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> |
2fe17c10 CH |
21 | #include <linux/falloc.h> |
22 | #include <linux/swap.h> | |
71b86f56 | 23 | #include <linux/crc32.h> |
33c3de32 | 24 | #include <linux/writeback.h> |
b3b94faa | 25 | #include <asm/uaccess.h> |
f057f6cd SW |
26 | #include <linux/dlm.h> |
27 | #include <linux/dlm_plock.h> | |
a27bb332 | 28 | #include <linux/aio.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 | |
965c8e59 | 48 | * @whence: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END) |
b3b94faa DT |
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 | ||
965c8e59 | 56 | static loff_t gfs2_llseek(struct file *file, loff_t offset, int whence) |
b3b94faa | 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 | ||
965c8e59 | 62 | switch (whence) { |
9453615a SW |
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) { | |
965c8e59 | 69 | error = generic_file_llseek(file, offset, whence); |
b3b94faa DT |
70 | gfs2_glock_dq_uninit(&i_gh); |
71 | } | |
9453615a SW |
72 | break; |
73 | case SEEK_CUR: | |
74 | case SEEK_SET: | |
965c8e59 | 75 | error = generic_file_llseek(file, offset, whence); |
9453615a SW |
76 | break; |
77 | default: | |
78 | error = -EINVAL; | |
79 | } | |
b3b94faa DT |
80 | |
81 | return error; | |
82 | } | |
83 | ||
b3b94faa | 84 | /** |
d81a8ef5 | 85 | * gfs2_readdir - Iterator for a directory |
b3b94faa | 86 | * @file: The directory to read from |
d81a8ef5 | 87 | * @ctx: What to feed directory entries to |
b3b94faa DT |
88 | * |
89 | * Returns: errno | |
90 | */ | |
91 | ||
d81a8ef5 | 92 | static int gfs2_readdir(struct file *file, struct dir_context *ctx) |
b3b94faa | 93 | { |
71b86f56 | 94 | struct inode *dir = file->f_mapping->host; |
feaa7bba | 95 | struct gfs2_inode *dip = GFS2_I(dir); |
b3b94faa | 96 | struct gfs2_holder d_gh; |
b3b94faa DT |
97 | int error; |
98 | ||
d81a8ef5 AV |
99 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh); |
100 | if (error) | |
b3b94faa | 101 | return error; |
b3b94faa | 102 | |
d81a8ef5 | 103 | error = gfs2_dir_read(dir, ctx, &file->f_ra); |
b3b94faa DT |
104 | |
105 | gfs2_glock_dq_uninit(&d_gh); | |
106 | ||
b3b94faa DT |
107 | return error; |
108 | } | |
109 | ||
128e5eba SW |
110 | /** |
111 | * fsflags_cvt | |
112 | * @table: A table of 32 u32 flags | |
113 | * @val: a 32 bit value to convert | |
114 | * | |
115 | * This function can be used to convert between fsflags values and | |
116 | * GFS2's own flags values. | |
117 | * | |
118 | * Returns: the converted flags | |
119 | */ | |
120 | static u32 fsflags_cvt(const u32 *table, u32 val) | |
121 | { | |
122 | u32 res = 0; | |
123 | while(val) { | |
124 | if (val & 1) | |
125 | res |= *table; | |
126 | table++; | |
127 | val >>= 1; | |
128 | } | |
129 | return res; | |
130 | } | |
b3b94faa | 131 | |
128e5eba SW |
132 | static const u32 fsflags_to_gfs2[32] = { |
133 | [3] = GFS2_DIF_SYNC, | |
134 | [4] = GFS2_DIF_IMMUTABLE, | |
135 | [5] = GFS2_DIF_APPENDONLY, | |
136 | [7] = GFS2_DIF_NOATIME, | |
137 | [12] = GFS2_DIF_EXHASH, | |
b9af7ca6 | 138 | [14] = GFS2_DIF_INHERIT_JDATA, |
23d0bb83 | 139 | [17] = GFS2_DIF_TOPDIR, |
71b86f56 SW |
140 | }; |
141 | ||
128e5eba SW |
142 | static const u32 gfs2_to_fsflags[32] = { |
143 | [gfs2fl_Sync] = FS_SYNC_FL, | |
144 | [gfs2fl_Immutable] = FS_IMMUTABLE_FL, | |
145 | [gfs2fl_AppendOnly] = FS_APPEND_FL, | |
146 | [gfs2fl_NoAtime] = FS_NOATIME_FL, | |
147 | [gfs2fl_ExHash] = FS_INDEX_FL, | |
23d0bb83 | 148 | [gfs2fl_TopLevel] = FS_TOPDIR_FL, |
128e5eba | 149 | [gfs2fl_InheritJdata] = FS_JOURNAL_DATA_FL, |
7ea9ea83 | 150 | }; |
71b86f56 | 151 | |
b09e593d | 152 | static int gfs2_get_flags(struct file *filp, u32 __user *ptr) |
71b86f56 | 153 | { |
496ad9aa | 154 | struct inode *inode = file_inode(filp); |
feaa7bba | 155 | struct gfs2_inode *ip = GFS2_I(inode); |
71b86f56 SW |
156 | struct gfs2_holder gh; |
157 | int error; | |
128e5eba | 158 | u32 fsflags; |
71b86f56 | 159 | |
719ee344 SW |
160 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh); |
161 | error = gfs2_glock_nq(&gh); | |
71b86f56 SW |
162 | if (error) |
163 | return error; | |
907b9bce | 164 | |
383f01fb SW |
165 | fsflags = fsflags_cvt(gfs2_to_fsflags, ip->i_diskflags); |
166 | if (!S_ISDIR(inode->i_mode) && ip->i_diskflags & GFS2_DIF_JDATA) | |
c9f6a6bb | 167 | fsflags |= FS_JOURNAL_DATA_FL; |
128e5eba | 168 | if (put_user(fsflags, ptr)) |
71b86f56 SW |
169 | error = -EFAULT; |
170 | ||
3cc3f710 | 171 | gfs2_glock_dq(&gh); |
71b86f56 SW |
172 | gfs2_holder_uninit(&gh); |
173 | return error; | |
174 | } | |
175 | ||
6b124d8d SW |
176 | void gfs2_set_inode_flags(struct inode *inode) |
177 | { | |
178 | struct gfs2_inode *ip = GFS2_I(inode); | |
6b124d8d SW |
179 | unsigned int flags = inode->i_flags; |
180 | ||
9964afbb SW |
181 | flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|S_NOSEC); |
182 | if ((ip->i_eattr == 0) && !is_sxid(inode->i_mode)) | |
183 | inode->i_flags |= S_NOSEC; | |
383f01fb | 184 | if (ip->i_diskflags & GFS2_DIF_IMMUTABLE) |
