]>
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) 2007, 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/include/lustre/lustre_idl.h | |
33 | * | |
34 | * Lustre wire protocol definitions. | |
35 | */ | |
36 | ||
37 | /** \defgroup lustreidl lustreidl | |
38 | * | |
39 | * Lustre wire protocol definitions. | |
40 | * | |
41 | * ALL structs passing over the wire should be declared here. Structs | |
42 | * that are used in interfaces with userspace should go in lustre_user.h. | |
43 | * | |
44 | * All structs being declared here should be built from simple fixed-size | |
45 | * types (__u8, __u16, __u32, __u64) or be built from other types or | |
46 | * structs also declared in this file. Similarly, all flags and magic | |
47 | * values in those structs should also be declared here. This ensures | |
48 | * that the Lustre wire protocol is not influenced by external dependencies. | |
49 | * | |
50 | * The only other acceptable items in this file are VERY SIMPLE accessor | |
36dc51cc | 51 | * functions to avoid callers grubbing inside the structures. Nothing that |
d7e09d03 PT |
52 | * depends on external functions or definitions should be in here. |
53 | * | |
54 | * Structs must be properly aligned to put 64-bit values on an 8-byte | |
55 | * boundary. Any structs being added here must also be added to | |
56 | * utils/wirecheck.c and "make newwiretest" run to regenerate the | |
57 | * utils/wiretest.c sources. This allows us to verify that wire structs | |
58 | * have the proper alignment/size on all architectures. | |
59 | * | |
60 | * DO NOT CHANGE any of the structs, flags, values declared here and used | |
61 | * in released Lustre versions. Some structs may have padding fields that | |
62 | * can be used. Some structs might allow addition at the end (verify this | |
63 | * in the code to ensure that new/old clients that see this larger struct | |
64 | * do not fail, otherwise you need to implement protocol compatibility). | |
65 | * | |
d7e09d03 PT |
66 | * @{ |
67 | */ | |
68 | ||
69 | #ifndef _LUSTRE_IDL_H_ | |
70 | #define _LUSTRE_IDL_H_ | |
71 | ||
55f5a824 | 72 | #include "../../../include/linux/libcfs/libcfs.h" |
bbf00c3d | 73 | #include "../../../include/linux/lnet/types.h" |
d7e09d03 PT |
74 | |
75 | /* Defn's shared with user-space. */ | |
1accaadf GKH |
76 | #include "lustre_user.h" |
77 | #include "lustre_errno.h" | |
00c0a6ae | 78 | #include "../lustre_ver.h" |
2d58de78 | 79 | |
d7e09d03 PT |
80 | /* |
81 | * GENERAL STUFF | |
82 | */ | |
83 | /* FOO_REQUEST_PORTAL is for incoming requests on the FOO | |
84 | * FOO_REPLY_PORTAL is for incoming replies on the FOO | |
85 | * FOO_BULK_PORTAL is for incoming bulk on the FOO | |
86 | */ | |
87 | ||
37604896 TL |
88 | /* Lustre service names are following the format |
89 | * service name + MDT + seq name | |
90 | */ | |
91 | #define LUSTRE_MDT_MAXNAMELEN 80 | |
92 | ||
d7e09d03 PT |
93 | #define CONNMGR_REQUEST_PORTAL 1 |
94 | #define CONNMGR_REPLY_PORTAL 2 | |
df693dbe | 95 | /*#define OSC_REQUEST_PORTAL 3 */ |
d7e09d03 | 96 | #define OSC_REPLY_PORTAL 4 |
df693dbe | 97 | /*#define OSC_BULK_PORTAL 5 */ |
d7e09d03 PT |
98 | #define OST_IO_PORTAL 6 |
99 | #define OST_CREATE_PORTAL 7 | |
100 | #define OST_BULK_PORTAL 8 | |
df693dbe | 101 | /*#define MDC_REQUEST_PORTAL 9 */ |
d7e09d03 | 102 | #define MDC_REPLY_PORTAL 10 |
df693dbe | 103 | /*#define MDC_BULK_PORTAL 11 */ |
d7e09d03 | 104 | #define MDS_REQUEST_PORTAL 12 |
df693dbe | 105 | /*#define MDS_REPLY_PORTAL 13 */ |
d7e09d03 PT |
106 | #define MDS_BULK_PORTAL 14 |
107 | #define LDLM_CB_REQUEST_PORTAL 15 | |
108 | #define LDLM_CB_REPLY_PORTAL 16 | |
109 | #define LDLM_CANCEL_REQUEST_PORTAL 17 | |
110 | #define LDLM_CANCEL_REPLY_PORTAL 18 | |
df693dbe OD |
111 | /*#define PTLBD_REQUEST_PORTAL 19 */ |
112 | /*#define PTLBD_REPLY_PORTAL 20 */ | |
113 | /*#define PTLBD_BULK_PORTAL 21 */ | |
d7e09d03 PT |
114 | #define MDS_SETATTR_PORTAL 22 |
115 | #define MDS_READPAGE_PORTAL 23 | |
a3310525 | 116 | #define OUT_PORTAL 24 |
d7e09d03 PT |
117 | |
118 | #define MGC_REPLY_PORTAL 25 | |
119 | #define MGS_REQUEST_PORTAL 26 | |
120 | #define MGS_REPLY_PORTAL 27 | |
121 | #define OST_REQUEST_PORTAL 28 | |
122 | #define FLD_REQUEST_PORTAL 29 | |
123 | #define SEQ_METADATA_PORTAL 30 | |
124 | #define SEQ_DATA_PORTAL 31 | |
125 | #define SEQ_CONTROLLER_PORTAL 32 | |
126 | #define MGS_BULK_PORTAL 33 | |
127 | ||
a1e616b0 OD |
128 | /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com, |
129 | * n8851@cray.com | |
130 | */ | |
d7e09d03 PT |
131 | |
132 | /* packet types */ | |
133 | #define PTL_RPC_MSG_REQUEST 4711 | |
134 | #define PTL_RPC_MSG_ERR 4712 | |
135 | #define PTL_RPC_MSG_REPLY 4713 | |
136 | ||
137 | /* DON'T use swabbed values of MAGIC as magic! */ | |
d7e09d03 | 138 | #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3 |
d7e09d03 PT |
139 | #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B |
140 | ||
141 | #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2 | |
142 | ||
143 | #define PTLRPC_MSG_VERSION 0x00000003 | |
144 | #define LUSTRE_VERSION_MASK 0xffff0000 | |
145 | #define LUSTRE_OBD_VERSION 0x00010000 | |
146 | #define LUSTRE_MDS_VERSION 0x00020000 | |
147 | #define LUSTRE_OST_VERSION 0x00030000 | |
148 | #define LUSTRE_DLM_VERSION 0x00040000 | |
149 | #define LUSTRE_LOG_VERSION 0x00050000 | |
150 | #define LUSTRE_MGS_VERSION 0x00060000 | |
151 | ||
d7e09d03 PT |
152 | /** |
153 | * Describes a range of sequence, lsr_start is included but lsr_end is | |
154 | * not in the range. | |
155 | * Same structure is used in fld module where lsr_index field holds mdt id | |
156 | * of the home mdt. | |
157 | */ | |
158 | struct lu_seq_range { | |
159 | __u64 lsr_start; | |
160 | __u64 lsr_end; | |
161 | __u32 lsr_index; | |
162 | __u32 lsr_flags; | |
163 | }; | |
164 | ||
b78c2b9b | 165 | struct lu_seq_range_array { |
166 | __u32 lsra_count; | |
167 | __u32 lsra_padding; | |
168 | struct lu_seq_range lsra_lsr[0]; | |
169 | }; | |
170 | ||
d7e09d03 PT |
171 | #define LU_SEQ_RANGE_MDT 0x0 |
172 | #define LU_SEQ_RANGE_OST 0x1 | |
173 | #define LU_SEQ_RANGE_ANY 0x3 | |
174 | ||
175 | #define LU_SEQ_RANGE_MASK 0x3 | |
176 | ||
d7e09d03 | 177 | /** \defgroup lu_fid lu_fid |
a1e616b0 OD |
178 | * @{ |
179 | */ | |
d7e09d03 PT |
180 | |
181 | /** | |
182 | * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat. | |
183 | * Deprecated since HSM and SOM attributes are now stored in separate on-disk | |
184 | * xattr. | |
185 | */ | |
186 | enum lma_compat { | |
c0ac76d9 | 187 | LMAC_HSM = 0x00000001, |
a823acf5 | 188 | /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */ |
c0ac76d9 FY |
189 | LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */ |
190 | LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is | |
a1e616b0 OD |
191 | * under /O/<seq>/d<x>. |
192 | */ | |
d7e09d03 PT |
193 | }; |
194 | ||
195 | /** | |
196 | * Masks for all features that should be supported by a Lustre version to | |
197 | * access a specific file. | |
198 | * This information is stored in lustre_mdt_attrs::lma_incompat. | |
199 | */ | |
200 | enum lma_incompat { | |
c03a98b4 FY |
201 | LMAI_RELEASED = 0x00000001, /* file is released */ |
202 | LMAI_AGENT = 0x00000002, /* agent inode */ | |
203 | LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object | |
a1e616b0 OD |
204 | * is on the remote MDT |
205 | */ | |
d7e09d03 | 206 | }; |
c9f6bb96 | 207 | |
d7e09d03 PT |
208 | #define LMA_INCOMPAT_SUPP (LMAI_AGENT | LMAI_REMOTE_PARENT) |
209 | ||
d7e09d03 PT |
210 | /** |
211 | * fid constants | |
212 | */ | |
213 | enum { | |
66cc83e9 MP |
214 | /** LASTID file has zero OID */ |
215 | LUSTRE_FID_LASTID_OID = 0UL, | |
d7e09d03 PT |
216 | /** initial fid id value */ |
217 | LUSTRE_FID_INIT_OID = 1UL | |
218 | }; | |
219 | ||
