Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1 | /* -*- mode: c; c-basic-offset: 8; -*- |
| 2 | * vim: noexpandtab sw=8 ts=8 sts=0: |
| 3 | * |
| 4 | * dlmglue.c |
| 5 | * |
| 6 | * Code which implements an OCFS2 specific interface to our DLM. |
| 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/mm.h> |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 30 | #include <linux/crc32.h> |
| 31 | #include <linux/kthread.h> |
| 32 | #include <linux/pagemap.h> |
| 33 | #include <linux/debugfs.h> |
| 34 | #include <linux/seq_file.h> |
| 35 | |
| 36 | #include <cluster/heartbeat.h> |
| 37 | #include <cluster/nodemanager.h> |
| 38 | #include <cluster/tcp.h> |
| 39 | |
| 40 | #include <dlm/dlmapi.h> |
| 41 | |
| 42 | #define MLOG_MASK_PREFIX ML_DLM_GLUE |
| 43 | #include <cluster/masklog.h> |
| 44 | |
| 45 | #include "ocfs2.h" |
Joel Becker | d24fbcd | 2008-01-25 17:02:21 -0800 | [diff] [blame] | 46 | #include "ocfs2_lockingver.h" |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 47 | |
| 48 | #include "alloc.h" |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 49 | #include "dcache.h" |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 50 | #include "dlmglue.h" |
| 51 | #include "extent_map.h" |
Tiger Yang | 7f1a37e | 2006-11-15 15:48:42 +0800 | [diff] [blame] | 52 | #include "file.h" |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 53 | #include "heartbeat.h" |
| 54 | #include "inode.h" |
| 55 | #include "journal.h" |
| 56 | #include "slot_map.h" |
| 57 | #include "super.h" |
| 58 | #include "uptodate.h" |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 59 | |
| 60 | #include "buffer_head_io.h" |
| 61 | |
| 62 | struct ocfs2_mask_waiter { |
| 63 | struct list_head mw_item; |
| 64 | int mw_status; |
| 65 | struct completion mw_complete; |
| 66 | unsigned long mw_mask; |
| 67 | unsigned long mw_goal; |
| 68 | }; |
| 69 | |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 70 | static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres); |
| 71 | static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres); |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 72 | static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 73 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 74 | /* |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 75 | * Return value from ->downconvert_worker functions. |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 76 | * |
Mark Fasheh | b5e500e | 2006-09-13 22:01:16 -0700 | [diff] [blame] | 77 | * These control the precise actions of ocfs2_unblock_lock() |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 78 | * and ocfs2_process_blocked_lock() |
| 79 | * |
| 80 | */ |
| 81 | enum ocfs2_unblock_action { |
| 82 | UNBLOCK_CONTINUE = 0, /* Continue downconvert */ |
| 83 | UNBLOCK_CONTINUE_POST = 1, /* Continue downconvert, fire |
| 84 | * ->post_unlock callback */ |
| 85 | UNBLOCK_STOP_POST = 2, /* Do not downconvert, fire |
| 86 | * ->post_unlock() callback. */ |
| 87 | }; |
| 88 | |
| 89 | struct ocfs2_unblock_ctl { |
| 90 | int requeue; |
| 91 | enum ocfs2_unblock_action unblock_action; |
| 92 | }; |
| 93 | |
Mark Fasheh | 810d5ae | 2006-09-13 21:39:52 -0700 | [diff] [blame] | 94 | static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, |
| 95 | int new_level); |
| 96 | static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres); |
| 97 | |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 98 | static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, |
| 99 | int blocking); |
| 100 | |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 101 | static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres, |
| 102 | int blocking); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 103 | |
| 104 | static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, |
| 105 | struct ocfs2_lock_res *lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 106 | |
Adrian Bunk | 6cb129f | 2007-04-26 00:29:35 -0700 | [diff] [blame] | 107 | |
| 108 | #define mlog_meta_lvb(__level, __lockres) ocfs2_dump_meta_lvb_info(__level, __PRETTY_FUNCTION__, __LINE__, __lockres) |
| 109 | |
| 110 | /* This aids in debugging situations where a bad LVB might be involved. */ |
| 111 | static void ocfs2_dump_meta_lvb_info(u64 level, |
| 112 | const char *function, |
| 113 | unsigned int line, |
| 114 | struct ocfs2_lock_res *lockres) |
| 115 | { |
| 116 | struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; |
| 117 | |
| 118 | mlog(level, "LVB information for %s (called from %s:%u):\n", |
| 119 | lockres->l_name, function, line); |
| 120 | mlog(level, "version: %u, clusters: %u, generation: 0x%x\n", |
| 121 | lvb->lvb_version, be32_to_cpu(lvb->lvb_iclusters), |
| 122 | be32_to_cpu(lvb->lvb_igeneration)); |
| 123 | mlog(level, "size: %llu, uid %u, gid %u, mode 0x%x\n", |
| 124 | (unsigned long long)be64_to_cpu(lvb->lvb_isize), |
| 125 | be32_to_cpu(lvb->lvb_iuid), be32_to_cpu(lvb->lvb_igid), |
| 126 | be16_to_cpu(lvb->lvb_imode)); |
| 127 | mlog(level, "nlink %u, atime_packed 0x%llx, ctime_packed 0x%llx, " |
| 128 | "mtime_packed 0x%llx iattr 0x%x\n", be16_to_cpu(lvb->lvb_inlink), |
| 129 | (long long)be64_to_cpu(lvb->lvb_iatime_packed), |
| 130 | (long long)be64_to_cpu(lvb->lvb_ictime_packed), |
| 131 | (long long)be64_to_cpu(lvb->lvb_imtime_packed), |
| 132 | be32_to_cpu(lvb->lvb_iattr)); |
| 133 | } |
| 134 | |
| 135 | |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 136 | /* |
| 137 | * OCFS2 Lock Resource Operations |
| 138 | * |
| 139 | * These fine tune the behavior of the generic dlmglue locking infrastructure. |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 140 | * |
| 141 | * The most basic of lock types can point ->l_priv to their respective |
| 142 | * struct ocfs2_super and allow the default actions to manage things. |
| 143 | * |
| 144 | * Right now, each lock type also needs to implement an init function, |
| 145 | * and trivial lock/unlock wrappers. ocfs2_simple_drop_lockres() |
| 146 | * should be called when the lock is no longer needed (i.e., object |
| 147 | * destruction time). |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 148 | */ |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 149 | struct ocfs2_lock_res_ops { |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 150 | /* |
| 151 | * Translate an ocfs2_lock_res * into an ocfs2_super *. Define |
| 152 | * this callback if ->l_priv is not an ocfs2_super pointer |
| 153 | */ |
| 154 | struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *); |
Mark Fasheh | b5e500e | 2006-09-13 22:01:16 -0700 | [diff] [blame] | 155 | |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 156 | /* |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 157 | * Optionally called in the downconvert thread after a |
| 158 | * successful downconvert. The lockres will not be referenced |
| 159 | * after this callback is called, so it is safe to free |
| 160 | * memory, etc. |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 161 | * |
| 162 | * The exact semantics of when this is called are controlled |
| 163 | * by ->downconvert_worker() |
| 164 | */ |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 165 | void (*post_unlock)(struct ocfs2_super *, struct ocfs2_lock_res *); |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 166 | |
| 167 | /* |
Mark Fasheh | 16d5b95 | 2006-09-13 21:10:12 -0700 | [diff] [blame] | 168 | * Allow a lock type to add checks to determine whether it is |
| 169 | * safe to downconvert a lock. Return 0 to re-queue the |
| 170 | * downconvert at a later time, nonzero to continue. |
| 171 | * |
| 172 | * For most locks, the default checks that there are no |
| 173 | * incompatible holders are sufficient. |
| 174 | * |
| 175 | * Called with the lockres spinlock held. |
| 176 | */ |
| 177 | int (*check_downconvert)(struct ocfs2_lock_res *, int); |
| 178 | |
| 179 | /* |
Mark Fasheh | 5ef0d4e | 2006-09-13 21:21:52 -0700 | [diff] [blame] | 180 | * Allows a lock type to populate the lock value block. This |
| 181 | * is called on downconvert, and when we drop a lock. |
| 182 | * |
| 183 | * Locks that want to use this should set LOCK_TYPE_USES_LVB |
| 184 | * in the flags field. |
| 185 | * |
| 186 | * Called with the lockres spinlock held. |
| 187 | */ |
| 188 | void (*set_lvb)(struct ocfs2_lock_res *); |
| 189 | |
| 190 | /* |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 191 | * Called from the downconvert thread when it is determined |
| 192 | * that a lock will be downconverted. This is called without |
| 193 | * any locks held so the function can do work that might |
| 194 | * schedule (syncing out data, etc). |
| 195 | * |
| 196 | * This should return any one of the ocfs2_unblock_action |
| 197 | * values, depending on what it wants the thread to do. |
| 198 | */ |
| 199 | int (*downconvert_worker)(struct ocfs2_lock_res *, int); |
| 200 | |
| 201 | /* |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 202 | * LOCK_TYPE_* flags which describe the specific requirements |
| 203 | * of a lock type. Descriptions of each individual flag follow. |
| 204 | */ |
| 205 | int flags; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 206 | }; |
| 207 | |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 208 | /* |
| 209 | * Some locks want to "refresh" potentially stale data when a |
| 210 | * meaningful (PRMODE or EXMODE) lock level is first obtained. If this |
| 211 | * flag is set, the OCFS2_LOCK_NEEDS_REFRESH flag will be set on the |
| 212 | * individual lockres l_flags member from the ast function. It is |
| 213 | * expected that the locking wrapper will clear the |
| 214 | * OCFS2_LOCK_NEEDS_REFRESH flag when done. |
| 215 | */ |
| 216 | #define LOCK_TYPE_REQUIRES_REFRESH 0x1 |
| 217 | |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 218 | /* |
Mark Fasheh | 5ef0d4e | 2006-09-13 21:21:52 -0700 | [diff] [blame] | 219 | * Indicate that a lock type makes use of the lock value block. The |
| 220 | * ->set_lvb lock type callback must be defined. |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 221 | */ |
| 222 | #define LOCK_TYPE_USES_LVB 0x2 |
| 223 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 224 | static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = { |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 225 | .get_osb = ocfs2_get_inode_osb, |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 226 | .flags = 0, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 227 | }; |
| 228 | |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 229 | static struct ocfs2_lock_res_ops ocfs2_inode_inode_lops = { |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 230 | .get_osb = ocfs2_get_inode_osb, |
Mark Fasheh | 810d5ae | 2006-09-13 21:39:52 -0700 | [diff] [blame] | 231 | .check_downconvert = ocfs2_check_meta_downconvert, |
| 232 | .set_lvb = ocfs2_set_meta_lvb, |
Mark Fasheh | f1f5406 | 2007-10-18 15:13:59 -0700 | [diff] [blame] | 233 | .downconvert_worker = ocfs2_data_convert_worker, |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 234 | .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 235 | }; |
| 236 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 237 | static struct ocfs2_lock_res_ops ocfs2_super_lops = { |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 238 | .flags = LOCK_TYPE_REQUIRES_REFRESH, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 239 | }; |
| 240 | |
| 241 | static struct ocfs2_lock_res_ops ocfs2_rename_lops = { |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 242 | .flags = 0, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 243 | }; |
| 244 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 245 | static struct ocfs2_lock_res_ops ocfs2_dentry_lops = { |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 246 | .get_osb = ocfs2_get_dentry_osb, |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 247 | .post_unlock = ocfs2_dentry_post_unlock, |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 248 | .downconvert_worker = ocfs2_dentry_convert_worker, |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 249 | .flags = 0, |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 250 | }; |
| 251 | |
Tiger Yang | 5000863 | 2007-03-20 16:01:38 -0700 | [diff] [blame] | 252 | static struct ocfs2_lock_res_ops ocfs2_inode_open_lops = { |
| 253 | .get_osb = ocfs2_get_inode_osb, |
| 254 | .flags = 0, |
| 255 | }; |
| 256 | |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 257 | static struct ocfs2_lock_res_ops ocfs2_flock_lops = { |
| 258 | .get_osb = ocfs2_get_file_osb, |
| 259 | .flags = 0, |
| 260 | }; |
| 261 | |
Joel Becker | d24fbcd | 2008-01-25 17:02:21 -0800 | [diff] [blame] | 262 | /* |
| 263 | * This is the filesystem locking protocol version. |
| 264 | * |
| 265 | * Whenever the filesystem does new things with locks (adds or removes a |
| 266 | * lock, orders them differently, does different things underneath a lock), |
| 267 | * the version must be changed. The protocol is negotiated when joining |
| 268 | * the dlm domain. A node may join the domain if its major version is |
| 269 | * identical to all other nodes and its minor version is greater than |
| 270 | * or equal to all other nodes. When its minor version is greater than |
| 271 | * the other nodes, it will run at the minor version specified by the |
| 272 | * other nodes. |
| 273 | * |
| 274 | * If a locking change is made that will not be compatible with older |
| 275 | * versions, the major number must be increased and the minor version set |
| 276 | * to zero. If a change merely adds a behavior that can be disabled when |
| 277 | * speaking to older versions, the minor version must be increased. If a |
| 278 | * change adds a fully backwards compatible change (eg, LVB changes that |
| 279 | * are just ignored by older versions), the version does not need to be |
| 280 | * updated. |
| 281 | */ |
| 282 | const struct dlm_protocol_version ocfs2_locking_protocol = { |
| 283 | .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, |
| 284 | .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, |
| 285 | }; |
| 286 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 287 | static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) |
| 288 | { |
| 289 | return lockres->l_type == OCFS2_LOCK_TYPE_META || |
Tiger Yang | 5000863 | 2007-03-20 16:01:38 -0700 | [diff] [blame] | 290 | lockres->l_type == OCFS2_LOCK_TYPE_RW || |
| 291 | lockres->l_type == OCFS2_LOCK_TYPE_OPEN; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 292 | } |
| 293 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 294 | static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres) |
| 295 | { |
| 296 | BUG_ON(!ocfs2_is_inode_lock(lockres)); |
| 297 | |
| 298 | return (struct inode *) lockres->l_priv; |
| 299 | } |
| 300 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 301 | static inline struct ocfs2_dentry_lock *ocfs2_lock_res_dl(struct ocfs2_lock_res *lockres) |
| 302 | { |
| 303 | BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_DENTRY); |
| 304 | |
| 305 | return (struct ocfs2_dentry_lock *)lockres->l_priv; |
| 306 | } |
| 307 | |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 308 | static inline struct ocfs2_super *ocfs2_get_lockres_osb(struct ocfs2_lock_res *lockres) |
| 309 | { |
| 310 | if (lockres->l_ops->get_osb) |
| 311 | return lockres->l_ops->get_osb(lockres); |
| 312 | |
| 313 | return (struct ocfs2_super *)lockres->l_priv; |
| 314 | } |
| 315 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 316 | static int ocfs2_lock_create(struct ocfs2_super *osb, |
| 317 | struct ocfs2_lock_res *lockres, |
| 318 | int level, |
| 319 | int dlm_flags); |
| 320 | static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, |
| 321 | int wanted); |
| 322 | static void ocfs2_cluster_unlock(struct ocfs2_super *osb, |
| 323 | struct ocfs2_lock_res *lockres, |
| 324 | int level); |
| 325 | static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres); |
| 326 | static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres); |
| 327 | static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres); |
| 328 | static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level); |
| 329 | static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, |
| 330 | struct ocfs2_lock_res *lockres); |
| 331 | static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, |
| 332 | int convert); |
| 333 | #define ocfs2_log_dlm_error(_func, _stat, _lockres) do { \ |
| 334 | mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on " \ |
| 335 | "resource %s: %s\n", dlm_errname(_stat), _func, \ |
| 336 | _lockres->l_name, dlm_errmsg(_stat)); \ |
| 337 | } while (0) |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 338 | static int ocfs2_downconvert_thread(void *arg); |
| 339 | static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, |
| 340 | struct ocfs2_lock_res *lockres); |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 341 | static int ocfs2_inode_lock_update(struct inode *inode, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 342 | struct buffer_head **bh); |
| 343 | static void ocfs2_drop_osb_locks(struct ocfs2_super *osb); |
| 344 | static inline int ocfs2_highest_compat_lock_level(int level); |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 345 | static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, |
| 346 | int new_level); |
| 347 | static int ocfs2_downconvert_lock(struct ocfs2_super *osb, |
| 348 | struct ocfs2_lock_res *lockres, |
| 349 | int new_level, |
| 350 | int lvb); |
| 351 | static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, |
| 352 | struct ocfs2_lock_res *lockres); |
| 353 | static int ocfs2_cancel_convert(struct ocfs2_super *osb, |
| 354 | struct ocfs2_lock_res *lockres); |
| 355 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 356 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 357 | static void ocfs2_build_lock_name(enum ocfs2_lock_type type, |