6b124d8d | 185 | flags |= S_IMMUTABLE; |
383f01fb | 186 | if (ip->i_diskflags & GFS2_DIF_APPENDONLY) |
6b124d8d | 187 | flags |= S_APPEND; |
383f01fb | 188 | if (ip->i_diskflags & GFS2_DIF_NOATIME) |
6b124d8d | 189 | flags |= S_NOATIME; |
383f01fb | 190 | if (ip->i_diskflags & GFS2_DIF_SYNC) |
6b124d8d SW |
191 | flags |= S_SYNC; |
192 | inode->i_flags = flags; | |
193 | } | |
194 | ||
71b86f56 SW |
195 | /* Flags that can be set by user space */ |
196 | #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \ | |
71b86f56 SW |
197 | GFS2_DIF_IMMUTABLE| \ |
198 | GFS2_DIF_APPENDONLY| \ | |
199 | GFS2_DIF_NOATIME| \ | |
200 | GFS2_DIF_SYNC| \ | |
201 | GFS2_DIF_SYSTEM| \ | |
23d0bb83 | 202 | GFS2_DIF_TOPDIR| \ |
71b86f56 SW |
203 | GFS2_DIF_INHERIT_JDATA) |
204 | ||
205 | /** | |
206 | * gfs2_set_flags - set flags on an inode | |
207 | * @inode: The inode | |
208 | * @flags: The flags to set | |
209 | * @mask: Indicates which flags are valid | |
210 | * | |
211 | */ | |
b09e593d | 212 | static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask) |
71b86f56 | 213 | { |
496ad9aa | 214 | struct inode *inode = file_inode(filp); |
feaa7bba SW |
215 | struct gfs2_inode *ip = GFS2_I(inode); |
216 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
71b86f56 SW |
217 | struct buffer_head *bh; |
218 | struct gfs2_holder gh; | |
219 | int error; | |
55eccc6d | 220 | u32 new_flags, flags; |
71b86f56 | 221 | |
a561be71 | 222 | error = mnt_want_write_file(filp); |
52f341cf | 223 | if (error) |
71b86f56 SW |
224 | return error; |
225 | ||
f58ba889 MS |
226 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); |
227 | if (error) | |
228 | goto out_drop_write; | |
229 | ||
7df0e039 | 230 | error = -EACCES; |
2e149670 | 231 | if (!inode_owner_or_capable(inode)) |
7df0e039 SW |
232 | goto out; |
233 | ||
234 | error = 0; | |
383f01fb | 235 | flags = ip->i_diskflags; |
55eccc6d | 236 | new_flags = (flags & ~mask) | (reqflags & mask); |
71b86f56 SW |
237 | if ((new_flags ^ flags) == 0) |
238 | goto out; | |
239 | ||
240 | error = -EINVAL; | |
241 | if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET) | |
242 | goto out; | |
243 | ||
71b86f56 SW |
244 | error = -EPERM; |
245 | if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE)) | |
246 | goto out; | |
247 | if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY)) | |
248 | goto out; | |
907b9bce | 249 | if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) && |
b9cb9813 | 250 | !capable(CAP_LINUX_IMMUTABLE)) |
71b86f56 | 251 | goto out; |
b9cb9813 | 252 | if (!IS_IMMUTABLE(inode)) { |
10556cb2 | 253 | error = gfs2_permission(inode, MAY_WRITE); |
b9cb9813 SW |
254 | if (error) |
255 | goto out; | |
256 | } | |
5561093e SW |
257 | if ((flags ^ new_flags) & GFS2_DIF_JDATA) { |
258 | if (flags & GFS2_DIF_JDATA) | |
259 | gfs2_log_flush(sdp, ip->i_gl); | |
260 | error = filemap_fdatawrite(inode->i_mapping); | |
261 | if (error) | |
262 | goto out; | |
263 | error = filemap_fdatawait(inode->i_mapping); | |
264 | if (error) | |
265 | goto out; | |
266 | } | |
55eccc6d | 267 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); |
71b86f56 SW |
268 | if (error) |
269 | goto out; | |
55eccc6d SW |
270 | error = gfs2_meta_inode_buffer(ip, &bh); |
271 | if (error) | |
272 | goto out_trans_end; | |
350a9b0a | 273 | gfs2_trans_add_meta(ip->i_gl, bh); |
383f01fb | 274 | ip->i_diskflags = new_flags; |
539e5d6b | 275 | gfs2_dinode_out(ip, bh->b_data); |
71b86f56 | 276 | brelse(bh); |
6b124d8d | 277 | gfs2_set_inode_flags(inode); |
5561093e | 278 | gfs2_set_aops(inode); |
55eccc6d SW |
279 | out_trans_end: |
280 | gfs2_trans_end(sdp); | |
71b86f56 SW |
281 | out: |
282 | gfs2_glock_dq_uninit(&gh); | |
f58ba889 | 283 | out_drop_write: |
2a79f17e | 284 | mnt_drop_write_file(filp); |
71b86f56 SW |
285 | return error; |
286 | } | |
287 | ||
b09e593d | 288 | static int gfs2_set_flags(struct file *filp, u32 __user *ptr) |
71b86f56 | 289 | { |
496ad9aa | 290 | struct inode *inode = file_inode(filp); |
128e5eba | 291 | u32 fsflags, gfsflags; |
7df0e039 | 292 | |
128e5eba | 293 | if (get_user(fsflags, ptr)) |
71b86f56 | 294 | return -EFAULT; |
7df0e039 | 295 | |
128e5eba | 296 | gfsflags = fsflags_cvt(fsflags_to_gfs2, fsflags); |
b9af7ca6 | 297 | if (!S_ISDIR(inode->i_mode)) { |
23d0bb83 | 298 | gfsflags &= ~GFS2_DIF_TOPDIR; |
b9af7ca6 SW |
299 | if (gfsflags & GFS2_DIF_INHERIT_JDATA) |
300 | gfsflags ^= (GFS2_DIF_JDATA | GFS2_DIF_INHERIT_JDATA); | |
b9af7ca6 SW |
301 | return do_gfs2_set_flags(filp, gfsflags, ~0); |
302 | } | |
303 | return do_gfs2_set_flags(filp, gfsflags, ~GFS2_DIF_JDATA); | |
71b86f56 SW |
304 | } |
305 | ||
b09e593d | 306 | static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) |
71b86f56 SW |
307 | { |
308 | switch(cmd) { | |
128e5eba | 309 | case FS_IOC_GETFLAGS: |
b09e593d | 310 | return gfs2_get_flags(filp, (u32 __user *)arg); |
128e5eba | 311 | case FS_IOC_SETFLAGS: |
b09e593d | 312 | return gfs2_set_flags(filp, (u32 __user *)arg); |
66fc061b SW |
313 | case FITRIM: |
314 | return gfs2_fitrim(filp, (void __user *)arg); | |
