#include "xfs_inode.h"
#include "xfs_rw.h"
+kmem_zone_t *xfs_log_ticket_zone;
#define xlog_write_adv_cnt(ptr, len, off, bytes) \
{ (ptr) += (bytes); \
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,
/* local ticket functions */
-STATIC void xlog_state_ticket_alloc(xlog_t *log);
STATIC xlog_ticket_t *xlog_ticket_get(xlog_t *log,
int unit_bytes,
int count,
*/
xlog_trace_loggrant(log, ticket, "xfs_log_done: (non-permanent)");
xlog_ungrant_log_space(log, ticket);
- xlog_state_put_ticket(log, ticket);
+ xlog_ticket_put(log, ticket);
} else {
xlog_trace_loggrant(log, ticket, "xfs_log_done: (permanent)");
xlog_regrant_reserve_log_space(log, ticket);
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
void *iclog_hndl, /* iclog to hang callback off */
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 */
/* may sleep if need to allocate more tickets */
internal_ticket = xlog_ticket_get(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,
* 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);
+ /*
+ * 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;
+ }
+
/*
* 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 error;
}
}
/* End mounting message in xfs_log_mount_finish */
return 0;
+error:
+ xfs_log_unmount_dealloc(mp);
+ return error;
} /* xfs_log_mount */
/*
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;
- iclog->ic_refcnt++;
+ atomic_inc(&iclog->ic_refcnt);
spin_unlock(&log->l_icloglock);
xlog_state_want_sync(log, iclog);
- (void) xlog_state_release_iclog(log, iclog);
+ error = xlog_state_release_iclog(log, iclog);
spin_lock(&log->l_icloglock);
if (!(iclog->ic_state == XLOG_STATE_ACTIVE ||
if (tic) {
xlog_trace_loggrant(log, tic, "unmount rec");
xlog_ungrant_log_space(log, tic);
- xlog_state_put_ticket(log, tic);
+ xlog_ticket_put(log, tic);
}
} else {
/*
*/
spin_lock(&log->l_icloglock);
iclog = log->l_iclog;
- iclog->ic_refcnt++;
+ atomic_inc(&iclog->ic_refcnt);
spin_unlock(&log->l_icloglock);
xlog_state_want_sync(log, iclog);
- (void) xlog_state_release_iclog(log, iclog);
+ error = xlog_state_release_iclog(log, iclog);
spin_lock(&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_trans_ail_destroy(mp);
xlog_dealloc_log(mp->m_log);
}
if (XLOG_FORCED_SHUTDOWN(log))
return;
- ASSERT(!XFS_FORCED_SHUTDOWN(mp));
if (tail_lsn == 0) {
/* needed since sync_lsn is 64 bits */
* the tail. The details of this case are described below, but the end
* result is that we return the size of the log as the amount of space left.
*/
-int
+STATIC int
xlog_space_left(xlog_t *log, int cycle, int bytes)
{
int free_bytes;
size >>= 1;
}
- if (XFS_SB_VERSION_HASLOGV2(&mp->m_sb)) {
+ if (xfs_sb_version_haslogv2(&mp->m_sb)) {
/* # headers = size / 32K
* one header holds cycles from 32K of data
*/
log->l_grant_reserve_cycle = 1;
log->l_grant_write_cycle = 1;
- if (XFS_SB_VERSION_HASSECTOR(&mp->m_sb)) {
+ 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);
/* 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));
+ xfs_sb_version_haslogv2(&mp->m_sb));
ASSERT(mp->m_sb.sb_logsectlog >= BBSHIFT);
}
log->l_sectbb_mask = (1 << log->l_sectbb_log) - 1;
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 */
/* log record size must be multiple of BBSIZE; see xlog_rec_header_t */
ASSERT((XFS_BUF_SIZE(bp) & BBMASK) == 0);
XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
iclog->ic_bp = bp;
iclog->hic_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));
- INT_SET(head->h_magicno, ARCH_CONVERT, XLOG_HEADER_MAGIC_NUM);
- INT_SET(head->h_version, ARCH_CONVERT,
- XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? 2 : 1);
- INT_SET(head->h_size, ARCH_CONVERT, log->l_iclog_size);
+ 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);
+ head->h_size = cpu_to_be32(log->l_iclog_size);
/* new fields */
- INT_SET(head->h_fmt, ARCH_CONVERT, XLOG_FMT);
+ 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;
* pushes on an lsn which is further along in the log once we reach the high
* water mark. In this manner, we would be creating a low water mark.
