aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ocfs2/dlm
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ocfs2/dlm')
-rw-r--r--fs/ocfs2/dlm/Makefile2
-rw-r--r--fs/ocfs2/dlm/dlmcommon.h70
-rw-r--r--fs/ocfs2/dlm/dlmconvert.c2
-rw-r--r--fs/ocfs2/dlm/dlmdebug.c911
-rw-r--r--fs/ocfs2/dlm/dlmdebug.h86
-rw-r--r--fs/ocfs2/dlm/dlmdomain.c173
-rw-r--r--fs/ocfs2/dlm/dlmlock.c22
-rw-r--r--fs/ocfs2/dlm/dlmmaster.c224
-rw-r--r--fs/ocfs2/dlm/dlmrecovery.c57
-rw-r--r--fs/ocfs2/dlm/dlmthread.c6
10 files changed, 1270 insertions, 283 deletions
diff --git a/fs/ocfs2/dlm/Makefile b/fs/ocfs2/dlm/Makefile
index ce3f7c29d270..190361375700 100644
--- a/fs/ocfs2/dlm/Makefile
+++ b/fs/ocfs2/dlm/Makefile
@@ -1,6 +1,6 @@
1EXTRA_CFLAGS += -Ifs/ocfs2 1EXTRA_CFLAGS += -Ifs/ocfs2
2 2
3obj-$(CONFIG_OCFS2_FS) += ocfs2_dlm.o ocfs2_dlmfs.o 3obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_dlm.o ocfs2_dlmfs.o
4 4
5ocfs2_dlm-objs := dlmdomain.o dlmdebug.o dlmthread.o dlmrecovery.o \ 5ocfs2_dlm-objs := dlmdomain.o dlmdebug.o dlmthread.o dlmrecovery.o \
6 dlmmaster.o dlmast.o dlmconvert.o dlmlock.o dlmunlock.o dlmver.o 6 dlmmaster.o dlmast.o dlmconvert.o dlmlock.o dlmunlock.o dlmver.o
diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index 9843ee17ea27..d5a86fb81a49 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -49,6 +49,41 @@
49/* Intended to make it easier for us to switch out hash functions */ 49/* Intended to make it easier for us to switch out hash functions */
50#define dlm_lockid_hash(_n, _l) full_name_hash(_n, _l) 50#define dlm_lockid_hash(_n, _l) full_name_hash(_n, _l)
51 51
52enum dlm_mle_type {
53 DLM_MLE_BLOCK,
54 DLM_MLE_MASTER,
55 DLM_MLE_MIGRATION
56};
57
58struct dlm_lock_name {
59 u8 len;
60 u8 name[DLM_LOCKID_NAME_MAX];
61};
62
63struct dlm_master_list_entry {
64 struct list_head list;
65 struct list_head hb_events;
66 struct dlm_ctxt *dlm;
67 spinlock_t spinlock;
68 wait_queue_head_t wq;
69 atomic_t woken;
70 struct kref mle_refs;
71 int inuse;
72 unsigned long maybe_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
73 unsigned long vote_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
74 unsigned long response_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
75 unsigned long node_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
76 u8 master;
77 u8 new_master;
78 enum dlm_mle_type type;
79 struct o2hb_callback_func mle_hb_up;
80 struct o2hb_callback_func mle_hb_down;
81 union {
82 struct dlm_lock_resource *res;
83 struct dlm_lock_name name;
84 } u;
85};
86
52enum dlm_ast_type { 87enum dlm_ast_type {
53 DLM_AST = 0, 88 DLM_AST = 0,
54 DLM_BAST, 89 DLM_BAST,
@@ -101,6 +136,7 @@ struct dlm_ctxt
101 struct list_head purge_list; 136 struct list_head purge_list;
102 struct list_head pending_asts; 137 struct list_head pending_asts;
103 struct list_head pending_basts; 138 struct list_head pending_basts;
139 struct list_head tracking_list;
104 unsigned int purge_count; 140 unsigned int purge_count;
105 spinlock_t spinlock; 141 spinlock_t spinlock;
106 spinlock_t ast_lock; 142 spinlock_t ast_lock;
@@ -122,6 +158,9 @@ struct dlm_ctxt
122 atomic_t remote_resources; 158 atomic_t remote_resources;
123 atomic_t unknown_resources; 159 atomic_t unknown_resources;
124 160
161 struct dlm_debug_ctxt *dlm_debug_ctxt;
162 struct dentry *dlm_debugfs_subroot;
163
125 /* NOTE: Next three are protected by dlm_domain_lock */ 164 /* NOTE: Next three are protected by dlm_domain_lock */
126 struct kref dlm_refs; 165 struct kref dlm_refs;
127 enum dlm_ctxt_state dlm_state; 166 enum dlm_ctxt_state dlm_state;
@@ -176,6 +215,7 @@ struct dlm_mig_lockres_priv
176{ 215{
177 struct dlm_lock_resource *lockres; 216 struct dlm_lock_resource *lockres;
178 u8 real_master; 217 u8 real_master;
218 u8 extra_ref;
179}; 219};
180 220
181struct dlm_assert_master_priv 221struct dlm_assert_master_priv
@@ -269,6 +309,9 @@ struct dlm_lock_resource
269 struct list_head dirty; 309 struct list_head dirty;
270 struct list_head recovering; // dlm_recovery_ctxt.resources list 310 struct list_head recovering; // dlm_recovery_ctxt.resources list
271 311
312 /* Added during init and removed during release */
313 struct list_head tracking; /* dlm->tracking_list */
314
272 /* unused lock resources have their last_used stamped and are 315 /* unused lock resources have their last_used stamped and are
273 * put on a list for the dlm thread to run. */ 316 * put on a list for the dlm thread to run. */
274 unsigned long last_used; 317 unsigned long last_used;
@@ -602,17 +645,19 @@ enum dlm_query_join_response_code {
602 JOIN_PROTOCOL_MISMATCH, 645 JOIN_PROTOCOL_MISMATCH,
603}; 646};
604 647
648struct dlm_query_join_packet {
649 u8 code; /* Response code. dlm_minor and fs_minor
650 are only valid if this is JOIN_OK */
651 u8 dlm_minor; /* The minor version of the protocol the
652 dlm is speaking. */
653 u8 fs_minor; /* The minor version of the protocol the
654 filesystem is speaking. */
655 u8 reserved;
656};
657
605union dlm_query_join_response { 658union dlm_query_join_response {
606 u32 intval; 659 u32 intval;
607 struct { 660 struct dlm_query_join_packet packet;
608 u8 code; /* Response code. dlm_minor and fs_minor
609 are only valid if this is JOIN_OK */
610 u8 dlm_minor; /* The minor version of the protocol the
611 dlm is speaking. */
612 u8 fs_minor; /* The minor version of the protocol the
613 filesystem is speaking. */
614 u8 reserved;
615 } packet;
616}; 661};
617 662
618struct dlm_lock_request 663struct dlm_lock_request
@@ -960,9 +1005,16 @@ static inline void __dlm_wait_on_lockres(struct dlm_lock_resource *res)
960 DLM_LOCK_RES_MIGRATING)); 1005 DLM_LOCK_RES_MIGRATING));
961} 1006}
962 1007
1008/* create/destroy slab caches */
1009int dlm_init_master_caches(void);
1010void dlm_destroy_master_caches(void);
1011
1012int dlm_init_lock_cache(void);
1013void dlm_destroy_lock_cache(void);
963 1014
964int dlm_init_mle_cache(void); 1015int dlm_init_mle_cache(void);
965void dlm_destroy_mle_cache(void); 1016void dlm_destroy_mle_cache(void);
1017
966void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up); 1018void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up);
967int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, 1019int dlm_drop_lockres_ref(struct dlm_ctxt *dlm,
968 struct dlm_lock_resource *res); 1020 struct dlm_lock_resource *res);
diff --git a/fs/ocfs2/dlm/dlmconvert.c b/fs/ocfs2/dlm/dlmconvert.c
index ecb4d997221e..75997b4deaf3 100644
--- a/fs/ocfs2/dlm/dlmconvert.c
+++ b/fs/ocfs2/dlm/dlmconvert.c
@@ -487,7 +487,7 @@ int dlm_convert_lock_handler(struct o2net_msg *msg, u32 len, void *data,
487 "cookie=%u:%llu\n", 487 "cookie=%u:%llu\n",
488 dlm_get_lock_cookie_node(be64_to_cpu(cnv->cookie)), 488 dlm_get_lock_cookie_node(be64_to_cpu(cnv->cookie)),
489 dlm_get_lock_cookie_seq(be64_to_cpu(cnv->cookie))); 489 dlm_get_lock_cookie_seq(be64_to_cpu(cnv->cookie)));
490 __dlm_print_one_lock_resource(res); 490 dlm_print_one_lock_resource(res);
491 goto leave; 491 goto leave;
492 } 492 }
493 493
diff --git a/fs/ocfs2/dlm/dlmdebug.c b/fs/ocfs2/dlm/dlmdebug.c
index 64239b37e5d4..5f6d858770a2 100644
--- a/fs/ocfs2/dlm/dlmdebug.c
+++ b/fs/ocfs2/dlm/dlmdebug.c
@@ -5,7 +5,7 @@
5 * 5 *
6 * debug functionality for the dlm 6 * debug functionality for the dlm
7 * 7 *
8 * Copyright (C) 2004 Oracle. All rights reserved. 8 * Copyright (C) 2004, 2008 Oracle. All rights reserved.
