git subrepo pull --force sys/contrib/subrepo-openzfs

subrepo:
  subdir:   "sys/contrib/subrepo-openzfs"
  merged:   "fb4d45df99fa"
upstream:
  origin:   "https://github.com/CTSRD-CHERI/zfs.git"
  branch:   "cheri-hybrid"
  commit:   "fb4d45df99fa"
git-subrepo:
  version:  "0.4.3"
  origin:   "???"
  commit:   "???"

sys/contrib/subrepo-openzfs/.gitrepo

@@ -6,7 +6,7 @@
 [subrepo]
 	remote = https://github.com/CTSRD-CHERI/zfs.git
 	branch = cheri-hybrid
-	commit = 64e6835be758b36ef129a1ee548a501e93a27dd0
-	parent = b421b291504ade31a8ab0e240af5ca04af401e47
+	commit = fb4d45df99faf34df8c8d4afd25441e71ef0e4e2
+	parent = 179fd0a21cd2ae87e75dc0b4f00b00bf9c6ab0f9
 	method = merge
 	cmdver = 0.4.3

sys/contrib/subrepo-openzfs/cmd/zdb/zdb.c

@@ -2360,7 +2360,7 @@ static void
 snprintf_zstd_header(spa_t *spa, char *blkbuf, size_t buflen,
     const blkptr_t *bp)
 {
-	abd_t *pabd;
+	static abd_t *pabd = NULL;
 	void *buf;
 	zio_t *zio;
 	zfs_zstdhdr_t zstd_hdr;
@@ -2391,7 +2391,8 @@ snprintf_zstd_header(spa_t *spa, char *blkbuf, size_t buflen,
 		return;
 	}
 
-	pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
+	if (!pabd)
+		pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
 	zio = zio_root(spa, NULL, NULL, 0);
 
 	/* Decrypt but don't decompress so we can read the compression header */
@@ -8487,11 +8488,45 @@ zdb_parse_block_sizes(char *sizes, uint64_t *lsize, uint64_t *psize)
 #define	ZIO_COMPRESS_MASK(alg)	(1ULL << (ZIO_COMPRESS_##alg))
 
