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; | |
139 | inode->i_mapping->a_ops = &gfs2_file_aops; | |
140 | } else if (S_ISDIR(mode)) { | |
141 | inode->i_op = &gfs2_dir_iops; | |
142 | inode->i_fop = &gfs2_dir_fops; | |
143 | } else if (S_ISLNK(mode)) { | |
144 | inode->i_op = &gfs2_symlink_iops; | |
145 | } else { | |
146 | inode->i_op = &gfs2_dev_iops; | |
147 | } | |
148 | ||
149 | unlock_new_inode(inode); | |
150 | } | |
151 | ||
b3b94faa | 152 | /** |
feaa7bba SW |
153 | * gfs2_inode_lookup - Lookup an inode |
154 | * @sb: The super block | |
dbb7cae2 | 155 | * @no_addr: The inode number |
feaa7bba | 156 | * @type: The type of the inode |
7a9f53b3 | 157 | * @skip_freeing: set this not return an inode if it is currently being freed. |
b3b94faa | 158 | * |
feaa7bba | 159 | * Returns: A VFS inode, or an error |
b3b94faa DT |
160 | */ |
161 | ||
bb9bcf06 WC |
162 | struct inode *gfs2_inode_lookup(struct super_block *sb, |
163 | unsigned int type, | |
164 | u64 no_addr, | |
7a9f53b3 | 165 | u64 no_formal_ino, int skip_freeing) |
b3b94faa | 166 | { |
7a9f53b3 BM |
167 | struct inode *inode; |
168 | struct gfs2_inode *ip; | |
feaa7bba SW |
169 | struct gfs2_glock *io_gl; |
170 | int error; | |
b3b94faa | 171 | |
7a9f53b3 BM |
172 | if (skip_freeing) |
173 | inode = gfs2_iget_skip(sb, no_addr); | |
174 | else | |
175 | inode = gfs2_iget(sb, no_addr); | |
176 | ip = GFS2_I(inode); | |
177 | ||
26d83ded SW |
178 | if (!inode) |
179 | return ERR_PTR(-ENOBUFS); | |
180 | ||
feaa7bba SW |
181 | if (inode->i_state & I_NEW) { |
182 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
bba9dfd8 | 183 | inode->i_private = ip; |
bb9bcf06 | 184 | ip->i_no_formal_ino = no_formal_ino; |
b3b94faa | 185 | |
dbb7cae2 | 186 | error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl); |
feaa7bba SW |
187 | if (unlikely(error)) |
188 | goto fail; | |
189 | ip->i_gl->gl_object = ip; | |
b3b94faa | 190 | |
dbb7cae2 | 191 | error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl); |
feaa7bba SW |
192 | if (unlikely(error)) |
193 | goto fail_put; | |
b3b94faa | 194 | |
bfded27b | 195 | set_bit(GIF_INVALID, &ip->i_flags); |
feaa7bba SW |
196 | error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh); |
197 | if (unlikely(error)) | |
198 | goto fail_iopen; | |
d93cfa98 | 199 | ip->i_iopen_gh.gh_gl->gl_object = ip; |
b3b94faa | 200 | |
feaa7bba | 201 | gfs2_glock_put(io_gl); |
c8cdf479 | 202 | |
35dcc52e WC |
203 | if ((type == DT_UNKNOWN) && (no_formal_ino == 0)) |
204 | goto gfs2_nfsbypass; | |
205 | ||
206 | inode->i_mode = DT2IF(type); | |
207 | ||
c8cdf479 SW |
208 | /* |
209 | * We must read the inode in order to work out its type in | |
210 | * this case. Note that this doesn't happen often as we normally | |
211 | * know the type beforehand. This code path only occurs during | |
212 | * unlinked inode recovery (where it is safe to do this glock, | |
213 | * which is not true in the general case). | |
214 | */ | |
c8cdf479 SW |
215 | if (type == DT_UNKNOWN) { |
216 | struct gfs2_holder gh; | |
217 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
218 | if (unlikely(error)) | |
219 | goto fail_glock; | |
220 | /* Inode is now uptodate */ | |
c8cdf479 SW |
221 | gfs2_glock_dq_uninit(&gh); |
222 | } | |
223 | ||
35dcc52e | 224 | gfs2_set_iop(inode); |
feaa7bba | 225 | } |
b3b94faa | 226 | |
35dcc52e | 227 | gfs2_nfsbypass: |
b3b94faa | 228 | return inode; |
c8cdf479 SW |
229 | fail_glock: |
230 | gfs2_glock_dq(&ip->i_iopen_gh); | |
feaa7bba SW |
231 | fail_iopen: |
232 | gfs2_glock_put(io_gl); | |
233 | fail_put: | |
234 | ip->i_gl->gl_object = NULL; | |
235 | gfs2_glock_put(ip->i_gl); | |
236 | fail: | |
237 | iput(inode); | |
238 | return ERR_PTR(error); | |
b3b94faa DT |
239 | } |
240 | ||
af339c02 | 241 | static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) |
ea744d01 SW |
242 | { |
243 | struct gfs2_dinode_host *di = &ip->i_di; | |
244 | const struct gfs2_dinode *str = buf; | |
245 | ||
dbb7cae2 | 246 | if (ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)) { |
af339c02 SW |
247 | if (gfs2_consist_inode(ip)) |
248 | gfs2_dinode_print(ip); | |
249 | return -EIO; | |
250 | } | |
dbb7cae2 | 251 | ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino); |
b60623c2 | 252 | ip->i_inode.i_mode = be32_to_cpu(str->di_mode); |
e7f14f4d | 253 | ip->i_inode.i_rdev = 0; |
b60623c2 | 254 | switch (ip->i_inode.i_mode & S_IFMT) { |
e7f14f4d SW |
255 | case S_IFBLK: |
256 | case S_IFCHR: | |
257 | ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major), | |
258 | be32_to_cpu(str->di_minor)); | |
259 | break; | |
260 | }; | |
261 | ||
2933f925 SW |
262 | ip->i_inode.i_uid = be32_to_cpu(str->di_uid); |
263 | ip->i_inode.i_gid = be32_to_cpu(str->di_gid); | |
4f56110a SW |
264 | /* |
265 | * We will need to review setting the nlink count here in the | |
266 | * light of the forthcoming ro bind mount work. This is a reminder | |
267 | * to do that. | |
268 | */ | |
269 | ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink); | |
ea744d01 | 270 | di->di_size = be64_to_cpu(str->di_size); |
9e2dbdac | 271 | i_size_write(&ip->i_inode, di->di_size); |
ea744d01 | 272 | di->di_blocks = be64_to_cpu(str->di_blocks); |
9e2dbdac | 273 | gfs2_set_inode_blocks(&ip->i_inode); |
1a7b1eed | 274 | ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime); |
4bd91ba1 | 275 | ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec); |
1a7b1eed | 276 | ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime); |
4bd91ba1 | 277 | ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec); |
1a7b1eed | 278 | ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime); |
4bd91ba1 | 279 | ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec); |
ea744d01 SW |
280 | |
281 | di->di_goal_meta = be64_to_cpu(str->di_goal_meta); | |
282 | di->di_goal_data = be64_to_cpu(str->di_goal_data); | |
283 | di->di_generation = be64_to_cpu(str->di_generation); | |
