diff options
Diffstat (limited to 'sys/contrib/openzfs/module/os')
12 files changed, 479 insertions, 49 deletions
diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c index 393bfaa65ff5..ebc2c0eeb6d2 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c @@ -163,6 +163,13 @@ param_set_arc_int(SYSCTL_HANDLER_ARGS) return (0); } +static void +warn_deprecated_sysctl(const char *old, const char *new) +{ + printf("WARNING: sysctl vfs.zfs.%s is deprecated. Use vfs.zfs.%s instead.\n", + old, new); +} + int param_set_arc_max(SYSCTL_HANDLER_ARGS) { @@ -185,9 +192,17 @@ param_set_arc_max(SYSCTL_HANDLER_ARGS) if (val != 0) zfs_arc_max = arc_c_max; + if (arg2 != 0) + warn_deprecated_sysctl("arc_max", "arc.max"); + return (0); } +SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_max, + CTLTYPE_ULONG | CTLFLAG_RWTUN | CTLFLAG_MPSAFE, + NULL, 1, param_set_arc_max, "LU", + "Maximum ARC size in bytes (LEGACY)"); + int param_set_arc_min(SYSCTL_HANDLER_ARGS) { @@ -209,9 +224,17 @@ param_set_arc_min(SYSCTL_HANDLER_ARGS) if (val != 0) zfs_arc_min = arc_c_min; + if (arg2 != 0) + warn_deprecated_sysctl("arc_min", "arc.min"); + return (0); } +SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_min, + CTLTYPE_ULONG | CTLFLAG_RWTUN | CTLFLAG_MPSAFE, + NULL, 1, param_set_arc_min, "LU", + "Minimum ARC size in bytes (LEGACY)"); + extern uint_t zfs_arc_free_target; int @@ -232,9 +255,22 @@ param_set_arc_free_target(SYSCTL_HANDLER_ARGS) zfs_arc_free_target = val; + if (arg2 != 0) + warn_deprecated_sysctl("arc_free_target", "arc.free_target"); + return (0); } +/* + * NOTE: This sysctl is CTLFLAG_RW not CTLFLAG_RWTUN due to its dependency on + * pagedaemon initialization. + */ +SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_free_target, + CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, + NULL, 1, param_set_arc_free_target, "IU", + "Desired number of free pages below which ARC triggers reclaim" + " (LEGACY)"); + int param_set_arc_no_grow_shift(SYSCTL_HANDLER_ARGS) { @@ -250,9 +286,193 @@ param_set_arc_no_grow_shift(SYSCTL_HANDLER_ARGS) arc_no_grow_shift = val; + if (arg2 != 0) + warn_deprecated_sysctl("arc_no_grow_shift", "arc.no_grow_shift"); + return (0); } +SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_no_grow_shift, + CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_MPSAFE, + NULL, 1, param_set_arc_no_grow_shift, "I", + "log2(fraction of ARC which must be free to allow growing) (LEGACY)"); + +extern uint64_t l2arc_write_max; + +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_write_max, + CTLFLAG_RWTUN, &l2arc_write_max, 0, + "Max write bytes per interval (LEGACY)"); + +extern uint64_t l2arc_write_boost; + +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_write_boost, + CTLFLAG_RWTUN, &l2arc_write_boost, 0, + "Extra write bytes during device warmup (LEGACY)"); + +extern uint64_t l2arc_headroom; + +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_headroom, + CTLFLAG_RWTUN, &l2arc_headroom, 0, + "Number of max device writes to precache (LEGACY)"); + +extern uint64_t l2arc_headroom_boost; + +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_headroom_boost, + CTLFLAG_RWTUN, &l2arc_headroom_boost, 0, + "Compressed l2arc_headroom multiplier (LEGACY)"); + +extern uint64_t l2arc_feed_secs; + +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_feed_secs, + CTLFLAG_RWTUN, &l2arc_feed_secs, 0, + "Seconds between