh);
82 82
83static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle, 83static int __ocfs2_add_entry(handle_t *handle,
84 struct inode *dir, 84 struct inode *dir,
85 const char *name, int namelen, 85 const char *name, int namelen,
86 struct inode *inode, u64 blkno, 86 struct inode *inode, u64 blkno,
@@ -93,43 +93,37 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
93 dev_t dev, 93 dev_t dev,
94 struct buffer_head **new_fe_bh, 94 struct buffer_head **new_fe_bh,
95 struct buffer_head *parent_fe_bh, 95 struct buffer_head *parent_fe_bh,
96 struct ocfs2_journal_handle *handle, 96 handle_t *handle,
97 struct inode **ret_inode, 97 struct inode **ret_inode,
98 struct ocfs2_alloc_context *inode_ac); 98 struct ocfs2_alloc_context *inode_ac);
99 99
100static int ocfs2_fill_new_dir(struct ocfs2_super *osb, 100static int ocfs2_fill_new_dir(struct ocfs2_super *osb,
101 struct ocfs2_journal_handle *handle, 101 handle_t *handle,
102 struct inode *parent, 102 struct inode *parent,
103 struct inode *inode, 103 struct inode *inode,
104 struct buffer_head *fe_bh, 104 struct buffer_head *fe_bh,
105 struct ocfs2_alloc_context *data_ac); 105 struct ocfs2_alloc_context *data_ac);
106 106
107static int ocfs2_double_lock(struct ocfs2_super *osb,
108 struct ocfs2_journal_handle *handle,
109 struct buffer_head **bh1,
110 struct inode *inode1,
111 struct buffer_head **bh2,
112 struct inode *inode2);
113
114static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, 107static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
115 struct ocfs2_journal_handle *handle, 108 struct inode **ret_orphan_dir,
116 struct inode *inode, 109 struct inode *inode,
117 char *name, 110 char *name,
118 struct buffer_head **de_bh); 111 struct buffer_head **de_bh);
119 112
120static int ocfs2_orphan_add(struct ocfs2_super *osb, 113static int ocfs2_orphan_add(struct ocfs2_super *osb,
121 struct ocfs2_journal_handle *handle, 114 handle_t *handle,
122 struct inode *inode, 115 struct inode *inode,
123 struct ocfs2_dinode *fe, 116 struct ocfs2_dinode *fe,
124 char *name, 117 char *name,
125 struct buffer_head *de_bh); 118 struct buffer_head *de_bh,
119 struct inode *orphan_dir_inode);
126 120
127static int ocfs2_create_symlink_data(struct ocfs2_super *osb, 121static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
128 struct ocfs2_journal_handle *handle, 122 handle_t *handle,
129 struct inode *inode, 123 struct inode *inode,
130 const char *symname); 124 const char *symname);
131 125
132static inline int ocfs2_add_entry(struct ocfs2_journal_handle *handle, 126static inline int ocfs2_add_entry(handle_t *handle,
133 struct dentry *dentry, 127 struct dentry *dentry,
134 struct inode *inode, u64 blkno, 128 struct inode *inode, u64 blkno,
135 struct buffer_head *parent_fe_bh, 129 struct buffer_head *parent_fe_bh,
@@ -165,7 +159,7 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry,
165 mlog(0, "find name %.*s in directory %llu\n", dentry->d_name.len, 159 mlog(0, "find name %.*s in directory %llu\n", dentry->d_name.len,
166 dentry->d_name.name, (unsigned long long)OCFS2_I(dir)->ip_blkno); 160 dentry->d_name.name, (unsigned long long)OCFS2_I(dir)->ip_blkno);
167 161
168 status = ocfs2_meta_lock(dir, NULL, NULL, 0); 162 status = ocfs2_meta_lock(dir, NULL, 0);
169 if (status < 0) { 163 if (status < 0) {
170 if (status != -ENOENT) 164 if (status != -ENOENT)
171 mlog_errno(status); 165 mlog_errno(status);
@@ -242,7 +236,7 @@ bail:
242} 236}
243 237
244static int ocfs2_fill_new_dir(struct ocfs2_super *osb, 238static int ocfs2_fill_new_dir(struct ocfs2_super *osb,
245 struct ocfs2_journal_handle *handle, 239 handle_t *handle,
246 struct inode *parent, 240 struct inode *parent,
247 struct inode *inode, 241 struct inode *inode,
248 struct buffer_head *fe_bh, 242 struct buffer_head *fe_bh,
@@ -317,7 +311,7 @@ static int ocfs2_mknod(struct inode *dir,
317{ 311{
318 int status = 0; 312 int status = 0;
319 struct buffer_head *parent_fe_bh = NULL; 313 struct buffer_head *parent_fe_bh = NULL;
320 struct ocfs2_journal_handle *handle = NULL; 314 handle_t *handle = NULL;
321 struct ocfs2_super *osb; 315 struct ocfs2_super *osb;
322 struct ocfs2_dinode *dirfe; 316 struct ocfs2_dinode *dirfe;
323 struct buffer_head *new_fe_bh = NULL; 317 struct buffer_head *new_fe_bh = NULL;
@@ -333,18 +327,11 @@ static int ocfs2_mknod(struct inode *dir,
333 /* get our super block */ 327 /* get our super block */
334 osb = OCFS2_SB(dir->i_sb); 328 osb = OCFS2_SB(dir->i_sb);
335 329
336 handle = ocfs2_alloc_handle(osb); 330 status = ocfs2_meta_lock(dir, &parent_fe_bh, 1);
337 if (handle == NULL) {
338 status = -ENOMEM;
339 mlog_errno(status);
340 goto leave;
341 }
342
343 status = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1);
344 if (status < 0) { 331 if (status < 0) {
345 if (status != -ENOENT) 332 if (status != -ENOENT)
346 mlog_errno(status); 333 mlog_errno(status);
347 goto leave; 334 return status;
348 } 335 }
349 336
350 if (S_ISDIR(mode) && (dir->i_nlink >= OCFS2_LINK_MAX)) { 337 if (S_ISDIR(mode) && (dir->i_nlink >= OCFS2_LINK_MAX)) {
@@ -374,7 +361,7 @@ static int ocfs2_mknod(struct inode *dir,
374 } 361 }
375 362
376 /* reserve an inode spot */ 363 /* reserve an inode spot */
377 status = ocfs2_reserve_new_inode(osb, handle, &inode_ac); 364 status = ocfs2_reserve_new_inode(osb, &inode_ac);
378 if (status < 0) { 365 if (status < 0) {
379 if (status != -ENOSPC) 366 if (status != -ENOSPC)
380 mlog_errno(status); 367 mlog_errno(status);
@@ -384,7 +371,7 @@ static int ocfs2_mknod(struct inode *dir,
384 /* are we making a directory? If so, reserve a cluster for his 371 /* are we making a directory? If so, reserve a cluster for his
385 * 1st extent. */ 372 * 1st extent. */
386 if (S_ISDIR(mode)) { 373 if (S_ISDIR(mode)) {
387 status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac); 374 status = ocfs2_reserve_clusters(osb, 1, &data_ac);
388 if (status < 0) { 375 if (status < 0) {
389 if (status != -ENOSPC) 376 if (status != -ENOSPC)
390 mlog_errno(status); 377 mlog_errno(status);
@@ -392,7 +379,7 @@ static int ocfs2_mknod(struct inode *dir,
392 } 379 }
393 } 380 }
394 381
395 handle = ocfs2_start_trans(osb, handle, OCFS2_MKNOD_CREDITS); 382 handle = ocfs2_start_trans(osb, OCFS2_MKNOD_CREDITS);
396 if (IS_ERR(handle)) { 383 if (IS_ERR(handle)) {
397 status = PTR_ERR(handle); 384 status = PTR_ERR(handle);
398 handle = NULL; 385 handle = NULL;
@@ -453,7 +440,9 @@ static int ocfs2_mknod(struct inode *dir,
453 status = 0; 440 status = 0;
454leave: 441leave:
455 if (handle) 442 if (handle)
456 ocfs2_commit_trans(handle); 443 ocfs2_commit_trans(osb, handle);
444
445 ocfs2_meta_unlock(dir, 1);
457 446
458 if (status == -ENOSPC) 447 if (status == -ENOSPC)
459 mlog(0, "Disk is full\n"); 448 mlog(0, "Disk is full\n");
@@ -487,7 +476,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
487 dev_t dev, 476 dev_t dev,
488 struct buffer_head **new_fe_bh, 477 struct buffer_head **new_fe_bh,
489 struct buffer_head *parent_fe_bh, 478 struct buffer_head *parent_fe_bh,
490 struct ocfs2_journal_handle *handle, 479 handle_t *handle,
491 struct inode **ret_inode, 480 struct inode **ret_inode,
492 struct ocfs2_alloc_context *inode_ac) 481 struct ocfs2_alloc_context *inode_ac)
493{ 482{
@@ -598,9 +587,11 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
598 } 587 }
599 588
600 ocfs2_inode_set_new(osb, inode); 589 ocfs2_inode_set_new(osb, inode);
601 status = ocfs2_create_new_inode_locks(inode); 590 if (!ocfs2_mount_local(osb)) {
602 if (status < 0) 591 status = ocfs2_create_new_inode_locks(inode);
603 mlog_errno(status); 592 if (status < 0)
593 mlog_errno(status);
594 }
604 595
605 status = 0; /* error in ocfs2_create_new_inode_locks is not 596 status = 0; /* error in ocfs2_create_new_inode_locks is not
606 * critical */ 597 * critical */
@@ -653,7 +644,7 @@ static int ocfs2_link(struct dentry *old_dentry,
653 struct inode *dir, 644 struct inode *dir,
654 struct dentry *dentry) 645 struct dentry *dentry)
655{ 646{
656 struct ocfs2_journal_handle *handle = NULL; 647 handle_t *handle;
657 struct inode *inode = old_dentry->d_inode; 648 struct inode *inode = old_dentry->d_inode;
658 int err; 649 int err;
659 struct buffer_head *fe_bh = NULL; 650 struct buffer_head *fe_bh = NULL;
@@ -666,68 +657,60 @@ static int ocfs2_link(struct dentry *old_dentry,
666 old_dentry->d_name.len, old_dentry->d_name.name, 657 old_dentry->d_name.len, old_dentry->d_name.name,
667 dentry->d_name.len, dentry->d_name.name); 658 dentry->d_name.len, dentry->d_name.name);
668 659
669 if (S_ISDIR(inode->i_mode)) { 660 if (S_ISDIR(inode->i_mode))
670 err = -EPERM; 661 return -EPERM;
671 goto bail;
672 }
673
674 handle = ocfs2_alloc_handle(osb);
675 if (handle == NULL) {
676 err = -ENOMEM;
677 goto bail;
678 }
679 662
680 err = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1); 663 err = ocfs2_meta_lock(dir, &parent_fe_bh, 1);
681 if (err < 0) { 664 if (err < 0) {
682 if (err != -ENOENT) 665 if (err != -ENOENT)
683 mlog_errno(err); 666 mlog_errno(err);
684 goto bail; 667 return err;
685 } 668 }
686 669
687 if (!dir->i_nlink) { 670 if (!dir->i_nlink) {
688 err = -ENOENT; 671 err = -ENOENT;
689 goto bail; 672 goto out;
690 } 673 }
691 674
692 err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name, 675 err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
693 dentry->d_name.len); 676 dentry->d_name.len);
694 if (err) 677 if (err)
695 goto bail; 678 goto out;
696 679
697 err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh, 680 err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
698 dentry->d_name.name, 681 dentry->d_name.name,
699 dentry->d_name.len, &de_bh); 682 dentry->d_name.len, &de_bh);
700 if (err < 0) { 683 if (err < 0) {
701 mlog_errno(err); 684 mlog_errno(err);
702 goto bail; 685 goto out;
703 } 686 }
704 687
705 err = ocfs2_meta_lock(inode, handle, &fe_bh, 1); 688 err = ocfs2_meta_lock(inode, &fe_bh, 1);
706 if (err < 0) { 689 if (err < 0) {
707 if (err != -ENOENT) 690 if (err != -ENOENT)
708 mlog_errno(err); 691 mlog_errno(err);
709 goto bail; 692 goto out;
710 } 693 }
711 694
712 fe = (struct ocfs2_dinode *) fe_bh->b_data; 695 fe = (struct ocfs2_dinode *) fe_bh->b_data;
713 if (le16_to_cpu(fe->i_links_count) >= OCFS2_LINK_MAX) { 696 if (le16_to_cpu(fe->i_links_count) >= OCFS2_LINK_MAX) {
714 err = -EMLINK; 697 err = -EMLINK;
715 goto bail; 698 goto out_unlock_inode;
716 } 699 }
717 700
718 handle = ocfs2_start_trans(osb, handle, OCFS2_LINK_CREDITS); 701 handle = ocfs2_start_trans(osb, OCFS2_LINK_CREDITS);
719 if (IS_ERR(handle)) { 702 if (IS_ERR(handle)) {
720 err = PTR_ERR(handle); 703 err = PTR_ERR(handle);
721 handle = NULL; 704 handle = NULL;
722 mlog_errno(err); 705 mlog_errno(err);
723 goto bail; 706 goto out_unlock_inode;
724 } 707 }
725 708
726 err = ocfs2_journal_access(handle, inode, fe_bh, 709 err = ocfs2_journal_access(handle, inode, fe_bh,
727 OCFS2_JOURNAL_ACCESS_WRITE); 710 OCFS2_JOURNAL_ACCESS_WRITE);
728 if (err < 0) { 711 if (err < 0) {
729 mlog_errno(err); 712 mlog_errno(err);
730 goto bail; 713 goto out_commit;
731 } 714 }
732 715