 static boolean_t
+try_decompress_block(abd_t *pabd, uint64_t lsize, uint64_t psize,
+    int flags, int cfunc, void *lbuf, void *lbuf2)
+{
+	if (flags & ZDB_FLAG_VERBOSE) {
+		(void) fprintf(stderr,
+		    "Trying %05llx -> %05llx (%s)\n",
+		    (u_longlong_t)psize,
+		    (u_longlong_t)lsize,
+		    zio_compress_table[cfunc].ci_name);
+	}
+
+	/*
+	 * We set lbuf to all zeros and lbuf2 to all
+	 * ones, then decompress to both buffers and
+	 * compare their contents. This way we can
+	 * know if decompression filled exactly to
+	 * lsize or if it left some bytes unwritten.
+	 */
+
+	memset(lbuf, 0x00, lsize);
+	memset(lbuf2, 0xff, lsize);
+
+	if (zio_decompress_data(cfunc, pabd,
+	    lbuf, psize, lsize, NULL) == 0 &&
+	    zio_decompress_data(cfunc, pabd,
+	    lbuf2, psize, lsize, NULL) == 0 &&
+	    memcmp(lbuf, lbuf2, lsize) == 0)
+		return (B_TRUE);
+	return (B_FALSE);
+}
+
+static uint64_t
 zdb_decompress_block(abd_t *pabd, void *buf, void *lbuf, uint64_t lsize,
     uint64_t psize, int flags)
 {
 	(void) buf;
-	boolean_t exceeded = B_FALSE;
+	uint64_t orig_lsize = lsize;
+	boolean_t tryzle = ((getenv("ZDB_NO_ZLE") == NULL));
+	boolean_t found = B_FALSE;
 	/*
 	 * We don't know how the data was compressed, so just try
 	 * every decompress function at every inflated blocksize.
@@ -8502,10 +8537,18 @@ zdb_decompress_block(abd_t *pabd, void *buf, void *lbuf, uint64_t lsize,
 	uint64_t maxlsize = SPA_MAXBLOCKSIZE;
 	uint64_t mask = ZIO_COMPRESS_MASK(ON) | ZIO_COMPRESS_MASK(OFF) |
 	    ZIO_COMPRESS_MASK(INHERIT) | ZIO_COMPRESS_MASK(EMPTY) |
-	    (getenv("ZDB_NO_ZLE") ? ZIO_COMPRESS_MASK(ZLE) : 0);
+	    ZIO_COMPRESS_MASK(ZLE);
 	*cfuncp++ = ZIO_COMPRESS_LZ4;
 	*cfuncp++ = ZIO_COMPRESS_LZJB;
 	mask |= ZIO_COMPRESS_MASK(LZ4) | ZIO_COMPRESS_MASK(LZJB);
+	/*
+	 * Every gzip level has the same decompressor, no need to
+	 * run it 9 times per bruteforce attempt.
+	 */
+	mask |= ZIO_COMPRESS_MASK(GZIP_2) | ZIO_COMPRESS_MASK(GZIP_3);
+	mask |= ZIO_COMPRESS_MASK(GZIP_4) | ZIO_COMPRESS_MASK(GZIP_5);
+	mask |= ZIO_COMPRESS_MASK(GZIP_6) | ZIO_COMPRESS_MASK(GZIP_7);
+	mask |= ZIO_COMPRESS_MASK(GZIP_8) | ZIO_COMPRESS_MASK(GZIP_9);
 	for (int c = 0; c < ZIO_COMPRESS_FUNCTIONS; c++)
 		if (((1ULL << c) & mask) == 0)
 			*cfuncp++ = c;
@@ -8521,49 +8564,38 @@ zdb_decompress_block(abd_t *pabd, void *buf, void *lbuf, uint64_t lsize,
 		lsize += SPA_MINBLOCKSIZE;
 	else
 		maxlsize = lsize;
+
 	for (; lsize <= maxlsize; lsize += SPA_MINBLOCKSIZE) {
 		for (cfuncp = cfuncs; *cfuncp; cfuncp++) {
-			if (flags & ZDB_FLAG_VERBOSE) {
-				(void) fprintf(stderr,
-				    "Trying %05llx -> %05llx (%s)\n",
-				    (u_longlong_t)psize,
-				    (u_longlong_t)lsize,
-				    zio_compress_table[*cfuncp].\
-				    ci_name);
-			}
-
-			/*
-			 * We set lbuf to all zeros and lbuf2 to all
-			 * ones, then decompress to both buffers and
-			 * compare their contents. This way we can
-			 * know if decompression filled exactly to
-			 * lsize or if it left some bytes unwritten.
-			 */
-			memset(lbuf, 0x00, lsize);
-			memset(lbuf2, 0xff, lsize);
-
-			if (zio_decompress_data(*cfuncp, pabd,
-			    lbuf, psize, lsize, NULL) == 0 &&
-			    zio_decompress_data(*cfuncp, pabd,
-			    lbuf2, psize, lsize, NULL) == 0 &&
-			    memcmp(lbuf, lbuf2, lsize) == 0)
+			if (try_decompress_block(pabd, lsize, psize, flags,
+			    *cfuncp, lbuf, lbuf2)) {
+				found = B_TRUE;
 				break;
+			}
 		}
 		if (*cfuncp != 0)
 			break;
 	}
+	if (!found && tryzle) {
+		for (lsize = orig_lsize; lsize <= maxlsize;
+		    lsize += SPA_MINBLOCKSIZE) {
+			if (try_decompress_block(pabd, lsize, psize, flags,
+			    ZIO_COMPRESS_ZLE, lbuf, lbuf2)) {
+				*cfuncp = ZIO_COMPRESS_ZLE;
+				found = B_TRUE;
+				break;
+			}
+		}
+	}
 	umem_free(lbuf2, SPA_MAXBLOCKSIZE);
 
-	if (lsize > maxlsize) {
-		exceeded = B_TRUE;
-	}
 	if (*cfuncp == ZIO_COMPRESS_ZLE) {
 		printf("\nZLE decompression was selected. If you "
 		    "suspect the results are wrong,\ntry avoiding ZLE "
 		    "by setting and exporting ZDB_NO_ZLE=\"true\"\n");
 	}
 
-	return (exceeded);
+	return (lsize > maxlsize ? -1 : lsize);
 }
 
 /*
@@ -8742,9 +8774,9 @@ zdb_read_block(char *thing, spa_t *spa)
 	uint64_t orig_lsize = lsize;
 	buf = lbuf;
 	if (flags & ZDB_FLAG_DECOMPRESS) {
-		boolean_t failed = zdb_decompress_block(pabd, buf, lbuf,
+		lsize = zdb_decompress_block(pabd, buf, lbuf,
 		    lsize, psize, flags);
-		if (failed) {
+		if (lsize == -1) {
 			(void) printf("Decompress of %s failed\n", thing);
 			goto out;
 		}
@@ -8765,11 +8797,11 @@ zdb_read_block(char *thing, spa_t *spa)
 			abd_return_buf_copy(pabd, buf, lsize);
 			borrowed = B_FALSE;
 			buf = lbuf;
-			boolean_t failed = zdb_decompress_block(pabd, buf,
+			lsize = zdb_decompress_block(pabd, buf,
 			    lbuf, lsize, psize, flags);
 			b = (const blkptr_t *)(void *)
 			    ((uintptr_t)buf + (uintptr_t)blkptr_offset);
-			if (failed || zfs_blkptr_verify(spa, b,
+			if (lsize == -1 || zfs_blkptr_verify(spa, b,
 			    BLK_CONFIG_NEEDED, BLK_VERIFY_LOG) == B_FALSE) {
 				printf("invalid block pointer at this DVA\n");
 				goto out;

sys/contrib/subrepo-openzfs/cmd/zfs/zfs_main.c

@@ -7234,7 +7234,8 @@ share_mount(int op, int argc, char **argv)
 		pthread_mutex_init(&share_mount_state.sm_lock, NULL);
 
