diff --git a/config/kernel-blkdev.m4 b/config/kernel-blkdev.m4 index 887acee670ba..e04a2bd2c3b6 100644 --- a/config/kernel-blkdev.m4 +++ b/config/kernel-blkdev.m4 @@ -1,533 +1,611 @@ dnl # dnl # 2.6.38 API change, dnl # Added blkdev_get_by_path() dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH], [ ZFS_LINUX_TEST_SRC([blkdev_get_by_path], [ #include #include ], [ struct block_device *bdev __attribute__ ((unused)) = NULL; const char *path = "path"; fmode_t mode = 0; void *holder = NULL; bdev = blkdev_get_by_path(path, mode, holder); ]) ]) +dnl # +dnl # 6.5.x API change, +dnl # blkdev_get_by_path() takes 4 args +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH_4ARG], [ + ZFS_LINUX_TEST_SRC([blkdev_get_by_path_4arg], [ + #include + #include + ], [ + struct block_device *bdev __attribute__ ((unused)) = NULL; + const char *path = "path"; + fmode_t mode = 0; + void *holder = NULL; + struct blk_holder_ops h; + + bdev = blkdev_get_by_path(path, mode, holder, &h); + ]) +]) + AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_GET_BY_PATH], [ - AC_MSG_CHECKING([whether blkdev_get_by_path() exists]) + AC_MSG_CHECKING([whether blkdev_get_by_path() exists and takes 3 args]) ZFS_LINUX_TEST_RESULT([blkdev_get_by_path], [ AC_MSG_RESULT(yes) ], [ - ZFS_LINUX_TEST_ERROR([blkdev_get_by_path()]) + AC_MSG_RESULT(no) + AC_MSG_CHECKING([whether blkdev_get_by_path() exists and takes 4 args]) + ZFS_LINUX_TEST_RESULT([blkdev_get_by_path_4arg], [ + AC_DEFINE(HAVE_BLKDEV_GET_BY_PATH_4ARG, 1, + [blkdev_get_by_path() exists and takes 4 args]) + AC_MSG_RESULT(yes) + ], [ + ZFS_LINUX_TEST_ERROR([blkdev_get_by_path()]) + ]) + ]) +]) + +dnl # +dnl # 6.5.x API change +dnl # blk_mode_t was added as a type to supercede some places where fmode_t +dnl # is used +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BLK_MODE_T], [ + ZFS_LINUX_TEST_SRC([blk_mode_t], [ + #include + #include + ], [ + blk_mode_t m __attribute((unused)) = (blk_mode_t)0; + ]) +]) + +AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BLK_MODE_T], [ + AC_MSG_CHECKING([whether blk_mode_t is defined]) + ZFS_LINUX_TEST_RESULT([blk_mode_t], [ + AC_MSG_RESULT(yes) + AC_DEFINE(HAVE_BLK_MODE_T, 1, [blk_mode_t is defined]) + ], [ + AC_MSG_RESULT(no) ]) ]) dnl # dnl # 2.6.38 API change, dnl # Added blkdev_put() dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_PUT], [ ZFS_LINUX_TEST_SRC([blkdev_put], [ #include #include ], [ struct block_device *bdev = NULL; fmode_t mode = 0; blkdev_put(bdev, mode); ]) ]) +dnl # +dnl # 6.5.x API change. +dnl # blkdev_put() takes (void* holder) as arg 2 +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_PUT_HOLDER], [ + ZFS_LINUX_TEST_SRC([blkdev_put_holder], [ + #include + #include + ], [ + struct block_device *bdev = NULL; + void *holder = NULL; + + blkdev_put(bdev, holder); + ]) +]) + AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_PUT], [ AC_MSG_CHECKING([whether blkdev_put() exists]) ZFS_LINUX_TEST_RESULT([blkdev_put], [ AC_MSG_RESULT(yes) ], [ - ZFS_LINUX_TEST_ERROR([blkdev_put()]) + AC_MSG_CHECKING([whether blkdev_put() accepts void* as arg 2]) + ZFS_LINUX_TEST_RESULT([blkdev_put_holder], [ + AC_MSG_RESULT(yes) + AC_DEFINE(HAVE_BLKDEV_PUT_HOLDER, 1, + [blkdev_put() accepts void* as arg 2]) + ], [ + ZFS_LINUX_TEST_ERROR([blkdev_put()]) + ]) ]) ]) dnl # dnl # 4.1 API, exported blkdev_reread_part() symbol, back ported to the dnl # 3.10.0 CentOS 7.x enterprise kernels. dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_REREAD_PART], [ ZFS_LINUX_TEST_SRC([blkdev_reread_part], [ #include #include ], [ struct block_device *bdev = NULL; int error; error = blkdev_reread_part(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_REREAD_PART], [ AC_MSG_CHECKING([whether blkdev_reread_part() exists]) ZFS_LINUX_TEST_RESULT([blkdev_reread_part], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BLKDEV_REREAD_PART, 1, [blkdev_reread_part() exists]) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # check_disk_change() was removed in 5.10 dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_CHECK_DISK_CHANGE], [ ZFS_LINUX_TEST_SRC([check_disk_change], [ #include #include ], [ struct block_device *bdev = NULL; bool error; error = check_disk_change(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE], [ AC_MSG_CHECKING([whether check_disk_change() exists]) ZFS_LINUX_TEST_RESULT([check_disk_change], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_CHECK_DISK_CHANGE, 1, [check_disk_change() exists]) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # 6.5.x API change dnl # disk_check_media_change() was added dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_DISK_CHECK_MEDIA_CHANGE], [ ZFS_LINUX_TEST_SRC([disk_check_media_change], [ #include #include ], [ struct block_device *bdev = NULL; bool error; error = disk_check_media_change(bdev->bd_disk); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_DISK_CHECK_MEDIA_CHANGE], [ AC_MSG_CHECKING([whether disk_check_media_change() exists]) ZFS_LINUX_TEST_RESULT([disk_check_media_change], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_DISK_CHECK_MEDIA_CHANGE, 1, [disk_check_media_change() exists]) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # bdev_kobj() is introduced from 5.12 dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ], [ ZFS_LINUX_TEST_SRC([bdev_kobj], [ #include #include #include ], [ struct block_device *bdev = NULL; struct kobject *disk_kobj; disk_kobj = bdev_kobj(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ], [ AC_MSG_CHECKING([whether bdev_kobj() exists]) ZFS_LINUX_TEST_RESULT([bdev_kobj], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BDEV_KOBJ, 1, [bdev_kobj() exists]) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # part_to_dev() was removed in 5.12 dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV], [ ZFS_LINUX_TEST_SRC([part_to_dev], [ #include #include ], [ struct hd_struct *p = NULL; struct device *pdev; pdev = part_to_dev(p); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV], [ AC_MSG_CHECKING([whether part_to_dev() exists]) ZFS_LINUX_TEST_RESULT([part_to_dev], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_PART_TO_DEV, 1, [part_to_dev() exists]) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # 5.10 API, check_disk_change() is removed, in favor of dnl # bdev_check_media_change(), which doesn't force revalidation dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_CHECK_MEDIA_CHANGE], [ ZFS_LINUX_TEST_SRC([bdev_check_media_change], [ #include #include ], [ struct block_device *bdev = NULL; int error; error = bdev_check_media_change(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_CHECK_MEDIA_CHANGE], [ AC_MSG_CHECKING([whether bdev_check_media_change() exists]) ZFS_LINUX_TEST_RESULT([bdev_check_media_change], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BDEV_CHECK_MEDIA_CHANGE, 1, [bdev_check_media_change() exists]) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # 2.6.22 API change dnl # Single argument invalidate_bdev() dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_INVALIDATE_BDEV], [ ZFS_LINUX_TEST_SRC([invalidate_bdev], [ #include #include ],[ struct block_device *bdev = NULL; invalidate_bdev(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_INVALIDATE_BDEV], [ AC_MSG_CHECKING([whether invalidate_bdev() exists]) ZFS_LINUX_TEST_RESULT([invalidate_bdev], [ AC_MSG_RESULT(yes) ],[ ZFS_LINUX_TEST_ERROR([invalidate_bdev()]) ]) ]) dnl # dnl # 5.11 API, lookup_bdev() takes dev_t argument. dnl # 2.6.27 API, lookup_bdev() was first exported. dnl # 4.4.0-6.21 API, lookup_bdev() on Ubuntu takes mode argument. dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_LOOKUP_BDEV], [ ZFS_LINUX_TEST_SRC([lookup_bdev_devt], [ #include ], [ int error __attribute__ ((unused)); const char path[] = "/example/path"; dev_t dev; error = lookup_bdev(path, &dev); ]) ZFS_LINUX_TEST_SRC([lookup_bdev_1arg], [ #include #include ], [ struct block_device *bdev __attribute__ ((unused)); const char path[] = "/example/path"; bdev = lookup_bdev(path); ]) ZFS_LINUX_TEST_SRC([lookup_bdev_mode], [ #include ], [ struct block_device *bdev __attribute__ ((unused)); const char path[] = "/example/path"; bdev = lookup_bdev(path, FMODE_READ); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_LOOKUP_BDEV], [ AC_MSG_CHECKING([whether lookup_bdev() wants dev_t arg]) ZFS_LINUX_TEST_RESULT_SYMBOL([lookup_bdev_devt], [lookup_bdev], [fs/block_dev.c], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_DEVT_LOOKUP_BDEV, 1, [lookup_bdev() wants dev_t arg]) ], [ AC_MSG_RESULT(no) AC_MSG_CHECKING([whether lookup_bdev() wants 1 arg]) ZFS_LINUX_TEST_RESULT_SYMBOL([lookup_bdev_1arg], [lookup_bdev], [fs/block_dev.c], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_1ARG_LOOKUP_BDEV, 1, [lookup_bdev() wants 1 arg]) ], [ AC_MSG_RESULT(no) AC_MSG_CHECKING([whether lookup_bdev() wants mode arg]) ZFS_LINUX_TEST_RESULT_SYMBOL([lookup_bdev_mode], [lookup_bdev], [fs/block_dev.c], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_MODE_LOOKUP_BDEV, 1, [lookup_bdev() wants mode arg]) ], [ ZFS_LINUX_TEST_ERROR([lookup_bdev()]) ]) ]) ]) ]) dnl # dnl # 2.6.30 API change dnl # dnl # The bdev_physical_block_size() interface was added to provide a way dnl # to determine the smallest write which can be performed without a dnl # read-modify-write operation. dnl # dnl # Unfortunately, this interface isn't entirely reliable because dnl # drives are sometimes known to misreport this value. dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE], [ ZFS_LINUX_TEST_SRC([bdev_physical_block_size], [ #include ],[ struct block_device *bdev __attribute__ ((unused)) = NULL; bdev_physical_block_size(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE], [ AC_MSG_CHECKING([whether bdev_physical_block_size() is available]) ZFS_LINUX_TEST_RESULT([bdev_physical_block_size], [ AC_MSG_RESULT(yes) ],[ ZFS_LINUX_TEST_ERROR([bdev_physical_block_size()]) ]) ]) dnl # dnl # 2.6.30 API change dnl # Added bdev_logical_block_size(). dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE], [ ZFS_LINUX_TEST_SRC([bdev_logical_block_size], [ #include ],[ struct block_device *bdev __attribute__ ((unused)) = NULL; bdev_logical_block_size(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE], [ AC_MSG_CHECKING([whether bdev_logical_block_size() is available]) ZFS_LINUX_TEST_RESULT([bdev_logical_block_size], [ AC_MSG_RESULT(yes) ],[ ZFS_LINUX_TEST_ERROR([bdev_logical_block_size()]) ]) ]) dnl # dnl # 5.11 API change dnl # Added bdev_whole() helper. dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_WHOLE], [ ZFS_LINUX_TEST_SRC([bdev_whole], [ #include ],[ struct block_device *bdev = NULL; bdev = bdev_whole(bdev); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_WHOLE], [ AC_MSG_CHECKING([whether bdev_whole() is available]) ZFS_LINUX_TEST_RESULT([bdev_whole], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BDEV_WHOLE, 1, [bdev_whole() is available]) ],[ AC_MSG_RESULT(no) ]) ]) dnl # dnl # 5.20 API change, dnl # Removed bdevname(), snprintf(.., %pg) should be used. dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME], [ ZFS_LINUX_TEST_SRC([bdevname], [ #include #include ], [ struct block_device *bdev __attribute__ ((unused)) = NULL; char path[BDEVNAME_SIZE]; (void) bdevname(bdev, path); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEVNAME], [ AC_MSG_CHECKING([whether bdevname() exists]) ZFS_LINUX_TEST_RESULT([bdevname], [ AC_DEFINE(HAVE_BDEVNAME, 1, [bdevname() is available]) AC_MSG_RESULT(yes) ], [ AC_MSG_RESULT(no) ]) ]) dnl # dnl # 5.19 API: blkdev_issue_secure_erase() dnl # 3.10 API: blkdev_issue_discard(..., BLKDEV_DISCARD_SECURE) dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE], [ ZFS_LINUX_TEST_SRC([blkdev_issue_secure_erase], [ #include ],[ struct block_device *bdev = NULL; sector_t sector = 0; sector_t nr_sects = 0; int error __attribute__ ((unused)); error = blkdev_issue_secure_erase(bdev, sector, nr_sects, GFP_KERNEL); ]) ZFS_LINUX_TEST_SRC([blkdev_issue_discard_flags], [ #include ],[ struct block_device *bdev = NULL; sector_t sector = 0; sector_t nr_sects = 0; unsigned long flags = 0; int error __attribute__ ((unused)); error = blkdev_issue_discard(bdev, sector, nr_sects, GFP_KERNEL, flags); ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE], [ AC_MSG_CHECKING([whether blkdev_issue_secure_erase() is available]) ZFS_LINUX_TEST_RESULT([blkdev_issue_secure_erase], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BLKDEV_ISSUE_SECURE_ERASE, 1, [blkdev_issue_secure_erase() is available]) ],[ AC_MSG_RESULT(no) AC_MSG_CHECKING([whether blkdev_issue_discard() is available]) ZFS_LINUX_TEST_RESULT([blkdev_issue_discard_flags], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BLKDEV_ISSUE_DISCARD, 1, [blkdev_issue_discard() is available]) ],[ ZFS_LINUX_TEST_ERROR([blkdev_issue_discard()]) ]) ]) ]) dnl # dnl # 5.13 API change dnl # blkdev_get_by_path() no longer handles ERESTARTSYS dnl # dnl # Unfortunately we're forced to rely solely on the kernel version dnl # number in order to determine the expected behavior. This was an dnl # internal change to blkdev_get_by_dev(), see commit a8ed1a0607. dnl # AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS], [ AC_MSG_CHECKING([whether blkdev_get_by_path() handles ERESTARTSYS]) AS_VERSION_COMPARE([$LINUX_VERSION], [5.13.0], [ AC_MSG_RESULT(yes) AC_DEFINE(HAVE_BLKDEV_GET_ERESTARTSYS, 1, [blkdev_get_by_path() handles ERESTARTSYS]) ],[ AC_MSG_RESULT(no) ],[ AC_MSG_RESULT(no) ]) ]) dnl # dnl # 6.5.x API change dnl # BLK_STS_NEXUS replaced with BLK_STS_RESV_CONFLICT dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BLK_STS_RESV_CONFLICT], [ ZFS_LINUX_TEST_SRC([blk_sts_resv_conflict], [ #include ],[ blk_status_t s __attribute__ ((unused)) = BLK_STS_RESV_CONFLICT; ]) ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BLK_STS_RESV_CONFLICT], [ AC_MSG_CHECKING([whether BLK_STS_RESV_CONFLICT is defined]) ZFS_LINUX_TEST_RESULT([blk_sts_resv_conflict], [ AC_DEFINE(HAVE_BLK_STS_RESV_CONFLICT, 1, [BLK_STS_RESV_CONFLICT is defined]) AC_MSG_RESULT(yes) ], [ AC_MSG_RESULT(no) ]) ]) ]) AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV], [ ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH + ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH_4ARG ZFS_AC_KERNEL_SRC_BLKDEV_PUT + ZFS_AC_KERNEL_SRC_BLKDEV_PUT_HOLDER ZFS_AC_KERNEL_SRC_BLKDEV_REREAD_PART ZFS_AC_KERNEL_SRC_BLKDEV_INVALIDATE_BDEV ZFS_AC_KERNEL_SRC_BLKDEV_LOOKUP_BDEV ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE ZFS_AC_KERNEL_SRC_BLKDEV_CHECK_DISK_CHANGE ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_CHECK_MEDIA_CHANGE ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_WHOLE ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV ZFS_AC_KERNEL_SRC_BLKDEV_DISK_CHECK_MEDIA_CHANGE ZFS_AC_KERNEL_SRC_BLKDEV_BLK_STS_RESV_CONFLICT + ZFS_AC_KERNEL_SRC_BLKDEV_BLK_MODE_T ]) AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [ ZFS_AC_KERNEL_BLKDEV_GET_BY_PATH ZFS_AC_KERNEL_BLKDEV_PUT ZFS_AC_KERNEL_BLKDEV_REREAD_PART ZFS_AC_KERNEL_BLKDEV_INVALIDATE_BDEV ZFS_AC_KERNEL_BLKDEV_LOOKUP_BDEV ZFS_AC_KERNEL_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE ZFS_AC_KERNEL_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE ZFS_AC_KERNEL_BLKDEV_BDEV_CHECK_MEDIA_CHANGE ZFS_AC_KERNEL_BLKDEV_BDEV_WHOLE ZFS_AC_KERNEL_BLKDEV_BDEVNAME ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV ZFS_AC_KERNEL_BLKDEV_DISK_CHECK_MEDIA_CHANGE ZFS_AC_KERNEL_BLKDEV_BLK_STS_RESV_CONFLICT + ZFS_AC_KERNEL_BLKDEV_BLK_MODE_T ]) diff --git a/config/kernel-block-device-operations.m4 b/config/kernel-block-device-operations.m4 index 84e39dc8a2f6..d13c1337b1fb 100644 --- a/config/kernel-block-device-operations.m4 +++ b/config/kernel-block-device-operations.m4 @@ -1,102 +1,133 @@ dnl # dnl # 2.6.38 API change dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS], [ ZFS_LINUX_TEST_SRC([block_device_operations_check_events], [ #include unsigned int blk_check_events(struct gendisk *disk, unsigned int clearing) { (void) disk, (void) clearing; return (0); } static const struct block_device_operations bops __attribute__ ((unused)) = { .check_events = blk_check_events, }; ], [], []) ]) AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS], [ AC_MSG_CHECKING([whether bops->check_events() exists]) ZFS_LINUX_TEST_RESULT([block_device_operations_check_events], [ AC_MSG_RESULT(yes) ],[ ZFS_LINUX_TEST_ERROR([bops->check_events()]) ]) ]) dnl # dnl # 3.10.x API change dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID], [ ZFS_LINUX_TEST_SRC([block_device_operations_release_void], [ #include void blk_release(struct gendisk *g, fmode_t mode) { (void) g, (void) mode; return; } static const struct block_device_operations bops __attribute__ ((unused)) = { .open = NULL, .release = blk_release, .ioctl = NULL, .compat_ioctl = NULL, }; ], [], []) ]) +dnl # +dnl # 5.9.x API change +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG], [ + ZFS_LINUX_TEST_SRC([block_device_operations_release_void_1arg], [ + #include + + void blk_release(struct gendisk *g) { + (void) g; + return; + } + + static const struct block_device_operations + bops __attribute__ ((unused)) = { + .open = NULL, + .release = blk_release, + .ioctl = NULL, + .compat_ioctl = NULL, + }; + ], [], []) +]) + AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID], [ - AC_MSG_CHECKING([whether bops->release() is void]) + AC_MSG_CHECKING([whether bops->release() is void and takes 2 args]) ZFS_LINUX_TEST_RESULT([block_device_operations_release_void], [ AC_MSG_RESULT(yes) ],[ - ZFS_LINUX_TEST_ERROR([bops->release()]) + AC_MSG_RESULT(no) + AC_MSG_CHECKING([whether bops->release() is void and takes 1 arg]) + ZFS_LINUX_TEST_RESULT([block_device_operations_release_void_1arg], [ + AC_MSG_RESULT(yes) + AC_DEFINE([HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG], [1], + [Define if release() in block_device_operations takes 1 arg]) + ],[ + ZFS_LINUX_TEST_ERROR([bops->release()]) + ]) ]) ]) dnl # dnl # 5.13 API change dnl # block_device_operations->revalidate_disk() was removed dnl # AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [ ZFS_LINUX_TEST_SRC([block_device_operations_revalidate_disk], [ #include int blk_revalidate_disk(struct gendisk *disk) { (void) disk; return(0); } static const struct block_device_operations bops __attribute__ ((unused)) = { .revalidate_disk = blk_revalidate_disk, }; ], [], []) ]) AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [ AC_MSG_CHECKING([whether bops->revalidate_disk() exists]) ZFS_LINUX_TEST_RESULT([block_device_operations_revalidate_disk], [ AC_DEFINE([HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [1], [Define if revalidate_disk() in block_device_operations]) AC_MSG_RESULT(yes) ],[ AC_MSG_RESULT(no) ]) ]) AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS], [ ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID + ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK ]) AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS], [ ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK ]) diff --git a/include/os/linux/kernel/linux/blkdev_compat.h b/include/os/linux/kernel/linux/blkdev_compat.h index 109fb9ace80a..912919f4c8db 100644 --- a/include/os/linux/kernel/linux/blkdev_compat.h +++ b/include/os/linux/kernel/linux/blkdev_compat.h @@ -1,658 +1,664 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (C) 2011 Lawrence Livermore National Security, LLC. * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER). * Written by Brian Behlendorf . * LLNL-CODE-403049. */ #ifndef _ZFS_BLKDEV_H #define _ZFS_BLKDEV_H #include #include #include #include #include /* for SECTOR_* */ #ifndef HAVE_BLK_QUEUE_FLAG_SET static inline void blk_queue_flag_set(unsigned int flag, struct request_queue *q) { queue_flag_set(flag, q); } #endif #ifndef HAVE_BLK_QUEUE_FLAG_CLEAR static inline void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) { queue_flag_clear(flag, q); } #endif /* * 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 * our own trivial wrapper when the GPL-only version is detected. * * 2.6.36 - 4.6 API, * The blk_queue_flush() interface has replaced blk_queue_ordered() * interface. However, while the old interface was available to all the * new one is GPL-only. Thus if the GPL-only version is detected we * implement our own trivial helper. */ static inline void blk_queue_set_write_cache(struct request_queue *q, bool wc, bool fua) { #if defined(HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY) if (wc) 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 blk_queue_flag_clear(QUEUE_FLAG_FUA, q); #elif defined(HAVE_BLK_QUEUE_WRITE_CACHE) blk_queue_write_cache(q, wc, fua); #elif defined(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY) if (wc) q->flush_flags |= REQ_FLUSH; if (fua) q->flush_flags |= REQ_FUA; #elif defined(HAVE_BLK_QUEUE_FLUSH) blk_queue_flush(q, (wc ? REQ_FLUSH : 0) | (fua ? REQ_FUA : 0)); #else #error "Unsupported kernel" #endif } 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 #define BIO_BI_SECTOR(bio) (bio)->bi_iter.bi_sector #define BIO_BI_SIZE(bio) (bio)->bi_iter.bi_size #define BIO_BI_IDX(bio) (bio)->bi_iter.bi_idx #define BIO_BI_SKIP(bio) (bio)->bi_iter.bi_bvec_done #define bio_for_each_segment4(bv, bvp, b, i) \ bio_for_each_segment((bv), (b), (i)) typedef struct bvec_iter bvec_iterator_t; #else #define BIO_BI_SECTOR(bio) (bio)->bi_sector #define BIO_BI_SIZE(bio) (bio)->bi_size #define BIO_BI_IDX(bio) (bio)->bi_idx #define BIO_BI_SKIP(bio) (0) #define bio_for_each_segment4(bv, bvp, b, i) \ bio_for_each_segment((bvp), (b), (i)) typedef int bvec_iterator_t; #endif static inline void bio_set_flags_failfast(struct block_device *bdev, int *flags) { #ifdef CONFIG_BUG /* * Disable FAILFAST for loopback devices because of the * following incorrect BUG_ON() in loop_make_request(). * This support is also disabled for md devices because the * test suite layers md devices on top of loopback devices. * This may be removed when the loopback driver is fixed. * * BUG_ON(!lo || (rw != READ && rw != WRITE)); */ if ((MAJOR(bdev->bd_dev) == LOOP_MAJOR) || (MAJOR(bdev->bd_dev) == MD_MAJOR)) return; #ifdef BLOCK_EXT_MAJOR if (MAJOR(bdev->bd_dev) == BLOCK_EXT_MAJOR) return; #endif /* BLOCK_EXT_MAJOR */ #endif /* CONFIG_BUG */ *flags |= REQ_FAILFAST_MASK; } /* * Maximum disk label length, it may be undefined for some kernels. */ #if !defined(DISK_NAME_LEN) #define DISK_NAME_LEN 32 #endif /* DISK_NAME_LEN */ #ifdef HAVE_BIO_BI_STATUS static inline int bi_status_to_errno(blk_status_t status) { switch (status) { case BLK_STS_OK: return (0); case BLK_STS_NOTSUPP: return (EOPNOTSUPP); case BLK_STS_TIMEOUT: return (ETIMEDOUT); case BLK_STS_NOSPC: return (ENOSPC); case BLK_STS_TRANSPORT: return (ENOLINK); case BLK_STS_TARGET: return (EREMOTEIO); #ifdef HAVE_BLK_STS_RESV_CONFLICT case BLK_STS_RESV_CONFLICT: #else case BLK_STS_NEXUS: #endif return (EBADE); case BLK_STS_MEDIUM: return (ENODATA); case BLK_STS_PROTECTION: return (EILSEQ); case BLK_STS_RESOURCE: return (ENOMEM); case BLK_STS_AGAIN: return (EAGAIN); case BLK_STS_IOERR: return (EIO); default: return (EIO); } } static inline blk_status_t errno_to_bi_status(int error) { switch (error) { case 0: return (BLK_STS_OK); case EOPNOTSUPP: return (BLK_STS_NOTSUPP); case ETIMEDOUT: return (BLK_STS_TIMEOUT); case ENOSPC: return (BLK_STS_NOSPC); case ENOLINK: return (BLK_STS_TRANSPORT); case EREMOTEIO: return (BLK_STS_TARGET); case EBADE: #ifdef HAVE_BLK_STS_RESV_CONFLICT return (BLK_STS_RESV_CONFLICT); #else return (BLK_STS_NEXUS); #endif case ENODATA: return (BLK_STS_MEDIUM); case EILSEQ: return (BLK_STS_PROTECTION); case ENOMEM: return (BLK_STS_RESOURCE); case EAGAIN: return (BLK_STS_AGAIN); case EIO: return (BLK_STS_IOERR); default: return (BLK_STS_IOERR); } } #endif /* HAVE_BIO_BI_STATUS */ /* * 4.3 API change * The bio_endio() prototype changed slightly. These are helper * macro's to ensure the prototype and invocation are handled. */ #ifdef HAVE_1ARG_BIO_END_IO_T #ifdef HAVE_BIO_BI_STATUS #define BIO_END_IO_ERROR(bio) bi_status_to_errno(bio->bi_status) #define BIO_END_IO_PROTO(fn, x, z) static void fn(struct bio *x) #define BIO_END_IO(bio, error) bio_set_bi_status(bio, error) static inline void bio_set_bi_status(struct bio *bio, int error) { ASSERT3S(error, <=, 0); bio->bi_status = errno_to_bi_status(-error); bio_endio(bio); } #else #define BIO_END_IO_ERROR(bio) (-(bio->bi_error)) #define BIO_END_IO_PROTO(fn, x, z) static void fn(struct bio *x) #define BIO_END_IO(bio, error) bio_set_bi_error(bio, error) static inline void bio_set_bi_error(struct bio *bio, int error) { ASSERT3S(error, <=, 0); bio->bi_error = error; bio_endio(bio); } #endif /* HAVE_BIO_BI_STATUS */ #else #define BIO_END_IO_PROTO(fn, x, z) static void fn(struct bio *x, int z) #define BIO_END_IO(bio, error) bio_endio(bio, error); #endif /* HAVE_1ARG_BIO_END_IO_T */ /* * 5.15 MACRO, * GD_DEAD * * 2.6.36 - 5.14 MACRO, * GENHD_FL_UP * * Check the disk status and return B_TRUE if alive * otherwise B_FALSE */ static inline boolean_t zfs_check_disk_status(struct block_device *bdev) { #if defined(GENHD_FL_UP) return (!!