*/
-void
+STATIC void
xlog_grant_push_ail(xfs_mount_t *mp,
int need_bytes)
{
* is added immediately before calling bwrite().
*/
-int
+STATIC int
xlog_sync(xlog_t *log,
xlog_in_core_t *iclog)
{
xfs_caddr_t dptr; /* pointer to byte sized element */
xfs_buf_t *bp;
- int i, ops;
+ int i;
uint count; /* byte count of bwrite */
uint count_init; /* initial count before roundup */
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;
/* real byte length */
if (v2) {
- INT_SET(iclog->ic_header.h_len,
- ARCH_CONVERT,
- iclog->ic_offset + roundoff);
+ iclog->ic_header.h_len =
+ cpu_to_be32(iclog->ic_offset + roundoff);
} else {
- INT_SET(iclog->ic_header.h_len, ARCH_CONVERT, iclog->ic_offset);
+ iclog->ic_header.h_len =
+ cpu_to_be32(iclog->ic_offset);
}
- /* put ops count in correct order */
- ops = iclog->ic_header.h_num_logops;
- INT_SET(iclog->ic_header.h_num_logops, ARCH_CONVERT, ops);
-
bp = iclog->ic_bp;
ASSERT(XFS_BUF_FSPRIVATE2(bp, unsigned long) == (unsigned long)1);
XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)2);
- XFS_BUF_SET_ADDR(bp, BLOCK_LSN(INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT)));
+ XFS_BUF_SET_ADDR(bp, BLOCK_LSN(be64_to_cpu(iclog->ic_header.h_lsn)));
XFS_STATS_ADD(xs_log_blocks, BTOBB(count));
* a new cycle. Watch out for the header magic number
* case, though.
*/
- for (i=0; i<split; i += BBSIZE) {
- INT_MOD(*(uint *)dptr, ARCH_CONVERT, +1);
- if (INT_GET(*(uint *)dptr, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM)
- INT_MOD(*(uint *)dptr, ARCH_CONVERT, +1);
+ for (i = 0; i < split; i += BBSIZE) {
+ be32_add_cpu((__be32 *)dptr, 1);
+ if (be32_to_cpu(*(__be32 *)dptr) == XLOG_HEADER_MAGIC_NUM)
+ be32_add_cpu((__be32 *)dptr, 1);
dptr += BBSIZE;
}
/*
* Deallocate a log structure
*/
-void
+STATIC void
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;
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) {
{
spin_lock(&log->l_icloglock);
- 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);
* we don't update ic_offset until the end when we know exactly how many
* bytes have been written out.
*/
-int
+STATIC int
xlog_write(xfs_mount_t * mp,
xfs_log_iovec_t reg[],
int nentries,
/* start_lsn is the first lsn written to. That's all we need. */
if (! *start_lsn)
- *start_lsn = INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT);
+ *start_lsn = be64_to_cpu(iclog->ic_header.h_lsn);
/* This loop writes out as many regions as can fit in the amount
* of space which was allocated by xlog_state_get_iclog_space().
*/
if (ticket->t_flags & XLOG_TIC_INITED) {
logop_head = (xlog_op_header_t *)ptr;
- INT_SET(logop_head->oh_tid, ARCH_CONVERT, ticket->t_tid);
+ 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;
/* Copy log operation header directly into data section */
logop_head = (xlog_op_header_t *)ptr;
- INT_SET(logop_head->oh_tid, ARCH_CONVERT, ticket->t_tid);
+ logop_head->oh_tid = cpu_to_be32(ticket->t_tid);
logop_head->oh_clientid = ticket->t_clientid;
logop_head->oh_res2 = 0;
copy_off = partial_copy_len;
if (need_copy <= iclog->ic_size - log_offset) { /*complete write */
- INT_SET(logop_head->oh_len, ARCH_CONVERT, copy_len = need_copy);
+ 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;
- INT_SET(logop_head->oh_len, ARCH_CONVERT, copy_len);
+ 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;
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
* We don't need to cover the dummy.