733 inc_nlink(inode); 716 inc_nlink(inode);
@@ -741,7 +724,7 @@ static int ocfs2_link(struct dentry *old_dentry,
741 le16_add_cpu(&fe->i_links_count, -1); 724 le16_add_cpu(&fe->i_links_count, -1);
742 drop_nlink(inode); 725 drop_nlink(inode);
743 mlog_errno(err); 726 mlog_errno(err);
744 goto bail; 727 goto out_commit;
745 } 728 }
746 729
747 err = ocfs2_add_entry(handle, dentry, inode, 730 err = ocfs2_add_entry(handle, dentry, inode,
@@ -751,21 +734,27 @@ static int ocfs2_link(struct dentry *old_dentry,
751 le16_add_cpu(&fe->i_links_count, -1); 734 le16_add_cpu(&fe->i_links_count, -1);
752 drop_nlink(inode); 735 drop_nlink(inode);
753 mlog_errno(err); 736 mlog_errno(err);
754 goto bail; 737 goto out_commit;
755 } 738 }
756 739
757 err = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno); 740 err = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
758 if (err) { 741 if (err) {
759 mlog_errno(err); 742 mlog_errno(err);
760 goto bail; 743 goto out_commit;
761 } 744 }
762 745
763 atomic_inc(&inode->i_count); 746 atomic_inc(&inode->i_count);
764 dentry->d_op = &ocfs2_dentry_ops; 747 dentry->d_op = &ocfs2_dentry_ops;
765 d_instantiate(dentry, inode); 748 d_instantiate(dentry, inode);
766bail: 749
767 if (handle) 750out_commit:
768 ocfs2_commit_trans(handle); 751 ocfs2_commit_trans(osb, handle);
752out_unlock_inode:
753 ocfs2_meta_unlock(inode, 1);
754
755out:
756 ocfs2_meta_unlock(dir, 1);
757
769 if (de_bh) 758 if (de_bh)
770 brelse(de_bh); 759 brelse(de_bh);
771 if (fe_bh) 760 if (fe_bh)
@@ -812,13 +801,15 @@ static int ocfs2_unlink(struct inode *dir,
812 struct dentry *dentry) 801 struct dentry *dentry)
813{ 802{
814 int status; 803 int status;
804 int child_locked = 0;
815 struct inode *inode = dentry->d_inode; 805 struct inode *inode = dentry->d_inode;
806 struct inode *orphan_dir = NULL;
816 struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); 807 struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
817 u64 blkno; 808 u64 blkno;
818 struct ocfs2_dinode *fe = NULL; 809 struct ocfs2_dinode *fe = NULL;
819 struct buffer_head *fe_bh = NULL; 810 struct buffer_head *fe_bh = NULL;
820 struct buffer_head *parent_node_bh = NULL; 811 struct buffer_head *parent_node_bh = NULL;
821 struct ocfs2_journal_handle *handle = NULL; 812 handle_t *handle = NULL;
822 struct ocfs2_dir_entry *dirent = NULL; 813 struct ocfs2_dir_entry *dirent = NULL;
823 struct buffer_head *dirent_bh = NULL; 814 struct buffer_head *dirent_bh = NULL;
824 char orphan_name[OCFS2_ORPHAN_NAMELEN + 1]; 815 char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
@@ -833,22 +824,14 @@ static int ocfs2_unlink(struct inode *dir,
833 824
834 if (inode == osb->root_inode) { 825 if (inode == osb->root_inode) {
835 mlog(0, "Cannot delete the root directory\n"); 826 mlog(0, "Cannot delete the root directory\n");
836 status = -EPERM; 827 return -EPERM;
837 goto leave;
838 }
839
840 handle = ocfs2_alloc_handle(osb);
841 if (handle == NULL) {
842 status = -ENOMEM;
843 mlog_errno(status);
844 goto leave;
845 } 828 }
846 829
847 status = ocfs2_meta_lock(dir, handle, &parent_node_bh, 1); 830 status = ocfs2_meta_lock(dir, &parent_node_bh, 1);
848 if (status < 0) { 831 if (status < 0) {
849 if (status != -ENOENT) 832 if (status != -ENOENT)
850 mlog_errno(status); 833 mlog_errno(status);
851 goto leave; 834 return status;
852 } 835 }
853 836
854 status = ocfs2_find_files_on_disk(dentry->d_name.name, 837 status = ocfs2_find_files_on_disk(dentry->d_name.name,
@@ -869,12 +852,13 @@ static int ocfs2_unlink(struct inode *dir,
869 goto leave; 852 goto leave;
870 } 853 }
871 854
872 status = ocfs2_meta_lock(inode, handle, &fe_bh, 1); 855 status = ocfs2_meta_lock(inode, &fe_bh, 1);
873 if (status < 0) { 856 if (status < 0) {
874 if (status != -ENOENT) 857 if (status != -ENOENT)
875 mlog_errno(status); 858 mlog_errno(status);
876 goto leave; 859 goto leave;
877 } 860 }
861 child_locked = 1;
878 862
879 if (S_ISDIR(inode->i_mode)) { 863 if (S_ISDIR(inode->i_mode)) {
880 if (!ocfs2_empty_dir(inode)) { 864 if (!ocfs2_empty_dir(inode)) {
@@ -895,7 +879,7 @@ static int ocfs2_unlink(struct inode *dir,
895 } 879 }
896 880
897 if (inode_is_unlinkable(inode)) { 881 if (inode_is_unlinkable(inode)) {
898 status = ocfs2_prepare_orphan_dir(osb, handle, inode, 882 status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, inode,
899 orphan_name, 883 orphan_name,
900 &orphan_entry_bh); 884 &orphan_entry_bh);
901 if (status < 0) { 885 if (status < 0) {
@@ -904,7 +888,7 @@ static int ocfs2_unlink(struct inode *dir,
904 } 888 }
905 } 889 }
906 890
907 handle = ocfs2_start_trans(osb, handle, OCFS2_UNLINK_CREDITS); 891 handle = ocfs2_start_trans(osb, OCFS2_UNLINK_CREDITS);
908 if (IS_ERR(handle)) { 892 if (IS_ERR(handle)) {
909 status = PTR_ERR(handle); 893 status = PTR_ERR(handle);
910 handle = NULL; 894 handle = NULL;
@@ -923,7 +907,7 @@ static int ocfs2_unlink(struct inode *dir,
923 907
924 if (inode_is_unlinkable(inode)) { 908 if (inode_is_unlinkable(inode)) {
925 status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name, 909 status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name,
926 orphan_entry_bh); 910 orphan_entry_bh, orphan_dir);
927 if (status < 0) { 911 if (status < 0) {
928 mlog_errno(status); 912 mlog_errno(status);
929 goto leave; 913 goto leave;
@@ -960,7 +944,19 @@ static int ocfs2_unlink(struct inode *dir,
960 944
961leave: 945leave:
962 if (handle) 946 if (handle)
963 ocfs2_commit_trans(handle); 947 ocfs2_commit_trans(osb, handle);
948
949 if (child_locked)
950 ocfs2_meta_unlock(inode, 1);
951
952 ocfs2_meta_unlock(dir, 1);
953
954 if (orphan_dir) {
955 /* This was locked for us in ocfs2_prepare_orphan_dir() */
956 ocfs2_meta_unlock(orphan_dir, 1);
957 mutex_unlock(&orphan_dir->i_mutex);
958 iput(orphan_dir);
959 }
964 960
965 if (fe_bh) 961 if (fe_bh)
966 brelse(fe_bh); 962 brelse(fe_bh);
@@ -984,7 +980,6 @@ leave:
984 * if they have the same id, then the 1st one is the only one locked. 980 * if they have the same id, then the 1st one is the only one locked.
985 */ 981 */
986static int ocfs2_double_lock(struct ocfs2_super *osb, 982static int ocfs2_double_lock(struct ocfs2_super *osb,
987 struct ocfs2_journal_handle *handle,
988 struct buffer_head **bh1, 983 struct buffer_head **bh1,
989 struct inode *inode1, 984 struct inode *inode1,
990 struct buffer_head **bh2, 985 struct buffer_head **bh2,
@@ -1000,8 +995,6 @@ static int ocfs2_double_lock(struct ocfs2_super *osb,
1000 (unsigned long long)oi1->ip_blkno, 995 (unsigned long long)oi1->ip_blkno,
1001 (unsigned long long)oi2->ip_blkno); 996 (unsigned long long)oi2->ip_blkno);
1002 997
1003 BUG_ON(!handle);
1004
1005 if (*bh1) 998 if (*bh1)
1006 *bh1 = NULL; 999 *bh1 = NULL;
1007 if (*bh2) 1000 if (*bh2)
@@ -1021,25 +1014,41 @@ static int ocfs2_double_lock(struct ocfs2_super *osb,
1021 inode1 = tmpinode; 1014 inode1 = tmpinode;
1022 } 1015 }
1023 /* lock id2 */ 1016 /* lock id2 */
1024 status = ocfs2_meta_lock(inode2, handle, bh2, 1); 1017 status = ocfs2_meta_lock(inode2, bh2, 1);
1025 if (status < 0) { 1018 if (status < 0) {
1026 if (status != -ENOENT) 1019 if (status != -ENOENT)
1027 mlog_errno(status); 1020 mlog_errno(status);
1028 goto bail; 1021 goto bail;
1029 } 1022 }
1030 } 1023 }
1024
1031 /* lock id1 */ 1025 /* lock id1 */
1032 status = ocfs2_meta_lock(inode1, handle, bh1, 1); 1026 status = ocfs2_meta_lock(inode1, bh1, 1);
1033 if (status < 0) { 1027 if (status < 0) {
1028 /*
1029 * An error return must mean that no cluster locks
1030 * were held on function exit.
1031 */
1032 if (oi1->ip_blkno != oi2->ip_blkno)
1033 ocfs2_meta_unlock(inode2, 1);
1034
1034 if (status != -ENOENT) 1035 if (status != -ENOENT)
1035 mlog_errno(status); 1036 mlog_errno(status);
1036 goto bail;
1037 } 1037 }
1038
1038bail: 1039bail:
1039 mlog_exit(status); 1040 mlog_exit(status);
1040 return status; 1041 return status;
1041} 1042}
1042 1043
1044static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2)
1045{
1046 ocfs2_meta_unlock(inode1, 1);
1047
1048 if (inode1 != inode2)
1049 ocfs2_meta_unlock(inode2, 1);
1050}
1051
1043#define PARENT_INO(buffer) \ 1052#define PARENT_INO(buffer) \
1044 ((struct ocfs2_dir_entry *) \ 1053 ((struct ocfs2_dir_entry *) \
1045 ((char *)buffer + \ 1054 ((char *)buffer + \
@@ -1050,9 +1059,11 @@ static int ocfs2_rename(struct inode *old_dir,
1050 struct inode *new_dir, 1059 struct inode *new_dir,
1051 struct dentry *new_dentry) 1060 struct dentry *new_dentry)
1052{ 1061{
1053 int status = 0, rename_lock = 0; 1062 int status = 0, rename_lock = 0, parents_locked = 0;
1063 int old_child_locked = 0, new_child_locked = 0;
1054 struct inode *old_inode = old_dentry->d_inode; 1064 struct inode *old_inode = old_dentry->d_inode;
1055 struct inode *new_inode = new_dentry->d_inode; 1065 struct inode *new_inode = new_dentry->d_inode;
1066 struct inode *orphan_dir = NULL;
1056 struct ocfs2_dinode *newfe = NULL; 1067 struct ocfs2_dinode *newfe = NULL;
1057 char orphan_name[OCFS2_ORPHAN_NAMELEN + 1]; 1068 char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
1058 struct buffer_head *orphan_entry_bh = NULL; 1069 struct buffer_head *orphan_entry_bh = NULL;
@@ -1060,7 +1071,7 @@ static int ocfs2_rename(struct inode *old_dir,
1060 struct buffer_head *insert_entry_bh = NULL; 1071 struct buffer_head *insert_entry_bh = NULL;
1061 struct ocfs2_super *osb = NULL; 1072 struct ocfs2_super *osb = NULL;
1062 u64 newfe_blkno; 1073 u64 newfe_blkno;
1063 struct ocfs2_journal_handle *handle = NULL; 1074 handle_t *handle = NULL;
1064 struct buffer_head *old_dir_bh = NULL; 1075 struct buffer_head *old_dir_bh = NULL;
1065 struct buffer_head *new_dir_bh = NULL; 1076 struct buffer_head *new_dir_bh = NULL;
1066 struct ocfs2_dir_entry *old_de = NULL, *new_de = NULL; // dirent for old_dentry 1077 struct ocfs2_dir_entry *old_de = NULL, *new_de = NULL; // dirent for old_dentry
@@ -1105,21 +1116,14 @@ static int ocfs2_rename(struct inode *old_dir,
1105 rename_lock = 1; 1116 rename_lock = 1;
1106 } 1117 }
1107 1118
1108 handle = ocfs2_alloc_handle(osb);
1109 if (handle == NULL) {
1110 status = -ENOMEM;
1111 mlog_errno(status);
1112 goto bail;
1113 }
1114
1115 /* if old and new are the same, this'll just do one lock. */ 1119 /* if old and new are the same, this'll just do one lock. */
1116 status = ocfs2_double_lock(osb, handle, 1120 status = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
1117 &old_dir_bh, old_dir, 1121 &new_dir_bh, new_dir);
1118 &new_dir_bh, new_dir);
1119 if (status < 0) { 1122 if (status < 0) {
1120 mlog_errno(status); 1123 mlog_errno(status);
1121 goto bail; 1124 goto bail;
1122 } 1125 }
1126 parents_locked = 1;
1123 1127
1124 /* make sure both dirs have bhs 1128 /* make sure both dirs have bhs
1125 * get an extra ref on old_dir_bh if old==new */ 1129 * get an extra ref on old_dir_bh if old==new */
@@ -1140,12 +1144,13 @@ static int ocfs2_rename(struct inode *old_dir,
1140 * the vote thread on other nodes won't have to concurrently 1144 * the vote thread on other nodes won't have to concurrently
1141 * downconvert the inode and the dentry locks. 1145 * downconvert the inode and the dentry locks.
