#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_rw.h"
+#include "xfs_trace.h"
-
-#define xlog_write_adv_cnt(ptr, len, off, bytes) \
- { (ptr) += (bytes); \
- (len) -= (bytes); \
- (off) += (bytes);}
+kmem_zone_t *xfs_log_ticket_zone;
/* Local miscellaneous function prototypes */
-STATIC int xlog_bdstrat_cb(struct xfs_buf *);
-STATIC int xlog_commit_record(xfs_mount_t *mp, xlog_ticket_t *ticket,
+STATIC int xlog_commit_record(struct log *log, struct xlog_ticket *ticket,
xlog_in_core_t **, xfs_lsn_t *);
STATIC xlog_t * xlog_alloc_log(xfs_mount_t *mp,
xfs_buftarg_t *log_target,
STATIC int xlog_space_left(xlog_t *log, int cycle, int bytes);
STATIC int xlog_sync(xlog_t *log, xlog_in_core_t *iclog);
STATIC void xlog_dealloc_log(xlog_t *log);
-STATIC int xlog_write(xfs_mount_t *mp, xfs_log_iovec_t region[],
- int nentries, xfs_log_ticket_t tic,
- xfs_lsn_t *start_lsn,
- xlog_in_core_t **commit_iclog,
- uint flags);
+STATIC int xlog_write(struct log *log, struct xfs_log_vec *log_vector,
+ struct xlog_ticket *tic, xfs_lsn_t *start_lsn,
+ xlog_in_core_t **commit_iclog, uint flags);
/* local state machine functions */
STATIC void xlog_state_done_syncing(xlog_in_core_t *iclog, int);
xlog_ticket_t *ticket,
int *continued_write,
int *logoffsetp);
-STATIC void xlog_state_put_ticket(xlog_t *log,
- xlog_ticket_t *tic);
STATIC int xlog_state_release_iclog(xlog_t *log,
xlog_in_core_t *iclog);
STATIC void xlog_state_switch_iclogs(xlog_t *log,
xlog_in_core_t *iclog,
int eventual_size);
-STATIC int xlog_state_sync(xlog_t *log,
- xfs_lsn_t lsn,
- uint flags,
- int *log_flushed);
-STATIC int xlog_state_sync_all(xlog_t *log, uint flags, int *log_flushed);
STATIC void xlog_state_want_sync(xlog_t *log, xlog_in_core_t *iclog);
/* local functions to manipulate grant head */
/* local ticket functions */
-STATIC void xlog_state_ticket_alloc(xlog_t *log);
-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 void xlog_verify_dest_ptr(xlog_t *log, char *ptr);
STATIC void xlog_verify_grant_head(xlog_t *log, int equals);
STATIC void xlog_verify_iclog(xlog_t *log, xlog_in_core_t *iclog,
int count, boolean_t syncing);
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;
}
}
* out when the next write occurs.
*/
xfs_lsn_t
-xfs_log_done(xfs_mount_t *mp,
- xfs_log_ticket_t xtic,
- void **iclog,
- uint flags)
+xfs_log_done(
+ struct xfs_mount *mp,
+ struct xlog_ticket *ticket,
+ struct xlog_in_core **iclog,
+ uint flags)
{
- xlog_t *log = mp->m_log;
- xlog_ticket_t *ticket = (xfs_log_ticket_t) xtic;
- xfs_lsn_t lsn = 0;
+ struct log *log = mp->m_log;
+ xfs_lsn_t lsn = 0;
if (XLOG_FORCED_SHUTDOWN(log) ||
/*
* If we get an error, just continue and give back the log ticket.
*/
(((ticket->t_flags & XLOG_TIC_INITED) == 0) &&
- (xlog_commit_record(mp, ticket,
- (xlog_in_core_t **)iclog, &lsn)))) {
+ (xlog_commit_record(log, ticket, iclog, &lsn)))) {
lsn = (xfs_lsn_t) -1;
if (ticket->t_flags & XLOG_TIC_PERM_RESERV) {
flags |= XFS_LOG_REL_PERM_RESERV;
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_state_put_ticket(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 */
-
-
-/*
- * Force the in-core log to disk. If flags == XFS_LOG_SYNC,
- * the force is done synchronously.
- *
- * 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
- * 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.
- */
-int
-_xfs_log_force(
- xfs_mount_t *mp,
- xfs_lsn_t lsn,
- uint flags,
- int *log_flushed)
-{
- xlog_t *log = mp->m_log;
- int dummy;
-
- if (!log_flushed)
- log_flushed = &dummy;
-
- ASSERT(flags & XFS_LOG_FORCE);
-
- XFS_STATS_INC(xs_log_force);
-
- if (log->l_flags & XLOG_IO_ERROR)
- return XFS_ERROR(EIO);
- if (lsn == 0)
- return xlog_state_sync_all(log, flags, log_flushed);
- else
- return xlog_state_sync(log, lsn, flags, log_flushed);
-} /* xfs_log_force */
+}
/*
* Attaches a new iclog I/O completion callback routine during
* executing the callback at an appropriate time.
*/
int
-xfs_log_notify(xfs_mount_t *mp, /* mount of partition */
- void *iclog_hndl, /* iclog to hang callback off */
- xfs_log_callback_t *cb)
+xfs_log_notify(
+ struct xfs_mount *mp,
+ struct xlog_in_core *iclog,
+ xfs_log_callback_t *cb)
{
- xlog_t *log = mp->m_log;
- xlog_in_core_t *iclog = (xlog_in_core_t *)iclog_hndl;
int abortflg;
- cb->cb_next = NULL;
- spin_lock(&log->l_icloglock);
+ spin_lock(&iclog->ic_callback_lock);
abortflg = (iclog->ic_state & XLOG_STATE_IOERROR);
if (!abortflg) {
ASSERT_ALWAYS((iclog->ic_state == XLOG_STATE_ACTIVE) ||
*(iclog->ic_callback_tail) = cb;
iclog->ic_callback_tail = &(cb->cb_next);
}
- spin_unlock(&log->l_icloglock);
+ spin_unlock(&iclog->ic_callback_lock);
return abortflg;
-} /* xfs_log_notify */
+}
int
-xfs_log_release_iclog(xfs_mount_t *mp,
- void *iclog_hndl)
+xfs_log_release_iclog(
+ struct xfs_mount *mp,
+ struct xlog_in_core *iclog)
{
- xlog_t *log = mp->m_log;
- xlog_in_core_t *iclog = (xlog_in_core_t *)iclog_hndl;
-
- if (xlog_state_release_iclog(log, iclog)) {
+ if (xlog_state_release_iclog(mp->m_log, iclog)) {
xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
return EIO;
}
* reservation, we prevent over allocation problems.
*/
int
-xfs_log_reserve(xfs_mount_t *mp,
- int unit_bytes,
- int cnt,
- xfs_log_ticket_t *ticket,
- __uint8_t client,
- uint flags,
- uint t_type)
+xfs_log_reserve(
+ struct xfs_mount *mp,
+ int unit_bytes,
+ int cnt,
+ struct xlog_ticket **ticket,
+ __uint8_t client,
+ uint flags,
+ uint t_type)
{
- xlog_t *log = mp->m_log;
- xlog_ticket_t *internal_ticket;
- int retval = 0;
+ struct log *log = mp->m_log;
+ struct xlog_ticket *internal_ticket;
+ int retval = 0;
ASSERT(client == XFS_TRANSACTION || client == XFS_LOG);
ASSERT((flags & XFS_LOG_NOSLEEP) == 0);
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)");
+ internal_ticket = *ticket;
+
+ 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));
* Return error or zero.
*/
int
-xfs_log_mount(xfs_mount_t *mp,
- xfs_buftarg_t *log_target,
- xfs_daddr_t blk_offset,
- int num_bblks)
+xfs_log_mount(
+ xfs_mount_t *mp,
+ xfs_buftarg_t *log_target,
+ xfs_daddr_t blk_offset,
+ int num_bblks)
{
+ int error;
+
if (!(mp->m_flags & XFS_MOUNT_NORECOVERY))
cmn_err(CE_NOTE, "XFS mounting filesystem %s", mp->m_fsname);
else {
}
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.
+ */
+ error = xfs_trans_ail_init(mp);
+ if (error) {
+ cmn_err(CE_WARN, "XFS: AIL initialisation failed: error %d", error);
+ goto out_free_log;
+ }
+ mp->m_log->l_ailp = mp->m_ail;
/*
* skip log recovery on a norecovery mount. pretend it all
* just worked.
