Tag zfs-2.0.6

META file and changelog updated.

Signed-off-by: Tony Hutter <hutter2@llnl.gov>

META

@@ -1,10 +1,10 @@
 Meta:          1
 Name:          zfs
 Branch:        1.0
-Version:       2.0.5
+Version:       2.0.6
 Release:       1
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS
-Linux-Maximum: 5.12
+Linux-Maximum: 5.14
 Linux-Minimum: 3.10

cmd/zdb/zdb.c

@@ -161,12 +161,6 @@ static int dump_bpobj_cb(void *arg, const blkptr_t *bp, boolean_t free,
     dmu_tx_t *tx);
 
 typedef struct sublivelist_verify {
-	/* all ALLOC'd blkptr_t in one sub-livelist */
-	zfs_btree_t sv_all_allocs;
-
-	/* all FREE'd blkptr_t in one sub-livelist */
-	zfs_btree_t sv_all_frees;
-
 	/* FREE's that haven't yet matched to an ALLOC, in one sub-livelist */
 	zfs_btree_t sv_pair;
 
@@ -225,29 +219,68 @@ typedef struct sublivelist_verify_block {
 
 static void zdb_print_blkptr(const blkptr_t *bp, int flags);
 
+typedef struct sublivelist_verify_block_refcnt {
+	/* block pointer entry in livelist being verified */
+	blkptr_t svbr_blk;
+
+	/*
+	 * Refcount gets incremented to 1 when we encounter the first
+	 * FREE entry for the svfbr block pointer and a node for it
+	 * is created in our ZDB verification/tracking metadata.
+	 *
+	 * As we encounter more FREE entries we increment this counter
+	 * and similarly decrement it whenever we find the respective
+	 * ALLOC entries for this block.
+	 *
+	 * When the refcount gets to 0 it means that all the FREE and
+	 * ALLOC entries of this block have paired up and we no longer
+	 * need to track it in our verification logic (e.g. the node
+	 * containing this struct in our verification data structure
+	 * should be freed).
+	 *
+	 * [refer to sublivelist_verify_blkptr() for the actual code]
+	 */
+	uint32_t svbr_refcnt;
+} sublivelist_verify_block_refcnt_t;
+
+static int
+sublivelist_block_refcnt_compare(const void *larg, const void *rarg)
+{
+	const sublivelist_verify_block_refcnt_t *l = larg;
+	const sublivelist_verify_block_refcnt_t *r = rarg;
+	return (livelist_compare(&l->svbr_blk, &r->svbr_blk));
+}
+
 static int
 sublivelist_verify_blkptr(void *arg, const blkptr_t *bp, boolean_t free,
     dmu_tx_t *tx)
 {
 	ASSERT3P(tx, ==, NULL);
 	struct sublivelist_verify *sv = arg;
-	char blkbuf[BP_SPRINTF_LEN];
+	sublivelist_verify_block_refcnt_t current = {
+			.svbr_blk = *bp,
+
+			/*
+			 * Start with 1 in case this is the first free entry.
+			 * This field is not used for our B-Tree comparisons
+			 * anyway.
+			 */
+			.svbr_refcnt = 1,
+	};
+
 	zfs_btree_index_t where;
+	sublivelist_verify_block_refcnt_t *pair =
+	    zfs_btree_find(&sv->sv_pair, &current, &where);
 	if (free) {
-		zfs_btree_add(&sv->sv_pair, bp);
-		/* Check if the FREE is a duplicate */
-		if (zfs_btree_find(&sv->sv_all_frees, bp, &where) != NULL) {
-			snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), bp,
-			    free);
-			(void) printf("\tERROR: Duplicate FREE: %s\n", blkbuf);
+		if (pair == NULL) {
+			/* first free entry for this block pointer */
+			zfs_btree_add(&sv->sv_pair, &current);
 		} else {
-			zfs_btree_add_idx(&sv->sv_all_frees, bp, &where);
+			pair->svbr_refcnt++;
 		}
 	} else {
-		/* Check if the ALLOC has been freed */
-		if (zfs_btree_find(&sv->sv_pair, bp, &where) != NULL) {
-			zfs_btree_remove_idx(&sv->sv_pair, &where);
-		} else {
+		if (pair == NULL) {
+			/* block that is currently marked as allocated */
 			for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
 				if (DVA_IS_EMPTY(&bp->blk_dva[i]))
 					break;
@@ -262,16 +295,16 @@ sublivelist_verify_blkptr(void *arg, const blkptr_t *bp, boolean_t free,
 					    &svb, &where);
 				}
 			}
-		}
-		/* Check if the ALLOC is a duplicate */
-		if (zfs_btree_find(&sv->sv_all_allocs, bp, &where) != NULL) {
-			snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), bp,
-			    free);
-			(void) printf("\tERROR: Duplicate ALLOC: %s\n", blkbuf);
 		} else {
-			zfs_btree_add_idx(&sv->sv_all_allocs, bp, &where);
+			/* alloc matches a free entry */
+			pair->svbr_refcnt--;
+			if (pair->svbr_refcnt == 0) {
+				/* all allocs and frees have been matched */
+				zfs_btree_remove_idx(&sv->sv_pair, &where);
+			}
 		}
 	}
+
 	return (0);
 }
 
@@ -279,32 +312,22 @@ static int
 sublivelist_verify_func(void *args, dsl_deadlist_entry_t *dle)
 {
 	int err;
-	char blkbuf[BP_SPRINTF_LEN];
 	struct sublivelist_verify *sv = args;
 
-	zfs_btree_create(&sv->sv_all_allocs, livelist_compare,
-	    sizeof (blkptr_t));
-
-	zfs_btree_create(&sv->sv_all_frees, livelist_compare,
-	    sizeof (blkptr_t));
-
-	zfs_btree_create(&sv->sv_pair, livelist_compare,
-	    sizeof (blkptr_t));
+	zfs_btree_create(&sv->sv_pair, sublivelist_block_refcnt_compare,
+	    sizeof (sublivelist_verify_block_refcnt_t));
 
 	err = bpobj_iterate_nofree(&dle->dle_bpobj, sublivelist_verify_blkptr,
 	    sv, NULL);
 
-	zfs_btree_clear(&sv->sv_all_allocs);
-	zfs_btree_destroy(&sv->sv_all_allocs);
-
-	zfs_btree_clear(&sv->sv_all_frees);
-	zfs_btree_destroy(&sv->sv_all_frees);
-
-	blkptr_t *e;
+	sublivelist_verify_block_refcnt_t *e;
 	zfs_btree_index_t *cookie = NULL;
 	while ((e = zfs_btree_destroy_nodes(&sv->sv_pair, &cookie)) != NULL) {
-		snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), e, B_TRUE);
-		(void) printf("\tERROR: Unmatched FREE: %s\n", blkbuf);
+		char blkbuf[BP_SPRINTF_LEN];
+		snprintf_blkptr_compact(blkbuf, sizeof (blkbuf),
+		    &e->svbr_blk, B_TRUE);
+		(void) printf("\tERROR: %d unmatched FREE(s): %s\n",
+		    e->svbr_refcnt, blkbuf);
 	}
 	zfs_btree_destroy(&sv->sv_pair);
 
@@ -613,10 +636,14 @@ mv_populate_livelist_allocs(metaslab_verify_t *mv, sublivelist_verify_t *sv)
 /*
  * [Livelist Check]
  * Iterate through all the sublivelists and:
- * - report leftover frees
- * - report double ALLOCs/FREEs
+ * - report leftover frees (**)
  * - record leftover ALLOCs together with their TXG [see Cross Check]
  *
+ * (**) Note: Double ALLOCs are valid in datasets that have dedup
+ *      enabled. Similarly double FREEs are allowed as well but
+ *      only if they pair up with a corresponding ALLOC entry once
+ *      we our done with our sublivelist iteration.
+ *
  * [Spacemap Check]
  * for each metaslab:
  * - iterate over spacemap and then the metaslab's entries in the

config/kernel-acl.m4

@@ -162,6 +162,9 @@ dnl #
 dnl # 3.1 API change,
 dnl # Check if inode_operations contains the function get_acl
 dnl #
+dnl # 5.15 API change,
+dnl # Added the bool rcu argument to get_acl for rcu path walk.
+dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_INODE_OPERATIONS_GET_ACL], [
 	ZFS_LINUX_TEST_SRC([inode_operations_get_acl], [
 		#include <linux/fs.h>
@@ -174,14 +177,32 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_INODE_OPERATIONS_GET_ACL], [
 			.get_acl = get_acl_fn,
 		};
 	],[])
+
+	ZFS_LINUX_TEST_SRC([inode_operations_get_acl_rcu], [
+		#include <linux/fs.h>
+
+		struct posix_acl *get_acl_fn(struct inode *inode, int type,
+		    bool rcu) { return NULL; }
+
+		static const struct inode_operations
+		    iops __attribute__ ((unused)) = {
+			.get_acl = get_acl_fn,
+		};
+	],[])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_INODE_OPERATIONS_GET_ACL], [
 	AC_MSG_CHECKING([whether iops->get_acl() exists])
 	ZFS_LINUX_TEST_RESULT([inode_operations_get_acl], [
 		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_GET_ACL, 1, [iops->get_acl() exists])
 	],[
-		ZFS_LINUX_TEST_ERROR([iops->get_acl()])
+		ZFS_LINUX_TEST_RESULT([inode_operations_get_acl_rcu], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_GET_ACL_RCU, 1, [iops->get_acl() takes rcu])
+		],[
+			ZFS_LINUX_TEST_ERROR([iops->get_acl()])
+		])
 	])
 ])
 

