#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_rw.h"
+#include "xfs_trace.h"
kmem_zone_t *xfs_log_ticket_zone;
/* local ticket functions */
-STATIC xlog_ticket_t *xlog_ticket_get(xlog_t *log,
+STATIC xlog_ticket_t *xlog_ticket_alloc(xlog_t *log,
int unit_bytes,
int count,
char clientid,
uint flags);
-STATIC void xlog_ticket_put(xlog_t *log, xlog_ticket_t *ticket);
#if defined(DEBUG)
STATIC void xlog_verify_dest_ptr(xlog_t *log, __psint_t ptr);
STATIC int xlog_iclogs_empty(xlog_t *log);
-#if defined(XFS_LOG_TRACE)
-void
-xlog_trace_loggrant(xlog_t *log, xlog_ticket_t *tic, xfs_caddr_t string)
-{
- unsigned long cnts;
-
- if (!log->l_grant_trace) {
- log->l_grant_trace = ktrace_alloc(2048, KM_NOSLEEP);
- if (!log->l_grant_trace)
- return;
- }
- /* ticket counts are 1 byte each */
- cnts = ((unsigned long)tic->t_ocnt) | ((unsigned long)tic->t_cnt) << 8;
-
- ktrace_enter(log->l_grant_trace,
- (void *)tic,
- (void *)log->l_reserve_headq,
- (void *)log->l_write_headq,
- (void *)((unsigned long)log->l_grant_reserve_cycle),
- (void *)((unsigned long)log->l_grant_reserve_bytes),
- (void *)((unsigned long)log->l_grant_write_cycle),
- (void *)((unsigned long)log->l_grant_write_bytes),
- (void *)((unsigned long)log->l_curr_cycle),
- (void *)((unsigned long)log->l_curr_block),
- (void *)((unsigned long)CYCLE_LSN(log->l_tail_lsn)),
- (void *)((unsigned long)BLOCK_LSN(log->l_tail_lsn)),
- (void *)string,
- (void *)((unsigned long)tic->t_trans_type),
- (void *)cnts,
- (void *)((unsigned long)tic->t_curr_res),
- (void *)((unsigned long)tic->t_unit_res));
-}
-
-void
-xlog_trace_iclog(xlog_in_core_t *iclog, uint state)
-{
- if (!iclog->ic_trace)
- iclog->ic_trace = ktrace_alloc(256, KM_SLEEP);
- ktrace_enter(iclog->ic_trace,
- (void *)((unsigned long)state),
- (void *)((unsigned long)current_pid()),
- (void *)NULL, (void *)NULL, (void *)NULL, (void *)NULL,
- (void *)NULL, (void *)NULL, (void *)NULL, (void *)NULL,
- (void *)NULL, (void *)NULL, (void *)NULL, (void *)NULL,
- (void *)NULL, (void *)NULL);
-}
-#else
-#define xlog_trace_loggrant(log,tic,string)
-#define xlog_trace_iclog(iclog,state)
-#endif /* XFS_LOG_TRACE */
-
static void
xlog_ins_ticketq(struct xlog_ticket **qp, struct xlog_ticket *tic)
static void
xlog_grant_add_space_write(struct log *log, int bytes)
{
- log->l_grant_write_bytes += bytes;
- if (log->l_grant_write_bytes > log->l_logsize) {
- log->l_grant_write_bytes -= log->l_logsize;
+ int tmp = log->l_logsize - log->l_grant_write_bytes;
+ if (tmp > bytes)
+ log->l_grant_write_bytes += bytes;
+ else {
log->l_grant_write_cycle++;
+ log->l_grant_write_bytes = bytes - tmp;
}
}
static void
xlog_grant_add_space_reserve(struct log *log, int bytes)
{
- log->l_grant_reserve_bytes += bytes;
- if (log->l_grant_reserve_bytes > log->l_logsize) {
- log->l_grant_reserve_bytes -= log->l_logsize;
+ int tmp = log->l_logsize - log->l_grant_reserve_bytes;
+ if (tmp > bytes)
+ log->l_grant_reserve_bytes += bytes;
+ else {
log->l_grant_reserve_cycle++;
+ log->l_grant_reserve_bytes = bytes - tmp;
}
}
if ((ticket->t_flags & XLOG_TIC_PERM_RESERV) == 0 ||
(flags & XFS_LOG_REL_PERM_RESERV)) {
+ trace_xfs_log_done_nonperm(log, ticket);
+
/*
* Release ticket if not permanent reservation or a specific
* request has been made to release a permanent reservation.
*/
- xlog_trace_loggrant(log, ticket, "xfs_log_done: (non-permanent)");
xlog_ungrant_log_space(log, ticket);
- xlog_ticket_put(log, ticket);
+ xfs_log_ticket_put(ticket);
} else {
- xlog_trace_loggrant(log, ticket, "xfs_log_done: (permanent)");
- xlog_regrant_reserve_log_space(log, ticket);
- }
+ trace_xfs_log_done_perm(log, ticket);
- /* If this ticket was a permanent reservation and we aren't
- * trying to release it, reset the inited flags; so next time
- * we write, a start record will be written out.
- */
- if ((ticket->t_flags & XLOG_TIC_PERM_RESERV) &&
- (flags & XFS_LOG_REL_PERM_RESERV) == 0)
+ xlog_regrant_reserve_log_space(log, ticket);
+ /* If this ticket was a permanent reservation and we aren't
+ * trying to release it, reset the inited flags; so next time
+ * we write, a start record will be written out.
+ */
ticket->t_flags |= XLOG_TIC_INITED;
+ }
return lsn;
} /* xfs_log_done */
* Asynchronous forces are implemented by setting the WANT_SYNC
* bit in the appropriate in-core log and then returning.
*
- * Synchronous forces are implemented with a semaphore. All callers
- * to force a given lsn to disk will wait on a semaphore attached to the
+ * Synchronous forces are implemented with a signal variable. All callers
+ * to force a given lsn to disk will wait on a the sv attached to the
* specific in-core log. When given in-core log finally completes its
* write to disk, that thread will wake up all threads waiting on the
- * semaphore.
