zed: mark disks as REMOVED when they are removed

ZED does not take any action for disk removal events if there is no
spare VDEV available. Added zpool_vdev_remove_wanted() in libzfs
and vdev_remove_wanted() in vdev.c to remove the VDEV through ZED
on removal event.  This means that if you are running zed and
remove a disk, it will be properly marked as REMOVED.

Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Ameer Hamza <ahamza@ixsystems.com>
Closes openzfs#13797

cmd/zed/agents/zfs_agents.c

@@ -80,6 +80,7 @@ zfs_agent_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *arg)
 	char *path = NULL;
 	uint_t c, children;
 	nvlist_t **child;
+	uint64_t vdev_guid;
 
 	/*
 	 * First iterate over any children.
@@ -100,7 +101,7 @@ zfs_agent_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *arg)
 	    &child, &children) == 0) {
 		for (c = 0; c < children; c++) {
 			if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
-				gsp->gs_vdev_type = DEVICE_TYPE_L2ARC;
+				gsp->gs_vdev_type = DEVICE_TYPE_SPARE;
 				return (B_TRUE);
 			}
 		}
@@ -109,7 +110,7 @@ zfs_agent_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *arg)
 	    &child, &children) == 0) {
 		for (c = 0; c < children; c++) {
 			if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
-				gsp->gs_vdev_type = DEVICE_TYPE_SPARE;
+				gsp->gs_vdev_type = DEVICE_TYPE_L2ARC;
 				return (B_TRUE);
 			}
 		}
@@ -126,6 +127,21 @@ zfs_agent_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *arg)
 		    &gsp->gs_vdev_expandtime);
 		return (B_TRUE);
 	}
+	/*
+	 * Otherwise, on a vdev guid match, grab the devid and expansion
+	 * time. The devid might be missing on removal since its not part
+	 * of blkid cache and L2ARC VDEV does not contain pool guid in its
+	 * blkid, so this is a special case for L2ARC VDEV.
+	 */
+	else if (gsp->gs_vdev_guid != 0 && gsp->gs_devid == NULL &&
+	    nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID, &vdev_guid) == 0 &&
+	    gsp->gs_vdev_guid == vdev_guid) {
+		(void) nvlist_lookup_string(nvl, ZPOOL_CONFIG_DEVID,
+		    &gsp->gs_devid);
+		(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_EXPANSION_TIME,
+		    &gsp->gs_vdev_expandtime);
+		return (B_TRUE);
+	}
 
 	return (B_FALSE);
 }
@@ -148,7 +164,7 @@ zfs_agent_iter_pool(zpool_handle_t *zhp, void *arg)
 	/*
 	 * if a match was found then grab the pool guid
 	 */
-	if (gsp->gs_vdev_guid) {
+	if (gsp->gs_vdev_guid && gsp->gs_devid) {
 		(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
 		    &gsp->gs_pool_guid);
 	}
@@ -195,11 +211,13 @@ zfs_agent_post_event(const char *class, const char *subclass, nvlist_t *nvl)
 		uint64_t pool_guid = 0, vdev_guid = 0;
 		guid_search_t search = { 0 };
 		device_type_t devtype = DEVICE_TYPE_PRIMARY;
+		char *devid = NULL;
 
 		class = "resource.fs.zfs.removed";
 		subclass = "";
 
 		(void) nvlist_add_string(payload, FM_CLASS, class);
+		(void) nvlist_lookup_string(nvl, DEV_IDENTIFIER, &devid);
 		(void) nvlist_lookup_uint64(nvl, ZFS_EV_POOL_GUID, &pool_guid);
 		(void) nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &vdev_guid);
 
@@ -209,20 +227,24 @@ zfs_agent_post_event(const char *class, const char *subclass, nvlist_t *nvl)
 		(void) nvlist_add_int64_array(payload, FM_EREPORT_TIME, tod, 2);
 