*/
if (!(mp->m_flags & XFS_MOUNT_NORECOVERY)) {
- int error, readonly = (mp->m_flags & XFS_MOUNT_RDONLY);
+ int readonly = (mp->m_flags & XFS_MOUNT_RDONLY);
if (readonly)
mp->m_flags &= ~XFS_MOUNT_RDONLY;
mp->m_flags |= XFS_MOUNT_RDONLY;
if (error) {
cmn_err(CE_WARN, "XFS: log mount/recovery failed: error %d", error);
- xlog_dealloc_log(mp->m_log);
- return 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;
-} /* xfs_log_mount */
+
+out_destroy_ail:
+ xfs_trans_ail_destroy(mp);
+out_free_log:
+ xlog_dealloc_log(mp->m_log);
+out:
+ return error;
+}
/*
* 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
#ifdef DEBUG
xlog_in_core_t *first_iclog;
#endif
- xfs_log_iovec_t reg[1];
- xfs_log_ticket_t tic = NULL;
+ xlog_ticket_t *tic = NULL;
xfs_lsn_t lsn;
int error;
- /* the data section must be 32 bit size aligned */
- struct {
- __uint16_t magic;
- __uint16_t pad1;
- __uint32_t pad2; /* may as well make it 64 bits */
- } magic = { XLOG_UNMOUNT_TYPE, 0, 0 };
-
/*
* Don't write out unmount record on read-only mounts.
* Or, if we are doing a forced umount (typically because of IO errors).
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, XFS_LOG_SYNC, NULL);
+ ASSERT(error || !(XLOG_FORCED_SHUTDOWN(log)));
#ifdef DEBUG
first_iclog = iclog = log->l_iclog;
} while (iclog != first_iclog);
#endif
if (! (XLOG_FORCED_SHUTDOWN(log))) {
- reg[0].i_addr = (void*)&magic;
- reg[0].i_len = sizeof(magic);
- XLOG_VEC_SET_TYPE(®[0], XLOG_REG_TYPE_UNMOUNT);
-
error = xfs_log_reserve(mp, 600, 1, &tic,
XFS_LOG, 0, XLOG_UNMOUNT_REC_TYPE);
if (!error) {
+ /* the data section must be 32 bit size aligned */
+ struct {
+ __uint16_t magic;
+ __uint16_t pad1;
+ __uint32_t pad2; /* may as well make it 64 bits */
+ } magic = {
+ .magic = XLOG_UNMOUNT_TYPE,
+ };
+ struct xfs_log_iovec reg = {
+ .i_addr = (void *)&magic,
+ .i_len = sizeof(magic),
+ .i_type = XLOG_REG_TYPE_UNMOUNT,
+ };
+ struct xfs_log_vec vec = {
+ .lv_niovecs = 1,
+ .lv_iovecp = ®,
+ };
+
/* remove inited flag */
- ((xlog_ticket_t *)tic)->t_flags = 0;
- error = xlog_write(mp, reg, 1, tic, &lsn,
+ tic->t_flags = 0;
+ error = xlog_write(log, &vec, tic, &lsn,
NULL, XLOG_UNMOUNT_TRANS);
/*
* At this point, we're umounting anyway,
spin_lock(&log->l_icloglock);
iclog = log->l_iclog;
- iclog->ic_refcnt++;
- spin_unlock(&log->l_icloglock);
+ atomic_inc(&iclog->ic_refcnt);
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_state_put_ticket(log, tic);
+ xfs_log_ticket_put(tic);
}
} else {
/*
*/
spin_lock(&log->l_icloglock);
iclog = log->l_iclog;
- iclog->ic_refcnt++;
- spin_unlock(&log->l_icloglock);
+ atomic_inc(&iclog->ic_refcnt);
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 */
/*
* Deallocate log structures for unmount/relocation.
+ *
+ * We need to stop the aild from running before we destroy
+ * 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);
}
+void
+xfs_log_item_init(
+ struct xfs_mount *mp,
+ struct xfs_log_item *item,
+ int type,
+ struct xfs_item_ops *ops)
+{
+ item->li_mountp = mp;
+ item->li_ailp = mp->m_ail;
+ item->li_type = type;
+ item->li_ops = ops;
+}
+
/*
* Write region vectors to log. The write happens using the space reservation
* of the ticket (tic). It is not a requirement that all writes for a given
- * transaction occur with one call to xfs_log_write().
+ * transaction occur with one call to xfs_log_write(). However, it is important
+ * to note that the transaction reservation code makes an assumption about the
+ * number of log headers a transaction requires that may be violated if you
+ * don't pass all the transaction vectors in one call....
*/
int
-xfs_log_write(xfs_mount_t * mp,
- xfs_log_iovec_t reg[],
- int nentries,
- xfs_log_ticket_t tic,
- xfs_lsn_t *start_lsn)
+xfs_log_write(
+ struct xfs_mount *mp,
+ struct xfs_log_iovec reg[],
+ int nentries,
+ struct xlog_ticket *tic,
+ xfs_lsn_t *start_lsn)
{
- int error;
- xlog_t *log = mp->m_log;
+ struct log *log = mp->m_log;
+ int error;
+ struct xfs_log_vec vec = {
+ .lv_niovecs = nentries,
+ .lv_iovecp = reg,
+ };
if (XLOG_FORCED_SHUTDOWN(log))
return XFS_ERROR(EIO);
- if ((error = xlog_write(mp, reg, nentries, tic, start_lsn, NULL, 0))) {
+ error = xlog_write(log, &vec, tic, start_lsn, NULL, 0);
+ if (error)
xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
- }
return error;
-} /* xfs_log_write */
-
+}
void
xfs_log_move_tail(xfs_mount_t *mp,
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);
}
/*
* Determine if we have a transaction that has gone to disk
- * that needs to be covered. Log activity needs to be idle (no AIL and
- * nothing in the iclogs). And, we need to be in the right state indicating
- * something has gone out.
+ * that needs to be covered. To begin the transition to the idle state
+ * firstly the log needs to be idle (no AIL and nothing in the iclogs).
+ * 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.
+ *
+ * Because this is called as part of the sync process, we should also indicate
+ * that dummy transactions should be issued in anything but the covered or
+ * idle states. This ensures that the log tail is accurately reflected in
+ * the log at the end of the sync, hence if a crash occurrs avoids replay
+ * of transactions where the metadata is already on disk.
*/
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))
return 0;
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)
- && xlog_iclogs_empty(log)) {
- if (log->l_covered_state == XLOG_STATE_COVER_NEED)
- log->l_covered_state = XLOG_STATE_COVER_DONE;
- else {
- ASSERT(log->l_covered_state == XLOG_STATE_COVER_NEED2);
- log->l_covered_state = XLOG_STATE_COVER_DONE2;
+ 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_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;
+ else
+ log->l_covered_state = XLOG_STATE_COVER_DONE2;
}
+ /* FALLTHRU */
+ default:
needed = 1;
+ break;
}
spin_unlock(&log->l_icloglock);
return needed;
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);
}
/*
} /* xlog_iodone */
/*
- * The bdstrat callback function for log bufs. This gives us a central
- * place to trap bufs in case we get hit by a log I/O error and need to
- * shutdown. Actually, in practice, even when we didn't get a log error,
- * we transition the iclogs to IOERROR state *after* flushing all existing
- * iclogs to disk. This is because we don't want anymore new transactions to be
- * started or completed afterwards.
