Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3 | * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved. | |
4 | * | |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
7 | * of the GNU General Public License v.2. | |
8 | */ | |
9 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
15 | #include <linux/pagemap.h> | |
16 | #include <linux/uio.h> | |
17 | #include <linux/blkdev.h> | |
18 | #include <linux/mm.h> | |
19 | #include <linux/smp_lock.h> | |
18ec7d5c | 20 | #include <linux/fs.h> |
5c676f6d | 21 | #include <linux/gfs2_ondisk.h> |
71b86f56 SW |
22 | #include <linux/ext2_fs.h> |
23 | #include <linux/crc32.h> | |
7ea9ea83 | 24 | #include <linux/iflags.h> |
b3b94faa DT |
25 | #include <asm/semaphore.h> |
26 | #include <asm/uaccess.h> | |
27 | ||
28 | #include "gfs2.h" | |
5c676f6d SW |
29 | #include "lm_interface.h" |
30 | #include "incore.h" | |
b3b94faa DT |
31 | #include "bmap.h" |
32 | #include "dir.h" | |
33 | #include "glock.h" | |
34 | #include "glops.h" | |
35 | #include "inode.h" | |
b3b94faa DT |
36 | #include "lm.h" |
37 | #include "log.h" | |
38 | #include "meta_io.h" | |
39 | #include "ops_file.h" | |
40 | #include "ops_vm.h" | |
41 | #include "quota.h" | |
42 | #include "rgrp.h" | |
43 | #include "trans.h" | |
5c676f6d | 44 | #include "util.h" |
71b86f56 | 45 | #include "eaops.h" |
b3b94faa DT |
46 | |
47 | /* "bad" is for NFS support */ | |
48 | struct filldir_bad_entry { | |
49 | char *fbe_name; | |
50 | unsigned int fbe_length; | |
51 | uint64_t fbe_offset; | |
52 | struct gfs2_inum fbe_inum; | |
53 | unsigned int fbe_type; | |
54 | }; | |
55 | ||
56 | struct filldir_bad { | |
57 | struct gfs2_sbd *fdb_sbd; | |
58 | ||
59 | struct filldir_bad_entry *fdb_entry; | |
60 | unsigned int fdb_entry_num; | |
61 | unsigned int fdb_entry_off; | |
62 | ||
63 | char *fdb_name; | |
64 | unsigned int fdb_name_size; | |
65 | unsigned int fdb_name_off; | |
66 | }; | |
67 | ||
68 | /* For regular, non-NFS */ | |
69 | struct filldir_reg { | |
70 | struct gfs2_sbd *fdr_sbd; | |
71 | int fdr_prefetch; | |
72 | ||
73 | filldir_t fdr_filldir; | |
74 | void *fdr_opaque; | |
75 | }; | |
76 | ||
61a30dcb SW |
77 | /* |
78 | * Most fields left uninitialised to catch anybody who tries to | |
79 | * use them. f_flags set to prevent file_accessed() from touching | |
80 | * any other part of this. Its use is purely as a flag so that we | |
81 | * know (in readpage()) whether or not do to locking. | |
82 | */ | |
83 | struct file gfs2_internal_file_sentinal = { | |
84 | .f_flags = O_NOATIME|O_RDONLY, | |
85 | }; | |
86 | ||
18ec7d5c SW |
87 | static int gfs2_read_actor(read_descriptor_t *desc, struct page *page, |
88 | unsigned long offset, unsigned long size) | |
89 | { | |
90 | char *kaddr; | |
91 | unsigned long count = desc->count; | |
92 | ||
93 | if (size > count) | |
94 | size = count; | |
95 | ||
96 | kaddr = kmap(page); | |
97 | memcpy(desc->arg.buf, kaddr + offset, size); | |
98 | kunmap(page); | |
99 | ||
100 | desc->count = count - size; | |
101 | desc->written += size; | |
102 | desc->arg.buf += size; | |
103 | return size; | |
104 | } | |
105 | ||
106 | int gfs2_internal_read(struct gfs2_inode *ip, struct file_ra_state *ra_state, | |
107 | char *buf, loff_t *pos, unsigned size) | |
108 | { | |
109 | struct inode *inode = ip->i_vnode; | |
110 | read_descriptor_t desc; | |
111 | desc.written = 0; | |
112 | desc.arg.buf = buf; | |
113 | desc.count = size; | |
114 | desc.error = 0; | |
61a30dcb SW |
115 | do_generic_mapping_read(inode->i_mapping, ra_state, |
116 | &gfs2_internal_file_sentinal, pos, &desc, | |
117 | gfs2_read_actor); | |
18ec7d5c SW |
118 | return desc.written ? desc.written : desc.error; |
119 | } | |
b3b94faa DT |
120 | |
121 | /** | |
122 | * gfs2_llseek - seek to a location in a file | |
123 | * @file: the file | |
124 | * @offset: the offset | |
125 | * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END) | |
126 | * | |
127 | * SEEK_END requires the glock for the file because it references the | |
128 | * file's size. | |
129 | * | |
130 | * Returns: The new offset, or errno | |
131 | */ | |
132 | ||
133 | static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin) | |
134 | { | |
5c676f6d | 135 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