L2ARC writing (LEGACY)"); + +extern uint64_t l2arc_feed_min_ms; + +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_feed_min_ms, + CTLFLAG_RWTUN, &l2arc_feed_min_ms, 0, + "Min feed interval in milliseconds (LEGACY)"); + +extern int l2arc_noprefetch; + +SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_noprefetch, + CTLFLAG_RWTUN, &l2arc_noprefetch, 0, + "Skip caching prefetched buffers (LEGACY)"); + +extern int l2arc_feed_again; + +SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_feed_again, + CTLFLAG_RWTUN, &l2arc_feed_again, 0, + "Turbo L2ARC warmup (LEGACY)"); + +extern int l2arc_norw; + +SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_norw, + CTLFLAG_RWTUN, &l2arc_norw, 0, + "No reads during writes (LEGACY)"); + +static int +param_get_arc_state_size(SYSCTL_HANDLER_ARGS) +{ + arc_state_t *state = (arc_state_t *)arg1; + int64_t val; + + val = zfs_refcount_count(&state->arcs_size[ARC_BUFC_DATA]) + + zfs_refcount_count(&state->arcs_size[ARC_BUFC_METADATA]); + return (sysctl_handle_64(oidp, &val, 0, req)); +} + +extern arc_state_t ARC_anon; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, anon_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_anon, 0, param_get_arc_state_size, "Q", + "size of anonymous state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_metadata_esize, CTLFLAG_RD, + &ARC_anon.arcs_esize[ARC_BUFC_METADATA].rc_count, 0, + "size of evictable metadata in anonymous state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_data_esize, CTLFLAG_RD, + &ARC_anon.arcs_esize[ARC_BUFC_DATA].rc_count, 0, + "size of evictable data in anonymous state"); + +extern arc_state_t ARC_mru; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, mru_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_mru, 0, param_get_arc_state_size, "Q", + "size of mru state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_metadata_esize, CTLFLAG_RD, + &ARC_mru.arcs_esize[ARC_BUFC_METADATA].rc_count, 0, + "size of evictable metadata in mru state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_data_esize, CTLFLAG_RD, + &ARC_mru.arcs_esize[ARC_BUFC_DATA].rc_count, 0, + "size of evictable data in mru state"); + +extern arc_state_t ARC_mru_ghost; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, mru_ghost_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_mru_ghost, 0, param_get_arc_state_size, "Q", + "size of mru ghost state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_metadata_esize, CTLFLAG_RD, + &ARC_mru_ghost.arcs_esize[ARC_BUFC_METADATA].rc_count, 0, + "size of evictable metadata in mru ghost state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_data_esize, CTLFLAG_RD, + &ARC_mru_ghost.arcs_esize[ARC_BUFC_DATA].rc_count, 0, + "size of evictable data in mru ghost state"); + +extern arc_state_t ARC_mfu; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, mfu_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_mfu, 0, param_get_arc_state_size, "Q", + "size of mfu state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_metadata_esize, CTLFLAG_RD, + &ARC_mfu.arcs_esize[ARC_BUFC_METADATA].rc_count, 0, + "size of evictable metadata in mfu state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_data_esize, CTLFLAG_RD, + &ARC_mfu.arcs_esize[ARC_BUFC_DATA].rc_count, 0, + "size of evictable data in mfu state"); + +extern arc_state_t