- */
-STATIC int
-xlog_bdstrat_cb(struct xfs_buf *bp)
-{
- xlog_in_core_t *iclog;
-
- iclog = XFS_BUF_FSPRIVATE(bp, xlog_in_core_t *);
-
- if ((iclog->ic_state & XLOG_STATE_IOERROR) == 0) {
- /* note for irix bstrat will need struct bdevsw passed
- * Fix the following macro if the code ever is merged
- */
- XFS_bdstrat(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.
size >>= 1;
}
- if (XFS_SB_VERSION_HASLOGV2(&mp->m_sb)) {
- /* # headers = size / 32K
- * one header holds cycles from 32K of data
+ if (xfs_sb_version_haslogv2(&mp->m_sb)) {
+ /* # 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;
- if (XFS_SB_VERSION_HASSECTOR(&mp->m_sb)) {
+ 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);
ASSERT(XFS_BUF_ISBUSY(bp));
ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
spin_lock_init(&log->l_icloglock);
spin_lock_init(&log->l_grant_lock);
- initnsema(&log->l_flushsema, 0, "ic-flush");
- xlog_state_ticket_alloc(log); /* wait until after icloglock inited */
+ 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
head = &iclog->ic_header;
memset(head, 0, sizeof(xlog_rec_header_t));
head->h_magicno = cpu_to_be32(XLOG_HEADER_MAGIC_NUM);
head->h_version = cpu_to_be32(
- XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? 2 : 1);
+ xfs_sb_version_haslogv2(&log->l_mp->m_sb) ? 2 : 1);
head->h_size = cpu_to_be32(log->l_iclog_size);
/* new fields */
head->h_fmt = cpu_to_be32(XLOG_FMT);
memcpy(&head->h_fs_uuid, &mp->m_sb.sb_uuid, sizeof(uuid_t));
-
iclog->ic_size = XFS_BUF_SIZE(bp) - log->l_iclog_hsize;
iclog->ic_state = XLOG_STATE_ACTIVE;
iclog->ic_log = log;
+ 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 */
* ticket. Return the lsn of the commit record.
*/
STATIC int
-xlog_commit_record(xfs_mount_t *mp,
- xlog_ticket_t *ticket,
- xlog_in_core_t **iclog,
- xfs_lsn_t *commitlsnp)
+xlog_commit_record(
+ struct log *log,
+ struct xlog_ticket *ticket,
+ struct xlog_in_core **iclog,
+ xfs_lsn_t *commitlsnp)
{
- int error;
- xfs_log_iovec_t reg[1];
-
- reg[0].i_addr = NULL;
- reg[0].i_len = 0;
- XLOG_VEC_SET_TYPE(®[0], XLOG_REG_TYPE_COMMIT);
+ struct xfs_mount *mp = log->l_mp;
+ int error;
+ struct xfs_log_iovec reg = {
+ .i_addr = NULL,
+ .i_len = 0,
+ .i_type = XLOG_REG_TYPE_COMMIT,
+ };
+ struct xfs_log_vec vec = {
+ .lv_niovecs = 1,
+ .lv_iovecp = ®,
+ };
ASSERT_ALWAYS(iclog);
- if ((error = xlog_write(mp, reg, 1, ticket, commitlsnp,
- iclog, XLOG_COMMIT_TRANS))) {
+ error = xlog_write(log, &vec, ticket, commitlsnp, iclog,
+ XLOG_COMMIT_TRANS);
+ if (error)
xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
- }
return error;
-} /* xlog_commit_record */
-
+}
/*
* Push on the buffer cache code if we ever use more than 75% of the on-disk
*/
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 */
+/*
+ * The bdstrat callback function for log bufs. This gives us a central
+ * place to trap bufs in case we get hit by a log I/O error and need to
+ * shutdown. Actually, in practice, even when we didn't get a log error,
+ * we transition the iclogs to IOERROR state *after* flushing all existing
+ * iclogs to disk. This is because we don't want anymore new transactions to be
+ * started or completed afterwards.
+ */
+STATIC int
+xlog_bdstrat(
+ struct xfs_buf *bp)
+{
+ struct xlog_in_core *iclog;
+
+ iclog = XFS_BUF_FSPRIVATE(bp, xlog_in_core_t *);
+ if (iclog->ic_state & XLOG_STATE_IOERROR) {
+ XFS_BUF_ERROR(bp, EIO);
+ XFS_BUF_STALE(bp);
+ xfs_biodone(bp);
+ /*
+ * 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.
+ */
+ return 0;
+ }
+
+ bp->b_flags |= _XBF_RUN_QUEUES;
+ xfs_buf_iorequest(bp);
+ return 0;
+}
/*
* Flush out the in-core log (iclog) to the on-disk log in an asynchronous
int roundoff; /* roundoff to BB or stripe */
int split = 0; /* split write into two regions */
int error;
- int v2 = XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb);
+ int v2 = xfs_sb_version_haslogv2(&log->l_mp->m_sb);
XFS_STATS_INC(xs_log_writes);
- ASSERT(iclog->ic_refcnt == 0);
+ ASSERT(atomic_read(&iclog->ic_refcnt) == 0);
/* Add for LR header */
count_init = log->l_iclog_hsize + iclog->ic_offset;
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_WRITE(bp);
- if ((error = XFS_bwrite(bp))) {
+ if ((error = xlog_bdstrat(bp))) {
xfs_ioerror_alert("xlog_sync", log->l_mp, bp,
XFS_BUF_ADDR(bp));
return error;
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);
* case, though.
*/
for (i = 0; i < split; i += BBSIZE) {
- be32_add((__be32 *)dptr, 1);
+ be32_add_cpu((__be32 *)dptr, 1);
if (be32_to_cpu(*(__be32 *)dptr) == XLOG_HEADER_MAGIC_NUM)
- be32_add((__be32 *)dptr, 1);
+ be32_add_cpu((__be32 *)dptr, 1);
dptr += BBSIZE;
}
/* account for internal log which doesn't start at block #0 */
XFS_BUF_SET_ADDR(bp, XFS_BUF_ADDR(bp) + log->l_logBBstart);
XFS_BUF_WRITE(bp);
- if ((error = XFS_bwrite(bp))) {
+ if ((error = xlog_bdstrat(bp))) {
xfs_ioerror_alert("xlog_sync (split)", log->l_mp,
bp, XFS_BUF_ADDR(bp));
return error;
xlog_dealloc_log(xlog_t *log)
{
xlog_in_core_t *iclog, *next_iclog;
- xlog_ticket_t *tic, *next_tic;
int i;
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);
- /* XXXsup take a look at this again. */
- if ((log->l_ticket_cnt != log->l_ticket_tcnt) &&
- !XLOG_FORCED_SHUTDOWN(log)) {
- xfs_fs_cmn_err(CE_WARN, log->l_mp,
- "xlog_dealloc_log: (cnt: %d, total: %d)",
- log->l_ticket_cnt, log->l_ticket_tcnt);
- /* ASSERT(log->l_ticket_cnt == log->l_ticket_tcnt); */
-
- } else {
- tic = log->l_unmount_free;
- while (tic) {
- next_tic = tic->t_next;
- kmem_free(tic, NBPP);
- tic = next_tic;
- }
- }
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 */
/*
{
spin_lock(&log->l_icloglock);
- be32_add(&iclog->ic_header.h_num_logops, record_cnt);
+ be32_add_cpu(&iclog->ic_header.h_num_logops, record_cnt);
iclog->ic_offset += copy_bytes;
spin_unlock(&log->l_icloglock);
}
/*
+ * Calculate the potential space needed by the log vector. Each region gets
+ * its own xlog_op_header_t and may need to be double word aligned.
+ */
+static int
+xlog_write_calc_vec_length(
+ struct xlog_ticket *ticket,
+ struct xfs_log_vec *log_vector)
+{
+ struct xfs_log_vec *lv;
+ int headers = 0;
+ int len = 0;
+ int i;
+
+ /* acct for start rec of xact */
+ if (ticket->t_flags & XLOG_TIC_INITED)
+ headers++;
+
+ for (lv = log_vector; lv; lv = lv->lv_next) {
+ headers += lv->lv_niovecs;
+
+ for (i = 0; i < lv->lv_niovecs; i++) {
+ struct xfs_log_iovec *vecp = &lv->lv_iovecp[i];
+
+ len += vecp->i_len;
+ xlog_tic_add_region(ticket, vecp->i_len, vecp->i_type);
+ }
+ }
+
+ ticket->t_res_num_ophdrs += headers;
+ len += headers * sizeof(struct xlog_op_header);
+
+ return len;
+}
+
+/*
+ * If first write for transaction, insert start record We can't be trying to
+ * commit if we are inited. We can't have any "partial_copy" if we are inited.
