Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3a8a9a10 | 3 | * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
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 | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
b3b94faa DT |
10 | #include <linux/slab.h> |
11 | #include <linux/spinlock.h> | |
12 | #include <linux/completion.h> | |
13 | #include <linux/buffer_head.h> | |
14 | #include <linux/pagemap.h> | |
15 | #include <linux/uio.h> | |
16 | #include <linux/blkdev.h> | |
17 | #include <linux/mm.h> | |
f58ba889 | 18 | #include <linux/mount.h> |
18ec7d5c | 19 | #include <linux/fs.h> |
5c676f6d | 20 | #include <linux/gfs2_ondisk.h> |
71b86f56 SW |
21 | #include <linux/ext2_fs.h> |
22 | #include <linux/crc32.h> | |
33c3de32 | 23 | #include <linux/writeback.h> |
b3b94faa | 24 | #include <asm/uaccess.h> |
f057f6cd SW |
25 | #include <linux/dlm.h> |
26 | #include <linux/dlm_plock.h> | |
b3b94faa DT |
27 | |
28 | #include "gfs2.h" | |
5c676f6d | 29 | #include "incore.h" |
b3b94faa DT |
30 | #include "bmap.h" |
31 | #include "dir.h" | |
32 | #include "glock.h" | |
33 | #include "glops.h" | |
34 | #include "inode.h" | |
b3b94faa DT |
35 | #include "log.h" |
36 | #include "meta_io.h" | |
b3b94faa DT |
37 | #include "quota.h" |
38 | #include "rgrp.h" | |
39 | #include "trans.h" | |
5c676f6d | 40 | #include "util.h" |
b3b94faa | 41 | |
b3b94faa DT |
42 | /** |
43 | * gfs2_llseek - seek to a location in a file | |
44 | * @file: the file | |
45 | * @offset: the offset | |
46 | * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END) | |
47 | * | |
48 | * SEEK_END requires the glock for the file because it references the | |
49 | * file's size. | |
50 | * | |
51 | * Returns: The new offset, or errno | |
52 | */ | |
53 | ||
54 | static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin) | |
55 | { | |
feaa7bba | 56 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
b3b94faa DT |
57 | struct gfs2_holder i_gh; |
58 | loff_t error; | |
59 | ||
b3b94faa DT |
60 | if (origin == 2) { |
61 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
62 | &i_gh); | |
63 | if (!error) { | |
9465efc9 | 64 | error = generic_file_llseek_unlocked(file, offset, origin); |
b3b94faa DT |
65 | gfs2_glock_dq_uninit(&i_gh); |
66 | } | |
67 | } else | |
9465efc9 | 68 | error = generic_file_llseek_unlocked(file, offset, origin); |
b3b94faa DT |
69 | |
70 | return error; | |
71 | } | |
72 | ||
b3b94faa | 73 | /** |
f0e522a9 | 74 | * gfs2_readdir - Read directory entries from a directory |
b3b94faa DT |
75 | * @file: The directory to read from |
76 | * @dirent: Buffer for dirents | |
77 | * @filldir: Function used to do the copying | |
78 | * | |
79 | * Returns: errno | |
80 | */ | |
81 | ||
f0e522a9 | 82 | static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir) |
b3b94faa | 83 | { |
71b86f56 | 84 | struct inode *dir = file->f_mapping->host; |
feaa7bba | 85 | struct gfs2_inode *dip = GFS2_I(dir); |
b3b94faa | 86 | struct gfs2_holder d_gh; |
cd915493 | 87 | u64 offset = file->f_pos; |
b3b94faa DT |
88 | int error; |
89 | ||
719ee344 SW |
90 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh); |
91 | error = gfs2_glock_nq(&d_gh); | |
b3b94faa DT |
92 | if (error) { |
93 | gfs2_holder_uninit(&d_gh); | |
94 | return error; | |
95 | } | |
96 | ||
3699e3a4 | 97 | error = gfs2_dir_read(dir, &offset, dirent, filldir); |
b3b94faa DT |
98 | |
99 | gfs2_glock_dq_uninit(&d_gh); | |
100 | ||
b3b94faa DT |
101 | file->f_pos = offset; |
102 | ||
b3b94faa DT |
103 | return error; |
104 | } | |
105 | ||
128e5eba SW |
106 | /** |
107 | * fsflags_cvt | |
108 | * @table: A table of 32 u32 flags | |
109 | * @val: a 32 bit value to convert | |
110 | * | |