71b86f56 SW |
315 | } |
316 | return -ENOTTY; | |
317 | } | |
318 | ||
da1dfb6a SW |
319 | /** |
320 | * gfs2_size_hint - Give a hint to the size of a write request | |
321 | * @file: The struct file | |
322 | * @offset: The file offset of the write | |
323 | * @size: The length of the write | |
324 | * | |
325 | * When we are about to do a write, this function records the total | |
326 | * write size in order to provide a suitable hint to the lower layers | |
327 | * about how many blocks will be required. | |
328 | * | |
329 | */ | |
330 | ||
331 | static void gfs2_size_hint(struct file *filep, loff_t offset, size_t size) | |
332 | { | |
496ad9aa | 333 | struct inode *inode = file_inode(filep); |
da1dfb6a SW |
334 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
335 | struct gfs2_inode *ip = GFS2_I(inode); | |
336 | size_t blks = (size + sdp->sd_sb.sb_bsize - 1) >> sdp->sd_sb.sb_bsize_shift; | |
337 | int hint = min_t(size_t, INT_MAX, blks); | |
338 | ||
339 | atomic_set(&ip->i_res->rs_sizehint, hint); | |
340 | } | |
341 | ||
3cc3f710 SW |
342 | /** |
343 | * gfs2_allocate_page_backing - Use bmap to allocate blocks | |
344 | * @page: The (locked) page to allocate backing for | |
345 | * | |
346 | * We try to allocate all the blocks required for the page in | |
347 | * one go. This might fail for various reasons, so we keep | |
348 | * trying until all the blocks to back this page are allocated. | |
349 | * If some of the blocks are already allocated, thats ok too. | |
350 | */ | |
351 | ||
352 | static int gfs2_allocate_page_backing(struct page *page) | |
353 | { | |
354 | struct inode *inode = page->mapping->host; | |
355 | struct buffer_head bh; | |
356 | unsigned long size = PAGE_CACHE_SIZE; | |
357 | u64 lblock = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
358 | ||
359 | do { | |
360 | bh.b_state = 0; | |
361 | bh.b_size = size; | |
e9e1ef2b | 362 | gfs2_block_map(inode, lblock, &bh, 1); |
3cc3f710 SW |
363 | if (!buffer_mapped(&bh)) |
364 | return -EIO; | |
365 | size -= bh.b_size; | |
366 | lblock += (bh.b_size >> inode->i_blkbits); | |
367 | } while(size > 0); | |
368 | return 0; | |
369 | } | |
370 | ||
371 | /** | |
372 | * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable | |
373 | * @vma: The virtual memory area | |
374 | * @page: The page which is about to become writable | |
375 | * | |
376 | * When the page becomes writable, we need to ensure that we have | |
377 | * blocks allocated on disk to back that page. | |
378 | */ | |
379 | ||
c2ec175c | 380 | static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) |
3cc3f710 | 381 | { |
c2ec175c | 382 | struct page *page = vmf->page; |
496ad9aa | 383 | struct inode *inode = file_inode(vma->vm_file); |
3cc3f710 SW |
384 | struct gfs2_inode *ip = GFS2_I(inode); |
385 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
386 | unsigned long last_index; | |
c8f554b9 | 387 | u64 pos = page->index << PAGE_CACHE_SHIFT; |
3cc3f710 | 388 | unsigned int data_blocks, ind_blocks, rblocks; |
3cc3f710 | 389 | struct gfs2_holder gh; |
13d921e3 | 390 | loff_t size; |
3cc3f710 SW |
391 | int ret; |
392 | ||
39263d5e | 393 | sb_start_pagefault(inode->i_sb); |
13d921e3 | 394 | |
a63e9b2e JK |
395 | /* Update file times before taking page lock */ |
396 | file_update_time(vma->vm_file); | |
13d921e3 | 397 | |
2b3dcf35 BP |
398 | ret = get_write_access(inode); |
399 | if (ret) | |
400 | goto out; | |
401 | ||
0a305e49 BP |
402 | ret = gfs2_rs_alloc(ip); |
403 | if (ret) | |
2b3dcf35 | 404 | goto out_write_access; |
0a305e49 | 405 | |
da1dfb6a | 406 | gfs2_size_hint(vma->vm_file, pos, PAGE_CACHE_SIZE); |
8e2e0047 | 407 | |
719ee344 SW |
408 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); |
409 | ret = gfs2_glock_nq(&gh); | |
3cc3f710 | 410 | if (ret) |
2b3dcf35 | 411 | goto out_uninit; |
3cc3f710 | 412 | |
9c538837 SW |
413 | set_bit(GLF_DIRTY, &ip->i_gl->gl_flags); |
414 | set_bit(GIF_SW_PAGED, &ip->i_flags); | |
415 | ||
13d921e3 SW |
416 | if (!gfs2_write_alloc_required(ip, pos, PAGE_CACHE_SIZE)) { |
417 | lock_page(page); | |
418 | if (!PageUptodate(page) || page->mapping != inode->i_mapping) { | |
419 | ret = -EAGAIN; | |
420 | unlock_page(page); | |
421 | } | |
3cc3f710 | 422 | goto out_unlock; |
13d921e3 SW |
423 | } |
424 | ||
5407e242 BP |
425 | ret = gfs2_rindex_update(sdp); |
426 | if (ret) | |
6dbd8224 SW |
427 | goto out_unlock; |
428 | ||
d82661d9 | 429 | ret = gfs2_quota_lock_check(ip); |
3cc3f710 | 430 | if (ret) |
5407e242 | 431 | goto out_unlock; |
7ed122e4 | 432 | gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks); |
9dbe9610 | 433 | ret = gfs2_inplace_reserve(ip, data_blocks + ind_blocks, 0); |
3cc3f710 SW |
434 | if (ret) |
435 | goto out_quota_unlock; | |
436 | ||
437 | rblocks = RES_DINODE + ind_blocks; | |
438 | if (gfs2_is_jdata(ip)) | |
439 | rblocks += data_blocks ? data_blocks : 1; | |
bf97b673 | 440 | if (ind_blocks || data_blocks) { |
3cc3f710 | 441 | rblocks += RES_STATFS + RES_QUOTA; |