+ */
+static int
+xlog_write_start_rec(
+ struct xlog_op_header *ophdr,
+ struct xlog_ticket *ticket)
+{
+ if (!(ticket->t_flags & XLOG_TIC_INITED))
+ return 0;
+
+ ophdr->oh_tid = cpu_to_be32(ticket->t_tid);
+ ophdr->oh_clientid = ticket->t_clientid;
+ ophdr->oh_len = 0;
+ ophdr->oh_flags = XLOG_START_TRANS;
+ ophdr->oh_res2 = 0;
+
+ ticket->t_flags &= ~XLOG_TIC_INITED;
+
+ return sizeof(struct xlog_op_header);
+}
+
+static xlog_op_header_t *
+xlog_write_setup_ophdr(
+ struct log *log,
+ struct xlog_op_header *ophdr,
+ struct xlog_ticket *ticket,
+ uint flags)
+{
+ ophdr->oh_tid = cpu_to_be32(ticket->t_tid);
+ ophdr->oh_clientid = ticket->t_clientid;
+ ophdr->oh_res2 = 0;
+
+ /* are we copying a commit or unmount record? */
+ ophdr->oh_flags = flags;
+
+ /*
+ * We've seen logs corrupted with bad transaction client ids. This
+ * makes sure that XFS doesn't generate them on. Turn this into an EIO
+ * and shut down the filesystem.
+ */
+ switch (ophdr->oh_clientid) {
+ case XFS_TRANSACTION:
+ case XFS_VOLUME:
+ case XFS_LOG:
+ break;
+ default:
+ xfs_fs_cmn_err(CE_WARN, log->l_mp,
+ "Bad XFS transaction clientid 0x%x in ticket 0x%p",
+ ophdr->oh_clientid, ticket);
+ return NULL;
+ }
+
+ return ophdr;
+}
+
+/*
+ * Set up the parameters of the region copy into the log. This has
+ * to handle region write split across multiple log buffers - this
+ * state is kept external to this function so that this code can
+ * can be written in an obvious, self documenting manner.
+ */
+static int
+xlog_write_setup_copy(
+ struct xlog_ticket *ticket,
+ struct xlog_op_header *ophdr,
+ int space_available,
+ int space_required,
+ int *copy_off,
+ int *copy_len,
+ int *last_was_partial_copy,
+ int *bytes_consumed)
+{
+ int still_to_copy;
+
+ still_to_copy = space_required - *bytes_consumed;
+ *copy_off = *bytes_consumed;
+
+ if (still_to_copy <= space_available) {
+ /* write of region completes here */
+ *copy_len = still_to_copy;
+ ophdr->oh_len = cpu_to_be32(*copy_len);
+ if (*last_was_partial_copy)
+ ophdr->oh_flags |= (XLOG_END_TRANS|XLOG_WAS_CONT_TRANS);
+ *last_was_partial_copy = 0;
+ *bytes_consumed = 0;
+ return 0;
+ }
+
+ /* partial write of region, needs extra log op header reservation */
+ *copy_len = space_available;
+ ophdr->oh_len = cpu_to_be32(*copy_len);
+ ophdr->oh_flags |= XLOG_CONTINUE_TRANS;
+ if (*last_was_partial_copy)
+ ophdr->oh_flags |= XLOG_WAS_CONT_TRANS;
+ *bytes_consumed += *copy_len;
+ (*last_was_partial_copy)++;
+
+ /* account for new log op header */
+ ticket->t_curr_res -= sizeof(struct xlog_op_header);
+ ticket->t_res_num_ophdrs++;
+
+ return sizeof(struct xlog_op_header);
+}
+
+static int
+xlog_write_copy_finish(
+ struct log *log,
+ struct xlog_in_core *iclog,
+ uint flags,
+ int *record_cnt,
+ int *data_cnt,
+ int *partial_copy,
+ int *partial_copy_len,
+ int log_offset,
+ struct xlog_in_core **commit_iclog)
+{
+ if (*partial_copy) {
+ /*
+ * This iclog has already been marked WANT_SYNC by
+ * xlog_state_get_iclog_space.
+ */
+ xlog_state_finish_copy(log, iclog, *record_cnt, *data_cnt);
+ *record_cnt = 0;
+ *data_cnt = 0;
+ return xlog_state_release_iclog(log, iclog);
+ }
+
+ *partial_copy = 0;
+ *partial_copy_len = 0;
+
+ if (iclog->ic_size - log_offset <= sizeof(xlog_op_header_t)) {
+ /* no more space in this iclog - push it. */
+ xlog_state_finish_copy(log, iclog, *record_cnt, *data_cnt);
+ *record_cnt = 0;
+ *data_cnt = 0;
+
+ spin_lock(&log->l_icloglock);
+ xlog_state_want_sync(log, iclog);
+ spin_unlock(&log->l_icloglock);
+
+ if (!commit_iclog)
+ return xlog_state_release_iclog(log, iclog);
+ ASSERT(flags & XLOG_COMMIT_TRANS);
+ *commit_iclog = iclog;
+ }
+
+ return 0;
+}
+
+/*
* Write some region out to in-core log
*
* This will be called when writing externally provided regions or when
* bytes have been written out.
*/
STATIC int
-xlog_write(xfs_mount_t * mp,
- xfs_log_iovec_t reg[],
- int nentries,
- xfs_log_ticket_t tic,
- xfs_lsn_t *start_lsn,
- xlog_in_core_t **commit_iclog,
- uint flags)
-{
- xlog_t *log = mp->m_log;
- xlog_ticket_t *ticket = (xlog_ticket_t *)tic;
- xlog_in_core_t *iclog = NULL; /* ptr to current in-core log */
- xlog_op_header_t *logop_head; /* ptr to log operation header */
- __psint_t ptr; /* copy address into data region */
- int len; /* # xlog_write() bytes 2 still copy */
- int index; /* region index currently copying */
- int log_offset; /* offset (from 0) into data region */
- int start_rec_copy; /* # bytes to copy for start record */
- int partial_copy; /* did we split a region? */
- int partial_copy_len;/* # bytes copied if split region */
- int need_copy; /* # bytes need to memcpy this region */
- int copy_len; /* # bytes actually memcpy'ing */
- int copy_off; /* # bytes from entry start */
- int contwr; /* continued write of in-core log? */
- int error;
- int record_cnt = 0, data_cnt = 0;
-
- partial_copy_len = partial_copy = 0;
-
- /* Calculate potential maximum space. Each region gets its own
- * xlog_op_header_t and may need to be double word aligned.
- */
- len = 0;
- if (ticket->t_flags & XLOG_TIC_INITED) { /* acct for start rec of xact */
- len += sizeof(xlog_op_header_t);
- ticket->t_res_num_ophdrs++;
- }
-
- for (index = 0; index < nentries; index++) {
- len += sizeof(xlog_op_header_t); /* each region gets >= 1 */
- ticket->t_res_num_ophdrs++;
- len += reg[index].i_len;
- xlog_tic_add_region(ticket, reg[index].i_len, reg[index].i_type);
- }
- contwr = *start_lsn = 0;
-
- if (ticket->t_curr_res < len) {
- xlog_print_tic_res(mp, ticket);
+xlog_write(
+ struct log *log,
+ struct xfs_log_vec *log_vector,
+ struct xlog_ticket *ticket,
+ xfs_lsn_t *start_lsn,
+ struct xlog_in_core **commit_iclog,
+ uint flags)
+{
+ struct xlog_in_core *iclog = NULL;
+ struct xfs_log_iovec *vecp;
+ struct xfs_log_vec *lv;
+ int len;
+ int index;
+ int partial_copy = 0;
+ int partial_copy_len = 0;
+ int contwr = 0;
+ int record_cnt = 0;
+ int data_cnt = 0;
+ int error;
+
+ *start_lsn = 0;
+
+ len = xlog_write_calc_vec_length(ticket, log_vector);
+ if (ticket->t_curr_res < len) {
+ xlog_print_tic_res(log->l_mp, ticket);
#ifdef DEBUG
- xlog_panic(
- "xfs_log_write: reservation ran out. Need to up reservation");
+ xlog_panic(
+ "xfs_log_write: reservation ran out. Need to up reservation");
#else
- /* Customer configurable panic */
- xfs_cmn_err(XFS_PTAG_LOGRES, CE_ALERT, mp,
- "xfs_log_write: reservation ran out. Need to up reservation");
- /* If we did not panic, shutdown the filesystem */
- xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
+ /* Customer configurable panic */
+ xfs_cmn_err(XFS_PTAG_LOGRES, CE_ALERT, log->l_mp,
+ "xfs_log_write: reservation ran out. Need to up reservation");
+
+ /* If we did not panic, shutdown the filesystem */
+ xfs_force_shutdown(log->l_mp, SHUTDOWN_CORRUPT_INCORE);
#endif
- } else
- ticket->t_curr_res -= len;
+ }
- for (index = 0; index < nentries; ) {
- if ((error = xlog_state_get_iclog_space(log, len, &iclog, ticket,
- &contwr, &log_offset)))
- return error;
+ ticket->t_curr_res -= len;
- ASSERT(log_offset <= iclog->ic_size - 1);
- ptr = (__psint_t) ((char *)iclog->ic_datap+log_offset);
+ index = 0;
+ lv = log_vector;
+ vecp = lv->lv_iovecp;
+ while (lv && index < lv->lv_niovecs) {
+ void *ptr;
+ int log_offset;
- /* start_lsn is the first lsn written to. That's all we need. */
- if (! *start_lsn)
- *start_lsn = be64_to_cpu(iclog->ic_header.h_lsn);
+ error = xlog_state_get_iclog_space(log, len, &iclog, ticket,
+ &contwr, &log_offset);
+ if (error)
+ return error;
- /* This loop writes out as many regions as can fit in the amount
- * of space which was allocated by xlog_state_get_iclog_space().