111 | * This function can be used to convert between fsflags values and | |
112 | * GFS2's own flags values. | |
113 | * | |
114 | * Returns: the converted flags | |
115 | */ | |
116 | static u32 fsflags_cvt(const u32 *table, u32 val) | |
117 | { | |
118 | u32 res = 0; | |
119 | while(val) { | |
120 | if (val & 1) | |
121 | res |= *table; | |
122 | table++; | |
123 | val >>= 1; | |
124 | } | |
125 | return res; | |
126 | } | |
b3b94faa | 127 | |
128e5eba SW |
128 | static const u32 fsflags_to_gfs2[32] = { |
129 | [3] = GFS2_DIF_SYNC, | |
130 | [4] = GFS2_DIF_IMMUTABLE, | |
131 | [5] = GFS2_DIF_APPENDONLY, | |
132 | [7] = GFS2_DIF_NOATIME, | |
133 | [12] = GFS2_DIF_EXHASH, | |
b9af7ca6 | 134 | [14] = GFS2_DIF_INHERIT_JDATA, |
71b86f56 SW |
135 | }; |
136 | ||
128e5eba SW |
137 | static const u32 gfs2_to_fsflags[32] = { |
138 | [gfs2fl_Sync] = FS_SYNC_FL, | |
139 | [gfs2fl_Immutable] = FS_IMMUTABLE_FL, | |
140 | [gfs2fl_AppendOnly] = FS_APPEND_FL, | |
141 | [gfs2fl_NoAtime] = FS_NOATIME_FL, | |
142 | [gfs2fl_ExHash] = FS_INDEX_FL, | |
128e5eba | 143 | [gfs2fl_InheritJdata] = FS_JOURNAL_DATA_FL, |
7ea9ea83 | 144 | }; |
71b86f56 | 145 | |
b09e593d | 146 | static int gfs2_get_flags(struct file *filp, u32 __user *ptr) |
71b86f56 | 147 | { |
81454098 | 148 | struct inode *inode = filp->f_path.dentry->d_inode; |
feaa7bba | 149 | struct gfs2_inode *ip = GFS2_I(inode); |
71b86f56 SW |
150 | struct gfs2_holder gh; |
151 | int error; | |
128e5eba | 152 | u32 fsflags; |
71b86f56 | 153 | |
719ee344 SW |
154 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh); |
155 | error = gfs2_glock_nq(&gh); | |
71b86f56 SW |
156 | if (error) |
157 | return error; | |
907b9bce | 158 | |
383f01fb SW |
159 | fsflags = fsflags_cvt(gfs2_to_fsflags, ip->i_diskflags); |
160 | if (!S_ISDIR(inode->i_mode) && ip->i_diskflags & GFS2_DIF_JDATA) | |
c9f6a6bb | 161 | fsflags |= FS_JOURNAL_DATA_FL; |
128e5eba | 162 | if (put_user(fsflags, ptr)) |
71b86f56 SW |
163 | error = -EFAULT; |
164 | ||
3cc3f710 | 165 | gfs2_glock_dq(&gh); |
71b86f56 SW |
166 | gfs2_holder_uninit(&gh); |
167 | return error; | |
168 | } | |
169 | ||
6b124d8d SW |
170 | void gfs2_set_inode_flags(struct inode *inode) |
171 | { | |
172 | struct gfs2_inode *ip = GFS2_I(inode); | |
6b124d8d SW |
173 | unsigned int flags = inode->i_flags; |
174 | ||
175 | flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); | |
383f01fb | 176 | if (ip->i_diskflags & GFS2_DIF_IMMUTABLE) |
6b124d8d | 177 | flags |= S_IMMUTABLE; |
383f01fb | 178 | if (ip->i_diskflags & GFS2_DIF_APPENDONLY) |
6b124d8d | 179 | flags |= S_APPEND; |
383f01fb | 180 | if (ip->i_diskflags & GFS2_DIF_NOATIME) |
6b124d8d | 181 | flags |= S_NOATIME; |
383f01fb | 182 | if (ip->i_diskflags & GFS2_DIF_SYNC) |
6b124d8d SW |
183 | flags |= S_SYNC; |
184 | inode->i_flags = flags; | |
185 | } | |
186 | ||
71b86f56 SW |
187 | /* Flags that can be set by user space */ |
188 | #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \ | |
71b86f56 SW |
189 | GFS2_DIF_IMMUTABLE| \ |
190 | GFS2_DIF_APPENDONLY| \ | |
191 | GFS2_DIF_NOATIME| \ | |
192 | GFS2_DIF_SYNC| \ | |
193 | GFS2_DIF_SYSTEM| \ | |
71b86f56 SW |
194 | GFS2_DIF_INHERIT_JDATA) |
195 | ||
196 | /** | |
197 | * gfs2_set_flags - set flags on an inode | |
198 | * @inode: The inode | |
199 | * @flags: The flags to set | |
200 | * @mask: Indicates which flags are valid | |
201 | * | |
202 | */ | |
b09e593d | 203 | static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask) |
71b86f56 | 204 | { |
81454098 | 205 | struct inode *inode = filp->f_path.dentry->d_inode; |
feaa7bba SW |
206 | struct gfs2_inode *ip = GFS2_I(inode); |
207 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