| 358 | u64 blkno, |
| 359 | u32 generation, |
| 360 | char *name) |
| 361 | { |
| 362 | int len; |
| 363 | |
| 364 | mlog_entry_void(); |
| 365 | |
| 366 | BUG_ON(type >= OCFS2_NUM_LOCK_TYPES); |
| 367 | |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 368 | len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x", |
| 369 | ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD, |
| 370 | (long long)blkno, generation); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 371 | |
| 372 | BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1)); |
| 373 | |
| 374 | mlog(0, "built lock resource with name: %s\n", name); |
| 375 | |
| 376 | mlog_exit_void(); |
| 377 | } |
| 378 | |
Ingo Molnar | 34af946 | 2006-06-27 02:53:55 -0700 | [diff] [blame] | 379 | static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 380 | |
| 381 | static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res, |
| 382 | struct ocfs2_dlm_debug *dlm_debug) |
| 383 | { |
| 384 | mlog(0, "Add tracking for lockres %s\n", res->l_name); |
| 385 | |
| 386 | spin_lock(&ocfs2_dlm_tracking_lock); |
| 387 | list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking); |
| 388 | spin_unlock(&ocfs2_dlm_tracking_lock); |
| 389 | } |
| 390 | |
| 391 | static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res) |
| 392 | { |
| 393 | spin_lock(&ocfs2_dlm_tracking_lock); |
| 394 | if (!list_empty(&res->l_debug_list)) |
| 395 | list_del_init(&res->l_debug_list); |
| 396 | spin_unlock(&ocfs2_dlm_tracking_lock); |
| 397 | } |
| 398 | |
| 399 | static void ocfs2_lock_res_init_common(struct ocfs2_super *osb, |
| 400 | struct ocfs2_lock_res *res, |
| 401 | enum ocfs2_lock_type type, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 402 | struct ocfs2_lock_res_ops *ops, |
| 403 | void *priv) |
| 404 | { |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 405 | res->l_type = type; |
| 406 | res->l_ops = ops; |
| 407 | res->l_priv = priv; |
| 408 | |
| 409 | res->l_level = LKM_IVMODE; |
| 410 | res->l_requested = LKM_IVMODE; |
| 411 | res->l_blocking = LKM_IVMODE; |
| 412 | res->l_action = OCFS2_AST_INVALID; |
| 413 | res->l_unlock_action = OCFS2_UNLOCK_INVALID; |
| 414 | |
| 415 | res->l_flags = OCFS2_LOCK_INITIALIZED; |
| 416 | |
| 417 | ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug); |
| 418 | } |
| 419 | |
| 420 | void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res) |
| 421 | { |
| 422 | /* This also clears out the lock status block */ |
| 423 | memset(res, 0, sizeof(struct ocfs2_lock_res)); |
| 424 | spin_lock_init(&res->l_lock); |
| 425 | init_waitqueue_head(&res->l_event); |
| 426 | INIT_LIST_HEAD(&res->l_blocked_list); |
| 427 | INIT_LIST_HEAD(&res->l_mask_waiters); |
| 428 | } |
| 429 | |
| 430 | void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res, |
| 431 | enum ocfs2_lock_type type, |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 432 | unsigned int generation, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 433 | struct inode *inode) |
| 434 | { |
| 435 | struct ocfs2_lock_res_ops *ops; |
| 436 | |
| 437 | switch(type) { |
| 438 | case OCFS2_LOCK_TYPE_RW: |
| 439 | ops = &ocfs2_inode_rw_lops; |
| 440 | break; |
| 441 | case OCFS2_LOCK_TYPE_META: |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 442 | ops = &ocfs2_inode_inode_lops; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 443 | break; |
Tiger Yang | 5000863 | 2007-03-20 16:01:38 -0700 | [diff] [blame] | 444 | case OCFS2_LOCK_TYPE_OPEN: |
| 445 | ops = &ocfs2_inode_open_lops; |
| 446 | break; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 447 | default: |
| 448 | mlog_bug_on_msg(1, "type: %d\n", type); |
| 449 | ops = NULL; /* thanks, gcc */ |
| 450 | break; |
| 451 | }; |
| 452 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 453 | ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno, |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 454 | generation, res->l_name); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 455 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode); |
| 456 | } |
| 457 | |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 458 | static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres) |
| 459 | { |
| 460 | struct inode *inode = ocfs2_lock_res_inode(lockres); |
| 461 | |
| 462 | return OCFS2_SB(inode->i_sb); |
| 463 | } |
| 464 | |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 465 | static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres) |
| 466 | { |
| 467 | struct ocfs2_file_private *fp = lockres->l_priv; |
| 468 | |
| 469 | return OCFS2_SB(fp->fp_file->f_mapping->host->i_sb); |
| 470 | } |
| 471 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 472 | static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres) |
| 473 | { |
| 474 | __be64 inode_blkno_be; |
| 475 | |
| 476 | memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], |
| 477 | sizeof(__be64)); |
| 478 | |
| 479 | return be64_to_cpu(inode_blkno_be); |
| 480 | } |
| 481 | |
Mark Fasheh | 54a7e75 | 2006-09-12 21:49:13 -0700 | [diff] [blame] | 482 | static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres) |
| 483 | { |
| 484 | struct ocfs2_dentry_lock *dl = lockres->l_priv; |
| 485 | |
| 486 | return OCFS2_SB(dl->dl_inode->i_sb); |
| 487 | } |
| 488 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 489 | void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl, |
| 490 | u64 parent, struct inode *inode) |
| 491 | { |
| 492 | int len; |
| 493 | u64 inode_blkno = OCFS2_I(inode)->ip_blkno; |
| 494 | __be64 inode_blkno_be = cpu_to_be64(inode_blkno); |
| 495 | struct ocfs2_lock_res *lockres = &dl->dl_lockres; |
| 496 | |
| 497 | ocfs2_lock_res_init_once(lockres); |
| 498 | |
| 499 | /* |
| 500 | * Unfortunately, the standard lock naming scheme won't work |
| 501 | * here because we have two 16 byte values to use. Instead, |
| 502 | * we'll stuff the inode number as a binary value. We still |
| 503 | * want error prints to show something without garbling the |
| 504 | * display, so drop a null byte in there before the inode |
| 505 | * number. A future version of OCFS2 will likely use all |
| 506 | * binary lock names. The stringified names have been a |
| 507 | * tremendous aid in debugging, but now that the debugfs |
| 508 | * interface exists, we can mangle things there if need be. |
| 509 | * |
| 510 | * NOTE: We also drop the standard "pad" value (the total lock |
| 511 | * name size stays the same though - the last part is all |
| 512 | * zeros due to the memset in ocfs2_lock_res_init_once() |
| 513 | */ |
| 514 | len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START, |
| 515 | "%c%016llx", |
| 516 | ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY), |
| 517 | (long long)parent); |
| 518 | |
| 519 | BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1)); |
| 520 | |
| 521 | memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be, |
| 522 | sizeof(__be64)); |
| 523 | |
| 524 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres, |
| 525 | OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops, |
| 526 | dl); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 527 | } |
| 528 | |
| 529 | static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res, |
| 530 | struct ocfs2_super *osb) |
| 531 | { |
| 532 | /* Superblock lockres doesn't come from a slab so we call init |
| 533 | * once on it manually. */ |
| 534 | ocfs2_lock_res_init_once(res); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 535 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO, |
| 536 | 0, res->l_name); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 537 | ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 538 | &ocfs2_super_lops, osb); |
| 539 | } |
| 540 | |
| 541 | static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res, |
| 542 | struct ocfs2_super *osb) |
| 543 | { |
| 544 | /* Rename lockres doesn't come from a slab so we call init |
| 545 | * once on it manually. */ |
| 546 | ocfs2_lock_res_init_once(res); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 547 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name); |
| 548 | ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 549 | &ocfs2_rename_lops, osb); |
| 550 | } |
| 551 | |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 552 | void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres, |
| 553 | struct ocfs2_file_private *fp) |
| 554 | { |
| 555 | struct inode *inode = fp->fp_file->f_mapping->host; |
| 556 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
| 557 | |
| 558 | ocfs2_lock_res_init_once(lockres); |
| 559 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_FLOCK, oi->ip_blkno, |
| 560 | inode->i_generation, lockres->l_name); |
| 561 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres, |
| 562 | OCFS2_LOCK_TYPE_FLOCK, &ocfs2_flock_lops, |
| 563 | fp); |
| 564 | lockres->l_flags |= OCFS2_LOCK_NOCACHE; |
| 565 | } |
| 566 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 567 | void ocfs2_lock_res_free(struct ocfs2_lock_res *res) |
| 568 | { |
| 569 | mlog_entry_void(); |
| 570 | |
| 571 | if (!(res->l_flags & OCFS2_LOCK_INITIALIZED)) |
| 572 | return; |
| 573 | |
| 574 | ocfs2_remove_lockres_tracking(res); |
| 575 | |
| 576 | mlog_bug_on_msg(!list_empty(&res->l_blocked_list), |
| 577 | "Lockres %s is on the blocked list\n", |
| 578 | res->l_name); |
| 579 | mlog_bug_on_msg(!list_empty(&res->l_mask_waiters), |
| 580 | "Lockres %s has mask waiters pending\n", |
| 581 | res->l_name); |
| 582 | mlog_bug_on_msg(spin_is_locked(&res->l_lock), |
| 583 | "Lockres %s is locked\n", |
| 584 | res->l_name); |
| 585 | mlog_bug_on_msg(res->l_ro_holders, |
| 586 | "Lockres %s has %u ro holders\n", |
| 587 | res->l_name, res->l_ro_holders); |
| 588 | mlog_bug_on_msg(res->l_ex_holders, |
| 589 | "Lockres %s has %u ex holders\n", |
| 590 | res->l_name, res->l_ex_holders); |
| 591 | |
| 592 | /* Need to clear out the lock status block for the dlm */ |
| 593 | memset(&res->l_lksb, 0, sizeof(res->l_lksb)); |
| 594 | |
| 595 | res->l_flags = 0UL; |
| 596 | mlog_exit_void(); |
| 597 | } |
| 598 | |
| 599 | static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, |
| 600 | int level) |
| 601 | { |
| 602 | mlog_entry_void(); |
| 603 | |
| 604 | BUG_ON(!lockres); |
| 605 | |
| 606 | switch(level) { |
| 607 | case LKM_EXMODE: |
| 608 | lockres->l_ex_holders++; |
| 609 | break; |
| 610 | case LKM_PRMODE: |
| 611 | lockres->l_ro_holders++; |
| 612 | break; |
| 613 | default: |
| 614 | BUG(); |
| 615 | } |
| 616 | |
| 617 | mlog_exit_void(); |
| 618 | } |
| 619 | |
| 620 | static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres, |
| 621 | int level) |
| 622 | { |
| 623 | mlog_entry_void(); |
| 624 | |
| 625 | BUG_ON(!lockres); |
| 626 | |
| 627 | switch(level) { |
| 628 | case LKM_EXMODE: |
| 629 | BUG_ON(!lockres->l_ex_holders); |
| 630 | lockres->l_ex_holders--; |
| 631 | break; |
| 632 | case LKM_PRMODE: |
| 633 | BUG_ON(!lockres->l_ro_holders); |
| 634 | lockres->l_ro_holders--; |
| 635 | break; |
| 636 | default: |
| 637 | BUG(); |
| 638 | } |
| 639 | mlog_exit_void(); |
| 640 | } |
| 641 | |
| 642 | /* WARNING: This function lives in a world where the only three lock |
| 643 | * levels are EX, PR, and NL. It *will* have to be adjusted when more |
| 644 | * lock types are added. */ |
| 645 | static inline int ocfs2_highest_compat_lock_level(int level) |
| 646 | { |
| 647 | int new_level = LKM_EXMODE; |
| 648 | |
| 649 | if (level == LKM_EXMODE) |
| 650 | new_level = LKM_NLMODE; |
| 651 | else if (level == LKM_PRMODE) |
| 652 | new_level = LKM_PRMODE; |
| 653 | return new_level; |
| 654 | } |
| 655 | |
| 656 | static void lockres_set_flags(struct ocfs2_lock_res *lockres, |
| 657 | unsigned long newflags) |
| 658 | { |
Christoph Hellwig | 800deef | 2007-05-17 16:03:13 +0200 | [diff] [blame] | 659 | struct ocfs2_mask_waiter *mw, *tmp; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 660 | |
| 661 | assert_spin_locked(&lockres->l_lock); |
| 662 | |
| 663 | lockres->l_flags = newflags; |
| 664 | |
Christoph Hellwig | 800deef | 2007-05-17 16:03:13 +0200 | [diff] [blame] | 665 | list_for_each_entry_safe(mw, tmp, &lockres->l_mask_waiters, mw_item) { |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 666 | if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal) |
| 667 | continue; |
| 668 | |
| 669 | list_del_init(&mw->mw_item); |
| 670 | mw->mw_status = 0; |
| 671 | complete(&mw->mw_complete); |
| 672 | } |
| 673 | } |
| 674 | static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or) |
| 675 | { |
| 676 | lockres_set_flags(lockres, lockres->l_flags | or); |
| 677 | } |
| 678 | static void lockres_clear_flags(struct ocfs2_lock_res *lockres, |
| 679 | unsigned long clear) |
| 680 | { |
| 681 | lockres_set_flags(lockres, lockres->l_flags & ~clear); |
| 682 | } |
| 683 | |
| 684 | static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres) |
| 685 | { |
| 686 | mlog_entry_void(); |
| 687 | |
| 688 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); |
| 689 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); |
| 690 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); |
| 691 | BUG_ON(lockres->l_blocking <= LKM_NLMODE); |
| 692 | |
| 693 | lockres->l_level = lockres->l_requested; |
| 694 | if (lockres->l_level <= |
| 695 | ocfs2_highest_compat_lock_level(lockres->l_blocking)) { |
| 696 | lockres->l_blocking = LKM_NLMODE; |
| 697 | lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED); |
| 698 | } |
| 699 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); |
| 700 | |
| 701 | mlog_exit_void(); |
| 702 | } |
| 703 | |
| 704 | static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres) |
| 705 | { |
| 706 | mlog_entry_void(); |
| 707 | |
| 708 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); |
| 709 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); |
| 710 | |
| 711 | /* Convert from RO to EX doesn't really need anything as our |
| 712 | * information is already up to data. Convert from NL to |
| 713 | * *anything* however should mark ourselves as needing an |
| 714 | * update */ |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 715 | if (lockres->l_level == LKM_NLMODE && |
| 716 | lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 717 | lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
| 718 | |
| 719 | lockres->l_level = lockres->l_requested; |
| 720 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); |
| 721 | |
| 722 | mlog_exit_void(); |
| 723 | } |
| 724 | |
| 725 | static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres) |
| 726 | { |
| 727 | mlog_entry_void(); |
| 728 | |
Roel Kluin | 3cf0c50 | 2007-10-27 00:20:36 +0200 | [diff] [blame] | 729 | BUG_ON((!(lockres->l_flags & OCFS2_LOCK_BUSY))); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 730 | BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); |
| 731 | |
| 732 | if (lockres->l_requested > LKM_NLMODE && |
Mark Fasheh | f625c97 | 2006-09-12 21:24:53 -0700 | [diff] [blame] | 733 | !(lockres->l_flags & OCFS2_LOCK_LOCAL) && |
| 734 | lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 735 | lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
| 736 | |
| 737 | lockres->l_level = lockres->l_requested; |
| 738 | lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED); |
| 739 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); |
| 740 | |
| 741 | mlog_exit_void(); |
| 742 | } |
| 743 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 744 | static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, |
| 745 | int level) |
| 746 | { |
| 747 | int needs_downconvert = 0; |
| 748 | mlog_entry_void(); |
| 749 | |
| 750 | assert_spin_locked(&lockres->l_lock); |
| 751 | |
| 752 | lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); |
| 753 | |
| 754 | if (level > lockres->l_blocking) { |
| 755 | /* only schedule a downconvert if we haven't already scheduled |
| 756 | * one that goes low enough to satisfy the level we're |
| 757 | * blocking. this also catches the case where we get |
| 758 | * duplicate BASTs */ |
| 759 | if (ocfs2_highest_compat_lock_level(level) < |
| 760 | ocfs2_highest_compat_lock_level(lockres->l_blocking)) |
| 761 | needs_downconvert = 1; |
| 762 | |
| 763 | lockres->l_blocking = level; |
| 764 | } |
| 765 | |
| 766 | mlog_exit(needs_downconvert); |
| 767 | return needs_downconvert; |
| 768 | } |
| 769 | |
Mark Fasheh | aa2623a | 2006-09-12 21:58:23 -0700 | [diff] [blame] | 770 | static void ocfs2_blocking_ast(void *opaque, int level) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 771 | { |
Mark Fasheh | aa2623a | 2006-09-12 21:58:23 -0700 | [diff] [blame] | 772 | struct ocfs2_lock_res *lockres = opaque; |
| 773 | struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 774 | int needs_downconvert; |
| 775 | unsigned long flags; |
| 776 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 777 | BUG_ON(level <= LKM_NLMODE); |
| 778 | |
Mark Fasheh | aa2623a | 2006-09-12 21:58:23 -0700 | [diff] [blame] | 779 | mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n", |
| 780 | lockres->l_name, level, lockres->l_level, |
| 781 | ocfs2_lock_type_string(lockres->l_type)); |
| 782 | |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 783 | /* |
| 784 | * We can skip the bast for locks which don't enable caching - |
| 785 | * they'll be dropped at the earliest possible time anyway. |
| 786 | */ |
| 787 | if (lockres->l_flags & OCFS2_LOCK_NOCACHE) |
| 788 | return; |
| 789 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 790 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 791 | needs_downconvert = ocfs2_generic_handle_bast(lockres, level); |
| 792 | if (needs_downconvert) |
| 793 | ocfs2_schedule_blocked_lock(osb, lockres); |
| 794 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 795 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 796 | wake_up(&lockres->l_event); |
