zfs: merge openzfs/zfs@d2ccc2155

Notable upstream pull request merges:
 #16431 244ea5c48 Add missing kstats to dataset kstats

Obtained from:	OpenZFS
OpenZFS commit: d2ccc21552

sys/contrib/openzfs/config/kernel-blk-queue.m4

@@ -25,6 +25,8 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_PLUG], [
 dnl #
 dnl # 2.6.32 - 4.11: statically allocated bdi in request_queue
 dnl # 4.12: dynamically allocated bdi in request_queue
+dnl # 6.11: bdi no longer available through request_queue, so get it from
+dnl #       the gendisk attached to the queue
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_BDI], [
 	ZFS_LINUX_TEST_SRC([blk_queue_bdi], [
@@ -47,6 +49,30 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_BDI], [
 	])
 ])
 
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_DISK_BDI], [
+	ZFS_LINUX_TEST_SRC([blk_queue_disk_bdi], [
+		#include <linux/blkdev.h>
+		#include <linux/backing-dev.h>
+	], [
+		struct request_queue q;
+		struct gendisk disk;
+		struct backing_dev_info bdi __attribute__ ((unused));
+		q.disk = &disk;
+		q.disk->bdi = &bdi;
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_DISK_BDI], [
+	AC_MSG_CHECKING([whether backing_dev_info is available through queue gendisk])
+	ZFS_LINUX_TEST_RESULT([blk_queue_disk_bdi], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BLK_QUEUE_DISK_BDI, 1,
+		    [backing_dev_info is available through queue gendisk])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 dnl #
 dnl # 5.9: added blk_queue_update_readahead(),
 dnl # 5.15: renamed to disk_update_readahead()
@@ -407,6 +433,7 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_MQ], [
 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_DISK_BDI
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_UPDATE_READAHEAD
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_DISCARD
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE_SECURE_ERASE
@@ -421,6 +448,7 @@ 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_DISK_BDI
 	ZFS_AC_KERNEL_BLK_QUEUE_UPDATE_READAHEAD
 	ZFS_AC_KERNEL_BLK_QUEUE_DISCARD
 	ZFS_AC_KERNEL_BLK_QUEUE_SECURE_ERASE

sys/contrib/openzfs/config/kernel-make-request-fn.m4

@@ -58,6 +58,13 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_MAKE_REQUEST_FN], [
 		disk = blk_alloc_disk(lim, NUMA_NO_NODE);
 	])
 
+	ZFS_LINUX_TEST_SRC([blkdev_queue_limits_features], [
+		#include <linux/blkdev.h>
+	],[
+		struct queue_limits *lim = NULL;
+		lim->features = 0;
+	])
+
 	ZFS_LINUX_TEST_SRC([blk_cleanup_disk], [
 		#include <linux/blkdev.h>
 	],[
@@ -114,6 +121,20 @@ AC_DEFUN([ZFS_AC_KERNEL_MAKE_REQUEST_FN], [
 			AC_MSG_RESULT(yes)
 			AC_DEFINE([HAVE_BLK_ALLOC_DISK_2ARG], 1, [blk_alloc_disk() exists and takes 2 args])
 
+			dnl #
+			dnl # Linux 6.11 API change:
+			dnl # struct queue_limits gains a 'features' field,
+			dnl # used to set flushing options
+			dnl #
+			AC_MSG_CHECKING([whether struct queue_limits has a features field])
+			ZFS_LINUX_TEST_RESULT([blkdev_queue_limits_features], [
+				AC_MSG_RESULT(yes)
+				AC_DEFINE([HAVE_BLKDEV_QUEUE_LIMITS_FEATURES], 1,
+				    [struct queue_limits has a features field])
+			], [
+				AC_MSG_RESULT(no)
+			])
+
 			dnl #
 			dnl # 5.20 API change,
 			dnl # Removed blk_cleanup_disk(), put_disk() should be used.

