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 | |
18 | * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf | |
19 | * | |
20 | * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, | |
21 | * CA 95054 USA or visit www.sun.com if you need additional information or | |
22 | * have any questions. | |
23 | * | |
24 | * GPL HEADER END | |
25 | */ | |
26 | /* | |
27 | * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. | |
28 | * Use is subject to license terms. | |
29 | * | |
1dc563a6 | 30 | * Copyright (c) 2011, 2015, Intel Corporation. |
d7e09d03 PT |
31 | */ |
32 | /* | |
33 | * This file is part of Lustre, http://www.lustre.org/ | |
34 | * Lustre is a trademark of Sun Microsystems, Inc. | |
35 | * | |
36 | * lustre/llite/llite_lib.c | |
37 | * | |
38 | * Lustre Light Super operations | |
39 | */ | |
40 | ||
41 | #define DEBUG_SUBSYSTEM S_LLITE | |
42 | ||
43 | #include <linux/module.h> | |
a9c7db39 | 44 | #include <linux/statfs.h> |
d7e09d03 | 45 | #include <linux/types.h> |
d7e09d03 PT |
46 | #include <linux/mm.h> |
47 | ||
67a235f5 GKH |
48 | #include "../include/lustre_lite.h" |
49 | #include "../include/lustre_ha.h" | |
50 | #include "../include/lustre_dlm.h" | |
51 | #include "../include/lprocfs_status.h" | |
52 | #include "../include/lustre_disk.h" | |
53 | #include "../include/lustre_param.h" | |
54 | #include "../include/lustre_log.h" | |
55 | #include "../include/cl_object.h" | |
56 | #include "../include/obd_cksum.h" | |
d7e09d03 PT |
57 | #include "llite_internal.h" |
58 | ||
59 | struct kmem_cache *ll_file_data_slab; | |
ae7c0f48 | 60 | struct dentry *llite_root; |
fd0d04ba | 61 | struct kset *llite_kset; |
d7e09d03 | 62 | |
d7e09d03 PT |
63 | #ifndef log2 |
64 | #define log2(n) ffz(~(n)) | |
65 | #endif | |
66 | ||
fd0d04ba | 67 | static struct ll_sb_info *ll_init_sbi(struct super_block *sb) |
d7e09d03 PT |
68 | { |
69 | struct ll_sb_info *sbi = NULL; | |
70 | unsigned long pages; | |
71 | unsigned long lru_page_max; | |
72 | struct sysinfo si; | |
73 | class_uuid_t uuid; | |
74 | int i; | |
d7e09d03 | 75 | |
496a51bd | 76 | sbi = kzalloc(sizeof(*sbi), GFP_NOFS); |
d7e09d03 | 77 | if (!sbi) |
0a3bdb00 | 78 | return NULL; |
d7e09d03 PT |
79 | |
80 | spin_lock_init(&sbi->ll_lock); | |
81 | mutex_init(&sbi->ll_lco.lco_lock); | |
82 | spin_lock_init(&sbi->ll_pp_extent_lock); | |
83 | spin_lock_init(&sbi->ll_process_lock); | |
84 | sbi->ll_rw_stats_on = 0; | |
85 | ||
86 | si_meminfo(&si); | |
87 | pages = si.totalram - si.totalhigh; | |
5196e42c | 88 | lru_page_max = pages / 2; |
d7e09d03 | 89 | |
ac5b1481 | 90 | /* initialize ll_cache data */ |
d7e09d03 PT |
91 | atomic_set(&sbi->ll_cache.ccc_users, 0); |
92 | sbi->ll_cache.ccc_lru_max = lru_page_max; | |
93 | atomic_set(&sbi->ll_cache.ccc_lru_left, lru_page_max); | |
94 | spin_lock_init(&sbi->ll_cache.ccc_lru_lock); | |
95 | INIT_LIST_HEAD(&sbi->ll_cache.ccc_lru); | |
96 | ||
ac5b1481 PS |
97 | atomic_set(&sbi->ll_cache.ccc_unstable_nr, 0); |
98 | init_waitqueue_head(&sbi->ll_cache.ccc_unstable_waitq); | |
99 | ||
d7e09d03 PT |
100 | sbi->ll_ra_info.ra_max_pages_per_file = min(pages / 32, |
101 | SBI_DEFAULT_READAHEAD_MAX); | |
102 | sbi->ll_ra_info.ra_max_pages = sbi->ll_ra_info.ra_max_pages_per_file; | |
103 | sbi->ll_ra_info.ra_max_read_ahead_whole_pages = | |
104 | SBI_DEFAULT_READAHEAD_WHOLE_MAX; | |
d7e09d03 PT |
105 | |
106 | ll_generate_random_uuid(uuid); | |
107 | class_uuid_unparse(uuid, &sbi->ll_sb_uuid); | |
108 | CDEBUG(D_CONFIG, "generated uuid: %s\n", sbi->ll_sb_uuid.uuid); | |
109 | ||
d7e09d03 PT |
110 | sbi->ll_flags |= LL_SBI_VERBOSE; |
111 | sbi->ll_flags |= LL_SBI_CHECKSUM; | |
112 | ||
113 | sbi->ll_flags |= LL_SBI_LRU_RESIZE; | |
114 | ||
115 | for (i = 0; i <= LL_PROCESS_HIST_MAX; i++) { | |
116 | spin_lock_init(&sbi->ll_rw_extents_info.pp_extents[i]. | |
117 | pp_r_hist.oh_lock); | |
118 | spin_lock_init(&sbi->ll_rw_extents_info.pp_extents[i]. | |
119 | pp_w_hist.oh_lock); | |
120 | } | |
121 | ||
122 | /* metadata statahead is enabled by default */ | |
123 | sbi->ll_sa_max = LL_SA_RPC_DEF; | |
124 | atomic_set(&sbi->ll_sa_total, 0); | |
125 | atomic_set(&sbi->ll_sa_wrong, 0); | |
126 | atomic_set(&sbi->ll_agl_total, 0); | |
127 | sbi->ll_flags |= LL_SBI_AGL_ENABLED; | |
128 | ||
fd0d04ba OD |
129 | sbi->ll_sb = sb; |
130 | ||
0a3bdb00 | 131 | return sbi; |
d7e09d03 PT |
132 | } |
133 | ||
2d95f10e | 134 | static void ll_free_sbi(struct super_block *sb) |
d7e09d03 PT |
135 | { |
136 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
d7e09d03 | 137 | |
ad88aae0 | 138 | kfree(sbi); |
d7e09d03 PT |
139 | } |
140 | ||
d7e09d03 PT |
141 | static int client_common_fill_super(struct super_block *sb, char *md, char *dt, |
142 | struct vfsmount *mnt) | |
143 | { | |
ea7893bb | 144 | struct inode *root = NULL; |
d7e09d03 PT |
145 | struct ll_sb_info *sbi = ll_s2sbi(sb); |
146 | struct obd_device *obd; | |
d7e09d03 PT |
147 | struct obd_statfs *osfs = NULL; |
148 | struct ptlrpc_request *request = NULL; | |
149 | struct obd_connect_data *data = NULL; | |
150 | struct obd_uuid *uuid; | |
151 | struct md_op_data *op_data; | |
152 | struct lustre_md lmd; | |
21aef7d9 | 153 | u64 valid; |
d7e09d03 | 154 | int size, err, checksum; |
d7e09d03 PT |
155 | |
156 | obd = class_name2obd(md); | |
157 | if (!obd) { | |
158 | CERROR("MD %s: not setup or attached\n", md); | |
0a3bdb00 | 159 | return -EINVAL; |
d7e09d03 PT |
160 | } |
161 | ||
496a51bd JL |
162 | data = kzalloc(sizeof(*data), GFP_NOFS); |
163 | if (!data) | |
0a3bdb00 | 164 | return -ENOMEM; |
d7e09d03 | 165 | |
496a51bd JL |
166 | osfs = kzalloc(sizeof(*osfs), GFP_NOFS); |
167 | if (!osfs) { | |
97903a26 | 168 | kfree(data); |
0a3bdb00 | 169 | return -ENOMEM; |
d7e09d03 PT |
170 | } |
171 | ||
d7e09d03 PT |
172 | /* indicate the features supported by this client */ |
173 | data->ocd_connect_flags = OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | | |
174 | OBD_CONNECT_ATTRFID | | |
175 | OBD_CONNECT_VERSION | OBD_CONNECT_BRW_SIZE | | |
d7e09d03 PT |
176 | OBD_CONNECT_CANCELSET | OBD_CONNECT_FID | |
177 | OBD_CONNECT_AT | OBD_CONNECT_LOV_V3 | | |
178 | OBD_CONNECT_RMT_CLIENT | OBD_CONNECT_VBR | | |
179 | OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH| | |
180 | OBD_CONNECT_EINPROGRESS | | |
181 | OBD_CONNECT_JOBSTATS | OBD_CONNECT_LVB_TYPE | | |
7fc1f831 | 182 | OBD_CONNECT_LAYOUTLOCK | |
69342b78 AS |
183 | OBD_CONNECT_PINGLESS | |
184 | OBD_CONNECT_MAX_EASIZE | | |
63d42578 HZ |
185 | OBD_CONNECT_FLOCK_DEAD | |
186 | OBD_CONNECT_DISP_STRIPE; | |
d7e09d03 PT |
187 | |
188 | if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) | |
189 | data->ocd_connect_flags |= OBD_CONNECT_SOM; | |
190 | ||
191 | if (sbi->ll_flags & LL_SBI_LRU_RESIZE) | |
192 | data->ocd_connect_flags |= OBD_CONNECT_LRU_RESIZE; | |
193 | #ifdef CONFIG_FS_POSIX_ACL | |
194 | data->ocd_connect_flags |= OBD_CONNECT_ACL | OBD_CONNECT_UMASK; | |
195 | #endif | |
196 | ||
197 | if (OBD_FAIL_CHECK(OBD_FAIL_MDC_LIGHTWEIGHT)) | |
198 | /* flag mdc connection as lightweight, only used for test | |
c0894c6c OD |
199 | * purpose, use with care |
200 | */ | |
d7e09d03 PT |
201 | data->ocd_connect_flags |= OBD_CONNECT_LIGHTWEIGHT; |
202 | ||
203 | data->ocd_ibits_known = MDS_INODELOCK_FULL; | |
204 | data->ocd_version = LUSTRE_VERSION_CODE; | |
205 | ||
206 | if (sb->s_flags & MS_RDONLY) | |
207 | data->ocd_connect_flags |= OBD_CONNECT_RDONLY; | |
208 | if (sbi->ll_flags & LL_SBI_USER_XATTR) | |
209 | data->ocd_connect_flags |= OBD_CONNECT_XATTR; | |
210 | ||
d7e09d03 PT |
211 | if (sbi->ll_flags & LL_SBI_FLOCK) |
212 | sbi->ll_fop = &ll_file_operations_flock; | |
213 | else if (sbi->ll_flags & LL_SBI_LOCALFLOCK) | |
214 | sbi->ll_fop = &ll_file_operations; | |
215 | else | |
216 | sbi->ll_fop = &ll_file_operations_noflock; | |
217 | ||
218 | /* real client */ | |
219 | data->ocd_connect_flags |= OBD_CONNECT_REAL; | |
220 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) | |
221 | data->ocd_connect_flags |= OBD_CONNECT_RMT_CLIENT_FORCE; | |
222 | ||
223 | data->ocd_brw_size = MD_MAX_BRW_SIZE; | |
224 | ||
e6768831 TJ |
225 | err = obd_connect(NULL, &sbi->ll_md_exp, obd, &sbi->ll_sb_uuid, |
226 | data, NULL); | |
d7e09d03 | 227 | if (err == -EBUSY) { |
2d00bd17 JP |
228 | LCONSOLE_ERROR_MSG(0x14f, "An MDT (md %s) is performing recovery, of which this client is not a part. Please wait for recovery to complete, abort, or time out.\n", |
229 | md); | |
34e1f2bb | 230 | goto out; |
d7e09d03 PT |
231 | } else if (err) { |
232 | CERROR("cannot connect to %s: rc = %d\n", md, err); | |
34e1f2bb | 233 | goto out; |
d7e09d03 PT |
234 | } |
235 | ||
236 | sbi->ll_md_exp->exp_connect_data = *data; | |
237 | ||
238 | err = obd_fid_init(sbi->ll_md_exp->exp_obd, sbi->ll_md_exp, | |
239 | LUSTRE_SEQ_METADATA); | |
240 | if (err) { | |
2d00bd17 JP |
241 | CERROR("%s: Can't init metadata layer FID infrastructure, rc = %d\n", |
242 | sbi->ll_md_exp->exp_obd->obd_name, err); | |
34e1f2bb | 243 | goto out_md; |
d7e09d03 PT |
244 | } |
245 | ||
246 | /* For mount, we only need fs info from MDT0, and also in DNE, it | |
247 | * can make sure the client can be mounted as long as MDT0 is | |
c0894c6c OD |
248 | * available |
249 | */ | |
d7e09d03 | 250 | err = obd_statfs(NULL, sbi->ll_md_exp, osfs, |
e15ba45d OD |
251 | cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), |
252 | OBD_STATFS_FOR_MDT0); | |
d7e09d03 | 253 | if (err) |
34e1f2bb | 254 | goto out_md_fid; |
d7e09d03 PT |
255 | |
256 | /* This needs to be after statfs to ensure connect has finished. | |
257 | * Note that "data" does NOT contain the valid connect reply. | |
258 | * If connecting to a 1.8 server there will be no LMV device, so | |
259 | * we can access the MDC export directly and exp_connect_flags will | |
260 | * be non-zero, but if accessing an upgraded 2.1 server it will | |
261 | * have the correct flags filled in. | |
c0894c6c OD |
262 | * XXX: fill in the LMV exp_connect_flags from MDC(s). |
263 | */ | |
d7e09d03 PT |
264 | valid = exp_connect_flags(sbi->ll_md_exp) & CLIENT_CONNECT_MDT_REQD; |
265 | if (exp_connect_flags(sbi->ll_md_exp) != 0 && | |
266 | valid != CLIENT_CONNECT_MDT_REQD) { | |
267 | char *buf; | |
268 | ||
09cbfeaf | 269 | buf = kzalloc(PAGE_SIZE, GFP_KERNEL); |
db562e81 GEHP |
270 | if (!buf) { |
271 | err = -ENOMEM; | |
272 | goto out_md_fid; | |
273 | } | |
09cbfeaf | 274 | obd_connect_flags2str(buf, PAGE_SIZE, |
d7e09d03 | 275 | valid ^ CLIENT_CONNECT_MDT_REQD, ","); |
2d00bd17 | 276 | LCONSOLE_ERROR_MSG(0x170, "Server %s does not support feature(s) needed for correct operation of this client (%s). Please upgrade server or downgrade client.\n", |
d7e09d03 | 277 | sbi->ll_md_exp->exp_obd->obd_name, buf); |
97903a26 | 278 | kfree(buf); |
34e1f2bb JL |
279 | err = -EPROTO; |
280 | goto out_md_fid; | |
d7e09d03 PT |
281 | } |
282 | ||
283 | size = sizeof(*data); | |
284 | err = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_CONN_DATA), | |
285 | KEY_CONN_DATA, &size, data, NULL); | |
286 | if (err) { | |
287 | CERROR("%s: Get connect data failed: rc = %d\n", | |
288 | sbi->ll_md_exp->exp_obd->obd_name, err); | |
34e1f2bb | 289 | goto out_md_fid; |
d7e09d03 PT |
290 | } |
291 | ||
292 | LASSERT(osfs->os_bsize); | |
