linux/fs/gfs2/lock_dlm.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
   4 * Copyright 2004-2011 Red Hat, Inc.
   5 */
   6
   7#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   8
   9#include <linux/fs.h>
  10#include <linux/dlm.h>
  11#include <linux/slab.h>
  12#include <linux/types.h>
  13#include <linux/delay.h>
  14#include <linux/gfs2_ondisk.h>
  15#include <linux/sched/signal.h>
  16
  17#include "incore.h"
  18#include "glock.h"
  19#include "util.h"
  20#include "sys.h"
  21#include "trace_gfs2.h"
  22
  23/**
  24 * gfs2_update_stats - Update time based stats
  25 * @mv: Pointer to mean/variance structure to update
  26 * @sample: New data to include
  27 *
  28 * @delta is the difference between the current rtt sample and the
  29 * running average srtt. We add 1/8 of that to the srtt in order to
  30 * update the current srtt estimate. The variance estimate is a bit
  31 * more complicated. We subtract the current variance estimate from
  32 * the abs value of the @delta and add 1/4 of that to the running
  33 * total.  That's equivalent to 3/4 of the current variance
  34 * estimate plus 1/4 of the abs of @delta.
  35 *
  36 * Note that the index points at the array entry containing the smoothed
  37 * mean value, and the variance is always in the following entry
  38 *
  39 * Reference: TCP/IP Illustrated, vol 2, p. 831,832
  40 * All times are in units of integer nanoseconds. Unlike the TCP/IP case,
  41 * they are not scaled fixed point.
  42 */
  43
  44static inline void gfs2_update_stats(struct gfs2_lkstats *s, unsigned index,
  45                                     s64 sample)
  46{
  47        s64 delta = sample - s->stats[index];
  48        s->stats[index] += (delta >> 3);
  49        index++;
  50        s->stats[index] += (s64)(abs(delta) - s->stats[index]) >> 2;
  51}
  52
  53/**
  54 * gfs2_update_reply_times - Update locking statistics
  55 * @gl: The glock to update
  56 *
  57 * This assumes that gl->gl_dstamp has been set earlier.
  58 *
  59 * The rtt (lock round trip time) is an estimate of the time
  60 * taken to perform a dlm lock request. We update it on each
  61 * reply from the dlm.
  62 *
  63 * The blocking flag is set on the glock for all dlm requests
  64 * which may potentially block due to lock requests from other nodes.
  65 * DLM requests where the current lock state is exclusive, the
  66 * requested state is null (or unlocked) or where the TRY or
  67 * TRY_1CB flags are set are classified as non-blocking. All
  68 * other DLM requests are counted as (potentially) blocking.
  69 */
  70static inline void gfs2_update_reply_times(struct gfs2_glock *gl)
  71{
  72        struct gfs2_pcpu_lkstats *lks;
  73        const unsigned gltype = gl->gl_name.ln_type;
  74        unsigned index = test_bit(GLF_BLOCKING, &gl->gl_flags) ?
  75                         GFS2_LKS_SRTTB : GFS2_LKS_SRTT;
  76        s64 rtt;
  77
  78        preempt_disable();
  79        rtt = ktime_to_ns(ktime_sub(ktime_get_real(), gl->gl_dstamp));
  80        lks = this_cpu_ptr(gl->gl_name.ln_sbd->sd_lkstats);
  81        gfs2_update_stats(&gl->gl_stats, index, rtt);           /* Local */
  82        gfs2_update_stats(&lks->lkstats[gltype], index, rtt);   /* Global */
  83        preempt_enable();
  84
  85        trace_gfs2_glock_lock_time(gl, rtt);
  86}
  87
  88/**
  89 * gfs2_update_request_times - Update locking statistics
  90 * @gl: The glock to update
  91 *
  92 * The irt (lock inter-request times) measures the average time
  93 * between requests to the dlm. It is updated immediately before
  94 * each dlm call.
  95 */
  96
  97static inline void gfs2_update_request_times(struct gfs2_glock *gl)
  98{
  99        struct gfs2_pcpu_lkstats *lks;
 100        const unsigned gltype = gl->gl_name.ln_type;
 101        ktime_t dstamp;
 102        s64 irt;
 103
 104        preempt_disable();
 105        dstamp = gl->gl_dstamp;
 106        gl->gl_dstamp = ktime_get_real();
 107        irt = ktime_to_ns(ktime_sub(gl->gl_dstamp, dstamp));
 108        lks = this_cpu_ptr(gl->gl_name.ln_sbd->sd_lkstats);
 109        gfs2_update_stats(&gl->gl_stats, GFS2_LKS_SIRT, irt);           /* Local */
 110        gfs2_update_stats(&lks->lkstats[gltype], GFS2_LKS_SIRT, irt);   /* Global */
 111        preempt_enable();
 112}
 113 
 114static void gdlm_ast(void *arg)
 115{
 116        struct gfs2_glock *gl = arg;
 117        unsigned ret = gl->gl_state;
 118
 119        gfs2_update_reply_times(gl);
 120        BUG_ON(gl->gl_lksb.sb_flags & DLM_SBF_DEMOTED);
 121
 122        if ((gl->gl_lksb.sb_flags & DLM_SBF_VALNOTVALID) && gl->gl_lksb.sb_lvbptr)
 123                memset(gl->gl_lksb.sb_lvbptr, 0, GDLM_LVB_SIZE);
 124
 125        switch (gl->gl_lksb.sb_status) {
 126        case -DLM_EUNLOCK: /* Unlocked, so glock can be freed */
 127                gfs2_glock_free(gl);
 128                return;
 129        case -DLM_ECANCEL: /* Cancel while getting lock */
 130                ret |= LM_OUT_CANCELED;
 131                goto out;
 132        case -EAGAIN: /* Try lock fails */
 133        case -EDEADLK: /* Deadlock detected */
 134                goto out;
 135        case -ETIMEDOUT: /* Canceled due to timeout */
 136                ret |= LM_OUT_ERROR;
 137                goto out;
 138        case 0: /* Success */
 139                break;
 140        default: /* Something unexpected */
 141                BUG();
 142        }
 143
 144        ret = gl->gl_req;
 145        if (gl->gl_lksb.sb_flags & DLM_SBF_ALTMODE) {
 146                if (gl->gl_req == LM_ST_SHARED)
 147                        ret = LM_ST_DEFERRED;
 148                else if (gl->gl_req == LM_ST_DEFERRED)
 149                        ret = LM_ST_SHARED;
 150                else
 151                        BUG();
 152        }
 153
 154        set_bit(GLF_INITIAL, &gl->gl_flags);
 155        gfs2_glock_complete(gl, ret);
 156        return;
 157out:
 158        if (!test_bit(GLF_INITIAL, &gl->gl_flags))
 159                gl->gl_lksb.sb_lkid = 0;
 160        gfs2_glock_complete(gl, ret);
 161}
 162
 163static void gdlm_bast(void *arg, int mode)
 164{
 165        struct gfs2_glock *gl = arg;
 166
 167        switch (mode) {
 168        case DLM_LOCK_EX:
 169                gfs2_glock_cb(gl, LM_ST_UNLOCKED);
 170                break;
 171        case DLM_LOCK_CW:
 172                gfs2_glock_cb(gl, LM_ST_DEFERRED);
 173                break;
 174        case DLM_LOCK_PR:
 175                gfs2_glock_cb(gl, LM_ST_SHARED);
 176                break;
 177        default:
