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 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
15 | #include <linux/posix_acl.h> | |
16 | #include <linux/sort.h> | |
5c676f6d | 17 | #include <linux/gfs2_ondisk.h> |
71b86f56 | 18 | #include <linux/crc32.h> |
7d308590 | 19 | #include <linux/lm_interface.h> |
fcb47e0b | 20 | #include <linux/security.h> |
b3b94faa DT |
21 | |
22 | #include "gfs2.h" | |
5c676f6d | 23 | #include "incore.h" |
b3b94faa DT |
24 | #include "acl.h" |
25 | #include "bmap.h" | |
26 | #include "dir.h" | |
27 | #include "eattr.h" | |
28 | #include "glock.h" | |
29 | #include "glops.h" | |
30 | #include "inode.h" | |
31 | #include "log.h" | |
32 | #include "meta_io.h" | |
33 | #include "ops_address.h" | |
b3b94faa DT |
34 | #include "ops_inode.h" |
35 | #include "quota.h" | |
36 | #include "rgrp.h" | |
37 | #include "trans.h" | |
5c676f6d | 38 | #include "util.h" |
b3b94faa | 39 | |
bb8d8a6f SW |
40 | struct gfs2_inum_range_host { |
41 | u64 ir_start; | |
42 | u64 ir_length; | |
43 | }; | |
44 | ||
feaa7bba SW |
45 | static int iget_test(struct inode *inode, void *opaque) |
46 | { | |
47 | struct gfs2_inode *ip = GFS2_I(inode); | |
dbb7cae2 | 48 | u64 *no_addr = opaque; |
feaa7bba | 49 | |
dbb7cae2 | 50 | if (ip->i_no_addr == *no_addr && |
1be38679 | 51 | inode->i_private != NULL) |
feaa7bba | 52 | return 1; |
b3b94faa | 53 | |
feaa7bba SW |
54 | return 0; |
55 | } | |
56 | ||
57 | static int iget_set(struct inode *inode, void *opaque) | |
b3b94faa | 58 | { |
feaa7bba | 59 | struct gfs2_inode *ip = GFS2_I(inode); |
dbb7cae2 | 60 | u64 *no_addr = opaque; |
b3b94faa | 61 | |
dbb7cae2 SW |
62 | inode->i_ino = (unsigned long)*no_addr; |
63 | ip->i_no_addr = *no_addr; | |
feaa7bba SW |
64 | return 0; |
65 | } | |
b3b94faa | 66 | |
dbb7cae2 | 67 | struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr) |
feaa7bba | 68 | { |
dbb7cae2 SW |
69 | unsigned long hash = (unsigned long)no_addr; |
70 | return ilookup5(sb, hash, iget_test, &no_addr); | |
feaa7bba | 71 | } |
b3b94faa | 72 | |
dbb7cae2 | 73 | static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr) |
feaa7bba | 74 | { |
dbb7cae2 SW |
75 | unsigned long hash = (unsigned long)no_addr; |
76 | return iget5_locked(sb, hash, iget_test, iget_set, &no_addr); | |
b3b94faa DT |
77 | } |
78 | ||
7a9f53b3 BM |
79 | struct gfs2_skip_data { |
80 | u64 no_addr; | |
81 | int skipped; | |
82 | }; | |
83 | ||
84 | static int iget_skip_test(struct inode *inode, void *opaque) | |
85 | { | |
86 | struct gfs2_inode *ip = GFS2_I(inode); | |
87 | struct gfs2_skip_data *data = opaque; | |
88 | ||
89 | if (ip->i_no_addr == data->no_addr && inode->i_private != NULL){ | |
90 | if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)){ | |
91 | data->skipped = 1; | |
92 | return 0; | |
93 | } | |
94 | return 1; | |
95 | } | |
96 | return 0; | |
97 | } | |
98 | ||
99 | static int iget_skip_set(struct inode *inode, void *opaque) | |
100 | { | |
101 | struct gfs2_inode *ip = GFS2_I(inode); | |
102 | struct gfs2_skip_data *data = opaque; | |
103 | ||
104 | if (data->skipped) | |
105 | return 1; | |
106 | inode->i_ino = (unsigned long)(data->no_addr); | |
107 | ip->i_no_addr = data->no_addr; | |
108 | return 0; | |
109 | } | |
110 | ||
111 | static struct inode *gfs2_iget_skip(struct super_block *sb, | |
112 | u64 no_addr) | |
113 | { | |
114 | struct gfs2_skip_data data; | |
115 | unsigned long hash = (unsigned long)no_addr; | |
116 | ||
117 | data.no_addr = no_addr; | |
118 | data.skipped = 0; | |
119 | return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data); | |
120 | } | |
121 | ||
35dcc52e WC |
122 | /** |
123 | * GFS2 lookup code fills in vfs inode contents based on info obtained | |
124 | * from directory entry inside gfs2_inode_lookup(). This has caused issues | |
125 | * with NFS code path since its get_dentry routine doesn't have the relevant | |
126 | * directory entry when gfs2_inode_lookup() is invoked. Part of the code | |
127 | * segment inside gfs2_inode_lookup code needs to get moved around. | |
128 | * | |
129 | * Clean up I_LOCK and I_NEW as well. | |
130 | **/ | |
131 | ||
132 | void gfs2_set_iop(struct inode *inode) | |
133 | { | |
134 | umode_t mode = inode->i_mode; | |
135 | ||
136 | if (S_ISREG(mode)) { | |
137 | inode->i_op = &gfs2_file_iops; | |
138 | inode->i_fop = &gfs2_file_fops; | |
35dcc52e WC |
139 | } else if (S_ISDIR(mode)) { |
140 | inode->i_op = &gfs2_dir_iops; | |
141 | inode->i_fop = &gfs2_dir_fops; | |
142 | } else if (S_ISLNK(mode)) { | |
143 | inode->i_op = &gfs2_symlink_iops; | |
144 | } else { | |
145 | inode->i_op = &gfs2_dev_iops; | |
146 | } | |
147 | ||
148 | unlock_new_inode(inode); | |
149 | } | |
150 | ||
b3b94faa | 151 | /** |
feaa7bba SW |
152 | * gfs2_inode_lookup - Lookup an inode |
153 | * @sb: The super block | |
dbb7cae2 | 154 | * @no_addr: The inode number |
feaa7bba | 155 | * @type: The type of the inode |
7a9f53b3 | 156 | * @skip_freeing: set this not return an inode if it is currently being freed. |
b3b94faa | 157 | * |
feaa7bba | 158 | * Returns: A VFS inode, or an error |
b3b94faa DT |
159 | */ |
160 | ||
bb9bcf06 WC |
161 | struct inode *gfs2_inode_lookup(struct super_block *sb, |
162 | unsigned int type, | |
163 | u64 no_addr, | |
7a9f53b3 | 164 | u64 no_formal_ino, int skip_freeing) |
b3b94faa | 165 | { |
7a9f53b3 BM |
166 | struct inode *inode; |
167 | struct gfs2_inode *ip; | |
feaa7bba SW |
168 | struct gfs2_glock *io_gl; |
169 | int error; | |
b3b94faa | 170 | |
7a9f53b3 BM |
171 | if (skip_freeing) |
172 | inode = gfs2_iget_skip(sb, no_addr); | |
173 | else | |
174 | inode = gfs2_iget(sb, no_addr); | |
175 | ip = GFS2_I(inode); | |
176 | ||
26d83ded SW |
177 | if (!inode) |
178 | return ERR_PTR(-ENOBUFS); | |
179 | ||
feaa7bba SW |
180 | if (inode->i_state & I_NEW) { |
181 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
bba9dfd8 | 182 | inode->i_private = ip; |
bb9bcf06 | 183 | ip->i_no_formal_ino = no_formal_ino; |
b3b94faa | 184 | |
dbb7cae2 | 185 | error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl); |
feaa7bba SW |
186 | if (unlikely(error)) |
187 | goto fail; | |
188 | ip->i_gl->gl_object = ip; | |
b3b94faa | 189 | |
dbb7cae2 | 190 | error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl); |
feaa7bba SW |
191 | if (unlikely(error)) |
192 | goto fail_put; | |
b3b94faa | 193 | |
bfded27b | 194 | set_bit(GIF_INVALID, &ip->i_flags); |
feaa7bba SW |
