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 | { | |
c97bfe43 | 134 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
35dcc52e WC |
135 | umode_t mode = inode->i_mode; |
136 | ||
137 | if (S_ISREG(mode)) { | |
138 | inode->i_op = &gfs2_file_iops; | |
c97bfe43 WC |
139 | if (sdp->sd_args.ar_localflocks) |
140 | inode->i_fop = &gfs2_file_fops_nolock; | |
141 | else | |
142 | inode->i_fop = &gfs2_file_fops; | |
35dcc52e WC |
143 | } else if (S_ISDIR(mode)) { |
144 | inode->i_op = &gfs2_dir_iops; | |
c97bfe43 WC |
145 | if (sdp->sd_args.ar_localflocks) |
146 | inode->i_fop = &gfs2_dir_fops_nolock; | |
147 | else | |
148 | inode->i_fop = &gfs2_dir_fops; | |
35dcc52e WC |
149 | } else if (S_ISLNK(mode)) { |
150 | inode->i_op = &gfs2_symlink_iops; | |
151 | } else { | |
152 | inode->i_op = &gfs2_dev_iops; | |
153 | } | |
154 | ||
155 | unlock_new_inode(inode); | |
156 | } | |
157 | ||
b3b94faa | 158 | /** |
feaa7bba SW |
159 | * gfs2_inode_lookup - Lookup an inode |
160 | * @sb: The super block | |
dbb7cae2 | 161 | * @no_addr: The inode number |
feaa7bba | 162 | * @type: The type of the inode |
7a9f53b3 | 163 | * @skip_freeing: set this not return an inode if it is currently being freed. |
b3b94faa | 164 | * |
feaa7bba | 165 | * Returns: A VFS inode, or an error |
b3b94faa DT |
166 | */ |
167 | ||
bb9bcf06 WC |
168 | struct inode *gfs2_inode_lookup(struct super_block *sb, |
169 | unsigned int type, | |
170 | u64 no_addr, | |
7a9f53b3 | 171 | u64 no_formal_ino, int skip_freeing) |
b3b94faa | 172 | { |
7a9f53b3 BM |
173 | struct inode *inode; |
174 | struct gfs2_inode *ip; | |
feaa7bba SW |
175 | struct gfs2_glock *io_gl; |
176 | int error; | |
b3b94faa | 177 | |
7a9f53b3 BM |
178 | if (skip_freeing) |
179 | inode = gfs2_iget_skip(sb, no_addr); | |
180 | else | |
181 | inode = gfs2_iget(sb, no_addr); | |
182 | ip = GFS2_I(inode); | |
183 | ||
26d83ded SW |
184 | if (!inode) |
185 | return ERR_PTR(-ENOBUFS); | |
186 | ||
feaa7bba SW |
187 | if (inode->i_state & I_NEW) { |
188 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
bba9dfd8 | 189 | inode->i_private = ip; |
bb9bcf06 | 190 | ip->i_no_formal_ino = no_formal_ino; |
b3b94faa | 191 | |
dbb7cae2 | 192 | error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl); |
feaa7bba SW |
193 | if (unlikely(error)) |
194 | goto fail; | |
195 | ip->i_gl->gl_object = ip; | |
b3b94faa | 196 | |
dbb7cae2 | 197 | error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl); |
feaa7bba SW |
198 | if (unlikely(error)) |
199 | goto fail_put; | |
b3b94faa | 200 | |
bfded27b | 201 | set_bit(GIF_INVALID, &ip->i_flags); |
feaa7bba SW |
202 | error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh); |
203 | if (unlikely(error)) | |
204 | goto fail_iopen; | |
d93cfa98 | 205 | ip->i_iopen_gh.gh_gl->gl_object = ip; |
b3b94faa | 206 | |
feaa7bba | 207 | gfs2_glock_put(io_gl); |
c8cdf479 | 208 | |
35dcc52e WC |
209 | if ((type == DT_UNKNOWN) && (no_formal_ino == 0)) |
210 | goto gfs2_nfsbypass; | |
211 | ||
212 | inode->i_mode = DT2IF(type); | |
213 | ||
c8cdf479 SW |
214 | /* |
215 | * We must read the inode in order to work out its type in | |
216 | * this case. Note that this doesn't happen often as we normally | |
217 | * know the type beforehand. This code path only occurs during | |
218 | * unlinked inode recovery (where it is safe to do this glock, | |
219 | * which is not true in the general case). | |
220 | */ | |
c8cdf479 SW |
221 | if (type == DT_UNKNOWN) { |
222 | struct gfs2_holder gh; | |
223 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
224 | if (unlikely(error)) | |
225 | goto fail_glock; | |
226 | /* Inode is now uptodate */ | |
c8cdf479 SW |
227 | gfs2_glock_dq_uninit(&gh); |
228 | } | |
229 | ||
35dcc52e | 230 | gfs2_set_iop(inode); |
feaa7bba | 231 | } |
b3b94faa | 232 | |
35dcc52e | 233 | gfs2_nfsbypass: |
b3b94faa | 234 | return inode; |
c8cdf479 SW |
235 | fail_glock: |
236 | gfs2_glock_dq(&ip->i_iopen_gh); | |
feaa7bba SW |
237 | fail_iopen: |
238 | gfs2_glock_put(io_gl); | |
239 | fail_put: | |
240 | ip->i_gl->gl_object = NULL; | |
241 | gfs2_glock_put(ip->i_gl); | |
242 | fail: | |
69840b0d | 243 | iget_failed(inode); |
feaa7bba | 244 | return ERR_PTR(error); |
b3b94faa DT |
245 | } |
246 | ||
af339c02 | 247 | static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) |
ea744d01 SW |
248 | { |
249 | struct gfs2_dinode_host *di = &ip->i_di; | |
250 | const struct gfs2_dinode *str = buf; | |
251 | ||
dbb7cae2 | 252 | if (ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)) { |
af339c02 SW |
253 | if (gfs2_consist_inode(ip)) |
254 | gfs2_dinode_print(ip); | |
255 | return -EIO; | |
256 | } | |
dbb7cae2 | 257 | ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino); |
b60623c2 | 258 | ip->i_inode.i_mode = be32_to_cpu(str->di_mode); |
e7f14f4d | 259 | ip->i_inode.i_rdev = 0; |
b60623c2 | 260 | switch (ip->i_inode.i_mode & S_IFMT) { |
e7f14f4d SW |
261 | case S_IFBLK: |
262 | case S_IFCHR: | |
263 | ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major), | |
264 | be32_to_cpu(str->di_minor)); | |
265 | break; | |
266 | }; | |
267 | ||
2933f925 SW |
268 | ip->i_inode.i_uid = be32_to_cpu(str->di_uid); |
269 | ip->i_inode.i_gid = be32_to_cpu(str->di_gid); | |
4f56110a SW |
270 | /* |
271 | * We will need to review setting the nlink count here in the | |
272 | * light of the forthcoming ro bind mount work. This is a reminder | |
273 | * to do that. | |
274 | */ | |
275 | ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink); | |
ea744d01 | 276 | di->di_size = be64_to_cpu(str->di_size); |
9e2dbdac | 277 | i_size_write(&ip->i_inode, di->di_size); |
ea744d01 | 278 | di->di_blocks = be64_to_cpu(str->di_blocks); |
9e2dbdac | 279 | gfs2_set_inode_blocks(&ip->i_inode); |
1a7b1eed | 280 | ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime); |
4bd91ba1 | 281 | ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec); |
1a7b1eed | 282 | ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime); |
4bd91ba1 | 283 | ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec); |