220 | /** returns fid object sequence */ | |
221 | static inline __u64 fid_seq(const struct lu_fid *fid) | |
222 | { | |
223 | return fid->f_seq; | |
224 | } | |
225 | ||
226 | /** returns fid object id */ | |
227 | static inline __u32 fid_oid(const struct lu_fid *fid) | |
228 | { | |
229 | return fid->f_oid; | |
230 | } | |
231 | ||
232 | /** returns fid object version */ | |
233 | static inline __u32 fid_ver(const struct lu_fid *fid) | |
234 | { | |
235 | return fid->f_ver; | |
236 | } | |
237 | ||
238 | static inline void fid_zero(struct lu_fid *fid) | |
239 | { | |
240 | memset(fid, 0, sizeof(*fid)); | |
241 | } | |
242 | ||
21aef7d9 | 243 | static inline __u64 fid_ver_oid(const struct lu_fid *fid) |
d7e09d03 PT |
244 | { |
245 | return ((__u64)fid_ver(fid) << 32 | fid_oid(fid)); | |
246 | } | |
247 | ||
2742c75e BF |
248 | /* copytool uses a 32b bitmask field to encode archive-Ids during register |
249 | * with MDT thru kuc. | |
250 | * archive num = 0 => all | |
251 | * archive num from 1 to 32 | |
252 | */ | |
253 | #define LL_HSM_MAX_ARCHIVE (sizeof(__u32) * 8) | |
254 | ||
d7e09d03 PT |
255 | /** |
256 | * Note that reserved SEQ numbers below 12 will conflict with ldiskfs | |
257 | * inodes in the IGIF namespace, so these reserved SEQ numbers can be | |
258 | * used for other purposes and not risk collisions with existing inodes. | |
259 | * | |
260 | * Different FID Format | |
25ed6a5e | 261 | * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs |
d7e09d03 PT |
262 | */ |
263 | enum fid_seq { | |
264 | FID_SEQ_OST_MDT0 = 0, | |
265 | FID_SEQ_LLOG = 1, /* unnamed llogs */ | |
266 | FID_SEQ_ECHO = 2, | |
267 | FID_SEQ_OST_MDT1 = 3, | |
268 | FID_SEQ_OST_MAX = 9, /* Max MDT count before OST_on_FID */ | |
269 | FID_SEQ_LLOG_NAME = 10, /* named llogs */ | |
270 | FID_SEQ_RSVD = 11, | |
271 | FID_SEQ_IGIF = 12, | |
272 | FID_SEQ_IGIF_MAX = 0x0ffffffffULL, | |
273 | FID_SEQ_IDIF = 0x100000000ULL, | |
274 | FID_SEQ_IDIF_MAX = 0x1ffffffffULL, | |
275 | /* Normal FID sequence starts from this value, i.e. 1<<33 */ | |
276 | FID_SEQ_START = 0x200000000ULL, | |
277 | /* sequence for local pre-defined FIDs listed in local_oid */ | |
278 | FID_SEQ_LOCAL_FILE = 0x200000001ULL, | |
279 | FID_SEQ_DOT_LUSTRE = 0x200000002ULL, | |
280 | /* sequence is used for local named objects FIDs generated | |
a1e616b0 OD |
281 | * by local_object_storage library |
282 | */ | |
d7e09d03 PT |
283 | FID_SEQ_LOCAL_NAME = 0x200000003ULL, |
284 | /* Because current FLD will only cache the fid sequence, instead | |
285 | * of oid on the client side, if the FID needs to be exposed to | |
286 | * clients sides, it needs to make sure all of fids under one | |
a1e616b0 OD |
287 | * sequence will be located in one MDT. |
288 | */ | |
d7e09d03 PT |
289 | FID_SEQ_SPECIAL = 0x200000004ULL, |
290 | FID_SEQ_QUOTA = 0x200000005ULL, | |
291 | FID_SEQ_QUOTA_GLB = 0x200000006ULL, | |
292 | FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */ | |
293 | FID_SEQ_NORMAL = 0x200000400ULL, | |
294 | FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL | |
295 | }; | |
296 | ||
297 | #define OBIF_OID_MAX_BITS 32 | |
298 | #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS) | |
299 | #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1) | |
300 | #define IDIF_OID_MAX_BITS 48 | |
301 | #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS) | |
302 | #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1) | |
303 | ||
304 | /** OID for FID_SEQ_SPECIAL */ | |
305 | enum special_oid { | |
306 | /* Big Filesystem Lock to serialize rename operations */ | |
307 | FID_OID_SPECIAL_BFL = 1UL, | |
308 | }; | |
309 | ||
310 | /** OID for FID_SEQ_DOT_LUSTRE */ | |
311 | enum dot_lustre_oid { | |
312 | FID_OID_DOT_LUSTRE = 1UL, | |
313 | FID_OID_DOT_LUSTRE_OBF = 2UL, | |
314 | }; | |
315 | ||
d8f183b3 | 316 | static inline bool fid_seq_is_mdt0(__u64 seq) |
d7e09d03 PT |
317 | { |
318 | return (seq == FID_SEQ_OST_MDT0); | |
319 | } | |
320 | ||
d8f183b3 | 321 | static inline bool fid_seq_is_mdt(__u64 seq) |
d7e09d03 PT |
322 | { |
323 | return seq == FID_SEQ_OST_MDT0 || seq >= FID_SEQ_NORMAL; | |
324 | }; | |
325 | ||
d8f183b3 | 326 | static inline bool fid_seq_is_echo(__u64 seq) |
d7e09d03 PT |
327 | { |
328 | return (seq == FID_SEQ_ECHO); | |
329 | } | |
330 | ||
d8f183b3 | 331 | static inline bool fid_is_echo(const struct lu_fid *fid) |
d7e09d03 PT |
332 | { |
333 | return fid_seq_is_echo(fid_seq(fid)); | |
334 | } | |
335 | ||
d8f183b3 | 336 | static inline bool fid_seq_is_llog(__u64 seq) |
d7e09d03 PT |
337 | { |
338 | return (seq == FID_SEQ_LLOG); | |
339 | } | |
340 | ||
d8f183b3 | 341 | static inline bool fid_is_llog(const struct lu_fid *fid) |
d7e09d03 | 342 | { |
66cc83e9 MP |
343 | /* file with OID == 0 is not llog but contains last oid */ |
344 | return fid_seq_is_llog(fid_seq(fid)) && fid_oid(fid) > 0; | |
d7e09d03 PT |
345 | } |
346 | ||
d8f183b3 | 347 | static inline bool fid_seq_is_rsvd(__u64 seq) |
d7e09d03 PT |
348 | { |
349 | return (seq > FID_SEQ_OST_MDT0 && seq <= FID_SEQ_RSVD); | |
350 | }; | |
351 | ||
d8f183b3 | 352 | static inline bool fid_seq_is_special(__u64 seq) |
d7e09d03 PT |
353 | { |
354 | return seq == FID_SEQ_SPECIAL; | |
355 | }; | |
356 | ||
d8f183b3 | 357 | static inline bool fid_seq_is_local_file(__u64 seq) |
d7e09d03 PT |
358 | { |
359 | return seq == FID_SEQ_LOCAL_FILE || | |
360 | seq == FID_SEQ_LOCAL_NAME; | |
361 | }; | |
362 | ||
d8f183b3 | 363 | static inline bool fid_seq_is_root(__u64 seq) |
d7e09d03 PT |
364 | { |
365 | return seq == FID_SEQ_ROOT; | |
366 | } | |
367 | ||
d8f183b3 | 368 | static inline bool fid_seq_is_dot(__u64 seq) |
d7e09d03 PT |
369 | { |
370 | return seq == FID_SEQ_DOT_LUSTRE; | |
371 | } | |
372 | ||
d8f183b3 | 373 | static inline bool fid_seq_is_default(__u64 seq) |
d7e09d03 PT |
374 | { |
375 | return seq == FID_SEQ_LOV_DEFAULT; | |
376 | } | |
377 | ||
d8f183b3 | 378 | static inline bool fid_is_mdt0(const struct lu_fid *fid) |
d7e09d03 PT |
379 | { |
380 | return fid_seq_is_mdt0(fid_seq(fid)); | |
381 | } | |
382 | ||
383 | static inline void lu_root_fid(struct lu_fid *fid) | |
384 | { | |
385 | fid->f_seq = FID_SEQ_ROOT; | |
386 | fid->f_oid = 1; | |
387 | fid->f_ver = 0; | |
388 | } | |
389 | ||
390 | /** | |
391 | * Check if a fid is igif or not. | |
392 | * \param fid the fid to be tested. | |
393 | * \return true if the fid is a igif; otherwise false. | |
394 | */ | |
d8f183b3 | 395 | static inline bool fid_seq_is_igif(__u64 seq) |
d7e09d03 PT |
396 | { |
397 | return seq >= FID_SEQ_IGIF && seq <= FID_SEQ_IGIF_MAX; | |
398 | } | |
399 | ||
d8f183b3 | 400 | static inline bool fid_is_igif(const struct lu_fid *fid) |
d7e09d03 PT |
401 | { |
402 | return fid_seq_is_igif(fid_seq(fid)); | |
403 | } | |
404 | ||
405 | /** | |
406 | * Check if a fid is idif or not. | |
407 | * \param fid the fid to be tested. | |
408 | * \return true if the fid is a idif; otherwise false. | |
409 | */ | |
d8f183b3 | 410 | static inline bool fid_seq_is_idif(__u64 seq) |
d7e09d03 PT |
411 | { |
412 | return seq >= FID_SEQ_IDIF && seq <= FID_SEQ_IDIF_MAX; | |
413 | } | |
414 | ||
d8f183b3 | 415 | static inline bool fid_is_idif(const struct lu_fid *fid) |
d7e09d03 PT |
416 | { |
417 | return fid_seq_is_idif(fid_seq(fid)); | |
418 | } | |
419 | ||
d8f183b3 | 420 | static inline bool fid_is_local_file(const struct lu_fid *fid) |
d7e09d03 PT |
421 | { |
422 | return fid_seq_is_local_file(fid_seq(fid)); | |
423 | } | |
424 | ||
d8f183b3 | 425 | static inline bool fid_seq_is_norm(__u64 seq) |
d7e09d03 PT |
426 | { |
427 | return (seq >= FID_SEQ_NORMAL); | |
428 | } | |
429 | ||
d8f183b3 | 430 | static inline bool fid_is_norm(const struct lu_fid *fid) |
d7e09d03 PT |
431 | { |
432 | return fid_seq_is_norm(fid_seq(fid)); | |
433 | } | |
434 | ||