 178                fs_err(gl->gl_name.ln_sbd, "unknown bast mode %d\n", mode);
 179                BUG();
 180        }
 181}
 182
 183/* convert gfs lock-state to dlm lock-mode */
 184
 185static int make_mode(struct gfs2_sbd *sdp, const unsigned int lmstate)
 186{
 187        switch (lmstate) {
 188        case LM_ST_UNLOCKED:
 189                return DLM_LOCK_NL;
 190        case LM_ST_EXCLUSIVE:
 191                return DLM_LOCK_EX;
 192        case LM_ST_DEFERRED:
 193                return DLM_LOCK_CW;
 194        case LM_ST_SHARED:
 195                return DLM_LOCK_PR;
 196        }
 197        fs_err(sdp, "unknown LM state %d\n", lmstate);
 198        BUG();
 199        return -1;
 200}
 201
 202static u32 make_flags(struct gfs2_glock *gl, const unsigned int gfs_flags,
 203                      const int req)
 204{
 205        u32 lkf = 0;
 206
 207        if (gl->gl_lksb.sb_lvbptr)
 208                lkf |= DLM_LKF_VALBLK;
 209
 210        if (gfs_flags & LM_FLAG_TRY)
 211                lkf |= DLM_LKF_NOQUEUE;
 212
 213        if (gfs_flags & LM_FLAG_TRY_1CB) {
 214                lkf |= DLM_LKF_NOQUEUE;
 215                lkf |= DLM_LKF_NOQUEUEBAST;
 216        }
 217
 218        if (gfs_flags & LM_FLAG_PRIORITY) {
 219                lkf |= DLM_LKF_NOORDER;
 220                lkf |= DLM_LKF_HEADQUE;
 221        }
 222
 223        if (gfs_flags & LM_FLAG_ANY) {
 224                if (req == DLM_LOCK_PR)
 225                        lkf |= DLM_LKF_ALTCW;
 226                else if (req == DLM_LOCK_CW)
 227                        lkf |= DLM_LKF_ALTPR;
 228                else
 229                        BUG();
 230        }
 231
 232        if (gl->gl_lksb.sb_lkid != 0) {
 233                lkf |= DLM_LKF_CONVERT;
 234                if (test_bit(GLF_BLOCKING, &gl->gl_flags))
 235                        lkf |= DLM_LKF_QUECVT;
 236        }
 237
 238        return lkf;
 239}
 240
 241static void gfs2_reverse_hex(char *c, u64 value)
 242{
 243        *c = '0';
 244        while (value) {
 245                *c-- = hex_asc[value & 0x0f];
 246                value >>= 4;
 247        }
 248}
 249
 250static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state,
 251                     unsigned int flags)
 252{
 253        struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct;
 254        int req;
 255        u32 lkf;
 256        char strname[GDLM_STRNAME_BYTES] = "";
 257
 258        req = make_mode(gl->gl_name.ln_sbd, req_state);
 259        lkf = make_flags(gl, flags, req);
 260        gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT);
 261        gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT);
 262        if (gl->gl_lksb.sb_lkid) {
 263                gfs2_update_request_times(gl);
 264        } else {
 265                memset(strname, ' ', GDLM_STRNAME_BYTES - 1);
 266                strname[GDLM_STRNAME_BYTES - 1] = '\0';
 267                gfs2_reverse_hex(strname + 7, gl->gl_name.ln_type);
 268                gfs2_reverse_hex(strname + 23, gl->gl_name.ln_number);
 269                gl->gl_dstamp = ktime_get_real();
 270        }
 271        /*
 272         * Submit the actual lock request.
 273         */
 274
 275        return dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, strname,
 276                        GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast);
 277}
 278
 279static void gdlm_put_lock(struct gfs2_glock *gl)
 280{
 281        struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 282        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 283        int lvb_needs_unlock = 0;
 284        int error;
 285
 286        if (gl->gl_lksb.sb_lkid == 0) {
 287                gfs2_glock_free(gl);
 288                return;
 289        }
 290
 291        clear_bit(GLF_BLOCKING, &gl->gl_flags);
 292        gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT);
 293        gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT);
 294        gfs2_update_request_times(gl);
 295
 296        /* don't want to skip dlm_unlock writing the lvb when lock is ex */
 297
 298        if (gl->gl_lksb.sb_lvbptr && (gl->gl_state == LM_ST_EXCLUSIVE))
 299                lvb_needs_unlock = 1;
 300
 301        if (test_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags) &&
 302            !lvb_needs_unlock) {
 303                gfs2_glock_free(gl);
 304                return;
 305        }
 306
 307        error = dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_VALBLK,
 308                           NULL, gl);
 309        if (error) {
 310                fs_err(sdp, "gdlm_unlock %x,%llx err=%d\n",
 311                       gl->gl_name.ln_type,
 312                       (unsigned long long)gl->gl_name.ln_number, error);
 313                return;
 314        }
 315}
 316
 317static void gdlm_cancel(struct gfs2_glock *gl)
 318{
 319        struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct;
 320        dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_CANCEL, NULL, gl);
 321}
 322
 323/*
 324 * dlm/gfs2 recovery coordination using dlm_recover callbacks
 325 *
 326 *  1. dlm_controld sees lockspace members change
 327 *  2. dlm_controld blocks dlm-kernel locking activity
 328 *  3. dlm_controld within dlm-kernel notifies gfs2 (recover_prep)
 329 *  4. dlm_controld starts and finishes its own user level recovery
 330 *  5. dlm_controld starts dlm-kernel dlm_recoverd to do kernel recovery
 331 *  6. dlm_recoverd notifies gfs2 of failed nodes (recover_slot)
 332 *  7. dlm_recoverd does its own lock recovery
 333 *  8. dlm_recoverd unblocks dlm-kernel locking activity
 334 *  9. dlm_recoverd notifies gfs2 when done (recover_done with new generation)
 335 * 10. gfs2_control updates control_lock lvb with new generation and jid bits
 336 * 11. gfs2_control enqueues journals for gfs2_recover to recover (maybe none)
 337 * 12. gfs2_recover dequeues and recovers journals of failed nodes
 338 * 13. gfs2_recover provides recovery results to gfs2_control (recovery_result)
 339 * 14. gfs2_control updates control_lock lvb jid bits for recovered journals
 340 * 15. gfs2_control unblocks normal locking when all journals are recovered
 341 *
 342 * - failures during recovery
 343 *
 344 * recover_prep() may set BLOCK_LOCKS (step 3) again before gfs2_control
 345 * clears BLOCK_LOCKS (step 15), e.g. another node fails while still
 346 * recovering for a prior failure.  gfs2_control needs a way to detect
 347 * this so it can leave BLOCK_LOCKS set in step 15.  This is managed using
 348 * the recover_block and recover_start values.