1142 */ 1146 */
1143 status = ocfs2_meta_lock(old_inode, handle, NULL, 1); 1147 status = ocfs2_meta_lock(old_inode, NULL, 1);
1144 if (status < 0) { 1148 if (status < 0) {
1145 if (status != -ENOENT) 1149 if (status != -ENOENT)
1146 mlog_errno(status); 1150 mlog_errno(status);
1147 goto bail; 1151 goto bail;
1148 } 1152 }
1153 old_child_locked = 1;
1149 1154
1150 status = ocfs2_remote_dentry_delete(old_dentry); 1155 status = ocfs2_remote_dentry_delete(old_dentry);
1151 if (status < 0) { 1156 if (status < 0) {
@@ -1231,12 +1236,13 @@ static int ocfs2_rename(struct inode *old_dir,
1231 goto bail; 1236 goto bail;
1232 } 1237 }
1233 1238
1234 status = ocfs2_meta_lock(new_inode, handle, &newfe_bh, 1); 1239 status = ocfs2_meta_lock(new_inode, &newfe_bh, 1);
1235 if (status < 0) { 1240 if (status < 0) {
1236 if (status != -ENOENT) 1241 if (status != -ENOENT)
1237 mlog_errno(status); 1242 mlog_errno(status);
1238 goto bail; 1243 goto bail;
1239 } 1244 }
1245 new_child_locked = 1;
1240 1246
1241 status = ocfs2_remote_dentry_delete(new_dentry); 1247 status = ocfs2_remote_dentry_delete(new_dentry);
1242 if (status < 0) { 1248 if (status < 0) {
@@ -1252,7 +1258,7 @@ static int ocfs2_rename(struct inode *old_dir,
1252 (unsigned long long)newfe_bh->b_blocknr : 0ULL); 1258 (unsigned long long)newfe_bh->b_blocknr : 0ULL);
1253 1259
1254 if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) { 1260 if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
1255 status = ocfs2_prepare_orphan_dir(osb, handle, 1261 status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
1256 new_inode, 1262 new_inode,
1257 orphan_name, 1263 orphan_name,
1258 &orphan_entry_bh); 1264 &orphan_entry_bh);
@@ -1280,7 +1286,7 @@ static int ocfs2_rename(struct inode *old_dir,
1280 } 1286 }
1281 } 1287 }
1282 1288
1283 handle = ocfs2_start_trans(osb, handle, OCFS2_RENAME_CREDITS); 1289 handle = ocfs2_start_trans(osb, OCFS2_RENAME_CREDITS);
1284 if (IS_ERR(handle)) { 1290 if (IS_ERR(handle)) {
1285 status = PTR_ERR(handle); 1291 status = PTR_ERR(handle);
1286 handle = NULL; 1292 handle = NULL;
@@ -1307,7 +1313,7 @@ static int ocfs2_rename(struct inode *old_dir,
1307 (newfe->i_links_count == cpu_to_le16(1))){ 1313 (newfe->i_links_count == cpu_to_le16(1))){
1308 status = ocfs2_orphan_add(osb, handle, new_inode, 1314 status = ocfs2_orphan_add(osb, handle, new_inode,
1309 newfe, orphan_name, 1315 newfe, orphan_name,
1310 orphan_entry_bh); 1316 orphan_entry_bh, orphan_dir);
1311 if (status < 0) { 1317 if (status < 0) {
1312 mlog_errno(status); 1318 mlog_errno(status);
1313 goto bail; 1319 goto bail;
@@ -1424,7 +1430,23 @@ bail:
1424 ocfs2_rename_unlock(osb); 1430 ocfs2_rename_unlock(osb);
1425 1431
1426 if (handle) 1432 if (handle)
1427 ocfs2_commit_trans(handle); 1433 ocfs2_commit_trans(osb, handle);
1434
1435 if (parents_locked)
1436 ocfs2_double_unlock(old_dir, new_dir);
1437
1438 if (old_child_locked)
1439 ocfs2_meta_unlock(old_inode, 1);
1440
1441 if (new_child_locked)
1442 ocfs2_meta_unlock(new_inode, 1);
1443
1444 if (orphan_dir) {
1445 /* This was locked for us in ocfs2_prepare_orphan_dir() */
1446 ocfs2_meta_unlock(orphan_dir, 1);
1447 mutex_unlock(&orphan_dir->i_mutex);
1448 iput(orphan_dir);
1449 }
1428 1450
1429 if (new_inode) 1451 if (new_inode)
1430 sync_mapping_buffers(old_inode->i_mapping); 1452 sync_mapping_buffers(old_inode->i_mapping);
@@ -1458,7 +1480,7 @@ bail:
1458 * data, including the null terminator. 1480 * data, including the null terminator.
1459 */ 1481 */
1460static int ocfs2_create_symlink_data(struct ocfs2_super *osb, 1482static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
1461 struct ocfs2_journal_handle *handle, 1483 handle_t *handle,
1462 struct inode *inode, 1484 struct inode *inode,
1463 const char *symname) 1485 const char *symname)
1464{ 1486{
@@ -1573,7 +1595,7 @@ static int ocfs2_symlink(struct inode *dir,
1573 struct buffer_head *parent_fe_bh = NULL; 1595 struct buffer_head *parent_fe_bh = NULL;
1574 struct ocfs2_dinode *fe = NULL; 1596 struct ocfs2_dinode *fe = NULL;
1575 struct ocfs2_dinode *dirfe; 1597 struct ocfs2_dinode *dirfe;
1576 struct ocfs2_journal_handle *handle = NULL; 1598 handle_t *handle = NULL;
1577 struct ocfs2_alloc_context *inode_ac = NULL; 1599 struct ocfs2_alloc_context *inode_ac = NULL;
1578 struct ocfs2_alloc_context *data_ac = NULL; 1600 struct ocfs2_alloc_context *data_ac = NULL;
1579 1601
@@ -1587,19 +1609,12 @@ static int ocfs2_symlink(struct inode *dir,
1587 1609
1588 credits = ocfs2_calc_symlink_credits(sb); 1610 credits = ocfs2_calc_symlink_credits(sb);
1589 1611
1590 handle = ocfs2_alloc_handle(osb);
1591 if (handle == NULL) {
1592 status = -ENOMEM;
1593 mlog_errno(status);
1594 goto bail;
1595 }
1596
1597 /* lock the parent directory */ 1612 /* lock the parent directory */
1598 status = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1); 1613 status = ocfs2_meta_lock(dir, &parent_fe_bh, 1);
1599 if (status < 0) { 1614 if (status < 0) {
1600 if (status != -ENOENT) 1615 if (status != -ENOENT)
1601 mlog_errno(status); 1616 mlog_errno(status);
1602 goto bail; 1617 return status;
1603 } 1618 }
1604 1619
1605 dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data; 1620 dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
@@ -1622,7 +1637,7 @@ static int ocfs2_symlink(struct inode *dir,
1622 goto bail; 1637 goto bail;
1623 } 1638 }
1624 1639
1625 status = ocfs2_reserve_new_inode(osb, handle, &inode_ac); 1640 status = ocfs2_reserve_new_inode(osb, &inode_ac);
1626 if (status < 0) { 1641 if (status < 0) {
1627 if (status != -ENOSPC) 1642 if (status != -ENOSPC)
1628 mlog_errno(status); 1643 mlog_errno(status);
@@ -1631,7 +1646,7 @@ static int ocfs2_symlink(struct inode *dir,
1631 1646
1632 /* don't reserve bitmap space for fast symlinks. */ 1647 /* don't reserve bitmap space for fast symlinks. */
1633 if (l > ocfs2_fast_symlink_chars(sb)) { 1648 if (l > ocfs2_fast_symlink_chars(sb)) {
1634 status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac); 1649 status = ocfs2_reserve_clusters(osb, 1, &data_ac);
1635 if (status < 0) { 1650 if (status < 0) {
1636 if (status != -ENOSPC) 1651 if (status != -ENOSPC)
1637 mlog_errno(status); 1652 mlog_errno(status);
@@ -1639,7 +1654,7 @@ static int ocfs2_symlink(struct inode *dir,
1639 } 1654 }
1640 } 1655 }
1641 1656
1642 handle = ocfs2_start_trans(osb, handle, credits); 1657 handle = ocfs2_start_trans(osb, credits);
1643 if (IS_ERR(handle)) { 1658 if (IS_ERR(handle)) {
1644 status = PTR_ERR(handle); 1659 status = PTR_ERR(handle);
1645 handle = NULL; 1660 handle = NULL;
@@ -1717,7 +1732,10 @@ static int ocfs2_symlink(struct inode *dir,
1717 d_instantiate(dentry, inode); 1732 d_instantiate(dentry, inode);
1718bail: 1733bail:
1719 if (handle) 1734 if (handle)
1720 ocfs2_commit_trans(handle); 1735 ocfs2_commit_trans(osb, handle);
1736
1737 ocfs2_meta_unlock(dir, 1);
1738
1721 if (new_fe_bh) 1739 if (new_fe_bh)
1722 brelse(new_fe_bh); 1740 brelse(new_fe_bh);
1723 if (parent_fe_bh) 1741 if (parent_fe_bh)
@@ -1768,7 +1786,7 @@ int ocfs2_check_dir_entry(struct inode * dir,
1768 * If you pass me insert_bh, I'll skip the search of the other dir 1786 * If you pass me insert_bh, I'll skip the search of the other dir
1769 * blocks and put the record in there. 1787 * blocks and put the record in there.