sys/contrib/openzfs/config/kernel-mm-page-size.m4

@@ -1,17 +0,0 @@
-AC_DEFUN([ZFS_AC_KERNEL_SRC_MM_PAGE_SIZE], [
-	ZFS_LINUX_TEST_SRC([page_size], [
-		#include <linux/mm.h>
-	],[
-		unsigned long s;
-		s = page_size(NULL);
-	])
-])
-AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_SIZE], [
-	AC_MSG_CHECKING([whether page_size() is available])
-	ZFS_LINUX_TEST_RESULT([page_size], [
-		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_MM_PAGE_SIZE, 1, [page_size() is available])
-	],[
-		AC_MSG_RESULT(no)
-	])
-])

sys/contrib/openzfs/config/kernel-mm-pagemap.m4

@@ -0,0 +1,36 @@
+AC_DEFUN([ZFS_AC_KERNEL_SRC_MM_PAGE_SIZE], [
+	ZFS_LINUX_TEST_SRC([page_size], [
+		#include <linux/mm.h>
+	],[
+		unsigned long s;
+		s = page_size(NULL);
+	])
+])
+AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_SIZE], [
+	AC_MSG_CHECKING([whether page_size() is available])
+	ZFS_LINUX_TEST_RESULT([page_size], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_MM_PAGE_SIZE, 1, [page_size() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_MM_PAGE_MAPPING], [
+	ZFS_LINUX_TEST_SRC([page_mapping], [
+		#include <linux/pagemap.h>
+	],[
+		struct page *p = NULL;
+		struct address_space *m = page_mapping(NULL);
+	])
+])
+AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_MAPPING], [
+	AC_MSG_CHECKING([whether page_mapping() is available])
+	ZFS_LINUX_TEST_RESULT([page_mapping], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_MM_PAGE_MAPPING, 1, [page_mapping() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

sys/contrib/openzfs/config/kernel-register_sysctl_table.m4

@@ -25,3 +25,62 @@ AC_DEFUN([ZFS_AC_KERNEL_REGISTER_SYSCTL_TABLE], [
 		AC_MSG_RESULT([no])
 	])
 ])
+
+dnl #
+dnl # Linux 6.11 register_sysctl() enforces that sysctl tables no longer
+dnl # supply a sentinel end-of-table element. 6.6 introduces
+dnl # register_sysctl_sz() to enable callers to choose, so we use it if
+dnl # available for backward compatibility.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_SZ], [
+	ZFS_LINUX_TEST_SRC([has_register_sysctl_sz], [
+		#include <linux/sysctl.h>
+	],[
+		struct ctl_table test_table[] __attribute__((unused)) = {0};
+		register_sysctl_sz("", test_table, 0);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_REGISTER_SYSCTL_SZ], [
+	AC_MSG_CHECKING([whether register_sysctl_sz exists])
+	ZFS_LINUX_TEST_RESULT([has_register_sysctl_sz], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_REGISTER_SYSCTL_SZ, 1,
+			[register_sysctl_sz exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])
+
+dnl #
+dnl # Linux 6.11 makes const the ctl_table arg of proc_handler
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_PROC_HANDLER_CTL_TABLE_CONST], [
+	ZFS_LINUX_TEST_SRC([has_proc_handler_ctl_table_const], [
+		#include <linux/sysctl.h>
+
+		static int test_handler(
+		    const struct ctl_table *ctl __attribute((unused)),
+		    int write __attribute((unused)),
+		    void *buffer __attribute((unused)),
+		    size_t *lenp __attribute((unused)),
+		    loff_t *ppos __attribute((unused)))
+		{
+			return (0);
+		}
+	], [
+		proc_handler *ph __attribute((unused)) =
+		    &test_handler;
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_PROC_HANDLER_CTL_TABLE_CONST], [
+	AC_MSG_CHECKING([whether proc_handler ctl_table arg is const])
+	ZFS_LINUX_TEST_RESULT([has_proc_handler_ctl_table_const], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_PROC_HANDLER_CTL_TABLE_CONST, 1,
+		    [proc_handler ctl_table arg is const])
+	], [
+		AC_MSG_RESULT([no])
+	])
+])

sys/contrib/openzfs/config/kernel.m4

@@ -167,9 +167,12 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_WRITEPAGE_T
 	ZFS_AC_KERNEL_SRC_RECLAIMED
 	ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_TABLE
+	ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_SZ
+	ZFS_AC_KERNEL_SRC_PROC_HANDLER_CTL_TABLE_CONST
 	ZFS_AC_KERNEL_SRC_COPY_SPLICE_READ
 	ZFS_AC_KERNEL_SRC_SYNC_BDEV
 	ZFS_AC_KERNEL_SRC_MM_PAGE_SIZE
+	ZFS_AC_KERNEL_SRC_MM_PAGE_MAPPING
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_SRC_CPU_HAS_FEATURE
@@ -319,9 +322,12 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_WRITEPAGE_T
 	ZFS_AC_KERNEL_RECLAIMED
 	ZFS_AC_KERNEL_REGISTER_SYSCTL_TABLE
+	ZFS_AC_KERNEL_REGISTER_SYSCTL_SZ
+	ZFS_AC_KERNEL_PROC_HANDLER_CTL_TABLE_CONST
 	ZFS_AC_KERNEL_COPY_SPLICE_READ
 	ZFS_AC_KERNEL_SYNC_BDEV
 	ZFS_AC_KERNEL_MM_PAGE_SIZE
+	ZFS_AC_KERNEL_MM_PAGE_MAPPING
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_CPU_HAS_FEATURE

