/* * We need to make sure the buffer pointer returned is naturally aligned for the * biggest basic data type we put into it. We have already accounted for this * padding when sizing the buffer. * * However, this padding does not get written into the log, and hence we have to * track the space used by the log vectors separately to prevent log space hangs * due to inaccurate accounting (i.e. a leak) of the used log space through the * CIL context ticket. * * We also add space for the xlog_op_header that describes this region in the * log. This prepends the data region we return to the caller to copy their data * into, so do all the static initialisation of the ophdr now. Because the ophdr * is not 8 byte aligned, we have to be careful to ensure that we align the * start of the buffer such that the region we return to the call is 8 byte * aligned and packed against the tail of the ophdr.
*/ void *
xlog_prepare_iovec( struct xfs_log_vec *lv, struct xfs_log_iovec **vecp,
uint type)
{ struct xfs_log_iovec *vec = *vecp; struct xlog_op_header *oph;
uint32_t len; void *buf;
/* * Return the space in the log between the tail and the head. In the case where * we have overrun available reservation space, return 0. The memory barrier * pairs with the smp_wmb() in xlog_cil_ail_insert() to ensure that grant head * vs tail space updates are seen in the correct order and hence avoid * transients as space is transferred from the grant heads to the AIL on commit * completion.
*/ static uint64_t
xlog_grant_space_left( struct xlog *log, struct xlog_grant_head *head)
{
int64_t free_bytes;
smp_rmb(); /* paired with smp_wmb in xlog_cil_ail_insert() */
free_bytes = log->l_logsize - READ_ONCE(log->l_tail_space) -
atomic64_read(&head->grant); if (free_bytes > 0) return free_bytes; return 0;
}
/* * Atomically get the log space required for a log ticket. * * Once a ticket gets put onto head->waiters, it will only return after the * needed reservation is satisfied. * * This function is structured so that it has a lock free fast path. This is * necessary because every new transaction reservation will come through this * path. Hence any lock will be globally hot if we take it unconditionally on * every pass. * * As tickets are only ever moved on and off head->waiters under head->lock, we * only need to take that lock if we are going to add the ticket to the queue * and sleep. We can avoid taking the lock if the ticket was never added to * head->waiters because the t_queue list head will be empty and we hold the * only reference to it so it can safely be checked unlocked.
*/ STATICint
xlog_grant_head_check( struct xlog *log, struct xlog_grant_head *head, struct xlog_ticket *tic, int *need_bytes)
{ int free_bytes; int error = 0;
ASSERT(!xlog_in_recovery(log));
/* * If there are other waiters on the queue then give them a chance at * logspace before us. Wake up the first waiters, if we do not wake * up all the waiters then go to sleep waiting for more free space, * otherwise try to get some space for this transaction.
*/
*need_bytes = xlog_ticket_reservation(log, head, tic);
free_bytes = xlog_grant_space_left(log, head); if (!list_empty_careful(&head->waiters)) {
spin_lock(&head->lock); if (!xlog_grant_head_wake(log, head, &free_bytes) ||
free_bytes < *need_bytes) {
error = xlog_grant_head_wait(log, head, tic,
*need_bytes);
}
spin_unlock(&head->lock);
} elseif (free_bytes < *need_bytes) {
spin_lock(&head->lock);
error = xlog_grant_head_wait(log, head, tic, *need_bytes);
spin_unlock(&head->lock);
}
return error;
}
bool
xfs_log_writable( struct xfs_mount *mp)
{ /* * Do not write to the log on norecovery mounts, if the data or log * devices are read-only, or if the filesystem is shutdown. Read-only * mounts allow internal writes for log recovery and unmount purposes, * so don't restrict that case.
*/ if (xfs_has_norecovery(mp)) returnfalse; if (xfs_readonly_buftarg(mp->m_ddev_targp)) returnfalse; if (xfs_readonly_buftarg(mp->m_log->l_targ)) returnfalse; if (xlog_is_shutdown(mp->m_log)) returnfalse; returntrue;
}
/* * Replenish the byte reservation required by moving the grant write head.
*/ int
xfs_log_regrant( struct xfs_mount *mp, struct xlog_ticket *tic)
{ struct xlog *log = mp->m_log; int need_bytes; int error = 0;
if (xlog_is_shutdown(log)) return -EIO;
XFS_STATS_INC(mp, xs_try_logspace);
/* * This is a new transaction on the ticket, so we need to change the * transaction ID so that the next transaction has a different TID in * the log. Just add one to the existing tid so that we can see chains * of rolling transactions in the log easily.
*/
tic->t_tid++;
tic->t_curr_res = tic->t_unit_res; if (tic->t_cnt > 0) return 0;
trace_xfs_log_regrant(log, tic);
error = xlog_grant_head_check(log, &log->l_write_head, tic,
&need_bytes); if (error) goto out_error;
out_error: /* * If we are failing, make sure the ticket doesn't have any current * reservations. We don't want to add this back when the ticket/ * transaction gets cancelled.
*/
tic->t_curr_res = 0;
tic->t_cnt = 0; /* ungrant will give back unit_res * t_cnt. */ return error;
}
/* * Reserve log space and return a ticket corresponding to the reservation. * * Each reservation is going to reserve extra space for a log record header. * When writes happen to the on-disk log, we don't subtract the length of the * log record header from any reservation. By wasting space in each * reservation, we prevent over allocation problems.
*/ int
xfs_log_reserve( struct xfs_mount *mp, int unit_bytes, int cnt, struct xlog_ticket **ticp, bool permanent)
{ struct xlog *log = mp->m_log; struct xlog_ticket *tic; int need_bytes; int error = 0;
out_error: /* * If we are failing, make sure the ticket doesn't have any current * reservations. We don't want to add this back when the ticket/ * transaction gets cancelled.