136 | struct gfs2_holder i_gh; |
137 | loff_t error; | |
138 | ||
b3b94faa DT |
139 | if (origin == 2) { |
140 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
141 | &i_gh); | |
142 | if (!error) { | |
143 | error = remote_llseek(file, offset, origin); | |
144 | gfs2_glock_dq_uninit(&i_gh); | |
145 | } | |
146 | } else | |
147 | error = remote_llseek(file, offset, origin); | |
148 | ||
149 | return error; | |
150 | } | |
151 | ||
b3b94faa | 152 | |
18ec7d5c SW |
153 | static ssize_t gfs2_direct_IO_read(struct kiocb *iocb, const struct iovec *iov, |
154 | loff_t offset, unsigned long nr_segs) | |
b3b94faa | 155 | { |
18ec7d5c SW |
156 | struct file *file = iocb->ki_filp; |
157 | struct address_space *mapping = file->f_mapping; | |
158 | ssize_t retval; | |
b3b94faa | 159 | |
18ec7d5c SW |
160 | retval = filemap_write_and_wait(mapping); |
161 | if (retval == 0) { | |
162 | retval = mapping->a_ops->direct_IO(READ, iocb, iov, offset, | |
163 | nr_segs); | |
b3b94faa | 164 | } |
18ec7d5c | 165 | return retval; |
b3b94faa DT |
166 | } |
167 | ||
168 | /** | |
18ec7d5c SW |
169 | * __gfs2_file_aio_read - The main GFS2 read function |
170 | * | |
171 | * N.B. This is almost, but not quite the same as __generic_file_aio_read() | |
172 | * the important subtle different being that inode->i_size isn't valid | |
173 | * unless we are holding a lock, and we do this _only_ on the O_DIRECT | |
174 | * path since otherwise locking is done entirely at the page cache | |
175 | * layer. | |
b3b94faa | 176 | */ |
18ec7d5c SW |
177 | static ssize_t __gfs2_file_aio_read(struct kiocb *iocb, |
178 | const struct iovec *iov, | |
179 | unsigned long nr_segs, loff_t *ppos) | |
b3b94faa | 180 | { |
18ec7d5c | 181 | struct file *filp = iocb->ki_filp; |
5c676f6d | 182 | struct gfs2_inode *ip = filp->f_mapping->host->u.generic_ip; |
b3b94faa | 183 | struct gfs2_holder gh; |
18ec7d5c SW |
184 | ssize_t retval; |
185 | unsigned long seg; | |
186 | size_t count; | |
187 | ||
188 | count = 0; | |
189 | for (seg = 0; seg < nr_segs; seg++) { | |
190 | const struct iovec *iv = &iov[seg]; | |
191 | ||
192 | /* | |
193 | * If any segment has a negative length, or the cumulative | |
194 | * length ever wraps negative then return -EINVAL. | |
195 | */ | |
d1665e41 SW |
196 | count += iv->iov_len; |
197 | if (unlikely((ssize_t)(count|iv->iov_len) < 0)) | |
198 | return -EINVAL; | |
199 | if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len)) | |
200 | continue; | |
201 | if (seg == 0) | |
202 | return -EFAULT; | |
203 | nr_segs = seg; | |
204 | count -= iv->iov_len; /* This segment is no good */ | |
205 | break; | |
18ec7d5c SW |
206 | } |
207 | ||
208 | /* coalesce the iovecs and go direct-to-BIO for O_DIRECT */ | |
209 | if (filp->f_flags & O_DIRECT) { | |
210 | loff_t pos = *ppos, size; | |
211 | struct address_space *mapping; | |
212 | struct inode *inode; | |
213 | ||
214 | mapping = filp->f_mapping; | |
215 | inode = mapping->host; | |
216 | retval = 0; | |
217 | if (!count) | |
218 | goto out; /* skip atime */ | |
219 | ||
220 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh); | |
221 | retval = gfs2_glock_nq_m_atime(1, &gh); | |
222 | if (retval) | |
223 | goto out; | |
d1665e41 SW |
224 | if (gfs2_is_stuffed(ip)) { |
225 | gfs2_glock_dq_m(1, &gh); | |
226 | gfs2_holder_uninit(&gh); | |
227 | goto fallback_to_normal; | |
228 | } | |
18ec7d5c SW |
229 | size = i_size_read(inode); |
230 | if (pos < size) { | |
d1665e41 | 231 | retval = gfs2_direct_IO_read(iocb, iov, pos, nr_segs); |
18ec7d5c SW |
232 | if (retval > 0 && !is_sync_kiocb(iocb)) |
233 | retval = -EIOCBQUEUED; | |
234 | if (retval > 0) | |
235 | *ppos = pos + retval; | |
b3b94faa | 236 | } |
18ec7d5c SW |
237 | file_accessed(filp); |
238 | gfs2_glock_dq_m(1, &gh); | |
239 | gfs2_holder_uninit(&gh); | |
b3b94faa | 240 | goto out; |
18ec7d5c | 241 | } |
b3b94faa | 242 | |
d1665e41 | 243 | fallback_to_normal: |
18ec7d5c SW |
244 | retval = 0; |
245 | if (count) { | |
246 | for (seg = 0; seg < nr_segs; seg++) { | |
247 | read_descriptor_t desc; | |
248 | ||
249 | desc.written = 0; | |
250 | desc.arg.buf = iov[seg].iov_base; | |
251 | desc.count = iov[seg].iov_len; | |
252 | if (desc.count == 0) | |
253 | continue; | |
254 | desc.error = 0; | |