config/kernel-blk-queue.m4

@@ -48,7 +48,45 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_BDI], [
 ])
 
 dnl #
-dnl # 2.6.32 - 4.x API,
+dnl # 5.9: added blk_queue_update_readahead(),
+dnl # 5.15: renamed to disk_update_readahead()
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_UPDATE_READAHEAD], [
+	ZFS_LINUX_TEST_SRC([blk_queue_update_readahead], [
+		#include <linux/blkdev.h>
+	],[
+		struct request_queue q;
+		blk_queue_update_readahead(&q);
+	])
+
+	ZFS_LINUX_TEST_SRC([disk_update_readahead], [
+		#include <linux/blkdev.h>
+	],[
+		struct gendisk disk;
+		disk_update_readahead(&disk);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_UPDATE_READAHEAD], [
+	AC_MSG_CHECKING([whether blk_queue_update_readahead() exists])
+	ZFS_LINUX_TEST_RESULT([blk_queue_update_readahead], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BLK_QUEUE_UPDATE_READAHEAD, 1,
+		    [blk_queue_update_readahead() exists])
+	],[
+		AC_MSG_CHECKING([whether disk_update_readahead() exists])
+		ZFS_LINUX_TEST_RESULT([disk_update_readahead], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_DISK_UPDATE_READAHEAD, 1,
+			    [disk_update_readahead() exists])
+		],[
+			AC_MSG_RESULT(no)
+		])
+	])
+])
+
+dnl #
+dnl # 2.6.32 API,
 dnl #   blk_queue_discard()
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_DISCARD], [
@@ -280,6 +318,7 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_MAX_SEGMENTS], [
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE], [
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_PLUG
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_BDI
+	ZFS_AC_KERNEL_SRC_BLK_QUEUE_UPDATE_READAHEAD
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_DISCARD
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_SECURE_ERASE
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_FLAG_SET
@@ -292,6 +331,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE], [
 AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE], [
 	ZFS_AC_KERNEL_BLK_QUEUE_PLUG
 	ZFS_AC_KERNEL_BLK_QUEUE_BDI
+	ZFS_AC_KERNEL_BLK_QUEUE_UPDATE_READAHEAD
 	ZFS_AC_KERNEL_BLK_QUEUE_DISCARD
 	ZFS_AC_KERNEL_BLK_QUEUE_SECURE_ERASE
 	ZFS_AC_KERNEL_BLK_QUEUE_FLAG_SET

config/kernel-make-request-fn.m4

@@ -42,6 +42,13 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_MAKE_REQUEST_FN], [
 		struct block_device_operations o;
 		o.submit_bio = NULL;
 	])
+
+	ZFS_LINUX_TEST_SRC([blk_alloc_disk], [
+		#include <linux/blkdev.h>
+	],[
+		struct gendisk *disk  __attribute__ ((unused));
+		disk = blk_alloc_disk(NUMA_NO_NODE);
+	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_MAKE_REQUEST_FN], [
@@ -56,6 +63,19 @@ AC_DEFUN([ZFS_AC_KERNEL_MAKE_REQUEST_FN], [
 
 		AC_DEFINE(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS, 1,
 		    [submit_bio is member of struct block_device_operations])
+
+		dnl #
+		dnl # Linux 5.14 API Change:
+		dnl # blk_alloc_queue() + alloc_disk() combo replaced by
+		dnl # a single call to blk_alloc_disk().
+		dnl #
+		AC_MSG_CHECKING([whether blk_alloc_disk() exists])
+		ZFS_LINUX_TEST_RESULT([blk_alloc_disk], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE([HAVE_BLK_ALLOC_DISK], 1, [blk_alloc_disk() exists])
+		], [
+			AC_MSG_RESULT(no)
+		])
 	],[
 		AC_MSG_RESULT(no)
 

config/kernel-siginfo.m4

@@ -0,0 +1,21 @@
+dnl #
+dnl # 4.20 API change
+dnl # Added kernel_siginfo_t
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SIGINFO], [
+	ZFS_LINUX_TEST_SRC([siginfo], [
+		#include <linux/signal_types.h>
+	],[
+		kernel_siginfo_t info __attribute__ ((unused));
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SIGINFO], [
+	AC_MSG_CHECKING([whether kernel_siginfo_t tyepedef exists])
+	ZFS_LINUX_TEST_RESULT([siginfo], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SIGINFO, 1, [kernel_siginfo_t exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-signal-stop.m4

@@ -0,0 +1,21 @@
+dnl #
+dnl # 4.4 API change
+dnl # Added kernel_signal_stop
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SIGNAL_STOP], [
+	ZFS_LINUX_TEST_SRC([signal_stop], [
+		#include <linux/sched/signal.h>
+	],[
+		kernel_signal_stop();
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SIGNAL_STOP], [
+	AC_MSG_CHECKING([whether signal_stop() exists])
+	ZFS_LINUX_TEST_RESULT([signal_stop], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SIGNAL_STOP, 1, [signal_stop() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-special-state.m4

@@ -0,0 +1,21 @@
+dnl #
+dnl # 4.17 API change
+dnl # Added set_special_state() function
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SET_SPECIAL_STATE], [
+	ZFS_LINUX_TEST_SRC([set_special_state], [
+		#include <linux/sched.h>
+	],[
+		set_special_state(TASK_STOPPED);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SET_SPECIAL_STATE], [
+	AC_MSG_CHECKING([whether set_special_state() exists])
+	ZFS_LINUX_TEST_RESULT([set_special_state], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SET_SPECIAL_STATE, 1, [set_special_state() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-stdarg.m4

@@ -0,0 +1,32 @@
+dnl #
+dnl # Linux 5.15 gets rid of -isystem and external <stdarg.h> inclusion
+dnl # and ships its own <linux/stdarg.h>. Check if this header file does
+dnl # exist and provide all necessary definitions for variable argument
+dnl # functions. Adjust the inclusion of <stdarg.h> according to the
+dnl # results.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_STANDALONE_LINUX_STDARG], [
+	ZFS_LINUX_TEST_SRC([has_standalone_linux_stdarg], [
+		#include <linux/stdarg.h>
+
+		#if !defined(va_start) || !defined(va_end) || \
+		    !defined(va_arg) || !defined(va_copy)
+		#error "<linux/stdarg.h> is invalid"
+		#endif
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_STANDALONE_LINUX_STDARG], [
+	dnl #
+	dnl # Linux 5.15 ships its own stdarg.h and doesn't allow to
+	dnl # include compiler headers.
+	dnl #
+	AC_MSG_CHECKING([whether standalone <linux/stdarg.h> exists])
+	ZFS_LINUX_TEST_RESULT([has_standalone_linux_stdarg], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_STANDALONE_LINUX_STDARG, 1,
+			[standalone <linux/stdarg.h> exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-vfs-set_page_dirty.m4

@@ -0,0 +1,34 @@
+dnl #
+dnl # Linux 5.14 adds a change to require set_page_dirty to be manually
+dnl # wired up in struct address_space_operations. Determine if this needs
+dnl # to be done. This patch set also introduced __set_page_dirty_nobuffers
+dnl # declaration in linux/pagemap.h, so these tests look for the presence
+dnl # of that function to tell the compiler to assign set_page_dirty in
+dnl # module/os/linux/zfs/zpl_file.c
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS], [
+	ZFS_LINUX_TEST_SRC([vfs_has_set_page_dirty_nobuffers], [
+		#include <linux/pagemap.h>
+		#include <linux/fs.h>
+
+		static const struct address_space_operations
+		    aops __attribute__ ((unused)) = {
+			.set_page_dirty = __set_page_dirty_nobuffers,
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS], [
+	dnl #
+	dnl # Linux 5.14 change requires set_page_dirty() to be assigned
+	dnl # in address_space_operations()
+	dnl #
+	AC_MSG_CHECKING([__set_page_dirty_nobuffers exists])
+	ZFS_LINUX_TEST_RESULT([vfs_has_set_page_dirty_nobuffers], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS, 1,
+			[__set_page_dirty_nobuffers exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel.m4

@@ -128,6 +128,11 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_MKNOD
 	ZFS_AC_KERNEL_SRC_SYMLINK
 	ZFS_AC_KERNEL_SRC_BIO_MAX_SEGS
+	ZFS_AC_KERNEL_SRC_SIGNAL_STOP
+	ZFS_AC_KERNEL_SRC_SIGINFO
+	ZFS_AC_KERNEL_SRC_SET_SPECIAL_STATE
+	ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS
+	ZFS_AC_KERNEL_SRC_STANDALONE_LINUX_STDARG
 
 	AC_MSG_CHECKING([for available kernel interfaces])
 	ZFS_LINUX_TEST_COMPILE_ALL([kabi])
@@ -229,6 +234,11 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_MKNOD
 	ZFS_AC_KERNEL_SYMLINK
 	ZFS_AC_KERNEL_BIO_MAX_SEGS
+	ZFS_AC_KERNEL_SIGNAL_STOP
+	ZFS_AC_KERNEL_SIGINFO
+	ZFS_AC_KERNEL_SET_SPECIAL_STATE
+	ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS
+	ZFS_AC_KERNEL_STANDALONE_LINUX_STDARG
 ])
 
 dnl #

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

@@ -92,11 +92,14 @@ blk_queue_set_write_cache(struct request_queue *q, bool wc, bool fua)
 static inline void
 blk_queue_set_read_ahead(struct request_queue *q, unsigned long ra_pages)
 {
+#if !defined(HAVE_BLK_QUEUE_UPDATE_READAHEAD) && \
+	!defined(HAVE_DISK_UPDATE_READAHEAD)
 #ifdef HAVE_BLK_QUEUE_BDI_DYNAMIC
 	q->backing_dev_info->ra_pages = ra_pages;
 #else
 	q->backing_dev_info.ra_pages = ra_pages;
 #endif
+#endif
 }
 