+ * sv.
*/
int
_xfs_log_force(
return xlog_state_sync_all(log, flags, log_flushed);
else
return xlog_state_sync(log, lsn, flags, log_flushed);
-} /* xfs_log_force */
+} /* _xfs_log_force */
+
+/*
+ * Wrapper for _xfs_log_force(), to be used when caller doesn't care
+ * about errors or whether the log was flushed or not. This is the normal
+ * interface to use when trying to unpin items or move the log forward.
+ */
+void
+xfs_log_force(
+ xfs_mount_t *mp,
+ xfs_lsn_t lsn,
+ uint flags)
+{
+ int error;
+ error = _xfs_log_force(mp, lsn, flags, NULL);
+ if (error) {
+ xfs_fs_cmn_err(CE_WARN, mp, "xfs_log_force: "
+ "error %d returned.", error);
+ }
+}
+
/*
* Attaches a new iclog I/O completion callback routine during
XFS_STATS_INC(xs_try_logspace);
+
if (*ticket != NULL) {
ASSERT(flags & XFS_LOG_PERM_RESERV);
internal_ticket = (xlog_ticket_t *)*ticket;
- xlog_trace_loggrant(log, internal_ticket, "xfs_log_reserve: existing ticket (permanent trans)");
+
+ trace_xfs_log_reserve(log, internal_ticket);
+
xlog_grant_push_ail(mp, internal_ticket->t_unit_res);
retval = xlog_regrant_write_log_space(log, internal_ticket);
} else {
/* may sleep if need to allocate more tickets */
- internal_ticket = xlog_ticket_get(log, unit_bytes, cnt,
+ internal_ticket = xlog_ticket_alloc(log, unit_bytes, cnt,
client, flags);
if (!internal_ticket)
return XFS_ERROR(ENOMEM);
internal_ticket->t_trans_type = t_type;
*ticket = internal_ticket;
- xlog_trace_loggrant(log, internal_ticket,
- (internal_ticket->t_flags & XLOG_TIC_PERM_RESERV) ?
- "xfs_log_reserve: create new ticket (permanent trans)" :
- "xfs_log_reserve: create new ticket");
+
+ trace_xfs_log_reserve(log, internal_ticket);
+
xlog_grant_push_ail(mp,
(internal_ticket->t_unit_res *
internal_ticket->t_cnt));
}
mp->m_log = xlog_alloc_log(mp, log_target, blk_offset, num_bblks);
+ if (IS_ERR(mp->m_log)) {
+ error = -PTR_ERR(mp->m_log);
+ goto out;
+ }
/*
* Initialize the AIL now we have a log.
*/
- spin_lock_init(&mp->m_ail_lock);
error = xfs_trans_ail_init(mp);
if (error) {
cmn_err(CE_WARN, "XFS: AIL initialisation failed: error %d", error);
- goto error;
+ goto out_free_log;
}
+ mp->m_log->l_ailp = mp->m_ail;
/*
* skip log recovery on a norecovery mount. pretend it all
mp->m_flags |= XFS_MOUNT_RDONLY;
if (error) {
cmn_err(CE_WARN, "XFS: log mount/recovery failed: error %d", error);
- goto error;
+ goto out_destroy_ail;
}
}
/* Normal transactions can now occur */
mp->m_log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
- /* End mounting message in xfs_log_mount_finish */
return 0;
-error:
- xfs_log_unmount_dealloc(mp);
+
+out_destroy_ail:
+ xfs_trans_ail_destroy(mp);
+out_free_log:
+ xlog_dealloc_log(mp->m_log);
+out:
return error;
-} /* xfs_log_mount */
+}
/*
* Finish the recovery of the file system. This is separate from
* mp - ubiquitous xfs mount point structure
*/
int
-xfs_log_mount_finish(xfs_mount_t *mp, int mfsi_flags)
+xfs_log_mount_finish(xfs_mount_t *mp)
{
int error;
if (!(mp->m_flags & XFS_MOUNT_NORECOVERY))
- error = xlog_recover_finish(mp->m_log, mfsi_flags);
+ error = xlog_recover_finish(mp->m_log);
else {
error = 0;
ASSERT(mp->m_flags & XFS_MOUNT_RDONLY);
}
/*
- * Unmount processing for the log.
- */
-int
-xfs_log_unmount(xfs_mount_t *mp)
-{
- int error;
-
- error = xfs_log_unmount_write(mp);
- xfs_log_unmount_dealloc(mp);
- return error;
-}
-
-/*
* Final log writes as part of unmount.
*
* Mark the filesystem clean as unmount happens. Note that during relocation
if (mp->m_flags & XFS_MOUNT_RDONLY)
return 0;
- xfs_log_force(mp, 0, XFS_LOG_FORCE|XFS_LOG_SYNC);
+ error = _xfs_log_force(mp, 0, XFS_LOG_FORCE|XFS_LOG_SYNC, NULL);
+ ASSERT(error || !(XLOG_FORCED_SHUTDOWN(log)));
#ifdef DEBUG
first_iclog = iclog = log->l_iclog;
spin_lock(&log->l_icloglock);
iclog = log->l_iclog;
atomic_inc(&iclog->ic_refcnt);
- spin_unlock(&log->l_icloglock);
xlog_state_want_sync(log, iclog);
- (void) xlog_state_release_iclog(log, iclog);
+ spin_unlock(&log->l_icloglock);
+ error = xlog_state_release_iclog(log, iclog);
spin_lock(&log->l_icloglock);
if (!(iclog->ic_state == XLOG_STATE_ACTIVE ||
iclog->ic_state == XLOG_STATE_DIRTY)) {
if (!XLOG_FORCED_SHUTDOWN(log)) {
- sv_wait(&iclog->ic_forcesema, PMEM,
+ sv_wait(&iclog->ic_force_wait, PMEM,
&log->l_icloglock, s);
} else {
spin_unlock(&log->l_icloglock);
spin_unlock(&log->l_icloglock);
}
if (tic) {
- xlog_trace_loggrant(log, tic, "unmount rec");
+ trace_xfs_log_umount_write(log, tic);
xlog_ungrant_log_space(log, tic);
- xlog_ticket_put(log, tic);
+ xfs_log_ticket_put(tic);
}
} else {
/*
spin_lock(&log->l_icloglock);
iclog = log->l_iclog;
atomic_inc(&iclog->ic_refcnt);
- spin_unlock(&log->l_icloglock);
xlog_state_want_sync(log, iclog);
- (void) xlog_state_release_iclog(log, iclog);
+ spin_unlock(&log->l_icloglock);
+ error = xlog_state_release_iclog(log, iclog);
spin_lock(&log->l_icloglock);
|| iclog->ic_state == XLOG_STATE_DIRTY
|| iclog->ic_state == XLOG_STATE_IOERROR) ) {
- sv_wait(&iclog->ic_forcesema, PMEM,
+ sv_wait(&iclog->ic_force_wait, PMEM,
&log->l_icloglock, s);
} else {
spin_unlock(&log->l_icloglock);
}
}
- return 0;
+ return error;
} /* xfs_log_unmount_write */
/*
* and deallocate the log as the aild references the log.