 349 *
 350 * recover_done() provides a new lockspace generation number each time it
 351 * is called (step 9).  This generation number is saved as recover_start.
 352 * When recover_prep() is called, it sets BLOCK_LOCKS and sets
 353 * recover_block = recover_start.  So, while recover_block is equal to
 354 * recover_start, BLOCK_LOCKS should remain set.  (recover_spin must
 355 * be held around the BLOCK_LOCKS/recover_block/recover_start logic.)
 356 *
 357 * - more specific gfs2 steps in sequence above
 358 *
 359 *  3. recover_prep sets BLOCK_LOCKS and sets recover_block = recover_start
 360 *  6. recover_slot records any failed jids (maybe none)
 361 *  9. recover_done sets recover_start = new generation number
 362 * 10. gfs2_control sets control_lock lvb = new gen + bits for failed jids
 363 * 12. gfs2_recover does journal recoveries for failed jids identified above
 364 * 14. gfs2_control clears control_lock lvb bits for recovered jids
 365 * 15. gfs2_control checks if recover_block == recover_start (step 3 occured
 366 *     again) then do nothing, otherwise if recover_start > recover_block
 367 *     then clear BLOCK_LOCKS.
 368 *
 369 * - parallel recovery steps across all nodes
 370 *
 371 * All nodes attempt to update the control_lock lvb with the new generation
 372 * number and jid bits, but only the first to get the control_lock EX will
 373 * do so; others will see that it's already done (lvb already contains new
 374 * generation number.)
 375 *
 376 * . All nodes get the same recover_prep/recover_slot/recover_done callbacks
 377 * . All nodes attempt to set control_lock lvb gen + bits for the new gen
 378 * . One node gets control_lock first and writes the lvb, others see it's done
 379 * . All nodes attempt to recover jids for which they see control_lock bits set
 380 * . One node succeeds for a jid, and that one clears the jid bit in the lvb
 381 * . All nodes will eventually see all lvb bits clear and unblock locks
 382 *
 383 * - is there a problem with clearing an lvb bit that should be set
 384 *   and missing a journal recovery?
 385 *
 386 * 1. jid fails
 387 * 2. lvb bit set for step 1
 388 * 3. jid recovered for step 1
 389 * 4. jid taken again (new mount)
 390 * 5. jid fails (for step 4)
 391 * 6. lvb bit set for step 5 (will already be set)
 392 * 7. lvb bit cleared for step 3
 393 *
 394 * This is not a problem because the failure in step 5 does not
 395 * require recovery, because the mount in step 4 could not have
 396 * progressed far enough to unblock locks and access the fs.  The
 397 * control_mount() function waits for all recoveries to be complete
 398 * for the latest lockspace generation before ever unblocking locks
 399 * and returning.  The mount in step 4 waits until the recovery in
 400 * step 1 is done.
 401 *
 402 * - special case of first mounter: first node to mount the fs
 403 *
 404 * The first node to mount a gfs2 fs needs to check all the journals
 405 * and recover any that need recovery before other nodes are allowed
 406 * to mount the fs.  (Others may begin mounting, but they must wait
 407 * for the first mounter to be done before taking locks on the fs
 408 * or accessing the fs.)  This has two parts:
 409 *
 410 * 1. The mounted_lock tells a node it's the first to mount the fs.
 411 * Each node holds the mounted_lock in PR while it's mounted.
 412 * Each node tries to acquire the mounted_lock in EX when it mounts.
 413 * If a node is granted the mounted_lock EX it means there are no
 414 * other mounted nodes (no PR locks exist), and it is the first mounter.
 415 * The mounted_lock is demoted to PR when first recovery is done, so
 416 * others will fail to get an EX lock, but will get a PR lock.
 417 *
 418 * 2. The control_lock blocks others in control_mount() while the first
 419 * mounter is doing first mount recovery of all journals.
 420 * A mounting node needs to acquire control_lock in EX mode before
 421 * it can proceed.  The first mounter holds control_lock in EX while doing
 422 * the first mount recovery, blocking mounts from other nodes, then demotes
 423 * control_lock to NL when it's done (others_may_mount/first_done),
 424 * allowing other nodes to continue mounting.
 425 *
 426 * first mounter:
 427 * control_lock EX/NOQUEUE success
 428 * mounted_lock EX/NOQUEUE success (no other PR, so no other mounters)
 429 * set first=1
 430 * do first mounter recovery
 431 * mounted_lock EX->PR
 432 * control_lock EX->NL, write lvb generation
 433 *
 434 * other mounter:
 435 * control_lock EX/NOQUEUE success (if fail -EAGAIN, retry)
 436 * mounted_lock EX/NOQUEUE fail -EAGAIN (expected due to other mounters PR)
 437 * mounted_lock PR/NOQUEUE success
 438 * read lvb generation
 439 * control_lock EX->NL
 440 * set first=0
 441 *
 442 * - mount during recovery
 443 *
 444 * If a node mounts while others are doing recovery (not first mounter),
 445 * the mounting node will get its initial recover_done() callback without
 446 * having seen any previous failures/callbacks.
 447 *
 448 * It must wait for all recoveries preceding its mount to be finished
 449 * before it unblocks locks.  It does this by repeating the "other mounter"
 450 * steps above until the lvb generation number is >= its mount generation
 451 * number (from initial recover_done) and all lvb bits are clear.