71b86f56 SW |
208 | struct buffer_head *bh; |
209 | struct gfs2_holder gh; | |
210 | int error; | |
55eccc6d | 211 | u32 new_flags, flags; |
71b86f56 | 212 | |
f58ba889 | 213 | error = mnt_want_write(filp->f_path.mnt); |
52f341cf | 214 | if (error) |
71b86f56 SW |
215 | return error; |
216 | ||
f58ba889 MS |
217 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); |
218 | if (error) | |
219 | goto out_drop_write; | |
220 | ||
7df0e039 SW |
221 | error = -EACCES; |
222 | if (!is_owner_or_cap(inode)) | |
223 | goto out; | |
224 | ||
225 | error = 0; | |
383f01fb | 226 | flags = ip->i_diskflags; |
55eccc6d | 227 | new_flags = (flags & ~mask) | (reqflags & mask); |
71b86f56 SW |
228 | if ((new_flags ^ flags) == 0) |
229 | goto out; | |
230 | ||
231 | error = -EINVAL; | |
232 | if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET) | |
233 | goto out; | |
234 | ||
71b86f56 SW |
235 | error = -EPERM; |
236 | if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE)) | |
237 | goto out; | |
238 | if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY)) | |
239 | goto out; | |
907b9bce | 240 | if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) && |
b9cb9813 | 241 | !capable(CAP_LINUX_IMMUTABLE)) |
71b86f56 | 242 | goto out; |
b9cb9813 | 243 | if (!IS_IMMUTABLE(inode)) { |
f58ba889 | 244 | error = gfs2_permission(inode, MAY_WRITE); |
b9cb9813 SW |
245 | if (error) |
246 | goto out; | |
247 | } | |
5561093e SW |
248 | if ((flags ^ new_flags) & GFS2_DIF_JDATA) { |
249 | if (flags & GFS2_DIF_JDATA) | |
250 | gfs2_log_flush(sdp, ip->i_gl); | |
251 | error = filemap_fdatawrite(inode->i_mapping); | |
252 | if (error) | |
253 | goto out; | |
254 | error = filemap_fdatawait(inode->i_mapping); | |
255 | if (error) | |
256 | goto out; | |
257 | } | |
55eccc6d | 258 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); |
71b86f56 SW |
259 | if (error) |
260 | goto out; | |
55eccc6d SW |
261 | error = gfs2_meta_inode_buffer(ip, &bh); |
262 | if (error) | |
263 | goto out_trans_end; | |
71b86f56 | 264 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
383f01fb | 265 | ip->i_diskflags = new_flags; |
539e5d6b | 266 | gfs2_dinode_out(ip, bh->b_data); |
71b86f56 | 267 | brelse(bh); |
6b124d8d | 268 | gfs2_set_inode_flags(inode); |
5561093e | 269 | gfs2_set_aops(inode); |
55eccc6d SW |
270 | out_trans_end: |
271 | gfs2_trans_end(sdp); | |
71b86f56 SW |
272 | out: |
273 | gfs2_glock_dq_uninit(&gh); | |
f58ba889 MS |
274 | out_drop_write: |
275 | mnt_drop_write(filp->f_path.mnt); | |
71b86f56 SW |
276 | return error; |
277 | } | |
278 | ||
b09e593d | 279 | static int gfs2_set_flags(struct file *filp, u32 __user *ptr) |
71b86f56 | 280 | { |
b9af7ca6 | 281 | struct inode *inode = filp->f_path.dentry->d_inode; |
128e5eba | 282 | u32 fsflags, gfsflags; |
7df0e039 | 283 | |
128e5eba | 284 | if (get_user(fsflags, ptr)) |
71b86f56 | 285 | return -EFAULT; |
7df0e039 | 286 | |
128e5eba | 287 | gfsflags = fsflags_cvt(fsflags_to_gfs2, fsflags); |
b9af7ca6 SW |
288 | if (!S_ISDIR(inode->i_mode)) { |
289 | if (gfsflags & GFS2_DIF_INHERIT_JDATA) | |
290 | gfsflags ^= (GFS2_DIF_JDATA | GFS2_DIF_INHERIT_JDATA); | |
b9af7ca6 SW |
291 | return do_gfs2_set_flags(filp, gfsflags, ~0); |
292 | } | |
293 | return do_gfs2_set_flags(filp, gfsflags, ~GFS2_DIF_JDATA); | |
71b86f56 SW |
294 | } |
295 | ||
b09e593d | 296 | static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) |
71b86f56 SW |
297 | { |
298 | switch(cmd) { | |
128e5eba | 299 | case FS_IOC_GETFLAGS: |
b09e593d | 300 | return gfs2_get_flags(filp, (u32 __user *)arg); |
128e5eba | 301 | case FS_IOC_SETFLAGS: |
b09e593d | 302 | return gfs2_set_flags(filp, (u32 __user *)arg); |
71b86f56 SW |