284 | ||
285 | di->di_flags = be32_to_cpu(str->di_flags); | |
6b124d8d | 286 | gfs2_set_inode_flags(&ip->i_inode); |
ea744d01 SW |
287 | di->di_height = be16_to_cpu(str->di_height); |
288 | ||
289 | di->di_depth = be16_to_cpu(str->di_depth); | |
290 | di->di_entries = be32_to_cpu(str->di_entries); | |
291 | ||
292 | di->di_eattr = be64_to_cpu(str->di_eattr); | |
af339c02 | 293 | return 0; |
ea744d01 SW |
294 | } |
295 | ||
b3b94faa DT |
296 | /** |
297 | * gfs2_inode_refresh - Refresh the incore copy of the dinode | |
298 | * @ip: The GFS2 inode | |
299 | * | |
300 | * Returns: errno | |
301 | */ | |
302 | ||
303 | int gfs2_inode_refresh(struct gfs2_inode *ip) | |
304 | { | |
305 | struct buffer_head *dibh; | |
306 | int error; | |
307 | ||
308 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
309 | if (error) | |
310 | return error; | |
311 | ||
feaa7bba | 312 | if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) { |
b3b94faa DT |
313 | brelse(dibh); |
314 | return -EIO; | |
315 | } | |
316 | ||
af339c02 | 317 | error = gfs2_dinode_in(ip, dibh->b_data); |
b3b94faa | 318 | brelse(dibh); |
bfded27b | 319 | clear_bit(GIF_INVALID, &ip->i_flags); |
b3b94faa | 320 | |
af339c02 | 321 | return error; |
b3b94faa DT |
322 | } |
323 | ||
feaa7bba | 324 | int gfs2_dinode_dealloc(struct gfs2_inode *ip) |
b3b94faa | 325 | { |
feaa7bba | 326 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
327 | struct gfs2_alloc *al; |
328 | struct gfs2_rgrpd *rgd; | |
329 | int error; | |
330 | ||
331 | if (ip->i_di.di_blocks != 1) { | |
332 | if (gfs2_consist_inode(ip)) | |
4cc14f0b | 333 | gfs2_dinode_print(ip); |
b3b94faa DT |
334 | return -EIO; |
335 | } | |
336 | ||
337 | al = gfs2_alloc_get(ip); | |
338 | ||
339 | error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
340 | if (error) | |
341 | goto out; | |
342 | ||
343 | error = gfs2_rindex_hold(sdp, &al->al_ri_gh); | |
344 | if (error) | |
345 | goto out_qs; | |
346 | ||
dbb7cae2 | 347 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr); |
b3b94faa DT |
348 | if (!rgd) { |
349 | gfs2_consist_inode(ip); | |
350 | error = -EIO; | |
351 | goto out_rindex_relse; | |
352 | } | |
353 | ||
354 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, | |
355 | &al->al_rgd_gh); | |
356 | if (error) | |
357 | goto out_rindex_relse; | |
358 | ||
420b9e5e | 359 | error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1); |
b3b94faa DT |
360 | if (error) |
361 | goto out_rg_gunlock; | |
362 | ||
363 | gfs2_trans_add_gl(ip->i_gl); | |
364 | ||
365 | gfs2_free_di(rgd, ip); | |
366 | ||
b3b94faa DT |
367 | gfs2_trans_end(sdp); |
368 | clear_bit(GLF_STICKY, &ip->i_gl->gl_flags); | |
369 | ||
feaa7bba | 370 | out_rg_gunlock: |
b3b94faa | 371 | gfs2_glock_dq_uninit(&al->al_rgd_gh); |
feaa7bba | 372 | out_rindex_relse: |
b3b94faa | 373 | gfs2_glock_dq_uninit(&al->al_ri_gh); |
feaa7bba | 374 | out_qs: |
b3b94faa | 375 | gfs2_quota_unhold(ip); |
36327521 | 376 | out: |
feaa7bba | 377 | gfs2_alloc_put(ip); |
b3b94faa DT |
378 | return error; |
379 | } | |
380 | ||
b3b94faa | 381 | /** |
87d21e07 | 382 | * gfs2_change_nlink - Change nlink count on inode |
b3b94faa DT |
383 | * @ip: The GFS2 inode |
384 | * @diff: The change in the nlink count required | |
385 | * | |
386 | * Returns: errno | |
387 | */ | |
87d21e07 | 388 | int gfs2_change_nlink(struct gfs2_inode *ip, int diff) |
b3b94faa DT |
389 | { |
390 | struct buffer_head *dibh; | |
cd915493 | 391 | u32 nlink; |
b3b94faa DT |
392 | int error; |
393 | ||
4f56110a SW |
394 | BUG_ON(diff != 1 && diff != -1); |
395 | nlink = ip->i_inode.i_nlink + diff; | |
b3b94faa DT |
396 | |
397 | /* If we are reducing the nlink count, but the new value ends up being | |
398 | bigger than the old one, we must have underflowed. */ | |
4f56110a | 399 | if (diff < 0 && nlink > ip->i_inode.i_nlink) { |
b3b94faa | 400 | if (gfs2_consist_inode(ip)) |
4cc14f0b | 401 | gfs2_dinode_print(ip); |
b3b94faa DT |
402 | return -EIO; |
403 | } | |
404 | ||
405 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
406 | if (error) | |
407 | return error; | |
408 | ||
4f56110a SW |
409 | if (diff > 0) |
410 | inc_nlink(&ip->i_inode); | |
411 | else | |
412 | drop_nlink(&ip->i_inode); | |
413 | ||
4bd91ba1 | 414 | ip->i_inode.i_ctime = CURRENT_TIME; |
b3b94faa | 415 | |
d4e9c4c3 | 416 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 417 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa | 418 | brelse(dibh); |
feaa7bba | 419 | mark_inode_dirty(&ip->i_inode); |
b3b94faa | 420 | |
87d21e07 | 421 | if (ip->i_inode.i_nlink == 0) |
ddee7608 | 422 | gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */ |
87d21e07 | 423 | |
5509826f WC |
424 | return error; |
425 | } | |
426 | ||
c752666c SW |
427 | struct inode *gfs2_lookup_simple(struct inode *dip, const char *name) |
428 | { | |
429 | struct qstr qstr; | |
6c93fd1e | 430 | struct inode *inode; |
71b86f56 | 431 | gfs2_str2qstr(&qstr, name); |
6c93fd1e RC |
432 | inode = gfs2_lookupi(dip, &qstr, 1, NULL); |
433 | /* gfs2_lookupi has inconsistent callers: vfs | |
434 | * related routines expect NULL for no entry found, | |
435 | * gfs2_lookup_simple callers expect ENOENT | |
436 | * and do not check for NULL. | |
437 | */ | |
438 | if (inode == NULL) | |
439 | return ERR_PTR(-ENOENT); | |
440 | else | |
441 | return inode; | |
c752666c SW |
442 | } |
443 | ||
444 | ||
b3b94faa DT |
445 | /** |
446 | * gfs2_lookupi - Look up a filename in a directory and return its inode | |
447 | * @d_gh: An initialized holder for the directory glock | |
448 | * @name: The name of the inode to look for | |
449 | * @is_root: If 1, ignore the caller's permissions | |
450 | * @i_gh: An uninitialized holder for the new inode glock | |
451 | * | |
d7c103d0 SW |
452 | * This can be called via the VFS filldir function when NFS is doing |
453 | * a readdirplus and the inode which its intending to stat isn't | |