*/
if (!changed &&
- (INT_GET(iclog->ic_header.h_num_logops, ARCH_CONVERT) == XLOG_COVER_OPS)) {
+ (be32_to_cpu(iclog->ic_header.h_num_logops) ==
+ XLOG_COVER_OPS)) {
changed = 1;
} else {
/*
lowest_lsn = 0;
do {
if (!(lsn_log->ic_state & (XLOG_STATE_ACTIVE|XLOG_STATE_DIRTY))) {
- lsn = INT_GET(lsn_log->ic_header.h_lsn, ARCH_CONVERT);
+ lsn = be64_to_cpu(lsn_log->ic_header.h_lsn);
if ((lsn && !lowest_lsn) ||
(XFS_LSN_CMP(lsn, lowest_lsn) < 0)) {
lowest_lsn = lsn;
*/
lowest_lsn = xlog_get_lowest_lsn(log);
- if (lowest_lsn && (
- XFS_LSN_CMP(
- lowest_lsn,
- INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT)
- )<0)) {
+ if (lowest_lsn &&
+ XFS_LSN_CMP(lowest_lsn,
+ be64_to_cpu(iclog->ic_header.h_lsn)) < 0) {
iclog = iclog->ic_next;
continue; /* Leave this iclog for
* another thread */
* No one else can be here except us.
*/
spin_lock(&log->l_grant_lock);
- ASSERT(XFS_LSN_CMP(
- log->l_last_sync_lsn,
- INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT)
- )<=0);
- log->l_last_sync_lsn = INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT);
+ ASSERT(XFS_LSN_CMP(log->l_last_sync_lsn,
+ be64_to_cpu(iclog->ic_header.h_lsn)) <= 0);
+ log->l_last_sync_lsn =
+ 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;
repeats = 0;
xfs_fs_cmn_err(CE_WARN, log->l_mp,
"%s: possible infinite loop (%d iterations)",
- __FUNCTION__, flushcnt);
+ __func__, flushcnt);
}
} while (!ioerrors && loopdidcallbacks);
* 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.
*/
-void
+STATIC void
xlog_state_done_syncing(
xlog_in_core_t *iclog,
int aborted)
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);
* needs to be incremented, depending on the amount of data which
* is copied.
*/
-int
+STATIC int
xlog_state_get_iclog_space(xlog_t *log,
int len,
xlog_in_core_t **iclogp,
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
xlog_tic_add_region(ticket,
log->l_iclog_hsize,
XLOG_REG_TYPE_LRHEADER);
- INT_SET(head->h_cycle, ARCH_CONVERT, log->l_curr_cycle);
- INT_SET(head->h_lsn, ARCH_CONVERT,
+ head->h_cycle = cpu_to_be32(log->l_curr_cycle);
+ head->h_lsn = cpu_to_be64(
xlog_assign_lsn(log->l_curr_cycle, log->l_curr_block));
ASSERT(log->l_curr_block >= 0);
}
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 (atomic_read(&iclog->ic_refcnt) == 1) {
spin_unlock(&log->l_icloglock);
if ((error = xlog_state_release_iclog(log, iclog)))
return error;
} else {
- iclog->ic_refcnt--;
+ atomic_dec(&iclog->ic_refcnt);
spin_unlock(&log->l_icloglock);
}
goto restart;
} /* xlog_ungrant_log_space */
-/*
- * Atomically put back used ticket.
- */
-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.
*
*
*/
-int
-xlog_state_release_iclog(xlog_t *log,
- xlog_in_core_t *iclog)
+STATIC int
+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;
- INT_SET(iclog->ic_header.h_tail_lsn, ARCH_CONVERT, log->l_tail_lsn);
+ 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 */
if (!eventual_size)
eventual_size = iclog->ic_offset;
iclog->ic_state = XLOG_STATE_WANT_SYNC;
- INT_SET(iclog->ic_header.h_prev_block, ARCH_CONVERT, log->l_prev_block);
+ iclog->ic_header.h_prev_block = cpu_to_be32(log->l_prev_block);
log->l_prev_block = log->l_curr_block;
log->l_prev_cycle = log->l_curr_cycle;
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);
* 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++;
- lsn = INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT);
+ 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);
return XFS_ERROR(EIO);
*log_flushed = 1;
spin_lock(&log->l_icloglock);
- if (INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT) == lsn &&
+ if (be64_to_cpu(iclog->ic_header.h_lsn) == lsn &&
iclog->ic_state != XLOG_STATE_DIRTY)
goto maybe_sleep;
else
* If filesystem activity goes to zero, the iclog will get flushed only by
* bdflush().