 		/* For a 'zfs share -a' operation start with a clean slate. */
-		zfs_truncate_shares(NULL);
+		if (op == OP_SHARE)
+			zfs_truncate_shares(NULL);
 
 		/*
 		 * libshare isn't mt-safe, so only do the operation in parallel

sys/contrib/subrepo-openzfs/config/rpm.am

@@ -83,6 +83,11 @@ srpm-common:
 		rpm-local || exit 1; \
 	LANG=C $(RPMBUILD) \
 		--define "_tmppath $$rpmbuild/TMP" \
+		--define "_builddir $$rpmbuild/BUILD" \
+		--define "_rpmdir $$rpmbuild/RPMS" \
+		--define "_srcrpmdir $$rpmbuild/SRPMS" \
+		--define "_specdir $$rpmbuild/SPECS" \
+		--define "_sourcedir $$rpmbuild/SOURCES" \
 		--define "_topdir $$rpmbuild" \
 		$(def) -bs $$rpmbuild/SPECS/$$rpmspec || exit 1; \
 	cp $$rpmbuild/SRPMS/$$rpmpkg . || exit 1; \
@@ -99,6 +104,11 @@ rpm-common:
 		rpm-local || exit 1; \
 	LANG=C ${RPMBUILD} \
 		--define "_tmppath $$rpmbuild/TMP" \
+		--define "_builddir $$rpmbuild/BUILD" \
+		--define "_rpmdir $$rpmbuild/RPMS" \
+		--define "_srcrpmdir $$rpmbuild/SRPMS" \
+		--define "_specdir $$rpmbuild/SPECS" \
+		--define "_sourcedir $$rpmbuild/SOURCES" \
 		--define "_topdir $$rpmbuild" \
 		$(def) --rebuild $$rpmpkg || exit 1; \
 	cp $$rpmbuild/RPMS/*/* . || exit 1; \

sys/contrib/subrepo-openzfs/include/os/freebsd/spl/sys/vnode.h

@@ -66,6 +66,7 @@ enum symfollow { NO_FOLLOW = NOFOLLOW };
 #include <sys/syscallsubr.h>
 #include <sys/vm.h>
 #include <vm/vm_object.h>
+#include <vm/vnode_pager.h>
 
 typedef	struct vop_vector	vnodeops_t;
 #define	VOP_FID		VOP_VPTOFH
@@ -100,11 +101,11 @@ vn_flush_cached_data(vnode_t *vp, boolean_t sync)
 #else
 	if (vp->v_object->flags & OBJ_MIGHTBEDIRTY) {
 #endif
-		int flags = sync ? OBJPC_SYNC : 0;
 		vn_lock(vp, LK_SHARED | LK_RETRY);
-		zfs_vmobject_wlock(vp->v_object);
-		vm_object_page_clean(vp->v_object, 0, 0, flags);
-		zfs_vmobject_wunlock(vp->v_object);
+		if (sync)
+			vnode_pager_clean_sync(vp);
+		else
+			vnode_pager_clean_async(vp);
 		VOP_UNLOCK1(vp);
 	}
 }

sys/contrib/subrepo-openzfs/include/sys/spa_impl.h

@@ -278,6 +278,7 @@ struct spa {
 
 	spa_aux_vdev_t	spa_spares;		/* hot spares */
 	spa_aux_vdev_t	spa_l2cache;		/* L2ARC cache devices */
+	boolean_t	spa_aux_sync_uber;	/* need to sync aux uber */
 	nvlist_t	*spa_label_features;	/* Features for reading MOS */
 	uint64_t	spa_config_object;	/* MOS object for pool config */
 	uint64_t	spa_config_generation;	/* config generation number */

sys/contrib/subrepo-openzfs/lib/libzfs/libzfs_import.c

@@ -291,10 +291,8 @@ zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
 
 	*inuse = B_FALSE;
 