1770 */ 1788 */
1771static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle, 1789static int __ocfs2_add_entry(handle_t *handle,
1772 struct inode *dir, 1790 struct inode *dir,
1773 const char *name, int namelen, 1791 const char *name, int namelen,
1774 struct inode *inode, u64 blkno, 1792 struct inode *inode, u64 blkno,
@@ -1854,7 +1872,7 @@ bail:
1854 * ocfs2_delete_entry deletes a directory entry by merging it with the 1872 * ocfs2_delete_entry deletes a directory entry by merging it with the
1855 * previous entry 1873 * previous entry
1856 */ 1874 */
1857static int ocfs2_delete_entry(struct ocfs2_journal_handle *handle, 1875static int ocfs2_delete_entry(handle_t *handle,
1858 struct inode *dir, 1876 struct inode *dir,
1859 struct ocfs2_dir_entry *de_del, 1877 struct ocfs2_dir_entry *de_del,
1860 struct buffer_head *bh) 1878 struct buffer_head *bh)
@@ -2085,19 +2103,19 @@ bail:
2085} 2103}
2086 2104
2087static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, 2105static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
2088 struct ocfs2_journal_handle *handle, 2106 struct inode **ret_orphan_dir,
2089 struct inode *inode, 2107 struct inode *inode,
2090 char *name, 2108 char *name,
2091 struct buffer_head **de_bh) 2109 struct buffer_head **de_bh)
2092{ 2110{
2093 struct inode *orphan_dir_inode = NULL; 2111 struct inode *orphan_dir_inode;
2094 struct buffer_head *orphan_dir_bh = NULL; 2112 struct buffer_head *orphan_dir_bh = NULL;
2095 int status = 0; 2113 int status = 0;
2096 2114
2097 status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name); 2115 status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name);
2098 if (status < 0) { 2116 if (status < 0) {
2099 mlog_errno(status); 2117 mlog_errno(status);
2100 goto leave; 2118 return status;
2101 } 2119 }
2102 2120
2103 orphan_dir_inode = ocfs2_get_system_file_inode(osb, 2121 orphan_dir_inode = ocfs2_get_system_file_inode(osb,
@@ -2106,11 +2124,12 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
2106 if (!orphan_dir_inode) { 2124 if (!orphan_dir_inode) {
2107 status = -ENOENT; 2125 status = -ENOENT;
2108 mlog_errno(status); 2126 mlog_errno(status);
2109 goto leave; 2127 return status;
2110 } 2128 }
2111 2129
2112 ocfs2_handle_add_inode(handle, orphan_dir_inode); 2130 mutex_lock(&orphan_dir_inode->i_mutex);
2113 status = ocfs2_meta_lock(orphan_dir_inode, handle, &orphan_dir_bh, 1); 2131
2132 status = ocfs2_meta_lock(orphan_dir_inode, &orphan_dir_bh, 1);
2114 if (status < 0) { 2133 if (status < 0) {
2115 mlog_errno(status); 2134 mlog_errno(status);
2116 goto leave; 2135 goto leave;
@@ -2120,13 +2139,19 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
2120 orphan_dir_bh, name, 2139 orphan_dir_bh, name,
2121 OCFS2_ORPHAN_NAMELEN, de_bh); 2140 OCFS2_ORPHAN_NAMELEN, de_bh);
2122 if (status < 0) { 2141 if (status < 0) {
2142 ocfs2_meta_unlock(orphan_dir_inode, 1);
2143
2123 mlog_errno(status); 2144 mlog_errno(status);
2124 goto leave; 2145 goto leave;
2125 } 2146 }
2126 2147
2148 *ret_orphan_dir = orphan_dir_inode;
2149
2127leave: 2150leave:
2128 if (orphan_dir_inode) 2151 if (status) {
2152 mutex_unlock(&orphan_dir_inode->i_mutex);
2129 iput(orphan_dir_inode); 2153 iput(orphan_dir_inode);
2154 }
2130 2155
2131 if (orphan_dir_bh) 2156 if (orphan_dir_bh)
2132 brelse(orphan_dir_bh); 2157 brelse(orphan_dir_bh);
@@ -2136,28 +2161,19 @@ leave:
2136} 2161}
2137 2162
2138static int ocfs2_orphan_add(struct ocfs2_super *osb, 2163static int ocfs2_orphan_add(struct ocfs2_super *osb,
2139 struct ocfs2_journal_handle *handle, 2164 handle_t *handle,
2140 struct inode *inode, 2165 struct inode *inode,
2141 struct ocfs2_dinode *fe, 2166 struct ocfs2_dinode *fe,
2142 char *name, 2167 char *name,
2143 struct buffer_head *de_bh) 2168 struct buffer_head *de_bh,
2169 struct inode *orphan_dir_inode)
2144{ 2170{
2145 struct inode *orphan_dir_inode = NULL;
2146 struct buffer_head *orphan_dir_bh = NULL; 2171 struct buffer_head *orphan_dir_bh = NULL;
2147 int status = 0; 2172 int status = 0;
2148 struct ocfs2_dinode *orphan_fe; 2173 struct ocfs2_dinode *orphan_fe;
2149 2174
2150 mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino); 2175 mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
2151 2176
2152 orphan_dir_inode = ocfs2_get_system_file_inode(osb,
2153 ORPHAN_DIR_SYSTEM_INODE,
2154 osb->slot_num);
2155 if (!orphan_dir_inode) {
2156 status = -ENOENT;
2157 mlog_errno(status);
2158 goto leave;
2159 }
2160
2161 status = ocfs2_read_block(osb, 2177 status = ocfs2_read_block(osb,
2162 OCFS2_I(orphan_dir_inode)->ip_blkno, 2178 OCFS2_I(orphan_dir_inode)->ip_blkno,
2163 &orphan_dir_bh, OCFS2_BH_CACHED, 2179 &orphan_dir_bh, OCFS2_BH_CACHED,
@@ -2209,9 +2225,6 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb,
2209 (unsigned long long)OCFS2_I(inode)->ip_blkno, osb->slot_num); 2225 (unsigned long long)OCFS2_I(inode)->ip_blkno, osb->slot_num);
2210 2226
2211leave: 2227leave:
2212 if (orphan_dir_inode)
2213 iput(orphan_dir_inode);
2214
2215 if (orphan_dir_bh) 2228 if (orphan_dir_bh)
2216 brelse(orphan_dir_bh); 2229 brelse(orphan_dir_bh);
2217 2230
@@ -2221,7 +2234,7 @@ leave:
2221 2234
2222/* unlike orphan_add, we expect the orphan dir to already be locked here. */ 2235/* unlike orphan_add, we expect the orphan dir to already be locked here. */
2223int ocfs2_orphan_del(struct ocfs2_super *osb, 2236int ocfs2_orphan_del(struct ocfs2_super *osb,
2224 struct ocfs2_journal_handle *handle, 2237 handle_t *handle,
2225 struct inode *orphan_dir_inode, 2238 struct inode *orphan_dir_inode,
2226 struct inode *inode, 2239 struct inode *inode,
2227 struct buffer_head *orphan_dir_bh) 2240 struct buffer_head *orphan_dir_bh)
@@ -2300,4 +2313,5 @@ struct inode_operations ocfs2_dir_iops = {
2300 .rename = ocfs2_rename, 2313 .rename = ocfs2_rename,
2301 .setattr = ocfs2_setattr, 2314 .setattr = ocfs2_setattr,
2302 .getattr = ocfs2_getattr, 2315 .getattr = ocfs2_getattr,
2316 .permission = ocfs2_permission,
2303}; 2317};
diff --git a/fs/ocfs2/namei.h b/fs/ocfs2/namei.h
index deaaa97dbf0b..8425944fcccd 100644
--- a/fs/ocfs2/namei.h
+++ b/fs/ocfs2/namei.h
@@ -39,7 +39,7 @@ struct buffer_head *ocfs2_find_entry(const char *name,
39 struct inode *dir, 39 struct inode *dir,
40 struct ocfs2_dir_entry **res_dir); 40 struct ocfs2_dir_entry **res_dir);
41int ocfs2_orphan_del(struct ocfs2_super *osb, 41int ocfs2_orphan_del(struct ocfs2_super *osb,
42 struct ocfs2_journal_handle *handle, 42 handle_t *handle,
43 struct inode *orphan_dir_inode, 43 struct inode *orphan_dir_inode,
44 struct inode *inode, 44 struct inode *inode,
45 struct buffer_head *orphan_dir_bh); 45 struct buffer_head *orphan_dir_bh);
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index 0462a7f4e21b..db8e77cd35d3 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -34,6 +34,7 @@
34#include <linux/workqueue.h> 34#include <linux/workqueue.h>
35#include <linux/kref.h> 35#include <linux/kref.h>
36#include <linux/mutex.h> 36#include <linux/mutex.h>
37#include <linux/jbd.h>
37 38
38#include "cluster/nodemanager.h" 39#include "cluster/nodemanager.h"
39#include "cluster/heartbeat.h" 40#include "cluster/heartbeat.h"
@@ -179,9 +180,9 @@ enum ocfs2_mount_options
179#define OCFS2_OSB_SOFT_RO 0x0001 180#define OCFS2_OSB_SOFT_RO 0x0001
180#define OCFS2_OSB_HARD_RO 0x0002 181#define OCFS2_OSB_HARD_RO 0x0002
181#define OCFS2_OSB_ERROR_FS 0x0004 182#define OCFS2_OSB_ERROR_FS 0x0004
183#define OCFS2_DEFAULT_ATIME_QUANTUM 60
182 184
183struct ocfs2_journal; 185struct ocfs2_journal;
184struct ocfs2_journal_handle;
185struct ocfs2_super 186struct ocfs2_super
186{ 187{
187 struct task_struct *commit_task; 188 struct task_struct *commit_task;
@@ -218,6 +219,7 @@ struct ocfs2_super
218 unsigned long osb_flags; 219 unsigned long osb_flags;
219 220
220 unsigned long s_mount_opt; 221 unsigned long s_mount_opt;
222 unsigned int s_atime_quantum;
221 223
222 u16 max_slots; 224 u16 max_slots;
223 s16 node_num; 225 s16 node_num;
@@ -283,7 +285,7 @@ struct ocfs2_super
283 /* Truncate log info */ 285 /* Truncate log info */
284 struct inode *osb_tl_inode; 286 struct inode *osb_tl_inode;
285 struct buffer_head *osb_tl_bh; 287 struct buffer_head *osb_tl_bh;
286 struct work_struct osb_truncate_log_wq; 288 struct delayed_work osb_truncate_log_wq;
287 289
288 struct ocfs2_node_map osb_recovering_orphan_dirs; 290 struct ocfs2_node_map osb_recovering_orphan_dirs;
289 unsigned int *osb_orphan_wipes; 291 unsigned int *osb_orphan_wipes;
@@ -347,6 +349,11 @@ static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
347 return ret; 349 return ret;
348} 350}
349 351
352static inline int ocfs2_mount_local(struct ocfs2_super *osb)
353{
354 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
355}
356
350#define OCFS2_IS_VALID_DINODE(ptr) \ 357#define OCFS2_IS_VALID_DINODE(ptr) \
351 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) 358 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
352 359
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 3330a5dc6be2..b5c68567077e 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -86,7 +86,7 @@
86 OCFS2_SB(sb)->s_feature_incompat &= ~(mask) 86 OCFS2_SB(sb)->s_feature_incompat &= ~(mask)
87 87
88#define OCFS2_FEATURE_COMPAT_SUPP 0 88#define OCFS2_FEATURE_COMPAT_SUPP 0
89#define OCFS2_FEATURE_INCOMPAT_SUPP 0 89#define OCFS2_FEATURE_INCOMPAT_SUPP OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT
90#define OCFS2_FEATURE_RO_COMPAT_SUPP 0 90#define OCFS2_FEATURE_RO_COMPAT_SUPP 0
91 91
92/* 92/*
@@ -96,6 +96,18 @@
96 */ 96 */
97#define OCFS2_FEATURE_INCOMPAT_HEARTBEAT_DEV 0x0002 97#define OCFS2_FEATURE_INCOMPAT_HEARTBEAT_DEV 0x0002
98 98
99/*
100 * tunefs sets this incompat flag before starting the resize and clears it
101 * at the end. This flag protects users from inadvertently mounting the fs
102 * after an aborted run without fsck-ing.
103 */
104#define OCFS2_FEATURE_INCOMPAT_RESIZE_INPROG 0x0004
105
106/* Used to denote a non-clustered volume */
107#define OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT 0x0008
108
109/* Support for sparse allocation in b-trees */
110#define OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC 0x0010
99 111
100/* 112/*
101 * Flags on ocfs2_dinode.i_flags 113 * Flags on ocfs2_dinode.i_flags
diff --git a/fs/ocfs2/slot_map.c b/fs/ocfs2/slot_map.c
index aa6f5aadedc4..2d3ac32cb74e 100644
--- a/fs/ocfs2/slot_map.c
+++ b/fs/ocfs2/slot_map.c
@@ -175,7 +175,7 @@ int ocfs2_init_slot_info(struct ocfs2_super *osb)
175 struct buffer_head *bh = NULL; 175 struct buffer_head *bh = NULL;
176 struct ocfs2_slot_info *si; 176 struct ocfs2_slot_info *si;
177 177
178 si = kcalloc(1, sizeof(struct ocfs2_slot_info), GFP_KERNEL); 178 si = kzalloc(sizeof(struct ocfs2_slot_info), GFP_KERNEL);
179 if (!si) { 179 if (!si) {
180 status = -ENOMEM; 180 status = -ENOMEM;
181 mlog_errno(status); 181 mlog_errno(status);
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 9d91e66f51a9..6dbb11762759 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -49,7 +49,7 @@
49static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg); 49static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg);
50static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe); 50static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe);
51static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl); 51static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl);
52static int ocfs2_block_group_fill(struct ocfs2_journal_handle *handle, 52static int ocfs2_block_group_fill(handle_t *handle,
53 struct inode *alloc_inode, 53 struct inode *alloc_inode,
54 struct buffer_head *bg_bh, 54 struct buffer_head *bg_bh,
55 u64 group_blkno, 55 u64 group_blkno,