454 | * already in cache. In this case we must not take the directory glock | |
455 | * again, since the readdir call will have already taken that lock. | |
b3b94faa DT |
456 | * |
457 | * Returns: errno | |
458 | */ | |
459 | ||
feaa7bba SW |
460 | struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name, |
461 | int is_root, struct nameidata *nd) | |
b3b94faa | 462 | { |
c9fd4307 | 463 | struct super_block *sb = dir->i_sb; |
feaa7bba | 464 | struct gfs2_inode *dip = GFS2_I(dir); |
b3b94faa | 465 | struct gfs2_holder d_gh; |
037bcbb7 | 466 | int error = 0; |
c752666c | 467 | struct inode *inode = NULL; |
d7c103d0 | 468 | int unlock = 0; |
b3b94faa DT |
469 | |
470 | if (!name->len || name->len > GFS2_FNAMESIZE) | |
c752666c | 471 | return ERR_PTR(-ENAMETOOLONG); |
b3b94faa | 472 | |
c752666c SW |
473 | if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) || |
474 | (name->len == 2 && memcmp(name->name, "..", 2) == 0 && | |
475 | dir == sb->s_root->d_inode)) { | |
320dd101 SW |
476 | igrab(dir); |
477 | return dir; | |
b3b94faa DT |
478 | } |
479 | ||
d7c103d0 SW |
480 | if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) { |
481 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh); | |
482 | if (error) | |
483 | return ERR_PTR(error); | |
484 | unlock = 1; | |
485 | } | |
b3b94faa DT |
486 | |
487 | if (!is_root) { | |
faf450ef | 488 | error = permission(dir, MAY_EXEC, NULL); |
b3b94faa DT |
489 | if (error) |
490 | goto out; | |
491 | } | |
492 | ||
dbb7cae2 SW |
493 | inode = gfs2_dir_search(dir, name); |
494 | if (IS_ERR(inode)) | |
495 | error = PTR_ERR(inode); | |
7359a19c | 496 | out: |
d7c103d0 SW |
497 | if (unlock) |
498 | gfs2_glock_dq_uninit(&d_gh); | |
c752666c SW |
499 | if (error == -ENOENT) |
500 | return NULL; | |
d7c103d0 | 501 | return inode ? inode : ERR_PTR(error); |
b3b94faa DT |
502 | } |
503 | ||
bb8d8a6f SW |
504 | static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf) |
505 | { | |
506 | const struct gfs2_inum_range *str = buf; | |
507 | ||
508 | ir->ir_start = be64_to_cpu(str->ir_start); | |
509 | ir->ir_length = be64_to_cpu(str->ir_length); | |
510 | } | |
511 | ||
512 | static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf) | |
513 | { | |
514 | struct gfs2_inum_range *str = buf; | |
515 | ||
516 | str->ir_start = cpu_to_be64(ir->ir_start); | |
517 | str->ir_length = cpu_to_be64(ir->ir_length); | |
518 | } | |
519 | ||
cd915493 | 520 | static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino) |
b3b94faa | 521 | { |
feaa7bba | 522 | struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode); |
b3b94faa | 523 | struct buffer_head *bh; |
e6972647 | 524 | struct gfs2_inum_range_host ir; |
b3b94faa DT |
525 | int error; |
526 | ||
527 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
528 | if (error) | |
529 | return error; | |
f55ab26a | 530 | mutex_lock(&sdp->sd_inum_mutex); |
b3b94faa DT |
531 | |
532 | error = gfs2_meta_inode_buffer(ip, &bh); | |
533 | if (error) { | |
f55ab26a | 534 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
535 | gfs2_trans_end(sdp); |
536 | return error; | |
537 | } | |
538 | ||
539 | gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode)); | |
540 | ||
541 | if (ir.ir_length) { | |
542 | *formal_ino = ir.ir_start++; | |
543 | ir.ir_length--; | |
d4e9c4c3 | 544 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
545 | gfs2_inum_range_out(&ir, |
546 | bh->b_data + sizeof(struct gfs2_dinode)); | |
547 | brelse(bh); | |
f55ab26a | 548 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
549 | gfs2_trans_end(sdp); |
550 | return 0; | |
551 | } | |
552 | ||
553 | brelse(bh); | |
554 | ||
f55ab26a | 555 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
556 | gfs2_trans_end(sdp); |
557 | ||
558 | return 1; | |
559 | } | |
560 | ||
cd915493 | 561 | static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino) |
b3b94faa | 562 | { |
feaa7bba SW |
563 | struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode); |
564 | struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode); | |
b3b94faa DT |
565 | struct gfs2_holder gh; |
566 | struct buffer_head *bh; | |
e6972647 | 567 | struct gfs2_inum_range_host ir; |
b3b94faa DT |
568 | int error; |
569 | ||
570 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
571 | if (error) | |
572 | return error; | |
573 | ||
574 | error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0); | |
575 | if (error) | |
576 | goto out; | |
f55ab26a | 577 | mutex_lock(&sdp->sd_inum_mutex); |
b3b94faa DT |
578 | |
579 | error = gfs2_meta_inode_buffer(ip, &bh); | |
580 | if (error) | |
581 | goto out_end_trans; | |
907b9bce | 582 | |
b3b94faa DT |
583 | gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode)); |
584 | ||
585 | if (!ir.ir_length) { | |
586 | struct buffer_head *m_bh; | |
cd915493 | 587 | u64 x, y; |
b44b84d7 | 588 | __be64 z; |
b3b94faa DT |
589 | |
590 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); | |
591 | if (error) | |
592 | goto out_brelse; | |
593 | ||
b44b84d7 AV |
594 | z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)); |
595 | x = y = be64_to_cpu(z); | |
b3b94faa DT |
596 | ir.ir_start = x; |
597 | ir.ir_length = GFS2_INUM_QUANTUM; | |
598 | x += GFS2_INUM_QUANTUM; | |
599 | if (x < y) | |
600 | gfs2_consist_inode(m_ip); | |
b44b84d7 | 601 | z = cpu_to_be64(x); |
d4e9c4c3 | 602 | gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1); |
b44b84d7 | 603 | *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z; |
b3b94faa DT |
604 | |
605 | brelse(m_bh); | |
606 | } | |
607 | ||
608 | *formal_ino = ir.ir_start++; | |
609 | ir.ir_length--; | |
610 | ||
d4e9c4c3 | 611 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
612 | gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode)); |
613 | ||
420b9e5e | 614 | out_brelse: |
b3b94faa | 615 | brelse(bh); |
420b9e5e | 616 | out_end_trans: |
