fs/dlm/recoverd.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /******************************************************************************
   3 *******************************************************************************
   4 **
   5 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
   6 **  Copyright (C) 2004-2011 Red Hat, Inc.  All rights reserved.
   7 **
   8 **
   9 *******************************************************************************
  10 ******************************************************************************/
  11
  12 #include "dlm_internal.h"
  13 #include "lockspace.h"
  14 #include "member.h"
  15 #include "dir.h"
  16 #include "ast.h"
  17 #include "recover.h"
  18 #include "lowcomms.h"
  19 #include "lock.h"
  20 #include "requestqueue.h"
  21 #include "recoverd.h"
  22
  23
  24 /* If the start for which we're re-enabling locking (seq) has been superseded
  25    by a newer stop (ls_recover_seq), we need to leave locking disabled.
  26
  27    We suspend dlm_recv threads here to avoid the race where dlm_recv a) sees
  28    locking stopped and b) adds a message to the requestqueue, but dlm_recoverd
  29    enables locking and clears the requestqueue between a and b. */
  30
  31 static int enable_locking(struct dlm_ls *ls, uint64_t seq)
  32 {
  33         int error = -EINTR;
  34
  35         down_write(&ls->ls_recv_active);
  36
  37         spin_lock(&ls->ls_recover_lock);
  38         if (ls->ls_recover_seq == seq) {
  39                 set_bit(LSFL_RUNNING, &ls->ls_flags);
  40                 /* unblocks processes waiting to enter the dlm */
  41                 up_write(&ls->ls_in_recovery);
  42                 clear_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
  43                 error = 0;
  44         }
  45         spin_unlock(&ls->ls_recover_lock);
  46
  47         up_write(&ls->ls_recv_active);
  48         return error;
  49 }
  50
  51 static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
  52 {
  53         unsigned long start;
  54         int error, neg = 0;
  55
  56         log_rinfo(ls, "dlm_recover %llu", (unsigned long long)rv->seq);
  57
  58         mutex_lock(&ls->ls_recoverd_active);
  59
  60         dlm_callback_suspend(ls);
  61
  62         dlm_clear_toss(ls);
  63
  64         /*
  65          * This list of root rsb's will be the basis of most of the recovery
  66          * routines.
  67          */
  68
  69         dlm_create_root_list(ls);
  70
  71         /*
  72          * Add or remove nodes from the lockspace's ls_nodes list.
  73          */
  74
  75         error = dlm_recover_members(ls, rv, &neg);
  76         if (error) {
  77                 log_rinfo(ls, "dlm_recover_members error %d", error);
  78                 goto fail;
  79         }
  80
  81         dlm_recover_dir_nodeid(ls);
  82
  83         ls->ls_recover_dir_sent_res = 0;
  84         ls->ls_recover_dir_sent_msg = 0;
  85         ls->ls_recover_locks_in = 0;
  86
  87         dlm_set_recover_status(ls, DLM_RS_NODES);
  88
  89         error = dlm_recover_members_wait(ls);
  90         if (error) {
  91                 log_rinfo(ls, "dlm_recover_members_wait error %d", error);
  92                 goto fail;
  93         }
  94
  95         start = jiffies;
  96
  97         /*
  98          * Rebuild our own share of the directory by collecting from all other
  99          * nodes their master rsb names that hash to us.
 100          */
 101
 102         error = dlm_recover_directory(ls);
 103         if (error) {
 104                 log_rinfo(ls, "dlm_recover_directory error %d", error);
 105                 goto fail;
 106         }
 107
 108         dlm_set_recover_status(ls, DLM_RS_DIR);
 109
 110         error = dlm_recover_directory_wait(ls);
 111         if (error) {
 112                 log_rinfo(ls, "dlm_recover_directory_wait error %d", error);
 113                 goto fail;
 114         }
 115
 116         log_rinfo(ls, "dlm_recover_directory %u out %u messages",
 117                   ls->ls_recover_dir_sent_res, ls->ls_recover_dir_sent_msg);
 118
 119         /*
 120          * We may have outstanding operations that are waiting for a reply from
 121          * a failed node.  Mark these to be resent after recovery.  Unlock and
 122          * cancel ops can just be completed.
