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 | 21 | #include <linux/ext2_fs.h> |
2fe17c10 CH |
22 | #include <linux/falloc.h> |
23 | #include <linux/swap.h> | |
71b86f56 | 24 | #include <linux/crc32.h> |
33c3de32 | 25 | #include <linux/writeback.h> |
b3b94faa | 26 | #include <asm/uaccess.h> |
f057f6cd SW |
27 | #include <linux/dlm.h> |
28 | #include <linux/dlm_plock.h> | |
b3b94faa DT |
29 | |
30 | #include "gfs2.h" | |
5c676f6d | 31 | #include "incore.h" |
b3b94faa DT |
32 | #include "bmap.h" |
33 | #include "dir.h" | |
34 | #include "glock.h" | |
35 | #include "glops.h" | |
36 | #include "inode.h" | |
b3b94faa DT |
37 | #include "log.h" |
38 | #include "meta_io.h" | |
b3b94faa DT |
39 | #include "quota.h" |
40 | #include "rgrp.h" | |
41 | #include "trans.h" | |
5c676f6d | 42 | #include "util.h" |
b3b94faa | 43 | |
b3b94faa DT |
44 | /** |
45 | * gfs2_llseek - seek to a location in a file | |
46 | * @file: the file | |
47 | * @offset: the offset | |
48 | * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END) | |
49 | * | |
50 | * SEEK_END requires the glock for the file because it references the | |
51 | * file's size. | |
52 | * | |
53 | * Returns: The new offset, or errno | |
54 | */ | |
55 | ||
56 | static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin) | |
57 | { | |
feaa7bba | 58 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
b3b94faa DT |
59 | struct gfs2_holder i_gh; |
60 | loff_t error; | |
61 | ||
9453615a SW |
62 | switch (origin) { |
63 | case SEEK_END: /* These reference inode->i_size */ | |
64 | case SEEK_DATA: | |
65 | case SEEK_HOLE: | |
b3b94faa DT |
66 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, |
67 | &i_gh); | |
68 | if (!error) { | |
9465efc9 | 69 | error = generic_file_llseek_unlocked(file, offset, origin); |
b3b94faa DT |
70 | gfs2_glock_dq_uninit(&i_gh); |
71 | } | |
9453615a SW |
72 | break; |
73 | case SEEK_CUR: | |
74 | case SEEK_SET: | |
9465efc9 | 75 | error = generic_file_llseek_unlocked(file, offset, origin); |
9453615a SW |
76 | break; |
77 | default: | |
78 | error = -EINVAL; | |
79 | } | |
b3b94faa DT |
80 | |
81 | return error; | |
82 | } | |
83 | ||
b3b94faa | 84 | /** |
f0e522a9 | 85 | * gfs2_readdir - Read directory entries from a directory |
b3b94faa DT |
86 | * @file: The directory to read from |
87 | * @dirent: Buffer for dirents | |
88 | * @filldir: Function used to do the copying | |
89 | * | |
90 | * Returns: errno | |
91 | */ | |
92 | ||
f0e522a9 | 93 | static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir) |
b3b94faa | 94 | { |
71b86f56 | 95 | struct inode *dir = file->f_mapping->host; |
feaa7bba | 96 | struct gfs2_inode *dip = GFS2_I(dir); |
b3b94faa | 97 | struct gfs2_holder d_gh; |
cd915493 | 98 | u64 offset = file->f_pos; |
b3b94faa DT |
99 | int error; |
100 | ||
719ee344 SW |
101 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh); |
102 | error = gfs2_glock_nq(&d_gh); | |
b3b94faa DT |
103 | if (error) { |
104 | gfs2_holder_uninit(&d_gh); | |
105 | return error; | |
106 | } | |
107 | ||
3699e3a4 | 108 | error = gfs2_dir_read(dir, &offset, dirent, filldir); |
b3b94faa DT |
109 | |
110 | gfs2_glock_dq_uninit(&d_gh); | |
111 | ||
b3b94faa DT |
112 | file->f_pos = offset; |
113 | ||
b3b94faa DT |
114 | return error; |
115 | } | |
116 | ||
128e5eba SW |
117 | /** |
118 | * fsflags_cvt | |
119 | * @table: A table of 32 u32 flags | |
120 | * @val: a 32 bit value to convert | |
121 | * | |
122 | * This function can be used to convert between fsflags values and | |
123 | * GFS2's own flags values. | |
124 | * | |
125 | * Returns: the converted flags | |
126 | */ | |
127 | static u32 fsflags_cvt(const u32 *table, u32 val) | |
128 | { | |
129 | u32 res = 0; | |
130 | while(val) { | |
131 | if (val & 1) | |
132 | res |= *table; | |
133 | table++; | |
134 | val >>= 1; | |
135 | } | |
136 | return res; | |
137 | } | |
b3b94faa | 138 | |
128e5eba SW |
139 | static const u32 fsflags_to_gfs2[32] = { |
140 | [3] = GFS2_DIF_SYNC, | |
141 | [4] = GFS2_DIF_IMMUTABLE, | |
142 | [5] = GFS2_DIF_APPENDONLY, | |
143 | [7] = GFS2_DIF_NOATIME, | |
144 | [12] = GFS2_DIF_EXHASH, | |
b9af7ca6 | 145 | [14] = GFS2_DIF_INHERIT_JDATA, |
71b86f56 SW |
146 | }; |
147 | ||
128e5eba SW |
148 | static const u32 gfs2_to_fsflags[32] = { |
149 | [gfs2fl_Sync] = FS_SYNC_FL, | |
150 | [gfs2fl_Immutable] = FS_IMMUTABLE_FL, | |
151 | [gfs2fl_AppendOnly] = FS_APPEND_FL, | |
152 | [gfs2fl_NoAtime] = FS_NOATIME_FL, | |
153 | [gfs2fl_ExHash] = FS_INDEX_FL, | |
128e5eba | 154 | [gfs2fl_InheritJdata] = FS_JOURNAL_DATA_FL, |
7ea9ea83 | 155 | }; |
71b86f56 | 156 | |
b09e593d | 157 | static int gfs2_get_flags(struct file *filp, u32 __user *ptr) |
71b86f56 | 158 | { |
81454098 | 159 | struct inode *inode = filp->f_path.dentry->d_inode; |
feaa7bba | 160 | struct gfs2_inode *ip = GFS2_I(inode); |
71b86f56 SW |
161 | struct gfs2_holder gh; |
162 | int error; | |
128e5eba | 163 | u32 fsflags; |
71b86f56 | 164 | |
719ee344 SW |
165 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh); |
166 | error = gfs2_glock_nq(&gh); | |
71b86f56 SW |
167 | if (error) |
168 | return error; | |
