Commit | Line | Data |
---|---|---|
d7e09d03 PT |
1 | /* |
2 | * GPL HEADER START | |
3 | * | |
4 | * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License version 2 only, | |
8 | * as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, but | |
11 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
13 | * General Public License version 2 for more details (a copy is included | |
14 | * in the LICENSE file that accompanied this code). | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * version 2 along with this program; If not, see | |
6a5b99a4 | 18 | * http://www.gnu.org/licenses/gpl-2.0.html |
d7e09d03 | 19 | * |
d7e09d03 PT |
20 | * GPL HEADER END |
21 | */ | |
22 | /* | |
23 | * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. | |
24 | * Use is subject to license terms. | |
25 | * | |
1dc563a6 | 26 | * Copyright (c) 2011, 2015, Intel Corporation. |
d7e09d03 PT |
27 | */ |
28 | /* | |
29 | * This file is part of Lustre, http://www.lustre.org/ | |
30 | * Lustre is a trademark of Sun Microsystems, Inc. | |
31 | * | |
32 | * lustre/llite/dir.c | |
33 | * | |
34 | * Directory code for lustre client. | |
35 | */ | |
36 | ||
37 | #include <linux/fs.h> | |
38 | #include <linux/pagemap.h> | |
39 | #include <linux/mm.h> | |
31982482 | 40 | #include <linux/uaccess.h> |
995c8b4a | 41 | #include <linux/buffer_head.h> /* for wait_on_buffer */ |
d7e09d03 | 42 | #include <linux/pagevec.h> |
2870cd10 | 43 | #include <linux/prefetch.h> |
d7e09d03 PT |
44 | |
45 | #define DEBUG_SUBSYSTEM S_LLITE | |
46 | ||
67a235f5 GKH |
47 | #include "../include/obd_support.h" |
48 | #include "../include/obd_class.h" | |
8877d3bf | 49 | #include "../include/lustre/lustre_ioctl.h" |
67a235f5 | 50 | #include "../include/lustre_lib.h" |
67a235f5 GKH |
51 | #include "../include/lustre_lite.h" |
52 | #include "../include/lustre_dlm.h" | |
53 | #include "../include/lustre_fid.h" | |
e2780478 | 54 | #include "../include/lustre_kernelcomm.h" |
d7e09d03 PT |
55 | #include "llite_internal.h" |
56 | ||
57 | /* | |
58 | * (new) readdir implementation overview. | |
59 | * | |
60 | * Original lustre readdir implementation cached exact copy of raw directory | |
61 | * pages on the client. These pages were indexed in client page cache by | |
62 | * logical offset in the directory file. This design, while very simple and | |
63 | * intuitive had some inherent problems: | |
64 | * | |
65 | * . it implies that byte offset to the directory entry serves as a | |
66 | * telldir(3)/seekdir(3) cookie, but that offset is not stable: in | |
67 | * ext3/htree directory entries may move due to splits, and more | |
68 | * importantly, | |
69 | * | |
70 | * . it is incompatible with the design of split directories for cmd3, | |
71 | * that assumes that names are distributed across nodes based on their | |
72 | * hash, and so readdir should be done in hash order. | |
73 | * | |
74 | * New readdir implementation does readdir in hash order, and uses hash of a | |
75 | * file name as a telldir/seekdir cookie. This led to number of complications: | |
76 | * | |
77 | * . hash is not unique, so it cannot be used to index cached directory | |
78 | * pages on the client (note, that it requires a whole pageful of hash | |
79 | * collided entries to cause two pages to have identical hashes); | |
80 | * | |
81 | * . hash is not unique, so it cannot, strictly speaking, be used as an | |
82 | * entry cookie. ext3/htree has the same problem and lustre implementation | |
83 | * mimics their solution: seekdir(hash) positions directory at the first | |
84 | * entry with the given hash. | |
85 | * | |
86 | * Client side. | |
87 | * | |
88 | * 0. caching | |
89 | * | |
90 | * Client caches directory pages using hash of the first entry as an index. As | |
91 | * noted above hash is not unique, so this solution doesn't work as is: | |
92 | * special processing is needed for "page hash chains" (i.e., sequences of | |
93 | * pages filled with entries all having the same hash value). | |
94 | * | |
95 | * First, such chains have to be detected. To this end, server returns to the | |
96 | * client the hash of the first entry on the page next to one returned. When | |
97 | * client detects that this hash is the same as hash of the first entry on the | |
98 | * returned page, page hash collision has to be handled. Pages in the | |
99 | * hash chain, except first one, are termed "overflow pages". | |
100 | * | |
101 | * Solution to index uniqueness problem is to not cache overflow | |
102 | * pages. Instead, when page hash collision is detected, all overflow pages | |
103 | * from emerging chain are immediately requested from the server and placed in | |
104 | * a special data structure (struct ll_dir_chain). This data structure is used | |
105 | * by ll_readdir() to process entries from overflow pages. When readdir | |
106 | * invocation finishes, overflow pages are discarded. If page hash collision | |
107 | * chain weren't completely processed, next call to readdir will again detect | |
108 | * page hash collision, again read overflow pages in, process next portion of | |
109 | * entries and again discard the pages. This is not as wasteful as it looks, | |
110 | * because, given reasonable hash, page hash collisions are extremely rare. | |
111 | * | |
112 | * 1. directory positioning | |
113 | * | |
114 | * When seekdir(hash) is called, original | |
115 | * | |
116 | * | |
117 | * | |
118 | * | |
119 | * | |
120 | * | |
121 | * | |
122 | * | |
123 | * Server. | |
124 | * | |
125 | * identification of and access to overflow pages | |
126 | * | |
127 | * page format | |
128 | * | |
129 | * Page in MDS_READPAGE RPC is packed in LU_PAGE_SIZE, and each page contains | |
130 | * a header lu_dirpage which describes the start/end hash, and whether this | |
131 | * page is empty (contains no dir entry) or hash collide with next page. | |
132 | * After client receives reply, several pages will be integrated into dir page | |
ea1754a0 KS |
133 | * in PAGE_SIZE (if PAGE_SIZE greater than LU_PAGE_SIZE), and the lu_dirpage |
134 | * for this integrated page will be adjusted. See lmv_adjust_dirpages(). | |
d7e09d03 PT |
135 | * |
136 | */ | |
4f76f0ec | 137 | struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data, |
138 | __u64 offset, struct ll_dir_chain *chain) | |
d7e09d03 | 139 | { |
4f76f0ec | 140 | struct md_callback cb_op; |
d7e09d03 | 141 | struct page *page; |
d7e09d03 | 142 | int rc; |
d7e09d03 | 143 | |
4f76f0ec | 144 | cb_op.md_blocking_ast = ll_md_blocking_ast; |
145 | rc = md_read_page(ll_i2mdexp(dir), op_data, &cb_op, offset, &page); | |
146 | if (rc) | |
147 | return ERR_PTR(rc); | |
d7e09d03 | 148 | |
4f76f0ec | 149 | return page; |
d7e09d03 PT |
150 | } |
151 | ||
77a782ab | 152 | void ll_release_page(struct inode *inode, struct page *page, bool remove) |
d7e09d03 PT |
153 | { |
154 | kunmap(page); | |
4f76f0ec | 155 | |
156 | /* | |
157 | * Always remove the page for striped dir, because the page is | |
158 | * built from temporarily in LMV layer | |
159 | */ | |
160 | if (inode && S_ISDIR(inode->i_mode) && | |
161 | ll_i2info(inode)->lli_lsm_md) { | |
162 | __free_page(page); | |
163 | return; | |
164 | } | |
165 | ||
d7e09d03 PT |
166 | if (remove) { |
167 | lock_page(page); | |
6e16818b | 168 | if (likely(page->mapping)) |
d7e09d03 PT |
169 | truncate_complete_page(page->mapping, page); |
170 | unlock_page(page); | |
171 | } | |
09cbfeaf | 172 | put_page(page); |
d7e09d03 PT |
173 | } |
174 | ||
0097dcab OD |
175 | /** |
176 | * return IF_* type for given lu_dirent entry. | |
177 | * IF_* flag shld be converted to particular OS file type in | |
178 | * platform llite module. | |
179 | */ | |
180 | static __u16 ll_dirent_type_get(struct lu_dirent *ent) | |
181 | { | |
182 | __u16 type = 0; | |
183 | struct luda_type *lt; | |
184 | int len = 0; | |
185 | ||
186 | if (le32_to_cpu(ent->lde_attrs) & LUDA_TYPE) { | |
187 | const unsigned int align = sizeof(struct luda_type) - 1; | |
188 | ||
189 | len = le16_to_cpu(ent->lde_namelen); | |
190 | len = (len + align) & ~align; | |
191 | lt = (void *)ent->lde_name + len; | |
192 | type = IFTODT(le16_to_cpu(lt->lt_type)); | |
193 | } | |
194 | return type; | |
195 | } | |
196 | ||
218ba485 | 197 | int ll_dir_read(struct inode *inode, __u64 *ppos, struct md_op_data *op_data, |
307bef74 | 198 | struct dir_context *ctx) |
d7e09d03 | 199 | { |