 #ifdef HAVE_BIO_BVEC_ITER

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

@@ -24,7 +24,11 @@
 #ifndef _SPL_CMN_ERR_H
 #define	_SPL_CMN_ERR_H
 
+#if defined(_KERNEL) && defined(HAVE_STANDALONE_LINUX_STDARG)
+#include <linux/stdarg.h>
+#else
 #include <stdarg.h>
+#endif
 
 #define	CE_CONT		0 /* continuation */
 #define	CE_NOTE		1 /* notice */

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

@@ -33,22 +33,6 @@
 #define	FORREAL		0	/* Usual side-effects */
 #define	JUSTLOOKING	1	/* Don't stop the process */
 
-/*
- * The "why" argument indicates the allowable side-effects of the call:
- *
- * FORREAL:  Extract the next pending signal from p_sig into p_cursig;
- * stop the process if a stop has been requested or if a traced signal
- * is pending.
- *
- * JUSTLOOKING:  Don't stop the process, just indicate whether or not
- * a signal might be pending (FORREAL is needed to tell for sure).
- */
-static __inline__ int
-issig(int why)
-{
-	ASSERT(why == FORREAL || why == JUSTLOOKING);
-
-	return (signal_pending(current));
-}
+extern int issig(int why);
 
 #endif /* SPL_SIGNAL_H */

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

@@ -70,4 +70,17 @@ extern struct task_struct *spl_kthread_create(int (*func)(void *),
 
 extern proc_t p0;
 
+#ifdef HAVE_SIGINFO
+typedef kernel_siginfo_t spl_kernel_siginfo_t;
+#else
+typedef siginfo_t spl_kernel_siginfo_t;
+#endif
+
+#ifdef HAVE_SET_SPECIAL_STATE
+#define	spl_set_special_state(x) set_special_state((x))
+#else
+#define	spl_set_special_state(x) __set_current_state((x))
+#endif
+
+
 #endif  /* _SPL_THREAD_H */

include/os/linux/zfs/sys/zpl.h

@@ -70,7 +70,11 @@ extern int zpl_set_acl(struct user_namespace *userns, struct inode *ip,
 extern int zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type);
 #endif /* HAVE_SET_ACL_USERNS */
 #endif /* HAVE_SET_ACL */
+#if defined(HAVE_GET_ACL_RCU)
+extern struct posix_acl *zpl_get_acl(struct inode *ip, int type, bool rcu);
+#elif defined(HAVE_GET_ACL)
 extern struct posix_acl *zpl_get_acl(struct inode *ip, int type);
+#endif
 extern int zpl_init_acl(struct inode *ip, struct inode *dir);
 extern int zpl_chmod_acl(struct inode *ip);
 #else

include/sys/dsl_synctask.h

@@ -41,10 +41,11 @@ typedef void (dsl_sigfunc_t)(void *, dmu_tx_t *);
 
 typedef enum zfs_space_check {
 	/*
-	 * Normal space check: if there is less than 3.2% free space,
-	 * the operation will fail.  Operations which are logically
-	 * creating things should use this (e.g. "zfs create", "zfs snapshot").
-	 * User writes (via the ZPL / ZVOL) also fail at this point.
+	 * Normal space check: if there is less than 3.2% free space (bounded
+	 * by spa_max_slop), the operation will fail.  Operations which are
+	 * logically creating things should use this (e.g. "zfs create", "zfs
+	 * snapshot").  User writes (via the ZPL / ZVOL) also fail at this
+	 * point.
 	 */
 	ZFS_SPACE_CHECK_NORMAL,
 

include/sys/fm/util.h

@@ -31,6 +31,7 @@ extern "C" {
 #endif
 
 #include <sys/nvpair.h>
+#include <sys/zfs_file.h>
 
 /*
  * Shared user/kernel definitions for class length, error channel name,
@@ -96,8 +97,8 @@ extern void fm_nvprint(nvlist_t *);
 extern void zfs_zevent_post_cb(nvlist_t *nvl, nvlist_t *detector);
 extern int zfs_zevent_post(nvlist_t *, nvlist_t *, zevent_cb_t *);
 extern void zfs_zevent_drain_all(int *);
-extern int zfs_zevent_fd_hold(int, minor_t *, zfs_zevent_t **);
-extern void zfs_zevent_fd_rele(int);
+extern zfs_file_t *zfs_zevent_fd_hold(int, minor_t *, zfs_zevent_t **);
+extern void zfs_zevent_fd_rele(zfs_file_t *);
 extern int zfs_zevent_next(zfs_zevent_t *, nvlist_t **, uint64_t *, uint64_t *);
 extern int zfs_zevent_wait(zfs_zevent_t *);
 extern int zfs_zevent_seek(zfs_zevent_t *, uint64_t);

include/sys/zfs_file.h

@@ -22,6 +22,8 @@
 #ifndef	_SYS_ZFS_FILE_H
 #define	_SYS_ZFS_FILE_H
 
+#include <sys/zfs_context.h>
+
 #ifndef _KERNEL
 typedef struct zfs_file {
 	int f_fd;
@@ -55,8 +57,8 @@ int zfs_file_fallocate(zfs_file_t *fp, int mode, loff_t offset, loff_t len);
 loff_t zfs_file_off(zfs_file_t *fp);
 int zfs_file_unlink(const char *);
 
-int zfs_file_get(int fd, zfs_file_t **fp);
-void zfs_file_put(int fd);
+zfs_file_t *zfs_file_get(int fd);
+void zfs_file_put(zfs_file_t *fp);
 void *zfs_file_private(zfs_file_t *fp);
 
 #endif /* _SYS_ZFS_FILE_H */

include/sys/zfs_ioctl.h

@@ -566,7 +566,7 @@ typedef struct zfsdev_state {
 } zfsdev_state_t;
 
 extern void *zfsdev_get_state(minor_t minor, enum zfsdev_state_type which);
-extern int zfsdev_getminor(int fd, minor_t *minorp);
+extern int zfsdev_getminor(zfs_file_t *fp, minor_t *minorp);
 extern minor_t zfsdev_minor_alloc(void);
 
 extern uint_t zfs_fsyncer_key;

include/sys/zfs_onexit.h

@@ -51,8 +51,8 @@ extern void zfs_onexit_destroy(zfs_onexit_t *zo);
 
 #endif
 
-extern int zfs_onexit_fd_hold(int fd, minor_t *minorp);
-extern void zfs_onexit_fd_rele(int fd);
+extern zfs_file_t *zfs_onexit_fd_hold(int fd, minor_t *minorp);
+extern void zfs_onexit_fd_rele(zfs_file_t *);
 extern int zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
     uint64_t *action_handle);
 

lib/libzpool/kernel.c

@@ -966,16 +966,16 @@ kmem_asprintf(const char *fmt, ...)
 }
 
 /* ARGSUSED */
-int
+zfs_file_t *
 zfs_onexit_fd_hold(int fd, minor_t *minorp)
 {
 	*minorp = 0;
-	return (0);
+	return (NULL);
 }
 
 /* ARGSUSED */
 void
-zfs_onexit_fd_rele(int fd)
+zfs_onexit_fd_rele(zfs_file_t *fp)
 {
 }
 
@@ -1385,28 +1385,26 @@ zfs_file_unlink(const char *path)
  * Get reference to file pointer
  *
  * fd - input file descriptor
- * fpp - pointer to file pointer
  *
- * Returns 0 on success EBADF on failure.
+ * Returns pointer to file struct or NULL.
  * Unsupported in user space.
  */
-int
-zfs_file_get(int fd, zfs_file_t **fpp)
+zfs_file_t *
+zfs_file_get(int fd)
 {
 	abort();
 
-	return (EOPNOTSUPP);
+	return (NULL);
 }
-
 /*
  * Drop reference to file pointer
  *
- * fd - input file descriptor
+ * fp - pointer to file struct
  *
  * Unsupported in user space.
  */
 void
-zfs_file_put(int fd)
+zfs_file_put(zfs_file_t *fp)
 {
 	abort();
 }

module/os/freebsd/zfs/crypto_os.c

@@ -172,11 +172,13 @@ zfs_crypto_dispatch(freebsd_crypt_session_t *session, 	struct cryptop *crp)
 			break;
 		mtx_lock(&session->fs_lock);
 		while (session->fs_done == false)
-			msleep(crp, &session->fs_lock, PRIBIO,
-			    "zfs_crypto", hz/5);
+			msleep(crp, &session->fs_lock, 0,
+			    "zfs_crypto", 0);
 		mtx_unlock(&session->fs_lock);
 
-		if (crp->crp_etype != EAGAIN) {
+		if (crp->crp_etype == ENOMEM) {
+			pause("zcrnomem", 1);
+		} else if (crp->crp_etype != EAGAIN) {
 			error = crp->crp_etype;
 			break;
 		}

module/os/freebsd/zfs/vdev_geom.c

@@ -379,7 +379,11 @@ vdev_geom_io(struct g_consumer *cp, int *cmds, void **datas, off_t *offsets,
 	int i, n_bios, j;
 	size_t bios_size;
 
+#if __FreeBSD_version > 1300130
+	maxio = maxphys - (maxphys % cp->provider->sectorsize);
+#else
 	maxio = MAXPHYS - (MAXPHYS % cp->provider->sectorsize);
+#endif
 	n_bios = 0;
 
 	/* How many bios are required for all commands ? */

module/os/freebsd/zfs/zfs_file_os.c

@@ -241,28 +241,21 @@ zfs_file_fsync(zfs_file_t *fp, int flags)
 	return (zfs_vop_fsync(fp->f_vnode));
 }
 