1a7b1eed | 284 | ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime); |
4bd91ba1 | 285 | ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec); |
ea744d01 SW |
286 | |
287 | di->di_goal_meta = be64_to_cpu(str->di_goal_meta); | |
288 | di->di_goal_data = be64_to_cpu(str->di_goal_data); | |
289 | di->di_generation = be64_to_cpu(str->di_generation); | |
290 | ||
291 | di->di_flags = be32_to_cpu(str->di_flags); | |
6b124d8d | 292 | gfs2_set_inode_flags(&ip->i_inode); |
ea744d01 SW |
293 | di->di_height = be16_to_cpu(str->di_height); |
294 | ||
295 | di->di_depth = be16_to_cpu(str->di_depth); | |
296 | di->di_entries = be32_to_cpu(str->di_entries); | |
297 | ||
298 | di->di_eattr = be64_to_cpu(str->di_eattr); | |
5561093e SW |
299 | if (S_ISREG(ip->i_inode.i_mode)) |
300 | gfs2_set_aops(&ip->i_inode); | |
301 | ||
af339c02 | 302 | return 0; |
ea744d01 SW |
303 | } |
304 | ||
b3b94faa DT |
305 | /** |
306 | * gfs2_inode_refresh - Refresh the incore copy of the dinode | |
307 | * @ip: The GFS2 inode | |
308 | * | |
309 | * Returns: errno | |
310 | */ | |
311 | ||
312 | int gfs2_inode_refresh(struct gfs2_inode *ip) | |
313 | { | |
314 | struct buffer_head *dibh; | |
315 | int error; | |
316 | ||
317 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
318 | if (error) | |
319 | return error; | |
320 | ||
feaa7bba | 321 | if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) { |
b3b94faa DT |
322 | brelse(dibh); |
323 | return -EIO; | |
324 | } | |
325 | ||
af339c02 | 326 | error = gfs2_dinode_in(ip, dibh->b_data); |
b3b94faa | 327 | brelse(dibh); |
bfded27b | 328 | clear_bit(GIF_INVALID, &ip->i_flags); |
b3b94faa | 329 | |
af339c02 | 330 | return error; |
b3b94faa DT |
331 | } |
332 | ||
feaa7bba | 333 | int gfs2_dinode_dealloc(struct gfs2_inode *ip) |
b3b94faa | 334 | { |
feaa7bba | 335 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
336 | struct gfs2_alloc *al; |
337 | struct gfs2_rgrpd *rgd; | |
338 | int error; | |
339 | ||
340 | if (ip->i_di.di_blocks != 1) { | |
341 | if (gfs2_consist_inode(ip)) | |
4cc14f0b | 342 | gfs2_dinode_print(ip); |
b3b94faa DT |
343 | return -EIO; |
344 | } | |
345 | ||
346 | al = gfs2_alloc_get(ip); | |
347 | ||
348 | error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
349 | if (error) | |
350 | goto out; | |
351 | ||
352 | error = gfs2_rindex_hold(sdp, &al->al_ri_gh); | |
353 | if (error) | |
354 | goto out_qs; | |
355 | ||
dbb7cae2 | 356 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr); |
b3b94faa DT |
357 | if (!rgd) { |
358 | gfs2_consist_inode(ip); | |
359 | error = -EIO; | |
360 | goto out_rindex_relse; | |
361 | } | |
362 | ||
363 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, | |
364 | &al->al_rgd_gh); | |
365 | if (error) | |
366 | goto out_rindex_relse; | |
367 | ||
420b9e5e | 368 | error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS + RES_QUOTA, 1); |
b3b94faa DT |
369 | if (error) |
370 | goto out_rg_gunlock; | |
371 | ||
2bcd610d SW |
372 | set_bit(GLF_DIRTY, &ip->i_gl->gl_flags); |
373 | set_bit(GLF_LFLUSH, &ip->i_gl->gl_flags); | |
b3b94faa DT |
374 | |
375 | gfs2_free_di(rgd, ip); | |
376 | ||
b3b94faa DT |
377 | gfs2_trans_end(sdp); |
378 | clear_bit(GLF_STICKY, &ip->i_gl->gl_flags); | |
379 | ||
feaa7bba | 380 | out_rg_gunlock: |
b3b94faa | 381 | gfs2_glock_dq_uninit(&al->al_rgd_gh); |
feaa7bba | 382 | out_rindex_relse: |
b3b94faa | 383 | gfs2_glock_dq_uninit(&al->al_ri_gh); |
feaa7bba | 384 | out_qs: |
b3b94faa | 385 | gfs2_quota_unhold(ip); |
36327521 | 386 | out: |
feaa7bba | 387 | gfs2_alloc_put(ip); |
b3b94faa DT |
388 | return error; |
389 | } | |
390 | ||
b3b94faa | 391 | /** |
87d21e07 | 392 | * gfs2_change_nlink - Change nlink count on inode |
b3b94faa DT |
393 | * @ip: The GFS2 inode |
394 | * @diff: The change in the nlink count required | |
395 | * | |
396 | * Returns: errno | |
397 | */ | |
87d21e07 | 398 | int gfs2_change_nlink(struct gfs2_inode *ip, int diff) |
b3b94faa DT |
399 | { |
400 | struct buffer_head *dibh; | |
cd915493 | 401 | u32 nlink; |
b3b94faa DT |
402 | int error; |
403 | ||
4f56110a SW |
404 | BUG_ON(diff != 1 && diff != -1); |
405 | nlink = ip->i_inode.i_nlink + diff; | |
b3b94faa DT |
406 | |
407 | /* If we are reducing the nlink count, but the new value ends up being | |
408 | bigger than the old one, we must have underflowed. */ | |
4f56110a | 409 | if (diff < 0 && nlink > ip->i_inode.i_nlink) { |
b3b94faa | 410 | if (gfs2_consist_inode(ip)) |
4cc14f0b | 411 | gfs2_dinode_print(ip); |
b3b94faa DT |
412 | return -EIO; |
413 | } | |
414 | ||
415 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
416 | if (error) | |
417 | return error; | |
418 | ||
4f56110a SW |
419 | if (diff > 0) |
420 | inc_nlink(&ip->i_inode); | |
421 | else | |
422 | drop_nlink(&ip->i_inode); | |
423 | ||
4bd91ba1 | 424 | ip->i_inode.i_ctime = CURRENT_TIME; |
b3b94faa | 425 | |
d4e9c4c3 | 426 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 427 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa | 428 | brelse(dibh); |
feaa7bba | 429 | mark_inode_dirty(&ip->i_inode); |
b3b94faa | 430 | |
87d21e07 | 431 | if (ip->i_inode.i_nlink == 0) |
ddee7608 | 432 | gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */ |
87d21e07 | 433 | |
5509826f WC |
434 | return error; |
435 | } | |
436 | ||
c752666c SW |
437 | struct inode *gfs2_lookup_simple(struct inode *dip, const char *name) |
438 | { | |
439 | struct qstr qstr; | |
6c93fd1e | 440 | struct inode *inode; |
71b86f56 | 441 | gfs2_str2qstr(&qstr, name); |
6c93fd1e RC |
442 | inode = gfs2_lookupi(dip, &qstr, 1, NULL); |
443 | /* gfs2_lookupi has inconsistent callers: vfs | |
444 | * related routines expect NULL for no entry found, | |
445 | * gfs2_lookup_simple callers expect ENOENT | |
446 | * and do not check for NULL. | |
447 | */ | |
448 | if (inode == NULL) | |
449 | return ERR_PTR(-ENOENT); | |
450 | else | |
451 | return inode; | |
c752666c SW |
452 | } |
453 | ||
454 | ||
b3b94faa DT |
455 | /** |