435 | /* convert an OST objid into an IDIF FID SEQ number */ | |
21aef7d9 | 436 | static inline __u64 fid_idif_seq(__u64 id, __u32 ost_idx) |
d7e09d03 PT |
437 | { |
438 | return FID_SEQ_IDIF | (ost_idx << 16) | ((id >> 32) & 0xffff); | |
439 | } | |
440 | ||
441 | /* convert a packed IDIF FID into an OST objid */ | |
21aef7d9 | 442 | static inline __u64 fid_idif_id(__u64 seq, __u32 oid, __u32 ver) |
d7e09d03 PT |
443 | { |
444 | return ((__u64)ver << 48) | ((seq & 0xffff) << 32) | oid; | |
445 | } | |
446 | ||
447 | /* extract ost index from IDIF FID */ | |
448 | static inline __u32 fid_idif_ost_idx(const struct lu_fid *fid) | |
449 | { | |
d7e09d03 PT |
450 | return (fid_seq(fid) >> 16) & 0xffff; |
451 | } | |
452 | ||
453 | /* extract OST sequence (group) from a wire ost_id (id/seq) pair */ | |
21aef7d9 | 454 | static inline __u64 ostid_seq(const struct ost_id *ostid) |
d7e09d03 PT |
455 | { |
456 | if (fid_seq_is_mdt0(ostid->oi.oi_seq)) | |
457 | return FID_SEQ_OST_MDT0; | |
458 | ||
22144626 | 459 | if (unlikely(fid_seq_is_default(ostid->oi.oi_seq))) |
d7e09d03 PT |
460 | return FID_SEQ_LOV_DEFAULT; |
461 | ||
462 | if (fid_is_idif(&ostid->oi_fid)) | |
463 | return FID_SEQ_OST_MDT0; | |
464 | ||
465 | return fid_seq(&ostid->oi_fid); | |
466 | } | |
467 | ||
468 | /* extract OST objid from a wire ost_id (id/seq) pair */ | |
21aef7d9 | 469 | static inline __u64 ostid_id(const struct ost_id *ostid) |
d7e09d03 | 470 | { |
22144626 | 471 | if (fid_seq_is_mdt0(ostid->oi.oi_seq)) |
d7e09d03 PT |
472 | return ostid->oi.oi_id & IDIF_OID_MASK; |
473 | ||
22144626 FY |
474 | if (unlikely(fid_seq_is_default(ostid->oi.oi_seq))) |
475 | return ostid->oi.oi_id; | |
476 | ||
d7e09d03 PT |
477 | if (fid_is_idif(&ostid->oi_fid)) |
478 | return fid_idif_id(fid_seq(&ostid->oi_fid), | |
479 | fid_oid(&ostid->oi_fid), 0); | |
480 | ||
481 | return fid_oid(&ostid->oi_fid); | |
482 | } | |
483 | ||
484 | static inline void ostid_set_seq(struct ost_id *oi, __u64 seq) | |
485 | { | |
486 | if (fid_seq_is_mdt0(seq) || fid_seq_is_default(seq)) { | |
487 | oi->oi.oi_seq = seq; | |
488 | } else { | |
489 | oi->oi_fid.f_seq = seq; | |
490 | /* Note: if f_oid + f_ver is zero, we need init it | |
491 | * to be 1, otherwise, ostid_seq will treat this | |
a1e616b0 OD |
492 | * as old ostid (oi_seq == 0) |
493 | */ | |
d7e09d03 PT |
494 | if (oi->oi_fid.f_oid == 0 && oi->oi_fid.f_ver == 0) |
495 | oi->oi_fid.f_oid = LUSTRE_FID_INIT_OID; | |
496 | } | |
497 | } | |
498 | ||
499 | static inline void ostid_set_seq_mdt0(struct ost_id *oi) | |
500 | { | |
501 | ostid_set_seq(oi, FID_SEQ_OST_MDT0); | |
502 | } | |
503 | ||
504 | static inline void ostid_set_seq_echo(struct ost_id *oi) | |
505 | { | |
506 | ostid_set_seq(oi, FID_SEQ_ECHO); | |
507 | } | |
508 | ||
509 | static inline void ostid_set_seq_llog(struct ost_id *oi) | |
510 | { | |
511 | ostid_set_seq(oi, FID_SEQ_LLOG); | |
512 | } | |
513 | ||
514 | /** | |
515 | * Note: we need check oi_seq to decide where to set oi_id, | |
516 | * so oi_seq should always be set ahead of oi_id. | |
517 | */ | |
518 | static inline void ostid_set_id(struct ost_id *oi, __u64 oid) | |
519 | { | |
22144626 | 520 | if (fid_seq_is_mdt0(oi->oi.oi_seq)) { |
d7e09d03 | 521 | if (oid >= IDIF_MAX_OID) { |
25782b53 AD |
522 | CERROR("Too large OID %#llx to set MDT0 " DOSTID "\n", |
523 | oid, POSTID(oi)); | |
d7e09d03 PT |
524 | return; |
525 | } | |
526 | oi->oi.oi_id = oid; | |
22144626 FY |
527 | } else if (fid_is_idif(&oi->oi_fid)) { |
528 | if (oid >= IDIF_MAX_OID) { | |
25782b53 | 529 | CERROR("Too large OID %#llx to set IDIF " DOSTID "\n", |
22144626 FY |
530 | oid, POSTID(oi)); |
531 | return; | |
532 | } | |
533 | oi->oi_fid.f_seq = fid_idif_seq(oid, | |
534 | fid_idif_ost_idx(&oi->oi_fid)); | |
535 | oi->oi_fid.f_oid = oid; | |
536 | oi->oi_fid.f_ver = oid >> 48; | |
d7e09d03 | 537 | } else { |
a5c954ef | 538 | if (oid >= OBIF_MAX_OID) { |
10457d4b | 539 | CERROR("Bad %llu to set " DOSTID "\n", oid, POSTID(oi)); |
d7e09d03 PT |
540 | return; |
541 | } | |
542 | oi->oi_fid.f_oid = oid; | |
543 | } | |
544 | } | |
545 | ||
22144626 | 546 | static inline int fid_set_id(struct lu_fid *fid, __u64 oid) |
d7e09d03 | 547 | { |
22144626 FY |
548 | if (unlikely(fid_seq_is_igif(fid->f_seq))) { |
549 | CERROR("bad IGIF, "DFID"\n", PFID(fid)); | |
550 | return -EBADF; | |
551 | } | |
552 | ||
553 | if (fid_is_idif(fid)) { | |
554 | if (oid >= IDIF_MAX_OID) { | |
25782b53 | 555 | CERROR("Too large OID %#llx to set IDIF " DFID "\n", |
22144626 FY |
556 | (unsigned long long)oid, PFID(fid)); |
557 | return -EBADF; | |
d7e09d03 | 558 | } |
22144626 FY |
559 | fid->f_seq = fid_idif_seq(oid, fid_idif_ost_idx(fid)); |
560 | fid->f_oid = oid; | |
561 | fid->f_ver = oid >> 48; | |
d7e09d03 | 562 | } else { |
a5c954ef | 563 | if (oid >= OBIF_MAX_OID) { |
25782b53 | 564 | CERROR("Too large OID %#llx to set REG " DFID "\n", |
22144626 FY |
565 | (unsigned long long)oid, PFID(fid)); |
566 | return -EBADF; | |
567 | } | |
568 | fid->f_oid = oid; | |
d7e09d03 | 569 | } |
22144626 | 570 | return 0; |
d7e09d03 PT |
571 | } |
572 | ||
573 | /** | |
574 | * Unpack an OST object id/seq (group) into a FID. This is needed for | |
575 | * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper | |
576 | * FIDs. Note that if an id/seq is already in FID/IDIF format it will | |
577 | * be passed through unchanged. Only legacy OST objects in "group 0" | |
578 | * will be mapped into the IDIF namespace so that they can fit into the | |
579 | * struct lu_fid fields without loss. For reference see: | |
25ed6a5e | 580 | * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs |
d7e09d03 PT |
581 | */ |
582 | static inline int ostid_to_fid(struct lu_fid *fid, struct ost_id *ostid, | |
583 | __u32 ost_idx) | |
584 | { | |
22144626 FY |
585 | __u64 seq = ostid_seq(ostid); |
586 | ||
d7e09d03 PT |
587 | if (ost_idx > 0xffff) { |
588 | CERROR("bad ost_idx, "DOSTID" ost_idx:%u\n", POSTID(ostid), | |
589 | ost_idx); | |
590 | return -EBADF; | |
591 | } | |
592 | ||
22144626 FY |
593 | if (fid_seq_is_mdt0(seq)) { |
594 | __u64 oid = ostid_id(ostid); | |
595 | ||
d7e09d03 PT |
596 | /* This is a "legacy" (old 1.x/2.early) OST object in "group 0" |
597 | * that we map into the IDIF namespace. It allows up to 2^48 | |
598 | * objects per OST, as this is the object namespace that has | |
599 | * been in production for years. This can handle create rates | |
a1e616b0 OD |
600 | * of 1M objects/s/OST for 9 years, or combinations thereof. |
601 | */ | |
22144626 | 602 | if (oid >= IDIF_MAX_OID) { |
defa220f OD |
603 | CERROR("bad MDT0 id, " DOSTID " ost_idx:%u\n", |
604 | POSTID(ostid), ost_idx); | |
605 | return -EBADF; | |
d7e09d03 | 606 | } |
22144626 | 607 | fid->f_seq = fid_idif_seq(oid, ost_idx); |
d7e09d03 | 608 | /* truncate to 32 bits by assignment */ |
22144626 | 609 | fid->f_oid = oid; |
d7e09d03 | 610 | /* in theory, not currently used */ |
22144626 FY |
611 | fid->f_ver = oid >> 48; |
612 | } else if (likely(!fid_seq_is_default(seq))) { | |
d7e09d03 PT |
613 | /* This is either an IDIF object, which identifies objects across |
614 | * all OSTs, or a regular FID. The IDIF namespace maps legacy | |
615 | * OST objects into the FID namespace. In both cases, we just | |
a1e616b0 OD |
616 | * pass the FID through, no conversion needed. |
617 | */ | |
d7e09d03 | 618 | if (ostid->oi_fid.f_ver != 0) { |
10457d4b OD |
619 | CERROR("bad MDT0 id, " DOSTID " ost_idx:%u\n", |
620 | POSTID(ostid), ost_idx); | |
d7e09d03 PT |
621 | return -EBADF; |
622 | } | |
623 | *fid = ostid->oi_fid; | |
624 | } | |
625 | ||
626 | return 0; | |
627 | } | |
628 | ||
629 | /* pack any OST FID into an ostid (id/seq) for the wire/disk */ | |
630 | static inline int fid_to_ostid(const struct lu_fid *fid, struct ost_id *ostid) | |
631 | { | |
