[illumos-gate merge]

commit b515258426fed6c7311fd3f1dea697cfbd4085c6
    5027 zfs large block support
commit f3477bb2505a35e160b09c9dbf5476112be4fa7d
    4457 we apparently change .comment of almost every userland object
commit 738f37bc3dcd61e8a893af0f2d466d76690b70ec
    5244 zio pipeline callers should explicitly invoke next stage
commit f7950bf1145637c6dc57742a8bb95631fd5c846f
    5243 zdb -b could be much faster
commit cffcfaee1e6b29ef9ceb7d80e4e053ffd029906b
    5270 ld(1) cannot handle CIE version 3 in .eh_frame

Conflicts:
	usr/src/man/man1m/zfs.1m

6 files changed, 50 insertions, 19 deletions

diff --git a/usr/src/lib/libzfs/common/libzfs.h b/usr/src/lib/libzfs/common/libzfs.h
index ef5224c763..eb8ae6ceb7 100644
--- a/usr/src/lib/libzfs/common/libzfs.h
+++ b/usr/src/lib/libzfs/common/libzfs.h
@@ -593,6 +593,9 @@ typedef struct sendflags {
 	/* show progress (ie. -v) */
 	boolean_t progress;
 
+	/* large blocks (>128K) are permitted */
+	boolean_t largeblock;
+
 	/* WRITE_EMBEDDED records of type DATA are permitted */
 	boolean_t embed_data;
 } sendflags_t;
diff --git a/usr/src/lib/libzfs/common/libzfs_dataset.c b/usr/src/lib/libzfs/common/libzfs_dataset.c
index 75a391de0f..1172abc647 100644
--- a/usr/src/lib/libzfs/common/libzfs_dataset.c
+++ b/usr/src/lib/libzfs/common/libzfs_dataset.c
@@ -1052,21 +1052,36 @@ zfs_valid_proplist(libzfs_handle_t *hdl, zfs_type_t type, nvlist_t *nvl,
 			break;
 		}
 
-		case ZFS_PROP_RECORDSIZE:
 		case ZFS_PROP_VOLBLOCKSIZE:
-			/* must be power of two within SPA_{MIN,MAX}BLOCKSIZE */
+		case ZFS_PROP_RECORDSIZE:
+		{
+			int maxbs = SPA_MAXBLOCKSIZE;
+			if (zhp != NULL) {
+				maxbs = zpool_get_prop_int(zhp->zpool_hdl,
+				    ZPOOL_PROP_MAXBLOCKSIZE, NULL);
+			}
+			/*
+			 * Volumes are limited to a volblocksize of 128KB,
+			 * because they typically service workloads with
+			 * small random writes, which incur a large performance
+			 * penalty with large blocks.
+			 */
+			if (prop == ZFS_PROP_VOLBLOCKSIZE)
+				maxbs = SPA_OLD_MAXBLOCKSIZE;
+			/*
+			 * The value must be a power of two between
+			 * SPA_MINBLOCKSIZE and maxbs.
+			 */
 			if (intval < SPA_MINBLOCKSIZE ||
-			    intval > SPA_MAXBLOCKSIZE || !ISP2(intval)) {
+			    intval > maxbs || !ISP2(intval)) {
 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
-				    "'%s' must be power of 2 from %u "
-				    "to %uk"), propname,
-				    (uint_t)SPA_MINBLOCKSIZE,
-				    (uint_t)SPA_MAXBLOCKSIZE >> 10);
+				    "'%s' must be power of 2 from 512B "
+				    "to %uKB"), propname, maxbs >> 10);
 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 				goto error;
 			}
 			break;
-
+		}
 		case ZFS_PROP_MLSLABEL:
 		{
 			/*
@@ -1441,7 +1456,8 @@ zfs_setprop_error(libzfs_handle_t *hdl, zfs_prop_t prop, int err,
 		break;
 