456 | * gfs2_lookupi - Look up a filename in a directory and return its inode | |
457 | * @d_gh: An initialized holder for the directory glock | |
458 | * @name: The name of the inode to look for | |
459 | * @is_root: If 1, ignore the caller's permissions | |
460 | * @i_gh: An uninitialized holder for the new inode glock | |
461 | * | |
d7c103d0 SW |
462 | * This can be called via the VFS filldir function when NFS is doing |
463 | * a readdirplus and the inode which its intending to stat isn't | |
464 | * already in cache. In this case we must not take the directory glock | |
465 | * again, since the readdir call will have already taken that lock. | |
b3b94faa DT |
466 | * |
467 | * Returns: errno | |
468 | */ | |
469 | ||
feaa7bba SW |
470 | struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name, |
471 | int is_root, struct nameidata *nd) | |
b3b94faa | 472 | { |
c9fd4307 | 473 | struct super_block *sb = dir->i_sb; |
feaa7bba | 474 | struct gfs2_inode *dip = GFS2_I(dir); |
b3b94faa | 475 | struct gfs2_holder d_gh; |
037bcbb7 | 476 | int error = 0; |
c752666c | 477 | struct inode *inode = NULL; |
d7c103d0 | 478 | int unlock = 0; |
b3b94faa DT |
479 | |
480 | if (!name->len || name->len > GFS2_FNAMESIZE) | |
c752666c | 481 | return ERR_PTR(-ENAMETOOLONG); |
b3b94faa | 482 | |
c752666c SW |
483 | if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) || |
484 | (name->len == 2 && memcmp(name->name, "..", 2) == 0 && | |
485 | dir == sb->s_root->d_inode)) { | |
320dd101 SW |
486 | igrab(dir); |
487 | return dir; | |
b3b94faa DT |
488 | } |
489 | ||
d7c103d0 SW |
490 | if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) { |
491 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh); | |
492 | if (error) | |
493 | return ERR_PTR(error); | |
494 | unlock = 1; | |
495 | } | |
b3b94faa DT |
496 | |
497 | if (!is_root) { | |
faf450ef | 498 | error = permission(dir, MAY_EXEC, NULL); |
b3b94faa DT |
499 | if (error) |
500 | goto out; | |
501 | } | |
502 | ||
dbb7cae2 SW |
503 | inode = gfs2_dir_search(dir, name); |
504 | if (IS_ERR(inode)) | |
505 | error = PTR_ERR(inode); | |
7359a19c | 506 | out: |
d7c103d0 SW |
507 | if (unlock) |
508 | gfs2_glock_dq_uninit(&d_gh); | |
c752666c SW |
509 | if (error == -ENOENT) |
510 | return NULL; | |
d7c103d0 | 511 | return inode ? inode : ERR_PTR(error); |
b3b94faa DT |
512 | } |
513 | ||
bb8d8a6f SW |
514 | static void gfs2_inum_range_in(struct gfs2_inum_range_host *ir, const void *buf) |
515 | { | |
516 | const struct gfs2_inum_range *str = buf; | |
517 | ||
518 | ir->ir_start = be64_to_cpu(str->ir_start); | |
519 | ir->ir_length = be64_to_cpu(str->ir_length); | |
520 | } | |
521 | ||
522 | static void gfs2_inum_range_out(const struct gfs2_inum_range_host *ir, void *buf) | |
523 | { | |
524 | struct gfs2_inum_range *str = buf; | |
525 | ||
526 | str->ir_start = cpu_to_be64(ir->ir_start); | |
527 | str->ir_length = cpu_to_be64(ir->ir_length); | |
528 | } | |
529 | ||
cd915493 | 530 | static int pick_formal_ino_1(struct gfs2_sbd *sdp, u64 *formal_ino) |
b3b94faa | 531 | { |
feaa7bba | 532 | struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode); |
b3b94faa | 533 | struct buffer_head *bh; |
e6972647 | 534 | struct gfs2_inum_range_host ir; |
b3b94faa DT |
535 | int error; |
536 | ||
537 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
538 | if (error) | |
539 | return error; | |
f55ab26a | 540 | mutex_lock(&sdp->sd_inum_mutex); |
b3b94faa DT |
541 | |
542 | error = gfs2_meta_inode_buffer(ip, &bh); | |
543 | if (error) { | |
f55ab26a | 544 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
545 | gfs2_trans_end(sdp); |
546 | return error; | |
547 | } | |
548 | ||
549 | gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode)); | |
550 | ||
551 | if (ir.ir_length) { | |
552 | *formal_ino = ir.ir_start++; | |
553 | ir.ir_length--; | |
d4e9c4c3 | 554 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
555 | gfs2_inum_range_out(&ir, |
556 | bh->b_data + sizeof(struct gfs2_dinode)); | |
557 | brelse(bh); | |
f55ab26a | 558 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
559 | gfs2_trans_end(sdp); |
560 | return 0; | |
561 | } | |
562 | ||
563 | brelse(bh); | |
564 | ||
f55ab26a | 565 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa DT |
566 | gfs2_trans_end(sdp); |
567 | ||
568 | return 1; | |
569 | } | |
570 | ||
cd915493 | 571 | static int pick_formal_ino_2(struct gfs2_sbd *sdp, u64 *formal_ino) |
b3b94faa | 572 | { |
feaa7bba SW |
573 | struct gfs2_inode *ip = GFS2_I(sdp->sd_ir_inode); |
574 | struct gfs2_inode *m_ip = GFS2_I(sdp->sd_inum_inode); | |
b3b94faa DT |
575 | struct gfs2_holder gh; |
576 | struct buffer_head *bh; | |
e6972647 | 577 | struct gfs2_inum_range_host ir; |
b3b94faa DT |
578 | int error; |
579 | ||
580 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
581 | if (error) | |
582 | return error; | |
583 | ||
584 | error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0); | |
585 | if (error) | |
586 | goto out; | |
f55ab26a | 587 | mutex_lock(&sdp->sd_inum_mutex); |
b3b94faa DT |
588 | |
589 | error = gfs2_meta_inode_buffer(ip, &bh); | |
590 | if (error) | |
591 | goto out_end_trans; | |
907b9bce | 592 | |
b3b94faa DT |
593 | gfs2_inum_range_in(&ir, bh->b_data + sizeof(struct gfs2_dinode)); |
594 | ||
595 | if (!ir.ir_length) { | |
596 | struct buffer_head *m_bh; | |
cd915493 | 597 | u64 x, y; |
b44b84d7 | 598 | __be64 z; |
b3b94faa DT |
599 | |
600 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); | |
601 | if (error) | |
602 | goto out_brelse; | |
603 | ||
b44b84d7 AV |
604 | z = *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)); |
605 | x = y = be64_to_cpu(z); | |
b3b94faa DT |
606 | ir.ir_start = x; |
607 | ir.ir_length = GFS2_INUM_QUANTUM; | |
608 | x += GFS2_INUM_QUANTUM; | |
609 | if (x < y) | |
610 | gfs2_consist_inode(m_ip); | |
b44b84d7 | 611 | z = cpu_to_be64(x); |
d4e9c4c3 | 612 | gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1); |
b44b84d7 | 613 | *(__be64 *)(m_bh->b_data + sizeof(struct gfs2_dinode)) = z; |