195 | error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh); |
196 | if (unlikely(error)) | |
197 | goto fail_iopen; | |
d93cfa98 | 198 | ip->i_iopen_gh.gh_gl->gl_object = ip; |
b3b94faa | 199 | |
feaa7bba | 200 | gfs2_glock_put(io_gl); |
c8cdf479 | 201 | |
35dcc52e WC |
202 | if ((type == DT_UNKNOWN) && (no_formal_ino == 0)) |
203 | goto gfs2_nfsbypass; | |
204 | ||
205 | inode->i_mode = DT2IF(type); | |
206 | ||
c8cdf479 SW |
207 | /* |
208 | * We must read the inode in order to work out its type in | |
209 | * this case. Note that this doesn't happen often as we normally | |
210 | * know the type beforehand. This code path only occurs during | |
211 | * unlinked inode recovery (where it is safe to do this glock, | |
212 | * which is not true in the general case). | |
213 | */ | |
c8cdf479 SW |
214 | if (type == DT_UNKNOWN) { |
215 | struct gfs2_holder gh; | |
216 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
217 | if (unlikely(error)) | |
218 | goto fail_glock; | |
219 | /* Inode is now uptodate */ | |
c8cdf479 SW |
220 | gfs2_glock_dq_uninit(&gh); |
221 | } | |
222 | ||
35dcc52e | 223 | gfs2_set_iop(inode); |
feaa7bba | 224 | } |
b3b94faa | 225 | |
35dcc52e | 226 | gfs2_nfsbypass: |
b3b94faa | 227 | return inode; |
c8cdf479 SW |
228 | fail_glock: |
229 | gfs2_glock_dq(&ip->i_iopen_gh); | |
feaa7bba SW |
230 | fail_iopen: |
231 | gfs2_glock_put(io_gl); | |
232 | fail_put: | |
233 | ip->i_gl->gl_object = NULL; | |
234 | gfs2_glock_put(ip->i_gl); | |
235 | fail: | |
236 | iput(inode); | |
237 | return ERR_PTR(error); | |
b3b94faa DT |
238 | } |
239 | ||
af339c02 | 240 | static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) |
ea744d01 SW |
241 | { |
242 | struct gfs2_dinode_host *di = &ip->i_di; | |
243 | const struct gfs2_dinode *str = buf; | |
244 | ||
dbb7cae2 | 245 | if (ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)) { |
af339c02 SW |
246 | if (gfs2_consist_inode(ip)) |
247 | gfs2_dinode_print(ip); | |
248 | return -EIO; | |
249 | } | |
dbb7cae2 | 250 | ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino); |
b60623c2 | 251 | ip->i_inode.i_mode = be32_to_cpu(str->di_mode); |
e7f14f4d | 252 | ip->i_inode.i_rdev = 0; |
b60623c2 | 253 | switch (ip->i_inode.i_mode & S_IFMT) { |
e7f14f4d SW |
254 | case S_IFBLK: |
255 | case S_IFCHR: | |
256 | ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major), | |
257 | be32_to_cpu(str->di_minor)); | |
258 | break; | |
259 | }; | |
260 | ||
2933f925 SW |
261 | ip->i_inode.i_uid = be32_to_cpu(str->di_uid); |
262 | ip->i_inode.i_gid = be32_to_cpu(str->di_gid); | |
4f56110a SW |
263 | /* |
264 | * We will need to review setting the nlink count here in the | |
265 | * light of the forthcoming ro bind mount work. This is a reminder | |
266 | * to do that. | |
267 | */ | |
268 | ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink); | |
ea744d01 | 269 | di->di_size = be64_to_cpu(str->di_size); |
9e2dbdac | 270 | i_size_write(&ip->i_inode, di->di_size); |
ea744d01 | 271 | di->di_blocks = be64_to_cpu(str->di_blocks); |
9e2dbdac | 272 | gfs2_set_inode_blocks(&ip->i_inode); |
1a7b1eed | 273 | ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime); |
4bd91ba1 | 274 | ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec); |
1a7b1eed | 275 | ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime); |
4bd91ba1 | 276 | ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec); |
1a7b1eed | 277 | ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime); |
4bd91ba1 | 278 | ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec); |
ea744d01 SW |
279 | |
280 | di->di_goal_meta = be64_to_cpu(str->di_goal_meta); | |
281 | di->di_goal_data = be64_to_cpu(str->di_goal_data); | |
282 | di->di_generation = be64_to_cpu(str->di_generation); | |
283 | ||
284 | di->di_flags = be32_to_cpu(str->di_flags); | |
6b124d8d | 285 | gfs2_set_inode_flags(&ip->i_inode); |
ea744d01 SW |
286 | di->di_height = be16_to_cpu(str->di_height); |
287 | ||
288 | di->di_depth = be16_to_cpu(str->di_depth); | |
289 | di->di_entries = be32_to_cpu(str->di_entries); | |
290 | ||
291 | di->di_eattr = be64_to_cpu(str->di_eattr); | |
5561093e SW |
292 | if (S_ISREG(ip->i_inode.i_mode)) |
293 | gfs2_set_aops(&ip->i_inode); | |
294 | ||
af339c02 | 295 | return 0; |
ea744d01 SW |
296 | } |
297 | ||
b3b94faa DT |
298 | /** |
299 | * gfs2_inode_refresh - Refresh the incore copy of the dinode | |
300 | * @ip: The GFS2 inode | |
301 | * | |
302 | * Returns: errno | |
303 | */ | |
304 | ||
305 | int gfs2_inode_refresh(struct gfs2_inode *ip) | |
306 | { | |
307 | struct buffer_head *dibh; | |
308 | int error; | |
309 | ||
310 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
311 | if (error) | |
312 | return error; | |
313 | ||
feaa7bba | 314 | if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) { |
b3b94faa DT |
315 | brelse(dibh); |
316 | return -EIO; | |
317 | } | |
318 | ||
af339c02 | 319 | error = gfs2_dinode_in(ip, dibh->b_data); |
b3b94faa | 320 | brelse(dibh); |
bfded27b | 321 | clear_bit(GIF_INVALID, &ip->i_flags); |
b3b94faa | 322 | |
af339c02 | 323 | return error; |
b3b94faa DT |
324 | } |
325 | ||
feaa7bba | 326 | int gfs2_dinode_dealloc(struct gfs2_inode *ip) |
b3b94faa | 327 | { |
feaa7bba | 328 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
329 | struct gfs2_alloc *al; |
330 | struct gfs2_rgrpd *rgd; | |
331 | int error; | |
332 | ||
333 | if (ip->i_di.di_blocks != 1) { | |
334 | if (gfs2_consist_inode(ip)) | |
4cc14f0b | 335 | gfs2_dinode_print(ip); |
b3b94faa DT |
336 | return -EIO; |
337 | } | |
338 | ||
339 | al = gfs2_alloc_get(ip); | |
340 | ||
341 | error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
342 | if (error) | |
343 | goto out; | |
344 | ||
345 | error = gfs2_rindex_hold(sdp, &al->al_ri_gh); | |
346 | if (error) | |
347 | goto out_qs; | |
348 | ||
dbb7cae2 | 349 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr); |
b3b94faa DT |
350 | if (!rgd) { |
351 | gfs2_consist_inode(ip); | |
352 | error = -EIO; | |
353 | goto out_rindex_relse; | |
354 | } | |
355 | ||
356 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, | |
357 | &al->al_rgd_gh); | |
358 | if (error) | |
359 | goto out_rindex_relse; | |
360 | ||
420b9e5e | 361 | error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1); |
b3b94faa DT |
362 | if (error) |
363 | goto out_rg_gunlock; | |
364 | ||
365 | gfs2_trans_add_gl(ip->i_gl); | |
366 | ||
367 | gfs2_free_di(rgd, ip); | |
368 | ||
b3b94faa DT |
369 | gfs2_trans_end(sdp); |