293 | sb->s_blocksize = osfs->os_bsize; | |
294 | sb->s_blocksize_bits = log2(osfs->os_bsize); | |
295 | sb->s_magic = LL_SUPER_MAGIC; | |
296 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
297 | sbi->ll_namelen = osfs->os_namelen; | |
d7e09d03 PT |
298 | |
299 | if ((sbi->ll_flags & LL_SBI_USER_XATTR) && | |
300 | !(data->ocd_connect_flags & OBD_CONNECT_XATTR)) { | |
2d00bd17 | 301 | LCONSOLE_INFO("Disabling user_xattr feature because it is not supported on the server\n"); |
d7e09d03 PT |
302 | sbi->ll_flags &= ~LL_SBI_USER_XATTR; |
303 | } | |
304 | ||
305 | if (data->ocd_connect_flags & OBD_CONNECT_ACL) { | |
d7e09d03 | 306 | sb->s_flags |= MS_POSIXACL; |
d7e09d03 PT |
307 | sbi->ll_flags |= LL_SBI_ACL; |
308 | } else { | |
309 | LCONSOLE_INFO("client wants to enable acl, but mdt not!\n"); | |
d7e09d03 | 310 | sb->s_flags &= ~MS_POSIXACL; |
d7e09d03 PT |
311 | sbi->ll_flags &= ~LL_SBI_ACL; |
312 | } | |
313 | ||
314 | if (data->ocd_connect_flags & OBD_CONNECT_RMT_CLIENT) { | |
315 | if (!(sbi->ll_flags & LL_SBI_RMT_CLIENT)) { | |
316 | sbi->ll_flags |= LL_SBI_RMT_CLIENT; | |
317 | LCONSOLE_INFO("client is set as remote by default.\n"); | |
318 | } | |
319 | } else { | |
320 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) { | |
321 | sbi->ll_flags &= ~LL_SBI_RMT_CLIENT; | |
2d00bd17 | 322 | LCONSOLE_INFO("client claims to be remote, but server rejected, forced to be local.\n"); |
d7e09d03 PT |
323 | } |
324 | } | |
325 | ||
d7e09d03 PT |
326 | if (data->ocd_connect_flags & OBD_CONNECT_64BITHASH) |
327 | sbi->ll_flags |= LL_SBI_64BIT_HASH; | |
328 | ||
329 | if (data->ocd_connect_flags & OBD_CONNECT_BRW_SIZE) | |
330 | sbi->ll_md_brw_size = data->ocd_brw_size; | |
331 | else | |
09cbfeaf | 332 | sbi->ll_md_brw_size = PAGE_SIZE; |
d7e09d03 PT |
333 | |
334 | if (data->ocd_connect_flags & OBD_CONNECT_LAYOUTLOCK) { | |
335 | LCONSOLE_INFO("Layout lock feature supported.\n"); | |
336 | sbi->ll_flags |= LL_SBI_LAYOUT_LOCK; | |
337 | } | |
338 | ||
7fc1f831 AP |
339 | if (data->ocd_ibits_known & MDS_INODELOCK_XATTR) { |
340 | if (!(data->ocd_connect_flags & OBD_CONNECT_MAX_EASIZE)) { | |
341 | LCONSOLE_INFO( | |
342 | "%s: disabling xattr cache due to unknown maximum xattr size.\n", | |
343 | dt); | |
344 | } else { | |
345 | sbi->ll_flags |= LL_SBI_XATTR_CACHE; | |
346 | sbi->ll_xattr_cache_enabled = 1; | |
347 | } | |
348 | } | |
349 | ||
d7e09d03 PT |
350 | obd = class_name2obd(dt); |
351 | if (!obd) { | |
352 | CERROR("DT %s: not setup or attached\n", dt); | |
34e1f2bb JL |
353 | err = -ENODEV; |
354 | goto out_md_fid; | |
d7e09d03 PT |
355 | } |
356 | ||
357 | data->ocd_connect_flags = OBD_CONNECT_GRANT | OBD_CONNECT_VERSION | | |
358 | OBD_CONNECT_REQPORTAL | OBD_CONNECT_BRW_SIZE | | |
359 | OBD_CONNECT_CANCELSET | OBD_CONNECT_FID | | |
360 | OBD_CONNECT_SRVLOCK | OBD_CONNECT_TRUNCLOCK| | |
361 | OBD_CONNECT_AT | OBD_CONNECT_RMT_CLIENT | | |
362 | OBD_CONNECT_OSS_CAPA | OBD_CONNECT_VBR| | |
363 | OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH | | |
364 | OBD_CONNECT_MAXBYTES | | |
365 | OBD_CONNECT_EINPROGRESS | | |
366 | OBD_CONNECT_JOBSTATS | OBD_CONNECT_LVB_TYPE | | |
367 | OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_PINGLESS; | |
368 | ||
369 | if (sbi->ll_flags & LL_SBI_SOM_PREVIEW) | |
370 | data->ocd_connect_flags |= OBD_CONNECT_SOM; | |
371 | ||
372 | if (!OBD_FAIL_CHECK(OBD_FAIL_OSC_CONNECT_CKSUM)) { | |
373 | /* OBD_CONNECT_CKSUM should always be set, even if checksums are | |
374 | * disabled by default, because it can still be enabled on the | |
40cc864a | 375 | * fly via /sys. As a consequence, we still need to come to an |
c0894c6c OD |
376 | * agreement on the supported algorithms at connect time |
377 | */ | |
d7e09d03 PT |
378 | data->ocd_connect_flags |= OBD_CONNECT_CKSUM; |
379 | ||
380 | if (OBD_FAIL_CHECK(OBD_FAIL_OSC_CKSUM_ADLER_ONLY)) | |
381 | data->ocd_cksum_types = OBD_CKSUM_ADLER; | |
382 | else | |
383 | data->ocd_cksum_types = cksum_types_supported_client(); | |
384 | } | |
385 | ||
386 | data->ocd_connect_flags |= OBD_CONNECT_LRU_RESIZE; | |
387 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) | |
388 | data->ocd_connect_flags |= OBD_CONNECT_RMT_CLIENT_FORCE; | |
389 | ||
2d00bd17 JP |
390 | CDEBUG(D_RPCTRACE, "ocd_connect_flags: %#llx ocd_version: %d ocd_grant: %d\n", |
391 | data->ocd_connect_flags, | |
d7e09d03 PT |
392 | data->ocd_version, data->ocd_grant); |
393 | ||
394 | obd->obd_upcall.onu_owner = &sbi->ll_lco; | |
395 | obd->obd_upcall.onu_upcall = cl_ocd_update; | |
396 | ||
397 | data->ocd_brw_size = DT_MAX_BRW_SIZE; | |
398 | ||
399 | err = obd_connect(NULL, &sbi->ll_dt_exp, obd, &sbi->ll_sb_uuid, data, | |
400 | NULL); | |
401 | if (err == -EBUSY) { | |
2d00bd17 JP |
402 | LCONSOLE_ERROR_MSG(0x150, "An OST (dt %s) is performing recovery, of which this client is not a part. Please wait for recovery to complete, abort, or time out.\n", |
403 | dt); | |
34e1f2bb | 404 | goto out_md; |
d7e09d03 PT |
405 | } else if (err) { |
406 | CERROR("%s: Cannot connect to %s: rc = %d\n", | |
407 | sbi->ll_dt_exp->exp_obd->obd_name, dt, err); | |
34e1f2bb | 408 | goto out_md; |
d7e09d03 PT |
409 | } |
410 | ||
411 | sbi->ll_dt_exp->exp_connect_data = *data; | |
412 | ||
413 | err = obd_fid_init(sbi->ll_dt_exp->exp_obd, sbi->ll_dt_exp, | |
414 | LUSTRE_SEQ_METADATA); | |
415 | if (err) { | |
2d00bd17 JP |
416 | CERROR("%s: Can't init data layer FID infrastructure, rc = %d\n", |
417 | sbi->ll_dt_exp->exp_obd->obd_name, err); | |
34e1f2bb | 418 | goto out_dt; |
d7e09d03 PT |
419 | } |
420 | ||
421 | mutex_lock(&sbi->ll_lco.lco_lock); | |
422 | sbi->ll_lco.lco_flags = data->ocd_connect_flags; | |
423 | sbi->ll_lco.lco_md_exp = sbi->ll_md_exp; | |
424 | sbi->ll_lco.lco_dt_exp = sbi->ll_dt_exp; | |
425 | mutex_unlock(&sbi->ll_lco.lco_lock); | |
426 | ||
427 | fid_zero(&sbi->ll_root_fid); | |
ef2e0f55 | 428 | err = md_getstatus(sbi->ll_md_exp, &sbi->ll_root_fid); |
d7e09d03 PT |
429 | if (err) { |
430 | CERROR("cannot mds_connect: rc = %d\n", err); | |
34e1f2bb | 431 | goto out_lock_cn_cb; |
d7e09d03 PT |
432 | } |
433 | if (!fid_is_sane(&sbi->ll_root_fid)) { | |
434 | CERROR("%s: Invalid root fid "DFID" during mount\n", | |
435 | sbi->ll_md_exp->exp_obd->obd_name, | |
436 | PFID(&sbi->ll_root_fid)); | |
34e1f2bb JL |
437 | err = -EINVAL; |
438 | goto out_lock_cn_cb; | |
d7e09d03 PT |
439 | } |
440 | CDEBUG(D_SUPER, "rootfid "DFID"\n", PFID(&sbi->ll_root_fid)); | |
441 | ||
442 | sb->s_op = &lustre_super_operations; | |
443 | #if THREAD_SIZE >= 8192 /*b=17630*/ | |
444 | sb->s_export_op = &lustre_export_operations; | |
445 | #endif | |
446 | ||
447 | /* make root inode | |
c0894c6c OD |
448 | * XXX: move this to after cbd setup? |
449 | */ | |
ef2e0f55 | 450 | valid = OBD_MD_FLGETATTR | OBD_MD_FLBLOCKS; |
d7e09d03 PT |
451 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) |
452 | valid |= OBD_MD_FLRMTPERM; | |
453 | else if (sbi->ll_flags & LL_SBI_ACL) | |
454 | valid |= OBD_MD_FLACL; | |
455 | ||
496a51bd JL |
456 | op_data = kzalloc(sizeof(*op_data), GFP_NOFS); |
457 | if (!op_data) { | |
34e1f2bb JL |
458 | err = -ENOMEM; |
459 | goto out_lock_cn_cb; | |
460 | } | |
d7e09d03 PT |
461 | |
462 | op_data->op_fid1 = sbi->ll_root_fid; | |
463 | op_data->op_mode = 0; | |
d7e09d03 PT |
464 | op_data->op_valid = valid; |
465 | ||
466 | err = md_getattr(sbi->ll_md_exp, op_data, &request); | |
97903a26 | 467 | kfree(op_data); |
d7e09d03 PT |
468 | if (err) { |
469 | CERROR("%s: md_getattr failed for root: rc = %d\n", | |
470 | sbi->ll_md_exp->exp_obd->obd_name, err); | |
34e1f2bb | 471 | goto out_lock_cn_cb; |
d7e09d03 PT |
472 | } |
473 | ||
474 | err = md_get_lustre_md(sbi->ll_md_exp, request, sbi->ll_dt_exp, | |
475 | sbi->ll_md_exp, &lmd); | |
476 | if (err) { | |
477 | CERROR("failed to understand root inode md: rc = %d\n", err); | |
478 | ptlrpc_req_finished(request); | |
34e1f2bb | 479 | goto out_lock_cn_cb; |
d7e09d03 PT |
480 | } |
481 | ||
482 | LASSERT(fid_is_sane(&sbi->ll_root_fid)); | |
483 | root = ll_iget(sb, cl_fid_build_ino(&sbi->ll_root_fid, | |
c1e2699d | 484 | sbi->ll_flags & LL_SBI_32BIT_API), |
d7e09d03 PT |
485 | &lmd); |
486 | md_free_lustre_md(sbi->ll_md_exp, &lmd); | |
487 | ptlrpc_req_finished(request); | |
488 | ||
020ecc6f | 489 | if (!(root)) { |
d7e09d03 PT |
490 | if (lmd.lsm) |
491 | obd_free_memmd(sbi->ll_dt_exp, &lmd.lsm); | |
492 | #ifdef CONFIG_FS_POSIX_ACL | |
493 | if (lmd.posix_acl) { | |
494 | posix_acl_release(lmd.posix_acl); | |
495 | lmd.posix_acl = NULL; | |
496 | } | |
497 | #endif | |
020ecc6f | 498 | err = -EBADF; |
d7e09d03 | 499 | CERROR("lustre_lite: bad iget4 for root\n"); |
34e1f2bb | 500 | goto out_root; |
d7e09d03 PT |
501 | } |
502 | ||
503 | err = ll_close_thread_start(&sbi->ll_lcq); | |
504 | if (err) { | |
505 | CERROR("cannot start close thread: rc %d\n", err); | |
34e1f2bb | 506 | goto out_root; |
d7e09d03 PT |
507 | } |
508 | ||
509 | #ifdef CONFIG_FS_POSIX_ACL | |
510 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) { | |
511 | rct_init(&sbi->ll_rct); | |
512 | et_init(&sbi->ll_et); | |
513 | } | |
514 | #endif | |
515 | ||
516 | checksum = sbi->ll_flags & LL_SBI_CHECKSUM; | |
517 | err = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_CHECKSUM), | |
518 | KEY_CHECKSUM, sizeof(checksum), &checksum, | |
519 | NULL); | |
520 | cl_sb_init(sb); | |
521 | ||
522 | err = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_CACHE_SET), | |
523 | KEY_CACHE_SET, sizeof(sbi->ll_cache), | |
524 | &sbi->ll_cache, NULL); | |
525 | ||
526 | sb->s_root = d_make_root(root); | |
6e16818b | 527 | if (!sb->s_root) { |
d7e09d03 | 528 | CERROR("%s: can't make root dentry\n", |
e15ba45d | 529 | ll_get_fsname(sb, NULL, 0)); |
34e1f2bb | 530 | err = -ENOMEM; |
caf382fe | 531 | goto out_lock_cn_cb; |
d7e09d03 PT |
532 | } |
533 | ||
d7e09d03 PT |
534 | sbi->ll_sdev_orig = sb->s_dev; |
535 | ||
536 | /* We set sb->s_dev equal on all lustre clients in order to support | |
537 | * NFS export clustering. NFSD requires that the FSID be the same | |
c0894c6c OD |
538 | * on all clients. |
539 | */ | |
d7e09d03 | 540 | /* s_dev is also used in lt_compare() to compare two fs, but that is |
c0894c6c OD |
541 | * only a node-local comparison. |
542 | */ | |
d7e09d03 | 543 | uuid = obd_get_uuid(sbi->ll_md_exp); |
6e16818b | 544 | if (uuid) { |
d7e09d03 | 545 | sb->s_dev = get_uuid2int(uuid->uuid, strlen(uuid->uuid)); |
bd994071 FY |
546 | get_uuid2fsid(uuid->uuid, strlen(uuid->uuid), &sbi->ll_fsid); |
547 | } | |
d7e09d03 | 548 | |
081825f5 JL |
549 | kfree(data); |
550 | kfree(osfs); | |
d7e09d03 | 551 | |
46dfb5aa GM |
552 | if (llite_root) { |
553 | err = ldebugfs_register_mountpoint(llite_root, sb, dt, md); | |
554 | if (err < 0) { | |
555 | CERROR("%s: could not register mount in debugfs: " | |
556 | "rc = %d\n", ll_get_fsname(sb, NULL, 0), err); | |
557 | err = 0; | |
558 | } | |
559 | } | |
560 | ||
0a3bdb00 | 561 | return err; |
d7e09d03 | 562 | out_root: |
ddafd514 | 563 | iput(root); |
d7e09d03 PT |
564 | out_lock_cn_cb: |
565 | obd_fid_fini(sbi->ll_dt_exp->exp_obd); | |
566 | out_dt: | |
567 | obd_disconnect(sbi->ll_dt_exp); | |
568 | sbi->ll_dt_exp = NULL; | |
569 | /* Make sure all OScs are gone, since cl_cache is accessing sbi. */ | |
570 | obd_zombie_barrier(); | |