255 | do_generic_file_read(filp,ppos,&desc,file_read_actor); | |
256 | retval += desc.written; | |
257 | if (desc.error) { | |
258 | retval = retval ?: desc.error; | |
259 | break; | |
260 | } | |
261 | } | |
262 | } | |
263 | out: | |
264 | return retval; | |
b3b94faa DT |
265 | } |
266 | ||
267 | /** | |
268 | * gfs2_read - Read bytes from a file | |
269 | * @file: The file to read from | |
270 | * @buf: The buffer to copy into | |
271 | * @size: The amount of data requested | |
272 | * @offset: The current file offset | |
273 | * | |
274 | * Outputs: Offset - updated according to number of bytes read | |
275 | * | |
276 | * Returns: The number of bytes read, errno on failure | |
277 | */ | |
278 | ||
18ec7d5c | 279 | static ssize_t gfs2_read(struct file *filp, char __user *buf, size_t size, |
b3b94faa DT |
280 | loff_t *offset) |
281 | { | |
b3b94faa | 282 | struct iovec local_iov = { .iov_base = buf, .iov_len = size }; |
18ec7d5c SW |
283 | struct kiocb kiocb; |
284 | ssize_t ret; | |
b3b94faa | 285 | |
18ec7d5c SW |
286 | init_sync_kiocb(&kiocb, filp); |
287 | ret = __gfs2_file_aio_read(&kiocb, &local_iov, 1, offset); | |
288 | if (-EIOCBQUEUED == ret) | |
289 | ret = wait_on_sync_kiocb(&kiocb); | |
290 | return ret; | |
b3b94faa DT |
291 | } |
292 | ||
18ec7d5c SW |
293 | static ssize_t gfs2_file_readv(struct file *filp, const struct iovec *iov, |
294 | unsigned long nr_segs, loff_t *ppos) | |
b3b94faa | 295 | { |
18ec7d5c SW |
296 | struct kiocb kiocb; |
297 | ssize_t ret; | |
b3b94faa | 298 | |
18ec7d5c SW |
299 | init_sync_kiocb(&kiocb, filp); |
300 | ret = __gfs2_file_aio_read(&kiocb, iov, nr_segs, ppos); | |
301 | if (-EIOCBQUEUED == ret) | |
302 | ret = wait_on_sync_kiocb(&kiocb); | |
303 | return ret; | |
b3b94faa DT |
304 | } |
305 | ||
18ec7d5c SW |
306 | static ssize_t gfs2_file_aio_read(struct kiocb *iocb, char __user *buf, |
307 | size_t count, loff_t pos) | |
b3b94faa | 308 | { |
18ec7d5c | 309 | struct iovec local_iov = { .iov_base = buf, .iov_len = count }; |
b3b94faa | 310 | |
18ec7d5c SW |
311 | BUG_ON(iocb->ki_pos != pos); |
312 | return __gfs2_file_aio_read(iocb, &local_iov, 1, &iocb->ki_pos); | |
b3b94faa DT |
313 | } |
314 | ||
b3b94faa DT |
315 | |
316 | /** | |
317 | * filldir_reg_func - Report a directory entry to the caller of gfs2_dir_read() | |
318 | * @opaque: opaque data used by the function | |
319 | * @name: the name of the directory entry | |
320 | * @length: the length of the name | |
321 | * @offset: the entry's offset in the directory | |
322 | * @inum: the inode number the entry points to | |
323 | * @type: the type of inode the entry points to | |
324 | * | |
325 | * Returns: 0 on success, 1 if buffer full | |
326 | */ | |
327 | ||
328 | static int filldir_reg_func(void *opaque, const char *name, unsigned int length, | |
329 | uint64_t offset, struct gfs2_inum *inum, | |
330 | unsigned int type) | |
331 | { | |
332 | struct filldir_reg *fdr = (struct filldir_reg *)opaque; | |
333 | struct gfs2_sbd *sdp = fdr->fdr_sbd; | |
334 | int error; | |
335 | ||
336 | error = fdr->fdr_filldir(fdr->fdr_opaque, name, length, offset, | |
337 | inum->no_formal_ino, type); | |
338 | if (error) | |
339 | return 1; | |
340 | ||
341 | if (fdr->fdr_prefetch && !(length == 1 && *name == '.')) { | |
342 | gfs2_glock_prefetch_num(sdp, | |
343 | inum->no_addr, &gfs2_inode_glops, | |
344 | LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY); | |
345 | gfs2_glock_prefetch_num(sdp, | |
346 | inum->no_addr, &gfs2_iopen_glops, | |
347 | LM_ST_SHARED, LM_FLAG_TRY); | |
348 | } | |
349 | ||
350 | return 0; | |
351 | } | |
352 | ||
353 | /** | |
354 | * readdir_reg - Read directory entries from a directory | |
355 | * @file: The directory to read from | |
356 | * @dirent: Buffer for dirents | |
357 | * @filldir: Function used to do the copying | |
358 | * | |
359 | * Returns: errno | |
360 | */ | |
361 | ||
362 | static int readdir_reg(struct file *file, void *dirent, filldir_t filldir) | |
363 | { | |
71b86f56 SW |
364 | struct inode *dir = file->f_mapping->host; |
365 | struct gfs2_inode *dip = dir->u.generic_ip; | |
b3b94faa DT |
366 | struct filldir_reg fdr; |
367 | struct gfs2_holder d_gh; | |
368 | uint64_t offset = file->f_pos; | |
369 | int error; | |
