Lines Matching defs:lvb
406 * 10. gfs2_control updates control_lock lvb with new generation and jid bits
410 * 14. gfs2_control updates control_lock lvb jid bits for recovered journals
433 * 10. gfs2_control sets control_lock lvb = new gen + bits for failed jids
435 * 14. gfs2_control clears control_lock lvb bits for recovered jids
442 * All nodes attempt to update the control_lock lvb with the new generation
444 * do so; others will see that it's already done (lvb already contains new
448 * . All nodes attempt to set control_lock lvb gen + bits for the new gen
449 * . One node gets control_lock first and writes the lvb, others see it's done
451 * . One node succeeds for a jid, and that one clears the jid bit in the lvb
452 * . All nodes will eventually see all lvb bits clear and unblock locks
454 * - is there a problem with clearing an lvb bit that should be set
458 * 2. lvb bit set for step 1
462 * 6. lvb bit set for step 5 (will already be set)
463 * 7. lvb bit cleared for step 3
503 * control_lock EX->NL, write lvb generation
509 * read lvb generation
521 * steps above until the lvb generation number is >= its mount generation
522 * number (from initial recover_done) and all lvb bits are clear.
524 * - control_lock lvb format
554 static int all_jid_bits_clear(char *lvb)
556 return !memchr_inv(lvb + JID_BITMAP_OFFSET, 0,
694 * Propagate recover_submit[] and recover_result[] to lvb:
698 * set lvb bit for jids in recover_submit[] if the lvb has not
701 * clear lvb bit for jids in recover_result[] if the result of
727 * Clear lvb bits for jids we've successfully recovered.
733 * the jid and cleared the lvb bit for it.
751 * Failed slots before start_gen are already set in lvb.
761 * Failed slots before start_gen are not yet set in lvb.
772 latest generation to the lvb */
795 * Everyone will see jid bits set in the lvb, run gfs2_recover_set(),
796 * and clear a jid bit in the lvb if the recovery is a success.
798 * be cleared in the lvb, and everyone will clear BLOCK_LOCKS.
813 * No more jid bits set in lvb, all recovery is done, unblock locks
890 * control_lock lvb keeps track of any pending journal recoveries.
931 * If not, then we need to wait for the control_lock lvb to be
966 * lvb generation to be >= the generation from our first recover_done
967 * and all lvb bits to be clear (no pending journal recoveries.)
982 /* wait for mounted nodes to update control_lock lvb to our
991 "mount %u lvb %u flags %lx\n", block_gen,
1000 /* wait for mounted nodes to update control_lock lvb to the
1003 "lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
1012 "lvb %u flags %lx\n", block_gen, start_gen, mount_gen,