907b9bce | 169 | |
383f01fb SW |
170 | fsflags = fsflags_cvt(gfs2_to_fsflags, ip->i_diskflags); |
171 | if (!S_ISDIR(inode->i_mode) && ip->i_diskflags & GFS2_DIF_JDATA) | |
c9f6a6bb | 172 | fsflags |= FS_JOURNAL_DATA_FL; |
128e5eba | 173 | if (put_user(fsflags, ptr)) |
71b86f56 SW |
174 | error = -EFAULT; |
175 | ||
3cc3f710 | 176 | gfs2_glock_dq(&gh); |
71b86f56 SW |
177 | gfs2_holder_uninit(&gh); |
178 | return error; | |
179 | } | |
180 | ||
6b124d8d SW |
181 | void gfs2_set_inode_flags(struct inode *inode) |
182 | { | |
183 | struct gfs2_inode *ip = GFS2_I(inode); | |
6b124d8d SW |
184 | unsigned int flags = inode->i_flags; |
185 | ||
9964afbb SW |
186 | flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|S_NOSEC); |
187 | if ((ip->i_eattr == 0) && !is_sxid(inode->i_mode)) | |
188 | inode->i_flags |= S_NOSEC; | |
383f01fb | 189 | if (ip->i_diskflags & GFS2_DIF_IMMUTABLE) |
6b124d8d | 190 | flags |= S_IMMUTABLE; |
383f01fb | 191 | if (ip->i_diskflags & GFS2_DIF_APPENDONLY) |
6b124d8d | 192 | flags |= S_APPEND; |
383f01fb | 193 | if (ip->i_diskflags & GFS2_DIF_NOATIME) |
6b124d8d | 194 | flags |= S_NOATIME; |
383f01fb | 195 | if (ip->i_diskflags & GFS2_DIF_SYNC) |
6b124d8d SW |
196 | flags |= S_SYNC; |
197 | inode->i_flags = flags; | |
198 | } | |
199 | ||
71b86f56 SW |
200 | /* Flags that can be set by user space */ |
201 | #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \ | |
71b86f56 SW |
202 | GFS2_DIF_IMMUTABLE| \ |
203 | GFS2_DIF_APPENDONLY| \ | |
204 | GFS2_DIF_NOATIME| \ | |
205 | GFS2_DIF_SYNC| \ | |
206 | GFS2_DIF_SYSTEM| \ | |
71b86f56 SW |
207 | GFS2_DIF_INHERIT_JDATA) |
208 | ||
209 | /** | |
210 | * gfs2_set_flags - set flags on an inode | |
211 | * @inode: The inode | |
212 | * @flags: The flags to set | |
213 | * @mask: Indicates which flags are valid | |
214 | * | |
215 | */ | |
b09e593d | 216 | static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask) |
71b86f56 | 217 | { |
81454098 | 218 | struct inode *inode = filp->f_path.dentry->d_inode; |
feaa7bba SW |
219 | struct gfs2_inode *ip = GFS2_I(inode); |
220 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
71b86f56 SW |
221 | struct buffer_head *bh; |
222 | struct gfs2_holder gh; | |
223 | int error; | |
55eccc6d | 224 | u32 new_flags, flags; |
71b86f56 | 225 | |
f58ba889 | 226 | error = mnt_want_write(filp->f_path.mnt); |
52f341cf | 227 | if (error) |
71b86f56 SW |
228 | return error; |
229 | ||
f58ba889 MS |
230 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); |
231 | if (error) | |
232 | goto out_drop_write; | |
233 | ||
7df0e039 | 234 | error = -EACCES; |
2e149670 | 235 | if (!inode_owner_or_capable(inode)) |
7df0e039 SW |
236 | goto out; |
237 | ||
238 | error = 0; | |
383f01fb | 239 | flags = ip->i_diskflags; |
55eccc6d | 240 | new_flags = (flags & ~mask) | (reqflags & mask); |
71b86f56 SW |
241 | if ((new_flags ^ flags) == 0) |
242 | goto out; | |
243 | ||
244 | error = -EINVAL; | |
245 | if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET) | |
246 | goto out; | |
247 | ||
71b86f56 SW |
248 | error = -EPERM; |
249 | if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE)) | |
250 | goto out; | |
251 | if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY)) | |
252 | goto out; | |
907b9bce | 253 | if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) && |
b9cb9813 | 254 | !capable(CAP_LINUX_IMMUTABLE)) |
71b86f56 | 255 | goto out; |
b9cb9813 | 256 | if (!IS_IMMUTABLE(inode)) { |
10556cb2 | 257 | error = gfs2_permission(inode, MAY_WRITE); |
b9cb9813 SW |
258 | if (error) |
259 | goto out; | |
260 | } | |
5561093e SW |
261 | if ((flags ^ new_flags) & GFS2_DIF_JDATA) { |
262 | if (flags & GFS2_DIF_JDATA) | |
263 | gfs2_log_flush(sdp, ip->i_gl); | |
264 | error = filemap_fdatawrite(inode->i_mapping); | |
265 | if (error) | |
266 | goto out; | |
267 | error = filemap_fdatawait(inode->i_mapping); | |
268 | if (error) | |
269 | goto out; | |
270 | } | |
55eccc6d | 271 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); |
71b86f56 SW |
272 | if (error) |
273 | goto out; | |
55eccc6d SW |
274 | error = gfs2_meta_inode_buffer(ip, &bh); |
275 | if (error) | |
276 | goto out_trans_end; | |
71b86f56 | 277 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
383f01fb | 278 | ip->i_diskflags = new_flags; |
539e5d6b | 279 | gfs2_dinode_out(ip, bh->b_data); |
71b86f56 | 280 | brelse(bh); |
6b124d8d | 281 | gfs2_set_inode_flags(inode); |
5561093e | 282 | gfs2_set_aops(inode); |
55eccc6d SW |
283 | out_trans_end: |
284 | gfs2_trans_end(sdp); | |
71b86f56 SW |
285 | out: |
286 | gfs2_glock_dq_uninit(&gh); | |
f58ba889 MS |
287 | out_drop_write: |
288 | mnt_drop_write(filp->f_path.mnt); | |
71b86f56 SW |
289 | return error; |
290 | } | |
291 | ||
b09e593d | 292 | static int gfs2_set_flags(struct file *filp, u32 __user *ptr) |
71b86f56 | 293 | { |
b9af7ca6 | 294 | struct inode *inode = filp->f_path.dentry->d_inode; |
128e5eba | 295 | u32 fsflags, gfsflags; |
7df0e039 | 296 | |
128e5eba | 297 | if (get_user(fsflags, ptr)) |
71b86f56 | 298 | return -EFAULT; |
7df0e039 | 299 | |
128e5eba | 300 | gfsflags = fsflags_cvt(fsflags_to_gfs2, fsflags); |
