4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (C) 2011 Lawrence Livermore National Security, LLC.
24 * Copyright (C) 2015 Jörg Thalheim.
30 #include <sys/taskq.h>
31 #include <linux/backing-dev.h>
35 * Added insert_inode_locked() helper function, prior to this most callers
36 * used insert_inode_hash(). The older method doesn't check for collisions
37 * in the inode_hashtable but it still acceptible for use.
39 #ifndef HAVE_INSERT_INODE_LOCKED
41 insert_inode_locked(struct inode
*ip
)
43 insert_inode_hash(ip
);
46 #endif /* HAVE_INSERT_INODE_LOCKED */
50 * Add truncate_setsize() if it is not exported by the Linux kernel.
52 * Truncate the inode and pages associated with the inode. The pages are
53 * unmapped and removed from cache.
55 #ifndef HAVE_TRUNCATE_SETSIZE
57 truncate_setsize(struct inode
*ip
, loff_t
new)
59 struct address_space
*mapping
= ip
->i_mapping
;
61 i_size_write(ip
, new);
63 unmap_mapping_range(mapping
, new + PAGE_SIZE
- 1, 0, 1);
64 truncate_inode_pages(mapping
, new);
65 unmap_mapping_range(mapping
, new + PAGE_SIZE
- 1, 0, 1);
67 #endif /* HAVE_TRUNCATE_SETSIZE */
70 * 2.6.32 - 2.6.33, bdi_setup_and_register() is not available.
71 * 2.6.34 - 3.19, bdi_setup_and_register() takes 3 arguments.
72 * 4.0 - 4.11, bdi_setup_and_register() takes 2 arguments.
73 * 4.12 - x.y, super_setup_bdi_name() new interface.
75 #if defined(HAVE_SUPER_SETUP_BDI_NAME)
76 extern atomic_long_t zfs_bdi_seq
;
79 zpl_bdi_setup(struct super_block
*sb
, char *name
)
81 return super_setup_bdi_name(sb
, "%.28s-%ld", name
,
82 atomic_long_inc_return(&zfs_bdi_seq
));
85 zpl_bdi_destroy(struct super_block
*sb
)
88 #elif defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER)
90 zpl_bdi_setup(struct super_block
*sb
, char *name
)
92 struct backing_dev_info
*bdi
;
95 bdi
= kmem_zalloc(sizeof (struct backing_dev_info
), KM_SLEEP
);
96 error
= bdi_setup_and_register(bdi
, name
);
98 kmem_free(bdi
, sizeof (struct backing_dev_info
));
107 zpl_bdi_destroy(struct super_block
*sb
)
109 struct backing_dev_info
*bdi
= sb
->s_bdi
;
112 kmem_free(bdi
, sizeof (struct backing_dev_info
));
115 #elif defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER)
117 zpl_bdi_setup(struct super_block
*sb
, char *name
)
119 struct backing_dev_info
*bdi
;
122 bdi
= kmem_zalloc(sizeof (struct backing_dev_info
), KM_SLEEP
);
123 error
= bdi_setup_and_register(bdi
, name
, BDI_CAP_MAP_COPY
);
125 kmem_free(sb
->s_bdi
, sizeof (struct backing_dev_info
));
134 zpl_bdi_destroy(struct super_block
*sb
)
136 struct backing_dev_info
*bdi
= sb
->s_bdi
;
139 kmem_free(bdi
, sizeof (struct backing_dev_info
));
143 extern atomic_long_t zfs_bdi_seq
;
146 zpl_bdi_setup(struct super_block
*sb
, char *name
)
148 struct backing_dev_info
*bdi
;
151 bdi
= kmem_zalloc(sizeof (struct backing_dev_info
), KM_SLEEP
);
153 bdi
->capabilities
= BDI_CAP_MAP_COPY
;
155 error
= bdi_init(bdi
);
157 kmem_free(bdi
, sizeof (struct backing_dev_info
));
161 error
= bdi_register(bdi
, NULL
, "%.28s-%ld", name
,
162 atomic_long_inc_return(&zfs_bdi_seq
));
165 kmem_free(bdi
, sizeof (struct backing_dev_info
));
174 zpl_bdi_destroy(struct super_block
*sb
)
176 struct backing_dev_info
*bdi
= sb
->s_bdi
;
179 kmem_free(bdi
, sizeof (struct backing_dev_info
));
186 * LOOKUP_RCU flag introduced to distinguish rcu-walk from ref-walk cases.
189 #define LOOKUP_RCU 0x0
190 #endif /* LOOKUP_RCU */
193 * 3.2-rc1 API change,
194 * Add set_nlink() if it is not exported by the Linux kernel.
