aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--fs/ocfs2/Makefile3
-rw-r--r--fs/ocfs2/cluster/tcp_internal.h5
-rw-r--r--fs/ocfs2/dcache.c8
-rw-r--r--fs/ocfs2/dlmglue.c164
-rw-r--r--fs/ocfs2/dlmglue.h5
-rw-r--r--fs/ocfs2/heartbeat.c7
-rw-r--r--fs/ocfs2/inode.c36
-rw-r--r--fs/ocfs2/journal.c15
-rw-r--r--fs/ocfs2/namei.c10
-rw-r--r--fs/ocfs2/ocfs2.h25
-rw-r--r--fs/ocfs2/slot_map.c19
-rw-r--r--fs/ocfs2/slot_map.h2
-rw-r--r--fs/ocfs2/super.c43
-rw-r--r--fs/ocfs2/vote.c756
-rw-r--r--fs/ocfs2/vote.h48
15 files changed, 179 insertions, 967 deletions
diff --git a/fs/ocfs2/Makefile b/fs/ocfs2/Makefile
index 9fb8132f19b0..d2057e7fbda7 100644
--- a/fs/ocfs2/Makefile
+++ b/fs/ocfs2/Makefile
@@ -27,8 +27,7 @@ ocfs2-objs := \
27 symlink.o \ 27 symlink.o \
28 sysfile.o \ 28 sysfile.o \
29 uptodate.o \ 29 uptodate.o \
30 ver.o \ 30 ver.o
31 vote.o
32 31
33obj-$(CONFIG_OCFS2_FS) += cluster/ 32obj-$(CONFIG_OCFS2_FS) += cluster/
34obj-$(CONFIG_OCFS2_FS) += dlm/ 33obj-$(CONFIG_OCFS2_FS) += dlm/
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index 9606111fe89d..79bd6665b3ca 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -38,6 +38,9 @@
38 * locking semantics of the file system using the protocol. It should 38 * locking semantics of the file system using the protocol. It should
39 * be somewhere else, I'm sure, but right now it isn't. 39 * be somewhere else, I'm sure, but right now it isn't.
40 * 40 *
41 * New in version 9:
42 * - All votes removed
43 *
41 * New in version 8: 44 * New in version 8:
42 * - Replace delete inode votes with a cluster lock 45 * - Replace delete inode votes with a cluster lock
43 * 46 *
@@ -60,7 +63,7 @@
60 * - full 64 bit i_size in the metadata lock lvbs 63 * - full 64 bit i_size in the metadata lock lvbs
61 * - introduction of "rw" lock and pushing meta/data locking down 64 * - introduction of "rw" lock and pushing meta/data locking down
62 */ 65 */
63#define O2NET_PROTOCOL_VERSION 8ULL 66#define O2NET_PROTOCOL_VERSION 9ULL
64struct o2net_handshake { 67struct o2net_handshake {
65 __be64 protocol_version; 68 __be64 protocol_version;
66 __be64 connector_id; 69 __be64 connector_id;
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
index 9923278ea6d4..b1cc7c381e88 100644
--- a/fs/ocfs2/dcache.c
+++ b/fs/ocfs2/dcache.c
@@ -128,9 +128,9 @@ static int ocfs2_match_dentry(struct dentry *dentry,
128/* 128/*
129 * Walk the inode alias list, and find a dentry which has a given 129 * Walk the inode alias list, and find a dentry which has a given
130 * parent. ocfs2_dentry_attach_lock() wants to find _any_ alias as it 130 * parent. ocfs2_dentry_attach_lock() wants to find _any_ alias as it
131 * is looking for a dentry_lock reference. The vote thread is looking 131 * is looking for a dentry_lock reference. The downconvert thread is
132 * to unhash aliases, so we allow it to skip any that already have 132 * looking to unhash aliases, so we allow it to skip any that already
133 * that property. 133 * have that property.
134 */ 134 */
135struct dentry *ocfs2_find_local_alias(struct inode *inode, 135struct dentry *ocfs2_find_local_alias(struct inode *inode,
136 u64 parent_blkno, 136 u64 parent_blkno,
@@ -266,7 +266,7 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry,
266 dl->dl_count = 0; 266 dl->dl_count = 0;
267 /* 267 /*
268 * Does this have to happen below, for all attaches, in case 268 * Does this have to happen below, for all attaches, in case
269 * the struct inode gets blown away by votes? 269 * the struct inode gets blown away by the downconvert thread?
270 */ 270 */
271 dl->dl_inode = igrab(inode); 271 dl->dl_inode = igrab(inode);
272 dl->dl_parent_blkno = parent_blkno; 272 dl->dl_parent_blkno = parent_blkno;
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index 4e97dcceaf8f..b3068ade3f7b 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -55,7 +55,6 @@
55#include "slot_map.h" 55#include "slot_map.h"
56#include "super.h" 56#include "super.h"
57#include "uptodate.h" 57#include "uptodate.h"
58#include "vote.h"
59 58
60#include "buffer_head_io.h" 59#include "buffer_head_io.h"
61 60
@@ -153,10 +152,10 @@ struct ocfs2_lock_res_ops {
153 struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *); 152 struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *);
154 153
155 /* 154 /*
156 * Optionally called in the downconvert (or "vote") thread 155 * Optionally called in the downconvert thread after a
157 * after a successful downconvert. The lockres will not be 156 * successful downconvert. The lockres will not be referenced
158 * referenced after this callback is called, so it is safe to 157 * after this callback is called, so it is safe to free
159 * free memory, etc. 158 * memory, etc.
160 * 159 *
161 * The exact semantics of when this is called are controlled 160 * The exact semantics of when this is called are controlled
162 * by ->downconvert_worker() 161 * by ->downconvert_worker()
@@ -310,8 +309,9 @@ static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
310 "resource %s: %s\n", dlm_errname(_stat), _func, \ 309 "resource %s: %s\n", dlm_errname(_stat), _func, \
311 _lockres->l_name, dlm_errmsg(_stat)); \ 310 _lockres->l_name, dlm_errmsg(_stat)); \
312} while (0) 311} while (0)
313static void ocfs2_vote_on_unlock(struct ocfs2_super *osb, 312static int ocfs2_downconvert_thread(void *arg);
314 struct ocfs2_lock_res *lockres); 313static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb,
314 struct ocfs2_lock_res *lockres);
315static int ocfs2_meta_lock_update(struct inode *inode, 315static int ocfs2_meta_lock_update(struct inode *inode,
316 struct buffer_head **bh); 316 struct buffer_head **bh);
317static void ocfs2_drop_osb_locks(struct ocfs2_super *osb); 317static void ocfs2_drop_osb_locks(struct ocfs2_super *osb);
@@ -732,7 +732,7 @@ static void ocfs2_blocking_ast(void *opaque, int level)
732 732
733 wake_up(&lockres->l_event); 733 wake_up(&lockres->l_event);
734 734
735 ocfs2_kick_vote_thread(osb); 735 ocfs2_wake_downconvert_thread(osb);
736} 736}
737 737
738static void ocfs2_locking_ast(void *opaque) 738static void ocfs2_locking_ast(void *opaque)
@@ -1089,7 +1089,7 @@ static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
1089 mlog_entry_void(); 1089 mlog_entry_void();
1090 spin_lock_irqsave(&lockres->l_lock, flags); 1090 spin_lock_irqsave(&lockres->l_lock, flags);
1091 ocfs2_dec_holders(lockres, level); 1091 ocfs2_dec_holders(lockres, level);
1092 ocfs2_vote_on_unlock(osb, lockres); 1092 ocfs2_downconvert_on_unlock(osb, lockres);
1093 spin_unlock_irqrestore(&lockres->l_lock, flags); 1093 spin_unlock_irqrestore(&lockres->l_lock, flags);
1094 mlog_exit_void(); 1094 mlog_exit_void();
1095} 1095}
@@ -1372,15 +1372,15 @@ int ocfs2_data_lock_with_page(struct inode *inode,
1372 return ret; 1372 return ret;
1373} 1373}
1374 1374
1375static void ocfs2_vote_on_unlock(struct ocfs2_super *osb, 1375static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb,
1376 struct ocfs2_lock_res *lockres) 1376 struct ocfs2_lock_res *lockres)
1377{ 1377{
1378 int kick = 0; 1378 int kick = 0;
1379 1379
1380 mlog_entry_void(); 1380 mlog_entry_void();
1381 1381
1382 /* If we know that another node is waiting on our lock, kick 1382 /* If we know that another node is waiting on our lock, kick
1383 * the vote thread * pre-emptively when we reach a release 1383 * the downconvert thread * pre-emptively when we reach a release
1384 * condition. */ 1384 * condition. */
1385 if (lockres->l_flags & OCFS2_LOCK_BLOCKED) { 1385 if (lockres->l_flags & OCFS2_LOCK_BLOCKED) {
1386 switch(lockres->l_blocking) { 1386 switch(lockres->l_blocking) {
@@ -1398,7 +1398,7 @@ static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
1398 } 1398 }
1399 1399
1400 if (kick) 1400 if (kick)
1401 ocfs2_kick_vote_thread(osb); 1401 ocfs2_wake_downconvert_thread(osb);
1402 1402
1403 mlog_exit_void(); 1403 mlog_exit_void();
1404} 1404}
@@ -1832,19 +1832,20 @@ bail:
1832} 1832}
1833 1833
1834/* 1834/*
1835 * This is working around a lock inversion between tasks acquiring DLM locks 1835 * This is working around a lock inversion between tasks acquiring DLM
1836 * while holding a page lock and the vote thread which blocks dlm lock acquiry 1836 * locks while holding a page lock and the downconvert thread which
1837 * while acquiring page locks. 1837 * blocks dlm lock acquiry while acquiring page locks.