*/
void
-xfs_log_unmount_dealloc(xfs_mount_t *mp)
+xfs_log_unmount(xfs_mount_t *mp)
{
xfs_trans_ail_destroy(mp);
xlog_dealloc_log(mp->m_log);
break;
tail_lsn = 0;
free_bytes -= tic->t_unit_res;
- sv_signal(&tic->t_sema);
+ sv_signal(&tic->t_wait);
tic = tic->t_next;
} while (tic != log->l_write_headq);
}
break;
tail_lsn = 0;
free_bytes -= need_bytes;
- sv_signal(&tic->t_sema);
+ sv_signal(&tic->t_wait);
tic = tic->t_next;
} while (tic != log->l_reserve_headq);
}
int
xfs_log_need_covered(xfs_mount_t *mp)
{
- int needed = 0, gen;
+ int needed = 0;
xlog_t *log = mp->m_log;
if (!xfs_fs_writable(mp))
spin_lock(&log->l_icloglock);
if (((log->l_covered_state == XLOG_STATE_COVER_NEED) ||
(log->l_covered_state == XLOG_STATE_COVER_NEED2))
- && !xfs_trans_first_ail(mp, &gen)
+ && !xfs_trans_ail_tail(log->l_ailp)
&& xlog_iclogs_empty(log)) {
if (log->l_covered_state == XLOG_STATE_COVER_NEED)
log->l_covered_state = XLOG_STATE_COVER_DONE;
xfs_lsn_t tail_lsn;
xlog_t *log = mp->m_log;
- tail_lsn = xfs_trans_tail_ail(mp);
+ tail_lsn = xfs_trans_ail_tail(mp->m_ail);
spin_lock(&log->l_grant_lock);
if (tail_lsn != 0) {
log->l_tail_lsn = tail_lsn;
ASSERT(XFS_BUF_FSPRIVATE2(bp, unsigned long) == (unsigned long) 2);
XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
aborted = 0;
-
- /*
- * Some versions of cpp barf on the recursive definition of
- * ic_log -> hic_fields.ic_log and expand ic_log twice when
- * it is passed through two macros. Workaround broken cpp.
- */
l = iclog->ic_log;
/*
- * If the ordered flag has been removed by a lower
- * layer, it means the underlyin device no longer supports
+ * If the _XFS_BARRIER_FAILED flag was set by a lower
+ * layer, it means the underlying device no longer supports
* barrier I/O. Warn loudly and turn off barriers.
*/
- if ((l->l_mp->m_flags & XFS_MOUNT_BARRIER) && !XFS_BUF_ORDERED(bp)) {
+ if (bp->b_flags & _XFS_BARRIER_FAILED) {
+ bp->b_flags &= ~_XFS_BARRIER_FAILED;
l->l_mp->m_flags &= ~XFS_MOUNT_BARRIER;
xfs_fs_cmn_err(CE_WARN, l->l_mp,
"xlog_iodone: Barriers are no longer supported"
" by device. Disabling barriers\n");
- xfs_buftrace("XLOG_IODONE BARRIERS OFF", bp);
}
/*
return 0;
}
- xfs_buftrace("XLOG__BDSTRAT IOERROR", bp);
XFS_BUF_ERROR(bp, EIO);
XFS_BUF_STALE(bp);
xfs_biodone(bp);
return XFS_ERROR(EIO);
-
-
}
/*
* Return size of each in-core log record buffer.
*
- * All machines get 8 x 32KB buffers by default, unless tuned otherwise.
+ * 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.
}
if (xfs_sb_version_haslogv2(&mp->m_sb)) {
- /* # headers = size / 32K
- * one header holds cycles from 32K of data
+ /* # headers = size / 32k
+ * one header holds cycles from 32k of data
*/
xhdrs = mp->m_logbsize / XLOG_HEADER_CYCLE_SIZE;
goto done;
}
- /* All machines use 32KB buffers by default. */
+ /* All machines use 32kB buffers by default. */
log->l_iclog_size = XLOG_BIG_RECORD_BSIZE;
log->l_iclog_size_log = XLOG_BIG_RECORD_BSHIFT;
log->l_iclog_hsize = BBSIZE;
log->l_iclog_heads = 1;
- /*
- * For 16KB, we use 3 32KB buffers. For 32KB block sizes, we use
- * 4 32KB buffers. For 64KB block sizes, we use 8 32KB buffers.