d7e09d03 | 200 | struct ll_sb_info *sbi = ll_i2sbi(inode); |
218ba485 | 201 | __u64 pos = *ppos; |
67537558 | 202 | int is_api32 = ll_need_32bit_api(sbi); |
203 | int is_hash64 = sbi->ll_flags & LL_SBI_64BIT_HASH; | |
d7e09d03 PT |
204 | struct page *page; |
205 | struct ll_dir_chain chain; | |
13810b0f | 206 | bool done = false; |
d7e09d03 | 207 | int rc = 0; |
d7e09d03 PT |
208 | |
209 | ll_dir_chain_init(&chain); | |
210 | ||
3978732f | 211 | page = ll_get_dir_page(inode, op_data, pos, &chain); |
d7e09d03 PT |
212 | |
213 | while (rc == 0 && !done) { | |
214 | struct lu_dirpage *dp; | |
215 | struct lu_dirent *ent; | |
b1f048c1 | 216 | __u64 hash; |
217 | __u64 next; | |
218 | ||
219 | if (IS_ERR(page)) { | |
220 | rc = PTR_ERR(page); | |
221 | break; | |
222 | } | |
223 | ||
224 | hash = MDS_DIR_END_OFF; | |
225 | dp = page_address(page); | |
226 | for (ent = lu_dirent_start(dp); ent && !done; | |
227 | ent = lu_dirent_next(ent)) { | |
228 | __u16 type; | |
229 | int namelen; | |
230 | struct lu_fid fid; | |
231 | __u64 lhash; | |
232 | __u64 ino; | |
d7e09d03 | 233 | |
b1f048c1 | 234 | hash = le64_to_cpu(ent->lde_hash); |
235 | if (hash < pos) | |
d7e09d03 | 236 | /* |
b1f048c1 | 237 | * Skip until we find target hash |
238 | * value. | |
d7e09d03 | 239 | */ |
b1f048c1 | 240 | continue; |
d7e09d03 | 241 | |
b1f048c1 | 242 | namelen = le16_to_cpu(ent->lde_namelen); |
243 | if (namelen == 0) | |
244 | /* | |
245 | * Skip dummy record. | |
0b09d381 | 246 | */ |
b1f048c1 | 247 | continue; |
248 | ||
67537558 | 249 | if (is_api32 && is_hash64) |
b1f048c1 | 250 | lhash = hash >> 32; |
251 | else | |
252 | lhash = hash; | |
253 | fid_le_to_cpu(&fid, &ent->lde_fid); | |
67537558 | 254 | ino = cl_fid_build_ino(&fid, is_api32); |
b1f048c1 | 255 | type = ll_dirent_type_get(ent); |
256 | ctx->pos = lhash; | |
257 | /* For 'll_nfs_get_name_filldir()', it will try | |
258 | * to access the 'ent' through its 'lde_name', | |
259 | * so the parameter 'name' for 'ctx->actor()' | |
260 | * must be part of the 'ent'. | |
261 | */ | |
262 | done = !dir_emit(ctx, ent->lde_name, | |
263 | namelen, ino, type); | |
264 | } | |
fb659ca1 | 265 | |
fee8cc87 | 266 | if (done) { |
b1f048c1 | 267 | pos = hash; |
77a782ab | 268 | ll_release_page(inode, page, false); |
fee8cc87 | 269 | break; |
270 | } | |
271 | ||
272 | next = le64_to_cpu(dp->ldp_hash_end); | |
273 | pos = next; | |
274 | if (pos == MDS_DIR_END_OFF) { | |
275 | /* | |
276 | * End of directory reached. | |
277 | */ | |
278 | done = 1; | |
77a782ab | 279 | ll_release_page(inode, page, false); |
26f5c084 | 280 | } else { |
fee8cc87 | 281 | /* |
282 | * Normal case: continue to the next | |
283 | * page. | |
284 | */ | |
006e4dcd | 285 | ll_release_page(inode, page, |
fee8cc87 | 286 | le32_to_cpu(dp->ldp_flags) & |
287 | LDF_COLLIDE); | |
288 | next = pos; | |
3978732f | 289 | page = ll_get_dir_page(inode, op_data, pos, |
fee8cc87 | 290 | &chain); |
d7e09d03 PT |
291 | } |
292 | } | |
293 | ||
0b09d381 | 294 | ctx->pos = pos; |
d7e09d03 | 295 | ll_dir_chain_fini(&chain); |
0a3bdb00 | 296 | return rc; |
d7e09d03 PT |
297 | } |
298 | ||
0b09d381 | 299 | static int ll_readdir(struct file *filp, struct dir_context *ctx) |
d7e09d03 | 300 | { |
2a8a3597 | 301 | struct inode *inode = file_inode(filp); |
d7e09d03 PT |
302 | struct ll_file_data *lfd = LUSTRE_FPRIVATE(filp); |
303 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
0fcd869f | 304 | __u64 pos = lfd ? lfd->lfd_pos : 0; |
d7e09d03 PT |
305 | int hash64 = sbi->ll_flags & LL_SBI_64BIT_HASH; |
306 | int api32 = ll_need_32bit_api(sbi); | |
307bef74 | 307 | struct md_op_data *op_data; |
d7e09d03 | 308 | int rc; |
d7e09d03 | 309 | |
969d25d4 | 310 | CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p) pos/size %lu/%llu 32bit_api %d\n", |
97a075cd JN |
311 | PFID(ll_inode2fid(inode)), inode, (unsigned long)pos, |
312 | i_size_read(inode), api32); | |
d7e09d03 | 313 | |
0fcd869f | 314 | if (pos == MDS_DIR_END_OFF) { |
d7e09d03 PT |
315 | /* |
316 | * end-of-file. | |
317 | */ | |
34e1f2bb JL |
318 | rc = 0; |
319 | goto out; | |
320 | } | |
d7e09d03 | 321 | |
307bef74 | 322 | op_data = ll_prep_md_op_data(NULL, inode, inode, NULL, 0, 0, |
323 | LUSTRE_OPC_ANY, inode); | |
324 | if (IS_ERR(op_data)) { | |
325 | rc = PTR_ERR(op_data); | |
326 | goto out; | |
327 | } | |
328 | ||
ef21b1fb | 329 | if (unlikely(op_data->op_mea1)) { |
330 | /* | |
331 | * This is only needed for striped dir to fill .., | |
332 | * see lmv_read_page | |
333 | */ | |
334 | if (file_dentry(filp)->d_parent && | |
335 | file_dentry(filp)->d_parent->d_inode) { | |
336 | __u64 ibits = MDS_INODELOCK_UPDATE; | |
337 | struct inode *parent; | |
338 | ||
339 | parent = file_dentry(filp)->d_parent->d_inode; | |
340 | if (ll_have_md_lock(parent, &ibits, LCK_MINMODE)) | |
341 | op_data->op_fid3 = *ll_inode2fid(parent); | |
342 | } | |
343 | ||
344 | /* | |
345 | * If it can not find in cache, do lookup .. on the master | |
346 | * object | |
347 | */ | |
348 | if (fid_is_zero(&op_data->op_fid3)) { | |
349 | rc = ll_dir_get_parent_fid(inode, &op_data->op_fid3); | |
350 | if (rc) { | |
351 | ll_finish_md_op_data(op_data); | |
352 | return rc; | |
353 | } | |
354 | } | |
355 | } | |
bce1bbf4 | 356 | op_data->op_max_pages = sbi->ll_md_brw_pages; |
0fcd869f | 357 | ctx->pos = pos; |
218ba485 | 358 | rc = ll_dir_read(inode, &pos, op_data, ctx); |
359 | pos = ctx->pos; | |
0fcd869f | 360 | if (lfd) |
218ba485 | 361 | lfd->lfd_pos = pos; |
362 | ||
363 | if (pos == MDS_DIR_END_OFF) { | |
d7e09d03 | 364 | if (api32) |
218ba485 | 365 | pos = LL_DIR_END_OFF_32BIT; |
d7e09d03 | 366 | else |
218ba485 | 367 | pos = LL_DIR_END_OFF; |
d7e09d03 PT |
368 | } else { |
369 | if (api32 && hash64) | |
218ba485 | 370 | pos >>= 32; |
d7e09d03 | 371 | } |
218ba485 | 372 | ctx->pos = pos; |
307bef74 | 373 | ll_finish_md_op_data(op_data); |
d7e09d03 | 374 | filp->f_version = inode->i_version; |
d7e09d03 PT |
375 | |
376 | out: | |
377 | if (!rc) | |
378 | ll_stats_ops_tally(sbi, LPROC_LL_READDIR, 1); | |
379 | ||
0a3bdb00 | 380 | return rc; |
d7e09d03 PT |
381 | } |
382 | ||
2d95f10e | 383 | static int ll_send_mgc_param(struct obd_export *mgc, char *string) |
d7e09d03 PT |
384 | { |
385 | struct mgs_send_param *msp; | |
386 | int rc = 0; | |
387 | ||
496a51bd | 388 | msp = kzalloc(sizeof(*msp), GFP_NOFS); |
d7e09d03 PT |
389 | if (!msp) |
390 | return -ENOMEM; | |
391 | ||
9563fe8a | 392 | strlcpy(msp->mgs_param, string, sizeof(msp->mgs_param)); |
d7e09d03 PT |
393 | rc = obd_set_info_async(NULL, mgc, sizeof(KEY_SET_INFO), KEY_SET_INFO, |
394 | sizeof(struct mgs_send_param), msp, NULL); | |
395 | if (rc) | |
396 | CERROR("Failed to set parameter: %d\n", rc); | |
97903a26 | 397 | kfree(msp); |
d7e09d03 PT |
398 | |
399 | return rc; | |
400 | } | |
401 | ||
d07d4cb8 | 402 | /** |
403 | * Create striped directory with specified stripe(@lump) | |
404 | * | |
405 | * param[in] parent the parent of the directory. | |
406 | * param[in] lump the specified stripes. | |
407 | * param[in] dirname the name of the directory. | |
408 | * param[in] mode the specified mode of the directory. | |
409 | * | |
410 | * retval =0 if striped directory is being created successfully. | |
411 | * <0 if the creation is failed. | |
412 | */ | |
413 | static int ll_dir_setdirstripe(struct inode *parent, struct lmv_user_md *lump, | |
414 | const char *dirname, umode_t mode) | |
d7e09d03 PT |
415 | { |
416 | struct ptlrpc_request *request = NULL; | |
417 | struct md_op_data *op_data; | |
d07d4cb8 | 418 | struct ll_sb_info *sbi = ll_i2sbi(parent); |
d7e09d03 PT |
419 | int err; |
420 | ||
2de35386 | 421 | if (unlikely(lump->lum_magic != LMV_USER_MAGIC)) |
422 | return -EINVAL; | |
423 | ||
2de35386 | 424 | CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p) name %s stripe_offset %d, stripe_count: %u\n", |
d07d4cb8 | 425 | PFID(ll_inode2fid(parent)), parent, dirname, |
2de35386 | 426 | (int)lump->lum_stripe_offset, lump->lum_stripe_count); |
427 | ||
428 | if (lump->lum_magic != cpu_to_le32(LMV_USER_MAGIC)) | |
429 | lustre_swab_lmv_user_md(lump); | |
430 | ||
d07d4cb8 | 431 | if (!IS_POSIXACL(parent) || !exp_connect_umask(ll_i2mdexp(parent))) |
432 | mode &= ~current_umask(); | |
433 | mode = (mode & (S_IRWXUGO | S_ISVTX)) | S_IFDIR; | |
434 | op_data = ll_prep_md_op_data(NULL, parent, NULL, dirname, | |
435 | strlen(dirname), mode, LUSTRE_OPC_MKDIR, | |