b9af7ca6 SW |
301 | if (!S_ISDIR(inode->i_mode)) { |
302 | if (gfsflags & GFS2_DIF_INHERIT_JDATA) | |
303 | gfsflags ^= (GFS2_DIF_JDATA | GFS2_DIF_INHERIT_JDATA); | |
b9af7ca6 SW |
304 | return do_gfs2_set_flags(filp, gfsflags, ~0); |
305 | } | |
306 | return do_gfs2_set_flags(filp, gfsflags, ~GFS2_DIF_JDATA); | |
71b86f56 SW |
307 | } |
308 | ||
b09e593d | 309 | static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) |
71b86f56 SW |
310 | { |
311 | switch(cmd) { | |
128e5eba | 312 | case FS_IOC_GETFLAGS: |
b09e593d | 313 | return gfs2_get_flags(filp, (u32 __user *)arg); |
128e5eba | 314 | case FS_IOC_SETFLAGS: |
b09e593d | 315 | return gfs2_set_flags(filp, (u32 __user *)arg); |
71b86f56 SW |
316 | } |
317 | return -ENOTTY; | |
318 | } | |
319 | ||
3cc3f710 SW |
320 | /** |
321 | * gfs2_allocate_page_backing - Use bmap to allocate blocks | |
322 | * @page: The (locked) page to allocate backing for | |
323 | * | |
324 | * We try to allocate all the blocks required for the page in | |
325 | * one go. This might fail for various reasons, so we keep | |
326 | * trying until all the blocks to back this page are allocated. | |
327 | * If some of the blocks are already allocated, thats ok too. | |
328 | */ | |
329 | ||
330 | static int gfs2_allocate_page_backing(struct page *page) | |
331 | { | |
332 | struct inode *inode = page->mapping->host; | |
333 | struct buffer_head bh; | |
334 | unsigned long size = PAGE_CACHE_SIZE; | |
335 | u64 lblock = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
336 | ||
337 | do { | |
338 | bh.b_state = 0; | |
339 | bh.b_size = size; | |
e9e1ef2b | 340 | gfs2_block_map(inode, lblock, &bh, 1); |
3cc3f710 SW |
341 | if (!buffer_mapped(&bh)) |
342 | return -EIO; | |
343 | size -= bh.b_size; | |
344 | lblock += (bh.b_size >> inode->i_blkbits); | |
345 | } while(size > 0); | |
346 | return 0; | |
347 | } | |
348 | ||
349 | /** | |
350 | * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable | |
351 | * @vma: The virtual memory area | |
352 | * @page: The page which is about to become writable | |
353 | * | |
354 | * When the page becomes writable, we need to ensure that we have | |
355 | * blocks allocated on disk to back that page. | |
356 | */ | |
357 | ||
c2ec175c | 358 | static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) |
3cc3f710 | 359 | { |
c2ec175c | 360 | struct page *page = vmf->page; |
3cc3f710 SW |
361 | struct inode *inode = vma->vm_file->f_path.dentry->d_inode; |
362 | struct gfs2_inode *ip = GFS2_I(inode); | |
363 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
364 | unsigned long last_index; | |
c8f554b9 | 365 | u64 pos = page->index << PAGE_CACHE_SHIFT; |
3cc3f710 | 366 | unsigned int data_blocks, ind_blocks, rblocks; |
3cc3f710 SW |
367 | struct gfs2_holder gh; |
368 | struct gfs2_alloc *al; | |
369 | int ret; | |
370 | ||
719ee344 SW |
371 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); |
372 | ret = gfs2_glock_nq(&gh); | |
3cc3f710 SW |
373 | if (ret) |
374 | goto out; | |
375 | ||
9c538837 SW |
376 | set_bit(GLF_DIRTY, &ip->i_gl->gl_flags); |
377 | set_bit(GIF_SW_PAGED, &ip->i_flags); | |
378 | ||
461cb419 | 379 | if (!gfs2_write_alloc_required(ip, pos, PAGE_CACHE_SIZE)) |
3cc3f710 | 380 | goto out_unlock; |
6dbd8224 | 381 | ret = -ENOMEM; |
3cc3f710 | 382 | al = gfs2_alloc_get(ip); |
6dbd8224 SW |
383 | if (al == NULL) |
384 | goto out_unlock; | |
385 | ||
d82661d9 | 386 | ret = gfs2_quota_lock_check(ip); |
3cc3f710 SW |
387 | if (ret) |
388 | goto out_alloc_put; | |
7ed122e4 | 389 | gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks); |
3cc3f710 SW |
390 | al->al_requested = data_blocks + ind_blocks; |
391 | ret = gfs2_inplace_reserve(ip); | |
392 | if (ret) | |
393 | goto out_quota_unlock; | |
394 | ||
395 | rblocks = RES_DINODE + ind_blocks; | |
396 | if (gfs2_is_jdata(ip)) | |
397 | rblocks += data_blocks ? data_blocks : 1; | |
bf97b673 | 398 | if (ind_blocks || data_blocks) { |
3cc3f710 | 399 | rblocks += RES_STATFS + RES_QUOTA; |
bf97b673 BM |
400 | rblocks += gfs2_rg_blocks(al); |
401 | } | |
3cc3f710 SW |
402 | ret = gfs2_trans_begin(sdp, rblocks, 0); |
403 | if (ret) | |
404 | goto out_trans_fail; | |
405 | ||
406 | lock_page(page); | |
407 | ret = -EINVAL; | |
408 | last_index = ip->i_inode.i_size >> PAGE_CACHE_SHIFT; | |
409 | if (page->index > last_index) | |
410 | goto out_unlock_page; | |
b7fe2e39 | 411 | ret = 0; |
3cc3f710 SW |
412 | if (!PageUptodate(page) || page->mapping != ip->i_inode.i_mapping) |
413 | goto out_unlock_page; | |
414 | if (gfs2_is_stuffed(ip)) { | |
415 | ret = gfs2_unstuff_dinode(ip, page); | |
416 | if (ret) | |
417 | goto out_unlock_page; | |
418 | } | |
419 | ret = gfs2_allocate_page_backing(page); | |
420 | ||
421 | out_unlock_page: | |
422 | unlock_page(page); | |
423 | gfs2_trans_end(sdp); | |
424 | out_trans_fail: | |
425 | gfs2_inplace_release(ip); | |
426 | out_quota_unlock: | |
427 | gfs2_quota_unlock(ip); | |
428 | out_alloc_put: | |
429 | gfs2_alloc_put(ip); | |
430 | out_unlock: | |
431 | gfs2_glock_dq(&gh); | |
432 | out: | |
433 | gfs2_holder_uninit(&gh); | |
e56985da SW |
434 | if (ret == -ENOMEM) |