370 | clear_bit(GLF_STICKY, &ip->i_gl->gl_flags); | |
371 | ||
feaa7bba | 372 | out_rg_gunlock: |
b3b94faa | 373 | gfs2_glock_dq_uninit(&al->al_rgd_gh); |
feaa7bba | 374 | out_rindex_relse: |
b3b94faa | 375 | gfs2_glock_dq_uninit(&al->al_ri_gh); |
feaa7bba | 376 | out_qs: |
b3b94faa | 377 | gfs2_quota_unhold(ip); |
36327521 | 378 | out: |
feaa7bba | 379 | gfs2_alloc_put(ip); |
b3b94faa DT |
380 | return error; |
381 | } | |
382 | ||
b3b94faa | 383 | /** |
87d21e07 | 384 | * gfs2_change_nlink - Change nlink count on inode |
b3b94faa DT |
385 | * @ip: The GFS2 inode |
386 | * @diff: The change in the nlink count required | |
387 | * | |
388 | * Returns: errno | |
389 | */ | |
87d21e07 | 390 | int gfs2_change_nlink(struct gfs2_inode *ip, int diff) |
b3b94faa DT |
391 | { |
392 | struct buffer_head *dibh; | |
cd915493 | 393 | u32 nlink; |
b3b94faa DT |
394 | int error; |
395 | ||
4f56110a SW |
396 | BUG_ON(diff != 1 && diff != -1); |
397 | nlink = ip->i_inode.i_nlink + diff; | |
b3b94faa DT |
398 | |
399 | /* If we are reducing the nlink count, but the new value ends up being | |
400 | bigger than the old one, we must have underflowed. */ | |
4f56110a | 401 | if (diff < 0 && nlink > ip->i_inode.i_nlink) { |
b3b94faa | 402 | if (gfs2_consist_inode(ip)) |
4cc14f0b | 403 | gfs2_dinode_print(ip); |
b3b94faa DT |
404 | return -EIO; |
405 | } | |
406 | ||
407 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
408 | if (error) | |
409 | return error; | |
410 | ||
4f56110a SW |
411 | if (diff > 0) |
412 | inc_nlink(&ip->i_inode); | |
413 | else | |
414 | drop_nlink(&ip->i_inode); | |
415 | ||
4bd91ba1 | 416 | ip->i_inode.i_ctime = CURRENT_TIME; |
b3b94faa | 417 | |
d4e9c4c3 | 418 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 419 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa | 420 | brelse(dibh); |
feaa7bba | 421 | mark_inode_dirty(&ip->i_inode); |
b3b94faa | 422 | |
87d21e07 | 423 | if (ip->i_inode.i_nlink == 0) |
ddee7608 | 424 | gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */ |
87d21e07 | 425 | |
5509826f WC |
426 | return error; |
427 | } | |
428 | ||
c752666c SW |
429 | struct inode *gfs2_lookup_simple(struct inode *dip, const char *name) |
430 | { | |
431 | struct qstr qstr; | |
6c93fd1e | 432 | struct inode *inode; |
71b86f56 | 433 | gfs2_str2qstr(&qstr, name); |
6c93fd1e RC |
434 | inode = gfs2_lookupi(dip, &qstr, 1, NULL); |
435 | /* gfs2_lookupi has inconsistent callers: vfs | |
436 | * related routines expect NULL for no entry found, | |
437 | * gfs2_lookup_simple callers expect ENOENT | |
438 | * and do not check for NULL. | |
439 | */ | |
440 | if (inode == NULL) | |
441 | return ERR_PTR(-ENOENT); | |
442 | else | |
443 | return inode; | |
c752666c SW |
444 | } |
445 | ||
446 | ||
b3b94faa DT |
447 | /** |
448 | * gfs2_lookupi - Look up a filename in a directory and return its inode | |
449 | * @d_gh: An initialized holder for the directory glock | |
450 | * @name: The name of the inode to look for | |
451 | * @is_root: If 1, ignore the caller's permissions | |
452 | * @i_gh: An uninitialized holder for the new inode glock | |
453 | * | |
d7c103d0 SW |
454 | * This can be called via the VFS filldir function when NFS is doing |
455 | * a readdirplus and the inode which its intending to stat isn't | |
456 | * already in cache. In this case we must not take the directory glock | |
457 | * again, since the readdir call will have already taken that lock. | |
b3b94faa DT |
458 | * |
459 | * Returns: errno | |
460 | */ | |
461 | ||
feaa7bba SW |
462 | struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name, |
463 | int is_root, struct nameidata *nd) | |
b3b94faa | 464 | { |
c9fd4307 | 465 | struct super_block *sb = dir->i_sb; |
feaa7bba | 466 | struct gfs2_inode *dip = GFS2_I(dir); |
b3b94faa | 467 | struct gfs2_holder d_gh; |
037bcbb7 | 468 | int error = 0; |
c752666c | 469 | struct inode *inode = NULL; |
d7c103d0 | 470 | int unlock = 0; |
b3b94faa DT |
471 | |
472 | if (!name->len || name->len > GFS2_FNAMESIZE) | |
c752666c | 473 | return ERR_PTR(-ENAMETOOLONG); |
b3b94faa | 474 | |
c752666c SW |
475 | if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) || |
476 | (name->len == 2 && memcmp(name->name, "..", 2) == 0 && | |
477 | dir == sb->s_root->d_inode)) { | |
320dd101 SW |
478 | igrab(dir); |
479 | return dir; | |
b3b94faa DT |
480 | } |
481 | ||
d7c103d0 SW |
482 | if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) { |
483 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh); | |
484 | if (error) | |
485 | return ERR_PTR(error); | |
486 | unlock = 1; | |
487 | } | |
b3b94faa DT |
488 | |
489 | if (!is_root) { | |
faf450ef | 490 | error = permission(dir, MAY_EXEC, NULL); |
b3b94faa DT |
491 | if (error) |
492 | goto out; | |
493 | } | |
494 | ||
dbb7cae2 SW |
495 | inode = gfs2_dir_search(dir, name); |
496 | if (IS_ERR(inode)) | |
497 | error = PTR_ERR(inode); | |
7359a19c | 498 | out: |
d7c103d0 SW |
499 | if (unlock) |
500 | gfs2_glock_dq_uninit(&d_gh); | |
c752666c SW |
501 | if (error == -ENOENT) |
502 | return NULL; | |
d7c103d0 | 503 | return inode ? inode : ERR_PTR(error); |
b3b94faa DT |
504 | } |
505 | ||
bb8d8a6f SW |
506 | static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf) |
507 | { | |
508 | const struct gfs2_inum_range *str = buf; | |
509 | ||
510 | ir->ir_start = be64_to_cpu(str->ir_start); | |
511 | ir->ir_length = be64_to_cpu(str->ir_length); | |
512 | } | |
513 | ||
514 | static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf) | |
515 | { | |
516 | struct gfs2_inum_range *str = buf; | |
517 | ||
518 | str->ir_start = cpu_to_be64(ir->ir_start); | |
519 | str->ir_length = cpu_to_be64(ir->ir_length); | |
520 | } | |
521 | ||
cd915493 | 522 | static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino) |
b3b94faa | 523 | { |
feaa7bba | 524 | struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode); |
b3b94faa | 525 | struct buffer_head *bh; |
e6972647 | 526 | struct gfs2_inum_range_host ir; |
b3b94faa DT |
527 | int error; |
528 | ||
529 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
530 | if (error) | |
531 | return error; | |
f55ab26a | 532 | mutex_lock(&sdp->sd_inum_mutex); |
b3b94faa DT |
533 | |
534 | error = gfs2_meta_inode_buffer(ip, &bh); | |
535 | if (error) { | |
f55ab26a | 536 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
537 | gfs2_trans_end(sdp); |
538 | return error; | |
539 | } | |
540 | ||