sys/contrib/openzfs/include/os/linux/kernel/linux/blkdev_compat.h

@@ -57,6 +57,11 @@ blk_queue_flag_clear(unsigned int flag, struct request_queue *q)
 #endif
 
 /*
+ * 6.11 API
+ * Setting the flush flags directly is no longer possible; flush flags are set
+ * on the queue_limits structure and passed to blk_disk_alloc(). In this case
+ * we remove this function entirely.
+ *
  * 4.7 API,
  * The blk_queue_write_cache() interface has replaced blk_queue_flush()
  * interface.  However, the new interface is GPL-only thus we implement
@@ -68,31 +73,33 @@ blk_queue_flag_clear(unsigned int flag, struct request_queue *q)
  * new one is GPL-only.   Thus if the GPL-only version is detected we
  * implement our own trivial helper.
  */
+#if !defined(HAVE_BLK_ALLOC_DISK_2ARG) || \
+	!defined(HAVE_BLKDEV_QUEUE_LIMITS_FEATURES)
 static inline void
-blk_queue_set_write_cache(struct request_queue *q, bool wc, bool fua)
+blk_queue_set_write_cache(struct request_queue *q, bool on)
 {
 #if defined(HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY)
-	if (wc)
+	if (on) {
 		blk_queue_flag_set(QUEUE_FLAG_WC, q);
-	else
-		blk_queue_flag_clear(QUEUE_FLAG_WC, q);
-	if (fua)
 		blk_queue_flag_set(QUEUE_FLAG_FUA, q);
-	else
+	} else {
+		blk_queue_flag_clear(QUEUE_FLAG_WC, q);
 		blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
+	}
 #elif defined(HAVE_BLK_QUEUE_WRITE_CACHE)
-	blk_queue_write_cache(q, wc, fua);
+	blk_queue_write_cache(q, on, on);
 #elif defined(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY)
-	if (wc)
-		q->flush_flags |= REQ_FLUSH;
-	if (fua)
-		q->flush_flags |= REQ_FUA;
+	if (on)
+		q->flush_flags |= REQ_FLUSH | REQ_FUA;
+	else
+		q->flush_flags &= ~(REQ_FLUSH | REQ_FUA);
 #elif defined(HAVE_BLK_QUEUE_FLUSH)
-	blk_queue_flush(q, (wc ? REQ_FLUSH : 0) | (fua ? REQ_FUA : 0));
+	blk_queue_flush(q, on ? (REQ_FLUSH | REQ_FUA) : 0);
 #else
 #error "Unsupported kernel"
 #endif
 }
+#endif /* !HAVE_BLK_ALLOC_DISK_2ARG || !HAVE_BLKDEV_QUEUE_LIMITS_FEATURES */
 
 /*
  * Detect if a device has a write cache. Used to set the intial value for the
@@ -126,8 +133,10 @@ 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
+#if defined(HAVE_BLK_QUEUE_BDI_DYNAMIC)
 	q->backing_dev_info->ra_pages = ra_pages;
+#elif defined(HAVE_BLK_QUEUE_DISK_BDI)
+	q->disk->bdi->ra_pages = ra_pages;
 #else
 	q->backing_dev_info.ra_pages = ra_pages;
 #endif

sys/contrib/openzfs/include/os/linux/kernel/linux/mm_compat.h

@@ -21,16 +21,23 @@
 
 /*
  * Copyright (c) 2023, 2024, Klara Inc.
+ * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
  */
 
 #ifndef _ZFS_MM_COMPAT_H
 #define	_ZFS_MM_COMPAT_H
 
 #include <linux/mm.h>
+#include <linux/pagemap.h>
 
 /* 5.4 introduced page_size(). Older kernels can use a trivial macro instead */
 #ifndef HAVE_MM_PAGE_SIZE
 #define	page_size(p) ((unsigned long)(PAGE_SIZE << compound_order(p)))
 #endif
 