| 797 | |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 798 | ocfs2_wake_downconvert_thread(osb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 799 | } |
| 800 | |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 801 | static void ocfs2_locking_ast(void *opaque) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 802 | { |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 803 | struct ocfs2_lock_res *lockres = opaque; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 804 | struct dlm_lockstatus *lksb = &lockres->l_lksb; |
| 805 | unsigned long flags; |
| 806 | |
| 807 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 808 | |
| 809 | if (lksb->status != DLM_NORMAL) { |
| 810 | mlog(ML_ERROR, "lockres %s: lksb status value of %u!\n", |
| 811 | lockres->l_name, lksb->status); |
| 812 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 813 | return; |
| 814 | } |
| 815 | |
| 816 | switch(lockres->l_action) { |
| 817 | case OCFS2_AST_ATTACH: |
| 818 | ocfs2_generic_handle_attach_action(lockres); |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 819 | lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 820 | break; |
| 821 | case OCFS2_AST_CONVERT: |
| 822 | ocfs2_generic_handle_convert_action(lockres); |
| 823 | break; |
| 824 | case OCFS2_AST_DOWNCONVERT: |
| 825 | ocfs2_generic_handle_downconvert_action(lockres); |
| 826 | break; |
| 827 | default: |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 828 | mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u " |
| 829 | "lockres flags = 0x%lx, unlock action: %u\n", |
| 830 | lockres->l_name, lockres->l_action, lockres->l_flags, |
| 831 | lockres->l_unlock_action); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 832 | BUG(); |
| 833 | } |
| 834 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 835 | /* set it to something invalid so if we get called again we |
| 836 | * can catch it. */ |
| 837 | lockres->l_action = OCFS2_AST_INVALID; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 838 | |
| 839 | wake_up(&lockres->l_event); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 840 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 841 | } |
| 842 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 843 | static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, |
| 844 | int convert) |
| 845 | { |
| 846 | unsigned long flags; |
| 847 | |
| 848 | mlog_entry_void(); |
| 849 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 850 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); |
| 851 | if (convert) |
| 852 | lockres->l_action = OCFS2_AST_INVALID; |
| 853 | else |
| 854 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; |
| 855 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 856 | |
| 857 | wake_up(&lockres->l_event); |
| 858 | mlog_exit_void(); |
| 859 | } |
| 860 | |
| 861 | /* Note: If we detect another process working on the lock (i.e., |
| 862 | * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller |
| 863 | * to do the right thing in that case. |
| 864 | */ |
| 865 | static int ocfs2_lock_create(struct ocfs2_super *osb, |
| 866 | struct ocfs2_lock_res *lockres, |
| 867 | int level, |
| 868 | int dlm_flags) |
| 869 | { |
| 870 | int ret = 0; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 871 | enum dlm_status status = DLM_NORMAL; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 872 | unsigned long flags; |
| 873 | |
| 874 | mlog_entry_void(); |
| 875 | |
| 876 | mlog(0, "lock %s, level = %d, flags = %d\n", lockres->l_name, level, |
| 877 | dlm_flags); |
| 878 | |
| 879 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 880 | if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) || |
| 881 | (lockres->l_flags & OCFS2_LOCK_BUSY)) { |
| 882 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 883 | goto bail; |
| 884 | } |
| 885 | |
| 886 | lockres->l_action = OCFS2_AST_ATTACH; |
| 887 | lockres->l_requested = level; |
| 888 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); |
| 889 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 890 | |
| 891 | status = dlmlock(osb->dlm, |
| 892 | level, |
| 893 | &lockres->l_lksb, |
| 894 | dlm_flags, |
| 895 | lockres->l_name, |
Mark Fasheh | f068106 | 2006-09-08 11:40:10 -0700 | [diff] [blame] | 896 | OCFS2_LOCK_ID_MAX_LEN - 1, |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 897 | ocfs2_locking_ast, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 898 | lockres, |
Mark Fasheh | aa2623a | 2006-09-12 21:58:23 -0700 | [diff] [blame] | 899 | ocfs2_blocking_ast); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 900 | if (status != DLM_NORMAL) { |
| 901 | ocfs2_log_dlm_error("dlmlock", status, lockres); |
| 902 | ret = -EINVAL; |
| 903 | ocfs2_recover_from_dlm_error(lockres, 1); |
| 904 | } |
| 905 | |
| 906 | mlog(0, "lock %s, successfull return from dlmlock\n", lockres->l_name); |
| 907 | |
| 908 | bail: |
| 909 | mlog_exit(ret); |
| 910 | return ret; |
| 911 | } |
| 912 | |
| 913 | static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres, |
| 914 | int flag) |
| 915 | { |
| 916 | unsigned long flags; |
| 917 | int ret; |
| 918 | |
| 919 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 920 | ret = lockres->l_flags & flag; |
| 921 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 922 | |
| 923 | return ret; |
| 924 | } |
| 925 | |
| 926 | static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres) |
| 927 | |
| 928 | { |
| 929 | wait_event(lockres->l_event, |
| 930 | !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY)); |
| 931 | } |
| 932 | |
| 933 | static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres) |
| 934 | |
| 935 | { |
| 936 | wait_event(lockres->l_event, |
| 937 | !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING)); |
| 938 | } |
| 939 | |
| 940 | /* predict what lock level we'll be dropping down to on behalf |
| 941 | * of another node, and return true if the currently wanted |
| 942 | * level will be compatible with it. */ |
| 943 | static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, |
| 944 | int wanted) |
| 945 | { |
| 946 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); |
| 947 | |
| 948 | return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking); |
| 949 | } |
| 950 | |
| 951 | static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw) |
| 952 | { |
| 953 | INIT_LIST_HEAD(&mw->mw_item); |
| 954 | init_completion(&mw->mw_complete); |
| 955 | } |
| 956 | |
| 957 | static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw) |
| 958 | { |
| 959 | wait_for_completion(&mw->mw_complete); |
| 960 | /* Re-arm the completion in case we want to wait on it again */ |
| 961 | INIT_COMPLETION(mw->mw_complete); |
| 962 | return mw->mw_status; |
| 963 | } |
| 964 | |
| 965 | static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres, |
| 966 | struct ocfs2_mask_waiter *mw, |
| 967 | unsigned long mask, |
| 968 | unsigned long goal) |
| 969 | { |
| 970 | BUG_ON(!list_empty(&mw->mw_item)); |
| 971 | |
| 972 | assert_spin_locked(&lockres->l_lock); |
| 973 | |
| 974 | list_add_tail(&mw->mw_item, &lockres->l_mask_waiters); |
| 975 | mw->mw_mask = mask; |
| 976 | mw->mw_goal = goal; |
| 977 | } |
| 978 | |
| 979 | /* returns 0 if the mw that was removed was already satisfied, -EBUSY |
| 980 | * if the mask still hadn't reached its goal */ |
| 981 | static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres, |
| 982 | struct ocfs2_mask_waiter *mw) |
| 983 | { |
| 984 | unsigned long flags; |
| 985 | int ret = 0; |
| 986 | |
| 987 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 988 | if (!list_empty(&mw->mw_item)) { |
| 989 | if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal) |
| 990 | ret = -EBUSY; |
| 991 | |
| 992 | list_del_init(&mw->mw_item); |
| 993 | init_completion(&mw->mw_complete); |
| 994 | } |
| 995 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 996 | |
| 997 | return ret; |
| 998 | |
| 999 | } |
| 1000 | |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 1001 | static int ocfs2_wait_for_mask_interruptible(struct ocfs2_mask_waiter *mw, |
| 1002 | struct ocfs2_lock_res *lockres) |
| 1003 | { |
| 1004 | int ret; |
| 1005 | |
| 1006 | ret = wait_for_completion_interruptible(&mw->mw_complete); |
| 1007 | if (ret) |
| 1008 | lockres_remove_mask_waiter(lockres, mw); |
| 1009 | else |
| 1010 | ret = mw->mw_status; |
| 1011 | /* Re-arm the completion in case we want to wait on it again */ |
| 1012 | INIT_COMPLETION(mw->mw_complete); |
| 1013 | return ret; |
| 1014 | } |
| 1015 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1016 | static int ocfs2_cluster_lock(struct ocfs2_super *osb, |
| 1017 | struct ocfs2_lock_res *lockres, |
| 1018 | int level, |
| 1019 | int lkm_flags, |
| 1020 | int arg_flags) |
| 1021 | { |
| 1022 | struct ocfs2_mask_waiter mw; |
| 1023 | enum dlm_status status; |
| 1024 | int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR); |
| 1025 | int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */ |
| 1026 | unsigned long flags; |
| 1027 | |
| 1028 | mlog_entry_void(); |
| 1029 | |
| 1030 | ocfs2_init_mask_waiter(&mw); |
| 1031 | |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 1032 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) |
| 1033 | lkm_flags |= LKM_VALBLK; |
| 1034 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1035 | again: |
| 1036 | wait = 0; |
| 1037 | |
| 1038 | if (catch_signals && signal_pending(current)) { |
| 1039 | ret = -ERESTARTSYS; |
| 1040 | goto out; |
| 1041 | } |
| 1042 | |
| 1043 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1044 | |
| 1045 | mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING, |
| 1046 | "Cluster lock called on freeing lockres %s! flags " |
| 1047 | "0x%lx\n", lockres->l_name, lockres->l_flags); |
| 1048 | |
| 1049 | /* We only compare against the currently granted level |
| 1050 | * here. If the lock is blocked waiting on a downconvert, |
| 1051 | * we'll get caught below. */ |
| 1052 | if (lockres->l_flags & OCFS2_LOCK_BUSY && |
| 1053 | level > lockres->l_level) { |
| 1054 | /* is someone sitting in dlm_lock? If so, wait on |
| 1055 | * them. */ |
| 1056 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); |
| 1057 | wait = 1; |
| 1058 | goto unlock; |
| 1059 | } |
| 1060 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1061 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED && |
| 1062 | !ocfs2_may_continue_on_blocked_lock(lockres, level)) { |
| 1063 | /* is the lock is currently blocked on behalf of |
| 1064 | * another node */ |
| 1065 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0); |
| 1066 | wait = 1; |
| 1067 | goto unlock; |
| 1068 | } |
| 1069 | |
| 1070 | if (level > lockres->l_level) { |
| 1071 | if (lockres->l_action != OCFS2_AST_INVALID) |
| 1072 | mlog(ML_ERROR, "lockres %s has action %u pending\n", |
| 1073 | lockres->l_name, lockres->l_action); |
| 1074 | |
Mark Fasheh | 019d1b2 | 2007-10-05 12:09:05 -0700 | [diff] [blame] | 1075 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { |
| 1076 | lockres->l_action = OCFS2_AST_ATTACH; |
| 1077 | lkm_flags &= ~LKM_CONVERT; |
| 1078 | } else { |
| 1079 | lockres->l_action = OCFS2_AST_CONVERT; |
| 1080 | lkm_flags |= LKM_CONVERT; |
| 1081 | } |
| 1082 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1083 | lockres->l_requested = level; |
| 1084 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); |
| 1085 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1086 | |
| 1087 | BUG_ON(level == LKM_IVMODE); |
| 1088 | BUG_ON(level == LKM_NLMODE); |
| 1089 | |
| 1090 | mlog(0, "lock %s, convert from %d to level = %d\n", |
| 1091 | lockres->l_name, lockres->l_level, level); |
| 1092 | |
| 1093 | /* call dlm_lock to upgrade lock now */ |
| 1094 | status = dlmlock(osb->dlm, |
| 1095 | level, |
| 1096 | &lockres->l_lksb, |
Mark Fasheh | 019d1b2 | 2007-10-05 12:09:05 -0700 | [diff] [blame] | 1097 | lkm_flags, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1098 | lockres->l_name, |
Mark Fasheh | f068106 | 2006-09-08 11:40:10 -0700 | [diff] [blame] | 1099 | OCFS2_LOCK_ID_MAX_LEN - 1, |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 1100 | ocfs2_locking_ast, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1101 | lockres, |
Mark Fasheh | aa2623a | 2006-09-12 21:58:23 -0700 | [diff] [blame] | 1102 | ocfs2_blocking_ast); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1103 | if (status != DLM_NORMAL) { |
| 1104 | if ((lkm_flags & LKM_NOQUEUE) && |
| 1105 | (status == DLM_NOTQUEUED)) |
| 1106 | ret = -EAGAIN; |
| 1107 | else { |
| 1108 | ocfs2_log_dlm_error("dlmlock", status, |
| 1109 | lockres); |
| 1110 | ret = -EINVAL; |
| 1111 | } |
| 1112 | ocfs2_recover_from_dlm_error(lockres, 1); |
| 1113 | goto out; |
| 1114 | } |
| 1115 | |
| 1116 | mlog(0, "lock %s, successfull return from dlmlock\n", |
| 1117 | lockres->l_name); |
| 1118 | |
| 1119 | /* At this point we've gone inside the dlm and need to |
| 1120 | * complete our work regardless. */ |
| 1121 | catch_signals = 0; |
| 1122 | |
| 1123 | /* wait for busy to clear and carry on */ |
| 1124 | goto again; |
| 1125 | } |
| 1126 | |
| 1127 | /* Ok, if we get here then we're good to go. */ |
| 1128 | ocfs2_inc_holders(lockres, level); |
| 1129 | |
| 1130 | ret = 0; |
| 1131 | unlock: |
| 1132 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1133 | out: |
| 1134 | /* |
| 1135 | * This is helping work around a lock inversion between the page lock |
| 1136 | * and dlm locks. One path holds the page lock while calling aops |
| 1137 | * which block acquiring dlm locks. The voting thread holds dlm |
| 1138 | * locks while acquiring page locks while down converting data locks. |
| 1139 | * This block is helping an aop path notice the inversion and back |
| 1140 | * off to unlock its page lock before trying the dlm lock again. |
| 1141 | */ |
| 1142 | if (wait && arg_flags & OCFS2_LOCK_NONBLOCK && |
| 1143 | mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) { |
| 1144 | wait = 0; |
| 1145 | if (lockres_remove_mask_waiter(lockres, &mw)) |
| 1146 | ret = -EAGAIN; |
| 1147 | else |
| 1148 | goto again; |
| 1149 | } |
| 1150 | if (wait) { |
| 1151 | ret = ocfs2_wait_for_mask(&mw); |
| 1152 | if (ret == 0) |
| 1153 | goto again; |
| 1154 | mlog_errno(ret); |
| 1155 | } |
| 1156 | |
| 1157 | mlog_exit(ret); |
| 1158 | return ret; |
| 1159 | } |
| 1160 | |
| 1161 | static void ocfs2_cluster_unlock(struct ocfs2_super *osb, |
| 1162 | struct ocfs2_lock_res *lockres, |
| 1163 | int level) |
| 1164 | { |
| 1165 | unsigned long flags; |
| 1166 | |
| 1167 | mlog_entry_void(); |
| 1168 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1169 | ocfs2_dec_holders(lockres, level); |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 1170 | ocfs2_downconvert_on_unlock(osb, lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1171 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1172 | mlog_exit_void(); |
| 1173 | } |
| 1174 | |
Adrian Bunk | da66116 | 2006-11-20 03:24:28 +0100 | [diff] [blame] | 1175 | static int ocfs2_create_new_lock(struct ocfs2_super *osb, |
| 1176 | struct ocfs2_lock_res *lockres, |
| 1177 | int ex, |
| 1178 | int local) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1179 | { |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 1180 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1181 | unsigned long flags; |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1182 | int lkm_flags = local ? LKM_LOCAL : 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1183 | |
| 1184 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1185 | BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); |
| 1186 | lockres_or_flags(lockres, OCFS2_LOCK_LOCAL); |
| 1187 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1188 | |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1189 | return ocfs2_lock_create(osb, lockres, level, lkm_flags); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1190 | } |
| 1191 | |
| 1192 | /* Grants us an EX lock on the data and metadata resources, skipping |
| 1193 | * the normal cluster directory lookup. Use this ONLY on newly created |
| 1194 | * inodes which other nodes can't possibly see, and which haven't been |
| 1195 | * hashed in the inode hash yet. This can give us a good performance |
| 1196 | * increase as it'll skip the network broadcast normally associated |
| 1197 | * with creating a new lock resource. */ |
| 1198 | int ocfs2_create_new_inode_locks(struct inode *inode) |
| 1199 | { |
| 1200 | int ret; |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 1201 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1202 | |
| 1203 | BUG_ON(!inode); |
| 1204 | BUG_ON(!ocfs2_inode_is_new(inode)); |
| 1205 | |
| 1206 | mlog_entry_void(); |
| 1207 | |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1208 | mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1209 | |
| 1210 | /* NOTE: That we don't increment any of the holder counts, nor |
| 1211 | * do we add anything to a journal handle. Since this is |
| 1212 | * supposed to be a new inode which the cluster doesn't know |
| 1213 | * about yet, there is no need to. As far as the LVB handling |
| 1214 | * is concerned, this is basically like acquiring an EX lock |
| 1215 | * on a resource which has an invalid one -- we'll set it |
| 1216 | * valid when we release the EX. */ |
| 1217 | |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1218 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_rw_lockres, 1, 1); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1219 | if (ret) { |
| 1220 | mlog_errno(ret); |
| 1221 | goto bail; |
| 1222 | } |
| 1223 | |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1224 | /* |
| 1225 | * We don't want to use LKM_LOCAL on a meta data lock as they |
| 1226 | * don't use a generation in their lock names. |
| 1227 | */ |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1228 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_inode_lockres, 1, 0); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1229 | if (ret) { |