*/
tic->t_curr_res = 0;
tic->t_cnt = 0; /* ungrant will give back unit_res * t_cnt. */ return error;
}
/* * Run all the pending iclog callbacks and wake log force waiters and iclog * space waiters so they can process the newly set shutdown state. We really * don't care what order we process callbacks here because the log is shut down * and so state cannot change on disk anymore. However, we cannot wake waiters * until the callbacks have been processed because we may be in unmount and * we must ensure that all AIL operations the callbacks perform have completed * before we tear down the AIL. * * We avoid processing actively referenced iclogs so that we don't run callbacks * while the iclog owner might still be preparing the iclog for IO submssion. * These will be caught by xlog_state_iclog_release() and call this function * again to process any callbacks that may have been added to that iclog.
*/ staticvoid
xlog_state_shutdown_callbacks( struct xlog *log)
{ struct xlog_in_core *iclog;
LIST_HEAD(cb_list);
iclog = log->l_iclog; do { if (atomic_read(&iclog->ic_refcnt)) { /* Reference holder will re-run iclog callbacks. */ continue;
}
list_splice_init(&iclog->ic_callbacks, &cb_list);
spin_unlock(&log->l_icloglock);
xlog_cil_process_committed(&cb_list);
spin_lock(&log->l_icloglock);
wake_up_all(&iclog->ic_write_wait);
wake_up_all(&iclog->ic_force_wait);
} while ((iclog = iclog->ic_next) != log->l_iclog);
wake_up_all(&log->l_flush_wait);
}
/* * Flush iclog to disk if this is the last reference to the given iclog and the * it is in the WANT_SYNC state. * * If XLOG_ICL_NEED_FUA is already set on the iclog, we need to ensure that the * log tail is updated correctly. NEED_FUA indicates that the iclog will be * written to stable storage, and implies that a commit record is contained * within the iclog. We need to ensure that the log tail does not move beyond * the tail that the first commit record in the iclog ordered against, otherwise * correct recovery of that checkpoint becomes dependent on future operations * performed on this iclog. * * Hence if NEED_FUA is set and the current iclog tail lsn is empty, write the * current tail into iclog. Once the iclog tail is set, future operations must * not modify it, otherwise they potentially violate ordering constraints for * the checkpoint commit that wrote the initial tail lsn value. The tail lsn in * the iclog will get zeroed on activation of the iclog after sync, so we * always capture the tail lsn on the iclog on the first NEED_FUA release * regardless of the number of active reference counts on this iclog.
*/ int
xlog_state_release_iclog( struct xlog *log, struct xlog_in_core *iclog, struct xlog_ticket *ticket)
{ bool last_ref;
lockdep_assert_held(&log->l_icloglock);
trace_xlog_iclog_release(iclog, _RET_IP_); /* * Grabbing the current log tail needs to be atomic w.r.t. the writing * of the tail LSN into the iclog so we guarantee that the log tail does * not move between the first time we know that the iclog needs to be * made stable and when we eventually submit it.
*/ if ((iclog->ic_state == XLOG_STATE_WANT_SYNC ||
(iclog->ic_flags & XLOG_ICL_NEED_FUA)) &&
!iclog->ic_header.h_tail_lsn) {
iclog->ic_header.h_tail_lsn =
cpu_to_be64(atomic64_read(&log->l_tail_lsn));
}
if (xlog_is_shutdown(log)) { /* * If there are no more references to this iclog, process the * pending iclog callbacks that were waiting on the release of * this iclog.
*/ if (last_ref)
xlog_state_shutdown_callbacks(log); return -EIO;
}
/* * Mount a log filesystem * * mp - ubiquitous xfs mount point structure * log_target - buftarg of on-disk log device * blk_offset - Start block # where block size is 512 bytes (BBSIZE) * num_bblocks - Number of BBSIZE blocks in on-disk log * * Return error or zero.
*/ int
xfs_log_mount(
xfs_mount_t *mp, struct xfs_buftarg *log_target,
xfs_daddr_t blk_offset, int num_bblks)
{ struct xlog *log; int error = 0; int min_logfsbs;
if (!xfs_has_norecovery(mp)) {
xfs_notice(mp, "Mounting V%d Filesystem %pU",
XFS_SB_VERSION_NUM(&mp->m_sb),
&mp->m_sb.sb_uuid);
} else {
xfs_notice(mp, "Mounting V%d filesystem %pU in no-recovery mode. Filesystem will be inconsistent.",
XFS_SB_VERSION_NUM(&mp->m_sb),
&mp->m_sb.sb_uuid);
ASSERT(xfs_is_readonly(mp));
}
/* * Now that we have set up the log and it's internal geometry * parameters, we can validate the given log space and drop a critical * message via syslog if the log size is too small. A log that is too * small can lead to unexpected situations in transaction log space * reservation stage. The superblock verifier has already validated all * the other log geometry constraints, so we don't have to check those * here. * * Note: For v4 filesystems, we can't just reject the mount if the * validation fails. This would mean that people would have to * downgrade their kernel just to remedy the situation as there is no * way to grow the log (short of black magic surgery with xfs_db). * * We can, however, reject mounts for V5 format filesystems, as the * mkfs binary being used to make the filesystem should never create a * filesystem with a log that is too small.
*/
min_logfsbs = xfs_log_calc_minimum_size(mp); if (mp->m_sb.sb_logblocks < min_logfsbs) {
xfs_warn(mp, "Log size %d blocks too small, minimum size is %d blocks",
mp->m_sb.sb_logblocks, min_logfsbs);
/* * Log check errors are always fatal on v5; or whenever bad * metadata leads to a crash.
*/ if (xfs_has_crc(mp)) {
xfs_crit(mp, "AAIEEE! Log failed size checks. Abort!");
ASSERT(0);
error = -EINVAL; goto out_free_log;
}
xfs_crit(mp, "Log size out of supported range.");
xfs_crit(mp, "Continuing onwards, but if log hangs are experienced then please report this message in the bug report.");
}
/* * Initialize the AIL now we have a log.
