14978 ZFS autoexpand property should work for root pools

Reviewed by: Andy Fiddaman <illumos@fiddaman.net>
Reviewed by: Toomas Soome <tsoome@me.com>
Approved by: Dan McDonald <danmcd@mnx.io>

9 files changed, 132 insertions, 65 deletions

diff --git a/usr/src/cmd/syseventd/modules/zfs_mod/zfs_mod.c b/usr/src/cmd/syseventd/modules/zfs_mod/zfs_mod.c
index 82c296a669..6a25af90f9 100644
--- a/usr/src/cmd/syseventd/modules/zfs_mod/zfs_mod.c
+++ b/usr/src/cmd/syseventd/modules/zfs_mod/zfs_mod.c
@@ -23,6 +23,7 @@
  * Copyright (c) 2012 by Delphix. All rights reserved.
  * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
  * Copyright 2022 OmniOS Community Edition (OmniOSce) Association.
+ * Copyright 2022 Oxide Computer Company
  */
 
 /*
@@ -564,6 +565,13 @@ zfsdle_vdev_online(zpool_handle_t *zhp, void *data)
 		char *path, fullpath[MAXPATHLEN];
 		uint64_t wholedisk = 0ULL;
 
+		/*
+		 * If the /dev path of the device is invalid because the disk
+		 * has been moved to a new location, we need to try to refresh
+		 * that path before onlining the device.
+		 */
+		zpool_vdev_refresh_path(g_zfshdl, zhp, tgt);
+
 		verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
 		    &path) == 0);
 		verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
diff --git a/usr/src/lib/libzfs/common/libzfs.h b/usr/src/lib/libzfs/common/libzfs.h
index 0516845d24..df89daf19a 100644
--- a/usr/src/lib/libzfs/common/libzfs.h
+++ b/usr/src/lib/libzfs/common/libzfs.h
@@ -316,6 +316,8 @@ extern nvlist_t *zpool_find_vdev_by_physpath(zpool_handle_t *, const char *,
     boolean_t *, boolean_t *, boolean_t *);
 extern int zpool_label_disk(libzfs_handle_t *, zpool_handle_t *, const char *,
     zpool_boot_label_t, uint64_t, int *);
+extern void zpool_vdev_refresh_path(libzfs_handle_t *, zpool_handle_t *,
+    nvlist_t *);
 
 /*
  * Functions to manage pool properties
diff --git a/usr/src/lib/libzfs/common/libzfs_pool.c b/usr/src/lib/libzfs/common/libzfs_pool.c
index 6e634a81cb..6f072e4935 100644
--- a/usr/src/lib/libzfs/common/libzfs_pool.c
+++ b/usr/src/lib/libzfs/common/libzfs_pool.c
@@ -28,6 +28,7 @@
  * Copyright (c) 2017 Datto Inc.
  * Copyright (c) 2017, Intel Corporation.
  * Copyright 2022 OmniOS Community Edition (OmniOSce) Association.
+ * Copyright 2022 Oxide Computer Company
  */
 
 #include <ctype.h>
@@ -3960,17 +3961,70 @@ set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
 }
 
 /*
+ * This routine is responsible for identifying when disks have been
+ * reconfigured in a new location.  The kernel will have opened the device by
+ * devid, but the path will still refer to the old location.  To catch this, we
+ * first do a path -> devid translation (which is fast for the common case).
+ * If the devid matches, we're done.  If not, we do a reverse devid -> path
+ * translation and issue the appropriate ioctl() to update the path of the
+ * vdev.
+ */
+void
+zpool_vdev_refresh_path(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv)
+{
+	char *path = NULL;
+	char *newpath = NULL;
+	char *physpath = NULL;
+	char *devid = NULL;
+
+	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) {
+		return;
+	}
+
+	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
+		/*
+		 * This vdev has a devid.  We can use it to check the current
+		 * path.
+		 */
+		char *newdevid = path_to_devid(path);
+
+		if (newdevid == NULL || strcmp(devid, newdevid) != 0) {
+			newpath = devid_to_path(devid);
+		}
+
+		if (newdevid != NULL) {
+			devid_str_free(newdevid);
+		}
+
+	} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
+	    &physpath) == 0) {
+		/*
+		 * This vdev does not have a devid, but it does have a physical
+		 * path.  Attempt to translate this to a /dev path.
+		 */
+		newpath = path_from_physpath(hdl, path, physpath);
+	}
+
+	if (newpath == NULL) {
+		/*
+		 * No path update is required.
+		 */
+		return;
+	}
+
+	set_path(zhp, nv, newpath);
+	fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, newpath);
+
+	free(newpath);
+}
+
+/*
  * Given a vdev, return the name to display in iostat.  If the vdev has a path,
  * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
- * We also check if this is a whole disk, in which case we strip off the
- * trailing 's0' slice name.
+ * We will confirm that the path and name of the vdev are current, and update
+ * them if not.  We also check if this is a whole disk, in which case we strip
+ * off the trailing 's0' slice name.
  *
- * This routine is also responsible for identifying when disks have been
- * reconfigured in a new location.  The kernel will have opened the device by
- * devid, but the path will still refer to the old location.  To catch this, we
- * first do a path -> devid translation (which is fast for the common case).  If
- * the devid matches, we're done.  If not, we do a reverse devid -> path
- * translation and issue the appropriate ioctl() to update the path of the vdev.
  * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
  * of these checks.
  */
@@ -4013,9 +4067,6 @@ zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
 	} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
 		vdev_stat_t *vs;
 		uint_t vsc;
-		char *newpath = NULL;
-		char *physpath = NULL;
-		char *devid = NULL;
 
