zfs.patch

Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c	(working copy)
@@ -344,6 +344,7 @@ typedef struct arc_stats {
 	kstat_named_t arcstat_l2_evict_lock_retry;
 	kstat_named_t arcstat_l2_evict_reading;
 	kstat_named_t arcstat_l2_free_on_write;
+	kstat_named_t arcstat_l2_cdata_free_on_write;
 	kstat_named_t arcstat_l2_abort_lowmem;
 	kstat_named_t arcstat_l2_cksum_bad;
 	kstat_named_t arcstat_l2_io_error;
@@ -421,6 +422,7 @@ static arc_stats_t arc_stats = {
 	{ "l2_evict_lock_retry",	KSTAT_DATA_UINT64 },
 	{ "l2_evict_reading",		KSTAT_DATA_UINT64 },
 	{ "l2_free_on_write",		KSTAT_DATA_UINT64 },
+	{ "l2_cdata_free_on_write",	KSTAT_DATA_UINT64 },
 	{ "l2_abort_lowmem",		KSTAT_DATA_UINT64 },
 	{ "l2_cksum_bad",		KSTAT_DATA_UINT64 },
 	{ "l2_io_error",		KSTAT_DATA_UINT64 },
@@ -1629,6 +1631,21 @@ arc_buf_add_ref(arc_buf_t *buf, void* tag)
 	    data, metadata, hits);
 }
 
+static void
+arc_buf_free_on_write(void *data, size_t size,
+    void (*free_func)(void *, size_t))
+{
+	l2arc_data_free_t *df;
+
+	df = kmem_alloc(sizeof (l2arc_data_free_t), KM_SLEEP);
+	df->l2df_data = data;
+	df->l2df_size = size;
+	df->l2df_func = free_func;
+	mutex_enter(&l2arc_free_on_write_mtx);
+	list_insert_head(l2arc_free_on_write, df);
+	mutex_exit(&l2arc_free_on_write_mtx);
+}
+
 /*
  * Free the arc data buffer.  If it is an l2arc write in progress,
  * the buffer is placed on l2arc_free_on_write to be freed later.
@@ -1639,14 +1656,7 @@ arc_buf_data_free(arc_buf_t *buf, void (*free_func
 	arc_buf_hdr_t *hdr = buf->b_hdr;
 