f55ab26a | 617 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa | 618 | gfs2_trans_end(sdp); |
420b9e5e | 619 | out: |
b3b94faa | 620 | gfs2_glock_dq_uninit(&gh); |
b3b94faa DT |
621 | return error; |
622 | } | |
623 | ||
cd915493 | 624 | static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum) |
b3b94faa DT |
625 | { |
626 | int error; | |
627 | ||
628 | error = pick_formal_ino_1(sdp, inum); | |
629 | if (error <= 0) | |
630 | return error; | |
631 | ||
632 | error = pick_formal_ino_2(sdp, inum); | |
633 | ||
634 | return error; | |
635 | } | |
636 | ||
637 | /** | |
638 | * create_ok - OK to create a new on-disk inode here? | |
639 | * @dip: Directory in which dinode is to be created | |
640 | * @name: Name of new dinode | |
641 | * @mode: | |
642 | * | |
643 | * Returns: errno | |
644 | */ | |
645 | ||
feaa7bba | 646 | static int create_ok(struct gfs2_inode *dip, const struct qstr *name, |
b3b94faa DT |
647 | unsigned int mode) |
648 | { | |
649 | int error; | |
650 | ||
faf450ef | 651 | error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL); |
b3b94faa DT |
652 | if (error) |
653 | return error; | |
654 | ||
655 | /* Don't create entries in an unlinked directory */ | |
4f56110a | 656 | if (!dip->i_inode.i_nlink) |
b3b94faa DT |
657 | return -EPERM; |
658 | ||
dbb7cae2 | 659 | error = gfs2_dir_check(&dip->i_inode, name, NULL); |
b3b94faa DT |
660 | switch (error) { |
661 | case -ENOENT: | |
662 | error = 0; | |
663 | break; | |
664 | case 0: | |
665 | return -EEXIST; | |
666 | default: | |
667 | return error; | |
668 | } | |
669 | ||
cd915493 | 670 | if (dip->i_di.di_entries == (u32)-1) |
b3b94faa | 671 | return -EFBIG; |
4f56110a | 672 | if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1) |
b3b94faa DT |
673 | return -EMLINK; |
674 | ||
675 | return 0; | |
676 | } | |
677 | ||
678 | static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode, | |
679 | unsigned int *uid, unsigned int *gid) | |
680 | { | |
feaa7bba | 681 | if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir && |
2933f925 | 682 | (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) { |
b3b94faa DT |
683 | if (S_ISDIR(*mode)) |
684 | *mode |= S_ISUID; | |
2933f925 | 685 | else if (dip->i_inode.i_uid != current->fsuid) |
b3b94faa | 686 | *mode &= ~07111; |
2933f925 | 687 | *uid = dip->i_inode.i_uid; |
b3b94faa DT |
688 | } else |
689 | *uid = current->fsuid; | |
690 | ||
b60623c2 | 691 | if (dip->i_inode.i_mode & S_ISGID) { |
b3b94faa DT |
692 | if (S_ISDIR(*mode)) |
693 | *mode |= S_ISGID; | |
2933f925 | 694 | *gid = dip->i_inode.i_gid; |
b3b94faa DT |
695 | } else |
696 | *gid = current->fsgid; | |
697 | } | |
698 | ||
dbb7cae2 | 699 | static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation) |
b3b94faa | 700 | { |
feaa7bba | 701 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
702 | int error; |
703 | ||
704 | gfs2_alloc_get(dip); | |
705 | ||
706 | dip->i_alloc.al_requested = RES_DINODE; | |
707 | error = gfs2_inplace_reserve(dip); | |
708 | if (error) | |
709 | goto out; | |
710 | ||
feaa7bba | 711 | error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0); |
b3b94faa DT |
712 | if (error) |
713 | goto out_ipreserv; | |
714 | ||
dbb7cae2 | 715 | *no_addr = gfs2_alloc_di(dip, generation); |
b3b94faa DT |
716 | |
717 | gfs2_trans_end(sdp); | |
718 | ||
4340fe62 | 719 | out_ipreserv: |
b3b94faa | 720 | gfs2_inplace_release(dip); |
4340fe62 | 721 | out: |
b3b94faa | 722 | gfs2_alloc_put(dip); |
b3b94faa DT |
723 | return error; |
724 | } | |
725 | ||
726 | /** | |
727 | * init_dinode - Fill in a new dinode structure | |
728 | * @dip: the directory this inode is being created in | |
729 | * @gl: The glock covering the new inode | |
730 | * @inum: the inode number | |
731 | * @mode: the file permissions | |
732 | * @uid: | |
733 | * @gid: | |
734 | * | |
735 | */ | |
736 | ||
737 | static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, | |
629a21e7 | 738 | const struct gfs2_inum_host *inum, unsigned int mode, |
4340fe62 | 739 | unsigned int uid, unsigned int gid, |
e9bd2b3b | 740 | const u64 *generation, dev_t dev, struct buffer_head **bhp) |
b3b94faa | 741 | { |
feaa7bba | 742 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b96ca4fa | 743 | struct gfs2_dinode *di; |
b3b94faa | 744 | struct buffer_head *dibh; |
4bd91ba1 | 745 | struct timespec tv = CURRENT_TIME; |
b3b94faa DT |
746 | |
747 | dibh = gfs2_meta_new(gl, inum->no_addr); | |
d4e9c4c3 | 748 | gfs2_trans_add_bh(gl, dibh, 1); |
b3b94faa DT |
749 | gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI); |
750 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); | |
b96ca4fa SW |
751 | di = (struct gfs2_dinode *)dibh->b_data; |
752 | ||
2442a098 SW |
753 | di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino); |
754 | di->di_num.no_addr = cpu_to_be64(inum->no_addr); | |
b96ca4fa SW |
755 | di->di_mode = cpu_to_be32(mode); |
756 | di->di_uid = cpu_to_be32(uid); | |
757 | di->di_gid = cpu_to_be32(gid); | |
294caaa3 SW |
758 | di->di_nlink = 0; |
759 | di->di_size = 0; | |
b96ca4fa | 760 | di->di_blocks = cpu_to_be64(1); |
4bd91ba1 | 761 | di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec); |
e7f14f4d SW |
762 | di->di_major = cpu_to_be32(MAJOR(dev)); |
763 | di->di_minor = cpu_to_be32(MINOR(dev)); | |
b96ca4fa | 764 | di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr); |
4340fe62 | 765 | di->di_generation = cpu_to_be64(*generation); |
294caaa3 | 766 | di->di_flags = 0; |
b3b94faa DT |
767 | |
768 | if (S_ISREG(mode)) { | |
769 | if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) || | |
770 | gfs2_tune_get(sdp, gt_new_files_jdata)) | |
b96ca4fa | 771 | di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA); |
b3b94faa DT |
772 | if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) || |
773 | gfs2_tune_get(sdp, gt_new_files_directio)) | |
b96ca4fa | 774 | di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO); |