b3b94faa DT |
614 | |
615 | brelse(m_bh); | |
616 | } | |
617 | ||
618 | *formal_ino = ir.ir_start++; | |
619 | ir.ir_length--; | |
620 | ||
d4e9c4c3 | 621 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
622 | gfs2_inum_range_out(&ir, bh->b_data + sizeof(struct gfs2_dinode)); |
623 | ||
420b9e5e | 624 | out_brelse: |
b3b94faa | 625 | brelse(bh); |
420b9e5e | 626 | out_end_trans: |
f55ab26a | 627 | mutex_unlock(&sdp->sd_inum_mutex); |
b3b94faa | 628 | gfs2_trans_end(sdp); |
420b9e5e | 629 | out: |
b3b94faa | 630 | gfs2_glock_dq_uninit(&gh); |
b3b94faa DT |
631 | return error; |
632 | } | |
633 | ||
cd915493 | 634 | static int pick_formal_ino(struct gfs2_sbd *sdp, u64 *inum) |
b3b94faa DT |
635 | { |
636 | int error; | |
637 | ||
638 | error = pick_formal_ino_1(sdp, inum); | |
639 | if (error <= 0) | |
640 | return error; | |
641 | ||
642 | error = pick_formal_ino_2(sdp, inum); | |
643 | ||
644 | return error; | |
645 | } | |
646 | ||
647 | /** | |
648 | * create_ok - OK to create a new on-disk inode here? | |
649 | * @dip: Directory in which dinode is to be created | |
650 | * @name: Name of new dinode | |
651 | * @mode: | |
652 | * | |
653 | * Returns: errno | |
654 | */ | |
655 | ||
feaa7bba | 656 | static int create_ok(struct gfs2_inode *dip, const struct qstr *name, |
b3b94faa DT |
657 | unsigned int mode) |
658 | { | |
659 | int error; | |
660 | ||
faf450ef | 661 | error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL); |
b3b94faa DT |
662 | if (error) |
663 | return error; | |
664 | ||
665 | /* Don't create entries in an unlinked directory */ | |
4f56110a | 666 | if (!dip->i_inode.i_nlink) |
b3b94faa DT |
667 | return -EPERM; |
668 | ||
dbb7cae2 | 669 | error = gfs2_dir_check(&dip->i_inode, name, NULL); |
b3b94faa DT |
670 | switch (error) { |
671 | case -ENOENT: | |
672 | error = 0; | |
673 | break; | |
674 | case 0: | |
675 | return -EEXIST; | |
676 | default: | |
677 | return error; | |
678 | } | |
679 | ||
cd915493 | 680 | if (dip->i_di.di_entries == (u32)-1) |
b3b94faa | 681 | return -EFBIG; |
4f56110a | 682 | if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1) |
b3b94faa DT |
683 | return -EMLINK; |
684 | ||
685 | return 0; | |
686 | } | |
687 | ||
688 | static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode, | |
689 | unsigned int *uid, unsigned int *gid) | |
690 | { | |
feaa7bba | 691 | if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir && |
2933f925 | 692 | (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) { |
b3b94faa DT |
693 | if (S_ISDIR(*mode)) |
694 | *mode |= S_ISUID; | |
2933f925 | 695 | else if (dip->i_inode.i_uid != current->fsuid) |
b3b94faa | 696 | *mode &= ~07111; |
2933f925 | 697 | *uid = dip->i_inode.i_uid; |
b3b94faa DT |
698 | } else |
699 | *uid = current->fsuid; | |
700 | ||
b60623c2 | 701 | if (dip->i_inode.i_mode & S_ISGID) { |
b3b94faa DT |
702 | if (S_ISDIR(*mode)) |
703 | *mode |= S_ISGID; | |
2933f925 | 704 | *gid = dip->i_inode.i_gid; |
b3b94faa DT |
705 | } else |
706 | *gid = current->fsgid; | |
707 | } | |
708 | ||
dbb7cae2 | 709 | static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation) |
b3b94faa | 710 | { |
feaa7bba | 711 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
712 | int error; |
713 | ||
6dbd8224 SW |
714 | if (gfs2_alloc_get(dip) == NULL) |
715 | return -ENOMEM; | |
b3b94faa | 716 | |
6dbd8224 | 717 | dip->i_alloc->al_requested = RES_DINODE; |
b3b94faa DT |
718 | error = gfs2_inplace_reserve(dip); |
719 | if (error) | |
720 | goto out; | |
721 | ||
feaa7bba | 722 | error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0); |
b3b94faa DT |
723 | if (error) |
724 | goto out_ipreserv; | |
725 | ||
dbb7cae2 | 726 | *no_addr = gfs2_alloc_di(dip, generation); |
b3b94faa DT |
727 | |
728 | gfs2_trans_end(sdp); | |
729 | ||
4340fe62 | 730 | out_ipreserv: |
b3b94faa | 731 | gfs2_inplace_release(dip); |
4340fe62 | 732 | out: |
b3b94faa | 733 | gfs2_alloc_put(dip); |
b3b94faa DT |
734 | return error; |
735 | } | |
736 | ||
737 | /** | |
738 | * init_dinode - Fill in a new dinode structure | |
739 | * @dip: the directory this inode is being created in | |
740 | * @gl: The glock covering the new inode | |
741 | * @inum: the inode number | |
742 | * @mode: the file permissions | |
743 | * @uid: | |
744 | * @gid: | |
745 | * | |
746 | */ | |
747 | ||
748 | static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, | |
629a21e7 | 749 | const struct gfs2_inum_host *inum, unsigned int mode, |
4340fe62 | 750 | unsigned int uid, unsigned int gid, |
e9bd2b3b | 751 | const u64 *generation, dev_t dev, struct buffer_head **bhp) |
b3b94faa | 752 | { |
feaa7bba | 753 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b96ca4fa | 754 | struct gfs2_dinode *di; |
b3b94faa | 755 | struct buffer_head *dibh; |
4bd91ba1 | 756 | struct timespec tv = CURRENT_TIME; |
b3b94faa DT |
757 | |
758 | dibh = gfs2_meta_new(gl, inum->no_addr); | |
d4e9c4c3 | 759 | gfs2_trans_add_bh(gl, dibh, 1); |
b3b94faa DT |
760 | gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI); |
761 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); | |
b96ca4fa SW |
762 | di = (struct gfs2_dinode *)dibh->b_data; |
763 | ||
2442a098 SW |
764 | di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino); |
765 | di->di_num.no_addr = cpu_to_be64(inum->no_addr); | |
b96ca4fa SW |
766 | di->di_mode = cpu_to_be32(mode); |
767 | di->di_uid = cpu_to_be32(uid); | |
768 | di->di_gid = cpu_to_be32(gid); | |
294caaa3 SW |
769 | di->di_nlink = 0; |
770 | di->di_size = 0; | |
b96ca4fa | 771 | di->di_blocks = cpu_to_be64(1); |
4bd91ba1 | 772 | di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec); |
e7f14f4d SW |
773 | di->di_major = cpu_to_be32(MAJOR(dev)); |
774 | di->di_minor = cpu_to_be32(MINOR(dev)); | |
b96ca4fa | 775 | di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr); |
4340fe62 | 776 | di->di_generation = cpu_to_be64(*generation); |
294caaa3 | 777 | di->di_flags = 0; |
b3b94faa DT |
778 | |
779 | if (S_ISREG(mode)) { | |