1838 * 1838 *
1839 * ** These _with_page variantes are only intended to be called from aop 1839 * ** These _with_page variantes are only intended to be called from aop
1840 * methods that hold page locks and return a very specific *positive* error 1840 * methods that hold page locks and return a very specific *positive* error
1841 * code that aop methods pass up to the VFS -- test for errors with != 0. ** 1841 * code that aop methods pass up to the VFS -- test for errors with != 0. **
1842 * 1842 *
1843 * The DLM is called such that it returns -EAGAIN if it would have blocked 1843 * The DLM is called such that it returns -EAGAIN if it would have
1844 * waiting for the vote thread. In that case we unlock our page so the vote 1844 * blocked waiting for the downconvert thread. In that case we unlock
1845 * thread can make progress. Once we've done this we have to return 1845 * our page so the downconvert thread can make progress. Once we've
1846 * AOP_TRUNCATED_PAGE so the aop method that called us can bubble that back up 1846 * done this we have to return AOP_TRUNCATED_PAGE so the aop method
1847 * into the VFS who will then immediately retry the aop call. 1847 * that called us can bubble that back up into the VFS who will then
1848 * immediately retry the aop call.
1848 * 1849 *
1849 * We do a blocking lock and immediate unlock before returning, though, so that 1850 * We do a blocking lock and immediate unlock before returning, though, so that
1850 * the lock has a great chance of being cached on this node by the time the VFS 1851 * the lock has a great chance of being cached on this node by the time the VFS
@@ -2320,11 +2321,11 @@ int ocfs2_dlm_init(struct ocfs2_super *osb)
2320 goto bail; 2321 goto bail;
2321 } 2322 }
2322 2323
2323 /* launch vote thread */ 2324 /* launch downconvert thread */
2324 osb->vote_task = kthread_run(ocfs2_vote_thread, osb, "ocfs2vote"); 2325 osb->dc_task = kthread_run(ocfs2_downconvert_thread, osb, "ocfs2dc");
2325 if (IS_ERR(osb->vote_task)) { 2326 if (IS_ERR(osb->dc_task)) {
2326 status = PTR_ERR(osb->vote_task); 2327 status = PTR_ERR(osb->dc_task);
2327 osb->vote_task = NULL; 2328 osb->dc_task = NULL;
2328 mlog_errno(status); 2329 mlog_errno(status);
2329 goto bail; 2330 goto bail;
2330 } 2331 }
@@ -2353,8 +2354,8 @@ local:
2353bail: 2354bail:
2354 if (status < 0) { 2355 if (status < 0) {
2355 ocfs2_dlm_shutdown_debug(osb); 2356 ocfs2_dlm_shutdown_debug(osb);
2356 if (osb->vote_task) 2357 if (osb->dc_task)
2357 kthread_stop(osb->vote_task); 2358 kthread_stop(osb->dc_task);
2358 } 2359 }
2359 2360
2360 mlog_exit(status); 2361 mlog_exit(status);
@@ -2369,9 +2370,9 @@ void ocfs2_dlm_shutdown(struct ocfs2_super *osb)
2369 2370
2370 ocfs2_drop_osb_locks(osb); 2371 ocfs2_drop_osb_locks(osb);
2371 2372
2372 if (osb->vote_task) { 2373 if (osb->dc_task) {
2373 kthread_stop(osb->vote_task); 2374 kthread_stop(osb->dc_task);
2374 osb->vote_task = NULL; 2375 osb->dc_task = NULL;
2375 } 2376 }
2376 2377
2377 ocfs2_lock_res_free(&osb->osb_super_lockres); 2378 ocfs2_lock_res_free(&osb->osb_super_lockres);
@@ -2527,7 +2528,7 @@ out:
2527 2528
2528/* Mark the lockres as being dropped. It will no longer be 2529/* Mark the lockres as being dropped. It will no longer be
2529 * queued if blocking, but we still may have to wait on it 2530 * queued if blocking, but we still may have to wait on it
2530 * being dequeued from the vote thread before we can consider 2531 * being dequeued from the downconvert thread before we can consider
2531 * it safe to drop. 2532 * it safe to drop.
2532 * 2533 *
2533 * You can *not* attempt to call cluster_lock on this lockres anymore. */ 2534 * You can *not* attempt to call cluster_lock on this lockres anymore. */
@@ -2903,7 +2904,7 @@ static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres)
2903 2904
2904/* 2905/*
2905 * Does the final reference drop on our dentry lock. Right now this 2906 * Does the final reference drop on our dentry lock. Right now this
2906 * happens in the vote thread, but we could choose to simplify the 2907 * happens in the downconvert thread, but we could choose to simplify the
2907 * dlmglue API and push these off to the ocfs2_wq in the future. 2908 * dlmglue API and push these off to the ocfs2_wq in the future.
2908 */ 2909 */
2909static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, 2910static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
@@ -3042,7 +3043,7 @@ void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
3042 mlog(0, "lockres %s blocked.\n", lockres->l_name); 3043 mlog(0, "lockres %s blocked.\n", lockres->l_name);
3043 3044
3044 /* Detect whether a lock has been marked as going away while 3045 /* Detect whether a lock has been marked as going away while
3045 * the vote thread was processing other things. A lock can 3046 * the downconvert thread was processing other things. A lock can
3046 * still be marked with OCFS2_LOCK_FREEING after this check, 3047 * still be marked with OCFS2_LOCK_FREEING after this check,
3047 * but short circuiting here will still save us some 3048 * but short circuiting here will still save us some
3048 * performance. */ 3049 * performance. */
@@ -3091,13 +3092,104 @@ static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
3091 3092
3092 lockres_or_flags(lockres, OCFS2_LOCK_QUEUED); 3093 lockres_or_flags(lockres, OCFS2_LOCK_QUEUED);
3093 3094
3094 spin_lock(&osb->vote_task_lock); 3095 spin_lock(&osb->dc_task_lock);
3095 if (list_empty(&lockres->l_blocked_list)) { 3096 if (list_empty(&lockres->l_blocked_list)) {
3096 list_add_tail(&lockres->l_blocked_list, 3097 list_add_tail(&lockres->l_blocked_list,
3097 &osb->blocked_lock_list); 3098 &osb->blocked_lock_list);
3098 osb->blocked_lock_count++; 3099 osb->blocked_lock_count++;
3099 } 3100 }
3100 spin_unlock(&osb->vote_task_lock); 3101 spin_unlock(&osb->dc_task_lock);
3101 3102
3102 mlog_exit_void(); 3103 mlog_exit_void();
3103} 3104}
3105
3106static void ocfs2_downconvert_thread_do_work(struct ocfs2_super *osb)
3107{
3108 unsigned long processed;
3109 struct ocfs2_lock_res *lockres;
3110
3111 mlog_entry_void();
3112
3113 spin_lock(&osb->dc_task_lock);
3114 /* grab this early so we know to try again if a state change and
3115 * wake happens part-way through our work */
3116 osb->dc_work_sequence = osb->dc_wake_sequence;
3117
3118 processed = osb->blocked_lock_count;
3119 while (processed) {
3120 BUG_ON(list_empty(&osb->blocked_lock_list));
3121
3122 lockres = list_entry(osb->blocked_lock_list.next,
3123 struct ocfs2_lock_res, l_blocked_list);
3124 list_del_init(&lockres->l_blocked_list);
3125 osb->blocked_lock_count--;
3126 spin_unlock(&osb->dc_task_lock);
3127
3128 BUG_ON(!processed);
3129 processed--;
3130
3131 ocfs2_process_blocked_lock(osb, lockres);
3132
3133 spin_lock(&osb->dc_task_lock);
3134 }
3135 spin_unlock(&osb->dc_task_lock);
3136
3137 mlog_exit_void();
3138}
3139
3140static int ocfs2_downconvert_thread_lists_empty(struct ocfs2_super *osb)
3141{
3142 int empty = 0;
3143
3144 spin_lock(&osb->dc_task_lock);
3145 if (list_empty(&osb->blocked_lock_list))
3146 empty = 1;
3147
3148 spin_unlock(&osb->dc_task_lock);
3149 return empty;
3150}
3151
3152static int ocfs2_downconvert_thread_should_wake(struct ocfs2_super *osb)
3153{
3154 int should_wake = 0;
3155
3156 spin_lock(&osb->dc_task_lock);
3157 if (osb->dc_work_sequence != osb->dc_wake_sequence)
3158 should_wake = 1;
3159 spin_unlock(&osb->dc_task_lock);
3160
3161 return should_wake;
3162}
3163
3164int ocfs2_downconvert_thread(void *arg)
3165{
3166 int status = 0;
3167 struct ocfs2_super *osb = arg;
3168
3169 /* only quit once we've been asked to stop and there is no more
3170 * work available */
3171 while (!(kthread_should_stop() &&
3172 ocfs2_downconvert_thread_lists_empty(osb))) {
3173
3174 wait_event_interruptible(osb->dc_event,
3175 ocfs2_downconvert_thread_should_wake(osb) ||
3176 kthread_should_stop());
