aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-10-21 22:01:34 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-10-21 22:01:34 -0400
commitf3270b16e00f0614fa418dcc50883da5949375b4 (patch)
tree2f76b9679cbc9e8da39c8a67cd2bc00084d441e7
parent3044100e58c84e133791c8b60a2f5bef69d732e4 (diff)
parent2decd65a2630633cee04d0b83fdcee46ad2989a1 (diff)
Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2: (48 commits) ocfs2: Avoid to evaluate xattr block flags again. ocfs2/cluster: Release debugfs file elapsed_time_in_ms ocfs2: Add a mount option "coherency=*" to handle cluster coherency for O_DIRECT writes. Initialize max_slots early When I tried to compile I got the following warning: fs/ocfs2/slot_map.c: In function ‘ocfs2_init_slot_info’: fs/ocfs2/slot_map.c:360: warning: ‘bytes’ may be used uninitialized in this function fs/ocfs2/slot_map.c:360: note: ‘bytes’ was declared here Compiler: gcc version 4.4.3 (GCC) on Mandriva I'm not sure why this warning occurs, I think compiler don't know that variable "bytes" is initialized when it is sent by reference to ocfs2_slot_map_physical_size and it throws that ugly warning. However, a simple initialization of "bytes" variable with 0 will fix it. ocfs2: validate bg_free_bits_count after update ocfs2/cluster: Bump up dlm protocol to version 1.1 ocfs2/cluster: Show per region heartbeat elapsed time ocfs2/cluster: Add mlogs for heartbeat up/down events ocfs2/cluster: Create debugfs dir/files for each region ocfs2/cluster: Create debugfs files for live, quorum and failed region bitmaps ocfs2/cluster: Maintain bitmap of failed regions ocfs2/cluster: Maintain bitmap of quorum regions ocfs2/cluster: Track bitmap of live heartbeat regions ocfs2/cluster: Track number of global heartbeat regions ocfs2/cluster: Maintain live node bitmap per heartbeat region ocfs2/cluster: Reorganize o2hb debugfs init ocfs2/cluster: Check slots for unconfigured live nodes ocfs2/cluster: Print messages when adding/removing nodes ocfs2/cluster: Print messages when adding/removing heartbeat regions ...
Diffstat
-rw-r--r--Documentation/filesystems/ocfs2.txt7
-rw-r--r--fs/ext3/super.c4
-rw-r--r--fs/ext4/super.c8
-rw-r--r--fs/jbd2/journal.c4
-rw-r--r--fs/libfs.c29
-rw-r--r--fs/ocfs2/aops.c9
-rw-r--r--fs/ocfs2/aops.h3
-rw-r--r--fs/ocfs2/cluster/heartbeat.c532
-rw-r--r--fs/ocfs2/cluster/heartbeat.h4
-rw-r--r--fs/ocfs2/cluster/masklog.h3
-rw-r--r--fs/ocfs2/cluster/nodemanager.c5
-rw-r--r--fs/ocfs2/cluster/ocfs2_nodemanager.h6
-rw-r--r--fs/ocfs2/cluster/tcp.c5
-rw-r--r--fs/ocfs2/dcache.c33
-rw-r--r--fs/ocfs2/dcache.h1
-rw-r--r--fs/ocfs2/dlm/dlmcommon.h29
-rw-r--r--fs/ocfs2/dlm/dlmdebug.c12
-rw-r--r--fs/ocfs2/dlm/dlmdomain.c400
-rw-r--r--fs/ocfs2/dlmglue.c8
-rw-r--r--fs/ocfs2/file.c73
-rw-r--r--fs/ocfs2/inode.c1
-rw-r--r--fs/ocfs2/inode.h12
-rw-r--r--fs/ocfs2/ioctl.c356
-rw-r--r--fs/ocfs2/journal.c9
-rw-r--r--fs/ocfs2/journal.h3
-rw-r--r--fs/ocfs2/mmap.c7
-rw-r--r--fs/ocfs2/namei.c3
-rw-r--r--fs/ocfs2/ocfs2.h63
-rw-r--r--fs/ocfs2/ocfs2_fs.h46
-rw-r--r--fs/ocfs2/ocfs2_ioctl.h95
-rw-r--r--fs/ocfs2/refcounttree.c43
-rw-r--r--fs/ocfs2/refcounttree.h7
-rw-r--r--fs/ocfs2/slot_map.c2
-rw-r--r--fs/ocfs2/stack_o2cb.c2
-rw-r--r--fs/ocfs2/suballoc.c16
-rw-r--r--fs/ocfs2/super.c163
-rw-r--r--fs/ocfs2/sysfile.c60
-rw-r--r--fs/ocfs2/xattr.c2
-rw-r--r--include/linux/fs.h2
39 files changed, 1894 insertions, 173 deletions
diff --git a/Documentation/filesystems/ocfs2.txt b/Documentation/filesystems/ocfs2.txt
index 1f7ae144f6d8..5393e6611691 100644
--- a/Documentation/filesystems/ocfs2.txt
+++ b/Documentation/filesystems/ocfs2.txt
@@ -87,3 +87,10 @@ dir_resv_level= (*) By default, directory reservations will scale with file
87 reservations - users should rarely need to change this 87 reservations - users should rarely need to change this
88 value. If allocation reservations are turned off, this 88 value. If allocation reservations are turned off, this
89 option will have no effect. 89 option will have no effect.
90coherency=full (*) Disallow concurrent O_DIRECT writes, cluster inode
91 lock will be taken to force other nodes drop cache,
92 therefore full cluster coherency is guaranteed even
93 for O_DIRECT writes.
94coherency=buffered Allow concurrent O_DIRECT writes without EX lock among
95 nodes, which gains high performance at risk of getting
96 stale data on other nodes.
diff --git a/fs/ext3/super.c b/fs/ext3/super.c
index 5dbf4dba03c4..a367dd044280 100644
--- a/fs/ext3/super.c
+++ b/fs/ext3/super.c
@@ -1849,8 +1849,8 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1849 goto failed_mount; 1849 goto failed_mount;
1850 } 1850 }
1851 1851
1852 if (le32_to_cpu(es->s_blocks_count) > 1852 if (generic_check_addressable(sb->s_blocksize_bits,
1853 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) { 1853 le32_to_cpu(es->s_blocks_count))) {
1854 ext3_msg(sb, KERN_ERR, 1854 ext3_msg(sb, KERN_ERR,
1855 "error: filesystem is too large to mount safely"); 1855 "error: filesystem is too large to mount safely");
1856 if (sizeof(sector_t) < 8) 1856 if (sizeof(sector_t) < 8)
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 26147746c272..7f47c366bf15 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -2831,15 +2831,13 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2831 * Test whether we have more sectors than will fit in sector_t, 2831 * Test whether we have more sectors than will fit in sector_t,
2832 * and whether the max offset is addressable by the page cache. 2832 * and whether the max offset is addressable by the page cache.
2833 */ 2833 */
2834 if ((ext4_blocks_count(es) > 2834 ret = generic_check_addressable(sb->s_blocksize_bits,
2835 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) || 2835 ext4_blocks_count(es));
2836 (ext4_blocks_count(es) > 2836 if (ret) {
2837 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2838 ext4_msg(sb, KERN_ERR, "filesystem" 2837 ext4_msg(sb, KERN_ERR, "filesystem"
2839 " too large to mount safely on this system"); 2838 " too large to mount safely on this system");
2840 if (sizeof(sector_t) < 8) 2839 if (sizeof(sector_t) < 8)
2841 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 2840 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2842 ret = -EFBIG;
2843 goto failed_mount; 2841 goto failed_mount;
2844 } 2842 }
2845 2843
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index 0e8014ea6b94..262419f83d80 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -1371,6 +1371,10 @@ int jbd2_journal_check_used_features (journal_t *journal, unsigned long compat,
1371 1371
1372 if (!compat && !ro && !incompat) 1372 if (!compat && !ro && !incompat)
1373 return 1; 1373 return 1;
1374 /* Load journal superblock if it is not loaded yet. */
1375 if (journal->j_format_version == 0 &&
1376 journal_get_superblock(journal) != 0)
1377 return 0;
1374 if (journal->j_format_version == 1) 1378 if (journal->j_format_version == 1)
1375 return 0; 1379 return 0;
1376 1380
diff --git a/fs/libfs.c b/fs/libfs.c
index 0a9da95317f7..62baa0387d6e 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -913,6 +913,35 @@ int generic_file_fsync(struct file *file, int datasync)
913} 913}
914EXPORT_SYMBOL(generic_file_fsync); 914EXPORT_SYMBOL(generic_file_fsync);
915 915
916/**
917 * generic_check_addressable - Check addressability of file system
918 * @blocksize_bits: log of file system block size
919 * @num_blocks: number of blocks in file system
920 *
921 * Determine whether a file system with @num_blocks blocks (and a
922 * block size of 2**@blocksize_bits) is addressable by the sector_t
923 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
924 */
925int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
926{
927 u64 last_fs_block = num_blocks - 1;
928 u64 last_fs_page =
929 last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
930
931 if (unlikely(num_blocks == 0))
932 return 0;
933
934 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
935 return -EINVAL;
936
937 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
938 (last_fs_page > (pgoff_t)(~0ULL))) {
939 return -EFBIG;
940 }
941 return 0;
942}
943EXPORT_SYMBOL(generic_check_addressable);
944
916/* 945/*
917 * No-op implementation of ->fsync for in-memory filesystems. 946 * No-op implementation of ->fsync for in-memory filesystems.
918 */ 947 */
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 0de69c9a08be..5cfeee118158 100644
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -883,8 +883,8 @@ struct ocfs2_write_ctxt {
883 * out in so that future reads from that region will get 883 * out in so that future reads from that region will get
884 * zero's. 884 * zero's.
885 */ 885 */
886 struct page *w_pages[OCFS2_MAX_CTXT_PAGES];
887 unsigned int w_num_pages; 886 unsigned int w_num_pages;
887 struct page *w_pages[OCFS2_MAX_CTXT_PAGES];
888 struct page *w_target_page; 888 struct page *w_target_page;
889 889
890 /* 890 /*
@@ -1642,7 +1642,8 @@ static int ocfs2_zero_tail(struct inode *inode, struct buffer_head *di_bh,
1642 return ret; 1642 return ret;
1643} 1643}
1644 1644
1645int ocfs2_write_begin_nolock(struct address_space *mapping, 1645int ocfs2_write_begin_nolock(struct file *filp,
1646 struct address_space *mapping,
1646 loff_t pos, unsigned len, unsigned flags, 1647 loff_t pos, unsigned len, unsigned flags,
1647 struct page **pagep, void **fsdata, 1648 struct page **pagep, void **fsdata,
1648 struct buffer_head *di_bh, struct page *mmap_page) 1649 struct buffer_head *di_bh, struct page *mmap_page)
@@ -1692,7 +1693,7 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
1692 mlog_errno(ret); 1693 mlog_errno(ret);
1693 goto out; 1694 goto out;
1694 } else if (ret == 1) { 1695 } else if (ret == 1) {
1695 ret = ocfs2_refcount_cow(inode, di_bh, 1696 ret = ocfs2_refcount_cow(inode, filp, di_bh,
1696 wc->w_cpos, wc->w_clen, UINT_MAX); 1697 wc->w_cpos, wc->w_clen, UINT_MAX);
1697 if (ret) { 1698 if (ret) {
1698 mlog_errno(ret); 1699 mlog_errno(ret);
@@ -1854,7 +1855,7 @@ static int ocfs2_write_begin(struct file *file, struct address_space *mapping,
1854 */ 1855 */
1855 down_write(&OCFS2_I(inode)->ip_alloc_sem); 1856 down_write(&OCFS2_I(inode)->ip_alloc_sem);
1856 1857
1857 ret = ocfs2_write_begin_nolock(mapping, pos, len, flags, pagep, 1858 ret = ocfs2_write_begin_nolock(file, mapping, pos, len, flags, pagep,
1858 fsdata, di_bh, NULL); 1859 fsdata, di_bh, NULL);
1859 if (ret) { 1860 if (ret) {
1860 mlog_errno(ret); 1861 mlog_errno(ret);
diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h
index c48e93ffc513..7606f663da6d 100644
--- a/fs/ocfs2/aops.h
+++ b/fs/ocfs2/aops.h
@@ -48,7 +48,8 @@ int ocfs2_write_end_nolock(struct address_space *mapping,
48 loff_t pos, unsigned len, unsigned copied, 48 loff_t pos, unsigned len, unsigned copied,
49 struct page *page, void *fsdata); 49 struct page *page, void *fsdata);
50 50
51int ocfs2_write_begin_nolock(struct address_space *mapping, 51int ocfs2_write_begin_nolock(struct file *filp,
52 struct address_space *mapping,
52 loff_t pos, unsigned len, unsigned flags, 53 loff_t pos, unsigned len, unsigned flags,
53 struct page **pagep, void **fsdata, 54 struct page **pagep, void **fsdata,
54 struct buffer_head *di_bh, struct page *mmap_page); 55 struct buffer_head *di_bh, struct page *mmap_page);
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index 41d5f1f92d56..52c7557f3e25 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -62,10 +62,51 @@ static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
62static LIST_HEAD(o2hb_node_events); 62static LIST_HEAD(o2hb_node_events);
63static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue); 63static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
64 64
65/*
66 * In global heartbeat, we maintain a series of region bitmaps.
67 * - o2hb_region_bitmap allows us to limit the region number to max region.
68 * - o2hb_live_region_bitmap tracks live regions (seen steady iterations).
69 * - o2hb_quorum_region_bitmap tracks live regions that have seen all nodes
70 * heartbeat on it.
71 * - o2hb_failed_region_bitmap tracks the regions that have seen io timeouts.
72 */
73static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
74static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
75static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
76static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
77
78#define O2HB_DB_TYPE_LIVENODES 0
79#define O2HB_DB_TYPE_LIVEREGIONS 1
80#define O2HB_DB_TYPE_QUORUMREGIONS 2
81#define O2HB_DB_TYPE_FAILEDREGIONS 3
82#define O2HB_DB_TYPE_REGION_LIVENODES 4
83#define O2HB_DB_TYPE_REGION_NUMBER 5
84#define O2HB_DB_TYPE_REGION_ELAPSED_TIME 6
85struct o2hb_debug_buf {
86 int db_type;
87 int db_size;
88 int db_len;
89 void *db_data;
90};
91
92static struct o2hb_debug_buf *o2hb_db_livenodes;
93static struct o2hb_debug_buf *o2hb_db_liveregions;
94static struct o2hb_debug_buf *o2hb_db_quorumregions;
95static struct o2hb_debug_buf *o2hb_db_failedregions;
96
65#define O2HB_DEBUG_DIR "o2hb" 97#define O2HB_DEBUG_DIR "o2hb"
66#define O2HB_DEBUG_LIVENODES "livenodes" 98#define O2HB_DEBUG_LIVENODES "livenodes"
99#define O2HB_DEBUG_LIVEREGIONS "live_regions"
100#define O2HB_DEBUG_QUORUMREGIONS "quorum_regions"
101#define O2HB_DEBUG_FAILEDREGIONS "failed_regions"
102#define O2HB_DEBUG_REGION_NUMBER "num"
103#define O2HB_DEBUG_REGION_ELAPSED_TIME "elapsed_time_in_ms"
104
67static struct dentry *o2hb_debug_dir; 105static struct dentry *o2hb_debug_dir;
68static struct dentry *o2hb_debug_livenodes; 106static struct dentry *o2hb_debug_livenodes;
107static struct dentry *o2hb_debug_liveregions;
108static struct dentry *o2hb_debug_quorumregions;
109static struct dentry *o2hb_debug_failedregions;
69 110
70static LIST_HEAD(o2hb_all_regions); 111static LIST_HEAD(o2hb_all_regions);
71 112
@@ -77,7 +118,19 @@ static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
77 118
78#define O2HB_DEFAULT_BLOCK_BITS 9 119#define O2HB_DEFAULT_BLOCK_BITS 9
79 120
121enum o2hb_heartbeat_modes {
122 O2HB_HEARTBEAT_LOCAL = 0,
123 O2HB_HEARTBEAT_GLOBAL,
124 O2HB_HEARTBEAT_NUM_MODES,
125};
126
127char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = {
128 "local", /* O2HB_HEARTBEAT_LOCAL */
129 "global", /* O2HB_HEARTBEAT_GLOBAL */
130};
131
80unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD; 132unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
133unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL;
81 134
82/* Only sets a new threshold if there are no active regions. 135/* Only sets a new threshold if there are no active regions.
83 * 136 *
@@ -94,6 +147,22 @@ static void o2hb_dead_threshold_set(unsigned int threshold)
94 } 147 }
95} 148}
96 149
150static int o2hb_global_hearbeat_mode_set(unsigned int hb_mode)
151{
152 int ret = -1;
153
154 if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) {
155 spin_lock(&o2hb_live_lock);
156 if (list_empty(&o2hb_all_regions)) {
157 o2hb_heartbeat_mode = hb_mode;
158 ret = 0;
159 }
160 spin_unlock(&o2hb_live_lock);
161 }
162
163 return ret;
164}
165
97struct o2hb_node_event { 166struct o2hb_node_event {
98 struct list_head hn_item; 167 struct list_head hn_item;
99 enum o2hb_callback_type hn_event_type; 168 enum o2hb_callback_type hn_event_type;
@@ -135,6 +204,18 @@ struct o2hb_region {
135 struct block_device *hr_bdev; 204 struct block_device *hr_bdev;
136 struct o2hb_disk_slot *hr_slots; 205 struct o2hb_disk_slot *hr_slots;
137 206
207 /* live node map of this region */
208 unsigned long hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
209 unsigned int hr_region_num;
210
211 struct dentry *hr_debug_dir;
212 struct dentry *hr_debug_livenodes;
213 struct dentry *hr_debug_regnum;
214 struct dentry *hr_debug_elapsed_time;
215 struct o2hb_debug_buf *hr_db_livenodes;
216 struct o2hb_debug_buf *hr_db_regnum;
217 struct o2hb_debug_buf *hr_db_elapsed_time;
218
138 /* let the person setting up hb wait for it to return until it 219 /* let the person setting up hb wait for it to return until it
139 * has reached a 'steady' state. This will be fixed when we have 220 * has reached a 'steady' state. This will be fixed when we have
140 * a more complete api that doesn't lead to this sort of fragility. */ 221 * a more complete api that doesn't lead to this sort of fragility. */
@@ -163,8 +244,19 @@ struct o2hb_bio_wait_ctxt {
163 int wc_error; 244 int wc_error;
164}; 245};
165 246
247static int o2hb_pop_count(void *map, int count)
248{
249 int i = -1, pop = 0;
250
251 while ((i = find_next_bit(map, count, i + 1)) < count)
252 pop++;
253 return pop;
254}
255
166static void o2hb_write_timeout(struct work_struct *work) 256static void o2hb_write_timeout(struct work_struct *work)
167{ 257{
258 int failed, quorum;
259 unsigned long flags;
168 struct o2hb_region *reg = 260 struct o2hb_region *reg =
169 container_of(work, struct o2hb_region, 261 container_of(work, struct o2hb_region,
170 hr_write_timeout_work.work); 262 hr_write_timeout_work.work);
@@ -172,6 +264,28 @@ static void o2hb_write_timeout(struct work_struct *work)
172 mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u " 264 mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
173 "milliseconds\n", reg->hr_dev_name, 265 "milliseconds\n", reg->hr_dev_name,
174 jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); 266 jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
267
268 if (o2hb_global_heartbeat_active()) {
269 spin_lock_irqsave(&o2hb_live_lock, flags);
270 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
271 set_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
272 failed = o2hb_pop_count(&o2hb_failed_region_bitmap,
273 O2NM_MAX_REGIONS);
274 quorum = o2hb_pop_count(&o2hb_quorum_region_bitmap,
275 O2NM_MAX_REGIONS);
276 spin_unlock_irqrestore(&o2hb_live_lock, flags);
277
278 mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n",
279 quorum, failed);
280
281 /*
282 * Fence if the number of failed regions >= half the number
283 * of quorum regions
284 */
285 if ((failed << 1) < quorum)
286 return;
287 }
288
175 o2quo_disk_timeout(); 289 o2quo_disk_timeout();
176} 290}
177 291
@@ -180,6 +294,11 @@ static void o2hb_arm_write_timeout(struct o2hb_region *reg)
180 mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n", 294 mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n",
181 O2HB_MAX_WRITE_TIMEOUT_MS); 295 O2HB_MAX_WRITE_TIMEOUT_MS);
182 296
297 if (o2hb_global_heartbeat_active()) {
298 spin_lock(&o2hb_live_lock);
299 clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
300 spin_unlock(&o2hb_live_lock);
301 }
183 cancel_delayed_work(&reg->hr_write_timeout_work); 302 cancel_delayed_work(&reg->hr_write_timeout_work);
184 reg->hr_last_timeout_start = jiffies; 303 reg->hr_last_timeout_start = jiffies;
185 schedule_delayed_work(&reg->hr_write_timeout_work, 304 schedule_delayed_work(&reg->hr_write_timeout_work,
@@ -513,6 +632,8 @@ static void o2hb_queue_node_event(struct o2hb_node_event *event,
513{ 632{
514 assert_spin_locked(&o2hb_live_lock); 633 assert_spin_locked(&o2hb_live_lock);
515 634
635 BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB));
636
516 event->hn_event_type = type; 637 event->hn_event_type = type;
517 event->hn_node = node; 638 event->hn_node = node;
518 event->hn_node_num = node_num; 639 event->hn_node_num = node_num;
@@ -554,6 +675,35 @@ static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
554 o2nm_node_put(node); 675 o2nm_node_put(node);
555} 676}
556 677
678static void o2hb_set_quorum_device(struct o2hb_region *reg,
679 struct o2hb_disk_slot *slot)
680{
681 assert_spin_locked(&o2hb_live_lock);
682
683 if (!o2hb_global_heartbeat_active())
684 return;
685
686 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
687 return;
688
689 /*
690 * A region can be added to the quorum only when it sees all
691 * live nodes heartbeat on it. In other words, the region has been
692 * added to all nodes.