780 | if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_JDATA) || | |
781 | gfs2_tune_get(sdp, gt_new_files_jdata)) | |
b96ca4fa | 782 | di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA); |
b3b94faa DT |
783 | if ((dip->i_di.di_flags & GFS2_DIF_INHERIT_DIRECTIO) || |
784 | gfs2_tune_get(sdp, gt_new_files_directio)) | |
b96ca4fa | 785 | di->di_flags |= cpu_to_be32(GFS2_DIF_DIRECTIO); |
b3b94faa | 786 | } else if (S_ISDIR(mode)) { |
568f4c96 SW |
787 | di->di_flags |= cpu_to_be32(dip->i_di.di_flags & |
788 | GFS2_DIF_INHERIT_DIRECTIO); | |
789 | di->di_flags |= cpu_to_be32(dip->i_di.di_flags & | |
790 | GFS2_DIF_INHERIT_JDATA); | |
b3b94faa DT |
791 | } |
792 | ||
b96ca4fa | 793 | di->__pad1 = 0; |
a9583c79 | 794 | di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0); |
294caaa3 | 795 | di->di_height = 0; |
b96ca4fa SW |
796 | di->__pad2 = 0; |
797 | di->__pad3 = 0; | |
294caaa3 SW |
798 | di->di_depth = 0; |
799 | di->di_entries = 0; | |
b96ca4fa | 800 | memset(&di->__pad4, 0, sizeof(di->__pad4)); |
294caaa3 | 801 | di->di_eattr = 0; |
4bd91ba1 SW |
802 | di->di_atime_nsec = cpu_to_be32(tv.tv_nsec); |
803 | di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec); | |
804 | di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec); | |
b96ca4fa | 805 | memset(&di->di_reserved, 0, sizeof(di->di_reserved)); |
e9bd2b3b WC |
806 | |
807 | set_buffer_uptodate(dibh); | |
b96ca4fa | 808 | |
e9bd2b3b | 809 | *bhp = dibh; |
b3b94faa DT |
810 | } |
811 | ||
812 | static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, | |
629a21e7 | 813 | unsigned int mode, const struct gfs2_inum_host *inum, |
e9bd2b3b | 814 | const u64 *generation, dev_t dev, struct buffer_head **bhp) |
b3b94faa | 815 | { |
feaa7bba | 816 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
817 | unsigned int uid, gid; |
818 | int error; | |
819 | ||
820 | munge_mode_uid_gid(dip, &mode, &uid, &gid); | |
b3b94faa DT |
821 | gfs2_alloc_get(dip); |
822 | ||
823 | error = gfs2_quota_lock(dip, uid, gid); | |
824 | if (error) | |
825 | goto out; | |
826 | ||
827 | error = gfs2_quota_check(dip, uid, gid); | |
828 | if (error) | |
829 | goto out_quota; | |
830 | ||
feaa7bba | 831 | error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0); |
b3b94faa DT |
832 | if (error) |
833 | goto out_quota; | |
834 | ||
e9bd2b3b | 835 | init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp); |
b3b94faa | 836 | gfs2_quota_change(dip, +1, uid, gid); |
b3b94faa DT |
837 | gfs2_trans_end(sdp); |
838 | ||
feaa7bba | 839 | out_quota: |
b3b94faa | 840 | gfs2_quota_unlock(dip); |
feaa7bba | 841 | out: |
b3b94faa | 842 | gfs2_alloc_put(dip); |
b3b94faa DT |
843 | return error; |
844 | } | |
845 | ||
feaa7bba SW |
846 | static int link_dinode(struct gfs2_inode *dip, const struct qstr *name, |
847 | struct gfs2_inode *ip) | |
b3b94faa | 848 | { |
feaa7bba | 849 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
850 | struct gfs2_alloc *al; |
851 | int alloc_required; | |
852 | struct buffer_head *dibh; | |
853 | int error; | |
854 | ||
855 | al = gfs2_alloc_get(dip); | |
856 | ||
857 | error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
858 | if (error) | |
859 | goto fail; | |
860 | ||
feaa7bba | 861 | error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name); |
c752666c | 862 | if (alloc_required < 0) |
1b8177ec | 863 | goto fail_quota_locks; |
b3b94faa | 864 | if (alloc_required) { |
2933f925 | 865 | error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid); |
b3b94faa DT |
866 | if (error) |
867 | goto fail_quota_locks; | |
868 | ||
869 | al->al_requested = sdp->sd_max_dirres; | |
870 | ||
871 | error = gfs2_inplace_reserve(dip); | |
872 | if (error) | |
873 | goto fail_quota_locks; | |
874 | ||
320dd101 | 875 | error = gfs2_trans_begin(sdp, sdp->sd_max_dirres + |
bb8d8a6f | 876 | al->al_rgd->rd_length + |
907b9bce | 877 | 2 * RES_DINODE + |
b3b94faa DT |
878 | RES_STATFS + RES_QUOTA, 0); |
879 | if (error) | |
880 | goto fail_ipreserv; | |
881 | } else { | |
feaa7bba | 882 | error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0); |
b3b94faa DT |
883 | if (error) |
884 | goto fail_quota_locks; | |
885 | } | |
886 | ||
dbb7cae2 | 887 | error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode)); |
b3b94faa DT |
888 | if (error) |
889 | goto fail_end_trans; | |
890 | ||
891 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
892 | if (error) | |
893 | goto fail_end_trans; | |
4f56110a | 894 | ip->i_inode.i_nlink = 1; |
d4e9c4c3 | 895 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 896 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa | 897 | brelse(dibh); |
b3b94faa DT |
898 | return 0; |
899 | ||
320dd101 | 900 | fail_end_trans: |
b3b94faa DT |
901 | gfs2_trans_end(sdp); |
902 | ||
320dd101 | 903 | fail_ipreserv: |
6dbd8224 | 904 | if (dip->i_alloc->al_rgd) |
b3b94faa DT |
905 | gfs2_inplace_release(dip); |
906 | ||
320dd101 | 907 | fail_quota_locks: |
b3b94faa DT |
908 | gfs2_quota_unlock(dip); |
909 | ||
320dd101 | 910 | fail: |
b3b94faa | 911 | gfs2_alloc_put(dip); |
b3b94faa DT |
912 | return error; |
913 | } | |
914 | ||
fcb47e0b RH |
915 | static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip) |
916 | { | |
917 | int err; | |
918 | size_t len; | |
919 | void *value; | |
920 | char *name; | |
921 | struct gfs2_ea_request er; | |
922 | ||
923 | err = security_inode_init_security(&ip->i_inode, &dip->i_inode, | |
924 | &name, &value, &len); | |
925 | ||
926 | if (err) { | |
927 | if (err == -EOPNOTSUPP) | |
928 | return 0; | |
929 | return err; | |
930 | } | |
931 | ||
932 | memset(&er, 0, sizeof(struct gfs2_ea_request)); | |
933 | ||
934 | er.er_type = GFS2_EATYPE_SECURITY; | |
935 | er.er_name = name; | |
936 | er.er_data = value; | |
937 | er.er_name_len = strlen(name); | |
938 | er.er_data_len = len; | |
939 | ||
940 | err = gfs2_ea_set_i(ip, &er); | |
941 | ||
942 | kfree(value); | |
943 | kfree(name); | |
944 | ||
945 | return err; | |
946 | } | |
947 | ||
b3b94faa DT |
948 | /** |