*/
error = xfs_trans_ail_init(mp); if (error) {
xfs_warn(mp, "AIL initialisation failed: error %d", error); goto out_free_log;
}
log->l_ailp = mp->m_ail;
/* * skip log recovery on a norecovery mount. pretend it all * just worked.
*/ if (!xfs_has_norecovery(mp)) {
error = xlog_recover(log); if (error) {
xfs_warn(mp, "log mount/recovery failed: error %d",
error);
xlog_recover_cancel(log); goto out_destroy_ail;
}
}
error = xfs_sysfs_init(&log->l_kobj, &xfs_log_ktype, &mp->m_kobj, "log"); if (error) goto out_destroy_ail;
/* Normal transactions can now occur */
clear_bit(XLOG_ACTIVE_RECOVERY, &log->l_opstate);
/* * Now the log has been fully initialised and we know were our * space grant counters are, we can initialise the permanent ticket * needed for delayed logging to work.
*/
xlog_cil_init_post_recovery(log);
/* * Finish the recovery of the file system. This is separate from the * xfs_log_mount() call, because it depends on the code in xfs_mountfs() to read * in the root and real-time bitmap inodes between calling xfs_log_mount() and * here. * * If we finish recovery successfully, start the background log work. If we are * not doing recovery, then we have a RO filesystem and we don't need to start * it.
*/ int
xfs_log_mount_finish( struct xfs_mount *mp)
{ struct xlog *log = mp->m_log; int error = 0;
if (xfs_has_norecovery(mp)) {
ASSERT(xfs_is_readonly(mp)); return 0;
}
/* * During the second phase of log recovery, we need iget and * iput to behave like they do for an active filesystem. * xfs_fs_drop_inode needs to be able to prevent the deletion * of inodes before we're done replaying log items on those * inodes. Turn it off immediately after recovery finishes * so that we don't leak the quota inodes if subsequent mount * activities fail. * * We let all inodes involved in redo item processing end up on * the LRU instead of being evicted immediately so that if we do * something to an unlinked inode, the irele won't cause * premature truncation and freeing of the inode, which results * in log recovery failure. We have to evict the unreferenced * lru inodes after clearing SB_ACTIVE because we don't * otherwise clean up the lru if there's a subsequent failure in * xfs_mountfs, which leads to us leaking the inodes if nothing * else (e.g. quotacheck) references the inodes before the * mount failure occurs.
*/
mp->m_super->s_flags |= SB_ACTIVE;
xfs_log_work_queue(mp); if (xlog_recovery_needed(log))
error = xlog_recover_finish(log);
mp->m_super->s_flags &= ~SB_ACTIVE;
evict_inodes(mp->m_super);
/* * Drain the buffer LRU after log recovery. This is required for v4 * filesystems to avoid leaving around buffers with NULL verifier ops, * but we do it unconditionally to make sure we're always in a clean * cache state after mount. * * Don't push in the error case because the AIL may have pending intents * that aren't removed until recovery is cancelled.
*/ if (xlog_recovery_needed(log)) { if (!error) {
xfs_log_force(mp, XFS_LOG_SYNC);
xfs_ail_push_all_sync(mp->m_ail);
}
xfs_notice(mp, "Ending recovery (logdev: %s)",
mp->m_logname ? mp->m_logname : "internal");
} else {
xfs_info(mp, "Ending clean mount");
}
xfs_buftarg_drain(mp->m_ddev_targp);
clear_bit(XLOG_RECOVERY_NEEDED, &log->l_opstate);
/* Make sure the log is dead if we're returning failure. */
ASSERT(!error || xlog_is_shutdown(log));
return error;
}
/* * The mount has failed. Cancel the recovery if it hasn't completed and destroy * the log.
*/ void
xfs_log_mount_cancel( struct xfs_mount *mp)
{
xlog_recover_cancel(mp->m_log);
xfs_log_unmount(mp);
}
/* * Flush out the iclog to disk ensuring that device caches are flushed and * the iclog hits stable storage before any completion waiters are woken.
*/ staticinlineint
xlog_force_iclog( struct xlog_in_core *iclog)
{
atomic_inc(&iclog->ic_refcnt);
iclog->ic_flags |= XLOG_ICL_NEED_FLUSH | XLOG_ICL_NEED_FUA; if (iclog->ic_state == XLOG_STATE_ACTIVE)
xlog_state_switch_iclogs(iclog->ic_log, iclog, 0); return xlog_state_release_iclog(iclog->ic_log, iclog, NULL);
}
/* * Cycle all the iclogbuf locks to make sure all log IO completion * is done before we tear down these buffers.
*/ staticvoid
xlog_wait_iclog_completion(struct xlog *log)
{ int i; struct xlog_in_core *iclog = log->l_iclog;
for (i = 0; i < log->l_iclog_bufs; i++) {
down(&iclog->ic_sema);
up(&iclog->ic_sema);
iclog = iclog->ic_next;
}
}
/* * Wait for the iclog and all prior iclogs to be written disk as required by the * log force state machine. Waiting on ic_force_wait ensures iclog completions * have been ordered and callbacks run before we are woken here, hence * guaranteeing that all the iclogs up to this one are on stable storage.
*/ int
xlog_wait_on_iclog( struct xlog_in_core *iclog)
__releases(iclog->ic_log->l_icloglock)
{ struct xlog *log = iclog->ic_log;
if (xlog_is_shutdown(log)) return -EIO; return 0;
}
/* * Write out an unmount record using the ticket provided. We have to account for * the data space used in the unmount ticket as this write is not done from a * transaction context that has already done the accounting for us.