541 | gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode)); | |
542 | ||
543 | if (ir.ir_length) { | |
544 | *formal_ino = ir.ir_start++; | |
545 | ir.ir_length--; | |
d4e9c4c3 | 546 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
547 | gfs2_inum_range_out(&ir, |
548 | bh->b_data + sizeof(struct gfs2_dinode)); | |
549 | brelse(bh); | |
f55ab26a | 550 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
551 | gfs2_trans_end(sdp); |
552 | return 0; | |
553 | } | |
554 | ||
555 | brelse(bh); | |
556 | ||
f55ab26a | 557 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
558 | gfs2_trans_end(sdp); |
559 | ||
560 | return 1; | |
561 | } | |
562 | ||
cd915493 | 563 | static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino) |
b3b94faa | 564 | { |
feaa7bba SW |
565 | struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode); |
566 | struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode); | |
b3b94faa DT |
567 | struct gfs2_holder gh; |
568 | struct buffer_head *bh; | |
e6972647 | 569 | struct gfs2_inum_range_host ir; |
b3b94faa DT |
570 | int error; |
571 | ||
572 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
573 | if (error) | |
574 | return error; | |
575 | ||
576 | error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0); | |
577 | if (error) | |
578 | goto out; | |
f55ab26a | 579 | mutex_lock(&sdp->sd_inum_mutex); |
b3b94faa DT |
580 | |
581 | error = gfs2_meta_inode_buffer(ip, &bh); | |
582 | if (error) | |
583 | goto out_end_trans; | |
907b9bce | 584 | |
b3b94faa DT |
585 | gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode)); |
586 | ||
587 | if (!ir.ir_length) { | |
588 | struct buffer_head *m_bh; | |
cd915493 | 589 | u64 x, y; |
b44b84d7 | 590 | __be64 z; |
b3b94faa DT |
591 | |
592 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); | |
593 | if (error) | |
594 | goto out_brelse; | |
595 | ||
b44b84d7 AV |
596 | z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)); |
597 | x = y = be64_to_cpu(z); | |
b3b94faa DT |
598 | ir.ir_start = x; |
599 | ir.ir_length = GFS2_INUM_QUANTUM; | |
600 | x += GFS2_INUM_QUANTUM; | |
601 | if (x < y) | |
602 | gfs2_consist_inode(m_ip); | |
b44b84d7 | 603 | z = cpu_to_be64(x); |
d4e9c4c3 | 604 | gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1); |
b44b84d7 | 605 | *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z; |
b3b94faa DT |
606 | |
607 | brelse(m_bh); | |
608 | } | |
609 | ||
610 | *formal_ino = ir.ir_start++; | |
611 | ir.ir_length--; | |
612 | ||
d4e9c4c3 | 613 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
614 | gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode)); |
615 | ||
420b9e5e | 616 | out_brelse: |
b3b94faa | 617 | brelse(bh); |
420b9e5e | 618 | out_end_trans: |
f55ab26a | 619 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa | 620 | gfs2_trans_end(sdp); |
420b9e5e | 621 | out: |
b3b94faa | 622 | gfs2_glock_dq_uninit(&gh); |
b3b94faa DT |
623 | return error; |
624 | } | |
625 | ||
cd915493 | 626 | static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum) |
b3b94faa DT |
627 | { |
628 | int error; | |
629 | ||
630 | error = pick_formal_ino_1(sdp, inum); | |
631 | if (error <= 0) | |
632 | return error; | |
633 | ||
634 | error = pick_formal_ino_2(sdp, inum); | |
635 | ||
636 | return error; | |
637 | } | |
638 | ||
639 | /** | |
640 | * create_ok - OK to create a new on-disk inode here? | |
641 | * @dip: Directory in which dinode is to be created | |
642 | * @name: Name of new dinode | |
643 | * @mode: | |
644 | * | |
645 | * Returns: errno | |
646 | */ | |
647 | ||
feaa7bba | 648 | static int create_ok(struct gfs2_inode *dip, const struct qstr *name, |
b3b94faa DT |
649 | unsigned int mode) |
650 | { | |
651 | int error; | |
652 | ||
faf450ef | 653 | error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL); |
b3b94faa DT |
654 | if (error) |
655 | return error; | |
656 | ||
657 | /* Don't create entries in an unlinked directory */ | |
4f56110a | 658 | if (!dip->i_inode.i_nlink) |
b3b94faa DT |
659 | return -EPERM; |
660 | ||
dbb7cae2 | 661 | error = gfs2_dir_check(&dip->i_inode, name, NULL); |
b3b94faa DT |
662 | switch (error) { |
663 | case -ENOENT: | |
664 | error = 0; | |
665 | break; | |
666 | case 0: | |
667 | return -EEXIST; | |
668 | default: | |
669 | return error; | |
670 | } | |
671 | ||
cd915493 | 672 | if (dip->i_di.di_entries == (u32)-1) |
b3b94faa | 673 | return -EFBIG; |
4f56110a | 674 | if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1) |
b3b94faa DT |
675 | return -EMLINK; |
676 | ||
677 | return 0; | |
678 | } | |
679 | ||
680 | static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode, | |
681 | unsigned int *uid, unsigned int *gid) | |
682 | { | |
feaa7bba | 683 | if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir && |
2933f925 | 684 | (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) { |
b3b94faa DT |
685 | if (S_ISDIR(*mode)) |
686 | *mode |= S_ISUID; | |
2933f925 | 687 | else if (dip->i_inode.i_uid != current->fsuid) |
b3b94faa | 688 | *mode &= ~07111; |
2933f925 | 689 | *uid = dip->i_inode.i_uid; |
b3b94faa DT |
690 | } else |
691 | *uid = current->fsuid; | |
692 | ||
b60623c2 | 693 | if (dip->i_inode.i_mode & S_ISGID) { |
b3b94faa DT |
694 | if (S_ISDIR(*mode)) |
695 | *mode |= S_ISGID; | |
2933f925 | 696 | *gid = dip->i_inode.i_gid; |
b3b94faa DT |
697 | } else |
698 | *gid = current->fsgid; | |
699 | } | |
700 | ||
dbb7cae2 | 701 | static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation) |
b3b94faa | 702 | { |
feaa7bba | 703 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
704 | int error; |
705 | ||
706 | gfs2_alloc_get(dip); | |
707 | ||
708 | dip->i_alloc.al_requested = RES_DINODE; | |
709 | error = gfs2_inplace_reserve(dip); | |
710 | if (error) | |
711 | goto out; | |
712 | ||
feaa7bba | 713 | error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0); |
b3b94faa DT |
714 | if (error) |
715 | goto out_ipreserv; | |
716 | ||
dbb7cae2 | 717 | *no_addr = gfs2_alloc_di(dip, generation); |
b3b94faa DT |
718 | |
719 | gfs2_trans_end(sdp); | |
720 | ||
4340fe62 | 721 | out_ipreserv: |
b3b94faa | 722 | gfs2_inplace_release(dip); |
4340fe62 | 723 | out: |
b3b94faa | 724 | gfs2_alloc_put(dip); |
b3b94faa DT |
725 | return error; |
726 | } | |
727 | ||
728 | /** | |
729 | * init_dinode - Fill in a new dinode structure | |
730 | * @dip: the directory this inode is being created in | |
731 | * @gl: The glock covering the new inode | |