| 1230 | mlog_errno(ret); |
| 1231 | goto bail; |
| 1232 | } |
| 1233 | |
Tiger Yang | 5000863 | 2007-03-20 16:01:38 -0700 | [diff] [blame] | 1234 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_open_lockres, 0, 0); |
| 1235 | if (ret) { |
| 1236 | mlog_errno(ret); |
| 1237 | goto bail; |
| 1238 | } |
| 1239 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1240 | bail: |
| 1241 | mlog_exit(ret); |
| 1242 | return ret; |
| 1243 | } |
| 1244 | |
| 1245 | int ocfs2_rw_lock(struct inode *inode, int write) |
| 1246 | { |
| 1247 | int status, level; |
| 1248 | struct ocfs2_lock_res *lockres; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1249 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1250 | |
| 1251 | BUG_ON(!inode); |
| 1252 | |
| 1253 | mlog_entry_void(); |
| 1254 | |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1255 | mlog(0, "inode %llu take %s RW lock\n", |
| 1256 | (unsigned long long)OCFS2_I(inode)->ip_blkno, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1257 | write ? "EXMODE" : "PRMODE"); |
| 1258 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1259 | if (ocfs2_mount_local(osb)) |
| 1260 | return 0; |
| 1261 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1262 | lockres = &OCFS2_I(inode)->ip_rw_lockres; |
| 1263 | |
| 1264 | level = write ? LKM_EXMODE : LKM_PRMODE; |
| 1265 | |
| 1266 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, 0, |
| 1267 | 0); |
| 1268 | if (status < 0) |
| 1269 | mlog_errno(status); |
| 1270 | |
| 1271 | mlog_exit(status); |
| 1272 | return status; |
| 1273 | } |
| 1274 | |
| 1275 | void ocfs2_rw_unlock(struct inode *inode, int write) |
| 1276 | { |
| 1277 | int level = write ? LKM_EXMODE : LKM_PRMODE; |
| 1278 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1279 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1280 | |
| 1281 | mlog_entry_void(); |
| 1282 | |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1283 | mlog(0, "inode %llu drop %s RW lock\n", |
| 1284 | (unsigned long long)OCFS2_I(inode)->ip_blkno, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1285 | write ? "EXMODE" : "PRMODE"); |
| 1286 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1287 | if (!ocfs2_mount_local(osb)) |
| 1288 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1289 | |
| 1290 | mlog_exit_void(); |
| 1291 | } |
| 1292 | |
Tiger Yang | 5000863 | 2007-03-20 16:01:38 -0700 | [diff] [blame] | 1293 | /* |
| 1294 | * ocfs2_open_lock always get PR mode lock. |
| 1295 | */ |
| 1296 | int ocfs2_open_lock(struct inode *inode) |
| 1297 | { |
| 1298 | int status = 0; |
| 1299 | struct ocfs2_lock_res *lockres; |
| 1300 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
| 1301 | |
| 1302 | BUG_ON(!inode); |
| 1303 | |
| 1304 | mlog_entry_void(); |
| 1305 | |
| 1306 | mlog(0, "inode %llu take PRMODE open lock\n", |
| 1307 | (unsigned long long)OCFS2_I(inode)->ip_blkno); |
| 1308 | |
| 1309 | if (ocfs2_mount_local(osb)) |
| 1310 | goto out; |
| 1311 | |
| 1312 | lockres = &OCFS2_I(inode)->ip_open_lockres; |
| 1313 | |
| 1314 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, |
| 1315 | LKM_PRMODE, 0, 0); |
| 1316 | if (status < 0) |
| 1317 | mlog_errno(status); |
| 1318 | |
| 1319 | out: |
| 1320 | mlog_exit(status); |
| 1321 | return status; |
| 1322 | } |
| 1323 | |
| 1324 | int ocfs2_try_open_lock(struct inode *inode, int write) |
| 1325 | { |
| 1326 | int status = 0, level; |
| 1327 | struct ocfs2_lock_res *lockres; |
| 1328 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
| 1329 | |
| 1330 | BUG_ON(!inode); |
| 1331 | |
| 1332 | mlog_entry_void(); |
| 1333 | |
| 1334 | mlog(0, "inode %llu try to take %s open lock\n", |
| 1335 | (unsigned long long)OCFS2_I(inode)->ip_blkno, |
| 1336 | write ? "EXMODE" : "PRMODE"); |
| 1337 | |
| 1338 | if (ocfs2_mount_local(osb)) |
| 1339 | goto out; |
| 1340 | |
| 1341 | lockres = &OCFS2_I(inode)->ip_open_lockres; |
| 1342 | |
| 1343 | level = write ? LKM_EXMODE : LKM_PRMODE; |
| 1344 | |
| 1345 | /* |
| 1346 | * The file system may already holding a PRMODE/EXMODE open lock. |
| 1347 | * Since we pass LKM_NOQUEUE, the request won't block waiting on |
| 1348 | * other nodes and the -EAGAIN will indicate to the caller that |
| 1349 | * this inode is still in use. |
| 1350 | */ |
| 1351 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, |
| 1352 | level, LKM_NOQUEUE, 0); |
| 1353 | |
| 1354 | out: |
| 1355 | mlog_exit(status); |
| 1356 | return status; |
| 1357 | } |
| 1358 | |
| 1359 | /* |
| 1360 | * ocfs2_open_unlock unlock PR and EX mode open locks. |
| 1361 | */ |
| 1362 | void ocfs2_open_unlock(struct inode *inode) |
| 1363 | { |
| 1364 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_open_lockres; |
| 1365 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
| 1366 | |
| 1367 | mlog_entry_void(); |
| 1368 | |
| 1369 | mlog(0, "inode %llu drop open lock\n", |
| 1370 | (unsigned long long)OCFS2_I(inode)->ip_blkno); |
| 1371 | |
| 1372 | if (ocfs2_mount_local(osb)) |
| 1373 | goto out; |
| 1374 | |
| 1375 | if(lockres->l_ro_holders) |
| 1376 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, |
| 1377 | LKM_PRMODE); |
| 1378 | if(lockres->l_ex_holders) |
| 1379 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, |
| 1380 | LKM_EXMODE); |
| 1381 | |
| 1382 | out: |
| 1383 | mlog_exit_void(); |
| 1384 | } |
| 1385 | |
Mark Fasheh | cf8e06f | 2007-12-20 16:43:10 -0800 | [diff] [blame] | 1386 | static int ocfs2_flock_handle_signal(struct ocfs2_lock_res *lockres, |
| 1387 | int level) |
| 1388 | { |
| 1389 | int ret; |
| 1390 | struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); |
| 1391 | unsigned long flags; |
| 1392 | struct ocfs2_mask_waiter mw; |
| 1393 | |
| 1394 | ocfs2_init_mask_waiter(&mw); |
| 1395 | |
| 1396 | retry_cancel: |
| 1397 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1398 | if (lockres->l_flags & OCFS2_LOCK_BUSY) { |
| 1399 | ret = ocfs2_prepare_cancel_convert(osb, lockres); |
| 1400 | if (ret) { |
| 1401 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1402 | ret = ocfs2_cancel_convert(osb, lockres); |
| 1403 | if (ret < 0) { |
| 1404 | mlog_errno(ret); |
| 1405 | goto out; |
| 1406 | } |
| 1407 | goto retry_cancel; |
| 1408 | } |
| 1409 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); |
| 1410 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1411 | |
| 1412 | ocfs2_wait_for_mask(&mw); |
| 1413 | goto retry_cancel; |
| 1414 | } |
| 1415 | |
| 1416 | ret = -ERESTARTSYS; |
| 1417 | /* |
| 1418 | * We may still have gotten the lock, in which case there's no |
| 1419 | * point to restarting the syscall. |
| 1420 | */ |
| 1421 | if (lockres->l_level == level) |
| 1422 | ret = 0; |
| 1423 | |
| 1424 | mlog(0, "Cancel returning %d. flags: 0x%lx, level: %d, act: %d\n", ret, |
| 1425 | lockres->l_flags, lockres->l_level, lockres->l_action); |
| 1426 | |
| 1427 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1428 | |
| 1429 | out: |
| 1430 | return ret; |
| 1431 | } |
| 1432 | |
| 1433 | /* |
| 1434 | * ocfs2_file_lock() and ocfs2_file_unlock() map to a single pair of |
| 1435 | * flock() calls. The locking approach this requires is sufficiently |
| 1436 | * different from all other cluster lock types that we implement a |
| 1437 | * seperate path to the "low-level" dlm calls. In particular: |
| 1438 | * |
| 1439 | * - No optimization of lock levels is done - we take at exactly |
| 1440 | * what's been requested. |
| 1441 | * |
| 1442 | * - No lock caching is employed. We immediately downconvert to |
| 1443 | * no-lock at unlock time. This also means flock locks never go on |
| 1444 | * the blocking list). |
| 1445 | * |
| 1446 | * - Since userspace can trivially deadlock itself with flock, we make |
| 1447 | * sure to allow cancellation of a misbehaving applications flock() |
| 1448 | * request. |
| 1449 | * |
| 1450 | * - Access to any flock lockres doesn't require concurrency, so we |
| 1451 | * can simplify the code by requiring the caller to guarantee |
| 1452 | * serialization of dlmglue flock calls. |
| 1453 | */ |
| 1454 | int ocfs2_file_lock(struct file *file, int ex, int trylock) |
| 1455 | { |
| 1456 | int ret, level = ex ? LKM_EXMODE : LKM_PRMODE; |
| 1457 | unsigned int lkm_flags = trylock ? LKM_NOQUEUE : 0; |
| 1458 | unsigned long flags; |
| 1459 | struct ocfs2_file_private *fp = file->private_data; |
| 1460 | struct ocfs2_lock_res *lockres = &fp->fp_flock; |
| 1461 | struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb); |
| 1462 | struct ocfs2_mask_waiter mw; |
| 1463 | |
| 1464 | ocfs2_init_mask_waiter(&mw); |
| 1465 | |
| 1466 | if ((lockres->l_flags & OCFS2_LOCK_BUSY) || |
| 1467 | (lockres->l_level > LKM_NLMODE)) { |
| 1468 | mlog(ML_ERROR, |
| 1469 | "File lock \"%s\" has busy or locked state: flags: 0x%lx, " |
| 1470 | "level: %u\n", lockres->l_name, lockres->l_flags, |
| 1471 | lockres->l_level); |
| 1472 | return -EINVAL; |
| 1473 | } |
| 1474 | |
| 1475 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1476 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { |
| 1477 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); |
| 1478 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1479 | |
| 1480 | /* |
| 1481 | * Get the lock at NLMODE to start - that way we |
| 1482 | * can cancel the upconvert request if need be. |
| 1483 | */ |
| 1484 | ret = ocfs2_lock_create(osb, lockres, LKM_NLMODE, 0); |
| 1485 | if (ret < 0) { |
| 1486 | mlog_errno(ret); |
| 1487 | goto out; |
| 1488 | } |
| 1489 | |
| 1490 | ret = ocfs2_wait_for_mask(&mw); |
| 1491 | if (ret) { |
| 1492 | mlog_errno(ret); |
| 1493 | goto out; |
| 1494 | } |
| 1495 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1496 | } |
| 1497 | |
| 1498 | lockres->l_action = OCFS2_AST_CONVERT; |
| 1499 | lkm_flags |= LKM_CONVERT; |
| 1500 | lockres->l_requested = level; |
| 1501 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); |
| 1502 | |
| 1503 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); |
| 1504 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1505 | |
| 1506 | ret = dlmlock(osb->dlm, level, &lockres->l_lksb, lkm_flags, |
| 1507 | lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1, |
| 1508 | ocfs2_locking_ast, lockres, ocfs2_blocking_ast); |
| 1509 | if (ret != DLM_NORMAL) { |
| 1510 | if (trylock && ret == DLM_NOTQUEUED) |
| 1511 | ret = -EAGAIN; |
| 1512 | else { |
| 1513 | ocfs2_log_dlm_error("dlmlock", ret, lockres); |
| 1514 | ret = -EINVAL; |
| 1515 | } |
| 1516 | |
| 1517 | ocfs2_recover_from_dlm_error(lockres, 1); |
| 1518 | lockres_remove_mask_waiter(lockres, &mw); |
| 1519 | goto out; |
| 1520 | } |
| 1521 | |
| 1522 | ret = ocfs2_wait_for_mask_interruptible(&mw, lockres); |
| 1523 | if (ret == -ERESTARTSYS) { |
| 1524 | /* |
| 1525 | * Userspace can cause deadlock itself with |
| 1526 | * flock(). Current behavior locally is to allow the |
| 1527 | * deadlock, but abort the system call if a signal is |
| 1528 | * received. We follow this example, otherwise a |
| 1529 | * poorly written program could sit in kernel until |
| 1530 | * reboot. |
| 1531 | * |
| 1532 | * Handling this is a bit more complicated for Ocfs2 |
| 1533 | * though. We can't exit this function with an |
| 1534 | * outstanding lock request, so a cancel convert is |
| 1535 | * required. We intentionally overwrite 'ret' - if the |
| 1536 | * cancel fails and the lock was granted, it's easier |
| 1537 | * to just bubble sucess back up to the user. |
| 1538 | */ |
| 1539 | ret = ocfs2_flock_handle_signal(lockres, level); |
| 1540 | } |
| 1541 | |
| 1542 | out: |
| 1543 | |
| 1544 | mlog(0, "Lock: \"%s\" ex: %d, trylock: %d, returns: %d\n", |
| 1545 | lockres->l_name, ex, trylock, ret); |
| 1546 | return ret; |
| 1547 | } |
| 1548 | |
| 1549 | void ocfs2_file_unlock(struct file *file) |
| 1550 | { |
| 1551 | int ret; |
| 1552 | unsigned long flags; |
| 1553 | struct ocfs2_file_private *fp = file->private_data; |
| 1554 | struct ocfs2_lock_res *lockres = &fp->fp_flock; |
| 1555 | struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb); |
| 1556 | struct ocfs2_mask_waiter mw; |
| 1557 | |
| 1558 | ocfs2_init_mask_waiter(&mw); |
| 1559 | |
| 1560 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) |
| 1561 | return; |
| 1562 | |
| 1563 | if (lockres->l_level == LKM_NLMODE) |
| 1564 | return; |
| 1565 | |
| 1566 | mlog(0, "Unlock: \"%s\" flags: 0x%lx, level: %d, act: %d\n", |
| 1567 | lockres->l_name, lockres->l_flags, lockres->l_level, |
| 1568 | lockres->l_action); |
| 1569 | |
| 1570 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1571 | /* |
| 1572 | * Fake a blocking ast for the downconvert code. |
| 1573 | */ |
| 1574 | lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); |
| 1575 | lockres->l_blocking = LKM_EXMODE; |
| 1576 | |
| 1577 | ocfs2_prepare_downconvert(lockres, LKM_NLMODE); |
| 1578 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); |
| 1579 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1580 | |
| 1581 | ret = ocfs2_downconvert_lock(osb, lockres, LKM_NLMODE, 0); |
| 1582 | if (ret) { |
| 1583 | mlog_errno(ret); |
| 1584 | return; |
| 1585 | } |
| 1586 | |
| 1587 | ret = ocfs2_wait_for_mask(&mw); |
| 1588 | if (ret) |
| 1589 | mlog_errno(ret); |
| 1590 | } |
| 1591 | |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 1592 | static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, |
| 1593 | struct ocfs2_lock_res *lockres) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1594 | { |
| 1595 | int kick = 0; |
| 1596 | |
| 1597 | mlog_entry_void(); |
| 1598 | |
| 1599 | /* If we know that another node is waiting on our lock, kick |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 1600 | * the downconvert thread * pre-emptively when we reach a release |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1601 | * condition. */ |
| 1602 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) { |
| 1603 | switch(lockres->l_blocking) { |
| 1604 | case LKM_EXMODE: |
| 1605 | if (!lockres->l_ex_holders && !lockres->l_ro_holders) |
| 1606 | kick = 1; |
| 1607 | break; |
| 1608 | case LKM_PRMODE: |
| 1609 | if (!lockres->l_ex_holders) |
| 1610 | kick = 1; |
| 1611 | break; |
| 1612 | default: |
| 1613 | BUG(); |
| 1614 | } |
| 1615 | } |
| 1616 | |
| 1617 | if (kick) |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 1618 | ocfs2_wake_downconvert_thread(osb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1619 | |
| 1620 | mlog_exit_void(); |
| 1621 | } |
| 1622 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1623 | #define OCFS2_SEC_BITS 34 |
| 1624 | #define OCFS2_SEC_SHIFT (64 - 34) |
| 1625 | #define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1) |
| 1626 | |
| 1627 | /* LVB only has room for 64 bits of time here so we pack it for |
| 1628 | * now. */ |
| 1629 | static u64 ocfs2_pack_timespec(struct timespec *spec) |
| 1630 | { |
| 1631 | u64 res; |
| 1632 | u64 sec = spec->tv_sec; |
| 1633 | u32 nsec = spec->tv_nsec; |
| 1634 | |
| 1635 | res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); |
| 1636 | |
| 1637 | return res; |
| 1638 | } |
| 1639 | |
| 1640 | /* Call this with the lockres locked. I am reasonably sure we don't |
| 1641 | * need ip_lock in this function as anyone who would be changing those |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1642 | * values is supposed to be blocked in ocfs2_inode_lock right now. */ |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1643 | static void __ocfs2_stuff_meta_lvb(struct inode *inode) |
| 1644 | { |
| 1645 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1646 | struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1647 | struct ocfs2_meta_lvb *lvb; |
| 1648 | |
| 1649 | mlog_entry_void(); |
| 1650 | |
| 1651 | lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; |
| 1652 | |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1653 | /* |
| 1654 | * Invalidate the LVB of a deleted inode - this way other |
| 1655 | * nodes are forced to go to disk and discover the new inode |
| 1656 | * status. |
| 1657 | */ |
| 1658 | if (oi->ip_flags & OCFS2_INODE_DELETED) { |
| 1659 | lvb->lvb_version = 0; |
| 1660 | goto out; |
| 1661 | } |
| 1662 | |
Mark Fasheh | 4d3b83f | 2006-09-12 15:22:18 -0700 | [diff] [blame] | 1663 | lvb->lvb_version = OCFS2_LVB_VERSION; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1664 | lvb->lvb_isize = cpu_to_be64(i_size_read(inode)); |
| 1665 | lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters); |
| 1666 | lvb->lvb_iuid = cpu_to_be32(inode->i_uid); |
| 1667 | lvb->lvb_igid = cpu_to_be32(inode->i_gid); |
| 1668 | lvb->lvb_imode = cpu_to_be16(inode->i_mode); |
| 1669 | lvb->lvb_inlink = cpu_to_be16(inode->i_nlink); |
| 1670 | lvb->lvb_iatime_packed = |
| 1671 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime)); |
| 1672 | lvb->lvb_ictime_packed = |
| 1673 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime)); |
| 1674 | lvb->lvb_imtime_packed = |
| 1675 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime)); |
Herbert Poetzl | ca4d147 | 2006-07-03 17:27:12 -0700 | [diff] [blame] | 1676 | lvb->lvb_iattr = cpu_to_be32(oi->ip_attr); |
Mark Fasheh | 15b1e36 | 2007-09-07 13:58:15 -0700 | [diff] [blame] | 1677 | lvb->lvb_idynfeatures = cpu_to_be16(oi->ip_dyn_features); |
Mark Fasheh | f9e2d82 | 2006-09-12 15:35:49 -0700 | [diff] [blame] | 1678 | lvb->lvb_igeneration = cpu_to_be32(inode->i_generation); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1679 | |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1680 | out: |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1681 | mlog_meta_lvb(0, lockres); |
| 1682 | |
| 1683 | mlog_exit_void(); |
| 1684 | } |
| 1685 | |
| 1686 | static void ocfs2_unpack_timespec(struct timespec *spec, |
| 1687 | u64 packed_time) |
| 1688 | { |
| 1689 | spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT; |
| 1690 | spec->tv_nsec = packed_time & OCFS2_NSEC_MASK; |
