PSARC 2002/240 ZFS

6338653 Integrate ZFS
PSARC 2004/652 - DKIOCFLUSH
5096886 Write caching disks need mechanism to flush cache to physical media

10 files changed, 228 insertions, 27 deletions

diff --git a/usr/src/uts/common/os/autoconf.c b/usr/src/uts/common/os/autoconf.c
index 9be275c03b..6127073b28 100644
--- a/usr/src/uts/common/os/autoconf.c
+++ b/usr/src/uts/common/os/autoconf.c
@@ -208,11 +208,11 @@ getlongprop_buf(int id, char *name, char *buf, int maxlen)
 {
 	int size;
 
-	size = prom_getproplen((dnode_t)id, name);
+	size = prom_getproplen((pnode_t)id, name);
 	if (size <= 0 || (size > maxlen - 1))
 		return (-1);
 
-	if (-1 == prom_getprop((dnode_t)id, name, buf))
+	if (-1 == prom_getprop((pnode_t)id, name, buf))
 		return (-1);
 
 	/*
@@ -246,14 +246,14 @@ get_neighbors(dev_info_t *di, int flag)
 	snid = cnid = 0;
 	switch (flag) {
 		case DDI_WALK_PRUNESIB:
-			cnid = (int)prom_childnode((dnode_t)nid);
+			cnid = (int)prom_childnode((pnode_t)nid);
 			break;
 		case DDI_WALK_PRUNECHILD:
-			snid = (int)prom_nextnode((dnode_t)nid);
+			snid = (int)prom_nextnode((pnode_t)nid);
 			break;
 		case 0:
-			snid = (int)prom_nextnode((dnode_t)nid);
-			cnid = (int)prom_childnode((dnode_t)nid);
+			snid = (int)prom_nextnode((pnode_t)nid);
+			cnid = (int)prom_childnode((pnode_t)nid);
 			break;
 		default:
 			return (DDI_WALK_TERMINATE);
@@ -265,7 +265,7 @@ get_neighbors(dev_info_t *di, int flag)
 		 * add the first sibling that passes check_status()
 		 */
 		for (; snid && (snid != -1);
-		    snid = (int)prom_nextnode((dnode_t)snid)) {
+		    snid = (int)prom_nextnode((pnode_t)snid)) {
 			if (getlongprop_buf(snid, OBP_NAME, buf,
 			    sizeof (buf)) > 0) {
 				if (check_status(snid, buf, parent) ==
@@ -286,9 +286,9 @@ get_neighbors(dev_info_t *di, int flag)
 			if (check_status(cnid, buf, di) == DDI_SUCCESS) {
 				(void) ddi_add_child(di, buf, cnid, -1);
 			} else {
-				for (cnid = (int)prom_nextnode((dnode_t)cnid);
+				for (cnid = (int)prom_nextnode((pnode_t)cnid);
 				    cnid && (cnid != -1);
-				    cnid = (int)prom_nextnode((dnode_t)cnid)) {
+				    cnid = (int)prom_nextnode((pnode_t)cnid)) {
 					if (getlongprop_buf(cnid, OBP_NAME,
 					    buf, sizeof (buf)) > 0) {
 						if (check_status(cnid, buf, di)
@@ -352,7 +352,7 @@ static void
 create_devinfo_tree(void)
 {
 	major_t major;
-	dnode_t nodeid;
+	pnode_t nodeid;
 
 	i_ddi_node_cache_init();
 #if defined(__sparc)
diff --git a/usr/src/uts/common/os/devcfg.c b/usr/src/uts/common/os/devcfg.c
index c11cbcdf86..deabd79f37 100644
--- a/usr/src/uts/common/os/devcfg.c
+++ b/usr/src/uts/common/os/devcfg.c
@@ -80,7 +80,7 @@ struct mt_config_handle {
 };
 