303 | } |
304 | return -ENOTTY; | |
305 | } | |
306 | ||
3cc3f710 SW |
307 | /** |
308 | * gfs2_allocate_page_backing - Use bmap to allocate blocks | |
309 | * @page: The (locked) page to allocate backing for | |
310 | * | |
311 | * We try to allocate all the blocks required for the page in | |
312 | * one go. This might fail for various reasons, so we keep | |
313 | * trying until all the blocks to back this page are allocated. | |
314 | * If some of the blocks are already allocated, thats ok too. | |
315 | */ | |
316 | ||
317 | static int gfs2_allocate_page_backing(struct page *page) | |
318 | { | |
319 | struct inode *inode = page->mapping->host; | |
320 | struct buffer_head bh; | |
321 | unsigned long size = PAGE_CACHE_SIZE; | |
322 | u64 lblock = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
323 | ||
324 | do { | |
325 | bh.b_state = 0; | |
326 | bh.b_size = size; | |
e9e1ef2b | 327 | gfs2_block_map(inode, lblock, &bh, 1); |
3cc3f710 SW |
328 | if (!buffer_mapped(&bh)) |
329 | return -EIO; | |
330 | size -= bh.b_size; | |
331 | lblock += (bh.b_size >> inode->i_blkbits); | |
332 | } while(size > 0); | |
333 | return 0; | |
334 | } | |
335 | ||
336 | /** | |
337 | * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable | |
338 | * @vma: The virtual memory area | |
339 | * @page: The page which is about to become writable | |
340 | * | |
341 | * When the page becomes writable, we need to ensure that we have | |
342 | * blocks allocated on disk to back that page. | |
343 | */ | |
344 | ||
c2ec175c | 345 | static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) |
3cc3f710 | 346 | { |
c2ec175c | 347 | struct page *page = vmf->page; |
3cc3f710 SW |
348 | struct inode *inode = vma->vm_file->f_path.dentry->d_inode; |
349 | struct gfs2_inode *ip = GFS2_I(inode); | |
350 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
351 | unsigned long last_index; | |
c8f554b9 | 352 | u64 pos = page->index << PAGE_CACHE_SHIFT; |
3cc3f710 SW |
353 | unsigned int data_blocks, ind_blocks, rblocks; |
354 | int alloc_required = 0; | |
355 | struct gfs2_holder gh; | |
356 | struct gfs2_alloc *al; | |
357 | int ret; | |
358 | ||
719ee344 SW |
359 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); |
360 | ret = gfs2_glock_nq(&gh); | |
3cc3f710 SW |
361 | if (ret) |
362 | goto out; | |
363 | ||
9c538837 SW |
364 | set_bit(GLF_DIRTY, &ip->i_gl->gl_flags); |
365 | set_bit(GIF_SW_PAGED, &ip->i_flags); | |
366 | ||
3cc3f710 SW |
367 | ret = gfs2_write_alloc_required(ip, pos, PAGE_CACHE_SIZE, &alloc_required); |
368 | if (ret || !alloc_required) | |
369 | goto out_unlock; | |
6dbd8224 | 370 | ret = -ENOMEM; |
3cc3f710 | 371 | al = gfs2_alloc_get(ip); |
6dbd8224 SW |
372 | if (al == NULL) |
373 | goto out_unlock; | |
374 | ||
d82661d9 | 375 | ret = gfs2_quota_lock_check(ip); |
3cc3f710 SW |
376 | if (ret) |
377 | goto out_alloc_put; | |
7ed122e4 | 378 | gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks); |
3cc3f710 SW |
379 | al->al_requested = data_blocks + ind_blocks; |
380 | ret = gfs2_inplace_reserve(ip); | |
381 | if (ret) | |
382 | goto out_quota_unlock; | |
383 | ||
384 | rblocks = RES_DINODE + ind_blocks; | |
385 | if (gfs2_is_jdata(ip)) | |
386 | rblocks += data_blocks ? data_blocks : 1; | |
387 | if (ind_blocks || data_blocks) | |
388 | rblocks += RES_STATFS + RES_QUOTA; | |
389 | ret = gfs2_trans_begin(sdp, rblocks, 0); | |
390 | if (ret) | |
391 | goto out_trans_fail; | |
392 | ||
393 | lock_page(page); | |
394 | ret = -EINVAL; | |
395 | last_index = ip->i_inode.i_size >> PAGE_CACHE_SHIFT; | |
396 | if (page->index > last_index) | |
397 | goto out_unlock_page; | |
b7fe2e39 | 398 | ret = 0; |
3cc3f710 SW |