9 * 9 *
10 * This program is free software; you can redistribute it and/or 10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public 11 * modify it under the terms of the GNU General Public
@@ -30,6 +30,7 @@
30#include <linux/utsname.h> 30#include <linux/utsname.h>
31#include <linux/sysctl.h> 31#include <linux/sysctl.h>
32#include <linux/spinlock.h> 32#include <linux/spinlock.h>
33#include <linux/debugfs.h>
33 34
34#include "cluster/heartbeat.h" 35#include "cluster/heartbeat.h"
35#include "cluster/nodemanager.h" 36#include "cluster/nodemanager.h"
@@ -37,17 +38,16 @@
37 38
38#include "dlmapi.h" 39#include "dlmapi.h"
39#include "dlmcommon.h" 40#include "dlmcommon.h"
40
41#include "dlmdomain.h" 41#include "dlmdomain.h"
42#include "dlmdebug.h"
42 43
43#define MLOG_MASK_PREFIX ML_DLM 44#define MLOG_MASK_PREFIX ML_DLM
44#include "cluster/masklog.h" 45#include "cluster/masklog.h"
45 46
47int stringify_lockname(const char *lockname, int locklen, char *buf, int len);
48
46void dlm_print_one_lock_resource(struct dlm_lock_resource *res) 49void dlm_print_one_lock_resource(struct dlm_lock_resource *res)
47{ 50{
48 mlog(ML_NOTICE, "lockres: %.*s, owner=%u, state=%u\n",
49 res->lockname.len, res->lockname.name,
50 res->owner, res->state);
51 spin_lock(&res->spinlock); 51 spin_lock(&res->spinlock);
52 __dlm_print_one_lock_resource(res); 52 __dlm_print_one_lock_resource(res);
53 spin_unlock(&res->spinlock); 53 spin_unlock(&res->spinlock);
@@ -58,7 +58,7 @@ static void dlm_print_lockres_refmap(struct dlm_lock_resource *res)
58 int bit; 58 int bit;
59 assert_spin_locked(&res->spinlock); 59 assert_spin_locked(&res->spinlock);
60 60
61 mlog(ML_NOTICE, " refmap nodes: [ "); 61 printk(" refmap nodes: [ ");
62 bit = 0; 62 bit = 0;
63 while (1) { 63 while (1) {
64 bit = find_next_bit(res->refmap, O2NM_MAX_NODES, bit); 64 bit = find_next_bit(res->refmap, O2NM_MAX_NODES, bit);
@@ -70,63 +70,66 @@ static void dlm_print_lockres_refmap(struct dlm_lock_resource *res)
70 printk("], inflight=%u\n", res->inflight_locks); 70 printk("], inflight=%u\n", res->inflight_locks);
71} 71}
72 72
73static void __dlm_print_lock(struct dlm_lock *lock)
74{
75 spin_lock(&lock->spinlock);
76
77 printk(" type=%d, conv=%d, node=%u, cookie=%u:%llu, "
78 "ref=%u, ast=(empty=%c,pend=%c), bast=(empty=%c,pend=%c), "
79 "pending=(conv=%c,lock=%c,cancel=%c,unlock=%c)\n",
80 lock->ml.type, lock->ml.convert_type, lock->ml.node,
81 dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
82 dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
83 atomic_read(&lock->lock_refs.refcount),
84 (list_empty(&lock->ast_list) ? 'y' : 'n'),
85 (lock->ast_pending ? 'y' : 'n'),
86 (list_empty(&lock->bast_list) ? 'y' : 'n'),
87 (lock->bast_pending ? 'y' : 'n'),
88 (lock->convert_pending ? 'y' : 'n'),
89 (lock->lock_pending ? 'y' : 'n'),
90 (lock->cancel_pending ? 'y' : 'n'),
91 (lock->unlock_pending ? 'y' : 'n'));
92
93 spin_unlock(&lock->spinlock);
94}
95
73void __dlm_print_one_lock_resource(struct dlm_lock_resource *res) 96void __dlm_print_one_lock_resource(struct dlm_lock_resource *res)
74{ 97{
75 struct list_head *iter2; 98 struct list_head *iter2;
76 struct dlm_lock *lock; 99 struct dlm_lock *lock;
100 char buf[DLM_LOCKID_NAME_MAX];
77 101
78 assert_spin_locked(&res->spinlock); 102 assert_spin_locked(&res->spinlock);
79 103
80 mlog(ML_NOTICE, "lockres: %.*s, owner=%u, state=%u\n", 104 stringify_lockname(res->lockname.name, res->lockname.len,
81 res->lockname.len, res->lockname.name, 105 buf, sizeof(buf) - 1);
82 res->owner, res->state); 106 printk("lockres: %s, owner=%u, state=%u\n",
83 mlog(ML_NOTICE, " last used: %lu, on purge list: %s\n", 107 buf, res->owner, res->state);
84 res->last_used, list_empty(&res->purge) ? "no" : "yes"); 108 printk(" last used: %lu, refcnt: %u, on purge list: %s\n",
109 res->last_used, atomic_read(&res->refs.refcount),
110 list_empty(&res->purge) ? "no" : "yes");
111 printk(" on dirty list: %s, on reco list: %s, "
112 "migrating pending: %s\n",
113 list_empty(&res->dirty) ? "no" : "yes",
114 list_empty(&res->recovering) ? "no" : "yes",
115 res->migration_pending ? "yes" : "no");
116 printk(" inflight locks: %d, asts reserved: %d\n",
117 res->inflight_locks, atomic_read(&res->asts_reserved));
85 dlm_print_lockres_refmap(res); 118 dlm_print_lockres_refmap(res);
86 mlog(ML_NOTICE, " granted queue: \n"); 119 printk(" granted queue:\n");
87 list_for_each(iter2, &res->granted) { 120 list_for_each(iter2, &res->granted) {
88 lock = list_entry(iter2, struct dlm_lock, list); 121 lock = list_entry(iter2, struct dlm_lock, list);
89 spin_lock(&lock->spinlock); 122 __dlm_print_lock(lock);
90 mlog(ML_NOTICE, " type=%d, conv=%d, node=%u, "
91 "cookie=%u:%llu, ast=(empty=%c,pend=%c), bast=(empty=%c,pend=%c)\n",
92 lock->ml.type, lock->ml.convert_type, lock->ml.node,
93 dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
94 dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
95 list_empty(&lock->ast_list) ? 'y' : 'n',
96 lock->ast_pending ? 'y' : 'n',
97 list_empty(&lock->bast_list) ? 'y' : 'n',
98 lock->bast_pending ? 'y' : 'n');
99 spin_unlock(&lock->spinlock);
100 } 123 }
101 mlog(ML_NOTICE, " converting queue: \n"); 124 printk(" converting queue:\n");
102 list_for_each(iter2, &res->converting) { 125 list_for_each(iter2, &res->converting) {
103 lock = list_entry(iter2, struct dlm_lock, list); 126 lock = list_entry(iter2, struct dlm_lock, list);
104 spin_lock(&lock->spinlock); 127 __dlm_print_lock(lock);
105 mlog(ML_NOTICE, " type=%d, conv=%d, node=%u, "
106 "cookie=%u:%llu, ast=(empty=%c,pend=%c), bast=(empty=%c,pend=%c)\n",
107 lock->ml.type, lock->ml.convert_type, lock->ml.node,
108 dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
109 dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
110 list_empty(&lock->ast_list) ? 'y' : 'n',
111 lock->ast_pending ? 'y' : 'n',
112 list_empty(&lock->bast_list) ? 'y' : 'n',
113 lock->bast_pending ? 'y' : 'n');
114 spin_unlock(&lock->spinlock);
115 } 128 }
116 mlog(ML_NOTICE, " blocked queue: \n"); 129 printk(" blocked queue:\n");
117 list_for_each(iter2, &res->blocked) { 130 list_for_each(iter2, &res->blocked) {
118 lock = list_entry(iter2, struct dlm_lock, list); 131 lock = list_entry(iter2, struct dlm_lock, list);
119 spin_lock(&lock->spinlock); 132 __dlm_print_lock(lock);
120 mlog(ML_NOTICE, " type=%d, conv=%d, node=%u, "
121 "cookie=%u:%llu, ast=(empty=%c,pend=%c), bast=(empty=%c,pend=%c)\n",
122 lock->ml.type, lock->ml.convert_type, lock->ml.node,
123 dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
124 dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
125 list_empty(&lock->ast_list) ? 'y' : 'n',
126 lock->ast_pending ? 'y' : 'n',
127 list_empty(&lock->bast_list) ? 'y' : 'n',
128 lock->bast_pending ? 'y' : 'n');
129 spin_unlock(&lock->spinlock);
130 } 133 }
131} 134}
132 135
@@ -136,31 +139,6 @@ void dlm_print_one_lock(struct dlm_lock *lockid)
136} 139}
137EXPORT_SYMBOL_GPL(dlm_print_one_lock); 140EXPORT_SYMBOL_GPL(dlm_print_one_lock);
138 141
139#if 0
140void dlm_dump_lock_resources(struct dlm_ctxt *dlm)
141{
142 struct dlm_lock_resource *res;
143 struct hlist_node *iter;
144 struct hlist_head *bucket;
145 int i;
146
147 mlog(ML_NOTICE, "struct dlm_ctxt: %s, node=%u, key=%u\n",
148 dlm->name, dlm->node_num, dlm->key);
149 if (!dlm || !dlm->name) {
150 mlog(ML_ERROR, "dlm=%p\n", dlm);
151 return;
152 }
153
154 spin_lock(&dlm->spinlock);
155 for (i=0; i<DLM_HASH_BUCKETS; i++) {
156 bucket = dlm_lockres_hash(dlm, i);
157 hlist_for_each_entry(res, iter, bucket, hash_node)
158 dlm_print_one_lock_resource(res);
159 }
160 spin_unlock(&dlm->spinlock);
161}
162#endif /* 0 */
163
164static const char *dlm_errnames[] = { 142static const char *dlm_errnames[] = {
165 [DLM_NORMAL] = "DLM_NORMAL", 143 [DLM_NORMAL] = "DLM_NORMAL",
166 [DLM_GRANTED] = "DLM_GRANTED", 144 [DLM_GRANTED] = "DLM_GRANTED",
@@ -266,3 +244,792 @@ const char *dlm_errname(enum dlm_status err)
266 return dlm_errnames[err]; 244 return dlm_errnames[err];
267} 245}
268EXPORT_SYMBOL_GPL(dlm_errname); 246EXPORT_SYMBOL_GPL(dlm_errname);
247
248/* NOTE: This function converts a lockname into a string. It uses knowledge
249 * of the format of the lockname that should be outside the purview of the dlm.
250 * We are adding only to make dlm debugging slightly easier.
251 *
252 * For more on lockname formats, please refer to dlmglue.c and ocfs2_lockid.h.
253 */
254int stringify_lockname(const char *lockname, int locklen, char *buf, int len)
255{
256 int out = 0;
257 __be64 inode_blkno_be;
258
259#define OCFS2_DENTRY_LOCK_INO_START 18
260 if (*lockname == 'N') {
261 memcpy((__be64 *)&inode_blkno_be,
262 (char *)&lockname[OCFS2_DENTRY_LOCK_INO_START],
263 sizeof(__be64));
264 out += snprintf(buf + out, len - out, "%.*s%08x",
265 OCFS2_DENTRY_LOCK_INO_START - 1, lockname,
266 (unsigned int)be64_to_cpu(inode_blkno_be));
267 } else
268 out += snprintf(buf + out, len - out, "%.*s",
269 locklen, lockname);
270 return out;
271}
272
273static int stringify_nodemap(unsigned long *nodemap, int maxnodes,
274 char *buf, int len)
275{
276 int out = 0;
277 int i = -1;
278
279 while ((i = find_next_bit(nodemap, maxnodes, i + 1)) < maxnodes)
280 out += snprintf(buf + out, len - out, "%d ", i);
281
282 return out;
283}
284
285static int dump_mle(struct dlm_master_list_entry *mle, char *buf, int len)
286{
287 int out = 0;
288 unsigned int namelen;
289 const char *name;
290 char *mle_type;
291
292 if (mle->type != DLM_MLE_MASTER) {
293 namelen = mle->u.name.len;
294 name = mle->u.name.name;
295 } else {
296 namelen = mle->u.res->lockname.len;
297 name = mle->u.res->lockname.name;
298 }
299
300 if (mle->type == DLM_MLE_BLOCK)
301 mle_type = "BLK";
302 else if (mle->type == DLM_MLE_MASTER)
303 mle_type = "MAS";
304 else
305 mle_type = "MIG";
306
307 out += stringify_lockname(name, namelen, buf + out, len - out);
308 out += snprintf(buf + out, len - out,
309 "\t%3s\tmas=%3u\tnew=%3u\tevt=%1d\tuse=%1d\tref=%3d\n",
310 mle_type, mle->master, mle->new_master,
311 !list_empty(&mle->hb_events),
312 !!mle->inuse,