196 * i_nlink is read-only in Linux 3.2, but it can be set directly in
199 #ifndef HAVE_SET_NLINK
201 set_nlink(struct inode
*inode
, unsigned int nlink
)
203 inode
->i_nlink
= nlink
;
205 #endif /* HAVE_SET_NLINK */
209 * The VFS .create, .mkdir and .mknod callbacks were updated to take a
210 * umode_t type rather than an int. To cleanly handle both definitions
211 * the zpl_umode_t type is introduced and set accordingly.
213 #ifdef HAVE_MKDIR_UMODE_T
214 typedef umode_t zpl_umode_t
;
216 typedef int zpl_umode_t
;
221 * The clear_inode() function replaces end_writeback() and introduces an
222 * ordering change regarding when the inode_sync_wait() occurs. See the
223 * configure check in config/kernel-clear-inode.m4 for full details.
225 #if defined(HAVE_EVICT_INODE) && !defined(HAVE_CLEAR_INODE)
226 #define clear_inode(ip) end_writeback(ip)
227 #endif /* HAVE_EVICT_INODE && !HAVE_CLEAR_INODE */
231 * The sget() helper function now takes the mount flags as an argument.
233 #ifdef HAVE_5ARG_SGET
234 #define zpl_sget(type, cmp, set, fl, mtd) sget(type, cmp, set, fl, mtd)
236 #define zpl_sget(type, cmp, set, fl, mtd) sget(type, cmp, set, mtd)
237 #endif /* HAVE_5ARG_SGET */
239 #if defined(SEEK_HOLE) && defined(SEEK_DATA) && !defined(HAVE_LSEEK_EXECUTE)
247 if (offset
< 0 && !(filp
->f_mode
& FMODE_UNSIGNED_OFFSET
))
250 if (offset
> maxsize
)
253 if (offset
!= filp
->f_pos
) {
254 spin_lock(&filp
->f_lock
);
255 filp
->f_pos
= offset
;
257 spin_unlock(&filp
->f_lock
);
262 #endif /* SEEK_HOLE && SEEK_DATA && !HAVE_LSEEK_EXECUTE */
264 #if defined(CONFIG_FS_POSIX_ACL)
266 * These functions safely approximates the behavior of posix_acl_release()
267 * which cannot be used because it calls the GPL-only symbol kfree_rcu().
268 * The in-kernel version, which can access the RCU, frees the ACLs after
269 * the grace period expires. Because we're unsure how long that grace
270 * period may be this implementation conservatively delays for 60 seconds.
271 * This is several orders of magnitude larger than expected grace period.
272 * At 60 seconds the kernel will also begin issuing RCU stall warnings.
274 #include <linux/posix_acl.h>
276 #if defined(HAVE_POSIX_ACL_RELEASE) && !defined(HAVE_POSIX_ACL_RELEASE_GPL_ONLY)
277 #define zpl_posix_acl_release(arg) posix_acl_release(arg)
279 void zpl_posix_acl_release_impl(struct posix_acl
*);
282 zpl_posix_acl_release(struct posix_acl
*acl
)
284 if ((acl
== NULL
) || (acl
== ACL_NOT_CACHED
))
287 if (atomic_dec_and_test(&acl
->a_refcount
))
288 zpl_posix_acl_release_impl(acl
);
290 #endif /* HAVE_POSIX_ACL_RELEASE */
292 #ifdef HAVE_SET_CACHED_ACL_USABLE
293 #define zpl_set_cached_acl(ip, ty, n) set_cached_acl(ip, ty, n)
294 #define zpl_forget_cached_acl(ip, ty) forget_cached_acl(ip, ty)
297 zpl_set_cached_acl(struct inode
*ip
, int type
, struct posix_acl
*newer
) {
298 struct posix_acl
*older
= NULL
;
300 spin_lock(&ip
->i_lock
);
302 if ((newer
!= ACL_NOT_CACHED
) && (newer
!= NULL
))
303 posix_acl_dup(newer
);
306 case ACL_TYPE_ACCESS
:
308 rcu_assign_pointer(ip
->i_acl
, newer
);
310 case ACL_TYPE_DEFAULT
:
311 older
= ip
->i_default_acl
;
312 rcu_assign_pointer(ip
->i_default_acl
, newer
);
316 spin_unlock(&ip
->i_lock
);
318 zpl_posix_acl_release(older
);
322 zpl_forget_cached_acl(struct inode
*ip
, int type
) {
323 zpl_set_cached_acl(ip
, type
, (struct posix_acl
*)ACL_NOT_CACHED
);
325 #endif /* HAVE_SET_CACHED_ACL_USABLE */
327 #ifndef HAVE___POSIX_ACL_CHMOD
328 #ifdef HAVE_POSIX_ACL_CHMOD
329 #define __posix_acl_chmod(acl, gfp, mode) posix_acl_chmod(acl, gfp, mode)