- */
- if (mp->m_sb.sb_blocksize >= 16*1024) {
- log->l_iclog_size = XLOG_BIG_RECORD_BSIZE;
- log->l_iclog_size_log = XLOG_BIG_RECORD_BSHIFT;
- if (mp->m_logbufs <= 0) {
- switch (mp->m_sb.sb_blocksize) {
- case 16*1024: /* 16 KB */
- log->l_iclog_bufs = 3;
- break;
- case 32*1024: /* 32 KB */
- log->l_iclog_bufs = 4;
- break;
- case 64*1024: /* 64 KB */
- log->l_iclog_bufs = 8;
- break;
- default:
- xlog_panic("XFS: Invalid blocksize");
- break;
- }
- }
- }
-
-done: /* are we being asked to make the sizes selected above visible? */
+done:
+ /* are we being asked to make the sizes selected above visible? */
if (mp->m_logbufs == 0)
mp->m_logbufs = log->l_iclog_bufs;
if (mp->m_logbsize == 0)
xfs_buf_t *bp;
int i;
int iclogsize;
+ int error = ENOMEM;
- log = (xlog_t *)kmem_zalloc(sizeof(xlog_t), KM_SLEEP);
+ log = kmem_zalloc(sizeof(xlog_t), KM_MAYFAIL);
+ if (!log) {
+ xlog_warn("XFS: Log allocation failed: No memory!");
+ goto out;
+ }
log->l_mp = mp;
log->l_targ = log_target;
log->l_grant_reserve_cycle = 1;
log->l_grant_write_cycle = 1;
+ error = EFSCORRUPTED;
if (xfs_sb_version_hassector(&mp->m_sb)) {
log->l_sectbb_log = mp->m_sb.sb_logsectlog - BBSHIFT;
- ASSERT(log->l_sectbb_log <= mp->m_sectbb_log);
+ if (log->l_sectbb_log < 0 ||
+ log->l_sectbb_log > mp->m_sectbb_log) {
+ xlog_warn("XFS: Log sector size (0x%x) out of range.",
+ log->l_sectbb_log);
+ goto out_free_log;
+ }
+
/* for larger sector sizes, must have v2 or external log */
- ASSERT(log->l_sectbb_log == 0 ||
- log->l_logBBstart == 0 ||
- xfs_sb_version_haslogv2(&mp->m_sb));
- ASSERT(mp->m_sb.sb_logsectlog >= BBSHIFT);
+ if (log->l_sectbb_log != 0 &&
+ (log->l_logBBstart != 0 &&
+ !xfs_sb_version_haslogv2(&mp->m_sb))) {
+ xlog_warn("XFS: log sector size (0x%x) invalid "
+ "for configuration.", log->l_sectbb_log);
+ goto out_free_log;
+ }
+ if (mp->m_sb.sb_logsectlog < BBSHIFT) {
+ xlog_warn("XFS: Log sector log (0x%x) too small.",
+ mp->m_sb.sb_logsectlog);
+ goto out_free_log;
+ }
}
log->l_sectbb_mask = (1 << log->l_sectbb_log) - 1;
xlog_get_iclog_buffer_size(mp, log);
+ error = ENOMEM;
bp = xfs_buf_get_empty(log->l_iclog_size, mp->m_logdev_targp);
+ if (!bp)
+ goto out_free_log;
XFS_BUF_SET_IODONE_FUNC(bp, xlog_iodone);
XFS_BUF_SET_BDSTRAT_FUNC(bp, xlog_bdstrat_cb);
XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
spin_lock_init(&log->l_icloglock);
spin_lock_init(&log->l_grant_lock);
- initnsema(&log->l_flushsema, 0, "ic-flush");
+ sv_init(&log->l_flush_wait, 0, "flush_wait");
/* log record size must be multiple of BBSIZE; see xlog_rec_header_t */
ASSERT((XFS_BUF_SIZE(bp) & BBMASK) == 0);
iclogsize = log->l_iclog_size;
ASSERT(log->l_iclog_size >= 4096);
for (i=0; i < log->l_iclog_bufs; i++) {
- *iclogp = (xlog_in_core_t *)
- kmem_zalloc(sizeof(xlog_in_core_t), KM_SLEEP);
+ *iclogp = kmem_zalloc(sizeof(xlog_in_core_t), KM_MAYFAIL);
+ if (!*iclogp)
+ goto out_free_iclog;
+
iclog = *iclogp;
iclog->ic_prev = prev_iclog;
prev_iclog = iclog;
bp = xfs_buf_get_noaddr(log->l_iclog_size, mp->m_logdev_targp);
+ if (!bp)
+ goto out_free_iclog;
if (!XFS_BUF_CPSEMA(bp))
ASSERT(0);
XFS_BUF_SET_IODONE_FUNC(bp, xlog_iodone);
XFS_BUF_SET_BDSTRAT_FUNC(bp, xlog_bdstrat_cb);
XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
iclog->ic_bp = bp;
- iclog->hic_data = bp->b_addr;
+ iclog->ic_data = bp->b_addr;
#ifdef DEBUG
log->l_iclog_bak[i] = (xfs_caddr_t)&(iclog->ic_header);
#endif
atomic_set(&iclog->ic_refcnt, 0);
spin_lock_init(&iclog->ic_callback_lock);
iclog->ic_callback_tail = &(iclog->ic_callback);
- iclog->ic_datap = (char *)iclog->hic_data + log->l_iclog_hsize;
+ iclog->ic_datap = (char *)iclog->ic_data + log->l_iclog_hsize;
ASSERT(XFS_BUF_ISBUSY(iclog->ic_bp));
ASSERT(XFS_BUF_VALUSEMA(iclog->ic_bp) <= 0);
- sv_init(&iclog->ic_forcesema, SV_DEFAULT, "iclog-force");
- sv_init(&iclog->ic_writesema, SV_DEFAULT, "iclog-write");
+ sv_init(&iclog->ic_force_wait, SV_DEFAULT, "iclog-force");
+ sv_init(&iclog->ic_write_wait, SV_DEFAULT, "iclog-write");
iclogp = &iclog->ic_next;
}
log->l_iclog->ic_prev = prev_iclog; /* re-write 1st prev ptr */
return log;
+
+out_free_iclog:
+ for (iclog = log->l_iclog; iclog; iclog = prev_iclog) {
+ prev_iclog = iclog->ic_next;
+ if (iclog->ic_bp) {
+ sv_destroy(&iclog->ic_force_wait);
+ sv_destroy(&iclog->ic_write_wait);
+ xfs_buf_free(iclog->ic_bp);
+ }
+ kmem_free(iclog);
+ }
+ spinlock_destroy(&log->l_icloglock);
+ spinlock_destroy(&log->l_grant_lock);
+ xfs_buf_free(log->l_xbuf);
+out_free_log:
+ kmem_free(log);
+out:
+ return ERR_PTR(-error);
} /* xlog_alloc_log */
*/
if (threshold_lsn &&
!XLOG_FORCED_SHUTDOWN(log))
- xfs_trans_push_ail(mp, threshold_lsn);
+ xfs_trans_ail_push(log->l_ailp, threshold_lsn);
} /* xlog_grant_push_ail */
XFS_BUF_ZEROFLAGS(bp);
XFS_BUF_BUSY(bp);
XFS_BUF_ASYNC(bp);
+ bp->b_flags |= XBF_LOG_BUFFER;
/*
* Do an ordered write for the log block.