313 atomic_read(&mle->mle_refs.refcount));
314
315 out += snprintf(buf + out, len - out, "Maybe=");
316 out += stringify_nodemap(mle->maybe_map, O2NM_MAX_NODES,
317 buf + out, len - out);
318 out += snprintf(buf + out, len - out, "\n");
319
320 out += snprintf(buf + out, len - out, "Vote=");
321 out += stringify_nodemap(mle->vote_map, O2NM_MAX_NODES,
322 buf + out, len - out);
323 out += snprintf(buf + out, len - out, "\n");
324
325 out += snprintf(buf + out, len - out, "Response=");
326 out += stringify_nodemap(mle->response_map, O2NM_MAX_NODES,
327 buf + out, len - out);
328 out += snprintf(buf + out, len - out, "\n");
329
330 out += snprintf(buf + out, len - out, "Node=");
331 out += stringify_nodemap(mle->node_map, O2NM_MAX_NODES,
332 buf + out, len - out);
333 out += snprintf(buf + out, len - out, "\n");
334
335 out += snprintf(buf + out, len - out, "\n");
336
337 return out;
338}
339
340void dlm_print_one_mle(struct dlm_master_list_entry *mle)
341{
342 char *buf;
343
344 buf = (char *) get_zeroed_page(GFP_NOFS);
345 if (buf) {
346 dump_mle(mle, buf, PAGE_SIZE - 1);
347 free_page((unsigned long)buf);
348 }
349}
350
351#ifdef CONFIG_DEBUG_FS
352
353static struct dentry *dlm_debugfs_root = NULL;
354
355#define DLM_DEBUGFS_DIR "o2dlm"
356#define DLM_DEBUGFS_DLM_STATE "dlm_state"
357#define DLM_DEBUGFS_LOCKING_STATE "locking_state"
358#define DLM_DEBUGFS_MLE_STATE "mle_state"
359#define DLM_DEBUGFS_PURGE_LIST "purge_list"
360
361/* begin - utils funcs */
362static void dlm_debug_free(struct kref *kref)
363{
364 struct dlm_debug_ctxt *dc;
365
366 dc = container_of(kref, struct dlm_debug_ctxt, debug_refcnt);
367
368 kfree(dc);
369}
370
371void dlm_debug_put(struct dlm_debug_ctxt *dc)
372{
373 if (dc)
374 kref_put(&dc->debug_refcnt, dlm_debug_free);
375}
376
377static void dlm_debug_get(struct dlm_debug_ctxt *dc)
378{
379 kref_get(&dc->debug_refcnt);
380}
381
382static struct debug_buffer *debug_buffer_allocate(void)
383{
384 struct debug_buffer *db = NULL;
385
386 db = kzalloc(sizeof(struct debug_buffer), GFP_KERNEL);
387 if (!db)
388 goto bail;
389
390 db->len = PAGE_SIZE;
391 db->buf = kmalloc(db->len, GFP_KERNEL);
392 if (!db->buf)
393 goto bail;
394
395 return db;
396bail:
397 kfree(db);
398 return NULL;
399}
400
401static ssize_t debug_buffer_read(struct file *file, char __user *buf,
402 size_t nbytes, loff_t *ppos)
403{
404 struct debug_buffer *db = file->private_data;
405
406 return simple_read_from_buffer(buf, nbytes, ppos, db->buf, db->len);
407}
408
409static loff_t debug_buffer_llseek(struct file *file, loff_t off, int whence)
410{
411 struct debug_buffer *db = file->private_data;
412 loff_t new = -1;
413
414 switch (whence) {
415 case 0:
416 new = off;
417 break;
418 case 1:
419 new = file->f_pos + off;
420 break;
421 }
422
423 if (new < 0 || new > db->len)
424 return -EINVAL;
425
426 return (file->f_pos = new);
427}
428
429static int debug_buffer_release(struct inode *inode, struct file *file)
430{
431 struct debug_buffer *db = (struct debug_buffer *)file->private_data;
432
433 if (db)
434 kfree(db->buf);
435 kfree(db);
436
437 return 0;
438}
439/* end - util funcs */
440
441/* begin - purge list funcs */
442static int debug_purgelist_print(struct dlm_ctxt *dlm, struct debug_buffer *db)
443{
444 struct dlm_lock_resource *res;
445 int out = 0;
446 unsigned long total = 0;
447
448 out += snprintf(db->buf + out, db->len - out,
449 "Dumping Purgelist for Domain: %s\n", dlm->name);
450
451 spin_lock(&dlm->spinlock);
452 list_for_each_entry(res, &dlm->purge_list, purge) {
453 ++total;
454 if (db->len - out < 100)
455 continue;
456 spin_lock(&res->spinlock);
457 out += stringify_lockname(res->lockname.name,
458 res->lockname.len,
459 db->buf + out, db->len - out);
460 out += snprintf(db->buf + out, db->len - out, "\t%ld\n",
461 (jiffies - res->last_used)/HZ);
462 spin_unlock(&res->spinlock);
463 }
464 spin_unlock(&dlm->spinlock);
465
466 out += snprintf(db->buf + out, db->len - out,
467 "Total on list: %ld\n", total);
468
469 return out;
470}
471
472static int debug_purgelist_open(struct inode *inode, struct file *file)
473{
474 struct dlm_ctxt *dlm = inode->i_private;
475 struct debug_buffer *db;
476
477 db = debug_buffer_allocate();
478 if (!db)
479 goto bail;
480
481 db->len = debug_purgelist_print(dlm, db);
482
483 file->private_data = db;
484
485 return 0;
486bail:
487 return -ENOMEM;
488}
489
490static struct file_operations debug_purgelist_fops = {
491 .open = debug_purgelist_open,
492 .release = debug_buffer_release,
493 .read = debug_buffer_read,
494 .llseek = debug_buffer_llseek,
495};
496/* end - purge list funcs */
497
498/* begin - debug mle funcs */
499static int debug_mle_print(struct dlm_ctxt *dlm, struct debug_buffer *db)
500{
501 struct dlm_master_list_entry *mle;
502 int out = 0;
503 unsigned long total = 0;
504
505 out += snprintf(db->buf + out, db->len - out,
506 "Dumping MLEs for Domain: %s\n", dlm->name);
507
508 spin_lock(&dlm->master_lock);
509 list_for_each_entry(mle, &dlm->master_list, list) {
510 ++total;
511 if (db->len - out < 200)
512 continue;
513 out += dump_mle(mle, db->buf + out, db->len - out);
514 }
515 spin_unlock(&dlm->master_lock);
516
517 out += snprintf(db->buf + out, db->len - out,
518 "Total on list: %ld\n", total);
519 return out;
520}
521
522static int debug_mle_open(struct inode *inode, struct file *file)
523{
524 struct dlm_ctxt *dlm = inode->i_private;
525 struct debug_buffer *db;
526
527 db = debug_buffer_allocate();
528 if (!db)
529 goto bail;
530
531 db->len = debug_mle_print(dlm, db);
532
533 file->private_data = db;
534
535 return 0;
536bail:
537 return -ENOMEM;
538}
539
540static struct file_operations debug_mle_fops = {
541 .open = debug_mle_open,
542 .release = debug_buffer_release,
543 .read = debug_buffer_read,
544 .llseek = debug_buffer_llseek,
545};
546
547/* end - debug mle funcs */
548
549/* begin - debug lockres funcs */
550static int dump_lock(struct dlm_lock *lock, int list_type, char *buf, int len)
551{
552 int out;
553
554#define DEBUG_LOCK_VERSION 1
555 spin_lock(&lock->spinlock);
556 out = snprintf(buf, len, "LOCK:%d,%d,%d,%d,%d,%d:%lld,%d,%d,%d,%d,%d,"
557 "%d,%d,%d,%d\n",
558 DEBUG_LOCK_VERSION,
559 list_type, lock->ml.type, lock->ml.convert_type,
560 lock->ml.node,
561 dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
562 dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
563 !list_empty(&lock->ast_list),
564 !list_empty(&lock->bast_list),
565 lock->ast_pending, lock->bast_pending,
566 lock->convert_pending, lock->lock_pending,
567 lock->cancel_pending, lock->unlock_pending,
568 atomic_read(&lock->lock_refs.refcount));
569 spin_unlock(&lock->spinlock);
570
571 return out;
572}
573
574static int dump_lockres(struct dlm_lock_resource *res, char *buf, int len)
575{
576 struct dlm_lock *lock;
577 int i;
578 int out = 0;
579
580 out += snprintf(buf + out, len - out, "NAME:");
581 out += stringify_lockname(res->lockname.name, res->lockname.len,
582 buf + out, len - out);
583 out += snprintf(buf + out, len - out, "\n");
584
585#define DEBUG_LRES_VERSION 1
586 out += snprintf(buf + out, len - out,
587 "LRES:%d,%d,%d,%ld,%d,%d,%d,%d,%d,%d,%d\n",
588 DEBUG_LRES_VERSION,
589 res->owner, res->state, res->last_used,
590 !list_empty(&res->purge),
591 !list_empty(&res->dirty),
592 !list_empty(&res->recovering),
593 res->inflight_locks, res->migration_pending,
594 atomic_read(&res->asts_reserved),
595 atomic_read(&res->refs.refcount));
596
597 /* refmap */
598 out += snprintf(buf + out, len - out, "RMAP:");
599 out += stringify_nodemap(res->refmap, O2NM_MAX_NODES,
600 buf + out, len - out);
601 out += snprintf(buf + out, len - out, "\n");
602
603 /* lvb */
604 out += snprintf(buf + out, len - out, "LVBX:");
605 for (i = 0; i < DLM_LVB_LEN; i++)
606 out += snprintf(buf + out, len - out,
607 "%02x", (unsigned char)res->lvb[i]);
608 out += snprintf(buf + out, len - out, "\n");
609
610 /* granted */
611 list_for_each_entry(lock, &res->granted, list)
612 out += dump_lock(lock, 0, buf + out, len - out);
613
614 /* converting */
615 list_for_each_entry(lock, &res->converting, list)
616 out += dump_lock(lock, 1, buf + out, len - out);
617
618 /* blocked */
619 list_for_each_entry(lock, &res->blocked, list)
620 out += dump_lock(lock, 2, buf + out, len - out);
621
622 out += snprintf(buf + out, len - out, "\n");
623
624 return out;
625}
626
627static void *lockres_seq_start(struct seq_file *m, loff_t *pos)
628{
629 struct debug_lockres *dl = m->private;
630 struct dlm_ctxt *dlm = dl->dl_ctxt;
631 struct dlm_lock_resource *res = NULL;
632
633 spin_lock(&dlm->spinlock);
634
635 if (dl->dl_res) {
636 list_for_each_entry(res, &dl->dl_res->tracking, tracking) {
637 if (dl->dl_res) {
638 dlm_lockres_put(dl->dl_res);
639 dl->dl_res = NULL;
640 }
641 if (&res->tracking == &dlm->tracking_list) {
642 mlog(0, "End of list found, %p\n", res);
643 dl = NULL;
644 break;
645 }
646 dlm_lockres_get(res);
647 dl->dl_res = res;
648 break;
649 }
650 } else {
651 if (!list_empty(&dlm->tracking_list)) {
652 list_for_each_entry(res, &dlm->tracking_list, tracking)
653 break;
654 dlm_lockres_get(res);
655 dl->dl_res = res;
656 } else
657 dl = NULL;
658 }
659
660 if (dl) {
661 spin_lock(&dl->dl_res->spinlock);
662 dump_lockres(dl->dl_res, dl->dl_buf, dl->dl_len - 1);
663 spin_unlock(&dl->dl_res->spinlock);
664 }
665
666 spin_unlock(&dlm->spinlock);
667
668 return dl;
669}
670
671static void lockres_seq_stop(struct seq_file *m, void *v)
672{
673}
674
675static void *lockres_seq_next(struct seq_file *m, void *v, loff_t *pos)
676{
677 return NULL;
678}
679
680static int lockres_seq_show(struct seq_file *s, void *v)
681{
682 struct debug_lockres *dl = (struct debug_lockres *)v;
683
684 seq_printf(s, "%s", dl->dl_buf);
685
686 return 0;
687}
688
689static struct seq_operations debug_lockres_ops = {
690 .start = lockres_seq_start,
691 .stop = lockres_seq_stop,
692 .next = lockres_seq_next,
693 .show = lockres_seq_show,
694};
695
696static int debug_lockres_open(struct inode *inode, struct file *file)
697{