571 | out_md_fid: | |
572 | obd_fid_fini(sbi->ll_md_exp->exp_obd); | |
573 | out_md: | |
574 | obd_disconnect(sbi->ll_md_exp); | |
575 | sbi->ll_md_exp = NULL; | |
576 | out: | |
081825f5 JL |
577 | kfree(data); |
578 | kfree(osfs); | |
d7e09d03 PT |
579 | return err; |
580 | } | |
581 | ||
582 | int ll_get_max_mdsize(struct ll_sb_info *sbi, int *lmmsize) | |
583 | { | |
584 | int size, rc; | |
585 | ||
586 | *lmmsize = obd_size_diskmd(sbi->ll_dt_exp, NULL); | |
587 | size = sizeof(int); | |
588 | rc = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_MAX_EASIZE), | |
589 | KEY_MAX_EASIZE, &size, lmmsize, NULL); | |
590 | if (rc) | |
4f211c20 | 591 | CERROR("Get max mdsize error rc %d\n", rc); |
d7e09d03 | 592 | |
0a3bdb00 | 593 | return rc; |
44779340 BB |
594 | } |
595 | ||
596 | int ll_get_default_mdsize(struct ll_sb_info *sbi, int *lmmsize) | |
597 | { | |
598 | int size, rc; | |
599 | ||
600 | size = sizeof(int); | |
601 | rc = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_DEFAULT_EASIZE), | |
e15ba45d | 602 | KEY_DEFAULT_EASIZE, &size, lmmsize, NULL); |
44779340 BB |
603 | if (rc) |
604 | CERROR("Get default mdsize error rc %d\n", rc); | |
605 | ||
606 | return rc; | |
607 | } | |
608 | ||
2d95f10e | 609 | static void client_common_put_super(struct super_block *sb) |
d7e09d03 PT |
610 | { |
611 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
d7e09d03 PT |
612 | |
613 | #ifdef CONFIG_FS_POSIX_ACL | |
614 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) { | |
615 | et_fini(&sbi->ll_et); | |
616 | rct_fini(&sbi->ll_rct); | |
617 | } | |
618 | #endif | |
619 | ||
620 | ll_close_thread_shutdown(sbi->ll_lcq); | |
621 | ||
622 | cl_sb_fini(sb); | |
623 | ||
d7e09d03 PT |
624 | obd_fid_fini(sbi->ll_dt_exp->exp_obd); |
625 | obd_disconnect(sbi->ll_dt_exp); | |
626 | sbi->ll_dt_exp = NULL; | |
627 | /* wait till all OSCs are gone, since cl_cache is accessing sbi. | |
c0894c6c OD |
628 | * see LU-2543. |
629 | */ | |
d7e09d03 PT |
630 | obd_zombie_barrier(); |
631 | ||
ae7c0f48 | 632 | ldebugfs_unregister_mountpoint(sbi); |
d7e09d03 PT |
633 | |
634 | obd_fid_fini(sbi->ll_md_exp->exp_obd); | |
635 | obd_disconnect(sbi->ll_md_exp); | |
636 | sbi->ll_md_exp = NULL; | |
d7e09d03 PT |
637 | } |
638 | ||
639 | void ll_kill_super(struct super_block *sb) | |
640 | { | |
641 | struct ll_sb_info *sbi; | |
642 | ||
d7e09d03 PT |
643 | /* not init sb ?*/ |
644 | if (!(sb->s_flags & MS_ACTIVE)) | |
645 | return; | |
646 | ||
647 | sbi = ll_s2sbi(sb); | |
e6768831 TJ |
648 | /* we need to restore s_dev from changed for clustered NFS before |
649 | * put_super because new kernels have cached s_dev and change sb->s_dev | |
c0894c6c OD |
650 | * in put_super not affected real removing devices |
651 | */ | |
65fb55d1 | 652 | if (sbi) { |
d7e09d03 | 653 | sb->s_dev = sbi->ll_sdev_orig; |
65fb55d1 NY |
654 | sbi->ll_umounting = 1; |
655 | } | |
d7e09d03 PT |
656 | } |
657 | ||
d7e09d03 PT |
658 | static inline int ll_set_opt(const char *opt, char *data, int fl) |
659 | { | |
660 | if (strncmp(opt, data, strlen(opt)) != 0) | |
fbe7c6c7 | 661 | return 0; |
d7e09d03 | 662 | else |
fbe7c6c7 | 663 | return fl; |
d7e09d03 PT |
664 | } |
665 | ||
666 | /* non-client-specific mount options are parsed in lmd_parse */ | |
667 | static int ll_options(char *options, int *flags) | |
668 | { | |
669 | int tmp; | |
670 | char *s1 = options, *s2; | |
d7e09d03 PT |
671 | |
672 | if (!options) | |
0a3bdb00 | 673 | return 0; |
d7e09d03 PT |
674 | |
675 | CDEBUG(D_CONFIG, "Parsing opts %s\n", options); | |
676 | ||
677 | while (*s1) { | |
678 | CDEBUG(D_SUPER, "next opt=%s\n", s1); | |
679 | tmp = ll_set_opt("nolock", s1, LL_SBI_NOLCK); | |
680 | if (tmp) { | |
681 | *flags |= tmp; | |
682 | goto next; | |
683 | } | |
684 | tmp = ll_set_opt("flock", s1, LL_SBI_FLOCK); | |
685 | if (tmp) { | |
686 | *flags |= tmp; | |
687 | goto next; | |
688 | } | |
689 | tmp = ll_set_opt("localflock", s1, LL_SBI_LOCALFLOCK); | |
690 | if (tmp) { | |
691 | *flags |= tmp; | |
692 | goto next; | |
693 | } | |
694 | tmp = ll_set_opt("noflock", s1, LL_SBI_FLOCK|LL_SBI_LOCALFLOCK); | |
695 | if (tmp) { | |
696 | *flags &= ~tmp; | |
697 | goto next; | |
698 | } | |
699 | tmp = ll_set_opt("user_xattr", s1, LL_SBI_USER_XATTR); | |
700 | if (tmp) { | |
701 | *flags |= tmp; | |
702 | goto next; | |
703 | } | |
704 | tmp = ll_set_opt("nouser_xattr", s1, LL_SBI_USER_XATTR); | |
705 | if (tmp) { | |
706 | *flags &= ~tmp; | |
707 | goto next; | |
708 | } | |
d7e09d03 PT |
709 | tmp = ll_set_opt("remote_client", s1, LL_SBI_RMT_CLIENT); |
710 | if (tmp) { | |
711 | *flags |= tmp; | |
712 | goto next; | |
713 | } | |
714 | tmp = ll_set_opt("user_fid2path", s1, LL_SBI_USER_FID2PATH); | |
715 | if (tmp) { | |
716 | *flags |= tmp; | |
717 | goto next; | |
718 | } | |
719 | tmp = ll_set_opt("nouser_fid2path", s1, LL_SBI_USER_FID2PATH); | |
720 | if (tmp) { | |
721 | *flags &= ~tmp; | |
722 | goto next; | |
723 | } | |
724 | ||
725 | tmp = ll_set_opt("checksum", s1, LL_SBI_CHECKSUM); | |
726 | if (tmp) { | |
727 | *flags |= tmp; | |
728 | goto next; | |
729 | } | |
730 | tmp = ll_set_opt("nochecksum", s1, LL_SBI_CHECKSUM); | |
731 | if (tmp) { | |
732 | *flags &= ~tmp; | |
733 | goto next; | |
734 | } | |
735 | tmp = ll_set_opt("lruresize", s1, LL_SBI_LRU_RESIZE); | |
736 | if (tmp) { | |
737 | *flags |= tmp; | |
738 | goto next; | |
739 | } | |
740 | tmp = ll_set_opt("nolruresize", s1, LL_SBI_LRU_RESIZE); | |
741 | if (tmp) { | |
742 | *flags &= ~tmp; | |
743 | goto next; | |
744 | } | |
745 | tmp = ll_set_opt("lazystatfs", s1, LL_SBI_LAZYSTATFS); | |
746 | if (tmp) { | |
747 | *flags |= tmp; | |
748 | goto next; | |
749 | } | |
750 | tmp = ll_set_opt("nolazystatfs", s1, LL_SBI_LAZYSTATFS); | |
751 | if (tmp) { | |
752 | *flags &= ~tmp; | |
753 | goto next; | |
754 | } | |
755 | tmp = ll_set_opt("som_preview", s1, LL_SBI_SOM_PREVIEW); | |
756 | if (tmp) { | |
757 | *flags |= tmp; | |
758 | goto next; | |
759 | } | |
760 | tmp = ll_set_opt("32bitapi", s1, LL_SBI_32BIT_API); | |
761 | if (tmp) { | |
762 | *flags |= tmp; | |
763 | goto next; | |
764 | } | |
765 | tmp = ll_set_opt("verbose", s1, LL_SBI_VERBOSE); | |
766 | if (tmp) { | |
767 | *flags |= tmp; | |
768 | goto next; | |
769 | } | |
770 | tmp = ll_set_opt("noverbose", s1, LL_SBI_VERBOSE); | |
771 | if (tmp) { | |
772 | *flags &= ~tmp; | |
773 | goto next; | |
774 | } | |
775 | LCONSOLE_ERROR_MSG(0x152, "Unknown option '%s', won't mount.\n", | |
776 | s1); | |
0a3bdb00 | 777 | return -EINVAL; |
d7e09d03 PT |
778 | |
779 | next: | |
780 | /* Find next opt */ | |
781 | s2 = strchr(s1, ','); | |
6e16818b | 782 | if (!s2) |
d7e09d03 PT |
783 | break; |
784 | s1 = s2 + 1; | |
785 | } | |
0a3bdb00 | 786 | return 0; |
d7e09d03 PT |
787 | } |
788 | ||
789 | void ll_lli_init(struct ll_inode_info *lli) | |
790 | { | |
791 | lli->lli_inode_magic = LLI_INODE_MAGIC; | |
792 | lli->lli_flags = 0; | |
793 | lli->lli_ioepoch = 0; | |
794 | lli->lli_maxbytes = MAX_LFS_FILESIZE; | |
795 | spin_lock_init(&lli->lli_lock); | |
796 | lli->lli_posix_acl = NULL; | |
797 | lli->lli_remote_perms = NULL; | |
798 | mutex_init(&lli->lli_rmtperm_mutex); | |
799 | /* Do not set lli_fid, it has been initialized already. */ | |
800 | fid_zero(&lli->lli_pfid); | |
801 | INIT_LIST_HEAD(&lli->lli_close_list); | |
d7e09d03 PT |
802 | lli->lli_rmtperm_time = 0; |
803 | lli->lli_pending_och = NULL; | |
804 | lli->lli_mds_read_och = NULL; | |
805 | lli->lli_mds_write_och = NULL; | |
806 | lli->lli_mds_exec_och = NULL; | |
807 | lli->lli_open_fd_read_count = 0; | |
808 | lli->lli_open_fd_write_count = 0; | |
809 | lli->lli_open_fd_exec_count = 0; | |
810 | mutex_init(&lli->lli_och_mutex); | |
811 | spin_lock_init(&lli->lli_agl_lock); | |
812 | lli->lli_has_smd = false; | |
09aed8a5 JX |
813 | spin_lock_init(&lli->lli_layout_lock); |
814 | ll_layout_version_set(lli, LL_LAYOUT_GEN_NONE); | |
d7e09d03 PT |
815 | lli->lli_clob = NULL; |
816 | ||
7fc1f831 AP |
817 | init_rwsem(&lli->lli_xattrs_list_rwsem); |
818 | mutex_init(&lli->lli_xattrs_enq_lock); | |
819 | ||
d7e09d03 PT |
820 | LASSERT(lli->lli_vfs_inode.i_mode != 0); |
821 | if (S_ISDIR(lli->lli_vfs_inode.i_mode)) { | |
822 | mutex_init(&lli->lli_readdir_mutex); | |
823 | lli->lli_opendir_key = NULL; | |
824 | lli->lli_sai = NULL; | |
d7e09d03 PT |
825 | spin_lock_init(&lli->lli_sa_lock); |
826 | lli->lli_opendir_pid = 0; | |
827 | } else { | |
47a57bde | 828 | mutex_init(&lli->lli_size_mutex); |
d7e09d03 PT |
829 | lli->lli_symlink_name = NULL; |
830 | init_rwsem(&lli->lli_trunc_sem); | |
831 | mutex_init(&lli->lli_write_mutex); | |
832 | init_rwsem(&lli->lli_glimpse_sem); | |
833 | lli->lli_glimpse_time = 0; | |
834 | INIT_LIST_HEAD(&lli->lli_agl_list); | |
835 | lli->lli_agl_index = 0; | |
836 | lli->lli_async_rc = 0; | |
d7e09d03 PT |
837 | } |
838 | mutex_init(&lli->lli_layout_mutex); | |
839 | } | |
840 | ||
841 | static inline int ll_bdi_register(struct backing_dev_info *bdi) | |
842 | { | |
843 | static atomic_t ll_bdi_num = ATOMIC_INIT(0); | |
844 | ||
845 | bdi->name = "lustre"; | |
846 | return bdi_register(bdi, NULL, "lustre-%d", | |
847 | atomic_inc_return(&ll_bdi_num)); | |
848 | } | |
849 | ||
850 | int ll_fill_super(struct super_block *sb, struct vfsmount *mnt) | |
851 | { | |
852 | struct lustre_profile *lprof = NULL; | |
853 | struct lustre_sb_info *lsi = s2lsi(sb); | |
854 | struct ll_sb_info *sbi; | |
855 | char *dt = NULL, *md = NULL; | |
856 | char *profilenm = get_profile_name(sb); | |
857 | struct config_llog_instance *cfg; | |
d7e09d03 | 858 | int err; |
d7e09d03 PT |
859 | |
860 | CDEBUG(D_VFSTRACE, "VFS Op: sb %p\n", sb); | |
861 | ||
496a51bd JL |
862 | cfg = kzalloc(sizeof(*cfg), GFP_NOFS); |
863 | if (!cfg) | |
0a3bdb00 | 864 | return -ENOMEM; |
d7e09d03 PT |
865 | |
866 | try_module_get(THIS_MODULE); | |
867 | ||
868 | /* client additional sb info */ | |
fd0d04ba | 869 | lsi->lsi_llsbi = sbi = ll_init_sbi(sb); |
d7e09d03 PT |
870 | if (!sbi) { |
871 | module_put(THIS_MODULE); | |
97903a26 | 872 | kfree(cfg); |
0a3bdb00 | 873 | return -ENOMEM; |
d7e09d03 PT |
874 | } |
875 | ||
876 | err = ll_options(lsi->lsi_lmd->lmd_opts, &sbi->ll_flags); | |
877 | if (err) | |
34e1f2bb | 878 | goto out_free; |
d7e09d03 PT |
879 | |
880 | err = bdi_init(&lsi->lsi_bdi); | |
881 | if (err) | |
34e1f2bb | 882 | goto out_free; |
d7e09d03 | 883 | lsi->lsi_flags |= LSI_BDI_INITIALIZED; |
b4caecd4 | 884 | lsi->lsi_bdi.capabilities = 0; |
d7e09d03 PT |
885 | err = ll_bdi_register(&lsi->lsi_bdi); |
886 | if (err) | |
34e1f2bb | 887 | goto out_free; |
d7e09d03 PT |
888 | |
889 | sb->s_bdi = &lsi->lsi_bdi; | |
3ea8f3bc LS |
890 | /* kernel >= 2.6.38 store dentry operations in sb->s_d_op. */ |
891 | sb->s_d_op = &ll_d_ops; | |
d7e09d03 PT |
892 | |
893 | /* Generate a string unique to this super, in case some joker tries | |
c0894c6c OD |
894 | * to mount the same fs at two mount points. |
895 | * Use the address of the super itself. | |
896 | */ | |
d7e09d03 PT |
897 | cfg->cfg_instance = sb; |
898 | cfg->cfg_uuid = lsi->lsi_llsbi->ll_sb_uuid; | |
899 | cfg->cfg_callback = class_config_llog_handler; | |
900 | /* set up client obds */ | |
901 | err = lustre_process_log(sb, profilenm, cfg); | |
902 | if (err < 0) { | |