 452 *
 453 * - control_lock lvb format
 454 *
 455 * 4 bytes generation number: the latest dlm lockspace generation number
 456 * from recover_done callback.  Indicates the jid bitmap has been updated
 457 * to reflect all slot failures through that generation.
 458 * 4 bytes unused.
 459 * GDLM_LVB_SIZE-8 bytes of jid bit map. If bit N is set, it indicates
 460 * that jid N needs recovery.
 461 */
 462
 463#define JID_BITMAP_OFFSET 8 /* 4 byte generation number + 4 byte unused */
 464
 465static void control_lvb_read(struct lm_lockstruct *ls, uint32_t *lvb_gen,
 466                             char *lvb_bits)
 467{
 468        __le32 gen;
 469        memcpy(lvb_bits, ls->ls_control_lvb, GDLM_LVB_SIZE);
 470        memcpy(&gen, lvb_bits, sizeof(__le32));
 471        *lvb_gen = le32_to_cpu(gen);
 472}
 473
 474static void control_lvb_write(struct lm_lockstruct *ls, uint32_t lvb_gen,
 475                              char *lvb_bits)
 476{
 477        __le32 gen;
 478        memcpy(ls->ls_control_lvb, lvb_bits, GDLM_LVB_SIZE);
 479        gen = cpu_to_le32(lvb_gen);
 480        memcpy(ls->ls_control_lvb, &gen, sizeof(__le32));
 481}
 482
 483static int all_jid_bits_clear(char *lvb)
 484{
 485        return !memchr_inv(lvb + JID_BITMAP_OFFSET, 0,
 486                        GDLM_LVB_SIZE - JID_BITMAP_OFFSET);
 487}
 488
 489static void sync_wait_cb(void *arg)
 490{
 491        struct lm_lockstruct *ls = arg;
 492        complete(&ls->ls_sync_wait);
 493}
 494
 495static int sync_unlock(struct gfs2_sbd *sdp, struct dlm_lksb *lksb, char *name)
 496{
 497        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 498        int error;
 499
 500        error = dlm_unlock(ls->ls_dlm, lksb->sb_lkid, 0, lksb, ls);
 501        if (error) {
 502                fs_err(sdp, "%s lkid %x error %d\n",
 503                       name, lksb->sb_lkid, error);
 504                return error;
 505        }
 506
 507        wait_for_completion(&ls->ls_sync_wait);
 508
 509        if (lksb->sb_status != -DLM_EUNLOCK) {
 510                fs_err(sdp, "%s lkid %x status %d\n",
 511                       name, lksb->sb_lkid, lksb->sb_status);
 512                return -1;
 513        }
 514        return 0;
 515}
 516
 517static int sync_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags,
 518                     unsigned int num, struct dlm_lksb *lksb, char *name)
 519{
 520        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 521        char strname[GDLM_STRNAME_BYTES];
 522        int error, status;
 523
 524        memset(strname, 0, GDLM_STRNAME_BYTES);
 525        snprintf(strname, GDLM_STRNAME_BYTES, "%8x%16x", LM_TYPE_NONDISK, num);
 526
 527        error = dlm_lock(ls->ls_dlm, mode, lksb, flags,
 528                         strname, GDLM_STRNAME_BYTES - 1,
 529                         0, sync_wait_cb, ls, NULL);
 530        if (error) {
 531                fs_err(sdp, "%s lkid %x flags %x mode %d error %d\n",
 532                       name, lksb->sb_lkid, flags, mode, error);
 533                return error;
 534        }
 535
 536        wait_for_completion(&ls->ls_sync_wait);
 537
 538        status = lksb->sb_status;
 539
 540        if (status && status != -EAGAIN) {
 541                fs_err(sdp, "%s lkid %x flags %x mode %d status %d\n",
 542                       name, lksb->sb_lkid, flags, mode, status);
 543        }
 544
 545        return status;
 546}
 547
 548static int mounted_unlock(struct gfs2_sbd *sdp)
 549{
 550        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 551        return sync_unlock(sdp, &ls->ls_mounted_lksb, "mounted_lock");
 552}
 553
 554static int mounted_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags)
 555{
 556        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 557        return sync_lock(sdp, mode, flags, GFS2_MOUNTED_LOCK,
 558                         &ls->ls_mounted_lksb, "mounted_lock");
 559}
 560
 561static int control_unlock(struct gfs2_sbd *sdp)
 562{
 563        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 564        return sync_unlock(sdp, &ls->ls_control_lksb, "control_lock");
 565}
 566
 567static int control_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags)
 568{
 569        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 570        return sync_lock(sdp, mode, flags, GFS2_CONTROL_LOCK,
 571                         &ls->ls_control_lksb, "control_lock");
 572}
 573
 574static void gfs2_control_func(struct work_struct *work)
 575{
 576        struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_control_work.work);
 577        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 578        uint32_t block_gen, start_gen, lvb_gen, flags;
 579        int recover_set = 0;
 580        int write_lvb = 0;
 581        int recover_size;
 582        int i, error;
 583
 584        spin_lock(&ls->ls_recover_spin);
 585        /*
 586         * No MOUNT_DONE means we're still mounting; control_mount()
 587         * will set this flag, after which this thread will take over
 588         * all further clearing of BLOCK_LOCKS.
 589         *
 590         * FIRST_MOUNT means this node is doing first mounter recovery,
 591         * for which recovery control is handled by
 592         * control_mount()/control_first_done(), not this thread.
 593         */
 594        if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
 595             test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
 596                spin_unlock(&ls->ls_recover_spin);
 597                return;
 598        }
 599        block_gen = ls->ls_recover_block;
 600        start_gen = ls->ls_recover_start;
 601        spin_unlock(&ls->ls_recover_spin);
 602
 603        /*
 604         * Equal block_gen and start_gen implies we are between
 605         * recover_prep and recover_done callbacks, which means
 606         * dlm recovery is in progress and dlm locking is blocked.
 607         * There's no point trying to do any work until recover_done.