949 | * gfs2_createi - Create a new inode | |
950 | * @ghs: An array of two holders | |
951 | * @name: The name of the new file | |
952 | * @mode: the permissions on the new inode | |
953 | * | |
954 | * @ghs[0] is an initialized holder for the directory | |
955 | * @ghs[1] is the holder for the inode lock | |
956 | * | |
7359a19c | 957 | * If the return value is not NULL, the glocks on both the directory and the new |
b3b94faa DT |
958 | * file are held. A transaction has been started and an inplace reservation |
959 | * is held, as well. | |
960 | * | |
7359a19c | 961 | * Returns: An inode |
b3b94faa DT |
962 | */ |
963 | ||
feaa7bba | 964 | struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, |
e7f14f4d | 965 | unsigned int mode, dev_t dev) |
b3b94faa | 966 | { |
e1cc8603 | 967 | struct inode *inode = NULL; |
5c676f6d | 968 | struct gfs2_inode *dip = ghs->gh_gl->gl_object; |
feaa7bba SW |
969 | struct inode *dir = &dip->i_inode; |
970 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); | |
dbb7cae2 | 971 | struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 }; |
b3b94faa | 972 | int error; |
4340fe62 | 973 | u64 generation; |
f91a0d3e | 974 | struct buffer_head *bh = NULL; |
b3b94faa DT |
975 | |
976 | if (!name->len || name->len > GFS2_FNAMESIZE) | |
7359a19c | 977 | return ERR_PTR(-ENAMETOOLONG); |
b3b94faa | 978 | |
b3b94faa DT |
979 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs); |
980 | error = gfs2_glock_nq(ghs); | |
981 | if (error) | |
982 | goto fail; | |
983 | ||
984 | error = create_ok(dip, name, mode); | |
985 | if (error) | |
986 | goto fail_gunlock; | |
987 | ||
feaa7bba | 988 | error = pick_formal_ino(sdp, &inum.no_formal_ino); |
b3b94faa DT |
989 | if (error) |
990 | goto fail_gunlock; | |
991 | ||
dbb7cae2 | 992 | error = alloc_dinode(dip, &inum.no_addr, &generation); |
b3b94faa DT |
993 | if (error) |
994 | goto fail_gunlock; | |
995 | ||
28626e20 SW |
996 | error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops, |
997 | LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1); | |
998 | if (error) | |
999 | goto fail_gunlock; | |
b3b94faa | 1000 | |
e9bd2b3b | 1001 | error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh); |
b3b94faa DT |
1002 | if (error) |
1003 | goto fail_gunlock2; | |
1004 | ||
bb9bcf06 WC |
1005 | inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), |
1006 | inum.no_addr, | |
7a9f53b3 | 1007 | inum.no_formal_ino, 0); |
feaa7bba | 1008 | if (IS_ERR(inode)) |
b3b94faa DT |
1009 | goto fail_gunlock2; |
1010 | ||
feaa7bba | 1011 | error = gfs2_inode_refresh(GFS2_I(inode)); |
b3b94faa | 1012 | if (error) |
e1cc8603 | 1013 | goto fail_gunlock2; |
b3b94faa | 1014 | |
feaa7bba | 1015 | error = gfs2_acl_create(dip, GFS2_I(inode)); |
b3b94faa | 1016 | if (error) |
e1cc8603 | 1017 | goto fail_gunlock2; |
b3b94faa | 1018 | |
fcb47e0b RH |
1019 | error = gfs2_security_init(dip, GFS2_I(inode)); |
1020 | if (error) | |
e1cc8603 | 1021 | goto fail_gunlock2; |
fcb47e0b | 1022 | |
feaa7bba | 1023 | error = link_dinode(dip, name, GFS2_I(inode)); |
b3b94faa | 1024 | if (error) |
e1cc8603 | 1025 | goto fail_gunlock2; |
b3b94faa | 1026 | |
f91a0d3e SW |
1027 | if (bh) |
1028 | brelse(bh); | |
7359a19c SW |
1029 | if (!inode) |
1030 | return ERR_PTR(-ENOMEM); | |
1031 | return inode; | |
b3b94faa | 1032 | |
320dd101 | 1033 | fail_gunlock2: |
b3b94faa | 1034 | gfs2_glock_dq_uninit(ghs + 1); |
e1cc8603 SW |
1035 | if (inode) |
1036 | iput(inode); | |
320dd101 | 1037 | fail_gunlock: |
b3b94faa | 1038 | gfs2_glock_dq(ghs); |
320dd101 | 1039 | fail: |
f91a0d3e SW |
1040 | if (bh) |
1041 | brelse(bh); | |
7359a19c | 1042 | return ERR_PTR(error); |
b3b94faa DT |
1043 | } |
1044 | ||
b3b94faa DT |
1045 | /** |
1046 | * gfs2_rmdiri - Remove a directory | |
1047 | * @dip: The parent directory of the directory to be removed | |
1048 | * @name: The name of the directory to be removed | |
1049 | * @ip: The GFS2 inode of the directory to be removed | |
1050 | * | |
1051 | * Assumes Glocks on dip and ip are held | |
1052 | * | |
1053 | * Returns: errno | |
1054 | */ | |
1055 | ||
feaa7bba SW |
1056 | int gfs2_rmdiri(struct gfs2_inode *dip, const struct qstr *name, |
1057 | struct gfs2_inode *ip) | |
b3b94faa | 1058 | { |
b3b94faa DT |
1059 | struct qstr dotname; |
1060 | int error; | |
1061 | ||
1062 | if (ip->i_di.di_entries != 2) { | |
1063 | if (gfs2_consist_inode(ip)) | |
4cc14f0b | 1064 | gfs2_dinode_print(ip); |
b3b94faa DT |
1065 | return -EIO; |
1066 | } | |
1067 | ||
1068 | error = gfs2_dir_del(dip, name); | |
1069 | if (error) | |
1070 | return error; | |
1071 | ||
1072 | error = gfs2_change_nlink(dip, -1); | |
1073 | if (error) | |
1074 | return error; | |
1075 | ||
71b86f56 | 1076 | gfs2_str2qstr(&dotname, "."); |
b3b94faa DT |
1077 | error = gfs2_dir_del(ip, &dotname); |
1078 | if (error) | |
1079 | return error; | |
1080 | ||
feaa7bba | 1081 | gfs2_str2qstr(&dotname, ".."); |
b3b94faa DT |
1082 | error = gfs2_dir_del(ip, &dotname); |
1083 | if (error) | |
1084 | return error; | |
1085 | ||
4f56110a SW |
1086 | /* It looks odd, but it really should be done twice */ |
1087 | error = gfs2_change_nlink(ip, -1); | |
1088 | if (error) | |
1089 | return error; | |
1090 | ||
1091 | error = gfs2_change_nlink(ip, -1); | |
b3b94faa DT |
1092 | if (error) |
1093 | return error; | |
1094 | ||
b3b94faa DT |
1095 | return error; |
1096 | } | |
1097 | ||
1098 | /* | |
1099 | * gfs2_unlink_ok - check to see that a inode is still in a directory | |
1100 | * @dip: the directory | |
1101 | * @name: the name of the file | |
1102 | * @ip: the inode | |
1103 | * | |
1104 | * Assumes that the lock on (at least) @dip is held. | |
1105 | * | |
1106 | * Returns: 0 if the parent/child relationship is correct, errno if it isn't | |
1107 | */ | |
1108 | ||
feaa7bba | 1109 | int gfs2_unlink_ok(struct gfs2_inode *dip, const struct qstr *name, |
dbb7cae2 | 1110 | const struct gfs2_inode *ip) |
b3b94faa | 1111 | { |
b3b94faa DT |
1112 | int error; |
1113 | ||
feaa7bba | 1114 | if (IS_IMMUTABLE(&ip->i_inode) || IS_APPEND(&ip->i_inode)) |