b3b94faa | 775 | } else if (S_ISDIR(mode)) { |
568f4c96 SW |
776 | di->di_flags |= cpu_to_be32(dip->i_di.di_flags & |
777 | GFS2_DIF_INHERIT_DIRECTIO); | |
778 | di->di_flags |= cpu_to_be32(dip->i_di.di_flags & | |
779 | GFS2_DIF_INHERIT_JDATA); | |
b3b94faa DT |
780 | } |
781 | ||
b96ca4fa | 782 | di->__pad1 = 0; |
a9583c79 | 783 | di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0); |
294caaa3 | 784 | di->di_height = 0; |
b96ca4fa SW |
785 | di->__pad2 = 0; |
786 | di->__pad3 = 0; | |
294caaa3 SW |
787 | di->di_depth = 0; |
788 | di->di_entries = 0; | |
b96ca4fa | 789 | memset(&di->__pad4, 0, sizeof(di->__pad4)); |
294caaa3 | 790 | di->di_eattr = 0; |
4bd91ba1 SW |
791 | di->di_atime_nsec = cpu_to_be32(tv.tv_nsec); |
792 | di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec); | |
793 | di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec); | |
b96ca4fa | 794 | memset(&di->di_reserved, 0, sizeof(di->di_reserved)); |
e9bd2b3b WC |
795 | |
796 | set_buffer_uptodate(dibh); | |
b96ca4fa | 797 | |
e9bd2b3b | 798 | *bhp = dibh; |
b3b94faa DT |
799 | } |
800 | ||
801 | static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, | |
629a21e7 | 802 | unsigned int mode, const struct gfs2_inum_host *inum, |
e9bd2b3b | 803 | const u64 *generation, dev_t dev, struct buffer_head **bhp) |
b3b94faa | 804 | { |
feaa7bba | 805 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
806 | unsigned int uid, gid; |
807 | int error; | |
808 | ||
809 | munge_mode_uid_gid(dip, &mode, &uid, &gid); | |
b3b94faa DT |
810 | gfs2_alloc_get(dip); |
811 | ||
812 | error = gfs2_quota_lock(dip, uid, gid); | |
813 | if (error) | |
814 | goto out; | |
815 | ||
816 | error = gfs2_quota_check(dip, uid, gid); | |
817 | if (error) | |
818 | goto out_quota; | |
819 | ||
feaa7bba | 820 | error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0); |
b3b94faa DT |
821 | if (error) |
822 | goto out_quota; | |
823 | ||
e9bd2b3b | 824 | init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp); |
b3b94faa | 825 | gfs2_quota_change(dip, +1, uid, gid); |
b3b94faa DT |
826 | gfs2_trans_end(sdp); |
827 | ||
feaa7bba | 828 | out_quota: |
b3b94faa | 829 | gfs2_quota_unlock(dip); |
feaa7bba | 830 | out: |
b3b94faa | 831 | gfs2_alloc_put(dip); |
b3b94faa DT |
832 | return error; |
833 | } | |
834 | ||
feaa7bba SW |
835 | static int link_dinode(struct gfs2_inode *dip, const struct qstr *name, |
836 | struct gfs2_inode *ip) | |
b3b94faa | 837 | { |
feaa7bba | 838 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
839 | struct gfs2_alloc *al; |
840 | int alloc_required; | |
841 | struct buffer_head *dibh; | |
842 | int error; | |
843 | ||
844 | al = gfs2_alloc_get(dip); | |
845 | ||
846 | error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
847 | if (error) | |
848 | goto fail; | |
849 | ||
feaa7bba | 850 | error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name); |
c752666c SW |
851 | if (alloc_required < 0) |
852 | goto fail; | |
b3b94faa | 853 | if (alloc_required) { |
2933f925 | 854 | error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid); |
b3b94faa DT |
855 | if (error) |
856 | goto fail_quota_locks; | |
857 | ||
858 | al->al_requested = sdp->sd_max_dirres; | |
859 | ||
860 | error = gfs2_inplace_reserve(dip); | |
861 | if (error) | |
862 | goto fail_quota_locks; | |
863 | ||
320dd101 | 864 | error = gfs2_trans_begin(sdp, sdp->sd_max_dirres + |
bb8d8a6f | 865 | al->al_rgd->rd_length + |
907b9bce | 866 | 2 * RES_DINODE + |
b3b94faa DT |
867 | RES_STATFS + RES_QUOTA, 0); |
868 | if (error) | |
869 | goto fail_ipreserv; | |
870 | } else { | |
feaa7bba | 871 | error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0); |
b3b94faa DT |
872 | if (error) |
873 | goto fail_quota_locks; | |
874 | } | |
875 | ||
dbb7cae2 | 876 | error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode)); |
b3b94faa DT |
877 | if (error) |
878 | goto fail_end_trans; | |
879 | ||
880 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
881 | if (error) | |
882 | goto fail_end_trans; | |
4f56110a | 883 | ip->i_inode.i_nlink = 1; |
d4e9c4c3 | 884 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 885 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa | 886 | brelse(dibh); |
b3b94faa DT |
887 | return 0; |
888 | ||
320dd101 | 889 | fail_end_trans: |
b3b94faa DT |
890 | gfs2_trans_end(sdp); |
891 | ||
320dd101 | 892 | fail_ipreserv: |
b3b94faa DT |
893 | if (dip->i_alloc.al_rgd) |
894 | gfs2_inplace_release(dip); | |
895 | ||
320dd101 | 896 | fail_quota_locks: |
b3b94faa DT |
897 | gfs2_quota_unlock(dip); |
898 | ||
320dd101 | 899 | fail: |
b3b94faa | 900 | gfs2_alloc_put(dip); |
b3b94faa DT |
901 | return error; |
902 | } | |
903 | ||
fcb47e0b RH |
904 | static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip) |
905 | { | |
906 | int err; | |
907 | size_t len; | |
908 | void *value; | |
909 | char *name; | |
910 | struct gfs2_ea_request er; | |
911 | ||
912 | err = security_inode_init_security(&ip->i_inode, &dip->i_inode, | |
913 | &name, &value, &len); | |
914 | ||
915 | if (err) { | |
916 | if (err == -EOPNOTSUPP) | |
917 | return 0; | |
918 | return err; | |
919 | } | |
920 | ||
921 | memset(&er, 0, sizeof(struct gfs2_ea_request)); | |
922 | ||
923 | er.er_type = GFS2_EATYPE_SECURITY; | |
924 | er.er_name = name; | |
925 | er.er_data = value; | |
926 | er.er_name_len = strlen(name); | |
927 | er.er_data_len = len; | |
928 | ||
929 | err = gfs2_ea_set_i(ip, &er); | |
930 | ||
931 | kfree(value); | |
932 | kfree(name); | |
933 | ||
934 | return err; | |
935 | } | |
936 | ||
b3b94faa DT |
937 | /** |
938 | * gfs2_createi - Create a new inode | |
939 | * @ghs: An array of two holders | |
940 | * @name: The name of the new file | |
941 | * @mode: the permissions on the new inode | |
942 | * | |
943 | * @ghs[0] is an initialized holder for the directory | |
944 | * @ghs[1] is the holder for the inode lock | |