71f890f7 | 442 | rblocks += gfs2_rg_blocks(ip, data_blocks + ind_blocks); |
bf97b673 | 443 | } |
3cc3f710 SW |
444 | ret = gfs2_trans_begin(sdp, rblocks, 0); |
445 | if (ret) | |
446 | goto out_trans_fail; | |
447 | ||
448 | lock_page(page); | |
449 | ret = -EINVAL; | |
13d921e3 SW |
450 | size = i_size_read(inode); |
451 | last_index = (size - 1) >> PAGE_CACHE_SHIFT; | |
452 | /* Check page index against inode size */ | |
453 | if (size == 0 || (page->index > last_index)) | |
454 | goto out_trans_end; | |
455 | ||
456 | ret = -EAGAIN; | |
457 | /* If truncated, we must retry the operation, we may have raced | |
458 | * with the glock demotion code. | |
459 | */ | |
460 | if (!PageUptodate(page) || page->mapping != inode->i_mapping) | |
461 | goto out_trans_end; | |
462 | ||
463 | /* Unstuff, if required, and allocate backing blocks for page */ | |
b7fe2e39 | 464 | ret = 0; |
13d921e3 | 465 | if (gfs2_is_stuffed(ip)) |
3cc3f710 | 466 | ret = gfs2_unstuff_dinode(ip, page); |
13d921e3 SW |
467 | if (ret == 0) |
468 | ret = gfs2_allocate_page_backing(page); | |
3cc3f710 | 469 | |
13d921e3 SW |
470 | out_trans_end: |
471 | if (ret) | |
472 | unlock_page(page); | |
3cc3f710 SW |
473 | gfs2_trans_end(sdp); |
474 | out_trans_fail: | |
475 | gfs2_inplace_release(ip); | |
476 | out_quota_unlock: | |
477 | gfs2_quota_unlock(ip); | |
3cc3f710 SW |
478 | out_unlock: |
479 | gfs2_glock_dq(&gh); | |
2b3dcf35 | 480 | out_uninit: |
3cc3f710 | 481 | gfs2_holder_uninit(&gh); |
13d921e3 SW |
482 | if (ret == 0) { |
483 | set_page_dirty(page); | |
1d1d1a76 | 484 | wait_for_stable_page(page); |
13d921e3 | 485 | } |
2b3dcf35 BP |
486 | out_write_access: |
487 | put_write_access(inode); | |
488 | out: | |
39263d5e | 489 | sb_end_pagefault(inode->i_sb); |
13d921e3 | 490 | return block_page_mkwrite_return(ret); |
3cc3f710 SW |
491 | } |
492 | ||
f0f37e2f | 493 | static const struct vm_operations_struct gfs2_vm_ops = { |
3cc3f710 SW |
494 | .fault = filemap_fault, |
495 | .page_mkwrite = gfs2_page_mkwrite, | |
0b173bc4 | 496 | .remap_pages = generic_file_remap_pages, |
3cc3f710 SW |
497 | }; |
498 | ||
b3b94faa DT |
499 | /** |
500 | * gfs2_mmap - | |
501 | * @file: The file to map | |
502 | * @vma: The VMA which described the mapping | |
503 | * | |
48bf2b17 SW |
504 | * There is no need to get a lock here unless we should be updating |
505 | * atime. We ignore any locking errors since the only consequence is | |
506 | * a missed atime update (which will just be deferred until later). | |
507 | * | |
508 | * Returns: 0 | |
b3b94faa DT |
509 | */ |
510 | ||
511 | static int gfs2_mmap(struct file *file, struct vm_area_struct *vma) | |
512 | { | |
feaa7bba | 513 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
b3b94faa | 514 | |
b9c93bb7 SW |
515 | if (!(file->f_flags & O_NOATIME) && |
516 | !IS_NOATIME(&ip->i_inode)) { | |
48bf2b17 SW |
517 | struct gfs2_holder i_gh; |
518 | int error; | |
b3b94faa | 519 | |
3d162688 BM |
520 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, |
521 | &i_gh); | |
b9c93bb7 SW |
522 | if (error) |
523 | return error; | |
3d162688 BM |
524 | /* grab lock to update inode */ |
525 | gfs2_glock_dq_uninit(&i_gh); | |
526 | file_accessed(file); | |
48bf2b17 | 527 | } |
3cc3f710 | 528 | vma->vm_ops = &gfs2_vm_ops; |
b3b94faa | 529 | |
48bf2b17 | 530 | return 0; |
b3b94faa DT |
531 | } |
532 | ||
533 | /** | |
6d4ade98 SW |
534 | * gfs2_open_common - This is common to open and atomic_open |
535 | * @inode: The inode being opened | |
536 | * @file: The file being opened | |
b3b94faa | 537 | * |
6d4ade98 SW |
538 | * This maybe called under a glock or not depending upon how it has |
539 | * been called. We must always be called under a glock for regular | |
540 | * files, however. For other file types, it does not matter whether | |
541 | * we hold the glock or not. | |
542 | * | |
543 | * Returns: Error code or 0 for success | |
b3b94faa DT |
544 | */ |
545 | ||
6d4ade98 | 546 | int gfs2_open_common(struct inode *inode, struct file *file) |
b3b94faa | 547 | { |
b3b94faa | 548 | struct gfs2_file *fp; |
6d4ade98 SW |
549 | int ret; |
550 | ||
551 | if (S_ISREG(inode->i_mode)) { | |
552 | ret = generic_file_open(inode, file); | |
553 | if (ret) | |
554 | return ret; | |
555 | } | |
b3b94faa | 556 | |
6d4ade98 | 557 | fp = kzalloc(sizeof(struct gfs2_file), GFP_NOFS); |
b3b94faa DT |
558 | if (!fp) |
559 | return -ENOMEM; | |
560 | ||
f55ab26a | 561 | mutex_init(&fp->f_fl_mutex); |
b3b94faa | 562 | |
feaa7bba | 563 | gfs2_assert_warn(GFS2_SB(inode), !file->private_data); |
5c676f6d | 564 | file->private_data = fp; |
6d4ade98 SW |
565 | return 0; |
566 | } | |
567 | ||
568 | /** | |
569 | * gfs2_open - open a file | |
570 | * @inode: the inode to open | |
571 | * @file: the struct file for this opening | |
572 | * | |
573 | * After atomic_open, this function is only used for opening files | |
574 | * which are already cached. We must still get the glock for regular | |
575 | * files to ensure that we have the file size uptodate for the large | |