-int
-zfs_file_get(int fd, zfs_file_t **fpp)
+zfs_file_t *
+zfs_file_get(int fd)
 {
 	struct file *fp;
 
 	if (fget(curthread, fd, &cap_no_rights, &fp))
-		return (SET_ERROR(EBADF));
+		return (NULL);
 
-	*fpp = fp;
-	return (0);
+	return (fp);
 }
 
 void
-zfs_file_put(int fd)
+zfs_file_put(zfs_file_t *fp)
 {
-	struct file *fp;
-
-	/* No CAP_ rights required, as we're only releasing. */
-	if (fget(curthread, fd, &cap_no_rights, &fp) == 0) {
-		fdrop(fp, curthread);
-		fdrop(fp, curthread);
-	}
+	fdrop(fp, curthread);
 }
 
 loff_t

module/os/freebsd/zfs/zvol_os.c

@@ -1229,7 +1229,11 @@ zvol_rename_minor(zvol_state_t *zv, const char *newname)
 		args.mda_si_drv2 = zv;
 		if (make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, newname)
 		    == 0) {
+#if __FreeBSD_version > 1300130
+			dev->si_iosize_max = maxphys;
+#else
 			dev->si_iosize_max = MAXPHYS;
+#endif
 			zsd->zsd_cdev = dev;
 		}
 	}
@@ -1265,9 +1269,10 @@ zvol_free(zvol_state_t *zv)
 		struct zvol_state_dev *zsd = &zv->zv_zso->zso_dev;
 		struct cdev *dev = zsd->zsd_cdev;
 
-		ASSERT3P(dev->si_drv2, ==, NULL);
-
-		destroy_dev(dev);
+		if (dev != NULL) {
+			ASSERT3P(dev->si_drv2, ==, NULL);
+			destroy_dev(dev);
+		}
 	}
 
 	mutex_destroy(&zv->zv_state_lock);
@@ -1362,16 +1367,15 @@ zvol_create_minor_impl(const char *name)
 		args.mda_gid = GID_OPERATOR;
 		args.mda_mode = 0640;
 		args.mda_si_drv2 = zv;
-		error = make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, name);
-		if (error) {
-			kmem_free(zv->zv_zso, sizeof (struct zvol_state_os));
-			mutex_destroy(&zv->zv_state_lock);
-			kmem_free(zv, sizeof (*zv));
-			dmu_objset_disown(os, B_TRUE, FTAG);
-			goto out_doi;
+		if (make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, name)
+		    == 0) {
+#if __FreeBSD_version > 1300130
+			dev->si_iosize_max = maxphys;
+#else
+			dev->si_iosize_max = MAXPHYS;
+#endif
+			zsd->zsd_cdev = dev;
 		}
-		dev->si_iosize_max = MAXPHYS;
-		zsd->zsd_cdev = dev;
 	}
 	(void) strlcpy(zv->zv_name, name, MAXPATHLEN);
 	rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL);
@@ -1444,7 +1448,8 @@ zvol_clear_private(zvol_state_t *zv)
 		struct zvol_state_dev *zsd = &zv->zv_zso->zso_dev;
 		struct cdev *dev = zsd->zsd_cdev;
 
-		dev->si_drv2 = NULL;
+		if (dev != NULL)
+			dev->si_drv2 = NULL;
 	}
 }
 

module/os/linux/spl/spl-thread.c

@@ -158,3 +158,54 @@ spl_kthread_create(int (*func)(void *), void *data, const char namefmt[], ...)
 	} while (1);
 }
 EXPORT_SYMBOL(spl_kthread_create);
+
+/*
+ * The "why" argument indicates the allowable side-effects of the call:
+ *
+ * FORREAL:  Extract the next pending signal from p_sig into p_cursig;
+ * stop the process if a stop has been requested or if a traced signal
+ * is pending.
+ *
+ * JUSTLOOKING:  Don't stop the process, just indicate whether or not
+ * a signal might be pending (FORREAL is needed to tell for sure).
+ */
+int
+issig(int why)
+{
+	ASSERT(why == FORREAL || why == JUSTLOOKING);
+
+	if (!signal_pending(current))
+		return (0);
+
+	if (why != FORREAL)
+		return (1);
+
+	struct task_struct *task = current;
+	spl_kernel_siginfo_t __info;
+	sigset_t set;
+	siginitsetinv(&set, 1ULL << (SIGSTOP - 1) | 1ULL << (SIGTSTP - 1));
+	sigorsets(&set, &task->blocked, &set);
+
+	spin_lock_irq(&task->sighand->siglock);
+	int ret;
+	if ((ret = dequeue_signal(task, &set, &__info)) != 0) {
+#ifdef HAVE_SIGNAL_STOP
+		spin_unlock_irq(&task->sighand->siglock);
+		kernel_signal_stop();
+#else
+		if (current->jobctl & JOBCTL_STOP_DEQUEUED)
+			spl_set_special_state(TASK_STOPPED);
+
+		spin_unlock_irq(&current->sighand->siglock);
+
+		schedule();
+#endif
+		return (0);
+	}
+
+	spin_unlock_irq(&task->sighand->siglock);
+
+	return (1);
+}
+
+EXPORT_SYMBOL(issig);

module/os/linux/zfs/zfs_file_os.c

@@ -405,36 +405,22 @@ zfs_file_unlink(const char *path)
  * Get reference to file pointer
  *
  * fd - input file descriptor
- * fpp - pointer to file pointer
  *
- * Returns 0 on success EBADF on failure.
+ * Returns pointer to file struct or NULL
  */
-int
-zfs_file_get(int fd, zfs_file_t **fpp)
+zfs_file_t *
+zfs_file_get(int fd)
 {
-	zfs_file_t *fp;
-
-	fp = fget(fd);
-	if (fp == NULL)
-		return (EBADF);
-
-	*fpp = fp;
-
-	return (0);
+	return (fget(fd));
 }
 
 /*
  * Drop reference to file pointer
  *
- * fd - input file descriptor
+ * fp - input file struct pointer
  */
 void
-zfs_file_put(int fd)
+zfs_file_put(zfs_file_t *fp)
 {
-	struct file *fp;
-
-	if ((fp = fget(fd)) != NULL) {
-		fput(fp);
-		fput(fp);
-	}
+	fput(fp);
 }

module/os/linux/zfs/zpl_file.c

@@ -33,6 +33,9 @@
 #include <sys/zfs_vfsops.h>
 #include <sys/zfs_vnops.h>
 #include <sys/zfs_project.h>
+#ifdef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS
+#include <linux/pagemap.h>
+#endif
 
 /*
  * When using fallocate(2) to preallocate space, inflate the requested
@@ -1007,6 +1010,9 @@ const struct address_space_operations zpl_address_space_operations = {
 	.writepage	= zpl_writepage,
 	.writepages	= zpl_writepages,
 	.direct_IO	= zpl_direct_IO,
+#ifdef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS
+	.set_page_dirty = __set_page_dirty_nobuffers,
+#endif
 };
 
 const struct file_operations zpl_file_operations = {

module/os/linux/zfs/zpl_xattr.c

@@ -1012,13 +1012,12 @@ zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type)
 }
 #endif /* HAVE_SET_ACL */
 
-struct posix_acl *
-zpl_get_acl(struct inode *ip, int type)
+static struct posix_acl *
+zpl_get_acl_impl(struct inode *ip, int type)
 {
 	struct posix_acl *acl;
 	void *value = NULL;
 	char *name;
-	int size;
 
 	/*
 	 * As of Linux 3.14, the kernel get_acl will check this for us.
@@ -1042,7 +1041,7 @@ zpl_get_acl(struct inode *ip, int type)
 		return (ERR_PTR(-EINVAL));
 	}
 
-	size = zpl_xattr_get(ip, name, NULL, 0);
+	int size = zpl_xattr_get(ip, name, NULL, 0);
 	if (size > 0) {
 		value = kmem_alloc(size, KM_SLEEP);
 		size = zpl_xattr_get(ip, name, value, size);
@@ -1068,6 +1067,25 @@ zpl_get_acl(struct inode *ip, int type)
 	return (acl);
 }
 
+#if defined(HAVE_GET_ACL_RCU)
+struct posix_acl *
+zpl_get_acl(struct inode *ip, int type, bool rcu)
+{
+	if (rcu)
+		return (ERR_PTR(-ECHILD));
+
+	return (zpl_get_acl_impl(ip, type));
+}
+#elif defined(HAVE_GET_ACL)
+struct posix_acl *
+zpl_get_acl(struct inode *ip, int type)
+{
+	return (zpl_get_acl_impl(ip, type));
+}
+#else
+#error "Unsupported iops->get_acl() implementation"
+#endif /* HAVE_GET_ACL_RCU */
+
 int
 zpl_init_acl(struct inode *ip, struct inode *dir)
 {
@@ -1078,7 +1096,7 @@ zpl_init_acl(struct inode *ip, struct inode *dir)
 		return (0);
 
 	if (!S_ISLNK(ip->i_mode)) {
-		acl = zpl_get_acl(dir, ACL_TYPE_DEFAULT);
+		acl = zpl_get_acl_impl(dir, ACL_TYPE_DEFAULT);
 		if (IS_ERR(acl))
 			return (PTR_ERR(acl));
 		if (!acl) {
@@ -1127,7 +1145,7 @@ zpl_chmod_acl(struct inode *ip)
 	if (S_ISLNK(ip->i_mode))
 		return (-EOPNOTSUPP);
 
-	acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
+	acl = zpl_get_acl_impl(ip, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl) || !acl)
 		return (PTR_ERR(acl));
 