632 | if (unlikely(fid_seq_is_igif(fid->f_seq))) { | |
633 | CERROR("bad IGIF, "DFID"\n", PFID(fid)); | |
634 | return -EBADF; | |
635 | } | |
636 | ||
637 | if (fid_is_idif(fid)) { | |
638 | ostid_set_seq_mdt0(ostid); | |
639 | ostid_set_id(ostid, fid_idif_id(fid_seq(fid), fid_oid(fid), | |
640 | fid_ver(fid))); | |
641 | } else { | |
642 | ostid->oi_fid = *fid; | |
643 | } | |
644 | ||
645 | return 0; | |
646 | } | |
647 | ||
648 | /* Check whether the fid is for LAST_ID */ | |
d8f183b3 | 649 | static inline bool fid_is_last_id(const struct lu_fid *fid) |
d7e09d03 | 650 | { |
66cc83e9 | 651 | return (fid_oid(fid) == 0); |
d7e09d03 PT |
652 | } |
653 | ||
654 | /** | |
655 | * Get inode number from a igif. | |
656 | * \param fid a igif to get inode number from. | |
657 | * \return inode number for the igif. | |
658 | */ | |
659 | static inline ino_t lu_igif_ino(const struct lu_fid *fid) | |
660 | { | |
661 | return fid_seq(fid); | |
662 | } | |
663 | ||
d7e09d03 PT |
664 | /** |
665 | * Get inode generation from a igif. | |
666 | * \param fid a igif to get inode generation from. | |
667 | * \return inode generation for the igif. | |
668 | */ | |
669 | static inline __u32 lu_igif_gen(const struct lu_fid *fid) | |
670 | { | |
671 | return fid_oid(fid); | |
672 | } | |
673 | ||
674 | /** | |
675 | * Build igif from the inode number/generation. | |
676 | */ | |
677 | static inline void lu_igif_build(struct lu_fid *fid, __u32 ino, __u32 gen) | |
678 | { | |
679 | fid->f_seq = ino; | |
680 | fid->f_oid = gen; | |
681 | fid->f_ver = 0; | |
682 | } | |
683 | ||
684 | /* | |
685 | * Fids are transmitted across network (in the sender byte-ordering), | |
686 | * and stored on disk in big-endian order. | |
687 | */ | |
688 | static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src) | |
689 | { | |
d7e09d03 PT |
690 | dst->f_seq = cpu_to_le64(fid_seq(src)); |
691 | dst->f_oid = cpu_to_le32(fid_oid(src)); | |
692 | dst->f_ver = cpu_to_le32(fid_ver(src)); | |
693 | } | |
694 | ||
695 | static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src) | |
696 | { | |
d7e09d03 PT |
697 | dst->f_seq = le64_to_cpu(fid_seq(src)); |
698 | dst->f_oid = le32_to_cpu(fid_oid(src)); | |
699 | dst->f_ver = le32_to_cpu(fid_ver(src)); | |
700 | } | |
701 | ||
702 | static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src) | |
703 | { | |
d7e09d03 PT |
704 | dst->f_seq = cpu_to_be64(fid_seq(src)); |
705 | dst->f_oid = cpu_to_be32(fid_oid(src)); | |
706 | dst->f_ver = cpu_to_be32(fid_ver(src)); | |
707 | } | |
708 | ||
709 | static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src) | |
710 | { | |
d7e09d03 PT |
711 | dst->f_seq = be64_to_cpu(fid_seq(src)); |
712 | dst->f_oid = be32_to_cpu(fid_oid(src)); | |
713 | dst->f_ver = be32_to_cpu(fid_ver(src)); | |
714 | } | |
715 | ||
d8f183b3 | 716 | static inline bool fid_is_sane(const struct lu_fid *fid) |
d7e09d03 | 717 | { |
d2a13989 | 718 | return fid && |
d7e09d03 PT |
719 | ((fid_seq(fid) >= FID_SEQ_START && fid_ver(fid) == 0) || |
720 | fid_is_igif(fid) || fid_is_idif(fid) || | |
721 | fid_seq_is_rsvd(fid_seq(fid))); | |
722 | } | |
723 | ||
d8f183b3 | 724 | static inline bool lu_fid_eq(const struct lu_fid *f0, const struct lu_fid *f1) |
d7e09d03 | 725 | { |
ec83e611 | 726 | return memcmp(f0, f1, sizeof(*f0)) == 0; |
d7e09d03 PT |
727 | } |
728 | ||
729 | #define __diff_normalize(val0, val1) \ | |
730 | ({ \ | |
731 | typeof(val0) __val0 = (val0); \ | |
732 | typeof(val1) __val1 = (val1); \ | |
733 | \ | |
b2952d62 | 734 | (__val0 == __val1 ? 0 : __val0 > __val1 ? 1 : -1); \ |
d7e09d03 PT |
735 | }) |
736 | ||
737 | static inline int lu_fid_cmp(const struct lu_fid *f0, | |
738 | const struct lu_fid *f1) | |
739 | { | |
740 | return | |
741 | __diff_normalize(fid_seq(f0), fid_seq(f1)) ?: | |
742 | __diff_normalize(fid_oid(f0), fid_oid(f1)) ?: | |
743 | __diff_normalize(fid_ver(f0), fid_ver(f1)); | |
744 | } | |
745 | ||
c5b60ba7 | 746 | static inline void ostid_cpu_to_le(const struct ost_id *src_oi, |
d7e09d03 PT |
747 | struct ost_id *dst_oi) |
748 | { | |
749 | if (fid_seq_is_mdt0(ostid_seq(src_oi))) { | |
750 | dst_oi->oi.oi_id = cpu_to_le64(src_oi->oi.oi_id); | |
751 | dst_oi->oi.oi_seq = cpu_to_le64(src_oi->oi.oi_seq); | |
752 | } else { | |
753 | fid_cpu_to_le(&dst_oi->oi_fid, &src_oi->oi_fid); | |
754 | } | |
755 | } | |
756 | ||
c5b60ba7 | 757 | static inline void ostid_le_to_cpu(const struct ost_id *src_oi, |
d7e09d03 PT |
758 | struct ost_id *dst_oi) |
759 | { | |
760 | if (fid_seq_is_mdt0(ostid_seq(src_oi))) { | |
761 | dst_oi->oi.oi_id = le64_to_cpu(src_oi->oi.oi_id); | |
762 | dst_oi->oi.oi_seq = le64_to_cpu(src_oi->oi.oi_seq); | |
763 | } else { | |
764 | fid_le_to_cpu(&dst_oi->oi_fid, &src_oi->oi_fid); | |
765 | } | |
766 | } | |
767 | ||
768 | /** @} lu_fid */ | |
769 | ||
770 | /** \defgroup lu_dir lu_dir | |
a1e616b0 OD |
771 | * @{ |
772 | */ | |
d7e09d03 PT |
773 | |
774 | /** | |
775 | * Enumeration of possible directory entry attributes. | |
776 | * | |
777 | * Attributes follow directory entry header in the order they appear in this | |
778 | * enumeration. | |
779 | */ | |
780 | enum lu_dirent_attrs { | |
781 | LUDA_FID = 0x0001, | |
782 | LUDA_TYPE = 0x0002, | |
783 | LUDA_64BITHASH = 0x0004, | |
d7e09d03 PT |
784 | }; |
785 | ||
d7e09d03 PT |
786 | /** |
787 | * Layout of readdir pages, as transmitted on wire. | |
788 | */ | |
789 | struct lu_dirent { | |
790 | /** valid if LUDA_FID is set. */ | |
791 | struct lu_fid lde_fid; | |
792 | /** a unique entry identifier: a hash or an offset. */ | |
793 | __u64 lde_hash; | |
794 | /** total record length, including all attributes. */ | |
795 | __u16 lde_reclen; | |
796 | /** name length */ | |
797 | __u16 lde_namelen; | |
798 | /** optional variable size attributes following this entry. | |
799 | * taken from enum lu_dirent_attrs. | |
800 | */ | |
801 | __u32 lde_attrs; | |
802 | /** name is followed by the attributes indicated in ->ldp_attrs, in | |
803 | * their natural order. After the last attribute, padding bytes are | |
804 | * added to make ->lde_reclen a multiple of 8. | |
805 | */ | |
806 | char lde_name[0]; | |
807 | }; | |
808 | ||
809 | /* | |
810 | * Definitions of optional directory entry attributes formats. | |
811 | * | |
812 | * Individual attributes do not have their length encoded in a generic way. It | |
813 | * is assumed that consumer of an attribute knows its format. This means that | |
814 | * it is impossible to skip over an unknown attribute, except by skipping over all | |
815 | * remaining attributes (by using ->lde_reclen), which is not too | |
816 | * constraining, because new server versions will append new attributes at | |
817 | * the end of an entry. | |
818 | */ | |
819 | ||
820 | /** | |
821 | * Fid directory attribute: a fid of an object referenced by the entry. This | |
822 | * will be almost always requested by the client and supplied by the server. | |
823 | * | |
824 | * Aligned to 8 bytes. | |
825 | */ | |
826 | /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */ | |
827 | ||
828 | /** | |
829 | * File type. | |
830 | * | |
831 | * Aligned to 2 bytes. | |
832 | */ | |
833 | struct luda_type { | |
834 | __u16 lt_type; | |
835 | }; | |
836 | ||
79783924 PT |
837 | #ifndef IFSHIFT |
838 | #define IFSHIFT 12 | |
839 | #endif | |
840 | ||
841 | #ifndef IFTODT | |
842 | #define IFTODT(type) (((type) & S_IFMT) >> IFSHIFT) | |
843 | #endif | |
844 | #ifndef DTTOIF | |
845 | #define DTTOIF(dirtype) ((dirtype) << IFSHIFT) | |
846 | #endif | |
847 | ||
d7e09d03 PT |
848 | struct lu_dirpage { |
849 | __u64 ldp_hash_start; | |
850 | __u64 ldp_hash_end; | |
851 | __u32 ldp_flags; | |
852 | __u32 ldp_pad0; | |
853 | struct lu_dirent ldp_entries[0]; | |
854 | }; | |
855 | ||
856 | enum lu_dirpage_flags { | |
857 | /** | |