*/ staticint
xlog_write_unmount_record( struct xlog *log, struct xlog_ticket *ticket)
{ struct { struct xlog_op_header ophdr; struct xfs_unmount_log_format ulf;
} unmount_rec = {
.ophdr = {
.oh_clientid = XFS_LOG,
.oh_tid = cpu_to_be32(ticket->t_tid),
.oh_flags = XLOG_UNMOUNT_TRANS,
},
.ulf = {
.magic = XLOG_UNMOUNT_TYPE,
},
}; struct xfs_log_iovec reg = {
.i_addr = &unmount_rec,
.i_len = sizeof(unmount_rec),
.i_type = XLOG_REG_TYPE_UNMOUNT,
}; struct xfs_log_vec vec = {
.lv_niovecs = 1,
.lv_iovecp = ®,
};
LIST_HEAD(lv_chain);
list_add(&vec.lv_list, &lv_chain);
/* * Mark the filesystem clean by writing an unmount record to the head of the * log.
*/ staticvoid
xlog_unmount_write( struct xlog *log)
{ struct xfs_mount *mp = log->l_mp; struct xlog_in_core *iclog; struct xlog_ticket *tic = NULL; int error;
error = xlog_write_unmount_record(log, tic); /* * At this point, we're umounting anyway, so there's no point in * transitioning log state to shutdown. Just continue...
*/
out_err: if (error)
xfs_alert(mp, "%s: unmount record failed", __func__);
do {
ASSERT(iclog->ic_state == XLOG_STATE_ACTIVE);
ASSERT(iclog->ic_offset == 0);
} while ((iclog = iclog->ic_next) != log->l_iclog);
}
/* * Unmount record used to have a string "Unmount filesystem--" in the * data section where the "Un" was really a magic number (XLOG_UNMOUNT_TYPE). * We just write the magic number now since that particular field isn't * currently architecture converted and "Unmount" is a bit foo. * As far as I know, there weren't any dependencies on the old behaviour.
*/ staticvoid
xfs_log_unmount_write( struct xfs_mount *mp)
{ struct xlog *log = mp->m_log;
if (!xfs_log_writable(mp)) return;
xfs_log_force(mp, XFS_LOG_SYNC);
if (xlog_is_shutdown(log)) return;
/* * If we think the summary counters are bad, avoid writing the unmount * record to force log recovery at next mount, after which the summary * counters will be recalculated. Refer to xlog_check_unmount_rec for * more details.
*/ if (XFS_TEST_ERROR(xfs_fs_has_sickness(mp, XFS_SICK_FS_COUNTERS), mp,
XFS_ERRTAG_FORCE_SUMMARY_RECALC)) {
xfs_alert(mp, "%s: will fix summary counters at next mount",
__func__); return;
}
/* * Empty the log for unmount/freeze. * * To do this, we first need to shut down the background log work so it is not * trying to cover the log as we clean up. We then need to unpin all objects in * the log so we can then flush them out. Once they have completed their IO and * run the callbacks removing themselves from the AIL, we can cover the log.
*/ int
xfs_log_quiesce( struct xfs_mount *mp)
{ /* * Clear log incompat features since we're quiescing the log. Report * failures, though it's not fatal to have a higher log feature * protection level than the log contents actually require.
*/ if (xfs_clear_incompat_log_features(mp)) { int error;
error = xfs_sync_sb(mp, false); if (error)
xfs_warn(mp, "Failed to clear log incompat features on quiesce");
}
/* * The superblock buffer is uncached and while xfs_ail_push_all_sync() * will push it, xfs_buftarg_wait() will not wait for it. Further, * xfs_buf_iowait() cannot be used because it was pushed with the * XBF_ASYNC flag set, so we need to use a lock/unlock pair to wait for * the IO to complete.
*/
xfs_ail_push_all_sync(mp->m_ail);
xfs_buftarg_wait(mp->m_ddev_targp);
xfs_buf_lock(mp->m_sb_bp);
xfs_buf_unlock(mp->m_sb_bp);
/* * Shut down and release the AIL and Log. * * During unmount, we need to ensure we flush all the dirty metadata objects * from the AIL so that the log is empty before we write the unmount record to * the log. Once this is done, we can tear down the AIL and the log.
*/ void
xfs_log_unmount( struct xfs_mount *mp)
{
xfs_log_clean(mp);
/* * If shutdown has come from iclog IO context, the log * cleaning will have been skipped and so we need to wait * for the iclog to complete shutdown processing before we * tear anything down.
*/
xlog_wait_iclog_completion(mp->m_log);
/* * Wake up processes waiting for log space after we have moved the log tail.
*/ void
xfs_log_space_wake( struct xfs_mount *mp)
{ struct xlog *log = mp->m_log; int free_bytes;
if (xlog_is_shutdown(log)) return;
if (!list_empty_careful(&log->l_write_head.waiters)) {
ASSERT(!xlog_in_recovery(log));
/* * Determine if we have a transaction that has gone to disk that needs to be * covered. To begin the transition to the idle state firstly the log needs to * be idle. That means the CIL, the AIL and the iclogs needs to be empty before * we start attempting to cover the log. * * Only if we are then in a state where covering is needed, the caller is * informed that dummy transactions are required to move the log into the idle * state. * * If there are any items in the AIl or CIL, then we do not want to attempt to * cover the log as we may be in a situation where there isn't log space * available to run a dummy transaction and this can lead to deadlocks when the * tail of the log is pinned by an item that is modified in the CIL. Hence * there's no point in running a dummy transaction at this point because we * can't start trying to idle the log until both the CIL and AIL are empty.
*/ staticbool
xfs_log_need_covered( struct xfs_mount *mp)
{ struct xlog *log = mp->m_log; bool needed = false;
if (!xlog_cil_empty(log)) returnfalse;
spin_lock(&log->l_icloglock); switch (log->l_covered_state) { case XLOG_STATE_COVER_DONE: case XLOG_STATE_COVER_DONE2: case XLOG_STATE_COVER_IDLE: break; case XLOG_STATE_COVER_NEED: case XLOG_STATE_COVER_NEED2: if (xfs_ail_min_lsn(log->l_ailp)) break; if (!xlog_iclogs_empty(log)) break;
/* * Explicitly cover the log. This is similar to background log covering but * intended for usage in quiesce codepaths. The caller is responsible to ensure * the log is idle and suitable for covering. The CIL, iclog buffers and AIL * must all be empty.