399 | if (!PageUptodate(page) || page->mapping != ip->i_inode.i_mapping) |
400 | goto out_unlock_page; | |
401 | if (gfs2_is_stuffed(ip)) { | |
402 | ret = gfs2_unstuff_dinode(ip, page); | |
403 | if (ret) | |
404 | goto out_unlock_page; | |
405 | } | |
406 | ret = gfs2_allocate_page_backing(page); | |
407 | ||
408 | out_unlock_page: | |
409 | unlock_page(page); | |
410 | gfs2_trans_end(sdp); | |
411 | out_trans_fail: | |
412 | gfs2_inplace_release(ip); | |
413 | out_quota_unlock: | |
414 | gfs2_quota_unlock(ip); | |
415 | out_alloc_put: | |
416 | gfs2_alloc_put(ip); | |
417 | out_unlock: | |
418 | gfs2_glock_dq(&gh); | |
419 | out: | |
420 | gfs2_holder_uninit(&gh); | |
e56985da SW |
421 | if (ret == -ENOMEM) |
422 | ret = VM_FAULT_OOM; | |
423 | else if (ret) | |
c2ec175c | 424 | ret = VM_FAULT_SIGBUS; |
3cc3f710 SW |
425 | return ret; |
426 | } | |
427 | ||
f0f37e2f | 428 | static const struct vm_operations_struct gfs2_vm_ops = { |
3cc3f710 SW |
429 | .fault = filemap_fault, |
430 | .page_mkwrite = gfs2_page_mkwrite, | |
431 | }; | |
432 | ||
b3b94faa DT |
433 | /** |
434 | * gfs2_mmap - | |
435 | * @file: The file to map | |
436 | * @vma: The VMA which described the mapping | |
437 | * | |
48bf2b17 SW |
438 | * There is no need to get a lock here unless we should be updating |
439 | * atime. We ignore any locking errors since the only consequence is | |
440 | * a missed atime update (which will just be deferred until later). | |
441 | * | |
442 | * Returns: 0 | |
b3b94faa DT |
443 | */ |
444 | ||
445 | static int gfs2_mmap(struct file *file, struct vm_area_struct *vma) | |
446 | { | |
feaa7bba | 447 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
b3b94faa | 448 | |
48bf2b17 SW |
449 | if (!(file->f_flags & O_NOATIME)) { |
450 | struct gfs2_holder i_gh; | |
451 | int error; | |
b3b94faa | 452 | |
48bf2b17 SW |
453 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &i_gh); |
454 | error = gfs2_glock_nq(&i_gh); | |
455 | file_accessed(file); | |
456 | if (error == 0) | |
457 | gfs2_glock_dq_uninit(&i_gh); | |
458 | } | |
3cc3f710 | 459 | vma->vm_ops = &gfs2_vm_ops; |
48bf2b17 | 460 | vma->vm_flags |= VM_CAN_NONLINEAR; |
b3b94faa | 461 | |
48bf2b17 | 462 | return 0; |
b3b94faa DT |
463 | } |
464 | ||
465 | /** | |
466 | * gfs2_open - open a file | |
467 | * @inode: the inode to open | |
468 | * @file: the struct file for this opening | |
469 | * | |
470 | * Returns: errno | |
471 | */ | |
472 | ||
473 | static int gfs2_open(struct inode *inode, struct file *file) | |
474 | { | |
feaa7bba | 475 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa DT |
476 | struct gfs2_holder i_gh; |
477 | struct gfs2_file *fp; | |
478 | int error; | |
479 | ||
b3b94faa DT |
480 | fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL); |
481 | if (!fp) | |
482 | return -ENOMEM; | |
483 | ||
f55ab26a | 484 | mutex_init(&fp->f_fl_mutex); |
b3b94faa | 485 | |
feaa7bba | 486 | gfs2_assert_warn(GFS2_SB(inode), !file->private_data); |
5c676f6d | 487 | file->private_data = fp; |
b3b94faa | 488 | |
b60623c2 | 489 | if (S_ISREG(ip->i_inode.i_mode)) { |
b3b94faa DT |
490 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, |
491 | &i_gh); | |
492 | if (error) | |
493 | goto fail; | |
494 | ||
495 | if (!(file->f_flags & O_LARGEFILE) && | |
c9e98886 | 496 | ip->i_disksize > MAX_NON_LFS) { |
a9c62a18 | 497 | error = -EOVERFLOW; |
b3b94faa DT |
498 | goto fail_gunlock; |
499 | } | |
500 | ||
b3b94faa DT |
501 | gfs2_glock_dq_uninit(&i_gh); |
502 | } | |
503 | ||
504 | return 0; | |
505 | ||
420b9e5e | 506 | fail_gunlock: |
b3b94faa | 507 | gfs2_glock_dq_uninit(&i_gh); |
420b9e5e | 508 | fail: |
5c676f6d | 509 | file->private_data = NULL; |