(bdev->bd_disk->flags & GENHD_FL_UP)); #elif defined(GD_DEAD) return (!test_bit(GD_DEAD, &bdev->bd_disk->state)); #else /* * This is encountered if neither GENHD_FL_UP nor GD_DEAD is available in * the kernel - likely due to an MACRO change that needs to be chased down. */ #error "Unsupported kernel: no usable disk status check" #endif } /* * 4.1 API, * 3.10.0 CentOS 7.x API, * blkdev_reread_part() * * For older kernels trigger a re-reading of the partition table by calling * check_disk_change() which calls flush_disk() to invalidate the device. * * For newer kernels (as of 5.10), bdev_check_media_change is used, in favor of * check_disk_change(), with the modification that invalidation is no longer * forced. */ #ifdef HAVE_CHECK_DISK_CHANGE #define zfs_check_media_change(bdev) check_disk_change(bdev) #ifdef HAVE_BLKDEV_REREAD_PART #define vdev_bdev_reread_part(bdev) blkdev_reread_part(bdev) #else #define vdev_bdev_reread_part(bdev) check_disk_change(bdev) #endif /* HAVE_BLKDEV_REREAD_PART */ #else #ifdef HAVE_BDEV_CHECK_MEDIA_CHANGE static inline int zfs_check_media_change(struct block_device *bdev) { #ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK struct gendisk *gd = bdev->bd_disk; const struct block_device_operations *bdo = gd->fops; #endif if (!bdev_check_media_change(bdev)) return (0); #ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK /* * Force revalidation, to mimic the old behavior of * check_disk_change() */ if (bdo->revalidate_disk) bdo->revalidate_disk(gd); #endif return (0); } #define vdev_bdev_reread_part(bdev) zfs_check_media_change(bdev) #elif defined(HAVE_DISK_CHECK_MEDIA_CHANGE) #define vdev_bdev_reread_part(bdev) disk_check_media_change(bdev->bd_disk) #define zfs_check_media_change(bdev) disk_check_media_change(bdev->bd_disk) #else /* * This is encountered if check_disk_change() and bdev_check_media_change() * are not available in the kernel - likely due to an API change that needs * to be chased down. */ #error "Unsupported kernel: no usable disk change check" #endif /* HAVE_BDEV_CHECK_MEDIA_CHANGE */ #endif /* HAVE_CHECK_DISK_CHANGE */ /* * 2.6.27 API change * The function was exported for use, prior to this it existed but the * symbol was not exported. * * 4.4.0-6.21 API change for Ubuntu * lookup_bdev() gained a second argument, FMODE_*, to check inode permissions. * * 5.11 API change * Changed to take a dev_t argument which is set on success and return a * non-zero error code on failure. */ static inline int vdev_lookup_bdev(const char *path, dev_t *dev) { #if defined(HAVE_DEVT_LOOKUP_BDEV) return (lookup_bdev(path, dev)); #elif defined(HAVE_1ARG_LOOKUP_BDEV) struct block_device *bdev = lookup_bdev(path); if (IS_ERR(bdev)) return (PTR_ERR(bdev)); *dev = bdev->bd_dev; bdput(bdev); return (0); #elif defined(HAVE_MODE_LOOKUP_BDEV) struct block_device *bdev = lookup_bdev(path, FMODE_READ); if (IS_ERR(bdev)) return (PTR_ERR(bdev)); *dev = bdev->bd_dev; bdput(bdev); return (0); #else #error "Unsupported kernel" #endif } +#if defined(HAVE_BLK_MODE_T) +#define blk_mode_is_open_write(flag) ((flag) & BLK_OPEN_WRITE) +#else +#define blk_mode_is_open_write(flag) ((flag) & FMODE_WRITE) +#endif + /* * Kernels without bio_set_op_attrs use bi_rw for the bio flags. */ #if !defined(HAVE_BIO_SET_OP_ATTRS) static inline void bio_set_op_attrs(struct bio *bio, unsigned rw, unsigned flags) { #if defined(HAVE_BIO_BI_OPF) bio->bi_opf = rw | flags; #else bio->bi_rw |= rw | flags; #endif /* HAVE_BIO_BI_OPF */ } #endif /* * bio_set_flush - Set the appropriate flags in a bio to guarantee * data are on non-volatile media on completion. * * 2.6.37 - 4.8 API, * Introduce WRITE_FLUSH, WRITE_FUA, and WRITE_FLUSH_FUA flags as a * replacement for WRITE_BARRIER to allow expressing richer semantics * to the block layer. It's up to the block layer to implement the * semantics correctly. Use the WRITE_FLUSH_FUA flag combination. * * 4.8 - 4.9 API, * REQ_FLUSH was renamed to REQ_PREFLUSH. For consistency with previous * OpenZFS releases, prefer the WRITE_FLUSH_FUA flag set if it's available. * * 4.10 API, * The read/write flags and their modifiers, including WRITE_FLUSH, * WRITE_FUA and WRITE_FLUSH_FUA were removed from fs.h in * torvalds/linux@70fd7614 and replaced by direct flag modification * of the REQ_ flags in bio->bi_opf. Use REQ_PREFLUSH. */ static inline void bio_set_flush(struct bio *bio) { #if defined(HAVE_REQ_PREFLUSH) /* >= 4.10 */ bio_set_op_attrs(bio, 0, REQ_PREFLUSH | REQ_OP_WRITE); #elif defined(WRITE_FLUSH_FUA) /* >= 2.6.37 and <= 4.9 */ bio_set_op_attrs(bio, 0, WRITE_FLUSH_FUA); #else #error "Allowing the build will cause bio_set_flush requests to be ignored." #endif } /* * 4.8 API, * REQ_OP_FLUSH * * 4.8-rc0 - 4.8-rc1, * REQ_PREFLUSH * * 2.6.36 - 4.7 API, * REQ_FLUSH * * in all cases but may have a performance impact for some kernels. It * has the advantage of minimizing kernel specific changes in the zvol code. * */ static inline boolean_t bio_is_flush(struct bio *bio) { #if defined(HAVE_REQ_OP_FLUSH) && defined(HAVE_BIO_BI_OPF) return ((bio_op(bio) == REQ_OP_FLUSH) || (bio->bi_opf & REQ_PREFLUSH)); #elif defined(HAVE_REQ_PREFLUSH) && defined(HAVE_BIO_BI_OPF) return (bio->bi_opf & REQ_PREFLUSH); #elif defined(HAVE_REQ_PREFLUSH) && !defined(HAVE_BIO_BI_OPF) return (bio->bi_rw & REQ_PREFLUSH); #elif defined(HAVE_REQ_FLUSH) return (bio->bi_rw & REQ_FLUSH); #else #error "Unsupported kernel" #endif } /* * 4.8 API, * REQ_FUA flag moved to bio->bi_opf * * 2.6.x - 4.7 API, * REQ_FUA */ static inline boolean_t bio_is_fua(struct bio *bio) { #if defined(HAVE_BIO_BI_OPF) return (bio->bi_opf & REQ_FUA); #elif defined(REQ_FUA) return (bio->bi_rw & REQ_FUA); #else #error "Allowing the build will cause fua requests to be ignored." #endif } /* * 4.8 API, * REQ_OP_DISCARD * * 2.6.36 - 4.7 API, * REQ_DISCARD * * In all cases the normal I/O path is used for discards. The only * difference is how the kernel tags individual I/Os as discards. */ static inline boolean_t bio_is_discard(struct bio *bio) { #if defined(HAVE_REQ_OP_DISCARD) return (bio_op(bio) == REQ_OP_DISCARD); #elif defined(HAVE_REQ_DISCARD) return (bio->bi_rw & REQ_DISCARD); #else #error "Unsupported kernel" #endif } /* * 4.8 API, * REQ_OP_SECURE_ERASE * * 2.6.36 - 4.7 API, * REQ_SECURE */ static inline boolean_t bio_is_secure_erase(struct bio *bio) { #if defined(HAVE_REQ_OP_SECURE_ERASE) return (bio_op(bio) == REQ_OP_SECURE_ERASE); #elif defined(REQ_SECURE) return (bio->bi_rw & REQ_SECURE); #else return (0); #endif } /* * 2.6.33 API change * Discard granularity and alignment restrictions may now be set. For * older kernels which do not support this it is safe to skip it. */ static inline void blk_queue_discard_granularity(struct request_queue *q, unsigned int dg) { q->limits.discard_granularity = dg; } /* * 5.19 API, * bdev_max_discard_sectors() * * 2.6.32 API, * blk_queue_discard() */ static inline boolean_t bdev_discard_supported(struct block_device *bdev) { #if defined(HAVE_BDEV_MAX_DISCARD_SECTORS) return (!!bdev_max_discard_sectors(bdev)); #elif defined(HAVE_BLK_QUEUE_DISCARD) return (!!blk_queue_discard(bdev_get_queue(bdev))); #else #error "Unsupported kernel" #endif } /* * 5.19 API, * bdev_max_secure_erase_sectors() * * 4.8 API, * blk_queue_secure_erase() * * 2.6.36 - 4.7 API, * blk_queue_secdiscard() */ static inline boolean_t bdev_secure_discard_supported(struct block_device *bdev) { #if defined(HAVE_BDEV_MAX_SECURE_ERASE_SECTORS) return (!!bdev_max_secure_erase_sectors(bdev)); #elif defined(HAVE_BLK_QUEUE_SECURE_ERASE) return (!!blk_queue_secure_erase(bdev_get_queue(bdev))); #elif defined(HAVE_BLK_QUEUE_SECDISCARD) return (!!blk_queue_secdiscard(bdev_get_queue(bdev))); #else #error "Unsupported kernel" #endif } /* * A common holder for vdev_bdev_open() is used to relax the exclusive open * semantics slightly. Internal vdev disk callers may pass VDEV_HOLDER to * allow them to open the device multiple times. Other kernel callers and * user space processes which don't pass this value will get EBUSY. This is * currently required for the correct operation of hot spares. */ #define VDEV_HOLDER ((void *)0x2401de7) static inline unsigned long blk_generic_start_io_acct(struct request_queue *q __attribute__((unused)), struct gendisk *disk __attribute__((unused)), int rw __attribute__((unused)), struct bio *bio) { #if defined(HAVE_BDEV_IO_ACCT_63) return (bdev_start_io_acct(bio->bi_bdev, bio_op(bio), jiffies)); #elif defined(HAVE_BDEV_IO_ACCT_OLD) return (bdev_start_io_acct(bio->bi_bdev, bio_sectors(bio), bio_op(bio), jiffies)); #elif defined(HAVE_DISK_IO_ACCT) return (disk_start_io_acct(disk, bio_sectors(bio), bio_op(bio))); #elif defined(HAVE_BIO_IO_ACCT) return (bio_start_io_acct(bio)); #elif defined(HAVE_GENERIC_IO_ACCT_3ARG) unsigned long start_time = jiffies; generic_start_io_acct(rw, bio_sectors(bio), &disk->part0); return (start_time); #elif defined(HAVE_GENERIC_IO_ACCT_4ARG) unsigned long start_time = jiffies; generic_start_io_acct(q, rw, bio_sectors(bio), &disk->part0); return (start_time); #else /* Unsupported */ return (0); #endif } static inline void blk_generic_end_io_acct(struct request_queue *q __attribute__((unused)), struct gendisk *disk __attribute__((unused)), int rw __attribute__((unused)), struct bio *bio, unsigned long start_time) { #if defined(HAVE_BDEV_IO_ACCT_63) bdev_end_io_acct(bio->bi_bdev, bio_op(bio), bio_sectors(bio), start_time); #elif defined(HAVE_BDEV_IO_ACCT_OLD) bdev_end_io_acct(bio->bi_bdev, bio_op(bio), start_time); #elif defined(HAVE_DISK_IO_ACCT) disk_end_io_acct(disk, bio_op(bio), start_time); #elif defined(HAVE_BIO_IO_ACCT) bio_end_io_acct(bio, start_time); #elif defined(HAVE_GENERIC_IO_ACCT_3ARG) generic_end_io_acct(rw, &disk->part0, start_time); #elif defined(HAVE_GENERIC_IO_ACCT_4ARG) generic_end_io_acct(q, rw, &disk->part0, start_time); #endif } #ifndef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS static inline struct request_queue * blk_generic_alloc_queue(make_request_fn make_request, int node_id) { #if defined(HAVE_BLK_ALLOC_QUEUE_REQUEST_FN) return (blk_alloc_queue(make_request, node_id)); #elif defined(HAVE_BLK_ALLOC_QUEUE_REQUEST_FN_RH) return (blk_alloc_queue_rh(make_request, node_id)); #else struct request_queue *q = blk_alloc_queue(GFP_KERNEL); if (q != NULL) blk_queue_make_request(q, make_request); return (q); #endif } #endif /* !HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */ #endif /* _ZFS_BLKDEV_H */ diff --git a/module/os/linux/zfs/vdev_disk.c b/module/os/linux/zfs/vdev_disk.c index 60b111c59f23..daf47a190594 100644 --- a/module/os/linux/zfs/vdev_disk.c +++ b/module/os/linux/zfs/vdev_disk.c @@ -1,1045 +1,1094 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC. * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER). * Rewritten for Linux by Brian Behlendorf . * LLNL-CODE-403049. * Copyright (c) 2012, 2019 by Delphix. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #ifdef HAVE_LINUX_BLK_CGROUP_HEADER #include #endif typedef struct vdev_disk { struct block_device *vd_bdev; krwlock_t vd_lock; } vdev_disk_t; /* * Unique identifier for the exclusive vdev holder. */ static void *zfs_vdev_holder = VDEV_HOLDER; /* * Wait up to zfs_vdev_open_timeout_ms milliseconds before determining the * device is missing. The missing path may be transient since the links * can be briefly removed and recreated in response to udev events. */ static uint_t zfs_vdev_open_timeout_ms = 1000; /* * Size of the "reserved" partition, in blocks. */ #define EFI_MIN_RESV_SIZE (16 * 1024) /* * Virtual device vector for disks. */ typedef struct dio_request { zio_t *dr_zio; /* Parent ZIO */ atomic_t dr_ref; /* References */ int dr_error; /* Bio error */ int dr_bio_count; /* Count of bio's */ struct bio *dr_bio[0]; /* Attached bio's */ } dio_request_t; +#ifdef HAVE_BLK_MODE_T +static blk_mode_t +#else static fmode_t +#endif vdev_bdev_mode(spa_mode_t spa_mode) { +#ifdef HAVE_BLK_MODE_T + blk_mode_t mode = 0; + + if (spa_mode & SPA_MODE_READ) + mode |= BLK_OPEN_READ; + + if (spa_mode & SPA_MODE_WRITE) + mode |= BLK_OPEN_WRITE; +#else fmode_t mode = 0; if (spa_mode & SPA_MODE_READ) mode |= FMODE_READ; if (spa_mode & SPA_MODE_WRITE) mode |= FMODE_WRITE; +#endif return (mode); } /* * Returns the usable capacity (in bytes) for the partition or disk. */ static uint64_t bdev_capacity(struct block_device *bdev) { return (i_size_read(bdev->bd_inode)); } #if !defined(HAVE_BDEV_WHOLE) static inline struct block_device * bdev_whole(struct block_device *bdev) { return (bdev->bd_contains); } #endif #if defined(HAVE_BDEVNAME) #define vdev_bdevname(bdev, name) bdevname(bdev, name) #else static inline void vdev_bdevname(struct block_device *bdev, char *name) { snprintf(name, BDEVNAME_SIZE, "%pg", bdev); } #endif /* * Returns the maximum expansion capacity of the block device (in bytes). * * It is possible to expand a vdev when it has been created as a wholedisk * and the containing block device has increased in capacity. Or when the * partition containing the pool has been manually increased in size. * * This function is only responsible for calculating the potential expansion * size so it can be reported by 'zpool list'. The efi_use_whole_disk() is * responsible for verifying the expected partition layout in the wholedisk * case, and updating the partition table if appropriate. Once the partition * size has been increased the additional capacity will be visible using * bdev_capacity(). * * The returned maximum expansion capacity is always expected to be larger, or * at the very least equal, to its usable capacity to prevent overestimating * the pool expandsize. */ static uint64_t bdev_max_capacity(struct block_device *bdev, uint64_t wholedisk) { uint64_t psize; int64_t available; if (wholedisk && bdev != bdev_whole(bdev)) { /* * When reporting maximum expansion capacity for a wholedisk * deduct any capacity which is expected to be lost due to * alignment restrictions. Over reporting this value isn't * harmful and would only result in slightly less capacity * than expected post expansion. * The estimated available space may be slightly smaller than * bdev_capacity() for devices where the number of sectors is * not a multiple of the alignment size and the partition layout * is keeping less than PARTITION_END_ALIGNMENT bytes after the * "reserved" EFI partition: in such cases return the device * usable capacity. */ available = i_size_read(bdev_whole(bdev)->bd_inode) - ((EFI_MIN_RESV_SIZE + NEW_START_BLOCK + PARTITION_END_ALIGNMENT) << SECTOR_BITS); psize = MAX(available, bdev_capacity(bdev)); } else { psize = bdev_capacity(bdev); } return (psize); } static void vdev_disk_error(zio_t *zio) { /* * This function can be called in interrupt context, for instance while * handling IRQs coming from a misbehaving disk device; use printk() * which is safe from any context. */ printk(KERN_WARNING "zio pool=%s vdev=%s error=%d type=%d " "offset=%llu size=%llu flags=%x\n", spa_name(zio->io_spa), zio->io_vd->vdev_path, zio->io_error, zio->io_type, (u_longlong_t)zio->io_offset, (u_longlong_t)zio->io_size, zio->io_flags); } static void vdev_disk_kobj_evt_post(vdev_t *v) { vdev_disk_t *vd = v->vdev_tsd; if (vd && vd->vd_bdev) { spl_signal_kobj_evt(vd->vd_bdev); } else { vdev_dbgmsg(v, "vdev_disk_t is NULL for VDEV:%s\n", v->vdev_path); } } +#if !defined(HAVE_BLKDEV_GET_BY_PATH_4ARG) +/* + * Define a dummy struct blk_holder_ops for kernel versions + * prior to 6.5. + */ +struct blk_holder_ops {}; +#endif + +static struct block_device * +vdev_blkdev_get_by_path(const char *path, spa_mode_t mode, void *holder, + const struct blk_holder_ops *hops) +{ +#ifdef HAVE_BLKDEV_GET_BY_PATH_4ARG + return (blkdev_get_by_path(path, + vdev_bdev_mode(mode) | BLK_OPEN_EXCL, holder, hops)); +#else + return (blkdev_get_by_path(path, + vdev_bdev_mode(mode) | FMODE_EXCL, holder)); +#endif +} + +static void +vdev_blkdev_put(struct block_device *bdev, spa_mode_t mode, void *holder) +{ +#ifdef HAVE_BLKDEV_PUT_HOLDER + return (blkdev_put(bdev, holder)); +#else + return (blkdev_put(bdev, vdev_bdev_mode(mode) | FMODE_EXCL)); +#endif +} + static int vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize, uint64_t *logical_ashift, uint64_t *physical_ashift) { struct block_device *bdev; +#ifdef HAVE_BLK_MODE_T + blk_mode_t mode = vdev_bdev_mode(spa_mode(v->vdev_spa)); +#else fmode_t mode = vdev_bdev_mode(spa_mode(v->vdev_spa)); +#endif hrtime_t timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms); vdev_disk_t *vd; /* Must have a pathname and it must be absolute. */ if (v->vdev_path == NULL || v->vdev_path[0] != '/') { v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; vdev_dbgmsg(v, "invalid vdev_path"); return (SET_ERROR(EINVAL)); } /* * Reopen the device if it is currently open. When expanding a * partition force re-scanning the partition table if userland * did not take care of this already. We need to do this while closed * in order to get an accurate updated block device size. Then * since udev may need to recreate the device links increase the * open retry timeout before reporting the device as unavailable. */ vd = v->vdev_tsd; if (vd) { char disk_name[BDEVNAME_SIZE + 6] = "/dev/"; boolean_t reread_part = B_FALSE; rw_enter(&vd->vd_lock, RW_WRITER); bdev = vd->vd_bdev; vd->vd_bdev = NULL; if (bdev) { if (v->vdev_expanding && bdev != bdev_whole(bdev)) { vdev_bdevname(bdev_whole(bdev), disk_name + 5); /* * If userland has BLKPG_RESIZE_PARTITION, * then it should have updated the partition * table already. We can detect this by * comparing our current physical size * with that of the device. If they are * the same, then we must not have * BLKPG_RESIZE_PARTITION or it failed to * update the partition table online. We * fallback to rescanning the partition * table from the kernel below. However, * if the capacity already reflects the * updated partition, then we skip * rescanning the partition table here. */ if (v->vdev_psize == bdev_capacity(bdev)) reread_part = B_TRUE; } - blkdev_put(bdev, mode | FMODE_EXCL); + vdev_blkdev_put(bdev, mode, zfs_vdev_holder); } if (reread_part) { - bdev = blkdev_get_by_path(disk_name, mode | FMODE_EXCL, - zfs_vdev_holder); + bdev = vdev_blkdev_get_by_path(disk_name, mode, + zfs_vdev_holder, NULL); if (!IS_ERR(bdev)) { int error = vdev_bdev_reread_part(bdev); - blkdev_put(bdev, mode | FMODE_EXCL); + vdev_blkdev_put(bdev, mode, zfs_vdev_holder); if (error == 0) { timeout = MSEC2NSEC( zfs_vdev_open_timeout_ms * 2); } } } } else { vd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP); rw_init(&vd->vd_lock, NULL, RW_DEFAULT, NULL); rw_enter(&vd->vd_lock, RW_WRITER); } /* * Devices are always opened by the path provided at configuration * time. This means that if the provided path is a udev by-id path * then drives may be re-cabled without an issue. If the provided * path is a udev by-path path, then the physical location information * will be preserved. This can be critical for more complicated * configurations where drives are located in specific physical * locations to maximize the systems tolerance to component failure. * * Alternatively, you can provide your own udev rule to flexibly map * the drives as you see fit. It is not advised that you use the * /dev/[hd]d devices which may be reordered due to probing order. * Devices in the wrong locations will be detected by the higher * level vdev validation. * * The specified paths may be briefly removed and recreated in * response to udev events. This should be exceptionally unlikely * because the zpool command makes every effort to verify these paths * have already settled prior to reaching this point. Therefore, * a ENOENT failure at this point is highly likely to be transient * and it is reasonable to sleep and retry before giving up. In * practice delays have been observed to be on the order of 100ms. * * When ERESTARTSYS is returned it indicates the block device is * a zvol which could not be opened due to the deadlock detection * logic in zvol_open(). Extend the timeout and retry the open * subsequent attempts are expected to eventually succeed. */ hrtime_t start = gethrtime(); bdev = ERR_PTR(-ENXIO); while (IS_ERR(bdev) && ((gethrtime() - start) < timeout)) { - bdev = blkdev_get_by_path(v->vdev_path, mode | FMODE_EXCL, - zfs_vdev_holder); + bdev = vdev_blkdev_get_by_path(v->vdev_path, mode, + zfs_vdev_holder, NULL); if (unlikely(PTR_ERR(bdev) == -ENOENT)) { /* * There is no point of waiting since device is removed * explicitly */ if (v->vdev_removed) break; schedule_timeout(MSEC_TO_TICK(10)); } else if (unlikely(PTR_ERR(bdev) == -ERESTARTSYS)) { timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms * 10); continue; } else if (IS_ERR(bdev)) { break; } } if (IS_ERR(bdev)) { int error = -PTR_ERR(bdev); vdev_dbgmsg(v, "open error=%d timeout=%llu/%llu", error, (u_longlong_t)(gethrtime() - start), (u_longlong_t)timeout); vd->vd_bdev = NULL; v->vdev_tsd = vd; rw_exit(&vd->vd_lock); return (SET_ERROR(error)); } else { vd->vd_bdev = bdev; v->vdev_tsd = vd; rw_exit(&vd->vd_lock); } /* Determine the physical block size */ int physical_block_size = bdev_physical_block_size(vd->vd_bdev); /* Determine the logical block size */ int logical_block_size = bdev_logical_block_size(vd->vd_bdev); /* Clear the nowritecache bit, causes vdev_reopen() to try again. */ v->vdev_nowritecache = B_FALSE; /* Set when device reports it supports TRIM. */ v->vdev_has_trim = bdev_discard_supported(vd->vd_bdev); /* Set when device reports it supports secure TRIM. */ v->vdev_has_securetrim = bdev_secure_discard_supported(vd->vd_bdev); /* Inform the ZIO pipeline that we are non-rotational */ v->vdev_nonrot = blk_queue_nonrot(bdev_get_queue(vd->vd_bdev)); /* Physical volume size in bytes for the partition */ *psize = bdev_capacity(vd->vd_bdev); /* Physical volume size in bytes including possible expansion space */ *max_psize = bdev_max_capacity(vd->vd_bdev, v->vdev_wholedisk); /* Based on the minimum sector size set the block size */ *physical_ashift = highbit64(MAX(physical_block_size, SPA_MINBLOCKSIZE)) - 1; *logical_ashift = highbit64(MAX(logical_block_size, SPA_MINBLOCKSIZE)) - 