+/* 6.11 removed page_mapping(). A simple wrapper around folio_mapping() works */
+#ifndef HAVE_MM_PAGE_MAPPING
+#define	page_mapping(p) folio_mapping(page_folio(p))
+#endif
+
 #endif /* _ZFS_MM_COMPAT_H */

sys/contrib/openzfs/module/os/linux/spl/spl-proc.c

@@ -22,6 +22,9 @@
  *
  *  Solaris Porting Layer (SPL) Proc Implementation.
  */
+/*
+ * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
+ */
 
 #include <sys/systeminfo.h>
 #include <sys/kstat.h>
@@ -43,6 +46,12 @@ typedef struct ctl_table __no_const spl_ctl_table;
 typedef struct ctl_table spl_ctl_table;
 #endif
 
+#ifdef HAVE_PROC_HANDLER_CTL_TABLE_CONST
+#define	CONST_CTL_TABLE		const struct ctl_table
+#else
+#define	CONST_CTL_TABLE		struct ctl_table
+#endif
+
 static unsigned long table_min = 0;
 static unsigned long table_max = ~0;
 
@@ -60,7 +69,7 @@ struct proc_dir_entry *proc_spl_kstat = NULL;
 
 #ifdef DEBUG_KMEM
 static int
-proc_domemused(struct ctl_table *table, int write,
+proc_domemused(CONST_CTL_TABLE *table, int write,
     void __user *buffer, size_t *lenp, loff_t *ppos)
 {
 	int rc = 0;
@@ -88,7 +97,7 @@ proc_domemused(struct ctl_table *table, int write,
 #endif /* DEBUG_KMEM */
 
 static int
-proc_doslab(struct ctl_table *table, int write,
+proc_doslab(CONST_CTL_TABLE *table, int write,
     void __user *buffer, size_t *lenp, loff_t *ppos)
 {
 	int rc = 0;
@@ -135,7 +144,7 @@ proc_doslab(struct ctl_table *table, int write,
 }
 
 static int
-proc_dohostid(struct ctl_table *table, int write,
+proc_dohostid(CONST_CTL_TABLE *table, int write,
     void __user *buffer, size_t *lenp, loff_t *ppos)
 {
 	char *end, str[32];
@@ -688,6 +697,37 @@ static void spl_proc_cleanup(void)
 	}
 }
 
+#ifndef HAVE_REGISTER_SYSCTL_TABLE
+
+/*
+ * Traditionally, struct ctl_table arrays have been terminated by an "empty"
+ * sentinel element (specifically, one with .procname == NULL).
+ *
+ * Linux 6.6 began migrating away from this, adding register_sysctl_sz() so
+ * that callers could provide the size directly, and redefining
+ * register_sysctl() to just call register_sysctl_sz() with the array size. It
+ * retained support for the terminating element so that existing callers would
+ * continue to work.
+ *
+ * Linux 6.11 removed support for the terminating element, instead interpreting
+ * it as a real malformed element, and rejecting it.
+ *
+ * In order to continue support older kernels, we retain the terminating
+ * sentinel element for our sysctl tables, but instead detect availability of
+ * register_sysctl_sz(). If it exists, we pass it the array size -1, stopping
+ * the kernel from trying to process the terminator. For pre-6.6 kernels that
+ * don't have register_sysctl_sz(), we just use register_sysctl(), which can
+ * handle the terminating element as it always has.
+ */
+#ifdef HAVE_REGISTER_SYSCTL_SZ
+#define	spl_proc_register_sysctl(p, t)	\
+	register_sysctl_sz(p, t, ARRAY_SIZE(t)-1)
+#else
+#define	spl_proc_register_sysctl(p, t)	\
+	register_sysctl(p, t)
+#endif
+#endif
+
 int
 spl_proc_init(void)
 {
@@ -698,16 +738,17 @@ spl_proc_init(void)
 	if (spl_header == NULL)
 		return (-EUNATCH);
 #else
-	spl_header = register_sysctl("kernel/spl", spl_table);
+	spl_header = spl_proc_register_sysctl("kernel/spl", spl_table);
 	if (spl_header == NULL)
 		return (-EUNATCH);
 
-	spl_kmem = register_sysctl("kernel/spl/kmem", spl_kmem_table);
+	spl_kmem = spl_proc_register_sysctl("kernel/spl/kmem", spl_kmem_table);
 	if (spl_kmem == NULL) {
 		rc = -EUNATCH;
 		goto out;
 	}
-	spl_kstat = register_sysctl("kernel/spl/kstat", spl_kstat_table);
+	spl_kstat = spl_proc_register_sysctl("kernel/spl/kstat",
+	    spl_kstat_table);
 	if (spl_kstat == NULL) {
 		rc = -EUNATCH;
 		goto out;