 608         */
 609
 610        if (block_gen == start_gen)
 611                return;
 612
 613        /*
 614         * Propagate recover_submit[] and recover_result[] to lvb:
 615         * dlm_recoverd adds to recover_submit[] jids needing recovery
 616         * gfs2_recover adds to recover_result[] journal recovery results
 617         *
 618         * set lvb bit for jids in recover_submit[] if the lvb has not
 619         * yet been updated for the generation of the failure
 620         *
 621         * clear lvb bit for jids in recover_result[] if the result of
 622         * the journal recovery is SUCCESS
 623         */
 624
 625        error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
 626        if (error) {
 627                fs_err(sdp, "control lock EX error %d\n", error);
 628                return;
 629        }
 630
 631        control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits);
 632
 633        spin_lock(&ls->ls_recover_spin);
 634        if (block_gen != ls->ls_recover_block ||
 635            start_gen != ls->ls_recover_start) {
 636                fs_info(sdp, "recover generation %u block1 %u %u\n",
 637                        start_gen, block_gen, ls->ls_recover_block);
 638                spin_unlock(&ls->ls_recover_spin);
 639                control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
 640                return;
 641        }
 642
 643        recover_size = ls->ls_recover_size;
 644
 645        if (lvb_gen <= start_gen) {
 646                /*
 647                 * Clear lvb bits for jids we've successfully recovered.
 648                 * Because all nodes attempt to recover failed journals,
 649                 * a journal can be recovered multiple times successfully
 650                 * in succession.  Only the first will really do recovery,
 651                 * the others find it clean, but still report a successful
 652                 * recovery.  So, another node may have already recovered
 653                 * the jid and cleared the lvb bit for it.
 654                 */
 655                for (i = 0; i < recover_size; i++) {
 656                        if (ls->ls_recover_result[i] != LM_RD_SUCCESS)
 657                                continue;
 658
 659                        ls->ls_recover_result[i] = 0;
 660
 661                        if (!test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET))
 662                                continue;
 663
 664                        __clear_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET);
 665                        write_lvb = 1;
 666                }
 667        }
 668
 669        if (lvb_gen == start_gen) {
 670                /*
 671                 * Failed slots before start_gen are already set in lvb.
 672                 */
 673                for (i = 0; i < recover_size; i++) {
 674                        if (!ls->ls_recover_submit[i])
 675                                continue;
 676                        if (ls->ls_recover_submit[i] < lvb_gen)
 677                                ls->ls_recover_submit[i] = 0;
 678                }
 679        } else if (lvb_gen < start_gen) {
 680                /*
 681                 * Failed slots before start_gen are not yet set in lvb.
 682                 */
 683                for (i = 0; i < recover_size; i++) {
 684                        if (!ls->ls_recover_submit[i])
 685                                continue;
 686                        if (ls->ls_recover_submit[i] < start_gen) {
 687                                ls->ls_recover_submit[i] = 0;
 688                                __set_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET);
 689                        }
 690                }
 691                /* even if there are no bits to set, we need to write the
 692                   latest generation to the lvb */
 693                write_lvb = 1;
 694        } else {
 695                /*
 696                 * we should be getting a recover_done() for lvb_gen soon
 697                 */
 698        }
 699        spin_unlock(&ls->ls_recover_spin);
 700
 701        if (write_lvb) {
 702                control_lvb_write(ls, start_gen, ls->ls_lvb_bits);
 703                flags = DLM_LKF_CONVERT | DLM_LKF_VALBLK;
 704        } else {
 705                flags = DLM_LKF_CONVERT;
 706        }
 707
 708        error = control_lock(sdp, DLM_LOCK_NL, flags);
 709        if (error) {
 710                fs_err(sdp, "control lock NL error %d\n", error);
 711                return;
 712        }
 713
 714        /*
 715         * Everyone will see jid bits set in the lvb, run gfs2_recover_set(),
 716         * and clear a jid bit in the lvb if the recovery is a success.
 717         * Eventually all journals will be recovered, all jid bits will
 718         * be cleared in the lvb, and everyone will clear BLOCK_LOCKS.
 719         */
 720
 721        for (i = 0; i < recover_size; i++) {
 722                if (test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET)) {
 723                        fs_info(sdp, "recover generation %u jid %d\n",
 724                                start_gen, i);
 725                        gfs2_recover_set(sdp, i);
 726                        recover_set++;
 727                }
 728        }
 729        if (recover_set)
 730                return;
 731
 732        /*
 733         * No more jid bits set in lvb, all recovery is done, unblock locks
 734         * (unless a new recover_prep callback has occured blocking locks
 735         * again while working above)
 736         */
 737
 738        spin_lock(&ls->ls_recover_spin);
 739        if (ls->ls_recover_block == block_gen &&
 740            ls->ls_recover_start == start_gen) {
 741                clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
 742                spin_unlock(&ls->ls_recover_spin);
 743                fs_info(sdp, "recover generation %u done\n", start_gen);
 744                gfs2_glock_thaw(sdp);
 745        } else {
 746                fs_info(sdp, "recover generation %u block2 %u %u\n",
 747                        start_gen, block_gen, ls->ls_recover_block);
 748                spin_unlock(&ls->ls_recover_spin);
 749        }
 750}
 751
 752static int control_mount(struct gfs2_sbd *sdp)
 753{
 754        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 755        uint32_t start_gen, block_gen, mount_gen, lvb_gen;
 756        int mounted_mode;
 757        int retries = 0;
 758        int error;
 759
 760        memset(&ls->ls_mounted_lksb, 0, sizeof(struct dlm_lksb));
 761        memset(&ls->ls_control_lksb, 0, sizeof(struct dlm_lksb));
 762        memset(&ls->ls_control_lvb, 0, GDLM_LVB_SIZE);
 763        ls->ls_control_lksb.sb_lvbptr = ls->ls_control_lvb;
 764        init_completion(&ls->ls_sync_wait);
 765
 766        set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
 767
 768        error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_VALBLK);
 769        if (error) {
 770                fs_err(sdp, "control_mount control_lock NL error %d\n", error);
 771                return error;
 772        }
 773
 774        error = mounted_lock(sdp, DLM_LOCK_NL, 0);
 775        if (error) {
 776                fs_err(sdp, "control_mount mounted_lock NL error %d\n", error);
 777                control_unlock(sdp);
 778                return error;
 779        }
 780        mounted_mode = DLM_LOCK_NL;
 781
 782restart:
 783        if (retries++ && signal_pending(current)) {
 784                error = -EINTR;
 785                goto fail;
 786        }
 787
 788        /*
 789         * We always start with both locks in NL. control_lock is
 790         * demoted to NL below so we don't need to do it here.
 791         */
 792
 793        if (mounted_mode != DLM_LOCK_NL) {
 794                error = mounted_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
 795                if (error)
 796                        goto fail;
 797                mounted_mode = DLM_LOCK_NL;
 798        }
 799
 800        /*
 801         * Other nodes need to do some work in dlm recovery and gfs2_control
 802         * before the recover_done and control_lock will be ready for us below.