1; return (0); } static void vdev_disk_close(vdev_t *v) { vdev_disk_t *vd = v->vdev_tsd; if (v->vdev_reopening || vd == NULL) return; if (vd->vd_bdev != NULL) { - blkdev_put(vd->vd_bdev, - vdev_bdev_mode(spa_mode(v->vdev_spa)) | FMODE_EXCL); + vdev_blkdev_put(vd->vd_bdev, spa_mode(v->vdev_spa), + zfs_vdev_holder); } rw_destroy(&vd->vd_lock); kmem_free(vd, sizeof (vdev_disk_t)); v->vdev_tsd = NULL; } static dio_request_t * vdev_disk_dio_alloc(int bio_count) { dio_request_t *dr = kmem_zalloc(sizeof (dio_request_t) + sizeof (struct bio *) * bio_count, KM_SLEEP); atomic_set(&dr->dr_ref, 0); dr->dr_bio_count = bio_count; dr->dr_error = 0; for (int i = 0; i < dr->dr_bio_count; i++) dr->dr_bio[i] = NULL; return (dr); } static void vdev_disk_dio_free(dio_request_t *dr) { int i; for (i = 0; i < dr->dr_bio_count; i++) if (dr->dr_bio[i]) bio_put(dr->dr_bio[i]); kmem_free(dr, sizeof (dio_request_t) + sizeof (struct bio *) * dr->dr_bio_count); } static void vdev_disk_dio_get(dio_request_t *dr) { atomic_inc(&dr->dr_ref); } static int vdev_disk_dio_put(dio_request_t *dr) { int rc = atomic_dec_return(&dr->dr_ref); /* * Free the dio_request when the last reference is dropped and * ensure zio_interpret is called only once with the correct zio */ if (rc == 0) { zio_t *zio = dr->dr_zio; int error = dr->dr_error; vdev_disk_dio_free(dr); if (zio) { zio->io_error = error; ASSERT3S(zio->io_error, >=, 0); if (zio->io_error) vdev_disk_error(zio); zio_delay_interrupt(zio); } } return (rc); } BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error) { dio_request_t *dr = bio->bi_private; int rc; if (dr->dr_error == 0) { #ifdef HAVE_1ARG_BIO_END_IO_T dr->dr_error = BIO_END_IO_ERROR(bio); #else if (error) dr->dr_error = -(error); else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) dr->dr_error = EIO; #endif } /* Drop reference acquired by __vdev_disk_physio */ rc = vdev_disk_dio_put(dr); } static inline void vdev_submit_bio_impl(struct bio *bio) { #ifdef HAVE_1ARG_SUBMIT_BIO (void) submit_bio(bio); #else (void) submit_bio(bio_data_dir(bio), bio); #endif } /* * preempt_schedule_notrace is GPL-only which breaks the ZFS build, so * replace it with preempt_schedule under the following condition: */ #if defined(CONFIG_ARM64) && \ defined(CONFIG_PREEMPTION) && \ defined(CONFIG_BLK_CGROUP) #define preempt_schedule_notrace(x) preempt_schedule(x) #endif /* * As for the Linux 5.18 kernel bio_alloc() expects a block_device struct * as an argument removing the need to set it with bio_set_dev(). This * removes the need for all of the following compatibility code. */ #if !defined(HAVE_BIO_ALLOC_4ARG) #ifdef HAVE_BIO_SET_DEV #if defined(CONFIG_BLK_CGROUP) && defined(HAVE_BIO_SET_DEV_GPL_ONLY) /* * The Linux 5.5 kernel updated percpu_ref_tryget() which is inlined by * blkg_tryget() to use rcu_read_lock() instead of rcu_read_lock_sched(). * As a side effect the function was converted to GPL-only. Define our * own version when needed which uses rcu_read_lock_sched(). * * The Linux 5.17 kernel split linux/blk-cgroup.h into a private and a public * part, moving blkg_tryget into the private one. Define our own version. */ #if defined(HAVE_BLKG_TRYGET_GPL_ONLY) || !defined(HAVE_BLKG_TRYGET) static inline bool vdev_blkg_tryget(struct blkcg_gq *blkg) { struct percpu_ref *ref = &blkg->refcnt; unsigned long __percpu *count; bool rc; rcu_read_lock_sched(); if (__ref_is_percpu(ref, &count)) { this_cpu_inc(*count); rc = true; } else { #ifdef ZFS_PERCPU_REF_COUNT_IN_DATA rc = atomic_long_inc_not_zero(&ref->data->count); #else rc = atomic_long_inc_not_zero(&ref->count); #endif } rcu_read_unlock_sched(); return (rc); } #else #define vdev_blkg_tryget(bg) blkg_tryget(bg) #endif #ifdef HAVE_BIO_SET_DEV_MACRO /* * The Linux 5.0 kernel updated the bio_set_dev() macro so it calls the * GPL-only bio_associate_blkg() symbol thus inadvertently converting * the entire macro. Provide a minimal version which always assigns the * request queue's root_blkg to the bio. */ static inline void vdev_bio_associate_blkg(struct bio *bio) { #if defined(HAVE_BIO_BDEV_DISK) struct request_queue *q = bio->bi_bdev->bd_disk->queue; #else struct request_queue *q = bio->bi_disk->queue; #endif ASSERT3P(q, !=, NULL); ASSERT3P(bio->bi_blkg, ==, NULL); if (q->root_blkg && vdev_blkg_tryget(q->root_blkg)) bio->bi_blkg = q->root_blkg; } #define bio_associate_blkg vdev_bio_associate_blkg #else static inline void vdev_bio_set_dev(struct bio *bio, struct block_device *bdev) { #if defined(HAVE_BIO_BDEV_DISK) struct request_queue *q = bdev->bd_disk->queue; #else struct request_queue *q = bio->bi_disk->queue; #endif bio_clear_flag(bio, BIO_REMAPPED); if (bio->bi_bdev != bdev) bio_clear_flag(bio, BIO_THROTTLED); bio->bi_bdev = bdev; ASSERT3P(q, !=, NULL); ASSERT3P(bio->bi_blkg, ==, NULL); if (q->root_blkg && vdev_blkg_tryget(q->root_blkg)) bio->bi_blkg = q->root_blkg; } #define bio_set_dev vdev_bio_set_dev #endif #endif #else /* * Provide a bio_set_dev() helper macro for pre-Linux 4.14 kernels. */ static inline void bio_set_dev(struct bio *bio, struct block_device *bdev) { bio->bi_bdev = bdev; } #endif /* HAVE_BIO_SET_DEV */ #endif /* !HAVE_BIO_ALLOC_4ARG */ static inline void vdev_submit_bio(struct bio *bio) { struct bio_list *bio_list = current->bio_list; current->bio_list = NULL; vdev_submit_bio_impl(bio); current->bio_list = bio_list; } static inline struct bio * vdev_bio_alloc(struct block_device *bdev, gfp_t gfp_mask, unsigned short nr_vecs) { struct bio *bio; #ifdef HAVE_BIO_ALLOC_4ARG bio = bio_alloc(bdev, nr_vecs, 0, gfp_mask); #else bio = bio_alloc(gfp_mask, nr_vecs); if (likely(bio != NULL)) bio_set_dev(bio, bdev); #endif return (bio); } static inline unsigned int vdev_bio_max_segs(zio_t *zio, int bio_size, uint64_t abd_offset) { unsigned long nr_segs = abd_nr_pages_off(zio->io_abd, bio_size, abd_offset); #ifdef HAVE_BIO_MAX_SEGS return (bio_max_segs(nr_segs)); #else return (MIN(nr_segs, BIO_MAX_PAGES)); #endif } static int __vdev_disk_physio(struct block_device *bdev, zio_t *zio, size_t io_size, uint64_t io_offset, int rw, int flags) { dio_request_t *dr; uint64_t abd_offset; uint64_t bio_offset; int bio_size; int bio_count = 16; int error = 0; struct blk_plug plug; unsigned short nr_vecs; /* * Accessing outside the block device is never allowed. */ if (io_offset + io_size > bdev->bd_inode->i_size) { vdev_dbgmsg(zio->io_vd, "Illegal access %llu size %llu, device size %llu", io_offset, io_size, i_size_read(bdev->bd_inode)); return (SET_ERROR(EIO)); } retry: dr = vdev_disk_dio_alloc(bio_count); if (zio && !(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD))) bio_set_flags_failfast(bdev, &flags); dr->dr_zio = zio; /* * Since bio's can have up to BIO_MAX_PAGES=256 iovec's, each of which * is at least 512 bytes and at most PAGESIZE (typically 4K), one bio * can cover at least 128KB and at most 1MB. When the required number * of iovec's exceeds this, we are forced to break the IO in multiple * bio's and wait for them all to complete. This is likely if the * recordsize property is increased beyond 1MB. The default * bio_count=16 should typically accommodate the maximum-size zio of * 16MB. */ abd_offset = 0; bio_offset = io_offset; bio_size = io_size; for (int i = 0; i <= dr->dr_bio_count; i++) { /* Finished constructing bio's for given buffer */ if (bio_size <= 0) break; /* * If additional bio's are required, we have to retry, but * this should be rare - see the comment above. */ if (dr->dr_bio_count == i) { vdev_disk_dio_free(dr); bio_count *= 2; goto retry; } nr_vecs = vdev_bio_max_segs(zio, bio_size, abd_offset); dr->dr_bio[i] = vdev_bio_alloc(bdev, GFP_NOIO, nr_vecs); if (unlikely(dr->dr_bio[i] == NULL)) { vdev_disk_dio_free(dr); return (SET_ERROR(ENOMEM)); } /* Matching put called by vdev_disk_physio_completion */ vdev_disk_dio_get(dr); BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9; dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion; dr->dr_bio[i]->bi_private = dr; bio_set_op_attrs(dr->dr_bio[i], rw, flags); /* Remaining size is returned to become the new size */ bio_size = abd_bio_map_off(dr->dr_bio[i], zio->io_abd, bio_size, abd_offset); /* Advance in buffer and construct another bio if needed */ abd_offset += BIO_BI_SIZE(dr->dr_bio[i]); bio_offset += BIO_BI_SIZE(dr->dr_bio[i]); } /* Extra reference to protect dio_request during vdev_submit_bio */ vdev_disk_dio_get(dr); if (dr->dr_bio_count > 1) blk_start_plug(&plug); /* Submit all bio's associated with this dio */ for (int i = 0; i < dr->dr_bio_count; i++) { if (dr->dr_bio[i]) vdev_submit_bio(dr->dr_bio[i]); } if (dr->dr_bio_count > 1) blk_finish_plug(&plug); (void) vdev_disk_dio_put(dr); return (error); } BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, error) { zio_t *zio = bio->bi_private; #ifdef HAVE_1ARG_BIO_END_IO_T zio->io_error = BIO_END_IO_ERROR(bio); #else zio->io_error = -error; #endif if (zio->io_error && (zio->io_error == EOPNOTSUPP)) zio->io_vd->vdev_nowritecache = B_TRUE; bio_put(bio); ASSERT3S(zio->io_error, >=, 0); if (zio->io_error) vdev_disk_error(zio); zio_interrupt(zio); } static int vdev_disk_io_flush(struct block_device *bdev, zio_t *zio) { struct request_queue *q; struct bio *bio; q = bdev_get_queue(bdev); if (!q) return (SET_ERROR(ENXIO)); bio = vdev_bio_alloc(bdev, GFP_NOIO, 0); if (unlikely(bio == NULL)) return (SET_ERROR(ENOMEM)); bio->bi_end_io = vdev_disk_io_flush_completion; bio->bi_private = zio; bio_set_flush(bio); vdev_submit_bio(bio); invalidate_bdev(bdev); return (0); } static int vdev_disk_io_trim(zio_t *zio) { vdev_t *v = zio->io_vd; vdev_disk_t *vd = v->vdev_tsd; #if defined(HAVE_BLKDEV_ISSUE_SECURE_ERASE) if (zio->io_trim_flags & ZIO_TRIM_SECURE) { return (-blkdev_issue_secure_erase(vd->vd_bdev, zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS)); } else { return (-blkdev_issue_discard(vd->vd_bdev, zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS)); } #elif defined(HAVE_BLKDEV_ISSUE_DISCARD) unsigned long trim_flags = 0; #if defined(BLKDEV_DISCARD_SECURE) if (zio->io_trim_flags & ZIO_TRIM_SECURE) trim_flags |= BLKDEV_DISCARD_SECURE; #endif return (-blkdev_issue_discard(vd->vd_bdev, zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS, trim_flags)); #else #error "Unsupported kernel" #endif } static void vdev_disk_io_start(zio_t *zio) { vdev_t *v = zio->io_vd; vdev_disk_t *vd = v->vdev_tsd; int rw, error; /* * If the vdev is closed, it's likely in the REMOVED or FAULTED state. * Nothing to be done here but return failure. */ if (vd == NULL) { zio->io_error = ENXIO; zio_interrupt(zio); return; } rw_enter(&vd->vd_lock, RW_READER); /* * If the vdev is closed, it's likely due to a failed reopen and is * in the UNAVAIL state. Nothing to be done here but return failure. */ if (vd->vd_bdev == NULL) { rw_exit(&vd->vd_lock); zio->io_error = ENXIO; zio_interrupt(zio); return; } switch (zio->io_type) { case ZIO_TYPE_IOCTL: if (!vdev_readable(v)) { rw_exit(&vd->vd_lock); zio->io_error = SET_ERROR(ENXIO); zio_interrupt(zio); return; } switch (zio->io_cmd) { case DKIOCFLUSHWRITECACHE: if (zfs_nocacheflush) break; if (v->vdev_nowritecache) { zio->io_error = SET_ERROR(ENOTSUP); break; } error = vdev_disk_io_flush(vd->vd_bdev, zio); if (error == 0) { rw_exit(&vd->vd_lock); return; } zio->io_error = error; break; default: zio->io_error = SET_ERROR(ENOTSUP); } rw_exit(&vd->vd_lock); zio_execute(zio); return; case ZIO_TYPE_WRITE: rw = WRITE; break; case ZIO_TYPE_READ: rw = READ; break; case ZIO_TYPE_TRIM: zio->io_error = vdev_disk_io_trim(zio); rw_exit(&vd->vd_lock); zio_interrupt(zio); return; default: rw_exit(&vd->vd_lock); zio->io_error = SET_ERROR(ENOTSUP); zio_interrupt(zio); return; } zio->io_target_timestamp = zio_handle_io_delay(zio); error = __vdev_disk_physio(vd->vd_bdev, zio, zio->io_size, zio->io_offset, rw, 0); rw_exit(&vd->vd_lock); if (error) { zio->io_error = error; zio_interrupt(zio); return; } } static void vdev_disk_io_done(zio_t *zio) { /* * If the device returned EIO, we revalidate the media. If it is * determined the media has changed this triggers the asynchronous * removal of the device from the configuration. */ if (zio->io_error == EIO) { vdev_t *v = zio->io_vd; vdev_disk_t *vd = v->vdev_tsd; if (!zfs_check_disk_status(vd->vd_bdev)) { invalidate_bdev(vd->vd_bdev); v->vdev_remove_wanted = B_TRUE; spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); } } } static void vdev_disk_hold(vdev_t *vd) { ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); /* We must have a pathname, and it must be absolute. */ if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') return; /* * Only prefetch path and devid info if the device has * never been opened. */ if (vd->vdev_tsd != NULL) return; } static void vdev_disk_rele(vdev_t *vd) { ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); /* XXX: Implement me as a vnode rele for the device */ } vdev_ops_t vdev_disk_ops = { .vdev_op_init = NULL, .vdev_op_fini = NULL, .vdev_op_open = vdev_disk_open, .vdev_op_close = vdev_disk_close, .vdev_op_asize = vdev_default_asize, .vdev_op_min_asize = vdev_default_min_asize, .vdev_op_min_alloc = NULL, .vdev_op_io_start = vdev_disk_io_start, .vdev_op_io_done = vdev_disk_io_done, .vdev_op_state_change = NULL, .vdev_op_need_resilver = NULL, .vdev_op_hold = vdev_disk_hold, .vdev_op_rele = vdev_disk_rele, .vdev_op_remap = NULL, .vdev_op_xlate = vdev_default_xlate, .vdev_op_rebuild_asize = NULL, .vdev_op_metaslab_init = NULL, .vdev_op_config_generate = NULL, .vdev_op_nparity = NULL, .vdev_op_ndisks = NULL, .vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */ .vdev_op_leaf = B_TRUE, /* leaf vdev */ .vdev_op_kobj_evt_post = vdev_disk_kobj_evt_post }; /* * The zfs_vdev_scheduler module option has been deprecated. Setting this * value no longer has any effect. It has not yet been entirely removed * to allow the module to be loaded if this option is specified in the * /etc/modprobe.d/zfs.conf file. The following warning will be logged. */ static int param_set_vdev_scheduler(const char *val, zfs_kernel_param_t *kp) { int error = param_set_charp(val, kp); if (error == 0) { printk(KERN_INFO "The 'zfs_vdev_scheduler' module option " "is not supported.\n"); } return (error); } char *zfs_vdev_scheduler = "unused"; module_param_call(zfs_vdev_scheduler, param_set_vdev_scheduler, param_get_charp, &zfs_vdev_scheduler, 0644); MODULE_PARM_DESC(zfs_vdev_scheduler, "I/O scheduler"); int param_set_min_auto_ashift(const char *buf, zfs_kernel_param_t *kp) { uint64_t val; int error; error = kstrtoull(buf, 0, &val); if (error < 0) return (SET_ERROR(error)); if (val < ASHIFT_MIN || val > zfs_vdev_max_auto_ashift) return (SET_ERROR(-EINVAL)); error = param_set_ulong(buf, kp); if (error < 0) return (SET_ERROR(error)); return (0); } int param_set_max_auto_ashift(const char *buf, zfs_kernel_param_t *kp) { uint64_t val; int error; error = kstrtoull(buf, 0, &val); if (error < 0) return (SET_ERROR(error)); if (val > ASHIFT_MAX || val < zfs_vdev_min_auto_ashift) return (SET_ERROR(-EINVAL)); error = param_set_ulong(buf, kp); if (error < 0) return (SET_ERROR(error)); return (0); } ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, open_timeout_ms, UINT, ZMOD_RW, "Timeout before determining that a device is missing"); diff --git a/module/os/linux/zfs/zfs_vnops_os.c b/module/os/linux/zfs/zfs_vnops_os.c index af0d553d516e..9c87a0d8cff7 100644 --- a/module/os/linux/zfs/zfs_vnops_os.c +++ b/module/os/linux/zfs/zfs_vnops_os.c @@ -1,4076 +1,4076 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2018 by Delphix. All rights reserved. * Copyright (c) 2015 by Chunwei Chen. All rights reserved. * Copyright 2017 Nexenta Systems, Inc. */ /* Portions Copyright 2007 Jeremy Teo */ /* Portions Copyright 2010 Robert Milkowski */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Programming rules. * * Each vnode op performs some logical unit of work. To do this, the ZPL must * properly lock its in-core state, create a DMU transaction, do the work, * record this work in the intent log (ZIL), commit the DMU transaction, * and wait for the intent log to commit if it is a synchronous operation. * Moreover, the vnode ops must work in both normal and log replay context. * The ordering of events is important to avoid deadlocks and references * to freed memory. The example below illustrates the following Big Rules: * * (1) A check must be made in each zfs thread for a mounted file system. * This is done avoiding races using ZFS_ENTER(zfsvfs). * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros * can return EIO from the calling function. * * (2) zrele() should always be the last thing except for zil_commit() (if * necessary) and ZFS_EXIT(). This is for 3 reasons: First, if it's the * last reference, the vnode/znode can be freed, so the zp may point to * freed memory. Second, the last reference will call zfs_zinactive(), * which may induce a lot of work -- pushing cached pages (which acquires * range locks) and syncing out cached atime changes. Third, * zfs_zinactive() may require a new tx, which could deadlock the system * if you were already holding one. This deadlock occurs because the tx * currently being operated on prevents a txg from syncing, which * prevents the new tx from progressing, resulting in a deadlock. If you * must call zrele() within a tx, use zfs_zrele_async(). Note that iput() * is a synonym for zrele(). * * (3) All range locks must be grabbed before calling dmu_tx_assign(), * as they can span dmu_tx_assign() calls. * * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to * dmu_tx_assign(). This is critical because we don't want to block * while holding locks. * * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This * reduces lock contention and CPU usage when we must wait (note that if * throughput is constrained by the storage, nearly every transaction * must wait). * * Note, in particular, that if a lock is sometimes acquired before * the tx assigns, and sometimes after (e.g. z_lock), then failing * to use a non-blocking assign can deadlock the system. The scenario: * * Thread A has grabbed a lock before calling dmu_tx_assign(). * Thread B is in an already-assigned tx, and blocks for this lock. * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() * forever, because the previous txg can't quiesce until B's tx commits. * * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, * then drop all locks, call dmu_tx_wait(), and try again. On subsequent * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT, * to indicate that this operation has already called dmu_tx_wait(). * This will ensure that we don't retry forever, waiting a short bit * each time. * * (5) If the operation succeeded, generate the intent log entry for it * before dropping locks. This ensures that the ordering of events * in the intent log matches the order in which they actually occurred. * During ZIL replay the zfs_log_* functions will update the sequence * number to indicate the zil transaction has replayed. * * (6) At the end of each vnode op, the DMU tx must always commit, * regardless of whether there were any errors. * * (7) After dropping all locks, invoke zil_commit(zilog, foid) * to ensure that synchronous semantics are provided when necessary. * * In general, this is how things should be ordered in each vnode op: * * ZFS_ENTER(zfsvfs); // exit if unmounted * top: * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) * rw_enter(...); // grab any other locks you need * tx = dmu_tx_create(...); // get DMU tx * dmu_tx_hold_*(); // hold each object you might modify * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); * if (error) { * rw_exit(...); // drop locks * zfs_dirent_unlock(dl); // unlock directory entry * zrele(...); // release held znodes * if (error == ERESTART) { * waited = B_TRUE; * dmu_tx_wait(tx); * dmu_tx_abort(tx); * goto top; * } * dmu_tx_abort(tx); // abort DMU tx * ZFS_EXIT(zfsvfs); // finished in zfs * return (error); // really out of space * } * error = do_real_work(); // do whatever this VOP does * if (error == 0) * zfs_log_*(...); // on success, make ZIL entry * dmu_tx_commit(tx); // commit DMU tx -- error or not * rw_exit(...); // drop locks * zfs_dirent_unlock(dl); // unlock directory entry * zrele(...); // release held znodes * zil_commit(zilog, foid); // synchronous when necessary * ZFS_EXIT(zfsvfs); // finished in zfs * return (error); // done, report error */ /* * Virus scanning is unsupported. It would be possible to add a hook * here to performance the required virus scan. This could be done * entirely in the kernel or potentially as an update to invoke a * scanning utility. */ static int zfs_vscan(struct inode *ip, cred_t *cr, int async) { return (0); } /* ARGSUSED */ int zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* Honor ZFS_APPENDONLY file attribute */ - if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && + if (blk_mode_is_open_write(mode) && (zp->z_pflags & ZFS_APPENDONLY) && ((flag & O_APPEND) == 0)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } /* Virus scan eligible files on open */ if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { if (zfs_vscan(ip, cr, 0) != 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EACCES)); } } /* Keep a count of the synchronous opens in the znode */ if (flag & O_SYNC) atomic_inc_32(&zp->z_sync_cnt); ZFS_EXIT(zfsvfs); return (0); } /* ARGSUSED */ int zfs_close(struct inode *ip, int flag, cred_t *cr) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* Decrement the synchronous opens in the znode */ if (flag & O_SYNC) atomic_dec_32(&zp->z_sync_cnt); if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) VERIFY(zfs_vscan(ip, cr, 1) == 0); ZFS_EXIT(zfsvfs); return (0); } #if defined(_KERNEL) static int zfs_fillpage(struct inode *ip, struct page *pp); /* * When a file is memory mapped, we must keep the IO data synchronized * between the DMU cache and the memory mapped pages. Update all mapped * pages with the contents of the coresponding dmu buffer. */ void update_pages(znode_t *zp, int64_t start, int len, objset_t *os) { struct address_space *mp = ZTOI(zp)->i_mapping; int64_t off = start & (PAGE_SIZE - 1); for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { uint64_t nbytes = MIN(PAGE_SIZE - off, len); struct page *pp = find_lock_page(mp, start >> PAGE_SHIFT); if (pp) { if (mapping_writably_mapped(mp)) flush_dcache_page(pp); void *pb = kmap(pp); int error = dmu_read(os, zp->z_id, start + off, nbytes, pb + off, DMU_READ_PREFETCH); kunmap(pp); if (error) { SetPageError(pp); ClearPageUptodate(pp); } else { ClearPageError(pp); SetPageUptodate(pp); if (mapping_writably_mapped(mp)) flush_dcache_page(pp); mark_page_accessed(pp); } unlock_page(pp); put_page(pp); } len -= nbytes; off = 0; } } /* * When a file is memory mapped, we must keep the I/O data synchronized * between the DMU cache and the memory mapped pages. Preferentially read * from memory mapped pages, otherwise fallback to reading through the dmu. */ int mappedread(znode_t *zp, int nbytes, zfs_uio_t *uio) { struct inode *ip = ZTOI(zp); struct address_space *mp = ip->i_mapping; int64_t start = uio->uio_loffset; int64_t off = start & (PAGE_SIZE - 1); int len = nbytes; int error = 0; for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { uint64_t bytes = MIN(PAGE_SIZE - off, len); struct page *pp = find_lock_page(mp, start >> PAGE_SHIFT); if (pp) { /* * If filemap_fault() retries there exists a window * where the page will be unlocked and not up to date. * In this case we must try and fill the page. */ if (unlikely(!PageUptodate(pp))) { error = zfs_fillpage(ip, pp); if (error) { unlock_page(pp); put_page(pp); return (error); } } ASSERT(PageUptodate(pp) || PageDirty(pp)); unlock_page(pp); void *pb = kmap(pp); error = zfs_uiomove(pb + off, bytes, UIO_READ, uio); kunmap(pp); if (mapping_writably_mapped(mp)) flush_dcache_page(pp); mark_page_accessed(pp); put_page(pp); } else { error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), uio, bytes); } len -= bytes; off = 0; if (error) break; } return (error); } #endif /* _KERNEL */ unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; /* * Write the bytes to a file. * * IN: zp - znode of file to be written to * data - bytes to write * len - number of bytes to write * pos - offset to start writing at * * OUT: resid - remaining bytes to write * * RETURN: 0 if success * positive error code if failure. EIO is returned * for a short write when residp isn't provided. * * Timestamps: * zp - ctime|mtime updated if byte count > 0 */ int zfs_write_simple(znode_t *zp, const void *data, size_t len, loff_t pos, size_t *residp) { fstrans_cookie_t cookie; int error; struct iovec iov; iov.iov_base = (void *)data; iov.iov_len = len; zfs_uio_t uio; zfs_uio_iovec_init(&uio, &iov, 1, pos, UIO_SYSSPACE, len, 0); cookie = spl_fstrans_mark(); error = zfs_write(zp, &uio, 0, kcred); spl_fstrans_unmark(cookie); if (error == 0) { if (residp != NULL) *residp = zfs_uio_resid(&uio); else if (zfs_uio_resid(&uio) != 0) error = SET_ERROR(EIO); } return (error); } static void zfs_rele_async_task(void *arg) { iput(arg); } void zfs_zrele_async(znode_t *zp) { struct inode *ip = ZTOI(zp); objset_t *os = ITOZSB(ip)->z_os; ASSERT(atomic_read(&ip->i_count) > 0); ASSERT(os != NULL); /* * If decrementing the count would put us at 0, we can't do it inline * here, because that would be synchronous. Instead, dispatch an iput * to run later. * * For more information on the dangers of a synchronous iput, see the * header comment of this file. */ if (!atomic_add_unless(&ip->i_count, -1, 1)) { VERIFY(taskq_dispatch(dsl_pool_zrele_taskq(dmu_objset_pool(os)), zfs_rele_async_task, ip, TQ_SLEEP) != TASKQID_INVALID); } } /* * Lookup an entry in a directory, or an extended attribute directory. * If it exists, return a held inode reference for it. * * IN: zdp - znode of directory to search. * nm - name of entry to lookup. * flags - LOOKUP_XATTR set if looking for an attribute. * cr - credentials of caller. * direntflags - directory lookup flags * realpnp - returned pathname. * * OUT: zpp - znode of located entry, NULL if not found. * * RETURN: 0 on success, error code on failure. * * Timestamps: * NA */ /* ARGSUSED */ int zfs_lookup(znode_t *zdp, char *nm, znode_t **zpp, int flags, cred_t *cr, int *direntflags, pathname_t *realpnp) { zfsvfs_t *zfsvfs = ZTOZSB(zdp); int error = 0; /* * Fast path lookup, however we must skip DNLC lookup * for case folding or normalizing lookups because the * DNLC code only stores the passed in name. This means * creating 'a' and removing 'A' on a case insensitive * file system would work, but DNLC still thinks 'a' * exists and won't let you create it again on the next * pass through fast path. */ if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { if (!S_ISDIR(ZTOI(zdp)->i_mode)) { return (SET_ERROR(ENOTDIR)); } else if (zdp->z_sa_hdl == NULL) { return (SET_ERROR(EIO)); } if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { error = zfs_fastaccesschk_execute(zdp, cr); if (!error) { *zpp = zdp; zhold(*zpp); return (0); } return (error); } } ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zdp); *zpp = NULL; if (flags & LOOKUP_XATTR) { /* * We don't allow recursive attributes.. * Maybe someday we will. */ if (zdp->z_pflags & ZFS_XATTR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if ((error = zfs_get_xattrdir(zdp, zpp, cr, flags))) { ZFS_EXIT(zfsvfs); return (error); } /* * Do we have permission to get into attribute directory? */ if ((error = zfs_zaccess(*zpp, ACE_EXECUTE, 0, B_FALSE, cr))) { zrele(*zpp); *zpp = NULL; } ZFS_EXIT(zfsvfs); return (error); } if (!S_ISDIR(ZTOI(zdp)->i_mode)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENOTDIR)); } /* * Check accessibility of directory. */ if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { ZFS_EXIT(zfsvfs); return (error); } if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } error = zfs_dirlook(zdp, nm, zpp, flags, direntflags, realpnp); if ((error == 0) && (*zpp)) zfs_znode_update_vfs(*zpp); ZFS_EXIT(zfsvfs); return (error); } /* * Attempt to create a new entry in a directory. If the entry * already exists, truncate the file if permissible, else return * an error. Return the ip of the created or trunc'd file. * * IN: dzp - znode of directory to put new file entry in. * name - name of new file entry. * vap - attributes of new file. * excl - flag indicating exclusive or non-exclusive mode. * mode - mode to open file with. * cr - credentials of caller. * flag - file flag. * vsecp - ACL to be set * * OUT: zpp - znode of created or trunc'd entry. * * RETURN: 0 on success, error code on failure. * * Timestamps: * dzp - ctime|mtime updated if new entry created * zp - ctime|mtime always, atime if new */ /* ARGSUSED */ int zfs_create(znode_t *dzp, char *name, vattr_t *vap, int excl, int mode, znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp) { znode_t *zp; zfsvfs_t *zfsvfs = ZTOZSB(dzp); zilog_t *zilog; objset_t *os; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; uid_t uid; gid_t gid; zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; boolean_t have_acl = B_FALSE; boolean_t waited = B_FALSE; boolean_t skip_acl = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; /* * If we have an ephemeral id, ACL, or XVATTR then * make sure file system is at proper version */ gid = crgetgid(cr); uid = crgetuid(cr); if (zfsvfs->z_use_fuids == B_FALSE && (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (SET_ERROR(EINVAL)); if (name == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); os = zfsvfs->z_os; zilog = zfsvfs->z_log; if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (vap->va_mask & ATTR_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, crgetuid(cr), cr, vap->va_mode)) != 0) { ZFS_EXIT(zfsvfs); return (error); } } top: *zpp = NULL; if (*name == '\0') { /* * Null component name refers to the directory itself. */ zhold(dzp); zp = dzp; dl = NULL; error = 0; } else { /* possible igrab(zp) */ int zflg = 0; if (flag & FIGNORECASE) zflg |= ZCILOOK; error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); if (error) { if (have_acl) zfs_acl_ids_free(&acl_ids); if (strcmp(name, "..") == 0) error = SET_ERROR(EISDIR); ZFS_EXIT(zfsvfs); return (error); } } if (zp == NULL) { uint64_t txtype; uint64_t projid = ZFS_DEFAULT_PROJID; /* * Create a new file object and update the directory * to reference it. */ if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, skip_acl, cr))) { if (have_acl) zfs_acl_ids_free(&acl_ids); goto out; } /* * We only support the creation of regular files in * extended attribute directories. */ if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { if (have_acl) zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EINVAL); goto out; } if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, &acl_ids)) != 0) goto out; have_acl = B_TRUE; if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) projid = zfs_inherit_projid(dzp); if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EDQUOT); goto out; } tx = dmu_tx_create(os); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); error = zfs_link_create(dl, zp, tx, ZNEW); if (error != 0) { /* * Since, we failed to add the directory entry for it, * delete the newly created dnode. */ zfs_znode_delete(zp, tx); remove_inode_hash(ZTOI(zp)); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); goto out; } if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); if (flag & FIGNORECASE) txtype |= TX_CI; zfs_log_create(zilog, tx, txtype, dzp, zp, name, vsecp, acl_ids.z_fuidp, vap); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); } else { int aflags = (flag & O_APPEND) ? V_APPEND : 0; if (have_acl) zfs_acl_ids_free(&acl_ids); have_acl = B_FALSE; /* * A directory entry already exists for this name. */ /* * Can't truncate an existing file if in exclusive mode. */ if (excl) { error = SET_ERROR(EEXIST); goto out; } /* * Can't open a directory for writing. */ if (S_ISDIR(ZTOI(zp)->i_mode)) { error = SET_ERROR(EISDIR); goto out; } /* * Verify requested access to file. */ if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { goto out; } mutex_enter(&dzp->z_lock); dzp->z_seq++; mutex_exit(&dzp->z_lock); /* * Truncate regular files if requested. */ if (S_ISREG(ZTOI(zp)->i_mode) && (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { /* we can't hold any locks when calling zfs_freesp() */ if (dl) { zfs_dirent_unlock(dl); dl = NULL; } error = zfs_freesp(zp, 0, 0, mode, TRUE); } } out: if (dl) zfs_dirent_unlock(dl); if (error) { if (zp) zrele(zp); } else { zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); *zpp = zp; } if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* ARGSUSED */ int zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) { znode_t *zp = NULL, *dzp = ITOZ(dip); zfsvfs_t *zfsvfs = ITOZSB(dip); objset_t *os; dmu_tx_t *tx; int error; uid_t uid; gid_t gid; zfs_acl_ids_t acl_ids; uint64_t projid = ZFS_DEFAULT_PROJID; boolean_t fuid_dirtied; boolean_t have_acl = B_FALSE; boolean_t waited = B_FALSE; /* * If we have an ephemeral id, ACL, or XVATTR then * make sure file system is at proper version */ gid = crgetgid(cr); uid = crgetuid(cr); if (zfsvfs->z_use_fuids == B_FALSE && (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); os = zfsvfs->z_os; if (vap->va_mask & ATTR_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, crgetuid(cr), cr, vap->va_mode)) != 0) { ZFS_EXIT(zfsvfs); return (error); } } top: *ipp = NULL; /* * Create a new file object and update the directory * to reference it. */ if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { if (have_acl) zfs_acl_ids_free(&acl_ids); goto out; } if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, &acl_ids)) != 0) goto out; have_acl = B_TRUE; if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) projid = zfs_inherit_projid(dzp); if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EDQUOT); goto out; } tx = dmu_tx_create(os); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); /* Add to unlinked set */ zp->z_unlinked = B_TRUE; zfs_unlinked_add(zp, tx); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); out: if (error) { if (zp) zrele(zp); } else { zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); *ipp = ZTOI(zp); } ZFS_EXIT(zfsvfs); return (error); } /* * Remove an entry from a directory. * * IN: dzp - znode of directory to remove entry from. * name - name of entry to remove. * cr - credentials of caller. * flags - case flags. * * RETURN: 0 if success * error code if failure * * Timestamps: * dzp - ctime|mtime * ip - ctime (if nlink > 0) */ uint64_t null_xattr = 0; /*ARGSUSED*/ int zfs_remove(znode_t *dzp, char *name, cred_t *cr, int flags) { znode_t *zp; znode_t *xzp; zfsvfs_t *zfsvfs = ZTOZSB(dzp); zilog_t *zilog; uint64_t acl_obj, xattr_obj; uint64_t xattr_obj_unlinked = 0; uint64_t obj = 0; uint64_t links; zfs_dirlock_t *dl; dmu_tx_t *tx; boolean_t may_delete_now, delete_now = FALSE; boolean_t unlinked, toobig = FALSE; uint64_t txtype; pathname_t *realnmp = NULL; pathname_t realnm; int error; int zflg = ZEXISTS; boolean_t waited = B_FALSE; if (name == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (flags & FIGNORECASE) { zflg |= ZCILOOK; pn_alloc(&realnm); realnmp = &realnm; } top: xattr_obj = 0; xzp = NULL; /* * Attempt to lock directory; fail if entry doesn't exist. */ if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, realnmp))) { if (realnmp) pn_free(realnmp); ZFS_EXIT(zfsvfs); return (error); } if ((error = zfs_zaccess_delete(dzp, zp, cr))) { goto out; } /* * Need to use rmdir for removing directories. */ if (S_ISDIR(ZTOI(zp)->i_mode)) { error = SET_ERROR(EPERM); goto out; } mutex_enter(&zp->z_lock); may_delete_now = atomic_read(&ZTOI(zp)->i_count) == 1 && !zn_has_cached_data(zp, 0, LLONG_MAX); mutex_exit(&zp->z_lock); /* * We may delete the znode now, or we may put it in the unlinked set; * it depends on whether we're the last link, and on whether there are * other holds on the inode. So we dmu_tx_hold() the right things to * allow for either case. */ obj = zp->z_id; tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); zfs_sa_upgrade_txholds(tx, dzp); if (may_delete_now) { toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; /* if the file is too big, only hold_free a token amount */ dmu_tx_hold_free(tx, zp->z_id, 0, (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); } /* are there any extended attributes? */ error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj, sizeof (xattr_obj)); if (error == 0 && xattr_obj) { error = zfs_zget(zfsvfs, xattr_obj, &xzp); ASSERT0(error); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); } mutex_enter(&zp->z_lock); if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); mutex_exit(&zp->z_lock); /* charge as an update -- would be nice not to charge at all */ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); /* * Mark this transaction as typically resulting in a net free of space */ dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); zrele(zp); if (xzp) zrele(xzp); goto top; } if (realnmp) pn_free(realnmp); dmu_tx_abort(tx); zrele(zp); if (xzp) zrele(xzp); ZFS_EXIT(zfsvfs); return (error); } /* * Remove the directory entry. */ error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); if (error) { dmu_tx_commit(tx); goto out; } if (unlinked) { /* * Hold z_lock so that we can make sure that the ACL obj * hasn't changed. Could have been deleted due to * zfs_sa_upgrade(). */ mutex_enter(&zp->z_lock); (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); delete_now = may_delete_now && !toobig && atomic_read(&ZTOI(zp)->i_count) == 1 && !zn_has_cached_data(zp, 0, LLONG_MAX) && xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == acl_obj; } if (delete_now) { if (xattr_obj_unlinked) { ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); mutex_enter(&xzp->z_lock); xzp->z_unlinked = B_TRUE; clear_nlink(ZTOI(xzp)); links = 0; error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), &links, sizeof (links), tx); ASSERT3U(error, ==, 0); mutex_exit(&xzp->z_lock); zfs_unlinked_add(xzp, tx); if (zp->z_is_sa) error = sa_remove(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), tx); else error = sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &null_xattr, sizeof (uint64_t), tx); ASSERT0(error); } /* * Add to the unlinked set because a new reference could be * taken concurrently resulting in a deferred destruction. */ zfs_unlinked_add(zp, tx); mutex_exit(&zp->z_lock); } else if (unlinked) { mutex_exit(&zp->z_lock); zfs_unlinked_add(zp, tx); } txtype = TX_REMOVE; if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked); dmu_tx_commit(tx); out: if (realnmp) pn_free(realnmp); zfs_dirent_unlock(dl); zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); if (delete_now) zrele(zp); else zfs_zrele_async(zp); if (xzp) { zfs_znode_update_vfs(xzp); zfs_zrele_async(xzp); } if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Create a new directory and insert it into dzp using the name * provided. Return a pointer to the inserted directory. * * IN: dzp - znode of directory to add subdir to. * dirname - name of new directory. * vap - attributes of new directory. * cr - credentials of caller. * flags - case flags. * vsecp - ACL to be set * * OUT: zpp - znode of created directory. * * RETURN: 0 if success * error code if failure * * Timestamps: * dzp - ctime|mtime updated * zpp - ctime|mtime|atime updated */ /*ARGSUSED*/ int zfs_mkdir(znode_t *dzp, char *dirname, vattr_t *vap, znode_t **zpp, cred_t *cr, int flags, vsecattr_t *vsecp) { znode_t *zp; zfsvfs_t *zfsvfs = ZTOZSB(dzp); zilog_t *zilog; zfs_dirlock_t *dl; uint64_t txtype; dmu_tx_t *tx; int error; int zf = ZNEW; uid_t uid; gid_t gid = crgetgid(cr); zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; boolean_t waited = B_FALSE; ASSERT(S_ISDIR(vap->va_mode)); /* * If we have an ephemeral id, ACL, or XVATTR then * make sure file system is at proper version */ uid = crgetuid(cr); if (zfsvfs->z_use_fuids == B_FALSE && (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (SET_ERROR(EINVAL)); if (dirname == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (dzp->z_pflags & ZFS_XATTR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if (zfsvfs->z_utf8 && u8_validate(dirname, strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zf |= ZCILOOK; if (vap->va_mask & ATTR_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, crgetuid(cr), cr, vap->va_mode)) != 0) { ZFS_EXIT(zfsvfs); return (error); } } if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, &acl_ids)) != 0) { ZFS_EXIT(zfsvfs); return (error); } /* * First make sure the new directory doesn't exist. * * Existence is checked first to make sure we don't return * EACCES instead of EEXIST which can cause some applications * to fail. */ top: *zpp = NULL; if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, NULL, NULL))) { zfs_acl_ids_free(&acl_ids); ZFS_EXIT(zfsvfs); return (error); } if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (error); } if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (SET_ERROR(EDQUOT)); } /* * Add a new entry to the directory. */ tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } /* * Create new node. */ zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); /* * Now put new name in parent dir. */ error = zfs_link_create(dl, zp, tx, ZNEW); if (error != 0) { zfs_znode_delete(zp, tx); remove_inode_hash(ZTOI(zp)); goto out; } if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); *zpp = zp; txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, acl_ids.z_fuidp, vap); out: zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); zfs_dirent_unlock(dl); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); if (error != 0) { zrele(zp); } else { zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); } ZFS_EXIT(zfsvfs); return (error); } /* * Remove a directory subdir entry. If the current working * directory is the same as the subdir to be removed, the * remove will fail. * * IN: dzp - znode of directory to remove from. * name - name of directory to be removed. * cwd - inode of current working directory. * cr - credentials of caller. * flags - case flags * * RETURN: 0 on success, error code on failure. * * Timestamps: * dzp - ctime|mtime updated */ /*ARGSUSED*/ int zfs_rmdir(znode_t *dzp, char *name, znode_t *cwd, cred_t *cr, int flags) { znode_t *zp; zfsvfs_t *zfsvfs = ZTOZSB(dzp); zilog_t *zilog; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; int zflg = ZEXISTS; boolean_t waited = B_FALSE; if (name == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (flags & FIGNORECASE) zflg |= ZCILOOK; top: zp = NULL; /* * Attempt to lock directory; fail if entry doesn't exist. */ if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL))) { ZFS_EXIT(zfsvfs); return (error); } if ((error = zfs_zaccess_delete(dzp, zp, cr))) { goto out; } if (!S_ISDIR(ZTOI(zp)->i_mode)) { error = SET_ERROR(ENOTDIR); goto out; } if (zp == cwd) { error = SET_ERROR(EINVAL); goto out; } /* * Grab a lock on the directory to make sure that no one is * trying to add (or lookup) entries while we are removing it. */ rw_enter(&zp->z_name_lock, RW_WRITER); /* * Grab a lock on the parent pointer to make sure we play well * with the treewalk and directory rename code. */ rw_enter(&zp->z_parent_lock, RW_WRITER); tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); zfs_sa_upgrade_txholds(tx, zp); zfs_sa_upgrade_txholds(tx, dzp); dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { rw_exit(&zp->z_parent_lock); rw_exit(&zp->z_name_lock); zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); zrele(zp); goto top; } dmu_tx_abort(tx); zrele(zp); ZFS_EXIT(zfsvfs); return (error); } error = zfs_link_destroy(dl, zp, tx, zflg, NULL); if (error == 0) { uint64_t txtype = TX_RMDIR; if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT, B_FALSE); } dmu_tx_commit(tx); rw_exit(&zp->z_parent_lock); rw_exit(&zp->z_name_lock); out: zfs_dirent_unlock(dl); zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); zrele(zp); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Read directory entries from the given directory cursor position and emit * name and position for each entry. * * IN: ip - inode of directory to read. * ctx - directory entry context. * cr - credentials of caller. * * RETURN: 0 if success * error code if failure * * Timestamps: * ip - atime updated * * Note that the low 4 bits of the cookie returned by zap is always zero. * This allows us to use the low range for "special" directory entries: * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, * we use the offset 2 for the '.zfs' directory. */ /* ARGSUSED */ int zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); objset_t *os; zap_cursor_t zc; zap_attribute_t zap; int error; uint8_t prefetch; uint8_t type; int done = 0; uint64_t parent; uint64_t offset; /* must be unsigned; checks for < 1 */ ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0) goto out; /* * Quit if directory has been removed (posix) */ if (zp->z_unlinked) goto out; error = 0; os = zfsvfs->z_os; offset = ctx->pos; prefetch = zp->z_zn_prefetch; /* * Initialize the iterator cursor. */ if (offset <= 3) { /* * Start iteration from the beginning of the directory. */ zap_cursor_init(&zc, os, zp->z_id); } else { /* * The offset is a serialized cursor. */ zap_cursor_init_serialized(&zc, os, zp->z_id, offset); } /* * Transform to file-system independent format */ while (!done) { uint64_t objnum; /* * Special case `.', `..', and `.zfs'. */ if (offset == 0) { (void) strcpy(zap.za_name, "."); zap.za_normalization_conflict = 0; objnum = zp->z_id; type = DT_DIR; } else if (offset == 1) { (void) strcpy(zap.za_name, ".."); zap.za_normalization_conflict = 0; objnum = parent; type = DT_DIR; } else if (offset == 2 && zfs_show_ctldir(zp)) { (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); zap.za_normalization_conflict = 0; objnum = ZFSCTL_INO_ROOT; type = DT_DIR; } else { /* * Grab next entry. */ if ((error = zap_cursor_retrieve(&zc, &zap))) { if (error == ENOENT) break; else goto update; } /* * Allow multiple entries provided the first entry is * the object id. Non-zpl consumers may safely make * use of the additional space. * * XXX: This should be a feature flag for compatibility */ if (zap.za_integer_length != 8 || zap.za_num_integers == 0) { cmn_err(CE_WARN, "zap_readdir: bad directory " "entry, obj = %lld, offset = %lld, " "length = %d, num = %lld\n", (u_longlong_t)zp->z_id, (u_longlong_t)offset, zap.za_integer_length, (u_longlong_t)zap.za_num_integers); error = SET_ERROR(ENXIO); goto update; } objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); type = ZFS_DIRENT_TYPE(zap.za_first_integer); } done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), objnum, type); if (done) break; /* Prefetch znode */ if (prefetch) { dmu_prefetch(os, objnum, 0, 0, 0, ZIO_PRIORITY_SYNC_READ); } /* * Move to the next entry, fill in the previous offset. */ if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { zap_cursor_advance(&zc); offset = zap_cursor_serialize(&zc); } else { offset += 1; } ctx->pos = offset; } zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ update: zap_cursor_fini(&zc); if (error == ENOENT) error = 0; out: ZFS_EXIT(zfsvfs); return (error); } /* * Get the basic file attributes and place them in the provided kstat * structure. The inode is assumed to be the authoritative source * for most of the attributes. However, the znode currently has the * authoritative atime, blksize, and block count. * * IN: ip - inode of file. * * OUT: sp - kstat values. * * RETURN: 0 (always succeeds) */ /* ARGSUSED */ int zfs_getattr_fast(zidmap_t *user_ns, struct inode *ip, struct kstat *sp) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); uint32_t blksize; u_longlong_t nblocks; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); mutex_enter(&zp->z_lock); zpl_generic_fillattr(user_ns, ip, sp); /* * +1 link count for root inode with visible '.zfs' directory. */ if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) if (sp->nlink < ZFS_LINK_MAX) sp->nlink++; sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); sp->blksize = blksize; sp->blocks = nblocks; if (unlikely(zp->z_blksz == 0)) { /* * Block size hasn't been set; suggest maximal I/O transfers. */ sp->blksize = zfsvfs->z_max_blksz; } mutex_exit(&zp->z_lock); /* * Required to prevent NFS client from detecting different inode * numbers of snapshot root dentry before and after snapshot mount. */ if (zfsvfs->z_issnap) { if (ip->i_sb->s_root->d_inode == ip) sp->ino = ZFSCTL_INO_SNAPDIRS - dmu_objset_id(zfsvfs->z_os); } ZFS_EXIT(zfsvfs); return (0); } /* * For the operation of changing file's user/group/project, we need to * handle not only the main object that is assigned to the file directly, * but also the ones that are used by the file via hidden xattr directory. * * Because the xattr directory may contains many EA entries, as to it may * be impossible to change all of them via the transaction of changing the * main object's user/group/project attributes. Then we have to change them * via other multiple independent transactions one by one. It may be not good * solution, but we have no better idea yet. */ static int zfs_setattr_dir(znode_t *dzp) { struct inode *dxip = ZTOI(dzp); struct inode *xip = NULL; zfsvfs_t *zfsvfs = ZTOZSB(dzp); objset_t *os = zfsvfs->z_os; zap_cursor_t zc; zap_attribute_t zap; zfs_dirlock_t *dl; znode_t *zp = NULL; dmu_tx_t *tx = NULL; uint64_t uid, gid; sa_bulk_attr_t bulk[4]; int count; int err; zap_cursor_init(&zc, os, dzp->z_id); while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { count = 0; if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { err = ENXIO; break; } err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, ZEXISTS, NULL, NULL); if (err == ENOENT) goto next; if (err) break; xip = ZTOI(zp); if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && zp->z_projid == dzp->z_projid) goto next; tx = dmu_tx_create(os); if (!