@@ -1189,7 +1207,7 @@ __zpl_xattr_acl_get_access(struct inode *ip, const char *name,
 	if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
 		return (-EOPNOTSUPP);
 
-	acl = zpl_get_acl(ip, type);
+	acl = zpl_get_acl_impl(ip, type);
 	if (IS_ERR(acl))
 		return (PTR_ERR(acl));
 	if (acl == NULL)
@@ -1217,7 +1235,7 @@ __zpl_xattr_acl_get_default(struct inode *ip, const char *name,
 	if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
 		return (-EOPNOTSUPP);
 
-	acl = zpl_get_acl(ip, type);
+	acl = zpl_get_acl_impl(ip, type);
 	if (IS_ERR(acl))
 		return (PTR_ERR(acl));
 	if (acl == NULL)

module/os/linux/zfs/zvol_os.c

@@ -727,7 +727,7 @@ static struct block_device_operations zvol_ops = {
 	.getgeo			= zvol_getgeo,
 	.owner			= THIS_MODULE,
 #ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
-    .submit_bio		= zvol_submit_bio,
+	.submit_bio		= zvol_submit_bio,
 #endif
 };
 
@@ -760,13 +760,40 @@ zvol_alloc(dev_t dev, const char *name)
 	mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL);
 
 #ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
-	zso->zvo_queue = blk_alloc_queue(NUMA_NO_NODE);
+#ifdef HAVE_BLK_ALLOC_DISK
+	zso->zvo_disk = blk_alloc_disk(NUMA_NO_NODE);
+	if (zso->zvo_disk == NULL)
+		goto out_kmem;
+
+	zso->zvo_disk->minors = ZVOL_MINORS;
+	zso->zvo_queue = zso->zvo_disk->queue;
 #else
-	zso->zvo_queue = blk_generic_alloc_queue(zvol_request, NUMA_NO_NODE);
-#endif
+	zso->zvo_queue = blk_alloc_queue(NUMA_NO_NODE);
 	if (zso->zvo_queue == NULL)
 		goto out_kmem;
 
+	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
+	if (zso->zvo_disk == NULL) {
+		blk_cleanup_queue(zso->zvo_queue);
+		goto out_kmem;
+	}
+
+	zso->zvo_disk->queue = zso->zvo_queue;
+#endif /* HAVE_BLK_ALLOC_DISK */
+#else
+	zso->zvo_queue = blk_generic_alloc_queue(zvol_request, NUMA_NO_NODE);
+	if (zso->zvo_queue == NULL)
+		goto out_kmem;
+
+	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
+	if (zso->zvo_disk == NULL) {
+		blk_cleanup_queue(zso->zvo_queue);
+		goto out_kmem;
+	}
+
+	zso->zvo_disk->queue = zso->zvo_queue;
+#endif /* HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */
+
 	blk_queue_set_write_cache(zso->zvo_queue, B_TRUE, B_TRUE);
 
 	/* Limit read-ahead to a single page to prevent over-prefetching. */
@@ -775,10 +802,6 @@ zvol_alloc(dev_t dev, const char *name)
 	/* Disable write merging in favor of the ZIO pipeline. */
 	blk_queue_flag_set(QUEUE_FLAG_NOMERGES, zso->zvo_queue);
 
-	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
-	if (zso->zvo_disk == NULL)
-		goto out_queue;
-
 	zso->zvo_queue->queuedata = zv;
 	zso->zvo_dev = dev;
 	zv->zv_open_count = 0;
@@ -809,14 +832,11 @@ zvol_alloc(dev_t dev, const char *name)
 	zso->zvo_disk->first_minor = (dev & MINORMASK);
 	zso->zvo_disk->fops = &zvol_ops;
 	zso->zvo_disk->private_data = zv;
-	zso->zvo_disk->queue = zso->zvo_queue;
 	snprintf(zso->zvo_disk->disk_name, DISK_NAME_LEN, "%s%d",
 	    ZVOL_DEV_NAME, (dev & MINORMASK));
 
 	return (zv);
 
-out_queue:
-	blk_cleanup_queue(zso->zvo_queue);
 out_kmem:
 	kmem_free(zso, sizeof (struct zvol_state_os));
 	kmem_free(zv, sizeof (zvol_state_t));
@@ -847,8 +867,13 @@ zvol_free(zvol_state_t *zv)
 	zfs_rangelock_fini(&zv->zv_rangelock);
 
 	del_gendisk(zv->zv_zso->zvo_disk);
+#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
+	defined(HAVE_BLK_ALLOC_DISK)
+	blk_cleanup_disk(zv->zv_zso->zvo_disk);
+#else
 	blk_cleanup_queue(zv->zv_zso->zvo_queue);
 	put_disk(zv->zv_zso->zvo_disk);
+#endif
 
 	ida_simple_remove(&zvol_ida,
 	    MINOR(zv->zv_zso->zvo_dev) >> ZVOL_MINOR_BITS);

module/zfs/dbuf.c

@@ -1090,42 +1090,6 @@ dbuf_set_data(dmu_buf_impl_t *db, arc_buf_t *buf)
 	db->db.db_data = buf->b_data;
 }
 
-static arc_buf_t *
-dbuf_alloc_arcbuf_from_arcbuf(dmu_buf_impl_t *db, arc_buf_t *data)
-{
-	objset_t *os = db->db_objset;
-	spa_t *spa = os->os_spa;
-	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
-	enum zio_compress compress_type;
-	uint8_t complevel;
-	int psize, lsize;
-
-	psize = arc_buf_size(data);
-	lsize = arc_buf_lsize(data);
-	compress_type = arc_get_compression(data);
-	complevel = arc_get_complevel(data);
-
-	if (arc_is_encrypted(data)) {
-		boolean_t byteorder;
-		uint8_t salt[ZIO_DATA_SALT_LEN];
-		uint8_t iv[ZIO_DATA_IV_LEN];
-		uint8_t mac[ZIO_DATA_MAC_LEN];
-		dnode_t *dn = DB_DNODE(db);
-
-		arc_get_raw_params(data, &byteorder, salt, iv, mac);
-		data = arc_alloc_raw_buf(spa, db, dmu_objset_id(os),
-		    byteorder, salt, iv, mac, dn->dn_type, psize, lsize,
-		    compress_type, complevel);
-	} else if (compress_type != ZIO_COMPRESS_OFF) {
-		ASSERT3U(type, ==, ARC_BUFC_DATA);
-		data = arc_alloc_compressed_buf(spa, db,
-		    psize, lsize, compress_type, complevel);
-	} else {
-		data = arc_alloc_buf(spa, db, type, psize);
-	}
-	return (data);
-}
-
 static arc_buf_t *
 dbuf_alloc_arcbuf(dmu_buf_impl_t *db)
 {
@@ -1575,9 +1539,35 @@ dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg)
 		arc_space_consume(bonuslen, ARC_SPACE_BONUS);
 		bcopy(db->db.db_data, dr->dt.dl.dr_data, bonuslen);
 	} else if (zfs_refcount_count(&db->db_holds) > db->db_dirtycnt) {
-		arc_buf_t *buf = dbuf_alloc_arcbuf_from_arcbuf(db, db->db_buf);
-		dr->dt.dl.dr_data = buf;
-		bcopy(db->db.db_data, buf->b_data, arc_buf_size(buf));
+		dnode_t *dn = DB_DNODE(db);
+		int size = arc_buf_size(db->db_buf);
+		arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+		spa_t *spa = db->db_objset->os_spa;
+		enum zio_compress compress_type =
+		    arc_get_compression(db->db_buf);
+		uint8_t complevel = arc_get_complevel(db->db_buf);
+
+		if (arc_is_encrypted(db->db_buf)) {
+			boolean_t byteorder;
+			uint8_t salt[ZIO_DATA_SALT_LEN];
+			uint8_t iv[ZIO_DATA_IV_LEN];
+			uint8_t mac[ZIO_DATA_MAC_LEN];
+
+			arc_get_raw_params(db->db_buf, &byteorder, salt,
+			    iv, mac);
+			dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db,
+			    dmu_objset_id(dn->dn_objset), byteorder, salt, iv,
+			    mac, dn->dn_type, size, arc_buf_lsize(db->db_buf),
+			    compress_type, complevel);
+		} else if (compress_type != ZIO_COMPRESS_OFF) {
+			ASSERT3U(type, ==, ARC_BUFC_DATA);
+			dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db,
+			    size, arc_buf_lsize(db->db_buf), compress_type,
+			    complevel);
+		} else {
+			dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size);
+		}
+		bcopy(db->db.db_data, dr->dt.dl.dr_data->b_data, size);
 	} else {
 		db->db_buf = NULL;
 		dbuf_clear_data(db);
@@ -3280,10 +3270,30 @@ noinline static void
 dbuf_hold_copy(dnode_t *dn, dmu_buf_impl_t *db)
 {
 	dbuf_dirty_record_t *dr = db->db_data_pending;
-	arc_buf_t *newdata, *data = dr->dt.dl.dr_data;
+	arc_buf_t *data = dr->dt.dl.dr_data;
+	enum zio_compress compress_type = arc_get_compression(data);
+	uint8_t complevel = arc_get_complevel(data);
+
+	if (arc_is_encrypted(data)) {
+		boolean_t byteorder;
+		uint8_t salt[ZIO_DATA_SALT_LEN];
+		uint8_t iv[ZIO_DATA_IV_LEN];
+		uint8_t mac[ZIO_DATA_MAC_LEN];
+
+		arc_get_raw_params(data, &byteorder, salt, iv, mac);
+		dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db,
+		    dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac,
+		    dn->dn_type, arc_buf_size(data), arc_buf_lsize(data),
+		    compress_type, complevel));
+	} else if (compress_type != ZIO_COMPRESS_OFF) {
+		dbuf_set_data(db, arc_alloc_compressed_buf(
+		    dn->dn_objset->os_spa, db, arc_buf_size(data),
+		    arc_buf_lsize(data), compress_type, complevel));
+	} else {
+		dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db,
+		    DBUF_GET_BUFC_TYPE(db), db->db.db_size));
+	}
 