ARC_mfu_ghost; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, mfu_ghost_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_mfu_ghost, 0, param_get_arc_state_size, "Q", + "size of mfu ghost state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_metadata_esize, CTLFLAG_RD, + &ARC_mfu_ghost.arcs_esize[ARC_BUFC_METADATA].rc_count, 0, + "size of evictable metadata in mfu ghost state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_data_esize, CTLFLAG_RD, + &ARC_mfu_ghost.arcs_esize[ARC_BUFC_DATA].rc_count, 0, + "size of evictable data in mfu ghost state"); + +extern arc_state_t ARC_uncached; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, uncached_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_uncached, 0, param_get_arc_state_size, "Q", + "size of uncached state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, uncached_metadata_esize, CTLFLAG_RD, + &ARC_uncached.arcs_esize[ARC_BUFC_METADATA].rc_count, 0, + "size of evictable metadata in uncached state"); +SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, uncached_data_esize, CTLFLAG_RD, + &ARC_uncached.arcs_esize[ARC_BUFC_DATA].rc_count, 0, + "size of evictable data in uncached state"); + +extern arc_state_t ARC_l2c_only; + +SYSCTL_PROC(_vfs_zfs, OID_AUTO, l2c_only_size, + CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, + &ARC_l2c_only, 0, param_get_arc_state_size, "Q", + "size of l2c_only state"); + +/* dbuf.c */ + +/* dmu.c */ + +/* dmu_zfetch.c */ + +SYSCTL_NODE(_vfs_zfs, OID_AUTO, zfetch, CTLFLAG_RW, 0, "ZFS ZFETCH (LEGACY)"); + +extern uint32_t zfetch_max_distance; + +SYSCTL_UINT(_vfs_zfs_zfetch, OID_AUTO, max_distance, + CTLFLAG_RWTUN, &zfetch_max_distance, 0, + "Max bytes to prefetch per stream (LEGACY)"); + +extern uint32_t zfetch_max_idistance; + +SYSCTL_UINT(_vfs_zfs_zfetch, OID_AUTO, max_idistance, + CTLFLAG_RWTUN, &zfetch_max_idistance, 0, + "Max bytes to prefetch indirects for per stream (LEGACY)"); + +/* dsl_pool.c */ + +/* dnode.c */ + +/* dsl_scan.c */ + /* metaslab.c */ int @@ -313,6 +533,19 @@ SYSCTL_UINT(_vfs_zfs, OID_AUTO, condense_pct, "Condense on-disk spacemap when it is more than this many percents" " of in-memory counterpart"); +extern uint_t zfs_remove_max_segment; + +SYSCTL_UINT(_vfs_zfs, OID_AUTO, remove_max_segment, + CTLFLAG_RWTUN, &zfs_remove_max_segment, 0, + "Largest contiguous segment ZFS will attempt to allocate when removing" + " a device"); + +extern int zfs_removal_suspend_progress; + +SYSCTL_INT(_vfs_zfs, OID_AUTO, removal_suspend_progress, + CTLFLAG_RWTUN, &zfs_removal_suspend_progress, 0, + "Ensures certain actions can happen while in the middle of a removal"); + /* * Minimum size which forces the dynamic allocator to change * it's allocation strategy. Once the space map cannot satisfy @@ -532,9 +765,18 @@ param_set_min_auto_ashift(SYSCTL_HANDLER_ARGS) zfs_vdev_min_auto_ashift = val; + if (arg2 != 0) + warn_deprecated_sysctl("min_auto_ashift", + "vdev.min_auto_ashift"); + return (0); } +SYSCTL_PROC(_vfs_zfs, OID_AUTO, min_auto_ashift, + CTLTYPE_UINT | CTLFLAG_RWTUN | CTLFLAG_MPSAFE, NULL, 1, + param_set_min_auto_ashift, "IU", + "Min ashift used when creating new top-level vdev. (LEGACY)"); + int param_set_max_auto_ashift(SYSCTL_HANDLER_ARGS) { @@ -551,9 +793,19 @@ param_set_max_auto_ashift(SYSCTL_HANDLER_ARGS) zfs_vdev_max_auto_ashift = val; + if (arg2 != 0) + warn_deprecated_sysctl("max_auto_ashift", + "vdev.max_auto_ashift"); + return (0); } +SYSCTL_PROC(_vfs_zfs, OID_AUTO, max_auto_ashift, + CTLTYPE_UINT | CTLFLAG_RWTUN | CTLFLAG_MPSAFE, NULL, 1, + param_set_max_auto_ashift, "IU", + "Max ashift used when optimizing for logical -> physical sector size on" + " new top-level vdevs. (LEGACY)"); + /* * Since the DTL space map of a vdev is not expected to have a lot of * entries, we default its block size to 4K. @@ -575,6 +827,23 @@ SYSCTL_INT(_vfs_zfs, OID_AUTO, standard_sm_blksz, CTLFLAG_RDTUN, &zfs_vdev_standard_sm_blksz, 0, "Block size for standard space map. Power of 2 greater than 4096."); +extern int vdev_validate_skip; + +SYSCTL_INT(_vfs_zfs, OID_AUTO, validate_skip, + CTLFLAG_RDTUN, &vdev_validate_skip, 0, + "Enable to bypass vdev_validate()."); + +/* vdev_mirror.c */ + +/* vdev_queue.c */ + +extern uint_t zfs_vdev_max_active; + +SYSCTL_UINT(_vfs_zfs, OID_AUTO, top_maxinflight, + CTLFLAG_RWTUN, &zfs_vdev_max_active, 0, + "The maximum number of I/Os of all types active for each device." + " (LEGACY)"); + /* zio.c */ SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, exclude_metadata, diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c index 4de48e013ec4..d0a9c662e6f0 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c @@ -762,8 +762,7 @@ zfsctl_common_pathconf(struct vop_pathconf_args *ap) return (0); case _PC_MIN_HOLE_SIZE: - *ap->a_retval = (int)SPA_MINBLOCKSIZE; - return (0); + return (EINVAL); case _PC_ACL_EXTENDED: *ap->a_retval = 0; diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c index 411225786089..f34a2fd37a77 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c @@ -4116,6 +4116,7 @@ zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr, { znode_t *zp; zfsvfs_t *zfsvfs; + uint_t blksize, iosize; int error; switch (cmd) { @@ -4127,8 +4128,20 @@ zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr, *valp = 64; return (0); case _PC_MIN_HOLE_SIZE: - *valp = (int)SPA_MINBLOCKSIZE; - return (0); + iosize = vp->v_mount->mnt_stat.f_iosize; + if (vp->v_type == VREG) { + zp = VTOZ(vp); + blksize = zp->z_blksz; + if (zp->z_size <= blksize) + blksize = MAX(blksize, iosize); + *valp = (int)blksize; + return (0); + } + if (vp->v_type == VDIR) { + *valp = (int)iosize; + return (0); + } + return (EINVAL); case _PC_ACL_EXTENDED: #if 0 /* POSIX ACLs are not implemented for ZFS on FreeBSD yet. */ zp = VTOZ(vp); @@ -4210,8 +4223,20 @@ zfs_getpages(struct vnode *vp, vm_page_t *ma, int count, int *rbehind, zfs_vmobject_wlock(object); (void) vm_page_grab_pages(object, OFF_TO_IDX(start), - VM_ALLOC_NORMAL | VM_ALLOC_WAITOK | VM_ALLOC_ZERO, + VM_ALLOC_NORMAL | VM_ALLOC_WAITOK, ma, count); + if (!vm_page_all_valid(ma[count - 1])) { + /* + * Later in this function, we copy DMU data to + * invalid pages only. The last page may not be + * entirely filled though, if the file does not + * end on a page boundary. Therefore, we zero + * that last page here to make sure it does not + * contain garbage after the end of file. + */ + ASSERT(vm_page_none_valid(ma[count - 1])); + vm_page_zero_invalid(ma[count - 1], FALSE); + } zfs_vmobject_wunlock(object); } if (blksz == zp->z_blksz) diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zio_crypt.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zio_crypt.c index 91cf38016e00..8562c42b3220 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zio_crypt.