@@ -59,9 +59,6 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
59 struct inode *alloc_inode, 59 struct inode *alloc_inode,
60 struct buffer_head *bh); 60 struct buffer_head *bh);
61 61
62static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb,
63 struct ocfs2_alloc_context *ac);
64
65static int ocfs2_cluster_group_search(struct inode *inode, 62static int ocfs2_cluster_group_search(struct inode *inode,
66 struct buffer_head *group_bh, 63 struct buffer_head *group_bh,
67 u32 bits_wanted, u32 min_bits, 64 u32 bits_wanted, u32 min_bits,
@@ -72,6 +69,7 @@ static int ocfs2_block_group_search(struct inode *inode,
72 u16 *bit_off, u16 *bits_found); 69 u16 *bit_off, u16 *bits_found);
73static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, 70static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
74 struct ocfs2_alloc_context *ac, 71 struct ocfs2_alloc_context *ac,
72 handle_t *handle,
75 u32 bits_wanted, 73 u32 bits_wanted,
76 u32 min_bits, 74 u32 min_bits,
77 u16 *bit_off, 75 u16 *bit_off,
@@ -79,20 +77,20 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
79 u64 *bg_blkno); 77 u64 *bg_blkno);
80static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh, 78static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
81 int nr); 79 int nr);
82static inline int ocfs2_block_group_set_bits(struct ocfs2_journal_handle *handle, 80static inline int ocfs2_block_group_set_bits(handle_t *handle,
83 struct inode *alloc_inode, 81 struct inode *alloc_inode,
84 struct ocfs2_group_desc *bg, 82 struct ocfs2_group_desc *bg,
85 struct buffer_head *group_bh, 83 struct buffer_head *group_bh,
86 unsigned int bit_off, 84 unsigned int bit_off,
87 unsigned int num_bits); 85 unsigned int num_bits);
88static inline int ocfs2_block_group_clear_bits(struct ocfs2_journal_handle *handle, 86static inline int ocfs2_block_group_clear_bits(handle_t *handle,
89 struct inode *alloc_inode, 87 struct inode *alloc_inode,
90 struct ocfs2_group_desc *bg, 88 struct ocfs2_group_desc *bg,
91 struct buffer_head *group_bh, 89 struct buffer_head *group_bh,
92 unsigned int bit_off, 90 unsigned int bit_off,
93 unsigned int num_bits); 91 unsigned int num_bits);
94 92
95static int ocfs2_relink_block_group(struct ocfs2_journal_handle *handle, 93static int ocfs2_relink_block_group(handle_t *handle,
96 struct inode *alloc_inode, 94 struct inode *alloc_inode,
97 struct buffer_head *fe_bh, 95 struct buffer_head *fe_bh,
98 struct buffer_head *bg_bh, 96 struct buffer_head *bg_bh,
@@ -100,7 +98,7 @@ static int ocfs2_relink_block_group(struct ocfs2_journal_handle *handle,
100 u16 chain); 98 u16 chain);
101static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg, 99static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg,
102 u32 wanted); 100 u32 wanted);
103static int ocfs2_free_suballoc_bits(struct ocfs2_journal_handle *handle, 101static int ocfs2_free_suballoc_bits(handle_t *handle,
104 struct inode *alloc_inode, 102 struct inode *alloc_inode,
105 struct buffer_head *alloc_bh, 103 struct buffer_head *alloc_bh,
106 unsigned int start_bit, 104 unsigned int start_bit,
@@ -120,8 +118,16 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode,
120 118
121void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac) 119void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac)
122{ 120{
123 if (ac->ac_inode) 121 struct inode *inode = ac->ac_inode;
124 iput(ac->ac_inode); 122
123 if (inode) {
124 if (ac->ac_which != OCFS2_AC_USE_LOCAL)
125 ocfs2_meta_unlock(inode, 1);
126
127 mutex_unlock(&inode->i_mutex);
128
129 iput(inode);
130 }
125 if (ac->ac_bh) 131 if (ac->ac_bh)
126 brelse(ac->ac_bh); 132 brelse(ac->ac_bh);
127 kfree(ac); 133 kfree(ac);
@@ -190,7 +196,7 @@ static int ocfs2_check_group_descriptor(struct super_block *sb,
190 return 0; 196 return 0;
191} 197}
192 198
193static int ocfs2_block_group_fill(struct ocfs2_journal_handle *handle, 199static int ocfs2_block_group_fill(handle_t *handle,
194 struct inode *alloc_inode, 200 struct inode *alloc_inode,
195 struct buffer_head *bg_bh, 201 struct buffer_head *bg_bh,
196 u64 group_blkno, 202 u64 group_blkno,
@@ -273,7 +279,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
273 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data; 279 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data;
274 struct ocfs2_chain_list *cl; 280 struct ocfs2_chain_list *cl;
275 struct ocfs2_alloc_context *ac = NULL; 281 struct ocfs2_alloc_context *ac = NULL;
276 struct ocfs2_journal_handle *handle = NULL; 282 handle_t *handle = NULL;
277 u32 bit_off, num_bits; 283 u32 bit_off, num_bits;
278 u16 alloc_rec; 284 u16 alloc_rec;
279 u64 bg_blkno; 285 u64 bg_blkno;
@@ -284,16 +290,8 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
284 290
285 mlog_entry_void(); 291 mlog_entry_void();
286 292
287 handle = ocfs2_alloc_handle(osb);
288 if (!handle) {
289 status = -ENOMEM;
290 mlog_errno(status);
291 goto bail;
292 }
293
294 cl = &fe->id2.i_chain; 293 cl = &fe->id2.i_chain;
295 status = ocfs2_reserve_clusters(osb, 294 status = ocfs2_reserve_clusters(osb,
296 handle,
297 le16_to_cpu(cl->cl_cpg), 295 le16_to_cpu(cl->cl_cpg),
298 &ac); 296 &ac);
299 if (status < 0) { 297 if (status < 0) {
@@ -304,7 +302,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
304 302
305 credits = ocfs2_calc_group_alloc_credits(osb->sb, 303 credits = ocfs2_calc_group_alloc_credits(osb->sb,
306 le16_to_cpu(cl->cl_cpg)); 304 le16_to_cpu(cl->cl_cpg));
307 handle = ocfs2_start_trans(osb, handle, credits); 305 handle = ocfs2_start_trans(osb, credits);
308 if (IS_ERR(handle)) { 306 if (IS_ERR(handle)) {
309 status = PTR_ERR(handle); 307 status = PTR_ERR(handle);
310 handle = NULL; 308 handle = NULL;
@@ -389,7 +387,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
389 status = 0; 387 status = 0;
390bail: 388bail:
391 if (handle) 389 if (handle)
392 ocfs2_commit_trans(handle); 390 ocfs2_commit_trans(osb, handle);
393 391
394 if (ac) 392 if (ac)
395 ocfs2_free_alloc_context(ac); 393 ocfs2_free_alloc_context(ac);
@@ -402,27 +400,38 @@ bail:
402} 400}
403 401
404static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb, 402static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb,
405 struct ocfs2_alloc_context *ac) 403 struct ocfs2_alloc_context *ac,
404 int type,
405 u32 slot)
406{ 406{
407 int status; 407 int status;
408 u32 bits_wanted = ac->ac_bits_wanted; 408 u32 bits_wanted = ac->ac_bits_wanted;
409 struct inode *alloc_inode = ac->ac_inode; 409 struct inode *alloc_inode;
410 struct buffer_head *bh = NULL; 410 struct buffer_head *bh = NULL;
411 struct ocfs2_journal_handle *handle = ac->ac_handle;
412 struct ocfs2_dinode *fe; 411 struct ocfs2_dinode *fe;
413 u32 free_bits; 412 u32 free_bits;
414 413
415 mlog_entry_void(); 414 mlog_entry_void();
416 415
417 BUG_ON(handle->flags & OCFS2_HANDLE_STARTED); 416 alloc_inode = ocfs2_get_system_file_inode(osb, type, slot);
417 if (!alloc_inode) {
418 mlog_errno(-EINVAL);
419 return -EINVAL;
420 }
418 421
419 ocfs2_handle_add_inode(handle, alloc_inode); 422 mutex_lock(&alloc_inode->i_mutex);
420 status = ocfs2_meta_lock(alloc_inode, handle, &bh, 1); 423
424 status = ocfs2_meta_lock(alloc_inode, &bh, 1);
421 if (status < 0) { 425 if (status < 0) {
426 mutex_unlock(&alloc_inode->i_mutex);
427 iput(alloc_inode);
428
422 mlog_errno(status); 429 mlog_errno(status);
423 goto bail; 430 return status;
424 } 431 }
425 432
433 ac->ac_inode = alloc_inode;
434
426 fe = (struct ocfs2_dinode *) bh->b_data; 435 fe = (struct ocfs2_dinode *) bh->b_data;
427 if (!OCFS2_IS_VALID_DINODE(fe)) { 436 if (!OCFS2_IS_VALID_DINODE(fe)) {
428 OCFS2_RO_ON_INVALID_DINODE(alloc_inode->i_sb, fe); 437 OCFS2_RO_ON_INVALID_DINODE(alloc_inode->i_sb, fe);
@@ -473,14 +482,13 @@ bail:
473} 482}
474 483
475int ocfs2_reserve_new_metadata(struct ocfs2_super *osb, 484int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
476 struct ocfs2_journal_handle *handle,
477 struct ocfs2_dinode *fe, 485 struct ocfs2_dinode *fe,
478 struct ocfs2_alloc_context **ac) 486 struct ocfs2_alloc_context **ac)
479{ 487{
480 int status; 488 int status;
481 struct inode *alloc_inode = NULL; 489 u32 slot;
482 490
483 *ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL); 491 *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
484 if (!(*ac)) { 492 if (!(*ac)) {
485 status = -ENOMEM; 493 status = -ENOMEM;
486 mlog_errno(status); 494 mlog_errno(status);
@@ -488,28 +496,18 @@ int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
488 } 496 }
489 497
490 (*ac)->ac_bits_wanted = ocfs2_extend_meta_needed(fe); 498 (*ac)->ac_bits_wanted = ocfs2_extend_meta_needed(fe);
491 (*ac)->ac_handle = handle;
492 (*ac)->ac_which = OCFS2_AC_USE_META; 499 (*ac)->ac_which = OCFS2_AC_USE_META;
493 500
494#ifndef OCFS2_USE_ALL_METADATA_SUBALLOCATORS 501#ifndef OCFS2_USE_ALL_METADATA_SUBALLOCATORS
495 alloc_inode = ocfs2_get_system_file_inode(osb, 502 slot = 0;
496 EXTENT_ALLOC_SYSTEM_INODE,
497 0);
498#else 503#else
499 alloc_inode = ocfs2_get_system_file_inode(osb, 504 slot = osb->slot_num;
500 EXTENT_ALLOC_SYSTEM_INODE,
501 osb->slot_num);
502#endif 505#endif
503 if (!alloc_inode) {
504 status = -ENOMEM;
505 mlog_errno(status);
506 goto bail;
507 }
508 506
509 (*ac)->ac_inode = igrab(alloc_inode);
510 (*ac)->ac_group_search = ocfs2_block_group_search; 507 (*ac)->ac_group_search = ocfs2_block_group_search;
511 508
512 status = ocfs2_reserve_suballoc_bits(osb, (*ac)); 509 status = ocfs2_reserve_suballoc_bits(osb, (*ac),
510 EXTENT_ALLOC_SYSTEM_INODE, slot);
513 if (status < 0) { 511 if (status < 0) {
514 if (status != -ENOSPC) 512 if (status != -ENOSPC)
515 mlog_errno(status); 513 mlog_errno(status);
@@ -523,21 +521,16 @@ bail:
523 *ac = NULL; 521 *ac = NULL;
524 } 522 }
525 523
526 if (alloc_inode)
527 iput(alloc_inode);
528
529 mlog_exit(status); 524 mlog_exit(status);
530 return status; 525 return status;
531} 526}
532 527
533int ocfs2_reserve_new_inode(struct ocfs2_super *osb, 528int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
534 struct ocfs2_journal_handle *handle,
535 struct ocfs2_alloc_context **ac) 529 struct ocfs2_alloc_context **ac)
536{ 530{
537 int status; 531 int status;
538 struct inode *alloc_inode = NULL;
539 532
540 *ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL); 533 *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
541 if (!(*ac)) { 534 if (!(*ac)) {
542 status = -ENOMEM; 535 status = -ENOMEM;
543 mlog_errno(status); 536 mlog_errno(status);
@@ -545,22 +538,13 @@ int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
545 } 538 }
546 539
547 (*ac)->ac_bits_wanted = 1; 540 (*ac)->ac_bits_wanted = 1;
548 (*ac)->ac_handle = handle;
549 (*ac)->ac_which = OCFS2_AC_USE_INODE; 541 (*ac)->ac_which = OCFS2_AC_USE_INODE;
550 542
551 alloc_inode = ocfs2_get_system_file_inode(osb,
552 INODE_ALLOC_SYSTEM_INODE,
553 osb->slot_num);
554 if (!alloc_inode) {
555 status = -ENOMEM;
556 mlog_errno(status);
557 goto bail;
558 }
559
560 (*ac)->ac_inode = igrab(alloc_inode);
561 (*ac)->ac_group_search = ocfs2_block_group_search; 543 (*ac)->ac_group_search = ocfs2_block_group_search;
562 544
563 status = ocfs2_reserve_suballoc_bits(osb, *ac); 545 status = ocfs2_reserve_suballoc_bits(osb, *ac,
546 INODE_ALLOC_SYSTEM_INODE,
547 osb->slot_num);
564 if (status < 0) { 548 if (status < 0) {
565 if (status != -ENOSPC) 549 if (status != -ENOSPC)
566 mlog_errno(status); 550 mlog_errno(status);
@@ -574,9 +558,6 @@ bail:
574 *ac = NULL; 558 *ac = NULL;
575 } 559 }
576 560
577 if (alloc_inode)
578 iput(alloc_inode);
579
580 mlog_exit(status); 561 mlog_exit(status);
581 return status; 562 return status;
582} 563}
@@ -588,20 +569,17 @@ int ocfs2_reserve_cluster_bitmap_bits(struct ocfs2_super *osb,
