]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ce4ef7c0 BS |
2 | /* |
3 | * linux/fs/nfs/file.c | |
4 | * | |
5 | * Copyright (C) 1992 Rick Sladkey | |
6 | */ | |
f4ac1674 | 7 | #include <linux/fs.h> |
bea51b30 | 8 | #include <linux/file.h> |
f4ac1674 | 9 | #include <linux/falloc.h> |
f2aedb71 | 10 | #include <linux/mount.h> |
ce4ef7c0 | 11 | #include <linux/nfs_fs.h> |
0cfcd405 | 12 | #include <linux/nfs_ssc.h> |
5445b1fb | 13 | #include "delegation.h" |
ce4ef7c0 | 14 | #include "internal.h" |
5445b1fb | 15 | #include "iostat.h" |
a4ff1468 | 16 | #include "fscache.h" |
ce4ef7c0 BS |
17 | #include "pnfs.h" |
18 | ||
81b79afb TM |
19 | #include "nfstrace.h" |
20 | ||
1c6dcbe5 AS |
21 | #ifdef CONFIG_NFS_V4_2 |
22 | #include "nfs42.h" | |
23 | #endif | |
24 | ||
ce4ef7c0 BS |
25 | #define NFSDBG_FACILITY NFSDBG_FILE |
26 | ||
27 | static int | |
28 | nfs4_file_open(struct inode *inode, struct file *filp) | |
29 | { | |
30 | struct nfs_open_context *ctx; | |
be62a1a8 | 31 | struct dentry *dentry = file_dentry(filp); |
ce4ef7c0 BS |
32 | struct dentry *parent = NULL; |
33 | struct inode *dir; | |
34 | unsigned openflags = filp->f_flags; | |
b243874f | 35 | fmode_t f_mode; |
ce4ef7c0 BS |
36 | struct iattr attr; |
37 | int err; | |
38 | ||
ce4ef7c0 BS |
39 | /* |
40 | * If no cached dentry exists or if it's negative, NFSv4 handled the | |
41 | * opens in ->lookup() or ->create(). | |
42 | * | |
43 | * We only get this far for a cached positive dentry. We skipped | |
44 | * revalidation, so handle it here by dropping the dentry and returning | |
45 | * -EOPENSTALE. The VFS will retry the lookup/create/open. | |
46 | */ | |
47 | ||
6de1472f | 48 | dprintk("NFS: open file(%pd2)\n", dentry); |
ce4ef7c0 | 49 | |
18a60089 BC |
50 | err = nfs_check_flags(openflags); |
51 | if (err) | |
52 | return err; | |
53 | ||
b243874f | 54 | f_mode = filp->f_mode; |
ce4ef7c0 | 55 | if ((openflags & O_ACCMODE) == 3) |
b243874f | 56 | f_mode |= flags_to_mode(openflags); |
ce4ef7c0 BS |
57 | |
58 | /* We can't create new files here */ | |
59 | openflags &= ~(O_CREAT|O_EXCL); | |
60 | ||
61 | parent = dget_parent(dentry); | |
2b0143b5 | 62 | dir = d_inode(parent); |
ce4ef7c0 | 63 | |
b243874f | 64 | ctx = alloc_nfs_open_context(file_dentry(filp), f_mode, filp); |
ce4ef7c0 BS |
65 | err = PTR_ERR(ctx); |
66 | if (IS_ERR(ctx)) | |
67 | goto out; | |
68 | ||
69 | attr.ia_valid = ATTR_OPEN; | |
70 | if (openflags & O_TRUNC) { | |
71 | attr.ia_valid |= ATTR_SIZE; | |
72 | attr.ia_size = 0; | |
8b7d9d09 | 73 | filemap_write_and_wait(inode->i_mapping); |
ce4ef7c0 BS |
74 | } |
75 | ||
c5c3fb5f | 76 | inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, &attr, NULL); |
ce4ef7c0 BS |
77 | if (IS_ERR(inode)) { |
78 | err = PTR_ERR(inode); | |
79 | switch (err) { | |
ce4ef7c0 | 80 | default: |
90cf500e TM |
81 | goto out_put_ctx; |
82 | case -ENOENT: | |
83 | case -ESTALE: | |