693 */
694 if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap,
695 sizeof(o2hb_live_node_bitmap)))
696 return;
697
698 if (slot->ds_changed_samples < O2HB_LIVE_THRESHOLD)
699 return;
700
701 printk(KERN_NOTICE "o2hb: Region %s is now a quorum device\n",
702 config_item_name(&reg->hr_item));
703
704 set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
705}
706
557static int o2hb_check_slot(struct o2hb_region *reg, 707static int o2hb_check_slot(struct o2hb_region *reg,
558 struct o2hb_disk_slot *slot) 708 struct o2hb_disk_slot *slot)
559{ 709{
@@ -565,14 +715,22 @@ static int o2hb_check_slot(struct o2hb_region *reg,
565 u64 cputime; 715 u64 cputime;
566 unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS; 716 unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS;
567 unsigned int slot_dead_ms; 717 unsigned int slot_dead_ms;
718 int tmp;
568 719
569 memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes); 720 memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
570 721
571 /* Is this correct? Do we assume that the node doesn't exist 722 /*
572 * if we're not configured for him? */ 723 * If a node is no longer configured but is still in the livemap, we
724 * may need to clear that bit from the livemap.
725 */
573 node = o2nm_get_node_by_num(slot->ds_node_num); 726 node = o2nm_get_node_by_num(slot->ds_node_num);
574 if (!node) 727 if (!node) {
575 return 0; 728 spin_lock(&o2hb_live_lock);
729 tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap);
730 spin_unlock(&o2hb_live_lock);
731 if (!tmp)
732 return 0;
733 }
576 734
577 if (!o2hb_verify_crc(reg, hb_block)) { 735 if (!o2hb_verify_crc(reg, hb_block)) {
578 /* all paths from here will drop o2hb_live_lock for 736 /* all paths from here will drop o2hb_live_lock for
@@ -639,8 +797,12 @@ fire_callbacks:
639 mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n", 797 mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n",
640 slot->ds_node_num, (long long)slot->ds_last_generation); 798 slot->ds_node_num, (long long)slot->ds_last_generation);
641 799
800 set_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
801
642 /* first on the list generates a callback */ 802 /* first on the list generates a callback */
643 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { 803 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
804 mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes "
805 "bitmap\n", slot->ds_node_num);
644 set_bit(slot->ds_node_num, o2hb_live_node_bitmap); 806 set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
645 807
646 o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node, 808 o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
@@ -684,13 +846,18 @@ fire_callbacks:
684 mlog(ML_HEARTBEAT, "Node %d left my region\n", 846 mlog(ML_HEARTBEAT, "Node %d left my region\n",
685 slot->ds_node_num); 847 slot->ds_node_num);
686 848
849 clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
850
687 /* last off the live_slot generates a callback */ 851 /* last off the live_slot generates a callback */
688 list_del_init(&slot->ds_live_item); 852 list_del_init(&slot->ds_live_item);
689 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { 853 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
854 mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live "
855 "nodes bitmap\n", slot->ds_node_num);
690 clear_bit(slot->ds_node_num, o2hb_live_node_bitmap); 856 clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
691 857
692 o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node, 858 /* node can be null */
693 slot->ds_node_num); 859 o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB,
860 node, slot->ds_node_num);
694 861
695 changed = 1; 862 changed = 1;
696 } 863 }
@@ -706,11 +873,14 @@ fire_callbacks:
706 slot->ds_equal_samples = 0; 873 slot->ds_equal_samples = 0;
707 } 874 }
708out: 875out:
876 o2hb_set_quorum_device(reg, slot);
877
709 spin_unlock(&o2hb_live_lock); 878 spin_unlock(&o2hb_live_lock);
710 879
711 o2hb_run_event_list(&event); 880 o2hb_run_event_list(&event);
712 881
713 o2nm_node_put(node); 882 if (node)
883 o2nm_node_put(node);
714 return changed; 884 return changed;
715} 885}
716 886
@@ -737,6 +907,7 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
737{ 907{
738 int i, ret, highest_node, change = 0; 908 int i, ret, highest_node, change = 0;
739 unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)]; 909 unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
910 unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
740 struct o2hb_bio_wait_ctxt write_wc; 911 struct o2hb_bio_wait_ctxt write_wc;
741 912
742 ret = o2nm_configured_node_map(configured_nodes, 913 ret = o2nm_configured_node_map(configured_nodes,
@@ -746,6 +917,17 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
746 return ret; 917 return ret;
747 } 918 }
748 919
920 /*
921 * If a node is not configured but is in the livemap, we still need
922 * to read the slot so as to be able to remove it from the livemap.
923 */
924 o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
925 i = -1;
926 while ((i = find_next_bit(live_node_bitmap,
927 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
928 set_bit(i, configured_nodes);
929 }
930
749 highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES); 931 highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
750 if (highest_node >= O2NM_MAX_NODES) { 932 if (highest_node >= O2NM_MAX_NODES) {
751 mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n"); 933 mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n");
@@ -917,21 +1099,59 @@ static int o2hb_thread(void *data)
917#ifdef CONFIG_DEBUG_FS 1099#ifdef CONFIG_DEBUG_FS
918static int o2hb_debug_open(struct inode *inode, struct file *file) 1100static int o2hb_debug_open(struct inode *inode, struct file *file)
919{ 1101{
1102 struct o2hb_debug_buf *db = inode->i_private;
1103 struct o2hb_region *reg;
920 unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 1104 unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)];
921 char *buf = NULL; 1105 char *buf = NULL;
922 int i = -1; 1106 int i = -1;
923 int out = 0; 1107 int out = 0;
924 1108
1109 /* max_nodes should be the largest bitmap we pass here */
1110 BUG_ON(sizeof(map) < db->db_size);
1111
925 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 1112 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
926 if (!buf) 1113 if (!buf)
927 goto bail; 1114 goto bail;
928 1115
929 o2hb_fill_node_map(map, sizeof(map)); 1116 switch (db->db_type) {
1117 case O2HB_DB_TYPE_LIVENODES:
1118 case O2HB_DB_TYPE_LIVEREGIONS:
1119 case O2HB_DB_TYPE_QUORUMREGIONS:
1120 case O2HB_DB_TYPE_FAILEDREGIONS:
1121 spin_lock(&o2hb_live_lock);
1122 memcpy(map, db->db_data, db->db_size);
1123 spin_unlock(&o2hb_live_lock);
1124 break;
1125
1126 case O2HB_DB_TYPE_REGION_LIVENODES:
1127 spin_lock(&o2hb_live_lock);
1128 reg = (struct o2hb_region *)db->db_data;
1129 memcpy(map, reg->hr_live_node_bitmap, db->db_size);
1130 spin_unlock(&o2hb_live_lock);
1131 break;
1132
1133 case O2HB_DB_TYPE_REGION_NUMBER:
1134 reg = (struct o2hb_region *)db->db_data;
1135 out += snprintf(buf + out, PAGE_SIZE - out, "%d\n",
1136 reg->hr_region_num);
1137 goto done;
1138
1139 case O2HB_DB_TYPE_REGION_ELAPSED_TIME:
1140 reg = (struct o2hb_region *)db->db_data;
1141 out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
1142 jiffies_to_msecs(jiffies -
1143 reg->hr_last_timeout_start));
1144 goto done;
1145
1146 default:
1147 goto done;
1148 }
930 1149
931 while ((i = find_next_bit(map, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) 1150 while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len)
932 out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i); 1151 out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i);
933 out += snprintf(buf + out, PAGE_SIZE - out, "\n"); 1152 out += snprintf(buf + out, PAGE_SIZE - out, "\n");
934 1153
1154done:
935 i_size_write(inode, out); 1155 i_size_write(inode, out);
936 1156
937 file->private_data = buf; 1157 file->private_data = buf;
@@ -978,10 +1198,104 @@ static const struct file_operations o2hb_debug_fops = {
978 1198
979void o2hb_exit(void) 1199void o2hb_exit(void)
980{ 1200{
981 if (o2hb_debug_livenodes) 1201 kfree(o2hb_db_livenodes);
982 debugfs_remove(o2hb_debug_livenodes); 1202 kfree(o2hb_db_liveregions);
983 if (o2hb_debug_dir) 1203 kfree(o2hb_db_quorumregions);
984 debugfs_remove(o2hb_debug_dir); 1204 kfree(o2hb_db_failedregions);
1205 debugfs_remove(o2hb_debug_failedregions);
1206 debugfs_remove(o2hb_debug_quorumregions);
1207 debugfs_remove(o2hb_debug_liveregions);
1208 debugfs_remove(o2hb_debug_livenodes);
1209 debugfs_remove(o2hb_debug_dir);
1210}
1211
1212static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir,
1213 struct o2hb_debug_buf **db, int db_len,
1214 int type, int size, int len, void *data)
1215{
1216 *db = kmalloc(db_len, GFP_KERNEL);
1217 if (!*db)
1218 return NULL;
1219
1220 (*db)->db_type = type;
1221 (*db)->db_size = size;
1222 (*db)->db_len = len;
1223 (*db)->db_data = data;
1224
1225 return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db,
1226 &o2hb_debug_fops);
1227}
1228
1229static int o2hb_debug_init(void)
1230{
1231 int ret = -ENOMEM;
1232
1233 o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL);
1234 if (!o2hb_debug_dir) {
1235 mlog_errno(ret);
1236 goto bail;
1237 }
1238
1239 o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES,
1240 o2hb_debug_dir,
1241 &o2hb_db_livenodes,
1242 sizeof(*o2hb_db_livenodes),
1243 O2HB_DB_TYPE_LIVENODES,
1244 sizeof(o2hb_live_node_bitmap),
1245 O2NM_MAX_NODES,
1246 o2hb_live_node_bitmap);
1247 if (!o2hb_debug_livenodes) {
1248 mlog_errno(ret);
1249 goto bail;
1250 }
1251
1252 o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS,
1253 o2hb_debug_dir,
1254 &o2hb_db_liveregions,
1255 sizeof(*o2hb_db_liveregions),
1256 O2HB_DB_TYPE_LIVEREGIONS,
1257 sizeof(o2hb_live_region_bitmap),
1258 O2NM_MAX_REGIONS,
1259 o2hb_live_region_bitmap);
1260 if (!o2hb_debug_liveregions) {
1261 mlog_errno(ret);
1262 goto bail;
1263 }
1264
1265 o2hb_debug_quorumregions =
1266 o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS,
1267 o2hb_debug_dir,
1268 &o2hb_db_quorumregions,
1269 sizeof(*o2hb_db_quorumregions),
1270 O2HB_DB_TYPE_QUORUMREGIONS,
1271 sizeof(o2hb_quorum_region_bitmap),
1272 O2NM_MAX_REGIONS,
1273 o2hb_quorum_region_bitmap);
1274 if (!o2hb_debug_quorumregions) {
1275 mlog_errno(ret);
1276 goto bail;
1277 }
1278
1279 o2hb_debug_failedregions =
1280 o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS,
1281 o2hb_debug_dir,
1282 &o2hb_db_failedregions,
1283 sizeof(*o2hb_db_failedregions),
1284 O2HB_DB_TYPE_FAILEDREGIONS,
1285 sizeof(o2hb_failed_region_bitmap),
1286 O2NM_MAX_REGIONS,
1287 o2hb_failed_region_bitmap);
1288 if (!o2hb_debug_failedregions) {
1289 mlog_errno(ret);
1290 goto bail;
1291 }
1292
1293 ret = 0;
1294bail:
1295 if (ret)
1296 o2hb_exit();
1297
1298 return ret;
985} 1299}
986 1300
987int o2hb_init(void) 1301int o2hb_init(void)
@@ -997,24 +1311,12 @@ int o2hb_init(void)
997 INIT_LIST_HEAD(&o2hb_node_events); 1311 INIT_LIST_HEAD(&o2hb_node_events);
998 1312
999 memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap)); 1313 memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
1314 memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap));
1315 memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap));
1316 memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap));
1317 memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap));
1000 1318
1001 o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL); 1319 return o2hb_debug_init();
1002 if (!o2hb_debug_dir) {
1003 mlog_errno(-ENOMEM);
1004 return -ENOMEM;
1005 }
1006
1007 o2hb_debug_livenodes = debugfs_create_file(O2HB_DEBUG_LIVENODES,
1008 S_IFREG|S_IRUSR,
1009 o2hb_debug_dir, NULL,
1010 &o2hb_debug_fops);
1011 if (!o2hb_debug_livenodes) {
1012 mlog_errno(-ENOMEM);
1013 debugfs_remove(o2hb_debug_dir);
1014 return -ENOMEM;
1015 }
1016
1017 return 0;
1018} 1320}
1019 1321
1020/* if we're already in a callback then we're already serialized by the sem */ 1322/* if we're already in a callback then we're already serialized by the sem */
@@ -1078,6 +1380,13 @@ static void o2hb_region_release(struct config_item *item)
1078 if (reg->hr_slots) 1380 if (reg->hr_slots)
1079 kfree(reg->hr_slots); 1381 kfree(reg->hr_slots);
1080 1382
1383 kfree(reg->hr_db_regnum);
1384 kfree(reg->hr_db_livenodes);
1385 debugfs_remove(reg->hr_debug_livenodes);
1386 debugfs_remove(reg->hr_debug_regnum);
1387 debugfs_remove(reg->hr_debug_elapsed_time);
1388 debugfs_remove(reg->hr_debug_dir);
1389
1081 spin_lock(&o2hb_live_lock); 1390 spin_lock(&o2hb_live_lock);
1082 list_del(&reg->hr_all_item); 1391 list_del(&reg->hr_all_item);
1083 spin_unlock(&o2hb_live_lock); 1392 spin_unlock(&o2hb_live_lock);
@@ -1441,6 +1750,8 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1441 /* Ok, we were woken. Make sure it wasn't by drop_item() */ 1750 /* Ok, we were woken. Make sure it wasn't by drop_item() */
1442 spin_lock(&o2hb_live_lock); 1751 spin_lock(&o2hb_live_lock);
1443 hb_task = reg->hr_task; 1752 hb_task = reg->hr_task;
1753 if (o2hb_global_heartbeat_active())
1754 set_bit(reg->hr_region_num, o2hb_live_region_bitmap);
1444 spin_unlock(&o2hb_live_lock); 1755 spin_unlock(&o2hb_live_lock);
1445 1756
1446 if (hb_task) 1757 if (hb_task)
@@ -1448,6 +1759,10 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1448 else 1759 else
1449 ret = -EIO; 1760 ret = -EIO;
1450 1761
1762 if (hb_task && o2hb_global_heartbeat_active())
1763 printk(KERN_NOTICE "o2hb: Heartbeat started on region %s\n",
1764 config_item_name(&reg->hr_item));
1765
1451out: 1766out:
1452 if (filp) 1767 if (filp)
1453 fput(filp); 1768 fput(filp);
@@ -1586,21 +1901,94 @@ static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group
1586 : NULL; 1901 : NULL;
1587} 1902}
1588 1903
1904static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir)
1905{
1906 int ret = -ENOMEM;
1907
1908 reg->hr_debug_dir =
1909 debugfs_create_dir(config_item_name(&reg->hr_item), dir);
1910 if (!reg->hr_debug_dir) {
1911 mlog_errno(ret);
1912 goto bail;
1913 }
1914
1915 reg->hr_debug_livenodes =
1916 o2hb_debug_create(O2HB_DEBUG_LIVENODES,
1917 reg->hr_debug_dir,
1918 &(reg->hr_db_livenodes),
1919 sizeof(*(reg->hr_db_livenodes)),
1920 O2HB_DB_TYPE_REGION_LIVENODES,
1921 sizeof(reg->hr_live_node_bitmap),
1922 O2NM_MAX_NODES, reg);
1923 if (!reg->hr_debug_livenodes) {
1924 mlog_errno(ret);
1925 goto bail;
1926 }
1927
1928 reg->hr_debug_regnum =
1929 o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER,
1930 reg->hr_debug_dir,
1931 &(reg->hr_db_regnum),
1932 sizeof(*(reg->hr_db_regnum)),
1933 O2HB_DB_TYPE_REGION_NUMBER,
1934 0, O2NM_MAX_NODES, reg);
1935 if (!reg->hr_debug_regnum) {
1936 mlog_errno(ret);
1937 goto bail;
1938 }
1939
1940 reg->hr_debug_elapsed_time =
1941 o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME,
1942 reg->hr_debug_dir,
1943 &(reg->hr_db_elapsed_time),
1944 sizeof(*(reg->hr_db_elapsed_time)),
1945 O2HB_DB_TYPE_REGION_ELAPSED_TIME,
1946 0, 0, reg);
1947 if (!reg->hr_debug_elapsed_time) {
1948 mlog_errno(ret);
1949 goto bail;
1950 }
1951
1952 ret = 0;
1953bail:
1954 return ret;
1955}
1956
1589static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group, 1957static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
1590 const char *name) 1958 const char *name)
1591{ 1959{
1592 struct o2hb_region *reg = NULL; 1960 struct o2hb_region *reg = NULL;
1961 int ret;
1593 1962
1594 reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL); 1963 reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
1595 if (reg == NULL) 1964 if (reg == NULL)
1596 return ERR_PTR(-ENOMEM); 1965 return ERR_PTR(-ENOMEM);
1597 1966
1598 config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type); 1967 if (strlen(name) > O2HB_MAX_REGION_NAME_LEN)
1968 return ERR_PTR(-ENAMETOOLONG);
1599 1969
1600 spin_lock(&o2hb_live_lock); 1970 spin_lock(&o2hb_live_lock);
1971 reg->hr_region_num = 0;
1972 if (o2hb_global_heartbeat_active()) {
1973 reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap,
1974 O2NM_MAX_REGIONS);
1975 if (reg->hr_region_num >= O2NM_MAX_REGIONS) {
1976 spin_unlock(&o2hb_live_lock);
1977 return ERR_PTR(-EFBIG);
1978 }
1979 set_bit(reg->hr_region_num, o2hb_region_bitmap);
1980 }
1601 list_add_tail(&reg->hr_all_item, &o2hb_all_regions); 1981 list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
1602 spin_unlock(&o2hb_live_lock); 1982 spin_unlock(&o2hb_live_lock);
1603 1983