 803         * A delay here is not required but often avoids having to retry.
 804         */
 805
 806        msleep_interruptible(500);
 807
 808        /*
 809         * Acquire control_lock in EX and mounted_lock in either EX or PR.
 810         * control_lock lvb keeps track of any pending journal recoveries.
 811         * mounted_lock indicates if any other nodes have the fs mounted.
 812         */
 813
 814        error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE|DLM_LKF_VALBLK);
 815        if (error == -EAGAIN) {
 816                goto restart;
 817        } else if (error) {
 818                fs_err(sdp, "control_mount control_lock EX error %d\n", error);
 819                goto fail;
 820        }
 821
 822        /**
 823         * If we're a spectator, we don't want to take the lock in EX because
 824         * we cannot do the first-mount responsibility it implies: recovery.
 825         */
 826        if (sdp->sd_args.ar_spectator)
 827                goto locks_done;
 828
 829        error = mounted_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
 830        if (!error) {
 831                mounted_mode = DLM_LOCK_EX;
 832                goto locks_done;
 833        } else if (error != -EAGAIN) {
 834                fs_err(sdp, "control_mount mounted_lock EX error %d\n", error);
 835                goto fail;
 836        }
 837
 838        error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
 839        if (!error) {
 840                mounted_mode = DLM_LOCK_PR;
 841                goto locks_done;
 842        } else {
 843                /* not even -EAGAIN should happen here */
 844                fs_err(sdp, "control_mount mounted_lock PR error %d\n", error);
 845                goto fail;
 846        }
 847
 848locks_done:
 849        /*
 850         * If we got both locks above in EX, then we're the first mounter.
 851         * If not, then we need to wait for the control_lock lvb to be
 852         * updated by other mounted nodes to reflect our mount generation.
 853         *
 854         * In simple first mounter cases, first mounter will see zero lvb_gen,
 855         * but in cases where all existing nodes leave/fail before mounting
 856         * nodes finish control_mount, then all nodes will be mounting and
 857         * lvb_gen will be non-zero.
 858         */
 859
 860        control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits);
 861
 862        if (lvb_gen == 0xFFFFFFFF) {
 863                /* special value to force mount attempts to fail */
 864                fs_err(sdp, "control_mount control_lock disabled\n");
 865                error = -EINVAL;
 866                goto fail;
 867        }
 868
 869        if (mounted_mode == DLM_LOCK_EX) {
 870                /* first mounter, keep both EX while doing first recovery */
 871                spin_lock(&ls->ls_recover_spin);
 872                clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
 873                set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
 874                set_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
 875                spin_unlock(&ls->ls_recover_spin);
 876                fs_info(sdp, "first mounter control generation %u\n", lvb_gen);
 877                return 0;
 878        }
 879
 880        error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
 881        if (error)
 882                goto fail;
 883
 884        /*
 885         * We are not first mounter, now we need to wait for the control_lock
 886         * lvb generation to be >= the generation from our first recover_done
 887         * and all lvb bits to be clear (no pending journal recoveries.)
 888         */
 889
 890        if (!all_jid_bits_clear(ls->ls_lvb_bits)) {
 891                /* journals need recovery, wait until all are clear */
 892                fs_info(sdp, "control_mount wait for journal recovery\n");
 893                goto restart;
 894        }
 895
 896        spin_lock(&ls->ls_recover_spin);
 897        block_gen = ls->ls_recover_block;
 898        start_gen = ls->ls_recover_start;
 899        mount_gen = ls->ls_recover_mount;
 900
 901        if (lvb_gen < mount_gen) {
 902                /* wait for mounted nodes to update control_lock lvb to our
 903                   generation, which might include new recovery bits set */
 904                if (sdp->sd_args.ar_spectator) {
 905                        fs_info(sdp, "Recovery is required. Waiting for a "
 906                                "non-spectator to mount.\n");
 907                        msleep_interruptible(1000);
 908                } else {
 909                        fs_info(sdp, "control_mount wait1 block %u start %u "
 910                                "mount %u lvb %u flags %lx\n", block_gen,
 911                                start_gen, mount_gen, lvb_gen,
 912                                ls->ls_recover_flags);
 913                }
 914                spin_unlock(&ls->ls_recover_spin);
 915                goto restart;
 916        }
 917
 918        if (lvb_gen != start_gen) {
 919                /* wait for mounted nodes to update control_lock lvb to the
 920                   latest recovery generation */
 921                fs_info(sdp, "control_mount wait2 block %u start %u mount %u "
 922                        "lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
 923                        lvb_gen, ls->ls_recover_flags);
 924                spin_unlock(&ls->ls_recover_spin);
 925                goto restart;
 926        }
 927
 928        if (block_gen == start_gen) {
 929                /* dlm recovery in progress, wait for it to finish */
 930                fs_info(sdp, "control_mount wait3 block %u start %u mount %u "
 931                        "lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
 932                        lvb_gen, ls->ls_recover_flags);
 933                spin_unlock(&ls->ls_recover_spin);
 934                goto restart;
 935        }
 936
 937        clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
 938        set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
 939        memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
 940        memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
 941        spin_unlock(&ls->ls_recover_spin);
 942        return 0;
 943
 944fail:
 945        mounted_unlock(sdp);
 946        control_unlock(sdp);
 947        return error;
 948}
 949
 950static int control_first_done(struct gfs2_sbd *sdp)
 951{
 952        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
 953        uint32_t start_gen, block_gen;
 954        int error;
 955
 956restart:
 957        spin_lock(&ls->ls_recover_spin);
 958        start_gen = ls->ls_recover_start;
 959        block_gen = ls->ls_recover_block;
 960
 961        if (test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags) ||
 962            !test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
 963            !test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
 964                /* sanity check, should not happen */
 965                fs_err(sdp, "control_first_done start %u block %u flags %lx\n",
 966                       start_gen, block_gen, ls->ls_recover_flags);
 967                spin_unlock(&ls->ls_recover_spin);
 968                control_unlock(sdp);
 969                return -1;
 970        }
 971
 972        if (start_gen == block_gen) {
 973                /*
 974                 * Wait for the end of a dlm recovery cycle to switch from
 975                 * first mounter recovery.  We can ignore any recover_slot
 976                 * callbacks between the recover_prep and next recover_done
 977                 * because we are still the first mounter and any failed nodes
 978                 * have not fully mounted, so they don't need recovery.