-	if (zpool_read_label(fd, &config, NULL) != 0) {
-		(void) no_memory(hdl);
+	if (zpool_read_label(fd, &config, NULL) != 0)
 		return (-1);
-	}
 
 	if (config == NULL)
 		return (0);

sys/contrib/subrepo-openzfs/lib/libzutil/zutil_import.c

@@ -1056,10 +1056,21 @@ zpool_read_label(int fd, nvlist_t **config, int *num_labels)
 				case EINVAL:
 					break;
 				case EINPROGRESS:
-					// This shouldn't be possible to
-					// encounter, die if we do.
+					/*
+					 * This shouldn't be possible to
+					 * encounter, die if we do.
+					 */
 					ASSERT(B_FALSE);
 					zfs_fallthrough;
+				case EREMOTEIO:
+					/*
+					 * May be returned by an NVMe device
+					 * which is visible in /dev/ but due
+					 * to a low-level format change, or
+					 * other error, needs to be rescanned.
+					 * Try the slow method.
+					 */
+					zfs_fallthrough;
 				case EOPNOTSUPP:
 				case ENOSYS:
 					do_slow = B_TRUE;
@@ -1210,13 +1221,26 @@ label_paths(libpc_handle_t *hdl, nvlist_t *label, const char **path,
 	nvlist_t *nvroot;
 	uint64_t pool_guid;
 	uint64_t vdev_guid;
+	uint64_t state;
 
 	*path = NULL;
 	*devid = NULL;
+	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &vdev_guid) != 0)
+		return (ENOENT);
+
+	/*
+	 * In case of spare or l2cache, we directly return path/devid from the
+	 * label.
+	 */
+	if (!(nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, &state)) &&
+	    (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE)) {
+		(void) nvlist_lookup_string(label, ZPOOL_CONFIG_PATH, path);
+		(void) nvlist_lookup_string(label, ZPOOL_CONFIG_DEVID, devid);
+		return (0);
+	}
 
 	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvroot) ||
-	    nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &pool_guid) ||
-	    nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &vdev_guid))
+	    nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &pool_guid))
 		return (ENOENT);
 
 	return (label_paths_impl(hdl, nvroot, pool_guid, vdev_guid, path,

sys/contrib/subrepo-openzfs/module/os/linux/spl/spl-condvar.c

@@ -20,7 +20,7 @@
  *  You should have received a copy of the GNU General Public License along
  *  with the SPL.  If not, see <http://www.gnu.org/licenses/>.
  *
- *  Solaris Porting Layer (SPL) Credential Implementation.
+ *  Solaris Porting Layer (SPL) Condition Variables Implementation.
  */
 
 #include <sys/condvar.h>

sys/contrib/subrepo-openzfs/module/os/linux/spl/spl-kmem-cache.c

@@ -91,7 +91,8 @@ MODULE_PARM_DESC(spl_kmem_cache_max_size, "Maximum size of slab in MB");
  * of 16K was determined to be optimal for architectures using 4K pages and
  * to also work well on architecutres using larger 64K page sizes.
  */
-static unsigned int spl_kmem_cache_slab_limit = 16384;
+static unsigned int spl_kmem_cache_slab_limit =
+    SPL_MAX_KMEM_ORDER_NR_PAGES * PAGE_SIZE;
 module_param(spl_kmem_cache_slab_limit, uint, 0644);
 MODULE_PARM_DESC(spl_kmem_cache_slab_limit,
 	"Objects less than N bytes use the Linux slab");
@@ -783,7 +784,7 @@ spl_kmem_cache_create(const char *name, size_t size, size_t align,
 	} else {
 		unsigned long slabflags = 0;
 
-		if (size > (SPL_MAX_KMEM_ORDER_NR_PAGES * PAGE_SIZE))
+		if (size > spl_kmem_cache_slab_limit)
 			goto out;
 
 #if defined(SLAB_USERCOPY)

sys/contrib/subrepo-openzfs/module/os/linux/zfs/zfs_vnops_os.c

@@ -1856,7 +1856,7 @@ zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr, zidmap_t *mnt_ns)
 {
 	struct inode	*ip;
 	zfsvfs_t	*zfsvfs = ZTOZSB(zp);
-	objset_t	*os = zfsvfs->z_os;
+	objset_t	*os;
 	zilog_t		*zilog;
 	dmu_tx_t	*tx;
 	vattr_t		oldva;
@@ -1888,6 +1888,7 @@ zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr, zidmap_t *mnt_ns)
 	if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
 		return (err);
 	ip = ZTOI(zp);
+	os = zfsvfs->z_os;
 
 	/*
 	 * If this is a xvattr_t, then get a pointer to the structure of
@@ -2438,9 +2439,8 @@ zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr, zidmap_t *mnt_ns)
 
 	if ((mask & ATTR_ATIME) || zp->z_atime_dirty) {
 		zp->z_atime_dirty = B_FALSE;
-		inode_timespec_t tmp_atime;
+		inode_timespec_t tmp_atime = zpl_inode_get_atime(ip);
 		ZFS_TIME_ENCODE(&tmp_atime, atime);
-		zpl_inode_set_atime_to_ts(ZTOI(zp), tmp_atime);
 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
 		    &atime, sizeof (atime));
 	}