945 | * | |
7359a19c | 946 | * If the return value is not NULL, the glocks on both the directory and the new |
b3b94faa DT |
947 | * file are held. A transaction has been started and an inplace reservation |
948 | * is held, as well. | |
949 | * | |
7359a19c | 950 | * Returns: An inode |
b3b94faa DT |
951 | */ |
952 | ||
feaa7bba | 953 | struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, |
e7f14f4d | 954 | unsigned int mode, dev_t dev) |
b3b94faa | 955 | { |
e1cc8603 | 956 | struct inode *inode = NULL; |
5c676f6d | 957 | struct gfs2_inode *dip = ghs->gh_gl->gl_object; |
feaa7bba SW |
958 | struct inode *dir = &dip->i_inode; |
959 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); | |
dbb7cae2 | 960 | struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 }; |
b3b94faa | 961 | int error; |
4340fe62 | 962 | u64 generation; |
f91a0d3e | 963 | struct buffer_head *bh = NULL; |
b3b94faa DT |
964 | |
965 | if (!name->len || name->len > GFS2_FNAMESIZE) | |
7359a19c | 966 | return ERR_PTR(-ENAMETOOLONG); |
b3b94faa | 967 | |
b3b94faa DT |
968 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs); |
969 | error = gfs2_glock_nq(ghs); | |
970 | if (error) | |
971 | goto fail; | |
972 | ||
973 | error = create_ok(dip, name, mode); | |
974 | if (error) | |
975 | goto fail_gunlock; | |
976 | ||
feaa7bba | 977 | error = pick_formal_ino(sdp, &inum.no_formal_ino); |
b3b94faa DT |
978 | if (error) |
979 | goto fail_gunlock; | |
980 | ||
dbb7cae2 | 981 | error = alloc_dinode(dip, &inum.no_addr, &generation); |
b3b94faa DT |
982 | if (error) |
983 | goto fail_gunlock; | |
984 | ||
28626e20 SW |
985 | error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops, |
986 | LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1); | |
987 | if (error) | |
988 | goto fail_gunlock; | |
b3b94faa | 989 | |
e9bd2b3b | 990 | error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh); |
b3b94faa DT |
991 | if (error) |
992 | goto fail_gunlock2; | |
993 | ||
bb9bcf06 WC |
994 | inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), |
995 | inum.no_addr, | |
7a9f53b3 | 996 | inum.no_formal_ino, 0); |
feaa7bba | 997 | if (IS_ERR(inode)) |
b3b94faa DT |
998 | goto fail_gunlock2; |
999 | ||
feaa7bba | 1000 | error = gfs2_inode_refresh(GFS2_I(inode)); |
b3b94faa | 1001 | if (error) |
e1cc8603 | 1002 | goto fail_gunlock2; |
b3b94faa | 1003 | |
feaa7bba | 1004 | error = gfs2_acl_create(dip, GFS2_I(inode)); |
b3b94faa | 1005 | if (error) |
e1cc8603 | 1006 | goto fail_gunlock2; |
b3b94faa | 1007 | |
fcb47e0b RH |
1008 | error = gfs2_security_init(dip, GFS2_I(inode)); |
1009 | if (error) | |
e1cc8603 | 1010 | goto fail_gunlock2; |
fcb47e0b | 1011 | |
feaa7bba | 1012 | error = link_dinode(dip, name, GFS2_I(inode)); |
b3b94faa | 1013 | if (error) |
e1cc8603 | 1014 | goto fail_gunlock2; |
b3b94faa | 1015 | |
f91a0d3e SW |
1016 | if (bh) |
1017 | brelse(bh); | |
7359a19c SW |
1018 | if (!inode) |
1019 | return ERR_PTR(-ENOMEM); | |
1020 | return inode; | |
b3b94faa | 1021 | |
320dd101 | 1022 | fail_gunlock2: |
b3b94faa | 1023 | gfs2_glock_dq_uninit(ghs + 1); |
e1cc8603 SW |
1024 | if (inode) |
1025 | iput(inode); | |
320dd101 | 1026 | fail_gunlock: |
b3b94faa | 1027 | gfs2_glock_dq(ghs); |
320dd101 | 1028 | fail: |
f91a0d3e SW |
1029 | if (bh) |
1030 | brelse(bh); | |
7359a19c | 1031 | return ERR_PTR(error); |
b3b94faa DT |
1032 | } |
1033 | ||
b3b94faa DT |
1034 | /** |
1035 | * gfs2_rmdiri - Remove a directory | |
1036 | * @dip: The parent directory of the directory to be removed | |
1037 | * @name: The name of the directory to be removed | |
1038 | * @ip: The GFS2 inode of the directory to be removed | |
1039 | * | |
1040 | * Assumes Glocks on dip and ip are held | |
1041 | * | |
1042 | * Returns: errno | |
1043 | */ | |
1044 | ||
feaa7bba SW |
1045 | int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name, |
1046 | struct gfs2_inode *ip) | |
b3b94faa | 1047 | { |
b3b94faa DT |
1048 | struct qstr dotname; |
1049 | int error; | |
1050 | ||
1051 | if (ip->i_di.di_entries != 2) { | |
1052 | if (gfs2_consist_inode(ip)) | |
4cc14f0b | 1053 | gfs2_dinode_print(ip); |
b3b94faa DT |
1054 | return -EIO; |
1055 | } | |
1056 | ||
1057 | error = gfs2_dir_del(dip, name); | |
1058 | if (error) | |
1059 | return error; | |
1060 | ||
1061 | error = gfs2_change_nlink(dip, -1); | |
1062 | if (error) | |
1063 | return error; | |
1064 | ||
71b86f56 | 1065 | gfs2_str2qstr(&dotname, "."); |
b3b94faa DT |
1066 | error = gfs2_dir_del(ip, &dotname); |
1067 | if (error) | |
1068 | return error; | |
1069 | ||
feaa7bba | 1070 | gfs2_str2qstr(&dotname, ".."); |
b3b94faa DT |
1071 | error = gfs2_dir_del(ip, &dotname); |
1072 | if (error) | |
1073 | return error; | |
1074 | ||
4f56110a SW |
1075 | /* It looks odd, but it really should be done twice */ |
1076 | error = gfs2_change_nlink(ip, -1); | |
1077 | if (error) | |
1078 | return error; | |
1079 | ||
1080 | error = gfs2_change_nlink(ip, -1); | |
b3b94faa DT |
1081 | if (error) |
1082 | return error; | |
1083 | ||
b3b94faa DT |
1084 | return error; |
1085 | } | |
1086 | ||
1087 | /* | |
1088 | * gfs2_unlink_ok - check to see that a inode is still in a directory | |
1089 | * @dip: the directory | |
1090 | * @name: the name of the file | |
1091 | * @ip: the inode | |
1092 | * | |
1093 | * Assumes that the lock on (at least) @dip is held. | |
1094 | * | |
1095 | * Returns: 0 if the parent/child relationship is correct, errno if it isn't | |
1096 | */ | |
1097 | ||
feaa7bba | 1098 | int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name, |
dbb7cae2 | 1099 | const struct gfs2_inode *ip) |
b3b94faa | 1100 | { |
b3b94faa DT |
1101 | int error; |
1102 | ||
feaa7bba | 1103 | if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode)) |
b3b94faa DT |
1104 | return -EPERM; |
1105 | ||
b60623c2 | 1106 | if ((dip->i_inode.i_mode & S_ISVTX) && |
2933f925 SW |