* Its unnecessary to flush the first split block in the log wrap case.
XFS_BUF_ZEROFLAGS(bp);
XFS_BUF_BUSY(bp);
XFS_BUF_ASYNC(bp);
+ bp->b_flags |= XBF_LOG_BUFFER;
if (log->l_mp->m_flags & XFS_MOUNT_BARRIER)
XFS_BUF_ORDERED(bp);
dptr = XFS_BUF_PTR(bp);
iclog = log->l_iclog;
for (i=0; i<log->l_iclog_bufs; i++) {
- sv_destroy(&iclog->ic_forcesema);
- sv_destroy(&iclog->ic_writesema);
+ sv_destroy(&iclog->ic_force_wait);
+ sv_destroy(&iclog->ic_write_wait);
xfs_buf_free(iclog->ic_bp);
-#ifdef XFS_LOG_TRACE
- if (iclog->ic_trace != NULL) {
- ktrace_free(iclog->ic_trace);
- }
-#endif
next_iclog = iclog->ic_next;
- kmem_free(iclog, sizeof(xlog_in_core_t));
+ kmem_free(iclog);
iclog = next_iclog;
}
- freesema(&log->l_flushsema);
spinlock_destroy(&log->l_icloglock);
spinlock_destroy(&log->l_grant_lock);
xfs_buf_free(log->l_xbuf);
-#ifdef XFS_LOG_TRACE
- if (log->l_trace != NULL) {
- ktrace_free(log->l_trace);
- }
- if (log->l_grant_trace != NULL) {
- ktrace_free(log->l_grant_trace);
- }
-#endif
log->l_mp->m_log = NULL;
- kmem_free(log, sizeof(xlog_t));
+ kmem_free(log);
} /* xlog_dealloc_log */
/*
if (iclog->ic_size - log_offset <= sizeof(xlog_op_header_t)) {
xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
record_cnt = data_cnt = 0;
+ spin_lock(&log->l_icloglock);
xlog_state_want_sync(log, iclog);
+ spin_unlock(&log->l_icloglock);
if (commit_iclog) {
ASSERT(flags & XLOG_COMMIT_TRANS);
*commit_iclog = iclog;
/* Clean iclogs starting from the head. This ordering must be
* maintained, so an iclog doesn't become ACTIVE beyond one that
* is SYNCING. This is also required to maintain the notion that we use
- * a counting semaphore to hold off would be writers to the log when every
+ * a ordered wait queue to hold off would be writers to the log when every
* iclog is trying to sync to disk.
*
* State Change: DIRTY -> ACTIVE
int funcdidcallbacks; /* flag: function did callbacks */
int repeats; /* for issuing console warnings if
* looping too many times */
+ int wake = 0;
spin_lock(&log->l_icloglock);
first_iclog = iclog = log->l_iclog;
xlog_state_clean_log(log);
/* wake up threads waiting in xfs_log_force() */
- sv_broadcast(&iclog->ic_forcesema);
+ sv_broadcast(&iclog->ic_force_wait);
iclog = iclog->ic_next;
} while (first_iclog != iclog);
repeats = 0;
xfs_fs_cmn_err(CE_WARN, log->l_mp,
"%s: possible infinite loop (%d iterations)",
- __FUNCTION__, flushcnt);
+ __func__, flushcnt);
}
} while (!ioerrors && loopdidcallbacks);
}
#endif
- flushcnt = 0;
- if (log->l_iclog->ic_state & (XLOG_STATE_ACTIVE|XLOG_STATE_IOERROR)) {
- flushcnt = log->l_flushcnt;
- log->l_flushcnt = 0;
- }
+ if (log->l_iclog->ic_state & (XLOG_STATE_ACTIVE|XLOG_STATE_IOERROR))
+ wake = 1;
spin_unlock(&log->l_icloglock);
- while (flushcnt--)
- vsema(&log->l_flushsema);
-} /* xlog_state_do_callback */
+
+ if (wake)
+ sv_broadcast(&log->l_flush_wait);
+}
/*
* the second completion goes through.
*
* Callbacks could take time, so they are done outside the scope of the
- * global state machine log lock. Assume that the calls to cvsema won't
- * take a long time. At least we know it won't sleep.
+ * global state machine log lock.
*/
STATIC void
xlog_state_done_syncing(
* iclog buffer, we wake them all, one will get to do the
* I/O, the others get to wait for the result.
*/
- sv_broadcast(&iclog->ic_writesema);
+ sv_broadcast(&iclog->ic_write_wait);
spin_unlock(&log->l_icloglock);
xlog_state_do_callback(log, aborted, iclog); /* also cleans log */
} /* xlog_state_done_syncing */
/*
* If the head of the in-core log ring is not (ACTIVE or DIRTY), then we must
- * sleep. The flush semaphore is set to the number of in-core buffers and
- * decremented around disk syncing. Therefore, if all buffers are syncing,
- * this semaphore will cause new writes to sleep until a sync completes.
- * Otherwise, this code just does p() followed by v(). This approximates
- * a sleep/wakeup except we can't race.
+ * sleep. We wait on the flush queue on the head iclog as that should be
+ * the first iclog to complete flushing. Hence if all iclogs are syncing,
+ * we will wait here and all new writes will sleep until a sync completes.