1984 config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
1985
1986 ret = o2hb_debug_region_init(reg, o2hb_debug_dir);
1987 if (ret) {
1988 config_item_put(&reg->hr_item);
1989 return ERR_PTR(ret);
1990 }
1991
1604 return &reg->hr_item; 1992 return &reg->hr_item;
1605} 1993}
1606 1994
@@ -1612,6 +2000,10 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group,
1612 2000
1613 /* stop the thread when the user removes the region dir */ 2001 /* stop the thread when the user removes the region dir */
1614 spin_lock(&o2hb_live_lock); 2002 spin_lock(&o2hb_live_lock);
2003 if (o2hb_global_heartbeat_active()) {
2004 clear_bit(reg->hr_region_num, o2hb_region_bitmap);
2005 clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
2006 }
1615 hb_task = reg->hr_task; 2007 hb_task = reg->hr_task;
1616 reg->hr_task = NULL; 2008 reg->hr_task = NULL;
1617 spin_unlock(&o2hb_live_lock); 2009 spin_unlock(&o2hb_live_lock);
@@ -1628,6 +2020,9 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group,
1628 wake_up(&o2hb_steady_queue); 2020 wake_up(&o2hb_steady_queue);
1629 } 2021 }
1630 2022
2023 if (o2hb_global_heartbeat_active())
2024 printk(KERN_NOTICE "o2hb: Heartbeat stopped on region %s\n",
2025 config_item_name(&reg->hr_item));
1631 config_item_put(item); 2026 config_item_put(item);
1632} 2027}
1633 2028
@@ -1688,6 +2083,41 @@ static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group
1688 return count; 2083 return count;
1689} 2084}
1690 2085
2086static
2087ssize_t o2hb_heartbeat_group_mode_show(struct o2hb_heartbeat_group *group,
2088 char *page)
2089{
2090 return sprintf(page, "%s\n",
2091 o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]);
2092}
2093
2094static
2095ssize_t o2hb_heartbeat_group_mode_store(struct o2hb_heartbeat_group *group,
2096 const char *page, size_t count)
2097{
2098 unsigned int i;
2099 int ret;
2100 size_t len;
2101
2102 len = (page[count - 1] == '\n') ? count - 1 : count;
2103 if (!len)
2104 return -EINVAL;
2105
2106 for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) {
2107 if (strnicmp(page, o2hb_heartbeat_mode_desc[i], len))
2108 continue;
2109
2110 ret = o2hb_global_hearbeat_mode_set(i);
2111 if (!ret)
2112 printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n",
2113 o2hb_heartbeat_mode_desc[i]);
2114 return count;
2115 }
2116
2117 return -EINVAL;
2118
2119}
2120
1691static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = { 2121static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = {
1692 .attr = { .ca_owner = THIS_MODULE, 2122 .attr = { .ca_owner = THIS_MODULE,
1693 .ca_name = "dead_threshold", 2123 .ca_name = "dead_threshold",
@@ -1696,8 +2126,17 @@ static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold
1696 .store = o2hb_heartbeat_group_threshold_store, 2126 .store = o2hb_heartbeat_group_threshold_store,
1697}; 2127};
1698 2128
2129static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_mode = {
2130 .attr = { .ca_owner = THIS_MODULE,
2131 .ca_name = "mode",
2132 .ca_mode = S_IRUGO | S_IWUSR },
2133 .show = o2hb_heartbeat_group_mode_show,
2134 .store = o2hb_heartbeat_group_mode_store,
2135};
2136
1699static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = { 2137static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
1700 &o2hb_heartbeat_group_attr_threshold.attr, 2138 &o2hb_heartbeat_group_attr_threshold.attr,
2139 &o2hb_heartbeat_group_attr_mode.attr,
1701 NULL, 2140 NULL,
1702}; 2141};
1703 2142
@@ -1963,3 +2402,34 @@ void o2hb_stop_all_regions(void)
1963 spin_unlock(&o2hb_live_lock); 2402 spin_unlock(&o2hb_live_lock);
1964} 2403}
1965EXPORT_SYMBOL_GPL(o2hb_stop_all_regions); 2404EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
2405
2406int o2hb_get_all_regions(char *region_uuids, u8 max_regions)
2407{
2408 struct o2hb_region *reg;
2409 int numregs = 0;
2410 char *p;
2411
2412 spin_lock(&o2hb_live_lock);
2413
2414 p = region_uuids;
2415 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2416 mlog(0, "Region: %s\n", config_item_name(&reg->hr_item));
2417 if (numregs < max_regions) {
2418 memcpy(p, config_item_name(&reg->hr_item),
2419 O2HB_MAX_REGION_NAME_LEN);
2420 p += O2HB_MAX_REGION_NAME_LEN;
2421 }
2422 numregs++;
2423 }
2424
2425 spin_unlock(&o2hb_live_lock);
2426
2427 return numregs;
2428}
2429EXPORT_SYMBOL_GPL(o2hb_get_all_regions);
2430
2431int o2hb_global_heartbeat_active(void)
2432{
2433 return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL);
2434}
2435EXPORT_SYMBOL(o2hb_global_heartbeat_active);
diff --git a/fs/ocfs2/cluster/heartbeat.h b/fs/ocfs2/cluster/heartbeat.h
index 2f1649253b49..00ad8e8fea51 100644
--- a/fs/ocfs2/cluster/heartbeat.h
+++ b/fs/ocfs2/cluster/heartbeat.h
@@ -31,6 +31,8 @@
31 31
32#define O2HB_REGION_TIMEOUT_MS 2000 32#define O2HB_REGION_TIMEOUT_MS 2000
33 33
34#define O2HB_MAX_REGION_NAME_LEN 32
35
34/* number of changes to be seen as live */ 36/* number of changes to be seen as live */
35#define O2HB_LIVE_THRESHOLD 2 37#define O2HB_LIVE_THRESHOLD 2
36/* number of equal samples to be seen as dead */ 38/* number of equal samples to be seen as dead */
@@ -81,5 +83,7 @@ int o2hb_check_node_heartbeating(u8 node_num);
81int o2hb_check_node_heartbeating_from_callback(u8 node_num); 83int o2hb_check_node_heartbeating_from_callback(u8 node_num);
82int o2hb_check_local_node_heartbeating(void); 84int o2hb_check_local_node_heartbeating(void);
83void o2hb_stop_all_regions(void); 85void o2hb_stop_all_regions(void);
86int o2hb_get_all_regions(char *region_uuids, u8 numregions);
87int o2hb_global_heartbeat_active(void);
84 88
85#endif /* O2CLUSTER_HEARTBEAT_H */ 89#endif /* O2CLUSTER_HEARTBEAT_H */
diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h
index fd96e2a2fa56..ea2ed9f56c94 100644
--- a/fs/ocfs2/cluster/masklog.h
+++ b/fs/ocfs2/cluster/masklog.h
@@ -119,7 +119,8 @@
119#define ML_ERROR 0x0000000100000000ULL /* sent to KERN_ERR */ 119#define ML_ERROR 0x0000000100000000ULL /* sent to KERN_ERR */
120#define ML_NOTICE 0x0000000200000000ULL /* setn to KERN_NOTICE */ 120#define ML_NOTICE 0x0000000200000000ULL /* setn to KERN_NOTICE */
121#define ML_KTHREAD 0x0000000400000000ULL /* kernel thread activity */ 121#define ML_KTHREAD 0x0000000400000000ULL /* kernel thread activity */
122#define ML_RESERVATIONS 0x0000000800000000ULL /* ocfs2 alloc reservations */ 122#define ML_RESERVATIONS 0x0000000800000000ULL /* ocfs2 alloc reservations */
123#define ML_CLUSTER 0x0000001000000000ULL /* cluster stack */
123 124
124#define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE) 125#define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
125#define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT) 126#define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT)
diff --git a/fs/ocfs2/cluster/nodemanager.c b/fs/ocfs2/cluster/nodemanager.c
index ed0c9f367fed..bb240647ca5f 100644
--- a/fs/ocfs2/cluster/nodemanager.c
+++ b/fs/ocfs2/cluster/nodemanager.c
@@ -711,6 +711,8 @@ static struct config_item *o2nm_node_group_make_item(struct config_group *group,
711 config_item_init_type_name(&node->nd_item, name, &o2nm_node_type); 711 config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
712 spin_lock_init(&node->nd_lock); 712 spin_lock_init(&node->nd_lock);
713 713
714 mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name);
715
714 return &node->nd_item; 716 return &node->nd_item;
715} 717}
716 718
@@ -744,6 +746,9 @@ static void o2nm_node_group_drop_item(struct config_group *group,
744 } 746 }
745 write_unlock(&cluster->cl_nodes_lock); 747 write_unlock(&cluster->cl_nodes_lock);
746 748
749 mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n",
750 config_item_name(&node->nd_item));
751
747 config_item_put(item); 752 config_item_put(item);
748} 753}
749 754
diff --git a/fs/ocfs2/cluster/ocfs2_nodemanager.h b/fs/ocfs2/cluster/ocfs2_nodemanager.h
index 5b9854bad571..49b594325bec 100644
--- a/fs/ocfs2/cluster/ocfs2_nodemanager.h
+++ b/fs/ocfs2/cluster/ocfs2_nodemanager.h
@@ -36,4 +36,10 @@
36/* host name, group name, cluster name all 64 bytes */ 36/* host name, group name, cluster name all 64 bytes */
37#define O2NM_MAX_NAME_LEN 64 // __NEW_UTS_LEN 37#define O2NM_MAX_NAME_LEN 64 // __NEW_UTS_LEN
38 38
39/*
40 * Maximum number of global heartbeat regions allowed.
41 * **CAUTION** Changing this number will break dlm compatibility.
42 */
43#define O2NM_MAX_REGIONS 32
44
39#endif /* _OCFS2_NODEMANAGER_H */ 45#endif /* _OCFS2_NODEMANAGER_H */
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index cbe2f057cc28..9aa426e42123 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1696,6 +1696,9 @@ static void o2net_hb_node_down_cb(struct o2nm_node *node, int node_num,
1696{ 1696{
1697 o2quo_hb_down(node_num); 1697 o2quo_hb_down(node_num);
1698 1698
1699 if (!node)
1700 return;
1701
1699 if (node_num != o2nm_this_node()) 1702 if (node_num != o2nm_this_node())
1700 o2net_disconnect_node(node); 1703 o2net_disconnect_node(node);
1701 1704
@@ -1709,6 +1712,8 @@ static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num,
1709 1712
1710 o2quo_hb_up(node_num); 1713 o2quo_hb_up(node_num);
1711 1714
1715 BUG_ON(!node);
1716
1712 /* ensure an immediate connect attempt */ 1717 /* ensure an immediate connect attempt */
1713 nn->nn_last_connect_attempt = jiffies - 1718 nn->nn_last_connect_attempt = jiffies -
1714 (msecs_to_jiffies(o2net_reconnect_delay()) + 1); 1719 (msecs_to_jiffies(o2net_reconnect_delay()) + 1);
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index b4957c7d9fe2..edaded48e7e9 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -40,6 +40,14 @@
40#include "inode.h" 40#include "inode.h"
41#include "super.h" 41#include "super.h"
42 42
43void ocfs2_dentry_attach_gen(struct dentry *dentry)
44{
45 unsigned long gen =
46 OCFS2_I(dentry->d_parent->d_inode)->ip_dir_lock_gen;
47 BUG_ON(dentry->d_inode);
48 dentry->d_fsdata = (void *)gen;
49}
50
43 51
44static int ocfs2_dentry_revalidate(struct dentry *dentry, 52static int ocfs2_dentry_revalidate(struct dentry *dentry,
45 struct nameidata *nd) 53 struct nameidata *nd)
@@ -51,11 +59,20 @@ static int ocfs2_dentry_revalidate(struct dentry *dentry,
51 mlog_entry("(0x%p, '%.*s')\n", dentry, 59 mlog_entry("(0x%p, '%.*s')\n", dentry,
52 dentry->d_name.len, dentry->d_name.name); 60 dentry->d_name.len, dentry->d_name.name);
53 61
54 /* Never trust a negative dentry - force a new lookup. */ 62 /* For a negative dentry -
63 * check the generation number of the parent and compare with the
64 * one stored in the inode.
65 */
55 if (inode == NULL) { 66 if (inode == NULL) {
56 mlog(0, "negative dentry: %.*s\n", dentry->d_name.len, 67 unsigned long gen = (unsigned long) dentry->d_fsdata;
57 dentry->d_name.name); 68 unsigned long pgen =
58 goto bail; 69 OCFS2_I(dentry->d_parent->d_inode)->ip_dir_lock_gen;
70 mlog(0, "negative dentry: %.*s parent gen: %lu "
71 "dentry gen: %lu\n",
72 dentry->d_name.len, dentry->d_name.name, pgen, gen);
73 if (gen != pgen)
74 goto bail;
75 goto valid;
59 } 76 }
60 77
61 BUG_ON(!osb); 78 BUG_ON(!osb);
@@ -96,6 +113,7 @@ static int ocfs2_dentry_revalidate(struct dentry *dentry,
96 goto bail; 113 goto bail;
97 } 114 }
98 115
116valid:
99 ret = 1; 117 ret = 1;
100 118
101bail: 119bail:
@@ -227,6 +245,12 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,
227 if (!inode) 245 if (!inode)
228 return 0; 246 return 0;
229 247
248 if (!dentry->d_inode && dentry->d_fsdata) {
249 /* Converting a negative dentry to positive
250 Clear dentry->d_fsdata */
251 dentry->d_fsdata = dl = NULL;
252 }
253
230 if (dl) { 254 if (dl) {
231 mlog_bug_on_msg(dl->dl_parent_blkno != parent_blkno, 255 mlog_bug_on_msg(dl->dl_parent_blkno != parent_blkno,
232 " \"%.*s\": old parent: %llu, new: %llu\n", 256 " \"%.*s\": old parent: %llu, new: %llu\n",
@@ -452,6 +476,7 @@ static void ocfs2_dentry_iput(struct dentry *dentry, struct inode *inode)
452 476
453out: 477out:
454 iput(inode); 478 iput(inode);
479 ocfs2_dentry_attach_gen(dentry);
455} 480}
456 481
457/* 482/*
diff --git a/fs/ocfs2/dcache.h b/fs/ocfs2/dcache.h
index f5dd1789acf1..b79eff709958 100644
--- a/fs/ocfs2/dcache.h
+++ b/fs/ocfs2/dcache.h
@@ -64,5 +64,6 @@ void ocfs2_dentry_move(struct dentry *dentry, struct dentry *target,
64 struct inode *old_dir, struct inode *new_dir); 64 struct inode *old_dir, struct inode *new_dir);
65 65
66extern spinlock_t dentry_attach_lock; 66extern spinlock_t dentry_attach_lock;
67void ocfs2_dentry_attach_gen(struct dentry *dentry);
67 68
68#endif /* OCFS2_DCACHE_H */ 69#endif /* OCFS2_DCACHE_H */
diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index 765298908f1d..b36d0bf77a5a 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -445,7 +445,9 @@ enum {
445 DLM_LOCK_REQUEST_MSG, /* 515 */ 445 DLM_LOCK_REQUEST_MSG, /* 515 */
446 DLM_RECO_DATA_DONE_MSG, /* 516 */ 446 DLM_RECO_DATA_DONE_MSG, /* 516 */
447 DLM_BEGIN_RECO_MSG, /* 517 */ 447 DLM_BEGIN_RECO_MSG, /* 517 */
448 DLM_FINALIZE_RECO_MSG /* 518 */ 448 DLM_FINALIZE_RECO_MSG, /* 518 */
449 DLM_QUERY_REGION, /* 519 */
450 DLM_QUERY_NODEINFO, /* 520 */
449}; 451};
450 452
451struct dlm_reco_node_data 453struct dlm_reco_node_data
@@ -727,6 +729,31 @@ struct dlm_cancel_join
727 u8 domain[O2NM_MAX_NAME_LEN]; 729 u8 domain[O2NM_MAX_NAME_LEN];
728}; 730};
729 731
732struct dlm_query_region {
733 u8 qr_node;
734 u8 qr_numregions;
735 u8 qr_namelen;
736 u8 pad1;
737 u8 qr_domain[O2NM_MAX_NAME_LEN];
738 u8 qr_regions[O2HB_MAX_REGION_NAME_LEN * O2NM_MAX_REGIONS];
739};
740
741struct dlm_node_info {
742 u8 ni_nodenum;
743 u8 pad1;
744 u16 ni_ipv4_port;
745 u32 ni_ipv4_address;
746};
747
748struct dlm_query_nodeinfo {
749 u8 qn_nodenum;
750 u8 qn_numnodes;
751 u8 qn_namelen;
752 u8 pad1;
753 u8 qn_domain[O2NM_MAX_NAME_LEN];
754 struct dlm_node_info qn_nodes[O2NM_MAX_NODES];
755};
756
730struct dlm_exit_domain 757struct dlm_exit_domain
731{ 758{
732 u8 node_idx; 759 u8 node_idx;
diff --git a/fs/ocfs2/dlm/dlmdebug.c b/fs/ocfs2/dlm/dlmdebug.c
index 901ca52bf86b..272ec8631a51 100644
--- a/fs/ocfs2/dlm/dlmdebug.c
+++ b/fs/ocfs2/dlm/dlmdebug.c
@@ -493,7 +493,7 @@ static int debug_mle_print(struct dlm_ctxt *dlm, struct debug_buffer *db)
493 struct hlist_head *bucket; 493 struct hlist_head *bucket;
494 struct hlist_node *list; 494 struct hlist_node *list;
495 int i, out = 0; 495 int i, out = 0;
496 unsigned long total = 0, longest = 0, bktcnt; 496 unsigned long total = 0, longest = 0, bucket_count = 0;
497 497
498 out += snprintf(db->buf + out, db->len - out, 498 out += snprintf(db->buf + out, db->len - out,
499 "Dumping MLEs for Domain: %s\n", dlm->name); 499 "Dumping MLEs for Domain: %s\n", dlm->name);
@@ -505,13 +505,13 @@ static int debug_mle_print(struct dlm_ctxt *dlm, struct debug_buffer *db)
505 mle = hlist_entry(list, struct dlm_master_list_entry, 505 mle = hlist_entry(list, struct dlm_master_list_entry,
506 master_hash_node); 506 master_hash_node);
507 ++total; 507 ++total;
508 ++bktcnt; 508 ++bucket_count;
509 if (db->len - out < 200) 509 if (db->len - out < 200)
510 continue; 510 continue;
511 out += dump_mle(mle, db->buf + out, db->len - out); 511 out += dump_mle(mle, db->buf + out, db->len - out);
512 } 512 }
513 longest = max(longest, bktcnt); 513 longest = max(longest, bucket_count);
514 bktcnt = 0; 514 bucket_count = 0;
515 } 515 }
516 spin_unlock(&dlm->master_lock); 516 spin_unlock(&dlm->master_lock);
517 517
@@ -782,7 +782,9 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db)
782 782
783 /* Domain: xxxxxxxxxx Key: 0xdfbac769 */ 783 /* Domain: xxxxxxxxxx Key: 0xdfbac769 */
784 out += snprintf(db->buf + out, db->len - out, 784 out += snprintf(db->buf + out, db->len - out,
785 "Domain: %s Key: 0x%08x\n", dlm->name, dlm->key); 785 "Domain: %s Key: 0x%08x Protocol: %d.%d\n",
786 dlm->name, dlm->key, dlm->dlm_locking_proto.pv_major,
787 dlm->dlm_locking_proto.pv_minor);
786 788
787 /* Thread Pid: xxx Node: xxx State: xxxxx */ 789 /* Thread Pid: xxx Node: xxx State: xxxxx */
788 out += snprintf(db->buf + out, db->len - out, 790 out += snprintf(db->buf + out, db->len - out,
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 11a5c87fd7f7..58a93b953735 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -128,10 +128,14 @@ static DECLARE_WAIT_QUEUE_HEAD(dlm_domain_events);
128 * will have a negotiated version with the same major number and a minor 128 * will have a negotiated version with the same major number and a minor
129 * number equal or smaller. The dlm_ctxt->dlm_locking_proto field should 129 * number equal or smaller. The dlm_ctxt->dlm_locking_proto field should
130 * be used to determine what a running domain is actually using. 130 * be used to determine what a running domain is actually using.