*/
-int
+STATIC int
xlog_state_sync(xlog_t *log,
xfs_lsn_t lsn,
uint flags,
}
do {
- if (INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT) != lsn) {
- iclog = iclog->ic_next;
- continue;
+ if (be64_to_cpu(iclog->ic_header.h_lsn) != lsn) {
+ iclog = iclog->ic_next;
+ continue;
}
if (iclog->ic_state == XLOG_STATE_DIRTY) {
already_slept = 1;
goto try_again;
} else {
- iclog->ic_refcnt++;
+ atomic_inc(&iclog->ic_refcnt);
xlog_state_switch_iclogs(log, iclog, 0);
spin_unlock(&log->l_icloglock);
if (xlog_state_release_iclog(log, iclog))
* Called when we want to mark the current iclog as being ready to sync to
* disk.
*/
-void
+STATIC void
xlog_state_want_sync(xlog_t *log, xlog_in_core_t *iclog)
{
spin_lock(&log->l_icloglock);
*/
/*
- * Algorithm doesn't take into account page size. ;-(
- */
-STATIC void
-xlog_state_ticket_alloc(xlog_t *log)
-{
- 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++;
- }
- 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.
+ * Free a used ticket.
*/
STATIC void
xlog_ticket_put(xlog_t *log,
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++;
+ kmem_zone_free(xfs_log_ticket_zone, ticket);
} /* xlog_ticket_put */
/*
- * Grab ticket off freelist or allocation some more
+ * Allocate and initialise a new log ticket.
*/
-xlog_ticket_t *
+STATIC xlog_ticket_t *
xlog_ticket_get(xlog_t *log,
int unit_bytes,
int cnt,
xlog_ticket_t *tic;
uint num_headers;
- 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
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;
spin_unlock(&log->l_icloglock);
/* check log magic numbers */
- ptr = (xfs_caddr_t) &(iclog->ic_header);
- if (INT_GET(*(uint *)ptr, ARCH_CONVERT) != XLOG_HEADER_MAGIC_NUM)
+ if (be32_to_cpu(iclog->ic_header.h_magicno) != XLOG_HEADER_MAGIC_NUM)
xlog_panic("xlog_verify_iclog: invalid magic num");
- for (ptr += BBSIZE; ptr < ((xfs_caddr_t)&(iclog->ic_header))+count;
+ ptr = (xfs_caddr_t) &iclog->ic_header;
+ for (ptr += BBSIZE; ptr < ((xfs_caddr_t)&iclog->ic_header) + count;
ptr += BBSIZE) {
- if (INT_GET(*(uint *)ptr, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM)
+ if (be32_to_cpu(*(__be32 *)ptr) == XLOG_HEADER_MAGIC_NUM)
xlog_panic("xlog_verify_iclog: unexpected magic num");
}
/* check fields */
- len = INT_GET(iclog->ic_header.h_num_logops, ARCH_CONVERT);
+ len = be32_to_cpu(iclog->ic_header.h_num_logops);
ptr = iclog->ic_datap;
base_ptr = ptr;
ophead = (xlog_op_header_t *)ptr;
field_offset = (__psint_t)
((xfs_caddr_t)&(ophead->oh_len) - base_ptr);
if (syncing == B_FALSE || (field_offset & 0x1ff)) {
- op_len = INT_GET(ophead->oh_len, ARCH_CONVERT);
+ op_len = be32_to_cpu(ophead->oh_len);
} else {
idx = BTOBBT((__psint_t)&ophead->oh_len -
(__psint_t)iclog->ic_datap);
if (idx >= (XLOG_HEADER_CYCLE_SIZE / BBSIZE)) {
j = idx / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
k = idx % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
- op_len = INT_GET(xhdr[j].hic_xheader.xh_cycle_data[k], ARCH_CONVERT);
+ op_len = be32_to_cpu(xhdr[j].hic_xheader.xh_cycle_data[k]);
} else {
- op_len = INT_GET(iclog->ic_header.h_cycle_data[idx], ARCH_CONVERT);
+ op_len = be32_to_cpu(iclog->ic_header.h_cycle_data[idx]);
}
}
ptr += sizeof(xlog_op_header_t) + op_len;
* 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);
/*
projects / ~andy / linux / blobdiff