370 | ||
371 | fdr.fdr_sbd = dip->i_sbd; | |
372 | fdr.fdr_prefetch = 1; | |
373 | fdr.fdr_filldir = filldir; | |
374 | fdr.fdr_opaque = dirent; | |
375 | ||
376 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh); | |
377 | error = gfs2_glock_nq_atime(&d_gh); | |
378 | if (error) { | |
379 | gfs2_holder_uninit(&d_gh); | |
380 | return error; | |
381 | } | |
382 | ||
71b86f56 | 383 | error = gfs2_dir_read(dir, &offset, &fdr, filldir_reg_func); |
b3b94faa DT |
384 | |
385 | gfs2_glock_dq_uninit(&d_gh); | |
386 | ||
387 | file->f_pos = offset; | |
388 | ||
389 | return error; | |
390 | } | |
391 | ||
392 | /** | |
393 | * filldir_bad_func - Report a directory entry to the caller of gfs2_dir_read() | |
394 | * @opaque: opaque data used by the function | |
395 | * @name: the name of the directory entry | |
396 | * @length: the length of the name | |
397 | * @offset: the entry's offset in the directory | |
398 | * @inum: the inode number the entry points to | |
399 | * @type: the type of inode the entry points to | |
400 | * | |
401 | * For supporting NFS. | |
402 | * | |
403 | * Returns: 0 on success, 1 if buffer full | |
404 | */ | |
405 | ||
406 | static int filldir_bad_func(void *opaque, const char *name, unsigned int length, | |
407 | uint64_t offset, struct gfs2_inum *inum, | |
408 | unsigned int type) | |
409 | { | |
410 | struct filldir_bad *fdb = (struct filldir_bad *)opaque; | |
411 | struct gfs2_sbd *sdp = fdb->fdb_sbd; | |
412 | struct filldir_bad_entry *fbe; | |
413 | ||
414 | if (fdb->fdb_entry_off == fdb->fdb_entry_num || | |
415 | fdb->fdb_name_off + length > fdb->fdb_name_size) | |
416 | return 1; | |
417 | ||
418 | fbe = &fdb->fdb_entry[fdb->fdb_entry_off]; | |
419 | fbe->fbe_name = fdb->fdb_name + fdb->fdb_name_off; | |
420 | memcpy(fbe->fbe_name, name, length); | |
421 | fbe->fbe_length = length; | |
422 | fbe->fbe_offset = offset; | |
423 | fbe->fbe_inum = *inum; | |
424 | fbe->fbe_type = type; | |
425 | ||
426 | fdb->fdb_entry_off++; | |
427 | fdb->fdb_name_off += length; | |
428 | ||
429 | if (!(length == 1 && *name == '.')) { | |
430 | gfs2_glock_prefetch_num(sdp, | |
431 | inum->no_addr, &gfs2_inode_glops, | |
432 | LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY); | |
433 | gfs2_glock_prefetch_num(sdp, | |
434 | inum->no_addr, &gfs2_iopen_glops, | |
435 | LM_ST_SHARED, LM_FLAG_TRY); | |
436 | } | |
437 | ||
438 | return 0; | |
439 | } | |
440 | ||
441 | /** | |
442 | * readdir_bad - Read directory entries from a directory | |
443 | * @file: The directory to read from | |
444 | * @dirent: Buffer for dirents | |
445 | * @filldir: Function used to do the copying | |
446 | * | |
447 | * For supporting NFS. | |
448 | * | |
449 | * Returns: errno | |
450 | */ | |
451 | ||
452 | static int readdir_bad(struct file *file, void *dirent, filldir_t filldir) | |
453 | { | |
71b86f56 SW |
454 | struct inode *dir = file->f_mapping->host; |
455 | struct gfs2_inode *dip = dir->u.generic_ip; | |
b3b94faa DT |
456 | struct gfs2_sbd *sdp = dip->i_sbd; |
457 | struct filldir_reg fdr; | |
458 | unsigned int entries, size; | |
459 | struct filldir_bad *fdb; | |
460 | struct gfs2_holder d_gh; | |
461 | uint64_t offset = file->f_pos; | |
462 | unsigned int x; | |
463 | struct filldir_bad_entry *fbe; | |
464 | int error; | |
465 | ||
466 | entries = gfs2_tune_get(sdp, gt_entries_per_readdir); | |
467 | size = sizeof(struct filldir_bad) + | |
468 | entries * (sizeof(struct filldir_bad_entry) + GFS2_FAST_NAME_SIZE); | |
469 | ||
470 | fdb = kzalloc(size, GFP_KERNEL); | |
471 | if (!fdb) | |
472 | return -ENOMEM; | |
473 | ||
474 | fdb->fdb_sbd = sdp; | |
475 | fdb->fdb_entry = (struct filldir_bad_entry *)(fdb + 1); | |
476 | fdb->fdb_entry_num = entries; | |
477 | fdb->fdb_name = ((char *)fdb) + sizeof(struct filldir_bad) + | |
478 | entries * sizeof(struct filldir_bad_entry); | |
479 | fdb->fdb_name_size = entries * GFS2_FAST_NAME_SIZE; | |
480 | ||
481 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh); | |
482 | error = gfs2_glock_nq_atime(&d_gh); | |
483 | if (error) { | |
484 | gfs2_holder_uninit(&d_gh); | |
485 | goto out; | |
486 | } | |
487 | ||
71b86f56 | 488 | error = gfs2_dir_read(dir, &offset, fdb, filldir_bad_func); |
b3b94faa DT |
489 | |