698 struct dlm_ctxt *dlm = inode->i_private;
699 int ret = -ENOMEM;
700 struct seq_file *seq;
701 struct debug_lockres *dl = NULL;
702
703 dl = kzalloc(sizeof(struct debug_lockres), GFP_KERNEL);
704 if (!dl) {
705 mlog_errno(ret);
706 goto bail;
707 }
708
709 dl->dl_len = PAGE_SIZE;
710 dl->dl_buf = kmalloc(dl->dl_len, GFP_KERNEL);
711 if (!dl->dl_buf) {
712 mlog_errno(ret);
713 goto bail;
714 }
715
716 ret = seq_open(file, &debug_lockres_ops);
717 if (ret) {
718 mlog_errno(ret);
719 goto bail;
720 }
721
722 seq = (struct seq_file *) file->private_data;
723 seq->private = dl;
724
725 dlm_grab(dlm);
726 dl->dl_ctxt = dlm;
727
728 return 0;
729bail:
730 if (dl)
731 kfree(dl->dl_buf);
732 kfree(dl);
733 return ret;
734}
735
736static int debug_lockres_release(struct inode *inode, struct file *file)
737{
738 struct seq_file *seq = (struct seq_file *)file->private_data;
739 struct debug_lockres *dl = (struct debug_lockres *)seq->private;
740
741 if (dl->dl_res)
742 dlm_lockres_put(dl->dl_res);
743 dlm_put(dl->dl_ctxt);
744 kfree(dl->dl_buf);
745 return seq_release_private(inode, file);
746}
747
748static struct file_operations debug_lockres_fops = {
749 .open = debug_lockres_open,
750 .release = debug_lockres_release,
751 .read = seq_read,
752 .llseek = seq_lseek,
753};
754/* end - debug lockres funcs */
755
756/* begin - debug state funcs */
757static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db)
758{
759 int out = 0;
760 struct dlm_reco_node_data *node;
761 char *state;
762 int lres, rres, ures, tres;
763
764 lres = atomic_read(&dlm->local_resources);
765 rres = atomic_read(&dlm->remote_resources);
766 ures = atomic_read(&dlm->unknown_resources);
767 tres = lres + rres + ures;
768
769 spin_lock(&dlm->spinlock);
770
771 switch (dlm->dlm_state) {
772 case DLM_CTXT_NEW:
773 state = "NEW"; break;
774 case DLM_CTXT_JOINED:
775 state = "JOINED"; break;
776 case DLM_CTXT_IN_SHUTDOWN:
777 state = "SHUTDOWN"; break;
778 case DLM_CTXT_LEAVING:
779 state = "LEAVING"; break;
780 default:
781 state = "UNKNOWN"; break;
782 }
783
784 /* Domain: xxxxxxxxxx Key: 0xdfbac769 */
785 out += snprintf(db->buf + out, db->len - out,
786 "Domain: %s Key: 0x%08x\n", dlm->name, dlm->key);
787
788 /* Thread Pid: xxx Node: xxx State: xxxxx */
789 out += snprintf(db->buf + out, db->len - out,
790 "Thread Pid: %d Node: %d State: %s\n",
791 dlm->dlm_thread_task->pid, dlm->node_num, state);
792
793 /* Number of Joins: xxx Joining Node: xxx */
794 out += snprintf(db->buf + out, db->len - out,
795 "Number of Joins: %d Joining Node: %d\n",
796 dlm->num_joins, dlm->joining_node);
797
798 /* Domain Map: xx xx xx */
799 out += snprintf(db->buf + out, db->len - out, "Domain Map: ");
800 out += stringify_nodemap(dlm->domain_map, O2NM_MAX_NODES,
801 db->buf + out, db->len - out);
802 out += snprintf(db->buf + out, db->len - out, "\n");
803
804 /* Live Map: xx xx xx */
805 out += snprintf(db->buf + out, db->len - out, "Live Map: ");
806 out += stringify_nodemap(dlm->live_nodes_map, O2NM_MAX_NODES,
807 db->buf + out, db->len - out);
808 out += snprintf(db->buf + out, db->len - out, "\n");
809
810 /* Mastered Resources Total: xxx Locally: xxx Remotely: ... */
811 out += snprintf(db->buf + out, db->len - out,
812 "Mastered Resources Total: %d Locally: %d "
813 "Remotely: %d Unknown: %d\n",
814 tres, lres, rres, ures);
815
816 /* Lists: Dirty=Empty Purge=InUse PendingASTs=Empty ... */
817 out += snprintf(db->buf + out, db->len - out,
818 "Lists: Dirty=%s Purge=%s PendingASTs=%s "
819 "PendingBASTs=%s Master=%s\n",
820 (list_empty(&dlm->dirty_list) ? "Empty" : "InUse"),
821 (list_empty(&dlm->purge_list) ? "Empty" : "InUse"),
822 (list_empty(&dlm->pending_asts) ? "Empty" : "InUse"),
823 (list_empty(&dlm->pending_basts) ? "Empty" : "InUse"),
824 (list_empty(&dlm->master_list) ? "Empty" : "InUse"));
825
826 /* Purge Count: xxx Refs: xxx */
827 out += snprintf(db->buf + out, db->len - out,
828 "Purge Count: %d Refs: %d\n", dlm->purge_count,
829 atomic_read(&dlm->dlm_refs.refcount));
830
831 /* Dead Node: xxx */
832 out += snprintf(db->buf + out, db->len - out,
833 "Dead Node: %d\n", dlm->reco.dead_node);
834
835 /* What about DLM_RECO_STATE_FINALIZE? */
836 if (dlm->reco.state == DLM_RECO_STATE_ACTIVE)
837 state = "ACTIVE";
838 else
839 state = "INACTIVE";
840
841 /* Recovery Pid: xxxx Master: xxx State: xxxx */
842 out += snprintf(db->buf + out, db->len - out,
843 "Recovery Pid: %d Master: %d State: %s\n",
844 dlm->dlm_reco_thread_task->pid,
845 dlm->reco.new_master, state);
846
847 /* Recovery Map: xx xx */
848 out += snprintf(db->buf + out, db->len - out, "Recovery Map: ");
849 out += stringify_nodemap(dlm->recovery_map, O2NM_MAX_NODES,
850 db->buf + out, db->len - out);
851 out += snprintf(db->buf + out, db->len - out, "\n");
852
853 /* Recovery Node State: */
854 out += snprintf(db->buf + out, db->len - out, "Recovery Node State:\n");
855 list_for_each_entry(node, &dlm->reco.node_data, list) {
856 switch (node->state) {
857 case DLM_RECO_NODE_DATA_INIT:
858 state = "INIT";
859 break;
860 case DLM_RECO_NODE_DATA_REQUESTING:
861 state = "REQUESTING";
862 break;
863 case DLM_RECO_NODE_DATA_DEAD:
864 state = "DEAD";
865 break;
866 case DLM_RECO_NODE_DATA_RECEIVING:
867 state = "RECEIVING";
868 break;
869 case DLM_RECO_NODE_DATA_REQUESTED:
870 state = "REQUESTED";
871 break;
872 case DLM_RECO_NODE_DATA_DONE:
873 state = "DONE";
874 break;
875 case DLM_RECO_NODE_DATA_FINALIZE_SENT:
876 state = "FINALIZE-SENT";
877 break;
878 default:
879 state = "BAD";
880 break;
881 }
882 out += snprintf(db->buf + out, db->len - out, "\t%u - %s\n",
883 node->node_num, state);
884 }
885
886 spin_unlock(&dlm->spinlock);
887
888 return out;
889}
890
891static int debug_state_open(struct inode *inode, struct file *file)
892{
893 struct dlm_ctxt *dlm = inode->i_private;
894 struct debug_buffer *db = NULL;
895
896 db = debug_buffer_allocate();
897 if (!db)
898 goto bail;
899
900 db->len = debug_state_print(dlm, db);
901
902 file->private_data = db;
903
904 return 0;
905bail:
906 return -ENOMEM;
907}
908
909static struct file_operations debug_state_fops = {
910 .open = debug_state_open,
911 .release = debug_buffer_release,
912 .read = debug_buffer_read,
913 .llseek = debug_buffer_llseek,
914};
915/* end - debug state funcs */
916
917/* files in subroot */
918int dlm_debug_init(struct dlm_ctxt *dlm)
919{
920 struct dlm_debug_ctxt *dc = dlm->dlm_debug_ctxt;
921
922 /* for dumping dlm_ctxt */
923 dc->debug_state_dentry = debugfs_create_file(DLM_DEBUGFS_DLM_STATE,
924 S_IFREG|S_IRUSR,
925 dlm->dlm_debugfs_subroot,
926 dlm, &debug_state_fops);
927 if (!dc->debug_state_dentry) {
928 mlog_errno(-ENOMEM);
929 goto bail;
930 }
931
932 /* for dumping lockres */
933 dc->debug_lockres_dentry =
934 debugfs_create_file(DLM_DEBUGFS_LOCKING_STATE,
935 S_IFREG|S_IRUSR,
936 dlm->dlm_debugfs_subroot,
937 dlm, &debug_lockres_fops);
938 if (!dc->debug_lockres_dentry) {
939 mlog_errno(-ENOMEM);
940 goto bail;
941 }
942
943 /* for dumping mles */
944 dc->debug_mle_dentry = debugfs_create_file(DLM_DEBUGFS_MLE_STATE,
945 S_IFREG|S_IRUSR,
946 dlm->dlm_debugfs_subroot,
947 dlm, &debug_mle_fops);
948 if (!dc->debug_mle_dentry) {
949 mlog_errno(-ENOMEM);
950 goto bail;
951 }
952
953 /* for dumping lockres on the purge list */
954 dc->debug_purgelist_dentry =
955 debugfs_create_file(DLM_DEBUGFS_PURGE_LIST,
956 S_IFREG|S_IRUSR,
957 dlm->dlm_debugfs_subroot,
958 dlm, &debug_purgelist_fops);
959 if (!dc->debug_purgelist_dentry) {
960 mlog_errno(-ENOMEM);
961 goto bail;
962 }
963
964 dlm_debug_get(dc);
965 return 0;
966
967bail:
968 dlm_debug_shutdown(dlm);
969 return -ENOMEM;
970}
971
972void dlm_debug_shutdown(struct dlm_ctxt *dlm)
973{
974 struct dlm_debug_ctxt *dc = dlm->dlm_debug_ctxt;
975
976 if (dc) {
977 if (dc->debug_purgelist_dentry)
978 debugfs_remove(dc->debug_purgelist_dentry);
979 if (dc->debug_mle_dentry)
980 debugfs_remove(dc->debug_mle_dentry);
981 if (dc->debug_lockres_dentry)
982 debugfs_remove(dc->debug_lockres_dentry);
983 if (dc->debug_state_dentry)
984 debugfs_remove(dc->debug_state_dentry);
985 dlm_debug_put(dc);
986 }
987}
988
989/* subroot - domain dir */
990int dlm_create_debugfs_subroot(struct dlm_ctxt *dlm)
991{
992 dlm->dlm_debugfs_subroot = debugfs_create_dir(dlm->name,
993 dlm_debugfs_root);
994 if (!dlm->dlm_debugfs_subroot) {
995 mlog_errno(-ENOMEM);
996 goto bail;
997 }
998
999 dlm->dlm_debug_ctxt = kzalloc(sizeof(struct dlm_debug_ctxt),
1000 GFP_KERNEL);
1001 if (!dlm->dlm_debug_ctxt) {
1002 mlog_errno(-ENOMEM);
1003 goto bail;
1004 }
1005 kref_init(&dlm->dlm_debug_ctxt->debug_refcnt);
1006
1007 return 0;
1008bail:
1009 dlm_destroy_debugfs_subroot(dlm);
1010 return -ENOMEM;
1011}
1012
1013void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm)
1014{
1015 if (dlm->dlm_debugfs_subroot)
1016 debugfs_remove(dlm->dlm_debugfs_subroot);
1017}
1018
1019/* debugfs root */
1020int dlm_create_debugfs_root(void)
1021{
1022 dlm_debugfs_root = debugfs_create_dir(DLM_DEBUGFS_DIR, NULL);
1023 if (!dlm_debugfs_root) {
1024 mlog_errno(-ENOMEM);
1025 return -ENOMEM;
1026 }
1027 return 0;
1028}
1029
1030void dlm_destroy_debugfs_root(void)
1031{
1032 if (dlm_debugfs_root)
1033 debugfs_remove(dlm_debugfs_root);
1034}
1035#endif /* CONFIG_DEBUG_FS */
diff --git a/fs/ocfs2/dlm/dlmdebug.h b/fs/ocfs2/dlm/dlmdebug.h
new file mode 100644
index 000000000000..d34a62a3a625
--- /dev/null
+++ b/fs/ocfs2/dlm/dlmdebug.h
@@ -0,0 +1,86 @@
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * dlmdebug.h
5 *
6 * Copyright (C) 2008 Oracle. All rights reserved.
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public
10 * License as published by the Free Software Foundation; either
11 * version 2 of the License, or (at your option) any later version.
12 *
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * General Public License for more details.
17 *
18 * You should have received a copy of the GNU General Public
19 * License along with this program; if not, write to the
20 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
21 * Boston, MA 021110-1307, USA.
