xfs: convert l_tail_lsn to an atomic variable.

log->l_tail_lsn is currently protected by the log grant lock. The
lock is only needed for serialising readers against writers, so we
don't really need the lock if we make the l_tail_lsn variable an
atomic. Converting the l_tail_lsn variable to an atomic64_t means we
can start to peel back the grant lock from various operations.

Also, provide functions to safely crack an atomic LSN variable into
it's component pieces and to recombined the components into an
atomic variable. Use them where appropriate.

This also removes the need for explicitly holding a spinlock to read
the l_tail_lsn on 32 bit platforms.

Signed-off-by: Dave Chinner <dchinner@redhat.com>

diff --git a/fs/xfs/linux-2.6/xfs_trace.h b/fs/xfs/linux-2.6/xfs_trace.h
index 3ff6b35f92072b1fe1c0d96fc357fbaea06198d7..b180e1bf8257a0f1fce9da0e7c62b2284e21ee54 100644 (file)
--- a/fs/xfs/linux-2.6/xfs_trace.h
+++ b/fs/xfs/linux-2.6/xfs_trace.h
@@ -794,7 +794,7 @@ DECLARE_EVENT_CLASS(xfs_loggrant_class,
                                &__entry->grant_write_bytes);
                __entry->curr_cycle = log->l_curr_cycle;
                __entry->curr_block = log->l_curr_block;
-               __entry->tail_lsn = log->l_tail_lsn;
+               __entry->tail_lsn = atomic64_read(&log->l_tail_lsn);
        ),
        TP_printk("dev %d:%d type %s t_ocnt %u t_cnt %u t_curr_res %u "
                  "t_unit_res %u t_flags %s reserveq %s "

diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index 70790eb48336b78a683192338ae36efc3ef392d5..d118bf8044809734d75c77a99244014062ecebc6 100644 (file)
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -678,15 +678,11 @@ xfs_log_move_tail(xfs_mount_t     *mp,
        if (tail_lsn == 0)
                tail_lsn = atomic64_read(&log->l_last_sync_lsn);
 
-       spin_lock(&log->l_grant_lock);
-
-       /* Also an invalid lsn.  1 implies that we aren't passing in a valid
-        * tail_lsn.
-        */
-       if (tail_lsn != 1) {
-               log->l_tail_lsn = tail_lsn;
-       }
+       /* tail_lsn == 1 implies that we weren't passed a valid value.  */
+       if (tail_lsn != 1)
+               atomic64_set(&log->l_tail_lsn, tail_lsn);
 
+       spin_lock(&log->l_grant_lock);
        if (!list_empty(&log->l_writeq)) {
 #ifdef DEBUG
                if (log->l_flags & XLOG_ACTIVE_RECOVERY)
@@ -789,21 +785,19 @@ xfs_log_need_covered(xfs_mount_t *mp)
  * We may be holding the log iclog lock upon entering this routine.
  */
 xfs_lsn_t
-xlog_assign_tail_lsn(xfs_mount_t *mp)
+xlog_assign_tail_lsn(
+       struct xfs_mount        *mp)
 {
-       xfs_lsn_t tail_lsn;
-       xlog_t    *log = mp->m_log;
+       xfs_lsn_t               tail_lsn;
+       struct log              *log = mp->m_log;
 
        tail_lsn = xfs_trans_ail_tail(mp->m_ail);
-       spin_lock(&log->l_grant_lock);
        if (!tail_lsn)
                tail_lsn = atomic64_read(&log->l_last_sync_lsn);
-       log->l_tail_lsn = tail_lsn;
-       spin_unlock(&log->l_grant_lock);
 
+       atomic64_set(&log->l_tail_lsn, tail_lsn);
        return tail_lsn;
-}      /* xlog_assign_tail_lsn */
-
+}
 
 /*
  * Return the space in the log between the tail and the head.  The head
@@ -831,8 +825,8 @@ xlog_space_left(
        int             head_bytes;
 
        xlog_crack_grant_head(head, &head_cycle, &head_bytes);
-       tail_bytes = BBTOB(BLOCK_LSN(log->l_tail_lsn));
-       tail_cycle = CYCLE_LSN(log->l_tail_lsn);
+       xlog_crack_atomic_lsn(&log->l_tail_lsn, &tail_cycle, &tail_bytes);
+       tail_bytes = BBTOB(tail_bytes);
        if (tail_cycle == head_cycle && head_bytes >= tail_bytes)
                free_bytes = log->l_logsize - (head_bytes - tail_bytes);
        else if (tail_cycle + 1 < head_cycle)
@@ -1009,8 +1003,8 @@ xlog_alloc_log(xfs_mount_t        *mp,
 