490 | gfs2_glock_dq_uninit(&d_gh); | |
491 | ||
492 | fdr.fdr_sbd = sdp; | |
493 | fdr.fdr_prefetch = 0; | |
494 | fdr.fdr_filldir = filldir; | |
495 | fdr.fdr_opaque = dirent; | |
496 | ||
497 | for (x = 0; x < fdb->fdb_entry_off; x++) { | |
498 | fbe = &fdb->fdb_entry[x]; | |
499 | ||
500 | error = filldir_reg_func(&fdr, | |
501 | fbe->fbe_name, fbe->fbe_length, | |
502 | fbe->fbe_offset, | |
503 | &fbe->fbe_inum, fbe->fbe_type); | |
504 | if (error) { | |
505 | file->f_pos = fbe->fbe_offset; | |
506 | error = 0; | |
507 | goto out; | |
508 | } | |
509 | } | |
510 | ||
511 | file->f_pos = offset; | |
512 | ||
513 | out: | |
514 | kfree(fdb); | |
515 | ||
516 | return error; | |
517 | } | |
518 | ||
519 | /** | |
520 | * gfs2_readdir - Read directory entries from a directory | |
521 | * @file: The directory to read from | |
522 | * @dirent: Buffer for dirents | |
523 | * @filldir: Function used to do the copying | |
524 | * | |
525 | * Returns: errno | |
526 | */ | |
527 | ||
528 | static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir) | |
529 | { | |
530 | int error; | |
531 | ||
b3b94faa DT |
532 | if (strcmp(current->comm, "nfsd") != 0) |
533 | error = readdir_reg(file, dirent, filldir); | |
534 | else | |
535 | error = readdir_bad(file, dirent, filldir); | |
536 | ||
537 | return error; | |
538 | } | |
539 | ||
7ea9ea83 SW |
540 | static const u32 iflags_to_gfs2[32] = { |
541 | [iflag_Sync] = GFS2_DIF_SYNC, | |
542 | [iflag_Immutable] = GFS2_DIF_IMMUTABLE, | |
543 | [iflag_Append] = GFS2_DIF_APPENDONLY, | |
544 | [iflag_NoAtime] = GFS2_DIF_NOATIME, | |
545 | [iflag_Index] = GFS2_DIF_EXHASH, | |
546 | [iflag_JournalData] = GFS2_DIF_JDATA, | |
547 | [iflag_DirectIO] = GFS2_DIF_DIRECTIO, | |
548 | [iflag_InheritDirectIO] = GFS2_DIF_INHERIT_DIRECTIO, | |
549 | [iflag_InheritJdata] = GFS2_DIF_INHERIT_JDATA, | |
71b86f56 SW |
550 | }; |
551 | ||
7ea9ea83 SW |
552 | static const u32 gfs2_to_iflags[32] = { |
553 | [gfs2fl_Sync] = IFLAG_SYNC, | |
554 | [gfs2fl_Immutable] = IFLAG_IMMUTABLE, | |
555 | [gfs2fl_AppendOnly] = IFLAG_APPEND, | |
556 | [gfs2fl_NoAtime] = IFLAG_NOATIME, | |
557 | [gfs2fl_ExHash] = IFLAG_INDEX, | |
558 | [gfs2fl_Jdata] = IFLAG_JOURNAL_DATA, | |
559 | [gfs2fl_Directio] = IFLAG_DIRECTIO, | |
560 | [gfs2fl_InheritDirectio] = IFLAG_INHERITDIRECTIO, | |
561 | [gfs2fl_InheritJdata] = IFLAG_INHERITJDATA, | |
562 | }; | |
71b86f56 | 563 | |
7ea9ea83 | 564 | static int gfs2_get_flags(struct inode *inode, u32 __user *ptr) |
71b86f56 SW |
565 | { |
566 | struct gfs2_inode *ip = inode->u.generic_ip; | |
567 | struct gfs2_holder gh; | |
568 | int error; | |
7ea9ea83 | 569 | u32 iflags; |
71b86f56 SW |
570 | |
571 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh); | |
572 | error = gfs2_glock_nq_m_atime(1, &gh); | |
573 | if (error) | |
574 | return error; | |
55eccc6d | 575 | |
7ea9ea83 SW |
576 | iflags = iflags_cvt(gfs2_to_iflags, ip->i_di.di_flags); |
577 | if (put_user(iflags, ptr)) | |
71b86f56 SW |
578 | error = -EFAULT; |
579 | ||
580 | gfs2_glock_dq_m(1, &gh); | |
581 | gfs2_holder_uninit(&gh); | |
582 | return error; | |
583 | } | |
584 | ||
585 | /* Flags that can be set by user space */ | |
586 | #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \ | |
587 | GFS2_DIF_DIRECTIO| \ | |
588 | GFS2_DIF_IMMUTABLE| \ | |
589 | GFS2_DIF_APPENDONLY| \ | |
590 | GFS2_DIF_NOATIME| \ | |
591 | GFS2_DIF_SYNC| \ | |
592 | GFS2_DIF_SYSTEM| \ | |
593 | GFS2_DIF_INHERIT_DIRECTIO| \ | |
594 | GFS2_DIF_INHERIT_JDATA) | |
595 | ||
596 | /** | |
597 | * gfs2_set_flags - set flags on an inode | |
598 | * @inode: The inode | |
599 | * @flags: The flags to set | |
600 | * @mask: Indicates which flags are valid | |
601 | * | |
602 | */ | |
55eccc6d | 603 | static int do_gfs2_set_flags(struct inode *inode, u32 reqflags, u32 mask) |
71b86f56 SW |
604 | { |
605 | struct gfs2_inode *ip = inode->u.generic_ip; | |
55eccc6d | 606 | struct gfs2_sbd *sdp = ip->i_sbd; |
71b86f56 SW |
607 | struct buffer_head *bh; |
608 | struct gfs2_holder gh; | |
609 | int error; | |
55eccc6d | 610 | u32 new_flags, flags; |
71b86f56 SW |
611 | |
612 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
613 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
614 | if (error) | |
615 | return error; | |
616 | ||