588{ 569{
589 int status; 570 int status;
590 571
591 ac->ac_inode = ocfs2_get_system_file_inode(osb,
592 GLOBAL_BITMAP_SYSTEM_INODE,
593 OCFS2_INVALID_SLOT);
594 if (!ac->ac_inode) {
595 status = -EINVAL;
596 mlog(ML_ERROR, "Could not get bitmap inode!\n");
597 goto bail;
598 }
599 ac->ac_which = OCFS2_AC_USE_MAIN; 572 ac->ac_which = OCFS2_AC_USE_MAIN;
600 ac->ac_group_search = ocfs2_cluster_group_search; 573 ac->ac_group_search = ocfs2_cluster_group_search;
601 574
602 status = ocfs2_reserve_suballoc_bits(osb, ac); 575 status = ocfs2_reserve_suballoc_bits(osb, ac,
603 if (status < 0 && status != -ENOSPC) 576 GLOBAL_BITMAP_SYSTEM_INODE,
577 OCFS2_INVALID_SLOT);
578 if (status < 0 && status != -ENOSPC) {
604 mlog_errno(status); 579 mlog_errno(status);
580 goto bail;
581 }
582
605bail: 583bail:
606 return status; 584 return status;
607} 585}
@@ -610,7 +588,6 @@ bail:
610 * use so we figure it out for them, but unfortunately this clutters 588 * use so we figure it out for them, but unfortunately this clutters
611 * things a bit. */ 589 * things a bit. */
612int ocfs2_reserve_clusters(struct ocfs2_super *osb, 590int ocfs2_reserve_clusters(struct ocfs2_super *osb,
613 struct ocfs2_journal_handle *handle,
614 u32 bits_wanted, 591 u32 bits_wanted,
615 struct ocfs2_alloc_context **ac) 592 struct ocfs2_alloc_context **ac)
616{ 593{
@@ -618,9 +595,7 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb,
618 595
619 mlog_entry_void(); 596 mlog_entry_void();
620 597
621 BUG_ON(!handle); 598 *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
622
623 *ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
624 if (!(*ac)) { 599 if (!(*ac)) {
625 status = -ENOMEM; 600 status = -ENOMEM;
626 mlog_errno(status); 601 mlog_errno(status);
@@ -628,12 +603,10 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb,
628 } 603 }
629 604
630 (*ac)->ac_bits_wanted = bits_wanted; 605 (*ac)->ac_bits_wanted = bits_wanted;
631 (*ac)->ac_handle = handle;
632 606
633 status = -ENOSPC; 607 status = -ENOSPC;
634 if (ocfs2_alloc_should_use_local(osb, bits_wanted)) { 608 if (ocfs2_alloc_should_use_local(osb, bits_wanted)) {
635 status = ocfs2_reserve_local_alloc_bits(osb, 609 status = ocfs2_reserve_local_alloc_bits(osb,
636 handle,
637 bits_wanted, 610 bits_wanted,
638 *ac); 611 *ac);
639 if ((status < 0) && (status != -ENOSPC)) { 612 if ((status < 0) && (status != -ENOSPC)) {
@@ -774,7 +747,7 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
774 return status; 747 return status;
775} 748}
776 749
777static inline int ocfs2_block_group_set_bits(struct ocfs2_journal_handle *handle, 750static inline int ocfs2_block_group_set_bits(handle_t *handle,
778 struct inode *alloc_inode, 751 struct inode *alloc_inode,
779 struct ocfs2_group_desc *bg, 752 struct ocfs2_group_desc *bg,
780 struct buffer_head *group_bh, 753 struct buffer_head *group_bh,
@@ -845,7 +818,7 @@ static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl)
845 return best; 818 return best;
846} 819}
847 820
848static int ocfs2_relink_block_group(struct ocfs2_journal_handle *handle, 821static int ocfs2_relink_block_group(handle_t *handle,
849 struct inode *alloc_inode, 822 struct inode *alloc_inode,
850 struct buffer_head *fe_bh, 823 struct buffer_head *fe_bh,
851 struct buffer_head *bg_bh, 824 struct buffer_head *bg_bh,
@@ -1025,7 +998,7 @@ static int ocfs2_block_group_search(struct inode *inode,
1025} 998}
1026 999
1027static int ocfs2_alloc_dinode_update_counts(struct inode *inode, 1000static int ocfs2_alloc_dinode_update_counts(struct inode *inode,
1028 struct ocfs2_journal_handle *handle, 1001 handle_t *handle,
1029 struct buffer_head *di_bh, 1002 struct buffer_head *di_bh,
1030 u32 num_bits, 1003 u32 num_bits,
1031 u16 chain) 1004 u16 chain)
@@ -1055,6 +1028,7 @@ out:
1055} 1028}
1056 1029
1057static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac, 1030static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
1031 handle_t *handle,
1058 u32 bits_wanted, 1032 u32 bits_wanted,
1059 u32 min_bits, 1033 u32 min_bits,
1060 u16 *bit_off, 1034 u16 *bit_off,
@@ -1067,7 +1041,6 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
1067 struct buffer_head *group_bh = NULL; 1041 struct buffer_head *group_bh = NULL;
1068 struct ocfs2_group_desc *gd; 1042 struct ocfs2_group_desc *gd;
1069 struct inode *alloc_inode = ac->ac_inode; 1043 struct inode *alloc_inode = ac->ac_inode;
1070 struct ocfs2_journal_handle *handle = ac->ac_handle;
1071 1044
1072 ret = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb), gd_blkno, 1045 ret = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb), gd_blkno,
1073 &group_bh, OCFS2_BH_CACHED, alloc_inode); 1046 &group_bh, OCFS2_BH_CACHED, alloc_inode);
@@ -1115,6 +1088,7 @@ out:
1115} 1088}
1116 1089
1117static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, 1090static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
1091 handle_t *handle,
1118 u32 bits_wanted, 1092 u32 bits_wanted,
1119 u32 min_bits, 1093 u32 min_bits,
1120 u16 *bit_off, 1094 u16 *bit_off,
@@ -1126,7 +1100,6 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
1126 u16 chain, tmp_bits; 1100 u16 chain, tmp_bits;
1127 u32 tmp_used; 1101 u32 tmp_used;
1128 u64 next_group; 1102 u64 next_group;
1129 struct ocfs2_journal_handle *handle = ac->ac_handle;
1130 struct inode *alloc_inode = ac->ac_inode; 1103 struct inode *alloc_inode = ac->ac_inode;
1131 struct buffer_head *group_bh = NULL; 1104 struct buffer_head *group_bh = NULL;
1132 struct buffer_head *prev_group_bh = NULL; 1105 struct buffer_head *prev_group_bh = NULL;
@@ -1272,6 +1245,7 @@ bail:
1272/* will give out up to bits_wanted contiguous bits. */ 1245/* will give out up to bits_wanted contiguous bits. */
1273static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, 1246static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
1274 struct ocfs2_alloc_context *ac, 1247 struct ocfs2_alloc_context *ac,
1248 handle_t *handle,
1275 u32 bits_wanted, 1249 u32 bits_wanted,
1276 u32 min_bits, 1250 u32 min_bits,
1277 u16 *bit_off, 1251 u16 *bit_off,
@@ -1313,8 +1287,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
1313 * by jumping straight to the most recently used 1287 * by jumping straight to the most recently used
1314 * allocation group. This helps us mantain some 1288 * allocation group. This helps us mantain some
1315 * contiguousness across allocations. */ 1289 * contiguousness across allocations. */
1316 status = ocfs2_search_one_group(ac, bits_wanted, min_bits, 1290 status = ocfs2_search_one_group(ac, handle, bits_wanted,
1317 bit_off, num_bits, 1291 min_bits, bit_off, num_bits,
1318 hint_blkno, &bits_left); 1292 hint_blkno, &bits_left);
1319 if (!status) { 1293 if (!status) {
1320 /* Be careful to update *bg_blkno here as the 1294 /* Be careful to update *bg_blkno here as the
@@ -1336,7 +1310,7 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
1336 ac->ac_chain = victim; 1310 ac->ac_chain = victim;
1337 ac->ac_allow_chain_relink = 1; 1311 ac->ac_allow_chain_relink = 1;
1338 1312
1339 status = ocfs2_search_chain(ac, bits_wanted, min_bits, bit_off, 1313 status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, bit_off,
1340 num_bits, bg_blkno, &bits_left); 1314 num_bits, bg_blkno, &bits_left);
1341 if (!status) 1315 if (!status)
1342 goto set_hint; 1316 goto set_hint;
@@ -1360,7 +1334,7 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
1360 continue; 1334 continue;
1361 1335
1362 ac->ac_chain = i; 1336 ac->ac_chain = i;
1363 status = ocfs2_search_chain(ac, bits_wanted, min_bits, 1337 status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
1364 bit_off, num_bits, bg_blkno, 1338 bit_off, num_bits, bg_blkno,
1365 &bits_left); 1339 &bits_left);
1366 if (!status) 1340 if (!status)
@@ -1388,7 +1362,7 @@ bail:
1388} 1362}
1389 1363
1390int ocfs2_claim_metadata(struct ocfs2_super *osb, 1364int ocfs2_claim_metadata(struct ocfs2_super *osb,
1391 struct ocfs2_journal_handle *handle, 1365 handle_t *handle,
1392 struct ocfs2_alloc_context *ac, 1366 struct ocfs2_alloc_context *ac,
1393 u32 bits_wanted, 1367 u32 bits_wanted,
1394 u16 *suballoc_bit_start, 1368 u16 *suballoc_bit_start,
@@ -1401,10 +1375,10 @@ int ocfs2_claim_metadata(struct ocfs2_super *osb,
1401 BUG_ON(!ac); 1375 BUG_ON(!ac);
1402 BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted)); 1376 BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
1403 BUG_ON(ac->ac_which != OCFS2_AC_USE_META); 1377 BUG_ON(ac->ac_which != OCFS2_AC_USE_META);
1404 BUG_ON(ac->ac_handle != handle);
1405 1378
1406 status = ocfs2_claim_suballoc_bits(osb, 1379 status = ocfs2_claim_suballoc_bits(osb,
1407 ac, 1380 ac,
1381 handle,
1408 bits_wanted, 1382 bits_wanted,
1409 1, 1383 1,
1410 suballoc_bit_start, 1384 suballoc_bit_start,
@@ -1425,7 +1399,7 @@ bail:
1425} 1399}
1426 1400
1427int ocfs2_claim_new_inode(struct ocfs2_super *osb, 1401int ocfs2_claim_new_inode(struct ocfs2_super *osb,
1428 struct ocfs2_journal_handle *handle, 1402 handle_t *handle,
1429 struct ocfs2_alloc_context *ac, 1403 struct ocfs2_alloc_context *ac,
1430 u16 *suballoc_bit, 1404 u16 *suballoc_bit,
1431 u64 *fe_blkno) 1405 u64 *fe_blkno)
@@ -1440,10 +1414,10 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
1440 BUG_ON(ac->ac_bits_given != 0); 1414 BUG_ON(ac->ac_bits_given != 0);
1441 BUG_ON(ac->ac_bits_wanted != 1); 1415 BUG_ON(ac->ac_bits_wanted != 1);
1442 BUG_ON(ac->ac_which != OCFS2_AC_USE_INODE); 1416 BUG_ON(ac->ac_which != OCFS2_AC_USE_INODE);
1443 BUG_ON(ac->ac_handle != handle);
1444 1417
1445 status = ocfs2_claim_suballoc_bits(osb, 1418 status = ocfs2_claim_suballoc_bits(osb,
1446 ac, 1419 ac,
1420 handle,
1447 1, 1421 1,
1448 1, 1422 1,
1449 suballoc_bit, 1423 suballoc_bit,
@@ -1528,7 +1502,7 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode,
1528 * of any size. 1502 * of any size.
1529 */ 1503 */
1530int ocfs2_claim_clusters(struct ocfs2_super *osb, 1504int ocfs2_claim_clusters(struct ocfs2_super *osb,
1531 struct ocfs2_journal_handle *handle, 1505 handle_t *handle,
1532 struct ocfs2_alloc_context *ac, 1506 struct ocfs2_alloc_context *ac,
1533 u32 min_clusters, 1507 u32 min_clusters,
1534 u32 *cluster_start, 1508 u32 *cluster_start,
@@ -1546,7 +1520,6 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
1546 1520
1547 BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL 1521 BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL
1548 && ac->ac_which != OCFS2_AC_USE_MAIN); 1522 && ac->ac_which != OCFS2_AC_USE_MAIN);
1549 BUG_ON(ac->ac_handle != handle);
1550 1523
1551 if (ac->ac_which == OCFS2_AC_USE_LOCAL) { 1524 if (ac->ac_which == OCFS2_AC_USE_LOCAL) {
1552 status = ocfs2_claim_local_alloc_bits(osb, 1525 status = ocfs2_claim_local_alloc_bits(osb,
@@ -1572,6 +1545,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
1572 1545
1573 status = ocfs2_claim_suballoc_bits(osb, 1546 status = ocfs2_claim_suballoc_bits(osb,
1574 ac, 1547 ac,
1548 handle,
1575 bits_wanted, 1549 bits_wanted,
1576 min_clusters, 1550 min_clusters,
1577 &bg_bit_off, 1551 &bg_bit_off,
@@ -1598,7 +1572,7 @@ bail:
1598 return status; 1572 return status;
1599} 1573}
1600 1574
1601static inline int ocfs2_block_group_clear_bits(struct ocfs2_journal_handle *handle, 1575static inline int ocfs2_block_group_clear_bits(handle_t *handle,
1602 struct inode *alloc_inode, 1576 struct inode *alloc_inode,
1603 struct ocfs2_group_desc *bg, 1577 struct ocfs2_group_desc *bg,
1604 struct buffer_head *group_bh, 1578 struct buffer_head *group_bh,
@@ -1653,7 +1627,7 @@ bail:
1653/* 1627/*
1654 * expects the suballoc inode to already be locked. 1628 * expects the suballoc inode to already be locked.