-	newdata = dbuf_alloc_arcbuf_from_arcbuf(db, data);
-	dbuf_set_data(db, newdata);
 	rw_enter(&db->db_rwlock, RW_WRITER);
 	bcopy(data->b_data, db->db.db_data, arc_buf_size(data));
 	rw_exit(&db->db_rwlock);
@@ -4067,8 +4077,31 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
 		 * objects only modified in the syncing context (e.g.
 		 * DNONE_DNODE blocks).
 		 */
-		*datap = dbuf_alloc_arcbuf_from_arcbuf(db, db->db_buf);
-		bcopy(db->db.db_data, (*datap)->b_data, arc_buf_size(*datap));
+		int psize = arc_buf_size(*datap);
+		int lsize = arc_buf_lsize(*datap);
+		arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+		enum zio_compress compress_type = arc_get_compression(*datap);
+		uint8_t complevel = arc_get_complevel(*datap);
+
+		if (arc_is_encrypted(*datap)) {
+			boolean_t byteorder;
+			uint8_t salt[ZIO_DATA_SALT_LEN];
+			uint8_t iv[ZIO_DATA_IV_LEN];
+			uint8_t mac[ZIO_DATA_MAC_LEN];
+
+			arc_get_raw_params(*datap, &byteorder, salt, iv, mac);
+			*datap = arc_alloc_raw_buf(os->os_spa, db,
+			    dmu_objset_id(os), byteorder, salt, iv, mac,
+			    dn->dn_type, psize, lsize, compress_type,
+			    complevel);
+		} else if (compress_type != ZIO_COMPRESS_OFF) {
+			ASSERT3U(type, ==, ARC_BUFC_DATA);
+			*datap = arc_alloc_compressed_buf(os->os_spa, db,
+			    psize, lsize, compress_type, complevel);
+		} else {
+			*datap = arc_alloc_buf(os->os_spa, db, type, psize);
+		}
+		bcopy(db->db.db_data, (*datap)->b_data, psize);
 	}
 	db->db_data_pending = dr;
 

module/zfs/ddt.c

@@ -503,7 +503,7 @@ ddt_get_dedup_histogram(spa_t *spa, ddt_histogram_t *ddh)
 {
 	for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) {
 		ddt_t *ddt = spa->spa_ddt[c];
-		for (enum ddt_type type = 0; type < DDT_TYPES; type++) {
+		for (enum ddt_type type = 0; type < DDT_TYPES && ddt; type++) {
 			for (enum ddt_class class = 0; class < DDT_CLASSES;
 			    class++) {
 				ddt_histogram_add(ddh,

module/zfs/dnode.c

@@ -129,6 +129,7 @@ dnode_cons(void *arg, void *unused, int kmflag)
 	zfs_refcount_create(&dn->dn_tx_holds);
 	list_link_init(&dn->dn_link);
 
+	bzero(&dn->dn_next_type[0], sizeof (dn->dn_next_type));
 	bzero(&dn->dn_next_nblkptr[0], sizeof (dn->dn_next_nblkptr));
 	bzero(&dn->dn_next_nlevels[0], sizeof (dn->dn_next_nlevels));
 	bzero(&dn->dn_next_indblkshift[0], sizeof (dn->dn_next_indblkshift));

module/zfs/dsl_deadlist.c

@@ -909,15 +909,16 @@ dsl_deadlist_move_bpobj(dsl_deadlist_t *dl, bpobj_t *bpo, uint64_t mintxg,
 }
 
 typedef struct livelist_entry {
-	const blkptr_t *le_bp;
+	blkptr_t le_bp;
+	uint32_t le_refcnt;
 	avl_node_t le_node;
 } livelist_entry_t;
 
 static int
 livelist_compare(const void *larg, const void *rarg)
 {
-	const blkptr_t *l = ((livelist_entry_t *)larg)->le_bp;
-	const blkptr_t *r = ((livelist_entry_t *)rarg)->le_bp;
+	const blkptr_t *l = &((livelist_entry_t *)larg)->le_bp;
+	const blkptr_t *r = &((livelist_entry_t *)rarg)->le_bp;
 
 	/* Sort them according to dva[0] */
 	uint64_t l_dva0_vdev = DVA_GET_VDEV(&l->blk_dva[0]);
@@ -944,6 +945,11 @@ struct livelist_iter_arg {
  * Expects an AVL tree which is incrementally filled will FREE blkptrs
  * and used to match up ALLOC/FREE pairs. ALLOC'd blkptrs without a
  * corresponding FREE are stored in the supplied bplist.
+ *
+ * Note that multiple FREE and ALLOC entries for the same blkptr may
+ * be encountered when dedup is involved. For this reason we keep a
+ * refcount for all the FREE entries of each blkptr and ensure that
+ * each of those FREE entries has a corresponding ALLOC preceding it.
  */
 static int
 dsl_livelist_iterate(void *arg, const blkptr_t *bp, boolean_t bp_freed,
@@ -957,23 +963,47 @@ dsl_livelist_iterate(void *arg, const blkptr_t *bp, boolean_t bp_freed,
 
 	if ((t != NULL) && (zthr_has_waiters(t) || zthr_iscancelled(t)))
 		return (SET_ERROR(EINTR));
+
+	livelist_entry_t node;
+	node.le_bp = *bp;
+	livelist_entry_t *found = avl_find(avl, &node, NULL);
 	if (bp_freed) {
-		livelist_entry_t *node = kmem_alloc(sizeof (livelist_entry_t),
-		    KM_SLEEP);
-		blkptr_t *temp_bp = kmem_alloc(sizeof (blkptr_t), KM_SLEEP);
-		*temp_bp = *bp;
-		node->le_bp = temp_bp;
-		avl_add(avl, node);
-	} else {
-		livelist_entry_t node;
-		node.le_bp = bp;
-		livelist_entry_t *found = avl_find(avl, &node, NULL);
-		if (found != NULL) {
-			avl_remove(avl, found);
-			kmem_free((blkptr_t *)found->le_bp, sizeof (blkptr_t));
-			kmem_free(found, sizeof (livelist_entry_t));
+		if (found == NULL) {
+			/* first free entry for this blkptr */
+			livelist_entry_t *e =
+			    kmem_alloc(sizeof (livelist_entry_t), KM_SLEEP);
+			e->le_bp = *bp;
+			e->le_refcnt = 1;
+			avl_add(avl, e);
 		} else {
+			/* dedup block free */
+			ASSERT(BP_GET_DEDUP(bp));
+			ASSERT3U(BP_GET_CHECKSUM(bp), ==,
+			    BP_GET_CHECKSUM(&found->le_bp));
+			ASSERT3U(found->le_refcnt + 1, >, found->le_refcnt);
+			found->le_refcnt++;
+		}
+	} else {
+		if (found == NULL) {
+			/* block is currently marked as allocated */
 			bplist_append(to_free, bp);
+		} else {
+			/* alloc matches a free entry */
+			ASSERT3U(found->le_refcnt, !=, 0);
+			found->le_refcnt--;
+			if (found->le_refcnt == 0) {
+				/* all tracked free pairs have been matched */
+				avl_remove(avl, found);
+				kmem_free(found, sizeof (livelist_entry_t));
+			} else {
+				/*
+				 * This is definitely a deduped blkptr so
+				 * let's validate it.
+				 */
+				ASSERT(BP_GET_DEDUP(bp));
+				ASSERT3U(BP_GET_CHECKSUM(bp), ==,
+				    BP_GET_CHECKSUM(&found->le_bp));
+			}
 		}
 	}
 	return (0);
@@ -999,6 +1029,7 @@ dsl_process_sub_livelist(bpobj_t *bpobj, bplist_t *to_free, zthr_t *t,
 	};
 	int err = bpobj_iterate_nofree(bpobj, dsl_livelist_iterate, &arg, size);
 
+	VERIFY0(avl_numnodes(&avl));
 	avl_destroy(&avl);
 	return (err);
 }

module/zfs/fm.c

@@ -586,25 +586,29 @@ zfs_zevent_minor_to_state(minor_t minor, zfs_zevent_t **ze)
 	return (0);
 }
 
-int
+zfs_file_t *
 zfs_zevent_fd_hold(int fd, minor_t *minorp, zfs_zevent_t **ze)
 {
-	int error;
+	zfs_file_t *fp = zfs_file_get(fd);
+	if (fp == NULL)
+		return (NULL);
 
-	error = zfsdev_getminor(fd, minorp);
+	int error = zfsdev_getminor(fp, minorp);
 	if (error == 0)
 		error = zfs_zevent_minor_to_state(*minorp, ze);
 
-	if (error)
-		zfs_zevent_fd_rele(fd);
+	if (error) {
+		zfs_zevent_fd_rele(fp);
+		fp = NULL;
+	}
 
-	return (error);
+	return (fp);
 }
 
 void
-zfs_zevent_fd_rele(int fd)
+zfs_zevent_fd_rele(zfs_file_t *fp)
 {
-	zfs_file_put(fd);
+	zfs_file_put(fp);
 }
 
 /*

module/zfs/spa_misc.c

@@ -347,10 +347,10 @@ int spa_asize_inflation = 24;
 