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zio_crypt.c @@ -437,6 +437,7 @@ zio_crypt_key_wrap(crypto_key_t *cwkey, zio_crypt_key_t *key, uint8_t *iv, ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS); + memset(&cuio_s, 0, sizeof (cuio_s)); zfs_uio_init(&cuio, &cuio_s); keydata_len = zio_crypt_table[crypt].ci_keylen; @@ -519,6 +520,7 @@ zio_crypt_key_unwrap(crypto_key_t *cwkey, uint64_t crypt, uint64_t version, keydata_len = zio_crypt_table[crypt].ci_keylen; rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL); + memset(&cuio_s, 0, sizeof (cuio_s)); zfs_uio_init(&cuio, &cuio_s); /* diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c index 0dd2ecd7fd8d..3ddbfcb97184 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c @@ -183,6 +183,7 @@ static struct filterops zvol_filterops_vnode = { .f_isfd = 1, .f_detach = zvol_filter_detach, .f_event = zvol_filter_vnode, + .f_copy = knote_triv_copy, }; extern uint_t zfs_geom_probe_vdev_key; diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c b/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c index 45c2999a4bb1..b2eae5d00b10 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c @@ -25,6 +25,10 @@ * SUCH DAMAGE. */ +/* + * Copyright (c) 2025, Rob Norris <robn@despairlabs.com> + */ + #include <sys/types.h> #include <sys/sysmacros.h> #include <sys/kmem.h> @@ -56,6 +60,19 @@ typedef struct zone_dataset { } zone_dataset_t; #ifdef CONFIG_USER_NS + +/* + * Linux 6.18 moved the generic namespace type away from ns->ops->type onto + * ns_common itself. + */ +#ifdef HAVE_NS_COMMON_TYPE +#define ns_is_newuser(ns) \ + ((ns)->ns_type == CLONE_NEWUSER) +#else +#define ns_is_newuser(ns) \ + ((ns)->ops != NULL && (ns)->ops->type == CLONE_NEWUSER) +#endif + /* * Returns: * - 0 on success @@ -84,7 +101,7 @@ user_ns_get(int fd, struct user_namespace **userns) goto done; } ns = get_proc_ns(file_inode(nsfile)); - if (ns->ops->type != CLONE_NEWUSER) { + if (!ns_is_newuser(ns)) { error = ENOTTY; goto done; } diff --git a/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c b/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c index 8a8316f63c48..18f2426fbbfc 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c @@ -23,6 +23,7 @@ * Copyright (c) 2014 by Chunwei Chen. All rights reserved. * Copyright (c) 2019 by Delphix. All rights reserved. * Copyright (c) 2023, 2024, Klara Inc. + * Copyright (c) 2025, Rob Norris <robn@despairlabs.com> */ /* @@ -1109,6 +1110,14 @@ abd_return_buf_copy(abd_t *abd, void *buf, size_t n) #define ABD_ITER_PAGE_SIZE(page) (PAGESIZE) #endif +#ifndef nth_page +/* + * Since 6.18 nth_page() no longer exists, and is no longer required to iterate + * within a single SG entry, so we replace it with a simple addition. + */ +#define nth_page(p, n) ((p)+(n)) +#endif + void abd_iter_page(struct abd_iter *aiter) { diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c index daa4b5776837..934d74a112fd 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c @@ -2524,7 +2524,7 @@ zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode, * Also note: DOS R/O is ignored for directories. */ if ((v4_mode & WRITE_MASK_DATA) && - S_ISDIR(ZTOI(zp)->i_mode) && + !S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_READONLY)) { return (SET_ERROR(EPERM)); } diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c index 6106726651a3..e845ad69ad78 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c @@ -2033,10 +2033,7 @@ zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr, zidmap_t *mnt_ns) goto out3; } - if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { - err = SET_ERROR(EPERM); - goto out3; - } + /* ZFS_READONLY will be handled in zfs_zaccess() */ /* * Verify timestamps doesn't overflow 32 bits. diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c index d07317b0d910..02965ac8cbee 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c @@ -23,6 +23,7 @@ * Copyright (c) 2011, Lawrence Livermore National Security, LLC. * Copyright (c) 2015 by Chunwei Chen. All rights reserved. * Copyright (c) 2025, Klara, Inc. + * Copyright (c) 2025, Rob Norris <robn@despairlabs.com> */ @@ -478,6 +479,7 @@ zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data) return (ret); } +#ifdef HAVE_WRITE_CACHE_PAGES #ifdef HAVE_WRITEPAGE_T_FOLIO static int zpl_putfolio(struct folio *pp, struct writeback_control *wbc, void *data) @@ -499,6 +501,78 @@ zpl_write_cache_pages(struct address_space *mapping, #endif return (result); } +#else +static inline int +zpl_write_cache_pages(struct address_space *mapping, + struct writeback_control *wbc, void *data) +{ + pgoff_t start = wbc->range_start >> PAGE_SHIFT; + pgoff_t end = wbc->range_end >> PAGE_SHIFT; + + struct folio_batch fbatch; + folio_batch_init(&fbatch); + + /* + * This atomically (-ish) tags all DIRTY pages in the range with + * TOWRITE, allowing users to continue dirtying or undirtying pages + * while we get on with writeback, without us treading on each other. + */ + tag_pages_for_writeback(mapping, start, end); + + int err = 0; + unsigned int npages; + + /* + * Grab references to the TOWRITE pages just flagged. This may not get + * all of them, so we do it in a loop until there are none left. + */ + while ((npages = filemap_get_folios_tag(mapping, &start, end, + PAGECACHE_TAG_TOWRITE, &fbatch)) != 0) { + + /* Loop over each page and write it out. */ + struct folio *folio; + while ((folio = folio_batch_next(&fbatch)) != NULL) { + folio_lock(folio); + + /* + * If the folio has been remapped, or is no longer + * dirty, then there's nothing to do. + */ + if (folio->mapping != mapping || + !folio_test_dirty(folio)) { + folio_unlock(folio); + continue; + } + + /* + * If writeback is already in progress, wait for it to + * finish. We continue after this even if the page + * ends up clean; zfs_putpage() will skip it if no + * further work is required. + */ + while (folio_test_writeback(folio)) + folio_wait_bit(folio, PG_writeback); + + /* + * Write it out and collect any error. zfs_putpage() + * will clear the TOWRITE and DIRTY flags, and return + * with the page unlocked. + */ + int ferr = zpl_putpage(&folio->page, wbc, data); + if (err == 0 && ferr != 0) + err = ferr; + + /* Housekeeping for the caller. */ + wbc->nr_to_write -= folio_nr_pages(folio); + } + + /* Release any remaining references on the batch. */ + folio_batch_release(&fbatch); + } + + return (err); +} +#endif static int zpl_writepages(struct address_space *mapping, struct writeback_control *wbc) diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c index 444948d03cb3..347b352506e5 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c @@ -23,6 +23,7 @@ * Copyright (c) 2011, Lawrence Livermore National Security, LLC. * Copyright (c) 2023, Datto Inc. All rights reserved. * Copyright (c) 2025, Klara, Inc. + * Copyright (c) 2025, Rob Norris <robn@despairlabs.com> */ @@ -33,6 +34,7 @@ #include <sys/zpl.h> #include <linux/iversion.h> #include <linux/version.h> +#include <linux/vfs_compat.h> /* * What to do when the last reference to an inode is released. If 0, the kernel @@ -104,7 +106,7 @@ zpl_dirty_inode(struct inode *ip, int flags) * reporting memory pressure and requests OpenZFS release some memory (see * zfs_prune()). * - * When set to 1, we call generic_delete_node(), which always returns "destroy + * When set to 1, we call generic_delete_inode(), which always returns "destroy * immediately", resulting in inodes being destroyed immediately, releasing * their associated dnodes and dbufs to the dbuf cached and the ARC to be * evicted as normal. diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c index 967a018640e1..fe939150b641 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c @@ -21,7 +21,7 @@ */ /* * Copyright (c) 2012, 2020 by Delphix. All rights reserved. - * Copyright (c) 2024, Rob Norris <robn@despairlabs.com> + * Copyright (c) 2024, 2025, Rob Norris <robn@despairlabs.com> * Copyright (c) 2024, 2025, Klara, Inc. */ @@ -337,16 +337,14 @@ zvol_discard(zv_request_t *zvr) } /* - * Align the request to volume block boundaries when a secure erase is - * not required. This will prevent dnode_free_range() from zeroing out - * the unaligned parts which is slow (read-modify-write) and useless - * since we are not freeing any space by doing so. + * Align the request to volume block boundaries. This will prevent + * dnode_free_range() from zeroing out the unaligned parts which is + * slow (read-modify-write) and useless since we are not freeing any + * space by doing so. */ - if (!io_is_secure_erase(bio, rq)) { - start = P2ROUNDUP(start, zv->zv_volblocksize); - end = P2ALIGN_TYPED(end, zv->zv_volblocksize, uint64_t); - size = end - start; - } + start = P2ROUNDUP(start, zv->zv_volblocksize); + end = P2ALIGN_TYPED(end, zv->zv_volblocksize, uint64_t); + size = end - start; if (start >= end) goto unlock; @@ -467,6 +465,24 @@ zvol_read_task(void *arg) zv_request_task_free(task); } +/* + * Note: + * + * The kernel uses different enum names for the IO opcode, depending on the + * kernel version ('req_opf', 'req_op'). To sidestep this, use macros rather + * than inline functions for these checks. + */ +/* Should this IO go down the zvol write path? */ +#define ZVOL_OP_IS_WRITE(op) \ + (op == REQ_OP_WRITE || \ + op == REQ_OP_FLUSH || \ + op == REQ_OP_DISCARD) + +/* Is this IO type supported by zvols? */ +#define ZVOL_OP_IS_SUPPORTED(op) (op == REQ_OP_READ || ZVOL_OP_IS_WRITE(op)) + +/* Get the IO opcode */ +#define ZVOL_OP(bio, rq) (bio != NULL ? bio_op(bio) : req_op(rq)) /* * Process a BIO or request @@ -486,27 +502,32 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq, uint64_t size = io_size(bio, rq); int rw; - if (rq != NULL) { - /* - * Flush & trim requests go down the zvol_write codepath. Or - * more specifically: - * - * If request is a write, or if it's op_is_sync() and not a - * read, or if it's a flush, or if it's a discard, then send the - * request down the write path. - */ - if (op_is_write(rq->cmd_flags) || - (op_is_sync(rq->cmd_flags) && req_op(rq) != REQ_OP_READ) || - req_op(rq) == REQ_OP_FLUSH || - op_is_discard(rq->cmd_flags)) { - rw = WRITE; - } else { - rw = READ; - } + if (unlikely(!ZVOL_OP_IS_SUPPORTED(ZVOL_OP(bio, rq)))) { + zfs_dbgmsg("Unsupported zvol %s, op=%d, flags=0x%x", + rq != NULL ? "request" : "BIO", + ZVOL_OP(bio, rq), + rq != NULL ? rq->cmd_flags : bio->bi_opf); + ASSERT(ZVOL_OP_IS_SUPPORTED(ZVOL_OP(bio, rq))); + zvol_end_io(bio, rq, SET_ERROR(ENOTSUPP)); + goto out; + } + + if (ZVOL_OP_IS_WRITE(ZVOL_OP(bio, rq))) { + rw = WRITE; } else { - rw = bio_data_dir(bio); + rw = READ; } + /* + * Sanity check + * + * If we're a BIO, check our rw matches the kernel's + * bio_data_dir(bio) rw. We need to check because we support fewer + * IO operations, and want to verify that what we think are reads and + * writes from those operations match what the kernel thinks. + */ + ASSERT(rq != NULL || rw == bio_data_dir(bio)); + if (unlikely(zv->zv_flags & ZVOL_REMOVING)) { zvol_end_io(bio, rq, SET_ERROR(ENXIO)); goto out; @@ -610,7 +631,7 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq, * interfaces lack this functionality (they block waiting for * the i/o to complete). */ - if (io_is_discard(bio, rq) || io_is_secure_erase(bio, rq)) { + if (io_is_discard(bio, rq)) { if (force_sync) { zvol_discard(&zvr); } else { @@ -1011,12 +1032,12 @@ zvol_os_update_volsize(zvol_state_t *zv, uint64_t volsize) * tiny devices. For devices over 1 Mib a standard head and sector count * is used to keep the cylinders count reasonable. */ -static int -zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo) +static inline int +zvol_getgeo_impl(struct gendisk *disk, struct hd_geometry *geo) { + zvol_state_t *zv = atomic_load_ptr(&disk->private_data); sector_t sectors; - zvol_state_t *zv = atomic_load_ptr(&bdev->bd_disk->private_data); ASSERT3P(zv, !=, NULL); ASSERT3U(zv->zv_open_count, >, 0); @@ -1036,6 +1057,20 @@ zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo) return (0); } +#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK +static int +zvol_getgeo(struct gendisk *disk, struct hd_geometry *geo) +{ + return (zvol_getgeo_impl(disk, geo)); +} +#else +static int +zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo) +{ + return (zvol_getgeo_impl(bdev->bd_disk, geo)); +} +#endif + /* * Why have two separate block_device_operations structs? * @@ -1479,7 +1514,7 @@ zvol_os_remove_minor(zvol_state_t *zv) if (zso->use_blk_mq) blk_mq_free_tag_set(&zso->tag_set); - ida_simple_remove(&zvol_ida, MINOR(zso->zvo_dev) >> ZVOL_MINOR_BITS); + ida_free(&zvol_ida, MINOR(zso->zvo_dev) >> ZVOL_MINOR_BITS); kmem_free(zso, sizeof (struct zvol_state_os)); @@ -1634,7 +1669,7 @@ zvol_os_create_minor(const char *name) if (zvol_inhibit_dev) return (0); - idx = ida_simple_get(&zvol_ida, 0, 0, kmem_flags_convert(KM_SLEEP)); + idx = ida_alloc(&zvol_ida, kmem_flags_convert(KM_SLEEP)); if (idx < 0) return (SET_ERROR(-idx)); minor = idx << ZVOL_MINOR_BITS; @@ -1642,7 +1677,7 @@ zvol_os_create_minor(const char *name) /* too many partitions can cause an overflow */ zfs_dbgmsg("zvol: create minor overflow: %s, minor %u/%u", name, minor, MINOR(minor)); - ida_simple_remove(&zvol_ida, idx); + ida_free(&zvol_ida, idx); return (SET_ERROR(EINVAL)); } @@ -1650,7 +1685,7 @@ zvol_os_create_minor(const char *name) if (zv) { ASSERT(MUTEX_HELD(&zv->zv_state_lock)); mutex_exit(&zv->zv_state_lock); - ida_simple_remove(&zvol_ida, idx); + ida_free(&zvol_ida, idx); return (SET_ERROR(EEXIST)); } @@ -1750,7 +1785,7 @@ out_doi: rw_exit(&zvol_state_lock); error = zvol_os_add_disk(zv->zv_zso->zvo_disk); } else { - ida_simple_remove(&zvol_ida, idx); + ida_free(&zvol_ida, idx); } return (error); |