131 *
132 * New in version 1.1:
133 * - Message DLM_QUERY_REGION added to support global heartbeat
134 * - Message DLM_QUERY_NODEINFO added to allow online node removes
131 */ 135 */
132static const struct dlm_protocol_version dlm_protocol = { 136static const struct dlm_protocol_version dlm_protocol = {
133 .pv_major = 1, 137 .pv_major = 1,
134 .pv_minor = 0, 138 .pv_minor = 1,
135}; 139};
136 140
137#define DLM_DOMAIN_BACKOFF_MS 200 141#define DLM_DOMAIN_BACKOFF_MS 200
@@ -142,6 +146,8 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
142 void **ret_data); 146 void **ret_data);
143static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data, 147static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data,
144 void **ret_data); 148 void **ret_data);
149static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
150 void *data, void **ret_data);
145static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, 151static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
146 void **ret_data); 152 void **ret_data);
147static int dlm_protocol_compare(struct dlm_protocol_version *existing, 153static int dlm_protocol_compare(struct dlm_protocol_version *existing,
@@ -921,6 +927,370 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
921 return 0; 927 return 0;
922} 928}
923 929
930static int dlm_match_regions(struct dlm_ctxt *dlm,
931 struct dlm_query_region *qr)
932{
933 char *local = NULL, *remote = qr->qr_regions;
934 char *l, *r;
935 int localnr, i, j, foundit;
936 int status = 0;
937
938 if (!o2hb_global_heartbeat_active()) {
939 if (qr->qr_numregions) {
940 mlog(ML_ERROR, "Domain %s: Joining node %d has global "
941 "heartbeat enabled but local node %d does not\n",
942 qr->qr_domain, qr->qr_node, dlm->node_num);
943 status = -EINVAL;
944 }
945 goto bail;
946 }
947
948 if (o2hb_global_heartbeat_active() && !qr->qr_numregions) {
949 mlog(ML_ERROR, "Domain %s: Local node %d has global "
950 "heartbeat enabled but joining node %d does not\n",
951 qr->qr_domain, dlm->node_num, qr->qr_node);
952 status = -EINVAL;
953 goto bail;
954 }
955
956 r = remote;
957 for (i = 0; i < qr->qr_numregions; ++i) {
958 mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, r);
959 r += O2HB_MAX_REGION_NAME_LEN;
960 }
961
962 local = kmalloc(sizeof(qr->qr_regions), GFP_KERNEL);
963 if (!local) {
964 status = -ENOMEM;
965 goto bail;
966 }
967
968 localnr = o2hb_get_all_regions(local, O2NM_MAX_REGIONS);
969
970 /* compare local regions with remote */
971 l = local;
972 for (i = 0; i < localnr; ++i) {
973 foundit = 0;
974 r = remote;
975 for (j = 0; j <= qr->qr_numregions; ++j) {
976 if (!memcmp(l, r, O2HB_MAX_REGION_NAME_LEN)) {
977 foundit = 1;
978 break;
979 }
980 r += O2HB_MAX_REGION_NAME_LEN;
981 }
982 if (!foundit) {
983 status = -EINVAL;
984 mlog(ML_ERROR, "Domain %s: Region '%.*s' registered "
985 "in local node %d but not in joining node %d\n",
986 qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, l,
987 dlm->node_num, qr->qr_node);
988 goto bail;
989 }
990 l += O2HB_MAX_REGION_NAME_LEN;
991 }
992
993 /* compare remote with local regions */
994 r = remote;
995 for (i = 0; i < qr->qr_numregions; ++i) {
996 foundit = 0;
997 l = local;
998 for (j = 0; j < localnr; ++j) {
999 if (!memcmp(r, l, O2HB_MAX_REGION_NAME_LEN)) {
1000 foundit = 1;
1001 break;
1002 }
1003 l += O2HB_MAX_REGION_NAME_LEN;
1004 }
1005 if (!foundit) {
1006 status = -EINVAL;
1007 mlog(ML_ERROR, "Domain %s: Region '%.*s' registered "
1008 "in joining node %d but not in local node %d\n",
1009 qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, r,
1010 qr->qr_node, dlm->node_num);
1011 goto bail;
1012 }
1013 r += O2HB_MAX_REGION_NAME_LEN;
1014 }
1015
1016bail:
1017 kfree(local);
1018
1019 return status;
1020}
1021
1022static int dlm_send_regions(struct dlm_ctxt *dlm, unsigned long *node_map)
1023{
1024 struct dlm_query_region *qr = NULL;
1025 int status, ret = 0, i;
1026 char *p;
1027
1028 if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES)
1029 goto bail;
1030
1031 qr = kzalloc(sizeof(struct dlm_query_region), GFP_KERNEL);
1032 if (!qr) {
1033 ret = -ENOMEM;
1034 mlog_errno(ret);
1035 goto bail;
1036 }
1037
1038 qr->qr_node = dlm->node_num;
1039 qr->qr_namelen = strlen(dlm->name);
1040 memcpy(qr->qr_domain, dlm->name, qr->qr_namelen);
1041 /* if local hb, the numregions will be zero */
1042 if (o2hb_global_heartbeat_active())
1043 qr->qr_numregions = o2hb_get_all_regions(qr->qr_regions,
1044 O2NM_MAX_REGIONS);
1045
1046 p = qr->qr_regions;
1047 for (i = 0; i < qr->qr_numregions; ++i, p += O2HB_MAX_REGION_NAME_LEN)
1048 mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, p);
1049
1050 i = -1;
1051 while ((i = find_next_bit(node_map, O2NM_MAX_NODES,
1052 i + 1)) < O2NM_MAX_NODES) {
1053 if (i == dlm->node_num)
1054 continue;
1055
1056 mlog(0, "Sending regions to node %d\n", i);
1057
1058 ret = o2net_send_message(DLM_QUERY_REGION, DLM_MOD_KEY, qr,
1059 sizeof(struct dlm_query_region),
1060 i, &status);
1061 if (ret >= 0)
1062 ret = status;
1063 if (ret) {
1064 mlog(ML_ERROR, "Region mismatch %d, node %d\n",
1065 ret, i);
1066 break;
1067 }
1068 }
1069
1070bail:
1071 kfree(qr);
1072 return ret;
1073}
1074
1075static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
1076 void *data, void **ret_data)
1077{
1078 struct dlm_query_region *qr;
1079 struct dlm_ctxt *dlm = NULL;
1080 int status = 0;
1081 int locked = 0;
1082
1083 qr = (struct dlm_query_region *) msg->buf;
1084
1085 mlog(0, "Node %u queries hb regions on domain %s\n", qr->qr_node,
1086 qr->qr_domain);
1087
1088 status = -EINVAL;
1089
1090 spin_lock(&dlm_domain_lock);
1091 dlm = __dlm_lookup_domain_full(qr->qr_domain, qr->qr_namelen);
1092 if (!dlm) {
1093 mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
1094 "before join domain\n", qr->qr_node, qr->qr_domain);
1095 goto bail;
1096 }
1097
1098 spin_lock(&dlm->spinlock);
1099 locked = 1;
1100 if (dlm->joining_node != qr->qr_node) {
1101 mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
1102 "but joining node is %d\n", qr->qr_node, qr->qr_domain,
1103 dlm->joining_node);
1104 goto bail;
1105 }
1106
1107 /* Support for global heartbeat was added in 1.1 */
1108 if (dlm->dlm_locking_proto.pv_major == 1 &&
1109 dlm->dlm_locking_proto.pv_minor == 0) {
1110 mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
1111 "but active dlm protocol is %d.%d\n", qr->qr_node,
1112 qr->qr_domain, dlm->dlm_locking_proto.pv_major,
1113 dlm->dlm_locking_proto.pv_minor);
1114 goto bail;
1115 }
1116
1117 status = dlm_match_regions(dlm, qr);
1118
1119bail:
1120 if (locked)
1121 spin_unlock(&dlm->spinlock);
1122 spin_unlock(&dlm_domain_lock);
1123
1124 return status;
1125}
1126
1127static int dlm_match_nodes(struct dlm_ctxt *dlm, struct dlm_query_nodeinfo *qn)
1128{
1129 struct o2nm_node *local;
1130 struct dlm_node_info *remote;
1131 int i, j;
1132 int status = 0;
1133
1134 for (j = 0; j < qn->qn_numnodes; ++j)
1135 mlog(0, "Node %3d, %pI4:%u\n", qn->qn_nodes[j].ni_nodenum,
1136 &(qn->qn_nodes[j].ni_ipv4_address),
1137 ntohs(qn->qn_nodes[j].ni_ipv4_port));
1138
1139 for (i = 0; i < O2NM_MAX_NODES && !status; ++i) {
1140 local = o2nm_get_node_by_num(i);
1141 remote = NULL;
1142 for (j = 0; j < qn->qn_numnodes; ++j) {
1143 if (qn->qn_nodes[j].ni_nodenum == i) {
1144 remote = &(qn->qn_nodes[j]);
1145 break;
1146 }
1147 }
1148
1149 if (!local && !remote)
1150 continue;
1151
1152 if ((local && !remote) || (!local && remote))
1153 status = -EINVAL;
1154
1155 if (!status &&
1156 ((remote->ni_nodenum != local->nd_num) ||
1157 (remote->ni_ipv4_port != local->nd_ipv4_port) ||
1158 (remote->ni_ipv4_address != local->nd_ipv4_address)))
1159 status = -EINVAL;
1160
1161 if (status) {
1162 if (remote && !local)
1163 mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) "
1164 "registered in joining node %d but not in "
1165 "local node %d\n", qn->qn_domain,
1166 remote->ni_nodenum,
1167 &(remote->ni_ipv4_address),
1168 ntohs(remote->ni_ipv4_port),
1169 qn->qn_nodenum, dlm->node_num);
1170 if (local && !remote)
1171 mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) "
1172 "registered in local node %d but not in "
1173 "joining node %d\n", qn->qn_domain,
1174 local->nd_num, &(local->nd_ipv4_address),
1175 ntohs(local->nd_ipv4_port),
1176 dlm->node_num, qn->qn_nodenum);
1177 BUG_ON((!local && !remote));
1178 }
1179
1180 if (local)
1181 o2nm_node_put(local);
1182 }
1183
1184 return status;
1185}
1186
1187static int dlm_send_nodeinfo(struct dlm_ctxt *dlm, unsigned long *node_map)
1188{
1189 struct dlm_query_nodeinfo *qn = NULL;
1190 struct o2nm_node *node;
1191 int ret = 0, status, count, i;
1192
1193 if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES)
1194 goto bail;
1195
1196 qn = kzalloc(sizeof(struct dlm_query_nodeinfo), GFP_KERNEL);
1197 if (!qn) {
1198 ret = -ENOMEM;
1199 mlog_errno(ret);
1200 goto bail;
1201 }
1202
1203 for (i = 0, count = 0; i < O2NM_MAX_NODES; ++i) {
1204 node = o2nm_get_node_by_num(i);
1205 if (!node)
1206 continue;
1207 qn->qn_nodes[count].ni_nodenum = node->nd_num;
1208 qn->qn_nodes[count].ni_ipv4_port = node->nd_ipv4_port;
1209 qn->qn_nodes[count].ni_ipv4_address = node->nd_ipv4_address;
1210 mlog(0, "Node %3d, %pI4:%u\n", node->nd_num,
1211 &(node->nd_ipv4_address), ntohs(node->nd_ipv4_port));
1212 ++count;
1213 o2nm_node_put(node);
1214 }
1215
1216 qn->qn_nodenum = dlm->node_num;
1217 qn->qn_numnodes = count;
1218 qn->qn_namelen = strlen(dlm->name);
1219 memcpy(qn->qn_domain, dlm->name, qn->qn_namelen);
1220
1221 i = -1;
1222 while ((i = find_next_bit(node_map, O2NM_MAX_NODES,
1223 i + 1)) < O2NM_MAX_NODES) {
1224 if (i == dlm->node_num)
1225 continue;
1226
1227 mlog(0, "Sending nodeinfo to node %d\n", i);
1228
1229 ret = o2net_send_message(DLM_QUERY_NODEINFO, DLM_MOD_KEY,
1230 qn, sizeof(struct dlm_query_nodeinfo),
1231 i, &status);
1232 if (ret >= 0)
1233 ret = status;
1234 if (ret) {
1235 mlog(ML_ERROR, "node mismatch %d, node %d\n", ret, i);
1236 break;
1237 }
1238 }
1239
1240bail:
1241 kfree(qn);
1242 return ret;
1243}
1244
1245static int dlm_query_nodeinfo_handler(struct o2net_msg *msg, u32 len,
1246 void *data, void **ret_data)
1247{
1248 struct dlm_query_nodeinfo *qn;
1249 struct dlm_ctxt *dlm = NULL;
1250 int locked = 0, status = -EINVAL;
1251
1252 qn = (struct dlm_query_nodeinfo *) msg->buf;
1253
1254 mlog(0, "Node %u queries nodes on domain %s\n", qn->qn_nodenum,
1255 qn->qn_domain);
1256
1257 spin_lock(&dlm_domain_lock);
1258 dlm = __dlm_lookup_domain_full(qn->qn_domain, qn->qn_namelen);
1259 if (!dlm) {
1260 mlog(ML_ERROR, "Node %d queried nodes on domain %s before "
1261 "join domain\n", qn->qn_nodenum, qn->qn_domain);
1262 goto bail;
1263 }
1264
1265 spin_lock(&dlm->spinlock);
1266 locked = 1;
1267 if (dlm->joining_node != qn->qn_nodenum) {
1268 mlog(ML_ERROR, "Node %d queried nodes on domain %s but "
1269 "joining node is %d\n", qn->qn_nodenum, qn->qn_domain,
1270 dlm->joining_node);
1271 goto bail;
1272 }
1273
1274 /* Support for node query was added in 1.1 */
1275 if (dlm->dlm_locking_proto.pv_major == 1 &&
1276 dlm->dlm_locking_proto.pv_minor == 0) {
1277 mlog(ML_ERROR, "Node %d queried nodes on domain %s "
1278 "but active dlm protocol is %d.%d\n", qn->qn_nodenum,
1279 qn->qn_domain, dlm->dlm_locking_proto.pv_major,
1280 dlm->dlm_locking_proto.pv_minor);
1281 goto bail;
1282 }
1283
1284 status = dlm_match_nodes(dlm, qn);
1285
1286bail:
1287 if (locked)
1288 spin_unlock(&dlm->spinlock);
1289 spin_unlock(&dlm_domain_lock);
1290
1291 return status;
1292}
1293
924static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data, 1294static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data,
925 void **ret_data) 1295 void **ret_data)
926{ 1296{
@@ -1241,6 +1611,20 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm)
1241 set_bit(dlm->node_num, dlm->domain_map); 1611 set_bit(dlm->node_num, dlm->domain_map);
1242 spin_unlock(&dlm->spinlock); 1612 spin_unlock(&dlm->spinlock);
1243 1613
1614 /* Support for global heartbeat and node info was added in 1.1 */
1615 if (dlm_protocol.pv_major > 1 || dlm_protocol.pv_minor > 0) {
1616 status = dlm_send_nodeinfo(dlm, ctxt->yes_resp_map);
1617 if (status) {
1618 mlog_errno(status);
1619 goto bail;
1620 }
1621 status = dlm_send_regions(dlm, ctxt->yes_resp_map);
1622 if (status) {
1623 mlog_errno(status);
1624 goto bail;
1625 }
1626 }
1627
1244 dlm_send_join_asserts(dlm, ctxt->yes_resp_map); 1628 dlm_send_join_asserts(dlm, ctxt->yes_resp_map);
1245 1629
1246 /* Joined state *must* be set before the joining node 1630 /* Joined state *must* be set before the joining node
@@ -1807,7 +2191,21 @@ static int dlm_register_net_handlers(void)
1807 sizeof(struct dlm_cancel_join), 2191 sizeof(struct dlm_cancel_join),
1808 dlm_cancel_join_handler, 2192 dlm_cancel_join_handler,
1809 NULL, NULL, &dlm_join_handlers); 2193 NULL, NULL, &dlm_join_handlers);
2194 if (status)
2195 goto bail;
2196
2197 status = o2net_register_handler(DLM_QUERY_REGION, DLM_MOD_KEY,
2198 sizeof(struct dlm_query_region),
2199 dlm_query_region_handler,
2200 NULL, NULL, &dlm_join_handlers);
1810 2201
2202 if (status)
2203 goto bail;
2204
2205 status = o2net_register_handler(DLM_QUERY_NODEINFO, DLM_MOD_KEY,
2206 sizeof(struct dlm_query_nodeinfo),
2207 dlm_query_nodeinfo_handler,
2208 NULL, NULL, &dlm_join_handlers);
1811bail: 2209bail:
1812 if (status < 0) 2210 if (status < 0)
1813 dlm_unregister_net_handlers(); 2211 dlm_unregister_net_handlers();
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index 5e02a893f46e..e8d94d722ecb 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -3635,10 +3635,18 @@ static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
3635{ 3635{
3636 struct inode *inode; 3636 struct inode *inode;
3637 struct address_space *mapping; 3637 struct address_space *mapping;
3638 struct ocfs2_inode_info *oi;
3638 3639
3639 inode = ocfs2_lock_res_inode(lockres); 3640 inode = ocfs2_lock_res_inode(lockres);
3640 mapping = inode->i_mapping; 3641 mapping = inode->i_mapping;
3641 3642
3643 if (S_ISDIR(inode->i_mode)) {
3644 oi = OCFS2_I(inode);
3645 oi->ip_dir_lock_gen++;
3646 mlog(0, "generation: %u\n", oi->ip_dir_lock_gen);
3647 goto out;
3648 }
3649
3642 if (!S_ISREG(inode->i_mode)) 3650 if (!S_ISREG(inode->i_mode))
3643 goto out; 3651 goto out;
3644 3652
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 9a03c151b5ce..9e8cc4346b76 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -64,12 +64,6 @@
64 64
65#include "buffer_head_io.h" 65#include "buffer_head_io.h"
66 66
67static int ocfs2_sync_inode(struct inode *inode)
68{
69 filemap_fdatawrite(inode->i_mapping);
70 return sync_mapping_buffers(inode->i_mapping);
71}
72
73static int ocfs2_init_file_private(struct inode *inode, struct file *file) 67static int ocfs2_init_file_private(struct inode *inode, struct file *file)
74{ 68{
75 struct ocfs2_file_private *fp; 69 struct ocfs2_file_private *fp;
@@ -180,16 +174,12 @@ static int ocfs2_sync_file(struct file *file, int datasync)
180{ 174{
181 int err = 0; 175 int err = 0;
182 journal_t *journal; 176 journal_t *journal;
183 struct dentry *dentry = file->f_path.dentry;
184 struct inode *inode = file->f_mapping->host; 177 struct inode *inode = file->f_mapping->host;
185 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 178 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
186 179
187 mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", file, dentry, datasync, 180 mlog_entry("(0x%p, %d, 0x%p, '%.*s')\n", file, datasync,
188 dentry->d_name.len, dentry->d_name.name); 181 file->f_path.dentry, file->f_path.dentry->d_name.len,
189 182 file->f_path.dentry->d_name.name);
190 err = ocfs2_sync_inode(dentry->d_inode);
191 if (err)
192 goto bail;
193 183
194 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) { 184 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) {
195 /* 185 /*
@@ -370,7 +360,7 @@ static int ocfs2_cow_file_pos(struct inode *inode,
370 if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) 360 if (!(ext_flags & OCFS2_EXT_REFCOUNTED))
371 goto out; 361 goto out;
372 362
373 return ocfs2_refcount_cow(inode, fe_bh, cpos, 1, cpos+1); 363 return ocfs2_refcount_cow(inode, NULL, fe_bh, cpos, 1, cpos+1);
374 364
375out: 365out:
376 return status; 366 return status;
@@ -913,8 +903,8 @@ static int ocfs2_zero_extend_get_range(struct inode *inode,
913 zero_clusters = last_cpos - zero_cpos; 903 zero_clusters = last_cpos - zero_cpos;
914 904
915 if (needs_cow) { 905 if (needs_cow) {
916 rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos, zero_clusters, 906 rc = ocfs2_refcount_cow(inode, NULL, di_bh, zero_cpos,
917 UINT_MAX); 907 zero_clusters, UINT_MAX);
918 if (rc) { 908 if (rc) {
919 mlog_errno(rc); 909 mlog_errno(rc);
920 goto out; 910 goto out;
@@ -2062,6 +2052,7 @@ out:
2062} 2052}
2063 2053
2064static int ocfs2_prepare_inode_for_refcount(struct inode *inode, 2054static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
2055 struct file *file,
2065 loff_t pos, size_t count, 2056 loff_t pos, size_t count,
2066 int *meta_level) 2057 int *meta_level)
2067{ 2058{
@@ -2079,7 +2070,7 @@ static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
2079 2070
2080 *meta_level = 1; 2071 *meta_level = 1;
2081 2072
2082 ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX); 2073 ret = ocfs2_refcount_cow(inode, file, di_bh, cpos, clusters, UINT_MAX);
2083 if (ret) 2074 if (ret)
2084 mlog_errno(ret); 2075 mlog_errno(ret);
2085out: 2076out:
@@ -2087,7 +2078,7 @@ out:
2087 return ret; 2078 return ret;
2088} 2079}
2089 2080
2090static int ocfs2_prepare_inode_for_write(struct dentry *dentry, 2081static int ocfs2_prepare_inode_for_write(struct file *file,
2091 loff_t *ppos, 2082 loff_t *ppos,
2092 size_t count, 2083 size_t count,
2093 int appending, 2084 int appending,
@@ -2095,6 +2086,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
2095 int *has_refcount) 2086 int *has_refcount)
2096{ 2087{
2097 int ret = 0, meta_level = 0; 2088 int ret = 0, meta_level = 0;
2089 struct dentry *dentry = file->f_path.dentry;
2098 struct inode *inode = dentry->d_inode; 2090 struct inode *inode = dentry->d_inode;
2099 loff_t saved_pos, end; 2091 loff_t saved_pos, end;
2100 2092
@@ -2150,6 +2142,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
2150 meta_level = -1; 2142 meta_level = -1;
2151 2143
2152 ret = ocfs2_prepare_inode_for_refcount(inode, 2144 ret = ocfs2_prepare_inode_for_refcount(inode,
2145 file,
2153 saved_pos, 2146 saved_pos,
2154 count, 2147 count,
2155 &meta_level); 2148 &meta_level);
@@ -2232,6 +2225,8 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
2232 struct file *file = iocb->ki_filp; 2225 struct file *file = iocb->ki_filp;
2233 struct inode *inode = file->f_path.dentry->d_inode; 2226 struct inode *inode = file->f_path.dentry->d_inode;
2234 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); 2227 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2228 int full_coherency = !(osb->s_mount_opt &
2229 OCFS2_MOUNT_COHERENCY_BUFFERED);
2235 2230
2236 mlog_entry("(0x%p, %u, '%.*s')\n", file, 2231 mlog_entry("(0x%p, %u, '%.*s')\n", file,
2237 (unsigned int)nr_segs, 2232 (unsigned int)nr_segs,
@@ -2255,16 +2250,39 @@ relock:
2255 have_alloc_sem = 1; 2250 have_alloc_sem = 1;
2256 } 2251 }
2257 2252
2258 /* concurrent O_DIRECT writes are allowed */ 2253 /*
2259 rw_level = !direct_io; 2254 * Concurrent O_DIRECT writes are allowed with
2255 * mount_option "coherency=buffered".