 /*
  * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in
- * the pool to be consumed.  This ensures that we don't run the pool
- * completely out of space, due to unaccounted changes (e.g. to the MOS).
- * It also limits the worst-case time to allocate space.  If we have
- * less than this amount of free space, most ZPL operations (e.g. write,
+ * the pool to be consumed (bounded by spa_max_slop).  This ensures that we
+ * don't run the pool completely out of space, due to unaccounted changes (e.g.
+ * to the MOS).  It also limits the worst-case time to allocate space.  If we
+ * have less than this amount of free space, most ZPL operations (e.g.  write,
  * create) will return ENOSPC.
  *
  * Certain operations (e.g. file removal, most administrative actions) can
@@ -374,10 +374,15 @@ int spa_asize_inflation = 24;
  * 3.2%, in an effort to have it be at least spa_min_slop (128MB),
  * but we never allow it to be more than half the pool size.
  *
+ * Further, on very large pools, the slop space will be smaller than
+ * 3.2%, to avoid reserving much more space than we actually need; bounded
+ * by spa_max_slop (128GB).
+ *
  * See also the comments in zfs_space_check_t.
  */
 int spa_slop_shift = 5;
-uint64_t spa_min_slop = 128 * 1024 * 1024;
+uint64_t spa_min_slop = 128ULL * 1024 * 1024;
+uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024;
 int spa_allocators = 4;
 
 
@@ -1775,17 +1780,38 @@ spa_get_worst_case_asize(spa_t *spa, uint64_t lsize)
 }
 
 /*
- * Return the amount of slop space in bytes.  It is 1/32 of the pool (3.2%),
- * or at least 128MB, unless that would cause it to be more than half the
- * pool size.
+ * Return the amount of slop space in bytes.  It is typically 1/32 of the pool
+ * (3.2%).  On very small pools, it may be slightly larger than this.  On very
+ * large pools, it will be capped to the value of spa_max_slop.
  *
- * See the comment above spa_slop_shift for details.
+ * See the comment above spa_slop_shift for more details.
  */
 uint64_t
 spa_get_slop_space(spa_t *spa)
 {
-	uint64_t space = spa_get_dspace(spa);
-	return (MAX(space >> spa_slop_shift, MIN(space >> 1, spa_min_slop)));
+	uint64_t space = 0;
+	uint64_t slop = 0;
+
+	/*
+	 * Make sure spa_dedup_dspace has been set.
+	 */
+	if (spa->spa_dedup_dspace == ~0ULL)
+		spa_update_dspace(spa);
+
+	/*
+	 * spa_get_dspace() includes the space only logically "used" by
+	 * deduplicated data, so since it's not useful to reserve more
+	 * space with more deduplicated data, we subtract that out here.
+	 */
+	space = spa_get_dspace(spa) - spa->spa_dedup_dspace;
+	slop = MIN(space >> spa_slop_shift, spa_max_slop);
+
+	/*
+	 * Slop space should be at least spa_min_slop, but no more than half
+	 * the entire pool.
+	 */
+	slop = MAX(slop, MIN(space >> 1, spa_min_slop));
+	return (slop);
 }
 
 uint64_t
@@ -1823,7 +1849,14 @@ spa_update_dspace(spa_t *spa)
 		spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
 		vdev_t *vd =
 		    vdev_lookup_top(spa, spa->spa_vdev_removal->svr_vdev_id);
-		if (vd->vdev_mg->mg_class == spa_normal_class(spa)) {
+		/*
+		 * If the stars align, we can wind up here after
+		 * vdev_remove_complete() has cleared vd->vdev_mg but before
+		 * spa->spa_vdev_removal gets cleared, so we must check before
+		 * we dereference.
+		 */
+		if (vd->vdev_mg &&
+		    vd->vdev_mg->mg_class == spa_normal_class(spa)) {
 			spa->spa_dspace -= spa_deflate(spa) ?
 			    vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
 		}

module/zfs/zfs_ioctl.c

@@ -4834,8 +4834,8 @@ zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
 	*errors = fnvlist_alloc();
 	off = 0;
 
-	if ((error = zfs_file_get(input_fd, &input_fp)))
-		return (error);
+	if ((input_fp = zfs_file_get(input_fd)) == NULL)
+		return (SET_ERROR(EBADF));
 
 	noff = off = zfs_file_off(input_fp);
 	error = dmu_recv_begin(tofs, tosnap, begin_record, force,
@@ -5115,7 +5115,7 @@ zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
 		nvlist_free(inheritprops);
 	}
 out:
-	zfs_file_put(input_fd);
+	zfs_file_put(input_fp);
 	nvlist_free(origrecvd);
 	nvlist_free(origprops);
 
@@ -5445,8 +5445,8 @@ zfs_ioc_send(zfs_cmd_t *zc)
 		zfs_file_t *fp;
 		dmu_send_outparams_t out = {0};
 
-		if ((error = zfs_file_get(zc->zc_cookie, &fp)))
-			return (error);
+		if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
+			return (SET_ERROR(EBADF));
 
 		off = zfs_file_off(fp);
 		out.dso_outfunc = dump_bytes;
@@ -5456,7 +5456,7 @@ zfs_ioc_send(zfs_cmd_t *zc)
 		    zc->zc_fromobj, embedok, large_block_ok, compressok,
 		    rawok, savedok, zc->zc_cookie, &off, &out);
 
-		zfs_file_put(zc->zc_cookie);
+		zfs_file_put(fp);
 	}
 	return (error);
 }
@@ -6020,25 +6020,24 @@ zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
 {
 	char *snap_name;
 	char *hold_name;
-	int error;
 	minor_t minor;
 
-	error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
-	if (error != 0)
-		return (error);
+	zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
+	if (fp == NULL)
+		return (SET_ERROR(EBADF));
 
 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
 	    (u_longlong_t)ddi_get_lbolt64());
 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
 
-	error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
+	int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
 	    hold_name);
 	if (error == 0)
 		(void) strlcpy(zc->zc_value, snap_name,
 		    sizeof (zc->zc_value));
 	kmem_strfree(snap_name);
 	kmem_strfree(hold_name);
-	zfs_onexit_fd_rele(zc->zc_cleanup_fd);
+	zfs_onexit_fd_rele(fp);
 	return (error);
 }
 
@@ -6058,13 +6057,13 @@ zfs_ioc_diff(zfs_cmd_t *zc)
 	offset_t off;
 	int error;
 
-	if ((error = zfs_file_get(zc->zc_cookie, &fp)))
-		return (error);
+	if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
+		return (SET_ERROR(EBADF));
 
 	off = zfs_file_off(fp);
 	error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
 
-	zfs_file_put(zc->zc_cookie);
+	zfs_file_put(fp);
 
 	return (error);
 }
@@ -6100,6 +6099,7 @@ zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
 	int cleanup_fd = -1;
 	int error;
 	minor_t minor = 0;
+	zfs_file_t *fp = NULL;
 
 	holds = fnvlist_lookup_nvlist(args, "holds");
 
@@ -6117,14 +6117,16 @@ zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
 	}
 
 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
-		error = zfs_onexit_fd_hold(cleanup_fd, &minor);
-		if (error != 0)
-			return (SET_ERROR(error));
+		fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
+		if (fp == NULL)
+			return (SET_ERROR(EBADF));
 	}
 
 	error = dsl_dataset_user_hold(holds, minor, errlist);
-	if (minor != 0)
-		zfs_onexit_fd_rele(cleanup_fd);
+	if (fp != NULL) {
+		ASSERT3U(minor, !=, 0);
+		zfs_onexit_fd_rele(fp);
+	}
 	return (SET_ERROR(error));
 }
 
@@ -6187,9 +6189,9 @@ zfs_ioc_events_next(zfs_cmd_t *zc)
 	uint64_t dropped = 0;
 	int error;
 
-	error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
-	if (error != 0)
-		return (error);
+	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
+	if (fp == NULL)
+		return (SET_ERROR(EBADF));
 
 	do {
 		error = zfs_zevent_next(ze, &event,
@@ -6211,7 +6213,7 @@ zfs_ioc_events_next(zfs_cmd_t *zc)
 			break;
 	} while (1);
 
-	zfs_zevent_fd_rele(zc->zc_cleanup_fd);
+	zfs_zevent_fd_rele(fp);
 
 	return (error);
 }
@@ -6243,12 +6245,12 @@ zfs_ioc_events_seek(zfs_cmd_t *zc)
 	minor_t minor;
 	int error;
 
-	error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
-	if (error != 0)
-		return (error);
+	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
+	if (fp == NULL)
+		return (SET_ERROR(EBADF));
 
 	error = zfs_zevent_seek(ze, zc->zc_guid);
-	zfs_zevent_fd_rele(zc->zc_cleanup_fd);
+	zfs_zevent_fd_rele(fp);
 
 	return (error);
 }
@@ -6432,8 +6434,8 @@ zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
 
 	(void) nvlist_lookup_string(innvl, "redactbook", &redactbook);
 
-	if ((error = zfs_file_get(fd, &fp)))
-		return (error);
+	if ((fp = zfs_file_get(fd)) == NULL)
+		return (SET_ERROR(EBADF));
 
 	off = zfs_file_off(fp);
 
@@ -6445,7 +6447,7 @@ zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
 	    compressok, rawok, savedok, resumeobj, resumeoff,
 	    redactbook, fd, &off, &out);
 
-	zfs_file_put(fd);
+	zfs_file_put(fp);
 	return (error);
 }
 
@@ -7318,17 +7320,12 @@ pool_status_check(const char *name, zfs_ioc_namecheck_t type,
 }
 
 int
-zfsdev_getminor(int fd, minor_t *minorp)
+zfsdev_getminor(zfs_file_t *fp, minor_t *minorp)
 {
 	zfsdev_state_t *zs, *fpd;
-	zfs_file_t *fp;
-	int rc;
 