1655 */ 1629 */
1656static int ocfs2_free_suballoc_bits(struct ocfs2_journal_handle *handle, 1630static int ocfs2_free_suballoc_bits(handle_t *handle,
1657 struct inode *alloc_inode, 1631 struct inode *alloc_inode,
1658 struct buffer_head *alloc_bh, 1632 struct buffer_head *alloc_bh,
1659 unsigned int start_bit, 1633 unsigned int start_bit,
@@ -1737,7 +1711,7 @@ static inline u64 ocfs2_which_suballoc_group(u64 block, unsigned int bit)
1737 return group; 1711 return group;
1738} 1712}
1739 1713
1740int ocfs2_free_dinode(struct ocfs2_journal_handle *handle, 1714int ocfs2_free_dinode(handle_t *handle,
1741 struct inode *inode_alloc_inode, 1715 struct inode *inode_alloc_inode,
1742 struct buffer_head *inode_alloc_bh, 1716 struct buffer_head *inode_alloc_bh,
1743 struct ocfs2_dinode *di) 1717 struct ocfs2_dinode *di)
@@ -1750,7 +1724,7 @@ int ocfs2_free_dinode(struct ocfs2_journal_handle *handle,
1750 inode_alloc_bh, bit, bg_blkno, 1); 1724 inode_alloc_bh, bit, bg_blkno, 1);
1751} 1725}
1752 1726
1753int ocfs2_free_extent_block(struct ocfs2_journal_handle *handle, 1727int ocfs2_free_extent_block(handle_t *handle,
1754 struct inode *eb_alloc_inode, 1728 struct inode *eb_alloc_inode,
1755 struct buffer_head *eb_alloc_bh, 1729 struct buffer_head *eb_alloc_bh,
1756 struct ocfs2_extent_block *eb) 1730 struct ocfs2_extent_block *eb)
@@ -1763,7 +1737,7 @@ int ocfs2_free_extent_block(struct ocfs2_journal_handle *handle,
1763 bit, bg_blkno, 1); 1737 bit, bg_blkno, 1);
1764} 1738}
1765 1739
1766int ocfs2_free_clusters(struct ocfs2_journal_handle *handle, 1740int ocfs2_free_clusters(handle_t *handle,
1767 struct inode *bitmap_inode, 1741 struct inode *bitmap_inode,
1768 struct buffer_head *bitmap_bh, 1742 struct buffer_head *bitmap_bh,
1769 u64 start_blk, 1743 u64 start_blk,
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index c787838d1052..1a3c94cb9250 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -43,7 +43,6 @@ struct ocfs2_alloc_context {
43#define OCFS2_AC_USE_INODE 3 43#define OCFS2_AC_USE_INODE 3
44#define OCFS2_AC_USE_META 4 44#define OCFS2_AC_USE_META 4
45 u32 ac_which; 45 u32 ac_which;
46 struct ocfs2_journal_handle *ac_handle;
47 46
48 /* these are used by the chain search */ 47 /* these are used by the chain search */
49 u16 ac_chain; 48 u16 ac_chain;
@@ -60,45 +59,42 @@ static inline int ocfs2_alloc_context_bits_left(struct ocfs2_alloc_context *ac)
60} 59}
61 60
62int ocfs2_reserve_new_metadata(struct ocfs2_super *osb, 61int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
63 struct ocfs2_journal_handle *handle,
64 struct ocfs2_dinode *fe, 62 struct ocfs2_dinode *fe,
65 struct ocfs2_alloc_context **ac); 63 struct ocfs2_alloc_context **ac);
66int ocfs2_reserve_new_inode(struct ocfs2_super *osb, 64int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
67 struct ocfs2_journal_handle *handle,
68 struct ocfs2_alloc_context **ac); 65 struct ocfs2_alloc_context **ac);
69int ocfs2_reserve_clusters(struct ocfs2_super *osb, 66int ocfs2_reserve_clusters(struct ocfs2_super *osb,
70 struct ocfs2_journal_handle *handle,
71 u32 bits_wanted, 67 u32 bits_wanted,
72 struct ocfs2_alloc_context **ac); 68 struct ocfs2_alloc_context **ac);
73 69
74int ocfs2_claim_metadata(struct ocfs2_super *osb, 70int ocfs2_claim_metadata(struct ocfs2_super *osb,
75 struct ocfs2_journal_handle *handle, 71 handle_t *handle,
76 struct ocfs2_alloc_context *ac, 72 struct ocfs2_alloc_context *ac,
77 u32 bits_wanted, 73 u32 bits_wanted,
78 u16 *suballoc_bit_start, 74 u16 *suballoc_bit_start,
79 u32 *num_bits, 75 u32 *num_bits,
80 u64 *blkno_start); 76 u64 *blkno_start);
81int ocfs2_claim_new_inode(struct ocfs2_super *osb, 77int ocfs2_claim_new_inode(struct ocfs2_super *osb,
82 struct ocfs2_journal_handle *handle, 78 handle_t *handle,
83 struct ocfs2_alloc_context *ac, 79 struct ocfs2_alloc_context *ac,
84 u16 *suballoc_bit, 80 u16 *suballoc_bit,
85 u64 *fe_blkno); 81 u64 *fe_blkno);
86int ocfs2_claim_clusters(struct ocfs2_super *osb, 82int ocfs2_claim_clusters(struct ocfs2_super *osb,
87 struct ocfs2_journal_handle *handle, 83 handle_t *handle,
88 struct ocfs2_alloc_context *ac, 84 struct ocfs2_alloc_context *ac,
89 u32 min_clusters, 85 u32 min_clusters,
90 u32 *cluster_start, 86 u32 *cluster_start,
91 u32 *num_clusters); 87 u32 *num_clusters);
92 88
93int ocfs2_free_dinode(struct ocfs2_journal_handle *handle, 89int ocfs2_free_dinode(handle_t *handle,
94 struct inode *inode_alloc_inode, 90 struct inode *inode_alloc_inode,
95 struct buffer_head *inode_alloc_bh, 91 struct buffer_head *inode_alloc_bh,
96 struct ocfs2_dinode *di); 92 struct ocfs2_dinode *di);
97int ocfs2_free_extent_block(struct ocfs2_journal_handle *handle, 93int ocfs2_free_extent_block(handle_t *handle,
98 struct inode *eb_alloc_inode, 94 struct inode *eb_alloc_inode,
99 struct buffer_head *eb_alloc_bh, 95 struct buffer_head *eb_alloc_bh,
100 struct ocfs2_extent_block *eb); 96 struct ocfs2_extent_block *eb);
101int ocfs2_free_clusters(struct ocfs2_journal_handle *handle, 97int ocfs2_free_clusters(handle_t *handle,
102 struct inode *bitmap_inode, 98 struct inode *bitmap_inode,
103 struct buffer_head *bitmap_bh, 99 struct buffer_head *bitmap_bh,
104 u64 start_blk, 100 u64 start_blk,
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index 76b46ebbb10c..6e300a88a47e 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -68,9 +68,7 @@
68 68
69#include "buffer_head_io.h" 69#include "buffer_head_io.h"
70 70
71static kmem_cache_t *ocfs2_inode_cachep = NULL; 71static struct kmem_cache *ocfs2_inode_cachep = NULL;
72
73kmem_cache_t *ocfs2_lock_cache = NULL;
74 72
75/* OCFS2 needs to schedule several differnt types of work which 73/* OCFS2 needs to schedule several differnt types of work which
76 * require cluster locking, disk I/O, recovery waits, etc. Since these 74 * require cluster locking, disk I/O, recovery waits, etc. Since these
@@ -141,6 +139,7 @@ enum {
141 Opt_hb_local, 139 Opt_hb_local,
142 Opt_data_ordered, 140 Opt_data_ordered,
143 Opt_data_writeback, 141 Opt_data_writeback,
142 Opt_atime_quantum,
144 Opt_err, 143 Opt_err,
145}; 144};
146 145
@@ -154,6 +153,7 @@ static match_table_t tokens = {
154 {Opt_hb_local, OCFS2_HB_LOCAL}, 153 {Opt_hb_local, OCFS2_HB_LOCAL},
155 {Opt_data_ordered, "data=ordered"}, 154 {Opt_data_ordered, "data=ordered"},
156 {Opt_data_writeback, "data=writeback"}, 155 {Opt_data_writeback, "data=writeback"},
156 {Opt_atime_quantum, "atime_quantum=%u"},
157 {Opt_err, NULL} 157 {Opt_err, NULL}
158}; 158};
159 159
@@ -303,7 +303,7 @@ static struct inode *ocfs2_alloc_inode(struct super_block *sb)
303{ 303{
304 struct ocfs2_inode_info *oi; 304 struct ocfs2_inode_info *oi;
305 305
306 oi = kmem_cache_alloc(ocfs2_inode_cachep, SLAB_NOFS); 306 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
307 if (!oi) 307 if (!oi)
308 return NULL; 308 return NULL;
309 309
@@ -508,6 +508,27 @@ bail:
508 return status; 508 return status;
509} 509}
510 510
511static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
512{
513 if (ocfs2_mount_local(osb)) {
514 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
515 mlog(ML_ERROR, "Cannot heartbeat on a locally "
516 "mounted device.\n");
517 return -EINVAL;
518 }
519 }
520
521 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
522 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) {
523 mlog(ML_ERROR, "Heartbeat has to be started to mount "
524 "a read-write clustered device.\n");
525 return -EINVAL;
526 }
527 }
528
529 return 0;
530}
531
511static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 532static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
512{ 533{
513 struct dentry *root; 534 struct dentry *root;
@@ -516,16 +537,24 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
516 struct inode *inode = NULL; 537 struct inode *inode = NULL;
517 struct ocfs2_super *osb = NULL; 538 struct ocfs2_super *osb = NULL;
518 struct buffer_head *bh = NULL; 539 struct buffer_head *bh = NULL;
540 char nodestr[8];
519 541
520 mlog_entry("%p, %p, %i", sb, data, silent); 542 mlog_entry("%p, %p, %i", sb, data, silent);
521 543
522 /* for now we only have one cluster/node, make sure we see it 544 if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
523 * in the heartbeat universe */
524 if (!o2hb_check_local_node_heartbeating()) {
525 status = -EINVAL; 545 status = -EINVAL;
526 goto read_super_error; 546 goto read_super_error;
527 } 547 }
528 548
549 /* for now we only have one cluster/node, make sure we see it
550 * in the heartbeat universe */
551 if (parsed_opt & OCFS2_MOUNT_HB_LOCAL) {
552 if (!o2hb_check_local_node_heartbeating()) {
553 status = -EINVAL;
554 goto read_super_error;
555 }
556 }
557
529 /* probe for superblock */ 558 /* probe for superblock */
530 status = ocfs2_sb_probe(sb, &bh, &sector_size); 559 status = ocfs2_sb_probe(sb, &bh, &sector_size);
531 if (status < 0) { 560 if (status < 0) {
@@ -541,11 +570,6 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
541 } 570 }
542 brelse(bh); 571 brelse(bh);
543 bh = NULL; 572 bh = NULL;
544
545 if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
546 status = -EINVAL;
547 goto read_super_error;
548 }
549 osb->s_mount_opt = parsed_opt; 573 osb->s_mount_opt = parsed_opt;
550 574
551 sb->s_magic = OCFS2_SUPER_MAGIC; 575 sb->s_magic = OCFS2_SUPER_MAGIC;
@@ -588,21 +612,16 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
588 } 612 }
589 613
590 if (!ocfs2_is_hard_readonly(osb)) { 614 if (!ocfs2_is_hard_readonly(osb)) {
591 /* If this isn't a hard readonly mount, then we need
592 * to make sure that heartbeat is in a valid state,
593 * and that we mark ourselves soft readonly is -oro
594 * was specified. */
595 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
596 mlog(ML_ERROR, "No heartbeat for device (%s)\n",
597 sb->s_id);
598 status = -EINVAL;
599 goto read_super_error;
600 }
601
602 if (sb->s_flags & MS_RDONLY) 615 if (sb->s_flags & MS_RDONLY)
603 ocfs2_set_ro_flag(osb, 0); 616 ocfs2_set_ro_flag(osb, 0);
604 } 617 }
605 618
619 status = ocfs2_verify_heartbeat(osb);
620 if (status < 0) {
621 mlog_errno(status);
622 goto read_super_error;
623 }
624
606 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 625 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
607 ocfs2_debugfs_root); 626 ocfs2_debugfs_root);
608 if (!osb->osb_debug_root) { 627 if (!osb->osb_debug_root) {
@@ -635,9 +654,14 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
635 654
636 ocfs2_complete_mount_recovery(osb); 655 ocfs2_complete_mount_recovery(osb);
637 656
638 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %d, slot %d) " 657 if (ocfs2_mount_local(osb))
658 snprintf(nodestr, sizeof(nodestr), "local");
659 else
660 snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num);
661
662 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
639 "with %s data mode.\n", 663 "with %s data mode.\n",
640 osb->dev_str, osb->node_num, osb->slot_num, 664 osb->dev_str, nodestr, osb->slot_num,
641 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 665 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
642 "ordered"); 666 "ordered");
643 667
@@ -707,6 +731,7 @@ static int ocfs2_parse_options(struct super_block *sb,
707 while ((p = strsep(&options, ",")) != NULL) { 731 while ((p = strsep(&options, ",")) != NULL) {
708 int token, option; 732 int token, option;
709 substring_t args[MAX_OPT_ARGS]; 733 substring_t args[MAX_OPT_ARGS];
734 struct ocfs2_super * osb = OCFS2_SB(sb);
710 735
711 if (!*p) 736 if (!*p)
712 continue; 737 continue;
@@ -747,6 +772,16 @@ static int ocfs2_parse_options(struct super_block *sb,
747 case Opt_data_writeback: 772 case Opt_data_writeback:
748 *mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 773 *mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