1107 | dip->i_inode.i_uid != current->fsuid && |
1108 | ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER)) | |
b3b94faa DT |
1109 | return -EPERM; |
1110 | ||
feaa7bba | 1111 | if (IS_APPEND(&dip->i_inode)) |
b3b94faa DT |
1112 | return -EPERM; |
1113 | ||
faf450ef | 1114 | error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL); |
b3b94faa DT |
1115 | if (error) |
1116 | return error; | |
1117 | ||
dbb7cae2 | 1118 | error = gfs2_dir_check(&dip->i_inode, name, ip); |
b3b94faa DT |
1119 | if (error) |
1120 | return error; | |
1121 | ||
b3b94faa DT |
1122 | return 0; |
1123 | } | |
1124 | ||
1125 | /* | |
1126 | * gfs2_ok_to_move - check if it's ok to move a directory to another directory | |
1127 | * @this: move this | |
1128 | * @to: to here | |
1129 | * | |
1130 | * Follow @to back to the root and make sure we don't encounter @this | |
1131 | * Assumes we already hold the rename lock. | |
1132 | * | |
1133 | * Returns: errno | |
1134 | */ | |
1135 | ||
1136 | int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to) | |
1137 | { | |
feaa7bba | 1138 | struct inode *dir = &to->i_inode; |
c9fd4307 | 1139 | struct super_block *sb = dir->i_sb; |
7359a19c | 1140 | struct inode *tmp; |
b3b94faa DT |
1141 | struct qstr dotdot; |
1142 | int error = 0; | |
1143 | ||
71b86f56 | 1144 | gfs2_str2qstr(&dotdot, ".."); |
b3b94faa | 1145 | |
7359a19c | 1146 | igrab(dir); |
b3b94faa DT |
1147 | |
1148 | for (;;) { | |
feaa7bba | 1149 | if (dir == &this->i_inode) { |
b3b94faa DT |
1150 | error = -EINVAL; |
1151 | break; | |
1152 | } | |
c9fd4307 | 1153 | if (dir == sb->s_root->d_inode) { |
b3b94faa DT |
1154 | error = 0; |
1155 | break; | |
1156 | } | |
1157 | ||
c752666c SW |
1158 | tmp = gfs2_lookupi(dir, &dotdot, 1, NULL); |
1159 | if (IS_ERR(tmp)) { | |
1160 | error = PTR_ERR(tmp); | |
b3b94faa | 1161 | break; |
c752666c | 1162 | } |
b3b94faa | 1163 | |
7359a19c SW |
1164 | iput(dir); |
1165 | dir = tmp; | |
b3b94faa DT |
1166 | } |
1167 | ||
7359a19c | 1168 | iput(dir); |
b3b94faa DT |
1169 | |
1170 | return error; | |
1171 | } | |
1172 | ||
1173 | /** | |
1174 | * gfs2_readlinki - return the contents of a symlink | |
1175 | * @ip: the symlink's inode | |
1176 | * @buf: a pointer to the buffer to be filled | |
1177 | * @len: a pointer to the length of @buf | |
1178 | * | |
1179 | * If @buf is too small, a piece of memory is kmalloc()ed and needs | |
1180 | * to be freed by the caller. | |
1181 | * | |
1182 | * Returns: errno | |
1183 | */ | |
1184 | ||
1185 | int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len) | |
1186 | { | |
1187 | struct gfs2_holder i_gh; | |
1188 | struct buffer_head *dibh; | |
1189 | unsigned int x; | |
1190 | int error; | |
1191 | ||
1192 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh); | |
1193 | error = gfs2_glock_nq_atime(&i_gh); | |
1194 | if (error) { | |
1195 | gfs2_holder_uninit(&i_gh); | |
1196 | return error; | |
1197 | } | |
1198 | ||
1199 | if (!ip->i_di.di_size) { | |
1200 | gfs2_consist_inode(ip); | |
1201 | error = -EIO; | |
1202 | goto out; | |
1203 | } | |
1204 | ||
1205 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1206 | if (error) | |
1207 | goto out; | |
1208 | ||
1209 | x = ip->i_di.di_size + 1; | |
1210 | if (x > *len) { | |
1211 | *buf = kmalloc(x, GFP_KERNEL); | |
1212 | if (!*buf) { | |
1213 | error = -ENOMEM; | |
1214 | goto out_brelse; | |
1215 | } | |
1216 | } | |
1217 | ||
1218 | memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x); | |
1219 | *len = x; | |
1220 | ||
feaa7bba | 1221 | out_brelse: |
b3b94faa | 1222 | brelse(dibh); |
feaa7bba | 1223 | out: |
b3b94faa | 1224 | gfs2_glock_dq_uninit(&i_gh); |
b3b94faa DT |
1225 | return error; |
1226 | } | |
1227 | ||
1228 | /** | |
1229 | * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and | |
1230 | * conditionally update the inode's atime | |
1231 | * @gh: the holder to acquire | |
1232 | * | |
1233 | * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap | |
1234 | * Update if the difference between the current time and the inode's current | |
1235 | * atime is greater than an interval specified at mount. | |
1236 | * | |
1237 | * Returns: errno | |
1238 | */ | |
1239 | ||
1240 | int gfs2_glock_nq_atime(struct gfs2_holder *gh) | |
1241 | { | |
1242 | struct gfs2_glock *gl = gh->gh_gl; | |
1243 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
5c676f6d | 1244 | struct gfs2_inode *ip = gl->gl_object; |
4bd91ba1 | 1245 | s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum); |
b3b94faa DT |
1246 | unsigned int state; |
1247 | int flags; | |
1248 | int error; | |
4bd91ba1 | 1249 | struct timespec tv = CURRENT_TIME; |
b3b94faa DT |
1250 | |
1251 | if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) || | |
1252 | gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) || | |
1253 | gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops)) | |
1254 | return -EINVAL; | |
1255 | ||
1256 | state = gh->gh_state; | |
1257 | flags = gh->gh_flags; | |
1258 | ||
1259 | error = gfs2_glock_nq(gh); | |
1260 | if (error) | |
1261 | return error; | |
1262 | ||
1263 | if (test_bit(SDF_NOATIME, &sdp->sd_flags) || | |
1264 | (sdp->sd_vfs->s_flags & MS_RDONLY)) | |
1265 | return 0; | |
1266 | ||
4bd91ba1 | 1267 | if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) { |
b3b94faa | 1268 | gfs2_glock_dq(gh); |
fd88de56 SW |
1269 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY, |
1270 | gh); | |
b3b94faa DT |
1271 | error = gfs2_glock_nq(gh); |
1272 | if (error) | |
1273 | return error; | |
1274 | ||
1275 | /* Verify that atime hasn't been updated while we were | |
1276 | trying to get exclusive lock. */ | |
1277 | ||
4bd91ba1 SW |
1278 | tv = CURRENT_TIME; |
1279 | if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) { | |
b3b94faa | 1280 | struct buffer_head *dibh; |
48516ced | 1281 | struct gfs2_dinode *di; |
b3b94faa DT |
1282 | |
1283 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