*/ staticint
xfs_log_cover( struct xfs_mount *mp)
{ int error = 0; bool need_covered;
/* * xfs_log_need_covered() is not idempotent because it progresses the * state machine if the log requires covering. Therefore, we must call * this function once and use the result until we've issued an sb sync. * Do so first to make that abundantly clear. * * Fall into the covering sequence if the log needs covering or the * mount has lazy superblock accounting to sync to disk. The sb sync * used for covering accumulates the in-core counters, so covering * handles this for us.
*/
need_covered = xfs_log_need_covered(mp); if (!need_covered && !xfs_has_lazysbcount(mp)) return 0;
/* * To cover the log, commit the superblock twice (at most) in * independent checkpoints. The first serves as a reference for the * tail pointer. The sync transaction and AIL push empties the AIL and * updates the in-core tail to the LSN of the first checkpoint. The * second commit updates the on-disk tail with the in-core LSN, * covering the log. Push the AIL one more time to leave it empty, as * we found it.
*/ do {
error = xfs_sync_sb(mp, true); if (error) break;
xfs_ail_push_all_sync(mp->m_ail);
} while (xfs_log_need_covered(mp));
error = blk_status_to_errno(iclog->ic_bio.bi_status); #ifdef DEBUG /* treat writes with injected CRC errors as failed */ if (iclog->ic_fail_crc)
error = -EIO; #endif
/* * Race to shutdown the filesystem if we see an error.
*/ if (XFS_TEST_ERROR(error, log->l_mp, XFS_ERRTAG_IODONE_IOERR)) {
xfs_alert(log->l_mp, "log I/O error %d", error);
xlog_force_shutdown(log, SHUTDOWN_LOG_IO_ERROR);
}
/* * Drop the lock to signal that we are done. Nothing references the * iclog after this, so an unmount waiting on this lock can now tear it * down safely. As such, it is unsafe to reference the iclog after the * unlock as we could race with it being freed.
*/
up(&iclog->ic_sema);
}
/* * Return size of each in-core log record buffer. * * All machines get 8 x 32kB buffers by default, unless tuned otherwise. * * If the filesystem blocksize is too large, we may need to choose a * larger size since the directory code currently logs entire blocks.
*/ STATICvoid
xlog_get_iclog_buffer_size( struct xfs_mount *mp, struct xlog *log)
{ if (mp->m_logbufs <= 0)
mp->m_logbufs = XLOG_MAX_ICLOGS; if (mp->m_logbsize <= 0)
mp->m_logbsize = XLOG_BIG_RECORD_BSIZE;
/* * Clear the log incompat flags if we have the opportunity. * * This only happens if we're about to log the second dummy transaction as part * of covering the log.
*/ staticinlinevoid
xlog_clear_incompat( struct xlog *log)
{ struct xfs_mount *mp = log->l_mp;
if (!xfs_sb_has_incompat_log_feature(&mp->m_sb,
XFS_SB_FEAT_INCOMPAT_LOG_ALL)) return;
if (log->l_covered_state != XLOG_STATE_COVER_DONE2) return;
xfs_clear_incompat_log_features(mp);
}
/* * Every sync period we need to unpin all items in the AIL and push them to * disk. If there is nothing dirty, then we might need to cover the log to * indicate that the filesystem is idle.
*/ staticvoid
xfs_log_worker( struct work_struct *work)
{ struct xlog *log = container_of(to_delayed_work(work), struct xlog, l_work); struct xfs_mount *mp = log->l_mp;
/* dgc: errors ignored - not fatal and nowhere to report them */ if (xfs_fs_writable(mp, SB_FREEZE_WRITE) && xfs_log_need_covered(mp)) { /* * Dump a transaction into the log that contains no real change. * This is needed to stamp the current tail LSN into the log * during the covering operation. * * We cannot use an inode here for this - that will push dirty * state back up into the VFS and then periodic inode flushing * will prevent log covering from making progress. Hence we * synchronously log the superblock instead to ensure the * superblock is immediately unpinned and can be written back.
*/
xlog_clear_incompat(log);
xfs_sync_sb(mp, true);
} else
xfs_log_force(mp, 0);
/* start pushing all the metadata that is currently dirty */
xfs_ail_push_all(mp->m_ail);
/* queue us up again */
xfs_log_work_queue(mp);
}
/* * This routine initializes some of the log structure for a given mount point. * Its primary purpose is to fill in enough, so recovery can occur. However, * some other stuff may be filled in too.
*/ STATICstruct xlog *
xlog_alloc_log( struct xfs_mount *mp, struct xfs_buftarg *log_target,
xfs_daddr_t blk_offset, int num_bblks)
{ struct xlog *log;
xlog_rec_header_t *head;
xlog_in_core_t **iclogp;
xlog_in_core_t *iclog, *prev_iclog=NULL; int i; int error = -ENOMEM;
uint log2_size = 0;
log = kzalloc(sizeof(struct xlog), GFP_KERNEL | __GFP_RETRY_MAYFAIL); if (!log) {
xfs_warn(mp, "Log allocation failed: No memory!"); goto out;
}
log->l_prev_block = -1; /* log->l_tail_lsn = 0x100000000LL; cycle = 1; current block = 0 */
xlog_assign_atomic_lsn(&log->l_tail_lsn, 1, 0);
log->l_curr_cycle = 1; /* 0 is bad since this is initial value */
iclogp = &log->l_iclog; /* * The amount of memory to allocate for the iclog structure is * rather funky due to the way the structure is defined. It is * done this way so that we can use different sizes for machines * with different amounts of memory. See the definition of * xlog_in_core_t in xfs_log_priv.h for details.