 		/*
+		 * If devid is missing but vdev_guid is available, find devid
+		 * and pool_guid from vdev_guid.
 		 * For multipath, spare and l2arc devices ZFS_EV_VDEV_GUID or
 		 * ZFS_EV_POOL_GUID may be missing so find them.
 		 */
-		if (pool_guid == 0 || vdev_guid == 0) {
-			if ((nvlist_lookup_string(nvl, DEV_IDENTIFIER,
-			    &search.gs_devid) == 0) &&
-			    (zpool_iter(g_zfs_hdl, zfs_agent_iter_pool, &search)
-			    == 1)) {
-				if (pool_guid == 0)
-					pool_guid = search.gs_pool_guid;
-				if (vdev_guid == 0)
-					vdev_guid = search.gs_vdev_guid;
-				devtype = search.gs_vdev_type;
-			}
+		if (devid == NULL || pool_guid == 0 || vdev_guid == 0) {
+			if (devid == NULL)
+				search.gs_vdev_guid = vdev_guid;
+			else
+				search.gs_devid = devid;
+			zpool_iter(g_zfs_hdl, zfs_agent_iter_pool, &search);
+			if (devid == NULL)
+				devid = search.gs_devid;
+			if (pool_guid == 0)
+				pool_guid = search.gs_pool_guid;
+			if (vdev_guid == 0)
+				vdev_guid = search.gs_vdev_guid;
+			devtype = search.gs_vdev_type;
 		}
 
 		/*
@@ -235,7 +257,9 @@ zfs_agent_post_event(const char *class, const char *subclass, nvlist_t *nvl)
 		    search.gs_vdev_expandtime + 10 > tv.tv_sec) {
 			zed_log_msg(LOG_INFO, "agent post event: ignoring '%s' "
 			    "for recently expanded device '%s'", EC_DEV_REMOVE,
-			    search.gs_devid);
+			    devid);
+			fnvlist_free(payload);
+			free(event);
 			goto out;
 		}
 

cmd/zed/agents/zfs_retire.c

@@ -323,6 +323,9 @@ zfs_retire_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
 	boolean_t is_disk;
 	vdev_aux_t aux;
 	uint64_t state = 0;
+	int l2arc;
+	vdev_stat_t *vs;
+	unsigned int c;
 
 	fmd_hdl_debug(hdl, "zfs_retire_recv: '%s'", class);
 
@@ -352,13 +355,32 @@ zfs_retire_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
 
 		devname = zpool_vdev_name(NULL, zhp, vdev, B_FALSE);
 
-		/* Can't replace l2arc with a spare: offline the device */
-		if (nvlist_lookup_string(nvl, FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
-		    &devtype) == 0 && strcmp(devtype, VDEV_TYPE_L2CACHE) == 0) {
-			fmd_hdl_debug(hdl, "zpool_vdev_offline '%s'", devname);
-			zpool_vdev_offline(zhp, devname, B_TRUE);
-		} else if (!fmd_prop_get_int32(hdl, "spare_on_remove") ||
-		    replace_with_spare(hdl, zhp, vdev) == B_FALSE) {
+		nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
+		    (uint64_t **)&vs, &c);
+
+		/*
+		 * If state removed is requested for already removed vdev,
+		 * its a loopback event from spa_async_remove(). Just
+		 * ignore it.
+		 */
+		if (vs->vs_state == VDEV_STATE_REMOVED &&
+		    state == VDEV_STATE_REMOVED)
+			return;
+
+		l2arc = (nvlist_lookup_string(nvl,
+		    FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, &devtype) == 0 &&
+		    strcmp(devtype, VDEV_TYPE_L2CACHE) == 0);
+
+		/* Remove the vdev since device is unplugged */
+		if (l2arc || (strcmp(class, "resource.fs.zfs.removed") == 0)) {
+			int status = zpool_vdev_remove_wanted(zhp, devname);
+			fmd_hdl_debug(hdl, "zpool_vdev_remove_wanted '%s'"
+			    ", ret:%d", devname, status);
+		}
+
+		/* Replace the vdev with a spare if its not a l2arc */
+		if (!l2arc && (!fmd_prop_get_int32(hdl, "spare_on_remove") ||
+		    replace_with_spare(hdl, zhp, vdev) == B_FALSE)) {
 			/* Could not handle with spare */
 			fmd_hdl_debug(hdl, "no spare for '%s'", devname);
 		}

cmd/ztest.c

@@ -1165,7 +1165,7 @@ ztest_kill(ztest_shared_t *zs)
 	 * See comment above spa_write_cachefile().
 	 */
 	mutex_enter(&spa_namespace_lock);
-	spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE);
+	spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE, B_FALSE);
 	mutex_exit(&spa_namespace_lock);
 