 979                 */
 980                spin_unlock(&ls->ls_recover_spin);
 981                fs_info(sdp, "control_first_done wait gen %u\n", start_gen);
 982
 983                wait_on_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY,
 984                            TASK_UNINTERRUPTIBLE);
 985                goto restart;
 986        }
 987
 988        clear_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
 989        set_bit(DFL_FIRST_MOUNT_DONE, &ls->ls_recover_flags);
 990        memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
 991        memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
 992        spin_unlock(&ls->ls_recover_spin);
 993
 994        memset(ls->ls_lvb_bits, 0, GDLM_LVB_SIZE);
 995        control_lvb_write(ls, start_gen, ls->ls_lvb_bits);
 996
 997        error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT);
 998        if (error)
 999                fs_err(sdp, "control_first_done mounted PR error %d\n", error);
1000
1001        error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
1002        if (error)
1003                fs_err(sdp, "control_first_done control NL error %d\n", error);
1004
1005        return error;
1006}
1007
1008/*
1009 * Expand static jid arrays if necessary (by increments of RECOVER_SIZE_INC)
1010 * to accomodate the largest slot number.  (NB dlm slot numbers start at 1,
1011 * gfs2 jids start at 0, so jid = slot - 1)
1012 */
1013
1014#define RECOVER_SIZE_INC 16
1015
1016static int set_recover_size(struct gfs2_sbd *sdp, struct dlm_slot *slots,
1017                            int num_slots)
1018{
1019        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1020        uint32_t *submit = NULL;
1021        uint32_t *result = NULL;
1022        uint32_t old_size, new_size;
1023        int i, max_jid;
1024
1025        if (!ls->ls_lvb_bits) {
1026                ls->ls_lvb_bits = kzalloc(GDLM_LVB_SIZE, GFP_NOFS);
1027                if (!ls->ls_lvb_bits)
1028                        return -ENOMEM;
1029        }
1030
1031        max_jid = 0;
1032        for (i = 0; i < num_slots; i++) {
1033                if (max_jid < slots[i].slot - 1)
1034                        max_jid = slots[i].slot - 1;
1035        }
1036
1037        old_size = ls->ls_recover_size;
1038        new_size = old_size;
1039        while (new_size < max_jid + 1)
1040                new_size += RECOVER_SIZE_INC;
1041        if (new_size == old_size)
1042                return 0;
1043
1044        submit = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS);
1045        result = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS);
1046        if (!submit || !result) {
1047                kfree(submit);
1048                kfree(result);
1049                return -ENOMEM;
1050        }
1051
1052        spin_lock(&ls->ls_recover_spin);
1053        memcpy(submit, ls->ls_recover_submit, old_size * sizeof(uint32_t));
1054        memcpy(result, ls->ls_recover_result, old_size * sizeof(uint32_t));
1055        kfree(ls->ls_recover_submit);
1056        kfree(ls->ls_recover_result);
1057        ls->ls_recover_submit = submit;
1058        ls->ls_recover_result = result;
1059        ls->ls_recover_size = new_size;
1060        spin_unlock(&ls->ls_recover_spin);
1061        return 0;
1062}
1063
1064static void free_recover_size(struct lm_lockstruct *ls)
1065{
1066        kfree(ls->ls_lvb_bits);
1067        kfree(ls->ls_recover_submit);
1068        kfree(ls->ls_recover_result);
1069        ls->ls_recover_submit = NULL;
1070        ls->ls_recover_result = NULL;
1071        ls->ls_recover_size = 0;
1072        ls->ls_lvb_bits = NULL;
1073}
1074
1075/* dlm calls before it does lock recovery */
1076
1077static void gdlm_recover_prep(void *arg)
1078{
1079        struct gfs2_sbd *sdp = arg;
1080        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1081
1082        spin_lock(&ls->ls_recover_spin);
1083        ls->ls_recover_block = ls->ls_recover_start;
1084        set_bit(DFL_DLM_RECOVERY, &ls->ls_recover_flags);
1085
1086        if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
1087             test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
1088                spin_unlock(&ls->ls_recover_spin);
1089                return;
1090        }
1091        set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
1092        spin_unlock(&ls->ls_recover_spin);
1093}
1094
1095/* dlm calls after recover_prep has been completed on all lockspace members;
1096   identifies slot/jid of failed member */
1097
1098static void gdlm_recover_slot(void *arg, struct dlm_slot *slot)
1099{
1100        struct gfs2_sbd *sdp = arg;
1101        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1102        int jid = slot->slot - 1;
1103
1104        spin_lock(&ls->ls_recover_spin);
1105        if (ls->ls_recover_size < jid + 1) {
1106                fs_err(sdp, "recover_slot jid %d gen %u short size %d\n",
1107                       jid, ls->ls_recover_block, ls->ls_recover_size);
1108                spin_unlock(&ls->ls_recover_spin);
1109                return;
1110        }
1111
1112        if (ls->ls_recover_submit[jid]) {
1113                fs_info(sdp, "recover_slot jid %d gen %u prev %u\n",
1114                        jid, ls->ls_recover_block, ls->ls_recover_submit[jid]);
1115        }
1116        ls->ls_recover_submit[jid] = ls->ls_recover_block;
1117        spin_unlock(&ls->ls_recover_spin);
1118}
1119
1120/* dlm calls after recover_slot and after it completes lock recovery */
1121
1122static void gdlm_recover_done(void *arg, struct dlm_slot *slots, int num_slots,
1123                              int our_slot, uint32_t generation)
1124{
1125        struct gfs2_sbd *sdp = arg;
1126        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1127
1128        /* ensure the ls jid arrays are large enough */
1129        set_recover_size(sdp, slots, num_slots);
1130
1131        spin_lock(&ls->ls_recover_spin);
1132        ls->ls_recover_start = generation;
1133
1134        if (!ls->ls_recover_mount) {
1135                ls->ls_recover_mount = generation;
1136                ls->ls_jid = our_slot - 1;
1137        }
1138
1139        if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags))
1140                queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work, 0);
1141
1142        clear_bit(DFL_DLM_RECOVERY, &ls->ls_recover_flags);
1143        smp_mb__after_atomic();
1144        wake_up_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY);
1145        spin_unlock(&ls->ls_recover_spin);
1146}
1147
1148/* gfs2_recover thread has a journal recovery result */
1149
1150static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid,
1151                                 unsigned int result)
1152{
1153        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1154