d7e09d03 | 436 | lump); |
34e1f2bb JL |
437 | if (IS_ERR(op_data)) { |
438 | err = PTR_ERR(op_data); | |
439 | goto err_exit; | |
440 | } | |
d7e09d03 PT |
441 | |
442 | op_data->op_cli_flags |= CLI_SET_MEA; | |
443 | err = md_create(sbi->ll_md_exp, op_data, lump, sizeof(*lump), mode, | |
4b1a25f0 PT |
444 | from_kuid(&init_user_ns, current_fsuid()), |
445 | from_kgid(&init_user_ns, current_fsgid()), | |
d7e09d03 PT |
446 | cfs_curproc_cap_pack(), 0, &request); |
447 | ll_finish_md_op_data(op_data); | |
448 | if (err) | |
34e1f2bb | 449 | goto err_exit; |
d7e09d03 PT |
450 | err_exit: |
451 | ptlrpc_req_finished(request); | |
452 | return err; | |
453 | } | |
454 | ||
455 | int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, | |
456 | int set_default) | |
457 | { | |
458 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
459 | struct md_op_data *op_data; | |
460 | struct ptlrpc_request *req = NULL; | |
461 | int rc = 0; | |
462 | struct lustre_sb_info *lsi = s2lsi(inode->i_sb); | |
463 | struct obd_device *mgc = lsi->lsi_mgc; | |
464 | int lum_size; | |
d7e09d03 | 465 | |
6e16818b | 466 | if (lump) { |
d7e09d03 PT |
467 | /* |
468 | * This is coming from userspace, so should be in | |
469 | * local endian. But the MDS would like it in little | |
470 | * endian, so we swab it before we send it. | |
471 | */ | |
472 | switch (lump->lmm_magic) { | |
473 | case LOV_USER_MAGIC_V1: { | |
474 | if (lump->lmm_magic != cpu_to_le32(LOV_USER_MAGIC_V1)) | |
475 | lustre_swab_lov_user_md_v1(lump); | |
476 | lum_size = sizeof(struct lov_user_md_v1); | |
477 | break; | |
478 | } | |
479 | case LOV_USER_MAGIC_V3: { | |
480 | if (lump->lmm_magic != cpu_to_le32(LOV_USER_MAGIC_V3)) | |
481 | lustre_swab_lov_user_md_v3( | |
482 | (struct lov_user_md_v3 *)lump); | |
483 | lum_size = sizeof(struct lov_user_md_v3); | |
484 | break; | |
485 | } | |
6e23ea98 | 486 | case LMV_USER_MAGIC: { |
487 | if (lump->lmm_magic != cpu_to_le32(LMV_USER_MAGIC)) | |
488 | lustre_swab_lmv_user_md( | |
489 | (struct lmv_user_md *)lump); | |
490 | lum_size = sizeof(struct lmv_user_md); | |
491 | break; | |
492 | } | |
d7e09d03 | 493 | default: { |
2d00bd17 JP |
494 | CDEBUG(D_IOCTL, "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n", |
495 | lump->lmm_magic, LOV_USER_MAGIC_V1, | |
496 | LOV_USER_MAGIC_V3); | |
0a3bdb00 | 497 | return -EINVAL; |
d7e09d03 PT |
498 | } |
499 | } | |
500 | } else { | |
501 | lum_size = sizeof(struct lov_user_md_v1); | |
502 | } | |
503 | ||
504 | op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0, | |
505 | LUSTRE_OPC_ANY, NULL); | |
506 | if (IS_ERR(op_data)) | |
0a3bdb00 | 507 | return PTR_ERR(op_data); |
d7e09d03 | 508 | |
d7e09d03 PT |
509 | /* swabbing is done in lov_setstripe() on server side */ |
510 | rc = md_setattr(sbi->ll_md_exp, op_data, lump, lum_size, | |
511 | NULL, 0, &req, NULL); | |
512 | ll_finish_md_op_data(op_data); | |
513 | ptlrpc_req_finished(req); | |
514 | if (rc) { | |
515 | if (rc != -EPERM && rc != -EACCES) | |
516 | CERROR("mdc_setattr fails: rc = %d\n", rc); | |
517 | } | |
518 | ||
519 | /* In the following we use the fact that LOV_USER_MAGIC_V1 and | |
c0894c6c OD |
520 | * LOV_USER_MAGIC_V3 have the same initial fields so we do not |
521 | * need to make the distinction between the 2 versions | |
522 | */ | |
d7e09d03 PT |
523 | if (set_default && mgc->u.cli.cl_mgc_mgsexp) { |
524 | char *param = NULL; | |
525 | char *buf; | |
526 | ||
496a51bd | 527 | param = kzalloc(MGS_PARAM_MAXLEN, GFP_NOFS); |
57876fd8 JL |
528 | if (!param) |
529 | return -ENOMEM; | |
d7e09d03 PT |
530 | |
531 | buf = param; | |
532 | /* Get fsname and assume devname to be -MDT0000. */ | |
533 | ll_get_fsname(inode->i_sb, buf, MTI_NAME_MAXLEN); | |
534 | strcat(buf, "-MDT0000.lov"); | |
535 | buf += strlen(buf); | |
536 | ||
537 | /* Set root stripesize */ | |
538 | sprintf(buf, ".stripesize=%u", | |
539 | lump ? le32_to_cpu(lump->lmm_stripe_size) : 0); | |
540 | rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); | |
541 | if (rc) | |
34e1f2bb | 542 | goto end; |
d7e09d03 PT |
543 | |
544 | /* Set root stripecount */ | |
545 | sprintf(buf, ".stripecount=%hd", | |
546 | lump ? le16_to_cpu(lump->lmm_stripe_count) : 0); | |
547 | rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); | |
548 | if (rc) | |
34e1f2bb | 549 | goto end; |
d7e09d03 PT |
550 | |
551 | /* Set root stripeoffset */ | |
552 | sprintf(buf, ".stripeoffset=%hd", | |
553 | lump ? le16_to_cpu(lump->lmm_stripe_offset) : | |
554 | (typeof(lump->lmm_stripe_offset))(-1)); | |
555 | rc = ll_send_mgc_param(mgc->u.cli.cl_mgc_mgsexp, param); | |
556 | ||
557 | end: | |
57876fd8 | 558 | kfree(param); |
d7e09d03 | 559 | } |
0a3bdb00 | 560 | return rc; |
d7e09d03 PT |
561 | } |
562 | ||
6e23ea98 | 563 | /** |
564 | * This function will be used to get default LOV/LMV/Default LMV | |
565 | * @valid will be used to indicate which stripe it will retrieve | |
566 | * OBD_MD_MEA LMV stripe EA | |
567 | * OBD_MD_DEFAULT_MEA Default LMV stripe EA | |
568 | * otherwise Default LOV EA. | |
569 | * Each time, it can only retrieve 1 stripe EA | |
570 | **/ | |
571 | int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size, | |
572 | struct ptlrpc_request **request, u64 valid) | |
d7e09d03 PT |
573 | { |
574 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
575 | struct mdt_body *body; | |
576 | struct lov_mds_md *lmm = NULL; | |
577 | struct ptlrpc_request *req = NULL; | |
578 | int rc, lmmsize; | |
579 | struct md_op_data *op_data; | |
580 | ||
6e23ea98 | 581 | rc = ll_get_max_mdsize(sbi, &lmmsize); |
d7e09d03 | 582 | if (rc) |
0a3bdb00 | 583 | return rc; |
d7e09d03 PT |
584 | |
585 | op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, | |
586 | 0, lmmsize, LUSTRE_OPC_ANY, | |
587 | NULL); | |
588 | if (IS_ERR(op_data)) | |
0a3bdb00 | 589 | return PTR_ERR(op_data); |
d7e09d03 PT |
590 | |
591 | op_data->op_valid = OBD_MD_FLEASIZE | OBD_MD_FLDIREA; | |
592 | rc = md_getattr(sbi->ll_md_exp, op_data, &req); | |
593 | ll_finish_md_op_data(op_data); | |
594 | if (rc < 0) { | |
97a075cd JN |
595 | CDEBUG(D_INFO, "md_getattr failed on inode "DFID": rc %d\n", |
596 | PFID(ll_inode2fid(inode)), rc); | |
34e1f2bb | 597 | goto out; |
d7e09d03 PT |
598 | } |
599 | ||
600 | body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY); | |
d7e09d03 | 601 | |
2e1b5b8b | 602 | lmmsize = body->mbo_eadatasize; |
d7e09d03 | 603 | |
2e1b5b8b | 604 | if (!(body->mbo_valid & (OBD_MD_FLEASIZE | OBD_MD_FLDIREA)) || |
d7e09d03 | 605 | lmmsize == 0) { |
34e1f2bb JL |
606 | rc = -ENODATA; |
607 | goto out; | |
d7e09d03 PT |
608 | } |
609 | ||
610 | lmm = req_capsule_server_sized_get(&req->rq_pill, | |
611 | &RMF_MDT_MD, lmmsize); | |
6e23ea98 | 612 | LASSERT(lmm); |
d7e09d03 PT |
613 | |
614 | /* | |
615 | * This is coming from the MDS, so is probably in | |
616 | * little endian. We convert it to host endian before | |
617 | * passing it to userspace. | |
618 | */ | |
619 | /* We don't swab objects for directories */ | |
620 | switch (le32_to_cpu(lmm->lmm_magic)) { | |
621 | case LOV_MAGIC_V1: | |
1f6eaf83 | 622 | if (cpu_to_le32(LOV_MAGIC) != LOV_MAGIC) |
d7e09d03 PT |
623 | lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lmm); |
624 | break; | |
625 | case LOV_MAGIC_V3: | |
1f6eaf83 | 626 | if (cpu_to_le32(LOV_MAGIC) != LOV_MAGIC) |
d7e09d03 PT |
627 | lustre_swab_lov_user_md_v3((struct lov_user_md_v3 *)lmm); |
628 | break; | |
6e23ea98 | 629 | case LMV_USER_MAGIC: |
630 | if (cpu_to_le32(LMV_USER_MAGIC) != LMV_USER_MAGIC) | |
631 | lustre_swab_lmv_user_md((struct lmv_user_md *)lmm); | |
632 | break; | |
d7e09d03 PT |
633 | default: |
634 | CERROR("unknown magic: %lX\n", (unsigned long)lmm->lmm_magic); | |
635 | rc = -EPROTO; | |
636 | } | |
637 | out: | |
6e23ea98 | 638 | *plmm = lmm; |
639 | *plmm_size = lmmsize; | |
d7e09d03 PT |
640 | *request = req; |
641 | return rc; | |
642 | } | |
643 | ||
79496845 | 644 | int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid) |
d7e09d03 | 645 | { |
d7e09d03 | 646 | struct md_op_data *op_data; |
6e23ea98 | 647 | int mdt_index, rc; |
d7e09d03 | 648 | |
6e23ea98 | 649 | op_data = kzalloc(sizeof(*op_data), GFP_NOFS); |
650 | if (!op_data) | |
651 | return -ENOMEM; | |
d7e09d03 PT |
652 | |
653 | op_data->op_flags |= MF_GET_MDT_IDX; | |
6e23ea98 | 654 | op_data->op_fid1 = *fid; |
d7e09d03 | 655 | rc = md_getattr(sbi->ll_md_exp, op_data, NULL); |