(zp->z_pflags & ZFS_PROJID)) dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); else dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); err = dmu_tx_assign(tx, TXG_WAIT); if (err) break; mutex_enter(&dzp->z_lock); if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { xip->i_uid = dxip->i_uid; uid = zfs_uid_read(dxip); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, sizeof (uid)); } if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { xip->i_gid = dxip->i_gid; gid = zfs_gid_read(dxip); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, sizeof (gid)); } if (zp->z_projid != dzp->z_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)); } zp->z_projid = dzp->z_projid; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, sizeof (zp->z_projid)); } 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 { dmu_tx_abort(tx); } tx = NULL; if (err != 0 && err != ENOENT) break; next: if (zp) { zrele(zp); zp = NULL; zfs_dirent_unlock(dl); } zap_cursor_advance(&zc); } if (tx) dmu_tx_abort(tx); if (zp) { zrele(zp); zfs_dirent_unlock(dl); } zap_cursor_fini(&zc); return (err == ENOENT ? 0 : err); } /* * Set the file attributes to the values contained in the * vattr structure. * * IN: zp - znode of file to be modified. * vap - new attribute values. * If ATTR_XVATTR set, then optional attrs are being set * flags - ATTR_UTIME set if non-default time values provided. * - ATTR_NOACLCHECK (CIFS context only). * cr - credentials of caller. * * RETURN: 0 if success * error code if failure * * Timestamps: * ip - ctime updated, mtime updated if size changed. */ /* ARGSUSED */ int zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr) { struct inode *ip; zfsvfs_t *zfsvfs = ZTOZSB(zp); objset_t *os = zfsvfs->z_os; zilog_t *zilog; dmu_tx_t *tx; vattr_t oldva; xvattr_t *tmpxvattr; uint_t mask = vap->va_mask; uint_t saved_mask = 0; int trim_mask = 0; uint64_t new_mode; uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; uint64_t xattr_obj; uint64_t mtime[2], ctime[2], atime[2]; uint64_t projid = ZFS_INVALID_PROJID; znode_t *attrzp; int need_policy = FALSE; int err, err2 = 0; zfs_fuid_info_t *fuidp = NULL; xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ xoptattr_t *xoap; zfs_acl_t *aclp; boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; boolean_t fuid_dirtied = B_FALSE; boolean_t handle_eadir = B_FALSE; sa_bulk_attr_t *bulk, *xattr_bulk; int count = 0, xattr_count = 0, bulks = 8; if (mask == 0) return (0); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); ip = ZTOI(zp); /* * If this is a xvattr_t, then get a pointer to the structure of * optional attributes. If this is NULL, then we have a vattr_t. */ xoap = xva_getxoptattr(xvap); if (xoap != NULL && (mask & ATTR_XVATTR)) { if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { if (!dmu_objset_projectquota_enabled(os) || (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENOTSUP)); } projid = xoap->xoa_projid; if (unlikely(projid == ZFS_INVALID_PROJID)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) projid = ZFS_INVALID_PROJID; else need_policy = TRUE; } if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && (xoap->xoa_projinherit != ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) && (!dmu_objset_projectquota_enabled(os) || (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENOTSUP)); } } zilog = zfsvfs->z_log; /* * Make sure that if we have ephemeral uid/gid or xvattr specified * that file system is at proper version level */ if (zfsvfs->z_use_fuids == B_FALSE && (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || (mask & ATTR_XVATTR))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EISDIR)); } if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); xva_init(tmpxvattr); bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); /* * Immutable files can only alter immutable bit and atime */ if ((zp->z_pflags & ZFS_IMMUTABLE) && ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { err = SET_ERROR(EPERM); goto out3; } if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { err = SET_ERROR(EPERM); goto out3; } /* * Verify timestamps doesn't overflow 32 bits. * ZFS can handle large timestamps, but 32bit syscalls can't * handle times greater than 2039. This check should be removed * once large timestamps are fully supported. */ if (mask & (ATTR_ATIME | ATTR_MTIME)) { if (((mask & ATTR_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || ((mask & ATTR_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { err = SET_ERROR(EOVERFLOW); goto out3; } } top: attrzp = NULL; aclp = NULL; /* Can this be moved to before the top label? */ if (zfs_is_readonly(zfsvfs)) { err = SET_ERROR(EROFS); goto out3; } /* * First validate permissions */ if (mask & ATTR_SIZE) { err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); if (err) goto out3; /* * XXX - Note, we are not providing any open * mode flags here (like FNDELAY), so we may * block if there are locks present... this * should be addressed in openat(). */ /* XXX - would it be OK to generate a log record here? */ err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); if (err) goto out3; } if (mask & (ATTR_ATIME|ATTR_MTIME) || ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || XVA_ISSET_REQ(xvap, XAT_READONLY) || XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || XVA_ISSET_REQ(xvap, XAT_OFFLINE) || XVA_ISSET_REQ(xvap, XAT_SPARSE) || XVA_ISSET_REQ(xvap, XAT_CREATETIME) || XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, skipaclchk, cr); } if (mask & (ATTR_UID|ATTR_GID)) { int idmask = (mask & (ATTR_UID|ATTR_GID)); int take_owner; int take_group; /* * NOTE: even if a new mode is being set, * we may clear S_ISUID/S_ISGID bits. */ if (!(mask & ATTR_MODE)) vap->va_mode = zp->z_mode; /* * Take ownership or chgrp to group we are a member of */ take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); take_group = (mask & ATTR_GID) && zfs_groupmember(zfsvfs, vap->va_gid, cr); /* * If both ATTR_UID and ATTR_GID are set then take_owner and * take_group must both be set in order to allow taking * ownership. * * Otherwise, send the check through secpolicy_vnode_setattr() * */ if (((idmask == (ATTR_UID|ATTR_GID)) && take_owner && take_group) || ((idmask == ATTR_UID) && take_owner) || ((idmask == ATTR_GID) && take_group)) { if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, skipaclchk, cr) == 0) { /* * Remove setuid/setgid for non-privileged users */ (void) secpolicy_setid_clear(vap, cr); trim_mask = (mask & (ATTR_UID|ATTR_GID)); } else { need_policy = TRUE; } } else { need_policy = TRUE; } } mutex_enter(&zp->z_lock); oldva.va_mode = zp->z_mode; zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); if (mask & ATTR_XVATTR) { /* * Update xvattr mask to include only those attributes * that are actually changing. * * the bits will be restored prior to actually setting * the attributes so the caller thinks they were set. */ if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { if (xoap->xoa_appendonly != ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_APPENDONLY); XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); } } if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { if (xoap->xoa_projinherit != ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_PROJINHERIT); XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); } } if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { if (xoap->xoa_nounlink != ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_NOUNLINK); XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); } } if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { if (xoap->xoa_immutable != ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_IMMUTABLE); XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); } } if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { if (xoap->xoa_nodump != ((zp->z_pflags & ZFS_NODUMP) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_NODUMP); XVA_SET_REQ(tmpxvattr, XAT_NODUMP); } } if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { if (xoap->xoa_av_modified != ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); } } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { if ((!S_ISREG(ip->i_mode) && xoap->xoa_av_quarantined) || xoap->xoa_av_quarantined != ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); } } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { mutex_exit(&zp->z_lock); err = SET_ERROR(EPERM); goto out3; } if (need_policy == FALSE && (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { need_policy = TRUE; } } mutex_exit(&zp->z_lock); if (mask & ATTR_MODE) { if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { err = secpolicy_setid_setsticky_clear(ip, vap, &oldva, cr); if (err) goto out3; trim_mask |= ATTR_MODE; } else { need_policy = TRUE; } } if (need_policy) { /* * If trim_mask is set then take ownership * has been granted or write_acl is present and user * has the ability to modify mode. In that case remove * UID|GID and or MODE from mask so that * secpolicy_vnode_setattr() doesn't revoke it. */ if (trim_mask) { saved_mask = vap->va_mask; vap->va_mask &= ~trim_mask; } err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); if (err) goto out3; if (trim_mask) vap->va_mask |= saved_mask; } /* * secpolicy_vnode_setattr, or take ownership may have * changed va_mask */ mask = vap->va_mask; if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { handle_eadir = B_TRUE; err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj, sizeof (xattr_obj)); if (err == 0 && xattr_obj) { err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); if (err) goto out2; } if (mask & ATTR_UID) { new_kuid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, new_kuid)) { if (attrzp) zrele(attrzp); err = SET_ERROR(EDQUOT); goto out2; } } if (mask & ATTR_GID) { new_kgid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, new_kgid)) { if (attrzp) zrele(attrzp); err = SET_ERROR(EDQUOT); goto out2; } } if (projid != ZFS_INVALID_PROJID && zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { if (attrzp) zrele(attrzp); err = EDQUOT; goto out2; } } tx = dmu_tx_create(os); if (mask & ATTR_MODE) { uint64_t pmode = zp->z_mode; uint64_t acl_obj; new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); if (ZTOZSB(zp)->z_acl_mode == ZFS_ACL_RESTRICTED && !(zp->z_pflags & ZFS_ACL_TRIVIAL)) { err = EPERM; goto out; } if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))) goto out; mutex_enter(&zp->z_lock); if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { /* * Are we upgrading ACL from old V0 format * to V1 format? */ if (zfsvfs->z_version >= ZPL_VERSION_FUID && zfs_znode_acl_version(zp) == ZFS_ACL_VERSION_INITIAL) { dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes); } else { dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes); } } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes); } mutex_exit(&zp->z_lock); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); } else { if (((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID))) dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); else dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); } if (attrzp) { dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); } fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); zfs_sa_upgrade_txholds(tx, zp); err = dmu_tx_assign(tx, TXG_WAIT); if (err) goto out; count = 0; /* * Set each attribute requested. * We group settings according to the locks they need to acquire. * * Note: you cannot set ctime directly, although it will be * updated as a side-effect of calling this function. */ if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { /* * For the existed object that is upgraded from old system, * its on-disk layout has no slot for the project ID attribute. * But quota accounting logic needs to access related slots by * offset directly. So we need to adjust old objects' layout * to make the project ID to some unified and fixed offset. */ if (attrzp) err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); if (err == 0) err = sa_add_projid(zp->z_sa_hdl, tx, projid); if (unlikely(err == EEXIST)) err = 0; else if (err != 0) goto out; else projid = ZFS_INVALID_PROJID; } if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_enter(&zp->z_acl_lock); mutex_enter(&zp->z_lock); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); if (attrzp) { if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_enter(&attrzp->z_acl_lock); mutex_enter(&attrzp->z_lock); SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, sizeof (attrzp->z_pflags)); if (projid != ZFS_INVALID_PROJID) { attrzp->z_projid = projid; SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, sizeof (attrzp->z_projid)); } } if (mask & (ATTR_UID|ATTR_GID)) { if (mask & ATTR_UID) { ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); new_uid = zfs_uid_read(ZTOI(zp)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &new_uid, sizeof (new_uid)); if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_UID(zfsvfs), NULL, &new_uid, sizeof (new_uid)); ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); } } if (mask & ATTR_GID) { ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); new_gid = zfs_gid_read(ZTOI(zp)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &new_gid, sizeof (new_gid)); if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_GID(zfsvfs), NULL, &new_gid, sizeof (new_gid)); ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); } } if (!(mask & ATTR_MODE)) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &new_mode, sizeof (new_mode)); new_mode = zp->z_mode; } err = zfs_acl_chown_setattr(zp); ASSERT(err == 0); if (attrzp) { err = zfs_acl_chown_setattr(attrzp); ASSERT(err == 0); } } if (mask & ATTR_MODE) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &new_mode, sizeof (new_mode)); zp->z_mode = ZTOI(zp)->i_mode = new_mode; ASSERT3P(aclp, !=, NULL); err = zfs_aclset_common(zp, aclp, cr, tx); ASSERT0(err); if (zp->z_acl_cached) zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = aclp; aclp = NULL; } if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { zp->z_atime_dirty = B_FALSE; ZFS_TIME_ENCODE(&ip->i_atime, atime); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, sizeof (atime)); } if (mask & (ATTR_MTIME | ATTR_SIZE)) { ZFS_TIME_ENCODE(&vap->va_mtime, mtime); ZTOI(zp)->i_mtime = zpl_inode_timestamp_truncate( vap->va_mtime, ZTOI(zp)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, sizeof (mtime)); } if (mask & (ATTR_CTIME | ATTR_SIZE)) { ZFS_TIME_ENCODE(&vap->va_ctime, ctime); ZTOI(zp)->i_ctime = zpl_inode_timestamp_truncate(vap->va_ctime, ZTOI(zp)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, sizeof (ctime)); } 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)); } if (attrzp && mask) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, sizeof (ctime)); } /* * Do this after setting timestamps to prevent timestamp * update from toggling bit */ if (xoap && (mask & ATTR_XVATTR)) { /* * restore trimmed off masks * so that return masks can be set for caller. */ if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { XVA_SET_REQ(xvap, XAT_APPENDONLY); } if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { XVA_SET_REQ(xvap, XAT_NOUNLINK); } if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { XVA_SET_REQ(xvap, XAT_IMMUTABLE); } if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { XVA_SET_REQ(xvap, XAT_NODUMP); } if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { XVA_SET_REQ(xvap, XAT_AV_MODIFIED); } if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { XVA_SET_REQ(xvap, XAT_PROJINHERIT); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ASSERT(S_ISREG(ip->i_mode)); zfs_xvattr_set(zp, xvap, tx); } if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); if (mask != 0) zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); mutex_exit(&zp->z_lock); if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_exit(&zp->z_acl_lock); if (attrzp) { if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_exit(&attrzp->z_acl_lock); mutex_exit(&attrzp->z_lock); } out: if (err == 0 && xattr_count > 0) { err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, xattr_count, tx); ASSERT(err2 == 0); } if (aclp) zfs_acl_free(aclp); if (fuidp) { zfs_fuid_info_free(fuidp); fuidp = NULL; } if (err) { dmu_tx_abort(tx); if (attrzp) zrele(attrzp); if (err == ERESTART) goto top; } else { if (count > 0) err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); dmu_tx_commit(tx); if (attrzp) { if (err2 == 0 && handle_eadir) err2 = zfs_setattr_dir(attrzp); zrele(attrzp); } zfs_znode_update_vfs(zp); } out2: if (os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); out3: kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); kmem_free(tmpxvattr, sizeof (xvattr_t)); ZFS_EXIT(zfsvfs); return (err); } typedef struct zfs_zlock { krwlock_t *zl_rwlock; /* lock we acquired */ znode_t *zl_znode; /* znode we held */ struct zfs_zlock *zl_next; /* next in list */ } zfs_zlock_t; /* * Drop locks and release vnodes that were held by zfs_rename_lock(). */ static void zfs_rename_unlock(zfs_zlock_t **zlpp) { zfs_zlock_t *zl; while ((zl = *zlpp) != NULL) { if (zl->zl_znode != NULL) zfs_zrele_async(zl->zl_znode); rw_exit(zl->zl_rwlock); *zlpp = zl->zl_next; kmem_free(zl, sizeof (*zl)); } } /* * Search back through the directory tree, using the ".." entries. * Lock each directory in the chain to prevent concurrent renames. * Fail any attempt to move a directory into one of its own descendants. * XXX - z_parent_lock can overlap with map or grow locks */ static int zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) { zfs_zlock_t *zl; znode_t *zp = tdzp; uint64_t rootid = ZTOZSB(zp)->z_root; uint64_t oidp = zp->z_id; krwlock_t *rwlp = &szp->z_parent_lock; krw_t rw = RW_WRITER; /* * First pass write-locks szp and compares to zp->z_id. * Later passes read-lock zp and compare to zp->z_parent. */ do { if (!rw_tryenter(rwlp, rw)) { /* * Another thread is renaming in this path. * Note that if we are a WRITER, we don't have any * parent_locks held yet. */ if (rw == RW_READER && zp->z_id > szp->z_id) { /* * Drop our locks and restart */ zfs_rename_unlock(&zl); *zlpp = NULL; zp = tdzp; oidp = zp->z_id; rwlp = &szp->z_parent_lock; rw = RW_WRITER; continue; } else { /* * Wait for other thread to drop its locks */ rw_enter(rwlp, rw); } } zl = kmem_alloc(sizeof (*zl), KM_SLEEP); zl->zl_rwlock = rwlp; zl->zl_znode = NULL; zl->zl_next = *zlpp; *zlpp = zl; if (oidp == szp->z_id) /* We're a descendant of szp */ return (SET_ERROR(EINVAL)); if (oidp == rootid) /* We've hit the top */ return (0); if (rw == RW_READER) { /* i.e. not the first pass */ int error = zfs_zget(ZTOZSB(zp), oidp, &zp); if (error) return (error); zl->zl_znode = zp; } (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), &oidp, sizeof (oidp)); rwlp = &zp->z_parent_lock; rw = RW_READER; } while (zp->z_id != sdzp->z_id); return (0); } /* * Move an entry from the provided source directory to the target * directory. Change the entry name as indicated. * * IN: sdzp - Source directory containing the "old entry". * snm - Old entry name. * tdzp - Target directory to contain the "new entry". * tnm - New entry name. * cr - credentials of caller. * flags - case flags * * RETURN: 0 on success, error code on failure. * * Timestamps: * sdzp,tdzp - ctime|mtime updated */ /*ARGSUSED*/ int zfs_rename(znode_t *sdzp, char *snm, znode_t *tdzp, char *tnm, cred_t *cr, int flags) { znode_t *szp, *tzp; zfsvfs_t *zfsvfs = ZTOZSB(sdzp); zilog_t *zilog; zfs_dirlock_t *sdl, *tdl; dmu_tx_t *tx; zfs_zlock_t *zl; int cmp, serr, terr; int error = 0; int zflg = 0; boolean_t waited = B_FALSE; if (snm == NULL || tnm == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(sdzp); zilog = zfsvfs->z_log; ZFS_VERIFY_ZP(tdzp); /* * We check i_sb because snapshots and the ctldir must have different * super blocks. */ if (ZTOI(tdzp)->i_sb != ZTOI(sdzp)->i_sb || zfsctl_is_node(ZTOI(tdzp))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EXDEV)); } if (zfsvfs->z_utf8 && u8_validate(tnm, strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zflg |= ZCILOOK; top: szp = NULL; tzp = NULL; zl = NULL; /* * This is to prevent the creation of links into attribute space * by renaming a linked file into/outof an attribute directory. * See the comment in zfs_link() for why this is considered bad. */ if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Lock source and target directory entries. To prevent deadlock, * a lock ordering must be defined. We lock the directory with * the smallest object id first, or if it's a tie, the one with * the lexically first name. */ if (sdzp->z_id < tdzp->z_id) { cmp = -1; } else if (sdzp->z_id > tdzp->z_id) { cmp = 1; } else { /* * First compare the two name arguments without * considering any case folding. */ int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); ASSERT(error == 0 || !zfsvfs->z_utf8); if (cmp == 0) { /* * POSIX: "If the old argument and the new argument * both refer to links to the same existing file, * the rename() function shall return successfully * and perform no other action." */ ZFS_EXIT(zfsvfs); return (0); } /* * If the file system is case-folding, then we may * have some more checking to do. A case-folding file * system is either supporting mixed case sensitivity * access or is completely case-insensitive. Note * that the file system is always case preserving. * * In mixed sensitivity mode case sensitive behavior * is the default. FIGNORECASE must be used to * explicitly request case insensitive behavior. * * If the source and target names provided differ only * by case (e.g., a request to rename 'tim' to 'Tim'), * we will treat this as a special case in the * case-insensitive mode: as long as the source name * is an exact match, we will allow this to proceed as * a name-change request. */ if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || (zfsvfs->z_case == ZFS_CASE_MIXED && flags & FIGNORECASE)) && u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, &error) == 0) { /* * case preserving rename request, require exact * name matches */ zflg |= ZCIEXACT; zflg &= ~ZCILOOK; } } /* * If the source and destination directories are the same, we should * grab the z_name_lock of that directory only once. */ if (sdzp == tdzp) { zflg |= ZHAVELOCK; rw_enter(&sdzp->z_name_lock, RW_READER); } if (cmp < 0) { serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, ZEXISTS | zflg, NULL, NULL); terr = zfs_dirent_lock(&tdl, tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); } else { terr = zfs_dirent_lock(&tdl, tdzp, tnm, &tzp, zflg, NULL, NULL); serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, NULL, NULL); } if (serr) { /* * Source entry invalid or not there. */ if (!terr) { zfs_dirent_unlock(tdl); if (tzp) zrele(tzp); } if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (strcmp(snm, "..") == 0) serr = EINVAL; ZFS_EXIT(zfsvfs); return (serr); } if (terr) { zfs_dirent_unlock(sdl); zrele(szp); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (strcmp(tnm, "..") == 0) terr = EINVAL; ZFS_EXIT(zfsvfs); return (terr); } /* * If we are using project inheritance, means if the directory has * ZFS_PROJINHERIT set, then its descendant directories will inherit * not only the project ID, but also the ZFS_PROJINHERIT flag. Under * such case, we only allow renames into our tree when the project * IDs are the same. */ if (tdzp->z_pflags & ZFS_PROJINHERIT && tdzp->z_projid != szp->z_projid) { error = SET_ERROR(EXDEV); goto out; } /* * Must have write access at the source to remove the old entry * and write access at the target to create the new entry. * Note that if target and source are the same, this can be * done in a single check. */ if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) goto out; if (S_ISDIR(ZTOI(szp)->i_mode)) { /* * Check to make sure rename is valid. * Can't do a move like this: /usr/a/b to /usr/a/b/c/d */ if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) goto out; } /* * Does target exist? */ if (tzp) { /* * Source and target must be the same type. */ if (S_ISDIR(ZTOI(szp)->i_mode)) { if (!S_ISDIR(ZTOI(tzp)->i_mode)) { error = SET_ERROR(ENOTDIR); goto out; } } else { if (S_ISDIR(ZTOI(tzp)->i_mode)) { error = SET_ERROR(EISDIR); goto out; } } /* * POSIX dictates that when the source and target * entries refer to the same file object, rename * must do nothing and exit without error. */ if (szp->z_id == tzp->z_id) { error = 0; goto out; } } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); if (sdzp != tdzp) { dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, tdzp); } if (tzp) { dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, tzp); } zfs_sa_upgrade_txholds(tx, szp); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { if (zl != NULL) zfs_rename_unlock(&zl); zfs_dirent_unlock(sdl); zfs_dirent_unlock(tdl); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); zrele(szp); if (tzp) zrele(tzp); goto top; } dmu_tx_abort(tx); zrele(szp); if (tzp) zrele(tzp); ZFS_EXIT(zfsvfs); return (error); } if (tzp) /* Attempt to remove the existing target */ error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); if (error == 0) { error = zfs_link_create(tdl, szp, tx, ZRENAMING); if (error == 0) { szp->z_pflags |= ZFS_AV_MODIFIED; if (tdzp->z_pflags & ZFS_PROJINHERIT) szp->z_pflags |= ZFS_PROJINHERIT; error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), (void *)&szp->z_pflags, sizeof (uint64_t), tx); ASSERT0(error); error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); if (error == 0) { zfs_log_rename(zilog, tx, TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp); } else { /* * At this point, we have successfully created * the target name, but have failed to remove * the source name. Since the create was done * with the ZRENAMING flag, there are * complications; for one, the link count is * wrong. The easiest way to deal with this * is to remove the newly created target, and * return the original error. This must * succeed; fortunately, it is very unlikely to * fail, since we just created it. */ VERIFY3U(zfs_link_destroy(tdl, szp, tx, ZRENAMING, NULL), ==, 0); } } else { /* * If we had removed the existing target, subsequent * call to zfs_link_create() to add back the same entry * but, the new dnode (szp) should not fail. */ ASSERT(tzp == NULL); } } dmu_tx_commit(tx); out: if (zl != NULL) zfs_rename_unlock(&zl); zfs_dirent_unlock(sdl); zfs_dirent_unlock(tdl); zfs_znode_update_vfs(sdzp); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (sdzp != tdzp) zfs_znode_update_vfs(tdzp); zfs_znode_update_vfs(szp); zrele(szp); if (tzp) { zfs_znode_update_vfs(tzp); zrele(tzp); } if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Insert the indicated symbolic reference entry into the directory. * * IN: dzp - Directory to contain new symbolic link. * name - Name of directory entry in dip. * vap - Attributes of new entry. * link - Name for new symlink entry. * cr - credentials of caller. * flags - case flags * * OUT: zpp - Znode for new symbolic link. * * RETURN: 0 on success, error code on failure. * * Timestamps: * dip - ctime|mtime updated */ /*ARGSUSED*/ int zfs_symlink(znode_t *dzp, char *name, vattr_t *vap, char *link, znode_t **zpp, cred_t *cr, int flags) { znode_t *zp; zfs_dirlock_t *dl; dmu_tx_t *tx; zfsvfs_t *zfsvfs = ZTOZSB(dzp); zilog_t *zilog; uint64_t len = strlen(link); int error; int zflg = ZNEW; zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; uint64_t txtype = TX_SYMLINK; boolean_t waited = B_FALSE; ASSERT(S_ISLNK(vap->va_mode)); if (name == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zflg |= ZCILOOK; if (len > MAXPATHLEN) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENAMETOOLONG)); } if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids)) != 0) { ZFS_EXIT(zfsvfs); return (error); } top: *zpp = NULL; /* * Attempt to lock directory; fail if entry already exists. */ error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); if (error) { zfs_acl_ids_free(&acl_ids); ZFS_EXIT(zfsvfs); return (error); } if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (error); } if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (SET_ERROR(EDQUOT)); } tx = dmu_tx_create(zfsvfs->z_os); fuid_dirtied = zfsvfs->z_fuid_dirty; dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE + len); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } /* * Create a new object for the symlink. * for version 4 ZPL datasets the symlink will be an SA attribute */ zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); mutex_enter(&zp->z_lock); if (zp->z_is_sa) error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), link, len, tx); else zfs_sa_symlink(zp, link, len, tx); mutex_exit(&zp->z_lock); zp->z_size = len; (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), &zp->z_size, sizeof (zp->z_size), tx); /* * Insert the new object into the directory. */ error = zfs_link_create(dl, zp, tx, ZNEW); if (error != 0) { zfs_znode_delete(zp, tx); remove_inode_hash(ZTOI(zp)); } else { if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); } zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); zfs_dirent_unlock(dl); if (error == 0) { *zpp = zp; if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); } else { zrele(zp); } ZFS_EXIT(zfsvfs); return (error); } /* * Return, in the buffer contained in the provided uio structure, * the symbolic path referred to by ip. * * IN: ip - inode of symbolic link * uio - structure to contain the link path. * cr - credentials of caller. * * RETURN: 0 if success * error code if failure * * Timestamps: * ip - atime updated */ /* ARGSUSED */ int zfs_readlink(struct inode *ip, zfs_uio_t *uio, cred_t *cr) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); int error; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); mutex_enter(&zp->z_lock); if (zp->z_is_sa) error = sa_lookup_uio(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), uio); else error = zfs_sa_readlink(zp, uio); mutex_exit(&zp->z_lock); ZFS_EXIT(zfsvfs); return (error); } /* * Insert a new entry into directory tdzp referencing szp. * * IN: tdzp - Directory to contain new entry. * szp - znode of new entry. * name - name of new entry. * cr - credentials of caller. * flags - case flags. * * RETURN: 0 if success * error code if failure * * Timestamps: * tdzp - ctime|mtime updated * szp - ctime updated */ /* ARGSUSED */ int zfs_link(znode_t *tdzp, znode_t *szp, char *name, cred_t *cr, int flags) { struct inode *sip = ZTOI(szp); znode_t *tzp; zfsvfs_t *zfsvfs = ZTOZSB(tdzp); zilog_t *zilog; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; int zf = ZNEW; uint64_t parent; uid_t owner; boolean_t waited = B_FALSE; boolean_t is_tmpfile = 0; uint64_t txg; #ifdef HAVE_TMPFILE is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); #endif ASSERT(S_ISDIR(ZTOI(tdzp)->i_mode)); if (name == NULL) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(tdzp); zilog = zfsvfs->z_log; /* * POSIX dictates that we return EPERM here. * Better choices include ENOTSUP or EISDIR. */ if (S_ISDIR(sip->i_mode)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } ZFS_VERIFY_ZP(szp); /* * If we are using project inheritance, means if the directory has * ZFS_PROJINHERIT set, then its descendant directories will inherit * not only the project ID, but also the ZFS_PROJINHERIT flag. Under * such case, we only allow hard link creation in our tree when the * project IDs are the same. */ if (tdzp->z_pflags & ZFS_PROJINHERIT && tdzp->z_projid != szp->z_projid) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EXDEV)); } /* * We check i_sb because snapshots and the ctldir must have different * super blocks. */ if (sip->i_sb != ZTOI(tdzp)->i_sb || zfsctl_is_node(sip)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EXDEV)); } /* Prevent links to .zfs/shares files */ if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), &parent, sizeof (uint64_t))) != 0) { ZFS_EXIT(zfsvfs); return (error); } if (parent == zfsvfs->z_shares_dir) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zf |= ZCILOOK; /* * We do not support links between attributes and non-attributes * because of the potential security risk of creating links * into "normal" file space in order to circumvent restrictions * imposed in attribute space. */ if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), cr, ZFS_OWNER); if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { ZFS_EXIT(zfsvfs); return (error); } top: /* * Attempt to lock directory; fail if entry already exists. */ error = zfs_dirent_lock(&dl, tdzp, name, &tzp, zf, NULL, NULL); if (error) { ZFS_EXIT(zfsvfs); return (error); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name); if (is_tmpfile) dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); zfs_sa_upgrade_txholds(tx, szp); zfs_sa_upgrade_txholds(tx, tdzp); error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } /* unmark z_unlinked so zfs_link_create will not reject */ if (is_tmpfile) szp->z_unlinked = B_FALSE; error = zfs_link_create(dl, szp, tx, 0); if (error == 0) { uint64_t txtype = TX_LINK; /* * tmpfile is created to be in z_unlinkedobj, so remove it. * Also, we don't log in ZIL, because all previous file * operation on the tmpfile are ignored by ZIL. Instead we * always wait for txg to sync to make sure all previous * operation are sync safe. */ if (is_tmpfile) { VERIFY(zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); } else { if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_link(zilog, tx, txtype, tdzp, szp, name); } } else if (is_tmpfile) { /* restore z_unlinked since when linking failed */ szp->z_unlinked = B_TRUE; } txg = dmu_tx_get_txg(tx); dmu_tx_commit(tx); zfs_dirent_unlock(dl); if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); if (is_tmpfile && zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); zfs_znode_update_vfs(tdzp); zfs_znode_update_vfs(szp); ZFS_EXIT(zfsvfs); return (error); } static void zfs_putpage_sync_commit_cb(void *arg) { struct page *pp = arg; ClearPageError(pp); end_page_writeback(pp); } static void zfs_putpage_async_commit_cb(void *arg) { struct page *pp = arg; znode_t *zp = ITOZ(pp->mapping->host); ClearPageError(pp); end_page_writeback(pp); atomic_dec_32(&zp->z_async_writes_cnt); } /* * Push a page out to disk, once the page is on stable storage the * registered commit callback will be run as notification of completion. * * IN: ip - page mapped for inode. * pp - page to push (page is locked) * wbc - writeback control data * for_sync - does the caller intend to wait synchronously for the * page writeback to complete? * * RETURN: 0 if success * error code if failure * * Timestamps: * ip - ctime|mtime updated */ /* ARGSUSED */ int zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, boolean_t for_sync) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); loff_t offset; loff_t pgoff; unsigned int pglen; dmu_tx_t *tx; caddr_t va; int err = 0; uint64_t mtime[2], ctime[2]; sa_bulk_attr_t bulk[3]; int cnt = 0; struct address_space *mapping; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); ASSERT(PageLocked(pp)); pgoff = page_offset(pp); /* Page byte-offset in file */ offset = i_size_read(ip); /* File length in bytes */ pglen = MIN(PAGE_SIZE, /* Page length in bytes */ P2ROUNDUP(offset, PAGE_SIZE)-pgoff); /* Page is beyond end of file */ if (pgoff >= offset) { unlock_page(pp); ZFS_EXIT(zfsvfs); return (0); } /* Truncate page length to end of file */ if (pgoff + pglen > offset) pglen = offset - pgoff; #if 0 /* * FIXME: Allow mmap writes past its quota. The correct fix * is to register a page_mkwrite() handler to count the page * against its quota when it is about to be dirtied. */ if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, KUID_TO_SUID(ip->i_uid)) || zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, KGID_TO_SGID(ip->i_gid)) || (zp->z_projid != ZFS_DEFAULT_PROJID && zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, zp->z_projid))) { err = EDQUOT; } #endif /* * The ordering here is critical and must adhere to the following * rules in order to avoid deadlocking in either zfs_read() or * zfs_free_range() due to a lock inversion. * * 1) The page must be unlocked prior to acquiring the range lock. * This is critical because zfs_read() calls find_lock_page() * which may block on the page lock while holding the range lock. * * 2) Before setting or clearing write back on a page the range lock * must be held in order to prevent a lock inversion with the * zfs_free_range() function. * * This presents a problem because upon entering this function the * page lock is already held. To safely acquire the range lock the * page lock must be dropped. This creates a window where another * process could truncate, invalidate, dirty, or write out the page. * * Therefore, after successfully reacquiring the range and page locks * the current page state is checked. In the common case everything * will be as is expected and it can be written out. However, if * the page state has changed it must be handled accordingly. */ mapping = pp->mapping; redirty_page_for_writepage(wbc, pp); unlock_page(pp); zfs_locked_range_t *lr = zfs_rangelock_enter(&zp->z_rangelock, pgoff, pglen, RL_WRITER); lock_page(pp); /* Page mapping changed or it was no longer dirty, we're done */ if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { unlock_page(pp); zfs_rangelock_exit(lr); ZFS_EXIT(zfsvfs); return (0); } /* Another process started write block if required */ if (PageWriteback(pp)) { unlock_page(pp); zfs_rangelock_exit(lr); if (wbc->sync_mode != WB_SYNC_NONE) { /* * Speed up any non-sync page writebacks since * they may take several seconds to complete. * Refer to the comment in zpl_fsync() (when * HAVE_FSYNC_RANGE is defined) for details. */ if (atomic_load_32(&zp->z_async_writes_cnt) > 0) { zil_commit(zfsvfs->z_log, zp->z_id); } if (PageWriteback(pp)) #ifdef HAVE_PAGEMAP_FOLIO_WAIT_BIT folio_wait_bit(page_folio(pp), PG_writeback); #else wait_on_page_bit(pp, PG_writeback); #endif } ZFS_EXIT(zfsvfs); return (0); } /* Clear the dirty flag the required locks are held */ if (!clear_page_dirty_for_io(pp)) { unlock_page(pp); zfs_rangelock_exit(lr); ZFS_EXIT(zfsvfs); return (0); } /* * Counterpart for redirty_page_for_writepage() above. This page * was in fact not skipped and should not be counted as if it were. */ wbc->pages_skipped--; if (!for_sync) atomic_inc_32(&zp->z_async_writes_cnt); set_page_writeback(pp); unlock_page(pp); tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); err = dmu_tx_assign(tx, TXG_NOWAIT); if (err != 0) { if (err == ERESTART) dmu_tx_wait(tx); dmu_tx_abort(tx); #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO filemap_dirty_folio(page_mapping(pp), page_folio(pp)); #else __set_page_dirty_nobuffers(pp); #endif ClearPageError(pp); end_page_writeback(pp); if (!for_sync) atomic_dec_32(&zp->z_async_writes_cnt); zfs_rangelock_exit(lr); ZFS_EXIT(zfsvfs); return (err); } va = kmap(pp); ASSERT3U(pglen, <=, PAGE_SIZE); dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); kunmap(pp); SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); /* Preserve the mtime and ctime provided by the inode */ ZFS_TIME_ENCODE(&ip->i_mtime, mtime); ZFS_TIME_ENCODE(&ip->i_ctime, ctime); zp->z_atime_dirty = B_FALSE; zp->z_seq++; err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, for_sync ? zfs_putpage_sync_commit_cb : zfs_putpage_async_commit_cb, pp); dmu_tx_commit(tx); zfs_rangelock_exit(lr); if (wbc->sync_mode != WB_SYNC_NONE) { /* * Note that this is rarely called under writepages(), because * writepages() normally handles the entire commit for * performance reasons. */ zil_commit(zfsvfs->z_log, zp->z_id); } else if (!for_sync && atomic_load_32(&zp->z_sync_writes_cnt) > 0) { /* * If the caller does not intend to wait synchronously * for this page writeback to complete and there are active * synchronous calls on this file, do a commit so that * the latter don't accidentally end up waiting for * our writeback to complete. Refer to the comment in * zpl_fsync() (when HAVE_FSYNC_RANGE is defined) for details. */ zil_commit(zfsvfs->z_log, zp->z_id); } dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, pglen); ZFS_EXIT(zfsvfs); return (err); } /* * Update the system attributes when the inode has been dirtied. For the * moment we only update the mode, atime, mtime, and ctime. */ int zfs_dirty_inode(struct inode *ip, int flags) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); dmu_tx_t *tx; uint64_t mode, atime[2], mtime[2], ctime[2]; sa_bulk_attr_t bulk[4]; int error = 0; int cnt = 0; if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) return (0); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); #ifdef I_DIRTY_TIME /* * This is the lazytime semantic introduced in Linux 4.0 * This flag will only be called from update_time when lazytime is set. * (Note, I_DIRTY_SYNC will also set if not lazytime) * Fortunately mtime and ctime are managed within ZFS itself, so we * only need to dirty atime. */ if (flags == I_DIRTY_TIME) { zp->z_atime_dirty = B_TRUE; goto out; } #endif tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); goto out; } mutex_enter(&zp->z_lock); zp->z_atime_dirty = B_FALSE; SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); /* Preserve the mode, mtime and ctime provided by the inode */ ZFS_TIME_ENCODE(&ip->i_atime, atime); ZFS_TIME_ENCODE(&ip->i_mtime, mtime); ZFS_TIME_ENCODE(&ip->i_ctime, ctime); mode = ip->i_mode; zp->z_mode = mode; error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); mutex_exit(&zp->z_lock); dmu_tx_commit(tx); out: ZFS_EXIT(zfsvfs); return (error); } /*ARGSUSED*/ void zfs_inactive(struct inode *ip) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); uint64_t atime[2]; int error; int need_unlock = 0; /* Only read lock if we haven't already write locked, e.g. rollback */ if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { need_unlock = 1; rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); } if (zp->z_sa_hdl == NULL) { if (need_unlock) rw_exit(&zfsvfs->z_teardown_inactive_lock); return; } if (zp->z_atime_dirty && zp->z_unlinked == B_FALSE) { dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); } else { ZFS_TIME_ENCODE(&ip->i_atime, atime); mutex_enter(&zp->z_lock); (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), (void *)&atime, sizeof (atime), tx); zp->z_atime_dirty = B_FALSE; mutex_exit(&zp->z_lock); dmu_tx_commit(tx); } } zfs_zinactive(zp); if (need_unlock) rw_exit(&zfsvfs->z_teardown_inactive_lock); } /* * Fill pages with data from the disk. */ static int zfs_fillpage(struct inode *ip, struct page *pp) { zfsvfs_t *zfsvfs = ITOZSB(ip); loff_t i_size = i_size_read(ip); u_offset_t io_off = page_offset(pp); size_t io_len = PAGE_SIZE; ASSERT3U(io_off, <, i_size); if (io_off + io_len > i_size) io_len = i_size - io_off; void *va = kmap(pp); int error = dmu_read(zfsvfs->z_os, ITOZ(ip)->z_id, io_off, io_len, va, DMU_READ_PREFETCH); if (io_len != PAGE_SIZE) memset((char *)va + io_len, 0, PAGE_SIZE - io_len); kunmap(pp); if (error) { /* convert checksum errors into IO errors */ if (error == ECKSUM) error = SET_ERROR(EIO); SetPageError(pp); ClearPageUptodate(pp); } else { ClearPageError(pp); SetPageUptodate(pp); } return (error); } /* * Uses zfs_fillpage to read data from the file and fill the page. * * IN: ip - inode of file to get data from. * pp - page to read * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - atime updated */ /* ARGSUSED */ int zfs_getpage(struct inode *ip, struct page *pp) { zfsvfs_t *zfsvfs = ITOZSB(ip); znode_t *zp = ITOZ(ip); int error; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); error = zfs_fillpage(ip, pp); if (error == 0) dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, PAGE_SIZE); ZFS_EXIT(zfsvfs); return (error); } /* * Check ZFS specific permissions to memory map a section of a file. * * IN: ip - inode of the file to mmap * off - file offset * addrp - start address in memory region * len - length of memory region * vm_flags- address flags * * RETURN: 0 if success * error code if failure */ /*ARGSUSED*/ int zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, unsigned long vm_flags) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if ((vm_flags & VM_WRITE) && (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if ((vm_flags & (VM_READ | VM_EXEC)) && (zp->z_pflags & ZFS_AV_QUARANTINED)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EACCES)); } if (off < 0 || len > MAXOFFSET_T - off) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENXIO)); } ZFS_EXIT(zfsvfs); return (0); } /* * Free or allocate space in a file. Currently, this function only * supports the `F_FREESP' command. However, this command is somewhat * misnamed, as its functionality includes the ability to allocate as * well as free space. * * IN: zp - znode of file to free data in. * cmd - action to take (only F_FREESP supported). * bfp - section of file to free/alloc. * flag - current file open mode flags. * offset - current file offset. * cr - credentials of caller. * * RETURN: 0 on success, error code on failure. * * Timestamps: * zp - ctime|mtime updated */ /* ARGSUSED */ int zfs_space(znode_t *zp, int cmd, flock64_t *bfp, int flag, offset_t offset, cred_t *cr) { zfsvfs_t *zfsvfs = ZTOZSB(zp); uint64_t off, len; int error; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if (cmd != F_FREESP) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Callers might not be able to detect properly that we are read-only, * so check it explicitly here. */ if (zfs_is_readonly(zfsvfs)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EROFS)); } if (bfp->l_len < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Permissions aren't checked on Solaris because on this OS * zfs_space() can only be called with an opened file handle. * On Linux we can get here through truncate_range() which * operates directly on inodes, so we need to check access rights. */ if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { ZFS_EXIT(zfsvfs); return (error); } off = bfp->l_start; len = bfp->l_len; /* 0 means from off to end of file */ error = zfs_freesp(zp, off, len, flag, TRUE); ZFS_EXIT(zfsvfs); return (error); } /*ARGSUSED*/ int zfs_fid(struct inode *ip, fid_t *fidp) { znode_t *zp = ITOZ(ip); zfsvfs_t *zfsvfs = ITOZSB(ip); uint32_t gen; uint64_t gen64; uint64_t object = zp->z_id; zfid_short_t *zfid; int size, i, error; ZFS_ENTER(zfsvfs); if (fidp->fid_len < SHORT_FID_LEN) { fidp->fid_len = SHORT_FID_LEN; ZFS_EXIT(zfsvfs); return (SET_ERROR(ENOSPC)); } ZFS_VERIFY_ZP(zp); if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &gen64, sizeof (uint64_t))) != 0) { ZFS_EXIT(zfsvfs); return (error); } gen = (uint32_t)gen64; size = SHORT_FID_LEN; zfid = (zfid_short_t *)fidp; zfid->zf_len = size; for (i = 0; i < sizeof (zfid->zf_object); i++) zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); /* Must have a non-zero generation number to distinguish from .zfs */ if (gen == 0) gen = 1; for (i = 0; i < sizeof (zfid->zf_gen); i++) zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); ZFS_EXIT(zfsvfs); return (0); } #if defined(_KERNEL) EXPORT_SYMBOL(zfs_open); EXPORT_SYMBOL(zfs_close); EXPORT_SYMBOL(zfs_lookup); EXPORT_SYMBOL(zfs_create); EXPORT_SYMBOL(zfs_tmpfile); EXPORT_SYMBOL(zfs_remove); EXPORT_SYMBOL(zfs_mkdir); EXPORT_SYMBOL(zfs_rmdir); EXPORT_SYMBOL(zfs_readdir); EXPORT_SYMBOL(zfs_getattr_fast); EXPORT_SYMBOL(zfs_setattr); EXPORT_SYMBOL(zfs_rename); EXPORT_SYMBOL(zfs_symlink); EXPORT_SYMBOL(zfs_readlink); EXPORT_SYMBOL(zfs_link); EXPORT_SYMBOL(zfs_inactive); EXPORT_SYMBOL(zfs_space); EXPORT_SYMBOL(zfs_fid); EXPORT_SYMBOL(zfs_getpage); EXPORT_SYMBOL(zfs_putpage); EXPORT_SYMBOL(zfs_dirty_inode); EXPORT_SYMBOL(zfs_map); /* BEGIN CSTYLED */ module_param(zfs_delete_blocks, ulong, 0644); MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); /* END CSTYLED */ #endif diff --git a/module/os/linux/zfs/zpl_ctldir.c b/module/os/linux/zfs/zpl_ctldir.c index cf4da470f25f..302f0bb0ed8f 100644 --- a/module/os/linux/zfs/zpl_ctldir.c +++ b/module/os/linux/zfs/zpl_ctldir.c @@ -1,638 +1,638 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (C) 2011 Lawrence Livermore National Security, LLC. * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER). * LLNL-CODE-403049. * Rewritten for Linux by: * Rohan Puri * Brian Behlendorf */ #include #include #include #include #include /* * Common open routine. Disallow any write access. */ /* ARGSUSED */ static int zpl_common_open(struct inode *ip, struct file *filp) { - if (filp->f_mode & FMODE_WRITE) + if (blk_mode_is_open_write(filp->f_mode)) return (-EACCES); return (generic_file_open(ip, filp)); } /* * Get root directory contents. */ static int zpl_root_iterate(struct file *filp, zpl_dir_context_t *ctx) { zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp)); int error = 0; ZPL_ENTER(zfsvfs); if (!zpl_dir_emit_dots(filp, ctx)) goto out; if (ctx->pos == 2) { if (!zpl_dir_emit(ctx, ZFS_SNAPDIR_NAME, strlen(ZFS_SNAPDIR_NAME), ZFSCTL_INO_SNAPDIR, DT_DIR)) goto out; ctx->pos++; } if (ctx->pos == 3) { if (!zpl_dir_emit(ctx, ZFS_SHAREDIR_NAME, strlen(ZFS_SHAREDIR_NAME), ZFSCTL_INO_SHARES, DT_DIR)) goto out; ctx->pos++; } out: ZPL_EXIT(zfsvfs); return (error); } #if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED) static int zpl_root_readdir(struct file *filp, void *dirent, filldir_t filldir) { zpl_dir_context_t ctx = ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos); int error; error = zpl_root_iterate(filp, &ctx); filp->f_pos = ctx.pos; return (error); } #endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */ /* * Get root directory attributes. */ /* ARGSUSED */ static int #ifdef HAVE_IDMAP_IOPS_GETATTR zpl_root_getattr_impl(struct mnt_idmap *user_ns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #elif defined(HAVE_USERNS_IOPS_GETATTR) zpl_root_getattr_impl(struct user_namespace *user_ns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #else zpl_root_getattr_impl(const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #endif { struct inode *ip = path->dentry->d_inode; #if (defined(HAVE_USERNS_IOPS_GETATTR) || defined(HAVE_IDMAP_IOPS_GETATTR)) #ifdef HAVE_GENERIC_FILLATTR_USERNS generic_fillattr(user_ns, ip, stat); #elif defined(HAVE_GENERIC_FILLATTR_IDMAP) generic_fillattr(user_ns, ip, stat); #else (void) user_ns; #endif #else generic_fillattr(ip, stat); #endif stat->atime = current_time(ip); return (0); } ZPL_GETATTR_WRAPPER(zpl_root_getattr); static struct dentry * zpl_root_lookup(struct inode *dip, struct dentry *dentry, unsigned int flags) { cred_t *cr = CRED(); struct inode *ip; int error; crhold(cr); error = -zfsctl_root_lookup(dip, dname(dentry), &ip, 0, cr, NULL, NULL); ASSERT3S(error, <=, 0); crfree(cr); if (error) { if (error == -ENOENT) return (d_splice_alias(NULL, dentry)); else return (ERR_PTR(error)); } return (d_splice_alias(ip, dentry)); } /* * The '.zfs' control directory file and inode operations. */ const struct file_operations zpl_fops_root = { .open = zpl_common_open, .llseek = generic_file_llseek, .read = generic_read_dir, #ifdef HAVE_VFS_ITERATE_SHARED .iterate_shared = zpl_root_iterate, #elif defined(HAVE_VFS_ITERATE) .iterate = zpl_root_iterate, #else .readdir = zpl_root_readdir, #endif }; const struct inode_operations zpl_ops_root = { .lookup = zpl_root_lookup, .getattr = zpl_root_getattr, }; static struct vfsmount * zpl_snapdir_automount(struct path *path) { int error; error = -zfsctl_snapshot_mount(path, 0); if (error) return (ERR_PTR(error)); /* * Rather than returning the new vfsmount for the snapshot we must * return NULL to indicate a mount collision. This is done because * the user space mount calls do_add_mount() which adds the vfsmount * to the name space. If we returned the new mount here it would be * added again to the vfsmount list resulting in list corruption. */ return (NULL); } /* * Negative dentries must always be revalidated so newly created snapshots * can be detected and automounted. Normal dentries should be kept because * as of the 3.18 kernel revaliding the mountpoint dentry will result in * the snapshot being immediately unmounted. */ static int #ifdef HAVE_D_REVALIDATE_NAMEIDATA zpl_snapdir_revalidate(struct dentry *dentry, struct nameidata *i) #else zpl_snapdir_revalidate(struct dentry *dentry, unsigned int flags) #endif { return (!!dentry->d_inode); } dentry_operations_t zpl_dops_snapdirs = { /* * Auto mounting of snapshots is only supported for 2.6.37 and * newer kernels. Prior to this kernel the ops->follow_link() * callback was used as a hack to trigger the mount. The * resulting vfsmount was then explicitly grafted in to the * name space. While it might be possible to add compatibility * code to accomplish this it would require considerable care. */ .d_automount = zpl_snapdir_automount, .d_revalidate = zpl_snapdir_revalidate, }; static struct dentry * zpl_snapdir_lookup(struct inode *dip, struct dentry *dentry, unsigned int flags) { fstrans_cookie_t cookie; cred_t *cr = CRED(); struct inode *ip = NULL; int error; crhold(cr); cookie = spl_fstrans_mark(); error = -zfsctl_snapdir_lookup(dip, dname(dentry), &ip, 0, cr, NULL, NULL); ASSERT3S(error, <=, 0); spl_fstrans_unmark(cookie); crfree(cr); if (error && error != -ENOENT) return (ERR_PTR(error)); ASSERT(error == 0 || ip == NULL); d_clear_d_op(dentry); d_set_d_op(dentry, &zpl_dops_snapdirs); dentry->d_flags |= DCACHE_NEED_AUTOMOUNT; return (d_splice_alias(ip, dentry)); } static int zpl_snapdir_iterate(struct file *filp, zpl_dir_context_t *ctx) { zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp)); fstrans_cookie_t cookie; char snapname[MAXNAMELEN]; boolean_t case_conflict; uint64_t id, pos; int error = 0; ZPL_ENTER(zfsvfs); cookie = spl_fstrans_mark(); if (!zpl_dir_emit_dots(filp, ctx)) goto out; /* Start the position at 0 if it already emitted . and .. */ pos = (ctx->pos == 2 ? 0 : ctx->pos); while (error == 0) { dsl_pool_config_enter(dmu_objset_pool(zfsvfs->z_os), FTAG); error = -dmu_snapshot_list_next(zfsvfs->z_os, MAXNAMELEN, snapname, &id, &pos, &case_conflict); dsl_pool_config_exit(dmu_objset_pool(zfsvfs->z_os), FTAG); if (error) goto out; if (!zpl_dir_emit(ctx, snapname, strlen(snapname), ZFSCTL_INO_SHARES - id, DT_DIR)) goto out; ctx->pos = pos; } out: spl_fstrans_unmark(cookie); ZPL_EXIT(zfsvfs); if (error == -ENOENT) return (0); return (error); } #if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED) static int zpl_snapdir_readdir(struct file *filp, void *dirent, filldir_t filldir) { zpl_dir_context_t ctx = ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos); int error; error = zpl_snapdir_iterate(filp, &ctx); filp->f_pos = ctx.pos; return (error); } #endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */ static int #ifdef HAVE_IOPS_RENAME_USERNS zpl_snapdir_rename2(struct user_namespace *user_ns, struct inode *sdip, struct dentry *sdentry, struct inode *tdip, struct dentry *tdentry, unsigned int flags) #elif defined(HAVE_IOPS_RENAME_IDMAP) zpl_snapdir_rename2(struct mnt_idmap *user_ns, struct inode *sdip, struct dentry *sdentry, struct inode *tdip, struct dentry *tdentry, unsigned int flags) #else zpl_snapdir_rename2(struct inode *sdip, struct dentry *sdentry, struct inode *tdip, struct dentry *tdentry, unsigned int flags) #endif { cred_t *cr = CRED(); int error; /* We probably don't want to support renameat2(2) in ctldir */ if (flags) return (-EINVAL); crhold(cr); error = -zfsctl_snapdir_rename(sdip, dname(sdentry), tdip, dname(tdentry), cr, 0); ASSERT3S(error, <=, 0); crfree(cr); return (error); } #if (!defined(HAVE_RENAME_WANTS_FLAGS) && \ !defined(HAVE_IOPS_RENAME_USERNS) && \ !defined(HAVE_IOPS_RENAME_IDMAP)) static int zpl_snapdir_rename(struct inode *sdip, struct dentry *sdentry, struct inode *tdip, struct dentry *tdentry) { return (zpl_snapdir_rename2(sdip, sdentry, tdip, tdentry, 0)); } #endif static int zpl_snapdir_rmdir(struct inode *dip, struct dentry *dentry) { cred_t *cr = CRED(); int error; crhold(cr); error = -zfsctl_snapdir_remove(dip, dname(dentry), cr, 0); ASSERT3S(error, <=, 0); crfree(cr); return (error); } static int #ifdef HAVE_IOPS_MKDIR_USERNS zpl_snapdir_mkdir(struct user_namespace *user_ns, struct inode *dip, struct dentry *dentry, umode_t mode) #elif defined(HAVE_IOPS_MKDIR_IDMAP) zpl_snapdir_mkdir(struct mnt_idmap *user_ns, struct inode *dip, struct dentry *dentry, umode_t mode) #else zpl_snapdir_mkdir(struct inode *dip, struct dentry *dentry, umode_t mode) #endif { cred_t *cr = CRED(); vattr_t *vap; struct inode *ip; int error; crhold(cr); vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP); zpl_vap_init(vap, dip, mode | S_IFDIR, cr); error = -zfsctl_snapdir_mkdir(dip, dname(dentry), vap, &ip, cr, 0); if (error == 0) { d_clear_d_op(dentry); d_set_d_op(dentry, &zpl_dops_snapdirs); d_instantiate(dentry, ip); } kmem_free(vap, sizeof (vattr_t)); ASSERT3S(error, <=, 0); crfree(cr); return (error); } /* * Get snapshot directory attributes. */ /* ARGSUSED */ static int #ifdef HAVE_IDMAP_IOPS_GETATTR zpl_snapdir_getattr_impl(struct mnt_idmap *user_ns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #elif defined(HAVE_USERNS_IOPS_GETATTR) zpl_snapdir_getattr_impl(struct user_namespace *user_ns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #else zpl_snapdir_getattr_impl(const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #endif { struct inode *ip = path->dentry->d_inode; zfsvfs_t *zfsvfs = ITOZSB(ip); ZPL_ENTER(zfsvfs); #if (defined(HAVE_USERNS_IOPS_GETATTR) || defined(HAVE_IDMAP_IOPS_GETATTR)) #ifdef HAVE_GENERIC_FILLATTR_USERNS generic_fillattr(user_ns, ip, stat); #elif defined(HAVE_GENERIC_FILLATTR_IDMAP) generic_fillattr(user_ns, ip, stat); #else (void) user_ns; #endif #else generic_fillattr(ip, stat); #endif stat->nlink = stat->size = 2; stat->ctime = stat->mtime = dmu_objset_snap_cmtime(zfsvfs->z_os); stat->atime = current_time(ip); ZPL_EXIT(zfsvfs); return (0); } ZPL_GETATTR_WRAPPER(zpl_snapdir_getattr); /* * The '.zfs/snapshot' directory file operations. These mainly control * generating the list of available snapshots when doing an 'ls' in the * directory. See zpl_snapdir_readdir(). */ const struct file_operations zpl_fops_snapdir = { .open = zpl_common_open, .llseek = generic_file_llseek, .read = generic_read_dir, #ifdef HAVE_VFS_ITERATE_SHARED .iterate_shared = zpl_snapdir_iterate, #elif defined(HAVE_VFS_ITERATE) .iterate = zpl_snapdir_iterate, #else .readdir = zpl_snapdir_readdir, #endif }; /* * The '.zfs/snapshot' directory inode operations. These mainly control * creating an inode for a snapshot directory and initializing the needed * infrastructure to automount the snapshot. See zpl_snapdir_lookup(). */ const struct inode_operations zpl_ops_snapdir = { .lookup = zpl_snapdir_lookup, .getattr = zpl_snapdir_getattr, #if (defined(HAVE_RENAME_WANTS_FLAGS) || \ defined(HAVE_IOPS_RENAME_USERNS) || \ defined(HAVE_IOPS_RENAME_IDMAP)) .rename = zpl_snapdir_rename2, #else .rename = zpl_snapdir_rename, #endif .rmdir = zpl_snapdir_rmdir, .mkdir = zpl_snapdir_mkdir, }; static struct dentry * zpl_shares_lookup(struct inode *dip, struct dentry *dentry, unsigned int flags) { fstrans_cookie_t cookie; cred_t *cr = CRED(); struct inode *ip = NULL; int error; crhold(cr); cookie = spl_fstrans_mark(); error = -zfsctl_shares_lookup(dip, dname(dentry), &ip, 0, cr, NULL, NULL); ASSERT3S(error, <=, 0); spl_fstrans_unmark(cookie); crfree(cr); if (error) { if (error == -ENOENT) return (d_splice_alias(NULL, dentry)); else return (ERR_PTR(error)); } return (d_splice_alias(ip, dentry)); } static int zpl_shares_iterate(struct file *filp, zpl_dir_context_t *ctx) { fstrans_cookie_t cookie; cred_t *cr = CRED(); zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp)); znode_t *dzp; int error = 0; ZPL_ENTER(zfsvfs); cookie = spl_fstrans_mark(); if (zfsvfs->z_shares_dir == 0) { zpl_dir_emit_dots(filp, ctx); goto out; } error = -zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp); if (error) goto out; crhold(cr); error = -zfs_readdir(ZTOI(dzp), ctx, cr); crfree(cr); iput(ZTOI(dzp)); out: spl_fstrans_unmark(cookie); ZPL_EXIT(zfsvfs); ASSERT3S(error, <=, 0); return (error); } #if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED) static int zpl_shares_readdir(struct file *filp, void *dirent, filldir_t filldir) { zpl_dir_context_t ctx = ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos); int error; error = zpl_shares_iterate(filp, &ctx); filp->f_pos = ctx.pos; return (error); } #endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */ /* ARGSUSED */ static int #ifdef HAVE_USERNS_IOPS_GETATTR zpl_shares_getattr_impl(struct user_namespace *user_ns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #elif defined(HAVE_IDMAP_IOPS_GETATTR) zpl_shares_getattr_impl(struct mnt_idmap *user_ns, const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #else zpl_shares_getattr_impl(const struct path *path, struct kstat *stat, u32 request_mask, unsigned int query_flags) #endif { struct inode *ip = path->dentry->d_inode; zfsvfs_t *zfsvfs = ITOZSB(ip); znode_t *dzp; int error; ZPL_ENTER(zfsvfs); if (zfsvfs->z_shares_dir == 0) { #if (defined(HAVE_USERNS_IOPS_GETATTR) || defined(HAVE_IDMAP_IOPS_GETATTR)) #ifdef HAVE_GENERIC_FILLATTR_USERNS generic_fillattr(user_ns, path->dentry->d_inode, stat); #elif defined(HAVE_GENERIC_FILLATTR_IDMAP) generic_fillattr(user_ns, path->dentry->d_inode, stat); #else (void) user_ns; #endif #else generic_fillattr(path->dentry->d_inode, stat); #endif stat->nlink = stat->size = 2; stat->atime = current_time(ip); ZPL_EXIT(zfsvfs); return (0); } error = -zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp); if (error == 0) { #if (defined(HAVE_USERNS_IOPS_GETATTR) || defined(HAVE_IDMAP_IOPS_GETATTR)) error = -zfs_getattr_fast(user_ns, ZTOI(dzp), stat); #else error = -zfs_getattr_fast(kcred->user_ns, ZTOI(dzp), stat); #endif iput(ZTOI(dzp)); } ZPL_EXIT(zfsvfs); ASSERT3S(error, <=, 0); return (error); } ZPL_GETATTR_WRAPPER(zpl_shares_getattr); /* * The '.zfs/shares' directory file operations. */ const struct file_operations zpl_fops_shares = { .open = zpl_common_open, .llseek = generic_file_llseek, .read = generic_read_dir, #ifdef HAVE_VFS_ITERATE_SHARED .iterate_shared = zpl_shares_iterate, #elif defined(HAVE_VFS_ITERATE) .iterate = zpl_shares_iterate, #else .readdir = zpl_shares_readdir, #endif }; /* * The '.zfs/shares' directory inode operations. */ const struct inode_operations zpl_ops_shares = { .lookup = zpl_shares_lookup, .getattr = zpl_shares_getattr, }; diff --git a/module/os/linux/zfs/zvol_os.c b/module/os/linux/zfs/zvol_os.c index cef047bec6f6..36f40ba5753f 100644 --- a/module/os/linux/zfs/zvol_os.c +++ b/module/os/linux/zfs/zvol_os.c @@ -1,1231 +1,1251 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2012, 2020 by Delphix. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include unsigned int zvol_major = ZVOL_MAJOR; unsigned int zvol_request_sync = 0; unsigned int zvol_prefetch_bytes = (128 * 1024); unsigned long zvol_max_discard_blocks = 16384; unsigned int zvol_threads = 32; unsigned int zvol_open_timeout_ms = 1000; struct zvol_state_os { struct gendisk *zvo_disk; /* generic disk */ struct request_queue *zvo_queue; /* request queue */ dev_t zvo_dev; /* device id */ }; taskq_t *zvol_taskq; static struct ida zvol_ida; typedef struct zv_request_stack { zvol_state_t *zv; struct bio *bio; } zv_request_t; typedef struct zv_request_task { zv_request_t zvr; taskq_ent_t ent; } zv_request_task_t; static zv_request_task_t * zv_request_task_create(zv_request_t zvr) { zv_request_task_t *task; task = kmem_alloc(sizeof (zv_request_task_t), KM_SLEEP); taskq_init_ent(&task->ent); task->zvr = zvr; return (task); } static void zv_request_task_free(zv_request_task_t *task) { kmem_free(task, sizeof (*task)); } /* * Given a path, return TRUE if path is a ZVOL. */ static boolean_t zvol_is_zvol_impl(const char *path) { dev_t dev = 0; if (vdev_lookup_bdev(path, &dev) != 0) return (B_FALSE); if (MAJOR(dev) == zvol_major) return (B_TRUE); return (B_FALSE); } static void zvol_write(zv_request_t *zvr) { struct bio *bio = zvr->bio; int error = 0; zfs_uio_t uio; zfs_uio_bvec_init(&uio, bio); zvol_state_t *zv = zvr->zv; ASSERT3P(zv, !=, NULL); ASSERT3U(zv->zv_open_count, >, 0); ASSERT3P(zv->zv_zilog, !=, NULL); /* bio marked as FLUSH need to flush before write */ if (bio_is_flush(bio)) zil_commit(zv->zv_zilog, ZVOL_OBJ); /* Some requests are just for flush and nothing else. */ if (uio.uio_resid == 0) { rw_exit(&zv->zv_suspend_lock); BIO_END_IO(bio, 0); return; } struct request_queue *q = zv->zv_zso->zvo_queue; struct gendisk *disk = zv->zv_zso->zvo_disk; ssize_t start_resid = uio.uio_resid; unsigned long start_time; boolean_t acct = blk_queue_io_stat(q); if (acct) start_time = blk_generic_start_io_acct(q, disk, WRITE, bio); boolean_t sync = bio_is_fua(bio) || zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS; zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock, uio.uio_loffset, uio.uio_resid, RL_WRITER); uint64_t volsize = zv->zv_volsize; while (uio.uio_resid > 0 && uio.uio_loffset < volsize) { uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1); uint64_t off = uio.uio_loffset; dmu_tx_t *tx = dmu_tx_create(zv->zv_objset); if (bytes > volsize - off) /* don't write past the end */ bytes = volsize - off; dmu_tx_hold_write_by_dnode(tx, zv->zv_dn, off, bytes); /* This will only fail for ENOSPC */ error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); break; } error = dmu_write_uio_dnode(zv->zv_dn, &uio, bytes, tx); if (error == 0) { zvol_log_write(zv, tx, off, bytes, sync); } dmu_tx_commit(tx); if (error) break; } zfs_rangelock_exit(lr); int64_t nwritten = start_resid - uio.uio_resid; dataset_kstats_update_write_kstats(&zv->zv_kstat, nwritten); task_io_account_write(nwritten); if (sync) zil_commit(zv->zv_zilog, ZVOL_OBJ); rw_exit(&zv->zv_suspend_lock); if (acct) blk_generic_end_io_acct(q, disk, WRITE, bio, start_time); BIO_END_IO(bio, -error); } static void zvol_write_task(void *arg) { zv_request_task_t *task = arg; zvol_write(&task->zvr); zv_request_task_free(task); } static void zvol_discard(zv_request_t *zvr) { struct bio *bio = zvr->bio; zvol_state_t *zv = zvr->zv; uint64_t start = BIO_BI_SECTOR(bio) << 9; uint64_t size = BIO_BI_SIZE(bio); uint64_t end = start + size; boolean_t sync; int error = 0; dmu_tx_t *tx; ASSERT3P(zv, !=, NULL); ASSERT3U(zv->zv_open_count, >, 0); ASSERT3P(zv->zv_zilog, !=, NULL); struct request_queue *q = zv->zv_zso->zvo_queue; struct gendisk *disk = zv->zv_zso->zvo_disk; unsigned long start_time; boolean_t acct = blk_queue_io_stat(q); if (acct) start_time = blk_generic_start_io_acct(q, disk, WRITE, bio); sync = bio_is_fua(bio) || zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS; if (end > zv->zv_volsize) { error = SET_ERROR(EIO); goto unlock; } /* * 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. */ if (!bio_is_secure_erase(bio)) { start = P2ROUNDUP(start, zv->zv_volblocksize); end = P2ALIGN(end, zv->zv_volblocksize); size = end - start; } if (start >= end) goto unlock; zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock, start, size, RL_WRITER); tx = dmu_tx_create(zv->zv_objset); dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) { dmu_tx_abort(tx); } else { zvol_log_truncate(zv, tx, start, size, B_TRUE); dmu_tx_commit(tx); error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, start, size); } zfs_rangelock_exit(lr); if (error == 0 && sync) zil_commit(zv->zv_zilog, ZVOL_OBJ); unlock: rw_exit(&zv->zv_suspend_lock); if (acct) blk_generic_end_io_acct(q, disk, WRITE, bio, start_time); BIO_END_IO(bio, -error); } static void zvol_discard_task(void *arg) { zv_request_task_t *task = arg; zvol_discard(&task->zvr); zv_request_task_free(task); } static void zvol_read(zv_request_t *zvr) { struct bio *bio = zvr->bio; int error = 0; zfs_uio_t uio; zfs_uio_bvec_init(&uio, bio); zvol_state_t *zv = zvr->zv; ASSERT3P(zv, !