b3b94faa | 510 | kfree(fp); |
b3b94faa DT |
511 | return error; |
512 | } | |
513 | ||
514 | /** | |
515 | * gfs2_close - called to close a struct file | |
516 | * @inode: the inode the struct file belongs to | |
517 | * @file: the struct file being closed | |
518 | * | |
519 | * Returns: errno | |
520 | */ | |
521 | ||
522 | static int gfs2_close(struct inode *inode, struct file *file) | |
523 | { | |
5c676f6d | 524 | struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; |
b3b94faa DT |
525 | struct gfs2_file *fp; |
526 | ||
5c676f6d SW |
527 | fp = file->private_data; |
528 | file->private_data = NULL; | |
b3b94faa DT |
529 | |
530 | if (gfs2_assert_warn(sdp, fp)) | |
531 | return -EIO; | |
532 | ||
533 | kfree(fp); | |
534 | ||
535 | return 0; | |
536 | } | |
537 | ||
538 | /** | |
539 | * gfs2_fsync - sync the dirty data for a file (across the cluster) | |
540 | * @file: the file that points to the dentry (we ignore this) | |
541 | * @dentry: the dentry that points to the inode to sync | |
542 | * | |
33c3de32 SW |
543 | * The VFS will flush "normal" data for us. We only need to worry |
544 | * about metadata here. For journaled data, we just do a log flush | |
545 | * as we can't avoid it. Otherwise we can just bale out if datasync | |
546 | * is set. For stuffed inodes we must flush the log in order to | |
547 | * ensure that all data is on disk. | |
548 | * | |
34126f9f SW |
549 | * The call to write_inode_now() is there to write back metadata and |
550 | * the inode itself. It does also try and write the data, but thats | |
551 | * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite() | |
552 | * for us. | |
553 | * | |
b3b94faa DT |
554 | * Returns: errno |
555 | */ | |
556 | ||
557 | static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync) | |
558 | { | |
33c3de32 SW |
559 | struct inode *inode = dentry->d_inode; |
560 | int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC); | |
561 | int ret = 0; | |
33c3de32 SW |
562 | |
563 | if (gfs2_is_jdata(GFS2_I(inode))) { | |
564 | gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl); | |
565 | return 0; | |
566 | } | |
b3b94faa | 567 | |
33c3de32 SW |
568 | if (sync_state != 0) { |
569 | if (!datasync) | |
34126f9f | 570 | ret = write_inode_now(inode, 0); |
b3b94faa | 571 | |
33c3de32 SW |
572 | if (gfs2_is_stuffed(GFS2_I(inode))) |
573 | gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl); | |
574 | } | |
575 | ||
576 | return ret; | |
b3b94faa DT |
577 | } |
578 | ||
56aa616a SW |
579 | /** |
580 | * gfs2_file_aio_write - Perform a write to a file | |
581 | * @iocb: The io context | |
582 | * @iov: The data to write | |
583 | * @nr_segs: Number of @iov segments | |
584 | * @pos: The file position | |
585 | * | |
586 | * We have to do a lock/unlock here to refresh the inode size for | |
587 | * O_APPEND writes, otherwise we can land up writing at the wrong | |
588 | * offset. There is still a race, but provided the app is using its | |
589 | * own file locking, this will make O_APPEND work as expected. | |
590 | * | |
591 | */ | |
592 | ||
593 | static ssize_t gfs2_file_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
594 | unsigned long nr_segs, loff_t pos) | |
595 | { | |
596 | struct file *file = iocb->ki_filp; | |
597 | ||
598 | if (file->f_flags & O_APPEND) { | |
599 | struct dentry *dentry = file->f_dentry; | |
600 | struct gfs2_inode *ip = GFS2_I(dentry->d_inode); | |
601 | struct gfs2_holder gh; | |
602 | int ret; | |
603 | ||
604 | ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &gh); | |
605 | if (ret) | |
606 | return ret; | |
607 | gfs2_glock_dq_uninit(&gh); | |
608 | } | |
609 | ||
610 | return generic_file_aio_write(iocb, iov, nr_segs, pos); | |