576 | * file check which is in the common code. That is only an issue for | |
577 | * regular files though. | |
578 | * | |
579 | * Returns: errno | |
580 | */ | |
581 | ||
582 | static int gfs2_open(struct inode *inode, struct file *file) | |
583 | { | |
584 | struct gfs2_inode *ip = GFS2_I(inode); | |
585 | struct gfs2_holder i_gh; | |
586 | int error; | |
587 | bool need_unlock = false; | |
b3b94faa | 588 | |
b60623c2 | 589 | if (S_ISREG(ip->i_inode.i_mode)) { |
b3b94faa DT |
590 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, |
591 | &i_gh); | |
592 | if (error) | |
6d4ade98 SW |
593 | return error; |
594 | need_unlock = true; | |
595 | } | |
b3b94faa | 596 | |
6d4ade98 | 597 | error = gfs2_open_common(inode, file); |
b3b94faa | 598 | |
6d4ade98 | 599 | if (need_unlock) |
b3b94faa | 600 | gfs2_glock_dq_uninit(&i_gh); |
b3b94faa | 601 | |
b3b94faa DT |
602 | return error; |
603 | } | |
604 | ||
605 | /** | |
df3fd117 | 606 | * gfs2_release - called to close a struct file |
b3b94faa DT |
607 | * @inode: the inode the struct file belongs to |
608 | * @file: the struct file being closed | |
609 | * | |
610 | * Returns: errno | |
611 | */ | |
612 | ||
df3fd117 | 613 | static int gfs2_release(struct inode *inode, struct file *file) |
b3b94faa | 614 | { |
0a305e49 | 615 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa | 616 | |
8e2e0047 | 617 | kfree(file->private_data); |
5c676f6d | 618 | file->private_data = NULL; |
b3b94faa | 619 | |
2b3dcf35 BP |
620 | if (!(file->f_mode & FMODE_WRITE)) |
621 | return 0; | |
0a305e49 | 622 | |
2b3dcf35 | 623 | gfs2_rs_delete(ip); |
b3b94faa DT |
624 | return 0; |
625 | } | |
626 | ||
627 | /** | |
628 | * gfs2_fsync - sync the dirty data for a file (across the cluster) | |
02c24a82 JB |
629 | * @file: the file that points to the dentry |
630 | * @start: the start position in the file to sync | |
631 | * @end: the end position in the file to sync | |
dba898b0 | 632 | * @datasync: set if we can ignore timestamp changes |
b3b94faa | 633 | * |
2f0264d5 SW |
634 | * We split the data flushing here so that we don't wait for the data |
635 | * until after we've also sent the metadata to disk. Note that for | |
636 | * data=ordered, we will write & wait for the data at the log flush | |
637 | * stage anyway, so this is unlikely to make much of a difference | |
638 | * except in the data=writeback case. | |
639 | * | |
640 | * If the fdatawrite fails due to any reason except -EIO, we will | |
641 | * continue the remainder of the fsync, although we'll still report | |
642 | * the error at the end. This is to match filemap_write_and_wait_range() | |
643 | * behaviour. | |
34126f9f | 644 | * |
b3b94faa DT |
645 | * Returns: errno |
646 | */ | |
647 | ||
02c24a82 JB |
648 | static int gfs2_fsync(struct file *file, loff_t start, loff_t end, |
649 | int datasync) | |
b3b94faa | 650 | { |
2f0264d5 SW |
651 | struct address_space *mapping = file->f_mapping; |
652 | struct inode *inode = mapping->host; | |
33c3de32 | 653 | int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC); |
dba898b0 | 654 | struct gfs2_inode *ip = GFS2_I(inode); |
87654896 | 655 | int ret = 0, ret1 = 0; |
b3b94faa | 656 | |
2f0264d5 SW |
657 | if (mapping->nrpages) { |
658 | ret1 = filemap_fdatawrite_range(mapping, start, end); | |
659 | if (ret1 == -EIO) | |
660 | return ret1; | |
661 | } | |
02c24a82 | 662 | |
dba898b0 SW |
663 | if (datasync) |
664 | sync_state &= ~I_DIRTY_SYNC; | |
b3b94faa | 665 | |
dba898b0 SW |
666 | if (sync_state) { |
667 | ret = sync_inode_metadata(inode, 1); | |
b5b24d7a | 668 | if (ret) |
dba898b0 | 669 | return ret; |
f1818529 SW |
670 | if (gfs2_is_jdata(ip)) |
671 | filemap_write_and_wait(mapping); | |
b5b24d7a | 672 | gfs2_ail_flush(ip->i_gl, 1); |
33c3de32 SW |
673 | } |
674 | ||
2f0264d5 SW |
675 | if (mapping->nrpages) |
676 | ret = filemap_fdatawait_range(mapping, start, end); | |
677 | ||
678 | return ret ? ret : ret1; | |
b3b94faa DT |
679 | } |
680 | ||
56aa616a SW |
681 | /** |
682 | * gfs2_file_aio_write - Perform a write to a file | |
683 | * @iocb: The io context | |
684 | * @iov: The data to write | |
685 | * @nr_segs: Number of @iov segments | |
686 | * @pos: The file position | |
687 | * | |
688 | * We have to do a lock/unlock here to refresh the inode size for | |
689 | * O_APPEND writes, otherwise we can land up writing at the wrong | |
690 | * offset. There is still a race, but provided the app is using its | |
691 | * own file locking, this will make O_APPEND work as expected. | |
692 | * | |
693 | */ | |
694 | ||
695 | static ssize_t gfs2_file_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
696 | unsigned long nr_segs, loff_t pos) | |
697 | { | |
698 | struct file *file = iocb->ki_filp; | |
8e2e0047 | 699 | size_t writesize = iov_length(iov, nr_segs); |
496ad9aa | 700 | struct gfs2_inode *ip = GFS2_I(file_inode(file)); |
0a305e49 BP |
701 | int ret; |
702 | ||
703 | ret = gfs2_rs_alloc(ip); | |
704 | if (ret) | |
705 | return ret; | |
56aa616a | 706 | |
da1dfb6a SW |