        log->l_prev_block  = -1;
        /* log->l_tail_lsn = 0x100000000LL; cycle = 1; current block = 0 */
-       log->l_tail_lsn    = xlog_assign_lsn(1, 0);
-       atomic64_set(&log->l_last_sync_lsn, xlog_assign_lsn(1, 0));
+       xlog_assign_atomic_lsn(&log->l_tail_lsn, 1, 0);
+       xlog_assign_atomic_lsn(&log->l_last_sync_lsn, 1, 0);
        log->l_curr_cycle  = 1;     /* 0 is bad since this is initial value */
        xlog_assign_grant_head(&log->l_grant_reserve_head, 1, 0);
        xlog_assign_grant_head(&log->l_grant_write_head, 1, 0);
@@ -1189,7 +1183,6 @@ xlog_grant_push_ail(
 {
        xfs_lsn_t       threshold_lsn = 0;
        xfs_lsn_t       last_sync_lsn;
-       xfs_lsn_t       tail_lsn;
        int             free_blocks;
        int             free_bytes;
        int             threshold_block;
@@ -1198,7 +1191,6 @@ xlog_grant_push_ail(
 
        ASSERT(BTOBB(need_bytes) < log->l_logBBsize);
 
-       tail_lsn = log->l_tail_lsn;
        free_bytes = xlog_space_left(log, &log->l_grant_reserve_head);
        free_blocks = BTOBBT(free_bytes);
 
@@ -1213,8 +1205,9 @@ xlog_grant_push_ail(
        if (free_blocks >= free_threshold)
                return;
 
-       threshold_block = BLOCK_LSN(tail_lsn) + free_threshold;
-       threshold_cycle = CYCLE_LSN(tail_lsn);
+       xlog_crack_atomic_lsn(&log->l_tail_lsn, &threshold_cycle,
+                                               &threshold_block);
+       threshold_block += free_threshold;
        if (threshold_block >= log->l_logBBsize) {
                threshold_block -= log->l_logBBsize;
                threshold_cycle += 1;
@@ -2828,11 +2821,11 @@ xlog_state_release_iclog(
 
        if (iclog->ic_state == XLOG_STATE_WANT_SYNC) {
                /* update tail before writing to iclog */
-               xlog_assign_tail_lsn(log->l_mp);
+               xfs_lsn_t tail_lsn = 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);
+               iclog->ic_header.h_tail_lsn = cpu_to_be64(tail_lsn);
+               xlog_verify_tail_lsn(log, iclog, tail_lsn);
                /* cycle incremented when incrementing curr_block */
        }
        spin_unlock(&log->l_icloglock);
@@ -3435,7 +3428,7 @@ STATIC void
 xlog_verify_grant_tail(
        struct log      *log)
 {
-       xfs_lsn_t       tail_lsn = log->l_tail_lsn;
+       int             tail_cycle, tail_blocks;
        int             cycle, space;
 
        /*
@@ -3445,9 +3438,10 @@ xlog_verify_grant_tail(
         * check the byte count.
         */
        xlog_crack_grant_head(&log->l_grant_write_head, &cycle, &space);
-       if (CYCLE_LSN(tail_lsn) != cycle) {
-               ASSERT(cycle - 1 == CYCLE_LSN(tail_lsn));
-               ASSERT(space <= BBTOB(BLOCK_LSN(tail_lsn)));
+       xlog_crack_atomic_lsn(&log->l_tail_lsn, &tail_cycle, &tail_blocks);
+       if (tail_cycle != cycle) {
+               ASSERT(cycle - 1 == tail_cycle);
+               ASSERT(space <= BBTOB(tail_blocks));
        }
 }
 

diff --git a/fs/xfs/xfs_log_priv.h b/fs/xfs/xfs_log_priv.h
index 958f356df10ed51cb6c33297554adc127c3b7657..d34af1c21ed275036e94594853d2d5827c54f8ac 100644 (file)
--- a/fs/xfs/xfs_log_priv.h
+++ b/fs/xfs/xfs_log_priv.h
@@ -53,7 +53,6 @@ struct xfs_mount;
        BTOBB(XLOG_MAX_ICLOGS << (xfs_sb_version_haslogv2(&log->l_mp->m_sb) ? \
         XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
 