- */
- while (index < nentries) {
- ASSERT(reg[index].i_len % sizeof(__int32_t) == 0);
- ASSERT((__psint_t)ptr % sizeof(__int32_t) == 0);
- start_rec_copy = 0;
-
- /* If first write for transaction, insert start record.
- * We can't be trying to commit if we are inited. We can't
- * have any "partial_copy" if we are inited.
- */
- if (ticket->t_flags & XLOG_TIC_INITED) {
- logop_head = (xlog_op_header_t *)ptr;
- logop_head->oh_tid = cpu_to_be32(ticket->t_tid);
- logop_head->oh_clientid = ticket->t_clientid;
- logop_head->oh_len = 0;
- logop_head->oh_flags = XLOG_START_TRANS;
- logop_head->oh_res2 = 0;
- ticket->t_flags &= ~XLOG_TIC_INITED; /* clear bit */
- record_cnt++;
-
- start_rec_copy = sizeof(xlog_op_header_t);
- xlog_write_adv_cnt(ptr, len, log_offset, start_rec_copy);
- }
+ ASSERT(log_offset <= iclog->ic_size - 1);
+ ptr = iclog->ic_datap + log_offset;
- /* Copy log operation header directly into data section */
- logop_head = (xlog_op_header_t *)ptr;
- logop_head->oh_tid = cpu_to_be32(ticket->t_tid);
- logop_head->oh_clientid = ticket->t_clientid;
- logop_head->oh_res2 = 0;
+ /* start_lsn is the first lsn written to. That's all we need. */
+ if (!*start_lsn)
+ *start_lsn = be64_to_cpu(iclog->ic_header.h_lsn);
- /* header copied directly */
- xlog_write_adv_cnt(ptr, len, log_offset, sizeof(xlog_op_header_t));
+ /*
+ * This loop writes out as many regions as can fit in the amount
+ * of space which was allocated by xlog_state_get_iclog_space().
+ */
+ while (lv && index < lv->lv_niovecs) {
+ struct xfs_log_iovec *reg = &vecp[index];
+ struct xlog_op_header *ophdr;
+ int start_rec_copy;
+ int copy_len;
+ int copy_off;
+
+ ASSERT(reg->i_len % sizeof(__int32_t) == 0);
+ ASSERT((unsigned long)ptr % sizeof(__int32_t) == 0);
+
+ start_rec_copy = xlog_write_start_rec(ptr, ticket);
+ if (start_rec_copy) {
+ record_cnt++;
+ xlog_write_adv_cnt(&ptr, &len, &log_offset,
+ start_rec_copy);
+ }
- /* are we copying a commit or unmount record? */
- logop_head->oh_flags = flags;
+ ophdr = xlog_write_setup_ophdr(log, ptr, ticket, flags);
+ if (!ophdr)
+ return XFS_ERROR(EIO);
- /*
- * We've seen logs corrupted with bad transaction client
- * ids. This makes sure that XFS doesn't generate them on.
- * Turn this into an EIO and shut down the filesystem.
- */
- switch (logop_head->oh_clientid) {
- case XFS_TRANSACTION:
- case XFS_VOLUME:
- case XFS_LOG:
- break;
- default:
- xfs_fs_cmn_err(CE_WARN, mp,
- "Bad XFS transaction clientid 0x%x in ticket 0x%p",
- logop_head->oh_clientid, tic);
- return XFS_ERROR(EIO);
- }
+ xlog_write_adv_cnt(&ptr, &len, &log_offset,
+ sizeof(struct xlog_op_header));
+
+ len += xlog_write_setup_copy(ticket, ophdr,
+ iclog->ic_size-log_offset,
+ reg->i_len,
+ ©_off, ©_len,
+ &partial_copy,
+ &partial_copy_len);
+ xlog_verify_dest_ptr(log, ptr);
+
+ /* copy region */
+ ASSERT(copy_len >= 0);
+ memcpy(ptr, reg->i_addr + copy_off, copy_len);
+ xlog_write_adv_cnt(&ptr, &len, &log_offset, copy_len);
+
+ copy_len += start_rec_copy + sizeof(xlog_op_header_t);
+ record_cnt++;
+ data_cnt += contwr ? copy_len : 0;
+
+ error = xlog_write_copy_finish(log, iclog, flags,
+ &record_cnt, &data_cnt,
+ &partial_copy,
+ &partial_copy_len,
+ log_offset,
+ commit_iclog);
+ if (error)
+ return error;
- /* Partial write last time? => (partial_copy != 0)
- * need_copy is the amount we'd like to copy if everything could
- * fit in the current memcpy.
- */
- need_copy = reg[index].i_len - partial_copy_len;
-
- copy_off = partial_copy_len;
- if (need_copy <= iclog->ic_size - log_offset) { /*complete write */
- copy_len = need_copy;
- logop_head->oh_len = cpu_to_be32(copy_len);
- if (partial_copy)
- logop_head->oh_flags|= (XLOG_END_TRANS|XLOG_WAS_CONT_TRANS);
- partial_copy_len = partial_copy = 0;
- } else { /* partial write */
- copy_len = iclog->ic_size - log_offset;
- logop_head->oh_len = cpu_to_be32(copy_len);
- logop_head->oh_flags |= XLOG_CONTINUE_TRANS;
- if (partial_copy)
- logop_head->oh_flags |= XLOG_WAS_CONT_TRANS;
- partial_copy_len += copy_len;
- partial_copy++;
- len += sizeof(xlog_op_header_t); /* from splitting of region */
- /* account for new log op header */
- ticket->t_curr_res -= sizeof(xlog_op_header_t);
- ticket->t_res_num_ophdrs++;
- }
- xlog_verify_dest_ptr(log, ptr);
-
- /* copy region */
- ASSERT(copy_len >= 0);
- memcpy((xfs_caddr_t)ptr, reg[index].i_addr + copy_off, copy_len);
- xlog_write_adv_cnt(ptr, len, log_offset, copy_len);
-
- /* make copy_len total bytes copied, including headers */
- copy_len += start_rec_copy + sizeof(xlog_op_header_t);
- record_cnt++;
- data_cnt += contwr ? copy_len : 0;
- if (partial_copy) { /* copied partial region */
- /* already marked WANT_SYNC by xlog_state_get_iclog_space */
- xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
- record_cnt = data_cnt = 0;
- if ((error = xlog_state_release_iclog(log, iclog)))
- return error;
- break; /* don't increment index */
- } else { /* copied entire region */
- index++;
- partial_copy_len = partial_copy = 0;
-
- 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;
- xlog_state_want_sync(log, iclog);
- if (commit_iclog) {
- ASSERT(flags & XLOG_COMMIT_TRANS);
- *commit_iclog = iclog;
- } else if ((error = xlog_state_release_iclog(log, iclog)))
- return error;
- if (index == nentries)
- return 0; /* we are done */
- else
- break;
+ /*
+ * if we had a partial copy, we need to get more iclog
+ * space but we don't want to increment the region
+ * index because there is still more is this region to
+ * write.
+ *
+ * If we completed writing this region, and we flushed
+ * the iclog (indicated by resetting of the record
+ * count), then we also need to get more log space. If
+ * this was the last record, though, we are done and
+ * can just return.