6e23ea98 | 656 | mdt_index = op_data->op_mds; |
657 | kvfree(op_data); | |
658 | if (rc < 0) | |
0a3bdb00 | 659 | return rc; |
6e23ea98 | 660 | |
661 | return mdt_index; | |
662 | } | |
663 | ||
664 | /* | |
665 | * Get MDT index for the inode. | |
666 | */ | |
667 | int ll_get_mdt_idx(struct inode *inode) | |
668 | { | |
669 | return ll_get_mdt_idx_by_fid(ll_i2sbi(inode), ll_inode2fid(inode)); | |
d7e09d03 PT |
670 | } |
671 | ||
672 | /** | |
673 | * Generic handler to do any pre-copy work. | |
674 | * | |
9c379663 | 675 | * It sends a first hsm_progress (with extent length == 0) to coordinator as a |
d7e09d03 PT |
676 | * first information for it that real work has started. |
677 | * | |
678 | * Moreover, for a ARCHIVE request, it will sample the file data version and | |
679 | * store it in \a copy. | |
680 | * | |
681 | * \return 0 on success. | |
682 | */ | |
683 | static int ll_ioc_copy_start(struct super_block *sb, struct hsm_copy *copy) | |
684 | { | |
685 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
686 | struct hsm_progress_kernel hpk; | |
687 | int rc; | |
d7e09d03 PT |
688 | |
689 | /* Forge a hsm_progress based on data from copy. */ | |
690 | hpk.hpk_fid = copy->hc_hai.hai_fid; | |
691 | hpk.hpk_cookie = copy->hc_hai.hai_cookie; | |
692 | hpk.hpk_extent.offset = copy->hc_hai.hai_extent.offset; | |
693 | hpk.hpk_extent.length = 0; | |
694 | hpk.hpk_flags = 0; | |
695 | hpk.hpk_errval = 0; | |
696 | hpk.hpk_data_version = 0; | |
697 | ||
d7e09d03 PT |
698 | /* For archive request, we need to read the current file version. */ |
699 | if (copy->hc_hai.hai_action == HSMA_ARCHIVE) { | |
700 | struct inode *inode; | |
701 | __u64 data_version = 0; | |
702 | ||
703 | /* Get inode for this fid */ | |
704 | inode = search_inode_for_lustre(sb, ©->hc_hai.hai_fid); | |
705 | if (IS_ERR(inode)) { | |
706 | hpk.hpk_flags |= HP_FLAG_RETRY; | |
707 | /* hpk_errval is >= 0 */ | |
708 | hpk.hpk_errval = -PTR_ERR(inode); | |
34e1f2bb JL |
709 | rc = PTR_ERR(inode); |
710 | goto progress; | |
d7e09d03 PT |
711 | } |
712 | ||
713 | /* Read current file data version */ | |
e1798006 | 714 | rc = ll_data_version(inode, &data_version, LL_DV_RD_FLUSH); |
d7e09d03 PT |
715 | iput(inode); |
716 | if (rc != 0) { | |
717 | CDEBUG(D_HSM, "Could not read file data version of " | |
55f5a824 | 718 | DFID" (rc = %d). Archive request (%#llx) could not be done.\n", |
d7e09d03 PT |
719 | PFID(©->hc_hai.hai_fid), rc, |
720 | copy->hc_hai.hai_cookie); | |
721 | hpk.hpk_flags |= HP_FLAG_RETRY; | |
722 | /* hpk_errval must be >= 0 */ | |
723 | hpk.hpk_errval = -rc; | |
34e1f2bb | 724 | goto progress; |
d7e09d03 PT |
725 | } |
726 | ||
9c379663 | 727 | /* Store in the hsm_copy for later copytool use. |
c0894c6c OD |
728 | * Always modified even if no lsm. |
729 | */ | |
d7e09d03 PT |
730 | copy->hc_data_version = data_version; |
731 | } | |
732 | ||
733 | progress: | |
c5300d55 AD |
734 | /* On error, the request should be considered as completed */ |
735 | if (hpk.hpk_errval > 0) | |
736 | hpk.hpk_flags |= HP_FLAG_COMPLETED; | |
d7e09d03 PT |
737 | rc = obd_iocontrol(LL_IOC_HSM_PROGRESS, sbi->ll_md_exp, sizeof(hpk), |
738 | &hpk, NULL); | |
739 | ||
0a3bdb00 | 740 | return rc; |
d7e09d03 PT |
741 | } |
742 | ||
743 | /** | |
744 | * Generic handler to do any post-copy work. | |
745 | * | |
746 | * It will send the last hsm_progress update to coordinator to inform it | |
747 | * that copy is finished and whether it was successful or not. | |
748 | * | |
749 | * Moreover, | |
750 | * - for ARCHIVE request, it will sample the file data version and compare it | |
751 | * with the version saved in ll_ioc_copy_start(). If they do not match, copy | |
752 | * will be considered as failed. | |
753 | * - for RESTORE request, it will sample the file data version and send it to | |
754 | * coordinator which is useful if the file was imported as 'released'. | |
755 | * | |
756 | * \return 0 on success. | |
757 | */ | |
758 | static int ll_ioc_copy_end(struct super_block *sb, struct hsm_copy *copy) | |
759 | { | |
760 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
761 | struct hsm_progress_kernel hpk; | |
762 | int rc; | |
d7e09d03 PT |
763 | |
764 | /* If you modify the logic here, also check llapi_hsm_copy_end(). */ | |
765 | /* Take care: copy->hc_hai.hai_action, len, gid and data are not | |
766 | * initialized if copy_end was called with copy == NULL. | |
767 | */ | |
768 | ||
769 | /* Forge a hsm_progress based on data from copy. */ | |
770 | hpk.hpk_fid = copy->hc_hai.hai_fid; | |
771 | hpk.hpk_cookie = copy->hc_hai.hai_cookie; | |
772 | hpk.hpk_extent = copy->hc_hai.hai_extent; | |
773 | hpk.hpk_flags = copy->hc_flags | HP_FLAG_COMPLETED; | |
774 | hpk.hpk_errval = copy->hc_errval; | |
775 | hpk.hpk_data_version = 0; | |
776 | ||
777 | /* For archive request, we need to check the file data was not changed. | |
778 | * | |
779 | * For restore request, we need to send the file data version, this is | |
780 | * useful when the file was created using hsm_import. | |
781 | */ | |
782 | if (((copy->hc_hai.hai_action == HSMA_ARCHIVE) || | |
783 | (copy->hc_hai.hai_action == HSMA_RESTORE)) && | |
784 | (copy->hc_errval == 0)) { | |
785 | struct inode *inode; | |
786 | __u64 data_version = 0; | |
787 | ||
788 | /* Get lsm for this fid */ | |
789 | inode = search_inode_for_lustre(sb, ©->hc_hai.hai_fid); | |
790 | if (IS_ERR(inode)) { | |
791 | hpk.hpk_flags |= HP_FLAG_RETRY; | |
792 | /* hpk_errval must be >= 0 */ | |
793 | hpk.hpk_errval = -PTR_ERR(inode); | |
34e1f2bb JL |
794 | rc = PTR_ERR(inode); |
795 | goto progress; | |
d7e09d03 PT |
796 | } |
797 | ||
e1798006 | 798 | rc = ll_data_version(inode, &data_version, LL_DV_RD_FLUSH); |
d7e09d03 PT |
799 | iput(inode); |
800 | if (rc) { | |
2d00bd17 | 801 | CDEBUG(D_HSM, "Could not read file data version. Request could not be confirmed.\n"); |
d7e09d03 PT |
802 | if (hpk.hpk_errval == 0) |
803 | hpk.hpk_errval = -rc; | |
34e1f2bb | 804 | goto progress; |
d7e09d03 PT |
805 | } |
806 | ||
9c379663 | 807 | /* Store in the hsm_copy for later copytool use. |
c0894c6c OD |
808 | * Always modified even if no lsm. |
809 | */ | |
d7e09d03 PT |
810 | hpk.hpk_data_version = data_version; |
811 | ||
812 | /* File could have been stripped during archiving, so we need | |
c0894c6c OD |
813 | * to check anyway. |
814 | */ | |
d7e09d03 PT |
815 | if ((copy->hc_hai.hai_action == HSMA_ARCHIVE) && |
816 | (copy->hc_data_version != data_version)) { | |
2d00bd17 | 817 | CDEBUG(D_HSM, "File data version mismatched. File content was changed during archiving. " |
55f5a824 | 818 | DFID", start:%#llx current:%#llx\n", |
d7e09d03 PT |
819 | PFID(©->hc_hai.hai_fid), |
820 | copy->hc_data_version, data_version); | |
821 | /* File was changed, send error to cdt. Do not ask for | |
822 | * retry because if a file is modified frequently, | |
823 | * the cdt will loop on retried archive requests. | |
824 | * The policy engine will ask for a new archive later | |
825 | * when the file will not be modified for some tunable | |
c0894c6c OD |
826 | * time |
827 | */ | |
d7e09d03 PT |
828 | /* we do not notify caller */ |
829 | hpk.hpk_flags &= ~HP_FLAG_RETRY; | |
830 | /* hpk_errval must be >= 0 */ | |
831 | hpk.hpk_errval = EBUSY; | |
832 | } | |
d7e09d03 PT |
833 | } |
834 | ||
835 | progress: | |
836 | rc = obd_iocontrol(LL_IOC_HSM_PROGRESS, sbi->ll_md_exp, sizeof(hpk), | |
837 | &hpk, NULL); | |
838 | ||
0a3bdb00 | 839 | return rc; |
d7e09d03 PT |
840 | } |
841 | ||
b0337d6c JH |
842 | static int copy_and_ioctl(int cmd, struct obd_export *exp, |
843 | const void __user *data, size_t size) | |
d7e09d03 | 844 | { |
b0337d6c | 845 | void *copy; |
d7e09d03 PT |
846 | int rc; |
847 | ||
c6c7a170 TH |
848 | copy = memdup_user(data, size); |
849 | if (IS_ERR(copy)) | |
850 | return PTR_ERR(copy); | |
b0337d6c JH |
851 | |
852 | rc = obd_iocontrol(cmd, exp, size, copy, NULL); | |
97903a26 | 853 | kfree(copy); |
b0337d6c | 854 | |
d7e09d03 PT |
855 | return rc; |
856 | } | |
857 | ||
858 | static int quotactl_ioctl(struct ll_sb_info *sbi, struct if_quotactl *qctl) | |
859 | { | |
860 | int cmd = qctl->qc_cmd; | |
861 | int type = qctl->qc_type; | |
862 | int id = qctl->qc_id; | |
863 | int valid = qctl->qc_valid; | |
864 | int rc = 0; | |
d7e09d03 PT |
865 | |
866 | switch (cmd) { | |
867 | case LUSTRE_Q_INVALIDATE: | |
868 | case LUSTRE_Q_FINVALIDATE: | |