2256 */
2257 rw_level = (!direct_io || full_coherency);
2258
2260 ret = ocfs2_rw_lock(inode, rw_level); 2259 ret = ocfs2_rw_lock(inode, rw_level);
2261 if (ret < 0) { 2260 if (ret < 0) {
2262 mlog_errno(ret); 2261 mlog_errno(ret);
2263 goto out_sems; 2262 goto out_sems;
2264 } 2263 }
2265 2264
2265 /*
2266 * O_DIRECT writes with "coherency=full" need to take EX cluster
2267 * inode_lock to guarantee coherency.
2268 */
2269 if (direct_io && full_coherency) {
2270 /*
2271 * We need to take and drop the inode lock to force
2272 * other nodes to drop their caches. Buffered I/O
2273 * already does this in write_begin().
2274 */
2275 ret = ocfs2_inode_lock(inode, NULL, 1);
2276 if (ret < 0) {
2277 mlog_errno(ret);
2278 goto out_sems;
2279 }
2280
2281 ocfs2_inode_unlock(inode, 1);
2282 }
2283
2266 can_do_direct = direct_io; 2284 can_do_direct = direct_io;
2267 ret = ocfs2_prepare_inode_for_write(file->f_path.dentry, ppos, 2285 ret = ocfs2_prepare_inode_for_write(file, ppos,
2268 iocb->ki_left, appending, 2286 iocb->ki_left, appending,
2269 &can_do_direct, &has_refcount); 2287 &can_do_direct, &has_refcount);
2270 if (ret < 0) { 2288 if (ret < 0) {
@@ -2312,17 +2330,6 @@ relock:
2312 written = generic_file_direct_write(iocb, iov, &nr_segs, *ppos, 2330 written = generic_file_direct_write(iocb, iov, &nr_segs, *ppos,
2313 ppos, count, ocount); 2331 ppos, count, ocount);
2314 if (written < 0) { 2332 if (written < 0) {
2315 /*
2316 * direct write may have instantiated a few
2317 * blocks outside i_size. Trim these off again.
2318 * Don't need i_size_read because we hold i_mutex.
2319 *
2320 * XXX(truncate): this looks buggy because ocfs2 did not
2321 * actually implement ->truncate. Take a look at
2322 * the new truncate sequence and update this accordingly
2323 */
2324 if (*ppos + count > inode->i_size)
2325 truncate_setsize(inode, inode->i_size);
2326 ret = written; 2333 ret = written;
2327 goto out_dio; 2334 goto out_dio;
2328 } 2335 }
@@ -2394,7 +2401,7 @@ static int ocfs2_splice_to_file(struct pipe_inode_info *pipe,
2394{ 2401{
2395 int ret; 2402 int ret;
2396 2403
2397 ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, &sd->pos, 2404 ret = ocfs2_prepare_inode_for_write(out, &sd->pos,
2398 sd->total_len, 0, NULL, NULL); 2405 sd->total_len, 0, NULL, NULL);
2399 if (ret < 0) { 2406 if (ret < 0) {
2400 mlog_errno(ret); 2407 mlog_errno(ret);
diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
index eece3e05d9d0..f935fd6600dd 100644
--- a/fs/ocfs2/inode.c
+++ b/fs/ocfs2/inode.c
@@ -335,6 +335,7 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
335 else 335 else
336 inode->i_fop = &ocfs2_dops_no_plocks; 336 inode->i_fop = &ocfs2_dops_no_plocks;
337 i_size_write(inode, le64_to_cpu(fe->i_size)); 337 i_size_write(inode, le64_to_cpu(fe->i_size));
338 OCFS2_I(inode)->ip_dir_lock_gen = 1;
338 break; 339 break;
339 case S_IFLNK: 340 case S_IFLNK:
340 if (ocfs2_inode_is_fast_symlink(inode)) 341 if (ocfs2_inode_is_fast_symlink(inode))
diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h
index 6de5a869db30..1c508b149b3a 100644
--- a/fs/ocfs2/inode.h
+++ b/fs/ocfs2/inode.h
@@ -46,30 +46,28 @@ struct ocfs2_inode_info
46 /* These fields are protected by ip_lock */ 46 /* These fields are protected by ip_lock */
47 spinlock_t ip_lock; 47 spinlock_t ip_lock;
48 u32 ip_open_count; 48 u32 ip_open_count;
49 u32 ip_clusters;
50 struct list_head ip_io_markers; 49 struct list_head ip_io_markers;
50 u32 ip_clusters;
51 51
52 u16 ip_dyn_features;
52 struct mutex ip_io_mutex; 53 struct mutex ip_io_mutex;
53
54 u32 ip_flags; /* see below */ 54 u32 ip_flags; /* see below */
55 u32 ip_attr; /* inode attributes */ 55 u32 ip_attr; /* inode attributes */
56 u16 ip_dyn_features;
57 56
58 /* protected by recovery_lock. */ 57 /* protected by recovery_lock. */
59 struct inode *ip_next_orphan; 58 struct inode *ip_next_orphan;
60 59
61 u32 ip_dir_start_lookup;
62
63 struct ocfs2_caching_info ip_metadata_cache; 60 struct ocfs2_caching_info ip_metadata_cache;
64
65 struct ocfs2_extent_map ip_extent_map; 61 struct ocfs2_extent_map ip_extent_map;
66
67 struct inode vfs_inode; 62 struct inode vfs_inode;
68 struct jbd2_inode ip_jinode; 63 struct jbd2_inode ip_jinode;
69 64
65 u32 ip_dir_start_lookup;
66
70 /* Only valid if the inode is the dir. */ 67 /* Only valid if the inode is the dir. */
71 u32 ip_last_used_slot; 68 u32 ip_last_used_slot;
72 u64 ip_last_used_group; 69 u64 ip_last_used_group;
70 u32 ip_dir_lock_gen;
73 71
74 struct ocfs2_alloc_reservation ip_la_data_resv; 72 struct ocfs2_alloc_reservation ip_la_data_resv;
75}; 73};
diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
index 7d9d9c132cef..7a4868196152 100644
--- a/fs/ocfs2/ioctl.c
+++ b/fs/ocfs2/ioctl.c
@@ -26,6 +26,26 @@
26 26
27#include <linux/ext2_fs.h> 27#include <linux/ext2_fs.h>
28 28
29#define o2info_from_user(a, b) \
30 copy_from_user(&(a), (b), sizeof(a))
31#define o2info_to_user(a, b) \
32 copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
33
34/*
35 * This call is void because we are already reporting an error that may
36 * be -EFAULT. The error will be returned from the ioctl(2) call. It's
37 * just a best-effort to tell userspace that this request caused the error.
38 */
39static inline void __o2info_set_request_error(struct ocfs2_info_request *kreq,
40 struct ocfs2_info_request __user *req)
41{
42 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
43 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
44}
45
46#define o2info_set_request_error(a, b) \
47 __o2info_set_request_error((struct ocfs2_info_request *)&(a), b)
48
29static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags) 49static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
30{ 50{
31 int status; 51 int status;
@@ -109,6 +129,328 @@ bail:
109 return status; 129 return status;
110} 130}
111 131
132int ocfs2_info_handle_blocksize(struct inode *inode,
133 struct ocfs2_info_request __user *req)
134{
135 int status = -EFAULT;
136 struct ocfs2_info_blocksize oib;
137
138 if (o2info_from_user(oib, req))
139 goto bail;
140
141 oib.ib_blocksize = inode->i_sb->s_blocksize;
142 oib.ib_req.ir_flags |= OCFS2_INFO_FL_FILLED;
143
144 if (o2info_to_user(oib, req))
145 goto bail;
146
147 status = 0;
148bail:
149 if (status)
150 o2info_set_request_error(oib, req);
151
152 return status;
153}
154
155int ocfs2_info_handle_clustersize(struct inode *inode,
156 struct ocfs2_info_request __user *req)
157{
158 int status = -EFAULT;
159 struct ocfs2_info_clustersize oic;
160 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
161
162 if (o2info_from_user(oic, req))
163 goto bail;
164
165 oic.ic_clustersize = osb->s_clustersize;
166 oic.ic_req.ir_flags |= OCFS2_INFO_FL_FILLED;
167
168 if (o2info_to_user(oic, req))
169 goto bail;
170
171 status = 0;
172bail:
173 if (status)
174 o2info_set_request_error(oic, req);
175
176 return status;
177}
178
179int ocfs2_info_handle_maxslots(struct inode *inode,
180 struct ocfs2_info_request __user *req)
181{
182 int status = -EFAULT;
183 struct ocfs2_info_maxslots oim;
184 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
185
186 if (o2info_from_user(oim, req))
187 goto bail;
188
189 oim.im_max_slots = osb->max_slots;
190 oim.im_req.ir_flags |= OCFS2_INFO_FL_FILLED;
191
192 if (o2info_to_user(oim, req))
193 goto bail;
194
195 status = 0;
196bail:
197 if (status)
198 o2info_set_request_error(oim, req);
199
200 return status;
201}
202
203int ocfs2_info_handle_label(struct inode *inode,
204 struct ocfs2_info_request __user *req)
205{
206 int status = -EFAULT;
207 struct ocfs2_info_label oil;
208 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
209
210 if (o2info_from_user(oil, req))
211 goto bail;
212
213 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
214 oil.il_req.ir_flags |= OCFS2_INFO_FL_FILLED;
215
216 if (o2info_to_user(oil, req))
217 goto bail;
218
219 status = 0;
220bail:
221 if (status)
222 o2info_set_request_error(oil, req);
223
224 return status;
225}
226
227int ocfs2_info_handle_uuid(struct inode *inode,
228 struct ocfs2_info_request __user *req)
229{
230 int status = -EFAULT;
231 struct ocfs2_info_uuid oiu;
232 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
233
234 if (o2info_from_user(oiu, req))
235 goto bail;
236
237 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
238 oiu.iu_req.ir_flags |= OCFS2_INFO_FL_FILLED;
239
240 if (o2info_to_user(oiu, req))
241 goto bail;
242
243 status = 0;
244bail:
245 if (status)
246 o2info_set_request_error(oiu, req);
247
248 return status;
249}
250
251int ocfs2_info_handle_fs_features(struct inode *inode,
252 struct ocfs2_info_request __user *req)
253{
254 int status = -EFAULT;
255 struct ocfs2_info_fs_features oif;
256 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
257
258 if (o2info_from_user(oif, req))
259 goto bail;
260
261 oif.if_compat_features = osb->s_feature_compat;
262 oif.if_incompat_features = osb->s_feature_incompat;
263 oif.if_ro_compat_features = osb->s_feature_ro_compat;
264 oif.if_req.ir_flags |= OCFS2_INFO_FL_FILLED;
265
266 if (o2info_to_user(oif, req))
267 goto bail;
268
269 status = 0;
270bail:
271 if (status)
272 o2info_set_request_error(oif, req);
273
274 return status;
275}
276
277int ocfs2_info_handle_journal_size(struct inode *inode,
278 struct ocfs2_info_request __user *req)
279{
280 int status = -EFAULT;
281 struct ocfs2_info_journal_size oij;
282 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
283
284 if (o2info_from_user(oij, req))
285 goto bail;
286
287 oij.ij_journal_size = osb->journal->j_inode->i_size;
288
289 oij.ij_req.ir_flags |= OCFS2_INFO_FL_FILLED;
290
291 if (o2info_to_user(oij, req))
292 goto bail;
293
294 status = 0;
295bail:
296 if (status)
297 o2info_set_request_error(oij, req);
298
299 return status;
300}
301
302int ocfs2_info_handle_unknown(struct inode *inode,
303 struct ocfs2_info_request __user *req)
304{
305 int status = -EFAULT;
306 struct ocfs2_info_request oir;
307
308 if (o2info_from_user(oir, req))
309 goto bail;
310
311 oir.ir_flags &= ~OCFS2_INFO_FL_FILLED;
312
313 if (o2info_to_user(oir, req))
314 goto bail;
315
316 status = 0;
317bail:
318 if (status)
319 o2info_set_request_error(oir, req);
320
321 return status;
322}
323
324/*
325 * Validate and distinguish OCFS2_IOC_INFO requests.
326 *
327 * - validate the magic number.
328 * - distinguish different requests.
329 * - validate size of different requests.
330 */
331int ocfs2_info_handle_request(struct inode *inode,
332 struct ocfs2_info_request __user *req)
333{
334 int status = -EFAULT;
335 struct ocfs2_info_request oir;
336
337 if (o2info_from_user(oir, req))
338 goto bail;
339
340 status = -EINVAL;
341 if (oir.ir_magic != OCFS2_INFO_MAGIC)
342 goto bail;
343
344 switch (oir.ir_code) {
345 case OCFS2_INFO_BLOCKSIZE:
346 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
347 status = ocfs2_info_handle_blocksize(inode, req);
348 break;
349 case OCFS2_INFO_CLUSTERSIZE:
350 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
351 status = ocfs2_info_handle_clustersize(inode, req);
352 break;
353 case OCFS2_INFO_MAXSLOTS:
354 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
355 status = ocfs2_info_handle_maxslots(inode, req);
356 break;
357 case OCFS2_INFO_LABEL:
358 if (oir.ir_size == sizeof(struct ocfs2_info_label))
359 status = ocfs2_info_handle_label(inode, req);
360 break;
361 case OCFS2_INFO_UUID:
362 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
363 status = ocfs2_info_handle_uuid(inode, req);
364 break;
365 case OCFS2_INFO_FS_FEATURES:
366 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
367 status = ocfs2_info_handle_fs_features(inode, req);
368 break;
369 case OCFS2_INFO_JOURNAL_SIZE:
370 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
371 status = ocfs2_info_handle_journal_size(inode, req);
372 break;
373 default:
374 status = ocfs2_info_handle_unknown(inode, req);
375 break;
376 }
377
378bail:
379 return status;
380}
381
382int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
383 u64 *req_addr, int compat_flag)
384{
385 int status = -EFAULT;
386 u64 __user *bp = NULL;
387
388 if (compat_flag) {
389#ifdef CONFIG_COMPAT
390 /*
391 * pointer bp stores the base address of a pointers array,
392 * which collects all addresses of separate request.