 		/*
 		 * If the device is dead (faulted, offline, etc) then don't
@@ -4030,42 +4081,12 @@ zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
 			goto after_open;
 		}
 
-		if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
-			/*
-			 * This vdev has a devid.  We can use it to check the
-			 * current path.
-			 */
-			char *newdevid = path_to_devid(path);
-
-			if (newdevid == NULL || strcmp(devid, newdevid) != 0) {
-				newpath = devid_to_path(devid);
-			}
-
-			if (newdevid != NULL)
-				devid_str_free(newdevid);
-
-		} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
-		    &physpath) == 0) {
-			/*
-			 * This vdev does not have a devid, but it does have a
-			 * physical path.  Attempt to translate this to a /dev
-			 * path.
-			 */
-			newpath = path_from_physpath(hdl, path, physpath);
-		}
-
-		if (newpath != NULL) {
-			/*
-			 * Update the path appropriately.
-			 */
-			set_path(zhp, nv, newpath);
-			if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH,
-			    newpath) == 0) {
-				verify(nvlist_lookup_string(nv,
-				    ZPOOL_CONFIG_PATH, &path) == 0);
-			}
-			free(newpath);
-		}
+		/*
+		 * Refresh the /dev path for this vdev if required, then ensure
+		 * we're using the latest path value:
+		 */
+		zpool_vdev_refresh_path(hdl, zhp, nv);
+		path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH);
 
 		if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
 			char *rp = realpath(path, NULL);
diff --git a/usr/src/lib/libzfs/common/mapfile-vers b/usr/src/lib/libzfs/common/mapfile-vers
index 196751b44c..4c2c6e67fe 100644
--- a/usr/src/lib/libzfs/common/mapfile-vers
+++ b/usr/src/lib/libzfs/common/mapfile-vers
@@ -279,6 +279,7 @@ SYMBOL_VERSION SUNWprivate_1.1 {
 	zpool_vdev_name;
 	zpool_vdev_offline;
 	zpool_vdev_online;
+	zpool_vdev_refresh_path;
 	zpool_vdev_remove;
 	zpool_vdev_remove_cancel;
 	zpool_vdev_split;
diff --git a/usr/src/lib/libzpool/Makefile.com b/usr/src/lib/libzpool/Makefile.com
index d52d56b1ff..47ff1d0b76 100644
--- a/usr/src/lib/libzpool/Makefile.com
+++ b/usr/src/lib/libzpool/Makefile.com
@@ -68,7 +68,7 @@ CSTD=	$(CSTD_GNU99)
 
 CFLAGS +=	$(CCGDEBUG) $(CCVERBOSE) $(CNOGLOBAL)
 CFLAGS64 +=	$(CCGDEBUG) $(CCVERBOSE) $(CNOGLOBAL)
-LDLIBS +=	-lcmdutils -lumem -lavl -lnvpair -lz -lc -lsysevent -lmd \
+LDLIBS +=	-lcmdutils -lumem -lavl -lnvpair -lz -lc -lmd \
 		-lfakekernel -lzutil
 NATIVE_LIBS +=	libz.so
 CPPFLAGS.first =	-I$(SRC)/lib/libfakekernel/common
diff --git a/usr/src/uts/common/fs/zfs/spa.c b/usr/src/uts/common/fs/zfs/spa.c
index 70d690d587..d390e39195 100644
--- a/usr/src/uts/common/fs/zfs/spa.c
+++ b/usr/src/uts/common/fs/zfs/spa.c
@@ -32,6 +32,7 @@
  * Copyright (c) 2017, Intel Corporation.
  * Copyright 2020 Joshua M. Clulow <josh@sysmgr.org>
  * Copyright 2021 OmniOS Community Edition (OmniOSce) Association.
+ * Copyright 2022 Oxide Computer Company
  */
 
 /*
@@ -5593,6 +5594,12 @@ spa_import_rootpool(char *devpath, char *devid, uint64_t pool_guid,
 		goto out;
 	}
 
+	/*
+	 * The root disk may have been expanded while the system was offline.
+	 * Kick off an async task to check for and handle expansion.
+	 */
+	spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);
+
 	error = 0;
 out:
 	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
@@ -7517,8 +7524,6 @@ spa_async_probe(spa_t *spa, vdev_t *vd)
 static void
 spa_async_autoexpand(spa_t *spa, vdev_t *vd)
 {
-	sysevent_id_t eid;
-	nvlist_t *attr;
 	char *physpath;
 
 	if (!spa->spa_autoexpand)
@@ -7535,13 +7540,8 @@ spa_async_autoexpand(spa_t *spa, vdev_t *vd)
 	physpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
 	(void) snprintf(physpath, MAXPATHLEN, "/devices%s", vd->vdev_physpath);
 