903 | CERROR("Unable to process log: %d\n", err); | |
34e1f2bb | 904 | goto out_free; |
d7e09d03 PT |
905 | } |
906 | ||
907 | /* Profile set with LCFG_MOUNTOPT so we can find our mdc and osc obds */ | |
908 | lprof = class_get_profile(profilenm); | |
6e16818b | 909 | if (!lprof) { |
2d00bd17 JP |
910 | LCONSOLE_ERROR_MSG(0x156, "The client profile '%s' could not be read from the MGS. Does that filesystem exist?\n", |
911 | profilenm); | |
34e1f2bb JL |
912 | err = -EINVAL; |
913 | goto out_free; | |
d7e09d03 PT |
914 | } |
915 | CDEBUG(D_CONFIG, "Found profile %s: mdc=%s osc=%s\n", profilenm, | |
916 | lprof->lp_md, lprof->lp_dt); | |
917 | ||
95745e9b | 918 | dt = kasprintf(GFP_NOFS, "%s-%p", lprof->lp_dt, cfg->cfg_instance); |
34e1f2bb JL |
919 | if (!dt) { |
920 | err = -ENOMEM; | |
921 | goto out_free; | |
922 | } | |
d7e09d03 | 923 | |
ef2e1a44 | 924 | md = kasprintf(GFP_NOFS, "%s-%p", lprof->lp_md, cfg->cfg_instance); |
34e1f2bb JL |
925 | if (!md) { |
926 | err = -ENOMEM; | |
927 | goto out_free; | |
928 | } | |
d7e09d03 PT |
929 | |
930 | /* connections, registrations, sb setup */ | |
931 | err = client_common_fill_super(sb, md, dt, mnt); | |
932 | ||
933 | out_free: | |
0550db92 | 934 | kfree(md); |
935 | kfree(dt); | |
d7e09d03 PT |
936 | if (err) |
937 | ll_put_super(sb); | |
938 | else if (sbi->ll_flags & LL_SBI_VERBOSE) | |
939 | LCONSOLE_WARN("Mounted %s\n", profilenm); | |
940 | ||
97903a26 | 941 | kfree(cfg); |
0a3bdb00 | 942 | return err; |
d7e09d03 PT |
943 | } /* ll_fill_super */ |
944 | ||
d7e09d03 PT |
945 | void ll_put_super(struct super_block *sb) |
946 | { | |
7d4bae45 | 947 | struct config_llog_instance cfg, params_cfg; |
d7e09d03 PT |
948 | struct obd_device *obd; |
949 | struct lustre_sb_info *lsi = s2lsi(sb); | |
950 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
951 | char *profilenm = get_profile_name(sb); | |
ac5b1481 | 952 | int ccc_count, next, force = 1, rc = 0; |
d7e09d03 PT |
953 | |
954 | CDEBUG(D_VFSTRACE, "VFS Op: sb %p - %s\n", sb, profilenm); | |
955 | ||
d7e09d03 PT |
956 | cfg.cfg_instance = sb; |
957 | lustre_end_log(sb, profilenm, &cfg); | |
958 | ||
7d4bae45 AB |
959 | params_cfg.cfg_instance = sb; |
960 | lustre_end_log(sb, PARAMS_FILENAME, ¶ms_cfg); | |
961 | ||
d7e09d03 PT |
962 | if (sbi->ll_md_exp) { |
963 | obd = class_exp2obd(sbi->ll_md_exp); | |
964 | if (obd) | |
965 | force = obd->obd_force; | |
966 | } | |
967 | ||
ac5b1481 PS |
968 | /* Wait for unstable pages to be committed to stable storage */ |
969 | if (!force) { | |
970 | struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL); | |
971 | ||
972 | rc = l_wait_event(sbi->ll_cache.ccc_unstable_waitq, | |
973 | !atomic_read(&sbi->ll_cache.ccc_unstable_nr), | |
974 | &lwi); | |
975 | } | |
976 | ||
977 | ccc_count = atomic_read(&sbi->ll_cache.ccc_unstable_nr); | |
978 | if (!force && rc != -EINTR) | |
979 | LASSERTF(!ccc_count, "count: %i\n", ccc_count); | |
980 | ||
d7e09d03 | 981 | /* We need to set force before the lov_disconnect in |
c0894c6c OD |
982 | * lustre_common_put_super, since l_d cleans up osc's as well. |
983 | */ | |
d7e09d03 PT |
984 | if (force) { |
985 | next = 0; | |
986 | while ((obd = class_devices_in_group(&sbi->ll_sb_uuid, | |
987 | &next)) != NULL) { | |
988 | obd->obd_force = force; | |
989 | } | |
990 | } | |
991 | ||
992 | if (sbi->ll_lcq) { | |
993 | /* Only if client_common_fill_super succeeded */ | |
994 | client_common_put_super(sb); | |
995 | } | |
996 | ||
997 | next = 0; | |
a15dbf99 | 998 | while ((obd = class_devices_in_group(&sbi->ll_sb_uuid, &next))) |
d7e09d03 | 999 | class_manual_cleanup(obd); |
d7e09d03 PT |
1000 | |
1001 | if (sbi->ll_flags & LL_SBI_VERBOSE) | |
1002 | LCONSOLE_WARN("Unmounted %s\n", profilenm ? profilenm : ""); | |
1003 | ||
1004 | if (profilenm) | |
1005 | class_del_profile(profilenm); | |
1006 | ||
1007 | if (lsi->lsi_flags & LSI_BDI_INITIALIZED) { | |
1008 | bdi_destroy(&lsi->lsi_bdi); | |
1009 | lsi->lsi_flags &= ~LSI_BDI_INITIALIZED; | |
1010 | } | |
1011 | ||
1012 | ll_free_sbi(sb); | |
1013 | lsi->lsi_llsbi = NULL; | |
1014 | ||
1015 | lustre_common_put_super(sb); | |
1016 | ||
26f98e82 JX |
1017 | cl_env_cache_purge(~0); |
1018 | ||
d7e09d03 | 1019 | module_put(THIS_MODULE); |
d7e09d03 PT |
1020 | } /* client_put_super */ |
1021 | ||
1022 | struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock) | |
1023 | { | |
1024 | struct inode *inode = NULL; | |
1025 | ||
1026 | /* NOTE: we depend on atomic igrab() -bzzz */ | |
1027 | lock_res_and_lock(lock); | |
1028 | if (lock->l_resource->lr_lvb_inode) { | |
aff9d8e8 | 1029 | struct ll_inode_info *lli; |
cf29a7b6 | 1030 | |
d7e09d03 PT |
1031 | lli = ll_i2info(lock->l_resource->lr_lvb_inode); |
1032 | if (lli->lli_inode_magic == LLI_INODE_MAGIC) { | |
1033 | inode = igrab(lock->l_resource->lr_lvb_inode); | |
1034 | } else { | |
1035 | inode = lock->l_resource->lr_lvb_inode; | |
1036 | LDLM_DEBUG_LIMIT(inode->i_state & I_FREEING ? D_INFO : | |
2d00bd17 | 1037 | D_WARNING, lock, "lr_lvb_inode %p is bogus: magic %08x", |
d7e09d03 PT |
1038 | lock->l_resource->lr_lvb_inode, |
1039 | lli->lli_inode_magic); | |
1040 | inode = NULL; | |
1041 | } | |
1042 | } | |
1043 | unlock_res_and_lock(lock); | |
1044 | return inode; | |
1045 | } | |
1046 | ||
d7e09d03 PT |
1047 | void ll_clear_inode(struct inode *inode) |
1048 | { | |
1049 | struct ll_inode_info *lli = ll_i2info(inode); | |
1050 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
d7e09d03 | 1051 | |
97a075cd JN |
1052 | CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n", |
1053 | PFID(ll_inode2fid(inode)), inode); | |
d7e09d03 PT |
1054 | |
1055 | if (S_ISDIR(inode->i_mode)) { | |
1056 | /* these should have been cleared in ll_file_release */ | |
6e16818b OD |
1057 | LASSERT(!lli->lli_opendir_key); |
1058 | LASSERT(!lli->lli_sai); | |
d7e09d03 PT |
1059 | LASSERT(lli->lli_opendir_pid == 0); |
1060 | } | |
1061 | ||
ae5ef67b | 1062 | spin_lock(&lli->lli_lock); |
d7e09d03 | 1063 | ll_i2info(inode)->lli_flags &= ~LLIF_MDS_SIZE_LOCK; |
ae5ef67b | 1064 | spin_unlock(&lli->lli_lock); |
d7e09d03 PT |
1065 | md_null_inode(sbi->ll_md_exp, ll_inode2fid(inode)); |
1066 | ||
1067 | LASSERT(!lli->lli_open_fd_write_count); | |
1068 | LASSERT(!lli->lli_open_fd_read_count); | |
1069 | LASSERT(!lli->lli_open_fd_exec_count); | |
1070 | ||
1071 | if (lli->lli_mds_write_och) | |
1072 | ll_md_real_close(inode, FMODE_WRITE); | |
1073 | if (lli->lli_mds_exec_och) | |
1074 | ll_md_real_close(inode, FMODE_EXEC); | |
1075 | if (lli->lli_mds_read_och) | |
1076 | ll_md_real_close(inode, FMODE_READ); | |
1077 | ||
a5cb8880 | 1078 | if (S_ISLNK(inode->i_mode)) { |
97903a26 | 1079 | kfree(lli->lli_symlink_name); |
d7e09d03 PT |
1080 | lli->lli_symlink_name = NULL; |
1081 | } | |
1082 | ||
7fc1f831 AP |
1083 | ll_xattr_cache_destroy(inode); |
1084 | ||
d7e09d03 | 1085 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) { |
6e16818b | 1086 | LASSERT(!lli->lli_posix_acl); |
d7e09d03 PT |
1087 | if (lli->lli_remote_perms) { |
1088 | free_rmtperm_hash(lli->lli_remote_perms); | |
1089 | lli->lli_remote_perms = NULL; | |
1090 | } | |
1091 | } | |
1092 | #ifdef CONFIG_FS_POSIX_ACL | |
1093 | else if (lli->lli_posix_acl) { | |
1094 | LASSERT(atomic_read(&lli->lli_posix_acl->a_refcount) == 1); | |
6e16818b | 1095 | LASSERT(!lli->lli_remote_perms); |
d7e09d03 PT |
1096 | posix_acl_release(lli->lli_posix_acl); |
1097 | lli->lli_posix_acl = NULL; | |
1098 | } | |
1099 | #endif | |
1100 | lli->lli_inode_magic = LLI_INODE_DEAD; | |
1101 | ||
d7e09d03 PT |
1102 | if (!S_ISDIR(inode->i_mode)) |
1103 | LASSERT(list_empty(&lli->lli_agl_list)); | |
1104 | ||
1105 | /* | |
1106 | * XXX This has to be done before lsm is freed below, because | |
1107 | * cl_object still uses inode lsm. | |
1108 | */ | |
1109 | cl_inode_fini(inode); | |
1110 | lli->lli_has_smd = false; | |
d7e09d03 PT |
1111 | } |
1112 | ||
b81f9b6d OD |
1113 | #define TIMES_SET_FLAGS (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET) |
1114 | ||
3d3ab8cc | 1115 | static int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data, |
e15ba45d | 1116 | struct md_open_data **mod) |
d7e09d03 PT |
1117 | { |
1118 | struct lustre_md md; | |
2b0143b5 | 1119 | struct inode *inode = d_inode(dentry); |
d7e09d03 PT |
1120 | struct ll_sb_info *sbi = ll_i2sbi(inode); |
1121 | struct ptlrpc_request *request = NULL; | |
1122 | int rc, ia_valid; | |
d7e09d03 PT |
1123 | |
1124 | op_data = ll_prep_md_op_data(op_data, inode, NULL, NULL, 0, 0, | |
1125 | LUSTRE_OPC_ANY, NULL); | |
1126 | if (IS_ERR(op_data)) | |
0a3bdb00 | 1127 | return PTR_ERR(op_data); |
d7e09d03 PT |
1128 | |
1129 | rc = md_setattr(sbi->ll_md_exp, op_data, NULL, 0, NULL, 0, | |
1130 | &request, mod); | |
1131 | if (rc) { | |
1132 | ptlrpc_req_finished(request); | |
1133 | if (rc == -ENOENT) { | |
1134 | clear_nlink(inode); | |
1135 | /* Unlinked special device node? Or just a race? | |
c0894c6c OD |
1136 | * Pretend we did everything. |
1137 | */ | |
d7e09d03 PT |
1138 | if (!S_ISREG(inode->i_mode) && |
1139 | !S_ISDIR(inode->i_mode)) { | |
1140 | ia_valid = op_data->op_attr.ia_valid; | |
1141 | op_data->op_attr.ia_valid &= ~TIMES_SET_FLAGS; | |
1142 | rc = simple_setattr(dentry, &op_data->op_attr); | |
1143 | op_data->op_attr.ia_valid = ia_valid; | |
1144 | } | |
1145 | } else if (rc != -EPERM && rc != -EACCES && rc != -ETXTBSY) { | |
1146 | CERROR("md_setattr fails: rc = %d\n", rc); | |
1147 | } | |
0a3bdb00 | 1148 | return rc; |
d7e09d03 PT |
1149 | } |
1150 | ||
1151 | rc = md_get_lustre_md(sbi->ll_md_exp, request, sbi->ll_dt_exp, | |
1152 | sbi->ll_md_exp, &md); | |
1153 | if (rc) { | |
1154 | ptlrpc_req_finished(request); | |
0a3bdb00 | 1155 | return rc; |
d7e09d03 PT |
1156 | } |
1157 | ||
251c4317 | 1158 | ia_valid = op_data->op_attr.ia_valid; |
ef2e0f55 | 1159 | /* inode size will be in cl_setattr_ost, can't do it now since dirty |
c0894c6c OD |
1160 | * cache is not cleared yet. |
1161 | */ | |
251c4317 JH |
1162 | op_data->op_attr.ia_valid &= ~(TIMES_SET_FLAGS | ATTR_SIZE); |
1163 | rc = simple_setattr(dentry, &op_data->op_attr); | |
1164 | op_data->op_attr.ia_valid = ia_valid; | |
1165 | ||
d7e09d03 PT |
1166 | /* Extract epoch data if obtained. */ |
1167 | op_data->op_handle = md.body->handle; | |
1168 | op_data->op_ioepoch = md.body->ioepoch; | |
1169 | ||
1170 | ll_update_inode(inode, &md); | |
1171 | ptlrpc_req_finished(request); | |
1172 | ||
0a3bdb00 | 1173 | return rc; |
d7e09d03 PT |
1174 | } |
1175 | ||
1176 | /* Close IO epoch and send Size-on-MDS attribute update. */ | |
1177 | static int ll_setattr_done_writing(struct inode *inode, | |
1178 | struct md_op_data *op_data, | |
1179 | struct md_open_data *mod) | |
1180 | { | |
1181 | struct ll_inode_info *lli = ll_i2info(inode); | |
1182 | int rc = 0; | |
d7e09d03 | 1183 | |
d7e09d03 | 1184 | if (!S_ISREG(inode->i_mode)) |
0a3bdb00 | 1185 | return 0; |
d7e09d03 | 1186 | |
b0f5aad5 | 1187 | CDEBUG(D_INODE, "Epoch %llu closed on "DFID" for truncate\n", |
d7e09d03 PT |
1188 | op_data->op_ioepoch, PFID(&lli->lli_fid)); |
1189 | ||
1190 | op_data->op_flags = MF_EPOCH_CLOSE; | |
1191 | ll_done_writing_attr(inode, op_data); | |
1192 | ll_pack_inode2opdata(inode, op_data, NULL); | |
1193 | ||