b3b94faa DT |
1115 | return -EPERM; |
1116 | ||
b60623c2 | 1117 | if ((dip->i_inode.i_mode & S_ISVTX) && |
2933f925 SW |
1118 | dip->i_inode.i_uid != current->fsuid && |
1119 | ip->i_inode.i_uid != current->fsuid && !capable(CAP_FOWNER)) | |
b3b94faa DT |
1120 | return -EPERM; |
1121 | ||
feaa7bba | 1122 | if (IS_APPEND(&dip->i_inode)) |
b3b94faa DT |
1123 | return -EPERM; |
1124 | ||
faf450ef | 1125 | error = permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, NULL); |
b3b94faa DT |
1126 | if (error) |
1127 | return error; | |
1128 | ||
dbb7cae2 | 1129 | error = gfs2_dir_check(&dip->i_inode, name, ip); |
b3b94faa DT |
1130 | if (error) |
1131 | return error; | |
1132 | ||
b3b94faa DT |
1133 | return 0; |
1134 | } | |
1135 | ||
1136 | /* | |
1137 | * gfs2_ok_to_move - check if it's ok to move a directory to another directory | |
1138 | * @this: move this | |
1139 | * @to: to here | |
1140 | * | |
1141 | * Follow @to back to the root and make sure we don't encounter @this | |
1142 | * Assumes we already hold the rename lock. | |
1143 | * | |
1144 | * Returns: errno | |
1145 | */ | |
1146 | ||
1147 | int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to) | |
1148 | { | |
feaa7bba | 1149 | struct inode *dir = &to->i_inode; |
c9fd4307 | 1150 | struct super_block *sb = dir->i_sb; |
7359a19c | 1151 | struct inode *tmp; |
b3b94faa DT |
1152 | struct qstr dotdot; |
1153 | int error = 0; | |
1154 | ||
71b86f56 | 1155 | gfs2_str2qstr(&dotdot, ".."); |
b3b94faa | 1156 | |
7359a19c | 1157 | igrab(dir); |
b3b94faa DT |
1158 | |
1159 | for (;;) { | |
feaa7bba | 1160 | if (dir == &this->i_inode) { |
b3b94faa DT |
1161 | error = -EINVAL; |
1162 | break; | |
1163 | } | |
c9fd4307 | 1164 | if (dir == sb->s_root->d_inode) { |
b3b94faa DT |
1165 | error = 0; |
1166 | break; | |
1167 | } | |
1168 | ||
c752666c SW |
1169 | tmp = gfs2_lookupi(dir, &dotdot, 1, NULL); |
1170 | if (IS_ERR(tmp)) { | |
1171 | error = PTR_ERR(tmp); | |
b3b94faa | 1172 | break; |
c752666c | 1173 | } |
b3b94faa | 1174 | |
7359a19c SW |
1175 | iput(dir); |
1176 | dir = tmp; | |
b3b94faa DT |
1177 | } |
1178 | ||
7359a19c | 1179 | iput(dir); |
b3b94faa DT |
1180 | |
1181 | return error; | |
1182 | } | |
1183 | ||
1184 | /** | |
1185 | * gfs2_readlinki - return the contents of a symlink | |
1186 | * @ip: the symlink's inode | |
1187 | * @buf: a pointer to the buffer to be filled | |
1188 | * @len: a pointer to the length of @buf | |
1189 | * | |
1190 | * If @buf is too small, a piece of memory is kmalloc()ed and needs | |
1191 | * to be freed by the caller. | |
1192 | * | |
1193 | * Returns: errno | |
1194 | */ | |
1195 | ||
1196 | int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len) | |
1197 | { | |
1198 | struct gfs2_holder i_gh; | |
1199 | struct buffer_head *dibh; | |
1200 | unsigned int x; | |
1201 | int error; | |
1202 | ||
1203 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh); | |
1204 | error = gfs2_glock_nq_atime(&i_gh); | |
1205 | if (error) { | |
1206 | gfs2_holder_uninit(&i_gh); | |
1207 | return error; | |
1208 | } | |
1209 | ||
1210 | if (!ip->i_di.di_size) { | |
1211 | gfs2_consist_inode(ip); | |
1212 | error = -EIO; | |
1213 | goto out; | |
1214 | } | |
1215 | ||
1216 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1217 | if (error) | |
1218 | goto out; | |
1219 | ||
1220 | x = ip->i_di.di_size + 1; | |
1221 | if (x > *len) { | |
1222 | *buf = kmalloc(x, GFP_KERNEL); | |
1223 | if (!*buf) { | |
1224 | error = -ENOMEM; | |
1225 | goto out_brelse; | |
1226 | } | |
1227 | } | |
1228 | ||
1229 | memcpy(*buf, dibh->b_data + sizeof(struct gfs2_dinode), x); | |
1230 | *len = x; | |
1231 | ||
feaa7bba | 1232 | out_brelse: |
b3b94faa | 1233 | brelse(dibh); |
feaa7bba | 1234 | out: |
b3b94faa | 1235 | gfs2_glock_dq_uninit(&i_gh); |
b3b94faa DT |
1236 | return error; |
1237 | } | |
1238 | ||
1239 | /** | |
1240 | * gfs2_glock_nq_atime - Acquire a hold on an inode's glock, and | |
1241 | * conditionally update the inode's atime | |
1242 | * @gh: the holder to acquire | |
1243 | * | |
1244 | * Tests atime (access time) for gfs2_read, gfs2_readdir and gfs2_mmap | |
1245 | * Update if the difference between the current time and the inode's current | |
1246 | * atime is greater than an interval specified at mount. | |
1247 | * | |
1248 | * Returns: errno | |
1249 | */ | |
1250 | ||
1251 | int gfs2_glock_nq_atime(struct gfs2_holder *gh) | |
1252 | { | |
1253 | struct gfs2_glock *gl = gh->gh_gl; | |
1254 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
5c676f6d | 1255 | struct gfs2_inode *ip = gl->gl_object; |
4bd91ba1 | 1256 | s64 quantum = gfs2_tune_get(sdp, gt_atime_quantum); |
b3b94faa DT |
1257 | unsigned int state; |
1258 | int flags; | |
1259 | int error; | |
4bd91ba1 | 1260 | struct timespec tv = CURRENT_TIME; |
b3b94faa DT |
1261 | |
1262 | if (gfs2_assert_warn(sdp, gh->gh_flags & GL_ATIME) || | |
1263 | gfs2_assert_warn(sdp, !(gh->gh_flags & GL_ASYNC)) || | |
1264 | gfs2_assert_warn(sdp, gl->gl_ops == &gfs2_inode_glops)) | |
1265 | return -EINVAL; | |
1266 | ||
1267 | state = gh->gh_state; | |
1268 | flags = gh->gh_flags; | |
1269 | ||
1270 | error = gfs2_glock_nq(gh); | |
1271 | if (error) | |
1272 | return error; | |
1273 | ||
1274 | if (test_bit(SDF_NOATIME, &sdp->sd_flags) || | |
1275 | (sdp->sd_vfs->s_flags & MS_RDONLY)) | |
1276 | return 0; | |
1277 | ||
4bd91ba1 | 1278 | if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) { |
b3b94faa | 1279 | gfs2_glock_dq(gh); |
fd88de56 SW |
1280 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, gh->gh_flags & ~LM_FLAG_ANY, |
1281 | gh); | |
b3b94faa DT |
1282 | error = gfs2_glock_nq(gh); |
1283 | if (error) | |
1284 | return error; | |
1285 | ||
1286 | /* Verify that atime hasn't been updated while we were | |
1287 | trying to get exclusive lock. */ | |
1288 | ||
4bd91ba1 SW |
1289 | tv = CURRENT_TIME; |
1290 | if (tv.tv_sec - ip->i_inode.i_atime.tv_sec >= quantum) { | |
b3b94faa | 1291 | struct buffer_head *dibh; |