+ */
+ if (partial_copy)
+ break;
+
+ if (++index == lv->lv_niovecs) {
+ lv = lv->lv_next;
+ index = 0;
+ if (lv)
+ vecp = lv->lv_iovecp;
+ }
+ if (record_cnt == 0) {
+ if (!lv)
+ return 0;
+ break;
+ }
}
- } /* if (partial_copy) */
- } /* while (index < nentries) */
- } /* for (index = 0; index < nentries; ) */
- ASSERT(len == 0);
+ }
+
+ ASSERT(len == 0);
+
+ xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
+ if (!commit_iclog)
+ return xlog_state_release_iclog(log, iclog);
- xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
- if (commit_iclog) {
ASSERT(flags & XLOG_COMMIT_TRANS);
*commit_iclog = iclog;
return 0;
- }
- return xlog_state_release_iclog(log, iclog);
-} /* xlog_write */
+}
/*****************************************************************************
/* 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
if (iclog->ic_state == XLOG_STATE_DIRTY) {
iclog->ic_state = XLOG_STATE_ACTIVE;
iclog->ic_offset = 0;
- iclog->ic_callback = NULL; /* don't need to free */
+ ASSERT(iclog->ic_callback == NULL);
/*
* If the number of ops in this iclog indicate it just
* contains the dummy transaction, we can
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;
be64_to_cpu(iclog->ic_header.h_lsn);
spin_unlock(&log->l_grant_lock);
- /*
- * Keep processing entries in the callback list
- * until we come around and it is empty. We
- * need to atomically see that the list is
- * empty and change the state to DIRTY so that
- * we don't miss any more callbacks being added.
- */
- spin_lock(&log->l_icloglock);
} else {
+ spin_unlock(&log->l_icloglock);
ioerrors++;
}
- cb = iclog->ic_callback;
+ /*
+ * Keep processing entries in the callback list until
+ * we come around and it is empty. We need to
+ * atomically see that the list is empty and change the
+ * state to DIRTY so that we don't miss any more
+ * callbacks being added.
+ */
+ spin_lock(&iclog->ic_callback_lock);
+ cb = iclog->ic_callback;
while (cb) {
iclog->ic_callback_tail = &(iclog->ic_callback);
iclog->ic_callback = NULL;
- spin_unlock(&log->l_icloglock);
+ spin_unlock(&iclog->ic_callback_lock);
/* perform callbacks in the order given */
for (; cb; cb = cb_next) {
cb_next = cb->cb_next;
cb->cb_func(cb->cb_arg, aborted);
}
- spin_lock(&log->l_icloglock);
+ spin_lock(&iclog->ic_callback_lock);
cb = iclog->ic_callback;
}
loopdidcallbacks++;
funcdidcallbacks++;
+ spin_lock(&log->l_icloglock);
ASSERT(iclog->ic_callback == NULL);
+ spin_unlock(&iclog->ic_callback_lock);
if (!(iclog->ic_state & XLOG_STATE_IOERROR))
iclog->ic_state = XLOG_STATE_DIRTY;
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(
ASSERT(iclog->ic_state == XLOG_STATE_SYNCING ||
iclog->ic_state == XLOG_STATE_IOERROR);
- ASSERT(iclog->ic_refcnt == 0);
+ ASSERT(atomic_read(&iclog->ic_refcnt) == 0);
ASSERT(iclog->ic_bwritecnt == 1 || iclog->ic_bwritecnt == 2);
* 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;
- iclog->ic_refcnt++; /* prevents sync */
+ atomic_inc(&iclog->ic_refcnt); /* prevents sync */
log_offset = iclog->ic_offset;
/* On the 1st write to an iclog, figure out lsn. This works
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 (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 {
- 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);
/*
- * Atomically put back used ticket.
- */
-STATIC void
-xlog_state_put_ticket(xlog_t *log,
- xlog_ticket_t *tic)
-{
- spin_lock(&log->l_icloglock);
- xlog_ticket_put(log, tic);
- spin_unlock(&log->l_icloglock);
-} /* xlog_state_put_ticket */
-
-/*
* Flush iclog to disk if this is the last reference to the given iclog and
* the WANT_SYNC bit is set.
*
*
*/
STATIC int
-xlog_state_release_iclog(xlog_t *log,
- xlog_in_core_t *iclog)
+xlog_state_release_iclog(
+ xlog_t *log,
+ xlog_in_core_t *iclog)
{
int sync = 0; /* do we sync? */
- xlog_assign_tail_lsn(log->l_mp);
+ if (iclog->ic_state & XLOG_STATE_IOERROR)
+ return XFS_ERROR(EIO);
- spin_lock(&log->l_icloglock);
+ ASSERT(atomic_read(&iclog->ic_refcnt) > 0);
+ if (!atomic_dec_and_lock(&iclog->ic_refcnt, &log->l_icloglock))
+ return 0;
if (iclog->ic_state & XLOG_STATE_IOERROR) {
spin_unlock(&log->l_icloglock);
return XFS_ERROR(EIO);
}
-
- ASSERT(iclog->ic_refcnt > 0);
ASSERT(iclog->ic_state == XLOG_STATE_ACTIVE ||
iclog->ic_state == XLOG_STATE_WANT_SYNC);
- if (--iclog->ic_refcnt == 0 &&
- iclog->ic_state == XLOG_STATE_WANT_SYNC) {
+ if (iclog->ic_state == XLOG_STATE_WANT_SYNC) {
+ /* update tail before writing to iclog */
+ xlog_assign_tail_lsn(log->l_mp);
sync++;
iclog->ic_state = XLOG_STATE_SYNCING;
iclog->ic_header.h_tail_lsn = cpu_to_be64(log->l_tail_lsn);
xlog_verify_tail_lsn(log, iclog, log->l_tail_lsn);
/* cycle incremented when incrementing curr_block */
}
-
spin_unlock(&log->l_icloglock);
/*
* this iclog has consistent data, so we ignore IOERROR
* flags after this point.
*/
- if (sync) {
+ if (sync)
return xlog_sync(log, iclog);
- }
return 0;
-
} /* xlog_state_release_iclog */
log->l_curr_block += BTOBB(eventual_size)+BTOBB(log->l_iclog_hsize);
/* Round up to next log-sunit */
- if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) &&
+ if (xfs_sb_version_haslogv2(&log->l_mp->m_sb) &&
log->l_mp->m_sb.sb_logsunit > 1) {
__uint32_t sunit_bb = BTOBB(log->l_mp->m_sb.sb_logsunit);
log->l_curr_block = roundup(log->l_curr_block, sunit_bb);
log->l_iclog = iclog->ic_next;
} /* xlog_state_switch_iclogs */
-
/*
* Write out all data in the in-core log as of this exact moment in time.
*
* 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
* b) when we return from flushing out this iclog, it is still
* not in the active nor dirty state.
*/
-STATIC int
-xlog_state_sync_all(xlog_t *log, uint flags, int *log_flushed)
+int
+_xfs_log_force(
+ struct xfs_mount *mp,
+ uint flags,
+ int *log_flushed)
{
- xlog_in_core_t *iclog;
- xfs_lsn_t lsn;
+ struct log *log = mp->m_log;
+ struct xlog_in_core *iclog;
+ xfs_lsn_t lsn;
+
+ XFS_STATS_INC(xs_log_force);
spin_lock(&log->l_icloglock);
* previous iclog and go to sleep.
*/
if (iclog->ic_state == XLOG_STATE_DIRTY ||
- (iclog->ic_refcnt == 0 && iclog->ic_offset == 0)) {
+ (atomic_read(&iclog->ic_refcnt) == 0
+ && iclog->ic_offset == 0)) {
iclog = iclog->ic_prev;
if (iclog->ic_state == XLOG_STATE_ACTIVE ||
iclog->ic_state == XLOG_STATE_DIRTY)
else
goto maybe_sleep;
} else {
- if (iclog->ic_refcnt == 0) {
+ if (atomic_read(&iclog->ic_refcnt) == 0) {
/* We are the only one with access to this
* iclog. Flush it out now. There should
* be a roundoff of zero to show that someone
* has already taken care of the roundoff from
* the previous sync.