435 | ret = VM_FAULT_OOM; | |
436 | else if (ret) | |
c2ec175c | 437 | ret = VM_FAULT_SIGBUS; |
3cc3f710 SW |
438 | return ret; |
439 | } | |
440 | ||
f0f37e2f | 441 | static const struct vm_operations_struct gfs2_vm_ops = { |
3cc3f710 SW |
442 | .fault = filemap_fault, |
443 | .page_mkwrite = gfs2_page_mkwrite, | |
444 | }; | |
445 | ||
b3b94faa DT |
446 | /** |
447 | * gfs2_mmap - | |
448 | * @file: The file to map | |
449 | * @vma: The VMA which described the mapping | |
450 | * | |
48bf2b17 SW |
451 | * There is no need to get a lock here unless we should be updating |
452 | * atime. We ignore any locking errors since the only consequence is | |
453 | * a missed atime update (which will just be deferred until later). | |
454 | * | |
455 | * Returns: 0 | |
b3b94faa DT |
456 | */ |
457 | ||
458 | static int gfs2_mmap(struct file *file, struct vm_area_struct *vma) | |
459 | { | |
feaa7bba | 460 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
b3b94faa | 461 | |
b9c93bb7 SW |
462 | if (!(file->f_flags & O_NOATIME) && |
463 | !IS_NOATIME(&ip->i_inode)) { | |
48bf2b17 SW |
464 | struct gfs2_holder i_gh; |
465 | int error; | |
b3b94faa | 466 | |
b9c93bb7 | 467 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &i_gh); |
48bf2b17 | 468 | error = gfs2_glock_nq(&i_gh); |
b9c93bb7 SW |
469 | if (error == 0) { |
470 | file_accessed(file); | |
471 | gfs2_glock_dq(&i_gh); | |
472 | } | |
473 | gfs2_holder_uninit(&i_gh); | |
474 | if (error) | |
475 | return error; | |
48bf2b17 | 476 | } |
3cc3f710 | 477 | vma->vm_ops = &gfs2_vm_ops; |
48bf2b17 | 478 | vma->vm_flags |= VM_CAN_NONLINEAR; |
b3b94faa | 479 | |
48bf2b17 | 480 | return 0; |
b3b94faa DT |
481 | } |
482 | ||
483 | /** | |
484 | * gfs2_open - open a file | |
485 | * @inode: the inode to open | |
486 | * @file: the struct file for this opening | |
487 | * | |
488 | * Returns: errno | |
489 | */ | |
490 | ||
491 | static int gfs2_open(struct inode *inode, struct file *file) | |
492 | { | |
feaa7bba | 493 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa DT |
494 | struct gfs2_holder i_gh; |
495 | struct gfs2_file *fp; | |
496 | int error; | |
497 | ||
b3b94faa DT |
498 | fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL); |
499 | if (!fp) | |
500 | return -ENOMEM; | |
501 | ||
f55ab26a | 502 | mutex_init(&fp->f_fl_mutex); |
b3b94faa | 503 | |
feaa7bba | 504 | gfs2_assert_warn(GFS2_SB(inode), !file->private_data); |
5c676f6d | 505 | file->private_data = fp; |
b3b94faa | 506 | |
b60623c2 | 507 | if (S_ISREG(ip->i_inode.i_mode)) { |
b3b94faa DT |
508 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, |
509 | &i_gh); | |
510 | if (error) | |
511 | goto fail; | |
512 | ||
513 | if (!(file->f_flags & O_LARGEFILE) && | |
a2e0f799 | 514 | i_size_read(inode) > MAX_NON_LFS) { |
a9c62a18 | 515 | error = -EOVERFLOW; |
b3b94faa DT |
516 | goto fail_gunlock; |
517 | } | |
518 | ||
b3b94faa DT |
519 | gfs2_glock_dq_uninit(&i_gh); |
520 | } | |
521 | ||
522 | return 0; | |
523 | ||
420b9e5e | 524 | fail_gunlock: |
b3b94faa | 525 | gfs2_glock_dq_uninit(&i_gh); |
420b9e5e | 526 | fail: |
5c676f6d | 527 | file->private_data = NULL; |
b3b94faa | 528 | kfree(fp); |
b3b94faa DT |
529 | return error; |
530 | } | |
531 | ||
532 | /** | |
533 | * gfs2_close - called to close a struct file | |
534 | * @inode: the inode the struct file belongs to | |
535 | * @file: the struct file being closed | |
536 | * | |
537 | * Returns: errno | |
538 | */ | |
539 | ||
540 | static int gfs2_close(struct inode *inode, struct file *file) | |
541 | { | |
5c676f6d | 542 | struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; |
b3b94faa DT |
543 | struct gfs2_file *fp; |
544 | ||
5c676f6d SW |
545 | fp = file->private_data; |
546 | file->private_data = NULL; | |
b3b94faa DT |
547 | |
548 | if (gfs2_assert_warn(sdp, fp)) | |
549 | return -EIO; | |
550 | ||
551 | kfree(fp); | |
552 | ||
553 | return 0; | |
554 | } | |
555 | ||
556 | /** | |
557 | * gfs2_fsync - sync the dirty data for a file (across the cluster) | |
02c24a82 JB |
558 | * @file: the file that points to the dentry |
559 | * @start: the start position in the file to sync | |
560 | * @end: the end position in the file to sync | |
dba898b0 | 561 | * @datasync: set if we can ignore timestamp changes |
b3b94faa | 562 | * |
2f0264d5 SW |
563 | * We split the data flushing here so that we don't wait for the data |
564 | * until after we've also sent the metadata to disk. Note that for | |
565 | * data=ordered, we will write & wait for the data at the log flush | |
566 | * stage anyway, so this is unlikely to make much of a difference | |
567 | * except in the data=writeback case. | |
568 | * | |
569 | * If the fdatawrite fails due to any reason except -EIO, we will | |
570 | * continue the remainder of the fsync, although we'll still report | |
571 | * the error at the end. This is to match filemap_write_and_wait_range() | |
572 | * behaviour. | |
34126f9f | 573 | * |
b3b94faa DT |
574 | * Returns: errno |
575 | */ | |
576 | ||
02c24a82 JB |
577 | static int gfs2_fsync(struct file *file, loff_t start, loff_t end, |
578 | int datasync) | |