55eccc6d SW |
617 | flags = ip->i_di.di_flags; |
618 | new_flags = (flags & ~mask) | (reqflags & mask); | |
71b86f56 SW |
619 | if ((new_flags ^ flags) == 0) |
620 | goto out; | |
621 | ||
622 | error = -EINVAL; | |
623 | if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET) | |
624 | goto out; | |
625 | ||
626 | if (S_ISDIR(inode->i_mode)) { | |
627 | if ((new_flags ^ flags) & (GFS2_DIF_JDATA | GFS2_DIF_DIRECTIO)) | |
628 | goto out; | |
629 | } else if (S_ISREG(inode->i_mode)) { | |
630 | if ((new_flags ^ flags) & (GFS2_DIF_INHERIT_DIRECTIO| | |
631 | GFS2_DIF_INHERIT_JDATA)) | |
632 | goto out; | |
633 | } else | |
634 | goto out; | |
635 | ||
636 | error = -EPERM; | |
637 | if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE)) | |
638 | goto out; | |
639 | if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY)) | |
640 | goto out; | |
641 | error = gfs2_repermission(inode, MAY_WRITE, NULL); | |
642 | if (error) | |
643 | goto out; | |
644 | ||
55eccc6d | 645 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); |
71b86f56 SW |
646 | if (error) |
647 | goto out; | |
55eccc6d SW |
648 | error = gfs2_meta_inode_buffer(ip, &bh); |
649 | if (error) | |
650 | goto out_trans_end; | |
71b86f56 SW |
651 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
652 | ip->i_di.di_flags = new_flags; | |
653 | gfs2_dinode_out(&ip->i_di, bh->b_data); | |
654 | brelse(bh); | |
55eccc6d SW |
655 | out_trans_end: |
656 | gfs2_trans_end(sdp); | |
71b86f56 SW |
657 | out: |
658 | gfs2_glock_dq_uninit(&gh); | |
659 | return error; | |
660 | } | |
661 | ||
7ea9ea83 | 662 | static int gfs2_set_flags(struct inode *inode, u32 __user *ptr) |
71b86f56 | 663 | { |
7ea9ea83 SW |
664 | u32 iflags, gfsflags; |
665 | if (get_user(iflags, ptr)) | |
71b86f56 | 666 | return -EFAULT; |
7ea9ea83 SW |
667 | gfsflags = iflags_cvt(iflags_to_gfs2, iflags); |
668 | return do_gfs2_set_flags(inode, gfsflags, ~0); | |
71b86f56 SW |
669 | } |
670 | ||
671 | int gfs2_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, | |
672 | unsigned long arg) | |
673 | { | |
674 | switch(cmd) { | |
7ea9ea83 SW |
675 | case IFLAGS_GET_IOC: |
676 | return gfs2_get_flags(inode, (u32 __user *)arg); | |
677 | case IFLAGS_SET_IOC: | |
678 | return gfs2_set_flags(inode, (u32 __user *)arg); | |
71b86f56 SW |
679 | } |
680 | return -ENOTTY; | |
681 | } | |
682 | ||
683 | ||
b3b94faa DT |
684 | /** |
685 | * gfs2_mmap - | |
686 | * @file: The file to map | |
687 | * @vma: The VMA which described the mapping | |
688 | * | |
689 | * Returns: 0 or error code | |
690 | */ | |
691 | ||
692 | static int gfs2_mmap(struct file *file, struct vm_area_struct *vma) | |
693 | { | |
5c676f6d | 694 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
695 | struct gfs2_holder i_gh; |
696 | int error; | |
697 | ||
b3b94faa DT |
698 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh); |
699 | error = gfs2_glock_nq_atime(&i_gh); | |
700 | if (error) { | |
701 | gfs2_holder_uninit(&i_gh); | |
702 | return error; | |
703 | } | |
704 | ||
18ec7d5c SW |
705 | /* This is VM_MAYWRITE instead of VM_WRITE because a call |
706 | to mprotect() can turn on VM_WRITE later. */ | |
707 | ||
708 | if ((vma->vm_flags & (VM_MAYSHARE | VM_MAYWRITE)) == | |
709 | (VM_MAYSHARE | VM_MAYWRITE)) | |
710 | vma->vm_ops = &gfs2_vm_ops_sharewrite; | |
711 | else | |
712 | vma->vm_ops = &gfs2_vm_ops_private; | |
b3b94faa DT |
713 | |
714 | gfs2_glock_dq_uninit(&i_gh); | |
715 | ||
716 | return error; | |
717 | } | |
718 | ||
719 | /** | |
720 | * gfs2_open - open a file | |
721 | * @inode: the inode to open | |
722 | * @file: the struct file for this opening | |
723 | * | |
724 | * Returns: errno | |
725 | */ | |
726 | ||
727 | static int gfs2_open(struct inode *inode, struct file *file) | |
728 | { | |
5c676f6d | 729 | struct gfs2_inode *ip = inode->u.generic_ip; |
b3b94faa DT |
730 | struct gfs2_holder i_gh; |
731 | struct gfs2_file *fp; | |
732 | int error; | |
733 | ||
b3b94faa DT |
734 | fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL); |
735 | if (!fp) | |
736 | return -ENOMEM; | |
737 | ||
f55ab26a | 738 | mutex_init(&fp->f_fl_mutex); |
b3b94faa | 739 | |
5c676f6d SW |
740 | gfs2_assert_warn(ip->i_sbd, !file->private_data); |
741 | file->private_data = fp; | |