611 | } | |
612 | ||
f057f6cd SW |
613 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM |
614 | ||
60446067 ME |
615 | /** |
616 | * gfs2_setlease - acquire/release a file lease | |
617 | * @file: the file pointer | |
618 | * @arg: lease type | |
619 | * @fl: file lock | |
620 | * | |
f057f6cd SW |
621 | * We don't currently have a way to enforce a lease across the whole |
622 | * cluster; until we do, disable leases (by just returning -EINVAL), | |
623 | * unless the administrator has requested purely local locking. | |
624 | * | |
60446067 ME |
625 | * Returns: errno |
626 | */ | |
627 | ||
628 | static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl) | |
629 | { | |
f057f6cd | 630 | return -EINVAL; |
da755fdb SW |
631 | } |
632 | ||
b3b94faa DT |
633 | /** |
634 | * gfs2_lock - acquire/release a posix lock on a file | |
635 | * @file: the file pointer | |
636 | * @cmd: either modify or retrieve lock state, possibly wait | |
637 | * @fl: type and range of lock | |
638 | * | |
639 | * Returns: errno | |
640 | */ | |
641 | ||
642 | static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl) | |
643 | { | |
feaa7bba SW |
644 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
645 | struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host); | |
f057f6cd | 646 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
b3b94faa | 647 | |
b3b94faa DT |
648 | if (!(fl->fl_flags & FL_POSIX)) |
649 | return -ENOLCK; | |
720e7749 | 650 | if (__mandatory_lock(&ip->i_inode) && fl->fl_type != F_UNLCK) |
b3b94faa DT |
651 | return -ENOLCK; |
652 | ||
586759f0 ME |
653 | if (cmd == F_CANCELLK) { |
654 | /* Hack: */ | |
655 | cmd = F_SETLK; | |
656 | fl->fl_type = F_UNLCK; | |
657 | } | |
f057f6cd SW |
658 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) |
659 | return -EIO; | |
b3b94faa | 660 | if (IS_GETLK(cmd)) |
f057f6cd | 661 | return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl); |
b3b94faa | 662 | else if (fl->fl_type == F_UNLCK) |
f057f6cd | 663 | return dlm_posix_unlock(ls->ls_dlm, ip->i_no_addr, file, fl); |
b3b94faa | 664 | else |
f057f6cd | 665 | return dlm_posix_lock(ls->ls_dlm, ip->i_no_addr, file, cmd, fl); |
b3b94faa DT |
666 | } |
667 | ||
b3b94faa DT |
668 | static int do_flock(struct file *file, int cmd, struct file_lock *fl) |
669 | { | |
5c676f6d | 670 | struct gfs2_file *fp = file->private_data; |
b3b94faa | 671 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
81454098 | 672 | struct gfs2_inode *ip = GFS2_I(file->f_path.dentry->d_inode); |
b3b94faa DT |
673 | struct gfs2_glock *gl; |
674 | unsigned int state; | |
675 | int flags; | |
676 | int error = 0; | |
677 | ||
678 | state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
6802e340 | 679 | flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; |
b3b94faa | 680 | |
f55ab26a | 681 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
682 | |
683 | gl = fl_gh->gh_gl; | |
684 | if (gl) { | |
685 | if (fl_gh->gh_state == state) | |
686 | goto out; | |
b3b94faa | 687 | flock_lock_file_wait(file, |
907b9bce | 688 | &(struct file_lock){.fl_type = F_UNLCK}); |
b4c20166 AD |
689 | gfs2_glock_dq_wait(fl_gh); |
690 | gfs2_holder_reinit(state, flags, fl_gh); | |
b3b94faa | 691 | } else { |
6802e340 SW |
692 | error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr, |
693 | &gfs2_flock_glops, CREATE, &gl); | |
b3b94faa DT |
694 | if (error) |
695 | goto out; | |
b4c20166 AD |
696 | gfs2_holder_init(gl, state, flags, fl_gh); |
697 | gfs2_glock_put(gl); | |
b3b94faa | 698 | } |
b3b94faa DT |
699 | error = gfs2_glock_nq(fl_gh); |
700 | if (error) { | |
701 | gfs2_holder_uninit(fl_gh); | |
702 | if (error == GLR_TRYFAILED) | |