393 */
394 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
395#else
396 BUG();
397#endif
398 } else
399 bp = (u64 __user *)(unsigned long)(info->oi_requests);
400
401 if (o2info_from_user(*req_addr, bp + idx))
402 goto bail;
403
404 status = 0;
405bail:
406 return status;
407}
408
409/*
410 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
411 *
412 * ocfs2_info_handle() recevies a large info aggregation, grab and
413 * validate the request count from header, then break it into small
414 * pieces, later specific handlers can handle them one by one.
415 *
416 * Idea here is to make each separate request small enough to ensure
417 * a better backward&forward compatibility, since a small piece of
418 * request will be less likely to be broken if disk layout get changed.
419 */
420int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
421 int compat_flag)
422{
423 int i, status = 0;
424 u64 req_addr;
425 struct ocfs2_info_request __user *reqp;
426
427 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
428 (!info->oi_requests)) {
429 status = -EINVAL;
430 goto bail;
431 }
432
433 for (i = 0; i < info->oi_count; i++) {
434
435 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
436 if (status)
437 break;
438
439 reqp = (struct ocfs2_info_request *)(unsigned long)req_addr;
440 if (!reqp) {
441 status = -EINVAL;
442 goto bail;
443 }
444
445 status = ocfs2_info_handle_request(inode, reqp);
446 if (status)
447 break;
448 }
449
450bail:
451 return status;
452}
453
112long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 454long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
113{ 455{
114 struct inode *inode = filp->f_path.dentry->d_inode; 456 struct inode *inode = filp->f_path.dentry->d_inode;
@@ -120,6 +462,7 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
120 struct reflink_arguments args; 462 struct reflink_arguments args;
121 const char *old_path, *new_path; 463 const char *old_path, *new_path;
122 bool preserve; 464 bool preserve;
465 struct ocfs2_info info;
123 466
124 switch (cmd) { 467 switch (cmd) {
125 case OCFS2_IOC_GETFLAGS: 468 case OCFS2_IOC_GETFLAGS:
@@ -174,6 +517,12 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
174 preserve = (args.preserve != 0); 517 preserve = (args.preserve != 0);
175 518
176 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve); 519 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
520 case OCFS2_IOC_INFO:
521 if (copy_from_user(&info, (struct ocfs2_info __user *)arg,
522 sizeof(struct ocfs2_info)))
523 return -EFAULT;
524
525 return ocfs2_info_handle(inode, &info, 0);
177 default: 526 default:
178 return -ENOTTY; 527 return -ENOTTY;
179 } 528 }
@@ -185,6 +534,7 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
185 bool preserve; 534 bool preserve;
186 struct reflink_arguments args; 535 struct reflink_arguments args;
187 struct inode *inode = file->f_path.dentry->d_inode; 536 struct inode *inode = file->f_path.dentry->d_inode;
537 struct ocfs2_info info;
188 538
189 switch (cmd) { 539 switch (cmd) {
190 case OCFS2_IOC32_GETFLAGS: 540 case OCFS2_IOC32_GETFLAGS:
@@ -209,6 +559,12 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
209 559
210 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path), 560 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
211 compat_ptr(args.new_path), preserve); 561 compat_ptr(args.new_path), preserve);
562 case OCFS2_IOC_INFO:
563 if (copy_from_user(&info, (struct ocfs2_info __user *)arg,
564 sizeof(struct ocfs2_info)))
565 return -EFAULT;
566
567 return ocfs2_info_handle(inode, &info, 1);
212 default: 568 default:
213 return -ENOIOCTLCMD; 569 return -ENOIOCTLCMD;
214 } 570 }
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index 9b57c0350ff9..faa2303dbf0a 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -301,7 +301,6 @@ static int ocfs2_commit_cache(struct ocfs2_super *osb)
301{ 301{
302 int status = 0; 302 int status = 0;
303 unsigned int flushed; 303 unsigned int flushed;
304 unsigned long old_id;
305 struct ocfs2_journal *journal = NULL; 304 struct ocfs2_journal *journal = NULL;
306 305
307 mlog_entry_void(); 306 mlog_entry_void();
@@ -326,7 +325,7 @@ static int ocfs2_commit_cache(struct ocfs2_super *osb)
326 goto finally; 325 goto finally;
327 } 326 }
328 327
329 old_id = ocfs2_inc_trans_id(journal); 328 ocfs2_inc_trans_id(journal);
330 329
331 flushed = atomic_read(&journal->j_num_trans); 330 flushed = atomic_read(&journal->j_num_trans);
332 atomic_set(&journal->j_num_trans, 0); 331 atomic_set(&journal->j_num_trans, 0);
@@ -342,9 +341,6 @@ finally:
342 return status; 341 return status;
343} 342}
344 343
345/* pass it NULL and it will allocate a new handle object for you. If
346 * you pass it a handle however, it may still return error, in which
347 * case it has free'd the passed handle for you. */
348handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int max_buffs) 344handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int max_buffs)
349{ 345{
350 journal_t *journal = osb->journal->j_journal; 346 journal_t *journal = osb->journal->j_journal;
@@ -1888,6 +1884,8 @@ void ocfs2_queue_orphan_scan(struct ocfs2_super *osb)
1888 1884
1889 os = &osb->osb_orphan_scan; 1885 os = &osb->osb_orphan_scan;
1890 1886
1887 mlog(0, "Begin orphan scan\n");
1888
1891 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) 1889 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
1892 goto out; 1890 goto out;
1893 1891
@@ -1920,6 +1918,7 @@ void ocfs2_queue_orphan_scan(struct ocfs2_super *osb)
1920unlock: 1918unlock:
1921 ocfs2_orphan_scan_unlock(osb, seqno); 1919 ocfs2_orphan_scan_unlock(osb, seqno);
1922out: 1920out:
1921 mlog(0, "Orphan scan completed\n");
1923 return; 1922 return;
1924} 1923}
1925 1924
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index b5baaa8e710f..43e56b97f9c0 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -67,11 +67,12 @@ struct ocfs2_journal {
67 struct buffer_head *j_bh; /* Journal disk inode block */ 67 struct buffer_head *j_bh; /* Journal disk inode block */
68 atomic_t j_num_trans; /* Number of transactions 68 atomic_t j_num_trans; /* Number of transactions
69 * currently in the system. */ 69 * currently in the system. */
70 spinlock_t j_lock;
70 unsigned long j_trans_id; 71 unsigned long j_trans_id;
71 struct rw_semaphore j_trans_barrier; 72 struct rw_semaphore j_trans_barrier;
72 wait_queue_head_t j_checkpointed; 73 wait_queue_head_t j_checkpointed;
73 74
74 spinlock_t j_lock; 75 /* both fields protected by j_lock*/
75 struct list_head j_la_cleanups; 76 struct list_head j_la_cleanups;
76 struct work_struct j_recovery_work; 77 struct work_struct j_recovery_work;
77}; 78};
diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c
index 4c18f4ad93b4..7e32db9c2c99 100644
--- a/fs/ocfs2/mmap.c
+++ b/fs/ocfs2/mmap.c
@@ -59,10 +59,11 @@ static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf)
59 return ret; 59 return ret;
60} 60}
61 61
62static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, 62static int __ocfs2_page_mkwrite(struct file *file, struct buffer_head *di_bh,
63 struct page *page) 63 struct page *page)
64{ 64{
65 int ret; 65 int ret;
66 struct inode *inode = file->f_path.dentry->d_inode;
66 struct address_space *mapping = inode->i_mapping; 67 struct address_space *mapping = inode->i_mapping;
67 loff_t pos = page_offset(page); 68 loff_t pos = page_offset(page);
68 unsigned int len = PAGE_CACHE_SIZE; 69 unsigned int len = PAGE_CACHE_SIZE;
@@ -111,7 +112,7 @@ static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh,
111 if (page->index == last_index) 112 if (page->index == last_index)
112 len = ((size - 1) & ~PAGE_CACHE_MASK) + 1; 113 len = ((size - 1) & ~PAGE_CACHE_MASK) + 1;
113 114
114 ret = ocfs2_write_begin_nolock(mapping, pos, len, 0, &locked_page, 115 ret = ocfs2_write_begin_nolock(file, mapping, pos, len, 0, &locked_page,
115 &fsdata, di_bh, page); 116 &fsdata, di_bh, page);
116 if (ret) { 117 if (ret) {
117 if (ret != -ENOSPC) 118 if (ret != -ENOSPC)
@@ -159,7 +160,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
159 */ 160 */
160 down_write(&OCFS2_I(inode)->ip_alloc_sem); 161 down_write(&OCFS2_I(inode)->ip_alloc_sem);
161 162
162 ret = __ocfs2_page_mkwrite(inode, di_bh, page); 163 ret = __ocfs2_page_mkwrite(vma->vm_file, di_bh, page);
163 164
164 up_write(&OCFS2_I(inode)->ip_alloc_sem); 165 up_write(&OCFS2_I(inode)->ip_alloc_sem);
165 166
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index a00dda2e4f16..e7bde21149ae 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -171,7 +171,8 @@ bail_add:
171 ret = ERR_PTR(status); 171 ret = ERR_PTR(status);
172 goto bail_unlock; 172 goto bail_unlock;
173 } 173 }
174 } 174 } else
175 ocfs2_dentry_attach_gen(dentry);
175 176
176bail_unlock: 177bail_unlock:
177 /* Don't drop the cluster lock until *after* the d_add -- 178 /* Don't drop the cluster lock until *after* the d_add --
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index c67003b6b5a2..d8408217e3bd 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -150,26 +150,33 @@ typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
150struct ocfs2_lock_res { 150struct ocfs2_lock_res {
151 void *l_priv; 151 void *l_priv;
152 struct ocfs2_lock_res_ops *l_ops; 152 struct ocfs2_lock_res_ops *l_ops;
153 spinlock_t l_lock; 153
154 154
155 struct list_head l_blocked_list; 155 struct list_head l_blocked_list;
156 struct list_head l_mask_waiters; 156 struct list_head l_mask_waiters;
157 157
158 enum ocfs2_lock_type l_type;
159 unsigned long l_flags; 158 unsigned long l_flags;
160 char l_name[OCFS2_LOCK_ID_MAX_LEN]; 159 char l_name[OCFS2_LOCK_ID_MAX_LEN];
161 int l_level;
162 unsigned int l_ro_holders; 160 unsigned int l_ro_holders;
163 unsigned int l_ex_holders; 161 unsigned int l_ex_holders;
164 struct ocfs2_dlm_lksb l_lksb; 162 unsigned char l_level;
163
164 /* Data packed - type enum ocfs2_lock_type */
165 unsigned char l_type;
165 166
166 /* used from AST/BAST funcs. */ 167 /* used from AST/BAST funcs. */
167 enum ocfs2_ast_action l_action; 168 /* Data packed - enum type ocfs2_ast_action */
168 enum ocfs2_unlock_action l_unlock_action; 169 unsigned char l_action;
169 int l_requested; 170 /* Data packed - enum type ocfs2_unlock_action */
170 int l_blocking; 171 unsigned char l_unlock_action;
172 unsigned char l_requested;
173 unsigned char l_blocking;
171 unsigned int l_pending_gen; 174 unsigned int l_pending_gen;
172 175
176 spinlock_t l_lock;
177
178 struct ocfs2_dlm_lksb l_lksb;
179
173 wait_queue_head_t l_event; 180 wait_queue_head_t l_event;
174 181
175 struct list_head l_debug_list; 182 struct list_head l_debug_list;
@@ -243,7 +250,7 @@ enum ocfs2_local_alloc_state
243 250
244enum ocfs2_mount_options 251enum ocfs2_mount_options
245{ 252{
246 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ 253 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Local heartbeat */
247 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ 254 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
248 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ 255 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
249 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ 256 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
@@ -256,6 +263,10 @@ enum ocfs2_mount_options
256 control lists */ 263 control lists */
257 OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */ 264 OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */
258 OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */ 265 OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */
266 OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12, /* Allow concurrent O_DIRECT
267 writes */
268 OCFS2_MOUNT_HB_NONE = 1 << 13, /* No heartbeat */
269 OCFS2_MOUNT_HB_GLOBAL = 1 << 14, /* Global heartbeat */
259}; 270};
260 271
261#define OCFS2_OSB_SOFT_RO 0x0001 272#define OCFS2_OSB_SOFT_RO 0x0001
@@ -277,7 +288,8 @@ struct ocfs2_super
277 struct super_block *sb; 288 struct super_block *sb;
278 struct inode *root_inode; 289 struct inode *root_inode;
279 struct inode *sys_root_inode; 290 struct inode *sys_root_inode;
280 struct inode *system_inodes[NUM_SYSTEM_INODES]; 291 struct inode *global_system_inodes[NUM_GLOBAL_SYSTEM_INODES];
292 struct inode **local_system_inodes;
281 293
282 struct ocfs2_slot_info *slot_info; 294 struct ocfs2_slot_info *slot_info;
283 295
@@ -368,6 +380,8 @@ struct ocfs2_super
368 struct ocfs2_alloc_stats alloc_stats; 380 struct ocfs2_alloc_stats alloc_stats;
369 char dev_str[20]; /* "major,minor" of the device */ 381 char dev_str[20]; /* "major,minor" of the device */
370 382
383 u8 osb_stackflags;
384
371 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 385 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
372 struct ocfs2_cluster_connection *cconn; 386 struct ocfs2_cluster_connection *cconn;
373 struct ocfs2_lock_res osb_super_lockres; 387 struct ocfs2_lock_res osb_super_lockres;
@@ -601,10 +615,35 @@ static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
601 return ret; 615 return ret;
602} 616}
603 617
604static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) 618static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb)
605{ 619{
606 return (osb->s_feature_incompat & 620 return (osb->s_feature_incompat &
607 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK); 621 (OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK |
622 OCFS2_FEATURE_INCOMPAT_CLUSTERINFO));
623}
624
625static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
626{
627 if (ocfs2_clusterinfo_valid(osb) &&
628 memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
629 OCFS2_STACK_LABEL_LEN))
630 return 1;
631 return 0;
632}
633
634static inline int ocfs2_o2cb_stack(struct ocfs2_super *osb)
635{
636 if (ocfs2_clusterinfo_valid(osb) &&
637 !memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
638 OCFS2_STACK_LABEL_LEN))
639 return 1;
640 return 0;
641}
642
643static inline int ocfs2_cluster_o2cb_global_heartbeat(struct ocfs2_super *osb)
644{
645 return ocfs2_o2cb_stack(osb) &&
646 (osb->osb_stackflags & OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT);
608} 647}
609 648
610static inline int ocfs2_mount_local(struct ocfs2_super *osb) 649static inline int ocfs2_mount_local(struct ocfs2_super *osb)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index fa31d05e41b7..c2e4f8222e2f 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -101,7 +101,8 @@
101 | OCFS2_FEATURE_INCOMPAT_META_ECC \ 101 | OCFS2_FEATURE_INCOMPAT_META_ECC \
102 | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \ 102 | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
103 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \ 103 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \
104 | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG) 104 | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG \
105 | OCFS2_FEATURE_INCOMPAT_CLUSTERINFO)
105#define OCFS2_FEATURE_RO_COMPAT_SUPP (OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \ 106#define OCFS2_FEATURE_RO_COMPAT_SUPP (OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
106 | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \ 107 | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
107 | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) 108 | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
@@ -170,6 +171,13 @@
170#define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG 0x2000 171#define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG 0x2000
171 172
172/* 173/*
174 * Incompat bit to indicate useable clusterinfo with stackflags for all
175 * cluster stacks (userspace adnd o2cb). If this bit is set,
176 * INCOMPAT_USERSPACE_STACK becomes superfluous and thus should not be set.
177 */
178#define OCFS2_FEATURE_INCOMPAT_CLUSTERINFO 0x4000
179
180/*
173 * backup superblock flag is used to indicate that this volume 181 * backup superblock flag is used to indicate that this volume
174 * has backup superblocks. 182 * has backup superblocks.
175 */ 183 */
@@ -292,10 +300,13 @@
292#define OCFS2_VOL_UUID_LEN 16 300#define OCFS2_VOL_UUID_LEN 16
293#define OCFS2_MAX_VOL_LABEL_LEN 64 301#define OCFS2_MAX_VOL_LABEL_LEN 64
294 302
295/* The alternate, userspace stack fields */ 303/* The cluster stack fields */
296#define OCFS2_STACK_LABEL_LEN 4 304#define OCFS2_STACK_LABEL_LEN 4
297#define OCFS2_CLUSTER_NAME_LEN 16 305#define OCFS2_CLUSTER_NAME_LEN 16
298 306
307/* Classic (historically speaking) cluster stack */
308#define OCFS2_CLASSIC_CLUSTER_STACK "o2cb"
309
299/* Journal limits (in bytes) */ 310/* Journal limits (in bytes) */
300#define OCFS2_MIN_JOURNAL_SIZE (4 * 1024 * 1024) 311#define OCFS2_MIN_JOURNAL_SIZE (4 * 1024 * 1024)
301 312
@@ -305,6 +316,11 @@
305 */ 316 */
306#define OCFS2_MIN_XATTR_INLINE_SIZE 256 317#define OCFS2_MIN_XATTR_INLINE_SIZE 256
307 318
319/*
320 * Cluster info flags (ocfs2_cluster_info.ci_stackflags)
321 */
322#define OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT (0x01)
323
308struct ocfs2_system_inode_info { 324struct ocfs2_system_inode_info {
309 char *si_name; 325 char *si_name;
310 int si_iflags; 326 int si_iflags;
@@ -322,6 +338,7 @@ enum {
322 USER_QUOTA_SYSTEM_INODE, 338 USER_QUOTA_SYSTEM_INODE,
323 GROUP_QUOTA_SYSTEM_INODE, 339 GROUP_QUOTA_SYSTEM_INODE,
324#define OCFS2_LAST_GLOBAL_SYSTEM_INODE GROUP_QUOTA_SYSTEM_INODE 340#define OCFS2_LAST_GLOBAL_SYSTEM_INODE GROUP_QUOTA_SYSTEM_INODE
341#define OCFS2_FIRST_LOCAL_SYSTEM_INODE ORPHAN_DIR_SYSTEM_INODE
325 ORPHAN_DIR_SYSTEM_INODE, 342 ORPHAN_DIR_SYSTEM_INODE,
326 EXTENT_ALLOC_SYSTEM_INODE, 343 EXTENT_ALLOC_SYSTEM_INODE,
327 INODE_ALLOC_SYSTEM_INODE, 344 INODE_ALLOC_SYSTEM_INODE,
@@ -330,8 +347,12 @@ enum {
330 TRUNCATE_LOG_SYSTEM_INODE, 347 TRUNCATE_LOG_SYSTEM_INODE,
331 LOCAL_USER_QUOTA_SYSTEM_INODE, 348 LOCAL_USER_QUOTA_SYSTEM_INODE,
332 LOCAL_GROUP_QUOTA_SYSTEM_INODE, 349 LOCAL_GROUP_QUOTA_SYSTEM_INODE,
350#define OCFS2_LAST_LOCAL_SYSTEM_INODE LOCAL_GROUP_QUOTA_SYSTEM_INODE
333 NUM_SYSTEM_INODES 351 NUM_SYSTEM_INODES
334}; 352};
353#define NUM_GLOBAL_SYSTEM_INODES OCFS2_LAST_GLOBAL_SYSTEM_INODE
354#define NUM_LOCAL_SYSTEM_INODES \
355 (NUM_SYSTEM_INODES - OCFS2_FIRST_LOCAL_SYSTEM_INODE)
335 356
336static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = { 357static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
337 /* Global system inodes (single copy) */ 358 /* Global system inodes (single copy) */
@@ -360,6 +381,7 @@ static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
360/* Parameter passed from mount.ocfs2 to module */ 381/* Parameter passed from mount.ocfs2 to module */
361#define OCFS2_HB_NONE "heartbeat=none" 382#define OCFS2_HB_NONE "heartbeat=none"
362#define OCFS2_HB_LOCAL "heartbeat=local" 383#define OCFS2_HB_LOCAL "heartbeat=local"
384#define OCFS2_HB_GLOBAL "heartbeat=global"
363 385
364/* 386/*
365 * OCFS2 directory file types. Only the low 3 bits are used. The 387 * OCFS2 directory file types. Only the low 3 bits are used. The
@@ -566,9 +588,21 @@ struct ocfs2_slot_map_extended {
566 */ 588 */
567}; 589};
568 590
591/*
592 * ci_stackflags is only valid if the incompat bit
593 * OCFS2_FEATURE_INCOMPAT_CLUSTERINFO is set.