3177
3178 mlog(0, "downconvert_thread: awoken\n");
3179
3180 ocfs2_downconvert_thread_do_work(osb);
3181 }
3182
3183 osb->dc_task = NULL;
3184 return status;
3185}
3186
3187void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb)
3188{
3189 spin_lock(&osb->dc_task_lock);
3190 /* make sure the voting thread gets a swipe at whatever changes
3191 * the caller may have made to the voting state */
3192 osb->dc_wake_sequence++;
3193 spin_unlock(&osb->dc_task_lock);
3194 wake_up(&osb->dc_event);
3195}
diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h
index 87a785e41205..931f6ee55146 100644
--- a/fs/ocfs2/dlmglue.h
+++ b/fs/ocfs2/dlmglue.h
@@ -54,7 +54,7 @@ struct ocfs2_meta_lvb {
54#define OCFS2_META_LOCK_RECOVERY (0x01) 54#define OCFS2_META_LOCK_RECOVERY (0x01)
55/* Instruct the dlm not to queue ourselves on the other node. */ 55/* Instruct the dlm not to queue ourselves on the other node. */
56#define OCFS2_META_LOCK_NOQUEUE (0x02) 56#define OCFS2_META_LOCK_NOQUEUE (0x02)
57/* don't block waiting for the vote thread, instead return -EAGAIN */ 57/* don't block waiting for the downconvert thread, instead return -EAGAIN */
58#define OCFS2_LOCK_NONBLOCK (0x04) 58#define OCFS2_LOCK_NONBLOCK (0x04)
59 59
60int ocfs2_dlm_init(struct ocfs2_super *osb); 60int ocfs2_dlm_init(struct ocfs2_super *osb);
@@ -112,9 +112,10 @@ void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres);
112void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, 112void ocfs2_simple_drop_lockres(struct ocfs2_super *osb,
113 struct ocfs2_lock_res *lockres); 113 struct ocfs2_lock_res *lockres);
114 114
115/* for the vote thread */ 115/* for the downconvert thread */
116void ocfs2_process_blocked_lock(struct ocfs2_super *osb, 116void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
117 struct ocfs2_lock_res *lockres); 117 struct ocfs2_lock_res *lockres);
118void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb);
118 119
119struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void); 120struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void);
120void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug); 121void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug);
diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c
index 6239fc52790c..c0efd9489fe8 100644
--- a/fs/ocfs2/heartbeat.c
+++ b/fs/ocfs2/heartbeat.c
@@ -41,7 +41,6 @@
41#include "heartbeat.h" 41#include "heartbeat.h"
42#include "inode.h" 42#include "inode.h"
43#include "journal.h" 43#include "journal.h"
44#include "vote.h"
45 44
46#include "buffer_head_io.h" 45#include "buffer_head_io.h"
47 46
@@ -58,9 +57,7 @@ static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
58void ocfs2_init_node_maps(struct ocfs2_super *osb) 57void ocfs2_init_node_maps(struct ocfs2_super *osb)
59{ 58{
60 spin_lock_init(&osb->node_map_lock); 59 spin_lock_init(&osb->node_map_lock);
61 ocfs2_node_map_init(&osb->mounted_map);
62 ocfs2_node_map_init(&osb->recovery_map); 60 ocfs2_node_map_init(&osb->recovery_map);
63 ocfs2_node_map_init(&osb->umount_map);
64 ocfs2_node_map_init(&osb->osb_recovering_orphan_dirs); 61 ocfs2_node_map_init(&osb->osb_recovering_orphan_dirs);
65} 62}
66 63
@@ -82,8 +79,6 @@ static void ocfs2_do_node_down(int node_num,
82 } 79 }
83 80
84 ocfs2_recovery_thread(osb, node_num); 81 ocfs2_recovery_thread(osb, node_num);
85
86 ocfs2_remove_node_from_vote_queues(osb, node_num);
87} 82}
88 83
89/* Called from the dlm when it's about to evict a node. We may also 84/* Called from the dlm when it's about to evict a node. We may also
@@ -268,8 +263,6 @@ int ocfs2_recovery_map_set(struct ocfs2_super *osb,
268 263
269 spin_lock(&osb->node_map_lock); 264 spin_lock(&osb->node_map_lock);
270 265
271 __ocfs2_node_map_clear_bit(&osb->mounted_map, num);
272
273 if (!test_bit(num, osb->recovery_map.map)) { 266 if (!test_bit(num, osb->recovery_map.map)) {
274 __ocfs2_node_map_set_bit(&osb->recovery_map, num); 267 __ocfs2_node_map_set_bit(&osb->recovery_map, num);
275 set = 1; 268 set = 1;
diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
index ebb2bbe30f35..86cf073996b5 100644
--- a/fs/ocfs2/inode.c
+++ b/fs/ocfs2/inode.c
@@ -49,7 +49,6 @@
49#include "symlink.h" 49#include "symlink.h"
50#include "sysfile.h" 50#include "sysfile.h"
51#include "uptodate.h" 51#include "uptodate.h"
52#include "vote.h"
53 52
54#include "buffer_head_io.h" 53#include "buffer_head_io.h"
55 54
@@ -718,8 +717,8 @@ static int ocfs2_wipe_inode(struct inode *inode,
718 } 717 }
719 718
720 /* we do this while holding the orphan dir lock because we 719 /* we do this while holding the orphan dir lock because we
721 * don't want recovery being run from another node to vote for 720 * don't want recovery being run from another node to try an
722 * an inode delete on us -- this will result in two nodes 721 * inode delete underneath us -- this will result in two nodes
723 * truncating the same file! */ 722 * truncating the same file! */
724 status = ocfs2_truncate_for_delete(osb, inode, di_bh); 723 status = ocfs2_truncate_for_delete(osb, inode, di_bh);
725 if (status < 0) { 724 if (status < 0) {
@@ -744,7 +743,7 @@ bail:
744} 743}
745 744
746/* There is a series of simple checks that should be done before a 745/* There is a series of simple checks that should be done before a
747 * vote is even considered. Encapsulate those in this function. */ 746 * trylock is even considered. Encapsulate those in this function. */
748static int ocfs2_inode_is_valid_to_delete(struct inode *inode) 747static int ocfs2_inode_is_valid_to_delete(struct inode *inode)
749{ 748{
750 int ret = 0; 749 int ret = 0;
@@ -758,14 +757,14 @@ static int ocfs2_inode_is_valid_to_delete(struct inode *inode)
758 goto bail; 757 goto bail;
759 } 758 }
760 759
761 /* If we're coming from process_vote we can't go into our own 760 /* If we're coming from downconvert_thread we can't go into our own
762 * voting [hello, deadlock city!], so unforuntately we just 761 * voting [hello, deadlock city!], so unforuntately we just
763 * have to skip deleting this guy. That's OK though because 762 * have to skip deleting this guy. That's OK though because
764 * the node who's doing the actual deleting should handle it 763 * the node who's doing the actual deleting should handle it
765 * anyway. */ 764 * anyway. */
766 if (current == osb->vote_task) { 765 if (current == osb->dc_task) {
767 mlog(0, "Skipping delete of %lu because we're currently " 766 mlog(0, "Skipping delete of %lu because we're currently "
768 "in process_vote\n", inode->i_ino); 767 "in downconvert\n", inode->i_ino);
769 goto bail; 768 goto bail;
770 } 769 }
771 770
@@ -779,10 +778,9 @@ static int ocfs2_inode_is_valid_to_delete(struct inode *inode)
779 goto bail_unlock; 778 goto bail_unlock;
780 } 779 }
781 780
782 /* If we have voted "yes" on the wipe of this inode for 781 /* If we have allowd wipe of this inode for another node, it
783 * another node, it will be marked here so we can safely skip 782 * will be marked here so we can safely skip it. Recovery will
784 * it. Recovery will cleanup any inodes we might inadvertantly 783 * cleanup any inodes we might inadvertantly skip here. */
785 * skip here. */
786 if (oi->ip_flags & OCFS2_INODE_SKIP_DELETE) { 784 if (oi->ip_flags & OCFS2_INODE_SKIP_DELETE) {
787 mlog(0, "Skipping delete of %lu because another node " 785 mlog(0, "Skipping delete of %lu because another node "
788 "has done this for us.\n", inode->i_ino); 786 "has done this for us.\n", inode->i_ino);
@@ -929,7 +927,7 @@ void ocfs2_delete_inode(struct inode *inode)
929 927
930 /* Lock down the inode. This gives us an up to date view of 928 /* Lock down the inode. This gives us an up to date view of
931 * it's metadata (for verification), and allows us to 929 * it's metadata (for verification), and allows us to
932 * serialize delete_inode votes. 930 * serialize delete_inode on multiple nodes.