| 1691 | } |
| 1692 | |
| 1693 | static void ocfs2_refresh_inode_from_lvb(struct inode *inode) |
| 1694 | { |
| 1695 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1696 | struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1697 | struct ocfs2_meta_lvb *lvb; |
| 1698 | |
| 1699 | mlog_entry_void(); |
| 1700 | |
| 1701 | mlog_meta_lvb(0, lockres); |
| 1702 | |
| 1703 | lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; |
| 1704 | |
| 1705 | /* We're safe here without the lockres lock... */ |
| 1706 | spin_lock(&oi->ip_lock); |
| 1707 | oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters); |
| 1708 | i_size_write(inode, be64_to_cpu(lvb->lvb_isize)); |
| 1709 | |
Herbert Poetzl | ca4d147 | 2006-07-03 17:27:12 -0700 | [diff] [blame] | 1710 | oi->ip_attr = be32_to_cpu(lvb->lvb_iattr); |
Mark Fasheh | 15b1e36 | 2007-09-07 13:58:15 -0700 | [diff] [blame] | 1711 | oi->ip_dyn_features = be16_to_cpu(lvb->lvb_idynfeatures); |
Herbert Poetzl | ca4d147 | 2006-07-03 17:27:12 -0700 | [diff] [blame] | 1712 | ocfs2_set_inode_flags(inode); |
| 1713 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1714 | /* fast-symlinks are a special case */ |
| 1715 | if (S_ISLNK(inode->i_mode) && !oi->ip_clusters) |
| 1716 | inode->i_blocks = 0; |
| 1717 | else |
Mark Fasheh | 8110b07 | 2007-03-22 16:53:23 -0700 | [diff] [blame] | 1718 | inode->i_blocks = ocfs2_inode_sector_count(inode); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1719 | |
| 1720 | inode->i_uid = be32_to_cpu(lvb->lvb_iuid); |
| 1721 | inode->i_gid = be32_to_cpu(lvb->lvb_igid); |
| 1722 | inode->i_mode = be16_to_cpu(lvb->lvb_imode); |
| 1723 | inode->i_nlink = be16_to_cpu(lvb->lvb_inlink); |
| 1724 | ocfs2_unpack_timespec(&inode->i_atime, |
| 1725 | be64_to_cpu(lvb->lvb_iatime_packed)); |
| 1726 | ocfs2_unpack_timespec(&inode->i_mtime, |
| 1727 | be64_to_cpu(lvb->lvb_imtime_packed)); |
| 1728 | ocfs2_unpack_timespec(&inode->i_ctime, |
| 1729 | be64_to_cpu(lvb->lvb_ictime_packed)); |
| 1730 | spin_unlock(&oi->ip_lock); |
| 1731 | |
| 1732 | mlog_exit_void(); |
| 1733 | } |
| 1734 | |
Mark Fasheh | f9e2d82 | 2006-09-12 15:35:49 -0700 | [diff] [blame] | 1735 | static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode, |
| 1736 | struct ocfs2_lock_res *lockres) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1737 | { |
| 1738 | struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; |
| 1739 | |
Mark Fasheh | f9e2d82 | 2006-09-12 15:35:49 -0700 | [diff] [blame] | 1740 | if (lvb->lvb_version == OCFS2_LVB_VERSION |
| 1741 | && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1742 | return 1; |
| 1743 | return 0; |
| 1744 | } |
| 1745 | |
| 1746 | /* Determine whether a lock resource needs to be refreshed, and |
| 1747 | * arbitrate who gets to refresh it. |
| 1748 | * |
| 1749 | * 0 means no refresh needed. |
| 1750 | * |
| 1751 | * > 0 means you need to refresh this and you MUST call |
| 1752 | * ocfs2_complete_lock_res_refresh afterwards. */ |
| 1753 | static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres) |
| 1754 | { |
| 1755 | unsigned long flags; |
| 1756 | int status = 0; |
| 1757 | |
| 1758 | mlog_entry_void(); |
| 1759 | |
| 1760 | refresh_check: |
| 1761 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1762 | if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) { |
| 1763 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1764 | goto bail; |
| 1765 | } |
| 1766 | |
| 1767 | if (lockres->l_flags & OCFS2_LOCK_REFRESHING) { |
| 1768 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1769 | |
| 1770 | ocfs2_wait_on_refreshing_lock(lockres); |
| 1771 | goto refresh_check; |
| 1772 | } |
| 1773 | |
| 1774 | /* Ok, I'll be the one to refresh this lock. */ |
| 1775 | lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING); |
| 1776 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1777 | |
| 1778 | status = 1; |
| 1779 | bail: |
| 1780 | mlog_exit(status); |
| 1781 | return status; |
| 1782 | } |
| 1783 | |
| 1784 | /* If status is non zero, I'll mark it as not being in refresh |
| 1785 | * anymroe, but i won't clear the needs refresh flag. */ |
| 1786 | static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres, |
| 1787 | int status) |
| 1788 | { |
| 1789 | unsigned long flags; |
| 1790 | mlog_entry_void(); |
| 1791 | |
| 1792 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 1793 | lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING); |
| 1794 | if (!status) |
| 1795 | lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
| 1796 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 1797 | |
| 1798 | wake_up(&lockres->l_event); |
| 1799 | |
| 1800 | mlog_exit_void(); |
| 1801 | } |
| 1802 | |
| 1803 | /* may or may not return a bh if it went to disk. */ |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1804 | static int ocfs2_inode_lock_update(struct inode *inode, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1805 | struct buffer_head **bh) |
| 1806 | { |
| 1807 | int status = 0; |
| 1808 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1809 | struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1810 | struct ocfs2_dinode *fe; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1811 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1812 | |
| 1813 | mlog_entry_void(); |
| 1814 | |
Mark Fasheh | be9e986 | 2007-04-18 15:22:08 -0700 | [diff] [blame] | 1815 | if (ocfs2_mount_local(osb)) |
| 1816 | goto bail; |
| 1817 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1818 | spin_lock(&oi->ip_lock); |
| 1819 | if (oi->ip_flags & OCFS2_INODE_DELETED) { |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1820 | mlog(0, "Orphaned inode %llu was deleted while we " |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1821 | "were waiting on a lock. ip_flags = 0x%x\n", |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1822 | (unsigned long long)oi->ip_blkno, oi->ip_flags); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1823 | spin_unlock(&oi->ip_lock); |
| 1824 | status = -ENOENT; |
| 1825 | goto bail; |
| 1826 | } |
| 1827 | spin_unlock(&oi->ip_lock); |
| 1828 | |
Mark Fasheh | be9e986 | 2007-04-18 15:22:08 -0700 | [diff] [blame] | 1829 | if (!ocfs2_should_refresh_lock_res(lockres)) |
| 1830 | goto bail; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1831 | |
| 1832 | /* This will discard any caching information we might have had |
| 1833 | * for the inode metadata. */ |
| 1834 | ocfs2_metadata_cache_purge(inode); |
| 1835 | |
Mark Fasheh | 8341897 | 2007-04-23 18:53:12 -0700 | [diff] [blame] | 1836 | ocfs2_extent_map_trunc(inode, 0); |
| 1837 | |
Mark Fasheh | be9e986 | 2007-04-18 15:22:08 -0700 | [diff] [blame] | 1838 | if (ocfs2_meta_lvb_is_trustable(inode, lockres)) { |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1839 | mlog(0, "Trusting LVB on inode %llu\n", |
| 1840 | (unsigned long long)oi->ip_blkno); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1841 | ocfs2_refresh_inode_from_lvb(inode); |
| 1842 | } else { |
| 1843 | /* Boo, we have to go to disk. */ |
| 1844 | /* read bh, cast, ocfs2_refresh_inode */ |
| 1845 | status = ocfs2_read_block(OCFS2_SB(inode->i_sb), oi->ip_blkno, |
| 1846 | bh, OCFS2_BH_CACHED, inode); |
| 1847 | if (status < 0) { |
| 1848 | mlog_errno(status); |
| 1849 | goto bail_refresh; |
| 1850 | } |
| 1851 | fe = (struct ocfs2_dinode *) (*bh)->b_data; |
| 1852 | |
| 1853 | /* This is a good chance to make sure we're not |
| 1854 | * locking an invalid object. |
| 1855 | * |
| 1856 | * We bug on a stale inode here because we checked |
| 1857 | * above whether it was wiped from disk. The wiping |
| 1858 | * node provides a guarantee that we receive that |
| 1859 | * message and can mark the inode before dropping any |
| 1860 | * locks associated with it. */ |
| 1861 | if (!OCFS2_IS_VALID_DINODE(fe)) { |
| 1862 | OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe); |
| 1863 | status = -EIO; |
| 1864 | goto bail_refresh; |
| 1865 | } |
| 1866 | mlog_bug_on_msg(inode->i_generation != |
| 1867 | le32_to_cpu(fe->i_generation), |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1868 | "Invalid dinode %llu disk generation: %u " |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1869 | "inode->i_generation: %u\n", |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1870 | (unsigned long long)oi->ip_blkno, |
| 1871 | le32_to_cpu(fe->i_generation), |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1872 | inode->i_generation); |
| 1873 | mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) || |
| 1874 | !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)), |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1875 | "Stale dinode %llu dtime: %llu flags: 0x%x\n", |
| 1876 | (unsigned long long)oi->ip_blkno, |
| 1877 | (unsigned long long)le64_to_cpu(fe->i_dtime), |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1878 | le32_to_cpu(fe->i_flags)); |
| 1879 | |
| 1880 | ocfs2_refresh_inode(inode, fe); |
| 1881 | } |
| 1882 | |
| 1883 | status = 0; |
| 1884 | bail_refresh: |
Mark Fasheh | be9e986 | 2007-04-18 15:22:08 -0700 | [diff] [blame] | 1885 | ocfs2_complete_lock_res_refresh(lockres, status); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1886 | bail: |
| 1887 | mlog_exit(status); |
| 1888 | return status; |
| 1889 | } |
| 1890 | |
| 1891 | static int ocfs2_assign_bh(struct inode *inode, |
| 1892 | struct buffer_head **ret_bh, |
| 1893 | struct buffer_head *passed_bh) |
| 1894 | { |
| 1895 | int status; |
| 1896 | |
| 1897 | if (passed_bh) { |
| 1898 | /* Ok, the update went to disk for us, use the |
| 1899 | * returned bh. */ |
| 1900 | *ret_bh = passed_bh; |
| 1901 | get_bh(*ret_bh); |
| 1902 | |
| 1903 | return 0; |
| 1904 | } |
| 1905 | |
| 1906 | status = ocfs2_read_block(OCFS2_SB(inode->i_sb), |
| 1907 | OCFS2_I(inode)->ip_blkno, |
| 1908 | ret_bh, |
| 1909 | OCFS2_BH_CACHED, |
| 1910 | inode); |
| 1911 | if (status < 0) |
| 1912 | mlog_errno(status); |
| 1913 | |
| 1914 | return status; |
| 1915 | } |
| 1916 | |
| 1917 | /* |
| 1918 | * returns < 0 error if the callback will never be called, otherwise |
| 1919 | * the result of the lock will be communicated via the callback. |
| 1920 | */ |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1921 | int ocfs2_inode_lock_full(struct inode *inode, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1922 | struct buffer_head **ret_bh, |
| 1923 | int ex, |
| 1924 | int arg_flags) |
| 1925 | { |
| 1926 | int status, level, dlm_flags, acquired; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1927 | struct ocfs2_lock_res *lockres = NULL; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1928 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
| 1929 | struct buffer_head *local_bh = NULL; |
| 1930 | |
| 1931 | BUG_ON(!inode); |
| 1932 | |
| 1933 | mlog_entry_void(); |
| 1934 | |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 1935 | mlog(0, "inode %llu, take %s META lock\n", |
| 1936 | (unsigned long long)OCFS2_I(inode)->ip_blkno, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1937 | ex ? "EXMODE" : "PRMODE"); |
| 1938 | |
| 1939 | status = 0; |
| 1940 | acquired = 0; |
| 1941 | /* We'll allow faking a readonly metadata lock for |
| 1942 | * rodevices. */ |
| 1943 | if (ocfs2_is_hard_readonly(osb)) { |
| 1944 | if (ex) |
| 1945 | status = -EROFS; |
| 1946 | goto bail; |
| 1947 | } |
| 1948 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1949 | if (ocfs2_mount_local(osb)) |
| 1950 | goto local; |
| 1951 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1952 | if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) |
| 1953 | wait_event(osb->recovery_event, |
| 1954 | ocfs2_node_map_is_empty(osb, &osb->recovery_map)); |
| 1955 | |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1956 | lockres = &OCFS2_I(inode)->ip_inode_lockres; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1957 | level = ex ? LKM_EXMODE : LKM_PRMODE; |
| 1958 | dlm_flags = 0; |
| 1959 | if (arg_flags & OCFS2_META_LOCK_NOQUEUE) |
| 1960 | dlm_flags |= LKM_NOQUEUE; |
| 1961 | |
| 1962 | status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags); |
| 1963 | if (status < 0) { |
| 1964 | if (status != -EAGAIN && status != -EIOCBRETRY) |
| 1965 | mlog_errno(status); |
| 1966 | goto bail; |
| 1967 | } |
| 1968 | |
| 1969 | /* Notify the error cleanup path to drop the cluster lock. */ |
| 1970 | acquired = 1; |
| 1971 | |
| 1972 | /* We wait twice because a node may have died while we were in |
| 1973 | * the lower dlm layers. The second time though, we've |
| 1974 | * committed to owning this lock so we don't allow signals to |
| 1975 | * abort the operation. */ |
| 1976 | if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) |
| 1977 | wait_event(osb->recovery_event, |
| 1978 | ocfs2_node_map_is_empty(osb, &osb->recovery_map)); |
| 1979 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1980 | local: |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1981 | /* |
| 1982 | * We only see this flag if we're being called from |
| 1983 | * ocfs2_read_locked_inode(). It means we're locking an inode |
| 1984 | * which hasn't been populated yet, so clear the refresh flag |
| 1985 | * and let the caller handle it. |
| 1986 | */ |
| 1987 | if (inode->i_state & I_NEW) { |
| 1988 | status = 0; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 1989 | if (lockres) |
| 1990 | ocfs2_complete_lock_res_refresh(lockres, 0); |
Mark Fasheh | 24c19ef | 2006-09-22 17:28:19 -0700 | [diff] [blame] | 1991 | goto bail; |
| 1992 | } |
| 1993 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1994 | /* This is fun. The caller may want a bh back, or it may |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1995 | * not. ocfs2_inode_lock_update definitely wants one in, but |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1996 | * may or may not read one, depending on what's in the |
| 1997 | * LVB. The result of all of this is that we've *only* gone to |
| 1998 | * disk if we have to, so the complexity is worthwhile. */ |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 1999 | status = ocfs2_inode_lock_update(inode, &local_bh); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2000 | if (status < 0) { |
| 2001 | if (status != -ENOENT) |
| 2002 | mlog_errno(status); |
| 2003 | goto bail; |
| 2004 | } |
| 2005 | |
| 2006 | if (ret_bh) { |
| 2007 | status = ocfs2_assign_bh(inode, ret_bh, local_bh); |
| 2008 | if (status < 0) { |
| 2009 | mlog_errno(status); |
| 2010 | goto bail; |
| 2011 | } |
| 2012 | } |
| 2013 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2014 | bail: |
| 2015 | if (status < 0) { |
| 2016 | if (ret_bh && (*ret_bh)) { |
| 2017 | brelse(*ret_bh); |
| 2018 | *ret_bh = NULL; |
| 2019 | } |
| 2020 | if (acquired) |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2021 | ocfs2_inode_unlock(inode, ex); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2022 | } |
| 2023 | |
| 2024 | if (local_bh) |
| 2025 | brelse(local_bh); |
| 2026 | |
| 2027 | mlog_exit(status); |
| 2028 | return status; |
| 2029 | } |
| 2030 | |
| 2031 | /* |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 2032 | * This is working around a lock inversion between tasks acquiring DLM |
| 2033 | * locks while holding a page lock and the downconvert thread which |
| 2034 | * blocks dlm lock acquiry while acquiring page locks. |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2035 | * |
| 2036 | * ** These _with_page variantes are only intended to be called from aop |
| 2037 | * methods that hold page locks and return a very specific *positive* error |
| 2038 | * code that aop methods pass up to the VFS -- test for errors with != 0. ** |
| 2039 | * |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 2040 | * The DLM is called such that it returns -EAGAIN if it would have |
| 2041 | * blocked waiting for the downconvert thread. In that case we unlock |
| 2042 | * our page so the downconvert thread can make progress. Once we've |
| 2043 | * done this we have to return AOP_TRUNCATED_PAGE so the aop method |
| 2044 | * that called us can bubble that back up into the VFS who will then |
| 2045 | * immediately retry the aop call. |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2046 | * |
| 2047 | * We do a blocking lock and immediate unlock before returning, though, so that |
| 2048 | * the lock has a great chance of being cached on this node by the time the VFS |
| 2049 | * calls back to retry the aop. This has a potential to livelock as nodes |
| 2050 | * ping locks back and forth, but that's a risk we're willing to take to avoid |
| 2051 | * the lock inversion simply. |
| 2052 | */ |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2053 | int ocfs2_inode_lock_with_page(struct inode *inode, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2054 | struct buffer_head **ret_bh, |
| 2055 | int ex, |
| 2056 | struct page *page) |
| 2057 | { |
| 2058 | int ret; |
| 2059 | |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2060 | ret = ocfs2_inode_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2061 | if (ret == -EAGAIN) { |
| 2062 | unlock_page(page); |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2063 | if (ocfs2_inode_lock(inode, ret_bh, ex) == 0) |
| 2064 | ocfs2_inode_unlock(inode, ex); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2065 | ret = AOP_TRUNCATED_PAGE; |
| 2066 | } |
| 2067 | |
| 2068 | return ret; |
| 2069 | } |
| 2070 | |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2071 | int ocfs2_inode_lock_atime(struct inode *inode, |
Tiger Yang | 7f1a37e | 2006-11-15 15:48:42 +0800 | [diff] [blame] | 2072 | struct vfsmount *vfsmnt, |
| 2073 | int *level) |
| 2074 | { |
| 2075 | int ret; |
| 2076 | |
| 2077 | mlog_entry_void(); |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2078 | ret = ocfs2_inode_lock(inode, NULL, 0); |
Tiger Yang | 7f1a37e | 2006-11-15 15:48:42 +0800 | [diff] [blame] | 2079 | if (ret < 0) { |
| 2080 | mlog_errno(ret); |