b3b94faa | 579 | { |
2f0264d5 SW |
580 | struct address_space *mapping = file->f_mapping; |
581 | struct inode *inode = mapping->host; | |
33c3de32 | 582 | int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC); |
dba898b0 | 583 | struct gfs2_inode *ip = GFS2_I(inode); |
2f0264d5 | 584 | int ret, ret1 = 0; |
b3b94faa | 585 | |
2f0264d5 SW |
586 | if (mapping->nrpages) { |
587 | ret1 = filemap_fdatawrite_range(mapping, start, end); | |
588 | if (ret1 == -EIO) | |
589 | return ret1; | |
590 | } | |
02c24a82 | 591 | |
dba898b0 SW |
592 | if (datasync) |
593 | sync_state &= ~I_DIRTY_SYNC; | |
b3b94faa | 594 | |
dba898b0 | 595 | if (sync_state) { |
2f0264d5 | 596 | mutex_lock(&inode->i_mutex); |
dba898b0 | 597 | ret = sync_inode_metadata(inode, 1); |
02c24a82 JB |
598 | if (ret) { |
599 | mutex_unlock(&inode->i_mutex); | |
dba898b0 | 600 | return ret; |
02c24a82 | 601 | } |
f1818529 SW |
602 | if (gfs2_is_jdata(ip)) |
603 | filemap_write_and_wait(mapping); | |
dba898b0 | 604 | gfs2_ail_flush(ip->i_gl); |
2f0264d5 | 605 | mutex_unlock(&inode->i_mutex); |
33c3de32 SW |
606 | } |
607 | ||
2f0264d5 SW |
608 | if (mapping->nrpages) |
609 | ret = filemap_fdatawait_range(mapping, start, end); | |
610 | ||
611 | return ret ? ret : ret1; | |
b3b94faa DT |
612 | } |
613 | ||
56aa616a SW |
614 | /** |
615 | * gfs2_file_aio_write - Perform a write to a file | |
616 | * @iocb: The io context | |
617 | * @iov: The data to write | |
618 | * @nr_segs: Number of @iov segments | |
619 | * @pos: The file position | |
620 | * | |
621 | * We have to do a lock/unlock here to refresh the inode size for | |
622 | * O_APPEND writes, otherwise we can land up writing at the wrong | |
623 | * offset. There is still a race, but provided the app is using its | |
624 | * own file locking, this will make O_APPEND work as expected. | |
625 | * | |
626 | */ | |
627 | ||
628 | static ssize_t gfs2_file_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
629 | unsigned long nr_segs, loff_t pos) | |
630 | { | |
631 | struct file *file = iocb->ki_filp; | |
632 | ||
633 | if (file->f_flags & O_APPEND) { | |
634 | struct dentry *dentry = file->f_dentry; | |
635 | struct gfs2_inode *ip = GFS2_I(dentry->d_inode); | |
636 | struct gfs2_holder gh; | |
637 | int ret; | |
638 | ||
639 | ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &gh); | |
640 | if (ret) | |
641 | return ret; | |
642 | gfs2_glock_dq_uninit(&gh); | |
643 | } | |
644 | ||
645 | return generic_file_aio_write(iocb, iov, nr_segs, pos); | |
646 | } | |
647 | ||
0ee53206 BM |
648 | static int empty_write_end(struct page *page, unsigned from, |
649 | unsigned to, int mode) | |
2fe17c10 | 650 | { |
0ee53206 BM |
651 | struct inode *inode = page->mapping->host; |
652 | struct gfs2_inode *ip = GFS2_I(inode); | |
653 | struct buffer_head *bh; | |
654 | unsigned offset, blksize = 1 << inode->i_blkbits; | |
655 | pgoff_t end_index = i_size_read(inode) >> PAGE_CACHE_SHIFT; | |
2fe17c10 | 656 | |
e4a7b7b0 | 657 | zero_user(page, from, to-from); |
2fe17c10 CH |
658 | mark_page_accessed(page); |
659 | ||
0ee53206 BM |
660 | if (page->index < end_index || !(mode & FALLOC_FL_KEEP_SIZE)) { |
661 | if (!gfs2_is_writeback(ip)) | |
662 | gfs2_page_add_databufs(ip, page, from, to); | |
663 | ||
664 | block_commit_write(page, from, to); | |
665 | return 0; | |
666 | } | |
667 | ||
668 | offset = 0; | |
669 | bh = page_buffers(page); | |
670 | while (offset < to) { | |
671 | if (offset >= from) { | |
672 | set_buffer_uptodate(bh); | |
673 | mark_buffer_dirty(bh); | |
674 | clear_buffer_new(bh); | |
675 | write_dirty_buffer(bh, WRITE); | |
676 | } | |
677 | offset += blksize; | |
678 | bh = bh->b_this_page; | |
679 | } | |
2fe17c10 | 680 | |
0ee53206 BM |
681 | offset = 0; |
682 | bh = page_buffers(page); | |
683 | while (offset < to) { | |
684 | if (offset >= from) { | |
685 | wait_on_buffer(bh); | |
686 | if (!buffer_uptodate(bh)) | |
687 | return -EIO; | |
688 | } | |
689 | offset += blksize; | |
690 | bh = bh->b_this_page; | |
691 | } | |
692 | return 0; | |
2fe17c10 CH |
693 | } |
694 | ||
e4a7b7b0 | 695 | static int needs_empty_write(sector_t block, struct inode *inode) |
2fe17c10 | 696 | { |
2fe17c10 | 697 | int error; |
e4a7b7b0 | 698 | struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 }; |
2fe17c10 | 699 | |
e4a7b7b0 BM |
700 | bh_map.b_size = 1 << inode->i_blkbits; |
701 | error = gfs2_block_map(inode, block, &bh_map, 0); | |
702 | if (unlikely(error)) | |
703 | return error; | |
704 | return !buffer_mapped(&bh_map); | |
705 | } | |
2fe17c10 | 706 | |
0ee53206 BM |
707 | static int write_empty_blocks(struct page *page, unsigned from, unsigned to, |
708 | int mode) | |
e4a7b7b0 BM |
709 | { |
710 | struct inode *inode = page->mapping->host; | |
711 | unsigned start, end, next, blksize; | |
712 | sector_t block = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
713 | int ret; | |
2fe17c10 | 714 | |
e4a7b7b0 | 715 | blksize = 1 << inode->i_blkbits; |
2fe17c10 CH |
716 | next = end = 0; |
717 | while (next < from) { | |
e4a7b7b0 BM |
718 | next += blksize; |
719 | block++; | |
2fe17c10 CH |
720 | } |