 	case ERANGE:
-		if (prop == ZFS_PROP_COMPRESSION) {
+		if (prop == ZFS_PROP_COMPRESSION ||
+		    prop == ZFS_PROP_RECORDSIZE) {
 			(void) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "property setting is not allowed on "
 			    "bootable datasets"));
@@ -3156,9 +3172,7 @@ zfs_create(libzfs_handle_t *hdl, const char *path, zfs_type_t type,
 		case EDOM:
 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 			    "volume block size must be power of 2 from "
-			    "%u to %uk"),
-			    (uint_t)SPA_MINBLOCKSIZE,
-			    (uint_t)SPA_MAXBLOCKSIZE >> 10);
+			    "512B to 128KB"));
 
 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
 
diff --git a/usr/src/lib/libzfs/common/libzfs_sendrecv.c b/usr/src/lib/libzfs/common/libzfs_sendrecv.c
index 6697b52831..c4944438aa 100644
--- a/usr/src/lib/libzfs/common/libzfs_sendrecv.c
+++ b/usr/src/lib/libzfs/common/libzfs_sendrecv.c
@@ -206,7 +206,7 @@ static void *
 cksummer(void *arg)
 {
 	dedup_arg_t *dda = arg;
-	char *buf = malloc(1<<20);
+	char *buf = zfs_alloc(dda->dedup_hdl, SPA_MAXBLOCKSIZE);
 	dmu_replay_record_t thedrr;
 	dmu_replay_record_t *drr = &thedrr;
 	struct drr_begin *drrb = &thedrr.drr_u.drr_begin;
@@ -271,9 +271,9 @@ cksummer(void *arg)
 			    DMU_COMPOUNDSTREAM && drr->drr_payloadlen != 0) {
 				int sz = drr->drr_payloadlen;
 
-				if (sz > 1<<20) {
-					free(buf);
-					buf = malloc(sz);
+				if (sz > SPA_MAXBLOCKSIZE) {
+					buf = zfs_realloc(dda->dedup_hdl, buf,
+					    SPA_MAXBLOCKSIZE, sz);
 				}
 				(void) ssread(buf, sz, ofp);
 				if (ferror(stdin))
@@ -806,7 +806,7 @@ typedef struct send_dump_data {
 	char prevsnap[ZFS_MAXNAMELEN];
 	uint64_t prevsnap_obj;
 	boolean_t seenfrom, seento, replicate, doall, fromorigin;
-	boolean_t verbose, dryrun, parsable, progress, embed_data;
+	boolean_t verbose, dryrun, parsable, progress, embed_data, large_block;
 	int outfd;
 	boolean_t err;
 	nvlist_t *fss;
@@ -1153,6 +1153,8 @@ dump_snapshot(zfs_handle_t *zhp, void *arg)
 		}
 
 		enum lzc_send_flags flags = 0;
+		if (sdd->large_block)
+			flags |= LZC_SEND_FLAG_LARGE_BLOCK;
 		if (sdd->embed_data)
 			flags |= LZC_SEND_FLAG_EMBED_DATA;
 
@@ -1501,6 +1503,7 @@ zfs_send(zfs_handle_t *zhp, const char *fromsnap, const char *tosnap,
 	sdd.parsable = flags->parsable;
 	sdd.progress = flags->progress;
 	sdd.dryrun = flags->dryrun;
+	sdd.large_block = flags->largeblock;
 	sdd.embed_data = flags->embed_data;
 	sdd.filter_cb = filter_func;
 	sdd.filter_cb_arg = cb_arg;
@@ -2506,7 +2509,7 @@ static int
 recv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap)
 {
 	dmu_replay_record_t *drr;
-	void *buf = malloc(1<<20);
+	void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);
 	char errbuf[1024];
 