1284 | if (error == -EROFS) | |
1285 | return 0; | |
1286 | if (error) | |
1287 | goto fail; | |
1288 | ||
1289 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1290 | if (error) | |
1291 | goto fail_end_trans; | |
1292 | ||
4bd91ba1 | 1293 | ip->i_inode.i_atime = tv; |
b3b94faa | 1294 | |
d4e9c4c3 | 1295 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
48516ced | 1296 | di = (struct gfs2_dinode *)dibh->b_data; |
1a7b1eed | 1297 | di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec); |
4bd91ba1 | 1298 | di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec); |
b3b94faa DT |
1299 | brelse(dibh); |
1300 | ||
1301 | gfs2_trans_end(sdp); | |
1302 | } | |
1303 | ||
1304 | /* If someone else has asked for the glock, | |
1305 | unlock and let them have it. Then reacquire | |
1306 | in the original state. */ | |
1307 | if (gfs2_glock_is_blocking(gl)) { | |
1308 | gfs2_glock_dq(gh); | |
1309 | gfs2_holder_reinit(state, flags, gh); | |
1310 | return gfs2_glock_nq(gh); | |
1311 | } | |
1312 | } | |
1313 | ||
1314 | return 0; | |
1315 | ||
feaa7bba | 1316 | fail_end_trans: |
b3b94faa | 1317 | gfs2_trans_end(sdp); |
feaa7bba | 1318 | fail: |
b3b94faa | 1319 | gfs2_glock_dq(gh); |
b3b94faa DT |
1320 | return error; |
1321 | } | |
1322 | ||
b3b94faa DT |
1323 | static int |
1324 | __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |
1325 | { | |
1326 | struct buffer_head *dibh; | |
1327 | int error; | |
1328 | ||
1329 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1330 | if (!error) { | |
feaa7bba SW |
1331 | error = inode_setattr(&ip->i_inode, attr); |
1332 | gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error); | |
d4e9c4c3 | 1333 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 1334 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa DT |
1335 | brelse(dibh); |
1336 | } | |
1337 | return error; | |
1338 | } | |
1339 | ||
1340 | /** | |
1341 | * gfs2_setattr_simple - | |
1342 | * @ip: | |
1343 | * @attr: | |
1344 | * | |
1345 | * Called with a reference on the vnode. | |
1346 | * | |
1347 | * Returns: errno | |
1348 | */ | |
1349 | ||
1350 | int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |
1351 | { | |
1352 | int error; | |
1353 | ||
5c676f6d | 1354 | if (current->journal_info) |
b3b94faa DT |
1355 | return __gfs2_setattr_simple(ip, attr); |
1356 | ||
feaa7bba | 1357 | error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0); |
b3b94faa DT |
1358 | if (error) |
1359 | return error; | |
1360 | ||
1361 | error = __gfs2_setattr_simple(ip, attr); | |
feaa7bba | 1362 | gfs2_trans_end(GFS2_SB(&ip->i_inode)); |
b3b94faa DT |
1363 | return error; |
1364 | } | |
1365 | ||
bb8d8a6f SW |
1366 | void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf) |
1367 | { | |
1368 | const struct gfs2_dinode_host *di = &ip->i_di; | |
1369 | struct gfs2_dinode *str = buf; | |
1370 | ||
1371 | str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC); | |
1372 | str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI); | |
1373 | str->di_header.__pad0 = 0; | |
1374 | str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI); | |
1375 | str->di_header.__pad1 = 0; | |
1376 | str->di_num.no_addr = cpu_to_be64(ip->i_no_addr); | |
1377 | str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino); | |
1378 | str->di_mode = cpu_to_be32(ip->i_inode.i_mode); | |
1379 | str->di_uid = cpu_to_be32(ip->i_inode.i_uid); | |
1380 | str->di_gid = cpu_to_be32(ip->i_inode.i_gid); | |
1381 | str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink); | |
1382 | str->di_size = cpu_to_be64(di->di_size); | |
1383 | str->di_blocks = cpu_to_be64(di->di_blocks); | |
1384 | str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec); | |
1385 | str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec); | |
1386 | str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec); | |
1387 | ||
1388 | str->di_goal_meta = cpu_to_be64(di->di_goal_meta); | |
1389 | str->di_goal_data = cpu_to_be64(di->di_goal_data); | |
1390 | str->di_generation = cpu_to_be64(di->di_generation); | |
1391 | ||
1392 | str->di_flags = cpu_to_be32(di->di_flags); | |
1393 | str->di_height = cpu_to_be16(di->di_height); | |
1394 | str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) && | |
1395 | !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ? | |
1396 | GFS2_FORMAT_DE : 0); | |
1397 | str->di_depth = cpu_to_be16(di->di_depth); | |
1398 | str->di_entries = cpu_to_be32(di->di_entries); | |
1399 | ||
1400 | str->di_eattr = cpu_to_be64(di->di_eattr); | |
4bd91ba1 SW |
1401 | str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec); |
1402 | str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec); | |
1403 | str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec); | |
bb8d8a6f SW |
1404 | } |
1405 | ||
1406 | void gfs2_dinode_print(const struct gfs2_inode *ip) | |
1407 | { | |
1408 | const struct gfs2_dinode_host *di = &ip->i_di; | |
1409 | ||
1410 | printk(KERN_INFO " no_formal_ino = %llu\n", | |
1411 | (unsigned long long)ip->i_no_formal_ino); | |
1412 | printk(KERN_INFO " no_addr = %llu\n", | |
1413 | (unsigned long long)ip->i_no_addr); | |
1414 | printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size); | |
1415 | printk(KERN_INFO " di_blocks = %llu\n", | |
1416 | (unsigned long long)di->di_blocks); | |
1417 | printk(KERN_INFO " di_goal_meta = %llu\n", | |
1418 | (unsigned long long)di->di_goal_meta); | |
1419 | printk(KERN_INFO " di_goal_data = %llu\n", | |
1420 | (unsigned long long)di->di_goal_data); | |
1421 | printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags); | |
1422 | printk(KERN_INFO " di_height = %u\n", di->di_height); | |
1423 | printk(KERN_INFO " di_depth = %u\n", di->di_depth); | |
1424 | printk(KERN_INFO " di_entries = %u\n", di->di_entries); | |
1425 | printk(KERN_INFO " di_eattr = %llu\n", | |
1426 | (unsigned long long)di->di_eattr); | |
1427 | } | |
1428 |