721 | start = next; | |
722 | do { | |
e4a7b7b0 BM |
723 | next += blksize; |
724 | ret = needs_empty_write(block, inode); | |
725 | if (unlikely(ret < 0)) | |
726 | return ret; | |
727 | if (ret == 0) { | |
2fe17c10 | 728 | if (end) { |
e4a7b7b0 BM |
729 | ret = __block_write_begin(page, start, end - start, |
730 | gfs2_block_map); | |
731 | if (unlikely(ret)) | |
732 | return ret; | |
0ee53206 BM |
733 | ret = empty_write_end(page, start, end, mode); |
734 | if (unlikely(ret)) | |
735 | return ret; | |
2fe17c10 CH |
736 | end = 0; |
737 | } | |
738 | start = next; | |
739 | } | |
740 | else | |
741 | end = next; | |
e4a7b7b0 | 742 | block++; |
2fe17c10 CH |
743 | } while (next < to); |
744 | ||
745 | if (end) { | |
e4a7b7b0 BM |
746 | ret = __block_write_begin(page, start, end - start, gfs2_block_map); |
747 | if (unlikely(ret)) | |
748 | return ret; | |
0ee53206 BM |
749 | ret = empty_write_end(page, start, end, mode); |
750 | if (unlikely(ret)) | |
751 | return ret; | |
2fe17c10 CH |
752 | } |
753 | ||
754 | return 0; | |
755 | } | |
756 | ||
757 | static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len, | |
758 | int mode) | |
759 | { | |
760 | struct gfs2_inode *ip = GFS2_I(inode); | |
761 | struct buffer_head *dibh; | |
762 | int error; | |
763 | u64 start = offset >> PAGE_CACHE_SHIFT; | |
764 | unsigned int start_offset = offset & ~PAGE_CACHE_MASK; | |
765 | u64 end = (offset + len - 1) >> PAGE_CACHE_SHIFT; | |
766 | pgoff_t curr; | |
767 | struct page *page; | |
768 | unsigned int end_offset = (offset + len) & ~PAGE_CACHE_MASK; | |
769 | unsigned int from, to; | |
770 | ||
771 | if (!end_offset) | |
772 | end_offset = PAGE_CACHE_SIZE; | |
773 | ||
774 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
775 | if (unlikely(error)) | |
776 | goto out; | |
777 | ||
778 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
779 | ||
780 | if (gfs2_is_stuffed(ip)) { | |
781 | error = gfs2_unstuff_dinode(ip, NULL); | |
782 | if (unlikely(error)) | |
783 | goto out; | |
784 | } | |
785 | ||
786 | curr = start; | |
787 | offset = start << PAGE_CACHE_SHIFT; | |
788 | from = start_offset; | |
789 | to = PAGE_CACHE_SIZE; | |
790 | while (curr <= end) { | |
791 | page = grab_cache_page_write_begin(inode->i_mapping, curr, | |
792 | AOP_FLAG_NOFS); | |
793 | if (unlikely(!page)) { | |
794 | error = -ENOMEM; | |
795 | goto out; | |
796 | } | |
797 | ||
798 | if (curr == end) | |
799 | to = end_offset; | |
0ee53206 | 800 | error = write_empty_blocks(page, from, to, mode); |
2fe17c10 CH |
801 | if (!error && offset + to > inode->i_size && |
802 | !(mode & FALLOC_FL_KEEP_SIZE)) { | |
803 | i_size_write(inode, offset + to); | |
804 | } | |
805 | unlock_page(page); | |
806 | page_cache_release(page); | |
807 | if (error) | |
808 | goto out; | |
809 | curr++; | |
810 | offset += PAGE_CACHE_SIZE; | |
811 | from = 0; | |
812 | } | |
813 | ||
2fe17c10 CH |
814 | mark_inode_dirty(inode); |
815 | ||
816 | brelse(dibh); | |
817 | ||
818 | out: | |
819 | return error; | |
820 | } | |
821 | ||
822 | static void calc_max_reserv(struct gfs2_inode *ip, loff_t max, loff_t *len, | |
823 | unsigned int *data_blocks, unsigned int *ind_blocks) | |
824 | { | |
825 | const struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
826 | unsigned int max_blocks = ip->i_alloc->al_rgd->rd_free_clone; | |
827 | unsigned int tmp, max_data = max_blocks - 3 * (sdp->sd_max_height - 1); | |
828 | ||
829 | for (tmp = max_data; tmp > sdp->sd_diptrs;) { | |
830 | tmp = DIV_ROUND_UP(tmp, sdp->sd_inptrs); | |
831 | max_data -= tmp; | |
832 | } | |
833 | /* This calculation isn't the exact reverse of gfs2_write_calc_reserve, | |
834 | so it might end up with fewer data blocks */ | |
835 | if (max_data <= *data_blocks) | |
836 | return; | |
837 | *data_blocks = max_data; | |
838 | *ind_blocks = max_blocks - max_data; | |
839 | *len = ((loff_t)max_data - 3) << sdp->sd_sb.sb_bsize_shift; | |
840 | if (*len > max) { | |
841 | *len = max; | |
842 | gfs2_write_calc_reserv(ip, max, data_blocks, ind_blocks); | |
843 | } | |
844 | } | |
845 | ||
846 | static long gfs2_fallocate(struct file *file, int mode, loff_t offset, | |
847 | loff_t len) | |
848 | { | |
849 | struct inode *inode = file->f_path.dentry->d_inode; | |
850 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
851 | struct gfs2_inode *ip = GFS2_I(inode); | |
852 | unsigned int data_blocks = 0, ind_blocks = 0, rblocks; | |
853 | loff_t bytes, max_bytes; | |
854 | struct gfs2_alloc *al; | |
855 | int error; | |
6905d9e4 | 856 | loff_t bsize_mask = ~((loff_t)sdp->sd_sb.sb_bsize - 1); |
2fe17c10 CH |
857 | loff_t next = (offset + len - 1) >> sdp->sd_sb.sb_bsize_shift; |
858 | next = (next + 1) << sdp->sd_sb.sb_bsize_shift; | |
859 | ||
860 | /* We only support the FALLOC_FL_KEEP_SIZE mode */ | |
861 | if (mode & ~FALLOC_FL_KEEP_SIZE) | |
862 | return -EOPNOTSUPP; | |
863 | ||
6905d9e4 | 864 | offset &= bsize_mask; |
2fe17c10 CH |
865 | |
866 | len = next - offset; | |
867 | bytes = sdp->sd_max_rg_data * sdp->sd_sb.sb_bsize / 2; | |
868 | if (!bytes) | |
869 | bytes = UINT_MAX; | |
6905d9e4 BM |
870 | bytes &= bsize_mask; |