 123          */
 124
 125         dlm_recover_waiters_pre(ls);
 126
 127         error = dlm_recovery_stopped(ls);
 128         if (error) {
 129                 error = -EINTR;
 130                 goto fail;
 131         }
 132
 133         if (neg || dlm_no_directory(ls)) {
 134                 /*
 135                  * Clear lkb's for departed nodes.
 136                  */
 137
 138                 dlm_recover_purge(ls);
 139
 140                 /*
 141                  * Get new master nodeid's for rsb's that were mastered on
 142                  * departed nodes.
 143                  */
 144
 145                 error = dlm_recover_masters(ls);
 146                 if (error) {
 147                         log_rinfo(ls, "dlm_recover_masters error %d", error);
 148                         goto fail;
 149                 }
 150
 151                 /*
 152                  * Send our locks on remastered rsb's to the new masters.
 153                  */
 154
 155                 error = dlm_recover_locks(ls);
 156                 if (error) {
 157                         log_rinfo(ls, "dlm_recover_locks error %d", error);
 158                         goto fail;
 159                 }
 160
 161                 dlm_set_recover_status(ls, DLM_RS_LOCKS);
 162
 163                 error = dlm_recover_locks_wait(ls);
 164                 if (error) {
 165                         log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
 166                         goto fail;
 167                 }
 168
 169                 log_rinfo(ls, "dlm_recover_locks %u in",
 170                           ls->ls_recover_locks_in);
 171
 172                 /*
 173                  * Finalize state in master rsb's now that all locks can be
 174                  * checked.  This includes conversion resolution and lvb
 175                  * settings.
 176                  */
 177
 178                 dlm_recover_rsbs(ls);
 179         } else {
 180                 /*
 181                  * Other lockspace members may be going through the "neg" steps
 182                  * while also adding us to the lockspace, in which case they'll
 183                  * be doing the recover_locks (RS_LOCKS) barrier.
 184                  */
 185                 dlm_set_recover_status(ls, DLM_RS_LOCKS);
 186
 187                 error = dlm_recover_locks_wait(ls);
 188                 if (error) {
 189                         log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
 190                         goto fail;
 191                 }
 192         }
 193
 194         dlm_release_root_list(ls);
 195
 196         /*
 197          * Purge directory-related requests that are saved in requestqueue.
 198          * All dir requests from before recovery are invalid now due to the dir
 199          * rebuild and will be resent by the requesting nodes.
 200          */
 201
 202         dlm_purge_requestqueue(ls);
 203
 204         dlm_set_recover_status(ls, DLM_RS_DONE);
 205
 206         error = dlm_recover_done_wait(ls);
 207         if (error) {
 208                 log_rinfo(ls, "dlm_recover_done_wait error %d", error);
 209                 goto fail;
 210         }
 211
 212         dlm_clear_members_gone(ls);
 213
 214         dlm_adjust_timeouts(ls);
 215
 216         dlm_callback_resume(ls);
 217
 218         error = enable_locking(ls, rv->seq);
 219         if (error) {
 220                 log_rinfo(ls, "enable_locking error %d", error);
 221                 goto fail;
 222         }
 223
 224         error = dlm_process_requestqueue(ls);
 225         if (error) {
 226                 log_rinfo(ls, "dlm_process_requestqueue error %d", error);
 227                 goto fail;
 228         }
 229
 230         error = dlm_recover_waiters_post(ls);
 231         if (error) {
 232                 log_rinfo(ls, "dlm_recover_waiters_post error %d", error);
 233                 goto fail;
 234         }
 235
 236         dlm_recover_grant(ls);
 237
 238         log_rinfo(ls, "dlm_recover %llu generation %u done: %u ms",