330 #define __posix_acl_create(acl, gfp, mode) posix_acl_create(acl, gfp, mode)
333 __posix_acl_chmod(struct posix_acl
**acl
, int flags
, umode_t umode
) {
334 struct posix_acl
*oldacl
= *acl
;
338 *acl
= posix_acl_clone(*acl
, flags
);
339 zpl_posix_acl_release(oldacl
);
344 error
= posix_acl_chmod_masq(*acl
, mode
);
346 zpl_posix_acl_release(*acl
);
354 __posix_acl_create(struct posix_acl
**acl
, int flags
, umode_t
*umodep
) {
355 struct posix_acl
*oldacl
= *acl
;
356 mode_t mode
= *umodep
;
359 *acl
= posix_acl_clone(*acl
, flags
);
360 zpl_posix_acl_release(oldacl
);
365 error
= posix_acl_create_masq(*acl
, &mode
);
369 zpl_posix_acl_release(*acl
);
375 #endif /* HAVE_POSIX_ACL_CHMOD */
376 #endif /* HAVE___POSIX_ACL_CHMOD */
378 #ifdef HAVE_POSIX_ACL_EQUIV_MODE_UMODE_T
379 typedef umode_t zpl_equivmode_t
;
381 typedef mode_t zpl_equivmode_t
;
382 #endif /* HAVE_POSIX_ACL_EQUIV_MODE_UMODE_T */
386 * posix_acl_valid() now must be passed a namespace, the namespace from
387 * from super block associated with the given inode is used for this purpose.
389 #ifdef HAVE_POSIX_ACL_VALID_WITH_NS
390 #define zpl_posix_acl_valid(ip, acl) posix_acl_valid(ip->i_sb->s_user_ns, acl)
392 #define zpl_posix_acl_valid(ip, acl) posix_acl_valid(acl)
395 #endif /* CONFIG_FS_POSIX_ACL */
399 * The is_owner_or_cap() function was renamed to inode_owner_or_capable().
401 #ifdef HAVE_INODE_OWNER_OR_CAPABLE
402 #define zpl_inode_owner_or_capable(ip) inode_owner_or_capable(ip)
404 #define zpl_inode_owner_or_capable(ip) is_owner_or_cap(ip)
405 #endif /* HAVE_INODE_OWNER_OR_CAPABLE */
409 * struct access f->f_dentry->d_inode was replaced by accessor function
412 #ifndef HAVE_FILE_INODE
413 static inline struct inode
*file_inode(const struct file
*f
)
415 return (f
->f_dentry
->d_inode
);
417 #endif /* HAVE_FILE_INODE */
422 #ifdef HAVE_FOLLOW_DOWN_ONE
423 #define zpl_follow_down_one(path) follow_down_one(path)
424 #define zpl_follow_up(path) follow_up(path)
426 #define zpl_follow_down_one(path) follow_down(path)
427 #define zpl_follow_up(path) follow_up(path)
433 #ifndef HAVE_SETATTR_PREPARE
435 setattr_prepare(struct dentry
*dentry
, struct iattr
*ia
)
437 return (inode_change_ok(dentry
->d_inode
, ia
));
443 * These macros are defined by kernel 4.11. We define them so that the same
444 * code builds under kernels < 4.11 and >= 4.11. The macros are set to 0 so
445 * that it will create obvious failures if they are accidentally used when built
446 * against a kernel >= 4.11.
449 #ifndef STATX_BASIC_STATS
450 #define STATX_BASIC_STATS 0
453 #ifndef AT_STATX_SYNC_AS_STAT
454 #define AT_STATX_SYNC_AS_STAT 0
459 * 4.11 takes struct path *, < 4.11 takes vfsmount *
462 #ifdef HAVE_VFSMOUNT_IOPS_GETATTR
463 #define ZPL_GETATTR_WRAPPER(func) \
465 func(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) \
467 struct path path = { .mnt = mnt, .dentry = dentry }; \
468 return func##_impl(&path, stat, STATX_BASIC_STATS, \
469 AT_STATX_SYNC_AS_STAT); \
471 #elif defined(HAVE_PATH_IOPS_GETATTR)
472 #define ZPL_GETATTR_WRAPPER(func) \
474 func(const struct path *path, struct kstat *stat, u32 request_mask, \
475 unsigned int query_flags) \
477 return (func##_impl(path, stat, request_mask, query_flags)); \
485 * Preferred interface to get the current FS time.
487 #if !defined(HAVE_CURRENT_TIME)
488 static inline struct timespec
489 current_time(struct inode
*ip
)
491 return (timespec_trunc(current_kernel_time(), ip
->i_sb
->s_time_gran
));
495 #endif /* _ZFS_VFS_H */