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

@@ -69,6 +69,7 @@
 #include <sys/zpl.h>
 #include <sys/zil.h>
 #include <sys/sa_impl.h>
+#include <linux/mm_compat.h>
 
 /*
  * Programming rules.
@@ -1788,24 +1789,36 @@ zfs_setattr_dir(znode_t *dzp)
 			    &gid, sizeof (gid));
 		}
 
-		if (zp->z_projid != dzp->z_projid) {
+
+		uint64_t projid = dzp->z_projid;
+		if (zp->z_projid != projid) {
 			if (!(zp->z_pflags & ZFS_PROJID)) {
-				zp->z_pflags |= ZFS_PROJID;
-				SA_ADD_BULK_ATTR(bulk, count,
-				    SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags,
-				    sizeof (zp->z_pflags));
+				err = sa_add_projid(zp->z_sa_hdl, tx, projid);
+				if (unlikely(err == EEXIST)) {
+					err = 0;
+				} else if (err != 0) {
+					goto sa_add_projid_err;
+				} else {
+					projid = ZFS_INVALID_PROJID;
+				}
 			}
 
-			zp->z_projid = dzp->z_projid;
-			SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs),
-			    NULL, &zp->z_projid, sizeof (zp->z_projid));
+			if (projid != ZFS_INVALID_PROJID) {
+				zp->z_projid = projid;
+				SA_ADD_BULK_ATTR(bulk, count,
+				    SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid,
+				    sizeof (zp->z_projid));
+			}
 		}
 
+sa_add_projid_err:
 		mutex_exit(&dzp->z_lock);
 
 		if (likely(count > 0)) {
 			err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
 			dmu_tx_commit(tx);
+		} else if (projid == ZFS_INVALID_PROJID) {
+			dmu_tx_commit(tx);
 		} else {
 			dmu_tx_abort(tx);
 		}

sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c

@@ -20,6 +20,7 @@
  */
 /*
  * Copyright (c) 2012, 2020 by Delphix. All rights reserved.
+ * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
  * Copyright (c) 2024, Klara, Inc.
  */
 
@@ -1089,11 +1090,42 @@ static const struct block_device_operations zvol_ops = {
 #endif
 };
 
+/*
+ * Since 6.9, Linux has been removing queue limit setters in favour of an
+ * initial queue_limits struct applied when the device is open. Since 6.11,
+ * queue_limits is being extended to allow more things to be applied when the
+ * device is open. Setters are also being removed for this.
+ *
+ * For OpenZFS, this means that depending on kernel version, some options may
+ * be set up before the device is open, and some applied to an open device
+ * (queue) after the fact.
+ *
+ * We manage this complexity by having our own limits struct,
+ * zvol_queue_limits_t, in which we carry any queue config that we're
+ * interested in setting. This structure is the same on all kernels.
+ *
+ * These limits are then applied to the queue at device open time by the most
+ * appropriate method for the kernel.
+ *
+ * zvol_queue_limits_convert() is used on 6.9+ (where the two-arg form of
+ * blk_alloc_disk() exists). This converts our limits struct to a proper Linux
+ * struct queue_limits, and passes it in. Any fields added in later kernels are
+ * (obviously) not set up here.
+ *
+ * zvol_queue_limits_apply() is called on all kernel versions after the queue
+ * is created, and applies any remaining config. Before 6.9 that will be
+ * everything, via setter methods. After 6.9 that will be whatever couldn't be
+ * put into struct queue_limits. (This implies that zvol_queue_limits_apply()
+ * will always be a no-op on the latest kernel we support).
+ */
 typedef struct zvol_queue_limits {
 	unsigned int	zql_max_hw_sectors;
 	unsigned short	zql_max_segments;
 	unsigned int	zql_max_segment_size;
 	unsigned int	zql_io_opt;
+	unsigned int	zql_physical_block_size;
+	unsigned int	zql_max_discard_sectors;
+	unsigned int	zql_discard_granularity;
 } zvol_queue_limits_t;
 
 static void
@@ -1162,6 +1194,11 @@ zvol_queue_limits_init(zvol_queue_limits_t *limits, zvol_state_t *zv,
 	}
 
 	limits->zql_io_opt = zv->zv_volblocksize;
+
+	limits->zql_physical_block_size = zv->zv_volblocksize;
+	limits->zql_max_discard_sectors =
+	    (zvol_max_discard_blocks * zv->zv_volblocksize) >> 9;
+	limits->zql_discard_granularity = zv->zv_volblocksize;
 }
 