869 | case Q_QUOTAON: | |
870 | case Q_QUOTAOFF: | |
871 | case Q_SETQUOTA: | |
872 | case Q_SETINFO: | |
341f1f0a | 873 | if (!capable(CFS_CAP_SYS_ADMIN)) |
0a3bdb00 | 874 | return -EPERM; |
d7e09d03 PT |
875 | break; |
876 | case Q_GETQUOTA: | |
4b1a25f0 | 877 | if (((type == USRQUOTA && |
8b9e418c | 878 | !uid_eq(current_euid(), make_kuid(&init_user_ns, id))) || |
4b1a25f0 PT |
879 | (type == GRPQUOTA && |
880 | !in_egroup_p(make_kgid(&init_user_ns, id)))) && | |
341f1f0a | 881 | !capable(CFS_CAP_SYS_ADMIN)) |
0a3bdb00 | 882 | return -EPERM; |
d7e09d03 PT |
883 | break; |
884 | case Q_GETINFO: | |
885 | break; | |
886 | default: | |
887 | CERROR("unsupported quotactl op: %#x\n", cmd); | |
0a3bdb00 | 888 | return -ENOTTY; |
d7e09d03 PT |
889 | } |
890 | ||
891 | if (valid != QC_GENERAL) { | |
d7e09d03 PT |
892 | if (cmd == Q_GETINFO) |
893 | qctl->qc_cmd = Q_GETOINFO; | |
894 | else if (cmd == Q_GETQUOTA) | |
895 | qctl->qc_cmd = Q_GETOQUOTA; | |
896 | else | |
0a3bdb00 | 897 | return -EINVAL; |
d7e09d03 PT |
898 | |
899 | switch (valid) { | |
900 | case QC_MDTIDX: | |
901 | rc = obd_iocontrol(OBD_IOC_QUOTACTL, sbi->ll_md_exp, | |
902 | sizeof(*qctl), qctl, NULL); | |
903 | break; | |
904 | case QC_OSTIDX: | |
905 | rc = obd_iocontrol(OBD_IOC_QUOTACTL, sbi->ll_dt_exp, | |
906 | sizeof(*qctl), qctl, NULL); | |
907 | break; | |
908 | case QC_UUID: | |
909 | rc = obd_iocontrol(OBD_IOC_QUOTACTL, sbi->ll_md_exp, | |
910 | sizeof(*qctl), qctl, NULL); | |
911 | if (rc == -EAGAIN) | |
912 | rc = obd_iocontrol(OBD_IOC_QUOTACTL, | |
913 | sbi->ll_dt_exp, | |
914 | sizeof(*qctl), qctl, NULL); | |
915 | break; | |
916 | default: | |
917 | rc = -EINVAL; | |
918 | break; | |
919 | } | |
920 | ||
921 | if (rc) | |
0a3bdb00 | 922 | return rc; |
d7e09d03 PT |
923 | |
924 | qctl->qc_cmd = cmd; | |
925 | } else { | |
926 | struct obd_quotactl *oqctl; | |
927 | ||
496a51bd JL |
928 | oqctl = kzalloc(sizeof(*oqctl), GFP_NOFS); |
929 | if (!oqctl) | |
0a3bdb00 | 930 | return -ENOMEM; |
d7e09d03 PT |
931 | |
932 | QCTL_COPY(oqctl, qctl); | |
933 | rc = obd_quotactl(sbi->ll_md_exp, oqctl); | |
934 | if (rc) { | |
935 | if (rc != -EALREADY && cmd == Q_QUOTAON) { | |
936 | oqctl->qc_cmd = Q_QUOTAOFF; | |
937 | obd_quotactl(sbi->ll_md_exp, oqctl); | |
938 | } | |
97903a26 | 939 | kfree(oqctl); |
0a3bdb00 | 940 | return rc; |
d7e09d03 PT |
941 | } |
942 | /* If QIF_SPACE is not set, client should collect the | |
c0894c6c OD |
943 | * space usage from OSSs by itself |
944 | */ | |
d7e09d03 PT |
945 | if (cmd == Q_GETQUOTA && |
946 | !(oqctl->qc_dqblk.dqb_valid & QIF_SPACE) && | |
947 | !oqctl->qc_dqblk.dqb_curspace) { | |
948 | struct obd_quotactl *oqctl_tmp; | |
949 | ||
496a51bd JL |
950 | oqctl_tmp = kzalloc(sizeof(*oqctl_tmp), GFP_NOFS); |
951 | if (!oqctl_tmp) { | |
34e1f2bb JL |
952 | rc = -ENOMEM; |
953 | goto out; | |
954 | } | |
d7e09d03 PT |
955 | |
956 | oqctl_tmp->qc_cmd = Q_GETOQUOTA; | |
957 | oqctl_tmp->qc_id = oqctl->qc_id; | |
958 | oqctl_tmp->qc_type = oqctl->qc_type; | |
959 | ||
960 | /* collect space usage from OSTs */ | |
961 | oqctl_tmp->qc_dqblk.dqb_curspace = 0; | |
962 | rc = obd_quotactl(sbi->ll_dt_exp, oqctl_tmp); | |
963 | if (!rc || rc == -EREMOTEIO) { | |
964 | oqctl->qc_dqblk.dqb_curspace = | |
965 | oqctl_tmp->qc_dqblk.dqb_curspace; | |
966 | oqctl->qc_dqblk.dqb_valid |= QIF_SPACE; | |
967 | } | |
968 | ||
969 | /* collect space & inode usage from MDTs */ | |
970 | oqctl_tmp->qc_dqblk.dqb_curspace = 0; | |
971 | oqctl_tmp->qc_dqblk.dqb_curinodes = 0; | |
972 | rc = obd_quotactl(sbi->ll_md_exp, oqctl_tmp); | |
973 | if (!rc || rc == -EREMOTEIO) { | |
974 | oqctl->qc_dqblk.dqb_curspace += | |
975 | oqctl_tmp->qc_dqblk.dqb_curspace; | |
976 | oqctl->qc_dqblk.dqb_curinodes = | |
977 | oqctl_tmp->qc_dqblk.dqb_curinodes; | |
978 | oqctl->qc_dqblk.dqb_valid |= QIF_INODES; | |
979 | } else { | |
980 | oqctl->qc_dqblk.dqb_valid &= ~QIF_SPACE; | |
981 | } | |
982 | ||
97903a26 | 983 | kfree(oqctl_tmp); |
d7e09d03 PT |
984 | } |
985 | out: | |
986 | QCTL_COPY(qctl, oqctl); | |
97903a26 | 987 | kfree(oqctl); |
d7e09d03 PT |
988 | } |
989 | ||
0a3bdb00 | 990 | return rc; |
d7e09d03 PT |
991 | } |
992 | ||
a7503434 OD |
993 | /* This function tries to get a single name component, |
994 | * to send to the server. No actual path traversal involved, | |
c0894c6c OD |
995 | * so we limit to NAME_MAX |
996 | */ | |
a7503434 | 997 | static char *ll_getname(const char __user *filename) |
d7e09d03 PT |
998 | { |
999 | int ret = 0, len; | |
a7503434 | 1000 | char *tmp; |
d7e09d03 | 1001 | |
a7503434 | 1002 | tmp = kzalloc(NAME_MAX + 1, GFP_KERNEL); |
d7e09d03 PT |
1003 | if (!tmp) |
1004 | return ERR_PTR(-ENOMEM); | |
1005 | ||
a7503434 OD |
1006 | len = strncpy_from_user(tmp, filename, NAME_MAX + 1); |
1007 | if (len < 0) | |
1008 | ret = len; | |
1009 | else if (len == 0) | |
d7e09d03 | 1010 | ret = -ENOENT; |
a7503434 | 1011 | else if (len > NAME_MAX && tmp[NAME_MAX] != 0) |
d7e09d03 PT |
1012 | ret = -ENAMETOOLONG; |
1013 | ||
1014 | if (ret) { | |
a7503434 | 1015 | kfree(tmp); |
d7e09d03 PT |
1016 | tmp = ERR_PTR(ret); |
1017 | } | |
1018 | return tmp; | |
1019 | } | |
1020 | ||
a7503434 | 1021 | #define ll_putname(filename) kfree(filename) |
d7e09d03 PT |
1022 | |
1023 | static long ll_dir_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
1024 | { | |
2a8a3597 | 1025 | struct inode *inode = file_inode(file); |
d7e09d03 PT |
1026 | struct ll_sb_info *sbi = ll_i2sbi(inode); |
1027 | struct obd_ioctl_data *data; | |
1028 | int rc = 0; | |
d7e09d03 | 1029 | |
97a075cd JN |
1030 | CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), cmd=%#x\n", |
1031 | PFID(ll_inode2fid(inode)), inode, cmd); | |
d7e09d03 PT |
1032 | |
1033 | /* asm-ppc{,64} declares TCGETS, et. al. as type 't' not 'T' */ | |
1034 | if (_IOC_TYPE(cmd) == 'T' || _IOC_TYPE(cmd) == 't') /* tty ioctls */ | |
1035 | return -ENOTTY; | |
1036 | ||
1037 | ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_IOCTL, 1); | |
a58a38ac | 1038 | switch (cmd) { |
d7e09d03 PT |
1039 | case FSFILT_IOC_GETFLAGS: |
1040 | case FSFILT_IOC_SETFLAGS: | |
0a3bdb00 | 1041 | return ll_iocontrol(inode, file, cmd, arg); |
d7e09d03 PT |
1042 | case FSFILT_IOC_GETVERSION_OLD: |
1043 | case FSFILT_IOC_GETVERSION: | |
af00f6c5 | 1044 | return put_user(inode->i_generation, (int __user *)arg); |
d7e09d03 PT |
1045 | /* We need to special case any other ioctls we want to handle, |
1046 | * to send them to the MDS/OST as appropriate and to properly | |
1047 | * network encode the arg field. | |
1048 | case FSFILT_IOC_SETVERSION_OLD: | |
1049 | case FSFILT_IOC_SETVERSION: | |
1050 | */ | |
1051 | case LL_IOC_GET_MDTIDX: { | |
1052 | int mdtidx; | |
1053 | ||
1054 | mdtidx = ll_get_mdt_idx(inode); | |
1055 | if (mdtidx < 0) | |
0a3bdb00 | 1056 | return mdtidx; |
d7e09d03 | 1057 | |
af00f6c5 | 1058 | if (put_user((int)mdtidx, (int __user *)arg)) |
0a3bdb00 | 1059 | return -EFAULT; |
d7e09d03 PT |
1060 | |
1061 | return 0; | |
1062 | } | |
1063 | case IOC_MDC_LOOKUP: { | |
d7e09d03 PT |
1064 | int namelen, len = 0; |
1065 | char *buf = NULL; | |
1066 | char *filename; | |
d7e09d03 | 1067 | |
e3e8ff41 | 1068 | rc = obd_ioctl_getdata(&buf, &len, (void __user *)arg); |
d7e09d03 | 1069 | if (rc) |
0a3bdb00 | 1070 | return rc; |
d7e09d03 PT |
1071 | data = (void *)buf; |
1072 | ||
1073 | filename = data->ioc_inlbuf1; | |
1074 | namelen = strlen(filename); | |
1075 | ||
1076 | if (namelen < 1) { | |
1077 | CDEBUG(D_INFO, "IOC_MDC_LOOKUP missing filename\n"); | |
34e1f2bb JL |
1078 | rc = -EINVAL; |
1079 | goto out_free; | |
d7e09d03 PT |
1080 | } |
1081 | ||
1d82425f | 1082 | rc = ll_get_fid_by_name(inode, filename, namelen, NULL); |
d7e09d03 | 1083 | if (rc < 0) { |
1d82425f | 1084 | CERROR("%s: lookup %.*s failed: rc = %d\n", |
1085 | ll_get_fsname(inode->i_sb, NULL, 0), namelen, | |
1086 | filename, rc); | |
34e1f2bb | 1087 | goto out_free; |
d7e09d03 | 1088 | } |
d7e09d03 PT |
1089 | out_free: |
1090 | obd_ioctl_freedata(buf, len); | |
1091 | return rc; | |
1092 | } | |
1093 | case LL_IOC_LMV_SETSTRIPE: { | |
1094 | struct lmv_user_md *lum; | |
1095 | char *buf = NULL; | |
1096 | char *filename; | |
1097 | int namelen = 0; | |
1098 | int lumlen = 0; | |
d07d4cb8 | 1099 | umode_t mode; |
d7e09d03 PT |
1100 | int len; |
1101 | int rc; | |
1102 | ||
e3e8ff41 | 1103 | rc = obd_ioctl_getdata(&buf, &len, (void __user *)arg); |