-	VERIFY(nvlist_alloc(&attr, NV_UNIQUE_NAME, KM_SLEEP) == 0);
-	VERIFY(nvlist_add_string(attr, DEV_PHYS_PATH, physpath) == 0);
-
-	(void) ddi_log_sysevent(zfs_dip, SUNW_VENDOR, EC_DEV_STATUS,
-	    ESC_DEV_DLE, attr, &eid, DDI_SLEEP);
+	zfs_post_dle_sysevent(physpath);
 
-	nvlist_free(attr);
 	kmem_free(physpath, MAXPATHLEN);
 }
 
diff --git a/usr/src/uts/common/fs/zfs/spa_misc.c b/usr/src/uts/common/fs/zfs/spa_misc.c
index cb59eef824..6e834005a1 100644
--- a/usr/src/uts/common/fs/zfs/spa_misc.c
+++ b/usr/src/uts/common/fs/zfs/spa_misc.c
@@ -29,6 +29,7 @@
  * Copyright 2019 Joyent, Inc.
  * Copyright (c) 2017, Intel Corporation.
  * Copyright 2020 Joyent, Inc.
+ * Copyright 2022 Oxide Computer Company
  */
 
 #include <sys/zfs_context.h>
@@ -2598,3 +2599,44 @@ spa_suspend_async_destroy(spa_t *spa)
 
 	return (B_FALSE);
 }
+
+/*
+ * Generate a LUN expansion event.  This routine does not use
+ * ddi_log_sysevent() because that would require a dev_info_t, and we may not
+ * have one available.
+ */
+void
+zfs_post_dle_sysevent(const char *physpath)
+{
+#ifdef _KERNEL
+	sysevent_t *ev = sysevent_alloc(EC_DEV_STATUS, ESC_DEV_DLE,
+	    SUNW_KERN_PUB "zfs", SE_SLEEP);
+	sysevent_attr_list_t *attr = NULL;
+	sysevent_id_t eid;
+
+	VERIFY(ev != NULL);
+
+	/*
+	 * The only attribute is the /devices path of the expanding device:
+	 */
+	sysevent_value_t value = {
+		.value_type = SE_DATA_TYPE_STRING,
+		.value = {
+			.sv_string = (char *)physpath,
+		},
+	};
+	if (sysevent_add_attr(&attr, DEV_PHYS_PATH, &value, SE_SLEEP) != 0) {
+		sysevent_free(ev);
+		return;
+	}
+
+	if (sysevent_attach_attributes(ev, attr) != 0) {
+		sysevent_free_attr(attr);
+		sysevent_free(ev);
+		return;
+	}
+
+	(void) log_sysevent(ev, SE_SLEEP, &eid);
+	sysevent_free(ev);
+#endif
+}
diff --git a/usr/src/uts/common/fs/zfs/sys/spa.h b/usr/src/uts/common/fs/zfs/sys/spa.h
index af8057be8f..b896d51577 100644
--- a/usr/src/uts/common/fs/zfs/sys/spa.h
+++ b/usr/src/uts/common/fs/zfs/sys/spa.h
@@ -1104,6 +1104,7 @@ extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl,
     const char *name);
 extern void spa_event_post(sysevent_t *ev);
 extern void spa_event_discard(sysevent_t *ev);
+extern void zfs_post_dle_sysevent(const char *);
 
 #ifdef ZFS_DEBUG
 #define	dprintf_bp(bp, fmt, ...) do {				\
diff --git a/usr/src/uts/common/fs/zfs/zvol.c b/usr/src/uts/common/fs/zfs/zvol.c
index 1de192f87f..2bb311d28d 100644
--- a/usr/src/uts/common/fs/zfs/zvol.c
+++ b/usr/src/uts/common/fs/zfs/zvol.c
@@ -813,20 +813,12 @@ zvol_update_live_volsize(zvol_state_t *zv, uint64_t volsize)
 	 * Generate a LUN expansion event.
 	 */
 	if (error == 0) {
-		sysevent_id_t eid;
-		nvlist_t *attr;
 		char *physpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
 
 		(void) snprintf(physpath, MAXPATHLEN, "%s%u", ZVOL_PSEUDO_DEV,
 		    zv->zv_minor);
 
-		VERIFY(nvlist_alloc(&attr, NV_UNIQUE_NAME, KM_SLEEP) == 0);
-		VERIFY(nvlist_add_string(attr, DEV_PHYS_PATH, physpath) == 0);
-
-		(void) ddi_log_sysevent(zfs_dip, SUNW_VENDOR, EC_DEV_STATUS,
-		    ESC_DEV_DLE, attr, &eid, DDI_SLEEP);
-
-		nvlist_free(attr);
+		zfs_post_dle_sysevent(physpath);
 		kmem_free(physpath, MAXPATHLEN);
 	}
 	return (error);