84 | case -EISDIR: | |
85 | case -ENOTDIR: | |
86 | case -ELOOP: | |
ce4ef7c0 BS |
87 | goto out_drop; |
88 | } | |
89 | } | |
2b0143b5 | 90 | if (inode != d_inode(dentry)) |
ce4ef7c0 BS |
91 | goto out_drop; |
92 | ||
ce4ef7c0 | 93 | nfs_file_set_open_context(filp, ctx); |
f1fe29b4 | 94 | nfs_fscache_open_file(inode, filp); |
ce4ef7c0 | 95 | err = 0; |
5ee3d10f | 96 | filp->f_mode |= FMODE_CAN_ODIRECT; |
ce4ef7c0 BS |
97 | |
98 | out_put_ctx: | |
99 | put_nfs_open_context(ctx); | |
100 | out: | |
101 | dput(parent); | |
102 | return err; | |
103 | ||
104 | out_drop: | |
105 | d_drop(dentry); | |
106 | err = -EOPENSTALE; | |
107 | goto out_put_ctx; | |
108 | } | |
109 | ||
5445b1fb TM |
110 | /* |
111 | * Flush all dirty pages, and check for write errors. | |
112 | */ | |
113 | static int | |
114 | nfs4_file_flush(struct file *file, fl_owner_t id) | |
115 | { | |
116 | struct inode *inode = file_inode(file); | |
67dd23f9 | 117 | errseq_t since; |
5445b1fb TM |
118 | |
119 | dprintk("NFS: flush(%pD2)\n", file); | |
120 | ||
121 | nfs_inc_stats(inode, NFSIOS_VFSFLUSH); | |
122 | if ((file->f_mode & FMODE_WRITE) == 0) | |
123 | return 0; | |
124 | ||
125 | /* | |
126 | * If we're holding a write delegation, then check if we're required | |
127 | * to flush the i/o on close. If not, then just start the i/o now. | |
128 | */ | |
129 | if (!nfs4_delegation_flush_on_close(inode)) | |
130 | return filemap_fdatawrite(file->f_mapping); | |
131 | ||
132 | /* Flush writes to the server and return any errors */ | |
67dd23f9 SM |
133 | since = filemap_sample_wb_err(file->f_mapping); |
134 | nfs_wb_all(inode); | |
135 | return filemap_check_wb_err(file->f_mapping, since); | |
5445b1fb TM |
136 | } |
137 | ||
1c6dcbe5 | 138 | #ifdef CONFIG_NFS_V4_2 |
64bf5ff5 DC |
139 | static ssize_t __nfs4_copy_file_range(struct file *file_in, loff_t pos_in, |
140 | struct file *file_out, loff_t pos_out, | |
141 | size_t count, unsigned int flags) | |
2e72448b | 142 | { |
0491567b | 143 | struct nfs42_copy_notify_res *cn_resp = NULL; |
1d38f3f0 OK |
144 | struct nl4_server *nss = NULL; |
145 | nfs4_stateid *cnrs = NULL; | |
0491567b | 146 | ssize_t ret; |
12751010 | 147 | bool sync = false; |
0491567b | 148 | |
5dae222a | 149 | /* Only offload copy if superblock is the same */ |
8dff1df5 | 150 | if (file_in->f_op != &nfs4_file_operations) |
5dae222a | 151 | return -EXDEV; |
d8a6ad91 DW |
152 | if (!nfs_server_capable(file_inode(file_out), NFS_CAP_COPY) || |
153 | !nfs_server_capable(file_inode(file_in), NFS_CAP_COPY)) | |
0769663b | 154 | return -EOPNOTSUPP; |
837bb1d7 | 155 | if (file_inode(file_in) == file_inode(file_out)) |
0769663b | 156 | return -EOPNOTSUPP; |
12751010 OK |
157 | /* if the copy size if smaller than 2 RPC payloads, make it |
158 | * synchronous | |
159 | */ | |
160 | if (count <= 2 * NFS_SERVER(file_inode(file_in))->rsize) | |
161 | sync = true; | |