749 break; 774 break;
775 case Opt_atime_quantum:
776 if (match_int(&args[0], &option)) {
777 status = 0;
778 goto bail;
779 }
780 if (option >= 0)
781 osb->s_atime_quantum = option;
782 else
783 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
784 break;
750 default: 785 default:
751 mlog(ML_ERROR, 786 mlog(ML_ERROR,
752 "Unrecognized mount option \"%s\" " 787 "Unrecognized mount option \"%s\" "
@@ -867,7 +902,7 @@ static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
867 goto bail; 902 goto bail;
868 } 903 }
869 904
870 status = ocfs2_meta_lock(inode, NULL, &bh, 0); 905 status = ocfs2_meta_lock(inode, &bh, 0);
871 if (status < 0) { 906 if (status < 0) {
872 mlog_errno(status); 907 mlog_errno(status);
873 goto bail; 908 goto bail;
@@ -903,7 +938,7 @@ bail:
903} 938}
904 939
905static void ocfs2_inode_init_once(void *data, 940static void ocfs2_inode_init_once(void *data,
906 kmem_cache_t *cachep, 941 struct kmem_cache *cachep,
907 unsigned long flags) 942 unsigned long flags)
908{ 943{
909 struct ocfs2_inode_info *oi = data; 944 struct ocfs2_inode_info *oi = data;
@@ -914,9 +949,7 @@ static void ocfs2_inode_init_once(void *data,
914 oi->ip_open_count = 0; 949 oi->ip_open_count = 0;
915 spin_lock_init(&oi->ip_lock); 950 spin_lock_init(&oi->ip_lock);
916 ocfs2_extent_map_init(&oi->vfs_inode); 951 ocfs2_extent_map_init(&oi->vfs_inode);
917 INIT_LIST_HEAD(&oi->ip_handle_list);
918 INIT_LIST_HEAD(&oi->ip_io_markers); 952 INIT_LIST_HEAD(&oi->ip_io_markers);
919 oi->ip_handle = NULL;
920 oi->ip_created_trans = 0; 953 oi->ip_created_trans = 0;
921 oi->ip_last_trans = 0; 954 oi->ip_last_trans = 0;
922 oi->ip_dir_start_lookup = 0; 955 oi->ip_dir_start_lookup = 0;
@@ -948,14 +981,6 @@ static int ocfs2_initialize_mem_caches(void)
948 if (!ocfs2_inode_cachep) 981 if (!ocfs2_inode_cachep)
949 return -ENOMEM; 982 return -ENOMEM;
950 983
951 ocfs2_lock_cache = kmem_cache_create("ocfs2_lock",
952 sizeof(struct ocfs2_journal_lock),
953 0,
954 SLAB_HWCACHE_ALIGN,
955 NULL, NULL);
956 if (!ocfs2_lock_cache)
957 return -ENOMEM;
958
959 return 0; 984 return 0;
960} 985}
961 986
@@ -963,11 +988,8 @@ static void ocfs2_free_mem_caches(void)
963{ 988{
964 if (ocfs2_inode_cachep) 989 if (ocfs2_inode_cachep)
965 kmem_cache_destroy(ocfs2_inode_cachep); 990 kmem_cache_destroy(ocfs2_inode_cachep);
966 if (ocfs2_lock_cache)
967 kmem_cache_destroy(ocfs2_lock_cache);
968 991
969 ocfs2_inode_cachep = NULL; 992 ocfs2_inode_cachep = NULL;
970 ocfs2_lock_cache = NULL;
971} 993}
972 994
973static int ocfs2_get_sector(struct super_block *sb, 995static int ocfs2_get_sector(struct super_block *sb,
@@ -1001,7 +1023,11 @@ static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
1001 1023
1002 /* XXX hold a ref on the node while mounte? easy enough, if 1024 /* XXX hold a ref on the node while mounte? easy enough, if
1003 * desirable. */ 1025 * desirable. */
1004 osb->node_num = o2nm_this_node(); 1026 if (ocfs2_mount_local(osb))
1027 osb->node_num = 0;
1028 else
1029 osb->node_num = o2nm_this_node();
1030
1005 if (osb->node_num == O2NM_MAX_NODES) { 1031 if (osb->node_num == O2NM_MAX_NODES) {
1006 mlog(ML_ERROR, "could not find this host's node number\n"); 1032 mlog(ML_ERROR, "could not find this host's node number\n");
1007 status = -ENOENT; 1033 status = -ENOENT;
@@ -1086,6 +1112,9 @@ static int ocfs2_mount_volume(struct super_block *sb)
1086 goto leave; 1112 goto leave;
1087 } 1113 }
1088 1114
1115 if (ocfs2_mount_local(osb))
1116 goto leave;
1117
1089 /* This should be sent *after* we recovered our journal as it 1118 /* This should be sent *after* we recovered our journal as it
1090 * will cause other nodes to unmark us as needing 1119 * will cause other nodes to unmark us as needing
1091 * recovery. However, we need to send it *before* dropping the 1120 * recovery. However, we need to send it *before* dropping the
@@ -1116,6 +1145,7 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1116{ 1145{
1117 int tmp; 1146 int tmp;
1118 struct ocfs2_super *osb = NULL; 1147 struct ocfs2_super *osb = NULL;
1148 char nodestr[8];
1119 1149
1120 mlog_entry("(0x%p)\n", sb); 1150 mlog_entry("(0x%p)\n", sb);
1121 1151
@@ -1179,8 +1209,13 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1179 1209
1180 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1210 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1181 1211
1182 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %d)\n", 1212 if (ocfs2_mount_local(osb))
1183 osb->dev_str, osb->node_num); 1213 snprintf(nodestr, sizeof(nodestr), "local");
1214 else
1215 snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num);
1216
1217 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1218 osb->dev_str, nodestr);
1184 1219
1185 ocfs2_delete_osb(osb); 1220 ocfs2_delete_osb(osb);
1186 kfree(osb); 1221 kfree(osb);
@@ -1196,7 +1231,7 @@ static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uu
1196 1231
1197 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1232 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1198 1233
1199 osb->uuid_str = kcalloc(1, OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1234 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1200 if (osb->uuid_str == NULL) 1235 if (osb->uuid_str == NULL)
1201 return -ENOMEM; 1236 return -ENOMEM;
1202 1237
@@ -1227,7 +1262,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
1227 1262
1228 mlog_entry_void(); 1263 mlog_entry_void();
1229 1264
1230 osb = kcalloc(1, sizeof(struct ocfs2_super), GFP_KERNEL); 1265 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1231 if (!osb) { 1266 if (!osb) {
1232 status = -ENOMEM; 1267 status = -ENOMEM;
1233 mlog_errno(status); 1268 mlog_errno(status);
@@ -1280,6 +1315,8 @@ static int ocfs2_initialize_super(struct super_block *sb,
1280 init_waitqueue_head(&osb->checkpoint_event); 1315 init_waitqueue_head(&osb->checkpoint_event);
1281 atomic_set(&osb->needs_checkpoint, 0); 1316 atomic_set(&osb->needs_checkpoint, 0);
1282 1317
1318 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1319
1283 osb->node_num = O2NM_INVALID_NODE_NUM; 1320 osb->node_num = O2NM_INVALID_NODE_NUM;
1284 osb->slot_num = OCFS2_INVALID_SLOT; 1321 osb->slot_num = OCFS2_INVALID_SLOT;
1285 1322
@@ -1350,7 +1387,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
1350 */ 1387 */
1351 /* initialize our journal structure */ 1388 /* initialize our journal structure */
1352 1389
1353 journal = kcalloc(1, sizeof(struct ocfs2_journal), GFP_KERNEL); 1390 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
1354 if (!journal) { 1391 if (!journal) {
1355 mlog(ML_ERROR, "unable to alloc journal\n"); 1392 mlog(ML_ERROR, "unable to alloc journal\n");
1356 status = -ENOMEM; 1393 status = -ENOMEM;
@@ -1365,7 +1402,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
1365 spin_lock_init(&journal->j_lock); 1402 spin_lock_init(&journal->j_lock);
1366 journal->j_trans_id = (unsigned long) 1; 1403 journal->j_trans_id = (unsigned long) 1;
1367 INIT_LIST_HEAD(&journal->j_la_cleanups); 1404 INIT_LIST_HEAD(&journal->j_la_cleanups);
1368 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery, osb); 1405 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
1369 journal->j_state = OCFS2_JOURNAL_FREE; 1406 journal->j_state = OCFS2_JOURNAL_FREE;
1370 1407
1371 /* get some pseudo constants for clustersize bits */ 1408 /* get some pseudo constants for clustersize bits */
@@ -1536,6 +1573,7 @@ static int ocfs2_check_volume(struct ocfs2_super *osb)
1536{ 1573{
1537 int status = 0; 1574 int status = 0;
1538 int dirty; 1575 int dirty;
1576 int local;
1539 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 1577 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1540 * recover 1578 * recover
1541 * ourselves. */ 1579 * ourselves. */
@@ -1563,8 +1601,10 @@ static int ocfs2_check_volume(struct ocfs2_super *osb)
1563 "recovering volume.\n"); 1601 "recovering volume.\n");
1564 } 1602 }
1565 1603
1604 local = ocfs2_mount_local(osb);
1605
1566 /* will play back anything left in the journal. */ 1606 /* will play back anything left in the journal. */
1567 ocfs2_journal_load(osb->journal); 1607 ocfs2_journal_load(osb->journal, local);
1568 1608
1569 if (dirty) { 1609 if (dirty) {
1570 /* recover my local alloc if we didn't unmount cleanly. */ 1610 /* recover my local alloc if we didn't unmount cleanly. */
@@ -1674,7 +1714,7 @@ void __ocfs2_error(struct super_block *sb,
1674 va_list args; 1714 va_list args;
1675 1715
1676 va_start(args, fmt); 1716 va_start(args, fmt);
1677 vsprintf(error_buf, fmt, args); 1717 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1678 va_end(args); 1718 va_end(args);
1679 1719
1680 /* Not using mlog here because we want to show the actual 1720 /* Not using mlog here because we want to show the actual
@@ -1695,7 +1735,7 @@ void __ocfs2_abort(struct super_block* sb,
1695 va_list args; 1735 va_list args;
1696 1736
1697 va_start(args, fmt); 1737 va_start(args, fmt);
1698 vsprintf(error_buf, fmt, args); 1738 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1699 va_end(args); 1739 va_end(args);
1700 1740
1701 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n", 1741 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
diff --git a/fs/ocfs2/symlink.c b/fs/ocfs2/symlink.c
index c0f68aa6c175..957d6878b03e 100644
--- a/fs/ocfs2/symlink.c
+++ b/fs/ocfs2/symlink.c
@@ -126,6 +126,10 @@ static int ocfs2_readlink(struct dentry *dentry,
126 goto out; 126 goto out;
127 } 127 }
128 128
129 /*
130 * Without vfsmount we can't update atime now,
131 * but we will update atime here ultimately.
132 */
129 ret = vfs_readlink(dentry, buffer, buflen, link); 133 ret = vfs_readlink(dentry, buffer, buflen, link);
130 134
131 brelse(bh); 135 brelse(bh);
diff --git a/fs/ocfs2/uptodate.c b/fs/ocfs2/uptodate.c
index 9707ed7a3206..39814b900fc0 100644
--- a/fs/ocfs2/uptodate.c
+++ b/fs/ocfs2/uptodate.c
@@ -69,7 +69,7 @@ struct ocfs2_meta_cache_item {
69 sector_t c_block; 69 sector_t c_block;
70}; 70};
71 71
72static kmem_cache_t *ocfs2_uptodate_cachep = NULL; 72static struct kmem_cache *ocfs2_uptodate_cachep = NULL;
73 73
74void ocfs2_metadata_cache_init(struct inode *inode) 74void ocfs2_metadata_cache_init(struct inode *inode)
75{ 75{
diff --git a/fs/ocfs2/vote.c b/fs/ocfs2/vote.c
index 5b4dca79990b..0afd8b9af70f 100644
--- a/fs/ocfs2/vote.c
+++ b/fs/ocfs2/vote.c
@@ -479,7 +479,7 @@ static struct ocfs2_net_wait_ctxt *ocfs2_new_net_wait_ctxt(unsigned int response
479{ 479{
480 struct ocfs2_net_wait_ctxt *w; 480 struct ocfs2_net_wait_ctxt *w;
481 481
482 w = kcalloc(1, sizeof(*w), GFP_NOFS); 482 w = kzalloc(sizeof(*w), GFP_NOFS);
483 if (!w) { 483 if (!w) {
484 mlog_errno(-ENOMEM); 484 mlog_errno(-ENOMEM);
485 goto bail; 485 goto bail;
@@ -642,7 +642,7 @@ static struct ocfs2_vote_msg * ocfs2_new_vote_request(struct ocfs2_super *osb,
642 642
643 BUG_ON(!ocfs2_is_valid_vote_request(type)); 643 BUG_ON(!ocfs2_is_valid_vote_request(type));
644 644
645 request = kcalloc(1, sizeof(*request), GFP_NOFS); 645 request = kzalloc(sizeof(*request), GFP_NOFS);
646 if (!request) { 646 if (!request) {
647 mlog_errno(-ENOMEM); 647 mlog_errno(-ENOMEM);
648 } else { 648 } else {
@@ -1000,6 +1000,9 @@ int ocfs2_register_net_handlers(struct ocfs2_super *osb)
1000{ 1000{
1001 int status = 0; 1001 int status = 0;
1002 1002
1003 if (ocfs2_mount_local(osb))
1004 return 0;
1005
1003 status = o2net_register_handler(OCFS2_MESSAGE_TYPE_RESPONSE, 1006 status = o2net_register_handler(OCFS2_MESSAGE_TYPE_RESPONSE,
1004 osb->net_key, 1007 osb->net_key,
1005 sizeof(struct ocfs2_response_msg), 1008 sizeof(struct ocfs2_response_msg),
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-26 23:31:22 -0500