22 *
23 */
24
25#ifndef DLMDEBUG_H
26#define DLMDEBUG_H
27
28void dlm_print_one_mle(struct dlm_master_list_entry *mle);
29
30#ifdef CONFIG_DEBUG_FS
31
32struct dlm_debug_ctxt {
33 struct kref debug_refcnt;
34 struct dentry *debug_state_dentry;
35 struct dentry *debug_lockres_dentry;
36 struct dentry *debug_mle_dentry;
37 struct dentry *debug_purgelist_dentry;
38};
39
40struct debug_buffer {
41 int len;
42 char *buf;
43};
44
45struct debug_lockres {
46 int dl_len;
47 char *dl_buf;
48 struct dlm_ctxt *dl_ctxt;
49 struct dlm_lock_resource *dl_res;
50};
51
52int dlm_debug_init(struct dlm_ctxt *dlm);
53void dlm_debug_shutdown(struct dlm_ctxt *dlm);
54
55int dlm_create_debugfs_subroot(struct dlm_ctxt *dlm);
56void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm);
57
58int dlm_create_debugfs_root(void);
59void dlm_destroy_debugfs_root(void);
60
61#else
62
63static int dlm_debug_init(struct dlm_ctxt *dlm)
64{
65 return 0;
66}
67static void dlm_debug_shutdown(struct dlm_ctxt *dlm)
68{
69}
70static int dlm_create_debugfs_subroot(struct dlm_ctxt *dlm)
71{
72 return 0;
73}
74static void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm)
75{
76}
77static int dlm_create_debugfs_root(void)
78{
79 return 0;
80}
81static void dlm_destroy_debugfs_root(void)
82{
83}
84
85#endif /* CONFIG_DEBUG_FS */
86#endif /* DLMDEBUG_H */
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 638d2ebb892b..63f8125824e8 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -33,6 +33,7 @@
33#include <linux/spinlock.h> 33#include <linux/spinlock.h>
34#include <linux/delay.h> 34#include <linux/delay.h>
35#include <linux/err.h> 35#include <linux/err.h>
36#include <linux/debugfs.h>
36 37
37#include "cluster/heartbeat.h" 38#include "cluster/heartbeat.h"
38#include "cluster/nodemanager.h" 39#include "cluster/nodemanager.h"
@@ -40,8 +41,8 @@
40 41
41#include "dlmapi.h" 42#include "dlmapi.h"
42#include "dlmcommon.h" 43#include "dlmcommon.h"
43
44#include "dlmdomain.h" 44#include "dlmdomain.h"
45#include "dlmdebug.h"
45 46
46#include "dlmver.h" 47#include "dlmver.h"
47 48
@@ -298,6 +299,8 @@ static int dlm_wait_on_domain_helper(const char *domain)
298 299
299static void dlm_free_ctxt_mem(struct dlm_ctxt *dlm) 300static void dlm_free_ctxt_mem(struct dlm_ctxt *dlm)
300{ 301{
302 dlm_destroy_debugfs_subroot(dlm);
303
301 if (dlm->lockres_hash) 304 if (dlm->lockres_hash)
302 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES); 305 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES);
303 306
@@ -395,6 +398,7 @@ static void dlm_destroy_dlm_worker(struct dlm_ctxt *dlm)
395static void dlm_complete_dlm_shutdown(struct dlm_ctxt *dlm) 398static void dlm_complete_dlm_shutdown(struct dlm_ctxt *dlm)
396{ 399{
397 dlm_unregister_domain_handlers(dlm); 400 dlm_unregister_domain_handlers(dlm);
401 dlm_debug_shutdown(dlm);
398 dlm_complete_thread(dlm); 402 dlm_complete_thread(dlm);
399 dlm_complete_recovery_thread(dlm); 403 dlm_complete_recovery_thread(dlm);
400 dlm_destroy_dlm_worker(dlm); 404 dlm_destroy_dlm_worker(dlm);
@@ -644,6 +648,7 @@ int dlm_shutting_down(struct dlm_ctxt *dlm)
644void dlm_unregister_domain(struct dlm_ctxt *dlm) 648void dlm_unregister_domain(struct dlm_ctxt *dlm)
645{ 649{
646 int leave = 0; 650 int leave = 0;
651 struct dlm_lock_resource *res;
647 652
648 spin_lock(&dlm_domain_lock); 653 spin_lock(&dlm_domain_lock);
649 BUG_ON(dlm->dlm_state != DLM_CTXT_JOINED); 654 BUG_ON(dlm->dlm_state != DLM_CTXT_JOINED);
@@ -673,6 +678,15 @@ void dlm_unregister_domain(struct dlm_ctxt *dlm)
673 msleep(500); 678 msleep(500);
674 mlog(0, "%s: more migration to do\n", dlm->name); 679 mlog(0, "%s: more migration to do\n", dlm->name);
675 } 680 }
681
682 /* This list should be empty. If not, print remaining lockres */
683 if (!list_empty(&dlm->tracking_list)) {
684 mlog(ML_ERROR, "Following lockres' are still on the "
685 "tracking list:\n");
686 list_for_each_entry(res, &dlm->tracking_list, tracking)
687 dlm_print_one_lock_resource(res);
688 }
689
676 dlm_mark_domain_leaving(dlm); 690 dlm_mark_domain_leaving(dlm);
677 dlm_leave_domain(dlm); 691 dlm_leave_domain(dlm);
678 dlm_complete_dlm_shutdown(dlm); 692 dlm_complete_dlm_shutdown(dlm);
@@ -713,14 +727,46 @@ static int dlm_query_join_proto_check(char *proto_type, int node,
713 return rc; 727 return rc;
714} 728}
715 729
730/*
731 * struct dlm_query_join_packet is made up of four one-byte fields. They
732 * are effectively in big-endian order already. However, little-endian
733 * machines swap them before putting the packet on the wire (because
734 * query_join's response is a status, and that status is treated as a u32
735 * on the wire). Thus, a big-endian and little-endian machines will treat
736 * this structure differently.
737 *
738 * The solution is to have little-endian machines swap the structure when
739 * converting from the structure to the u32 representation. This will
740 * result in the structure having the correct format on the wire no matter
741 * the host endian format.
742 */
743static void dlm_query_join_packet_to_wire(struct dlm_query_join_packet *packet,
744 u32 *wire)
745{
746 union dlm_query_join_response response;
747
748 response.packet = *packet;
749 *wire = cpu_to_be32(response.intval);
750}
751
752static void dlm_query_join_wire_to_packet(u32 wire,
753 struct dlm_query_join_packet *packet)
754{
755 union dlm_query_join_response response;
756
757 response.intval = cpu_to_be32(wire);
758 *packet = response.packet;
759}
760
716static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data, 761static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
717 void **ret_data) 762 void **ret_data)
718{ 763{
719 struct dlm_query_join_request *query; 764 struct dlm_query_join_request *query;
720 union dlm_query_join_response response = { 765 struct dlm_query_join_packet packet = {
721 .packet.code = JOIN_DISALLOW, 766 .code = JOIN_DISALLOW,
722 }; 767 };
723 struct dlm_ctxt *dlm = NULL; 768 struct dlm_ctxt *dlm = NULL;
769 u32 response;
724 u8 nodenum; 770 u8 nodenum;
725 771
726 query = (struct dlm_query_join_request *) msg->buf; 772 query = (struct dlm_query_join_request *) msg->buf;
@@ -737,11 +783,11 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
737 mlog(0, "node %u is not in our live map yet\n", 783 mlog(0, "node %u is not in our live map yet\n",
738 query->node_idx); 784 query->node_idx);
739 785
740 response.packet.code = JOIN_DISALLOW; 786 packet.code = JOIN_DISALLOW;
741 goto respond; 787 goto respond;
742 } 788 }
743 789
744 response.packet.code = JOIN_OK_NO_MAP; 790 packet.code = JOIN_OK_NO_MAP;
745 791
746 spin_lock(&dlm_domain_lock); 792 spin_lock(&dlm_domain_lock);
747 dlm = __dlm_lookup_domain_full(query->domain, query->name_len); 793 dlm = __dlm_lookup_domain_full(query->domain, query->name_len);
@@ -760,7 +806,7 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
760 mlog(0, "disallow join as node %u does not " 806 mlog(0, "disallow join as node %u does not "
761 "have node %u in its nodemap\n", 807 "have node %u in its nodemap\n",
762 query->node_idx, nodenum); 808 query->node_idx, nodenum);
763 response.packet.code = JOIN_DISALLOW; 809 packet.code = JOIN_DISALLOW;
764 goto unlock_respond; 810 goto unlock_respond;
765 } 811 }
766 } 812 }
@@ -780,23 +826,23 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
780 /*If this is a brand new context and we 826 /*If this is a brand new context and we
781 * haven't started our join process yet, then 827 * haven't started our join process yet, then
782 * the other node won the race. */ 828 * the other node won the race. */
783 response.packet.code = JOIN_OK_NO_MAP; 829 packet.code = JOIN_OK_NO_MAP;
784 } else if (dlm->joining_node != DLM_LOCK_RES_OWNER_UNKNOWN) { 830 } else if (dlm->joining_node != DLM_LOCK_RES_OWNER_UNKNOWN) {
785 /* Disallow parallel joins. */ 831 /* Disallow parallel joins. */
786 response.packet.code = JOIN_DISALLOW; 832 packet.code = JOIN_DISALLOW;
787 } else if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) { 833 } else if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) {
788 mlog(0, "node %u trying to join, but recovery " 834 mlog(0, "node %u trying to join, but recovery "
789 "is ongoing.\n", bit); 835 "is ongoing.\n", bit);
790 response.packet.code = JOIN_DISALLOW; 836 packet.code = JOIN_DISALLOW;
791 } else if (test_bit(bit, dlm->recovery_map)) { 837 } else if (test_bit(bit, dlm->recovery_map)) {
792 mlog(0, "node %u trying to join, but it " 838 mlog(0, "node %u trying to join, but it "
793 "still needs recovery.\n", bit); 839 "still needs recovery.\n", bit);
794 response.packet.code = JOIN_DISALLOW; 840 packet.code = JOIN_DISALLOW;
795 } else if (test_bit(bit, dlm->domain_map)) { 841 } else if (test_bit(bit, dlm->domain_map)) {
796 mlog(0, "node %u trying to join, but it " 842 mlog(0, "node %u trying to join, but it "
797 "is still in the domain! needs recovery?\n", 843 "is still in the domain! needs recovery?\n",
798 bit); 844 bit);
799 response.packet.code = JOIN_DISALLOW; 845 packet.code = JOIN_DISALLOW;
800 } else { 846 } else {
801 /* Alright we're fully a part of this domain 847 /* Alright we're fully a part of this domain
802 * so we keep some state as to who's joining 848 * so we keep some state as to who's joining
@@ -807,19 +853,15 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
807 if (dlm_query_join_proto_check("DLM", bit, 853 if (dlm_query_join_proto_check("DLM", bit,
808 &dlm->dlm_locking_proto, 854 &dlm->dlm_locking_proto,
809 &query->dlm_proto)) { 855 &query->dlm_proto)) {
810 response.packet.code = 856 packet.code = JOIN_PROTOCOL_MISMATCH;
811 JOIN_PROTOCOL_MISMATCH;
812 } else if (dlm_query_join_proto_check("fs", bit, 857 } else if (dlm_query_join_proto_check("fs", bit,
813 &dlm->fs_locking_proto, 858 &dlm->fs_locking_proto,
814 &query->fs_proto)) { 859 &query->fs_proto)) {
815 response.packet.code = 860 packet.code = JOIN_PROTOCOL_MISMATCH;
816 JOIN_PROTOCOL_MISMATCH;
817 } else { 861 } else {
818 response.packet.dlm_minor = 862 packet.dlm_minor = query->dlm_proto.pv_minor;
819 query->dlm_proto.pv_minor; 863 packet.fs_minor = query->fs_proto.pv_minor;