594 */
569struct ocfs2_cluster_info { 595struct ocfs2_cluster_info {
570/*00*/ __u8 ci_stack[OCFS2_STACK_LABEL_LEN]; 596/*00*/ __u8 ci_stack[OCFS2_STACK_LABEL_LEN];
571 __le32 ci_reserved; 597 union {
598 __le32 ci_reserved;
599 struct {
600 __u8 ci_stackflags;
601 __u8 ci_reserved1;
602 __u8 ci_reserved2;
603 __u8 ci_reserved3;
604 };
605 };
572/*08*/ __u8 ci_cluster[OCFS2_CLUSTER_NAME_LEN]; 606/*08*/ __u8 ci_cluster[OCFS2_CLUSTER_NAME_LEN];
573/*18*/ 607/*18*/
574}; 608};
@@ -605,9 +639,9 @@ struct ocfs2_super_block {
605 * group header */ 639 * group header */
606/*50*/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */ 640/*50*/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */
607/*90*/ __u8 s_uuid[OCFS2_VOL_UUID_LEN]; /* 128-bit uuid */ 641/*90*/ __u8 s_uuid[OCFS2_VOL_UUID_LEN]; /* 128-bit uuid */
608/*A0*/ struct ocfs2_cluster_info s_cluster_info; /* Selected userspace 642/*A0*/ struct ocfs2_cluster_info s_cluster_info; /* Only valid if either
609 stack. Only valid 643 userspace or clusterinfo
610 with INCOMPAT flag. */ 644 INCOMPAT flag set. */
611/*B8*/ __le16 s_xattr_inline_size; /* extended attribute inline size 645/*B8*/ __le16 s_xattr_inline_size; /* extended attribute inline size
612 for this fs*/ 646 for this fs*/
613 __le16 s_reserved0; 647 __le16 s_reserved0;
diff --git a/fs/ocfs2/ocfs2_ioctl.h b/fs/ocfs2/ocfs2_ioctl.h
index 5d241505690b..b46f39bf7438 100644
--- a/fs/ocfs2/ocfs2_ioctl.h
+++ b/fs/ocfs2/ocfs2_ioctl.h
@@ -76,4 +76,99 @@ struct reflink_arguments {
76}; 76};
77#define OCFS2_IOC_REFLINK _IOW('o', 4, struct reflink_arguments) 77#define OCFS2_IOC_REFLINK _IOW('o', 4, struct reflink_arguments)
78 78
79/* Following definitions dedicated for ocfs2_info_request ioctls. */
80#define OCFS2_INFO_MAX_REQUEST (50)
81#define OCFS2_TEXT_UUID_LEN (OCFS2_VOL_UUID_LEN * 2)
82
83/* Magic number of all requests */
84#define OCFS2_INFO_MAGIC (0x4F32494E)
85
86/*
87 * Always try to separate info request into small pieces to
88 * guarantee the backward&forward compatibility.
89 */
90struct ocfs2_info {
91 __u64 oi_requests; /* Array of __u64 pointers to requests */
92 __u32 oi_count; /* Number of requests in info_requests */
93 __u32 oi_pad;
94};
95
96struct ocfs2_info_request {
97/*00*/ __u32 ir_magic; /* Magic number */
98 __u32 ir_code; /* Info request code */
99 __u32 ir_size; /* Size of request */
100 __u32 ir_flags; /* Request flags */
101/*10*/ /* Request specific fields */
102};
103
104struct ocfs2_info_clustersize {
105 struct ocfs2_info_request ic_req;
106 __u32 ic_clustersize;
107 __u32 ic_pad;
108};
109
110struct ocfs2_info_blocksize {
111 struct ocfs2_info_request ib_req;
112 __u32 ib_blocksize;
113 __u32 ib_pad;
114};
115
116struct ocfs2_info_maxslots {
117 struct ocfs2_info_request im_req;
118 __u32 im_max_slots;
119 __u32 im_pad;
120};
121
122struct ocfs2_info_label {
123 struct ocfs2_info_request il_req;
124 __u8 il_label[OCFS2_MAX_VOL_LABEL_LEN];
125} __attribute__ ((packed));
126
127struct ocfs2_info_uuid {
128 struct ocfs2_info_request iu_req;
129 __u8 iu_uuid_str[OCFS2_TEXT_UUID_LEN + 1];
130} __attribute__ ((packed));
131
132struct ocfs2_info_fs_features {
133 struct ocfs2_info_request if_req;
134 __u32 if_compat_features;
135 __u32 if_incompat_features;
136 __u32 if_ro_compat_features;
137 __u32 if_pad;
138};
139
140struct ocfs2_info_journal_size {
141 struct ocfs2_info_request ij_req;
142 __u64 ij_journal_size;
143};
144
145/* Codes for ocfs2_info_request */
146enum ocfs2_info_type {
147 OCFS2_INFO_CLUSTERSIZE = 1,
148 OCFS2_INFO_BLOCKSIZE,
149 OCFS2_INFO_MAXSLOTS,
150 OCFS2_INFO_LABEL,
151 OCFS2_INFO_UUID,
152 OCFS2_INFO_FS_FEATURES,
153 OCFS2_INFO_JOURNAL_SIZE,
154 OCFS2_INFO_NUM_TYPES
155};
156
157/* Flags for struct ocfs2_info_request */
158/* Filled by the caller */
159#define OCFS2_INFO_FL_NON_COHERENT (0x00000001) /* Cluster coherency not
160 required. This is a hint.
161 It is up to ocfs2 whether
162 the request can be fulfilled
163 without locking. */
164/* Filled by ocfs2 */
165#define OCFS2_INFO_FL_FILLED (0x40000000) /* Filesystem understood
166 this request and
167 filled in the answer */
168
169#define OCFS2_INFO_FL_ERROR (0x80000000) /* Error happened during
170 request handling. */
171
172#define OCFS2_IOC_INFO _IOR('o', 5, struct ocfs2_info)
173
79#endif /* OCFS2_IOCTL_H */ 174#endif /* OCFS2_IOCTL_H */
diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c
index efdd75607406..b5f9160e93e9 100644
--- a/fs/ocfs2/refcounttree.c
+++ b/fs/ocfs2/refcounttree.c
@@ -49,6 +49,7 @@
49 49
50struct ocfs2_cow_context { 50struct ocfs2_cow_context {
51 struct inode *inode; 51 struct inode *inode;
52 struct file *file;
52 u32 cow_start; 53 u32 cow_start;
53 u32 cow_len; 54 u32 cow_len;
54 struct ocfs2_extent_tree data_et; 55 struct ocfs2_extent_tree data_et;
@@ -2932,13 +2933,16 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle,
2932 u64 new_block = ocfs2_clusters_to_blocks(sb, new_cluster); 2933 u64 new_block = ocfs2_clusters_to_blocks(sb, new_cluster);
2933 struct page *page; 2934 struct page *page;
2934 pgoff_t page_index; 2935 pgoff_t page_index;
2935 unsigned int from, to; 2936 unsigned int from, to, readahead_pages;
2936 loff_t offset, end, map_end; 2937 loff_t offset, end, map_end;
2937 struct address_space *mapping = context->inode->i_mapping; 2938 struct address_space *mapping = context->inode->i_mapping;
2938 2939
2939 mlog(0, "old_cluster %u, new %u, len %u at offset %u\n", old_cluster, 2940 mlog(0, "old_cluster %u, new %u, len %u at offset %u\n", old_cluster,
2940 new_cluster, new_len, cpos); 2941 new_cluster, new_len, cpos);
2941 2942
2943 readahead_pages =
2944 (ocfs2_cow_contig_clusters(sb) <<
2945 OCFS2_SB(sb)->s_clustersize_bits) >> PAGE_CACHE_SHIFT;
2942 offset = ((loff_t)cpos) << OCFS2_SB(sb)->s_clustersize_bits; 2946 offset = ((loff_t)cpos) << OCFS2_SB(sb)->s_clustersize_bits;
2943 end = offset + (new_len << OCFS2_SB(sb)->s_clustersize_bits); 2947 end = offset + (new_len << OCFS2_SB(sb)->s_clustersize_bits);
2944 /* 2948 /*
@@ -2969,6 +2973,14 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle,
2969 if (PAGE_CACHE_SIZE <= OCFS2_SB(sb)->s_clustersize) 2973 if (PAGE_CACHE_SIZE <= OCFS2_SB(sb)->s_clustersize)
2970 BUG_ON(PageDirty(page)); 2974 BUG_ON(PageDirty(page));
2971 2975
2976 if (PageReadahead(page) && context->file) {
2977 page_cache_async_readahead(mapping,
2978 &context->file->f_ra,
2979 context->file,
2980 page, page_index,
2981 readahead_pages);
2982 }
2983
2972 if (!PageUptodate(page)) { 2984 if (!PageUptodate(page)) {
2973 ret = block_read_full_page(page, ocfs2_get_block); 2985 ret = block_read_full_page(page, ocfs2_get_block);
2974 if (ret) { 2986 if (ret) {
@@ -3409,12 +3421,35 @@ static int ocfs2_replace_cow(struct ocfs2_cow_context *context)
3409 return ret; 3421 return ret;
3410} 3422}
3411 3423
3424static void ocfs2_readahead_for_cow(struct inode *inode,
3425 struct file *file,
3426 u32 start, u32 len)
3427{
3428 struct address_space *mapping;
3429 pgoff_t index;
3430 unsigned long num_pages;
3431 int cs_bits = OCFS2_SB(inode->i_sb)->s_clustersize_bits;
3432
3433 if (!file)
3434 return;
3435
3436 mapping = file->f_mapping;
3437 num_pages = (len << cs_bits) >> PAGE_CACHE_SHIFT;
3438 if (!num_pages)
3439 num_pages = 1;
3440
3441 index = ((loff_t)start << cs_bits) >> PAGE_CACHE_SHIFT;
3442 page_cache_sync_readahead(mapping, &file->f_ra, file,
3443 index, num_pages);
3444}
3445
3412/* 3446/*
3413 * Starting at cpos, try to CoW write_len clusters. Don't CoW 3447 * Starting at cpos, try to CoW write_len clusters. Don't CoW
3414 * past max_cpos. This will stop when it runs into a hole or an 3448 * past max_cpos. This will stop when it runs into a hole or an
3415 * unrefcounted extent. 3449 * unrefcounted extent.
3416 */ 3450 */
3417static int ocfs2_refcount_cow_hunk(struct inode *inode, 3451static int ocfs2_refcount_cow_hunk(struct inode *inode,
3452 struct file *file,
3418 struct buffer_head *di_bh, 3453 struct buffer_head *di_bh,
3419 u32 cpos, u32 write_len, u32 max_cpos) 3454 u32 cpos, u32 write_len, u32 max_cpos)
3420{ 3455{
@@ -3443,6 +3478,8 @@ static int ocfs2_refcount_cow_hunk(struct inode *inode,
3443 3478
3444 BUG_ON(cow_len == 0); 3479 BUG_ON(cow_len == 0);
3445 3480
3481 ocfs2_readahead_for_cow(inode, file, cow_start, cow_len);
3482
3446 context = kzalloc(sizeof(struct ocfs2_cow_context), GFP_NOFS); 3483 context = kzalloc(sizeof(struct ocfs2_cow_context), GFP_NOFS);
3447 if (!context) { 3484 if (!context) {
3448 ret = -ENOMEM; 3485 ret = -ENOMEM;
@@ -3464,6 +3501,7 @@ static int ocfs2_refcount_cow_hunk(struct inode *inode,
3464 context->ref_root_bh = ref_root_bh; 3501 context->ref_root_bh = ref_root_bh;
3465 context->cow_duplicate_clusters = ocfs2_duplicate_clusters_by_page; 3502 context->cow_duplicate_clusters = ocfs2_duplicate_clusters_by_page;
3466 context->get_clusters = ocfs2_di_get_clusters; 3503 context->get_clusters = ocfs2_di_get_clusters;
3504 context->file = file;
3467 3505
3468 ocfs2_init_dinode_extent_tree(&context->data_et, 3506 ocfs2_init_dinode_extent_tree(&context->data_et,
3469 INODE_CACHE(inode), di_bh); 3507 INODE_CACHE(inode), di_bh);
@@ -3492,6 +3530,7 @@ out:
3492 * clusters between cpos and cpos+write_len are safe to modify. 3530 * clusters between cpos and cpos+write_len are safe to modify.
3493 */ 3531 */
3494int ocfs2_refcount_cow(struct inode *inode, 3532int ocfs2_refcount_cow(struct inode *inode,
3533 struct file *file,
3495 struct buffer_head *di_bh, 3534 struct buffer_head *di_bh,
3496 u32 cpos, u32 write_len, u32 max_cpos) 3535 u32 cpos, u32 write_len, u32 max_cpos)
3497{ 3536{
@@ -3511,7 +3550,7 @@ int ocfs2_refcount_cow(struct inode *inode,
3511 num_clusters = write_len; 3550 num_clusters = write_len;
3512 3551
3513 if (ext_flags & OCFS2_EXT_REFCOUNTED) { 3552 if (ext_flags & OCFS2_EXT_REFCOUNTED) {
3514 ret = ocfs2_refcount_cow_hunk(inode, di_bh, cpos, 3553 ret = ocfs2_refcount_cow_hunk(inode, file, di_bh, cpos,
3515 num_clusters, max_cpos); 3554 num_clusters, max_cpos);
3516 if (ret) { 3555 if (ret) {
3517 mlog_errno(ret); 3556 mlog_errno(ret);
diff --git a/fs/ocfs2/refcounttree.h b/fs/ocfs2/refcounttree.h
index 9983ba1570e2..c8ce46f7d8e3 100644
--- a/fs/ocfs2/refcounttree.h
+++ b/fs/ocfs2/refcounttree.h
@@ -21,14 +21,14 @@ struct ocfs2_refcount_tree {
21 struct rb_node rf_node; 21 struct rb_node rf_node;
22 u64 rf_blkno; 22 u64 rf_blkno;
23 u32 rf_generation; 23 u32 rf_generation;
24 struct kref rf_getcnt;
24 struct rw_semaphore rf_sem; 25 struct rw_semaphore rf_sem;
25 struct ocfs2_lock_res rf_lockres; 26 struct ocfs2_lock_res rf_lockres;
26 struct kref rf_getcnt;
27 int rf_removed; 27 int rf_removed;
28 28
29 /* the following 4 fields are used by caching_info. */ 29 /* the following 4 fields are used by caching_info. */
30 struct ocfs2_caching_info rf_ci;
31 spinlock_t rf_lock; 30 spinlock_t rf_lock;
31 struct ocfs2_caching_info rf_ci;
32 struct mutex rf_io_mutex; 32 struct mutex rf_io_mutex;
33 struct super_block *rf_sb; 33 struct super_block *rf_sb;
34}; 34};
@@ -52,7 +52,8 @@ int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
52 u32 clusters, 52 u32 clusters,
53 int *credits, 53 int *credits,
54 int *ref_blocks); 54 int *ref_blocks);
55int ocfs2_refcount_cow(struct inode *inode, struct buffer_head *di_bh, 55int ocfs2_refcount_cow(struct inode *inode,
56 struct file *filep, struct buffer_head *di_bh,
56 u32 cpos, u32 write_len, u32 max_cpos); 57 u32 cpos, u32 write_len, u32 max_cpos);
57 58
58typedef int (ocfs2_post_refcount_func)(struct inode *inode, 59typedef int (ocfs2_post_refcount_func)(struct inode *inode,
diff --git a/fs/ocfs2/slot_map.c b/fs/ocfs2/slot_map.c
index bfbd7e9e949f..ab4e0172cc1d 100644
--- a/fs/ocfs2/slot_map.c
+++ b/fs/ocfs2/slot_map.c
@@ -357,7 +357,7 @@ static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
357{ 357{
358 int status = 0; 358 int status = 0;
359 u64 blkno; 359 u64 blkno;
360 unsigned long long blocks, bytes; 360 unsigned long long blocks, bytes = 0;
361 unsigned int i; 361 unsigned int i;
362 struct buffer_head *bh; 362 struct buffer_head *bh;
363 363
diff --git a/fs/ocfs2/stack_o2cb.c b/fs/ocfs2/stack_o2cb.c
index 0d3049f696c5..19965b00c43c 100644
--- a/fs/ocfs2/stack_o2cb.c
+++ b/fs/ocfs2/stack_o2cb.c
@@ -283,6 +283,8 @@ static int o2cb_cluster_connect(struct ocfs2_cluster_connection *conn)
283 /* for now we only have one cluster/node, make sure we see it 283 /* for now we only have one cluster/node, make sure we see it
284 * in the heartbeat universe */ 284 * in the heartbeat universe */
285 if (!o2hb_check_local_node_heartbeating()) { 285 if (!o2hb_check_local_node_heartbeating()) {
286 if (o2hb_global_heartbeat_active())
287 mlog(ML_ERROR, "Global heartbeat not started\n");
286 rc = -EINVAL; 288 rc = -EINVAL;
287 goto out; 289 goto out;
288 } 290 }
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 849c2f0e0a0e..5fed60de7630 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -1380,6 +1380,14 @@ static inline int ocfs2_block_group_set_bits(handle_t *handle,
1380 } 1380 }
1381 1381
1382 le16_add_cpu(&bg->bg_free_bits_count, -num_bits); 1382 le16_add_cpu(&bg->bg_free_bits_count, -num_bits);
1383 if (le16_to_cpu(bg->bg_free_bits_count) > le16_to_cpu(bg->bg_bits)) {
1384 ocfs2_error(alloc_inode->i_sb, "Group descriptor # %llu has bit"
1385 " count %u but claims %u are freed. num_bits %d",
1386 (unsigned long long)le64_to_cpu(bg->bg_blkno),
1387 le16_to_cpu(bg->bg_bits),
1388 le16_to_cpu(bg->bg_free_bits_count), num_bits);
1389 return -EROFS;
1390 }
1383 while(num_bits--) 1391 while(num_bits--)
1384 ocfs2_set_bit(bit_off++, bitmap); 1392 ocfs2_set_bit(bit_off++, bitmap);
1385 1393
@@ -2419,6 +2427,14 @@ static int ocfs2_block_group_clear_bits(handle_t *handle,
2419 (unsigned long *) undo_bg->bg_bitmap); 2427 (unsigned long *) undo_bg->bg_bitmap);
2420 } 2428 }
2421 le16_add_cpu(&bg->bg_free_bits_count, num_bits); 2429 le16_add_cpu(&bg->bg_free_bits_count, num_bits);
2430 if (le16_to_cpu(bg->bg_free_bits_count) > le16_to_cpu(bg->bg_bits)) {
2431 ocfs2_error(alloc_inode->i_sb, "Group descriptor # %llu has bit"
2432 " count %u but claims %u are freed. num_bits %d",
2433 (unsigned long long)le64_to_cpu(bg->bg_blkno),
2434 le16_to_cpu(bg->bg_bits),
2435 le16_to_cpu(bg->bg_free_bits_count), num_bits);
2436 return -EROFS;
2437 }
2422 2438
2423 if (undo_fn) 2439 if (undo_fn)
2424 jbd_unlock_bh_state(group_bh); 2440 jbd_unlock_bh_state(group_bh);
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index fa1be1b304d1..a8a0ca44f88f 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -162,6 +162,7 @@ enum {
162 Opt_nointr, 162 Opt_nointr,
163 Opt_hb_none, 163 Opt_hb_none,
164 Opt_hb_local, 164 Opt_hb_local,
165 Opt_hb_global,
165 Opt_data_ordered, 166 Opt_data_ordered,
166 Opt_data_writeback, 167 Opt_data_writeback,
167 Opt_atime_quantum, 168 Opt_atime_quantum,
@@ -177,6 +178,8 @@ enum {
177 Opt_noacl, 178 Opt_noacl,
178 Opt_usrquota, 179 Opt_usrquota,
179 Opt_grpquota, 180 Opt_grpquota,
181 Opt_coherency_buffered,
182 Opt_coherency_full,
180 Opt_resv_level, 183 Opt_resv_level,
181 Opt_dir_resv_level, 184 Opt_dir_resv_level,
182 Opt_err, 185 Opt_err,
@@ -190,6 +193,7 @@ static const match_table_t tokens = {
190 {Opt_nointr, "nointr"}, 193 {Opt_nointr, "nointr"},
191 {Opt_hb_none, OCFS2_HB_NONE}, 194 {Opt_hb_none, OCFS2_HB_NONE},
192 {Opt_hb_local, OCFS2_HB_LOCAL}, 195 {Opt_hb_local, OCFS2_HB_LOCAL},
196 {Opt_hb_global, OCFS2_HB_GLOBAL},
193 {Opt_data_ordered, "data=ordered"}, 197 {Opt_data_ordered, "data=ordered"},
194 {Opt_data_writeback, "data=writeback"}, 198 {Opt_data_writeback, "data=writeback"},
195 {Opt_atime_quantum, "atime_quantum=%u"}, 199 {Opt_atime_quantum, "atime_quantum=%u"},
@@ -205,6 +209,8 @@ static const match_table_t tokens = {
205 {Opt_noacl, "noacl"}, 209 {Opt_noacl, "noacl"},
206 {Opt_usrquota, "usrquota"}, 210 {Opt_usrquota, "usrquota"},
207 {Opt_grpquota, "grpquota"}, 211 {Opt_grpquota, "grpquota"},
212 {Opt_coherency_buffered, "coherency=buffered"},
213 {Opt_coherency_full, "coherency=full"},
208 {Opt_resv_level, "resv_level=%u"}, 214 {Opt_resv_level, "resv_level=%u"},
209 {Opt_dir_resv_level, "dir_resv_level=%u"}, 215 {Opt_dir_resv_level, "dir_resv_level=%u"},
210 {Opt_err, NULL} 216 {Opt_err, NULL}
@@ -514,11 +520,11 @@ static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
514 520
515 mlog_entry_void(); 521 mlog_entry_void();
516 522
517 for (i = 0; i < NUM_SYSTEM_INODES; i++) { 523 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
518 inode = osb->system_inodes[i]; 524 inode = osb->global_system_inodes[i];