b3b94faa DT |
742 | |
743 | if (S_ISREG(ip->i_di.di_mode)) { | |
744 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
745 | &i_gh); | |
746 | if (error) | |
747 | goto fail; | |
748 | ||
749 | if (!(file->f_flags & O_LARGEFILE) && | |
750 | ip->i_di.di_size > MAX_NON_LFS) { | |
751 | error = -EFBIG; | |
752 | goto fail_gunlock; | |
753 | } | |
754 | ||
755 | /* Listen to the Direct I/O flag */ | |
756 | ||
757 | if (ip->i_di.di_flags & GFS2_DIF_DIRECTIO) | |
758 | file->f_flags |= O_DIRECT; | |
759 | ||
b3b94faa DT |
760 | gfs2_glock_dq_uninit(&i_gh); |
761 | } | |
762 | ||
763 | return 0; | |
764 | ||
765 | fail_gunlock: | |
766 | gfs2_glock_dq_uninit(&i_gh); | |
767 | ||
768 | fail: | |
5c676f6d | 769 | file->private_data = NULL; |
b3b94faa DT |
770 | kfree(fp); |
771 | ||
772 | return error; | |
773 | } | |
774 | ||
775 | /** | |
776 | * gfs2_close - called to close a struct file | |
777 | * @inode: the inode the struct file belongs to | |
778 | * @file: the struct file being closed | |
779 | * | |
780 | * Returns: errno | |
781 | */ | |
782 | ||
783 | static int gfs2_close(struct inode *inode, struct file *file) | |
784 | { | |
5c676f6d | 785 | struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; |
b3b94faa DT |
786 | struct gfs2_file *fp; |
787 | ||
5c676f6d SW |
788 | fp = file->private_data; |
789 | file->private_data = NULL; | |
b3b94faa DT |
790 | |
791 | if (gfs2_assert_warn(sdp, fp)) | |
792 | return -EIO; | |
793 | ||
794 | kfree(fp); | |
795 | ||
796 | return 0; | |
797 | } | |
798 | ||
799 | /** | |
800 | * gfs2_fsync - sync the dirty data for a file (across the cluster) | |
801 | * @file: the file that points to the dentry (we ignore this) | |
802 | * @dentry: the dentry that points to the inode to sync | |
803 | * | |
804 | * Returns: errno | |
805 | */ | |
806 | ||
807 | static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync) | |
808 | { | |
5c676f6d | 809 | struct gfs2_inode *ip = dentry->d_inode->u.generic_ip; |
b3b94faa | 810 | |
b3b94faa DT |
811 | gfs2_log_flush_glock(ip->i_gl); |
812 | ||
813 | return 0; | |
814 | } | |
815 | ||
816 | /** | |
817 | * gfs2_lock - acquire/release a posix lock on a file | |
818 | * @file: the file pointer | |
819 | * @cmd: either modify or retrieve lock state, possibly wait | |
820 | * @fl: type and range of lock | |
821 | * | |
822 | * Returns: errno | |
823 | */ | |
824 | ||
825 | static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl) | |
826 | { | |
5c676f6d | 827 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
828 | struct gfs2_sbd *sdp = ip->i_sbd; |
829 | struct lm_lockname name = | |
830 | { .ln_number = ip->i_num.no_addr, | |
831 | .ln_type = LM_TYPE_PLOCK }; | |
832 | ||
b3b94faa DT |
833 | if (!(fl->fl_flags & FL_POSIX)) |
834 | return -ENOLCK; | |
835 | if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID) | |
836 | return -ENOLCK; | |
837 | ||
838 | if (sdp->sd_args.ar_localflocks) { | |
839 | if (IS_GETLK(cmd)) { | |
8628de05 SW |
840 | struct file_lock tmp; |
841 | int ret; | |
842 | ret = posix_test_lock(file, fl, &tmp); | |
b3b94faa | 843 | fl->fl_type = F_UNLCK; |
8628de05 SW |
844 | if (ret) |
845 | memcpy(fl, &tmp, sizeof(struct file_lock)); | |
b3b94faa DT |
846 | return 0; |
847 | } else { | |
8628de05 | 848 | return posix_lock_file_wait(file, fl); |
b3b94faa DT |
849 | } |
850 | } | |
851 | ||
852 | if (IS_GETLK(cmd)) | |
853 | return gfs2_lm_plock_get(sdp, &name, file, fl); | |
854 | else if (fl->fl_type == F_UNLCK) | |
855 | return gfs2_lm_punlock(sdp, &name, file, fl); | |
856 | else | |
857 | return gfs2_lm_plock(sdp, &name, file, cmd, fl); | |
858 | } | |
859 | ||
860 | /** | |
861 | * gfs2_sendfile - Send bytes to a file or socket | |
862 | * @in_file: The file to read from | |
863 | * @out_file: The file to write to | |
864 | * @count: The amount of data | |
865 | * @offset: The beginning file offset | |
866 | * | |
867 | * Outputs: offset - updated according to number of bytes read | |
868 | * | |
869 | * Returns: The number of bytes sent, errno on failure | |
870 | */ | |
871 | ||
872 | static ssize_t gfs2_sendfile(struct file *in_file, loff_t *offset, size_t count, | |
873 | read_actor_t actor, void *target) | |
874 | { | |