 #ifdef HAVE_BLK_ALLOC_DISK_2ARG
@@ -1174,18 +1211,34 @@ zvol_queue_limits_convert(zvol_queue_limits_t *limits,
 	qlimits->max_segments = limits->zql_max_segments;
 	qlimits->max_segment_size = limits->zql_max_segment_size;
 	qlimits->io_opt = limits->zql_io_opt;
+	qlimits->physical_block_size = limits->zql_physical_block_size;
+	qlimits->max_discard_sectors = limits->zql_max_discard_sectors;
+	qlimits->discard_granularity = limits->zql_discard_granularity;
+#ifdef HAVE_BLKDEV_QUEUE_LIMITS_FEATURES
+	qlimits->features =
+	    BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_IO_STAT;
+#endif
 }
-#else
+#endif
+
 static void
 zvol_queue_limits_apply(zvol_queue_limits_t *limits,
     struct request_queue *queue)
 {
+#ifndef HAVE_BLK_ALLOC_DISK_2ARG
 	blk_queue_max_hw_sectors(queue, limits->zql_max_hw_sectors);
 	blk_queue_max_segments(queue, limits->zql_max_segments);
 	blk_queue_max_segment_size(queue, limits->zql_max_segment_size);
 	blk_queue_io_opt(queue, limits->zql_io_opt);
-}
+	blk_queue_physical_block_size(queue, limits->zql_physical_block_size);
+	blk_queue_max_discard_sectors(queue, limits->zql_max_discard_sectors);
+	blk_queue_discard_granularity(queue, limits->zql_discard_granularity);
 #endif
+#ifndef HAVE_BLKDEV_QUEUE_LIMITS_FEATURES
+	blk_queue_set_write_cache(queue, B_TRUE);
+	blk_queue_flag_set(QUEUE_FLAG_IO_STAT, queue);
+#endif
+}
 
 static int
 zvol_alloc_non_blk_mq(struct zvol_state_os *zso, zvol_queue_limits_t *limits)
@@ -1208,6 +1261,10 @@ zvol_alloc_non_blk_mq(struct zvol_state_os *zso, zvol_queue_limits_t *limits)
 		return (1);
 	}
 
+#ifndef HAVE_BLKDEV_QUEUE_LIMITS_FEATURES
+	blk_queue_set_write_cache(zso->zvo_queue, B_TRUE);
+#endif
+
 	zso->zvo_disk = disk;
 	zso->zvo_disk->minors = ZVOL_MINORS;
 	zso->zvo_queue = zso->zvo_disk->queue;
@@ -1223,7 +1280,6 @@ zvol_alloc_non_blk_mq(struct zvol_state_os *zso, zvol_queue_limits_t *limits)
 	}
 
 	zso->zvo_disk->queue = zso->zvo_queue;
-	zvol_queue_limits_apply(limits, zso->zvo_queue);
 #endif /* HAVE_BLK_ALLOC_DISK */
 #else
 	zso->zvo_queue = blk_generic_alloc_queue(zvol_request, NUMA_NO_NODE);
@@ -1237,8 +1293,10 @@ zvol_alloc_non_blk_mq(struct zvol_state_os *zso, zvol_queue_limits_t *limits)
 	}
 
 	zso->zvo_disk->queue = zso->zvo_queue;
-	zvol_queue_limits_apply(limits, zso->zvo_queue);
 #endif /* HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */
+
+	zvol_queue_limits_apply(limits, zso->zvo_queue);
+
 	return (0);
 
 }
@@ -1260,7 +1318,6 @@ zvol_alloc_blk_mq(zvol_state_t *zv, zvol_queue_limits_t *limits)
 		return (1);
 	}
 	zso->zvo_queue = zso->zvo_disk->queue;
-	zvol_queue_limits_apply(limits, zso->zvo_queue);
 	zso->zvo_disk->minors = ZVOL_MINORS;
 #elif defined(HAVE_BLK_ALLOC_DISK_2ARG)
 	struct queue_limits qlimits;
@@ -1291,10 +1348,11 @@ zvol_alloc_blk_mq(zvol_state_t *zv, zvol_queue_limits_t *limits)
 
 	/* Our queue is now created, assign it to our disk */
 	zso->zvo_disk->queue = zso->zvo_queue;
-	zvol_queue_limits_apply(limits, zso->zvo_queue);
+#endif
 