820 response.packet.fs_minor = 864 packet.code = JOIN_OK;
821 query->fs_proto.pv_minor;
822 response.packet.code = JOIN_OK;
823 __dlm_set_joining_node(dlm, query->node_idx); 865 __dlm_set_joining_node(dlm, query->node_idx);
824 } 866 }
825 } 867 }
@@ -830,9 +872,10 @@ unlock_respond:
830 spin_unlock(&dlm_domain_lock); 872 spin_unlock(&dlm_domain_lock);
831 873
832respond: 874respond:
833 mlog(0, "We respond with %u\n", response.packet.code); 875 mlog(0, "We respond with %u\n", packet.code);
834 876
835 return response.intval; 877 dlm_query_join_packet_to_wire(&packet, &response);
878 return response;
836} 879}
837 880
838static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data, 881static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
@@ -937,7 +980,7 @@ static int dlm_send_join_cancels(struct dlm_ctxt *dlm,
937 sizeof(unsigned long))) { 980 sizeof(unsigned long))) {
938 mlog(ML_ERROR, 981 mlog(ML_ERROR,
939 "map_size %u != BITS_TO_LONGS(O2NM_MAX_NODES) %u\n", 982 "map_size %u != BITS_TO_LONGS(O2NM_MAX_NODES) %u\n",
940 map_size, BITS_TO_LONGS(O2NM_MAX_NODES)); 983 map_size, (unsigned)BITS_TO_LONGS(O2NM_MAX_NODES));
941 return -EINVAL; 984 return -EINVAL;
942 } 985 }
943 986
@@ -968,7 +1011,8 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
968{ 1011{
969 int status; 1012 int status;
970 struct dlm_query_join_request join_msg; 1013 struct dlm_query_join_request join_msg;
971 union dlm_query_join_response join_resp; 1014 struct dlm_query_join_packet packet;
1015 u32 join_resp;
972 1016
973 mlog(0, "querying node %d\n", node); 1017 mlog(0, "querying node %d\n", node);
974 1018
@@ -984,11 +1028,12 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
984 1028
985 status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg, 1029 status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg,
986 sizeof(join_msg), node, 1030 sizeof(join_msg), node,
987 &join_resp.intval); 1031 &join_resp);
988 if (status < 0 && status != -ENOPROTOOPT) { 1032 if (status < 0 && status != -ENOPROTOOPT) {
989 mlog_errno(status); 1033 mlog_errno(status);
990 goto bail; 1034 goto bail;
991 } 1035 }
1036 dlm_query_join_wire_to_packet(join_resp, &packet);
992 1037
993 /* -ENOPROTOOPT from the net code means the other side isn't 1038 /* -ENOPROTOOPT from the net code means the other side isn't
994 listening for our message type -- that's fine, it means 1039 listening for our message type -- that's fine, it means
@@ -997,10 +1042,10 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
997 if (status == -ENOPROTOOPT) { 1042 if (status == -ENOPROTOOPT) {
998 status = 0; 1043 status = 0;
999 *response = JOIN_OK_NO_MAP; 1044 *response = JOIN_OK_NO_MAP;
1000 } else if (join_resp.packet.code == JOIN_DISALLOW || 1045 } else if (packet.code == JOIN_DISALLOW ||
1001 join_resp.packet.code == JOIN_OK_NO_MAP) { 1046 packet.code == JOIN_OK_NO_MAP) {
1002 *response = join_resp.packet.code; 1047 *response = packet.code;
1003 } else if (join_resp.packet.code == JOIN_PROTOCOL_MISMATCH) { 1048 } else if (packet.code == JOIN_PROTOCOL_MISMATCH) {
1004 mlog(ML_NOTICE, 1049 mlog(ML_NOTICE,
1005 "This node requested DLM locking protocol %u.%u and " 1050 "This node requested DLM locking protocol %u.%u and "
1006 "filesystem locking protocol %u.%u. At least one of " 1051 "filesystem locking protocol %u.%u. At least one of "
@@ -1012,14 +1057,12 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
1012 dlm->fs_locking_proto.pv_minor, 1057 dlm->fs_locking_proto.pv_minor,
1013 node); 1058 node);
1014 status = -EPROTO; 1059 status = -EPROTO;
1015 *response = join_resp.packet.code; 1060 *response = packet.code;
1016 } else if (join_resp.packet.code == JOIN_OK) { 1061 } else if (packet.code == JOIN_OK) {
1017 *response = join_resp.packet.code; 1062 *response = packet.code;
1018 /* Use the same locking protocol as the remote node */ 1063 /* Use the same locking protocol as the remote node */
1019 dlm->dlm_locking_proto.pv_minor = 1064 dlm->dlm_locking_proto.pv_minor = packet.dlm_minor;
1020 join_resp.packet.dlm_minor; 1065 dlm->fs_locking_proto.pv_minor = packet.fs_minor;
1021 dlm->fs_locking_proto.pv_minor =
1022 join_resp.packet.fs_minor;
1023 mlog(0, 1066 mlog(0,
1024 "Node %d responds JOIN_OK with DLM locking protocol " 1067 "Node %d responds JOIN_OK with DLM locking protocol "
1025 "%u.%u and fs locking protocol %u.%u\n", 1068 "%u.%u and fs locking protocol %u.%u\n",
@@ -1031,11 +1074,11 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
1031 } else { 1074 } else {
1032 status = -EINVAL; 1075 status = -EINVAL;
1033 mlog(ML_ERROR, "invalid response %d from node %u\n", 1076 mlog(ML_ERROR, "invalid response %d from node %u\n",
1034 join_resp.packet.code, node); 1077 packet.code, node);
1035 } 1078 }
1036 1079
1037 mlog(0, "status %d, node %d response is %d\n", status, node, 1080 mlog(0, "status %d, node %d response is %d\n", status, node,
1038 *response); 1081 *response);
1039 1082
1040bail: 1083bail:
1041 return status; 1084 return status;
@@ -1376,6 +1419,12 @@ static int dlm_join_domain(struct dlm_ctxt *dlm)
1376 goto bail; 1419 goto bail;
1377 } 1420 }
1378 1421
1422 status = dlm_debug_init(dlm);
1423 if (status < 0) {
1424 mlog_errno(status);
1425 goto bail;
1426 }
1427
1379 status = dlm_launch_thread(dlm); 1428 status = dlm_launch_thread(dlm);
1380 if (status < 0) { 1429 if (status < 0) {
1381 mlog_errno(status); 1430 mlog_errno(status);
@@ -1443,6 +1492,7 @@ bail:
1443 1492
1444 if (status) { 1493 if (status) {
1445 dlm_unregister_domain_handlers(dlm); 1494 dlm_unregister_domain_handlers(dlm);
1495 dlm_debug_shutdown(dlm);
1446 dlm_complete_thread(dlm); 1496 dlm_complete_thread(dlm);
1447 dlm_complete_recovery_thread(dlm); 1497 dlm_complete_recovery_thread(dlm);
1448 dlm_destroy_dlm_worker(dlm); 1498 dlm_destroy_dlm_worker(dlm);
@@ -1455,6 +1505,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
1455 u32 key) 1505 u32 key)
1456{ 1506{
1457 int i; 1507 int i;
1508 int ret;
1458 struct dlm_ctxt *dlm = NULL; 1509 struct dlm_ctxt *dlm = NULL;
1459 1510
1460 dlm = kzalloc(sizeof(*dlm), GFP_KERNEL); 1511 dlm = kzalloc(sizeof(*dlm), GFP_KERNEL);
@@ -1487,6 +1538,15 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
1487 dlm->key = key; 1538 dlm->key = key;
1488 dlm->node_num = o2nm_this_node(); 1539 dlm->node_num = o2nm_this_node();
1489 1540
1541 ret = dlm_create_debugfs_subroot(dlm);
1542 if (ret < 0) {
1543 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES);
1544 kfree(dlm->name);
1545 kfree(dlm);
1546 dlm = NULL;
1547 goto leave;
1548 }
1549
1490 spin_lock_init(&dlm->spinlock); 1550 spin_lock_init(&dlm->spinlock);
1491 spin_lock_init(&dlm->master_lock); 1551 spin_lock_init(&dlm->master_lock);
1492 spin_lock_init(&dlm->ast_lock); 1552 spin_lock_init(&dlm->ast_lock);
@@ -1497,6 +1557,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
1497 INIT_LIST_HEAD(&dlm->reco.node_data); 1557 INIT_LIST_HEAD(&dlm->reco.node_data);
1498 INIT_LIST_HEAD(&dlm->purge_list); 1558 INIT_LIST_HEAD(&dlm->purge_list);
1499 INIT_LIST_HEAD(&dlm->dlm_domain_handlers); 1559 INIT_LIST_HEAD(&dlm->dlm_domain_handlers);
1560 INIT_LIST_HEAD(&dlm->tracking_list);
1500 dlm->reco.state = 0; 1561 dlm->reco.state = 0;
1501 1562
1502 INIT_LIST_HEAD(&dlm->pending_asts); 1563 INIT_LIST_HEAD(&dlm->pending_asts);
@@ -1787,21 +1848,49 @@ static int __init dlm_init(void)
1787 dlm_print_version(); 1848 dlm_print_version();
1788 1849
1789 status = dlm_init_mle_cache(); 1850 status = dlm_init_mle_cache();
1790 if (status) 1851 if (status) {
1791 return -1; 1852 mlog(ML_ERROR, "Could not create o2dlm_mle slabcache\n");
1853 goto error;
1854 }
1855
1856 status = dlm_init_master_caches();
1857 if (status) {
1858 mlog(ML_ERROR, "Could not create o2dlm_lockres and "
1859 "o2dlm_lockname slabcaches\n");
1860 goto error;
1861 }
1862
1863 status = dlm_init_lock_cache();
1864 if (status) {
1865 mlog(ML_ERROR, "Count not create o2dlm_lock slabcache\n");
1866 goto error;
1867 }
1792 1868
1793 status = dlm_register_net_handlers(); 1869 status = dlm_register_net_handlers();
1794 if (status) { 1870 if (status) {
1795 dlm_destroy_mle_cache(); 1871 mlog(ML_ERROR, "Unable to register network handlers\n");
1796 return -1; 1872 goto error;
1797 } 1873 }
1798 1874
1875 status = dlm_create_debugfs_root();
1876 if (status)
1877 goto error;
1878
1799 return 0; 1879 return 0;
1880error:
1881 dlm_unregister_net_handlers();
1882 dlm_destroy_lock_cache();
1883 dlm_destroy_master_caches();
1884 dlm_destroy_mle_cache();
1885 return -1;
1800} 1886}
1801 1887
1802static void __exit dlm_exit (void) 1888static void __exit dlm_exit (void)
1803{ 1889{
1890 dlm_destroy_debugfs_root();
1804 dlm_unregister_net_handlers(); 1891 dlm_unregister_net_handlers();
1892 dlm_destroy_lock_cache();
1893 dlm_destroy_master_caches();
1805 dlm_destroy_mle_cache(); 1894 dlm_destroy_mle_cache();
1806} 1895}
1807 1896
diff --git a/fs/ocfs2/dlm/dlmlock.c b/fs/ocfs2/dlm/dlmlock.c
index 52578d907d9a..83a9f2972ac8 100644
--- a/fs/ocfs2/dlm/dlmlock.c
+++ b/fs/ocfs2/dlm/dlmlock.c
@@ -53,6 +53,8 @@
53#define MLOG_MASK_PREFIX ML_DLM 53#define MLOG_MASK_PREFIX ML_DLM
54#include "cluster/masklog.h" 54#include "cluster/masklog.h"
55 55
56static struct kmem_cache *dlm_lock_cache = NULL;
57
56static DEFINE_SPINLOCK(dlm_cookie_lock); 58static DEFINE_SPINLOCK(dlm_cookie_lock);
57static u64 dlm_next_cookie = 1; 59static u64 dlm_next_cookie = 1;
58 60
@@ -64,6 +66,22 @@ static void dlm_init_lock(struct dlm_lock *newlock, int type,
64static void dlm_lock_release(struct kref *kref); 66static void dlm_lock_release(struct kref *kref);
65static void dlm_lock_detach_lockres(struct dlm_lock *lock); 67static void dlm_lock_detach_lockres(struct dlm_lock *lock);
66 68
69int dlm_init_lock_cache(void)
70{
71 dlm_lock_cache = kmem_cache_create("o2dlm_lock",
72 sizeof(struct dlm_lock),
73 0, SLAB_HWCACHE_ALIGN, NULL);
74 if (dlm_lock_cache == NULL)
75 return -ENOMEM;
76 return 0;
77}
78
79void dlm_destroy_lock_cache(void)
80{
81 if (dlm_lock_cache)
82 kmem_cache_destroy(dlm_lock_cache);
83}
84
67/* Tell us whether we can grant a new lock request. 85/* Tell us whether we can grant a new lock request.