-
 static inline xfs_lsn_t xlog_assign_lsn(uint cycle, uint block)
 {
        return ((xfs_lsn_t)cycle << 32) | block;
@@ -505,8 +504,6 @@ typedef struct log {
                                                 * log entries" */
        xlog_in_core_t          *l_iclog;       /* head log queue       */
        spinlock_t              l_icloglock;    /* grab to change iclog state */
-       xfs_lsn_t               l_tail_lsn;     /* lsn of 1st LR with unflushed
-                                                * buffers */
        int                     l_curr_cycle;   /* Cycle number of log writes */
        int                     l_prev_cycle;   /* Cycle number before last
                                                 * block increment */
@@ -521,12 +518,15 @@ typedef struct log {
        int64_t                 l_grant_write_head;
 
        /*
-        * l_last_sync_lsn is an atomic so it can be set and read without
-        * needing to hold specific locks. To avoid operations contending with
-        * other hot objects, place it on a separate cacheline.
+        * l_last_sync_lsn and l_tail_lsn are atomics so they can be set and
+        * read without needing to hold specific locks. To avoid operations
+        * contending with other hot objects, place each of them on a separate
+        * cacheline.
         */
        /* lsn of last LR on disk */
        atomic64_t              l_last_sync_lsn ____cacheline_aligned_in_smp;
+       /* lsn of 1st LR with unflushed * buffers */
+       atomic64_t              l_tail_lsn ____cacheline_aligned_in_smp;
 
        /* The following field are used for debugging; need to hold icloglock */
 #ifdef DEBUG
@@ -565,6 +565,31 @@ 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);
 
+/*
+ * When we crack an atomic LSN, we sample it first so that the value will not
+ * change while we are cracking it into the component values. This means we
+ * will always get consistent component values to work from. This should always
+ * be used to smaple and crack LSNs taht are stored and updated in atomic
+ * variables.
+ */
+static inline void
+xlog_crack_atomic_lsn(atomic64_t *lsn, uint *cycle, uint *block)
+{
+       xfs_lsn_t val = atomic64_read(lsn);
+
+       *cycle = CYCLE_LSN(val);
+       *block = BLOCK_LSN(val);
+}
+
+/*
+ * Calculate and assign a value to an atomic LSN variable from component pieces.
+ */
+static inline void
+xlog_assign_atomic_lsn(atomic64_t *lsn, uint cycle, uint block)
+{
+       atomic64_set(lsn, xlog_assign_lsn(cycle, block));
+}
+
 /*
  * When we crack the grrant head, we sample it first so that the value will not
  * change while we are cracking it into the component values. This means we

diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 18e1e18d7147173731df2a279f7fd43313016454..204d8e5fa7faf63ada5fc2135fefc2b7330d34ec 100644 (file)
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -936,7 +936,7 @@ xlog_find_tail(
        log->l_curr_cycle = be32_to_cpu(rhead->h_cycle);
        if (found == 2)
                log->l_curr_cycle++;
-       log->l_tail_lsn = be64_to_cpu(rhead->h_tail_lsn);
+       atomic64_set(&log->l_tail_lsn, be64_to_cpu(rhead->h_tail_lsn));
        atomic64_set(&log->l_last_sync_lsn, be64_to_cpu(rhead->h_lsn));
        xlog_assign_grant_head(&log->l_grant_reserve_head, log->l_curr_cycle,
                                        BBTOB(log->l_curr_block));
@@ -971,7 +971,7 @@ xlog_find_tail(
        }
        after_umount_blk = (i + hblks + (int)
                BTOBB(be32_to_cpu(rhead->h_len))) % log->l_logBBsize;
-       tail_lsn = log->l_tail_lsn;
+       tail_lsn = atomic64_read(&log->l_tail_lsn);
        if (*head_blk == after_umount_blk &&
            be32_to_cpu(rhead->h_num_logops) == 1) {
                umount_data_blk = (i + hblks) % log->l_logBBsize;
@@ -986,12 +986,10 @@ xlog_find_tail(
                         * log records will point recovery to after the
                         * current unmount record.
                         */
-                       log->l_tail_lsn =
-                               xlog_assign_lsn(log->l_curr_cycle,
-                                               after_umount_blk);
-                       atomic64_set(&log->l_last_sync_lsn,
-                               xlog_assign_lsn(log->l_curr_cycle,
-                                               after_umount_blk));
+                       xlog_assign_atomic_lsn(&log->l_tail_lsn,
+                                       log->l_curr_cycle, after_umount_blk);
+                       xlog_assign_atomic_lsn(&log->l_last_sync_lsn,
+                                       log->l_curr_cycle, after_umount_blk);
                        *tail_blk = after_umount_blk;
 
                        /*