858 | * dirpage contains no entry. | |
859 | */ | |
860 | LDF_EMPTY = 1 << 0, | |
861 | /** | |
862 | * last entry's lde_hash equals ldp_hash_end. | |
863 | */ | |
864 | LDF_COLLIDE = 1 << 1 | |
865 | }; | |
866 | ||
867 | static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp) | |
868 | { | |
869 | if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY) | |
870 | return NULL; | |
871 | else | |
872 | return dp->ldp_entries; | |
873 | } | |
874 | ||
875 | static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent) | |
876 | { | |
877 | struct lu_dirent *next; | |
878 | ||
879 | if (le16_to_cpu(ent->lde_reclen) != 0) | |
880 | next = ((void *)ent) + le16_to_cpu(ent->lde_reclen); | |
881 | else | |
882 | next = NULL; | |
883 | ||
884 | return next; | |
885 | } | |
886 | ||
8060c901 | 887 | static inline size_t lu_dirent_calc_size(size_t namelen, __u16 attr) |
d7e09d03 | 888 | { |
8060c901 | 889 | size_t size; |
d7e09d03 PT |
890 | |
891 | if (attr & LUDA_TYPE) { | |
8060c901 | 892 | const size_t align = sizeof(struct luda_type) - 1; |
50ffcb7e | 893 | |
d7e09d03 PT |
894 | size = (sizeof(struct lu_dirent) + namelen + align) & ~align; |
895 | size += sizeof(struct luda_type); | |
da5ecb4d | 896 | } else { |
d7e09d03 | 897 | size = sizeof(struct lu_dirent) + namelen; |
da5ecb4d | 898 | } |
d7e09d03 PT |
899 | |
900 | return (size + 7) & ~7; | |
901 | } | |
902 | ||
d7e09d03 PT |
903 | #define MDS_DIR_END_OFF 0xfffffffffffffffeULL |
904 | ||
905 | /** | |
906 | * MDS_READPAGE page size | |
907 | * | |
908 | * This is the directory page size packed in MDS_READPAGE RPC. | |
ea1754a0 | 909 | * It's different than PAGE_SIZE because the client needs to |
d7e09d03 PT |
910 | * access the struct lu_dirpage header packed at the beginning of |
911 | * the "page" and without this there isn't any way to know find the | |
ea1754a0 | 912 | * lu_dirpage header is if client and server PAGE_SIZE differ. |
d7e09d03 PT |
913 | */ |
914 | #define LU_PAGE_SHIFT 12 | |
915 | #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT) | |
916 | #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1)) | |
917 | ||
09cbfeaf | 918 | #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT)) |
d7e09d03 PT |
919 | |
920 | /** @} lu_dir */ | |
921 | ||
922 | struct lustre_handle { | |
923 | __u64 cookie; | |
924 | }; | |
c9f6bb96 | 925 | |
d7e09d03 PT |
926 | #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL |
927 | ||
d8f183b3 | 928 | static inline bool lustre_handle_is_used(const struct lustre_handle *lh) |
d7e09d03 PT |
929 | { |
930 | return lh->cookie != 0ull; | |
931 | } | |
932 | ||
d8f183b3 JH |
933 | static inline bool lustre_handle_equal(const struct lustre_handle *lh1, |
934 | const struct lustre_handle *lh2) | |
d7e09d03 PT |
935 | { |
936 | return lh1->cookie == lh2->cookie; | |
937 | } | |
938 | ||
939 | static inline void lustre_handle_copy(struct lustre_handle *tgt, | |
ac8f0a5c | 940 | const struct lustre_handle *src) |
d7e09d03 PT |
941 | { |
942 | tgt->cookie = src->cookie; | |
943 | } | |
944 | ||
945 | /* flags for lm_flags */ | |
946 | #define MSGHDR_AT_SUPPORT 0x1 | |
947 | #define MSGHDR_CKSUM_INCOMPAT18 0x2 | |
948 | ||
949 | #define lustre_msg lustre_msg_v2 | |
950 | /* we depend on this structure to be 8-byte aligned */ | |
951 | /* this type is only endian-adjusted in lustre_unpack_msg() */ | |
952 | struct lustre_msg_v2 { | |
953 | __u32 lm_bufcount; | |
954 | __u32 lm_secflvr; | |
955 | __u32 lm_magic; | |
956 | __u32 lm_repsize; | |
957 | __u32 lm_cksum; | |
958 | __u32 lm_flags; | |
959 | __u32 lm_padding_2; | |
960 | __u32 lm_padding_3; | |
961 | __u32 lm_buflens[0]; | |
962 | }; | |
963 | ||
964 | /* without gss, ptlrpc_body is put at the first buffer. */ | |
965 | #define PTLRPC_NUM_VERSIONS 4 | |
c9fe1f7f | 966 | |
d7e09d03 PT |
967 | struct ptlrpc_body_v3 { |
968 | struct lustre_handle pb_handle; | |
969 | __u32 pb_type; | |
970 | __u32 pb_version; | |
971 | __u32 pb_opc; | |
972 | __u32 pb_status; | |
e33115e3 GP |
973 | __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */ |
974 | __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */ | |
975 | __u16 pb_padding0; | |
976 | __u32 pb_padding1; | |
d7e09d03 PT |
977 | __u64 pb_last_committed; |
978 | __u64 pb_transno; | |
979 | __u32 pb_flags; | |
980 | __u32 pb_op_flags; | |
981 | __u32 pb_conn_cnt; | |
982 | __u32 pb_timeout; /* for req, the deadline, for rep, the service est */ | |
983 | __u32 pb_service_time; /* for rep, actual service time */ | |
984 | __u32 pb_limit; | |
985 | __u64 pb_slv; | |
986 | /* VBR: pre-versions */ | |
987 | __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS]; | |
8bcaef92 | 988 | __u64 pb_mbits; /**< match bits for bulk request */ |
d7e09d03 | 989 | /* padding for future needs */ |
8bcaef92 LZ |
990 | __u64 pb_padding64_0; |
991 | __u64 pb_padding64_1; | |
992 | __u64 pb_padding64_2; | |
c9fe1f7f | 993 | char pb_jobid[LUSTRE_JOBID_SIZE]; |
d7e09d03 | 994 | }; |
c9f6bb96 | 995 | |
d7e09d03 PT |
996 | #define ptlrpc_body ptlrpc_body_v3 |
997 | ||
998 | struct ptlrpc_body_v2 { | |
999 | struct lustre_handle pb_handle; | |
1000 | __u32 pb_type; | |
1001 | __u32 pb_version; | |
1002 | __u32 pb_opc; | |
1003 | __u32 pb_status; | |
e33115e3 GP |
1004 | __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */ |
1005 | __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */ | |
1006 | __u16 pb_padding0; | |
1007 | __u32 pb_padding1; | |
d7e09d03 PT |
1008 | __u64 pb_last_committed; |
1009 | __u64 pb_transno; | |
1010 | __u32 pb_flags; | |
1011 | __u32 pb_op_flags; | |
1012 | __u32 pb_conn_cnt; | |
1013 | __u32 pb_timeout; /* for req, the deadline, for rep, the service est */ | |
1014 | __u32 pb_service_time; /* for rep, actual service time, also used for | |
a1e616b0 OD |
1015 | * net_latency of req |
1016 | */ | |
d7e09d03 PT |
1017 | __u32 pb_limit; |
1018 | __u64 pb_slv; | |
1019 | /* VBR: pre-versions */ | |
1020 | __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS]; | |
8bcaef92 | 1021 | __u64 pb_mbits; /**< unused in V2 */ |
d7e09d03 | 1022 | /* padding for future needs */ |
8bcaef92 LZ |
1023 | __u64 pb_padding64_0; |
1024 | __u64 pb_padding64_1; | |
1025 | __u64 pb_padding64_2; | |
d7e09d03 PT |
1026 | }; |
1027 | ||
d7e09d03 PT |
1028 | /* message body offset for lustre_msg_v2 */ |
1029 | /* ptlrpc body offset in all request/reply messages */ | |
1030 | #define MSG_PTLRPC_BODY_OFF 0 | |
1031 | ||
1032 | /* normal request/reply message record offset */ | |
1033 | #define REQ_REC_OFF 1 | |
1034 | #define REPLY_REC_OFF 1 | |
1035 | ||
1036 | /* ldlm request message body offset */ | |
1037 | #define DLM_LOCKREQ_OFF 1 /* lockreq offset */ | |
1038 | #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */ | |
1039 | ||
1040 | /* ldlm intent lock message body offset */ | |
1041 | #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */ | |
1042 | #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */ | |
1043 | ||
1044 | /* ldlm reply message body offset */ | |
1045 | #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */ | |
1046 | #define DLM_REPLY_REC_OFF 2 /* reply record offset */ | |
1047 | ||
1048 | /** only use in req->rq_{req,rep}_swab_mask */ | |
1049 | #define MSG_PTLRPC_HEADER_OFF 31 | |
1050 | ||
1051 | /* Flags that are operation-specific go in the top 16 bits. */ | |
1052 | #define MSG_OP_FLAG_MASK 0xffff0000 | |
1053 | #define MSG_OP_FLAG_SHIFT 16 | |
1054 | ||
1055 | /* Flags that apply to all requests are in the bottom 16 bits */ | |
1056 | #define MSG_GEN_FLAG_MASK 0x0000ffff | |
1057 | #define MSG_LAST_REPLAY 0x0001 | |
1058 | #define MSG_RESENT 0x0002 | |
1059 | #define MSG_REPLAY 0x0004 | |
1060 | /* #define MSG_AT_SUPPORT 0x0008 | |
1061 | * This was used in early prototypes of adaptive timeouts, and while there | |