*
* The in-core logs are used in a circular fashion. They are not used
* out-of-order even when an iclog past the head is free.
}
iclog = log->l_iclog;
- if (! (iclog->ic_state == XLOG_STATE_ACTIVE)) {
- log->l_flushcnt++;
- spin_unlock(&log->l_icloglock);
- xlog_trace_iclog(iclog, XLOG_TRACE_SLEEP_FLUSH);
+ if (iclog->ic_state != XLOG_STATE_ACTIVE) {
XFS_STATS_INC(xs_log_noiclogs);
- /* Ensure that log writes happen */
- psema(&log->l_flushsema, PINOD);
+
+ /* Wait for log writes to have flushed */
+ sv_wait(&log->l_flush_wait, 0, &log->l_icloglock, 0);
goto restart;
}
- ASSERT(iclog->ic_state == XLOG_STATE_ACTIVE);
+
head = &iclog->ic_header;
atomic_inc(&iclog->ic_refcnt); /* prevents sync */
if (iclog->ic_size - iclog->ic_offset < 2*sizeof(xlog_op_header_t)) {
xlog_state_switch_iclogs(log, iclog, iclog->ic_size);
- /* If I'm the only one writing to this iclog, sync it to disk */
- if (atomic_read(&iclog->ic_refcnt) == 1) {
+ /*
+ * If I'm the only one writing to this iclog, sync it to disk.
+ * We need to do an atomic compare and decrement here to avoid
+ * racing with concurrent atomic_dec_and_lock() calls in
+ * xlog_state_release_iclog() when there is more than one
+ * reference to the iclog.
+ */
+ if (!atomic_add_unless(&iclog->ic_refcnt, -1, 1)) {
+ /* we are the only one */
spin_unlock(&log->l_icloglock);
- if ((error = xlog_state_release_iclog(log, iclog)))
+ error = xlog_state_release_iclog(log, iclog);
+ if (error)
return error;
} else {
- atomic_dec(&iclog->ic_refcnt);
spin_unlock(&log->l_icloglock);
}
goto restart;
/* Is there space or do we need to sleep? */
spin_lock(&log->l_grant_lock);
- xlog_trace_loggrant(log, tic, "xlog_grant_log_space: enter");
+
+ trace_xfs_log_grant_enter(log, tic);
/* something is already sleeping; insert new transaction at end */
if (log->l_reserve_headq) {
xlog_ins_ticketq(&log->l_reserve_headq, tic);
- xlog_trace_loggrant(log, tic,
- "xlog_grant_log_space: sleep 1");
+
+ trace_xfs_log_grant_sleep1(log, tic);
+
/*
* Gotta check this before going to sleep, while we're
* holding the grant lock.
goto error_return;
XFS_STATS_INC(xs_sleep_logspace);
- sv_wait(&tic->t_sema, PINOD|PLTWAIT, &log->l_grant_lock, s);
+ sv_wait(&tic->t_wait, PINOD|PLTWAIT, &log->l_grant_lock, s);
/*
* If we got an error, and the filesystem is shutting down,
* we'll catch it down below. So just continue...
*/
- xlog_trace_loggrant(log, tic,
- "xlog_grant_log_space: wake 1");
+ trace_xfs_log_grant_wake1(log, tic);
spin_lock(&log->l_grant_lock);
}
if (tic->t_flags & XFS_LOG_PERM_RESERV)
if (free_bytes < need_bytes) {
if ((tic->t_flags & XLOG_TIC_IN_Q) == 0)
xlog_ins_ticketq(&log->l_reserve_headq, tic);
- xlog_trace_loggrant(log, tic,
- "xlog_grant_log_space: sleep 2");
- XFS_STATS_INC(xs_sleep_logspace);
- sv_wait(&tic->t_sema, PINOD|PLTWAIT, &log->l_grant_lock, s);
- if (XLOG_FORCED_SHUTDOWN(log)) {
- spin_lock(&log->l_grant_lock);
- goto error_return;
- }
+ trace_xfs_log_grant_sleep2(log, tic);
- xlog_trace_loggrant(log, tic,
- "xlog_grant_log_space: wake 2");
+ spin_unlock(&log->l_grant_lock);
xlog_grant_push_ail(log->l_mp, need_bytes);
spin_lock(&log->l_grant_lock);
+
+ XFS_STATS_INC(xs_sleep_logspace);
+ sv_wait(&tic->t_wait, PINOD|PLTWAIT, &log->l_grant_lock, s);
+
+ spin_lock(&log->l_grant_lock);
+ if (XLOG_FORCED_SHUTDOWN(log))
+ goto error_return;
+
+ trace_xfs_log_grant_wake2(log, tic);
+
goto redo;
} else if (tic->t_flags & XLOG_TIC_IN_Q)
xlog_del_ticketq(&log->l_reserve_headq, tic);
ASSERT(log->l_grant_write_bytes <= BBTOB(BLOCK_LSN(tail_lsn)));
}
#endif
- xlog_trace_loggrant(log, tic, "xlog_grant_log_space: exit");
+ trace_xfs_log_grant_exit(log, tic);
xlog_verify_grant_head(log, 1);
spin_unlock(&log->l_grant_lock);
return 0;
error_return:
if (tic->t_flags & XLOG_TIC_IN_Q)
xlog_del_ticketq(&log->l_reserve_headq, tic);
- xlog_trace_loggrant(log, tic, "xlog_grant_log_space: err_ret");
+
+ trace_xfs_log_grant_error(log, tic);
+
/*
* If we are failing, make sure the ticket doesn't have any
* current reservations. We don't want to add this back when
#endif
spin_lock(&log->l_grant_lock);
- xlog_trace_loggrant(log, tic, "xlog_regrant_write_log_space: enter");
+
+ trace_xfs_log_regrant_write_enter(log, tic);
if (XLOG_FORCED_SHUTDOWN(log))
goto error_return;
* for more free space, otherwise try to get some space for
* this transaction.