 struct devi_nodeid {
-	dnode_t nodeid;
+	pnode_t nodeid;
 	dev_info_t *dip;
 	struct devi_nodeid *next;
 };
@@ -196,7 +196,7 @@ i_ddi_node_cache_init()
  * The allocated node has a reference count of 0.
  */
 dev_info_t *
-i_ddi_alloc_node(dev_info_t *pdip, char *node_name, dnode_t nodeid,
+i_ddi_alloc_node(dev_info_t *pdip, char *node_name, pnode_t nodeid,
     int instance, ddi_prop_t *sys_prop, int flag)
 {
 	struct dev_info *devi;
@@ -1666,7 +1666,7 @@ ndi_devi_tryenter(dev_info_t *dip, int *circular)
  * not allowed to sleep.
  */
 int
-ndi_devi_alloc(dev_info_t *parent, char *node_name, dnode_t nodeid,
+ndi_devi_alloc(dev_info_t *parent, char *node_name, pnode_t nodeid,
     dev_info_t **ret_dip)
 {
 	ASSERT(node_name != NULL);
@@ -1686,7 +1686,7 @@ ndi_devi_alloc(dev_info_t *parent, char *node_name, dnode_t nodeid,
  * This routine may sleep and should not be called at interrupt time
  */
 void
-ndi_devi_alloc_sleep(dev_info_t *parent, char *node_name, dnode_t nodeid,
+ndi_devi_alloc_sleep(dev_info_t *parent, char *node_name, pnode_t nodeid,
     dev_info_t **ret_dip)
 {
 	ASSERT(node_name != NULL);
@@ -3652,7 +3652,7 @@ init_spec_child(dev_info_t *pdip, struct hwc_spec *specp, uint_t flags)
 		return;
 	}
 
-	dip = i_ddi_alloc_node(pdip, node_name, (dnode_t)DEVI_PSEUDO_NODEID,
+	dip = i_ddi_alloc_node(pdip, node_name, (pnode_t)DEVI_PSEUDO_NODEID,
 	    -1, specp->hwc_devi_sys_prop_ptr, KM_SLEEP);
 
 	if (dip == NULL)
@@ -5367,7 +5367,7 @@ path_to_major(char *path)
 {
 	dev_info_t *dip;
 	char *p, *q;
-	dnode_t nodeid;
+	pnode_t nodeid;
 	major_t maj;
 
 	/*
@@ -6390,7 +6390,7 @@ mt_config_driver(struct mt_config_handle *hdl)
  * NOTE: This function will return NULL for .conf nodeids.
  */
 dev_info_t *
-e_ddi_nodeid_to_dip(dnode_t nodeid)
+e_ddi_nodeid_to_dip(pnode_t nodeid)
 {
 	dev_info_t		*dip = NULL;
 	struct devi_nodeid	*prev, *elem;
diff --git a/usr/src/uts/common/os/kmem.c b/usr/src/uts/common/os/kmem.c
index 7ae9d5f16b..be7ba8b0cc 100644
--- a/usr/src/uts/common/os/kmem.c
+++ b/usr/src/uts/common/os/kmem.c
@@ -1682,6 +1682,21 @@ kmem_cache_magazine_enable(kmem_cache_t *cp)
 }
 
 /*
+ * Reap (almost) everything right now.  See kmem_cache_magazine_purge()
+ * for explanation of the back-to-back kmem_depot_ws_update() calls.
+ */
+void
+kmem_cache_reap_now(kmem_cache_t *cp)
+{
+	kmem_depot_ws_update(cp);
+	kmem_depot_ws_update(cp);
+
+	(void) taskq_dispatch(kmem_taskq,
+	    (task_func_t *)kmem_depot_ws_reap, cp, TQ_SLEEP);
+	taskq_wait(kmem_taskq);
+}
+
+/*
  * Recompute a cache's magazine size.  The trade-off is that larger magazines
  * provide a higher transfer rate with the depot, while smaller magazines
  * reduce memory consumption.  Magazine resizing is an expensive operation;
@@ -1978,6 +1993,15 @@ kmem_maxavail(void)
 	return ((size_t)ptob(MAX(MIN(pmem, vmem), 0)));
 }
 
+/*
+ * Indicate whether memory-intensive kmem debugging is enabled.
+ */
+int
+kmem_debugging(void)
+{
+	return (kmem_flags & (KMF_AUDIT | KMF_REDZONE));
+}
+
 kmem_cache_t *
 kmem_cache_create(
 	char *name,		/* descriptive name for this cache */
diff --git a/usr/src/uts/common/os/list.c b/usr/src/uts/common/os/list.c
index b511d60bdc..8194b79147 100644
--- a/usr/src/uts/common/os/list.c
+++ b/usr/src/uts/common/os/list.c
@@ -20,7 +20,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -180,3 +180,16 @@ list_move_tail(list_t *dst, list_t *src)
 	/* empty src list */
 	srcnode->list_next = srcnode->list_prev = srcnode;
 }
+
+int
+list_link_active(list_node_t *link)
+{
+	ASSERT((link->list_next == NULL) == (link->list_prev == NULL));
+	return (link->list_next != NULL);
+}
+
+int
+list_is_empty(list_t *list)
+{
+	return (list_empty(list));
+}
diff --git a/usr/src/uts/common/os/policy.c b/usr/src/uts/common/os/policy.c
index 8c0b61a6c5..a5b011a3f8 100644
--- a/usr/src/uts/common/os/policy.c
+++ b/usr/src/uts/common/os/policy.c
@@ -1761,3 +1761,15 @@ secpolicy_gart_map(const cred_t *cr)
 	}
 	return (0);
 }
+
+/*
+ * secpolicy_zfs
+ *
+ * Determine if the user has permission to manipulate ZFS datasets (not pools).
+ * Equivalent to the SYS_MOUNT privilege.
+ */
+int
+secpolicy_zfs(const cred_t *cr)
+{
+	return (PRIV_POLICY(cr, PRIV_SYS_MOUNT, B_FALSE, EPERM, NULL));
+}
diff --git a/usr/src/uts/common/os/printf.c b/usr/src/uts/common/os/printf.c
index d21e8cb029..603da31b62 100644
--- a/usr/src/uts/common/os/printf.c
+++ b/usr/src/uts/common/os/printf.c
@@ -20,7 +20,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2003 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -298,6 +298,23 @@ assfail(const char *a, const char *f, int l)
 	return (0);
 }
 