 	(void) raise(SIGKILL);

config/kernel-blkdev.m4

@@ -103,6 +103,57 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE], [
 	])
 ])
 
+dnl #
+dnl # bdev_kobj() is introduced from 5.12
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ], [
+	ZFS_LINUX_TEST_SRC([bdev_kobj], [
+		#include <linux/fs.h>
+		#include <linux/blkdev.h>
+		#include <linux/kobject.h>
+	], [
+		struct block_device *bdev = NULL;
+		struct kobject *disk_kobj;
+		disk_kobj = bdev_kobj(bdev);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ], [
+	AC_MSG_CHECKING([whether bdev_kobj() exists])
+	ZFS_LINUX_TEST_RESULT([bdev_kobj], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BDEV_KOBJ, 1,
+		    [bdev_kobj() exists])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # part_to_dev() was removed in 5.12
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV], [
+	ZFS_LINUX_TEST_SRC([part_to_dev], [
+		#include <linux/fs.h>
+		#include <linux/blkdev.h>
+	], [
+		struct hd_struct *p = NULL;
+		struct device *pdev;
+		pdev = part_to_dev(p);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV], [
+	AC_MSG_CHECKING([whether part_to_dev() exists])
+	ZFS_LINUX_TEST_RESULT([part_to_dev], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_PART_TO_DEV, 1,
+		    [part_to_dev() exists])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])
+
 dnl #
 dnl # 5.10 API, check_disk_change() is removed, in favor of
 dnl # bdev_check_media_change(), which doesn't force revalidation
@@ -405,6 +456,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV], [
 	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_WHOLE
 	ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME
 	ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE
+	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ
+	ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
@@ -421,4 +474,6 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
 	ZFS_AC_KERNEL_BLKDEV_BDEVNAME
 	ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS
 	ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE
+	ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ
+	ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV
 ])

include/libzfs.h

@@ -310,6 +310,7 @@ _LIBZFS_H int zpool_vdev_indirect_size(zpool_handle_t *, const char *,
     uint64_t *);
 _LIBZFS_H int zpool_vdev_split(zpool_handle_t *, char *, nvlist_t **,
     nvlist_t *, splitflags_t);
+_LIBZFS_H int zpool_vdev_remove_wanted(zpool_handle_t *, const char *);
 
 _LIBZFS_H int zpool_vdev_fault(zpool_handle_t *, uint64_t, vdev_aux_t);
 _LIBZFS_H int zpool_vdev_degrade(zpool_handle_t *, uint64_t, vdev_aux_t);

include/os/linux/Makefile.am

@@ -71,6 +71,7 @@ kernel_spl_sys_HEADERS = \
 	%D%/spl/sys/kmem_cache.h \
 	%D%/spl/sys/kstat.h \
 	%D%/spl/sys/list.h \
+	%D%/spl/sys/misc.h \
 	%D%/spl/sys/mod_os.h \
 	%D%/spl/sys/mutex.h \
 	%D%/spl/sys/param.h \

include/os/linux/kernel/linux/blkdev_compat.h

@@ -261,6 +261,32 @@ bio_set_bi_error(struct bio *bio, int error)
 #define	BIO_END_IO(bio, error)		bio_endio(bio, error);
 #endif /* HAVE_1ARG_BIO_END_IO_T */
 