1194 | rc = md_done_writing(ll_i2sbi(inode)->ll_md_exp, op_data, mod); | |
9d927bc5 | 1195 | if (rc == -EAGAIN) |
d7e09d03 | 1196 | /* MDS has instructed us to obtain Size-on-MDS attribute |
c0894c6c OD |
1197 | * from OSTs and send setattr to back to MDS. |
1198 | */ | |
d7e09d03 | 1199 | rc = ll_som_update(inode, op_data); |
97a075cd JN |
1200 | else if (rc) { |
1201 | CERROR("%s: inode "DFID" mdc truncate failed: rc = %d\n", | |
1202 | ll_i2sbi(inode)->ll_md_exp->exp_obd->obd_name, | |
1203 | PFID(ll_inode2fid(inode)), rc); | |
1204 | } | |
0a3bdb00 | 1205 | return rc; |
d7e09d03 PT |
1206 | } |
1207 | ||
d7e09d03 PT |
1208 | /* If this inode has objects allocated to it (lsm != NULL), then the OST |
1209 | * object(s) determine the file size and mtime. Otherwise, the MDS will | |
1210 | * keep these values until such a time that objects are allocated for it. | |
1211 | * We do the MDS operations first, as it is checking permissions for us. | |
1212 | * We don't to the MDS RPC if there is nothing that we want to store there, | |
1213 | * otherwise there is no harm in updating mtime/atime on the MDS if we are | |
1214 | * going to do an RPC anyways. | |
1215 | * | |
1216 | * If we are doing a truncate, we will send the mtime and ctime updates | |
1217 | * to the OST with the punch RPC, otherwise we do an explicit setattr RPC. | |
1218 | * I don't believe it is possible to get e.g. ATTR_MTIME_SET and ATTR_SIZE | |
1219 | * at the same time. | |
a720b790 JL |
1220 | * |
1221 | * In case of HSMimport, we only set attr on MDS. | |
d7e09d03 | 1222 | */ |
a720b790 | 1223 | int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import) |
d7e09d03 | 1224 | { |
2b0143b5 | 1225 | struct inode *inode = d_inode(dentry); |
d7e09d03 PT |
1226 | struct ll_inode_info *lli = ll_i2info(inode); |
1227 | struct md_op_data *op_data = NULL; | |
1228 | struct md_open_data *mod = NULL; | |
5ea17d6c | 1229 | bool file_is_released = false; |
d7e09d03 | 1230 | int rc = 0, rc1 = 0; |
d7e09d03 | 1231 | |
97a075cd JN |
1232 | CDEBUG(D_VFSTRACE, "%s: setattr inode "DFID"(%p) from %llu to %llu, valid %x, hsm_import %d\n", |
1233 | ll_get_fsname(inode->i_sb, NULL, 0), PFID(&lli->lli_fid), inode, | |
1234 | i_size_read(inode), attr->ia_size, attr->ia_valid, hsm_import); | |
d7e09d03 PT |
1235 | |
1236 | if (attr->ia_valid & ATTR_SIZE) { | |
1237 | /* Check new size against VFS/VM file size limit and rlimit */ | |
1238 | rc = inode_newsize_ok(inode, attr->ia_size); | |
1239 | if (rc) | |
0a3bdb00 | 1240 | return rc; |
d7e09d03 PT |
1241 | |
1242 | /* The maximum Lustre file size is variable, based on the | |
1243 | * OST maximum object size and number of stripes. This | |
c0894c6c OD |
1244 | * needs another check in addition to the VFS check above. |
1245 | */ | |
d7e09d03 | 1246 | if (attr->ia_size > ll_file_maxbytes(inode)) { |
1d8cb70c | 1247 | CDEBUG(D_INODE, "file "DFID" too large %llu > %llu\n", |
d7e09d03 PT |
1248 | PFID(&lli->lli_fid), attr->ia_size, |
1249 | ll_file_maxbytes(inode)); | |
0a3bdb00 | 1250 | return -EFBIG; |
d7e09d03 PT |
1251 | } |
1252 | ||
1253 | attr->ia_valid |= ATTR_MTIME | ATTR_CTIME; | |
1254 | } | |
1255 | ||
1256 | /* POSIX: check before ATTR_*TIME_SET set (from inode_change_ok) */ | |
1257 | if (attr->ia_valid & TIMES_SET_FLAGS) { | |
4b1a25f0 | 1258 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && |
2eb90a75 | 1259 | !capable(CFS_CAP_FOWNER)) |
0a3bdb00 | 1260 | return -EPERM; |
d7e09d03 PT |
1261 | } |
1262 | ||
1263 | /* We mark all of the fields "set" so MDS/OST does not re-set them */ | |
1264 | if (attr->ia_valid & ATTR_CTIME) { | |
0f1c743b | 1265 | attr->ia_ctime = CURRENT_TIME; |
d7e09d03 PT |
1266 | attr->ia_valid |= ATTR_CTIME_SET; |
1267 | } | |
1268 | if (!(attr->ia_valid & ATTR_ATIME_SET) && | |
1269 | (attr->ia_valid & ATTR_ATIME)) { | |
0f1c743b | 1270 | attr->ia_atime = CURRENT_TIME; |
d7e09d03 PT |
1271 | attr->ia_valid |= ATTR_ATIME_SET; |
1272 | } | |
1273 | if (!(attr->ia_valid & ATTR_MTIME_SET) && | |
1274 | (attr->ia_valid & ATTR_MTIME)) { | |
0f1c743b | 1275 | attr->ia_mtime = CURRENT_TIME; |
d7e09d03 PT |
1276 | attr->ia_valid |= ATTR_MTIME_SET; |
1277 | } | |
1278 | ||
1279 | if (attr->ia_valid & (ATTR_MTIME | ATTR_CTIME)) | |
8d7eed54 | 1280 | CDEBUG(D_INODE, "setting mtime %lu, ctime %lu, now = %llu\n", |
d7e09d03 | 1281 | LTIME_S(attr->ia_mtime), LTIME_S(attr->ia_ctime), |
8d7eed54 | 1282 | (s64)ktime_get_real_seconds()); |
d7e09d03 | 1283 | |
d7e09d03 | 1284 | /* We always do an MDS RPC, even if we're only changing the size; |
c0894c6c OD |
1285 | * only the MDS knows whether truncate() should fail with -ETXTBUSY |
1286 | */ | |
d7e09d03 | 1287 | |
496a51bd JL |
1288 | op_data = kzalloc(sizeof(*op_data), GFP_NOFS); |
1289 | if (!op_data) | |
0a3bdb00 | 1290 | return -ENOMEM; |
d7e09d03 | 1291 | |
81e053c7 | 1292 | if (!S_ISDIR(inode->i_mode)) |
5955102c | 1293 | inode_unlock(inode); |
d7e09d03 | 1294 | |
5ea17d6c JL |
1295 | /* truncate on a released file must failed with -ENODATA, |
1296 | * so size must not be set on MDS for released file | |
1297 | * but other attributes must be set | |
1298 | */ | |
1299 | if (S_ISREG(inode->i_mode)) { | |
1300 | struct lov_stripe_md *lsm; | |
1301 | __u32 gen; | |
1302 | ||
1303 | ll_layout_refresh(inode, &gen); | |
1304 | lsm = ccc_inode_lsm_get(inode); | |
1305 | if (lsm && lsm->lsm_pattern & LOV_PATTERN_F_RELEASED) | |
1306 | file_is_released = true; | |
1307 | ccc_inode_lsm_put(inode, lsm); | |
1b1594da JX |
1308 | |
1309 | if (!hsm_import && attr->ia_valid & ATTR_SIZE) { | |
1310 | if (file_is_released) { | |
1311 | rc = ll_layout_restore(inode, 0, attr->ia_size); | |
1312 | if (rc < 0) | |
1313 | goto out; | |
1314 | ||
1315 | file_is_released = false; | |
1316 | ll_layout_refresh(inode, &gen); | |
1317 | } | |
1318 | ||
1319 | /* | |
1320 | * If we are changing file size, file content is | |
1321 | * modified, flag it. | |
1322 | */ | |
1323 | attr->ia_valid |= MDS_OPEN_OWNEROVERRIDE; | |
1324 | spin_lock(&lli->lli_lock); | |
1325 | lli->lli_flags |= LLIF_DATA_MODIFIED; | |
1326 | spin_unlock(&lli->lli_lock); | |
1327 | op_data->op_bias |= MDS_DATA_MODIFIED; | |
1328 | } | |
5ea17d6c JL |
1329 | } |
1330 | ||
1b1594da JX |
1331 | memcpy(&op_data->op_attr, attr, sizeof(*attr)); |
1332 | ||
1333 | /* Open epoch for truncate. */ | |
1334 | if (exp_connect_som(ll_i2mdexp(inode)) && !hsm_import && | |
1335 | (attr->ia_valid & (ATTR_SIZE | ATTR_MTIME | ATTR_MTIME_SET))) | |
1336 | op_data->op_flags = MF_EPOCH_OPEN; | |
5ea17d6c | 1337 | |
d7e09d03 PT |
1338 | rc = ll_md_setattr(dentry, op_data, &mod); |
1339 | if (rc) | |
34e1f2bb | 1340 | goto out; |
d7e09d03 PT |
1341 | |
1342 | /* RPC to MDT is sent, cancel data modification flag */ | |
5210a63a | 1343 | if (op_data->op_bias & MDS_DATA_MODIFIED) { |
d7e09d03 PT |
1344 | spin_lock(&lli->lli_lock); |
1345 | lli->lli_flags &= ~LLIF_DATA_MODIFIED; | |
1346 | spin_unlock(&lli->lli_lock); | |
1347 | } | |
1348 | ||
1349 | ll_ioepoch_open(lli, op_data->op_ioepoch); | |
1b1594da | 1350 | if (!S_ISREG(inode->i_mode) || file_is_released) { |
34e1f2bb JL |
1351 | rc = 0; |
1352 | goto out; | |
1353 | } | |
d7e09d03 PT |
1354 | |
1355 | if (attr->ia_valid & (ATTR_SIZE | | |
1356 | ATTR_ATIME | ATTR_ATIME_SET | | |
53bd4a00 | 1357 | ATTR_MTIME | ATTR_MTIME_SET)) { |
d7e09d03 PT |
1358 | /* For truncate and utimes sending attributes to OSTs, setting |
1359 | * mtime/atime to the past will be performed under PW [0:EOF] | |
1360 | * extent lock (new_size:EOF for truncate). It may seem | |
1361 | * excessive to send mtime/atime updates to OSTs when not | |
1362 | * setting times to past, but it is necessary due to possible | |
c0894c6c OD |
1363 | * time de-synchronization between MDT inode and OST objects |
1364 | */ | |
178ba1e0 BJ |
1365 | if (attr->ia_valid & ATTR_SIZE) |
1366 | down_write(&lli->lli_trunc_sem); | |
ef2e0f55 | 1367 | rc = cl_setattr_ost(inode, attr); |
178ba1e0 BJ |
1368 | if (attr->ia_valid & ATTR_SIZE) |
1369 | up_write(&lli->lli_trunc_sem); | |
53bd4a00 | 1370 | } |
d7e09d03 | 1371 | out: |
83d6b8fe SB |
1372 | if (op_data->op_ioepoch) { |
1373 | rc1 = ll_setattr_done_writing(inode, op_data, mod); | |
1374 | if (!rc) | |
1375 | rc = rc1; | |
d7e09d03 | 1376 | } |
83d6b8fe SB |
1377 | ll_finish_md_op_data(op_data); |
1378 | ||
d7e09d03 | 1379 | if (!S_ISDIR(inode->i_mode)) { |
5955102c | 1380 | inode_lock(inode); |
a720b790 | 1381 | if ((attr->ia_valid & ATTR_SIZE) && !hsm_import) |
d7e09d03 PT |
1382 | inode_dio_wait(inode); |
1383 | } | |
1384 | ||
1385 | ll_stats_ops_tally(ll_i2sbi(inode), (attr->ia_valid & ATTR_SIZE) ? | |
1386 | LPROC_LL_TRUNC : LPROC_LL_SETATTR, 1); | |
1387 | ||
251c4317 | 1388 | return rc; |
d7e09d03 PT |
1389 | } |
1390 | ||
1391 | int ll_setattr(struct dentry *de, struct iattr *attr) | |
1392 | { | |
2b0143b5 | 1393 | int mode = d_inode(de)->i_mode; |
d7e09d03 PT |
1394 | |
1395 | if ((attr->ia_valid & (ATTR_CTIME|ATTR_SIZE|ATTR_MODE)) == | |
1396 | (ATTR_CTIME|ATTR_SIZE|ATTR_MODE)) | |
1397 | attr->ia_valid |= MDS_OPEN_OWNEROVERRIDE; | |
1398 | ||
1399 | if (((attr->ia_valid & (ATTR_MODE|ATTR_FORCE|ATTR_SIZE)) == | |
1400 | (ATTR_SIZE|ATTR_MODE)) && | |
1401 | (((mode & S_ISUID) && !(attr->ia_mode & S_ISUID)) || | |
1402 | (((mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) && | |
1403 | !(attr->ia_mode & S_ISGID)))) | |
1404 | attr->ia_valid |= ATTR_FORCE; | |
1405 | ||
98639249 NC |
1406 | if ((attr->ia_valid & ATTR_MODE) && |
1407 | (mode & S_ISUID) && | |
d7e09d03 PT |
1408 | !(attr->ia_mode & S_ISUID) && |
1409 | !(attr->ia_valid & ATTR_KILL_SUID)) | |
1410 | attr->ia_valid |= ATTR_KILL_SUID; | |
1411 | ||
98639249 NC |
1412 | if ((attr->ia_valid & ATTR_MODE) && |
1413 | ((mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) && | |
d7e09d03 PT |
1414 | !(attr->ia_mode & S_ISGID) && |
1415 | !(attr->ia_valid & ATTR_KILL_SGID)) | |
1416 | attr->ia_valid |= ATTR_KILL_SGID; | |
1417 | ||
a720b790 | 1418 | return ll_setattr_raw(de, attr, false); |
d7e09d03 PT |
1419 | } |
1420 | ||
1421 | int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, | |
1422 | __u64 max_age, __u32 flags) | |
1423 | { | |
1424 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
1425 | struct obd_statfs obd_osfs; | |
1426 | int rc; | |
d7e09d03 PT |
1427 | |
1428 | rc = obd_statfs(NULL, sbi->ll_md_exp, osfs, max_age, flags); | |
1429 | if (rc) { | |
1430 | CERROR("md_statfs fails: rc = %d\n", rc); | |
0a3bdb00 | 1431 | return rc; |
d7e09d03 PT |
1432 | } |
1433 | ||
1434 | osfs->os_type = sb->s_magic; | |
1435 | ||
b0f5aad5 | 1436 | CDEBUG(D_SUPER, "MDC blocks %llu/%llu objects %llu/%llu\n", |
1d8cb70c GD |
1437 | osfs->os_bavail, osfs->os_blocks, osfs->os_ffree, |
1438 | osfs->os_files); | |
d7e09d03 PT |
1439 | |
1440 | if (sbi->ll_flags & LL_SBI_LAZYSTATFS) | |
1441 | flags |= OBD_STATFS_NODELAY; | |
1442 | ||
1443 | rc = obd_statfs_rqset(sbi->ll_dt_exp, &obd_osfs, max_age, flags); | |
1444 | if (rc) { | |
1445 | CERROR("obd_statfs fails: rc = %d\n", rc); | |
0a3bdb00 | 1446 | return rc; |
d7e09d03 PT |
1447 | } |
1448 | ||
b0f5aad5 | 1449 | CDEBUG(D_SUPER, "OSC blocks %llu/%llu objects %llu/%llu\n", |
d7e09d03 PT |
1450 | obd_osfs.os_bavail, obd_osfs.os_blocks, obd_osfs.os_ffree, |
1451 | obd_osfs.os_files); | |