48516ced | 1292 | struct gfs2_dinode *di; |
b3b94faa DT |
1293 | |
1294 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
1295 | if (error == -EROFS) | |
1296 | return 0; | |
1297 | if (error) | |
1298 | goto fail; | |
1299 | ||
1300 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1301 | if (error) | |
1302 | goto fail_end_trans; | |
1303 | ||
4bd91ba1 | 1304 | ip->i_inode.i_atime = tv; |
b3b94faa | 1305 | |
d4e9c4c3 | 1306 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
48516ced | 1307 | di = (struct gfs2_dinode *)dibh->b_data; |
1a7b1eed | 1308 | di->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec); |
4bd91ba1 | 1309 | di->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec); |
b3b94faa DT |
1310 | brelse(dibh); |
1311 | ||
1312 | gfs2_trans_end(sdp); | |
1313 | } | |
1314 | ||
1315 | /* If someone else has asked for the glock, | |
1316 | unlock and let them have it. Then reacquire | |
1317 | in the original state. */ | |
1318 | if (gfs2_glock_is_blocking(gl)) { | |
1319 | gfs2_glock_dq(gh); | |
1320 | gfs2_holder_reinit(state, flags, gh); | |
1321 | return gfs2_glock_nq(gh); | |
1322 | } | |
1323 | } | |
1324 | ||
1325 | return 0; | |
1326 | ||
feaa7bba | 1327 | fail_end_trans: |
b3b94faa | 1328 | gfs2_trans_end(sdp); |
feaa7bba | 1329 | fail: |
b3b94faa | 1330 | gfs2_glock_dq(gh); |
b3b94faa DT |
1331 | return error; |
1332 | } | |
1333 | ||
b3b94faa DT |
1334 | static int |
1335 | __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |
1336 | { | |
1337 | struct buffer_head *dibh; | |
1338 | int error; | |
1339 | ||
1340 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
1341 | if (!error) { | |
feaa7bba SW |
1342 | error = inode_setattr(&ip->i_inode, attr); |
1343 | gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error); | |
d4e9c4c3 | 1344 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
539e5d6b | 1345 | gfs2_dinode_out(ip, dibh->b_data); |
b3b94faa DT |
1346 | brelse(dibh); |
1347 | } | |
1348 | return error; | |
1349 | } | |
1350 | ||
1351 | /** | |
1352 | * gfs2_setattr_simple - | |
1353 | * @ip: | |
1354 | * @attr: | |
1355 | * | |
1356 | * Called with a reference on the vnode. | |
1357 | * | |
1358 | * Returns: errno | |
1359 | */ | |
1360 | ||
1361 | int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |
1362 | { | |
1363 | int error; | |
1364 | ||
5c676f6d | 1365 | if (current->journal_info) |
b3b94faa DT |
1366 | return __gfs2_setattr_simple(ip, attr); |
1367 | ||
feaa7bba | 1368 | error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0); |
b3b94faa DT |
1369 | if (error) |
1370 | return error; | |
1371 | ||
1372 | error = __gfs2_setattr_simple(ip, attr); | |
feaa7bba | 1373 | gfs2_trans_end(GFS2_SB(&ip->i_inode)); |
b3b94faa DT |
1374 | return error; |
1375 | } | |
1376 | ||
bb8d8a6f SW |
1377 | void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf) |
1378 | { | |
1379 | const struct gfs2_dinode_host *di = &ip->i_di; | |
1380 | struct gfs2_dinode *str = buf; | |
1381 | ||
1382 | str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC); | |
1383 | str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI); | |
1384 | str->di_header.__pad0 = 0; | |
1385 | str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI); | |
1386 | str->di_header.__pad1 = 0; | |
1387 | str->di_num.no_addr = cpu_to_be64(ip->i_no_addr); | |
1388 | str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino); | |
1389 | str->di_mode = cpu_to_be32(ip->i_inode.i_mode); | |
1390 | str->di_uid = cpu_to_be32(ip->i_inode.i_uid); | |
1391 | str->di_gid = cpu_to_be32(ip->i_inode.i_gid); | |
1392 | str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink); | |
1393 | str->di_size = cpu_to_be64(di->di_size); | |
1394 | str->di_blocks = cpu_to_be64(di->di_blocks); | |
1395 | str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec); | |
1396 | str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec); | |
1397 | str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec); | |
1398 | ||
1399 | str->di_goal_meta = cpu_to_be64(di->di_goal_meta); | |
1400 | str->di_goal_data = cpu_to_be64(di->di_goal_data); | |
1401 | str->di_generation = cpu_to_be64(di->di_generation); | |
1402 | ||
1403 | str->di_flags = cpu_to_be32(di->di_flags); | |
1404 | str->di_height = cpu_to_be16(di->di_height); | |
1405 | str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) && | |
1406 | !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ? | |
1407 | GFS2_FORMAT_DE : 0); | |
1408 | str->di_depth = cpu_to_be16(di->di_depth); | |
1409 | str->di_entries = cpu_to_be32(di->di_entries); | |
1410 | ||
1411 | str->di_eattr = cpu_to_be64(di->di_eattr); | |
4bd91ba1 SW |
1412 | str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec); |
1413 | str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec); | |
1414 | str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec); | |
bb8d8a6f SW |
1415 | } |
1416 | ||
1417 | void gfs2_dinode_print(const struct gfs2_inode *ip) | |
1418 | { | |
1419 | const struct gfs2_dinode_host *di = &ip->i_di; | |
1420 | ||
1421 | printk(KERN_INFO " no_formal_ino = %llu\n", | |
1422 | (unsigned long long)ip->i_no_formal_ino); | |
1423 | printk(KERN_INFO " no_addr = %llu\n", | |
1424 | (unsigned long long)ip->i_no_addr); | |
1425 | printk(KERN_INFO " di_size = %llu\n", (unsigned long long)di->di_size); | |
1426 | printk(KERN_INFO " di_blocks = %llu\n", | |
1427 | (unsigned long long)di->di_blocks); | |
1428 | printk(KERN_INFO " di_goal_meta = %llu\n", | |
1429 | (unsigned long long)di->di_goal_meta); | |
1430 | printk(KERN_INFO " di_goal_data = %llu\n", | |
1431 | (unsigned long long)di->di_goal_data); | |
1432 | printk(KERN_INFO " di_flags = 0x%.8X\n", di->di_flags); | |
1433 | printk(KERN_INFO " di_height = %u\n", di->di_height); | |
1434 | printk(KERN_INFO " di_depth = %u\n", di->di_depth); | |
1435 | printk(KERN_INFO " di_entries = %u\n", di->di_entries); | |
1436 | printk(KERN_INFO " di_eattr = %llu\n", | |
1437 | (unsigned long long)di->di_eattr); | |
1438 | } | |
1439 |