1062 | * shouldn't be any users of that code there also isn't a need for using this | |
a1e616b0 OD |
1063 | * bits. Defer usage until at least 1.10 to avoid potential conflict. |
1064 | */ | |
d7e09d03 PT |
1065 | #define MSG_DELAY_REPLAY 0x0010 |
1066 | #define MSG_VERSION_REPLAY 0x0020 | |
1067 | #define MSG_REQ_REPLAY_DONE 0x0040 | |
1068 | #define MSG_LOCK_REPLAY_DONE 0x0080 | |
1069 | ||
1070 | /* | |
1071 | * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT) | |
1072 | */ | |
1073 | ||
1074 | #define MSG_CONNECT_RECOVERING 0x00000001 | |
1075 | #define MSG_CONNECT_RECONNECT 0x00000002 | |
1076 | #define MSG_CONNECT_REPLAYABLE 0x00000004 | |
df693dbe | 1077 | /*#define MSG_CONNECT_PEER 0x8 */ |
d7e09d03 PT |
1078 | #define MSG_CONNECT_LIBCLIENT 0x00000010 |
1079 | #define MSG_CONNECT_INITIAL 0x00000020 | |
1080 | #define MSG_CONNECT_ASYNC 0x00000040 | |
1081 | #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */ | |
1082 | #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */ | |
1083 | ||
1084 | /* Connect flags */ | |
a1e616b0 OD |
1085 | #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/ |
1086 | #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */ | |
1087 | #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */ | |
1088 | #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */ | |
1089 | #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */ | |
1090 | #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */ | |
1091 | #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */ | |
1092 | #define OBD_CONNECT_ACL 0x80ULL /*access control lists */ | |
1093 | #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */ | |
d7e09d03 | 1094 | #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/ |
a1e616b0 OD |
1095 | #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */ |
1096 | #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */ | |
1097 | #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/ | |
d7e09d03 PT |
1098 | #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated. |
1099 | *We do not support JOIN FILE | |
1100 | *anymore, reserve this flags | |
1101 | *just for preventing such bit | |
a1e616b0 OD |
1102 | *to be reused. |
1103 | */ | |
1104 | #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/ | |
1105 | #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/ | |
341f1f0a FY |
1106 | #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used |
1107 | * in production. Removed in | |
1108 | * 2.9. Keep this flag to | |
1109 | * avoid reuse. | |
1110 | */ | |
1111 | #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force, | |
1112 | * never used in production. | |
1113 | * Removed in 2.9. Keep this | |
1114 | * flag to avoid reuse | |
1115 | */ | |
a1e616b0 OD |
1116 | #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */ |
1117 | #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */ | |
1118 | #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */ | |
1119 | #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */ | |
1120 | #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */ | |
1121 | #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */ | |
1122 | #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */ | |
d7e09d03 | 1123 | #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */ |
a1e616b0 | 1124 | #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */ |
d7e09d03 PT |
1125 | #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */ |
1126 | #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */ | |
a1e616b0 OD |
1127 | #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/ |
1128 | #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */ | |
1129 | #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */ | |
1130 | #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */ | |
d7e09d03 PT |
1131 | #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */ |
1132 | #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */ | |
1133 | #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */ | |
1134 | #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */ | |
1135 | #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */ | |
1136 | #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits | |
a1e616b0 OD |
1137 | * directory hash |
1138 | */ | |
d7e09d03 PT |
1139 | #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */ |
1140 | #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */ | |
1141 | #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */ | |
1142 | #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */ | |
1143 | #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS | |
a1e616b0 OD |
1144 | * RPC error properly |
1145 | */ | |
d7e09d03 | 1146 | #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for |
a1e616b0 OD |
1147 | * finer space reservation |
1148 | */ | |
d7e09d03 | 1149 | #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8 |
a1e616b0 OD |
1150 | * policy and 2.x server |
1151 | */ | |
d7e09d03 PT |
1152 | #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */ |
1153 | #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */ | |
1154 | #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */ | |
1155 | #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */ | |
1156 | #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */ | |
69342b78 | 1157 | #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* flock deadlock detection */ |
63d42578 | 1158 | #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/*create stripe disposition*/ |
98f2ec61 LS |
1159 | #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack |
1160 | * name in request | |
1161 | */ | |
865b734e | 1162 | #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */ |
ac094f38 | 1163 | #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */ |
872cbbdf GP |
1164 | #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify |
1165 | * RPCs in parallel | |
1166 | */ | |
4edc630a | 1167 | #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL/* striped DNE dir */ |
9ed8e66a | 1168 | #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */ |
0b83bc16 | 1169 | #define OBD_CONNECT_LOCK_AHEAD 0x1000000000000000ULL /* lock ahead */ |
8bcaef92 LZ |
1170 | /** bulk matchbits is sent within ptlrpc_body */ |
1171 | #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL | |
59b6693a | 1172 | #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */ |
3ce173a1 | 1173 | #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */ |
69342b78 | 1174 | |
d7e09d03 PT |
1175 | /* XXX README XXX: |
1176 | * Please DO NOT add flag values here before first ensuring that this same | |
1177 | * flag value is not in use on some other branch. Please clear any such | |
1178 | * changes with senior engineers before starting to use a new flag. Then, | |
1179 | * submit a small patch against EVERY branch that ONLY adds the new flag, | |
1180 | * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the | |
1181 | * flag to check_obd_connect_data(), and updates wiretests accordingly, so it | |
faf2ca92 OD |
1182 | * can be approved and landed easily to reserve the flag for future use. |
1183 | */ | |
d7e09d03 PT |
1184 | |
1185 | /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS | |
1186 | * connection. It is a temporary bug fix for Imperative Recovery interop | |
1187 | * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for | |
faf2ca92 OD |
1188 | * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. |
1189 | */ | |
d7e09d03 PT |
1190 | #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS |
1191 | ||
1192 | #define OCD_HAS_FLAG(ocd, flg) \ | |
1193 | (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg)) | |
1194 | ||
d7e09d03 PT |
1195 | /* Features required for this version of the client to work with server */ |