1452 | ||
1453 | osfs->os_bsize = obd_osfs.os_bsize; | |
1454 | osfs->os_blocks = obd_osfs.os_blocks; | |
1455 | osfs->os_bfree = obd_osfs.os_bfree; | |
1456 | osfs->os_bavail = obd_osfs.os_bavail; | |
1457 | ||
1458 | /* If we don't have as many objects free on the OST as inodes | |
1459 | * on the MDS, we reduce the total number of inodes to | |
1460 | * compensate, so that the "inodes in use" number is correct. | |
1461 | */ | |
1462 | if (obd_osfs.os_ffree < osfs->os_ffree) { | |
1463 | osfs->os_files = (osfs->os_files - osfs->os_ffree) + | |
1464 | obd_osfs.os_ffree; | |
1465 | osfs->os_ffree = obd_osfs.os_ffree; | |
1466 | } | |
1467 | ||
0a3bdb00 | 1468 | return rc; |
d7e09d03 | 1469 | } |
c9f6bb96 | 1470 | |
d7e09d03 PT |
1471 | int ll_statfs(struct dentry *de, struct kstatfs *sfs) |
1472 | { | |
1473 | struct super_block *sb = de->d_sb; | |
1474 | struct obd_statfs osfs; | |
1475 | int rc; | |
1476 | ||
b0f5aad5 | 1477 | CDEBUG(D_VFSTRACE, "VFS Op: at %llu jiffies\n", get_jiffies_64()); |
d7e09d03 PT |
1478 | ll_stats_ops_tally(ll_s2sbi(sb), LPROC_LL_STAFS, 1); |
1479 | ||
1480 | /* Some amount of caching on the client is allowed */ | |
1481 | rc = ll_statfs_internal(sb, &osfs, | |
1482 | cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), | |
1483 | 0); | |
1484 | if (rc) | |
1485 | return rc; | |
1486 | ||
1487 | statfs_unpack(sfs, &osfs); | |
1488 | ||
1489 | /* We need to downshift for all 32-bit kernels, because we can't | |
1490 | * tell if the kernel is being called via sys_statfs64() or not. | |
1491 | * Stop before overflowing f_bsize - in which case it is better | |
c0894c6c OD |
1492 | * to just risk EOVERFLOW if caller is using old sys_statfs(). |
1493 | */ | |
d7e09d03 PT |
1494 | if (sizeof(long) < 8) { |
1495 | while (osfs.os_blocks > ~0UL && sfs->f_bsize < 0x40000000) { | |
1496 | sfs->f_bsize <<= 1; | |
1497 | ||
1498 | osfs.os_blocks >>= 1; | |
1499 | osfs.os_bfree >>= 1; | |
1500 | osfs.os_bavail >>= 1; | |
1501 | } | |
1502 | } | |
1503 | ||
1504 | sfs->f_blocks = osfs.os_blocks; | |
1505 | sfs->f_bfree = osfs.os_bfree; | |
1506 | sfs->f_bavail = osfs.os_bavail; | |
bd994071 | 1507 | sfs->f_fsid = ll_s2sbi(sb)->ll_fsid; |
d7e09d03 PT |
1508 | return 0; |
1509 | } | |
1510 | ||
1511 | void ll_inode_size_lock(struct inode *inode) | |
1512 | { | |
1513 | struct ll_inode_info *lli; | |
1514 | ||
1515 | LASSERT(!S_ISDIR(inode->i_mode)); | |
1516 | ||
1517 | lli = ll_i2info(inode); | |
47a57bde | 1518 | mutex_lock(&lli->lli_size_mutex); |
d7e09d03 PT |
1519 | } |
1520 | ||
1521 | void ll_inode_size_unlock(struct inode *inode) | |
1522 | { | |
1523 | struct ll_inode_info *lli; | |
1524 | ||
1525 | lli = ll_i2info(inode); | |
47a57bde | 1526 | mutex_unlock(&lli->lli_size_mutex); |
d7e09d03 PT |
1527 | } |
1528 | ||
1529 | void ll_update_inode(struct inode *inode, struct lustre_md *md) | |
1530 | { | |
1531 | struct ll_inode_info *lli = ll_i2info(inode); | |
1532 | struct mdt_body *body = md->body; | |
1533 | struct lov_stripe_md *lsm = md->lsm; | |
1534 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
1535 | ||
629ecb5b | 1536 | LASSERT((lsm != NULL) == ((body->valid & OBD_MD_FLEASIZE) != 0)); |
6e16818b | 1537 | if (lsm) { |
d7e09d03 PT |
1538 | if (!lli->lli_has_smd && |
1539 | !(sbi->ll_flags & LL_SBI_LAYOUT_LOCK)) | |
1540 | cl_file_inode_init(inode, md); | |
1541 | ||
1542 | lli->lli_maxbytes = lsm->lsm_maxbytes; | |
1543 | if (lli->lli_maxbytes > MAX_LFS_FILESIZE) | |
1544 | lli->lli_maxbytes = MAX_LFS_FILESIZE; | |
1545 | } | |
1546 | ||
1547 | if (sbi->ll_flags & LL_SBI_RMT_CLIENT) { | |
1548 | if (body->valid & OBD_MD_FLRMTPERM) | |
1549 | ll_update_remote_perm(inode, md->remote_perm); | |
1550 | } | |
1551 | #ifdef CONFIG_FS_POSIX_ACL | |
1552 | else if (body->valid & OBD_MD_FLACL) { | |
1553 | spin_lock(&lli->lli_lock); | |
1554 | if (lli->lli_posix_acl) | |
1555 | posix_acl_release(lli->lli_posix_acl); | |
1556 | lli->lli_posix_acl = md->posix_acl; | |
1557 | spin_unlock(&lli->lli_lock); | |
1558 | } | |
1559 | #endif | |
c1e2699d | 1560 | inode->i_ino = cl_fid_build_ino(&body->fid1, |
1561 | sbi->ll_flags & LL_SBI_32BIT_API); | |
d7e09d03 PT |
1562 | inode->i_generation = cl_fid_build_gen(&body->fid1); |
1563 | ||
1564 | if (body->valid & OBD_MD_FLATIME) { | |
1565 | if (body->atime > LTIME_S(inode->i_atime)) | |
1566 | LTIME_S(inode->i_atime) = body->atime; | |
d2995737 | 1567 | lli->lli_atime = body->atime; |
d7e09d03 PT |
1568 | } |
1569 | if (body->valid & OBD_MD_FLMTIME) { | |
1570 | if (body->mtime > LTIME_S(inode->i_mtime)) { | |
b0f5aad5 GKH |
1571 | CDEBUG(D_INODE, "setting ino %lu mtime from %lu to %llu\n", |
1572 | inode->i_ino, LTIME_S(inode->i_mtime), | |
1573 | body->mtime); | |
d7e09d03 PT |
1574 | LTIME_S(inode->i_mtime) = body->mtime; |
1575 | } | |
d2995737 | 1576 | lli->lli_mtime = body->mtime; |
d7e09d03 PT |
1577 | } |
1578 | if (body->valid & OBD_MD_FLCTIME) { | |
1579 | if (body->ctime > LTIME_S(inode->i_ctime)) | |
1580 | LTIME_S(inode->i_ctime) = body->ctime; | |
d2995737 | 1581 | lli->lli_ctime = body->ctime; |
d7e09d03 PT |
1582 | } |
1583 | if (body->valid & OBD_MD_FLMODE) | |
1584 | inode->i_mode = (inode->i_mode & S_IFMT)|(body->mode & ~S_IFMT); | |
1585 | if (body->valid & OBD_MD_FLTYPE) | |
1586 | inode->i_mode = (inode->i_mode & ~S_IFMT)|(body->mode & S_IFMT); | |
1587 | LASSERT(inode->i_mode != 0); | |
566be54d | 1588 | if (S_ISREG(inode->i_mode)) |
e6768831 TJ |
1589 | inode->i_blkbits = min(PTLRPC_MAX_BRW_BITS + 1, |
1590 | LL_MAX_BLKSIZE_BITS); | |
566be54d | 1591 | else |
d7e09d03 | 1592 | inode->i_blkbits = inode->i_sb->s_blocksize_bits; |
d7e09d03 | 1593 | if (body->valid & OBD_MD_FLUID) |
4b1a25f0 | 1594 | inode->i_uid = make_kuid(&init_user_ns, body->uid); |
d7e09d03 | 1595 | if (body->valid & OBD_MD_FLGID) |
4b1a25f0 | 1596 | inode->i_gid = make_kgid(&init_user_ns, body->gid); |
d7e09d03 PT |
1597 | if (body->valid & OBD_MD_FLFLAGS) |
1598 | inode->i_flags = ll_ext_to_inode_flags(body->flags); | |
1599 | if (body->valid & OBD_MD_FLNLINK) | |
1600 | set_nlink(inode, body->nlink); | |
1601 | if (body->valid & OBD_MD_FLRDEV) | |
1602 | inode->i_rdev = old_decode_dev(body->rdev); | |
1603 | ||
1604 | if (body->valid & OBD_MD_FLID) { | |
1605 | /* FID shouldn't be changed! */ | |
1606 | if (fid_is_sane(&lli->lli_fid)) { | |
1607 | LASSERTF(lu_fid_eq(&lli->lli_fid, &body->fid1), | |
97a075cd | 1608 | "Trying to change FID "DFID" to the "DFID", inode "DFID"(%p)\n", |
d7e09d03 | 1609 | PFID(&lli->lli_fid), PFID(&body->fid1), |
97a075cd | 1610 | PFID(ll_inode2fid(inode)), inode); |
da5ecb4d | 1611 | } else { |
d7e09d03 | 1612 | lli->lli_fid = body->fid1; |
da5ecb4d | 1613 | } |
d7e09d03 PT |
1614 | } |
1615 | ||
1616 | LASSERT(fid_seq(&lli->lli_fid) != 0); | |
1617 | ||
1618 | if (body->valid & OBD_MD_FLSIZE) { | |
1619 | if (exp_connect_som(ll_i2mdexp(inode)) && | |
1620 | S_ISREG(inode->i_mode)) { | |
1621 | struct lustre_handle lockh; | |
52ee0d20 | 1622 | enum ldlm_mode mode; |
d7e09d03 PT |
1623 | |
1624 | /* As it is possible a blocking ast has been processed | |
1625 | * by this time, we need to check there is an UPDATE | |
1626 | * lock on the client and set LLIF_MDS_SIZE_LOCK holding | |
c0894c6c OD |
1627 | * it. |
1628 | */ | |
d7e09d03 | 1629 | mode = ll_take_md_lock(inode, MDS_INODELOCK_UPDATE, |
7fc1f831 AP |
1630 | &lockh, LDLM_FL_CBPENDING, |
1631 | LCK_CR | LCK_CW | | |
1632 | LCK_PR | LCK_PW); | |
d7e09d03 PT |
1633 | if (mode) { |
1634 | if (lli->lli_flags & (LLIF_DONE_WRITING | | |
1635 | LLIF_EPOCH_PENDING | | |
1636 | LLIF_SOM_DIRTY)) { | |
97a075cd JN |
1637 | CERROR("%s: inode "DFID" flags %u still has size authority! do not trust the size got from MDS\n", |
1638 | sbi->ll_md_exp->exp_obd->obd_name, | |
1639 | PFID(ll_inode2fid(inode)), | |
1640 | lli->lli_flags); | |
d7e09d03 PT |
1641 | } else { |
1642 | /* Use old size assignment to avoid | |
c0894c6c OD |
1643 | * deadlock bz14138 & bz14326 |
1644 | */ | |
d7e09d03 | 1645 | i_size_write(inode, body->size); |
ae5ef67b | 1646 | spin_lock(&lli->lli_lock); |
d7e09d03 | 1647 | lli->lli_flags |= LLIF_MDS_SIZE_LOCK; |
ae5ef67b | 1648 | spin_unlock(&lli->lli_lock); |
d7e09d03 PT |
1649 | } |
1650 | ldlm_lock_decref(&lockh, mode); | |
1651 | } | |
1652 | } else { | |
1653 | /* Use old size assignment to avoid | |
c0894c6c OD |
1654 | * deadlock bz14138 & bz14326 |
1655 | */ | |
d7e09d03 PT |
1656 | i_size_write(inode, body->size); |
1657 | ||
1658 | CDEBUG(D_VFSTRACE, "inode=%lu, updating i_size %llu\n", | |
1659 | inode->i_ino, (unsigned long long)body->size); | |
1660 | } | |
1661 | ||
1662 | if (body->valid & OBD_MD_FLBLOCKS) | |
1663 | inode->i_blocks = body->blocks; | |
1664 | } | |
1665 | ||
5ea17d6c JL |
1666 | if (body->valid & OBD_MD_TSTATE) { |
1667 | if (body->t_state & MS_RESTORE) | |
1668 | lli->lli_flags |= LLIF_FILE_RESTORING; | |
1669 | } | |
d7e09d03 PT |
1670 | } |
1671 | ||
1672 | void ll_read_inode2(struct inode *inode, void *opaque) | |
1673 | { | |
1674 | struct lustre_md *md = opaque; | |
1675 | struct ll_inode_info *lli = ll_i2info(inode); | |
d7e09d03 PT |
1676 | |
1677 | CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n", | |
1678 | PFID(&lli->lli_fid), inode); | |
1679 | ||
1680 | LASSERT(!lli->lli_has_smd); | |
1681 | ||
1682 | /* Core attributes from the MDS first. This is a new inode, and | |
1683 | * the VFS doesn't zero times in the core inode so we have to do | |
1684 | * it ourselves. They will be overwritten by either MDS or OST | |
c0894c6c OD |
1685 | * attributes - we just need to make sure they aren't newer. |
1686 | */ | |
d7e09d03 PT |
1687 | LTIME_S(inode->i_mtime) = 0; |
1688 | LTIME_S(inode->i_atime) = 0; | |
1689 | LTIME_S(inode->i_ctime) = 0; | |
1690 | inode->i_rdev = 0; | |
1691 | ll_update_inode(inode, md); | |
1692 | ||
1693 | /* OIDEBUG(inode); */ | |
1694 | ||
d7e09d03 PT |
1695 | if (S_ISREG(inode->i_mode)) { |
1696 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
cf29a7b6 | 1697 | |
d7e09d03 PT |
1698 | inode->i_op = &ll_file_inode_operations; |
1699 | inode->i_fop = sbi->ll_fop; | |
1700 | inode->i_mapping->a_ops = (struct address_space_operations *)&ll_aops; | |
d7e09d03 PT |
1701 | } else if (S_ISDIR(inode->i_mode)) { |
1702 | inode->i_op = &ll_dir_inode_operations; | |
1703 | inode->i_fop = &ll_dir_operations; | |
d7e09d03 PT |
1704 | } else if (S_ISLNK(inode->i_mode)) { |
1705 | inode->i_op = &ll_fast_symlink_inode_operations; | |
d7e09d03 PT |
1706 | } else { |
1707 | inode->i_op = &ll_special_inode_operations; | |
1708 | ||
1709 | init_special_inode(inode, inode->i_mode, | |
1710 | inode->i_rdev); | |
d7e09d03 PT |
1711 | } |
1712 | } | |
1713 | ||
1714 | void ll_delete_inode(struct inode *inode) | |
1715 | { | |
1929c433 | 1716 | struct ll_inode_info *lli = ll_i2info(inode); |
d7e09d03 | 1717 | |
6e16818b | 1718 | if (S_ISREG(inode->i_mode) && lli->lli_clob) |
d7e09d03 | 1719 | /* discard all dirty pages before truncating them, required by |
c0894c6c OD |
1720 | * osc_extent implementation at LU-1030. |
1721 | */ | |
65fb55d1 NY |
1722 | cl_sync_file_range(inode, 0, OBD_OBJECT_EOF, |
1723 | CL_FSYNC_DISCARD, 1); | |
d7e09d03 | 1724 | |
91b0abe3 | 1725 | truncate_inode_pages_final(&inode->i_data); |
d7e09d03 PT |
1726 | |