*/
ASSERT(log->l_iclog_size >= 4096); for (i = 0; i < log->l_iclog_bufs; i++) {
size_t bvec_size = howmany(log->l_iclog_size, PAGE_SIZE) * sizeof(struct bio_vec);
/* * Stamp cycle number in every block
*/ STATICvoid
xlog_pack_data( struct xlog *log, struct xlog_in_core *iclog, int roundoff)
{ int i, j, k; int size = iclog->ic_offset + roundoff;
__be32 cycle_lsn; char *dp;
dp = iclog->ic_datap; for (i = 0; i < BTOBB(size); i++) { if (i >= (XLOG_HEADER_CYCLE_SIZE / BBSIZE)) break;
iclog->ic_header.h_cycle_data[i] = *(__be32 *)dp;
*(__be32 *)dp = cycle_lsn;
dp += BBSIZE;
}
if (xfs_has_logv2(log->l_mp)) {
xlog_in_core_2_t *xhdr = iclog->ic_data;
for ( ; i < BTOBB(size); i++) {
j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
xhdr[j].hic_xheader.xh_cycle_data[k] = *(__be32 *)dp;
*(__be32 *)dp = cycle_lsn;
dp += BBSIZE;
}
for (i = 1; i < log->l_iclog_heads; i++)
xhdr[i].hic_xheader.xh_cycle = cycle_lsn;
}
}
/* * Calculate the checksum for a log buffer. * * This is a little more complicated than it should be because the various * headers and the actual data are non-contiguous.
*/
__le32
xlog_cksum( struct xlog *log, struct xlog_rec_header *rhead, char *dp, unsignedint hdrsize, unsignedint size)
{
uint32_t crc;
/* first generate the crc for the record header ... */
crc = xfs_start_cksum_update((char *)rhead, hdrsize,
offsetof(struct xlog_rec_header, h_crc));
/* ... then for additional cycle data for v2 logs ... */ if (xfs_has_logv2(log->l_mp)) { union xlog_in_core2 *xhdr = (union xlog_in_core2 *)rhead; int i; int xheads;
/* * We lock the iclogbufs here so that we can serialise against I/O * completion during unmount. We might be processing a shutdown * triggered during unmount, and that can occur asynchronously to the * unmount thread, and hence we need to ensure that completes before * tearing down the iclogbufs. Hence we need to hold the buffer lock * across the log IO to archieve that.
*/
down(&iclog->ic_sema); if (xlog_is_shutdown(log)) { /* * It would seem logical to return EIO here, but we rely on * the log state machine to propagate I/O errors instead of * doing it here. We kick of the state machine and unlock * the buffer manually, the code needs to be kept in sync * with the I/O completion path.
*/ goto sync;
}
/* * We use REQ_SYNC | REQ_IDLE here to tell the block layer the are more * IOs coming immediately after this one. This prevents the block layer * writeback throttle from throttling log writes behind background * metadata writeback and causing priority inversions.
*/
bio_init(&iclog->ic_bio, log->l_targ->bt_bdev, iclog->ic_bvec,
howmany(count, PAGE_SIZE),
REQ_OP_WRITE | REQ_META | REQ_SYNC | REQ_IDLE);
iclog->ic_bio.bi_iter.bi_sector = log->l_logBBstart + bno;
iclog->ic_bio.bi_end_io = xlog_bio_end_io;
iclog->ic_bio.bi_private = iclog;
if (iclog->ic_flags & XLOG_ICL_NEED_FLUSH) {
iclog->ic_bio.bi_opf |= REQ_PREFLUSH; /* * For external log devices, we also need to flush the data * device cache first to ensure all metadata writeback covered * by the LSN in this iclog is on stable storage. This is slow, * but it *must* complete before we issue the external log IO. * * If the flush fails, we cannot conclude that past metadata * writeback from the log succeeded. Repeating the flush is * not possible, hence we must shut down with log IO error to * avoid shutdown re-entering this path and erroring out again.
*/ if (log->l_targ != log->l_mp->m_ddev_targp &&
blkdev_issue_flush(log->l_mp->m_ddev_targp->bt_bdev)) goto shutdown;
} if (iclog->ic_flags & XLOG_ICL_NEED_FUA)
iclog->ic_bio.bi_opf |= REQ_FUA;
if (is_vmalloc_addr(iclog->ic_data)) { if (!bio_add_vmalloc(&iclog->ic_bio, iclog->ic_data, count)) goto shutdown;
} else {
bio_add_virt_nofail(&iclog->ic_bio, iclog->ic_data, count);
}
/* * If this log buffer would straddle the end of the log we will have * to split it up into two bios, so that we can continue at the start.
*/ if (bno + BTOBB(count) > log->l_logBBsize) { struct bio *split;
/* * We need to bump cycle number for the part of the iclog that is * written to the start of the log. Watch out for the header magic * number case, though.
*/ staticvoid
xlog_split_iclog( struct xlog *log, void *data,
uint64_t bno, unsignedint count)
{ unsignedint split_offset = BBTOB(log->l_logBBsize - bno); unsignedint i;
for (i = split_offset; i < count; i += BBSIZE) {
uint32_t cycle = get_unaligned_be32(data + i);
if (++cycle == XLOG_HEADER_MAGIC_NUM)
cycle++;
put_unaligned_be32(cycle, data + i);
}
}
/* * Flush out the in-core log (iclog) to the on-disk log in an asynchronous * fashion. Previously, we should have moved the current iclog * ptr in the log to point to the next available iclog. This allows further * write to continue while this code syncs out an iclog ready to go. * Before an in-core log can be written out, the data section must be scanned * to save away the 1st word of each BBSIZE block into the header. We replace * it with the current cycle count. Each BBSIZE block is tagged with the * cycle count because there in an implicit assumption that drives will * guarantee that entire 512 byte blocks get written at once. In other words, * we can't have part of a 512 byte block written and part not written. By * tagging each block, we will know which blocks are valid when recovering * after an unclean shutdown. * * This routine is single threaded on the iclog. No other thread can be in * this routine with the same iclog. Changing contents of iclog can there- * fore be done without grabbing the state machine lock. Updating the global * log will require grabbing the lock though. * * The entire log manager uses a logical block numbering scheme. Only * xlog_write_iclog knows about the fact that the log may not start with * block zero on a given device.