707 | gfs2_size_hint(file, pos, writesize); |
708 | ||
56aa616a | 709 | if (file->f_flags & O_APPEND) { |
56aa616a | 710 | struct gfs2_holder gh; |
56aa616a SW |
711 | |
712 | ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &gh); | |
713 | if (ret) | |
714 | return ret; | |
715 | gfs2_glock_dq_uninit(&gh); | |
716 | } | |
717 | ||
718 | return generic_file_aio_write(iocb, iov, nr_segs, pos); | |
719 | } | |
720 | ||
2fe17c10 CH |
721 | static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len, |
722 | int mode) | |
723 | { | |
724 | struct gfs2_inode *ip = GFS2_I(inode); | |
725 | struct buffer_head *dibh; | |
726 | int error; | |
58a7d5fb | 727 | loff_t size = len; |
64dd153c BM |
728 | unsigned int nr_blks; |
729 | sector_t lblock = offset >> inode->i_blkbits; | |
2fe17c10 CH |
730 | |
731 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
732 | if (unlikely(error)) | |
64dd153c | 733 | return error; |
2fe17c10 | 734 | |
350a9b0a | 735 | gfs2_trans_add_meta(ip->i_gl, dibh); |
2fe17c10 CH |
736 | |
737 | if (gfs2_is_stuffed(ip)) { | |
738 | error = gfs2_unstuff_dinode(ip, NULL); | |
739 | if (unlikely(error)) | |
740 | goto out; | |
741 | } | |
742 | ||
64dd153c BM |
743 | while (len) { |
744 | struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 }; | |
745 | bh_map.b_size = len; | |
746 | set_buffer_zeronew(&bh_map); | |
2fe17c10 | 747 | |
64dd153c BM |
748 | error = gfs2_block_map(inode, lblock, &bh_map, 1); |
749 | if (unlikely(error)) | |
750 | goto out; | |
751 | len -= bh_map.b_size; | |
752 | nr_blks = bh_map.b_size >> inode->i_blkbits; | |
753 | lblock += nr_blks; | |
754 | if (!buffer_new(&bh_map)) | |
755 | continue; | |
756 | if (unlikely(!buffer_zeronew(&bh_map))) { | |
757 | error = -EIO; | |
2fe17c10 | 758 | goto out; |
64dd153c | 759 | } |
2fe17c10 | 760 | } |
58a7d5fb BM |
761 | if (offset + size > inode->i_size && !(mode & FALLOC_FL_KEEP_SIZE)) |
762 | i_size_write(inode, offset + size); | |
2fe17c10 | 763 | |
2fe17c10 CH |
764 | mark_inode_dirty(inode); |
765 | ||
2fe17c10 | 766 | out: |
64dd153c | 767 | brelse(dibh); |
2fe17c10 CH |
768 | return error; |
769 | } | |
770 | ||
771 | static void calc_max_reserv(struct gfs2_inode *ip, loff_t max, loff_t *len, | |
772 | unsigned int *data_blocks, unsigned int *ind_blocks) | |
773 | { | |
774 | const struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
54335b1f | 775 | unsigned int max_blocks = ip->i_rgd->rd_free_clone; |
2fe17c10 CH |
776 | unsigned int tmp, max_data = max_blocks - 3 * (sdp->sd_max_height - 1); |
777 | ||
778 | for (tmp = max_data; tmp > sdp->sd_diptrs;) { | |
779 | tmp = DIV_ROUND_UP(tmp, sdp->sd_inptrs); | |
780 | max_data -= tmp; | |
781 | } | |
782 | /* This calculation isn't the exact reverse of gfs2_write_calc_reserve, | |
783 | so it might end up with fewer data blocks */ | |
784 | if (max_data <= *data_blocks) | |
785 | return; | |
786 | *data_blocks = max_data; | |
787 | *ind_blocks = max_blocks - max_data; | |
788 | *len = ((loff_t)max_data - 3) << sdp->sd_sb.sb_bsize_shift; | |
789 | if (*len > max) { | |
790 | *len = max; | |
791 | gfs2_write_calc_reserv(ip, max, data_blocks, ind_blocks); | |
792 | } | |
793 | } | |
794 | ||
795 | static long gfs2_fallocate(struct file *file, int mode, loff_t offset, | |
796 | loff_t len) | |
797 | { | |
496ad9aa | 798 | struct inode *inode = file_inode(file); |
2fe17c10 CH |
799 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
800 | struct gfs2_inode *ip = GFS2_I(inode); | |
801 | unsigned int data_blocks = 0, ind_blocks = 0, rblocks; | |
802 | loff_t bytes, max_bytes; | |
2fe17c10 | 803 | int error; |
4442f2e0 SW |
804 | const loff_t pos = offset; |
805 | const loff_t count = len; | |
6905d9e4 | 806 | loff_t bsize_mask = ~((loff_t)sdp->sd_sb.sb_bsize - 1); |
2fe17c10 | 807 | loff_t next = (offset + len - 1) >> sdp->sd_sb.sb_bsize_shift; |
64dd153c | 808 | loff_t max_chunk_size = UINT_MAX & bsize_mask; |
2fe17c10 CH |
809 | next = (next + 1) << sdp->sd_sb.sb_bsize_shift; |
810 | ||
811 | /* We only support the FALLOC_FL_KEEP_SIZE mode */ | |
812 | if (mode & ~FALLOC_FL_KEEP_SIZE) | |
813 | return -EOPNOTSUPP; | |
814 | ||
6905d9e4 | 815 | offset &= bsize_mask; |
2fe17c10 CH |
816 | |
817 | len = next - offset; | |
818 | bytes = sdp->sd_max_rg_data * sdp->sd_sb.sb_bsize / 2; | |
819 | if (!bytes) | |
820 | bytes = UINT_MAX; | |
6905d9e4 BM |
821 | bytes &= bsize_mask; |
822 | if (bytes == 0) | |
823 | bytes = sdp->sd_sb.sb_bsize; | |
2fe17c10 | 824 | |
0a305e49 BP |
825 | error = gfs2_rs_alloc(ip); |
826 | if (error) | |
827 | return error; | |
828 | ||
2fe17c10 CH |
829 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &ip->i_gh); |
830 | error = gfs2_glock_nq(&ip->i_gh); | |
831 | if (unlikely(error)) | |
832 | goto out_uninit; | |
833 | ||
da1dfb6a | 834 | gfs2_size_hint(file, offset, len); |
8e2e0047 | 835 | |
2fe17c10 CH |
836 | while (len > 0) { |
837 | if (len < bytes) | |
838 | bytes = len; | |
58a7d5fb BM |
839 | if (!gfs2_write_alloc_required(ip, offset, bytes)) { |
840 | len -= bytes; | |