732 | * @inum: the inode number | |
733 | * @mode: the file permissions | |
734 | * @uid: | |
735 | * @gid: | |
736 | * | |
737 | */ | |
738 | ||
739 | static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, | |
629a21e7 | 740 | const struct gfs2_inum_host *inum, unsigned int mode, |
4340fe62 | 741 | unsigned int uid, unsigned int gid, |
e9bd2b3b | 742 | const u64 *generation, dev_t dev, struct buffer_head **bhp) |
b3b94faa | 743 | { |
feaa7bba | 744 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b96ca4fa | 745 | struct gfs2_dinode *di; |
b3b94faa | 746 | struct buffer_head *dibh; |
4bd91ba1 | 747 | struct timespec tv = CURRENT_TIME; |
b3b94faa DT |
748 | |
749 | dibh = gfs2_meta_new(gl, inum->no_addr); | |
d4e9c4c3 | 750 | gfs2_trans_add_bh(gl, dibh, 1); |
b3b94faa DT |
751 | gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI); |
752 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); | |
b96ca4fa SW |
753 | di = (struct gfs2_dinode *)dibh->b_data; |
754 | ||
2442a098 SW |
755 | di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino); |
756 | di->di_num.no_addr = cpu_to_be64(inum->no_addr); | |
b96ca4fa SW |
757 | di->di_mode = cpu_to_be32(mode); |
758 | di->di_uid = cpu_to_be32(uid); | |
759 | di->di_gid = cpu_to_be32(gid); | |
294caaa3 SW |
760 | di->di_nlink = 0; |
761 | di->di_size = 0; | |
b96ca4fa | 762 | di->di_blocks = cpu_to_be64(1); |
4bd91ba1 | 763 | di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec); |
e7f14f4d SW |
764 | di->di_major = cpu_to_be32(MAJOR(dev)); |
765 | di->di_minor = cpu_to_be32(MINOR(dev)); | |
b96ca4fa | 766 | di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr); |
4340fe62 | 767 | di->di_generation = cpu_to_be64(*generation); |
294caaa3 | 768 | di->di_flags = 0; |
b3b94faa DT |
769 | |
770 | if (S_ISREG(mode)) { | |
771 | if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) || | |
772 | gfs2_tune_get(sdp, gt_new_files_jdata)) | |
b96ca4fa | 773 | di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA); |
b3b94faa DT |
774 | if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) || |
775 | gfs2_tune_get(sdp, gt_new_files_directio)) | |
b96ca4fa | 776 | di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO); |
b3b94faa | 777 | } else if (S_ISDIR(mode)) { |
568f4c96 SW |
778 | di->di_flags |= cpu_to_be32(dip->i_di.di_flags & |
779 | GFS2_DIF_INHERIT_DIRECTIO); | |
780 | di->di_flags |= cpu_to_be32(dip->i_di.di_flags & | |
781 | GFS2_DIF_INHERIT_JDATA); | |
b3b94faa DT |
782 | } |
783 | ||
b96ca4fa | 784 | di->__pad1 = 0; |
a9583c79 | 785 | di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0); |
294caaa3 | 786 | di->di_height = 0; |
b96ca4fa SW |
787 | di->__pad2 = 0; |
788 | di->__pad3 = 0; | |
294caaa3 SW |
789 | di->di_depth = 0; |
790 | di->di_entries = 0; | |
b96ca4fa | 791 | memset(&di->__pad4, 0, sizeof(di->__pad4)); |
294caaa3 | 792 | di->di_eattr = 0; |
4bd91ba1 SW |
793 | di->di_atime_nsec = cpu_to_be32(tv.tv_nsec); |
794 | di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec); | |
795 | di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec); | |
b96ca4fa | 796 | memset(&di->di_reserved, 0, sizeof(di->di_reserved)); |
e9bd2b3b WC |
797 | |
798 | set_buffer_uptodate(dibh); | |
b96ca4fa | 799 | |
e9bd2b3b | 800 | *bhp = dibh; |
b3b94faa DT |
801 | } |
802 | ||
803 | static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, | |
629a21e7 | 804 | unsigned int mode, const struct gfs2_inum_host *inum, |
e9bd2b3b | 805 | const u64 *generation, dev_t dev, struct buffer_head **bhp) |
b3b94faa | 806 | { |
feaa7bba | 807 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
808 | unsigned int uid, gid; |
809 | int error; | |
810 | ||
811 | munge_mode_uid_gid(dip, &mode, &uid, &gid); | |
b3b94faa DT |
812 | gfs2_alloc_get(dip); |
813 | ||
814 | error = gfs2_quota_lock(dip, uid, gid); | |
815 | if (error) | |
816 | goto out; | |
817 | ||
818 | error = gfs2_quota_check(dip, uid, gid); | |
819 | if (error) | |
820 | goto out_quota; | |
821 | ||
feaa7bba | 822 | error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0); |
b3b94faa DT |
823 | if (error) |
824 | goto out_quota; | |
825 | ||
e9bd2b3b | 826 | init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp); |
b3b94faa | 827 | gfs2_quota_change(dip, +1, uid, gid); |
b3b94faa DT |
828 | gfs2_trans_end(sdp); |
829 | ||
feaa7bba | 830 | out_quota: |
b3b94faa | 831 | gfs2_quota_unlock(dip); |
feaa7bba | 832 | out: |
b3b94faa | 833 | gfs2_alloc_put(dip); |
b3b94faa DT |
834 | return error; |
835 | } | |
836 | ||
feaa7bba SW |
837 | static int link_dinode(struct gfs2_inode *dip, const struct qstr *name, |
838 | struct gfs2_inode *ip) | |
b3b94faa | 839 | { |
feaa7bba | 840 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
841 | struct gfs2_alloc *al; |
842 | int alloc_required; | |
843 | struct buffer_head *dibh; | |
844 | int error; | |
845 | ||
846 | al = gfs2_alloc_get(dip); | |
847 | ||
848 | error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
849 | if (error) | |
850 | goto fail; | |
851 | ||
feaa7bba | 852 | error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name); |
c752666c SW |
853 | if (alloc_required < 0) |
854 | goto fail; | |
b3b94faa | 855 | if (alloc_required) { |
2933f925 | 856 | error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid); |
b3b94faa DT |
857 | if (error) |
858 | goto fail_quota_locks; | |
859 | ||
860 | al->al_requested = sdp->sd_max_dirres; | |
861 | ||
862 | error = gfs2_inplace_reserve(dip); | |
863 | if (error) | |
864 | goto fail_quota_locks; | |
865 | ||
320dd101 | 866 | error = gfs2_trans_begin(sdp, sdp->sd_max_dirres + |
bb8d8a6f | 867 | al->al_rgd->rd_length + |
907b9bce | 868 | 2 * RES_DINODE + |
b3b94faa DT |
869 | RES_STATFS + RES_QUOTA, 0); |
870 | if (error) | |
871 | goto fail_ipreserv; | |
872 | } else { | |
feaa7bba | 873 | error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0); |
b3b94faa DT |
874 | if (error) |
875 | goto fail_quota_locks; | |
876 | } | |
877 | ||
dbb7cae2 | 878 | error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode)); |
b3b94faa DT |
879 | if (error) |
880 | goto fail_end_trans; | |
881 | ||
882 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
883 | if (error) | |
884 | goto fail_end_trans; | |
4f56110a | 885 | ip->i_inode.i_nlink = 1; |
d4e9c4c3 | 886 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 887 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa | 888 | brelse(dibh); |
b3b94faa DT |