0e65a32c | 162 | retry: |
0491567b | 163 | if (!nfs42_files_from_same_server(file_in, file_out)) { |
ca7d1d1a DN |
164 | /* |
165 | * for inter copy, if copy size is too small | |
166 | * then fallback to generic copy. | |
12751010 | 167 | */ |
ca7d1d1a | 168 | if (sync) |
12751010 | 169 | return -EOPNOTSUPP; |
0491567b | 170 | cn_resp = kzalloc(sizeof(struct nfs42_copy_notify_res), |
4fb547be | 171 | GFP_KERNEL); |
0491567b OK |
172 | if (unlikely(cn_resp == NULL)) |
173 | return -ENOMEM; | |
174 | ||
175 | ret = nfs42_proc_copy_notify(file_in, file_out, cn_resp); | |
176 | if (ret) { | |
177 | ret = -EOPNOTSUPP; | |
178 | goto out; | |
179 | } | |
1d38f3f0 OK |
180 | nss = &cn_resp->cnr_src; |
181 | cnrs = &cn_resp->cnr_stateid; | |
0491567b | 182 | } |
1d38f3f0 | 183 | ret = nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count, |
12751010 | 184 | nss, cnrs, sync); |
0491567b | 185 | out: |
98c27f27 TR |
186 | kfree(cn_resp); |
187 | ||
0e65a32c OK |
188 | if (ret == -EAGAIN) |
189 | goto retry; | |
0491567b | 190 | return ret; |
2e72448b AS |
191 | } |
192 | ||
64bf5ff5 DC |
193 | static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in, |
194 | struct file *file_out, loff_t pos_out, | |
195 | size_t count, unsigned int flags) | |
196 | { | |
197 | ssize_t ret; | |
198 | ||
199 | ret = __nfs4_copy_file_range(file_in, pos_in, file_out, pos_out, count, | |
200 | flags); | |
5dae222a | 201 | if (ret == -EOPNOTSUPP || ret == -EXDEV) |
64bf5ff5 DC |
202 | ret = generic_copy_file_range(file_in, pos_in, file_out, |
203 | pos_out, count, flags); | |
204 | return ret; | |
205 | } | |
206 | ||
1c6dcbe5 AS |
207 | static loff_t nfs4_file_llseek(struct file *filep, loff_t offset, int whence) |
208 | { | |
209 | loff_t ret; | |
210 | ||
211 | switch (whence) { | |
212 | case SEEK_HOLE: | |
213 | case SEEK_DATA: | |
214 | ret = nfs42_proc_llseek(filep, offset, whence); | |
e67afa7e | 215 | if (ret != -EOPNOTSUPP) |
1c6dcbe5 | 216 | return ret; |
df561f66 | 217 | fallthrough; |
1c6dcbe5 AS |
218 | default: |
219 | return nfs_file_llseek(filep, offset, whence); | |
220 | } | |
221 | } | |
f4ac1674 AS |
222 | |
223 | static long nfs42_fallocate(struct file *filep, int mode, loff_t offset, loff_t len) | |
224 | { | |
225 | struct inode *inode = file_inode(filep); | |
226 | long ret; | |
227 | ||
228 | if (!S_ISREG(inode->i_mode)) | |
229 | return -EOPNOTSUPP; | |
230 | ||
624bd5b7 | 231 | if ((mode != 0) && (mode != (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE))) |
f4ac1674 AS |
232 | return -EOPNOTSUPP; |
233 | ||
234 | ret = inode_newsize_ok(inode, offset + len); | |
235 | if (ret < 0) | |
236 | return ret; | |
237 | ||
624bd5b7 | 238 | if (mode & FALLOC_FL_PUNCH_HOLE) |
f830f7dd AS |
239 | return nfs42_proc_deallocate(filep, offset, len); |
240 | return nfs42_proc_allocate(filep, offset, len); | |
f4ac1674 | 241 | } |
bea51b30 | 242 | |
42ec3d4c DW |