+	zvol_queue_limits_apply(limits, zso->zvo_queue);
 #endif
-#endif
+
 	return (0);
 }
 
@@ -1360,8 +1418,6 @@ zvol_alloc(dev_t dev, const char *name)
 	if (ret != 0)
 		goto out_kmem;
 
-	blk_queue_set_write_cache(zso->zvo_queue, B_TRUE, B_TRUE);
-
 	/* Limit read-ahead to a single page to prevent over-prefetching. */
 	blk_queue_set_read_ahead(zso->zvo_queue, 1);
 
@@ -1370,9 +1426,6 @@ zvol_alloc(dev_t dev, const char *name)
 		blk_queue_flag_set(QUEUE_FLAG_NOMERGES, zso->zvo_queue);
 	}
 
-	/* Enable /proc/diskstats */
-	blk_queue_flag_set(QUEUE_FLAG_IO_STAT, zso->zvo_queue);
-
 	zso->zvo_queue->queuedata = zv;
 	zso->zvo_dev = dev;
 	zv->zv_open_count = 0;
@@ -1639,14 +1692,6 @@ zvol_os_create_minor(const char *name)
 
 	set_capacity(zv->zv_zso->zvo_disk, zv->zv_volsize >> 9);
 
-
-
-	blk_queue_physical_block_size(zv->zv_zso->zvo_queue,
-	    zv->zv_volblocksize);
-	blk_queue_max_discard_sectors(zv->zv_zso->zvo_queue,
-	    (zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
-	blk_queue_discard_granularity(zv->zv_zso->zvo_queue,
-	    zv->zv_volblocksize);
 #ifdef QUEUE_FLAG_DISCARD
 	blk_queue_flag_set(QUEUE_FLAG_DISCARD, zv->zv_zso->zvo_queue);
 #endif

sys/contrib/openzfs/tests/zfs-tests/tests/functional/alloc_class/alloc_class.kshlib

@@ -55,7 +55,7 @@ function display_status
 	((ret |= $?))
 
 	typeset mntpnt=$(get_prop mountpoint $pool)
-	dd if=/dev/random of=$mntpnt/testfile.$$ &
+	dd if=/dev/urandom of=$mntpnt/testfile.$$ &
 	typeset pid=$!
 
 	zpool iostat -v 1 3 > /dev/null

sys/contrib/openzfs/tests/zfs-tests/tests/functional/block_cloning/block_cloning_rlimit_fsize.ksh

@@ -54,7 +54,7 @@ log_must truncate -s 1G $VDEV
 
 log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $VDEV
 
-log_must dd if=/dev/random of=/$TESTPOOL/file1 bs=1 count=1000
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=1 count=1000
 
 ulimit -f 2
 log_must clonefile -f /$TESTPOOL/file1 /$TESTPOOL/file2 0 0 all

sys/contrib/openzfs/tests/zfs-tests/tests/functional/fault/suspend_resume_single.ksh

@@ -42,7 +42,7 @@ log_onexit cleanup
 log_assert "ensure single-disk pool resumes properly after suspend and clear"
 
 # create a file, and take a checksum, so we can compare later
-log_must dd if=/dev/random of=$DATAFILE bs=128K count=1
+log_must dd if=/dev/urandom of=$DATAFILE bs=128K count=1
 typeset sum1=$(cat $DATAFILE | md5sum)
 
 # make a debug device that we can "unplug"

sys/contrib/openzfs/tests/zfs-tests/tests/functional/upgrade/upgrade_projectquota_001_pos.ksh

@@ -63,6 +63,7 @@ log_must mkfiles $TESTDIR/fs2/tf $((RANDOM % 100 + 1))
 log_must zfs create $TESTPOOL/fs3
 log_must mkdir $TESTDIR/fs3/dir
 log_must mkfiles $TESTDIR/fs3/tf $((RANDOM % 100 + 1))
+log_must set_xattr_stdin passwd $TESTDIR/fs3/dir < /etc/passwd
 