+void
+assfail3(const char *a, uintmax_t lv, const char *op, uintmax_t rv,
+    const char *f, int l)
+{
+	if (aask)  {
+		printf("ASSERTION CAUGHT: %s (0x%llx %s 0x%llx), file: %s, "
+		    "line: %d", a, (u_longlong_t)lv, op, (u_longlong_t)rv,
+		    f, l);
+		debug_enter(NULL);
+	}
+
+	if (!aok && !panicstr)
+		panic("assertion failed: %s (0x%llx %s 0x%llx), file: %s, "
+		    "line: %d", a, (u_longlong_t)lv, op, (u_longlong_t)rv,
+		    f, l);
+}
+
 int
 strlog(short mid, short sid, char level, ushort_t sl, char *fmt, ...)
 {
diff --git a/usr/src/uts/common/os/sunddi.c b/usr/src/uts/common/os/sunddi.c
index 92f331c157..081e9d9005 100644
--- a/usr/src/uts/common/os/sunddi.c
+++ b/usr/src/uts/common/os/sunddi.c
@@ -4500,7 +4500,7 @@ impl_ddi_bus_prop_op(dev_t dev, dev_info_t *dip, dev_info_t *ch_dip,
 	if (((dev == DDI_DEV_T_NONE) || (dev == DDI_DEV_T_ANY)) &&
 	    ndi_dev_is_prom_node(ch_dip) &&
 	    ((mod_flags & DDI_PROP_NOTPROM) == 0)) {
-		len = prom_getproplen((dnode_t)DEVI(ch_dip)->devi_nodeid, name);
+		len = prom_getproplen((pnode_t)DEVI(ch_dip)->devi_nodeid, name);
 		if (len == -1) {
 			return (DDI_PROP_NOT_FOUND);
 		}
@@ -4555,7 +4555,7 @@ impl_ddi_bus_prop_op(dev_t dev, dev_info_t *dip, dev_info_t *ch_dip,
 		/*
 		 * Call the PROM function to do the copy.
 		 */
-		(void) prom_getprop((dnode_t)DEVI(ch_dip)->devi_nodeid,
+		(void) prom_getprop((pnode_t)DEVI(ch_dip)->devi_nodeid,
 			name, buffer);
 
 		*lengthp = len; /* return the actual length to the caller */
diff --git a/usr/src/uts/common/os/sunndi.c b/usr/src/uts/common/os/sunndi.c
index 4758668769..4def654094 100644
--- a/usr/src/uts/common/os/sunndi.c
+++ b/usr/src/uts/common/os/sunndi.c
@@ -957,7 +957,7 @@ i_dc_devi_create(struct devctl_iocdata *dcp, dev_info_t *pdip,
 	/*
 	 * construct a new dev_info node with a user-provided nodename
 	 */
-	ndi_devi_alloc_sleep(pdip, cname, (dnode_t)DEVI_SID_NODEID, &cdip);
+	ndi_devi_alloc_sleep(pdip, cname, (pnode_t)DEVI_SID_NODEID, &cdip);
 
 	/*
 	 * create hardware properties for each member in the property
diff --git a/usr/src/uts/common/os/swapgeneric.c b/usr/src/uts/common/os/swapgeneric.c
index a180ec293c..9da38ab18b 100644
--- a/usr/src/uts/common/os/swapgeneric.c
+++ b/usr/src/uts/common/os/swapgeneric.c
@@ -877,7 +877,7 @@ load_boot_platform_modules(char *drv)
  * to locate a given nodeid in the device tree.
  */
 struct i_path_findnode {
-	dnode_t nodeid;
+	pnode_t nodeid;
 	dev_info_t *dip;
 };
 