243 | static loff_t nfs42_remap_file_range(struct file *src_file, loff_t src_off, |
244 | struct file *dst_file, loff_t dst_off, loff_t count, | |
2e5dfc99 | 245 | unsigned int remap_flags) |
bea51b30 PT |
246 | { |
247 | struct inode *dst_inode = file_inode(dst_file); | |
811b7b85 | 248 | struct nfs_server *server = NFS_SERVER(dst_inode); |
04b38d60 | 249 | struct inode *src_inode = file_inode(src_file); |
811b7b85 | 250 | unsigned int bs = server->clone_blksize; |
21fad313 | 251 | bool same_inode = false; |
bea51b30 PT |
252 | int ret; |
253 | ||
9026b3a9 DW |
254 | /* NFS does not support deduplication. */ |
255 | if (remap_flags & REMAP_FILE_DEDUP) | |
256 | return -EOPNOTSUPP; | |
257 | ||
258 | if (remap_flags & ~REMAP_FILE_ADVISORY) | |
2e5dfc99 DW |
259 | return -EINVAL; |
260 | ||
f5fdf124 MZ |
261 | if (IS_SWAPFILE(dst_inode) || IS_SWAPFILE(src_inode)) |
262 | return -ETXTBSY; | |
263 | ||
811b7b85 PT |
264 | /* check alignment w.r.t. clone_blksize */ |
265 | ret = -EINVAL; | |
266 | if (bs) { | |
267 | if (!IS_ALIGNED(src_off, bs) || !IS_ALIGNED(dst_off, bs)) | |
04b38d60 | 268 | goto out; |
811b7b85 | 269 | if (!IS_ALIGNED(count, bs) && i_size_read(src_inode) != (src_off + count)) |
04b38d60 | 270 | goto out; |
811b7b85 PT |
271 | } |
272 | ||
04b38d60 CH |
273 | if (src_inode == dst_inode) |
274 | same_inode = true; | |
21fad313 | 275 | |
bea51b30 | 276 | /* XXX: do we lock at all? what if server needs CB_RECALL_LAYOUT? */ |
21fad313 | 277 | if (same_inode) { |
5955102c | 278 | inode_lock(src_inode); |
21fad313 | 279 | } else if (dst_inode < src_inode) { |
5955102c AV |
280 | inode_lock_nested(dst_inode, I_MUTEX_PARENT); |
281 | inode_lock_nested(src_inode, I_MUTEX_CHILD); | |
bea51b30 | 282 | } else { |
5955102c AV |
283 | inode_lock_nested(src_inode, I_MUTEX_PARENT); |
284 | inode_lock_nested(dst_inode, I_MUTEX_CHILD); | |
bea51b30 PT |
285 | } |
286 | ||
287 | /* flush all pending writes on both src and dst so that server | |
288 | * has the latest data */ | |
289 | ret = nfs_sync_inode(src_inode); | |
290 | if (ret) | |
291 | goto out_unlock; | |
292 | ret = nfs_sync_inode(dst_inode); | |
293 | if (ret) | |
294 | goto out_unlock; | |
295 | ||
04b38d60 | 296 | ret = nfs42_proc_clone(src_file, dst_file, src_off, dst_off, count); |
bea51b30 PT |
297 | |
298 | /* truncate inode page cache of the dst range so that future reads can fetch | |
299 | * new data from server */ | |
300 | if (!ret) | |
301 | truncate_inode_pages_range(&dst_inode->i_data, dst_off, dst_off + count - 1); | |
302 | ||
303 | out_unlock: | |
21fad313 | 304 | if (same_inode) { |
5955102c | 305 | inode_unlock(src_inode); |
21fad313 | 306 | } else if (dst_inode < src_inode) { |
5955102c AV |
307 | inode_unlock(src_inode); |
308 | inode_unlock(dst_inode); | |
bea51b30 | 309 | } else { |
5955102c AV |
310 | inode_unlock(dst_inode); |
311 | inode_unlock(src_inode); | |
