zfs-2.2.6 patchset

META

@@ -1,10 +1,10 @@
 Meta:          1
 Name:          zfs
 Branch:        1.0
-Version:       2.2.5
+Version:       2.2.6
 Release:       1
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS
-Linux-Maximum: 6.9
+Linux-Maximum: 6.10
 Linux-Minimum: 3.10

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

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.

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)
+	])
+])

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])
+	])
+])

config/kernel.m4

@@ -166,9 +166,12 @@ 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 "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_SRC_CPU_HAS_FEATURE
@@ -317,9 +320,12 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_WRITEPAGE_T
 	ZFS_AC_KERNEL_RECLAIMED
 	ZFS_AC_KERNEL_REGISTER_SYSCTL_TABLE
+	ZFS_AC_KERNEL_REGISTER_SYSCTL_SZ
+	ZFS_AC_KERNEL_PROC_HANDLER_CTL_TABLE_CONST
 	ZFS_AC_KERNEL_COPY_SPLICE_READ
 	ZFS_AC_KERNEL_SYNC_BDEV
 	ZFS_AC_KERNEL_MM_PAGE_SIZE
+	ZFS_AC_KERNEL_MM_PAGE_MAPPING
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_CPU_HAS_FEATURE

contrib/bash_completion.d/.gitignore

@@ -1 +1,2 @@
 /zfs
+/zpool

contrib/bash_completion.d/Makefile.am

@@ -1,5 +1,9 @@
-nodist_bashcompletion_DATA  = %D%/zfs
-SUBSTFILES                 += $(nodist_bashcompletion_DATA)
+nodist_bashcompletion_DATA  = %D%/zfs %D%/zpool
+COMPLETION_FILES            = %D%/zfs
+SUBSTFILES                 += $(COMPLETION_FILES)
 
-SHELLCHECKSCRIPTS   += $(nodist_bashcompletion_DATA)
-$(call SHELLCHECK_OPTS,$(nodist_bashcompletion_DATA)): SHELLCHECK_SHELL = bash
+SHELLCHECKSCRIPTS   += $(COMPLETION_FILES)
+$(call SHELLCHECK_OPTS,$(COMPLETION_FILES)): SHELLCHECK_SHELL = bash
+
+%D%/zpool: %D%/zfs
+	$(LN_S) zfs $@

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
@@ -68,39 +73,43 @@ 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 */
 
 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
+#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

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 */

man/man4/zfs.4

@@ -121,20 +121,26 @@ Controls whether buffers present on special vdevs are eligible for caching
 into L2ARC.
 If set to 1, exclude dbufs on special vdevs from being cached to L2ARC.
 .
-.It Sy l2arc_mfuonly Ns = Ns Sy 0 Ns | Ns 1 Pq  int
+.It Sy l2arc_mfuonly Ns = Ns Sy 0 Ns | Ns 1 Ns | Ns 2 Pq int
 Controls whether only MFU metadata and data are cached from ARC into L2ARC.
 This may be desired to avoid wasting space on L2ARC when reading/writing large
 amounts of data that are not expected to be accessed more than once.
 .Pp
-The default is off,
+The default is 0,
 meaning both MRU and MFU data and metadata are cached.
-When turning off this feature, some MRU buffers will still be present
-in ARC and eventually cached on L2ARC.
+When turning off this feature (setting it to 0), some MRU buffers will
+still be present in ARC and eventually cached on L2ARC.
 .No If Sy l2arc_noprefetch Ns = Ns Sy 0 ,
 some prefetched buffers will be cached to L2ARC, and those might later
 transition to MRU, in which case the
 .Sy l2arc_mru_asize No arcstat will not be Sy 0 .
 .Pp
+Setting it to 1 means to L2 cache only MFU data and metadata.
+.Pp
+Setting it to 2 means to L2 cache all metadata (MRU+MFU) but
+only MFU data (ie: MRU data are not cached). This can be the right setting
+to cache as much metadata as possible even when having high data turnover.
+.Pp
 Regardless of
 .Sy l2arc_noprefetch ,
 some MFU buffers might be evicted from ARC,

module/Kbuild.in

@@ -16,8 +16,8 @@ src = @abs_srcdir@
 obj = @abs_builddir@
 else
 zfs_include = $(srctree)/include/zfs
-icp_include = $(srctree)/$(src)/icp/include
-zstd_include = $(srctree)/$(src)/zstd/include
+icp_include = $(src)/icp/include
+zstd_include = $(src)/zstd/include
 ZFS_MODULE_CFLAGS += -include $(zfs_include)/zfs_config.h
 endif