889 | return 0; |
890 | ||
320dd101 | 891 | fail_end_trans: |
b3b94faa DT |
892 | gfs2_trans_end(sdp); |
893 | ||
320dd101 | 894 | fail_ipreserv: |
b3b94faa DT |
895 | if (dip->i_alloc.al_rgd) |
896 | gfs2_inplace_release(dip); | |
897 | ||
320dd101 | 898 | fail_quota_locks: |
b3b94faa DT |
899 | gfs2_quota_unlock(dip); |
900 | ||
320dd101 | 901 | fail: |
b3b94faa | 902 | gfs2_alloc_put(dip); |
b3b94faa DT |
903 | return error; |
904 | } | |
905 | ||
fcb47e0b RH |
906 | static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip) |
907 | { | |
908 | int err; | |
909 | size_t len; | |
910 | void *value; | |
911 | char *name; | |
912 | struct gfs2_ea_request er; | |
913 | ||
914 | err = security_inode_init_security(&ip->i_inode, &dip->i_inode, | |
915 | &name, &value, &len); | |
916 | ||
917 | if (err) { | |
918 | if (err == -EOPNOTSUPP) | |
919 | return 0; | |
920 | return err; | |
921 | } | |
922 | ||
923 | memset(&er, 0, sizeof(struct gfs2_ea_request)); | |
924 | ||
925 | er.er_type = GFS2_EATYPE_SECURITY; | |
926 | er.er_name = name; | |
927 | er.er_data = value; | |
928 | er.er_name_len = strlen(name); | |
929 | er.er_data_len = len; | |
930 | ||
931 | err = gfs2_ea_set_i(ip, &er); | |
932 | ||
933 | kfree(value); | |
934 | kfree(name); | |
935 | ||
936 | return err; | |
937 | } | |
938 | ||
b3b94faa DT |
939 | /** |
940 | * gfs2_createi - Create a new inode | |
941 | * @ghs: An array of two holders | |
942 | * @name: The name of the new file | |
943 | * @mode: the permissions on the new inode | |
944 | * | |
945 | * @ghs[0] is an initialized holder for the directory | |
946 | * @ghs[1] is the holder for the inode lock | |
947 | * | |
7359a19c | 948 | * If the return value is not NULL, the glocks on both the directory and the new |
b3b94faa DT |
949 | * file are held. A transaction has been started and an inplace reservation |
950 | * is held, as well. | |
951 | * | |
7359a19c | 952 | * Returns: An inode |
b3b94faa DT |
953 | */ |
954 | ||
feaa7bba | 955 | struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, |
e7f14f4d | 956 | unsigned int mode, dev_t dev) |
b3b94faa | 957 | { |
e1cc8603 | 958 | struct inode *inode = NULL; |
5c676f6d | 959 | struct gfs2_inode *dip = ghs->gh_gl->gl_object; |
feaa7bba SW |
960 | struct inode *dir = &dip->i_inode; |
961 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); | |
dbb7cae2 | 962 | struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 }; |
b3b94faa | 963 | int error; |
4340fe62 | 964 | u64 generation; |
f91a0d3e | 965 | struct buffer_head *bh = NULL; |
b3b94faa DT |
966 | |
967 | if (!name->len || name->len > GFS2_FNAMESIZE) | |
7359a19c | 968 | return ERR_PTR(-ENAMETOOLONG); |
b3b94faa | 969 | |
b3b94faa DT |
970 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs); |
971 | error = gfs2_glock_nq(ghs); | |
972 | if (error) | |
973 | goto fail; | |
974 | ||
975 | error = create_ok(dip, name, mode); | |
976 | if (error) | |
977 | goto fail_gunlock; | |
978 | ||
feaa7bba | 979 | error = pick_formal_ino(sdp, &inum.no_formal_ino); |
b3b94faa DT |
980 | if (error) |
981 | goto fail_gunlock; | |
982 | ||
dbb7cae2 | 983 | error = alloc_dinode(dip, &inum.no_addr, &generation); |
b3b94faa DT |
984 | if (error) |
985 | goto fail_gunlock; | |
986 | ||
28626e20 SW |
987 | error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops, |
988 | LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1); | |
989 | if (error) | |
990 | goto fail_gunlock; | |
b3b94faa | 991 | |
e9bd2b3b | 992 | error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh); |
b3b94faa DT |
993 | if (error) |
994 | goto fail_gunlock2; | |
995 | ||
bb9bcf06 WC |
996 | inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), |
997 | inum.no_addr, | |
7a9f53b3 | 998 | inum.no_formal_ino, 0); |
feaa7bba | 999 | if (IS_ERR(inode)) |
b3b94faa DT |
1000 | goto fail_gunlock2; |
1001 | ||
feaa7bba | 1002 | error = gfs2_inode_refresh(GFS2_I(inode)); |
b3b94faa | 1003 | if (error) |
e1cc8603 | 1004 | goto fail_gunlock2; |
b3b94faa | 1005 | |
feaa7bba | 1006 | error = gfs2_acl_create(dip, GFS2_I(inode)); |
b3b94faa | 1007 | if (error) |
e1cc8603 | 1008 | goto fail_gunlock2; |
b3b94faa | 1009 | |
fcb47e0b RH |
1010 | error = gfs2_security_init(dip, GFS2_I(inode)); |
1011 | if (error) | |
e1cc8603 | 1012 | goto fail_gunlock2; |
fcb47e0b | 1013 | |
feaa7bba | 1014 | error = link_dinode(dip, name, GFS2_I(inode)); |
b3b94faa | 1015 | if (error) |
e1cc8603 | 1016 | goto fail_gunlock2; |
b3b94faa | 1017 | |
f91a0d3e SW |
1018 | if (bh) |
1019 | brelse(bh); | |
7359a19c SW |
1020 | if (!inode) |
1021 | return ERR_PTR(-ENOMEM); | |
1022 | return inode; | |
b3b94faa | 1023 | |
320dd101 | 1024 | fail_gunlock2: |
b3b94faa | 1025 | gfs2_glock_dq_uninit(ghs + 1); |
e1cc8603 SW |
1026 | if (inode) |
1027 | iput(inode); | |
320dd101 | 1028 | fail_gunlock: |
b3b94faa | 1029 | gfs2_glock_dq(ghs); |
320dd101 | 1030 | fail: |
f91a0d3e SW |
1031 | if (bh) |
1032 | brelse(bh); | |
7359a19c | 1033 | return ERR_PTR(error); |
b3b94faa DT |
1034 | } |
1035 | ||
b3b94faa DT |
1036 | /** |
1037 | * gfs2_rmdiri - Remove a directory | |
1038 | * @dip: The parent directory of the directory to be removed | |
1039 | * @name: The name of the directory to be removed | |
1040 | * @ip: The GFS2 inode of the directory to be removed | |
1041 | * | |
1042 | * Assumes Glocks on dip and ip are held | |
1043 | * | |
1044 | * Returns: errno | |
1045 | */ | |
1046 | ||
feaa7bba SW |
1047 | int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name, |
1048 | struct gfs2_inode *ip) | |
b3b94faa | 1049 | { |
b3b94faa DT |
1050 | struct qstr dotname; |
1051 | int error; | |
1052 | ||
1053 | if (ip->i_di.di_entries != 2) { | |
1054 | if (gfs2_consist_inode(ip)) | |
4cc14f0b | 1055 | gfs2_dinode_print(ip); |
b3b94faa DT |
1056 | return -EIO; |
1057 | } | |
1058 | ||
1059 | error = gfs2_dir_del(dip, name); | |
1060 | if (error) | |
1061 | return error; | |
1062 | ||
1063 | error = gfs2_change_nlink(dip, -1); | |
1064 | if (error) | |
1065 | return error; | |
1066 | ||
71b86f56 | 1067 | gfs2_str2qstr(&dotname, "."); |
b3b94faa DT |
1068 | error = gfs2_dir_del(ip, &dotname); |
1069 | if (error) | |
1070 | return error; | |
1071 | ||
feaa7bba | 1072 | gfs2_str2qstr(&dotname, ".."); |
b3b94faa DT |
1073 | error = gfs2_dir_del(ip, &dotname); |
1074 | if (error) | |
1075 | return error; | |
1076 | ||
4f56110a SW |
1077 | /* It looks odd, but it really should be done twice */ |