933 * 931 *
934 * Even though we might be doing a truncate, we don't take the 932 * Even though we might be doing a truncate, we don't take the
935 * allocation lock here as it won't be needed - nobody will 933 * allocation lock here as it won't be needed - nobody will
@@ -947,15 +945,15 @@ void ocfs2_delete_inode(struct inode *inode)
947 * before we go ahead and wipe the inode. */ 945 * before we go ahead and wipe the inode. */
948 status = ocfs2_query_inode_wipe(inode, di_bh, &wipe); 946 status = ocfs2_query_inode_wipe(inode, di_bh, &wipe);
949 if (!wipe || status < 0) { 947 if (!wipe || status < 0) {
950 /* Error and inode busy vote both mean we won't be 948 /* Error and remote inode busy both mean we won't be
951 * removing the inode, so they take almost the same 949 * removing the inode, so they take almost the same
952 * path. */ 950 * path. */
953 if (status < 0) 951 if (status < 0)
954 mlog_errno(status); 952 mlog_errno(status);
955 953
956 /* Someone in the cluster has voted to not wipe this 954 /* Someone in the cluster has disallowed a wipe of
957 * inode, or it was never completely orphaned. Write 955 * this inode, or it was never completely
958 * out the pages and exit now. */ 956 * orphaned. Write out the pages and exit now. */
959 ocfs2_cleanup_delete_inode(inode, 1); 957 ocfs2_cleanup_delete_inode(inode, 1);
960 goto bail_unlock_inode; 958 goto bail_unlock_inode;
961 } 959 }
@@ -1008,12 +1006,12 @@ void ocfs2_clear_inode(struct inode *inode)
1008 mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL, 1006 mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL,
1009 "Inode=%lu\n", inode->i_ino); 1007 "Inode=%lu\n", inode->i_ino);
1010 1008
1011 /* For remove delete_inode vote, we hold open lock before, 1009 /* To preven remote deletes we hold open lock before, now it
1012 * now it is time to unlock PR and EX open locks. */ 1010 * is time to unlock PR and EX open locks. */
1013 ocfs2_open_unlock(inode); 1011 ocfs2_open_unlock(inode);
1014 1012
1015 /* Do these before all the other work so that we don't bounce 1013 /* Do these before all the other work so that we don't bounce
1016 * the vote thread while waiting to destroy the locks. */ 1014 * the downconvert thread while waiting to destroy the locks. */
1017 ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres); 1015 ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres);
1018 ocfs2_mark_lockres_freeing(&oi->ip_meta_lockres); 1016 ocfs2_mark_lockres_freeing(&oi->ip_meta_lockres);
1019 ocfs2_mark_lockres_freeing(&oi->ip_data_lockres); 1017 ocfs2_mark_lockres_freeing(&oi->ip_data_lockres);
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index 8d81f6c1b877..f2ebe2eb3c21 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -44,7 +44,6 @@
44#include "localalloc.h" 44#include "localalloc.h"
45#include "slot_map.h" 45#include "slot_map.h"
46#include "super.h" 46#include "super.h"
47#include "vote.h"
48#include "sysfile.h" 47#include "sysfile.h"
49 48
50#include "buffer_head_io.h" 49#include "buffer_head_io.h"
@@ -103,7 +102,7 @@ static int ocfs2_commit_cache(struct ocfs2_super *osb)
103 mlog(0, "commit_thread: flushed transaction %lu (%u handles)\n", 102 mlog(0, "commit_thread: flushed transaction %lu (%u handles)\n",
104 journal->j_trans_id, flushed); 103 journal->j_trans_id, flushed);
105 104
106 ocfs2_kick_vote_thread(osb); 105 ocfs2_wake_downconvert_thread(osb);
107 wake_up(&journal->j_checkpointed); 106 wake_up(&journal->j_checkpointed);
108finally: 107finally:
109 mlog_exit(status); 108 mlog_exit(status);
@@ -883,8 +882,8 @@ restart:
883 ocfs2_super_unlock(osb, 1); 882 ocfs2_super_unlock(osb, 1);
884 883
885 /* We always run recovery on our own orphan dir - the dead 884 /* We always run recovery on our own orphan dir - the dead
886 * node(s) may have voted "no" on an inode delete earlier. A 885 * node(s) may have disallowd a previos inode delete. Re-processing
887 * revote is therefore required. */ 886 * is therefore required. */
888 ocfs2_queue_recovery_completion(osb->journal, osb->slot_num, NULL, 887 ocfs2_queue_recovery_completion(osb->journal, osb->slot_num, NULL,
889 NULL); 888 NULL);
890 889
@@ -1380,10 +1379,10 @@ static int ocfs2_recover_orphans(struct ocfs2_super *osb,
1380 iter = oi->ip_next_orphan; 1379 iter = oi->ip_next_orphan;
1381 1380
1382 spin_lock(&oi->ip_lock); 1381 spin_lock(&oi->ip_lock);
1383 /* Delete voting may have set these on the assumption 1382 /* The remote delete code may have set these on the
1384 * that the other node would wipe them successfully. 1383 * assumption that the other node would wipe them
1385 * If they are still in the node's orphan dir, we need 1384 * successfully. If they are still in the node's
1386 * to reset that state. */ 1385 * orphan dir, we need to reset that state. */
1387 oi->ip_flags &= ~(OCFS2_INODE_DELETED|OCFS2_INODE_SKIP_DELETE); 1386 oi->ip_flags &= ~(OCFS2_INODE_DELETED|OCFS2_INODE_SKIP_DELETE);
1388 1387
1389 /* Set the proper information to get us going into 1388 /* Set the proper information to get us going into
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index 989ac2718587..6295fd6ae469 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -60,7 +60,6 @@
60#include "symlink.h" 60#include "symlink.h"
61#include "sysfile.h" 61#include "sysfile.h"
62#include "uptodate.h" 62#include "uptodate.h"
63#include "vote.h"
64 63
65#include "buffer_head_io.h" 64#include "buffer_head_io.h"
66 65
@@ -176,7 +175,7 @@ bail_unlock:
176 /* Don't drop the cluster lock until *after* the d_add -- 175 /* Don't drop the cluster lock until *after* the d_add --
177 * unlink on another node will message us to remove that 176 * unlink on another node will message us to remove that
178 * dentry under this lock so otherwise we can race this with 177 * dentry under this lock so otherwise we can race this with
179 * the vote thread and have a stale dentry. */ 178 * the downconvert thread and have a stale dentry. */
180 ocfs2_meta_unlock(dir, 0); 179 ocfs2_meta_unlock(dir, 0);
181 180
182bail: 181bail:
@@ -765,7 +764,7 @@ static int ocfs2_unlink(struct inode *dir,
765 764
766 status = ocfs2_remote_dentry_delete(dentry); 765 status = ocfs2_remote_dentry_delete(dentry);
767 if (status < 0) { 766 if (status < 0) {
768 /* This vote should succeed under all normal 767 /* This remote delete should succeed under all normal
769 * circumstances. */ 768 * circumstances. */
770 mlog_errno(status); 769 mlog_errno(status);
771 goto leave; 770 goto leave;
@@ -1031,8 +1030,9 @@ static int ocfs2_rename(struct inode *old_dir,
1031 1030
1032 /* 1031 /*
1033 * Aside from allowing a meta data update, the locking here 1032 * Aside from allowing a meta data update, the locking here
1034 * also ensures that the vote thread on other nodes won't have 1033 * also ensures that the downconvert thread on other nodes
1035 * to concurrently downconvert the inode and the dentry locks. 1034 * won't have to concurrently downconvert the inode and the
1035 * dentry locks.
