Illumos #2619 and #2747

[zfs.git] / module / zfs / zio.c

diff --git a/module/zfs/zio.c b/module/zfs/zio.c
index fe2bdc8..66f228b 100644 (file)
--- a/module/zfs/zio.c
+++ b/module/zfs/zio.c
@@ -259,7 +259,7 @@ zio_buf_alloc(size_t size)
 
        ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
 
-       return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
+       return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE | KM_NODEBUG));
 }
 
 /*
@@ -275,7 +275,8 @@ zio_data_buf_alloc(size_t size)
 
        ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
 
-       return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
+       return (kmem_cache_alloc(zio_data_buf_cache[c],
+           KM_PUSHPAGE | KM_NODEBUG));
 }
 
 void
@@ -703,7 +704,7 @@ zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
            zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
            zp->zp_compress >= ZIO_COMPRESS_OFF &&
            zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
-           zp->zp_type < DMU_OT_NUMTYPES &&
+           DMU_OT_IS_VALID(zp->zp_type) &&
            zp->zp_level < 32 &&
            zp->zp_copies > 0 &&
            zp->zp_copies <= spa_max_replication(spa) &&
@@ -987,7 +988,7 @@ zio_read_bp_init(zio_t *zio)
                zio_push_transform(zio, cbuf, psize, psize, zio_decompress);
        }
 
-       if (!dmu_ot[BP_GET_TYPE(bp)].ot_metadata && BP_GET_LEVEL(bp) == 0)
+       if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0)
                zio->io_flags |= ZIO_FLAG_DONT_CACHE;
 
        if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP)
@@ -1315,7 +1316,7 @@ zio_wait(zio_t *zio)
 
        mutex_enter(&zio->io_lock);
        while (zio->io_executor != NULL)
-               cv_wait(&zio->io_cv, &zio->io_lock);
+               cv_wait_io(&zio->io_cv, &zio->io_lock);
        mutex_exit(&zio->io_lock);
 
        error = zio->io_error;
@@ -1860,6 +1861,11 @@ zio_write_gang_block(zio_t *pio)
         */
        pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
 
+       /*
+        * We didn't allocate this bp, so make sure it doesn't get unmarked.
+        */
+       pio->io_flags &= ~ZIO_FLAG_FASTWRITE;
+
        zio_nowait(zio);
 
        return (ZIO_PIPELINE_CONTINUE);
@@ -2269,6 +2275,7 @@ zio_dva_allocate(zio_t *zio)
        flags |= (zio->io_flags & ZIO_FLAG_NODATA) ? METASLAB_GANG_AVOID : 0;
        flags |= (zio->io_flags & ZIO_FLAG_GANG_CHILD) ?
            METASLAB_GANG_CHILD : 0;
+       flags |= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0;
        error = metaslab_alloc(spa, mc, zio->io_size, bp,
            zio->io_prop.zp_copies, zio->io_txg, NULL, flags);
 
@@ -2332,8 +2339,8 @@ zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp)
  * Try to allocate an intent log block.  Return 0 on success, errno on failure.
  */
 int
-zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp,
-    uint64_t size, boolean_t use_slog)
+zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, uint64_t size,
+    boolean_t use_slog)
 {
        int error = 1;
 
@@ -2346,14 +2353,14 @@ zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp,
         */
        if (use_slog) {
                error = metaslab_alloc(spa, spa_log_class(spa), size,
-                   new_bp, 1, txg, old_bp,
-                   METASLAB_HINTBP_AVOID | METASLAB_GANG_AVOID);
+                   new_bp, 1, txg, NULL,
+                   METASLAB_FASTWRITE | METASLAB_GANG_AVOID);
        }
 
        if (error) {
                error = metaslab_alloc(spa, spa_normal_class(spa), size,
-                   new_bp, 1, txg, old_bp,
-                   METASLAB_HINTBP_AVOID | METASLAB_GANG_AVOID);
+                   new_bp, 1, txg, NULL,
+                   METASLAB_FASTWRITE | METASLAB_GANG_AVOID);
        }
 
        if (error == 0) {
@@ -3059,6 +3066,11 @@ zio_done(zio_t *zio)
                zfs_ereport_free_checksum(zcr);
        }
 
+       if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
+           !BP_IS_HOLE(zio->io_bp)) {
+               metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp);
+       }
+
        /*
         * It is the responsibility of the done callback to ensure that this
         * particular zio is no longer discoverable for adoption, and as
@@ -3119,6 +3131,48 @@ static zio_pipe_stage_t *zio_pipeline[] = {
        zio_done
 };
 
+/* dnp is the dnode for zb1->zb_object */
+boolean_t
+zbookmark_is_before(const dnode_phys_t *dnp, const zbookmark_t *zb1,
+    const zbookmark_t *zb2)
+{
+       uint64_t zb1nextL0, zb2thisobj;
+
+       ASSERT(zb1->zb_objset == zb2->zb_objset);
+       ASSERT(zb2->zb_level == 0);
+
+       /*
+        * A bookmark in the deadlist is considered to be after
+        * everything else.
+        */
+       if (zb2->zb_object == DMU_DEADLIST_OBJECT)
+               return (B_TRUE);
+
+       /* The objset_phys_t isn't before anything. */
+       if (dnp == NULL)
+               return (B_FALSE);
+
+       zb1nextL0 = (zb1->zb_blkid + 1) <<
+           ((zb1->zb_level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT));
+
+       zb2thisobj = zb2->zb_object ? zb2->zb_object :
+           zb2->zb_blkid << (DNODE_BLOCK_SHIFT - DNODE_SHIFT);
+
+       if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
+               uint64_t nextobj = zb1nextL0 *
+                   (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT) >> DNODE_SHIFT;
+               return (nextobj <= zb2thisobj);
+       }
+
+       if (zb1->zb_object < zb2thisobj)
+               return (B_TRUE);
+       if (zb1->zb_object > zb2thisobj)
+               return (B_FALSE);
+       if (zb2->zb_object == DMU_META_DNODE_OBJECT)
+               return (B_FALSE);
+       return (zb1nextL0 <= zb2->zb_blkid);
+}
+
 #if defined(_KERNEL) && defined(HAVE_SPL)
 /* Fault injection */
 EXPORT_SYMBOL(zio_injection_enabled);