1078 | error = gfs2_change_nlink(ip, -1); | |
1079 | if (error) | |
1080 | return error; | |
1081 | ||
1082 | error = gfs2_change_nlink(ip, -1); | |
b3b94faa DT |
1083 | if (error) |
1084 | return error; | |
1085 | ||
b3b94faa DT |
1086 | return error; |
1087 | } | |
1088 | ||
1089 | /* | |
1090 | * gfs2_unlink_ok - check to see that a inode is still in a directory | |
1091 | * @dip: the directory | |
1092 | * @name: the name of the file | |
1093 | * @ip: the inode | |
1094 | * | |
1095 | * Assumes that the lock on (at least) @dip is held. | |
1096 | * | |
1097 | * Returns: 0 if the parent/child relationship is correct, errno if it isn't | |
1098 | */ | |
1099 | ||
feaa7bba | 1100 | int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name, |
dbb7cae2 | 1101 | const struct gfs2_inode *ip) |
b3b94faa | 1102 | { |
b3b94faa DT |
1103 | int error; |
1104 | ||
feaa7bba | 1105 | if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode)) |
b3b94faa DT |
1106 | return -EPERM; |
1107 | ||
b60623c2 | 1108 | if ((dip->i_inode.i_mode & S_ISVTX) && |
2933f925 SW |
1109 | dip->i_inode.i_uid != current->fsuid && |
1110 | ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER)) | |
b3b94faa DT |
1111 | return -EPERM; |
1112 | ||
feaa7bba | 1113 | if (IS_APPEND(&dip->i_inode)) |
b3b94faa DT |
1114 | return -EPERM; |
1115 | ||
faf450ef | 1116 | error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL); |
b3b94faa DT |
1117 | if (error) |
1118 | return error; | |
1119 | ||
dbb7cae2 | 1120 | error = gfs2_dir_check(&dip->i_inode, name, ip); |
b3b94faa DT |
1121 | if (error) |
1122 | return error; | |
1123 | ||
b3b94faa DT |
1124 | return 0; |
1125 | } | |
1126 | ||
1127 | /* | |
1128 | * gfs2_ok_to_move - check if it's ok to move a directory to another directory | |
1129 | * @this: move this | |
1130 | * @to: to here | |
1131 | * | |
1132 | * Follow @to back to the root and make sure we don't encounter @this | |
1133 | * Assumes we already hold the rename lock. | |
1134 | * | |
1135 | * Returns: errno | |
1136 | */ | |
1137 | ||
1138 | int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to) | |
1139 | { | |
feaa7bba | 1140 | struct inode *dir = &to->i_inode; |
c9fd4307 | 1141 | struct super_block *sb = dir->i_sb; |
7359a19c | 1142 | struct inode *tmp; |
b3b94faa DT |
1143 | struct qstr dotdot; |
1144 | int error = 0; | |
1145 | ||
71b86f56 | 1146 | gfs2_str2qstr(&dotdot, ".."); |
b3b94faa | 1147 | |
7359a19c | 1148 | igrab(dir); |
b3b94faa DT |
1149 | |
1150 | for (;;) { | |
feaa7bba | 1151 | if (dir == &this->i_inode) { |
b3b94faa DT |
1152 | error = -EINVAL; |
1153 | break; | |
1154 | } | |
c9fd4307 | 1155 | if (dir == sb->s_root->d_inode) { |
b3b94faa DT |
1156 | error = 0; |
1157 | break; | |
1158 | } | |
1159 | ||
c752666c SW |
1160 | tmp = gfs2_lookupi(dir, &dotdot, 1, NULL); |
1161 | if (IS_ERR(tmp)) { | |
1162 | error = PTR_ERR(tmp); | |
b3b94faa | 1163 | break; |
c752666c | 1164 | } |
b3b94faa | 1165 | |
7359a19c SW |
1166 | iput(dir); |
1167 | dir = tmp; | |
b3b94faa DT |
1168 | } |
1169 | ||
7359a19c | 1170 | iput(dir); |
b3b94faa DT |
1171 | |
1172 | return error; | |
1173 | } | |
1174 | ||
1175 | /** | |
1176 | * gfs2_readlinki - return the contents of a symlink | |
1177 | * @ip: the symlink's inode | |
1178 | * @buf: a pointer to the buffer to be filled | |
1179 | * @len: a pointer to the length of @buf | |
1180 | * | |
1181 | * If @buf is too small, a piece of memory is kmalloc()ed and needs | |
1182 | * to be freed by the caller. | |
1183 | * | |
1184 | * Returns: errno | |
1185 | */ | |
1186 | ||
1187 | int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len) | |
1188 | { | |
1189 | struct gfs2_holder i_gh; | |
1190 | struct buffer_head *dibh; | |
1191 | unsigned int x; | |
1192 | int error; | |
1193 | ||
1194 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh); | |
1195 | error = gfs2_glock_nq_atime(&i_gh); | |
1196 | if (error) { | |
1197 | gfs2_holder_uninit(&i_gh); | |
1198 | return error; | |
1199 | } | |
1200 | ||
1201 | if (!ip->i_di.di_size) { | |
1202 | gfs2_consist_inode(ip); | |
1203 | error = -EIO; | |
1204 | goto out; | |
1205 | } | |
1206 | ||
1207 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1208 | if (error) | |
1209 | goto out; | |
1210 | ||
1211 | x = ip->i_di.di_size + 1; | |
1212 | if (x > *len) { | |
1213 | *buf = kmalloc(x, GFP_KERNEL); | |
1214 | if (!*buf) { | |
1215 | error = -ENOMEM; | |
1216 | goto out_brelse; | |
1217 | } | |
1218 | } | |
1219 | ||
1220 | memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x); | |
1221 | *len = x; | |
1222 | ||
feaa7bba | 1223 | out_brelse: |
b3b94faa | 1224 | brelse(dibh); |
feaa7bba | 1225 | out: |
b3b94faa | 1226 | gfs2_glock_dq_uninit(&i_gh); |
b3b94faa DT |
1227 | return error; |
1228 | } | |
1229 | ||
1230 | /** | |
1231 | * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and | |
1232 | * conditionally update the inode's atime | |
1233 | * @gh: the holder to acquire | |
1234 | * | |
1235 | * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap | |
1236 | * Update if the difference between the current time and the inode's current | |
1237 | * atime is greater than an interval specified at mount. | |
1238 | * | |
1239 | * Returns: errno | |
1240 | */ | |
1241 | ||
1242 | int gfs2_glock_nq_atime(struct gfs2_holder *gh) | |
1243 | { | |
1244 | struct gfs2_glock *gl = gh->gh_gl; | |
1245 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
5c676f6d | 1246 | struct gfs2_inode *ip = gl->gl_object; |
4bd91ba1 | 1247 | s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum); |
b3b94faa DT |
1248 | unsigned int state; |
1249 | int flags; | |
1250 | int error; | |
4bd91ba1 | 1251 | struct timespec tv = CURRENT_TIME; |
b3b94faa DT |
1252 | |
1253 | if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) || | |
1254 | gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) || | |
1255 | gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops)) | |
1256 | return -EINVAL; | |
1257 | ||
1258 | state = gh->gh_state; | |
1259 | flags = gh->gh_flags; | |
1260 | ||
1261 | error = gfs2_glock_nq(gh); | |
1262 | if (error) | |
1263 | return error; | |
1264 | ||
1265 | if (test_bit(SDF_NOATIME, &sdp->sd_flags) || | |
1266 | (sdp->sd_vfs->s_flags & MS_RDONLY)) | |
1267 | return 0; | |
1268 | ||
4bd91ba1 | 1269 | if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) { |
b3b94faa | 1270 | gfs2_glock_dq(gh); |
fd88de56 SW |