bea51b30 | 312 | } |
04b38d60 | 313 | out: |
42ec3d4c | 314 | return ret < 0 ? ret : count; |
bea51b30 | 315 | } |
ec4b0925 OK |
316 | |
317 | static int read_name_gen = 1; | |
318 | #define SSC_READ_NAME_BODY "ssc_read_%d" | |
319 | ||
0cfcd405 DN |
320 | static struct file *__nfs42_ssc_open(struct vfsmount *ss_mnt, |
321 | struct nfs_fh *src_fh, nfs4_stateid *stateid) | |
ec4b0925 | 322 | { |
156cd285 | 323 | struct nfs_fattr *fattr = nfs_alloc_fattr(); |
ec4b0925 OK |
324 | struct file *filep, *res; |
325 | struct nfs_server *server; | |
326 | struct inode *r_ino = NULL; | |
327 | struct nfs_open_context *ctx; | |
328 | struct nfs4_state_owner *sp; | |
f751c545 | 329 | char *read_name = NULL; |
ec4b0925 OK |
330 | int len, status = 0; |
331 | ||
f6957b71 | 332 | server = NFS_SB(ss_mnt->mnt_sb); |
ec4b0925 | 333 | |
156cd285 TM |
334 | if (!fattr) |
335 | return ERR_PTR(-ENOMEM); | |
ec4b0925 | 336 | |
2ef61e0e | 337 | status = nfs4_proc_getattr(server, src_fh, fattr, NULL); |
ec4b0925 OK |
338 | if (status < 0) { |
339 | res = ERR_PTR(status); | |
340 | goto out; | |
341 | } | |
342 | ||
fcfc8be1 OK |
343 | if (!S_ISREG(fattr->mode)) { |
344 | res = ERR_PTR(-EBADF); | |
345 | goto out; | |
346 | } | |
347 | ||
ec4b0925 OK |
348 | res = ERR_PTR(-ENOMEM); |
349 | len = strlen(SSC_READ_NAME_BODY) + 16; | |
4fb547be | 350 | read_name = kzalloc(len, GFP_KERNEL); |
ec4b0925 OK |
351 | if (read_name == NULL) |
352 | goto out; | |
353 | snprintf(read_name, len, SSC_READ_NAME_BODY, read_name_gen++); | |
354 | ||
f6957b71 | 355 | r_ino = nfs_fhget(ss_mnt->mnt_sb, src_fh, fattr); |
ec4b0925 OK |
356 | if (IS_ERR(r_ino)) { |
357 | res = ERR_CAST(r_ino); | |
f751c545 | 358 | goto out_free_name; |
ec4b0925 OK |
359 | } |
360 | ||
aa97a3ef | 361 | filep = alloc_file_pseudo(r_ino, ss_mnt, read_name, O_RDONLY, |
ec4b0925 OK |
362 | r_ino->i_fop); |
363 | if (IS_ERR(filep)) { | |
364 | res = ERR_CAST(filep); | |
fcfc8be1 | 365 | iput(r_ino); |
f751c545 | 366 | goto out_free_name; |
ec4b0925 | 367 | } |
ec4b0925 OK |
368 | |
369 | ctx = alloc_nfs_open_context(filep->f_path.dentry, filep->f_mode, | |
370 | filep); | |
371 | if (IS_ERR(ctx)) { | |
372 | res = ERR_CAST(ctx); | |
373 | goto out_filep; | |
374 | } | |
375 | ||
376 | res = ERR_PTR(-EINVAL); | |
377 | sp = nfs4_get_state_owner(server, ctx->cred, GFP_KERNEL); | |
378 | if (sp == NULL) | |
379 | goto out_ctx; | |
380 | ||
381 | ctx->state = nfs4_get_open_state(r_ino, sp); | |
382 | if (ctx->state == NULL) | |
383 | goto out_stateowner; | |
384 | ||
0b9018b9 | 385 | set_bit(NFS_SRV_SSC_COPY_STATE, &ctx->state->flags); |
ec4b0925 OK |
386 | memcpy(&ctx->state->open_stateid.other, &stateid->other, |
387 | NFS4_STATEID_OTHER_SIZE); | |
388 | update_open_stateid(ctx->state, stateid, NULL, filep->f_mode); | |
fe8eb820 | 389 | set_bit(NFS_OPEN_STATE, &ctx->state->flags); |
ec4b0925 OK |
390 | |