1036 */ 1036 */
1037 status = ocfs2_meta_lock(old_inode, &old_inode_bh, 1); 1037 status = ocfs2_meta_lock(old_inode, &old_inode_bh, 1);
1038 if (status < 0) { 1038 if (status < 0) {
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index 60a23e1906b0..f8f866144c6a 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -189,9 +189,7 @@ struct ocfs2_super
189 struct ocfs2_slot_info *slot_info; 189 struct ocfs2_slot_info *slot_info;
190 190
191 spinlock_t node_map_lock; 191 spinlock_t node_map_lock;
192 struct ocfs2_node_map mounted_map;
193 struct ocfs2_node_map recovery_map; 192 struct ocfs2_node_map recovery_map;
194 struct ocfs2_node_map umount_map;
195 193
196 u64 root_blkno; 194 u64 root_blkno;
197 u64 system_dir_blkno; 195 u64 system_dir_blkno;
@@ -254,28 +252,15 @@ struct ocfs2_super
254 252
255 wait_queue_head_t recovery_event; 253 wait_queue_head_t recovery_event;
256 254
257 spinlock_t vote_task_lock; 255 spinlock_t dc_task_lock;
258 struct task_struct *vote_task; 256 struct task_struct *dc_task;
259 wait_queue_head_t vote_event; 257 wait_queue_head_t dc_event;
260 unsigned long vote_wake_sequence; 258 unsigned long dc_wake_sequence;
261 unsigned long vote_work_sequence; 259 unsigned long dc_work_sequence;
262 260
263 struct list_head blocked_lock_list; 261 struct list_head blocked_lock_list;
264 unsigned long blocked_lock_count; 262 unsigned long blocked_lock_count;
265 263
266 struct list_head vote_list;
267 int vote_count;
268
269 u32 net_key;
270 spinlock_t net_response_lock;
271 unsigned int net_response_ids;
272 struct list_head net_response_list;
273
274 struct o2hb_callback_func osb_hb_up;
275 struct o2hb_callback_func osb_hb_down;
276
277 struct list_head osb_net_handlers;
278
279 wait_queue_head_t osb_mount_event; 264 wait_queue_head_t osb_mount_event;
280 265
281 /* Truncate log info */ 266 /* Truncate log info */
diff --git a/fs/ocfs2/slot_map.c b/fs/ocfs2/slot_map.c
index af4882b62cfa..3a50ce555e64 100644
--- a/fs/ocfs2/slot_map.c
+++ b/fs/ocfs2/slot_map.c
@@ -48,25 +48,6 @@ static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
48 s16 slot_num, 48 s16 slot_num,
49 s16 node_num); 49 s16 node_num);
50 50
51/* Use the slot information we've collected to create a map of mounted
52 * nodes. Should be holding an EX on super block. assumes slot info is
53 * up to date. Note that we call this *after* we find a slot, so our
54 * own node should be set in the map too... */
55void ocfs2_populate_mounted_map(struct ocfs2_super *osb)
56{
57 int i;
58 struct ocfs2_slot_info *si = osb->slot_info;
59
60 spin_lock(&si->si_lock);
61
62 for (i = 0; i < si->si_size; i++)
63 if (si->si_global_node_nums[i] != OCFS2_INVALID_SLOT)
64 ocfs2_node_map_set_bit(osb, &osb->mounted_map,
65 si->si_global_node_nums[i]);
66
67 spin_unlock(&si->si_lock);
68}
69
70/* post the slot information on disk into our slot_info struct. */ 51/* post the slot information on disk into our slot_info struct. */
71void ocfs2_update_slot_info(struct ocfs2_slot_info *si) 52void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
72{ 53{
diff --git a/fs/ocfs2/slot_map.h b/fs/ocfs2/slot_map.h
index d8c8ceed031b..1025872aaade 100644
--- a/fs/ocfs2/slot_map.h
+++ b/fs/ocfs2/slot_map.h
@@ -52,8 +52,6 @@ s16 ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
52void ocfs2_clear_slot(struct ocfs2_slot_info *si, 52void ocfs2_clear_slot(struct ocfs2_slot_info *si,
53 s16 slot_num); 53 s16 slot_num);
54 54
55void ocfs2_populate_mounted_map(struct ocfs2_super *osb);
56
57static inline int ocfs2_is_empty_slot(struct ocfs2_slot_info *si, 55static inline int ocfs2_is_empty_slot(struct ocfs2_slot_info *si,
58 int slot_num) 56 int slot_num)
59{ 57{
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index 64b81b341ece..1996820488cc 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -65,7 +65,6 @@
65#include "sysfile.h" 65#include "sysfile.h"
66#include "uptodate.h" 66#include "uptodate.h"
67#include "ver.h" 67#include "ver.h"
68#include "vote.h"
69 68
70#include "buffer_head_io.h" 69#include "buffer_head_io.h"
71 70
@@ -1123,13 +1122,6 @@ static int ocfs2_mount_volume(struct super_block *sb)
1123 goto leave; 1122 goto leave;
1124 } 1123 }
1125 1124
1126 /* requires vote_thread to be running. */
1127 status = ocfs2_register_net_handlers(osb);
1128 if (status < 0) {
1129 mlog_errno(status);
1130 goto leave;
1131 }
1132
1133 status = ocfs2_super_lock(osb, 1); 1125 status = ocfs2_super_lock(osb, 1);
1134 if (status < 0) { 1126 if (status < 0) {
1135 mlog_errno(status); 1127 mlog_errno(status);
@@ -1144,8 +1136,6 @@ static int ocfs2_mount_volume(struct super_block *sb)
1144 goto leave; 1136 goto leave;
1145 } 1137 }
1146 1138
1147 ocfs2_populate_mounted_map(osb);
1148
1149 /* load all node-local system inodes */ 1139 /* load all node-local system inodes */
1150 status = ocfs2_init_local_system_inodes(osb); 1140 status = ocfs2_init_local_system_inodes(osb);
1151 if (status < 0) { 1141 if (status < 0) {
@@ -1168,15 +1158,6 @@ static int ocfs2_mount_volume(struct super_block *sb)
1168 if (ocfs2_mount_local(osb)) 1158 if (ocfs2_mount_local(osb))
1169 goto leave; 1159 goto leave;
1170 1160
1171 /* This should be sent *after* we recovered our journal as it
1172 * will cause other nodes to unmark us as needing
1173 * recovery. However, we need to send it *before* dropping the
1174 * super block lock as otherwise their recovery threads might
1175 * try to clean us up while we're live! */
1176 status = ocfs2_request_mount_vote(osb);
1177 if (status < 0)
1178 mlog_errno(status);
1179
1180leave: 1161leave:
1181 if (unlock_super) 1162 if (unlock_super)
1182 ocfs2_super_unlock(osb, 1); 1163 ocfs2_super_unlock(osb, 1);
@@ -1234,10 +1215,6 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1234 mlog_errno(tmp); 1215 mlog_errno(tmp);
1235 return; 1216 return;
1236 } 1217 }
1237
1238 tmp = ocfs2_request_umount_vote(osb);
1239 if (tmp < 0)
1240 mlog_errno(tmp);
1241 } 1218 }
1242 1219
1243 if (osb->slot_num != OCFS2_INVALID_SLOT) 1220 if (osb->slot_num != OCFS2_INVALID_SLOT)
@@ -1248,11 +1225,8 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1248 1225
1249 ocfs2_release_system_inodes(osb); 1226 ocfs2_release_system_inodes(osb);
1250 1227
1251 if (osb->dlm) { 1228 if (osb->dlm)
1252 ocfs2_unregister_net_handlers(osb);
1253
1254 ocfs2_dlm_shutdown(osb); 1229 ocfs2_dlm_shutdown(osb);
1255 }
1256 1230
1257 debugfs_remove(osb->osb_debug_root); 1231 debugfs_remove(osb->osb_debug_root);
1258 1232
@@ -1336,19 +1310,13 @@ static int ocfs2_initialize_super(struct super_block *sb,
1336 osb->s_sectsize_bits = blksize_bits(sector_size); 1310 osb->s_sectsize_bits = blksize_bits(sector_size);
1337 BUG_ON(!osb->s_sectsize_bits); 1311 BUG_ON(!osb->s_sectsize_bits);
1338 1312
1339 osb->net_response_ids = 0;
1340 spin_lock_init(&osb->net_response_lock);
1341 INIT_LIST_HEAD(&osb->net_response_list);
1342
1343 INIT_LIST_HEAD(&osb->osb_net_handlers);
1344 init_waitqueue_head(&osb->recovery_event); 1313 init_waitqueue_head(&osb->recovery_event);
1345 spin_lock_init(&osb->vote_task_lock); 1314 spin_lock_init(&osb->dc_task_lock);
1346 init_waitqueue_head(&osb->vote_event); 1315 init_waitqueue_head(&osb->dc_event);