1271 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY, |
1272 | gh); | |
b3b94faa DT |
1273 | error = gfs2_glock_nq(gh); |
1274 | if (error) | |
1275 | return error; | |
1276 | ||
1277 | /* Verify that atime hasn't been updated while we were | |
1278 | trying to get exclusive lock. */ | |
1279 | ||
4bd91ba1 SW |
1280 | tv = CURRENT_TIME; |
1281 | if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) { | |
b3b94faa | 1282 | struct buffer_head *dibh; |
48516ced | 1283 | struct gfs2_dinode *di; |
b3b94faa DT |
1284 | |
1285 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
1286 | if (error == -EROFS) | |
1287 | return 0; | |
1288 | if (error) | |
1289 | goto fail; | |
1290 | ||
1291 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1292 | if (error) | |
1293 | goto fail_end_trans; | |
1294 | ||
4bd91ba1 | 1295 | ip->i_inode.i_atime = tv; |
b3b94faa | 1296 | |
d4e9c4c3 | 1297 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
48516ced | 1298 | di = (struct gfs2_dinode *)dibh->b_data; |
1a7b1eed | 1299 | di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec); |
4bd91ba1 | 1300 | di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec); |
b3b94faa DT |
1301 | brelse(dibh); |
1302 | ||
1303 | gfs2_trans_end(sdp); | |
1304 | } | |
1305 | ||
1306 | /* If someone else has asked for the glock, | |
1307 | unlock and let them have it. Then reacquire | |
1308 | in the original state. */ | |
1309 | if (gfs2_glock_is_blocking(gl)) { | |
1310 | gfs2_glock_dq(gh); | |
1311 | gfs2_holder_reinit(state, flags, gh); | |
1312 | return gfs2_glock_nq(gh); | |
1313 | } | |
1314 | } | |
1315 | ||
1316 | return 0; | |
1317 | ||
feaa7bba | 1318 | fail_end_trans: |
b3b94faa | 1319 | gfs2_trans_end(sdp); |
feaa7bba | 1320 | fail: |
b3b94faa | 1321 | gfs2_glock_dq(gh); |
b3b94faa DT |
1322 | return error; |
1323 | } | |
1324 | ||
b3b94faa DT |
1325 | static int |
1326 | __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |
1327 | { | |
1328 | struct buffer_head *dibh; | |
1329 | int error; | |
1330 | ||
1331 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1332 | if (!error) { | |
feaa7bba SW |
1333 | error = inode_setattr(&ip->i_inode, attr); |
1334 | gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error); | |
d4e9c4c3 | 1335 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 1336 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa DT |
1337 | brelse(dibh); |
1338 | } | |
1339 | return error; | |
1340 | } | |
1341 | ||
1342 | /** | |
1343 | * gfs2_setattr_simple - | |
1344 | * @ip: | |
1345 | * @attr: | |
1346 | * | |
1347 | * Called with a reference on the vnode. | |
1348 | * | |
1349 | * Returns: errno | |
1350 | */ | |
1351 | ||
1352 | int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |
1353 | { | |
1354 | int error; | |
1355 | ||
5c676f6d | 1356 | if (current->journal_info) |
b3b94faa DT |
1357 | return __gfs2_setattr_simple(ip, attr); |
1358 | ||
feaa7bba | 1359 | error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0); |
b3b94faa DT |
1360 | if (error) |
1361 | return error; | |
1362 | ||
1363 | error = __gfs2_setattr_simple(ip, attr); | |
feaa7bba | 1364 | gfs2_trans_end(GFS2_SB(&ip->i_inode)); |
b3b94faa DT |
1365 | return error; |
1366 | } | |
1367 | ||
bb8d8a6f SW |
1368 | void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf) |
1369 | { | |
1370 | const struct gfs2_dinode_host *di = &ip->i_di; | |
1371 | struct gfs2_dinode *str = buf; | |
1372 | ||
1373 | str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC); | |
1374 | str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI); | |
1375 | str->di_header.__pad0 = 0; | |
1376 | str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI); | |
1377 | str->di_header.__pad1 = 0; | |
1378 | str->di_num.no_addr = cpu_to_be64(ip->i_no_addr); | |
1379 | str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino); | |
1380 | str->di_mode = cpu_to_be32(ip->i_inode.i_mode); | |
1381 | str->di_uid = cpu_to_be32(ip->i_inode.i_uid); | |
1382 | str->di_gid = cpu_to_be32(ip->i_inode.i_gid); | |
1383 | str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink); | |
1384 | str->di_size = cpu_to_be64(di->di_size); | |
1385 | str->di_blocks = cpu_to_be64(di->di_blocks); | |
1386 | str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec); | |
1387 | str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec); | |
1388 | str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec); | |
1389 | ||
1390 | str->di_goal_meta = cpu_to_be64(di->di_goal_meta); | |
1391 | str->di_goal_data = cpu_to_be64(di->di_goal_data); | |
1392 | str->di_generation = cpu_to_be64(di->di_generation); | |
1393 | ||
1394 | str->di_flags = cpu_to_be32(di->di_flags); | |
1395 | str->di_height = cpu_to_be16(di->di_height); | |
1396 | str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) && | |
1397 | !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ? | |
1398 | GFS2_FORMAT_DE : 0); | |
1399 | str->di_depth = cpu_to_be16(di->di_depth); | |
1400 | str->di_entries = cpu_to_be32(di->di_entries); | |
1401 | ||
1402 | str->di_eattr = cpu_to_be64(di->di_eattr); | |
4bd91ba1 SW |
1403 | str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec); |
1404 | str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec); | |
1405 | str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec); | |
bb8d8a6f SW |
1406 | } |
1407 | ||
1408 | void gfs2_dinode_print(const struct gfs2_inode *ip) | |
1409 | { | |
1410 | const struct gfs2_dinode_host *di = &ip->i_di; | |
1411 | ||
1412 | printk(KERN_INFO " no_formal_ino = %llu\n", | |
1413 | (unsigned long long)ip->i_no_formal_ino); | |
1414 | printk(KERN_INFO " no_addr = %llu\n", | |
1415 | (unsigned long long)ip->i_no_addr); | |
1416 | printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size); | |
1417 | printk(KERN_INFO " di_blocks = %llu\n", | |
1418 | (unsigned long long)di->di_blocks); | |
1419 | printk(KERN_INFO " di_goal_meta = %llu\n", | |
1420 | (unsigned long long)di->di_goal_meta); | |
1421 | printk(KERN_INFO " di_goal_data = %llu\n", | |
1422 | (unsigned long long)di->di_goal_data); | |
1423 | printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags); | |
1424 | printk(KERN_INFO " di_height = %u\n", di->di_height); | |
1425 | printk(KERN_INFO " di_depth = %u\n", di->di_depth); | |
1426 | printk(KERN_INFO " di_entries = %u\n", di->di_entries); | |
1427 | printk(KERN_INFO " di_eattr = %llu\n", | |
1428 | (unsigned long long)di->di_eattr); | |
1429 | } | |
1430 |