| 2081 | return ret; |
| 2082 | } |
| 2083 | |
| 2084 | /* |
| 2085 | * If we should update atime, we will get EX lock, |
| 2086 | * otherwise we just get PR lock. |
| 2087 | */ |
| 2088 | if (ocfs2_should_update_atime(inode, vfsmnt)) { |
| 2089 | struct buffer_head *bh = NULL; |
| 2090 | |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2091 | ocfs2_inode_unlock(inode, 0); |
| 2092 | ret = ocfs2_inode_lock(inode, &bh, 1); |
Tiger Yang | 7f1a37e | 2006-11-15 15:48:42 +0800 | [diff] [blame] | 2093 | if (ret < 0) { |
| 2094 | mlog_errno(ret); |
| 2095 | return ret; |
| 2096 | } |
| 2097 | *level = 1; |
| 2098 | if (ocfs2_should_update_atime(inode, vfsmnt)) |
| 2099 | ocfs2_update_inode_atime(inode, bh); |
| 2100 | if (bh) |
| 2101 | brelse(bh); |
| 2102 | } else |
| 2103 | *level = 0; |
| 2104 | |
| 2105 | mlog_exit(ret); |
| 2106 | return ret; |
| 2107 | } |
| 2108 | |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2109 | void ocfs2_inode_unlock(struct inode *inode, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2110 | int ex) |
| 2111 | { |
| 2112 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2113 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_inode_lockres; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2114 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2115 | |
| 2116 | mlog_entry_void(); |
| 2117 | |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 2118 | mlog(0, "inode %llu drop %s META lock\n", |
| 2119 | (unsigned long long)OCFS2_I(inode)->ip_blkno, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2120 | ex ? "EXMODE" : "PRMODE"); |
| 2121 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2122 | if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)) && |
| 2123 | !ocfs2_mount_local(osb)) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2124 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); |
| 2125 | |
| 2126 | mlog_exit_void(); |
| 2127 | } |
| 2128 | |
| 2129 | int ocfs2_super_lock(struct ocfs2_super *osb, |
| 2130 | int ex) |
| 2131 | { |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2132 | int status = 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2133 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
| 2134 | struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; |
| 2135 | struct buffer_head *bh; |
| 2136 | struct ocfs2_slot_info *si = osb->slot_info; |
| 2137 | |
| 2138 | mlog_entry_void(); |
| 2139 | |
| 2140 | if (ocfs2_is_hard_readonly(osb)) |
| 2141 | return -EROFS; |
| 2142 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2143 | if (ocfs2_mount_local(osb)) |
| 2144 | goto bail; |
| 2145 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2146 | status = ocfs2_cluster_lock(osb, lockres, level, 0, 0); |
| 2147 | if (status < 0) { |
| 2148 | mlog_errno(status); |
| 2149 | goto bail; |
| 2150 | } |
| 2151 | |
| 2152 | /* The super block lock path is really in the best position to |
| 2153 | * know when resources covered by the lock need to be |
| 2154 | * refreshed, so we do it here. Of course, making sense of |
| 2155 | * everything is up to the caller :) */ |
| 2156 | status = ocfs2_should_refresh_lock_res(lockres); |
| 2157 | if (status < 0) { |
| 2158 | mlog_errno(status); |
| 2159 | goto bail; |
| 2160 | } |
| 2161 | if (status) { |
| 2162 | bh = si->si_bh; |
| 2163 | status = ocfs2_read_block(osb, bh->b_blocknr, &bh, 0, |
| 2164 | si->si_inode); |
| 2165 | if (status == 0) |
| 2166 | ocfs2_update_slot_info(si); |
| 2167 | |
| 2168 | ocfs2_complete_lock_res_refresh(lockres, status); |
| 2169 | |
| 2170 | if (status < 0) |
| 2171 | mlog_errno(status); |
| 2172 | } |
| 2173 | bail: |
| 2174 | mlog_exit(status); |
| 2175 | return status; |
| 2176 | } |
| 2177 | |
| 2178 | void ocfs2_super_unlock(struct ocfs2_super *osb, |
| 2179 | int ex) |
| 2180 | { |
| 2181 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
| 2182 | struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; |
| 2183 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2184 | if (!ocfs2_mount_local(osb)) |
| 2185 | ocfs2_cluster_unlock(osb, lockres, level); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2186 | } |
| 2187 | |
| 2188 | int ocfs2_rename_lock(struct ocfs2_super *osb) |
| 2189 | { |
| 2190 | int status; |
| 2191 | struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; |
| 2192 | |
| 2193 | if (ocfs2_is_hard_readonly(osb)) |
| 2194 | return -EROFS; |
| 2195 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2196 | if (ocfs2_mount_local(osb)) |
| 2197 | return 0; |
| 2198 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2199 | status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0); |
| 2200 | if (status < 0) |
| 2201 | mlog_errno(status); |
| 2202 | |
| 2203 | return status; |
| 2204 | } |
| 2205 | |
| 2206 | void ocfs2_rename_unlock(struct ocfs2_super *osb) |
| 2207 | { |
| 2208 | struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; |
| 2209 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2210 | if (!ocfs2_mount_local(osb)) |
| 2211 | ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2212 | } |
| 2213 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2214 | int ocfs2_dentry_lock(struct dentry *dentry, int ex) |
| 2215 | { |
| 2216 | int ret; |
| 2217 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
| 2218 | struct ocfs2_dentry_lock *dl = dentry->d_fsdata; |
| 2219 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
| 2220 | |
| 2221 | BUG_ON(!dl); |
| 2222 | |
| 2223 | if (ocfs2_is_hard_readonly(osb)) |
| 2224 | return -EROFS; |
| 2225 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2226 | if (ocfs2_mount_local(osb)) |
| 2227 | return 0; |
| 2228 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2229 | ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0); |
| 2230 | if (ret < 0) |
| 2231 | mlog_errno(ret); |
| 2232 | |
| 2233 | return ret; |
| 2234 | } |
| 2235 | |
| 2236 | void ocfs2_dentry_unlock(struct dentry *dentry, int ex) |
| 2237 | { |
| 2238 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
| 2239 | struct ocfs2_dentry_lock *dl = dentry->d_fsdata; |
| 2240 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
| 2241 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2242 | if (!ocfs2_mount_local(osb)) |
| 2243 | ocfs2_cluster_unlock(osb, &dl->dl_lockres, level); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2244 | } |
| 2245 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2246 | /* Reference counting of the dlm debug structure. We want this because |
| 2247 | * open references on the debug inodes can live on after a mount, so |
| 2248 | * we can't rely on the ocfs2_super to always exist. */ |
| 2249 | static void ocfs2_dlm_debug_free(struct kref *kref) |
| 2250 | { |
| 2251 | struct ocfs2_dlm_debug *dlm_debug; |
| 2252 | |
| 2253 | dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt); |
| 2254 | |
| 2255 | kfree(dlm_debug); |
| 2256 | } |
| 2257 | |
| 2258 | void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug) |
| 2259 | { |
| 2260 | if (dlm_debug) |
| 2261 | kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free); |
| 2262 | } |
| 2263 | |
| 2264 | static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug) |
| 2265 | { |
| 2266 | kref_get(&debug->d_refcnt); |
| 2267 | } |
| 2268 | |
| 2269 | struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void) |
| 2270 | { |
| 2271 | struct ocfs2_dlm_debug *dlm_debug; |
| 2272 | |
| 2273 | dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL); |
| 2274 | if (!dlm_debug) { |
| 2275 | mlog_errno(-ENOMEM); |
| 2276 | goto out; |
| 2277 | } |
| 2278 | |
| 2279 | kref_init(&dlm_debug->d_refcnt); |
| 2280 | INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking); |
| 2281 | dlm_debug->d_locking_state = NULL; |
| 2282 | out: |
| 2283 | return dlm_debug; |
| 2284 | } |
| 2285 | |
| 2286 | /* Access to this is arbitrated for us via seq_file->sem. */ |
| 2287 | struct ocfs2_dlm_seq_priv { |
| 2288 | struct ocfs2_dlm_debug *p_dlm_debug; |
| 2289 | struct ocfs2_lock_res p_iter_res; |
| 2290 | struct ocfs2_lock_res p_tmp_res; |
| 2291 | }; |
| 2292 | |
| 2293 | static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start, |
| 2294 | struct ocfs2_dlm_seq_priv *priv) |
| 2295 | { |
| 2296 | struct ocfs2_lock_res *iter, *ret = NULL; |
| 2297 | struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug; |
| 2298 | |
| 2299 | assert_spin_locked(&ocfs2_dlm_tracking_lock); |
| 2300 | |
| 2301 | list_for_each_entry(iter, &start->l_debug_list, l_debug_list) { |
| 2302 | /* discover the head of the list */ |
| 2303 | if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) { |
| 2304 | mlog(0, "End of list found, %p\n", ret); |
| 2305 | break; |
| 2306 | } |
| 2307 | |
| 2308 | /* We track our "dummy" iteration lockres' by a NULL |
| 2309 | * l_ops field. */ |
| 2310 | if (iter->l_ops != NULL) { |
| 2311 | ret = iter; |
| 2312 | break; |
| 2313 | } |
| 2314 | } |
| 2315 | |
| 2316 | return ret; |
| 2317 | } |
| 2318 | |
| 2319 | static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos) |
| 2320 | { |
| 2321 | struct ocfs2_dlm_seq_priv *priv = m->private; |
| 2322 | struct ocfs2_lock_res *iter; |
| 2323 | |
| 2324 | spin_lock(&ocfs2_dlm_tracking_lock); |
| 2325 | iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv); |
| 2326 | if (iter) { |
| 2327 | /* Since lockres' have the lifetime of their container |
| 2328 | * (which can be inodes, ocfs2_supers, etc) we want to |
| 2329 | * copy this out to a temporary lockres while still |
| 2330 | * under the spinlock. Obviously after this we can't |
| 2331 | * trust any pointers on the copy returned, but that's |
| 2332 | * ok as the information we want isn't typically held |
| 2333 | * in them. */ |
| 2334 | priv->p_tmp_res = *iter; |
| 2335 | iter = &priv->p_tmp_res; |
| 2336 | } |
| 2337 | spin_unlock(&ocfs2_dlm_tracking_lock); |
| 2338 | |
| 2339 | return iter; |
| 2340 | } |
| 2341 | |
| 2342 | static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v) |
| 2343 | { |
| 2344 | } |
| 2345 | |
| 2346 | static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos) |
| 2347 | { |
| 2348 | struct ocfs2_dlm_seq_priv *priv = m->private; |
| 2349 | struct ocfs2_lock_res *iter = v; |
| 2350 | struct ocfs2_lock_res *dummy = &priv->p_iter_res; |
| 2351 | |
| 2352 | spin_lock(&ocfs2_dlm_tracking_lock); |
| 2353 | iter = ocfs2_dlm_next_res(iter, priv); |
| 2354 | list_del_init(&dummy->l_debug_list); |
| 2355 | if (iter) { |
| 2356 | list_add(&dummy->l_debug_list, &iter->l_debug_list); |
| 2357 | priv->p_tmp_res = *iter; |
| 2358 | iter = &priv->p_tmp_res; |
| 2359 | } |
| 2360 | spin_unlock(&ocfs2_dlm_tracking_lock); |
| 2361 | |
| 2362 | return iter; |
| 2363 | } |
| 2364 | |
| 2365 | /* So that debugfs.ocfs2 can determine which format is being used */ |
| 2366 | #define OCFS2_DLM_DEBUG_STR_VERSION 1 |
| 2367 | static int ocfs2_dlm_seq_show(struct seq_file *m, void *v) |
| 2368 | { |
| 2369 | int i; |
| 2370 | char *lvb; |
| 2371 | struct ocfs2_lock_res *lockres = v; |
| 2372 | |
| 2373 | if (!lockres) |
| 2374 | return -EINVAL; |
| 2375 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2376 | seq_printf(m, "0x%x\t", OCFS2_DLM_DEBUG_STR_VERSION); |
| 2377 | |
| 2378 | if (lockres->l_type == OCFS2_LOCK_TYPE_DENTRY) |
| 2379 | seq_printf(m, "%.*s%08x\t", OCFS2_DENTRY_LOCK_INO_START - 1, |
| 2380 | lockres->l_name, |
| 2381 | (unsigned int)ocfs2_get_dentry_lock_ino(lockres)); |
| 2382 | else |
| 2383 | seq_printf(m, "%.*s\t", OCFS2_LOCK_ID_MAX_LEN, lockres->l_name); |
| 2384 | |
| 2385 | seq_printf(m, "%d\t" |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2386 | "0x%lx\t" |
| 2387 | "0x%x\t" |
| 2388 | "0x%x\t" |
| 2389 | "%u\t" |
| 2390 | "%u\t" |
| 2391 | "%d\t" |
| 2392 | "%d\t", |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2393 | lockres->l_level, |
| 2394 | lockres->l_flags, |
| 2395 | lockres->l_action, |
| 2396 | lockres->l_unlock_action, |
| 2397 | lockres->l_ro_holders, |
| 2398 | lockres->l_ex_holders, |
| 2399 | lockres->l_requested, |
| 2400 | lockres->l_blocking); |
| 2401 | |
| 2402 | /* Dump the raw LVB */ |
| 2403 | lvb = lockres->l_lksb.lvb; |
| 2404 | for(i = 0; i < DLM_LVB_LEN; i++) |
| 2405 | seq_printf(m, "0x%x\t", lvb[i]); |
| 2406 | |
| 2407 | /* End the line */ |
| 2408 | seq_printf(m, "\n"); |
| 2409 | return 0; |
| 2410 | } |
| 2411 | |
Jan Engelhardt | 90d9977 | 2008-01-22 20:52:20 +0100 | [diff] [blame^] | 2412 | static const struct seq_operations ocfs2_dlm_seq_ops = { |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2413 | .start = ocfs2_dlm_seq_start, |
| 2414 | .stop = ocfs2_dlm_seq_stop, |
| 2415 | .next = ocfs2_dlm_seq_next, |
| 2416 | .show = ocfs2_dlm_seq_show, |
| 2417 | }; |
| 2418 | |
| 2419 | static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file) |
| 2420 | { |
| 2421 | struct seq_file *seq = (struct seq_file *) file->private_data; |
| 2422 | struct ocfs2_dlm_seq_priv *priv = seq->private; |
| 2423 | struct ocfs2_lock_res *res = &priv->p_iter_res; |
| 2424 | |
| 2425 | ocfs2_remove_lockres_tracking(res); |
| 2426 | ocfs2_put_dlm_debug(priv->p_dlm_debug); |
| 2427 | return seq_release_private(inode, file); |
| 2428 | } |
| 2429 | |
| 2430 | static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file) |
| 2431 | { |
| 2432 | int ret; |
| 2433 | struct ocfs2_dlm_seq_priv *priv; |
| 2434 | struct seq_file *seq; |
| 2435 | struct ocfs2_super *osb; |
| 2436 | |
| 2437 | priv = kzalloc(sizeof(struct ocfs2_dlm_seq_priv), GFP_KERNEL); |
| 2438 | if (!priv) { |
| 2439 | ret = -ENOMEM; |
| 2440 | mlog_errno(ret); |
| 2441 | goto out; |
| 2442 | } |
Theodore Ts'o | 8e18e29 | 2006-09-27 01:50:46 -0700 | [diff] [blame] | 2443 | osb = inode->i_private; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2444 | ocfs2_get_dlm_debug(osb->osb_dlm_debug); |
| 2445 | priv->p_dlm_debug = osb->osb_dlm_debug; |
| 2446 | INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list); |
| 2447 | |
| 2448 | ret = seq_open(file, &ocfs2_dlm_seq_ops); |
| 2449 | if (ret) { |
| 2450 | kfree(priv); |
| 2451 | mlog_errno(ret); |
| 2452 | goto out; |
| 2453 | } |
| 2454 | |
| 2455 | seq = (struct seq_file *) file->private_data; |
| 2456 | seq->private = priv; |
| 2457 | |
| 2458 | ocfs2_add_lockres_tracking(&priv->p_iter_res, |
| 2459 | priv->p_dlm_debug); |
| 2460 | |
| 2461 | out: |
| 2462 | return ret; |
| 2463 | } |
| 2464 | |
Arjan van de Ven | 4b6f5d2 | 2006-03-28 01:56:42 -0800 | [diff] [blame] | 2465 | static const struct file_operations ocfs2_dlm_debug_fops = { |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2466 | .open = ocfs2_dlm_debug_open, |
| 2467 | .release = ocfs2_dlm_debug_release, |
| 2468 | .read = seq_read, |
| 2469 | .llseek = seq_lseek, |
| 2470 | }; |
| 2471 | |
| 2472 | static int ocfs2_dlm_init_debug(struct ocfs2_super *osb) |
| 2473 | { |
| 2474 | int ret = 0; |
| 2475 | struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug; |
| 2476 | |
| 2477 | dlm_debug->d_locking_state = debugfs_create_file("locking_state", |
| 2478 | S_IFREG|S_IRUSR, |
| 2479 | osb->osb_debug_root, |
| 2480 | osb, |
| 2481 | &ocfs2_dlm_debug_fops); |
| 2482 | if (!dlm_debug->d_locking_state) { |
| 2483 | ret = -EINVAL; |
| 2484 | mlog(ML_ERROR, |
| 2485 | "Unable to create locking state debugfs file.\n"); |
| 2486 | goto out; |
| 2487 | } |
| 2488 | |
| 2489 | ocfs2_get_dlm_debug(dlm_debug); |
| 2490 | out: |
| 2491 | return ret; |
| 2492 | } |
| 2493 | |
| 2494 | static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb) |
| 2495 | { |
| 2496 | struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug; |
| 2497 | |
| 2498 | if (dlm_debug) { |
| 2499 | debugfs_remove(dlm_debug->d_locking_state); |
| 2500 | ocfs2_put_dlm_debug(dlm_debug); |
| 2501 | } |
| 2502 | } |
| 2503 | |
| 2504 | int ocfs2_dlm_init(struct ocfs2_super *osb) |
| 2505 | { |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2506 | int status = 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2507 | u32 dlm_key; |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2508 | struct dlm_ctxt *dlm = NULL; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2509 | |
| 2510 | mlog_entry_void(); |
| 2511 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2512 | if (ocfs2_mount_local(osb)) |
| 2513 | goto local; |
| 2514 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2515 | status = ocfs2_dlm_init_debug(osb); |
| 2516 | if (status < 0) { |
| 2517 | mlog_errno(status); |
| 2518 | goto bail; |
| 2519 | } |
| 2520 | |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 2521 | /* launch downconvert thread */ |
| 2522 | osb->dc_task = kthread_run(ocfs2_downconvert_thread, osb, "ocfs2dc"); |
| 2523 | if (IS_ERR(osb->dc_task)) { |
| 2524 | status = PTR_ERR(osb->dc_task); |
| 2525 | osb->dc_task = NULL; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2526 | mlog_errno(status); |
| 2527 | goto bail; |
| 2528 | } |
| 2529 | |
| 2530 | /* used by the dlm code to make message headers unique, each |
| 2531 | * node in this domain must agree on this. */ |
| 2532 | dlm_key = crc32_le(0, osb->uuid_str, strlen(osb->uuid_str)); |
| 2533 | |
| 2534 | /* for now, uuid == domain */ |
Joel Becker | d24fbcd | 2008-01-25 17:02:21 -0800 | [diff] [blame] | 2535 | dlm = dlm_register_domain(osb->uuid_str, dlm_key, |
| 2536 | &osb->osb_locking_proto); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2537 | if (IS_ERR(dlm)) { |
| 2538 | status = PTR_ERR(dlm); |
| 2539 | mlog_errno(status); |
| 2540 | goto bail; |
| 2541 | } |
| 2542 | |
Sunil Mushran | c271c5c | 2006-12-05 17:56:35 -0800 | [diff] [blame] | 2543 | dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb); |
| 2544 | |
| 2545 | local: |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2546 | ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb); |
| 2547 | ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb); |
| 2548 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2549 | osb->dlm = dlm; |