1347 osb->vote_work_sequence = 0; 1316 osb->dc_work_sequence = 0;
1348 osb->vote_wake_sequence = 0; 1317 osb->dc_wake_sequence = 0;
1349 INIT_LIST_HEAD(&osb->blocked_lock_list); 1318 INIT_LIST_HEAD(&osb->blocked_lock_list);
1350 osb->blocked_lock_count = 0; 1319 osb->blocked_lock_count = 0;
1351 INIT_LIST_HEAD(&osb->vote_list);
1352 spin_lock_init(&osb->osb_lock); 1320 spin_lock_init(&osb->osb_lock);
1353 1321
1354 atomic_set(&osb->alloc_stats.moves, 0); 1322 atomic_set(&osb->alloc_stats.moves, 0);
@@ -1488,7 +1456,6 @@ static int ocfs2_initialize_super(struct super_block *sb,
1488 } 1456 }
1489 1457
1490 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); 1458 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1491 osb->net_key = le32_to_cpu(uuid_net_key);
1492 1459
1493 strncpy(osb->vol_label, di->id2.i_super.s_label, 63); 1460 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1494 osb->vol_label[63] = '\0'; 1461 osb->vol_label[63] = '\0';
diff --git a/fs/ocfs2/vote.c b/fs/ocfs2/vote.c
deleted file mode 100644
index c05358538f2b..000000000000
--- a/fs/ocfs2/vote.c
+++ /dev/null
@@ -1,756 +0,0 @@
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * vote.c
5 *
6 * description here
7 *
8 * Copyright (C) 2003, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#include <linux/types.h>
27#include <linux/slab.h>
28#include <linux/highmem.h>
29#include <linux/kthread.h>
30
31#include <cluster/heartbeat.h>
32#include <cluster/nodemanager.h>
33#include <cluster/tcp.h>
34
35#include <dlm/dlmapi.h>
36
37#define MLOG_MASK_PREFIX ML_VOTE
38#include <cluster/masklog.h>
39
40#include "ocfs2.h"
41
42#include "alloc.h"
43#include "dlmglue.h"
44#include "extent_map.h"
45#include "heartbeat.h"
46#include "inode.h"
47#include "journal.h"
48#include "slot_map.h"
49#include "vote.h"
50
51#include "buffer_head_io.h"
52
53#define OCFS2_MESSAGE_TYPE_VOTE (0x1)
54#define OCFS2_MESSAGE_TYPE_RESPONSE (0x2)
55struct ocfs2_msg_hdr
56{
57 __be32 h_response_id; /* used to lookup message handle on sending
58 * node. */
59 __be32 h_request;
60 __be64 h_blkno;
61 __be32 h_generation;
62 __be32 h_node_num; /* node sending this particular message. */
63};
64
65struct ocfs2_vote_msg
66{
67 struct ocfs2_msg_hdr v_hdr;
68 __be32 v_reserved1;
69} __attribute__ ((packed));
70
71/* Responses are given these values to maintain backwards
72 * compatibility with older ocfs2 versions */
73#define OCFS2_RESPONSE_OK (0)
74#define OCFS2_RESPONSE_BUSY (-16)
75#define OCFS2_RESPONSE_BAD_MSG (-22)
76
77struct ocfs2_response_msg
78{
79 struct ocfs2_msg_hdr r_hdr;
80 __be32 r_response;
81} __attribute__ ((packed));
82
83struct ocfs2_vote_work {
84 struct list_head w_list;
85 struct ocfs2_vote_msg w_msg;
86};
87
88enum ocfs2_vote_request {
89 OCFS2_VOTE_REQ_INVALID = 0,
90 OCFS2_VOTE_REQ_MOUNT,
91 OCFS2_VOTE_REQ_UMOUNT,
92 OCFS2_VOTE_REQ_LAST
93};
94
95static inline int ocfs2_is_valid_vote_request(int request)
96{
97 return OCFS2_VOTE_REQ_INVALID < request &&
98 request < OCFS2_VOTE_REQ_LAST;
99}
100
101typedef void (*ocfs2_net_response_callback)(void *priv,
102 struct ocfs2_response_msg *resp);
103struct ocfs2_net_response_cb {
104 ocfs2_net_response_callback rc_cb;
105 void *rc_priv;
106};
107
108struct ocfs2_net_wait_ctxt {
109 struct list_head n_list;
110 u32 n_response_id;
111 wait_queue_head_t n_event;
112 struct ocfs2_node_map n_node_map;
113 int n_response; /* an agreggate response. 0 if
114 * all nodes are go, < 0 on any
115 * negative response from any
116 * node or network error. */
117 struct ocfs2_net_response_cb *n_callback;
118};
119
120static void ocfs2_process_mount_request(struct ocfs2_super *osb,
121 unsigned int node_num)
122{
123 mlog(0, "MOUNT vote from node %u\n", node_num);
124 /* The other node only sends us this message when he has an EX
125 * on the superblock, so our recovery threads (if having been
126 * launched) are waiting on it.*/
127 ocfs2_recovery_map_clear(osb, node_num);
128 ocfs2_node_map_set_bit(osb, &osb->mounted_map, node_num);
129
130 /* We clear the umount map here because a node may have been
131 * previously mounted, safely unmounted but never stopped
132 * heartbeating - in which case we'd have a stale entry. */
133 ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
134}
135
136static void ocfs2_process_umount_request(struct ocfs2_super *osb,
137 unsigned int node_num)
138{
139 mlog(0, "UMOUNT vote from node %u\n", node_num);
140 ocfs2_node_map_clear_bit(osb, &osb->mounted_map, node_num);
141 ocfs2_node_map_set_bit(osb, &osb->umount_map, node_num);
142}
143
144static void ocfs2_process_vote(struct ocfs2_super *osb,
145 struct ocfs2_vote_msg *msg)
146{
147 int net_status, vote_response;
148 unsigned int node_num;
149 u64 blkno;
150 enum ocfs2_vote_request request;
151 struct ocfs2_msg_hdr *hdr = &msg->v_hdr;
152 struct ocfs2_response_msg response;
153
154 /* decode the network mumbo jumbo into local variables. */
155 request = be32_to_cpu(hdr->h_request);
156 blkno = be64_to_cpu(hdr->h_blkno);
157 node_num = be32_to_cpu(hdr->h_node_num);
158
159 mlog(0, "processing vote: request = %u, blkno = %llu, node_num = %u\n",
160 request, (unsigned long long)blkno, node_num);
161
162 if (!ocfs2_is_valid_vote_request(request)) {
163 mlog(ML_ERROR, "Invalid vote request %d from node %u\n",
164 request, node_num);
165 vote_response = OCFS2_RESPONSE_BAD_MSG;
166 goto respond;
167 }
168
169 vote_response = OCFS2_RESPONSE_OK;
170
171 switch (request) {
172 case OCFS2_VOTE_REQ_UMOUNT:
173 ocfs2_process_umount_request(osb, node_num);
174 goto respond;
175 case OCFS2_VOTE_REQ_MOUNT:
176 ocfs2_process_mount_request(osb, node_num);
177 goto respond;
178 default:
179 /* avoids a gcc warning */
180 break;
181 }
182
183respond:
184 /* Response struture is small so we just put it on the stack
185 * and stuff it inline. */
186 memset(&response, 0, sizeof(struct ocfs2_response_msg));
187 response.r_hdr.h_response_id = hdr->h_response_id;
188 response.r_hdr.h_blkno = hdr->h_blkno;
189 response.r_hdr.h_generation = hdr->h_generation;
190 response.r_hdr.h_node_num = cpu_to_be32(osb->node_num);
191 response.r_response = cpu_to_be32(vote_response);
192
193 net_status = o2net_send_message(OCFS2_MESSAGE_TYPE_RESPONSE,
194 osb->net_key,
195 &response,
196 sizeof(struct ocfs2_response_msg),
197 node_num,
198 NULL);
199 /* We still want to error print for ENOPROTOOPT here. The
200 * sending node shouldn't have unregistered his net handler
201 * without sending an unmount vote 1st */
202 if (net_status < 0
203 && net_status != -ETIMEDOUT
204 && net_status != -ENOTCONN)
205 mlog(ML_ERROR, "message to node %u fails with error %d!\n",
206 node_num, net_status);
207}
208
209static void ocfs2_vote_thread_do_work(struct ocfs2_super *osb)
210{
211 unsigned long processed;
212 struct ocfs2_lock_res *lockres;
213 struct ocfs2_vote_work *work;
214
215 mlog_entry_void();
216
217 spin_lock(&osb->vote_task_lock);
218 /* grab this early so we know to try again if a state change and
219 * wake happens part-way through our work */
220 osb->vote_work_sequence = osb->vote_wake_sequence;
221
222 processed = osb->blocked_lock_count;
223 while (processed) {
224 BUG_ON(list_empty(&osb->blocked_lock_list));
225
226 lockres = list_entry(osb->blocked_lock_list.next,