68 * locking: 86 * locking:
69 * caller needs: res->spinlock 87 * caller needs: res->spinlock
@@ -353,7 +371,7 @@ static void dlm_lock_release(struct kref *kref)
353 mlog(0, "freeing kernel-allocated lksb\n"); 371 mlog(0, "freeing kernel-allocated lksb\n");
354 kfree(lock->lksb); 372 kfree(lock->lksb);
355 } 373 }
356 kfree(lock); 374 kmem_cache_free(dlm_lock_cache, lock);
357} 375}
358 376
359/* associate a lock with it's lockres, getting a ref on the lockres */ 377/* associate a lock with it's lockres, getting a ref on the lockres */
@@ -412,7 +430,7 @@ struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie,
412 struct dlm_lock *lock; 430 struct dlm_lock *lock;
413 int kernel_allocated = 0; 431 int kernel_allocated = 0;
414 432
415 lock = kzalloc(sizeof(*lock), GFP_NOFS); 433 lock = (struct dlm_lock *) kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
416 if (!lock) 434 if (!lock)
417 return NULL; 435 return NULL;
418 436
diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
index a54d33d95ada..efc015c6128a 100644
--- a/fs/ocfs2/dlm/dlmmaster.c
+++ b/fs/ocfs2/dlm/dlmmaster.c
@@ -48,47 +48,11 @@
48#include "dlmapi.h" 48#include "dlmapi.h"
49#include "dlmcommon.h" 49#include "dlmcommon.h"
50#include "dlmdomain.h" 50#include "dlmdomain.h"
51#include "dlmdebug.h"
51 52
52#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_MASTER) 53#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_MASTER)
53#include "cluster/masklog.h" 54#include "cluster/masklog.h"
54 55
55enum dlm_mle_type {
56 DLM_MLE_BLOCK,
57 DLM_MLE_MASTER,
58 DLM_MLE_MIGRATION
59};
60
61struct dlm_lock_name
62{
63 u8 len;
64 u8 name[DLM_LOCKID_NAME_MAX];
65};
66
67struct dlm_master_list_entry
68{
69 struct list_head list;
70 struct list_head hb_events;
71 struct dlm_ctxt *dlm;
72 spinlock_t spinlock;
73 wait_queue_head_t wq;
74 atomic_t woken;
75 struct kref mle_refs;
76 int inuse;
77 unsigned long maybe_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
78 unsigned long vote_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
79 unsigned long response_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
80 unsigned long node_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
81 u8 master;
82 u8 new_master;
83 enum dlm_mle_type type;
84 struct o2hb_callback_func mle_hb_up;
85 struct o2hb_callback_func mle_hb_down;
86 union {
87 struct dlm_lock_resource *res;
88 struct dlm_lock_name name;
89 } u;
90};
91
92static void dlm_mle_node_down(struct dlm_ctxt *dlm, 56static void dlm_mle_node_down(struct dlm_ctxt *dlm,
93 struct dlm_master_list_entry *mle, 57 struct dlm_master_list_entry *mle,
94 struct o2nm_node *node, 58 struct o2nm_node *node,
@@ -128,98 +92,10 @@ static inline int dlm_mle_equal(struct dlm_ctxt *dlm,
128 return 1; 92 return 1;
129} 93}
130 94
131#define dlm_print_nodemap(m) _dlm_print_nodemap(m,#m) 95static struct kmem_cache *dlm_lockres_cache = NULL;
132static void _dlm_print_nodemap(unsigned long *map, const char *mapname) 96static struct kmem_cache *dlm_lockname_cache = NULL;
133{
134 int i;
135 printk("%s=[ ", mapname);
136 for (i=0; i<O2NM_MAX_NODES; i++)
137 if (test_bit(i, map))
138 printk("%d ", i);
139 printk("]");
140}
141
142static void dlm_print_one_mle(struct dlm_master_list_entry *mle)
143{
144 int refs;
145 char *type;
146 char attached;
147 u8 master;
148 unsigned int namelen;
149 const char *name;
150 struct kref *k;
151 unsigned long *maybe = mle->maybe_map,
152 *vote = mle->vote_map,
153 *resp = mle->response_map,
154 *node = mle->node_map;
155
156 k = &mle->mle_refs;
157 if (mle->type == DLM_MLE_BLOCK)
158 type = "BLK";
159 else if (mle->type == DLM_MLE_MASTER)
160 type = "MAS";
161 else
162 type = "MIG";
163 refs = atomic_read(&k->refcount);
164 master = mle->master;
165 attached = (list_empty(&mle->hb_events) ? 'N' : 'Y');
166
167 if (mle->type != DLM_MLE_MASTER) {
168 namelen = mle->u.name.len;
169 name = mle->u.name.name;
170 } else {
171 namelen = mle->u.res->lockname.len;
172 name = mle->u.res->lockname.name;
173 }
174
175 mlog(ML_NOTICE, "%.*s: %3s refs=%3d mas=%3u new=%3u evt=%c inuse=%d ",
176 namelen, name, type, refs, master, mle->new_master, attached,
177 mle->inuse);
178 dlm_print_nodemap(maybe);
179 printk(", ");
180 dlm_print_nodemap(vote);
181 printk(", ");
182 dlm_print_nodemap(resp);
183 printk(", ");
184 dlm_print_nodemap(node);
185 printk(", ");
186 printk("\n");
187}
188
189#if 0
190/* Code here is included but defined out as it aids debugging */
191
192static void dlm_dump_mles(struct dlm_ctxt *dlm)
193{
194 struct dlm_master_list_entry *mle;
195
196 mlog(ML_NOTICE, "dumping all mles for domain %s:\n", dlm->name);
197 spin_lock(&dlm->master_lock);
198 list_for_each_entry(mle, &dlm->master_list, list)
199 dlm_print_one_mle(mle);
200 spin_unlock(&dlm->master_lock);
201}
202
203int dlm_dump_all_mles(const char __user *data, unsigned int len)
204{
205 struct dlm_ctxt *dlm;
206
207 spin_lock(&dlm_domain_lock);
208 list_for_each_entry(dlm, &dlm_domains, list) {
209 mlog(ML_NOTICE, "found dlm: %p, name=%s\n", dlm, dlm->name);
210 dlm_dump_mles(dlm);
211 }
212 spin_unlock(&dlm_domain_lock);
213 return len;
214}
215EXPORT_SYMBOL_GPL(dlm_dump_all_mles);
216
217#endif /* 0 */
218
219
220static struct kmem_cache *dlm_mle_cache = NULL; 97static struct kmem_cache *dlm_mle_cache = NULL;
221 98
222
223static void dlm_mle_release(struct kref *kref); 99static void dlm_mle_release(struct kref *kref);
224static void dlm_init_mle(struct dlm_master_list_entry *mle, 100static void dlm_init_mle(struct dlm_master_list_entry *mle,
225 enum dlm_mle_type type, 101 enum dlm_mle_type type,
@@ -507,7 +383,7 @@ static void dlm_mle_node_up(struct dlm_ctxt *dlm,
507 383
508int dlm_init_mle_cache(void) 384int dlm_init_mle_cache(void)
509{ 385{
510 dlm_mle_cache = kmem_cache_create("dlm_mle_cache", 386 dlm_mle_cache = kmem_cache_create("o2dlm_mle",
511 sizeof(struct dlm_master_list_entry), 387 sizeof(struct dlm_master_list_entry),
512 0, SLAB_HWCACHE_ALIGN, 388 0, SLAB_HWCACHE_ALIGN,
513 NULL); 389 NULL);
@@ -560,6 +436,35 @@ static void dlm_mle_release(struct kref *kref)
560 * LOCK RESOURCE FUNCTIONS 436 * LOCK RESOURCE FUNCTIONS
561 */ 437 */
562 438
439int dlm_init_master_caches(void)
440{
441 dlm_lockres_cache = kmem_cache_create("o2dlm_lockres",
442 sizeof(struct dlm_lock_resource),
443 0, SLAB_HWCACHE_ALIGN, NULL);
444 if (!dlm_lockres_cache)
445 goto bail;
446
447 dlm_lockname_cache = kmem_cache_create("o2dlm_lockname",
448 DLM_LOCKID_NAME_MAX, 0,
449 SLAB_HWCACHE_ALIGN, NULL);
450 if (!dlm_lockname_cache)
451 goto bail;
452
453 return 0;
454bail:
455 dlm_destroy_master_caches();
456 return -ENOMEM;
457}
458
459void dlm_destroy_master_caches(void)
460{
461 if (dlm_lockname_cache)
462 kmem_cache_destroy(dlm_lockname_cache);
463
464 if (dlm_lockres_cache)
465 kmem_cache_destroy(dlm_lockres_cache);
466}
467
563static void dlm_set_lockres_owner(struct dlm_ctxt *dlm, 468static void dlm_set_lockres_owner(struct dlm_ctxt *dlm,
564 struct dlm_lock_resource *res, 469 struct dlm_lock_resource *res,
565 u8 owner) 470 u8 owner)
@@ -610,6 +515,14 @@ static void dlm_lockres_release(struct kref *kref)
610 mlog(0, "destroying lockres %.*s\n", res->lockname.len, 515 mlog(0, "destroying lockres %.*s\n", res->lockname.len,
611 res->lockname.name); 516 res->lockname.name);
612 517
518 if (!list_empty(&res->tracking))
519 list_del_init(&res->tracking);
520 else {
521 mlog(ML_ERROR, "Resource %.*s not on the Tracking list\n",
522 res->lockname.len, res->lockname.name);
523 dlm_print_one_lock_resource(res);
524 }
525
613 if (!hlist_unhashed(&res->hash_node) || 526 if (!hlist_unhashed(&res->hash_node) ||
614 !list_empty(&res->granted) || 527 !list_empty(&res->granted) ||
615 !list_empty(&res->converting) || 528 !list_empty(&res->converting) ||
@@ -642,9 +555,9 @@ static void dlm_lockres_release(struct kref *kref)
642 BUG_ON(!list_empty(&res->recovering)); 555 BUG_ON(!list_empty(&res->recovering));
643 BUG_ON(!list_empty(&res->purge)); 556 BUG_ON(!list_empty(&res->purge));
644 557
645 kfree(res->lockname.name); 558 kmem_cache_free(dlm_lockname_cache, (void *)res->lockname.name);
646 559
647 kfree(res); 560 kmem_cache_free(dlm_lockres_cache, res);
648} 561}
649 562
650void dlm_lockres_put(struct dlm_lock_resource *res) 563void dlm_lockres_put(struct dlm_lock_resource *res)
@@ -677,6 +590,7 @@ static void dlm_init_lockres(struct dlm_ctxt *dlm,
677 INIT_LIST_HEAD(&res->dirty); 590 INIT_LIST_HEAD(&res->dirty);
678 INIT_LIST_HEAD(&res->recovering); 591 INIT_LIST_HEAD(&res->recovering);
679 INIT_LIST_HEAD(&res->purge); 592 INIT_LIST_HEAD(&res->purge);
593 INIT_LIST_HEAD(&res->tracking);
680 atomic_set(&res->asts_reserved, 0); 594 atomic_set(&res->asts_reserved, 0);
681 res->migration_pending = 0; 595 res->migration_pending = 0;
682 res->inflight_locks = 0; 596 res->inflight_locks = 0;
@@ -692,6 +606,8 @@ static void dlm_init_lockres(struct dlm_ctxt *dlm,
692 606
693 res->last_used = 0; 607 res->last_used = 0;
694 608
609 list_add_tail(&res->tracking, &dlm->tracking_list);
610
695 memset(res->lvb, 0, DLM_LVB_LEN); 611 memset(res->lvb, 0, DLM_LVB_LEN);
696 memset(res->refmap, 0, sizeof(res->refmap)); 612 memset(res->refmap, 0, sizeof(res->refmap));
697} 613}
@@ -700,20 +616,28 @@ struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm,
700 const char *name, 616 const char *name,
701 unsigned int namelen) 617 unsigned int namelen)
702{ 618{
703 struct dlm_lock_resource *res; 619 struct dlm_lock_resource *res = NULL;
704 620
705 res = kmalloc(sizeof(struct dlm_lock_resource), GFP_NOFS); 621 res = (struct dlm_lock_resource *)
622 kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
706 if (!res) 623 if (!res)
707 return NULL; 624 goto error;
708 625
709 res->lockname.name = kmalloc(namelen, GFP_NOFS); 626 res->lockname.name = (char *)
710 if (!res->lockname.name) { 627 kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
711 kfree(res); 628 if (!res->lockname.name)
712 return NULL; 629 goto error;
713 }
714 630
715 dlm_init_lockres(dlm, res, name, namelen); 631 dlm_init_lockres(dlm, res, name, namelen);
716 return res; 632 return res;
633
634error:
635 if (res && res->lockname.name)
636 kmem_cache_free(dlm_lockname_cache, (void *)res->lockname.name);
637
638 if (res)
639 kmem_cache_free(dlm_lockres_cache, res);
640 return NULL;
717} 641}
718 642
719void __dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm, 643void __dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm,
@@ -1663,7 +1587,12 @@ way_up_top:
1663 dlm_put_mle(tmpmle); 1587 dlm_put_mle(tmpmle);
1664 } 1588 }
1665send_response: 1589send_response:
1666 1590 /*
1591 * __dlm_lookup_lockres() grabbed a reference to this lockres.
1592 * The reference is released by dlm_assert_master_worker() under
1593 * the call to dlm_dispatch_assert_master(). If
1594 * dlm_assert_master_worker() isn't called, we drop it here.
1595 */
1667 if (dispatch_assert) { 1596 if (dispatch_assert) {
1668 if (response != DLM_MASTER_RESP_YES) 1597 if (response != DLM_MASTER_RESP_YES)
1669 mlog(ML_ERROR, "invalid response %d\n", response); 1598 mlog(ML_ERROR, "invalid response %d\n", response);
@@ -1678,7 +1607,11 @@ send_response:
1678 if (ret < 0) { 1607 if (ret < 0) {
1679 mlog(ML_ERROR, "failed to dispatch assert master work\n"); 1608 mlog(ML_ERROR, "failed to dispatch assert master work\n");
1680 response = DLM_MASTER_RESP_ERROR; 1609 response = DLM_MASTER_RESP_ERROR;
1610 dlm_lockres_put(res);
1681 } 1611 }
1612 } else {
1613 if (res)
1614 dlm_lockres_put(res);
1682 } 1615 }
1683 1616
1684 dlm_put(dlm); 1617 dlm_put(dlm);
@@ -1695,9 +1628,9 @@ send_response:
1695 * can periodically run all locks owned by this node 1628 * can periodically run all locks owned by this node
1696 * and re-assert across the cluster... 1629 * and re-assert across the cluster...