| 2550 | |
| 2551 | status = 0; |
| 2552 | bail: |
| 2553 | if (status < 0) { |
| 2554 | ocfs2_dlm_shutdown_debug(osb); |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 2555 | if (osb->dc_task) |
| 2556 | kthread_stop(osb->dc_task); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2557 | } |
| 2558 | |
| 2559 | mlog_exit(status); |
| 2560 | return status; |
| 2561 | } |
| 2562 | |
| 2563 | void ocfs2_dlm_shutdown(struct ocfs2_super *osb) |
| 2564 | { |
| 2565 | mlog_entry_void(); |
| 2566 | |
| 2567 | dlm_unregister_eviction_cb(&osb->osb_eviction_cb); |
| 2568 | |
| 2569 | ocfs2_drop_osb_locks(osb); |
| 2570 | |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 2571 | if (osb->dc_task) { |
| 2572 | kthread_stop(osb->dc_task); |
| 2573 | osb->dc_task = NULL; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2574 | } |
| 2575 | |
| 2576 | ocfs2_lock_res_free(&osb->osb_super_lockres); |
| 2577 | ocfs2_lock_res_free(&osb->osb_rename_lockres); |
| 2578 | |
| 2579 | dlm_unregister_domain(osb->dlm); |
| 2580 | osb->dlm = NULL; |
| 2581 | |
| 2582 | ocfs2_dlm_shutdown_debug(osb); |
| 2583 | |
| 2584 | mlog_exit_void(); |
| 2585 | } |
| 2586 | |
Mark Fasheh | 2a45f2d | 2006-09-12 21:36:58 -0700 | [diff] [blame] | 2587 | static void ocfs2_unlock_ast(void *opaque, enum dlm_status status) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2588 | { |
| 2589 | struct ocfs2_lock_res *lockres = opaque; |
| 2590 | unsigned long flags; |
| 2591 | |
| 2592 | mlog_entry_void(); |
| 2593 | |
| 2594 | mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name, |
| 2595 | lockres->l_unlock_action); |
| 2596 | |
| 2597 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2598 | /* We tried to cancel a convert request, but it was already |
| 2599 | * granted. All we want to do here is clear our unlock |
| 2600 | * state. The wake_up call done at the bottom is redundant |
| 2601 | * (ocfs2_prepare_cancel_convert doesn't sleep on this) but doesn't |
| 2602 | * hurt anything anyway */ |
| 2603 | if (status == DLM_CANCELGRANT && |
| 2604 | lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) { |
| 2605 | mlog(0, "Got cancelgrant for %s\n", lockres->l_name); |
| 2606 | |
| 2607 | /* We don't clear the busy flag in this case as it |
| 2608 | * should have been cleared by the ast which the dlm |
| 2609 | * has called. */ |
| 2610 | goto complete_unlock; |
| 2611 | } |
| 2612 | |
| 2613 | if (status != DLM_NORMAL) { |
| 2614 | mlog(ML_ERROR, "Dlm passes status %d for lock %s, " |
| 2615 | "unlock_action %d\n", status, lockres->l_name, |
| 2616 | lockres->l_unlock_action); |
| 2617 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2618 | return; |
| 2619 | } |
| 2620 | |
| 2621 | switch(lockres->l_unlock_action) { |
| 2622 | case OCFS2_UNLOCK_CANCEL_CONVERT: |
| 2623 | mlog(0, "Cancel convert success for %s\n", lockres->l_name); |
| 2624 | lockres->l_action = OCFS2_AST_INVALID; |
| 2625 | break; |
| 2626 | case OCFS2_UNLOCK_DROP_LOCK: |
| 2627 | lockres->l_level = LKM_IVMODE; |
| 2628 | break; |
| 2629 | default: |
| 2630 | BUG(); |
| 2631 | } |
| 2632 | |
| 2633 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); |
| 2634 | complete_unlock: |
| 2635 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; |
| 2636 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2637 | |
| 2638 | wake_up(&lockres->l_event); |
| 2639 | |
| 2640 | mlog_exit_void(); |
| 2641 | } |
| 2642 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2643 | static int ocfs2_drop_lock(struct ocfs2_super *osb, |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 2644 | struct ocfs2_lock_res *lockres) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2645 | { |
| 2646 | enum dlm_status status; |
| 2647 | unsigned long flags; |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 2648 | int lkm_flags = 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2649 | |
| 2650 | /* We didn't get anywhere near actually using this lockres. */ |
| 2651 | if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED)) |
| 2652 | goto out; |
| 2653 | |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 2654 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) |
| 2655 | lkm_flags |= LKM_VALBLK; |
| 2656 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2657 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2658 | |
| 2659 | mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING), |
| 2660 | "lockres %s, flags 0x%lx\n", |
| 2661 | lockres->l_name, lockres->l_flags); |
| 2662 | |
| 2663 | while (lockres->l_flags & OCFS2_LOCK_BUSY) { |
| 2664 | mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = " |
| 2665 | "%u, unlock_action = %u\n", |
| 2666 | lockres->l_name, lockres->l_flags, lockres->l_action, |
| 2667 | lockres->l_unlock_action); |
| 2668 | |
| 2669 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2670 | |
| 2671 | /* XXX: Today we just wait on any busy |
| 2672 | * locks... Perhaps we need to cancel converts in the |
| 2673 | * future? */ |
| 2674 | ocfs2_wait_on_busy_lock(lockres); |
| 2675 | |
| 2676 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2677 | } |
| 2678 | |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 2679 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) { |
| 2680 | if (lockres->l_flags & OCFS2_LOCK_ATTACHED && |
| 2681 | lockres->l_level == LKM_EXMODE && |
| 2682 | !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) |
| 2683 | lockres->l_ops->set_lvb(lockres); |
| 2684 | } |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2685 | |
| 2686 | if (lockres->l_flags & OCFS2_LOCK_BUSY) |
| 2687 | mlog(ML_ERROR, "destroying busy lock: \"%s\"\n", |
| 2688 | lockres->l_name); |
| 2689 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) |
| 2690 | mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name); |
| 2691 | |
| 2692 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { |
| 2693 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2694 | goto out; |
| 2695 | } |
| 2696 | |
| 2697 | lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED); |
| 2698 | |
| 2699 | /* make sure we never get here while waiting for an ast to |
| 2700 | * fire. */ |
| 2701 | BUG_ON(lockres->l_action != OCFS2_AST_INVALID); |
| 2702 | |
| 2703 | /* is this necessary? */ |
| 2704 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); |
| 2705 | lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK; |
| 2706 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2707 | |
| 2708 | mlog(0, "lock %s\n", lockres->l_name); |
| 2709 | |
Mark Fasheh | b80fc01 | 2006-09-12 22:08:14 -0700 | [diff] [blame] | 2710 | status = dlmunlock(osb->dlm, &lockres->l_lksb, lkm_flags, |
Mark Fasheh | 2a45f2d | 2006-09-12 21:36:58 -0700 | [diff] [blame] | 2711 | ocfs2_unlock_ast, lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2712 | if (status != DLM_NORMAL) { |
| 2713 | ocfs2_log_dlm_error("dlmunlock", status, lockres); |
| 2714 | mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags); |
| 2715 | dlm_print_one_lock(lockres->l_lksb.lockid); |
| 2716 | BUG(); |
| 2717 | } |
| 2718 | mlog(0, "lock %s, successfull return from dlmunlock\n", |
| 2719 | lockres->l_name); |
| 2720 | |
| 2721 | ocfs2_wait_on_busy_lock(lockres); |
| 2722 | out: |
| 2723 | mlog_exit(0); |
| 2724 | return 0; |
| 2725 | } |
| 2726 | |
| 2727 | /* Mark the lockres as being dropped. It will no longer be |
| 2728 | * queued if blocking, but we still may have to wait on it |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 2729 | * being dequeued from the downconvert thread before we can consider |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2730 | * it safe to drop. |
| 2731 | * |
| 2732 | * You can *not* attempt to call cluster_lock on this lockres anymore. */ |
| 2733 | void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres) |
| 2734 | { |
| 2735 | int status; |
| 2736 | struct ocfs2_mask_waiter mw; |
| 2737 | unsigned long flags; |
| 2738 | |
| 2739 | ocfs2_init_mask_waiter(&mw); |
| 2740 | |
| 2741 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2742 | lockres->l_flags |= OCFS2_LOCK_FREEING; |
| 2743 | while (lockres->l_flags & OCFS2_LOCK_QUEUED) { |
| 2744 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0); |
| 2745 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2746 | |
| 2747 | mlog(0, "Waiting on lockres %s\n", lockres->l_name); |
| 2748 | |
| 2749 | status = ocfs2_wait_for_mask(&mw); |
| 2750 | if (status) |
| 2751 | mlog_errno(status); |
| 2752 | |
| 2753 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2754 | } |
| 2755 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2756 | } |
| 2757 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2758 | void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, |
| 2759 | struct ocfs2_lock_res *lockres) |
| 2760 | { |
| 2761 | int ret; |
| 2762 | |
| 2763 | ocfs2_mark_lockres_freeing(lockres); |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 2764 | ret = ocfs2_drop_lock(osb, lockres); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2765 | if (ret) |
| 2766 | mlog_errno(ret); |
| 2767 | } |
| 2768 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2769 | static void ocfs2_drop_osb_locks(struct ocfs2_super *osb) |
| 2770 | { |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2771 | ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres); |
| 2772 | ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2773 | } |
| 2774 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2775 | int ocfs2_drop_inode_locks(struct inode *inode) |
| 2776 | { |
| 2777 | int status, err; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2778 | |
| 2779 | mlog_entry_void(); |
| 2780 | |
| 2781 | /* No need to call ocfs2_mark_lockres_freeing here - |
| 2782 | * ocfs2_clear_inode has done it for us. */ |
| 2783 | |
| 2784 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), |
Tiger Yang | 5000863 | 2007-03-20 16:01:38 -0700 | [diff] [blame] | 2785 | &OCFS2_I(inode)->ip_open_lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2786 | if (err < 0) |
| 2787 | mlog_errno(err); |
| 2788 | |
| 2789 | status = err; |
| 2790 | |
| 2791 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), |
Mark Fasheh | e63aecb6 | 2007-10-18 15:30:42 -0700 | [diff] [blame] | 2792 | &OCFS2_I(inode)->ip_inode_lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2793 | if (err < 0) |
| 2794 | mlog_errno(err); |
| 2795 | if (err < 0 && !status) |
| 2796 | status = err; |
| 2797 | |
| 2798 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), |
Mark Fasheh | 0d5dc6c | 2006-09-14 14:44:51 -0700 | [diff] [blame] | 2799 | &OCFS2_I(inode)->ip_rw_lockres); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2800 | if (err < 0) |
| 2801 | mlog_errno(err); |
| 2802 | if (err < 0 && !status) |
| 2803 | status = err; |
| 2804 | |
| 2805 | mlog_exit(status); |
| 2806 | return status; |
| 2807 | } |
| 2808 | |
| 2809 | static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, |
| 2810 | int new_level) |
| 2811 | { |
| 2812 | assert_spin_locked(&lockres->l_lock); |
| 2813 | |
| 2814 | BUG_ON(lockres->l_blocking <= LKM_NLMODE); |
| 2815 | |
| 2816 | if (lockres->l_level <= new_level) { |
| 2817 | mlog(ML_ERROR, "lockres->l_level (%u) <= new_level (%u)\n", |
| 2818 | lockres->l_level, new_level); |
| 2819 | BUG(); |
| 2820 | } |
| 2821 | |
| 2822 | mlog(0, "lock %s, new_level = %d, l_blocking = %d\n", |
| 2823 | lockres->l_name, new_level, lockres->l_blocking); |
| 2824 | |
| 2825 | lockres->l_action = OCFS2_AST_DOWNCONVERT; |
| 2826 | lockres->l_requested = new_level; |
| 2827 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); |
| 2828 | } |
| 2829 | |
| 2830 | static int ocfs2_downconvert_lock(struct ocfs2_super *osb, |
| 2831 | struct ocfs2_lock_res *lockres, |
| 2832 | int new_level, |
| 2833 | int lvb) |
| 2834 | { |
| 2835 | int ret, dlm_flags = LKM_CONVERT; |
| 2836 | enum dlm_status status; |
| 2837 | |
| 2838 | mlog_entry_void(); |
| 2839 | |
| 2840 | if (lvb) |
| 2841 | dlm_flags |= LKM_VALBLK; |
| 2842 | |
| 2843 | status = dlmlock(osb->dlm, |
| 2844 | new_level, |
| 2845 | &lockres->l_lksb, |
| 2846 | dlm_flags, |
| 2847 | lockres->l_name, |
Mark Fasheh | f068106 | 2006-09-08 11:40:10 -0700 | [diff] [blame] | 2848 | OCFS2_LOCK_ID_MAX_LEN - 1, |
Mark Fasheh | e92d57d | 2006-09-12 21:34:35 -0700 | [diff] [blame] | 2849 | ocfs2_locking_ast, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2850 | lockres, |
Mark Fasheh | aa2623a | 2006-09-12 21:58:23 -0700 | [diff] [blame] | 2851 | ocfs2_blocking_ast); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2852 | if (status != DLM_NORMAL) { |
| 2853 | ocfs2_log_dlm_error("dlmlock", status, lockres); |
| 2854 | ret = -EINVAL; |
| 2855 | ocfs2_recover_from_dlm_error(lockres, 1); |
| 2856 | goto bail; |
| 2857 | } |
| 2858 | |
| 2859 | ret = 0; |
| 2860 | bail: |
| 2861 | mlog_exit(ret); |
| 2862 | return ret; |
| 2863 | } |
| 2864 | |
| 2865 | /* returns 1 when the caller should unlock and call dlmunlock */ |
| 2866 | static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, |
| 2867 | struct ocfs2_lock_res *lockres) |
| 2868 | { |
| 2869 | assert_spin_locked(&lockres->l_lock); |
| 2870 | |
| 2871 | mlog_entry_void(); |
| 2872 | mlog(0, "lock %s\n", lockres->l_name); |
| 2873 | |
| 2874 | if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) { |
| 2875 | /* If we're already trying to cancel a lock conversion |
| 2876 | * then just drop the spinlock and allow the caller to |
| 2877 | * requeue this lock. */ |
| 2878 | |
| 2879 | mlog(0, "Lockres %s, skip convert\n", lockres->l_name); |
| 2880 | return 0; |
| 2881 | } |
| 2882 | |
| 2883 | /* were we in a convert when we got the bast fire? */ |
| 2884 | BUG_ON(lockres->l_action != OCFS2_AST_CONVERT && |
| 2885 | lockres->l_action != OCFS2_AST_DOWNCONVERT); |
| 2886 | /* set things up for the unlockast to know to just |
| 2887 | * clear out the ast_action and unset busy, etc. */ |
| 2888 | lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT; |
| 2889 | |
| 2890 | mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY), |
| 2891 | "lock %s, invalid flags: 0x%lx\n", |
| 2892 | lockres->l_name, lockres->l_flags); |
| 2893 | |
| 2894 | return 1; |
| 2895 | } |
| 2896 | |
| 2897 | static int ocfs2_cancel_convert(struct ocfs2_super *osb, |
| 2898 | struct ocfs2_lock_res *lockres) |
| 2899 | { |
| 2900 | int ret; |
| 2901 | enum dlm_status status; |
| 2902 | |
| 2903 | mlog_entry_void(); |
| 2904 | mlog(0, "lock %s\n", lockres->l_name); |
| 2905 | |
| 2906 | ret = 0; |
| 2907 | status = dlmunlock(osb->dlm, |
| 2908 | &lockres->l_lksb, |
| 2909 | LKM_CANCEL, |
Mark Fasheh | 2a45f2d | 2006-09-12 21:36:58 -0700 | [diff] [blame] | 2910 | ocfs2_unlock_ast, |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2911 | lockres); |
| 2912 | if (status != DLM_NORMAL) { |
| 2913 | ocfs2_log_dlm_error("dlmunlock", status, lockres); |
| 2914 | ret = -EINVAL; |
| 2915 | ocfs2_recover_from_dlm_error(lockres, 0); |
| 2916 | } |
| 2917 | |
| 2918 | mlog(0, "lock %s return from dlmunlock\n", lockres->l_name); |
| 2919 | |
| 2920 | mlog_exit(ret); |
| 2921 | return ret; |
| 2922 | } |
| 2923 | |
Mark Fasheh | b5e500e | 2006-09-13 22:01:16 -0700 | [diff] [blame] | 2924 | static int ocfs2_unblock_lock(struct ocfs2_super *osb, |
| 2925 | struct ocfs2_lock_res *lockres, |
| 2926 | struct ocfs2_unblock_ctl *ctl) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2927 | { |
| 2928 | unsigned long flags; |
| 2929 | int blocking; |
| 2930 | int new_level; |
| 2931 | int ret = 0; |
Mark Fasheh | 5ef0d4e | 2006-09-13 21:21:52 -0700 | [diff] [blame] | 2932 | int set_lvb = 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2933 | |
| 2934 | mlog_entry_void(); |
| 2935 | |
| 2936 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2937 | |
| 2938 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); |
| 2939 | |
| 2940 | recheck: |
| 2941 | if (lockres->l_flags & OCFS2_LOCK_BUSY) { |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2942 | ctl->requeue = 1; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2943 | ret = ocfs2_prepare_cancel_convert(osb, lockres); |
| 2944 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2945 | if (ret) { |
| 2946 | ret = ocfs2_cancel_convert(osb, lockres); |
| 2947 | if (ret < 0) |
| 2948 | mlog_errno(ret); |
| 2949 | } |
| 2950 | goto leave; |
| 2951 | } |
| 2952 | |
| 2953 | /* if we're blocking an exclusive and we have *any* holders, |
| 2954 | * then requeue. */ |
| 2955 | if ((lockres->l_blocking == LKM_EXMODE) |
Mark Fasheh | f7fbfdd | 2006-09-13 21:02:29 -0700 | [diff] [blame] | 2956 | && (lockres->l_ex_holders || lockres->l_ro_holders)) |
| 2957 | goto leave_requeue; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2958 | |
| 2959 | /* If it's a PR we're blocking, then only |
| 2960 | * requeue if we've got any EX holders */ |
| 2961 | if (lockres->l_blocking == LKM_PRMODE && |
Mark Fasheh | f7fbfdd | 2006-09-13 21:02:29 -0700 | [diff] [blame] | 2962 | lockres->l_ex_holders) |
| 2963 | goto leave_requeue; |
| 2964 | |
| 2965 | /* |
| 2966 | * Can we get a lock in this state if the holder counts are |
| 2967 | * zero? The meta data unblock code used to check this. |
| 2968 | */ |
| 2969 | if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) |
| 2970 | && (lockres->l_flags & OCFS2_LOCK_REFRESHING)) |
| 2971 | goto leave_requeue; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2972 | |
Mark Fasheh | 16d5b95 | 2006-09-13 21:10:12 -0700 | [diff] [blame] | 2973 | new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking); |
| 2974 | |
| 2975 | if (lockres->l_ops->check_downconvert |
| 2976 | && !lockres->l_ops->check_downconvert(lockres, new_level)) |
| 2977 | goto leave_requeue; |
| 2978 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2979 | /* If we get here, then we know that there are no more |
| 2980 | * incompatible holders (and anyone asking for an incompatible |