841 | offset += bytes; | |
842 | continue; | |
843 | } | |
2fe17c10 CH |
844 | error = gfs2_quota_lock_check(ip); |
845 | if (error) | |
5407e242 | 846 | goto out_unlock; |
2fe17c10 CH |
847 | |
848 | retry: | |
849 | gfs2_write_calc_reserv(ip, bytes, &data_blocks, &ind_blocks); | |
850 | ||
9dbe9610 | 851 | error = gfs2_inplace_reserve(ip, data_blocks + ind_blocks, 0); |
2fe17c10 CH |
852 | if (error) { |
853 | if (error == -ENOSPC && bytes > sdp->sd_sb.sb_bsize) { | |
854 | bytes >>= 1; | |
6905d9e4 BM |
855 | bytes &= bsize_mask; |
856 | if (bytes == 0) | |
857 | bytes = sdp->sd_sb.sb_bsize; | |
2fe17c10 CH |
858 | goto retry; |
859 | } | |
860 | goto out_qunlock; | |
861 | } | |
862 | max_bytes = bytes; | |
64dd153c BM |
863 | calc_max_reserv(ip, (len > max_chunk_size)? max_chunk_size: len, |
864 | &max_bytes, &data_blocks, &ind_blocks); | |
2fe17c10 CH |
865 | |
866 | rblocks = RES_DINODE + ind_blocks + RES_STATFS + RES_QUOTA + | |
71f890f7 | 867 | RES_RG_HDR + gfs2_rg_blocks(ip, data_blocks + ind_blocks); |
2fe17c10 CH |
868 | if (gfs2_is_jdata(ip)) |
869 | rblocks += data_blocks ? data_blocks : 1; | |
870 | ||
871 | error = gfs2_trans_begin(sdp, rblocks, | |
872 | PAGE_CACHE_SIZE/sdp->sd_sb.sb_bsize); | |
873 | if (error) | |
874 | goto out_trans_fail; | |
875 | ||
876 | error = fallocate_chunk(inode, offset, max_bytes, mode); | |
877 | gfs2_trans_end(sdp); | |
878 | ||
879 | if (error) | |
880 | goto out_trans_fail; | |
881 | ||
882 | len -= max_bytes; | |
883 | offset += max_bytes; | |
884 | gfs2_inplace_release(ip); | |
885 | gfs2_quota_unlock(ip); | |
2fe17c10 | 886 | } |
4442f2e0 SW |
887 | |
888 | if (error == 0) | |
889 | error = generic_write_sync(file, pos, count); | |
2fe17c10 CH |
890 | goto out_unlock; |
891 | ||
892 | out_trans_fail: | |
893 | gfs2_inplace_release(ip); | |
894 | out_qunlock: | |
895 | gfs2_quota_unlock(ip); | |
2fe17c10 CH |
896 | out_unlock: |
897 | gfs2_glock_dq(&ip->i_gh); | |
898 | out_uninit: | |
899 | gfs2_holder_uninit(&ip->i_gh); | |
900 | return error; | |
901 | } | |
902 | ||
f057f6cd SW |
903 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM |
904 | ||
60446067 ME |
905 | /** |
906 | * gfs2_setlease - acquire/release a file lease | |
907 | * @file: the file pointer | |
908 | * @arg: lease type | |
909 | * @fl: file lock | |
910 | * | |
f057f6cd SW |
911 | * We don't currently have a way to enforce a lease across the whole |
912 | * cluster; until we do, disable leases (by just returning -EINVAL), | |
913 | * unless the administrator has requested purely local locking. | |
914 | * | |
1c8c601a | 915 | * Locking: called under i_lock |
b89f4321 | 916 | * |
60446067 ME |
917 | * Returns: errno |
918 | */ | |
919 | ||
920 | static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl) | |
921 | { | |
f057f6cd | 922 | return -EINVAL; |
da755fdb SW |
923 | } |
924 | ||
b3b94faa DT |
925 | /** |
926 | * gfs2_lock - acquire/release a posix lock on a file | |
927 | * @file: the file pointer | |
928 | * @cmd: either modify or retrieve lock state, possibly wait | |
929 | * @fl: type and range of lock | |
930 | * | |
931 | * Returns: errno | |
932 | */ | |
933 | ||
934 | static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl) | |
935 | { | |
feaa7bba SW |
936 | struct gfs2_inode *ip = GFS2_I(file->f_mapping->host); |
937 | struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host); | |
f057f6cd | 938 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
b3b94faa | 939 | |
b3b94faa DT |
940 | if (!(fl->fl_flags & FL_POSIX)) |
941 | return -ENOLCK; | |
720e7749 | 942 | if (__mandatory_lock(&ip->i_inode) && fl->fl_type != F_UNLCK) |
b3b94faa DT |
943 | return -ENOLCK; |
944 | ||
586759f0 ME |
945 | if (cmd == F_CANCELLK) { |
946 | /* Hack: */ | |
947 | cmd = F_SETLK; | |
948 | fl->fl_type = F_UNLCK; | |
949 | } | |
c2952d20 SW |
950 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) { |
951 | if (fl->fl_type == F_UNLCK) | |
952 | posix_lock_file_wait(file, fl); | |
f057f6cd | 953 | return -EIO; |
c2952d20 | 954 | } |
b3b94faa | 955 | if (IS_GETLK(cmd)) |
f057f6cd | 956 | return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl); |
b3b94faa | 957 | else if (fl->fl_type == F_UNLCK) |
f057f6cd | 958 | return dlm_posix_unlock(ls->ls_dlm, ip->i_no_addr, file, fl); |
b3b94faa | 959 | else |
f057f6cd | 960 | return dlm_posix_lock(ls->ls_dlm, ip->i_no_addr, file, cmd, fl); |
b3b94faa DT |
961 | } |
962 | ||
b3b94faa DT |
963 | static int do_flock(struct file *file, int cmd, struct file_lock *fl) |
964 | { | |
5c676f6d | 965 | struct gfs2_file *fp = file->private_data; |
b3b94faa | 966 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
496ad9aa | 967 | struct gfs2_inode *ip = GFS2_I(file_inode(file)); |
b3b94faa DT |
968 | struct gfs2_glock *gl; |
969 | unsigned int state; | |
970 | int flags; | |
971 | int error = 0; | |
972 | ||
973 | state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
6802e340 | 974 | flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; |
b3b94faa | 975 | |