*/
- iclog->ic_refcnt++;
+ atomic_inc(&iclog->ic_refcnt);
lsn = be64_to_cpu(iclog->ic_header.h_lsn);
xlog_state_switch_iclogs(log, iclog, 0);
spin_unlock(&log->l_icloglock);
if (xlog_state_release_iclog(log, iclog))
return XFS_ERROR(EIO);
- *log_flushed = 1;
+
+ if (log_flushed)
+ *log_flushed = 1;
spin_lock(&log->l_icloglock);
if (be64_to_cpu(iclog->ic_header.h_lsn) == lsn &&
iclog->ic_state != XLOG_STATE_DIRTY)
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
*/
if (iclog->ic_state & XLOG_STATE_IOERROR)
return XFS_ERROR(EIO);
- *log_flushed = 1;
-
+ if (log_flushed)
+ *log_flushed = 1;
} else {
no_sleep:
spin_unlock(&log->l_icloglock);
}
return 0;
-} /* xlog_state_sync_all */
+}
+
+/*
+ * 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,
+ uint flags)
+{
+ int error;
+ error = _xfs_log_force(mp, flags, NULL);
+ if (error) {
+ xfs_fs_cmn_err(CE_WARN, mp, "xfs_log_force: "
+ "error %d returned.", error);
+ }
+}
/*
- * Used by code which implements synchronous log forces.
+ * Force the in-core log to disk for a specific LSN.
*
* Find in-core log with lsn.
* If it is in the DIRTY state, just return.
* state and go to sleep or return.
* If it is in any other state, go to sleep or return.
*
- * If filesystem activity goes to zero, the iclog will get flushed only by
- * bdflush().
+ * 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
+ * sv.
*/
-STATIC int
-xlog_state_sync(xlog_t *log,
- xfs_lsn_t lsn,
- uint flags,
- int *log_flushed)
+int
+_xfs_log_force_lsn(
+ struct xfs_mount *mp,
+ xfs_lsn_t lsn,
+ uint flags,
+ int *log_flushed)
{
- xlog_in_core_t *iclog;
- int already_slept = 0;
-
-try_again:
- spin_lock(&log->l_icloglock);
- iclog = log->l_iclog;
+ struct log *log = mp->m_log;
+ struct xlog_in_core *iclog;
+ int already_slept = 0;
- if (iclog->ic_state & XLOG_STATE_IOERROR) {
- spin_unlock(&log->l_icloglock);
- return XFS_ERROR(EIO);
- }
+ ASSERT(lsn != 0);
- do {
- if (be64_to_cpu(iclog->ic_header.h_lsn) != lsn) {
- iclog = iclog->ic_next;
- continue;
- }
+ XFS_STATS_INC(xs_log_force);
- if (iclog->ic_state == XLOG_STATE_DIRTY) {
+try_again:
+ spin_lock(&log->l_icloglock);
+ iclog = log->l_iclog;
+ if (iclog->ic_state & XLOG_STATE_IOERROR) {
spin_unlock(&log->l_icloglock);
- return 0;
+ return XFS_ERROR(EIO);
}
- if (iclog->ic_state == XLOG_STATE_ACTIVE) {
- /*
- * We sleep here if we haven't already slept (e.g.
- * this is the first time we've looked at the correct
- * iclog buf) and the buffer before us is going to
- * be sync'ed. The reason for this is that if we
- * are doing sync transactions here, by waiting for
- * the previous I/O to complete, we can allow a few
- * more transactions into this iclog before we close
- * it down.
- *
- * Otherwise, we mark the buffer WANT_SYNC, and bump
- * up the refcnt so we can release the log (which drops
- * the ref count). The state switch keeps new transaction
- * commits from using this buffer. When the current commits
- * finish writing into the buffer, the refcount will drop to
- * zero and the buffer will go out then.
- */
- if (!already_slept &&
- (iclog->ic_prev->ic_state & (XLOG_STATE_WANT_SYNC |
- 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,
- &log->l_icloglock, s);
- *log_flushed = 1;
- already_slept = 1;
- goto try_again;
- } else {
- iclog->ic_refcnt++;
+ do {
+ if (be64_to_cpu(iclog->ic_header.h_lsn) != lsn) {
+ iclog = iclog->ic_next;
+ continue;
+ }
+
+ if (iclog->ic_state == XLOG_STATE_DIRTY) {
+ spin_unlock(&log->l_icloglock);
+ return 0;
+ }
+
+ if (iclog->ic_state == XLOG_STATE_ACTIVE) {
+ /*
+ * We sleep here if we haven't already slept (e.g.
+ * this is the first time we've looked at the correct
+ * iclog buf) and the buffer before us is going to
+ * be sync'ed. The reason for this is that if we
+ * are doing sync transactions here, by waiting for
+ * the previous I/O to complete, we can allow a few
+ * more transactions into this iclog before we close
+ * it down.
+ *
+ * Otherwise, we mark the buffer WANT_SYNC, and bump
+ * up the refcnt so we can release the log (which
+ * drops the ref count). The state switch keeps new
+ * transaction commits from using this buffer. When
+ * the current commits finish writing into the buffer,
+ * the refcount will drop to zero and the buffer will
+ * go out then.
+ */
+ if (!already_slept &&
+ (iclog->ic_prev->ic_state &
+ (XLOG_STATE_WANT_SYNC | XLOG_STATE_SYNCING))) {
+ ASSERT(!(iclog->ic_state & XLOG_STATE_IOERROR));
+
+ XFS_STATS_INC(xs_log_force_sleep);
+
+ sv_wait(&iclog->ic_prev->ic_write_wait,
+ PSWP, &log->l_icloglock, s);
+ if (log_flushed)
+ *log_flushed = 1;
+ already_slept = 1;
+ goto try_again;
+ }
+ atomic_inc(&iclog->ic_refcnt);
xlog_state_switch_iclogs(log, iclog, 0);
spin_unlock(&log->l_icloglock);
if (xlog_state_release_iclog(log, iclog))
return XFS_ERROR(EIO);
- *log_flushed = 1;
+ if (log_flushed)
+ *log_flushed = 1;
spin_lock(&log->l_icloglock);
}
- }
- if ((flags & XFS_LOG_SYNC) && /* sleep */
- !(iclog->ic_state & (XLOG_STATE_ACTIVE | XLOG_STATE_DIRTY))) {
+ if ((flags & XFS_LOG_SYNC) && /* sleep */
+ !(iclog->ic_state &
+ (XLOG_STATE_ACTIVE | XLOG_STATE_DIRTY))) {
+ /*
+ * Don't wait on completion if we know that we've
+ * gotten a log write error.
+ */
+ if (iclog->ic_state & XLOG_STATE_IOERROR) {
+ spin_unlock(&log->l_icloglock);
+ return XFS_ERROR(EIO);
+ }
+ XFS_STATS_INC(xs_log_force_sleep);
+ 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
+ * and the memory read should be atomic.
+ */
+ if (iclog->ic_state & XLOG_STATE_IOERROR)
+ return XFS_ERROR(EIO);
- /*
- * Don't wait on the forcesema if we know that we've
- * gotten a log write error.
- */
- if (iclog->ic_state & XLOG_STATE_IOERROR) {
+ if (log_flushed)
+ *log_flushed = 1;
+ } else { /* just return */
spin_unlock(&log->l_icloglock);
- return XFS_ERROR(EIO);
}
- XFS_STATS_INC(xs_log_force_sleep);
- sv_wait(&iclog->ic_forcesema, PSWP, &log->l_icloglock, s);
- /*
- * No need to grab the log lock here since we're
- * only deciding whether or not to return EIO
- * and the memory read should be atomic.
- */
- if (iclog->ic_state & XLOG_STATE_IOERROR)
- return XFS_ERROR(EIO);
- *log_flushed = 1;
- } else { /* just return */
- spin_unlock(&log->l_icloglock);
- }
- return 0;
- } while (iclog != log->l_iclog);
+ return 0;
+ } while (iclog != log->l_iclog);
+
+ spin_unlock(&log->l_icloglock);
+ return 0;
+}
- spin_unlock(&log->l_icloglock);
- return 0;
-} /* xlog_state_sync */
+/*
+ * Wrapper for _xfs_log_force_lsn(), 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_lsn(
+ xfs_mount_t *mp,
+ xfs_lsn_t lsn,
+ uint flags)
+{
+ int error;
+ error = _xfs_log_force_lsn(mp, lsn, flags, NULL);
+ if (error) {
+ xfs_fs_cmn_err(CE_WARN, mp, "xfs_log_force: "
+ "error %d returned.", error);
+ }
+}
/*
* Called when we want to mark the current iclog as being ready to sync to
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 */
-
+}
/*****************************************************************************
*/
/*
- * Algorithm doesn't take into account page size. ;-(
+ * Free a used ticket when its refcount falls to zero.