18ec7d5c | 875 | return generic_file_sendfile(in_file, offset, count, actor, target); |
b3b94faa DT |
876 | } |
877 | ||
878 | static int do_flock(struct file *file, int cmd, struct file_lock *fl) | |
879 | { | |
5c676f6d | 880 | struct gfs2_file *fp = file->private_data; |
b3b94faa | 881 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
55eccc6d | 882 | struct gfs2_inode *ip = file->f_dentry->d_inode->u.generic_ip; |
b3b94faa DT |
883 | struct gfs2_glock *gl; |
884 | unsigned int state; | |
885 | int flags; | |
886 | int error = 0; | |
887 | ||
888 | state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
889 | flags = ((IS_SETLKW(cmd)) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; | |
890 | ||
f55ab26a | 891 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
892 | |
893 | gl = fl_gh->gh_gl; | |
894 | if (gl) { | |
895 | if (fl_gh->gh_state == state) | |
896 | goto out; | |
897 | gfs2_glock_hold(gl); | |
898 | flock_lock_file_wait(file, | |
899 | &(struct file_lock){.fl_type = F_UNLCK}); | |
900 | gfs2_glock_dq_uninit(fl_gh); | |
901 | } else { | |
902 | error = gfs2_glock_get(ip->i_sbd, | |
903 | ip->i_num.no_addr, &gfs2_flock_glops, | |
904 | CREATE, &gl); | |
905 | if (error) | |
906 | goto out; | |
907 | } | |
908 | ||
909 | gfs2_holder_init(gl, state, flags, fl_gh); | |
910 | gfs2_glock_put(gl); | |
911 | ||
912 | error = gfs2_glock_nq(fl_gh); | |
913 | if (error) { | |
914 | gfs2_holder_uninit(fl_gh); | |
915 | if (error == GLR_TRYFAILED) | |
916 | error = -EAGAIN; | |
917 | } else { | |
918 | error = flock_lock_file_wait(file, fl); | |
919 | gfs2_assert_warn(ip->i_sbd, !error); | |
920 | } | |
921 | ||
922 | out: | |
f55ab26a | 923 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
924 | |
925 | return error; | |
926 | } | |
927 | ||
928 | static void do_unflock(struct file *file, struct file_lock *fl) | |
929 | { | |
5c676f6d | 930 | struct gfs2_file *fp = file->private_data; |
b3b94faa DT |
931 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
932 | ||
f55ab26a | 933 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
934 | flock_lock_file_wait(file, fl); |
935 | if (fl_gh->gh_gl) | |
936 | gfs2_glock_dq_uninit(fl_gh); | |
f55ab26a | 937 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
938 | } |
939 | ||
940 | /** | |
941 | * gfs2_flock - acquire/release a flock lock on a file | |
942 | * @file: the file pointer | |
943 | * @cmd: either modify or retrieve lock state, possibly wait | |
944 | * @fl: type and range of lock | |
945 | * | |
946 | * Returns: errno | |
947 | */ | |
948 | ||
949 | static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl) | |
950 | { | |
5c676f6d | 951 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
952 | struct gfs2_sbd *sdp = ip->i_sbd; |
953 | ||
b3b94faa DT |
954 | if (!(fl->fl_flags & FL_FLOCK)) |
955 | return -ENOLCK; | |
956 | if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID) | |
957 | return -ENOLCK; | |
958 | ||
959 | if (sdp->sd_args.ar_localflocks) | |
960 | return flock_lock_file_wait(file, fl); | |
961 | ||
962 | if (fl->fl_type == F_UNLCK) { | |
963 | do_unflock(file, fl); | |
964 | return 0; | |
965 | } else | |
966 | return do_flock(file, cmd, fl); | |
967 | } | |
968 | ||
969 | struct file_operations gfs2_file_fops = { | |
970 | .llseek = gfs2_llseek, | |
971 | .read = gfs2_read, | |
18ec7d5c SW |
972 | .readv = gfs2_file_readv, |
973 | .aio_read = gfs2_file_aio_read, | |
974 | .write = generic_file_write, | |
975 | .writev = generic_file_writev, | |
976 | .aio_write = generic_file_aio_write, | |
71b86f56 | 977 | .ioctl = gfs2_ioctl, |
b3b94faa DT |
978 | .mmap = gfs2_mmap, |
979 | .open = gfs2_open, | |
980 | .release = gfs2_close, | |
981 | .fsync = gfs2_fsync, | |
982 | .lock = gfs2_lock, | |
983 | .sendfile = gfs2_sendfile, | |
984 | .flock = gfs2_flock, | |
8628de05 SW |
985 | .splice_read = generic_file_splice_read, |
986 | .splice_write = generic_file_splice_write, | |
b3b94faa DT |
987 | }; |
988 | ||
989 | struct file_operations gfs2_dir_fops = { | |
990 | .readdir = gfs2_readdir, | |
71b86f56 | 991 | .ioctl = gfs2_ioctl, |
b3b94faa DT |
992 | .open = gfs2_open, |
993 | .release = gfs2_close, | |
994 | .fsync = gfs2_fsync, | |
995 | .lock = gfs2_lock, | |
996 | .flock = gfs2_flock, | |
997 | }; | |
998 |