@@ -952,7 +952,7 @@ netboot_over_ib(char *bootpath)
 
 	char		*temp;
 	boolean_t	ret = B_FALSE;
-	dnode_t		node = prom_finddevice(bootpath);
+	pnode_t		node = prom_finddevice(bootpath);
 	int		len;
 	char		devicetype[OBP_MAXDRVNAME];
 
diff --git a/usr/src/uts/common/os/zone.c b/usr/src/uts/common/os/zone.c
index b4abecaadf..71fe7305a2 100644
--- a/usr/src/uts/common/os/zone.c
+++ b/usr/src/uts/common/os/zone.c
@@ -173,7 +173,7 @@
  *   the following system calls (all subcodes of the primary "zone"
  *   system call):
  *   - zone_create: creates a zone with selected attributes (name,
- *     root path, privileges, resource controls)
+ *     root path, privileges, resource controls, ZFS datasets)
  *   - zone_enter: allows the current process to enter a zone
  *   - zone_getattr: reports attributes of a zone
  *   - zone_list: lists all zones active in the system
@@ -770,6 +770,23 @@ zone_free_zsd(zone_t *zone)
 }
 
 /*
+ * Frees memory associated with the zone dataset list.
+ */
+static void
+zone_free_datasets(zone_t *zone)
+{
+	zone_dataset_t *t, *next;
+
+	for (t = list_head(&zone->zone_datasets); t != NULL; t = next) {
+		next = list_next(&zone->zone_datasets, t);
+		list_remove(&zone->zone_datasets, t);
+		kmem_free(t->zd_dataset, strlen(t->zd_dataset) + 1);
+		kmem_free(t, sizeof (*t));
+	}
+	list_destroy(&zone->zone_datasets);
+}
+
+/*
  * zone.cpu-shares resource control support.
  */
 /*ARGSUSED*/
@@ -1055,6 +1072,7 @@ zone_free(zone_t *zone)
 	}
 
 	zone_free_zsd(zone);
+	zone_free_datasets(zone);
 
 	if (zone->zone_rootvp != NULL)
 		VN_RELE(zone->zone_rootvp);
@@ -2500,6 +2518,55 @@ zone_create_error(int er_error, int er_ext, int *er_out) {
 }
 
 /*
+ * Parses a comma-separated list of ZFS datasets into a per-zone dictionary.
+ */
+static int
+parse_zfs(zone_t *zone, caddr_t ubuf, size_t buflen)
+{
+	char *kbuf;
+	char *dataset, *next;
+	zone_dataset_t *zd;
+	size_t len;
+
+	if (ubuf == NULL || buflen == 0)
+		return (0);
+
+	if ((kbuf = kmem_alloc(buflen, KM_NOSLEEP)) == NULL)
+		return (ENOMEM);
+
+	if (copyin(ubuf, kbuf, buflen) != 0) {
+		kmem_free(kbuf, buflen);
+		return (EFAULT);
+	}
+
+	dataset = next = kbuf;
+	for (;;) {
+		zd = kmem_alloc(sizeof (zone_dataset_t), KM_SLEEP);
+
+		next = strchr(dataset, ',');
+
+		if (next == NULL)
+			len = strlen(dataset);
+		else
+			len = next - dataset;
+
+		zd->zd_dataset = kmem_alloc(len + 1, KM_SLEEP);
+		bcopy(dataset, zd->zd_dataset, len);
+		zd->zd_dataset[len] = '\0';
+
+		list_insert_head(&zone->zone_datasets, zd);
+
+		if (next == NULL)
+			break;
+
+		dataset = next + 1;
+	}
+
+	kmem_free(kbuf, buflen);
+	return (0);
+}
+
+/*
  * System call to create/initialize a new zone named 'zone_name', rooted
  * at 'zone_root', with a zone-wide privilege limit set of 'zone_privs',
  * and initialized with the zone-wide rctls described in 'rctlbuf'.
@@ -2510,7 +2577,7 @@ zone_create_error(int er_error, int er_ext, int *er_out) {
 static zoneid_t
 zone_create(const char *zone_name, const char *zone_root,
     const priv_set_t *zone_privs, caddr_t rctlbuf, size_t rctlbufsz,
-    int *extended_error)
+    caddr_t zfsbuf, size_t zfsbufsz, int *extended_error)
 {
 	struct zsched_arg zarg;
 	nvlist_t *rctls = NULL;
@@ -2543,6 +2610,8 @@ zone_create(const char *zone_name, const char *zone_root,
 	cv_init(&zone->zone_cv, NULL, CV_DEFAULT, NULL);
 	list_create(&zone->zone_zsd, sizeof (struct zsd_entry),
 	    offsetof(struct zsd_entry, zsd_linkage));
+	list_create(&zone->zone_datasets, sizeof (zone_dataset_t),
+	    offsetof(zone_dataset_t, zd_linkage));
 
 	if ((error = zone_set_name(zone, zone_name)) != 0) {
 		zone_free(zone);
@@ -2578,6 +2647,11 @@ zone_create(const char *zone_name, const char *zone_root,
 		return (zone_create_error(error, 0, extended_error));
 	}
 