 	ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
 
-	if ((rc = zfs_file_get(fd, &fp)))
-		return (rc);
-
 	fpd = zfs_file_private(fp);
 	if (fpd == NULL)
 		return (SET_ERROR(EBADF));

module/zfs/zfs_log.c

@@ -126,9 +126,11 @@ zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
 
 	/* Now pack the attributes up in a single uint64_t */
 	attrs = (uint64_t *)bitmap;
-	crtime = attrs + 1;
-	scanstamp = (caddr_t)(crtime + 2);
 	*attrs = 0;
+	crtime = attrs + 1;
+	bzero(crtime, 2 * sizeof (uint64_t));
+	scanstamp = (caddr_t)(crtime + 2);
+	bzero(scanstamp, AV_SCANSTAMP_SZ);
 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
 		*attrs |= (xoap->xoa_readonly == 0) ? 0 :
 		    XAT0_READONLY;

module/zfs/zfs_onexit.c

@@ -107,30 +107,33 @@ zfs_onexit_destroy(zfs_onexit_t *zo)
  * of this function must call zfs_onexit_fd_rele() when they're finished
  * using the minor number.
  */
-int
+zfs_file_t *
 zfs_onexit_fd_hold(int fd, minor_t *minorp)
 {
 	zfs_onexit_t *zo = NULL;
-	int error;
 
-	error = zfsdev_getminor(fd, minorp);
+	zfs_file_t *fp = zfs_file_get(fd);
+	if (fp == NULL)
+		return (NULL);
+
+	int error = zfsdev_getminor(fp, minorp);
 	if (error) {
-		zfs_onexit_fd_rele(fd);
-		return (error);
+		zfs_onexit_fd_rele(fp);
+		return (NULL);
 	}
 
 	zo = zfsdev_get_state(*minorp, ZST_ONEXIT);
 	if (zo == NULL) {
-		zfs_onexit_fd_rele(fd);
-		return (SET_ERROR(EBADF));
+		zfs_onexit_fd_rele(fp);
+		return (NULL);
 	}
-	return (0);
+	return (fp);
 }
 
 void
-zfs_onexit_fd_rele(int fd)
+zfs_onexit_fd_rele(zfs_file_t *fp)
 {
-	zfs_file_put(fd);
+	zfs_file_put(fp);
 }
 
 static int

module/zfs/zil.c

@@ -1617,7 +1617,7 @@ zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb)
 	lr_t *lrcb, *lrc;
 	lr_write_t *lrwb, *lrw;
 	char *lr_buf;
-	uint64_t dlen, dnow, lwb_sp, reclen, txg, max_log_data;
+	uint64_t dlen, dnow, dpad, lwb_sp, reclen, txg, max_log_data;
 
 	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
 	ASSERT3P(lwb, !=, NULL);
@@ -1651,8 +1651,9 @@ zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb)
 	if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) {
 		dlen = P2ROUNDUP_TYPED(
 		    lrw->lr_length, sizeof (uint64_t), uint64_t);
+		dpad = dlen - lrw->lr_length;
 	} else {
-		dlen = 0;
+		dlen = dpad = 0;
 	}
 	reclen = lrc->lrc_reclen;
 	zilog->zl_cur_used += (reclen + dlen);
@@ -1746,6 +1747,9 @@ cont:
 			error = zilog->zl_get_data(itx->itx_private,
 			    itx->itx_gen, lrwb, dbuf, lwb,
 			    lwb->lwb_write_zio);
+			if (dbuf != NULL && error == 0 && dnow == dlen)
+				/* Zero any padding bytes in the last block. */
+				bzero((char *)dbuf + lrwb->lr_length, dpad);
 
 			if (error == EIO) {
 				txg_wait_synced(zilog->zl_dmu_pool, txg);
@@ -1783,18 +1787,19 @@ cont:
 }
 
 itx_t *
-zil_itx_create(uint64_t txtype, size_t lrsize)
+zil_itx_create(uint64_t txtype, size_t olrsize)
 {
-	size_t itxsize;
+	size_t itxsize, lrsize;
 	itx_t *itx;
 
-	lrsize = P2ROUNDUP_TYPED(lrsize, sizeof (uint64_t), size_t);
+	lrsize = P2ROUNDUP_TYPED(olrsize, sizeof (uint64_t), size_t);
 	itxsize = offsetof(itx_t, itx_lr) + lrsize;
 
 	itx = zio_data_buf_alloc(itxsize);
 	itx->itx_lr.lrc_txtype = txtype;
 	itx->itx_lr.lrc_reclen = lrsize;
 	itx->itx_lr.lrc_seq = 0;	/* defensive */
+	bzero((char *)&itx->itx_lr + olrsize, lrsize - olrsize);
 	itx->itx_sync = B_TRUE;		/* default is synchronous */
 	itx->itx_callback = NULL;
 	itx->itx_callback_data = NULL;

tests/runfiles/common.run

@@ -165,8 +165,8 @@ tags = ['functional', 'cli_root', 'zfs_create']
 
 [tests/functional/cli_root/zfs_destroy]
 tests = ['zfs_clone_livelist_condense_and_disable',
-    'zfs_clone_livelist_condense_races', 'zfs_destroy_001_pos',
-    'zfs_destroy_002_pos', 'zfs_destroy_003_pos',
+    'zfs_clone_livelist_condense_races', 'zfs_clone_livelist_dedup',
+    'zfs_destroy_001_pos', 'zfs_destroy_002_pos', 'zfs_destroy_003_pos',
     'zfs_destroy_004_pos', 'zfs_destroy_005_neg', 'zfs_destroy_006_neg',
     'zfs_destroy_007_neg', 'zfs_destroy_008_pos', 'zfs_destroy_009_pos',
     'zfs_destroy_010_pos', 'zfs_destroy_011_pos', 'zfs_destroy_012_pos',

tests/zfs-tests/tests/functional/cli_root/zfs_destroy/zfs_clone_livelist_dedup.ksh

@@ -0,0 +1,88 @@
+#!/bin/ksh -p
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+
+#
+# Copyright (c) 2021 by Delphix. All rights reserved.
+#
+
+# DESCRIPTION
+# Verify zfs destroy test for clones with livelists that contain
+# dedup blocks. This test is a baseline regression test created
+# to ensure that past bugs that we've encountered between dedup
+# and the livelist logic don't resurface.
+
+# STRATEGY
+# 1. Create a clone from a test filesystem and enable dedup.
+# 2. Write some data and create a livelist.
+# 3. Copy the data within the clone to create dedup blocks.
+# 4. Remove some of the dedup data to create multiple free
+#    entries for the same block pointers.
+# 5. Process all the livelist entries by destroying the clone.
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/cli_root/zfs_destroy/zfs_destroy_common.kshlib
+
+function cleanup
+{
+	log_must zfs destroy -Rf $TESTPOOL/$TESTFS1
+	# Reset the minimum percent shared to 75
+	set_tunable32 LIVELIST_MIN_PERCENT_SHARED $ORIGINAL_MIN_SHARED
+}
+
+function test_dedup
+{
+	# Set a small percent shared threshold so the livelist is not disabled
+	set_tunable32 LIVELIST_MIN_PERCENT_SHARED 10
+	clone_dataset $TESTFS1 snap $TESTCLONE
+
+	# Enable dedup
+	log_must zfs set dedup=on $TESTPOOL/$TESTCLONE
+
+	# Create some data to be deduped
+	log_must dd if=/dev/urandom of="/$TESTPOOL/$TESTCLONE/data" bs=512 count=10k
+
+	# Create dedup blocks
+	# Note: We sync before and after so all dedup blocks belong to the
+	#       same TXG, otherwise they won't look identical to the livelist
+	#       iterator due to their logical birth TXG being different.
+	log_must zpool sync $TESTPOOL
+	log_must cp /$TESTPOOL/$TESTCLONE/data /$TESTPOOL/$TESTCLONE/data-dup-0
+	log_must cp /$TESTPOOL/$TESTCLONE/data /$TESTPOOL/$TESTCLONE/data-dup-1
+	log_must cp /$TESTPOOL/$TESTCLONE/data /$TESTPOOL/$TESTCLONE/data-dup-2
+	log_must cp /$TESTPOOL/$TESTCLONE/data /$TESTPOOL/$TESTCLONE/data-dup-3
+	log_must zpool sync $TESTPOOL
+	check_livelist_exists $TESTCLONE
+
+	# Introduce "double frees"
+	#   We want to introduce consecutive FREEs of the same block as this
+	#   was what triggered past panics.
+	# Note: Similarly to the previouys step we sync before and after our
+	#       our deletions so all the entries end up in the same TXG.
+	log_must zpool sync $TESTPOOL
+	log_must rm /$TESTPOOL/$TESTCLONE/data-dup-2
+	log_must rm /$TESTPOOL/$TESTCLONE/data-dup-3
+	log_must zpool sync $TESTPOOL
+	check_livelist_exists $TESTCLONE
+
+	log_must zfs destroy $TESTPOOL/$TESTCLONE
+	check_livelist_gone
+}
+
+ORIGINAL_MIN_SHARED=$(get_tunable LIVELIST_MIN_PERCENT_SHARED)
+
+log_onexit cleanup
+log_must zfs create $TESTPOOL/$TESTFS1
+log_must mkfile 5m /$TESTPOOL/$TESTFS1/atestfile
+log_must zfs snapshot $TESTPOOL/$TESTFS1@snap
+test_dedup
+
+log_pass "Clone's livelist processes dedup blocks as expected."