 239                   (unsigned long long)rv->seq, ls->ls_generation,
 240                   jiffies_to_msecs(jiffies - start));
 241         mutex_unlock(&ls->ls_recoverd_active);
 242
 243         dlm_lsop_recover_done(ls);
 244         return 0;
 245
 246  fail:
 247         dlm_release_root_list(ls);
 248         log_rinfo(ls, "dlm_recover %llu error %d",
 249                   (unsigned long long)rv->seq, error);
 250         mutex_unlock(&ls->ls_recoverd_active);
 251         return error;
 252 }
 253
 254 /* The dlm_ls_start() that created the rv we take here may already have been
 255    stopped via dlm_ls_stop(); in that case we need to leave the RECOVERY_STOP
 256    flag set. */
 257
 258 static void do_ls_recovery(struct dlm_ls *ls)
 259 {
 260         struct dlm_recover *rv = NULL;
 261
 262         spin_lock(&ls->ls_recover_lock);
 263         rv = ls->ls_recover_args;
 264         ls->ls_recover_args = NULL;
 265         if (rv && ls->ls_recover_seq == rv->seq)
 266                 clear_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
 267         spin_unlock(&ls->ls_recover_lock);
 268
 269         if (rv) {
 270                 ls_recover(ls, rv);
 271                 kfree(rv->nodes);
 272                 kfree(rv);
 273         }
 274 }
 275
 276 static int dlm_recoverd(void *arg)
 277 {
 278         struct dlm_ls *ls;
 279
 280         ls = dlm_find_lockspace_local(arg);
 281         if (!ls) {
 282                 log_print("dlm_recoverd: no lockspace %p", arg);
 283                 return -1;
 284         }
 285
 286         down_write(&ls->ls_in_recovery);
 287         set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
 288         wake_up(&ls->ls_recover_lock_wait);
 289
 290         while (1) {
 291                 /*
 292                  * We call kthread_should_stop() after set_current_state().
 293                  * This is because it works correctly if kthread_stop() is
 294                  * called just before set_current_state().
 295                  */
 296                 set_current_state(TASK_INTERRUPTIBLE);
 297                 if (kthread_should_stop()) {
 298                         set_current_state(TASK_RUNNING);
 299                         break;
 300                 }
 301                 if (!test_bit(LSFL_RECOVER_WORK, &ls->ls_flags) &&
 302                     !test_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
 303                         if (kthread_should_stop())
 304                                 break;
 305                         schedule();
 306                 }
 307                 set_current_state(TASK_RUNNING);
 308
 309                 if (test_and_clear_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
 310                         down_write(&ls->ls_in_recovery);
 311                         set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
 312                         wake_up(&ls->ls_recover_lock_wait);
 313                 }
 314
 315                 if (test_and_clear_bit(LSFL_RECOVER_WORK, &ls->ls_flags))
 316                         do_ls_recovery(ls);
 317         }
 318
 319         if (test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags))
 320                 up_write(&ls->ls_in_recovery);
 321
 322         dlm_put_lockspace(ls);
 323         return 0;
 324 }
 325
 326 int dlm_recoverd_start(struct dlm_ls *ls)
 327 {
 328         struct task_struct *p;
 329         int error = 0;
 330
 331         p = kthread_run(dlm_recoverd, ls, "dlm_recoverd");
 332         if (IS_ERR(p))
 333                 error = PTR_ERR(p);
 334         else
 335                 ls->ls_recoverd_task = p;
 336         return error;
 337 }
 338
 339 void dlm_recoverd_stop(struct dlm_ls *ls)
 340 {
 341         kthread_stop(ls->ls_recoverd_task);
 342 }
 343
 344 void dlm_recoverd_suspend(struct dlm_ls *ls)
 345 {
 346         wake_up(&ls->ls_wait_general);
 347         mutex_lock(&ls->ls_recoverd_active);
 348 }
 349
 350 void dlm_recoverd_resume(struct dlm_ls *ls)
 351 {
 352         mutex_unlock(&ls->ls_recoverd_active);
 353 }
 354