*/ STATICvoid
xlog_sync( struct xlog *log, struct xlog_in_core *iclog, struct xlog_ticket *ticket)
{ unsignedint count; /* byte count of bwrite */ unsignedint roundoff; /* roundoff to BB or stripe */
uint64_t bno; unsignedint size;
/* * If we have a ticket, account for the roundoff via the ticket * reservation to avoid touching the hot grant heads needlessly. * Otherwise, we have to move grant heads directly.
*/ if (ticket) {
ticket->t_curr_res -= roundoff;
} else {
xlog_grant_add_space(&log->l_reserve_head, roundoff);
xlog_grant_add_space(&log->l_write_head, roundoff);
}
/* put cycle number in every block */
xlog_pack_data(log, iclog, roundoff);
/* real byte length */
size = iclog->ic_offset; if (xfs_has_logv2(log->l_mp))
size += roundoff;
iclog->ic_header.h_len = cpu_to_be32(size);
/* Do we need to split this write into 2 parts? */ if (bno + BTOBB(count) > log->l_logBBsize)
xlog_split_iclog(log, &iclog->ic_header, bno, count);
/* calculcate the checksum */
iclog->ic_header.h_crc = xlog_cksum(log, &iclog->ic_header,
iclog->ic_datap, XLOG_REC_SIZE, size); /* * Intentionally corrupt the log record CRC based on the error injection * frequency, if defined. This facilitates testing log recovery in the * event of torn writes. Hence, set the IOABORT state to abort the log * write on I/O completion and shutdown the fs. The subsequent mount * detects the bad CRC and attempts to recover.
*/ #ifdef DEBUG if (XFS_TEST_ERROR(false, log->l_mp, XFS_ERRTAG_LOG_BAD_CRC)) {
iclog->ic_header.h_crc &= cpu_to_le32(0xAAAAAAAA);
iclog->ic_fail_crc = true;
xfs_warn(log->l_mp, "Intentionally corrupted log record at LSN 0x%llx. Shutdown imminent.",
be64_to_cpu(iclog->ic_header.h_lsn));
} #endif
xlog_verify_iclog(log, iclog, count);
xlog_write_iclog(log, iclog, bno, count);
}
/* * Deallocate a log structure
*/ STATICvoid
xlog_dealloc_log( struct xlog *log)
{
xlog_in_core_t *iclog, *next_iclog; int i;
/* * Destroy the CIL after waiting for iclog IO completion because an * iclog EIO error will try to shut down the log, which accesses the * CIL to wake up the waiters.
*/
xlog_cil_destroy(log);
iclog = log->l_iclog; for (i = 0; i < log->l_iclog_bufs; i++) {
next_iclog = iclog->ic_next;
kvfree(iclog->ic_data);
kfree(iclog);
iclog = next_iclog;
}
/* * Update counters atomically now that memcpy is done.
*/ staticinlinevoid
xlog_state_finish_copy( struct xlog *log, struct xlog_in_core *iclog, int record_cnt, int copy_bytes)
{
lockdep_assert_held(&log->l_icloglock);
/* * print out info relating to regions written which consume * the reservation
*/ void
xlog_print_tic_res( struct xfs_mount *mp, struct xlog_ticket *ticket)
{
xfs_warn(mp, "ticket reservation summary:");
xfs_warn(mp, " unit res = %d bytes", ticket->t_unit_res);
xfs_warn(mp, " current res = %d bytes", ticket->t_curr_res);
xfs_warn(mp, " original count = %d", ticket->t_ocnt);
xfs_warn(mp, " remaining count = %d", ticket->t_cnt);
}
/* * Print a summary of the transaction.
*/ void
xlog_print_trans( struct xfs_trans *tp)
{ struct xfs_mount *mp = tp->t_mountp; struct xfs_log_item *lip;
/* * Write log vectors into a single iclog which is guaranteed by the caller * to have enough space to write the entire log vector into.
*/ staticvoid
xlog_write_full( struct xfs_log_vec *lv, struct xlog_ticket *ticket, struct xlog_in_core *iclog,
uint32_t *log_offset,
uint32_t *len,
uint32_t *record_cnt,
uint32_t *data_cnt)
{ int index;
/* * Ordered log vectors have no regions to write so this * loop will naturally skip them.
*/ for (index = 0; index < lv->lv_niovecs; index++) { struct xfs_log_iovec *reg = &lv->lv_iovecp[index]; struct xlog_op_header *ophdr = reg->i_addr;
/* * Write log vectors into a single iclog which is smaller than the current chain * length. We write until we cannot fit a full record into the remaining space * and then stop. We return the log vector that is to be written that cannot * wholly fit in the iclog.
*/ staticint
xlog_write_partial( struct xfs_log_vec *lv, struct xlog_ticket *ticket, struct xlog_in_core **iclogp,
uint32_t *log_offset,
uint32_t *len,
uint32_t *record_cnt,
uint32_t *data_cnt)
{ struct xlog_in_core *iclog = *iclogp; struct xlog_op_header *ophdr; int index = 0;
uint32_t rlen; int error;
/* walk the logvec, copying until we run out of space in the iclog */ for (index = 0; index < lv->lv_niovecs; index++) { struct xfs_log_iovec *reg = &lv->lv_iovecp[index];
uint32_t reg_offset = 0;
/* * The first region of a continuation must have a non-zero * length otherwise log recovery will just skip over it and * start recovering from the next opheader it finds. Because we * mark the next opheader as a continuation, recovery will then * incorrectly add the continuation to the previous region and * that breaks stuff. * * Hence if there isn't space for region data after the * opheader, then we need to start afresh with a new iclog.