1196 | #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \ | |
1197 | OBD_CONNECT_FULL20) | |
1198 | ||
d7e09d03 PT |
1199 | /* This structure is used for both request and reply. |
1200 | * | |
1201 | * If we eventually have separate connect data for different types, which we | |
a1e616b0 OD |
1202 | * almost certainly will, then perhaps we stick a union in here. |
1203 | */ | |
d7e09d03 PT |
1204 | struct obd_connect_data { |
1205 | __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */ | |
1206 | __u32 ocd_version; /* lustre release version number */ | |
1207 | __u32 ocd_grant; /* initial cache grant amount (bytes) */ | |
1208 | __u32 ocd_index; /* LOV index to connect to */ | |
1209 | __u32 ocd_brw_size; /* Maximum BRW size in bytes */ | |
1210 | __u64 ocd_ibits_known; /* inode bits this client understands */ | |
1211 | __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */ | |
1212 | __u8 ocd_inodespace; /* log2 of the per-inode space consumption */ | |
1213 | __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */ | |
a1e616b0 | 1214 | __u32 ocd_unused; /* also fix lustre_swab_connect */ |
d7e09d03 PT |
1215 | __u64 ocd_transno; /* first transno from client to be replayed */ |
1216 | __u32 ocd_group; /* MDS group on OST */ | |
1217 | __u32 ocd_cksum_types; /* supported checksum algorithms */ | |
1218 | __u32 ocd_max_easize; /* How big LOV EA can be on MDS */ | |
1219 | __u32 ocd_instance; /* instance # of this target */ | |
1220 | __u64 ocd_maxbytes; /* Maximum stripe size in bytes */ | |
1221 | /* Fields after ocd_maxbytes are only accessible by the receiver | |
1222 | * if the corresponding flag in ocd_connect_flags is set. Accessing | |
1223 | * any field after ocd_maxbytes on the receiver without a valid flag | |
a1e616b0 OD |
1224 | * may result in out-of-bound memory access and kernel oops. |
1225 | */ | |
872cbbdf GP |
1226 | __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */ |
1227 | __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */ | |
1228 | __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */ | |
3ce173a1 | 1229 | __u64 ocd_connect_flags2; |
d7e09d03 PT |
1230 | __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */ |
1231 | __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */ | |
1232 | __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */ | |
1233 | __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */ | |
1234 | __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */ | |
1235 | __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */ | |
1236 | __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */ | |
1237 | __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */ | |
1238 | __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */ | |
1239 | __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */ | |
1240 | __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */ | |
1241 | __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */ | |
1242 | __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */ | |
1243 | }; | |
c9f6bb96 | 1244 | |
d7e09d03 PT |
1245 | /* XXX README XXX: |
1246 | * Please DO NOT use any fields here before first ensuring that this same | |
1247 | * field is not in use on some other branch. Please clear any such changes | |
1248 | * with senior engineers before starting to use a new field. Then, submit | |
1249 | * a small patch against EVERY branch that ONLY adds the new field along with | |
1250 | * the matching OBD_CONNECT flag, so that can be approved and landed easily to | |
a1e616b0 OD |
1251 | * reserve the flag for future use. |
1252 | */ | |
d7e09d03 | 1253 | |
d7e09d03 PT |
1254 | /* |
1255 | * Supported checksum algorithms. Up to 32 checksum types are supported. | |
1256 | * (32-bit mask stored in obd_connect_data::ocd_cksum_types) | |
1257 | * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new | |
1258 | * algorithm and also the OBD_FL_CKSUM* flags. | |
1259 | */ | |
d133210f | 1260 | enum cksum_type { |
b2952d62 MR |
1261 | OBD_CKSUM_CRC32 = 0x00000001, |
1262 | OBD_CKSUM_ADLER = 0x00000002, | |
1263 | OBD_CKSUM_CRC32C = 0x00000004, | |
d133210f | 1264 | }; |
d7e09d03 PT |
1265 | |
1266 | /* | |
1267 | * OST requests: OBDO & OBD request records | |
1268 | */ | |
1269 | ||
1270 | /* opcodes */ | |
efede67c | 1271 | enum ost_cmd { |
d7e09d03 PT |
1272 | OST_REPLY = 0, /* reply ? */ |
1273 | OST_GETATTR = 1, | |
1274 | OST_SETATTR = 2, | |
1275 | OST_READ = 3, | |
1276 | OST_WRITE = 4, | |
1277 | OST_CREATE = 5, | |
1278 | OST_DESTROY = 6, | |
1279 | OST_GET_INFO = 7, | |
1280 | OST_CONNECT = 8, | |
1281 | OST_DISCONNECT = 9, | |
1282 | OST_PUNCH = 10, | |
1283 | OST_OPEN = 11, | |
1284 | OST_CLOSE = 12, | |
1285 | OST_STATFS = 13, | |
1286 | OST_SYNC = 16, | |
1287 | OST_SET_INFO = 17, | |
e57721e7 | 1288 | OST_QUOTACHECK = 18, /* not used since 2.4 */ |
d7e09d03 PT |
1289 | OST_QUOTACTL = 19, |
1290 | OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */ | |
1291 | OST_LAST_OPC | |
efede67c | 1292 | }; |
d7e09d03 PT |
1293 | #define OST_FIRST_OPC OST_REPLY |
1294 | ||
1295 | enum obdo_flags { | |
1296 | OBD_FL_INLINEDATA = 0x00000001, | |
1297 | OBD_FL_OBDMDEXISTS = 0x00000002, | |
1298 | OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */ | |
a1e616b0 | 1299 | OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */ |
d7e09d03 | 1300 | OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/ |
b2952d62 | 1301 | OBD_FL_RECREATE_OBJS = 0x00000020, /* recreate missing obj */ |
d7e09d03 PT |
1302 | OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */ |
1303 | OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */ | |
1304 | OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */ | |
1305 | OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */ | |
1306 | OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */ | |
1307 | OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */ | |
1308 | OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */ | |
1309 | OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */ | |
1310 | OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */ | |
1311 | OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */ | |
1312 | OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */ | |
a1e616b0 | 1313 | OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client. |
d7e09d03 | 1314 | * XXX: obsoleted - reserved for old |
a1e616b0 OD |
1315 | * clients prior than 2.2 |
1316 | */ | |
d7e09d03 PT |
1317 | OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */ |
1318 | OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */ | |
e1798006 | 1319 | OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */ |
8bd3efb7 | 1320 | OBD_FL_SHORT_IO = 0x00400000, /* short io request */ |
d7e09d03 PT |
1321 | |
1322 | /* Note that while these checksum values are currently separate bits, | |
a1e616b0 OD |
1323 | * in 2.x we can actually allow all values from 1-31 if we wanted. |
1324 | */ | |
d7e09d03 PT |
1325 | OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER | |
1326 | OBD_FL_CKSUM_CRC32C, | |
1327 | ||
1328 | /* mask for local-only flag, which won't be sent over network */ | |
1329 | OBD_FL_LOCAL_MASK = 0xF0000000, | |
1330 | }; | |
1331 | ||
afebe4a5 FY |
1332 | /* |
1333 | * All LOV EA magics should have the same postfix, if some new version | |
1334 | * Lustre instroduces new LOV EA magic, then when down-grade to an old | |
1335 | * Lustre, even though the old version system does not recognizes such | |
1336 | * new magic, it still can distinguish the corrupted cases by checking | |
1337 | * the magic's postfix. | |
1338 | */ | |
1339 | #define LOV_MAGIC_MAGIC 0x0BD0 | |
1340 | #define LOV_MAGIC_MASK 0xFFFF | |
1341 | ||
1342 | #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC) | |