*/
-
+ need_bytes = tic->t_unit_res;
if ((ntic = log->l_write_headq)) {
free_bytes = xlog_space_left(log, log->l_grant_write_cycle,
log->l_grant_write_bytes);
if (free_bytes < ntic->t_unit_res)
break;
free_bytes -= ntic->t_unit_res;
- sv_signal(&ntic->t_sema);
+ sv_signal(&ntic->t_wait);
ntic = ntic->t_next;
} while (ntic != log->l_write_headq);
if ((tic->t_flags & XLOG_TIC_IN_Q) == 0)
xlog_ins_ticketq(&log->l_write_headq, tic);
- xlog_trace_loggrant(log, tic,
- "xlog_regrant_write_log_space: sleep 1");
+ trace_xfs_log_regrant_write_sleep1(log, tic);
+
+ spin_unlock(&log->l_grant_lock);
+ xlog_grant_push_ail(log->l_mp, need_bytes);
+ spin_lock(&log->l_grant_lock);
+
XFS_STATS_INC(xs_sleep_logspace);
- sv_wait(&tic->t_sema, PINOD|PLTWAIT,
+ sv_wait(&tic->t_wait, PINOD|PLTWAIT,
&log->l_grant_lock, s);
/* If we're shutting down, this tic is already
* off the queue */
- if (XLOG_FORCED_SHUTDOWN(log)) {
- spin_lock(&log->l_grant_lock);
+ spin_lock(&log->l_grant_lock);
+ if (XLOG_FORCED_SHUTDOWN(log))
goto error_return;
- }
- xlog_trace_loggrant(log, tic,
- "xlog_regrant_write_log_space: wake 1");
- xlog_grant_push_ail(log->l_mp, tic->t_unit_res);
- spin_lock(&log->l_grant_lock);
+ trace_xfs_log_regrant_write_wake1(log, tic);
}
}
- need_bytes = tic->t_unit_res;
-
redo:
if (XLOG_FORCED_SHUTDOWN(log))
goto error_return;
if (free_bytes < need_bytes) {
if ((tic->t_flags & XLOG_TIC_IN_Q) == 0)
xlog_ins_ticketq(&log->l_write_headq, tic);
+ spin_unlock(&log->l_grant_lock);
+ xlog_grant_push_ail(log->l_mp, need_bytes);
+ spin_lock(&log->l_grant_lock);
+
XFS_STATS_INC(xs_sleep_logspace);
- sv_wait(&tic->t_sema, PINOD|PLTWAIT, &log->l_grant_lock, s);
+ trace_xfs_log_regrant_write_sleep2(log, tic);
+
+ sv_wait(&tic->t_wait, PINOD|PLTWAIT, &log->l_grant_lock, s);
/* If we're shutting down, this tic is already off the queue */
- if (XLOG_FORCED_SHUTDOWN(log)) {
- spin_lock(&log->l_grant_lock);
+ spin_lock(&log->l_grant_lock);
+ if (XLOG_FORCED_SHUTDOWN(log))
goto error_return;
- }
- xlog_trace_loggrant(log, tic,
- "xlog_regrant_write_log_space: wake 2");
- xlog_grant_push_ail(log->l_mp, need_bytes);
- spin_lock(&log->l_grant_lock);
+ trace_xfs_log_regrant_write_wake2(log, tic);
goto redo;
} else if (tic->t_flags & XLOG_TIC_IN_Q)
xlog_del_ticketq(&log->l_write_headq, tic);
}
#endif
- xlog_trace_loggrant(log, tic, "xlog_regrant_write_log_space: exit");
+ trace_xfs_log_regrant_write_exit(log, tic);
+
xlog_verify_grant_head(log, 1);
spin_unlock(&log->l_grant_lock);
return 0;
error_return:
if (tic->t_flags & XLOG_TIC_IN_Q)
xlog_del_ticketq(&log->l_reserve_headq, tic);
- xlog_trace_loggrant(log, tic, "xlog_regrant_write_log_space: err_ret");
+
+ trace_xfs_log_regrant_write_error(log, tic);
+
/*
* If we are failing, make sure the ticket doesn't have any
* current reservations. We don't want to add this back when
xlog_regrant_reserve_log_space(xlog_t *log,
xlog_ticket_t *ticket)
{
- xlog_trace_loggrant(log, ticket,
- "xlog_regrant_reserve_log_space: enter");
+ trace_xfs_log_regrant_reserve_enter(log, ticket);
+
if (ticket->t_cnt > 0)
ticket->t_cnt--;
xlog_grant_sub_space(log, ticket->t_curr_res);
ticket->t_curr_res = ticket->t_unit_res;
xlog_tic_reset_res(ticket);
- xlog_trace_loggrant(log, ticket,
- "xlog_regrant_reserve_log_space: sub current res");
+
+ trace_xfs_log_regrant_reserve_sub(log, ticket);
+
xlog_verify_grant_head(log, 1);
/* just return if we still have some of the pre-reserved space */
}
xlog_grant_add_space_reserve(log, ticket->t_unit_res);
- xlog_trace_loggrant(log, ticket,
- "xlog_regrant_reserve_log_space: exit");
+
+ trace_xfs_log_regrant_reserve_exit(log, ticket);
+
xlog_verify_grant_head(log, 0);
spin_unlock(&log->l_grant_lock);
ticket->t_curr_res = ticket->t_unit_res;
ticket->t_cnt--;
spin_lock(&log->l_grant_lock);
- xlog_trace_loggrant(log, ticket, "xlog_ungrant_log_space: enter");
+ trace_xfs_log_ungrant_enter(log, ticket);
xlog_grant_sub_space(log, ticket->t_curr_res);
- xlog_trace_loggrant(log, ticket, "xlog_ungrant_log_space: sub current");
+ trace_xfs_log_ungrant_sub(log, ticket);
/* If this is a permanent reservation ticket, we may be able to free
* up more space based on the remaining count.
xlog_grant_sub_space(log, ticket->t_unit_res*ticket->t_cnt);
}
- xlog_trace_loggrant(log, ticket, "xlog_ungrant_log_space: exit");
+ trace_xfs_log_ungrant_exit(log, ticket);
+
xlog_verify_grant_head(log, 1);
spin_unlock(&log->l_grant_lock);
xfs_log_move_tail(log->l_mp, 1);
* 2. the current iclog is drity, and the previous iclog is in the
* active or dirty state.