1727 | /* Workaround for LU-118 */ | |
1728 | if (inode->i_data.nrpages) { | |
c4226c54 VT |
1729 | spin_lock_irq(&inode->i_data.tree_lock); |
1730 | spin_unlock_irq(&inode->i_data.tree_lock); | |
d7e09d03 | 1731 | LASSERTF(inode->i_data.nrpages == 0, |
97a075cd JN |
1732 | "inode="DFID"(%p) nrpages=%lu, see http://jira.whamcloud.com/browse/LU-118\n", |
1733 | PFID(ll_inode2fid(inode)), inode, | |
d7e09d03 PT |
1734 | inode->i_data.nrpages); |
1735 | } | |
1736 | /* Workaround end */ | |
1737 | ||
1738 | ll_clear_inode(inode); | |
1739 | clear_inode(inode); | |
d7e09d03 PT |
1740 | } |
1741 | ||
1742 | int ll_iocontrol(struct inode *inode, struct file *file, | |
1743 | unsigned int cmd, unsigned long arg) | |
1744 | { | |
1745 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
1746 | struct ptlrpc_request *req = NULL; | |
1747 | int rc, flags = 0; | |
d7e09d03 | 1748 | |
a58a38ac | 1749 | switch (cmd) { |
d7e09d03 PT |
1750 | case FSFILT_IOC_GETFLAGS: { |
1751 | struct mdt_body *body; | |
1752 | struct md_op_data *op_data; | |
1753 | ||
1754 | op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, | |
1755 | 0, 0, LUSTRE_OPC_ANY, | |
1756 | NULL); | |
1757 | if (IS_ERR(op_data)) | |
0a3bdb00 | 1758 | return PTR_ERR(op_data); |
d7e09d03 PT |
1759 | |
1760 | op_data->op_valid = OBD_MD_FLFLAGS; | |
1761 | rc = md_getattr(sbi->ll_md_exp, op_data, &req); | |
1762 | ll_finish_md_op_data(op_data); | |
1763 | if (rc) { | |
97a075cd JN |
1764 | CERROR("%s: failure inode "DFID": rc = %d\n", |
1765 | sbi->ll_md_exp->exp_obd->obd_name, | |
1766 | PFID(ll_inode2fid(inode)), rc); | |
0a3bdb00 | 1767 | return -abs(rc); |
d7e09d03 PT |
1768 | } |
1769 | ||
1770 | body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY); | |
1771 | ||
1772 | flags = body->flags; | |
1773 | ||
1774 | ptlrpc_req_finished(req); | |
1775 | ||
7ac5db21 | 1776 | return put_user(flags, (int __user *)arg); |
d7e09d03 PT |
1777 | } |
1778 | case FSFILT_IOC_SETFLAGS: { | |
1779 | struct lov_stripe_md *lsm; | |
45efd655 | 1780 | struct obd_info oinfo = { }; |
d7e09d03 PT |
1781 | struct md_op_data *op_data; |
1782 | ||
7ac5db21 | 1783 | if (get_user(flags, (int __user *)arg)) |
0a3bdb00 | 1784 | return -EFAULT; |
d7e09d03 PT |
1785 | |
1786 | op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0, | |
1787 | LUSTRE_OPC_ANY, NULL); | |
1788 | if (IS_ERR(op_data)) | |
0a3bdb00 | 1789 | return PTR_ERR(op_data); |
d7e09d03 | 1790 | |
bb41292b | 1791 | op_data->op_attr_flags = flags; |
d7e09d03 PT |
1792 | op_data->op_attr.ia_valid |= ATTR_ATTR_FLAG; |
1793 | rc = md_setattr(sbi->ll_md_exp, op_data, | |
1794 | NULL, 0, NULL, 0, &req, NULL); | |
1795 | ll_finish_md_op_data(op_data); | |
1796 | ptlrpc_req_finished(req); | |
1797 | if (rc) | |
0a3bdb00 | 1798 | return rc; |
d7e09d03 PT |
1799 | |
1800 | inode->i_flags = ll_ext_to_inode_flags(flags); | |
1801 | ||
1802 | lsm = ccc_inode_lsm_get(inode); | |
5dd16419 JX |
1803 | if (!lsm_has_objects(lsm)) { |
1804 | ccc_inode_lsm_put(inode, lsm); | |
0a3bdb00 | 1805 | return 0; |
5dd16419 | 1806 | } |
d7e09d03 | 1807 | |
21068c46 | 1808 | oinfo.oi_oa = kmem_cache_zalloc(obdo_cachep, GFP_NOFS); |
d7e09d03 PT |
1809 | if (!oinfo.oi_oa) { |
1810 | ccc_inode_lsm_put(inode, lsm); | |
0a3bdb00 | 1811 | return -ENOMEM; |
d7e09d03 PT |
1812 | } |
1813 | oinfo.oi_md = lsm; | |
1814 | oinfo.oi_oa->o_oi = lsm->lsm_oi; | |
1815 | oinfo.oi_oa->o_flags = flags; | |
1816 | oinfo.oi_oa->o_valid = OBD_MD_FLID | OBD_MD_FLFLAGS | | |
1817 | OBD_MD_FLGROUP; | |
d7e09d03 PT |
1818 | obdo_set_parent_fid(oinfo.oi_oa, &ll_i2info(inode)->lli_fid); |
1819 | rc = obd_setattr_rqset(sbi->ll_dt_exp, &oinfo, NULL); | |
2ba262fb | 1820 | kmem_cache_free(obdo_cachep, oinfo.oi_oa); |
d7e09d03 PT |
1821 | ccc_inode_lsm_put(inode, lsm); |
1822 | ||
1823 | if (rc && rc != -EPERM && rc != -EACCES) | |
1824 | CERROR("osc_setattr_async fails: rc = %d\n", rc); | |
1825 | ||
0a3bdb00 | 1826 | return rc; |
d7e09d03 PT |
1827 | } |
1828 | default: | |
0a3bdb00 | 1829 | return -ENOSYS; |
d7e09d03 PT |
1830 | } |
1831 | ||
0a3bdb00 | 1832 | return 0; |
d7e09d03 PT |
1833 | } |
1834 | ||
1835 | int ll_flush_ctx(struct inode *inode) | |
1836 | { | |
1837 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
1838 | ||
4b1a25f0 | 1839 | CDEBUG(D_SEC, "flush context for user %d\n", |
e15ba45d | 1840 | from_kuid(&init_user_ns, current_uid())); |
d7e09d03 PT |
1841 | |
1842 | obd_set_info_async(NULL, sbi->ll_md_exp, | |
1843 | sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX, | |
1844 | 0, NULL, NULL); | |
1845 | obd_set_info_async(NULL, sbi->ll_dt_exp, | |
1846 | sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX, | |
1847 | 0, NULL, NULL); | |
1848 | return 0; | |
1849 | } | |
1850 | ||
1851 | /* umount -f client means force down, don't save state */ | |
1852 | void ll_umount_begin(struct super_block *sb) | |
1853 | { | |
1854 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
1855 | struct obd_device *obd; | |
1856 | struct obd_ioctl_data *ioc_data; | |
d7e09d03 PT |
1857 | |
1858 | CDEBUG(D_VFSTRACE, "VFS Op: superblock %p count %d active %d\n", sb, | |
1859 | sb->s_count, atomic_read(&sb->s_active)); | |
1860 | ||
1861 | obd = class_exp2obd(sbi->ll_md_exp); | |
6e16818b | 1862 | if (!obd) { |
55f5a824 | 1863 | CERROR("Invalid MDC connection handle %#llx\n", |
d7e09d03 | 1864 | sbi->ll_md_exp->exp_handle.h_cookie); |
d7e09d03 PT |
1865 | return; |
1866 | } | |
1867 | obd->obd_force = 1; | |
1868 | ||
1869 | obd = class_exp2obd(sbi->ll_dt_exp); | |
6e16818b | 1870 | if (!obd) { |
55f5a824 | 1871 | CERROR("Invalid LOV connection handle %#llx\n", |
d7e09d03 | 1872 | sbi->ll_dt_exp->exp_handle.h_cookie); |
d7e09d03 PT |
1873 | return; |
1874 | } | |
1875 | obd->obd_force = 1; | |
1876 | ||
496a51bd | 1877 | ioc_data = kzalloc(sizeof(*ioc_data), GFP_NOFS); |
d7e09d03 PT |
1878 | if (ioc_data) { |
1879 | obd_iocontrol(IOC_OSC_SET_ACTIVE, sbi->ll_md_exp, | |
ec83e611 | 1880 | sizeof(*ioc_data), ioc_data, NULL); |
d7e09d03 PT |
1881 | |
1882 | obd_iocontrol(IOC_OSC_SET_ACTIVE, sbi->ll_dt_exp, | |
ec83e611 | 1883 | sizeof(*ioc_data), ioc_data, NULL); |
d7e09d03 | 1884 | |
97903a26 | 1885 | kfree(ioc_data); |
d7e09d03 PT |
1886 | } |
1887 | ||
d7e09d03 PT |
1888 | /* Really, we'd like to wait until there are no requests outstanding, |
1889 | * and then continue. For now, we just invalidate the requests, | |
1890 | * schedule() and sleep one second if needed, and hope. | |
1891 | */ | |
1892 | schedule(); | |
d7e09d03 PT |
1893 | } |
1894 | ||
1895 | int ll_remount_fs(struct super_block *sb, int *flags, char *data) | |
1896 | { | |
1897 | struct ll_sb_info *sbi = ll_s2sbi(sb); | |
1898 | char *profilenm = get_profile_name(sb); | |
1899 | int err; | |
1900 | __u32 read_only; | |
1901 | ||
1902 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { | |
1903 | read_only = *flags & MS_RDONLY; | |
1904 | err = obd_set_info_async(NULL, sbi->ll_md_exp, | |
1905 | sizeof(KEY_READ_ONLY), | |
1906 | KEY_READ_ONLY, sizeof(read_only), | |
1907 | &read_only, NULL); | |
1908 | if (err) { | |
1909 | LCONSOLE_WARN("Failed to remount %s %s (%d)\n", | |
1910 | profilenm, read_only ? | |
1911 | "read-only" : "read-write", err); | |
1912 | return err; | |
1913 | } | |
1914 | ||
1915 | if (read_only) | |
1916 | sb->s_flags |= MS_RDONLY; | |
1917 | else | |
1918 | sb->s_flags &= ~MS_RDONLY; | |
1919 | ||
1920 | if (sbi->ll_flags & LL_SBI_VERBOSE) | |
1921 | LCONSOLE_WARN("Remounted %s %s\n", profilenm, | |
1922 | read_only ? "read-only" : "read-write"); | |
1923 | } | |
1924 | return 0; | |
1925 | } | |
1926 | ||
44ecac68 FY |
1927 | /** |
1928 | * Cleanup the open handle that is cached on MDT-side. | |
1929 | * | |
1930 | * For open case, the client side open handling thread may hit error | |
1931 | * after the MDT grant the open. Under such case, the client should | |
1932 | * send close RPC to the MDT as cleanup; otherwise, the open handle | |
1933 | * on the MDT will be leaked there until the client umount or evicted. | |
1934 | * | |
1935 | * In further, if someone unlinked the file, because the open handle | |
1936 | * holds the reference on such file/object, then it will block the | |
1937 | * subsequent threads that want to locate such object via FID. | |
1938 | * | |
1939 | * \param[in] sb super block for this file-system | |
1940 | * \param[in] open_req pointer to the original open request | |
1941 | */ | |
1942 | void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req) | |
1943 | { | |
1944 | struct mdt_body *body; | |
1945 | struct md_op_data *op_data; | |
1946 | struct ptlrpc_request *close_req = NULL; | |
1947 | struct obd_export *exp = ll_s2sbi(sb)->ll_md_exp; | |
1948 | ||
1949 | body = req_capsule_server_get(&open_req->rq_pill, &RMF_MDT_BODY); | |
af13af52 | 1950 | op_data = kzalloc(sizeof(*op_data), GFP_NOFS); |
83ea341d | 1951 | if (!op_data) |
44ecac68 | 1952 | return; |
44ecac68 FY |
1953 | |
1954 | op_data->op_fid1 = body->fid1; | |
1955 | op_data->op_ioepoch = body->ioepoch; | |
1956 | op_data->op_handle = body->handle; | |
1957 | op_data->op_mod_time = get_seconds(); | |
1958 | md_close(exp, op_data, NULL, &close_req); | |
1959 | ptlrpc_req_finished(close_req); | |
1960 | ll_finish_md_op_data(op_data); | |
1961 | } | |
1962 | ||
d7e09d03 PT |
1963 | int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req, |
1964 | struct super_block *sb, struct lookup_intent *it) | |
1965 | { | |
1966 | struct ll_sb_info *sbi = NULL; | |
24af3e16 | 1967 | struct lustre_md md = { NULL }; |
d7e09d03 | 1968 | int rc; |
d7e09d03 PT |
1969 | |
1970 | LASSERT(*inode || sb); | |
1971 | sbi = sb ? ll_s2sbi(sb) : ll_i2sbi(*inode); | |
1972 | rc = md_get_lustre_md(sbi->ll_md_exp, req, sbi->ll_dt_exp, | |
1973 | sbi->ll_md_exp, &md); | |
1974 | if (rc) | |
44ecac68 | 1975 | goto cleanup; |
d7e09d03 PT |
1976 | |
1977 | if (*inode) { | |
1978 | ll_update_inode(*inode, &md); | |
1979 | } else { | |
6e16818b | 1980 | LASSERT(sb); |
d7e09d03 PT |
1981 | |
1982 | /* | |
1983 | * At this point server returns to client's same fid as client | |
1984 | * generated for creating. So using ->fid1 is okay here. | |
1985 | */ | |
1986 | LASSERT(fid_is_sane(&md.body->fid1)); | |
1987 | ||
1988 | *inode = ll_iget(sb, cl_fid_build_ino(&md.body->fid1, | |
c1e2699d | 1989 | sbi->ll_flags & LL_SBI_32BIT_API), |
d7e09d03 | 1990 | &md); |
4ae51a74 | 1991 | if (!*inode) { |
d7e09d03 PT |
1992 | #ifdef CONFIG_FS_POSIX_ACL |
1993 | if (md.posix_acl) { | |
1994 | posix_acl_release(md.posix_acl); | |
1995 | md.posix_acl = NULL; | |
1996 | } | |
1997 | #endif | |
020ecc6f | 1998 | rc = -ENOMEM; |
d7e09d03 | 1999 | CERROR("new_inode -fatal: rc %d\n", rc); |
34e1f2bb | 2000 | goto out; |
d7e09d03 PT |
2001 | } |
2002 | } | |
2003 | ||
2004 | /* Handling piggyback layout lock. | |
2005 | * Layout lock can be piggybacked by getattr and open request. | |
2006 | * The lsm can be applied to inode only if it comes with a layout lock | |
2007 | * otherwise correct layout may be overwritten, for example: | |
2008 | * 1. proc1: mdt returns a lsm but not granting layout | |
2009 | * 2. layout was changed by another client | |
2010 | * 3. proc2: refresh layout and layout lock granted | |
c0894c6c OD |
2011 | * 4. proc1: to apply a stale layout |