d7e09d03 | 1104 | if (rc) |
0a3bdb00 | 1105 | return rc; |
d7e09d03 PT |
1106 | |
1107 | data = (void *)buf; | |
6e16818b | 1108 | if (!data->ioc_inlbuf1 || !data->ioc_inlbuf2 || |
34e1f2bb JL |
1109 | data->ioc_inllen1 == 0 || data->ioc_inllen2 == 0) { |
1110 | rc = -EINVAL; | |
1111 | goto lmv_out_free; | |
1112 | } | |
d7e09d03 PT |
1113 | |
1114 | filename = data->ioc_inlbuf1; | |
1115 | namelen = data->ioc_inllen1; | |
1116 | ||
1117 | if (namelen < 1) { | |
1118 | CDEBUG(D_INFO, "IOC_MDC_LOOKUP missing filename\n"); | |
34e1f2bb JL |
1119 | rc = -EINVAL; |
1120 | goto lmv_out_free; | |
d7e09d03 PT |
1121 | } |
1122 | lum = (struct lmv_user_md *)data->ioc_inlbuf2; | |
1123 | lumlen = data->ioc_inllen2; | |
1124 | ||
1125 | if (lum->lum_magic != LMV_USER_MAGIC || | |
1126 | lumlen != sizeof(*lum)) { | |
1127 | CERROR("%s: wrong lum magic %x or size %d: rc = %d\n", | |
1128 | filename, lum->lum_magic, lumlen, -EFAULT); | |
34e1f2bb JL |
1129 | rc = -EINVAL; |
1130 | goto lmv_out_free; | |
d7e09d03 PT |
1131 | } |
1132 | ||
d07d4cb8 | 1133 | #if OBD_OCD_VERSION(2, 9, 50, 0) > LUSTRE_VERSION_CODE |
1134 | mode = data->ioc_type != 0 ? data->ioc_type : S_IRWXUGO; | |
1135 | #else | |
1136 | mode = data->ioc_type; | |
1137 | #endif | |
1138 | rc = ll_dir_setdirstripe(inode, lum, filename, mode); | |
d7e09d03 PT |
1139 | lmv_out_free: |
1140 | obd_ioctl_freedata(buf, len); | |
0a3bdb00 | 1141 | return rc; |
d7e09d03 | 1142 | } |
6e23ea98 | 1143 | case LL_IOC_LMV_SET_DEFAULT_STRIPE: { |
1144 | struct lmv_user_md __user *ulump; | |
1145 | struct lmv_user_md lum; | |
1146 | int rc; | |
1147 | ||
1148 | ulump = (struct lmv_user_md __user *)arg; | |
1149 | if (copy_from_user(&lum, ulump, sizeof(lum))) | |
1150 | return -EFAULT; | |
1151 | ||
1152 | if (lum.lum_magic != LMV_USER_MAGIC) | |
1153 | return -EINVAL; | |
1154 | ||
1155 | rc = ll_dir_setstripe(inode, (struct lov_user_md *)&lum, 0); | |
1156 | ||
1157 | return rc; | |
1158 | } | |
d7e09d03 PT |
1159 | case LL_IOC_LOV_SETSTRIPE: { |
1160 | struct lov_user_md_v3 lumv3; | |
1161 | struct lov_user_md_v1 *lumv1 = (struct lov_user_md_v1 *)&lumv3; | |
af00f6c5 OD |
1162 | struct lov_user_md_v1 __user *lumv1p = (void __user *)arg; |
1163 | struct lov_user_md_v3 __user *lumv3p = (void __user *)arg; | |
d7e09d03 PT |
1164 | |
1165 | int set_default = 0; | |
1166 | ||
1167 | LASSERT(sizeof(lumv3) == sizeof(*lumv3p)); | |
1168 | LASSERT(sizeof(lumv3.lmm_objects[0]) == | |
1169 | sizeof(lumv3p->lmm_objects[0])); | |
1170 | /* first try with v1 which is smaller than v3 */ | |
1171 | if (copy_from_user(lumv1, lumv1p, sizeof(*lumv1))) | |
0a3bdb00 | 1172 | return -EFAULT; |
d7e09d03 | 1173 | |
557732ad | 1174 | if (lumv1->lmm_magic == LOV_USER_MAGIC_V3) { |
d7e09d03 | 1175 | if (copy_from_user(&lumv3, lumv3p, sizeof(lumv3))) |
0a3bdb00 | 1176 | return -EFAULT; |
d7e09d03 PT |
1177 | } |
1178 | ||
f76c23da | 1179 | if (is_root_inode(inode)) |
d7e09d03 PT |
1180 | set_default = 1; |
1181 | ||
1182 | /* in v1 and v3 cases lumv1 points to data */ | |
1183 | rc = ll_dir_setstripe(inode, lumv1, set_default); | |
1184 | ||
0a3bdb00 | 1185 | return rc; |
d7e09d03 PT |
1186 | } |
1187 | case LL_IOC_LMV_GETSTRIPE: { | |
6e23ea98 | 1188 | struct lmv_user_md __user *ulmv; |
d7e09d03 | 1189 | struct lmv_user_md lum; |
6e23ea98 | 1190 | struct ptlrpc_request *request = NULL; |
1191 | struct lmv_user_md *tmp = NULL; | |
1192 | union lmv_mds_md *lmm = NULL; | |
1193 | u64 valid = 0; | |
1194 | int stripe_count; | |
1195 | int mdt_index; | |
d7e09d03 | 1196 | int lum_size; |
6e23ea98 | 1197 | int lmmsize; |
1198 | int rc; | |
1199 | int i; | |
d7e09d03 | 1200 | |
6e23ea98 | 1201 | ulmv = (struct lmv_user_md __user *)arg; |
1202 | if (copy_from_user(&lum, ulmv, sizeof(*ulmv))) | |
0a3bdb00 | 1203 | return -EFAULT; |
d7e09d03 | 1204 | |
6e23ea98 | 1205 | /* |
1206 | * lum_magic will indicate which stripe the ioctl will like | |
1207 | * to get, LMV_MAGIC_V1 is for normal LMV stripe, LMV_USER_MAGIC | |
1208 | * is for default LMV stripe | |
1209 | */ | |
1210 | if (lum.lum_magic == LMV_MAGIC_V1) | |
1211 | valid |= OBD_MD_MEA; | |
1212 | else if (lum.lum_magic == LMV_USER_MAGIC) | |
1213 | valid |= OBD_MD_DEFAULT_MEA; | |
1214 | else | |
0a3bdb00 | 1215 | return -EINVAL; |
d7e09d03 | 1216 | |
6e23ea98 | 1217 | rc = ll_dir_getstripe(inode, (void **)&lmm, &lmmsize, &request, |
1218 | valid); | |
8f18c8a4 | 1219 | if (rc) |
6e23ea98 | 1220 | goto finish_req; |
1221 | ||
1222 | /* Get default LMV EA */ | |
1223 | if (lum.lum_magic == LMV_USER_MAGIC) { | |
1224 | if (rc) | |
1225 | goto finish_req; | |
1226 | ||
1227 | if (lmmsize > sizeof(*ulmv)) { | |
1228 | rc = -EINVAL; | |
1229 | goto finish_req; | |
1230 | } | |
1231 | ||
1232 | if (copy_to_user(ulmv, lmm, lmmsize)) | |
1233 | rc = -EFAULT; | |
1234 | ||
1235 | goto finish_req; | |
1236 | } | |
1237 | ||
8f18c8a4 | 1238 | stripe_count = lmv_mds_md_stripe_count_get(lmm); |
6e23ea98 | 1239 | lum_size = lmv_user_md_size(stripe_count, LMV_MAGIC_V1); |
496a51bd JL |
1240 | tmp = kzalloc(lum_size, GFP_NOFS); |
1241 | if (!tmp) { | |
34e1f2bb | 1242 | rc = -ENOMEM; |
6e23ea98 | 1243 | goto finish_req; |
34e1f2bb | 1244 | } |
d7e09d03 | 1245 | |
6e23ea98 | 1246 | mdt_index = ll_get_mdt_idx(inode); |
1247 | if (mdt_index < 0) { | |
34e1f2bb | 1248 | rc = -ENOMEM; |
6e23ea98 | 1249 | goto out_tmp; |
1250 | } | |
8f18c8a4 | 1251 | tmp->lum_magic = LMV_MAGIC_V1; |
1252 | tmp->lum_stripe_count = 0; | |
6e23ea98 | 1253 | tmp->lum_stripe_offset = mdt_index; |
8f18c8a4 | 1254 | for (i = 0; i < stripe_count; i++) { |
1255 | struct lu_fid *fid; | |
1256 | ||
1257 | fid = &lmm->lmv_md_v1.lmv_stripe_fids[i]; | |
1258 | mdt_index = ll_get_mdt_idx_by_fid(sbi, fid); | |
6e23ea98 | 1259 | if (mdt_index < 0) { |
1260 | rc = mdt_index; | |
1261 | goto out_tmp; | |
1262 | } | |
1263 | tmp->lum_objects[i].lum_mds = mdt_index; | |
8f18c8a4 | 1264 | tmp->lum_objects[i].lum_fid = *fid; |
6e23ea98 | 1265 | tmp->lum_stripe_count++; |
34e1f2bb | 1266 | } |
d7e09d03 | 1267 | |
6e23ea98 | 1268 | if (copy_to_user(ulmv, tmp, lum_size)) { |
34e1f2bb | 1269 | rc = -EFAULT; |
6e23ea98 | 1270 | goto out_tmp; |
34e1f2bb | 1271 | } |
6e23ea98 | 1272 | out_tmp: |
fd5e2fd0 | 1273 | kfree(tmp); |
6e23ea98 | 1274 | finish_req: |
1275 | ptlrpc_req_finished(request); | |
0a3bdb00 | 1276 | return rc; |
d7e09d03 | 1277 | } |
6e23ea98 | 1278 | |
d7e09d03 | 1279 | case LL_IOC_LOV_SWAP_LAYOUTS: |
0a3bdb00 | 1280 | return -EPERM; |
8877d3bf | 1281 | case IOC_OBD_STATFS: |
4c6243ec | 1282 | return ll_obd_statfs(inode, (void __user *)arg); |
d7e09d03 PT |
1283 | case LL_IOC_LOV_GETSTRIPE: |
1284 | case LL_IOC_MDC_GETINFO: | |
1285 | case IOC_MDC_GETFILEINFO: | |
1286 | case IOC_MDC_GETFILESTRIPE: { | |
1287 | struct ptlrpc_request *request = NULL; | |
af00f6c5 | 1288 | struct lov_user_md __user *lump; |
d7e09d03 PT |
1289 | struct lov_mds_md *lmm = NULL; |
1290 | struct mdt_body *body; | |
1291 | char *filename = NULL; | |
1292 | int lmmsize; | |
1293 | ||
1294 | if (cmd == IOC_MDC_GETFILEINFO || | |
1295 | cmd == IOC_MDC_GETFILESTRIPE) { | |
d47bb83b | 1296 | filename = ll_getname((const char __user *)arg); |
d7e09d03 | 1297 | if (IS_ERR(filename)) |
0a3bdb00 | 1298 | return PTR_ERR(filename); |
d7e09d03 PT |
1299 | |
1300 | rc = ll_lov_getstripe_ea_info(inode, filename, &lmm, | |
1301 | &lmmsize, &request); | |
1302 | } else { | |
6e23ea98 | 1303 | rc = ll_dir_getstripe(inode, (void **)&lmm, &lmmsize, |
1304 | &request, 0); | |
d7e09d03 PT |
1305 | } |
1306 | ||
1307 | if (request) { | |
1308 | body = req_capsule_server_get(&request->rq_pill, | |
1309 | &RMF_MDT_BODY); | |
6e16818b | 1310 | LASSERT(body); |
d7e09d03 | 1311 | } else { |
34e1f2bb | 1312 | goto out_req; |
d7e09d03 PT |
1313 | } |
1314 | ||
1315 | if (rc < 0) { | |
1316 | if (rc == -ENODATA && (cmd == IOC_MDC_GETFILEINFO || | |
34e1f2bb JL |
1317 | cmd == LL_IOC_MDC_GETINFO)) { |
1318 | rc = 0; | |
1319 | goto skip_lmm; | |
da5ecb4d | 1320 | } else { |
34e1f2bb | 1321 | goto out_req; |
da5ecb4d | 1322 | } |
d7e09d03 PT |
1323 | } |
1324 | ||
1325 | if (cmd == IOC_MDC_GETFILESTRIPE || | |
1326 | cmd == LL_IOC_LOV_GETSTRIPE) { | |
af00f6c5 | 1327 | lump = (struct lov_user_md __user *)arg; |
d7e09d03 | 1328 | } else { |
af00f6c5 | 1329 | struct lov_user_mds_data __user *lmdp; |
79792190 | 1330 | |
af00f6c5 | 1331 | lmdp = (struct lov_user_mds_data __user *)arg; |
d7e09d03 PT |