*/
-STATIC void
-xlog_state_ticket_alloc(xlog_t *log)
+void
+xfs_log_ticket_put(
+ xlog_ticket_t *ticket)
{
- xlog_ticket_t *t_list;
- xlog_ticket_t *next;
- xfs_caddr_t buf;
- uint i = (NBPP / sizeof(xlog_ticket_t)) - 2;
-
- /*
- * The kmem_zalloc may sleep, so we shouldn't be holding the
- * global lock. XXXmiken: may want to use zone allocator.
- */
- buf = (xfs_caddr_t) kmem_zalloc(NBPP, KM_SLEEP);
-
- spin_lock(&log->l_icloglock);
-
- /* Attach 1st ticket to Q, so we can keep track of allocated memory */
- t_list = (xlog_ticket_t *)buf;
- t_list->t_next = log->l_unmount_free;
- log->l_unmount_free = t_list++;
- log->l_ticket_cnt++;
- log->l_ticket_tcnt++;
-
- /* Next ticket becomes first ticket attached to ticket free list */
- if (log->l_freelist != NULL) {
- ASSERT(log->l_tail != NULL);
- log->l_tail->t_next = t_list;
- } else {
- log->l_freelist = t_list;
- }
- log->l_ticket_cnt++;
- log->l_ticket_tcnt++;
-
- /* Cycle through rest of alloc'ed memory, building up free Q */
- for ( ; i > 0; i--) {
- next = t_list + 1;
- t_list->t_next = next;
- t_list = next;
- log->l_ticket_cnt++;
- log->l_ticket_tcnt++;
+ 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);
}
- t_list->t_next = NULL;
- log->l_tail = t_list;
- spin_unlock(&log->l_icloglock);
-} /* xlog_state_ticket_alloc */
-
+}
-/*
- * Put ticket into free list
- *
- * Assumption: log lock is held around this call.
- */
-STATIC void
-xlog_ticket_put(xlog_t *log,
- xlog_ticket_t *ticket)
+xlog_ticket_t *
+xfs_log_ticket_get(
+ xlog_ticket_t *ticket)
{
- sv_destroy(&ticket->t_sema);
-
- /*
- * Don't think caching will make that much difference. It's
- * more important to make debug easier.
- */
-#if 0
- /* real code will want to use LIFO for caching */
- ticket->t_next = log->l_freelist;
- log->l_freelist = ticket;
- /* no need to clear fields */
-#else
- /* When we debug, it is easier if tickets are cycled */
- ticket->t_next = NULL;
- if (log->l_tail) {
- log->l_tail->t_next = ticket;
- } else {
- ASSERT(log->l_freelist == NULL);
- log->l_freelist = ticket;
- }
- log->l_tail = ticket;
-#endif /* DEBUG */
- log->l_ticket_cnt++;
-} /* xlog_ticket_put */
-
+ ASSERT(atomic_read(&ticket->t_ref) > 0);
+ atomic_inc(&ticket->t_ref);
+ return ticket;
+}
/*
- * Grab ticket off freelist or allocation some more
+ * Allocate and initialise a new log ticket.
*/
STATIC xlog_ticket_t *
-xlog_ticket_get(xlog_t *log,
- int unit_bytes,
- int cnt,
- char client,
- uint xflags)
+xlog_ticket_alloc(
+ struct log *log,
+ int unit_bytes,
+ int cnt,
+ char client,
+ uint xflags)
{
- xlog_ticket_t *tic;
+ struct xlog_ticket *tic;
uint num_headers;
+ int iclog_space;
- alloc:
- if (log->l_freelist == NULL)
- xlog_state_ticket_alloc(log); /* potentially sleep */
-
- spin_lock(&log->l_icloglock);
- if (log->l_freelist == NULL) {
- spin_unlock(&log->l_icloglock);
- goto alloc;
- }
- tic = log->l_freelist;
- log->l_freelist = tic->t_next;
- if (log->l_freelist == NULL)
- log->l_tail = NULL;
- log->l_ticket_cnt--;
- spin_unlock(&log->l_icloglock);
+ tic = kmem_zone_zalloc(xfs_log_ticket_zone, KM_SLEEP|KM_MAYFAIL);
+ if (!tic)
+ return NULL;
/*
* Permanent reservations have up to 'cnt'-1 active log operations
/* for start-rec */
unit_bytes += sizeof(xlog_op_header_t);
- /* for LR headers */
- num_headers = ((unit_bytes + log->l_iclog_size-1) >> log->l_iclog_size_log);
+ /*
+ * for LR headers - the space for data in an iclog is the size minus
+ * the space used for the headers. If we use the iclog size, then we
+ * undercalculate the number of headers required.
+ *
+ * Furthermore - the addition of op headers for split-recs might
+ * increase the space required enough to require more log and op
+ * headers, so take that into account too.
+ *
+ * IMPORTANT: This reservation makes the assumption that if this
+ * transaction is the first in an iclog and hence has the LR headers
+ * accounted to it, then the remaining space in the iclog is
+ * exclusively for this transaction. i.e. if the transaction is larger
+ * than the iclog, it will be the only thing in that iclog.
+ * Fundamentally, this means we must pass the entire log vector to
+ * xlog_write to guarantee this.
+ */
+ iclog_space = log->l_iclog_size - log->l_iclog_hsize;
+ num_headers = howmany(unit_bytes, iclog_space);
+
+ /* for split-recs - ophdrs added when data split over LRs */
+ unit_bytes += sizeof(xlog_op_header_t) * num_headers;
+
+ /* add extra header reservations if we overrun */
+ while (!num_headers ||
+ howmany(unit_bytes, iclog_space) > num_headers) {
+ unit_bytes += sizeof(xlog_op_header_t);
+ num_headers++;
+ }
unit_bytes += log->l_iclog_hsize * num_headers;
/* for commit-rec LR header - note: padding will subsume the ophdr */
unit_bytes += log->l_iclog_hsize;
- /* for split-recs - ophdrs added when data split over LRs */
- unit_bytes += sizeof(xlog_op_header_t) * num_headers;
-
/* for roundoff padding for transaction data and one for commit record */
- if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) &&
+ if (xfs_sb_version_haslogv2(&log->l_mp->m_sb) &&
log->l_mp->m_sb.sb_logsunit > 1) {
/* log su roundoff */
unit_bytes += 2*log->l_mp->m_sb.sb_logsunit;
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_ocnt = cnt;
- tic->t_tid = (xlog_tid_t)((__psint_t)tic & 0xffffffff);
+ tic->t_tid = random32();
tic->t_clientid = client;
tic->t_flags = XLOG_TIC_INITED;
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 */
+}
/******************************************************************************
* part of the log in case we trash the log structure.
*/
void
-xlog_verify_dest_ptr(xlog_t *log,
- __psint_t ptr)
+xlog_verify_dest_ptr(
+ struct log *log,
+ char *ptr)
{
int i;
int good_ptr = 0;
- for (i=0; i < log->l_iclog_bufs; i++) {
- if (ptr >= (__psint_t)log->l_iclog_bak[i] &&
- ptr <= (__psint_t)log->l_iclog_bak[i]+log->l_iclog_size)
+ for (i = 0; i < log->l_iclog_bufs; i++) {
+ if (ptr >= log->l_iclog_bak[i] &&
+ ptr <= log->l_iclog_bak[i] + log->l_iclog_size)
good_ptr++;
}
- if (! good_ptr)
+
+ if (!good_ptr)
xlog_panic("xlog_verify_dest_ptr: invalid ptr");
-} /* xlog_verify_dest_ptr */
+}
STATIC void
xlog_verify_grant_head(xlog_t *log, int equals)
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;
xlog_ticket_t *tic;
xlog_t *log;
int retval;
- int dummy;
log = mp->m_log;
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);
}
spin_unlock(&log->l_grant_lock);
- if (! (log->l_iclog->ic_state & XLOG_STATE_IOERROR)) {
+ if (!(log->l_iclog->ic_state & XLOG_STATE_IOERROR)) {
ASSERT(!logerror);
/*
* Force the incore logs to disk before shutting the
* log down completely.
*/
- xlog_state_sync_all(log, XFS_LOG_FORCE|XFS_LOG_SYNC, &dummy);
+ _xfs_log_force(mp, XFS_LOG_SYNC, NULL);
+
spin_lock(&log->l_icloglock);
retval = xlog_state_ioerror(log);
spin_unlock(&log->l_icloglock);