227 struct ocfs2_lock_res, l_blocked_list);
228 list_del_init(&lockres->l_blocked_list);
229 osb->blocked_lock_count--;
230 spin_unlock(&osb->vote_task_lock);
231
232 BUG_ON(!processed);
233 processed--;
234
235 ocfs2_process_blocked_lock(osb, lockres);
236
237 spin_lock(&osb->vote_task_lock);
238 }
239
240 while (osb->vote_count) {
241 BUG_ON(list_empty(&osb->vote_list));
242 work = list_entry(osb->vote_list.next,
243 struct ocfs2_vote_work, w_list);
244 list_del(&work->w_list);
245 osb->vote_count--;
246 spin_unlock(&osb->vote_task_lock);
247
248 ocfs2_process_vote(osb, &work->w_msg);
249 kfree(work);
250
251 spin_lock(&osb->vote_task_lock);
252 }
253 spin_unlock(&osb->vote_task_lock);
254
255 mlog_exit_void();
256}
257
258static int ocfs2_vote_thread_lists_empty(struct ocfs2_super *osb)
259{
260 int empty = 0;
261
262 spin_lock(&osb->vote_task_lock);
263 if (list_empty(&osb->blocked_lock_list) &&
264 list_empty(&osb->vote_list))
265 empty = 1;
266
267 spin_unlock(&osb->vote_task_lock);
268 return empty;
269}
270
271static int ocfs2_vote_thread_should_wake(struct ocfs2_super *osb)
272{
273 int should_wake = 0;
274
275 spin_lock(&osb->vote_task_lock);
276 if (osb->vote_work_sequence != osb->vote_wake_sequence)
277 should_wake = 1;
278 spin_unlock(&osb->vote_task_lock);
279
280 return should_wake;
281}
282
283int ocfs2_vote_thread(void *arg)
284{
285 int status = 0;
286 struct ocfs2_super *osb = arg;
287
288 /* only quit once we've been asked to stop and there is no more
289 * work available */
290 while (!(kthread_should_stop() &&
291 ocfs2_vote_thread_lists_empty(osb))) {
292
293 wait_event_interruptible(osb->vote_event,
294 ocfs2_vote_thread_should_wake(osb) ||
295 kthread_should_stop());
296
297 mlog(0, "vote_thread: awoken\n");
298
299 ocfs2_vote_thread_do_work(osb);
300 }
301
302 osb->vote_task = NULL;
303 return status;
304}
305
306static struct ocfs2_net_wait_ctxt *ocfs2_new_net_wait_ctxt(unsigned int response_id)
307{
308 struct ocfs2_net_wait_ctxt *w;
309
310 w = kzalloc(sizeof(*w), GFP_NOFS);
311 if (!w) {
312 mlog_errno(-ENOMEM);
313 goto bail;
314 }
315
316 INIT_LIST_HEAD(&w->n_list);
317 init_waitqueue_head(&w->n_event);
318 ocfs2_node_map_init(&w->n_node_map);
319 w->n_response_id = response_id;
320 w->n_callback = NULL;
321bail:
322 return w;
323}
324
325static unsigned int ocfs2_new_response_id(struct ocfs2_super *osb)
326{
327 unsigned int ret;
328
329 spin_lock(&osb->net_response_lock);
330 ret = ++osb->net_response_ids;
331 spin_unlock(&osb->net_response_lock);
332
333 return ret;
334}
335
336static void ocfs2_dequeue_net_wait_ctxt(struct ocfs2_super *osb,
337 struct ocfs2_net_wait_ctxt *w)
338{
339 spin_lock(&osb->net_response_lock);
340 list_del(&w->n_list);
341 spin_unlock(&osb->net_response_lock);
342}
343
344static void ocfs2_queue_net_wait_ctxt(struct ocfs2_super *osb,
345 struct ocfs2_net_wait_ctxt *w)
346{
347 spin_lock(&osb->net_response_lock);
348 list_add_tail(&w->n_list,
349 &osb->net_response_list);
350 spin_unlock(&osb->net_response_lock);
351}
352
353static void __ocfs2_mark_node_responded(struct ocfs2_super *osb,
354 struct ocfs2_net_wait_ctxt *w,
355 int node_num)
356{
357 assert_spin_locked(&osb->net_response_lock);
358
359 ocfs2_node_map_clear_bit(osb, &w->n_node_map, node_num);
360 if (ocfs2_node_map_is_empty(osb, &w->n_node_map))
361 wake_up(&w->n_event);
362}
363
364/* Intended to be called from the node down callback, we fake remove
365 * the node from all our response contexts */
366void ocfs2_remove_node_from_vote_queues(struct ocfs2_super *osb,
367 int node_num)
368{
369 struct list_head *p;
370 struct ocfs2_net_wait_ctxt *w = NULL;
371
372 spin_lock(&osb->net_response_lock);
373
374 list_for_each(p, &osb->net_response_list) {
375 w = list_entry(p, struct ocfs2_net_wait_ctxt, n_list);
376
377 __ocfs2_mark_node_responded(osb, w, node_num);
378 }
379
380 spin_unlock(&osb->net_response_lock);
381}
382
383static int ocfs2_broadcast_vote(struct ocfs2_super *osb,
384 struct ocfs2_vote_msg *request,
385 unsigned int response_id,
386 int *response,
387 struct ocfs2_net_response_cb *callback)
388{
389 int status, i, remote_err;
390 struct ocfs2_net_wait_ctxt *w = NULL;
391 int dequeued = 0;
392
393 mlog_entry_void();
394
395 w = ocfs2_new_net_wait_ctxt(response_id);
396 if (!w) {
397 status = -ENOMEM;
398 mlog_errno(status);
399 goto bail;
400 }
401 w->n_callback = callback;
402
403 /* we're pretty much ready to go at this point, and this fills
404 * in n_response which we need anyway... */
405 ocfs2_queue_net_wait_ctxt(osb, w);
406
407 i = ocfs2_node_map_iterate(osb, &osb->mounted_map, 0);
408
409 while (i != O2NM_INVALID_NODE_NUM) {
410 if (i != osb->node_num) {
411 mlog(0, "trying to send request to node %i\n", i);
412 ocfs2_node_map_set_bit(osb, &w->n_node_map, i);
413
414 remote_err = 0;
415 status = o2net_send_message(OCFS2_MESSAGE_TYPE_VOTE,
416 osb->net_key,
417 request,
418 sizeof(*request),
419 i,
420 &remote_err);
421 if (status == -ETIMEDOUT) {
422 mlog(0, "remote node %d timed out!\n", i);
423 status = -EAGAIN;
424 goto bail;
425 }
426 if (remote_err < 0) {
427 status = remote_err;
428 mlog(0, "remote error %d on node %d!\n",
429 remote_err, i);
430 mlog_errno(status);
431 goto bail;
432 }
433 if (status < 0) {
434 mlog_errno(status);
435 goto bail;
436 }
437 }
438 i++;
439 i = ocfs2_node_map_iterate(osb, &osb->mounted_map, i);
440 mlog(0, "next is %d, i am %d\n", i, osb->node_num);
441 }
442 mlog(0, "done sending, now waiting on responses...\n");
443
444 wait_event(w->n_event, ocfs2_node_map_is_empty(osb, &w->n_node_map));
445
446 ocfs2_dequeue_net_wait_ctxt(osb, w);
447 dequeued = 1;
448
449 *response = w->n_response;
450 status = 0;
451bail:
452 if (w) {
453 if (!dequeued)
454 ocfs2_dequeue_net_wait_ctxt(osb, w);
455 kfree(w);
456 }
457
458 mlog_exit(status);
459 return status;
460}
461
462static struct ocfs2_vote_msg * ocfs2_new_vote_request(struct ocfs2_super *osb,
463 u64 blkno,
464 unsigned int generation,
465 enum ocfs2_vote_request type)
466{
467 struct ocfs2_vote_msg *request;
468 struct ocfs2_msg_hdr *hdr;
469
470 BUG_ON(!ocfs2_is_valid_vote_request(type));
471
472 request = kzalloc(sizeof(*request), GFP_NOFS);
473 if (!request) {
474 mlog_errno(-ENOMEM);
475 } else {
476 hdr = &request->v_hdr;
477 hdr->h_node_num = cpu_to_be32(osb->node_num);
478 hdr->h_request = cpu_to_be32(type);
479 hdr->h_blkno = cpu_to_be64(blkno);
480 hdr->h_generation = cpu_to_be32(generation);
481 }
482
483 return request;
484}
485
486/* Complete the buildup of a new vote request and process the
487 * broadcast return value. */
488static int ocfs2_do_request_vote(struct ocfs2_super *osb,
489 struct ocfs2_vote_msg *request,
490 struct ocfs2_net_response_cb *callback)
491{
492 int status, response = -EBUSY;
493 unsigned int response_id;
494 struct ocfs2_msg_hdr *hdr;
495
496 response_id = ocfs2_new_response_id(osb);
497
498 hdr = &request->v_hdr;
499 hdr->h_response_id = cpu_to_be32(response_id);
500
501 status = ocfs2_broadcast_vote(osb, request, response_id, &response,
502 callback);
503 if (status < 0) {
504 mlog_errno(status);
505 goto bail;
506 }
507
508 status = response;
509bail:
510
511 return status;
512}