1332 | lump = &lmdp->lmd_lmm; |
1333 | } | |
1334 | if (copy_to_user(lump, lmm, lmmsize)) { | |
34e1f2bb JL |
1335 | if (copy_to_user(lump, lmm, sizeof(*lump))) { |
1336 | rc = -EFAULT; | |
1337 | goto out_req; | |
1338 | } | |
d7e09d03 PT |
1339 | rc = -EOVERFLOW; |
1340 | } | |
eb73f514 | 1341 | skip_lmm: |
d7e09d03 | 1342 | if (cmd == IOC_MDC_GETFILEINFO || cmd == LL_IOC_MDC_GETINFO) { |
af00f6c5 | 1343 | struct lov_user_mds_data __user *lmdp; |
d7e09d03 PT |
1344 | lstat_t st = { 0 }; |
1345 | ||
1346 | st.st_dev = inode->i_sb->s_dev; | |
2e1b5b8b JH |
1347 | st.st_mode = body->mbo_mode; |
1348 | st.st_nlink = body->mbo_nlink; | |
1349 | st.st_uid = body->mbo_uid; | |
1350 | st.st_gid = body->mbo_gid; | |
1351 | st.st_rdev = body->mbo_rdev; | |
1352 | st.st_size = body->mbo_size; | |
09cbfeaf | 1353 | st.st_blksize = PAGE_SIZE; |
2e1b5b8b JH |
1354 | st.st_blocks = body->mbo_blocks; |
1355 | st.st_atime = body->mbo_atime; | |
1356 | st.st_mtime = body->mbo_mtime; | |
1357 | st.st_ctime = body->mbo_ctime; | |
1358 | st.st_ino = cl_fid_build_ino(&body->mbo_fid1, | |
d780846e | 1359 | sbi->ll_flags & |
1360 | LL_SBI_32BIT_API); | |
d7e09d03 | 1361 | |
af00f6c5 | 1362 | lmdp = (struct lov_user_mds_data __user *)arg; |
34e1f2bb JL |
1363 | if (copy_to_user(&lmdp->lmd_st, &st, sizeof(st))) { |
1364 | rc = -EFAULT; | |
1365 | goto out_req; | |
1366 | } | |
d7e09d03 PT |
1367 | } |
1368 | ||
eb73f514 | 1369 | out_req: |
d7e09d03 PT |
1370 | ptlrpc_req_finished(request); |
1371 | if (filename) | |
1372 | ll_putname(filename); | |
1373 | return rc; | |
1374 | } | |
1375 | case IOC_LOV_GETINFO: { | |
af00f6c5 | 1376 | struct lov_user_mds_data __user *lumd; |
d7e09d03 | 1377 | struct lov_stripe_md *lsm; |
af00f6c5 | 1378 | struct lov_user_md __user *lum; |
d7e09d03 PT |
1379 | struct lov_mds_md *lmm; |
1380 | int lmmsize; | |
1381 | lstat_t st; | |
1382 | ||
af00f6c5 | 1383 | lumd = (struct lov_user_mds_data __user *)arg; |
d7e09d03 PT |
1384 | lum = &lumd->lmd_lmm; |
1385 | ||
1386 | rc = ll_get_max_mdsize(sbi, &lmmsize); | |
1387 | if (rc) | |
0a3bdb00 | 1388 | return rc; |
d7e09d03 | 1389 | |
e958f49b | 1390 | lmm = libcfs_kvzalloc(lmmsize, GFP_NOFS); |
6e16818b | 1391 | if (!lmm) |
0a3bdb00 | 1392 | return -ENOMEM; |
34e1f2bb JL |
1393 | if (copy_from_user(lmm, lum, lmmsize)) { |
1394 | rc = -EFAULT; | |
1395 | goto free_lmm; | |
1396 | } | |
d7e09d03 PT |
1397 | |
1398 | switch (lmm->lmm_magic) { | |
1399 | case LOV_USER_MAGIC_V1: | |
1f6eaf83 | 1400 | if (cpu_to_le32(LOV_USER_MAGIC_V1) == LOV_USER_MAGIC_V1) |
d7e09d03 PT |
1401 | break; |
1402 | /* swab objects first so that stripes num will be sane */ | |
1403 | lustre_swab_lov_user_md_objects( | |
1404 | ((struct lov_user_md_v1 *)lmm)->lmm_objects, | |
1405 | ((struct lov_user_md_v1 *)lmm)->lmm_stripe_count); | |
1406 | lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lmm); | |
1407 | break; | |
1408 | case LOV_USER_MAGIC_V3: | |
1f6eaf83 | 1409 | if (cpu_to_le32(LOV_USER_MAGIC_V3) == LOV_USER_MAGIC_V3) |
d7e09d03 PT |
1410 | break; |
1411 | /* swab objects first so that stripes num will be sane */ | |
1412 | lustre_swab_lov_user_md_objects( | |
1413 | ((struct lov_user_md_v3 *)lmm)->lmm_objects, | |
1414 | ((struct lov_user_md_v3 *)lmm)->lmm_stripe_count); | |
1415 | lustre_swab_lov_user_md_v3((struct lov_user_md_v3 *)lmm); | |
1416 | break; | |
1417 | default: | |
34e1f2bb JL |
1418 | rc = -EINVAL; |
1419 | goto free_lmm; | |
d7e09d03 PT |
1420 | } |
1421 | ||
1422 | rc = obd_unpackmd(sbi->ll_dt_exp, &lsm, lmm, lmmsize); | |
34e1f2bb JL |
1423 | if (rc < 0) { |
1424 | rc = -ENOMEM; | |
1425 | goto free_lmm; | |
1426 | } | |
d7e09d03 PT |
1427 | |
1428 | /* Perform glimpse_size operation. */ | |
1429 | memset(&st, 0, sizeof(st)); | |
1430 | ||
1431 | rc = ll_glimpse_ioctl(sbi, lsm, &st); | |
1432 | if (rc) | |
34e1f2bb | 1433 | goto free_lsm; |
d7e09d03 | 1434 | |
34e1f2bb JL |
1435 | if (copy_to_user(&lumd->lmd_st, &st, sizeof(st))) { |
1436 | rc = -EFAULT; | |
1437 | goto free_lsm; | |
1438 | } | |
d7e09d03 | 1439 | |
eb73f514 | 1440 | free_lsm: |
d7e09d03 | 1441 | obd_free_memmd(sbi->ll_dt_exp, &lsm); |
eb73f514 | 1442 | free_lmm: |
e958f49b | 1443 | kvfree(lmm); |
d7e09d03 PT |
1444 | return rc; |
1445 | } | |
d7e09d03 PT |
1446 | case OBD_IOC_QUOTACHECK: { |
1447 | struct obd_quotactl *oqctl; | |
1448 | int error = 0; | |
1449 | ||
341f1f0a | 1450 | if (!capable(CFS_CAP_SYS_ADMIN)) |
0a3bdb00 | 1451 | return -EPERM; |
d7e09d03 | 1452 | |
496a51bd | 1453 | oqctl = kzalloc(sizeof(*oqctl), GFP_NOFS); |
d7e09d03 | 1454 | if (!oqctl) |
0a3bdb00 | 1455 | return -ENOMEM; |
d7e09d03 PT |
1456 | oqctl->qc_type = arg; |
1457 | rc = obd_quotacheck(sbi->ll_md_exp, oqctl); | |
1458 | if (rc < 0) { | |
1459 | CDEBUG(D_INFO, "md_quotacheck failed: rc %d\n", rc); | |
1460 | error = rc; | |
1461 | } | |
1462 | ||
1463 | rc = obd_quotacheck(sbi->ll_dt_exp, oqctl); | |
1464 | if (rc < 0) | |
1465 | CDEBUG(D_INFO, "obd_quotacheck failed: rc %d\n", rc); | |
1466 | ||
97903a26 | 1467 | kfree(oqctl); |
d7e09d03 PT |
1468 | return error ?: rc; |
1469 | } | |
1470 | case OBD_IOC_POLL_QUOTACHECK: { | |
1471 | struct if_quotacheck *check; | |
1472 | ||
341f1f0a | 1473 | if (!capable(CFS_CAP_SYS_ADMIN)) |
0a3bdb00 | 1474 | return -EPERM; |
d7e09d03 | 1475 | |
496a51bd | 1476 | check = kzalloc(sizeof(*check), GFP_NOFS); |
d7e09d03 | 1477 | if (!check) |
0a3bdb00 | 1478 | return -ENOMEM; |
d7e09d03 PT |
1479 | |
1480 | rc = obd_iocontrol(cmd, sbi->ll_md_exp, 0, (void *)check, | |
1481 | NULL); | |
1482 | if (rc) { | |
1483 | CDEBUG(D_QUOTA, "mdc ioctl %d failed: %d\n", cmd, rc); | |
af00f6c5 OD |
1484 | if (copy_to_user((void __user *)arg, check, |
1485 | sizeof(*check))) | |
d7e09d03 | 1486 | CDEBUG(D_QUOTA, "copy_to_user failed\n"); |
34e1f2bb | 1487 | goto out_poll; |
d7e09d03 PT |
1488 | } |
1489 | ||
1490 | rc = obd_iocontrol(cmd, sbi->ll_dt_exp, 0, (void *)check, | |
1491 | NULL); | |
1492 | if (rc) { | |
1493 | CDEBUG(D_QUOTA, "osc ioctl %d failed: %d\n", cmd, rc); | |
af00f6c5 OD |
1494 | if (copy_to_user((void __user *)arg, check, |
1495 | sizeof(*check))) | |
d7e09d03 | 1496 | CDEBUG(D_QUOTA, "copy_to_user failed\n"); |
34e1f2bb | 1497 | goto out_poll; |
d7e09d03 | 1498 | } |
eb73f514 | 1499 | out_poll: |
97903a26 | 1500 | kfree(check); |
0a3bdb00 | 1501 | return rc; |
d7e09d03 | 1502 | } |
8877d3bf | 1503 | case OBD_IOC_QUOTACTL: { |
d7e09d03 PT |
1504 | struct if_quotactl *qctl; |
1505 | ||
496a51bd | 1506 | qctl = kzalloc(sizeof(*qctl), GFP_NOFS); |
d7e09d03 | 1507 | if (!qctl) |
0a3bdb00 | 1508 | return -ENOMEM; |
d7e09d03 | 1509 | |
af00f6c5 | 1510 | if (copy_from_user(qctl, (void __user *)arg, sizeof(*qctl))) { |
34e1f2bb JL |
1511 | rc = -EFAULT; |
1512 | goto out_quotactl; | |
1513 | } | |
d7e09d03 PT |
1514 | |
1515 | rc = quotactl_ioctl(sbi, qctl); | |
1516 | ||
af00f6c5 OD |
1517 | if (rc == 0 && copy_to_user((void __user *)arg, qctl, |
1518 | sizeof(*qctl))) | |
d7e09d03 PT |
1519 | rc = -EFAULT; |
1520 | ||
eb73f514 | 1521 | out_quotactl: |
97903a26 | 1522 | kfree(qctl); |
0a3bdb00 | 1523 | return rc; |
d7e09d03 PT |
1524 | } |
1525 | case OBD_IOC_GETDTNAME: | |
1526 | case OBD_IOC_GETMDNAME: | |
0a3bdb00 | 1527 | return ll_get_obd_name(inode, cmd, arg); |
d7e09d03 | 1528 | case LL_IOC_FLUSHCTX: |
0a3bdb00 | 1529 | return ll_flush_ctx(inode); |
d7e09d03 PT |
1530 | case LL_IOC_GETOBDCOUNT: { |
1531 | int count, vallen; | |
1532 | struct obd_export *exp; | |
1533 | ||
af00f6c5 | 1534 | if (copy_from_user(&count, (int __user *)arg, sizeof(int))) |
0a3bdb00 | 1535 | return -EFAULT; |
d7e09d03 PT |
1536 | |
1537 | /* get ost count when count is zero, get mdt count otherwise */ | |
1538 | exp = count ? sbi->ll_md_exp : sbi->ll_dt_exp; | |
1539 | vallen = sizeof(count); | |
1540 | rc = obd_get_info(NULL, exp, sizeof(KEY_TGT_COUNT), | |
1541 | KEY_TGT_COUNT, &vallen, &count, NULL); | |
1542 | if (rc) { | |
1543 | CERROR("get target count failed: %d\n", rc); | |
0a3bdb00 | 1544 | return rc; |
d7e09d03 PT |
1545 | } |
1546 | ||
af00f6c5 | 1547 | if (copy_to_user((int __user *)arg, &count, sizeof(int))) |
0a3bdb00 | 1548 | return -EFAULT; |
d7e09d03 | 1549 | |
0a3bdb00 | 1550 | return 0; |
d7e09d03 PT |
1551 | } |
1552 | case LL_IOC_PATH2FID: | |
af00f6c5 OD |
1553 | if (copy_to_user((void __user *)arg, ll_inode2fid(inode), |
1554 | sizeof(struct lu_fid))) | |
0a3bdb00 GKH |
1555 | return -EFAULT; |