1697 */ 1630 */
1698int dlm_do_assert_master(struct dlm_ctxt *dlm, 1631static int dlm_do_assert_master(struct dlm_ctxt *dlm,
1699 struct dlm_lock_resource *res, 1632 struct dlm_lock_resource *res,
1700 void *nodemap, u32 flags) 1633 void *nodemap, u32 flags)
1701{ 1634{
1702 struct dlm_assert_master assert; 1635 struct dlm_assert_master assert;
1703 int to, tmpret; 1636 int to, tmpret;
@@ -2348,7 +2281,7 @@ int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
2348 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref " 2281 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref "
2349 "but it is already dropped!\n", dlm->name, 2282 "but it is already dropped!\n", dlm->name,
2350 res->lockname.len, res->lockname.name, node); 2283 res->lockname.len, res->lockname.name, node);
2351 __dlm_print_one_lock_resource(res); 2284 dlm_print_one_lock_resource(res);
2352 } 2285 }
2353 ret = 0; 2286 ret = 0;
2354 goto done; 2287 goto done;
@@ -2408,7 +2341,7 @@ static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data)
2408 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref " 2341 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref "
2409 "but it is already dropped!\n", dlm->name, 2342 "but it is already dropped!\n", dlm->name,
2410 res->lockname.len, res->lockname.name, node); 2343 res->lockname.len, res->lockname.name, node);
2411 __dlm_print_one_lock_resource(res); 2344 dlm_print_one_lock_resource(res);
2412 } 2345 }
2413 2346
2414 dlm_lockres_put(res); 2347 dlm_lockres_put(res);
@@ -2933,6 +2866,9 @@ static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
2933 dlm_lockres_clear_refmap_bit(lock->ml.node, res); 2866 dlm_lockres_clear_refmap_bit(lock->ml.node, res);
2934 list_del_init(&lock->list); 2867 list_del_init(&lock->list);
2935 dlm_lock_put(lock); 2868 dlm_lock_put(lock);
2869 /* In a normal unlock, we would have added a
2870 * DLM_UNLOCK_FREE_LOCK action. Force it. */
2871 dlm_lock_put(lock);
2936 } 2872 }
2937 } 2873 }
2938 queue++; 2874 queue++;
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index 91f747b8a538..bcb9260c3735 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -519,9 +519,9 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm)
519 return 0; 519 return 0;
520 520
521master_here: 521master_here:
522 mlog(0, "(%d) mastering recovery of %s:%u here(this=%u)!\n", 522 mlog(ML_NOTICE, "(%d) Node %u is the Recovery Master for the Dead Node "
523 task_pid_nr(dlm->dlm_reco_thread_task), 523 "%u for Domain %s\n", task_pid_nr(dlm->dlm_reco_thread_task),
524 dlm->name, dlm->reco.dead_node, dlm->node_num); 524 dlm->node_num, dlm->reco.dead_node, dlm->name);
525 525
526 status = dlm_remaster_locks(dlm, dlm->reco.dead_node); 526 status = dlm_remaster_locks(dlm, dlm->reco.dead_node);
527 if (status < 0) { 527 if (status < 0) {
@@ -1191,7 +1191,7 @@ static int dlm_add_lock_to_array(struct dlm_lock *lock,
1191 (ml->type == LKM_EXMODE || 1191 (ml->type == LKM_EXMODE ||
1192 memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN))) { 1192 memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN))) {
1193 mlog(ML_ERROR, "mismatched lvbs!\n"); 1193 mlog(ML_ERROR, "mismatched lvbs!\n");
1194 __dlm_print_one_lock_resource(lock->lockres); 1194 dlm_print_one_lock_resource(lock->lockres);
1195 BUG(); 1195 BUG();
1196 } 1196 }
1197 memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN); 1197 memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN);
@@ -1327,6 +1327,7 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
1327 (struct dlm_migratable_lockres *)msg->buf; 1327 (struct dlm_migratable_lockres *)msg->buf;
1328 int ret = 0; 1328 int ret = 0;
1329 u8 real_master; 1329 u8 real_master;
1330 u8 extra_refs = 0;
1330 char *buf = NULL; 1331 char *buf = NULL;
1331 struct dlm_work_item *item = NULL; 1332 struct dlm_work_item *item = NULL;
1332 struct dlm_lock_resource *res = NULL; 1333 struct dlm_lock_resource *res = NULL;
@@ -1404,16 +1405,28 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
1404 __dlm_insert_lockres(dlm, res); 1405 __dlm_insert_lockres(dlm, res);
1405 spin_unlock(&dlm->spinlock); 1406 spin_unlock(&dlm->spinlock);
1406 1407
1408 /* Add an extra ref for this lock-less lockres lest the
1409 * dlm_thread purges it before we get the chance to add
1410 * locks to it */
1411 dlm_lockres_get(res);
1412
1413 /* There are three refs that need to be put.
1414 * 1. Taken above.
1415 * 2. kref_init in dlm_new_lockres()->dlm_init_lockres().
1416 * 3. dlm_lookup_lockres()
1417 * The first one is handled at the end of this function. The
1418 * other two are handled in the worker thread after locks have
1419 * been attached. Yes, we don't wait for purge time to match
1420 * kref_init. The lockres will still have atleast one ref
1421 * added because it is in the hash __dlm_insert_lockres() */
1422 extra_refs++;
1423
1407 /* now that the new lockres is inserted, 1424 /* now that the new lockres is inserted,
1408 * make it usable by other processes */ 1425 * make it usable by other processes */
1409 spin_lock(&res->spinlock); 1426 spin_lock(&res->spinlock);
1410 res->state &= ~DLM_LOCK_RES_IN_PROGRESS; 1427 res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
1411 spin_unlock(&res->spinlock); 1428 spin_unlock(&res->spinlock);
1412 wake_up(&res->wq); 1429 wake_up(&res->wq);
1413
1414 /* add an extra ref for just-allocated lockres
1415 * otherwise the lockres will be purged immediately */
1416 dlm_lockres_get(res);
1417 } 1430 }
1418 1431
1419 /* at this point we have allocated everything we need, 1432 /* at this point we have allocated everything we need,
@@ -1443,12 +1456,17 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
1443 dlm_init_work_item(dlm, item, dlm_mig_lockres_worker, buf); 1456 dlm_init_work_item(dlm, item, dlm_mig_lockres_worker, buf);
1444 item->u.ml.lockres = res; /* already have a ref */ 1457 item->u.ml.lockres = res; /* already have a ref */
1445 item->u.ml.real_master = real_master; 1458 item->u.ml.real_master = real_master;
1459 item->u.ml.extra_ref = extra_refs;
1446 spin_lock(&dlm->work_lock); 1460 spin_lock(&dlm->work_lock);
1447 list_add_tail(&item->list, &dlm->work_list); 1461 list_add_tail(&item->list, &dlm->work_list);
1448 spin_unlock(&dlm->work_lock); 1462 spin_unlock(&dlm->work_lock);
1449 queue_work(dlm->dlm_worker, &dlm->dispatched_work); 1463 queue_work(dlm->dlm_worker, &dlm->dispatched_work);
1450 1464
1451leave: 1465leave:
1466 /* One extra ref taken needs to be put here */
1467 if (extra_refs)
1468 dlm_lockres_put(res);
1469
1452 dlm_put(dlm); 1470 dlm_put(dlm);
1453 if (ret < 0) { 1471 if (ret < 0) {
1454 if (buf) 1472 if (buf)
@@ -1464,17 +1482,19 @@ leave:
1464 1482
1465static void dlm_mig_lockres_worker(struct dlm_work_item *item, void *data) 1483static void dlm_mig_lockres_worker(struct dlm_work_item *item, void *data)
1466{ 1484{
1467 struct dlm_ctxt *dlm = data; 1485 struct dlm_ctxt *dlm;
1468 struct dlm_migratable_lockres *mres; 1486 struct dlm_migratable_lockres *mres;
1469 int ret = 0; 1487 int ret = 0;
1470 struct dlm_lock_resource *res; 1488 struct dlm_lock_resource *res;
1471 u8 real_master; 1489 u8 real_master;
1490 u8 extra_ref;
1472 1491
1473 dlm = item->dlm; 1492 dlm = item->dlm;
1474 mres = (struct dlm_migratable_lockres *)data; 1493 mres = (struct dlm_migratable_lockres *)data;
1475 1494
1476 res = item->u.ml.lockres; 1495 res = item->u.ml.lockres;
1477 real_master = item->u.ml.real_master; 1496 real_master = item->u.ml.real_master;
1497 extra_ref = item->u.ml.extra_ref;
1478 1498
1479 if (real_master == DLM_LOCK_RES_OWNER_UNKNOWN) { 1499 if (real_master == DLM_LOCK_RES_OWNER_UNKNOWN) {
1480 /* this case is super-rare. only occurs if 1500 /* this case is super-rare. only occurs if
@@ -1517,6 +1537,12 @@ again:
1517 } 1537 }
1518 1538
1519leave: 1539leave:
1540 /* See comment in dlm_mig_lockres_handler() */
1541 if (res) {
1542 if (extra_ref)
1543 dlm_lockres_put(res);
1544 dlm_lockres_put(res);
1545 }
1520 kfree(data); 1546 kfree(data);
1521 mlog_exit(ret); 1547 mlog_exit(ret);
1522} 1548}
@@ -1644,7 +1670,8 @@ int dlm_master_requery_handler(struct o2net_msg *msg, u32 len, void *data,
1644 /* retry!? */ 1670 /* retry!? */
1645 BUG(); 1671 BUG();
1646 } 1672 }
1647 } 1673 } else /* put.. incase we are not the master */
1674 dlm_lockres_put(res);
1648 spin_unlock(&res->spinlock); 1675 spin_unlock(&res->spinlock);
1649 } 1676 }
1650 spin_unlock(&dlm->spinlock); 1677 spin_unlock(&dlm->spinlock);
@@ -1921,6 +1948,7 @@ void dlm_move_lockres_to_recovery_list(struct dlm_ctxt *dlm,
1921 "Recovering res %s:%.*s, is already on recovery list!\n", 1948 "Recovering res %s:%.*s, is already on recovery list!\n",
1922 dlm->name, res->lockname.len, res->lockname.name); 1949 dlm->name, res->lockname.len, res->lockname.name);
1923 list_del_init(&res->recovering); 1950 list_del_init(&res->recovering);
1951 dlm_lockres_put(res);
1924 } 1952 }
1925 /* We need to hold a reference while on the recovery list */ 1953 /* We need to hold a reference while on the recovery list */
1926 dlm_lockres_get(res); 1954 dlm_lockres_get(res);
@@ -2130,11 +2158,16 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm,
2130 assert_spin_locked(&dlm->spinlock); 2158 assert_spin_locked(&dlm->spinlock);
2131 assert_spin_locked(&res->spinlock); 2159 assert_spin_locked(&res->spinlock);
2132 2160
2161 /* We do two dlm_lock_put(). One for removing from list and the other is
2162 * to force the DLM_UNLOCK_FREE_LOCK action so as to free the locks */
2163
2133 /* TODO: check pending_asts, pending_basts here */ 2164 /* TODO: check pending_asts, pending_basts here */
2134 list_for_each_entry_safe(lock, next, &res->granted, list) { 2165 list_for_each_entry_safe(lock, next, &res->granted, list) {
2135 if (lock->ml.node == dead_node) { 2166 if (lock->ml.node == dead_node) {
2136 list_del_init(&lock->list); 2167 list_del_init(&lock->list);
2137 dlm_lock_put(lock); 2168 dlm_lock_put(lock);
2169 /* Can't schedule DLM_UNLOCK_FREE_LOCK - do manually */
2170 dlm_lock_put(lock);
2138 freed++; 2171 freed++;
2139 } 2172 }
2140 } 2173 }
@@ -2142,6 +2175,8 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm,
2142 if (lock->ml.node == dead_node) { 2175 if (lock->ml.node == dead_node) {
2143 list_del_init(&lock->list); 2176 list_del_init(&lock->list);
2144 dlm_lock_put(lock); 2177 dlm_lock_put(lock);
2178 /* Can't schedule DLM_UNLOCK_FREE_LOCK - do manually */
2179 dlm_lock_put(lock);
2145 freed++; 2180 freed++;
2146 } 2181 }
2147 } 2182 }
@@ -2149,6 +2184,8 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm,
2149 if (lock->ml.node == dead_node) { 2184 if (lock->ml.node == dead_node) {
2150 list_del_init(&lock->list); 2185 list_del_init(&lock->list);
2151 dlm_lock_put(lock); 2186 dlm_lock_put(lock);
2187 /* Can't schedule DLM_UNLOCK_FREE_LOCK - do manually */
2188 dlm_lock_put(lock);
2152 freed++; 2189 freed++;
2153 } 2190 }
2154 } 2191 }
diff --git a/fs/ocfs2/dlm/dlmthread.c b/fs/ocfs2/dlm/dlmthread.c
index cebd089f8955..4060bb328bc8 100644
--- a/fs/ocfs2/dlm/dlmthread.c
+++ b/fs/ocfs2/dlm/dlmthread.c
@@ -176,12 +176,14 @@ static int dlm_purge_lockres(struct dlm_ctxt *dlm,
176 res->lockname.name, master); 176 res->lockname.name, master);
177 177
178 if (!master) { 178 if (!master) {
179 /* drop spinlock... retake below */
180 spin_unlock(&dlm->spinlock);
181
179 spin_lock(&res->spinlock); 182 spin_lock(&res->spinlock);
180 /* This ensures that clear refmap is sent after the set */ 183 /* This ensures that clear refmap is sent after the set */
181 __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG); 184 __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG);
182 spin_unlock(&res->spinlock); 185 spin_unlock(&res->spinlock);
183 /* drop spinlock to do messaging, retake below */ 186
184 spin_unlock(&dlm->spinlock);
185 /* clear our bit from the master's refmap, ignore errors */ 187 /* clear our bit from the master's refmap, ignore errors */
186 ret = dlm_drop_lockres_ref(dlm, res); 188 ret = dlm_drop_lockres_ref(dlm, res);
187 if (ret < 0) { 189 if (ret < 0) {
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-14 02:06:25 -0500