*
- * We may sleep (call psema) if:
+ * We may sleep if:
*
* 1. the current iclog is not in the active nor dirty state.
* 2. the current iclog dirty, and the previous iclog is not in the
return XFS_ERROR(EIO);
}
XFS_STATS_INC(xs_log_force_sleep);
- sv_wait(&iclog->ic_forcesema, PINOD, &log->l_icloglock, s);
+ sv_wait(&iclog->ic_force_wait, PINOD, &log->l_icloglock, s);
/*
* No need to grab the log lock here since we're
* only deciding whether or not to return EIO
XLOG_STATE_SYNCING))) {
ASSERT(!(iclog->ic_state & XLOG_STATE_IOERROR));
XFS_STATS_INC(xs_log_force_sleep);
- sv_wait(&iclog->ic_prev->ic_writesema, PSWP,
+ sv_wait(&iclog->ic_prev->ic_write_wait, PSWP,
&log->l_icloglock, s);
*log_flushed = 1;
already_slept = 1;
!(iclog->ic_state & (XLOG_STATE_ACTIVE | XLOG_STATE_DIRTY))) {
/*
- * Don't wait on the forcesema if we know that we've
+ * Don't wait on completion if we know that we've
* gotten a log write error.
*/
if (iclog->ic_state & XLOG_STATE_IOERROR) {
return XFS_ERROR(EIO);
}
XFS_STATS_INC(xs_log_force_sleep);
- sv_wait(&iclog->ic_forcesema, PSWP, &log->l_icloglock, s);
+ sv_wait(&iclog->ic_force_wait, PSWP, &log->l_icloglock, s);
/*
* No need to grab the log lock here since we're
* only deciding whether or not to return EIO
STATIC void
xlog_state_want_sync(xlog_t *log, xlog_in_core_t *iclog)
{
- spin_lock(&log->l_icloglock);
+ assert_spin_locked(&log->l_icloglock);
if (iclog->ic_state == XLOG_STATE_ACTIVE) {
xlog_state_switch_iclogs(log, iclog, 0);
ASSERT(iclog->ic_state &
(XLOG_STATE_WANT_SYNC|XLOG_STATE_IOERROR));
}
-
- spin_unlock(&log->l_icloglock);
-} /* xlog_state_want_sync */
-
+}
/*****************************************************************************
*/
/*
- * Free a used ticket.
+ * Free a used ticket when its refcount falls to zero.
*/
-STATIC void
-xlog_ticket_put(xlog_t *log,
- xlog_ticket_t *ticket)
+void
+xfs_log_ticket_put(
+ xlog_ticket_t *ticket)
{
- sv_destroy(&ticket->t_sema);
- kmem_zone_free(xfs_log_ticket_zone, ticket);
-} /* xlog_ticket_put */
+ ASSERT(atomic_read(&ticket->t_ref) > 0);
+ if (atomic_dec_and_test(&ticket->t_ref)) {
+ sv_destroy(&ticket->t_wait);
+ kmem_zone_free(xfs_log_ticket_zone, ticket);
+ }
+}
+xlog_ticket_t *
+xfs_log_ticket_get(
+ xlog_ticket_t *ticket)
+{
+ ASSERT(atomic_read(&ticket->t_ref) > 0);
+ atomic_inc(&ticket->t_ref);
+ return ticket;
+}
/*
* Allocate and initialise a new log ticket.
*/
STATIC xlog_ticket_t *
-xlog_ticket_get(xlog_t *log,
+xlog_ticket_alloc(xlog_t *log,
int unit_bytes,
int cnt,
char client,
unit_bytes += 2*BBSIZE;
}
+ atomic_set(&tic->t_ref, 1);
tic->t_unit_res = unit_bytes;
tic->t_curr_res = unit_bytes;
tic->t_cnt = cnt;
tic->t_trans_type = 0;
if (xflags & XFS_LOG_PERM_RESERV)
tic->t_flags |= XLOG_TIC_PERM_RESERV;
- sv_init(&(tic->t_sema), SV_DEFAULT, "logtick");
+ sv_init(&(tic->t_wait), SV_DEFAULT, "logtick");
xlog_tic_reset_res(tic);
return tic;
-} /* xlog_ticket_get */
+}
/******************************************************************************
ptr = iclog->ic_datap;
base_ptr = ptr;
ophead = (xlog_op_header_t *)ptr;
- xhdr = (xlog_in_core_2_t *)&iclog->ic_header;
+ xhdr = iclog->ic_data;
for (i = 0; i < len; i++) {
ophead = (xlog_op_header_t *)ptr;
if (!log ||
log->l_flags & XLOG_ACTIVE_RECOVERY) {
mp->m_flags |= XFS_MOUNT_FS_SHUTDOWN;
- XFS_BUF_DONE(mp->m_sb_bp);
+ if (mp->m_sb_bp)
+ XFS_BUF_DONE(mp->m_sb_bp);
return 0;
}
* before we mark the filesystem SHUTDOWN and wake
* everybody up to tell the bad news.
*/
- spin_lock(&log->l_grant_lock);
spin_lock(&log->l_icloglock);
+ spin_lock(&log->l_grant_lock);
mp->m_flags |= XFS_MOUNT_FS_SHUTDOWN;
- XFS_BUF_DONE(mp->m_sb_bp);
+ if (mp->m_sb_bp)
+ XFS_BUF_DONE(mp->m_sb_bp);
+
/*
* This flag is sort of redundant because of the mount flag, but
* it's good to maintain the separation between the log and the rest
*/
if ((tic = log->l_reserve_headq)) {
do {
- sv_signal(&tic->t_sema);
+ sv_signal(&tic->t_wait);
tic = tic->t_next;
} while (tic != log->l_reserve_headq);
}
if ((tic = log->l_write_headq)) {
do {
- sv_signal(&tic->t_sema);
+ sv_signal(&tic->t_wait);
tic = tic->t_next;
} while (tic != log->l_write_headq);
}