1556 | return 0; | |
d7e09d03 | 1557 | case LL_IOC_GET_CONNECT_FLAGS: { |
e09bee34 OD |
1558 | return obd_iocontrol(cmd, sbi->ll_md_exp, 0, NULL, |
1559 | (void __user *)arg); | |
d7e09d03 PT |
1560 | } |
1561 | case OBD_IOC_CHANGELOG_SEND: | |
1562 | case OBD_IOC_CHANGELOG_CLEAR: | |
bbaa9c10 NY |
1563 | if (!capable(CFS_CAP_SYS_ADMIN)) |
1564 | return -EPERM; | |
1565 | ||
af00f6c5 | 1566 | rc = copy_and_ioctl(cmd, sbi->ll_md_exp, (void __user *)arg, |
d7e09d03 | 1567 | sizeof(struct ioc_changelog)); |
0a3bdb00 | 1568 | return rc; |
d7e09d03 | 1569 | case OBD_IOC_FID2PATH: |
7ec89fa5 | 1570 | return ll_fid2path(inode, (void __user *)arg); |
d7e09d03 PT |
1571 | case LL_IOC_HSM_REQUEST: { |
1572 | struct hsm_user_request *hur; | |
6b2eb32e | 1573 | ssize_t totalsize; |
d7e09d03 | 1574 | |
af00f6c5 | 1575 | hur = memdup_user((void __user *)arg, sizeof(*hur)); |
4a07594e AH |
1576 | if (IS_ERR(hur)) |
1577 | return PTR_ERR(hur); | |
d7e09d03 PT |
1578 | |
1579 | /* Compute the whole struct size */ | |
1580 | totalsize = hur_len(hur); | |
97903a26 | 1581 | kfree(hur); |
6b2eb32e NC |
1582 | if (totalsize < 0) |
1583 | return -E2BIG; | |
e55c4476 JN |
1584 | |
1585 | /* Final size will be more than double totalsize */ | |
1586 | if (totalsize >= MDS_MAXREQSIZE / 3) | |
1587 | return -E2BIG; | |
1588 | ||
e958f49b | 1589 | hur = libcfs_kvzalloc(totalsize, GFP_NOFS); |
6e16818b | 1590 | if (!hur) |
0a3bdb00 | 1591 | return -ENOMEM; |
d7e09d03 PT |
1592 | |
1593 | /* Copy the whole struct */ | |
af00f6c5 | 1594 | if (copy_from_user(hur, (void __user *)arg, totalsize)) { |
e958f49b | 1595 | kvfree(hur); |
0a3bdb00 | 1596 | return -EFAULT; |
d7e09d03 PT |
1597 | } |
1598 | ||
48d23e61 JX |
1599 | if (hur->hur_request.hr_action == HUA_RELEASE) { |
1600 | const struct lu_fid *fid; | |
1601 | struct inode *f; | |
1602 | int i; | |
1603 | ||
1604 | for (i = 0; i < hur->hur_request.hr_itemcount; i++) { | |
1605 | fid = &hur->hur_user_item[i].hui_fid; | |
1606 | f = search_inode_for_lustre(inode->i_sb, fid); | |
1607 | if (IS_ERR(f)) { | |
1608 | rc = PTR_ERR(f); | |
1609 | break; | |
1610 | } | |
1611 | ||
1612 | rc = ll_hsm_release(f); | |
1613 | iput(f); | |
1614 | if (rc != 0) | |
1615 | break; | |
1616 | } | |
1617 | } else { | |
1618 | rc = obd_iocontrol(cmd, ll_i2mdexp(inode), totalsize, | |
1619 | hur, NULL); | |
1620 | } | |
d7e09d03 | 1621 | |
e958f49b | 1622 | kvfree(hur); |
d7e09d03 | 1623 | |
0a3bdb00 | 1624 | return rc; |
d7e09d03 PT |
1625 | } |
1626 | case LL_IOC_HSM_PROGRESS: { | |
1627 | struct hsm_progress_kernel hpk; | |
1628 | struct hsm_progress hp; | |
1629 | ||
af00f6c5 | 1630 | if (copy_from_user(&hp, (void __user *)arg, sizeof(hp))) |
0a3bdb00 | 1631 | return -EFAULT; |
d7e09d03 PT |
1632 | |
1633 | hpk.hpk_fid = hp.hp_fid; | |
1634 | hpk.hpk_cookie = hp.hp_cookie; | |
1635 | hpk.hpk_extent = hp.hp_extent; | |
1636 | hpk.hpk_flags = hp.hp_flags; | |
1637 | hpk.hpk_errval = hp.hp_errval; | |
1638 | hpk.hpk_data_version = 0; | |
1639 | ||
1640 | /* File may not exist in Lustre; all progress | |
c0894c6c OD |
1641 | * reported to Lustre root |
1642 | */ | |
d7e09d03 PT |
1643 | rc = obd_iocontrol(cmd, sbi->ll_md_exp, sizeof(hpk), &hpk, |
1644 | NULL); | |
0a3bdb00 | 1645 | return rc; |
d7e09d03 PT |
1646 | } |
1647 | case LL_IOC_HSM_CT_START: | |
1cc28b77 JH |
1648 | if (!capable(CFS_CAP_SYS_ADMIN)) |
1649 | return -EPERM; | |
1650 | ||
af00f6c5 | 1651 | rc = copy_and_ioctl(cmd, sbi->ll_md_exp, (void __user *)arg, |
d7e09d03 | 1652 | sizeof(struct lustre_kernelcomm)); |
0a3bdb00 | 1653 | return rc; |
d7e09d03 PT |
1654 | |
1655 | case LL_IOC_HSM_COPY_START: { | |
1656 | struct hsm_copy *copy; | |
1657 | int rc; | |
1658 | ||
af00f6c5 | 1659 | copy = memdup_user((char __user *)arg, sizeof(*copy)); |
4a07594e AH |
1660 | if (IS_ERR(copy)) |
1661 | return PTR_ERR(copy); | |
d7e09d03 PT |
1662 | |
1663 | rc = ll_ioc_copy_start(inode->i_sb, copy); | |
af00f6c5 | 1664 | if (copy_to_user((char __user *)arg, copy, sizeof(*copy))) |
d7e09d03 PT |
1665 | rc = -EFAULT; |
1666 | ||
97903a26 | 1667 | kfree(copy); |
0a3bdb00 | 1668 | return rc; |
d7e09d03 PT |
1669 | } |
1670 | case LL_IOC_HSM_COPY_END: { | |
1671 | struct hsm_copy *copy; | |
1672 | int rc; | |
1673 | ||
af00f6c5 | 1674 | copy = memdup_user((char __user *)arg, sizeof(*copy)); |
4a07594e AH |
1675 | if (IS_ERR(copy)) |
1676 | return PTR_ERR(copy); | |
d7e09d03 PT |
1677 | |
1678 | rc = ll_ioc_copy_end(inode->i_sb, copy); | |
af00f6c5 | 1679 | if (copy_to_user((char __user *)arg, copy, sizeof(*copy))) |
d7e09d03 PT |
1680 | rc = -EFAULT; |
1681 | ||
97903a26 | 1682 | kfree(copy); |
0a3bdb00 | 1683 | return rc; |
d7e09d03 | 1684 | } |
79496845 | 1685 | case LL_IOC_MIGRATE: { |
1686 | char *buf = NULL; | |
1687 | const char *filename; | |
1688 | int namelen = 0; | |
1689 | int len; | |
1690 | int rc; | |
1691 | int mdtidx; | |
1692 | ||
1693 | rc = obd_ioctl_getdata(&buf, &len, (void __user *)arg); | |
1694 | if (rc < 0) | |
1695 | return rc; | |
1696 | ||
1697 | data = (struct obd_ioctl_data *)buf; | |
1698 | if (!data->ioc_inlbuf1 || !data->ioc_inlbuf2 || | |
1699 | !data->ioc_inllen1 || !data->ioc_inllen2) { | |
1700 | rc = -EINVAL; | |
1701 | goto migrate_free; | |
1702 | } | |
1703 | ||
1704 | filename = data->ioc_inlbuf1; | |
1705 | namelen = data->ioc_inllen1; | |
1d82425f | 1706 | if (namelen < 1 || namelen != strlen(filename) + 1) { |
79496845 | 1707 | rc = -EINVAL; |
1708 | goto migrate_free; | |
1709 | } | |
1710 | ||
1711 | if (data->ioc_inllen2 != sizeof(mdtidx)) { | |
1712 | rc = -EINVAL; | |
1713 | goto migrate_free; | |
1714 | } | |
1715 | mdtidx = *(int *)data->ioc_inlbuf2; | |
1716 | ||
1d82425f | 1717 | rc = ll_migrate(inode, file, mdtidx, filename, namelen - 1); |
79496845 | 1718 | migrate_free: |
1719 | obd_ioctl_freedata(buf, len); | |
1720 | ||
1721 | return rc; | |
1722 | } | |
1723 | ||
d7e09d03 | 1724 | default: |
e09bee34 OD |
1725 | return obd_iocontrol(cmd, sbi->ll_dt_exp, 0, NULL, |
1726 | (void __user *)arg); | |
d7e09d03 PT |
1727 | } |
1728 | } | |
1729 | ||
1730 | static loff_t ll_dir_seek(struct file *file, loff_t offset, int origin) | |
1731 | { | |
1732 | struct inode *inode = file->f_mapping->host; | |
1733 | struct ll_file_data *fd = LUSTRE_FPRIVATE(file); | |
1734 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
1735 | int api32 = ll_need_32bit_api(sbi); | |
1736 | loff_t ret = -EINVAL; | |
d7e09d03 | 1737 | |
d7e09d03 | 1738 | switch (origin) { |
e17f5594 | 1739 | case SEEK_SET: |
1740 | break; | |
1741 | case SEEK_CUR: | |
1742 | offset += file->f_pos; | |
1743 | break; | |
1744 | case SEEK_END: | |
1745 | if (offset > 0) | |
34e1f2bb | 1746 | goto out; |
e17f5594 | 1747 | if (api32) |
1748 | offset += LL_DIR_END_OFF_32BIT; | |
1749 | else | |
1750 | offset += LL_DIR_END_OFF; | |
1751 | break; | |
1752 | default: | |
1753 | goto out; | |
d7e09d03 PT |
1754 | } |
1755 | ||
1756 | if (offset >= 0 && | |
1757 | ((api32 && offset <= LL_DIR_END_OFF_32BIT) || | |
1758 | (!api32 && offset <= LL_DIR_END_OFF))) { | |
1759 | if (offset != file->f_pos) { | |
1760 | if ((api32 && offset == LL_DIR_END_OFF_32BIT) || | |
1761 | (!api32 && offset == LL_DIR_END_OFF)) | |
1762 | fd->lfd_pos = MDS_DIR_END_OFF; | |
1763 | else if (api32 && sbi->ll_flags & LL_SBI_64BIT_HASH) | |
1764 | fd->lfd_pos = offset << 32; | |
1765 | else | |
1766 | fd->lfd_pos = offset; | |
1767 | file->f_pos = offset; | |
1768 | file->f_version = 0; | |
1769 | } | |
1770 | ret = offset; | |
1771 | } | |
34e1f2bb | 1772 | goto out; |
d7e09d03 PT |
1773 | |
1774 | out: | |
d7e09d03 PT |
1775 | return ret; |
1776 | } | |
1777 | ||
2d95f10e | 1778 | static int ll_dir_open(struct inode *inode, struct file *file) |
d7e09d03 | 1779 | { |
0a3bdb00 | 1780 | return ll_file_open(inode, file); |
d7e09d03 PT |
1781 | } |
1782 | ||
2d95f10e | 1783 | static int ll_dir_release(struct inode *inode, struct file *file) |
d7e09d03 | 1784 | { |
0a3bdb00 | 1785 | return ll_file_release(inode, file); |
d7e09d03 PT |
1786 | } |
1787 | ||
2d95f10e | 1788 | const struct file_operations ll_dir_operations = { |
d7e09d03 PT |
1789 | .llseek = ll_dir_seek, |
1790 | .open = ll_dir_open, | |
1791 | .release = ll_dir_release, | |
1792 | .read = generic_read_dir, | |
060ff688 | 1793 | .iterate_shared = ll_readdir, |
d7e09d03 PT |
1794 | .unlocked_ioctl = ll_dir_ioctl, |
1795 | .fsync = ll_fsync, | |
1796 | }; |