 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
diff --git a/usr/src/lib/libzfs_core/common/libzfs_core.c b/usr/src/lib/libzfs_core/common/libzfs_core.c
index 6f36568667..06221fab4a 100644
--- a/usr/src/lib/libzfs_core/common/libzfs_core.c
+++ b/usr/src/lib/libzfs_core/common/libzfs_core.c
@@ -455,6 +455,10 @@ lzc_get_holds(const char *snapname, nvlist_t **holdsp)
  *
  * "fd" is the file descriptor to write the send stream to.
  *
+ * If "flags" contains LZC_SEND_FLAG_LARGE_BLOCK, the stream is permitted
+ * to contain DRR_WRITE records with drr_length > 128K, and DRR_OBJECT
+ * records with drr_blksz > 128K.
+ *
  * If "flags" contains LZC_SEND_FLAG_EMBED_DATA, the stream is permitted
  * to contain DRR_WRITE_EMBEDDED records with drr_etype==BP_EMBEDDED_TYPE_DATA,
  * which the receiving system must support (as indicated by support
@@ -471,6 +475,8 @@ lzc_send(const char *snapname, const char *from, int fd,
 	fnvlist_add_int32(args, "fd", fd);
 	if (from != NULL)
 		fnvlist_add_string(args, "fromsnap", from);
+	if (flags & LZC_SEND_FLAG_LARGE_BLOCK)
+		fnvlist_add_boolean(args, "largeblockok");
 	if (flags & LZC_SEND_FLAG_EMBED_DATA)
 		fnvlist_add_boolean(args, "embedok");
 	err = lzc_ioctl(ZFS_IOC_SEND_NEW, snapname, args, NULL);
diff --git a/usr/src/lib/libzfs_core/common/libzfs_core.h b/usr/src/lib/libzfs_core/common/libzfs_core.h
index d7d767055d..bdd6c951ee 100644
--- a/usr/src/lib/libzfs_core/common/libzfs_core.h
+++ b/usr/src/lib/libzfs_core/common/libzfs_core.h
@@ -53,7 +53,8 @@ int lzc_release(nvlist_t *, nvlist_t **);
 int lzc_get_holds(const char *, nvlist_t **);
 
 enum lzc_send_flags {
-	LZC_SEND_FLAG_EMBED_DATA = 1 << 0
+	LZC_SEND_FLAG_EMBED_DATA = 1 << 0,
+	LZC_SEND_FLAG_LARGE_BLOCK = 1 << 1
 };
 
 int lzc_send(const char *, const char *, int, enum lzc_send_flags);
diff --git a/usr/src/lib/libzpool/common/taskq.c b/usr/src/lib/libzpool/common/taskq.c
index 2c5dfd86dc..a4ab58963d 100644
--- a/usr/src/lib/libzpool/common/taskq.c
+++ b/usr/src/lib/libzpool/common/taskq.c
@@ -25,6 +25,7 @@
 /*
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright 2012 Garrett D'Amore <garrett@damore.org>.  All rights reserved.
+ * Copyright (c) 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -33,8 +34,10 @@ int taskq_now;
 taskq_t *system_taskq;
 
 #define	TASKQ_ACTIVE	0x00010000
+#define	TASKQ_NAMELEN	31
 
 struct taskq {
+	char		tq_name[TASKQ_NAMELEN + 1];
 	kmutex_t	tq_lock;
 	krwlock_t	tq_threadlock;
 	kcondvar_t	tq_dispatch_cv;
@@ -247,6 +250,7 @@ taskq_create(const char *name, int nthreads, pri_t pri,
 	cv_init(&tq->tq_dispatch_cv, NULL, CV_DEFAULT, NULL);
 	cv_init(&tq->tq_wait_cv, NULL, CV_DEFAULT, NULL);
 	cv_init(&tq->tq_maxalloc_cv, NULL, CV_DEFAULT, NULL);
+	(void) strncpy(tq->tq_name, name, TASKQ_NAMELEN + 1);
 	tq->tq_flags = flags | TASKQ_ACTIVE;
 	tq->tq_active = nthreads;
 	tq->tq_nthreads = nthreads;