1155        if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
1156                return;
1157
1158        /* don't care about the recovery of own journal during mount */
1159        if (jid == ls->ls_jid)
1160                return;
1161
1162        spin_lock(&ls->ls_recover_spin);
1163        if (test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
1164                spin_unlock(&ls->ls_recover_spin);
1165                return;
1166        }
1167        if (ls->ls_recover_size < jid + 1) {
1168                fs_err(sdp, "recovery_result jid %d short size %d\n",
1169                       jid, ls->ls_recover_size);
1170                spin_unlock(&ls->ls_recover_spin);
1171                return;
1172        }
1173
1174        fs_info(sdp, "recover jid %d result %s\n", jid,
1175                result == LM_RD_GAVEUP ? "busy" : "success");
1176
1177        ls->ls_recover_result[jid] = result;
1178
1179        /* GAVEUP means another node is recovering the journal; delay our
1180           next attempt to recover it, to give the other node a chance to
1181           finish before trying again */
1182
1183        if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags))
1184                queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work,
1185                                   result == LM_RD_GAVEUP ? HZ : 0);
1186        spin_unlock(&ls->ls_recover_spin);
1187}
1188
1189static const struct dlm_lockspace_ops gdlm_lockspace_ops = {
1190        .recover_prep = gdlm_recover_prep,
1191        .recover_slot = gdlm_recover_slot,
1192        .recover_done = gdlm_recover_done,
1193};
1194
1195static int gdlm_mount(struct gfs2_sbd *sdp, const char *table)
1196{
1197        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1198        char cluster[GFS2_LOCKNAME_LEN];
1199        const char *fsname;
1200        uint32_t flags;
1201        int error, ops_result;
1202
1203        /*
1204         * initialize everything
1205         */
1206
1207        INIT_DELAYED_WORK(&sdp->sd_control_work, gfs2_control_func);
1208        spin_lock_init(&ls->ls_recover_spin);
1209        ls->ls_recover_flags = 0;
1210        ls->ls_recover_mount = 0;
1211        ls->ls_recover_start = 0;
1212        ls->ls_recover_block = 0;
1213        ls->ls_recover_size = 0;
1214        ls->ls_recover_submit = NULL;
1215        ls->ls_recover_result = NULL;
1216        ls->ls_lvb_bits = NULL;
1217
1218        error = set_recover_size(sdp, NULL, 0);
1219        if (error)
1220                goto fail;
1221
1222        /*
1223         * prepare dlm_new_lockspace args
1224         */
1225
1226        fsname = strchr(table, ':');
1227        if (!fsname) {
1228                fs_info(sdp, "no fsname found\n");
1229                error = -EINVAL;
1230                goto fail_free;
1231        }
1232        memset(cluster, 0, sizeof(cluster));
1233        memcpy(cluster, table, strlen(table) - strlen(fsname));
1234        fsname++;
1235
1236        flags = DLM_LSFL_FS | DLM_LSFL_NEWEXCL;
1237
1238        /*
1239         * create/join lockspace
1240         */
1241
1242        error = dlm_new_lockspace(fsname, cluster, flags, GDLM_LVB_SIZE,
1243                                  &gdlm_lockspace_ops, sdp, &ops_result,
1244                                  &ls->ls_dlm);
1245        if (error) {
1246                fs_err(sdp, "dlm_new_lockspace error %d\n", error);
1247                goto fail_free;
1248        }
1249
1250        if (ops_result < 0) {
1251                /*
1252                 * dlm does not support ops callbacks,
1253                 * old dlm_controld/gfs_controld are used, try without ops.
1254                 */
1255                fs_info(sdp, "dlm lockspace ops not used\n");
1256                free_recover_size(ls);
1257                set_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags);
1258                return 0;
1259        }
1260
1261        if (!test_bit(SDF_NOJOURNALID, &sdp->sd_flags)) {
1262                fs_err(sdp, "dlm lockspace ops disallow jid preset\n");
1263                error = -EINVAL;
1264                goto fail_release;
1265        }
1266
1267        /*
1268         * control_mount() uses control_lock to determine first mounter,
1269         * and for later mounts, waits for any recoveries to be cleared.
1270         */
1271
1272        error = control_mount(sdp);
1273        if (error) {
1274                fs_err(sdp, "mount control error %d\n", error);
1275                goto fail_release;
1276        }
1277
1278        ls->ls_first = !!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
1279        clear_bit(SDF_NOJOURNALID, &sdp->sd_flags);
1280        smp_mb__after_atomic();
1281        wake_up_bit(&sdp->sd_flags, SDF_NOJOURNALID);
1282        return 0;
1283
1284fail_release:
1285        dlm_release_lockspace(ls->ls_dlm, 2);
1286fail_free:
1287        free_recover_size(ls);
1288fail:
1289        return error;
1290}
1291
1292static void gdlm_first_done(struct gfs2_sbd *sdp)
1293{
1294        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1295        int error;
1296
1297        if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
1298                return;
1299
1300        error = control_first_done(sdp);
1301        if (error)
1302                fs_err(sdp, "mount first_done error %d\n", error);
1303}
1304
1305static void gdlm_unmount(struct gfs2_sbd *sdp)
1306{
1307        struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1308
1309        if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
1310                goto release;
1311
1312        /* wait for gfs2_control_wq to be done with this mount */
1313
1314        spin_lock(&ls->ls_recover_spin);
1315        set_bit(DFL_UNMOUNT, &ls->ls_recover_flags);
1316        spin_unlock(&ls->ls_recover_spin);
1317        flush_delayed_work(&sdp->sd_control_work);
1318
1319        /* mounted_lock and control_lock will be purged in dlm recovery */
1320release:
1321        if (ls->ls_dlm) {
1322                dlm_release_lockspace(ls->ls_dlm, 2);
1323                ls->ls_dlm = NULL;
1324        }
1325
1326        free_recover_size(ls);
1327}
1328
1329static const match_table_t dlm_tokens = {
1330        { Opt_jid, "jid=%d"},
1331        { Opt_id, "id=%d"},
1332        { Opt_first, "first=%d"},
1333        { Opt_nodir, "nodir=%d"},
1334        { Opt_err, NULL },
1335};
1336
1337const struct lm_lockops gfs2_dlm_ops = {
1338        .lm_proto_name = "lock_dlm",
1339        .lm_mount = gdlm_mount,
1340        .lm_first_done = gdlm_first_done,
1341        .lm_recovery_result = gdlm_recovery_result,
1342        .lm_unmount = gdlm_unmount,
1343        .lm_put_lock = gdlm_put_lock,
1344        .lm_lock = gdlm_lock,
1345        .lm_cancel = gdlm_cancel,
1346        .lm_tokens = &dlm_tokens,
1347};
1348
1349