 	if (HDR_L2_WRITING(hdr)) {
-		l2arc_data_free_t *df;
-		df = kmem_alloc(sizeof (l2arc_data_free_t), KM_SLEEP);
-		df->l2df_data = buf->b_data;
-		df->l2df_size = hdr->b_size;
-		df->l2df_func = free_func;
-		mutex_enter(&l2arc_free_on_write_mtx);
-		list_insert_head(l2arc_free_on_write, df);
-		mutex_exit(&l2arc_free_on_write_mtx);
+		arc_buf_free_on_write(buf->b_data, hdr->b_size, free_func);
 		ARCSTAT_BUMP(arcstat_l2_free_on_write);
 	} else {
 		free_func(buf->b_data, hdr->b_size);
@@ -1658,6 +1668,23 @@ arc_buf_data_free(arc_buf_t *buf, void (*free_func
  * arc_buf_t off of the the arc_buf_hdr_t's list and free it.
  */
 static void
+arc_buf_l2_cdata_free(arc_buf_hdr_t *hdr)
+{
+	l2arc_buf_hdr_t *l2hdr = hdr->b_l2hdr;
+
+	ASSERT(MUTEX_HELD(&l2arc_buflist_mtx));
+
+	if (l2hdr->b_tmp_cdata == NULL)
+		return;
+
+	ASSERT(HDR_L2_WRITING(hdr));
+	arc_buf_free_on_write(l2hdr->b_tmp_cdata, hdr->b_size,
+	    zio_data_buf_free);
+	ARCSTAT_BUMP(arcstat_l2_cdata_free_on_write);
+	l2hdr->b_tmp_cdata = NULL;
+}
+
+static void
 arc_buf_destroy(arc_buf_t *buf, boolean_t recycle, boolean_t remove)
 {
 	arc_buf_t **bufp;
@@ -1756,6 +1783,7 @@ arc_hdr_destroy(arc_buf_hdr_t *hdr)
 			trim_map_free(l2hdr->b_dev->l2ad_vdev, l2hdr->b_daddr,
 			    hdr->b_size, 0);
 			list_remove(l2hdr->b_dev->l2ad_buflist, hdr);
+			arc_buf_l2_cdata_free(hdr);
 			ARCSTAT_INCR(arcstat_l2_size, -hdr->b_size);
 			ARCSTAT_INCR(arcstat_l2_asize, -l2hdr->b_asize);
 			vdev_space_update(l2hdr->b_dev->l2ad_vdev,
@@ -3605,6 +3633,7 @@ arc_release(arc_buf_t *buf, void *tag)
 	l2hdr = hdr->b_l2hdr;
 	if (l2hdr) {
 		mutex_enter(&l2arc_buflist_mtx);
+		arc_buf_l2_cdata_free(hdr);
 		hdr->b_l2hdr = NULL;
 		list_remove(l2hdr->b_dev->l2ad_buflist, hdr);
 	}
@@ -4895,6 +4924,11 @@ top:
 				ARCSTAT_INCR(arcstat_l2_asize, -abl2->b_asize);
 				bytes_evicted += abl2->b_asize;
 				ab->b_l2hdr = NULL;
+				/*
+				 * We are destroying l2hdr, so ensure that
+				 * its compressed buffer, if any, is not leaked.
+				 */
+				ASSERT(abl2->b_tmp_cdata == NULL);
 				kmem_free(abl2, sizeof (l2arc_buf_hdr_t));
 				ARCSTAT_INCR(arcstat_l2_size, -ab->b_size);
 			}
@@ -5133,6 +5167,14 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev,
 		buf_data = l2hdr->b_tmp_cdata;
 		buf_sz = l2hdr->b_asize;
 
+		/*
+		 * If the data has not been compressed, then clear b_tmp_cdata
+		 * to make sure that it points only to a temporary compression
+		 * buffer.
+		 */
+		if (!L2ARC_IS_VALID_COMPRESS(l2hdr->b_compress))
+			l2hdr->b_tmp_cdata = NULL;
+
 		/* Compression may have squashed the buffer to zero length. */
 		if (buf_sz != 0) {
 			uint64_t buf_p_sz;
@@ -5323,7 +5365,8 @@ l2arc_release_cdata_buf(arc_buf_hdr_t *ab)
 {
 	l2arc_buf_hdr_t *l2hdr = ab->b_l2hdr;
 
-	if (l2hdr->b_compress == ZIO_COMPRESS_LZ4) {
+	ASSERT(L2ARC_IS_VALID_COMPRESS(l2hdr->b_compress));
+	if (l2hdr->b_compress != ZIO_COMPRESS_EMPTY) {
 		/*
 		 * If the data was compressed, then we've allocated a
 		 * temporary buffer for it, so now we need to release it.
@@ -5330,8 +5373,10 @@ l2arc_release_cdata_buf(arc_buf_hdr_t *ab)
 		 */
 		ASSERT(l2hdr->b_tmp_cdata != NULL);
 		zio_data_buf_free(l2hdr->b_tmp_cdata, ab->b_size);
+		l2hdr->b_tmp_cdata = NULL;
+	} else {
+		ASSERT(l2hdr->b_tmp_cdata == NULL);
 	}
-	l2hdr->b_tmp_cdata = NULL;
 }
 
 /*
Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/trim_map.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/trim_map.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/trim_map.c	(working copy)
@@ -155,11 +155,9 @@ trim_map_create(vdev_t *vd)
 {
 	trim_map_t *tm;
 
-	ASSERT(vd->vdev_ops->vdev_op_leaf);
+	ASSERT(zfs_trim_enabled && !vd->vdev_notrim &&
+		vd->vdev_ops->vdev_op_leaf);
 