+/*
+ * 5.15 MACRO,
+ *   GD_DEAD
+ *
+ * 2.6.36 - 5.14 MACRO,
+ *   GENHD_FL_UP
+ *
+ * Check the disk status and return B_TRUE if alive
+ * otherwise B_FALSE
+ */
+static inline boolean_t
+zfs_check_disk_status(struct block_device *bdev)
+{
+#if defined(GENHD_FL_UP)
+	return (!!(bdev->bd_disk->flags & GENHD_FL_UP));
+#elif defined(GD_DEAD)
+	return (!test_bit(GD_DEAD, &bdev->bd_disk->state));
+#else
+/*
+ * This is encountered if neither GENHD_FL_UP nor GD_DEAD is available in
+ * the kernel - likely due to an MACRO change that needs to be chased down.
+ */
+#error "Unsupported kernel: no usable disk status check"
+#endif
+}
+
 /*
  * 4.1 API,
  * 3.10.0 CentOS 7.x API,

include/os/linux/spl/sys/misc.h

@@ -0,0 +1,29 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or https://opensource.org/licenses/CDDL-1.0.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+#ifndef _OS_LINUX_SPL_MISC_H
+#define	_OS_LINUX_SPL_MISC_H
+
+#include <linux/kobject.h>
+
+extern void spl_signal_kobj_evt(struct block_device *bdev);
+
+#endif

include/sys/spa.h

@@ -838,7 +838,7 @@ extern kmutex_t spa_namespace_lock;
 #define	SPA_CONFIG_UPDATE_POOL	0
 #define	SPA_CONFIG_UPDATE_VDEVS	1
 
-extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
+extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t, boolean_t);
 extern void spa_config_load(void);
 extern nvlist_t *spa_all_configs(uint64_t *);
 extern void spa_config_set(spa_t *spa, nvlist_t *config);

include/sys/vdev.h

@@ -148,6 +148,7 @@ extern int vdev_degrade(spa_t *spa, uint64_t guid, vdev_aux_t aux);
 extern int vdev_online(spa_t *spa, uint64_t guid, uint64_t flags,
     vdev_state_t *);
 extern int vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags);
+extern int vdev_remove_wanted(spa_t *spa, uint64_t guid);
 extern void vdev_clear(spa_t *spa, vdev_t *vd);
 
 extern boolean_t vdev_is_dead(vdev_t *vd);
@@ -190,6 +191,8 @@ typedef enum vdev_config_flag {
 	VDEV_CONFIG_MISSING = 1 << 4
 } vdev_config_flag_t;
 
+extern void vdev_post_kobj_evt(vdev_t *vd);
+extern void vdev_clear_kobj_evt(vdev_t *vd);
 extern void vdev_top_config_generate(spa_t *spa, nvlist_t *config);
 extern nvlist_t *vdev_config_generate(spa_t *spa, vdev_t *vd,
     boolean_t getstats, vdev_config_flag_t flags);

include/sys/vdev_impl.h

@@ -69,6 +69,7 @@ extern uint32_t zfs_vdev_async_write_max_active;
  * Virtual device operations
  */
 typedef int	vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd);
+typedef void	vdev_kobj_post_evt_func_t(vdev_t *vd);
 typedef void	vdev_fini_func_t(vdev_t *vd);
 typedef int	vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
     uint64_t *ashift, uint64_t *pshift);
@@ -123,6 +124,7 @@ typedef const struct vdev_ops {
 	vdev_config_generate_func_t	*vdev_op_config_generate;
 	vdev_nparity_func_t		*vdev_op_nparity;
 	vdev_ndisks_func_t		*vdev_op_ndisks;
+	vdev_kobj_post_evt_func_t	*vdev_op_kobj_evt_post;
 	char				vdev_op_type[16];
 	boolean_t			vdev_op_leaf;
 } vdev_ops_t;
@@ -436,6 +438,7 @@ struct vdev {
 	boolean_t	vdev_isl2cache;	/* was a l2cache device		*/
 	boolean_t	vdev_copy_uberblocks;  /* post expand copy uberblocks */
 	boolean_t	vdev_resilver_deferred;  /* resilver deferred */
+	boolean_t	vdev_kobj_flag; /* kobj event record */
 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
 	vdev_cache_t	vdev_cache;	/* physical block cache		*/
 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/

include/sys/zfs_context.h

@@ -50,6 +50,7 @@ extern "C" {
 #include <sys/kmem.h>
 #include <sys/kmem_cache.h>
 #include <sys/vmem.h>
+#include <sys/misc.h>
 #include <sys/taskq.h>
 #include <sys/param.h>
 #include <sys/disp.h>