519 if (inode) { 525 if (inode) {
520 iput(inode); 526 iput(inode);
521 osb->system_inodes[i] = NULL; 527 osb->global_system_inodes[i] = NULL;
522 } 528 }
523 } 529 }
524 530
@@ -534,6 +540,20 @@ static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
534 osb->root_inode = NULL; 540 osb->root_inode = NULL;
535 } 541 }
536 542
543 if (!osb->local_system_inodes)
544 goto out;
545
546 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
547 if (osb->local_system_inodes[i]) {
548 iput(osb->local_system_inodes[i]);
549 osb->local_system_inodes[i] = NULL;
550 }
551 }
552
553 kfree(osb->local_system_inodes);
554 osb->local_system_inodes = NULL;
555
556out:
537 mlog_exit(0); 557 mlog_exit(0);
538} 558}
539 559
@@ -608,6 +628,7 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
608 int ret = 0; 628 int ret = 0;
609 struct mount_options parsed_options; 629 struct mount_options parsed_options;
610 struct ocfs2_super *osb = OCFS2_SB(sb); 630 struct ocfs2_super *osb = OCFS2_SB(sb);
631 u32 tmp;
611 632
612 lock_kernel(); 633 lock_kernel();
613 634
@@ -617,8 +638,9 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
617 goto out; 638 goto out;
618 } 639 }
619 640
620 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != 641 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
621 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 642 OCFS2_MOUNT_HB_NONE;
643 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
622 ret = -EINVAL; 644 ret = -EINVAL;
623 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); 645 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
624 goto out; 646 goto out;
@@ -809,23 +831,29 @@ bail:
809 831
810static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 832static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
811{ 833{
812 if (ocfs2_mount_local(osb)) { 834 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
813 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 835
836 if (osb->s_mount_opt & hb_enabled) {
837 if (ocfs2_mount_local(osb)) {
814 mlog(ML_ERROR, "Cannot heartbeat on a locally " 838 mlog(ML_ERROR, "Cannot heartbeat on a locally "
815 "mounted device.\n"); 839 "mounted device.\n");
816 return -EINVAL; 840 return -EINVAL;
817 } 841 }
818 } 842 if (ocfs2_userspace_stack(osb)) {
819
820 if (ocfs2_userspace_stack(osb)) {
821 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
822 mlog(ML_ERROR, "Userspace stack expected, but " 843 mlog(ML_ERROR, "Userspace stack expected, but "
823 "o2cb heartbeat arguments passed to mount\n"); 844 "o2cb heartbeat arguments passed to mount\n");
824 return -EINVAL; 845 return -EINVAL;
825 } 846 }
847 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
848 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
849 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
850 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
851 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
852 return -EINVAL;
853 }
826 } 854 }
827 855
828 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 856 if (!(osb->s_mount_opt & hb_enabled)) {
829 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && 857 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
830 !ocfs2_userspace_stack(osb)) { 858 !ocfs2_userspace_stack(osb)) {
831 mlog(ML_ERROR, "Heartbeat has to be started to mount " 859 mlog(ML_ERROR, "Heartbeat has to be started to mount "
@@ -1291,6 +1319,7 @@ static int ocfs2_parse_options(struct super_block *sb,
1291{ 1319{
1292 int status; 1320 int status;
1293 char *p; 1321 char *p;
1322 u32 tmp;
1294 1323
1295 mlog_entry("remount: %d, options: \"%s\"\n", is_remount, 1324 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1296 options ? options : "(none)"); 1325 options ? options : "(none)");
@@ -1322,7 +1351,10 @@ static int ocfs2_parse_options(struct super_block *sb,
1322 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1351 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1323 break; 1352 break;
1324 case Opt_hb_none: 1353 case Opt_hb_none:
1325 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; 1354 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1355 break;
1356 case Opt_hb_global:
1357 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1326 break; 1358 break;
1327 case Opt_barrier: 1359 case Opt_barrier:
1328 if (match_int(&args[0], &option)) { 1360 if (match_int(&args[0], &option)) {
@@ -1438,6 +1470,12 @@ static int ocfs2_parse_options(struct super_block *sb,
1438 case Opt_grpquota: 1470 case Opt_grpquota:
1439 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1471 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1440 break; 1472 break;
1473 case Opt_coherency_buffered:
1474 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1475 break;
1476 case Opt_coherency_full:
1477 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1478 break;
1441 case Opt_acl: 1479 case Opt_acl:
1442 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1480 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1443 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL; 1481 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
@@ -1477,6 +1515,15 @@ static int ocfs2_parse_options(struct super_block *sb,
1477 } 1515 }
1478 } 1516 }
1479 1517
1518 /* Ensure only one heartbeat mode */
1519 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1520 OCFS2_MOUNT_HB_NONE);
1521 if (hweight32(tmp) != 1) {
1522 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1523 status = 0;
1524 goto bail;
1525 }
1526
1480 status = 1; 1527 status = 1;
1481 1528
1482bail: 1529bail:
@@ -1490,10 +1537,14 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1490 unsigned long opts = osb->s_mount_opt; 1537 unsigned long opts = osb->s_mount_opt;
1491 unsigned int local_alloc_megs; 1538 unsigned int local_alloc_megs;
1492 1539
1493 if (opts & OCFS2_MOUNT_HB_LOCAL) 1540 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1494 seq_printf(s, ",_netdev,heartbeat=local"); 1541 seq_printf(s, ",_netdev");
1495 else 1542 if (opts & OCFS2_MOUNT_HB_LOCAL)
1496 seq_printf(s, ",heartbeat=none"); 1543 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1544 else
1545 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1546 } else
1547 seq_printf(s, ",%s", OCFS2_HB_NONE);
1497 1548
1498 if (opts & OCFS2_MOUNT_NOINTR) 1549 if (opts & OCFS2_MOUNT_NOINTR)
1499 seq_printf(s, ",nointr"); 1550 seq_printf(s, ",nointr");
@@ -1536,6 +1587,11 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1536 if (opts & OCFS2_MOUNT_GRPQUOTA) 1587 if (opts & OCFS2_MOUNT_GRPQUOTA)
1537 seq_printf(s, ",grpquota"); 1588 seq_printf(s, ",grpquota");
1538 1589
1590 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1591 seq_printf(s, ",coherency=buffered");
1592 else
1593 seq_printf(s, ",coherency=full");
1594
1539 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1595 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1540 seq_printf(s, ",nouser_xattr"); 1596 seq_printf(s, ",nouser_xattr");
1541 else 1597 else
@@ -1990,6 +2046,36 @@ static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uu
1990 return 0; 2046 return 0;
1991} 2047}
1992 2048
2049/* Make sure entire volume is addressable by our journal. Requires
2050 osb_clusters_at_boot to be valid and for the journal to have been
2051 initialized by ocfs2_journal_init(). */
2052static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2053{
2054 int status = 0;
2055 u64 max_block =
2056 ocfs2_clusters_to_blocks(osb->sb,
2057 osb->osb_clusters_at_boot) - 1;
2058
2059 /* 32-bit block number is always OK. */
2060 if (max_block <= (u32)~0ULL)
2061 goto out;
2062
2063 /* Volume is "huge", so see if our journal is new enough to
2064 support it. */
2065 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2066 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2067 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2068 JBD2_FEATURE_INCOMPAT_64BIT))) {
2069 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2070 "Enable the 'block64' journal option with tunefs.ocfs2");
2071 status = -EFBIG;
2072 goto out;
2073 }
2074
2075 out:
2076 return status;
2077}
2078
1993static int ocfs2_initialize_super(struct super_block *sb, 2079static int ocfs2_initialize_super(struct super_block *sb,
1994 struct buffer_head *bh, 2080 struct buffer_head *bh,
1995 int sector_size, 2081 int sector_size,
@@ -2002,6 +2088,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
2002 struct ocfs2_journal *journal; 2088 struct ocfs2_journal *journal;
2003 __le32 uuid_net_key; 2089 __le32 uuid_net_key;
2004 struct ocfs2_super *osb; 2090 struct ocfs2_super *osb;
2091 u64 total_blocks;
2005 2092
2006 mlog_entry_void(); 2093 mlog_entry_void();
2007 2094
@@ -2060,6 +2147,15 @@ static int ocfs2_initialize_super(struct super_block *sb,
2060 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 2147 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2061 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 2148 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2062 2149
2150 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2151 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2152 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2153 osb->max_slots);
2154 status = -EINVAL;
2155 goto bail;
2156 }
2157 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2158
2063 ocfs2_orphan_scan_init(osb); 2159 ocfs2_orphan_scan_init(osb);
2064 2160
2065 status = ocfs2_recovery_init(osb); 2161 status = ocfs2_recovery_init(osb);
@@ -2098,15 +2194,6 @@ static int ocfs2_initialize_super(struct super_block *sb,
2098 goto bail; 2194 goto bail;
2099 } 2195 }
2100 2196
2101 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2102 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2103 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2104 osb->max_slots);
2105 status = -EINVAL;
2106 goto bail;
2107 }
2108 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2109
2110 osb->slot_recovery_generations = 2197 osb->slot_recovery_generations =
2111 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2198 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2112 GFP_KERNEL); 2199 GFP_KERNEL);
@@ -2149,7 +2236,9 @@ static int ocfs2_initialize_super(struct super_block *sb,
2149 goto bail; 2236 goto bail;
2150 } 2237 }
2151 2238
2152 if (ocfs2_userspace_stack(osb)) { 2239 if (ocfs2_clusterinfo_valid(osb)) {
2240 osb->osb_stackflags =
2241 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2153 memcpy(osb->osb_cluster_stack, 2242 memcpy(osb->osb_cluster_stack,
2154 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2243 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2155 OCFS2_STACK_LABEL_LEN); 2244 OCFS2_STACK_LABEL_LEN);
@@ -2214,11 +2303,15 @@ static int ocfs2_initialize_super(struct super_block *sb,
2214 goto bail; 2303 goto bail;
2215 } 2304 }
2216 2305
2217 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) 2306 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2218 > (u32)~0UL) { 2307 le32_to_cpu(di->i_clusters));
2219 mlog(ML_ERROR, "Volume might try to write to blocks beyond " 2308
2220 "what jbd can address in 32 bits.\n"); 2309 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2221 status = -EINVAL; 2310 total_blocks);
2311 if (status) {
2312 mlog(ML_ERROR, "Volume too large "
2313 "to mount safely on this system");
2314 status = -EFBIG;
2222 goto bail; 2315 goto bail;
2223 } 2316 }
2224 2317
@@ -2380,6 +2473,12 @@ static int ocfs2_check_volume(struct ocfs2_super *osb)
2380 goto finally; 2473 goto finally;
2381 } 2474 }
2382 2475
2476 /* Now that journal has been initialized, check to make sure
2477 entire volume is addressable. */
2478 status = ocfs2_journal_addressable(osb);
2479 if (status)
2480 goto finally;
2481
2383 /* If the journal was unmounted cleanly then we don't want to 2482 /* If the journal was unmounted cleanly then we don't want to
2384 * recover anything. Otherwise, journal_load will do that 2483 * recover anything. Otherwise, journal_load will do that
2385 * dirty work for us :) */ 2484 * dirty work for us :) */
diff --git a/fs/ocfs2/sysfile.c b/fs/ocfs2/sysfile.c
index bfe7190cdbf1..902efb23b6a6 100644
--- a/fs/ocfs2/sysfile.c
+++ b/fs/ocfs2/sysfile.c
@@ -44,11 +44,6 @@ static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb,
44 int type, 44 int type,
45 u32 slot); 45 u32 slot);
46 46
47static inline int is_global_system_inode(int type);
48static inline int is_in_system_inode_array(struct ocfs2_super *osb,
49 int type,
50 u32 slot);
51
52#ifdef CONFIG_DEBUG_LOCK_ALLOC 47#ifdef CONFIG_DEBUG_LOCK_ALLOC
53static struct lock_class_key ocfs2_sysfile_cluster_lock_key[NUM_SYSTEM_INODES]; 48static struct lock_class_key ocfs2_sysfile_cluster_lock_key[NUM_SYSTEM_INODES];
54#endif 49#endif
@@ -59,11 +54,52 @@ static inline int is_global_system_inode(int type)
59 type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; 54 type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE;
60} 55}
61 56
62static inline int is_in_system_inode_array(struct ocfs2_super *osb, 57static struct inode **get_local_system_inode(struct ocfs2_super *osb,
63 int type, 58 int type,
64 u32 slot) 59 u32 slot)
65{ 60{
66 return slot == osb->slot_num || is_global_system_inode(type); 61 int index;
62 struct inode **local_system_inodes, **free = NULL;
63
64 BUG_ON(slot == OCFS2_INVALID_SLOT);
65 BUG_ON(type < OCFS2_FIRST_LOCAL_SYSTEM_INODE ||
66 type > OCFS2_LAST_LOCAL_SYSTEM_INODE);
67
68 spin_lock(&osb->osb_lock);
69 local_system_inodes = osb->local_system_inodes;
70 spin_unlock(&osb->osb_lock);
71
72 if (unlikely(!local_system_inodes)) {
73 local_system_inodes = kzalloc(sizeof(struct inode *) *
74 NUM_LOCAL_SYSTEM_INODES *
75 osb->max_slots,
76 GFP_NOFS);
77 if (!local_system_inodes) {
78 mlog_errno(-ENOMEM);
79 /*
80 * return NULL here so that ocfs2_get_sytem_file_inodes
81 * will try to create an inode and use it. We will try
82 * to initialize local_system_inodes next time.
83 */
84 return NULL;
85 }
86
87 spin_lock(&osb->osb_lock);
88 if (osb->local_system_inodes) {
89 /* Someone has initialized it for us. */
90 free = local_system_inodes;
91 local_system_inodes = osb->local_system_inodes;
92 } else
93 osb->local_system_inodes = local_system_inodes;
94 spin_unlock(&osb->osb_lock);
95 if (unlikely(free))
96 kfree(free);
97 }
98
99 index = (slot * NUM_LOCAL_SYSTEM_INODES) +
100 (type - OCFS2_FIRST_LOCAL_SYSTEM_INODE);
101
102 return &local_system_inodes[index];
67} 103}
68 104
69struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb, 105struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb,
@@ -74,8 +110,10 @@ struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb,
74 struct inode **arr = NULL; 110 struct inode **arr = NULL;
75 111
76 /* avoid the lookup if cached in local system file array */ 112 /* avoid the lookup if cached in local system file array */
77 if (is_in_system_inode_array(osb, type, slot)) 113 if (is_global_system_inode(type)) {
78 arr = &(osb->system_inodes[type]); 114 arr = &(osb->global_system_inodes[type]);
115 } else
116 arr = get_local_system_inode(osb, type, slot);
79 117
80 if (arr && ((inode = *arr) != NULL)) { 118 if (arr && ((inode = *arr) != NULL)) {
81 /* get a ref in addition to the array ref */ 119 /* get a ref in addition to the array ref */
diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index 06fa5e77c40e..67cd43914641 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -7081,7 +7081,7 @@ static int ocfs2_reflink_xattr_in_block(struct ocfs2_xattr_reflink *args,
7081 goto out; 7081 goto out;
7082 } 7082 }
7083 7083
7084 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) 7084 if (!indexed)
7085 ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); 7085 ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh);
7086 else 7086 else
7087 ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); 7087 ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 3168dcfb94f2..7d6f18fddfdb 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2378,6 +2378,8 @@ extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
2378 2378
2379extern int generic_file_fsync(struct file *, int); 2379extern int generic_file_fsync(struct file *, int);
2380 2380
2381extern int generic_check_addressable(unsigned, u64);
2382
2381#ifdef CONFIG_MIGRATION 2383#ifdef CONFIG_MIGRATION
2382extern int buffer_migrate_page(struct address_space *, 2384extern int buffer_migrate_page(struct address_space *,
2383 struct page *, struct page *); 2385 struct page *, struct page *);
generated by cgit v1.2.2 (git 2.25.0) at 2024-09-23 14:35:48 -0400