703 | error = -EAGAIN; | |
704 | } else { | |
705 | error = flock_lock_file_wait(file, fl); | |
feaa7bba | 706 | gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error); |
b3b94faa DT |
707 | } |
708 | ||
420b9e5e | 709 | out: |
f55ab26a | 710 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
711 | return error; |
712 | } | |
713 | ||
714 | static void do_unflock(struct file *file, struct file_lock *fl) | |
715 | { | |
5c676f6d | 716 | struct gfs2_file *fp = file->private_data; |
b3b94faa DT |
717 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
718 | ||
f55ab26a | 719 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
720 | flock_lock_file_wait(file, fl); |
721 | if (fl_gh->gh_gl) | |
722 | gfs2_glock_dq_uninit(fl_gh); | |
f55ab26a | 723 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
724 | } |
725 | ||
726 | /** | |
727 | * gfs2_flock - acquire/release a flock lock on a file | |
728 | * @file: the file pointer | |
729 | * @cmd: either modify or retrieve lock state, possibly wait | |
730 | * @fl: type and range of lock | |
731 | * | |
732 | * Returns: errno | |
733 | */ | |
734 | ||
735 | static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl) | |
736 | { | |
b3b94faa DT |
737 | if (!(fl->fl_flags & FL_FLOCK)) |
738 | return -ENOLCK; | |
a12af1eb AD |
739 | if (fl->fl_type & LOCK_MAND) |
740 | return -EOPNOTSUPP; | |
b3b94faa | 741 | |
b3b94faa DT |
742 | if (fl->fl_type == F_UNLCK) { |
743 | do_unflock(file, fl); | |
744 | return 0; | |
d00223f1 | 745 | } else { |
b3b94faa | 746 | return do_flock(file, cmd, fl); |
d00223f1 | 747 | } |
b3b94faa DT |
748 | } |
749 | ||
10d21988 | 750 | const struct file_operations gfs2_file_fops = { |
26c1a574 | 751 | .llseek = gfs2_llseek, |
d00223f1 | 752 | .read = do_sync_read, |
26c1a574 | 753 | .aio_read = generic_file_aio_read, |
d00223f1 | 754 | .write = do_sync_write, |
56aa616a | 755 | .aio_write = gfs2_file_aio_write, |
26c1a574 SW |
756 | .unlocked_ioctl = gfs2_ioctl, |
757 | .mmap = gfs2_mmap, | |
758 | .open = gfs2_open, | |
759 | .release = gfs2_close, | |
760 | .fsync = gfs2_fsync, | |
761 | .lock = gfs2_lock, | |
26c1a574 SW |
762 | .flock = gfs2_flock, |
763 | .splice_read = generic_file_splice_read, | |
764 | .splice_write = generic_file_splice_write, | |
60446067 | 765 | .setlease = gfs2_setlease, |
b3b94faa DT |
766 | }; |
767 | ||
10d21988 | 768 | const struct file_operations gfs2_dir_fops = { |
26c1a574 SW |
769 | .readdir = gfs2_readdir, |
770 | .unlocked_ioctl = gfs2_ioctl, | |
771 | .open = gfs2_open, | |
772 | .release = gfs2_close, | |
773 | .fsync = gfs2_fsync, | |
774 | .lock = gfs2_lock, | |
775 | .flock = gfs2_flock, | |
b3b94faa DT |
776 | }; |
777 | ||
f057f6cd SW |
778 | #endif /* CONFIG_GFS2_FS_LOCKING_DLM */ |
779 | ||
10d21988 | 780 | const struct file_operations gfs2_file_fops_nolock = { |
c97bfe43 WC |
781 | .llseek = gfs2_llseek, |
782 | .read = do_sync_read, | |
783 | .aio_read = generic_file_aio_read, | |
784 | .write = do_sync_write, | |
56aa616a | 785 | .aio_write = gfs2_file_aio_write, |
c97bfe43 WC |
786 | .unlocked_ioctl = gfs2_ioctl, |
787 | .mmap = gfs2_mmap, | |
788 | .open = gfs2_open, | |
789 | .release = gfs2_close, | |
790 | .fsync = gfs2_fsync, | |
791 | .splice_read = generic_file_splice_read, | |
792 | .splice_write = generic_file_splice_write, | |
f057f6cd | 793 | .setlease = generic_setlease, |
c97bfe43 WC |
794 | }; |
795 | ||
10d21988 | 796 | const struct file_operations gfs2_dir_fops_nolock = { |
c97bfe43 WC |
797 | .readdir = gfs2_readdir, |
798 | .unlocked_ioctl = gfs2_ioctl, | |
799 | .open = gfs2_open, | |
800 | .release = gfs2_close, | |
801 | .fsync = gfs2_fsync, | |
802 | }; | |
803 |