871 | if (bytes == 0) | |
872 | bytes = sdp->sd_sb.sb_bsize; | |
2fe17c10 CH |
873 | |
874 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &ip->i_gh); | |
875 | error = gfs2_glock_nq(&ip->i_gh); | |
876 | if (unlikely(error)) | |
877 | goto out_uninit; | |
878 | ||
879 | if (!gfs2_write_alloc_required(ip, offset, len)) | |
880 | goto out_unlock; | |
881 | ||
882 | while (len > 0) { | |
883 | if (len < bytes) | |
884 | bytes = len; | |
885 | al = gfs2_alloc_get(ip); | |
886 | if (!al) { | |
887 | error = -ENOMEM; | |
888 | goto out_unlock; | |
889 | } | |
890 | ||
891 | error = gfs2_quota_lock_check(ip); | |
892 | if (error) | |
893 | goto out_alloc_put; | |
894 | ||
895 | retry: | |
896 | gfs2_write_calc_reserv(ip, bytes, &data_blocks, &ind_blocks); | |
897 | ||
898 | al->al_requested = data_blocks + ind_blocks; | |
899 | error = gfs2_inplace_reserve(ip); | |
900 | if (error) { | |
901 | if (error == -ENOSPC && bytes > sdp->sd_sb.sb_bsize) { | |
902 | bytes >>= 1; | |
6905d9e4 BM |
903 | bytes &= bsize_mask; |
904 | if (bytes == 0) | |
905 | bytes = sdp->sd_sb.sb_bsize; | |
2fe17c10 CH |
906 | goto retry; |
907 | } | |
908 | goto out_qunlock; | |
909 | } | |
910 | max_bytes = bytes; | |
911 | calc_max_reserv(ip, len, &max_bytes, &data_blocks, &ind_blocks); | |
912 | al->al_requested = data_blocks + ind_blocks; | |
913 | ||
914 | rblocks = RES_DINODE + ind_blocks + RES_STATFS + RES_QUOTA + | |
915 | RES_RG_HDR + gfs2_rg_blocks(al); | |
916 | if (gfs2_is_jdata(ip)) | |
917 | rblocks += data_blocks ? data_blocks : 1; | |
918 | ||
919 | error = gfs2_trans_begin(sdp, rblocks, | |
920 | PAGE_CACHE_SIZE/sdp->sd_sb.sb_bsize); | |
921 | if (error) | |
922 | goto out_trans_fail; | |
923 | ||
924 | error = fallocate_chunk(inode, offset, max_bytes, mode); | |
925 | gfs2_trans_end(sdp); | |
926 | ||
927 | if (error) | |
928 | goto out_trans_fail; | |
929 | ||
930 | len -= max_bytes; | |
931 | offset += max_bytes; | |
932 | gfs2_inplace_release(ip); | |
933 | gfs2_quota_unlock(ip); | |
934 | gfs2_alloc_put(ip); | |
935 | } | |
936 | goto out_unlock; | |
937 | ||
938 | out_trans_fail: | |
939 | gfs2_inplace_release(ip); | |
940 | out_qunlock: | |
941 | gfs2_quota_unlock(ip); | |
942 | out_alloc_put: | |
943 | gfs2_alloc_put(ip); | |
944 | out_unlock: | |
945 | gfs2_glock_dq(&ip->i_gh); | |
946 | out_uninit: | |
947 | gfs2_holder_uninit(&ip->i_gh); | |
948 | return error; | |
949 | } | |
950 | ||
f057f6cd SW |
951 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM |
952 | ||
60446067 ME |
953 | /** |
954 | * gfs2_setlease - acquire/release a file lease | |
955 | * @file: the file pointer | |
956 | * @arg: lease type | |
957 | * @fl: file lock | |
958 | * | |
f057f6cd SW |
959 | * We don't currently have a way to enforce a lease across the whole |
960 | * cluster; until we do, disable leases (by just returning -EINVAL), | |
961 | * unless the administrator has requested purely local locking. | |
962 | * | |
b89f4321 AB |
963 | * Locking: called under lock_flocks |
964 | * | |
60446067 ME |
965 | * Returns: errno |
966 | */ | |
967 | ||
968 | static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl) | |
969 | { | |
f057f6cd | 970 | return -EINVAL; |
da755fdb SW |
971 | } |
972 | ||
b3b94faa DT |
973 | /** |
974 | * gfs2_lock - acquire/release a posix lock on a file | |
975 | * @file: the file pointer | |
976 | * @cmd: either modify or retrieve lock state, possibly wait | |
977 | * @fl: type and range of lock | |
978 | * | |
979 | * Returns: errno | |
980 | */ | |
981 | ||
982 | static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl) | |
983 | { | |
feaa7bba SW |
984 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
985 | struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host); | |
f057f6cd | 986 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
b3b94faa | 987 | |
b3b94faa DT |
988 | if (!(fl->fl_flags & FL_POSIX)) |
989 | return -ENOLCK; | |
720e7749 | 990 | if (__mandatory_lock(&ip->i_inode) && fl->fl_type != F_UNLCK) |
b3b94faa DT |
991 | return -ENOLCK; |
992 | ||
586759f0 ME |
993 | if (cmd == F_CANCELLK) { |
994 | /* Hack: */ | |
995 | cmd = F_SETLK; | |
996 | fl->fl_type = F_UNLCK; | |
997 | } | |
f057f6cd SW |
998 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) |
999 | return -EIO; | |
b3b94faa | 1000 | if (IS_GETLK(cmd)) |
f057f6cd | 1001 | return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl); |
b3b94faa | 1002 | else if (fl->fl_type == F_UNLCK) |
f057f6cd | 1003 | return dlm_posix_unlock(ls->ls_dlm, ip->i_no_addr, file, fl); |
b3b94faa | 1004 | else |
f057f6cd | 1005 | return dlm_posix_lock(ls->ls_dlm, ip->i_no_addr, file, cmd, fl); |
b3b94faa DT |
1006 | } |
1007 | ||
b3b94faa DT |
1008 | static int do_flock(struct file *file, int cmd, struct file_lock *fl) |
1009 | { | |
5c676f6d | 1010 | struct gfs2_file *fp = file->private_data; |
b3b94faa | 1011 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
81454098 | 1012 | struct gfs2_inode *ip = GFS2_I(file->f_path.dentry->d_inode); |
b3b94faa DT |
1013 | struct gfs2_glock *gl; |
1014 | unsigned int state; | |
1015 | int flags; | |