=, NULL); ASSERT3U(zv->zv_open_count, >, 0); struct request_queue *q = zv->zv_zso->zvo_queue; struct gendisk *disk = zv->zv_zso->zvo_disk; ssize_t start_resid = uio.uio_resid; unsigned long start_time; boolean_t acct = blk_queue_io_stat(q); if (acct) start_time = blk_generic_start_io_acct(q, disk, READ, bio); zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock, uio.uio_loffset, uio.uio_resid, RL_READER); uint64_t volsize = zv->zv_volsize; while (uio.uio_resid > 0 && uio.uio_loffset < volsize) { uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1); /* don't read past the end */ if (bytes > volsize - uio.uio_loffset) bytes = volsize - uio.uio_loffset; error = dmu_read_uio_dnode(zv->zv_dn, &uio, bytes); if (error) { /* convert checksum errors into IO errors */ if (error == ECKSUM) error = SET_ERROR(EIO); break; } } zfs_rangelock_exit(lr); int64_t nread = start_resid - uio.uio_resid; dataset_kstats_update_read_kstats(&zv->zv_kstat, nread); task_io_account_read(nread); rw_exit(&zv->zv_suspend_lock); if (acct) blk_generic_end_io_acct(q, disk, READ, bio, start_time); BIO_END_IO(bio, -error); } static void zvol_read_task(void *arg) { zv_request_task_t *task = arg; zvol_read(&task->zvr); zv_request_task_free(task); } #ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS #ifdef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID static void zvol_submit_bio(struct bio *bio) #else static blk_qc_t zvol_submit_bio(struct bio *bio) #endif #else static MAKE_REQUEST_FN_RET zvol_request(struct request_queue *q, struct bio *bio) #endif { #ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS #if defined(HAVE_BIO_BDEV_DISK) struct request_queue *q = bio->bi_bdev->bd_disk->queue; #else struct request_queue *q = bio->bi_disk->queue; #endif #endif zvol_state_t *zv = q->queuedata; fstrans_cookie_t cookie = spl_fstrans_mark(); uint64_t offset = BIO_BI_SECTOR(bio) << 9; uint64_t size = BIO_BI_SIZE(bio); int rw = bio_data_dir(bio); if (bio_has_data(bio) && offset + size > zv->zv_volsize) { printk(KERN_INFO "%s: bad access: offset=%llu, size=%lu\n", zv->zv_zso->zvo_disk->disk_name, (long long unsigned)offset, (long unsigned)size); BIO_END_IO(bio, -SET_ERROR(EIO)); goto out; } zv_request_t zvr = { .zv = zv, .bio = bio, }; zv_request_task_t *task; if (rw == WRITE) { if (unlikely(zv->zv_flags & ZVOL_RDONLY)) { BIO_END_IO(bio, -SET_ERROR(EROFS)); goto out; } /* * Prevents the zvol from being suspended, or the ZIL being * concurrently opened. Will be released after the i/o * completes. */ rw_enter(&zv->zv_suspend_lock, RW_READER); /* * Open a ZIL if this is the first time we have written to this * zvol. We protect zv->zv_zilog with zv_suspend_lock rather * than zv_state_lock so that we don't need to acquire an * additional lock in this path. */ if (zv->zv_zilog == NULL) { rw_exit(&zv->zv_suspend_lock); rw_enter(&zv->zv_suspend_lock, RW_WRITER); if (zv->zv_zilog == NULL) { zv->zv_zilog = zil_open(zv->zv_objset, zvol_get_data); zv->zv_flags |= ZVOL_WRITTEN_TO; /* replay / destroy done in zvol_create_minor */ VERIFY0((zv->zv_zilog->zl_header->zh_flags & ZIL_REPLAY_NEEDED)); } rw_downgrade(&zv->zv_suspend_lock); } /* * We don't want this thread to be blocked waiting for i/o to * complete, so we instead wait from a taskq callback. The * i/o may be a ZIL write (via zil_commit()), or a read of an * indirect block, or a read of a data block (if this is a * partial-block write). We will indicate that the i/o is * complete by calling BIO_END_IO() from the taskq callback. * * This design allows the calling thread to continue and * initiate more concurrent operations by calling * zvol_request() again. There are typically only a small * number of threads available to call zvol_request() (e.g. * one per iSCSI target), so keeping the latency of * zvol_request() low is important for performance. * * The zvol_request_sync module parameter allows this * behavior to be altered, for performance evaluation * purposes. If the callback blocks, setting * zvol_request_sync=1 will result in much worse performance. * * We can have up to zvol_threads concurrent i/o's being * processed for all zvols on the system. This is typically * a vast improvement over the zvol_request_sync=1 behavior * of one i/o at a time per zvol. However, an even better * design would be for zvol_request() to initiate the zio * directly, and then be notified by the zio_done callback, * which would call BIO_END_IO(). Unfortunately, the DMU/ZIL * interfaces lack this functionality (they block waiting for * the i/o to complete). */ if (bio_is_discard(bio) || bio_is_secure_erase(bio)) { if (zvol_request_sync) { zvol_discard(&zvr); } else { task = zv_request_task_create(zvr); taskq_dispatch_ent(zvol_taskq, zvol_discard_task, task, 0, &task->ent); } } else { if (zvol_request_sync) { zvol_write(&zvr); } else { task = zv_request_task_create(zvr); taskq_dispatch_ent(zvol_taskq, zvol_write_task, task, 0, &task->ent); } } } else { /* * The SCST driver, and possibly others, may issue READ I/Os * with a length of zero bytes. These empty I/Os contain no * data and require no additional handling. */ if (size == 0) { BIO_END_IO(bio, 0); goto out; } rw_enter(&zv->zv_suspend_lock, RW_READER); /* See comment in WRITE case above. */ if (zvol_request_sync) { zvol_read(&zvr); } else { task = zv_request_task_create(zvr); taskq_dispatch_ent(zvol_taskq, zvol_read_task, task, 0, &task->ent); } } out: spl_fstrans_unmark(cookie); #if (defined(HAVE_MAKE_REQUEST_FN_RET_QC) || \ defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS)) && \ !defined(HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID) return (BLK_QC_T_NONE); #endif } static int +#ifdef HAVE_BLK_MODE_T +zvol_open(struct gendisk *disk, blk_mode_t flag) +#else zvol_open(struct block_device *bdev, fmode_t flag) +#endif { zvol_state_t *zv; int error = 0; boolean_t drop_suspend = B_FALSE; #ifndef HAVE_BLKDEV_GET_ERESTARTSYS hrtime_t timeout = MSEC2NSEC(zvol_open_timeout_ms); hrtime_t start = gethrtime(); retry: #endif rw_enter(&zvol_state_lock, RW_READER); /* * Obtain a copy of private_data under the zvol_state_lock to make * sure that either the result of zvol free code path setting - * bdev->bd_disk->private_data to NULL is observed, or zvol_free() + * disk->private_data to NULL is observed, or zvol_os_free() * is not called on this zv because of the positive zv_open_count. */ +#ifdef HAVE_BLK_MODE_T + zv = disk->private_data; +#else zv = bdev->bd_disk->private_data; +#endif if (zv == NULL) { rw_exit(&zvol_state_lock); return (SET_ERROR(-ENXIO)); } mutex_enter(&zv->zv_state_lock); /* * Make sure zvol is not suspended during first open * (hold zv_suspend_lock) and respect proper lock acquisition * ordering - zv_suspend_lock before zv_state_lock */ if (zv->zv_open_count == 0) { if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) { mutex_exit(&zv->zv_state_lock); rw_enter(&zv->zv_suspend_lock, RW_READER); mutex_enter(&zv->zv_state_lock); /* check to see if zv_suspend_lock is needed */ if (zv->zv_open_count != 0) { rw_exit(&zv->zv_suspend_lock); } else { drop_suspend = B_TRUE; } } else { drop_suspend = B_TRUE; } } rw_exit(&zvol_state_lock); ASSERT(MUTEX_HELD(&zv->zv_state_lock)); if (zv->zv_open_count == 0) { boolean_t drop_namespace = B_FALSE; ASSERT(RW_READ_HELD(&zv->zv_suspend_lock)); /* * In all other call paths the spa_namespace_lock is taken * before the bdev->bd_mutex lock. However, on open(2) * the __blkdev_get() function calls fops->open() with the * bdev->bd_mutex lock held. This can result in a deadlock * when zvols from one pool are used as vdevs in another. * * To prevent a lock inversion deadlock we preemptively * take the spa_namespace_lock. Normally the lock will not * be contended and this is safe because spa_open_common() * handles the case where the caller already holds the * spa_namespace_lock. * * When the lock cannot be aquired after multiple retries * this must be the vdev on zvol deadlock case and we have * no choice but to return an error. For 5.12 and older * kernels returning -ERESTARTSYS will result in the * bdev->bd_mutex being dropped, then reacquired, and * fops->open() being called again. This process can be * repeated safely until both locks are acquired. For 5.13 * and newer the -ERESTARTSYS retry logic was removed from * the kernel so the only option is to return the error for * the caller to handle it. */ if (!mutex_owned(&spa_namespace_lock)) { if (!mutex_tryenter(&spa_namespace_lock)) { mutex_exit(&zv->zv_state_lock); rw_exit(&zv->zv_suspend_lock); #ifdef HAVE_BLKDEV_GET_ERESTARTSYS schedule(); return (SET_ERROR(-ERESTARTSYS)); #else if ((gethrtime() - start) > timeout) return (SET_ERROR(-ERESTARTSYS)); schedule_timeout(MSEC_TO_TICK(10)); goto retry; #endif } else { drop_namespace = B_TRUE; } } - error = -zvol_first_open(zv, !(flag & FMODE_WRITE)); + error = -zvol_first_open(zv, !(blk_mode_is_open_write(flag))); if (drop_namespace) mutex_exit(&spa_namespace_lock); } if (error == 0) { - if ((flag & FMODE_WRITE) && (zv->zv_flags & ZVOL_RDONLY)) { + if ((blk_mode_is_open_write(flag)) && + (zv->zv_flags & ZVOL_RDONLY)) { if (zv->zv_open_count == 0) zvol_last_close(zv); error = SET_ERROR(-EROFS); } else { zv->zv_open_count++; } } mutex_exit(&zv->zv_state_lock); if (drop_suspend) rw_exit(&zv->zv_suspend_lock); if (error == 0) +#ifdef HAVE_BLK_MODE_T + disk_check_media_change(disk); +#else zfs_check_media_change(bdev); +#endif return (error); } static void -zvol_release(struct gendisk *disk, fmode_t mode) +#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG +zvol_release(struct gendisk *disk) +#else +zvol_release(struct gendisk *disk, fmode_t unused) +#endif { +#if !defined(HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG) + (void) unused; +#endif zvol_state_t *zv; boolean_t drop_suspend = B_TRUE; rw_enter(&zvol_state_lock, RW_READER); zv = disk->private_data; mutex_enter(&zv->zv_state_lock); ASSERT3U(zv->zv_open_count, >, 0); /* * make sure zvol is not suspended during last close * (hold zv_suspend_lock) and respect proper lock acquisition * ordering - zv_suspend_lock before zv_state_lock */ if (zv->zv_open_count == 1) { if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) { mutex_exit(&zv->zv_state_lock); rw_enter(&zv->zv_suspend_lock, RW_READER); mutex_enter(&zv->zv_state_lock); /* check to see if zv_suspend_lock is needed */ if (zv->zv_open_count != 1) { rw_exit(&zv->zv_suspend_lock); drop_suspend = B_FALSE; } } } else { drop_suspend = B_FALSE; } rw_exit(&zvol_state_lock); ASSERT(MUTEX_HELD(&zv->zv_state_lock)); zv->zv_open_count--; if (zv->zv_open_count == 0) { ASSERT(RW_READ_HELD(&zv->zv_suspend_lock)); zvol_last_close(zv); } mutex_exit(&zv->zv_state_lock); if (drop_suspend) rw_exit(&zv->zv_suspend_lock); } static int zvol_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd, unsigned long arg) { zvol_state_t *zv = bdev->bd_disk->private_data; int error = 0; ASSERT3U(zv->zv_open_count, >, 0); switch (cmd) { case BLKFLSBUF: fsync_bdev(bdev); invalidate_bdev(bdev); rw_enter(&zv->zv_suspend_lock, RW_READER); if (!(zv->zv_flags & ZVOL_RDONLY)) txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); rw_exit(&zv->zv_suspend_lock); break; case BLKZNAME: mutex_enter(&zv->zv_state_lock); error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN); mutex_exit(&zv->zv_state_lock); break; default: error = -ENOTTY; break; } return (SET_ERROR(error)); } #ifdef CONFIG_COMPAT static int zvol_compat_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, unsigned long arg) { return (zvol_ioctl(bdev, mode, cmd, arg)); } #else #define zvol_compat_ioctl NULL #endif static unsigned int zvol_check_events(struct gendisk *disk, unsigned int clearing) { unsigned int mask = 0; rw_enter(&zvol_state_lock, RW_READER); zvol_state_t *zv = disk->private_data; if (zv != NULL) { mutex_enter(&zv->zv_state_lock); mask = zv->zv_changed ? DISK_EVENT_MEDIA_CHANGE : 0; zv->zv_changed = 0; mutex_exit(&zv->zv_state_lock); } rw_exit(&zvol_state_lock); return (mask); } static int zvol_revalidate_disk(struct gendisk *disk) { rw_enter(&zvol_state_lock, RW_READER); zvol_state_t *zv = disk->private_data; if (zv != NULL) { mutex_enter(&zv->zv_state_lock); set_capacity(zv->zv_zso->zvo_disk, zv->zv_volsize >> SECTOR_BITS); mutex_exit(&zv->zv_state_lock); } rw_exit(&zvol_state_lock); return (0); } static int zvol_update_volsize(zvol_state_t *zv, uint64_t volsize) { struct gendisk *disk = zv->zv_zso->zvo_disk; #if defined(HAVE_REVALIDATE_DISK_SIZE) revalidate_disk_size(disk, zvol_revalidate_disk(disk) == 0); #elif defined(HAVE_REVALIDATE_DISK) revalidate_disk(disk); #else zvol_revalidate_disk(disk); #endif return (0); } static void zvol_clear_private(zvol_state_t *zv) { /* * Cleared while holding zvol_state_lock as a writer * which will prevent zvol_open() from opening it. */ zv->zv_zso->zvo_disk->private_data = NULL; } /* * Provide a simple virtual geometry for legacy compatibility. For devices * smaller than 1 MiB a small head and sector count is used to allow very * 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) { zvol_state_t *zv = bdev->bd_disk->private_data; sector_t sectors; ASSERT3U(zv->zv_open_count, >, 0); sectors = get_capacity(zv->zv_zso->zvo_disk); if (sectors > 2048) { geo->heads = 16; geo->sectors = 63; } else { geo->heads = 2; geo->sectors = 4; } geo->start = 0; geo->cylinders = sectors / (geo->heads * geo->sectors); return (0); } static struct block_device_operations zvol_ops = { .open = zvol_open, .release = zvol_release, .ioctl = zvol_ioctl, .compat_ioctl = zvol_compat_ioctl, .check_events = zvol_check_events, #ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK .revalidate_disk = zvol_revalidate_disk, #endif .getgeo = zvol_getgeo, .owner = THIS_MODULE, #ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS .submit_bio = zvol_submit_bio, #endif }; /* * Allocate memory for a new zvol_state_t and setup the required * request queue and generic disk structures for the block device. */ static zvol_state_t * zvol_alloc(dev_t dev, const char *name) { zvol_state_t *zv; struct zvol_state_os *zso; uint64_t volmode; if (dsl_prop_get_integer(name, "volmode", &volmode, NULL) != 0) return (NULL); if (volmode == ZFS_VOLMODE_DEFAULT) volmode = zvol_volmode; if (volmode == ZFS_VOLMODE_NONE) return (NULL); zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP); zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP); zv->zv_zso = zso; zv->zv_volmode = volmode; list_link_init(&zv->zv_next); mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL); #ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS #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_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. */ blk_queue_set_read_ahead(zso->zvo_queue, 1); /* Disable write merging in favor of the ZIO pipeline. */ 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; strlcpy(zv->zv_name, name, MAXNAMELEN); zfs_rangelock_init(&zv->zv_rangelock, NULL, NULL); rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL); zso->zvo_disk->major = zvol_major; zso->zvo_disk->events = DISK_EVENT_MEDIA_CHANGE; /* * Setting ZFS_VOLMODE_DEV disables partitioning on ZVOL devices. * This is accomplished by limiting the number of minors for the * device to one and explicitly disabling partition scanning. */ if (volmode == ZFS_VOLMODE_DEV) { zso->zvo_disk->minors = 1; zso->zvo_disk->flags &= ~ZFS_GENHD_FL_EXT_DEVT; zso->zvo_disk->flags |= ZFS_GENHD_FL_NO_PART; } zso->zvo_disk->first_minor = (dev & MINORMASK); zso->zvo_disk->fops = &zvol_ops; zso->zvo_disk->private_data = zv; snprintf(zso->zvo_disk->disk_name, DISK_NAME_LEN, "%s%d", ZVOL_DEV_NAME, (dev & MINORMASK)); return (zv); out_kmem: kmem_free(zso, sizeof (struct zvol_state_os)); kmem_free(zv, sizeof (zvol_state_t)); return (NULL); } /* * Cleanup then free a zvol_state_t which was created by zvol_alloc(). * At this time, the structure is not opened by anyone, is taken off * the zvol_state_list, and has its private data set to NULL. * The zvol_state_lock is dropped. * * This function may take many milliseconds to complete (e.g. we've seen * it take over 256ms), due to the calls to "blk_cleanup_queue" and * "del_gendisk". Thus, consumers need to be careful to account for this * latency when calling this function. */ static void zvol_free(zvol_state_t *zv) { ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock)); ASSERT(!MUTEX_HELD(&zv->zv_state_lock)); ASSERT0(zv->zv_open_count); ASSERT3P(zv->zv_zso->zvo_disk->private_data, ==, NULL); rw_destroy(&zv->zv_suspend_lock); 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) #if defined(HAVE_BLK_CLEANUP_DISK) blk_cleanup_disk(zv->zv_zso->zvo_disk); #else put_disk(zv->zv_zso->zvo_disk); #endif #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); mutex_destroy(&zv->zv_state_lock); dataset_kstats_destroy(&zv->zv_kstat); kmem_free(zv->zv_zso, sizeof (struct zvol_state_os)); kmem_free(zv, sizeof (zvol_state_t)); } void zvol_wait_close(zvol_state_t *zv) { } /* * Create a block device minor node and setup the linkage between it * and the specified volume. Once this function returns the block * device is live and ready for use. */ static int zvol_os_create_minor(const char *name) { zvol_state_t *zv; objset_t *os; dmu_object_info_t *doi; uint64_t volsize; uint64_t len; unsigned minor = 0; int error = 0; int idx; uint64_t hash = zvol_name_hash(name); if (zvol_inhibit_dev) return (0); idx = ida_simple_get(&zvol_ida, 0, 0, kmem_flags_convert(KM_SLEEP)); if (idx < 0) return (SET_ERROR(-idx)); minor = idx << ZVOL_MINOR_BITS; zv = zvol_find_by_name_hash(name, hash, RW_NONE); if (zv) { ASSERT(MUTEX_HELD(&zv->zv_state_lock)); mutex_exit(&zv->zv_state_lock); ida_simple_remove(&zvol_ida, idx); return (SET_ERROR(EEXIST)); } doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP); error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, B_TRUE, FTAG, &os); if (error) goto out_doi; error = dmu_object_info(os, ZVOL_OBJ, doi); if (error) goto out_dmu_objset_disown; error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); if (error) goto out_dmu_objset_disown; zv = zvol_alloc(MKDEV(zvol_major, minor), name); if (zv == NULL) { error = SET_ERROR(EAGAIN); goto out_dmu_objset_disown; } zv->zv_hash = hash; if (dmu_objset_is_snapshot(os)) zv->zv_flags |= ZVOL_RDONLY; zv->zv_volblocksize = doi->doi_data_block_size; zv->zv_volsize = volsize; zv->zv_objset = os; set_capacity(zv->zv_zso->zvo_disk, zv->zv_volsize >> 9); blk_queue_max_hw_sectors(zv->zv_zso->zvo_queue, (DMU_MAX_ACCESS / 4) >> 9); blk_queue_max_segments(zv->zv_zso->zvo_queue, UINT16_MAX); blk_queue_max_segment_size(zv->zv_zso->zvo_queue, UINT_MAX); blk_queue_physical_block_size(zv->zv_zso->zvo_queue, zv->zv_volblocksize); blk_queue_io_opt(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 #ifdef QUEUE_FLAG_NONROT blk_queue_flag_set(QUEUE_FLAG_NONROT, zv->zv_zso->zvo_queue); #endif #ifdef QUEUE_FLAG_ADD_RANDOM blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, zv->zv_zso->zvo_queue); #endif /* This flag was introduced in kernel version 4.12. */ #ifdef QUEUE_FLAG_SCSI_PASSTHROUGH blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, zv->zv_zso->zvo_queue); #endif ASSERT3P(zv->zv_zilog, ==, NULL); zv->zv_zilog = zil_open(os, zvol_get_data); if (spa_writeable(dmu_objset_spa(os))) { if (zil_replay_disable) zil_destroy(zv->zv_zilog, B_FALSE); else zil_replay(os, zv, zvol_replay_vector); } zil_close(zv->zv_zilog); zv->zv_zilog = NULL; ASSERT3P(zv->zv_kstat.dk_kstats, ==, NULL); dataset_kstats_create(&zv->zv_kstat, zv->zv_objset); /* * When udev detects the addition of the device it will immediately * invoke blkid(8) to determine the type of content on the device. * Prefetching the blocks commonly scanned by blkid(8) will speed * up this process. */ len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE); if (len > 0) { dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ); dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len, ZIO_PRIORITY_SYNC_READ); } zv->zv_objset = NULL; out_dmu_objset_disown: dmu_objset_disown(os, B_TRUE, FTAG); out_doi: kmem_free(doi, sizeof (dmu_object_info_t)); /* * Keep in mind that once add_disk() is called, the zvol is * announced to the world, and zvol_open()/zvol_release() can * be called at any time. Incidentally, add_disk() itself calls * zvol_open()->zvol_first_open() and zvol_release()->zvol_last_close() * directly as well. */ if (error == 0) { rw_enter(&zvol_state_lock, RW_WRITER); zvol_insert(zv); rw_exit(&zvol_state_lock); #ifdef HAVE_ADD_DISK_RET error = add_disk(zv->zv_zso->zvo_disk); #else add_disk(zv->zv_zso->zvo_disk); #endif } else { ida_simple_remove(&zvol_ida, idx); } return (error); } static void zvol_rename_minor(zvol_state_t *zv, const char *newname) { int readonly = get_disk_ro(zv->zv_zso->zvo_disk); ASSERT(RW_LOCK_HELD(&zvol_state_lock)); ASSERT(MUTEX_HELD(&zv->zv_state_lock)); strlcpy(zv->zv_name, newname, sizeof (zv->zv_name)); /* move to new hashtable entry */ zv->zv_hash = zvol_name_hash(zv->zv_name); hlist_del(&zv->zv_hlink); hlist_add_head(&zv->zv_hlink, ZVOL_HT_HEAD(zv->zv_hash)); /* * The block device's read-only state is briefly changed causing * a KOBJ_CHANGE uevent to be issued. This ensures udev detects * the name change and fixes the symlinks. This does not change * ZVOL_RDONLY in zv->zv_flags so the actual read-only state never * changes. This would normally be done using kobject_uevent() but * that is a GPL-only symbol which is why we need this workaround. */ set_disk_ro(zv->zv_zso->zvo_disk, !readonly); set_disk_ro(zv->zv_zso->zvo_disk, readonly); } static void zvol_set_disk_ro_impl(zvol_state_t *zv, int flags) { set_disk_ro(zv->zv_zso->zvo_disk, flags); } static void zvol_set_capacity_impl(zvol_state_t *zv, uint64_t capacity) { set_capacity(zv->zv_zso->zvo_disk, capacity); } const static zvol_platform_ops_t zvol_linux_ops = { .zv_free = zvol_free, .zv_rename_minor = zvol_rename_minor, .zv_create_minor = zvol_os_create_minor, .zv_update_volsize = zvol_update_volsize, .zv_clear_private = zvol_clear_private, .zv_is_zvol = zvol_is_zvol_impl, .zv_set_disk_ro = zvol_set_disk_ro_impl, .zv_set_capacity = zvol_set_capacity_impl, }; int zvol_init(void) { int error; int threads = MIN(MAX(zvol_threads, 1), 1024); error = register_blkdev(zvol_major, ZVOL_DRIVER); if (error) { printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error); return (error); } zvol_taskq = taskq_create(ZVOL_DRIVER, threads, maxclsyspri, threads * 2, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC); if (zvol_taskq == NULL) { unregister_blkdev(zvol_major, ZVOL_DRIVER); return (-ENOMEM); } zvol_init_impl(); ida_init(&zvol_ida); zvol_register_ops(&zvol_linux_ops); return (0); } void zvol_fini(void) { zvol_fini_impl(); unregister_blkdev(zvol_major, ZVOL_DRIVER); taskq_destroy(zvol_taskq); ida_destroy(&zvol_ida); } /* BEGIN CSTYLED */ module_param(zvol_inhibit_dev, uint, 0644); MODULE_PARM_DESC(zvol_inhibit_dev, "Do not create zvol device nodes"); module_param(zvol_major, uint, 0444); MODULE_PARM_DESC(zvol_major, "Major number for zvol device"); module_param(zvol_threads, uint, 0444); MODULE_PARM_DESC(zvol_threads, "Max number of threads to handle I/O requests"); module_param(zvol_request_sync, uint, 0644); MODULE_PARM_DESC(zvol_request_sync, "Synchronously handle bio requests"); module_param(zvol_max_discard_blocks, ulong, 0444); MODULE_PARM_DESC(zvol_max_discard_blocks, "Max number of blocks to discard"); module_param(zvol_prefetch_bytes, uint, 0644); MODULE_PARM_DESC(zvol_prefetch_bytes, "Prefetch N bytes at zvol start+end"); module_param(zvol_volmode, uint, 0644); MODULE_PARM_DESC(zvol_volmode, "Default volmode property value"); /* END CSTYLED */