*/ if (iclog->ic_size - *log_offset <= sizeof(struct xlog_op_header)) {
error = xlog_write_get_more_iclog_space(ticket,
&iclog, log_offset, *len, record_cnt,
data_cnt); if (error) return error;
}
/* If we wrote the whole region, move to the next. */ if (rlen == reg->i_len) continue;
/* * We now have a partially written iovec, but it can span * multiple iclogs so we loop here. First we release the iclog * we currently have, then we get a new iclog and add a new * opheader. Then we continue copying from where we were until * we either complete the iovec or fill the iclog. If we * complete the iovec, then we increment the index and go right * back to the top of the outer loop. if we fill the iclog, we * run the inner loop again. * * This is complicated by the tail of a region using all the * space in an iclog and hence requiring us to release the iclog * and get a new one before returning to the outer loop. We must * always guarantee that we exit this inner loop with at least * space for log transaction opheaders left in the current * iclog, hence we cannot just terminate the loop at the end * of the of the continuation. So we loop while there is no * space left in the current iclog, and check for the end of the * continuation after getting a new iclog.
*/ do { /* * Ensure we include the continuation opheader in the * space we need in the new iclog by adding that size * to the length we require. This continuation opheader * needs to be accounted to the ticket as the space it * consumes hasn't been accounted to the lv we are * writing.
*/
error = xlog_write_get_more_iclog_space(ticket,
&iclog, log_offset,
*len + sizeof(struct xlog_op_header),
record_cnt, data_cnt); if (error) return error;
/* * If rlen fits in the iclog, then end the region * continuation. Otherwise we're going around again.
*/
reg_offset += rlen;
rlen = reg->i_len - reg_offset; if (rlen <= iclog->ic_size - *log_offset)
ophdr->oh_flags |= XLOG_END_TRANS; else
ophdr->oh_flags |= XLOG_CONTINUE_TRANS;
} while (ophdr->oh_flags & XLOG_CONTINUE_TRANS);
}
/* * No more iovecs remain in this logvec so return the next log vec to * the caller so it can go back to fast path copying.
*/
*iclogp = iclog; return 0;
}
/* * Write some region out to in-core log * * This will be called when writing externally provided regions or when * writing out a commit record for a given transaction. * * General algorithm: * 1. Find total length of this write. This may include adding to the * lengths passed in. * 2. Check whether we violate the tickets reservation. * 3. While writing to this iclog * A. Reserve as much space in this iclog as can get * B. If this is first write, save away start lsn * C. While writing this region: * 1. If first write of transaction, write start record * 2. Write log operation header (header per region) * 3. Find out if we can fit entire region into this iclog * 4. Potentially, verify destination memcpy ptr * 5. Memcpy (partial) region * 6. If partial copy, release iclog; otherwise, continue * copying more regions into current iclog * 4. Mark want sync bit (in simulation mode) * 5. Release iclog for potential flush to on-disk log. * * ERRORS: * 1. Panic if reservation is overrun. This should never happen since * reservation amounts are generated internal to the filesystem. * NOTES: * 1. Tickets are single threaded data structures. * 2. The XLOG_END_TRANS & XLOG_CONTINUE_TRANS flags are passed down to the * syncing routine. When a single log_write region needs to span * multiple in-core logs, the XLOG_CONTINUE_TRANS bit should be set * on all log operation writes which don't contain the end of the * region. The XLOG_END_TRANS bit is used for the in-core log * operation which contains the end of the continued log_write region. * 3. When xlog_state_get_iclog_space() grabs the rest of the current iclog, * we don't really know exactly how much space will be used. As a result, * we don't update ic_offset until the end when we know exactly how many * bytes have been written out.
*/ int
xlog_write( struct xlog *log, struct xfs_cil_ctx *ctx, struct list_head *lv_chain, struct xlog_ticket *ticket,
uint32_t len)
if (ticket->t_curr_res < 0) {
xfs_alert_tag(log->l_mp, XFS_PTAG_LOGRES, "ctx ticket reservation ran out. Need to up reservation");
xlog_print_tic_res(log->l_mp, ticket);
xlog_force_shutdown(log, SHUTDOWN_LOG_IO_ERROR);
}
/* * If we have a context pointer, pass it the first iclog we are * writing to so it can record state needed for iclog write * ordering.
*/ if (ctx)
xlog_cil_set_ctx_write_state(ctx, iclog);
list_for_each_entry(lv, lv_chain, lv_list) { /* * If the entire log vec does not fit in the iclog, punt it to * the partial copy loop which can handle this case.
*/ if (lv->lv_niovecs &&
lv->lv_bytes > iclog->ic_size - log_offset) {
error = xlog_write_partial(lv, ticket, &iclog,
&log_offset, &len, &record_cnt,
&data_cnt); if (error) { /* * We have no iclog to release, so just return * the error immediately.
*/ return error;
}
} else {
xlog_write_full(lv, ticket, iclog, &log_offset,
&len, &record_cnt, &data_cnt);
}
}
ASSERT(len == 0);
/* * We've already been guaranteed that the last writes will fit inside * the current iclog, and hence it will already have the space used by * those writes accounted to it. Hence we do not need to update the * iclog with the number of bytes written here.
*/
spin_lock(&log->l_icloglock);
xlog_state_finish_copy(log, iclog, record_cnt, 0);
error = xlog_state_release_iclog(log, iclog, ticket);
spin_unlock(&log->l_icloglock);
/* * If the number of ops in this iclog indicate it just contains the * dummy transaction, we can change state into IDLE (the second time * around). Otherwise we should change the state into NEED a dummy. * We don't need to cover the dummy.
*/ if (*iclogs_changed == 0 &&
iclog->ic_header.h_num_logops == cpu_to_be32(XLOG_COVER_OPS)) {
*iclogs_changed = 1;
} else { /*
--> --------------------
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.