-	if (!zfs_trim_enabled)
-		return;
-
 	tm = kmem_zalloc(sizeof (*tm), KM_SLEEP);
 	mutex_init(&tm->tm_lock, NULL, MUTEX_DEFAULT, NULL);
 	list_create(&tm->tm_head, sizeof (trim_seg_t),
Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c	(working copy)
@@ -1214,6 +1214,7 @@ vdev_open(vdev_t *vd)
 	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
 	vd->vdev_cant_read = B_FALSE;
 	vd->vdev_cant_write = B_FALSE;
+	vd->vdev_notrim = B_FALSE;
 	vd->vdev_min_asize = vdev_get_min_asize(vd);
 
 	/*
@@ -1283,10 +1284,8 @@ vdev_open(vdev_t *vd)
 	if (vd->vdev_ishole || vd->vdev_ops == &vdev_missing_ops)
 		return (0);
 
-	if (vd->vdev_ops->vdev_op_leaf) {
-		vd->vdev_notrim = B_FALSE;
+	if (zfs_trim_enabled && !vd->vdev_notrim && vd->vdev_ops->vdev_op_leaf)
 		trim_map_create(vd);
-	}
 
 	for (int c = 0; c < vd->vdev_children; c++) {
 		if (vd->vdev_child[c]->vdev_state != VDEV_STATE_HEALTHY) {
Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_disk.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_disk.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_disk.c	(working copy)
@@ -796,6 +796,8 @@ vdev_disk_io_start(zio_t *zio)
 		return (ZIO_PIPELINE_STOP);
 	}
 
+	ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
+
 	vb = kmem_alloc(sizeof (vdev_buf_t), KM_SLEEP);
 
 	vb->vb_io = zio;
Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_file.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_file.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_file.c	(working copy)
@@ -129,6 +129,8 @@ skip_open:
 		return (error);
 	}
 
+	vd->vdev_notrim = B_TRUE;
+
 	*max_psize = *psize = vattr.va_size;
 	*logical_ashift = SPA_MINBLOCKSHIFT;
 	*physical_ashift = SPA_MINBLOCKSHIFT;
@@ -185,6 +187,8 @@ vdev_file_io_start(zio_t *zio)
 		return (ZIO_PIPELINE_STOP);
 	}
 
+	ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
+
 	zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
 	    UIO_READ : UIO_WRITE, vp, zio->io_data, zio->io_size,
 	    zio->io_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c	(working copy)
@@ -832,6 +832,11 @@ vdev_geom_io_start(zio_t *zio)
 		return (ZIO_PIPELINE_STOP);
 	}
 sendreq:
+	ASSERT(zio->io_type == ZIO_TYPE_READ ||
+	    zio->io_type == ZIO_TYPE_WRITE ||
+	    zio->io_type == ZIO_TYPE_FREE ||
+	    zio->io_type == ZIO_TYPE_IOCTL);
+
 	cp = vd->vdev_tsd;
 	if (cp == NULL) {
 		zio->io_error = SET_ERROR(ENXIO);
Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c
===================================================================
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c	(revision 284174)
+++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c	(working copy)
@@ -713,8 +713,9 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_l
 	 * Don't TRIM if removing so that we don't interfere with zpool
 	 * disaster recovery.
 	 */
-	if (zfs_trim_enabled && vdev_trim_on_init && (reason == VDEV_LABEL_CREATE ||
-	    reason == VDEV_LABEL_SPARE || reason == VDEV_LABEL_L2CACHE))
+	if (zfs_trim_enabled && vdev_trim_on_init && !vd->vdev_notrim && 
+	    (reason == VDEV_LABEL_CREATE || reason == VDEV_LABEL_SPARE ||
+	    reason == VDEV_LABEL_L2CACHE))
 		zio_wait(zio_trim(NULL, spa, vd, 0, vd->vdev_psize));
 
 	/*