f55ab26a | 976 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
977 | |
978 | gl = fl_gh->gh_gl; | |
979 | if (gl) { | |
980 | if (fl_gh->gh_state == state) | |
981 | goto out; | |
b3b94faa | 982 | flock_lock_file_wait(file, |
907b9bce | 983 | &(struct file_lock){.fl_type = F_UNLCK}); |
b4c20166 AD |
984 | gfs2_glock_dq_wait(fl_gh); |
985 | gfs2_holder_reinit(state, flags, fl_gh); | |
b3b94faa | 986 | } else { |
6802e340 SW |
987 | error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr, |
988 | &gfs2_flock_glops, CREATE, &gl); | |
b3b94faa DT |
989 | if (error) |
990 | goto out; | |
b4c20166 AD |
991 | gfs2_holder_init(gl, state, flags, fl_gh); |
992 | gfs2_glock_put(gl); | |
b3b94faa | 993 | } |
b3b94faa DT |
994 | error = gfs2_glock_nq(fl_gh); |
995 | if (error) { | |
996 | gfs2_holder_uninit(fl_gh); | |
997 | if (error == GLR_TRYFAILED) | |
998 | error = -EAGAIN; | |
999 | } else { | |
1000 | error = flock_lock_file_wait(file, fl); | |
feaa7bba | 1001 | gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error); |
b3b94faa DT |
1002 | } |
1003 | ||
420b9e5e | 1004 | out: |
f55ab26a | 1005 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
1006 | return error; |
1007 | } | |
1008 | ||
1009 | static void do_unflock(struct file *file, struct file_lock *fl) | |
1010 | { | |
5c676f6d | 1011 | struct gfs2_file *fp = file->private_data; |
b3b94faa DT |
1012 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
1013 | ||
f55ab26a | 1014 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa | 1015 | flock_lock_file_wait(file, fl); |
0a33443b SW |
1016 | if (fl_gh->gh_gl) { |
1017 | gfs2_glock_dq_wait(fl_gh); | |
1018 | gfs2_holder_uninit(fl_gh); | |
1019 | } | |
f55ab26a | 1020 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
1021 | } |
1022 | ||
1023 | /** | |
1024 | * gfs2_flock - acquire/release a flock lock on a file | |
1025 | * @file: the file pointer | |
1026 | * @cmd: either modify or retrieve lock state, possibly wait | |
1027 | * @fl: type and range of lock | |
1028 | * | |
1029 | * Returns: errno | |
1030 | */ | |
1031 | ||
1032 | static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl) | |
1033 | { | |
b3b94faa DT |
1034 | if (!(fl->fl_flags & FL_FLOCK)) |
1035 | return -ENOLCK; | |
a12af1eb AD |
1036 | if (fl->fl_type & LOCK_MAND) |
1037 | return -EOPNOTSUPP; | |
b3b94faa | 1038 | |
b3b94faa DT |
1039 | if (fl->fl_type == F_UNLCK) { |
1040 | do_unflock(file, fl); | |
1041 | return 0; | |
d00223f1 | 1042 | } else { |
b3b94faa | 1043 | return do_flock(file, cmd, fl); |
d00223f1 | 1044 | } |
b3b94faa DT |
1045 | } |
1046 | ||
10d21988 | 1047 | const struct file_operations gfs2_file_fops = { |
26c1a574 | 1048 | .llseek = gfs2_llseek, |
d00223f1 | 1049 | .read = do_sync_read, |
26c1a574 | 1050 | .aio_read = generic_file_aio_read, |
d00223f1 | 1051 | .write = do_sync_write, |
56aa616a | 1052 | .aio_write = gfs2_file_aio_write, |
26c1a574 SW |
1053 | .unlocked_ioctl = gfs2_ioctl, |
1054 | .mmap = gfs2_mmap, | |
1055 | .open = gfs2_open, | |
df3fd117 | 1056 | .release = gfs2_release, |
26c1a574 SW |
1057 | .fsync = gfs2_fsync, |
1058 | .lock = gfs2_lock, | |
26c1a574 SW |
1059 | .flock = gfs2_flock, |
1060 | .splice_read = generic_file_splice_read, | |
1061 | .splice_write = generic_file_splice_write, | |
60446067 | 1062 | .setlease = gfs2_setlease, |
2fe17c10 | 1063 | .fallocate = gfs2_fallocate, |
b3b94faa DT |
1064 | }; |
1065 | ||
10d21988 | 1066 | const struct file_operations gfs2_dir_fops = { |
d81a8ef5 | 1067 | .iterate = gfs2_readdir, |
26c1a574 SW |
1068 | .unlocked_ioctl = gfs2_ioctl, |
1069 | .open = gfs2_open, | |
df3fd117 | 1070 | .release = gfs2_release, |
26c1a574 SW |
1071 | .fsync = gfs2_fsync, |
1072 | .lock = gfs2_lock, | |
1073 | .flock = gfs2_flock, | |
6038f373 | 1074 | .llseek = default_llseek, |
b3b94faa DT |
1075 | }; |
1076 | ||
f057f6cd SW |
1077 | #endif /* CONFIG_GFS2_FS_LOCKING_DLM */ |
1078 | ||
10d21988 | 1079 | const struct file_operations gfs2_file_fops_nolock = { |
c97bfe43 WC |
1080 | .llseek = gfs2_llseek, |
1081 | .read = do_sync_read, | |
1082 | .aio_read = generic_file_aio_read, | |
1083 | .write = do_sync_write, | |
56aa616a | 1084 | .aio_write = gfs2_file_aio_write, |
c97bfe43 WC |
1085 | .unlocked_ioctl = gfs2_ioctl, |
1086 | .mmap = gfs2_mmap, | |
1087 | .open = gfs2_open, | |
df3fd117 | 1088 | .release = gfs2_release, |
c97bfe43 WC |
1089 | .fsync = gfs2_fsync, |
1090 | .splice_read = generic_file_splice_read, | |
1091 | .splice_write = generic_file_splice_write, | |
f057f6cd | 1092 | .setlease = generic_setlease, |
2fe17c10 | 1093 | .fallocate = gfs2_fallocate, |
c97bfe43 WC |
1094 | }; |
1095 | ||
10d21988 | 1096 | const struct file_operations gfs2_dir_fops_nolock = { |
d81a8ef5 | 1097 | .iterate = gfs2_readdir, |
c97bfe43 WC |
1098 | .unlocked_ioctl = gfs2_ioctl, |
1099 | .open = gfs2_open, | |
df3fd117 | 1100 | .release = gfs2_release, |
c97bfe43 | 1101 | .fsync = gfs2_fsync, |
6038f373 | 1102 | .llseek = default_llseek, |
c97bfe43 WC |
1103 | }; |
1104 |