+	if ((error = parse_zfs(zone, zfsbuf, zfsbufsz)) != 0) {
+		zone_free(zone);
+		return (set_errno(error));
+	}
+
 	/*
 	 * Stop all lwps since that's what normally happens as part of fork().
 	 * This needs to happen before we grab any locks to avoid deadlock
@@ -3722,7 +3796,7 @@ zone_lookup(const char *zone_name)
 
 /* ARGSUSED */
 long
-zone(int cmd, void *arg1, void *arg2, void *arg3, void *arg4, void *arg5)
+zone(int cmd, void *arg1, void *arg2, void *arg3, void *arg4)
 {
 	zone_def zs;
 
@@ -3748,6 +3822,8 @@ zone(int cmd, void *arg1, void *arg2, void *arg3, void *arg4, void *arg5)
 			    (unsigned long)zs32.zone_privs;
 			zs.rctlbuf = (caddr_t)(unsigned long)zs32.rctlbuf;
 			zs.rctlbufsz = zs32.rctlbufsz;
+			zs.zfsbuf = (caddr_t)(unsigned long)zs32.zfsbuf;
+			zs.zfsbufsz = zs32.zfsbufsz;
 			zs.extended_error =
 			    (int *)(unsigned long)zs32.extended_error;
 #else
@@ -3757,6 +3833,7 @@ zone(int cmd, void *arg1, void *arg2, void *arg3, void *arg4, void *arg5)
 
 		return (zone_create(zs.zone_name, zs.zone_root,
 			zs.zone_privs, (caddr_t)zs.rctlbuf, zs.rctlbufsz,
+			(caddr_t)zs.zfsbuf, zs.zfsbufsz,
 			zs.extended_error));
 	case ZONE_BOOT:
 		return (zone_boot((zoneid_t)(uintptr_t)arg1,
@@ -4037,3 +4114,61 @@ zone_shutdown_global(void)
 	zone_status_set(global_zone, ZONE_IS_SHUTTING_DOWN);
 	mutex_exit(&zone_status_lock);
 }
+
+/*
+ * Returns true if the named dataset is visible in the current zone.
+ * The 'write' parameter is set to 1 if the dataset is also writable.
+ */
+int
+zone_dataset_visible(const char *dataset, int *write)
+{
+	zone_dataset_t *zd;
+	size_t len;
+	zone_t *zone = curproc->p_zone;
+
+	if (dataset[0] == '\0')
+		return (0);
+
+	/*
+	 * Walk the list once, looking for datasets which match exactly, or
+	 * specify a dataset underneath an exported dataset.  If found, return
+	 * true and note that it is writable.
+	 */
+	for (zd = list_head(&zone->zone_datasets); zd != NULL;
+	    zd = list_next(&zone->zone_datasets, zd)) {
+
+		len = strlen(zd->zd_dataset);
+		if (strlen(dataset) >= len &&
+		    bcmp(dataset, zd->zd_dataset, len) == 0 &&
+		    (zd->zd_dataset[len-1] == '/' ||
+		    dataset[len] == '\0' || dataset[len] == '/')) {
+			if (write)
+				*write = 1;
+			return (1);
+		}
+	}
+
+	/*
+	 * Walk the list a second time, searching for datasets which are parents
+	 * of exported datasets.  These should be visible, but read-only.
+	 *
+	 * Note that we also have to support forms such as 'pool/dataset/', with
+	 * a trailing slash.
+	 */
+	for (zd = list_head(&zone->zone_datasets); zd != NULL;
+	    zd = list_next(&zone->zone_datasets, zd)) {
+
+		len = strlen(dataset);
+		if (dataset[len - 1] == '/')
+			len--;	/* Ignore trailing slash */
+		if (len < strlen(zd->zd_dataset) &&
+		    bcmp(dataset, zd->zd_dataset, len) == 0 &&
+		    zd->zd_dataset[len] == '/') {
+			if (write)
+				*write = 0;
+			return (1);
+		}
+	}
+
+	return (0);
+}