 # Make sure project quota is disabled
 zfs projectspace -o used $TESTPOOL | grep -q "USED" &&
@@ -109,9 +110,23 @@ log_must chattr -p 100 $TESTDIR/fs3/dir
 log_must sleep 5 # upgrade done in the background so let's wait for a while
 zfs projectspace -o used $TESTPOOL/fs3 | grep -q "USED" ||
 	log_fail "project quota should be enabled for $TESTPOOL/fs3"
+dirino=$(stat -c '%i' $TESTDIR/fs3/dir)
+log_must zdb -ddddd $TESTPOOL/fs3 $dirino
+xattrdirino=$(zdb -ddddd $TESTPOOL/fs3 $dirino |grep -w "xattr" |awk '{print $2}')
+echo "xattrdirino: $xattrdirino"
+expectedcnt=1
+echo "expectedcnt: $expectedcnt"
+if [ "$xattrdirino" != "" ]; then
+	expectedcnt=$(($expectedcnt + 1))
+	echo "expectedcnt: $expectedcnt"
+	log_must zdb -ddddd $TESTPOOL/fs3 $xattrdirino
+	xattrinocnt=$(zdb -ddddd $TESTPOOL/fs3 $xattrdirino |grep -w "(type:" |wc -l)
+	echo "xattrinocnt: $xattrinocnt"
+	expectedcnt=$(($expectedcnt + $xattrinocnt))
+	echo "expectedcnt: $expectedcnt"
+fi
 cnt=$(get_prop projectobjused@100 $TESTPOOL/fs3)
-# if 'xattr=on', then 'cnt = 2'
-[[ $cnt -ne 1 ]] && [[ $cnt -ne 2 ]] &&
+[[ $cnt -ne $expectedcnt ]] &&
 	log_fail "projectquota accounting failed $cnt"
 
 # All in all, after having been through this, the dataset for testpool

sys/modules/zfs/zfs_config.h

@@ -189,6 +189,9 @@
 /* blkdev_put() accepts void* as arg 2 */
 /* #undef HAVE_BLKDEV_PUT_HOLDER */
 
+/* struct queue_limits has a features field */
+/* #undef HAVE_BLKDEV_QUEUE_LIMITS_FEATURES */
+
 /* blkdev_reread_part() exists */
 /* #undef HAVE_BLKDEV_REREAD_PART */
 
@@ -228,6 +231,9 @@
 /* blk_queue_discard() is available */
 /* #undef HAVE_BLK_QUEUE_DISCARD */
 
+/* backing_dev_info is available through queue gendisk */
+/* #undef HAVE_BLK_QUEUE_DISK_BDI */
+
 /* blk_queue_flag_clear() exists */
 /* #undef HAVE_BLK_QUEUE_FLAG_CLEAR */
 
@@ -703,6 +709,9 @@
 /* Define to 1 if you have the 'mlockall' function. */
 #define HAVE_MLOCKALL 1
 
+/* page_mapping() is available */
+/* #undef HAVE_MM_PAGE_MAPPING */
+
 /* page_size() is available */
 /* #undef HAVE_MM_PAGE_SIZE */
 
@@ -748,6 +757,9 @@
 /* posix_acl_valid() wants user namespace */
 /* #undef HAVE_POSIX_ACL_VALID_WITH_NS */
 
+/* proc_handler ctl_table arg is const */
+/* #undef HAVE_PROC_HANDLER_CTL_TABLE_CONST */
+
 /* proc_ops structure exists */
 /* #undef HAVE_PROC_OPS_STRUCT */
 
@@ -772,6 +784,9 @@
 /* register_shrinker is vararg */
 /* #undef HAVE_REGISTER_SHRINKER_VARARG */
 
+/* register_sysctl_sz exists */
+/* #undef HAVE_REGISTER_SYSCTL_SZ */
+
 /* register_sysctl_table exists */
 /* #undef HAVE_REGISTER_SYSCTL_TABLE */
 
@@ -1200,7 +1215,7 @@
 /* #undef ZFS_IS_GPL_COMPATIBLE */
 
 /* Define the project alias string. */
-#define ZFS_META_ALIAS "zfs-2.2.99-623-FreeBSD_g9c56b8ec7"
+#define ZFS_META_ALIAS "zfs-2.2.99-634-FreeBSD_gd2ccc2155"
 
 /* Define the project author. */
 #define ZFS_META_AUTHOR "OpenZFS"
@@ -1230,7 +1245,7 @@
 #define ZFS_META_NAME "zfs"
 
 /* Define the project release. */
-#define ZFS_META_RELEASE "623-FreeBSD_g9c56b8ec7"
+#define ZFS_META_RELEASE "634-FreeBSD_gd2ccc2155"
 
 /* Define the project version. */
 #define ZFS_META_VERSION "2.2.99"

sys/modules/zfs/zfs_gitrev.h

@@ -1 +1 @@
-#define	ZFS_META_GITREV "zfs-2.2.99-623-g9c56b8ec7"
+#define	ZFS_META_GITREV "zfs-2.2.99-634-gd2ccc2155"