391 | nfs_file_set_open_context(filep, ctx); | |
392 | put_nfs_open_context(ctx); | |
393 | ||
394 | file_ra_state_init(&filep->f_ra, filep->f_mapping->host->i_mapping); | |
395 | res = filep; | |
f751c545 OK |
396 | out_free_name: |
397 | kfree(read_name); | |
ec4b0925 | 398 | out: |
156cd285 | 399 | nfs_free_fattr(fattr); |
ec4b0925 OK |
400 | return res; |
401 | out_stateowner: | |
402 | nfs4_put_state_owner(sp); | |
403 | out_ctx: | |
404 | put_nfs_open_context(ctx); | |
405 | out_filep: | |
406 | fput(filep); | |
f751c545 | 407 | goto out_free_name; |
ec4b0925 | 408 | } |
0cfcd405 DN |
409 | |
410 | static void __nfs42_ssc_close(struct file *filep) | |
ec4b0925 OK |
411 | { |
412 | struct nfs_open_context *ctx = nfs_file_open_context(filep); | |
413 | ||
414 | ctx->state->flags = 0; | |
415 | } | |
0cfcd405 DN |
416 | |
417 | static const struct nfs4_ssc_client_ops nfs4_ssc_clnt_ops_tbl = { | |
418 | .sco_open = __nfs42_ssc_open, | |
419 | .sco_close = __nfs42_ssc_close, | |
420 | }; | |
421 | ||
422 | /** | |
423 | * nfs42_ssc_register_ops - Wrapper to register NFS_V4 ops in nfs_common | |
424 | * | |
425 | * Return values: | |
426 | * None | |
427 | */ | |
428 | void nfs42_ssc_register_ops(void) | |
429 | { | |
430 | nfs42_ssc_register(&nfs4_ssc_clnt_ops_tbl); | |
431 | } | |
432 | ||
433 | /** | |
434 | * nfs42_ssc_unregister_ops - wrapper to un-register NFS_V4 ops in nfs_common | |
435 | * | |
436 | * Return values: | |
437 | * None. | |
438 | */ | |
439 | void nfs42_ssc_unregister_ops(void) | |
440 | { | |
441 | nfs42_ssc_unregister(&nfs4_ssc_clnt_ops_tbl); | |
442 | } | |
6b7153da | 443 | #endif /* CONFIG_NFS_V4_2 */ |
bea51b30 | 444 | |
e93a5e93 TM |
445 | static int nfs4_setlease(struct file *file, long arg, struct file_lock **lease, |
446 | void **priv) | |
447 | { | |
448 | return nfs4_proc_setlease(file, arg, lease, priv); | |
449 | } | |
450 | ||
ce4ef7c0 | 451 | const struct file_operations nfs4_file_operations = { |
3aa2d199 | 452 | .read_iter = nfs_file_read, |
edaf4369 | 453 | .write_iter = nfs_file_write, |
ce4ef7c0 BS |
454 | .mmap = nfs_file_mmap, |
455 | .open = nfs4_file_open, | |
5445b1fb | 456 | .flush = nfs4_file_flush, |
ce4ef7c0 | 457 | .release = nfs_file_release, |
4ff79bc7 | 458 | .fsync = nfs_file_fsync, |
ce4ef7c0 BS |
459 | .lock = nfs_lock, |
460 | .flock = nfs_flock, | |
82c156f8 | 461 | .splice_read = generic_file_splice_read, |
4da54c21 | 462 | .splice_write = iter_file_splice_write, |
ce4ef7c0 | 463 | .check_flags = nfs_check_flags, |
e93a5e93 | 464 | .setlease = nfs4_setlease, |
6b7153da | 465 | #ifdef CONFIG_NFS_V4_2 |
2e72448b | 466 | .copy_file_range = nfs4_copy_file_range, |
6b7153da CH |
467 | .llseek = nfs4_file_llseek, |
468 | .fallocate = nfs42_fallocate, | |
2e5dfc99 | 469 | .remap_file_range = nfs42_remap_file_range, |
6b7153da CH |
470 | #else |
471 | .llseek = nfs_file_llseek, | |
472 | #endif | |
ce4ef7c0 | 473 | }; |