1016 | int error = 0; | |
1017 | ||
1018 | state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
6802e340 | 1019 | flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; |
b3b94faa | 1020 | |
f55ab26a | 1021 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
1022 | |
1023 | gl = fl_gh->gh_gl; | |
1024 | if (gl) { | |
1025 | if (fl_gh->gh_state == state) | |
1026 | goto out; | |
b3b94faa | 1027 | flock_lock_file_wait(file, |
907b9bce | 1028 | &(struct file_lock){.fl_type = F_UNLCK}); |
b4c20166 AD |
1029 | gfs2_glock_dq_wait(fl_gh); |
1030 | gfs2_holder_reinit(state, flags, fl_gh); | |
b3b94faa | 1031 | } else { |
6802e340 SW |
1032 | error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr, |
1033 | &gfs2_flock_glops, CREATE, &gl); | |
b3b94faa DT |
1034 | if (error) |
1035 | goto out; | |
b4c20166 AD |
1036 | gfs2_holder_init(gl, state, flags, fl_gh); |
1037 | gfs2_glock_put(gl); | |
b3b94faa | 1038 | } |
b3b94faa DT |
1039 | error = gfs2_glock_nq(fl_gh); |
1040 | if (error) { | |
1041 | gfs2_holder_uninit(fl_gh); | |
1042 | if (error == GLR_TRYFAILED) | |
1043 | error = -EAGAIN; | |
1044 | } else { | |
1045 | error = flock_lock_file_wait(file, fl); | |
feaa7bba | 1046 | gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error); |
b3b94faa DT |
1047 | } |
1048 | ||
420b9e5e | 1049 | out: |
f55ab26a | 1050 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
1051 | return error; |
1052 | } | |
1053 | ||
1054 | static void do_unflock(struct file *file, struct file_lock *fl) | |
1055 | { | |
5c676f6d | 1056 | struct gfs2_file *fp = file->private_data; |
b3b94faa DT |
1057 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
1058 | ||
f55ab26a | 1059 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa | 1060 | flock_lock_file_wait(file, fl); |
0a33443b SW |
1061 | if (fl_gh->gh_gl) { |
1062 | gfs2_glock_dq_wait(fl_gh); | |
1063 | gfs2_holder_uninit(fl_gh); | |
1064 | } | |
f55ab26a | 1065 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
1066 | } |
1067 | ||
1068 | /** | |
1069 | * gfs2_flock - acquire/release a flock lock on a file | |
1070 | * @file: the file pointer | |
1071 | * @cmd: either modify or retrieve lock state, possibly wait | |
1072 | * @fl: type and range of lock | |
1073 | * | |
1074 | * Returns: errno | |
1075 | */ | |
1076 | ||
1077 | static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl) | |
1078 | { | |
b3b94faa DT |
1079 | if (!(fl->fl_flags & FL_FLOCK)) |
1080 | return -ENOLCK; | |
a12af1eb AD |
1081 | if (fl->fl_type & LOCK_MAND) |
1082 | return -EOPNOTSUPP; | |
b3b94faa | 1083 | |
b3b94faa DT |
1084 | if (fl->fl_type == F_UNLCK) { |
1085 | do_unflock(file, fl); | |
1086 | return 0; | |
d00223f1 | 1087 | } else { |
b3b94faa | 1088 | return do_flock(file, cmd, fl); |
d00223f1 | 1089 | } |
b3b94faa DT |
1090 | } |
1091 | ||
10d21988 | 1092 | const struct file_operations gfs2_file_fops = { |
26c1a574 | 1093 | .llseek = gfs2_llseek, |
d00223f1 | 1094 | .read = do_sync_read, |
26c1a574 | 1095 | .aio_read = generic_file_aio_read, |
d00223f1 | 1096 | .write = do_sync_write, |
56aa616a | 1097 | .aio_write = gfs2_file_aio_write, |
26c1a574 SW |
1098 | .unlocked_ioctl = gfs2_ioctl, |
1099 | .mmap = gfs2_mmap, | |
1100 | .open = gfs2_open, | |
1101 | .release = gfs2_close, | |
1102 | .fsync = gfs2_fsync, | |
1103 | .lock = gfs2_lock, | |
26c1a574 SW |
1104 | .flock = gfs2_flock, |
1105 | .splice_read = generic_file_splice_read, | |
1106 | .splice_write = generic_file_splice_write, | |
60446067 | 1107 | .setlease = gfs2_setlease, |
2fe17c10 | 1108 | .fallocate = gfs2_fallocate, |
b3b94faa DT |
1109 | }; |
1110 | ||
10d21988 | 1111 | const struct file_operations gfs2_dir_fops = { |
26c1a574 SW |
1112 | .readdir = gfs2_readdir, |
1113 | .unlocked_ioctl = gfs2_ioctl, | |
1114 | .open = gfs2_open, | |
1115 | .release = gfs2_close, | |
1116 | .fsync = gfs2_fsync, | |
1117 | .lock = gfs2_lock, | |
1118 | .flock = gfs2_flock, | |
6038f373 | 1119 | .llseek = default_llseek, |
b3b94faa DT |
1120 | }; |
1121 | ||
f057f6cd SW |
1122 | #endif /* CONFIG_GFS2_FS_LOCKING_DLM */ |
1123 | ||
10d21988 | 1124 | const struct file_operations gfs2_file_fops_nolock = { |
c97bfe43 WC |
1125 | .llseek = gfs2_llseek, |
1126 | .read = do_sync_read, | |
1127 | .aio_read = generic_file_aio_read, | |
1128 | .write = do_sync_write, | |
56aa616a | 1129 | .aio_write = gfs2_file_aio_write, |
c97bfe43 WC |
1130 | .unlocked_ioctl = gfs2_ioctl, |
1131 | .mmap = gfs2_mmap, | |
1132 | .open = gfs2_open, | |
1133 | .release = gfs2_close, | |
1134 | .fsync = gfs2_fsync, | |
1135 | .splice_read = generic_file_splice_read, | |
1136 | .splice_write = generic_file_splice_write, | |
f057f6cd | 1137 | .setlease = generic_setlease, |
2fe17c10 | 1138 | .fallocate = gfs2_fallocate, |
c97bfe43 WC |
1139 | }; |
1140 | ||
10d21988 | 1141 | const struct file_operations gfs2_dir_fops_nolock = { |
c97bfe43 WC |
1142 | .readdir = gfs2_readdir, |
1143 | .unlocked_ioctl = gfs2_ioctl, | |
1144 | .open = gfs2_open, | |
1145 | .release = gfs2_close, | |
1146 | .fsync = gfs2_fsync, | |
6038f373 | 1147 | .llseek = default_llseek, |
c97bfe43 WC |
1148 | }; |
1149 |