2012 | */ | |
6e16818b | 2013 | if (it && it->d.lustre.it_lock_mode != 0) { |
d7e09d03 PT |
2014 | struct lustre_handle lockh; |
2015 | struct ldlm_lock *lock; | |
2016 | ||
2017 | lockh.cookie = it->d.lustre.it_lock_handle; | |
2018 | lock = ldlm_handle2lock(&lockh); | |
6e16818b | 2019 | LASSERT(lock); |
d7e09d03 PT |
2020 | if (ldlm_has_layout(lock)) { |
2021 | struct cl_object_conf conf; | |
2022 | ||
2023 | memset(&conf, 0, sizeof(conf)); | |
2024 | conf.coc_opc = OBJECT_CONF_SET; | |
2025 | conf.coc_inode = *inode; | |
2026 | conf.coc_lock = lock; | |
2027 | conf.u.coc_md = &md; | |
2028 | (void)ll_layout_conf(*inode, &conf); | |
2029 | } | |
2030 | LDLM_LOCK_PUT(lock); | |
2031 | } | |
2032 | ||
2033 | out: | |
6e16818b | 2034 | if (md.lsm) |
d7e09d03 PT |
2035 | obd_free_memmd(sbi->ll_dt_exp, &md.lsm); |
2036 | md_free_lustre_md(sbi->ll_md_exp, &md); | |
44ecac68 FY |
2037 | |
2038 | cleanup: | |
2039 | if (rc != 0 && it && it->it_op & IT_OPEN) | |
2040 | ll_open_cleanup(sb ? sb : (*inode)->i_sb, req); | |
2041 | ||
0a3bdb00 | 2042 | return rc; |
d7e09d03 PT |
2043 | } |
2044 | ||
4c6243ec | 2045 | int ll_obd_statfs(struct inode *inode, void __user *arg) |
d7e09d03 PT |
2046 | { |
2047 | struct ll_sb_info *sbi = NULL; | |
2048 | struct obd_export *exp; | |
2049 | char *buf = NULL; | |
2050 | struct obd_ioctl_data *data = NULL; | |
2051 | __u32 type; | |
d7e09d03 PT |
2052 | int len = 0, rc; |
2053 | ||
c650ba73 TR |
2054 | if (!inode) { |
2055 | rc = -EINVAL; | |
2056 | goto out_statfs; | |
2057 | } | |
2058 | ||
2059 | sbi = ll_i2sbi(inode); | |
2060 | if (!sbi) { | |
34e1f2bb JL |
2061 | rc = -EINVAL; |
2062 | goto out_statfs; | |
2063 | } | |
d7e09d03 PT |
2064 | |
2065 | rc = obd_ioctl_getdata(&buf, &len, arg); | |
2066 | if (rc) | |
34e1f2bb | 2067 | goto out_statfs; |
d7e09d03 | 2068 | |
bdbb0512 | 2069 | data = (void *)buf; |
d7e09d03 | 2070 | if (!data->ioc_inlbuf1 || !data->ioc_inlbuf2 || |
34e1f2bb JL |
2071 | !data->ioc_pbuf1 || !data->ioc_pbuf2) { |
2072 | rc = -EINVAL; | |
2073 | goto out_statfs; | |
2074 | } | |
d7e09d03 PT |
2075 | |
2076 | if (data->ioc_inllen1 != sizeof(__u32) || | |
2077 | data->ioc_inllen2 != sizeof(__u32) || | |
2078 | data->ioc_plen1 != sizeof(struct obd_statfs) || | |
34e1f2bb JL |
2079 | data->ioc_plen2 != sizeof(struct obd_uuid)) { |
2080 | rc = -EINVAL; | |
2081 | goto out_statfs; | |
2082 | } | |
d7e09d03 PT |
2083 | |
2084 | memcpy(&type, data->ioc_inlbuf1, sizeof(__u32)); | |
da5ecb4d | 2085 | if (type & LL_STATFS_LMV) { |
d7e09d03 | 2086 | exp = sbi->ll_md_exp; |
da5ecb4d | 2087 | } else if (type & LL_STATFS_LOV) { |
d7e09d03 | 2088 | exp = sbi->ll_dt_exp; |
da5ecb4d | 2089 | } else { |
34e1f2bb JL |
2090 | rc = -ENODEV; |
2091 | goto out_statfs; | |
2092 | } | |
d7e09d03 | 2093 | |
44164fc9 | 2094 | rc = obd_iocontrol(IOC_OBD_STATFS, exp, len, buf, NULL); |
d7e09d03 | 2095 | if (rc) |
34e1f2bb | 2096 | goto out_statfs; |
d7e09d03 PT |
2097 | out_statfs: |
2098 | if (buf) | |
2099 | obd_ioctl_freedata(buf, len); | |
2100 | return rc; | |
2101 | } | |
2102 | ||
2103 | int ll_process_config(struct lustre_cfg *lcfg) | |
2104 | { | |
2105 | char *ptr; | |
2106 | void *sb; | |
2107 | struct lprocfs_static_vars lvars; | |
2108 | unsigned long x; | |
2109 | int rc = 0; | |
2110 | ||
2111 | lprocfs_llite_init_vars(&lvars); | |
2112 | ||
2113 | /* The instance name contains the sb: lustre-client-aacfe000 */ | |
2114 | ptr = strrchr(lustre_cfg_string(lcfg, 0), '-'); | |
2115 | if (!ptr || !*(++ptr)) | |
2116 | return -EINVAL; | |
692f2b6c | 2117 | rc = kstrtoul(ptr, 16, &x); |
2118 | if (rc != 0) | |
d7e09d03 PT |
2119 | return -EINVAL; |
2120 | sb = (void *)x; | |
2121 | /* This better be a real Lustre superblock! */ | |
2122 | LASSERT(s2lsi((struct super_block *)sb)->lsi_lmd->lmd_magic == LMD_MAGIC); | |
2123 | ||
2124 | /* Note we have not called client_common_fill_super yet, so | |
c0894c6c OD |
2125 | * proc fns must be able to handle that! |
2126 | */ | |
d7e09d03 PT |
2127 | rc = class_process_proc_param(PARAM_LLITE, lvars.obd_vars, |
2128 | lcfg, sb); | |
2129 | if (rc > 0) | |
2130 | rc = 0; | |
fbe7c6c7 | 2131 | return rc; |
d7e09d03 PT |
2132 | } |
2133 | ||
2134 | /* this function prepares md_op_data hint for passing ot down to MD stack. */ | |
aff9d8e8 | 2135 | struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data, |
e15ba45d OD |
2136 | struct inode *i1, struct inode *i2, |
2137 | const char *name, int namelen, | |
2138 | int mode, __u32 opc, void *data) | |
d7e09d03 | 2139 | { |
d7e09d03 PT |
2140 | if (namelen > ll_i2sbi(i1)->ll_namelen) |
2141 | return ERR_PTR(-ENAMETOOLONG); | |
2142 | ||
6e16818b | 2143 | if (!op_data) |
496a51bd | 2144 | op_data = kzalloc(sizeof(*op_data), GFP_NOFS); |
d7e09d03 | 2145 | |
6e16818b | 2146 | if (!op_data) |
d7e09d03 PT |
2147 | return ERR_PTR(-ENOMEM); |
2148 | ||
2149 | ll_i2gids(op_data->op_suppgids, i1, i2); | |
2150 | op_data->op_fid1 = *ll_inode2fid(i1); | |
d7e09d03 | 2151 | |
ef2e0f55 | 2152 | if (i2) |
d7e09d03 | 2153 | op_data->op_fid2 = *ll_inode2fid(i2); |
ef2e0f55 | 2154 | else |
d7e09d03 | 2155 | fid_zero(&op_data->op_fid2); |
d7e09d03 PT |
2156 | |
2157 | op_data->op_name = name; | |
2158 | op_data->op_namelen = namelen; | |
2159 | op_data->op_mode = mode; | |
14e3f92a | 2160 | op_data->op_mod_time = ktime_get_real_seconds(); |
4b1a25f0 PT |
2161 | op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid()); |
2162 | op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid()); | |
d7e09d03 PT |
2163 | op_data->op_cap = cfs_curproc_cap_pack(); |
2164 | op_data->op_bias = 0; | |
2165 | op_data->op_cli_flags = 0; | |
6e16818b OD |
2166 | if ((opc == LUSTRE_OPC_CREATE) && name && |
2167 | filename_is_volatile(name, namelen, NULL)) | |
d7e09d03 PT |
2168 | op_data->op_bias |= MDS_CREATE_VOLATILE; |
2169 | op_data->op_opc = opc; | |
2170 | op_data->op_mds = 0; | |
2171 | op_data->op_data = data; | |
2172 | ||
2173 | /* If the file is being opened after mknod() (normally due to NFS) | |
2174 | * try to use the default stripe data from parent directory for | |
c0894c6c OD |
2175 | * allocating OST objects. Try to pass the parent FID to MDS. |
2176 | */ | |
d7e09d03 PT |
2177 | if (opc == LUSTRE_OPC_CREATE && i1 == i2 && S_ISREG(i2->i_mode) && |
2178 | !ll_i2info(i2)->lli_has_smd) { | |
2179 | struct ll_inode_info *lli = ll_i2info(i2); | |
2180 | ||
2181 | spin_lock(&lli->lli_lock); | |
2182 | if (likely(!lli->lli_has_smd && !fid_is_zero(&lli->lli_pfid))) | |
2183 | op_data->op_fid1 = lli->lli_pfid; | |
2184 | spin_unlock(&lli->lli_lock); | |
d7e09d03 PT |
2185 | } |
2186 | ||
2187 | /* When called by ll_setattr_raw, file is i1. */ | |
1f6eaf83 | 2188 | if (ll_i2info(i1)->lli_flags & LLIF_DATA_MODIFIED) |
d7e09d03 PT |
2189 | op_data->op_bias |= MDS_DATA_MODIFIED; |
2190 | ||
2191 | return op_data; | |
2192 | } | |
2193 | ||
2194 | void ll_finish_md_op_data(struct md_op_data *op_data) | |
2195 | { | |
97903a26 | 2196 | kfree(op_data); |
d7e09d03 PT |
2197 | } |
2198 | ||
2199 | int ll_show_options(struct seq_file *seq, struct dentry *dentry) | |
2200 | { | |
2201 | struct ll_sb_info *sbi; | |
2202 | ||
6e16818b | 2203 | LASSERT(seq && dentry); |
d7e09d03 PT |
2204 | sbi = ll_s2sbi(dentry->d_sb); |
2205 | ||
2206 | if (sbi->ll_flags & LL_SBI_NOLCK) | |
2207 | seq_puts(seq, ",nolock"); | |
2208 | ||
2209 | if (sbi->ll_flags & LL_SBI_FLOCK) | |
2210 | seq_puts(seq, ",flock"); | |
2211 | ||
2212 | if (sbi->ll_flags & LL_SBI_LOCALFLOCK) | |
2213 | seq_puts(seq, ",localflock"); | |
2214 | ||
2215 | if (sbi->ll_flags & LL_SBI_USER_XATTR) | |
2216 | seq_puts(seq, ",user_xattr"); | |
2217 | ||
2218 | if (sbi->ll_flags & LL_SBI_LAZYSTATFS) | |
2219 | seq_puts(seq, ",lazystatfs"); | |
2220 | ||
2221 | if (sbi->ll_flags & LL_SBI_USER_FID2PATH) | |
2222 | seq_puts(seq, ",user_fid2path"); | |
2223 | ||
0a3bdb00 | 2224 | return 0; |
d7e09d03 PT |
2225 | } |
2226 | ||
2227 | /** | |
2228 | * Get obd name by cmd, and copy out to user space | |
2229 | */ | |
2230 | int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg) | |
2231 | { | |
2232 | struct ll_sb_info *sbi = ll_i2sbi(inode); | |
2233 | struct obd_device *obd; | |
d7e09d03 PT |
2234 | |
2235 | if (cmd == OBD_IOC_GETDTNAME) | |
2236 | obd = class_exp2obd(sbi->ll_dt_exp); | |
2237 | else if (cmd == OBD_IOC_GETMDNAME) | |
2238 | obd = class_exp2obd(sbi->ll_md_exp); | |
2239 | else | |
0a3bdb00 | 2240 | return -EINVAL; |
d7e09d03 PT |
2241 | |
2242 | if (!obd) | |
0a3bdb00 | 2243 | return -ENOENT; |
d7e09d03 | 2244 | |
7ac5db21 OD |
2245 | if (copy_to_user((void __user *)arg, obd->obd_name, |
2246 | strlen(obd->obd_name) + 1)) | |
0a3bdb00 | 2247 | return -EFAULT; |
d7e09d03 | 2248 | |
0a3bdb00 | 2249 | return 0; |
d7e09d03 PT |
2250 | } |
2251 | ||
2252 | /** | |
2253 | * Get lustre file system name by \a sbi. If \a buf is provided(non-NULL), the | |
2254 | * fsname will be returned in this buffer; otherwise, a static buffer will be | |
2255 | * used to store the fsname and returned to caller. | |
2256 | */ | |
2257 | char *ll_get_fsname(struct super_block *sb, char *buf, int buflen) | |
2258 | { | |
2259 | static char fsname_static[MTI_NAME_MAXLEN]; | |
2260 | struct lustre_sb_info *lsi = s2lsi(sb); | |
2261 | char *ptr; | |
2262 | int len; | |
2263 | ||
6e16818b | 2264 | if (!buf) { |
d7e09d03 PT |
2265 | /* this means the caller wants to use static buffer |
2266 | * and it doesn't care about race. Usually this is | |
c0894c6c OD |
2267 | * in error reporting path |
2268 | */ | |
d7e09d03 PT |
2269 | buf = fsname_static; |
2270 | buflen = sizeof(fsname_static); | |
2271 | } | |
2272 | ||
2273 | len = strlen(lsi->lsi_lmd->lmd_profile); | |
2274 | ptr = strrchr(lsi->lsi_lmd->lmd_profile, '-'); | |
2275 | if (ptr && (strcmp(ptr, "-client") == 0)) | |
2276 | len -= 7; | |
2277 | ||
2278 | if (unlikely(len >= buflen)) | |
2279 | len = buflen - 1; | |
2280 | strncpy(buf, lsi->lsi_lmd->lmd_profile, len); | |
2281 | buf[len] = '\0'; | |
2282 | ||
2283 | return buf; | |
2284 | } | |
2285 | ||
d7e09d03 PT |
2286 | void ll_dirty_page_discard_warn(struct page *page, int ioret) |
2287 | { | |
2288 | char *buf, *path = NULL; | |
2289 | struct dentry *dentry = NULL; | |
8c7b0e1a | 2290 | struct vvp_object *obj = cl_inode2vvp(page->mapping->host); |
d7e09d03 PT |
2291 | |
2292 | /* this can be called inside spin lock so use GFP_ATOMIC. */ | |
2293 | buf = (char *)__get_free_page(GFP_ATOMIC); | |
6e16818b | 2294 | if (buf) { |
d7e09d03 | 2295 | dentry = d_find_alias(page->mapping->host); |
6e16818b | 2296 | if (dentry) |
1ad581eb | 2297 | path = dentry_path_raw(dentry, buf, PAGE_SIZE); |
d7e09d03 PT |
2298 | } |
2299 | ||
73b89907 | 2300 | CDEBUG(D_WARNING, |
2d00bd17 JP |
2301 | "%s: dirty page discard: %s/fid: " DFID "/%s may get corrupted (rc %d)\n", |
2302 | ll_get_fsname(page->mapping->host->i_sb, NULL, 0), | |
73b89907 | 2303 | s2lsi(page->mapping->host->i_sb)->lsi_lmd->lmd_dev, |
8c7b0e1a | 2304 | PFID(&obj->vob_header.coh_lu.loh_fid), |
73b89907 | 2305 | (path && !IS_ERR(path)) ? path : "", ioret); |
d7e09d03 | 2306 | |
6e16818b | 2307 | if (dentry) |
d7e09d03 PT |
2308 | dput(dentry); |
2309 | ||
6e16818b | 2310 | if (buf) |
d7e09d03 PT |
2311 | free_page((unsigned long)buf); |
2312 | } |