| 2981 | * lock is blocked). We can now downconvert the lock */ |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 2982 | if (!lockres->l_ops->downconvert_worker) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2983 | goto downconvert; |
| 2984 | |
| 2985 | /* Some lockres types want to do a bit of work before |
| 2986 | * downconverting a lock. Allow that here. The worker function |
| 2987 | * may sleep, so we save off a copy of what we're blocking as |
| 2988 | * it may change while we're not holding the spin lock. */ |
| 2989 | blocking = lockres->l_blocking; |
| 2990 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 2991 | |
Mark Fasheh | cc567d8 | 2006-09-13 21:52:21 -0700 | [diff] [blame] | 2992 | ctl->unblock_action = lockres->l_ops->downconvert_worker(lockres, blocking); |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 2993 | |
| 2994 | if (ctl->unblock_action == UNBLOCK_STOP_POST) |
| 2995 | goto leave; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 2996 | |
| 2997 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 2998 | if (blocking != lockres->l_blocking) { |
| 2999 | /* If this changed underneath us, then we can't drop |
| 3000 | * it just yet. */ |
| 3001 | goto recheck; |
| 3002 | } |
| 3003 | |
| 3004 | downconvert: |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3005 | ctl->requeue = 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3006 | |
Mark Fasheh | 5ef0d4e | 2006-09-13 21:21:52 -0700 | [diff] [blame] | 3007 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) { |
| 3008 | if (lockres->l_level == LKM_EXMODE) |
| 3009 | set_lvb = 1; |
| 3010 | |
| 3011 | /* |
| 3012 | * We only set the lvb if the lock has been fully |
| 3013 | * refreshed - otherwise we risk setting stale |
| 3014 | * data. Otherwise, there's no need to actually clear |
| 3015 | * out the lvb here as it's value is still valid. |
| 3016 | */ |
| 3017 | if (set_lvb && !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) |
| 3018 | lockres->l_ops->set_lvb(lockres); |
| 3019 | } |
| 3020 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3021 | ocfs2_prepare_downconvert(lockres, new_level); |
| 3022 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
Mark Fasheh | 5ef0d4e | 2006-09-13 21:21:52 -0700 | [diff] [blame] | 3023 | ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3024 | leave: |
| 3025 | mlog_exit(ret); |
| 3026 | return ret; |
Mark Fasheh | f7fbfdd | 2006-09-13 21:02:29 -0700 | [diff] [blame] | 3027 | |
| 3028 | leave_requeue: |
| 3029 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 3030 | ctl->requeue = 1; |
| 3031 | |
| 3032 | mlog_exit(0); |
| 3033 | return 0; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3034 | } |
| 3035 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3036 | static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, |
| 3037 | int blocking) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3038 | { |
| 3039 | struct inode *inode; |
| 3040 | struct address_space *mapping; |
| 3041 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3042 | inode = ocfs2_lock_res_inode(lockres); |
| 3043 | mapping = inode->i_mapping; |
| 3044 | |
Mark Fasheh | 1044e40 | 2008-02-28 17:16:03 -0800 | [diff] [blame] | 3045 | if (!S_ISREG(inode->i_mode)) |
Mark Fasheh | f1f5406 | 2007-10-18 15:13:59 -0700 | [diff] [blame] | 3046 | goto out; |
| 3047 | |
Mark Fasheh | 7f4a2a9 | 2006-12-11 11:06:36 -0800 | [diff] [blame] | 3048 | /* |
| 3049 | * We need this before the filemap_fdatawrite() so that it can |
| 3050 | * transfer the dirty bit from the PTE to the |
| 3051 | * page. Unfortunately this means that even for EX->PR |
| 3052 | * downconverts, we'll lose our mappings and have to build |
| 3053 | * them up again. |
| 3054 | */ |
| 3055 | unmap_mapping_range(mapping, 0, 0, 0); |
| 3056 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3057 | if (filemap_fdatawrite(mapping)) { |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 3058 | mlog(ML_ERROR, "Could not sync inode %llu for downconvert!", |
| 3059 | (unsigned long long)OCFS2_I(inode)->ip_blkno); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3060 | } |
| 3061 | sync_mapping_buffers(mapping); |
| 3062 | if (blocking == LKM_EXMODE) { |
| 3063 | truncate_inode_pages(mapping, 0); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3064 | } else { |
| 3065 | /* We only need to wait on the I/O if we're not also |
| 3066 | * truncating pages because truncate_inode_pages waits |
| 3067 | * for us above. We don't truncate pages if we're |
| 3068 | * blocking anything < EXMODE because we want to keep |
| 3069 | * them around in that case. */ |
| 3070 | filemap_fdatawait(mapping); |
| 3071 | } |
| 3072 | |
Mark Fasheh | f1f5406 | 2007-10-18 15:13:59 -0700 | [diff] [blame] | 3073 | out: |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3074 | return UNBLOCK_CONTINUE; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3075 | } |
| 3076 | |
Mark Fasheh | 810d5ae | 2006-09-13 21:39:52 -0700 | [diff] [blame] | 3077 | static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, |
| 3078 | int new_level) |
| 3079 | { |
| 3080 | struct inode *inode = ocfs2_lock_res_inode(lockres); |
| 3081 | int checkpointed = ocfs2_inode_fully_checkpointed(inode); |
| 3082 | |
| 3083 | BUG_ON(new_level != LKM_NLMODE && new_level != LKM_PRMODE); |
| 3084 | BUG_ON(lockres->l_level != LKM_EXMODE && !checkpointed); |
| 3085 | |
| 3086 | if (checkpointed) |
| 3087 | return 1; |
| 3088 | |
| 3089 | ocfs2_start_checkpoint(OCFS2_SB(inode->i_sb)); |
| 3090 | return 0; |
| 3091 | } |
| 3092 | |
| 3093 | static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres) |
| 3094 | { |
| 3095 | struct inode *inode = ocfs2_lock_res_inode(lockres); |
| 3096 | |
| 3097 | __ocfs2_stuff_meta_lvb(inode); |
| 3098 | } |
| 3099 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3100 | /* |
| 3101 | * Does the final reference drop on our dentry lock. Right now this |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 3102 | * happens in the downconvert thread, but we could choose to simplify the |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3103 | * dlmglue API and push these off to the ocfs2_wq in the future. |
| 3104 | */ |
| 3105 | static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, |
| 3106 | struct ocfs2_lock_res *lockres) |
| 3107 | { |
| 3108 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); |
| 3109 | ocfs2_dentry_lock_put(osb, dl); |
| 3110 | } |
| 3111 | |
| 3112 | /* |
| 3113 | * d_delete() matching dentries before the lock downconvert. |
| 3114 | * |
| 3115 | * At this point, any process waiting to destroy the |
| 3116 | * dentry_lock due to last ref count is stopped by the |
| 3117 | * OCFS2_LOCK_QUEUED flag. |
| 3118 | * |
| 3119 | * We have two potential problems |
| 3120 | * |
| 3121 | * 1) If we do the last reference drop on our dentry_lock (via dput) |
| 3122 | * we'll wind up in ocfs2_release_dentry_lock(), waiting on |
| 3123 | * the downconvert to finish. Instead we take an elevated |
| 3124 | * reference and push the drop until after we've completed our |
| 3125 | * unblock processing. |
| 3126 | * |
| 3127 | * 2) There might be another process with a final reference, |
| 3128 | * waiting on us to finish processing. If this is the case, we |
| 3129 | * detect it and exit out - there's no more dentries anyway. |
| 3130 | */ |
| 3131 | static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres, |
| 3132 | int blocking) |
| 3133 | { |
| 3134 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); |
| 3135 | struct ocfs2_inode_info *oi = OCFS2_I(dl->dl_inode); |
| 3136 | struct dentry *dentry; |
| 3137 | unsigned long flags; |
| 3138 | int extra_ref = 0; |
| 3139 | |
| 3140 | /* |
| 3141 | * This node is blocking another node from getting a read |
| 3142 | * lock. This happens when we've renamed within a |
| 3143 | * directory. We've forced the other nodes to d_delete(), but |
| 3144 | * we never actually dropped our lock because it's still |
| 3145 | * valid. The downconvert code will retain a PR for this node, |
| 3146 | * so there's no further work to do. |
| 3147 | */ |
| 3148 | if (blocking == LKM_PRMODE) |
| 3149 | return UNBLOCK_CONTINUE; |
| 3150 | |
| 3151 | /* |
| 3152 | * Mark this inode as potentially orphaned. The code in |
| 3153 | * ocfs2_delete_inode() will figure out whether it actually |
| 3154 | * needs to be freed or not. |
| 3155 | */ |
| 3156 | spin_lock(&oi->ip_lock); |
| 3157 | oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED; |
| 3158 | spin_unlock(&oi->ip_lock); |
| 3159 | |
| 3160 | /* |
| 3161 | * Yuck. We need to make sure however that the check of |
| 3162 | * OCFS2_LOCK_FREEING and the extra reference are atomic with |
| 3163 | * respect to a reference decrement or the setting of that |
| 3164 | * flag. |
| 3165 | */ |
| 3166 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 3167 | spin_lock(&dentry_attach_lock); |
| 3168 | if (!(lockres->l_flags & OCFS2_LOCK_FREEING) |
| 3169 | && dl->dl_count) { |
| 3170 | dl->dl_count++; |
| 3171 | extra_ref = 1; |
| 3172 | } |
| 3173 | spin_unlock(&dentry_attach_lock); |
| 3174 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 3175 | |
| 3176 | mlog(0, "extra_ref = %d\n", extra_ref); |
| 3177 | |
| 3178 | /* |
| 3179 | * We have a process waiting on us in ocfs2_dentry_iput(), |
| 3180 | * which means we can't have any more outstanding |
| 3181 | * aliases. There's no need to do any more work. |
| 3182 | */ |
| 3183 | if (!extra_ref) |
| 3184 | return UNBLOCK_CONTINUE; |
| 3185 | |
| 3186 | spin_lock(&dentry_attach_lock); |
| 3187 | while (1) { |
| 3188 | dentry = ocfs2_find_local_alias(dl->dl_inode, |
| 3189 | dl->dl_parent_blkno, 1); |
| 3190 | if (!dentry) |
| 3191 | break; |
| 3192 | spin_unlock(&dentry_attach_lock); |
| 3193 | |
| 3194 | mlog(0, "d_delete(%.*s);\n", dentry->d_name.len, |
| 3195 | dentry->d_name.name); |
| 3196 | |
| 3197 | /* |
| 3198 | * The following dcache calls may do an |
| 3199 | * iput(). Normally we don't want that from the |
| 3200 | * downconverting thread, but in this case it's ok |
| 3201 | * because the requesting node already has an |
| 3202 | * exclusive lock on the inode, so it can't be queued |
| 3203 | * for a downconvert. |
| 3204 | */ |
| 3205 | d_delete(dentry); |
| 3206 | dput(dentry); |
| 3207 | |
| 3208 | spin_lock(&dentry_attach_lock); |
| 3209 | } |
| 3210 | spin_unlock(&dentry_attach_lock); |
| 3211 | |
| 3212 | /* |
| 3213 | * If we are the last holder of this dentry lock, there is no |
| 3214 | * reason to downconvert so skip straight to the unlock. |
| 3215 | */ |
| 3216 | if (dl->dl_count == 1) |
| 3217 | return UNBLOCK_STOP_POST; |
| 3218 | |
| 3219 | return UNBLOCK_CONTINUE_POST; |
| 3220 | } |
| 3221 | |
Adrian Bunk | 0060005 | 2008-01-29 00:11:41 +0200 | [diff] [blame] | 3222 | static void ocfs2_process_blocked_lock(struct ocfs2_super *osb, |
| 3223 | struct ocfs2_lock_res *lockres) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3224 | { |
| 3225 | int status; |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3226 | struct ocfs2_unblock_ctl ctl = {0, 0,}; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3227 | unsigned long flags; |
| 3228 | |
| 3229 | /* Our reference to the lockres in this function can be |
| 3230 | * considered valid until we remove the OCFS2_LOCK_QUEUED |
| 3231 | * flag. */ |
| 3232 | |
| 3233 | mlog_entry_void(); |
| 3234 | |
| 3235 | BUG_ON(!lockres); |
| 3236 | BUG_ON(!lockres->l_ops); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3237 | |
| 3238 | mlog(0, "lockres %s blocked.\n", lockres->l_name); |
| 3239 | |
| 3240 | /* Detect whether a lock has been marked as going away while |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 3241 | * the downconvert thread was processing other things. A lock can |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3242 | * still be marked with OCFS2_LOCK_FREEING after this check, |
| 3243 | * but short circuiting here will still save us some |
| 3244 | * performance. */ |
| 3245 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 3246 | if (lockres->l_flags & OCFS2_LOCK_FREEING) |
| 3247 | goto unqueue; |
| 3248 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 3249 | |
Mark Fasheh | b5e500e | 2006-09-13 22:01:16 -0700 | [diff] [blame] | 3250 | status = ocfs2_unblock_lock(osb, lockres, &ctl); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3251 | if (status < 0) |
| 3252 | mlog_errno(status); |
| 3253 | |
| 3254 | spin_lock_irqsave(&lockres->l_lock, flags); |
| 3255 | unqueue: |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3256 | if (lockres->l_flags & OCFS2_LOCK_FREEING || !ctl.requeue) { |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3257 | lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED); |
| 3258 | } else |
| 3259 | ocfs2_schedule_blocked_lock(osb, lockres); |
| 3260 | |
| 3261 | mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name, |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3262 | ctl.requeue ? "yes" : "no"); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3263 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
| 3264 | |
Mark Fasheh | d680efe | 2006-09-08 14:14:34 -0700 | [diff] [blame] | 3265 | if (ctl.unblock_action != UNBLOCK_CONTINUE |
| 3266 | && lockres->l_ops->post_unlock) |
| 3267 | lockres->l_ops->post_unlock(osb, lockres); |
| 3268 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3269 | mlog_exit_void(); |
| 3270 | } |
| 3271 | |
| 3272 | static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, |
| 3273 | struct ocfs2_lock_res *lockres) |
| 3274 | { |
| 3275 | mlog_entry_void(); |
| 3276 | |
| 3277 | assert_spin_locked(&lockres->l_lock); |
| 3278 | |
| 3279 | if (lockres->l_flags & OCFS2_LOCK_FREEING) { |
| 3280 | /* Do not schedule a lock for downconvert when it's on |
| 3281 | * the way to destruction - any nodes wanting access |
| 3282 | * to the resource will get it soon. */ |
| 3283 | mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n", |
| 3284 | lockres->l_name, lockres->l_flags); |
| 3285 | return; |
| 3286 | } |
| 3287 | |
| 3288 | lockres_or_flags(lockres, OCFS2_LOCK_QUEUED); |
| 3289 | |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 3290 | spin_lock(&osb->dc_task_lock); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3291 | if (list_empty(&lockres->l_blocked_list)) { |
| 3292 | list_add_tail(&lockres->l_blocked_list, |
| 3293 | &osb->blocked_lock_list); |
| 3294 | osb->blocked_lock_count++; |
| 3295 | } |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 3296 | spin_unlock(&osb->dc_task_lock); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 3297 | |
| 3298 | mlog_exit_void(); |
| 3299 | } |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 3300 | |
| 3301 | static void ocfs2_downconvert_thread_do_work(struct ocfs2_super *osb) |
| 3302 | { |
| 3303 | unsigned long processed; |
| 3304 | struct ocfs2_lock_res *lockres; |
| 3305 | |
| 3306 | mlog_entry_void(); |
| 3307 | |
| 3308 | spin_lock(&osb->dc_task_lock); |
| 3309 | /* grab this early so we know to try again if a state change and |
| 3310 | * wake happens part-way through our work */ |
| 3311 | osb->dc_work_sequence = osb->dc_wake_sequence; |
| 3312 | |
| 3313 | processed = osb->blocked_lock_count; |
| 3314 | while (processed) { |
| 3315 | BUG_ON(list_empty(&osb->blocked_lock_list)); |
| 3316 | |
| 3317 | lockres = list_entry(osb->blocked_lock_list.next, |
| 3318 | struct ocfs2_lock_res, l_blocked_list); |
| 3319 | list_del_init(&lockres->l_blocked_list); |
| 3320 | osb->blocked_lock_count--; |
| 3321 | spin_unlock(&osb->dc_task_lock); |
| 3322 | |
| 3323 | BUG_ON(!processed); |
| 3324 | processed--; |
| 3325 | |
| 3326 | ocfs2_process_blocked_lock(osb, lockres); |
| 3327 | |
| 3328 | spin_lock(&osb->dc_task_lock); |
| 3329 | } |
| 3330 | spin_unlock(&osb->dc_task_lock); |
| 3331 | |
| 3332 | mlog_exit_void(); |
| 3333 | } |
| 3334 | |
| 3335 | static int ocfs2_downconvert_thread_lists_empty(struct ocfs2_super *osb) |
| 3336 | { |
| 3337 | int empty = 0; |
| 3338 | |
| 3339 | spin_lock(&osb->dc_task_lock); |
| 3340 | if (list_empty(&osb->blocked_lock_list)) |
| 3341 | empty = 1; |
| 3342 | |
| 3343 | spin_unlock(&osb->dc_task_lock); |
| 3344 | return empty; |
| 3345 | } |
| 3346 | |
| 3347 | static int ocfs2_downconvert_thread_should_wake(struct ocfs2_super *osb) |
| 3348 | { |
| 3349 | int should_wake = 0; |
| 3350 | |
| 3351 | spin_lock(&osb->dc_task_lock); |
| 3352 | if (osb->dc_work_sequence != osb->dc_wake_sequence) |
| 3353 | should_wake = 1; |
| 3354 | spin_unlock(&osb->dc_task_lock); |
| 3355 | |
| 3356 | return should_wake; |
| 3357 | } |
| 3358 | |
Adrian Bunk | 200bfae | 2008-02-17 10:20:38 +0200 | [diff] [blame] | 3359 | static int ocfs2_downconvert_thread(void *arg) |
Mark Fasheh | 34d024f | 2007-09-24 15:56:19 -0700 | [diff] [blame] | 3360 | { |
| 3361 | int status = 0; |
| 3362 | struct ocfs2_super *osb = arg; |
| 3363 | |
| 3364 | /* only quit once we've been asked to stop and there is no more |
| 3365 | * work available */ |
| 3366 | while (!(kthread_should_stop() && |
| 3367 | ocfs2_downconvert_thread_lists_empty(osb))) { |
| 3368 | |
| 3369 | wait_event_interruptible(osb->dc_event, |
| 3370 | ocfs2_downconvert_thread_should_wake(osb) || |
| 3371 | kthread_should_stop()); |
| 3372 | |
| 3373 | mlog(0, "downconvert_thread: awoken\n"); |
| 3374 | |
| 3375 | ocfs2_downconvert_thread_do_work(osb); |
| 3376 | } |
| 3377 | |
| 3378 | osb->dc_task = NULL; |
| 3379 | return status; |
| 3380 | } |
| 3381 | |
| 3382 | void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb) |
| 3383 | { |
| 3384 | spin_lock(&osb->dc_task_lock); |
| 3385 | /* make sure the voting thread gets a swipe at whatever changes |
| 3386 | * the caller may have made to the voting state */ |
| 3387 | osb->dc_wake_sequence++; |
| 3388 | spin_unlock(&osb->dc_task_lock); |
| 3389 | wake_up(&osb->dc_event); |
| 3390 | } |