513
514int ocfs2_request_mount_vote(struct ocfs2_super *osb)
515{
516 int status;
517 struct ocfs2_vote_msg *request = NULL;
518
519 request = ocfs2_new_vote_request(osb, 0ULL, 0, OCFS2_VOTE_REQ_MOUNT);
520 if (!request) {
521 status = -ENOMEM;
522 goto bail;
523 }
524
525 status = -EAGAIN;
526 while (status == -EAGAIN) {
527 if (!(osb->s_mount_opt & OCFS2_MOUNT_NOINTR) &&
528 signal_pending(current)) {
529 status = -ERESTARTSYS;
530 goto bail;
531 }
532
533 if (ocfs2_node_map_is_only(osb, &osb->mounted_map,
534 osb->node_num)) {
535 status = 0;
536 goto bail;
537 }
538
539 status = ocfs2_do_request_vote(osb, request, NULL);
540 }
541
542bail:
543 kfree(request);
544 return status;
545}
546
547int ocfs2_request_umount_vote(struct ocfs2_super *osb)
548{
549 int status;
550 struct ocfs2_vote_msg *request = NULL;
551
552 request = ocfs2_new_vote_request(osb, 0ULL, 0, OCFS2_VOTE_REQ_UMOUNT);
553 if (!request) {
554 status = -ENOMEM;
555 goto bail;
556 }
557
558 status = -EAGAIN;
559 while (status == -EAGAIN) {
560 /* Do not check signals on this vote... We really want
561 * this one to go all the way through. */
562
563 if (ocfs2_node_map_is_only(osb, &osb->mounted_map,
564 osb->node_num)) {
565 status = 0;
566 goto bail;
567 }
568
569 status = ocfs2_do_request_vote(osb, request, NULL);
570 }
571
572bail:
573 kfree(request);
574 return status;
575}
576
577/* TODO: This should eventually be a hash table! */
578static struct ocfs2_net_wait_ctxt * __ocfs2_find_net_wait_ctxt(struct ocfs2_super *osb,
579 u32 response_id)
580{
581 struct list_head *p;
582 struct ocfs2_net_wait_ctxt *w = NULL;
583
584 list_for_each(p, &osb->net_response_list) {
585 w = list_entry(p, struct ocfs2_net_wait_ctxt, n_list);
586 if (response_id == w->n_response_id)
587 break;
588 w = NULL;
589 }
590
591 return w;
592}
593
594/* Translate response codes into local node errno values */
595static inline int ocfs2_translate_response(int response)
596{
597 int ret;
598
599 switch (response) {
600 case OCFS2_RESPONSE_OK:
601 ret = 0;
602 break;
603
604 case OCFS2_RESPONSE_BUSY:
605 ret = -EBUSY;
606 break;
607
608 default:
609 ret = -EINVAL;
610 }
611
612 return ret;
613}
614
615static int ocfs2_handle_response_message(struct o2net_msg *msg,
616 u32 len,
617 void *data, void **ret_data)
618{
619 unsigned int response_id, node_num;
620 int response_status;
621 struct ocfs2_super *osb = data;
622 struct ocfs2_response_msg *resp;
623 struct ocfs2_net_wait_ctxt * w;
624 struct ocfs2_net_response_cb *resp_cb;
625
626 resp = (struct ocfs2_response_msg *) msg->buf;
627
628 response_id = be32_to_cpu(resp->r_hdr.h_response_id);
629 node_num = be32_to_cpu(resp->r_hdr.h_node_num);
630 response_status =
631 ocfs2_translate_response(be32_to_cpu(resp->r_response));
632
633 mlog(0, "received response message:\n");
634 mlog(0, "h_response_id = %u\n", response_id);
635 mlog(0, "h_request = %u\n", be32_to_cpu(resp->r_hdr.h_request));
636 mlog(0, "h_blkno = %llu\n",
637 (unsigned long long)be64_to_cpu(resp->r_hdr.h_blkno));
638 mlog(0, "h_generation = %u\n", be32_to_cpu(resp->r_hdr.h_generation));
639 mlog(0, "h_node_num = %u\n", node_num);
640 mlog(0, "r_response = %d\n", response_status);
641
642 spin_lock(&osb->net_response_lock);
643 w = __ocfs2_find_net_wait_ctxt(osb, response_id);
644 if (!w) {
645 mlog(0, "request not found!\n");
646 goto bail;
647 }
648 resp_cb = w->n_callback;
649
650 if (response_status && (!w->n_response)) {
651 /* we only really need one negative response so don't
652 * set it twice. */
653 w->n_response = response_status;
654 }
655
656 if (resp_cb) {
657 spin_unlock(&osb->net_response_lock);
658
659 resp_cb->rc_cb(resp_cb->rc_priv, resp);
660
661 spin_lock(&osb->net_response_lock);
662 }
663
664 __ocfs2_mark_node_responded(osb, w, node_num);
665bail:
666 spin_unlock(&osb->net_response_lock);
667
668 return 0;
669}
670
671static int ocfs2_handle_vote_message(struct o2net_msg *msg,
672 u32 len,
673 void *data, void **ret_data)
674{
675 int status;
676 struct ocfs2_super *osb = data;
677 struct ocfs2_vote_work *work;
678
679 work = kmalloc(sizeof(struct ocfs2_vote_work), GFP_NOFS);
680 if (!work) {
681 status = -ENOMEM;
682 mlog_errno(status);
683 goto bail;
684 }
685
686 INIT_LIST_HEAD(&work->w_list);
687 memcpy(&work->w_msg, msg->buf, sizeof(struct ocfs2_vote_msg));
688
689 mlog(0, "scheduling vote request:\n");
690 mlog(0, "h_response_id = %u\n",
691 be32_to_cpu(work->w_msg.v_hdr.h_response_id));
692 mlog(0, "h_request = %u\n", be32_to_cpu(work->w_msg.v_hdr.h_request));
693 mlog(0, "h_blkno = %llu\n",
694 (unsigned long long)be64_to_cpu(work->w_msg.v_hdr.h_blkno));
695 mlog(0, "h_generation = %u\n",
696 be32_to_cpu(work->w_msg.v_hdr.h_generation));
697 mlog(0, "h_node_num = %u\n",
698 be32_to_cpu(work->w_msg.v_hdr.h_node_num));
699
700 spin_lock(&osb->vote_task_lock);
701 list_add_tail(&work->w_list, &osb->vote_list);
702 osb->vote_count++;
703 spin_unlock(&osb->vote_task_lock);
704
705 ocfs2_kick_vote_thread(osb);
706
707 status = 0;
708bail:
709 return status;
710}
711
712void ocfs2_unregister_net_handlers(struct ocfs2_super *osb)
713{
714 if (!osb->net_key)
715 return;
716
717 o2net_unregister_handler_list(&osb->osb_net_handlers);
718
719 if (!list_empty(&osb->net_response_list))
720 mlog(ML_ERROR, "net response list not empty!\n");
721
722 osb->net_key = 0;
723}
724
725int ocfs2_register_net_handlers(struct ocfs2_super *osb)
726{
727 int status = 0;
728
729 if (ocfs2_mount_local(osb))
730 return 0;
731
732 status = o2net_register_handler(OCFS2_MESSAGE_TYPE_RESPONSE,
733 osb->net_key,
734 sizeof(struct ocfs2_response_msg),
735 ocfs2_handle_response_message,
736 osb, NULL, &osb->osb_net_handlers);
737 if (status) {
738 mlog_errno(status);
739 goto bail;
740 }
741
742 status = o2net_register_handler(OCFS2_MESSAGE_TYPE_VOTE,
743 osb->net_key,
744 sizeof(struct ocfs2_vote_msg),
745 ocfs2_handle_vote_message,
746 osb, NULL, &osb->osb_net_handlers);
747 if (status) {
748 mlog_errno(status);
749 goto bail;
750 }
751bail:
752 if (status < 0)
753 ocfs2_unregister_net_handlers(osb);
754
755 return status;
756}
diff --git a/fs/ocfs2/vote.h b/fs/ocfs2/vote.h
deleted file mode 100644
index 9ea46f62de31..000000000000
--- a/fs/ocfs2/vote.h
+++ /dev/null
@@ -1,48 +0,0 @@
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * vote.h
5 *
6 * description here
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26
27#ifndef VOTE_H
28#define VOTE_H
29
30int ocfs2_vote_thread(void *arg);
31static inline void ocfs2_kick_vote_thread(struct ocfs2_super *osb)
32{
33 spin_lock(&osb->vote_task_lock);
34 /* make sure the voting thread gets a swipe at whatever changes
35 * the caller may have made to the voting state */
36 osb->vote_wake_sequence++;
37 spin_unlock(&osb->vote_task_lock);
38 wake_up(&osb->vote_event);
39}
40
41int ocfs2_request_mount_vote(struct ocfs2_super *osb);
42int ocfs2_request_umount_vote(struct ocfs2_super *osb);
43int ocfs2_register_net_handlers(struct ocfs2_super *osb);
44void ocfs2_unregister_net_handlers(struct ocfs2_super *osb);
45
46void ocfs2_remove_node_from_vote_queues(struct ocfs2_super *osb,
47 int node_num);
48#endif
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-10 11:00:54 -0500