1343 | #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC) | |
1344 | #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC) | |
1345 | #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC) | |
dbf789ce JX |
1346 | /* reserved for specifying OSTs */ |
1347 | #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC) | |
afebe4a5 | 1348 | #define LOV_MAGIC LOV_MAGIC_V1 |
d7e09d03 PT |
1349 | |
1350 | /* | |
1351 | * magic for fully defined striping | |
1352 | * the idea is that we should have different magics for striping "hints" | |
1353 | * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct | |
1354 | * lov_mds_md_v[13]). at the moment the magics are used in wire protocol, | |
1355 | * we can't just change it w/o long way preparation, but we still need a | |
1356 | * mechanism to allow LOD to differentiate hint versus ready striping. | |
1357 | * so, at the moment we do a trick: MDT knows what to expect from request | |
1358 | * depending on the case (replay uses ready striping, non-replay req uses | |
1359 | * hints), so MDT replaces magic with appropriate one and now LOD can | |
1360 | * easily understand what's inside -bzzz | |
1361 | */ | |
1362 | #define LOV_MAGIC_V1_DEF 0x0CD10BD0 | |
1363 | #define LOV_MAGIC_V3_DEF 0x0CD30BD0 | |
1364 | ||
5dd16419 JX |
1365 | #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK) |
1366 | #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK) | |
d7e09d03 PT |
1367 | |
1368 | #define lov_ost_data lov_ost_data_v1 | |
1369 | struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/ | |
1370 | struct ost_id l_ost_oi; /* OST object ID */ | |
1371 | __u32 l_ost_gen; /* generation of this l_ost_idx */ | |
1372 | __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */ | |
1373 | }; | |
1374 | ||
1375 | #define lov_mds_md lov_mds_md_v1 | |
1376 | struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */ | |
1377 | __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */ | |
1378 | __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */ | |
1379 | struct ost_id lmm_oi; /* LOV object ID */ | |
1380 | __u32 lmm_stripe_size; /* size of stripe in bytes */ | |
1381 | /* lmm_stripe_count used to be __u32 */ | |
1382 | __u16 lmm_stripe_count; /* num stripes in use for this object */ | |
1383 | __u16 lmm_layout_gen; /* layout generation number */ | |
1384 | struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */ | |
1385 | }; | |
1386 | ||
1387 | /** | |
1388 | * Sigh, because pre-2.4 uses | |
1389 | * struct lov_mds_md_v1 { | |
1390 | * ........ | |
1391 | * __u64 lmm_object_id; | |
1392 | * __u64 lmm_object_seq; | |
1393 | * ...... | |
1394 | * } | |
1395 | * to identify the LOV(MDT) object, and lmm_object_seq will | |
1396 | * be normal_fid, which make it hard to combine these conversion | |
1397 | * to ostid_to FID. so we will do lmm_oi/fid conversion separately | |
1398 | * | |
1399 | * We can tell the lmm_oi by this way, | |
1400 | * 1.8: lmm_object_id = {inode}, lmm_object_gr = 0 | |
1401 | * 2.1: lmm_object_id = {oid < 128k}, lmm_object_seq = FID_SEQ_NORMAL | |
1402 | * 2.4: lmm_oi.f_seq = FID_SEQ_NORMAL, lmm_oi.f_oid = {oid < 128k}, | |
1403 | * lmm_oi.f_ver = 0 | |
1404 | * | |
1405 | * But currently lmm_oi/lsm_oi does not have any "real" usages, | |
1406 | * except for printing some information, and the user can always | |
1407 | * get the real FID from LMA, besides this multiple case check might | |
1408 | * make swab more complicate. So we will keep using id/seq for lmm_oi. | |
1409 | */ | |
1410 | ||
1411 | static inline void fid_to_lmm_oi(const struct lu_fid *fid, | |
1412 | struct ost_id *oi) | |
1413 | { | |
1414 | oi->oi.oi_id = fid_oid(fid); | |
1415 | oi->oi.oi_seq = fid_seq(fid); | |
1416 | } | |
1417 | ||
1418 | static inline void lmm_oi_set_seq(struct ost_id *oi, __u64 seq) | |
1419 | { | |
1420 | oi->oi.oi_seq = seq; | |
1421 | } | |
1422 | ||
22144626 FY |
1423 | static inline void lmm_oi_set_id(struct ost_id *oi, __u64 oid) |
1424 | { | |
1425 | oi->oi.oi_id = oid; | |
1426 | } | |
1427 | ||
ac8f0a5c | 1428 | static inline __u64 lmm_oi_id(const struct ost_id *oi) |
d7e09d03 PT |
1429 | { |
1430 | return oi->oi.oi_id; | |
1431 | } | |
1432 | ||
ac8f0a5c | 1433 | static inline __u64 lmm_oi_seq(const struct ost_id *oi) |
d7e09d03 PT |
1434 | { |
1435 | return oi->oi.oi_seq; | |
1436 | } | |
1437 | ||
1438 | static inline void lmm_oi_le_to_cpu(struct ost_id *dst_oi, | |
ac8f0a5c | 1439 | const struct ost_id *src_oi) |
d7e09d03 PT |
1440 | { |
1441 | dst_oi->oi.oi_id = le64_to_cpu(src_oi->oi.oi_id); | |
1442 | dst_oi->oi.oi_seq = le64_to_cpu(src_oi->oi.oi_seq); | |
1443 | } | |
1444 | ||
1445 | static inline void lmm_oi_cpu_to_le(struct ost_id *dst_oi, | |
ac8f0a5c | 1446 | const struct ost_id *src_oi) |
d7e09d03 PT |
1447 | { |
1448 | dst_oi->oi.oi_id = cpu_to_le64(src_oi->oi.oi_id); | |
1449 | dst_oi->oi.oi_seq = cpu_to_le64(src_oi->oi.oi_seq); | |
1450 | } | |
1451 | ||
ec83e611 JP |
1452 | #define MAX_MD_SIZE \ |
1453 | (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data)) | |
1454 | #define MIN_MD_SIZE \ | |
1455 | (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data)) | |
d7e09d03 PT |
1456 | |
1457 | #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access" | |
1458 | #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default" | |
1459 | #define XATTR_USER_PREFIX "user." | |
1460 | #define XATTR_TRUSTED_PREFIX "trusted." | |
1461 | #define XATTR_SECURITY_PREFIX "security." | |
1462 | #define XATTR_LUSTRE_PREFIX "lustre." | |
1463 | ||
1464 | #define XATTR_NAME_LOV "trusted.lov" | |
1465 | #define XATTR_NAME_LMA "trusted.lma" | |
1466 | #define XATTR_NAME_LMV "trusted.lmv" | |
2de35386 | 1467 | #define XATTR_NAME_DEFAULT_LMV "trusted.dmv" |
d7e09d03 PT |
1468 | #define XATTR_NAME_LINK "trusted.link" |
1469 | #define XATTR_NAME_FID "trusted.fid" | |
1470 | #define XATTR_NAME_VERSION "trusted.version" | |
1471 | #define XATTR_NAME_SOM "trusted.som" | |
1472 | #define XATTR_NAME_HSM "trusted.hsm" | |
1473 | #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_namespace" | |
1474 | ||
1475 | struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */ | |
1476 | __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */ | |
1477 | __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */ | |
1478 | struct ost_id lmm_oi; /* LOV object ID */ | |
1479 | __u32 lmm_stripe_size; /* size of stripe in bytes */ | |
1480 | /* lmm_stripe_count used to be __u32 */ | |
1481 | __u16 lmm_stripe_count; /* num stripes in use for this object */ | |
1482 | __u16 lmm_layout_gen; /* layout generation number */ | |
aaf06e29 | 1483 | char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */ |
d7e09d03 PT |
1484 | struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */ |
1485 | }; | |
1486 | ||
bc06a678 | 1487 | static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic) |
1488 | { | |
1489 | if (lmm_magic == LOV_MAGIC_V3) | |
1490 | return sizeof(struct lov_mds_md_v3) + | |
1491 | stripes * sizeof(struct lov_ost_data_v1); | |
1492 | else | |
1493 | return sizeof(struct lov_mds_md_v1) + | |
1494 | stripes * sizeof(struct lov_ost_data_v1); | |
1495 | } | |
1496 | ||
081b7265 JH |
1497 | static inline __u32 |
1498 | lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic) | |
1499 | { | |
1500 | switch (lmm_magic) { | |
1501 | case LOV_MAGIC_V1: { | |
1502 | struct lov_mds_md_v1 lmm; | |
1503 | ||
1504 | if (buf_size < sizeof(lmm)) | |
1505 | return 0; | |
1506 | ||
1507 | return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]); | |
1508 | } | |
1509 | case LOV_MAGIC_V3: { | |
1510 | struct lov_mds_md_v3 lmm; | |
1511 | ||
1512 | if (buf_size < sizeof(lmm)) | |
1513 | return 0; | |
1514 | ||
1515 | return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]); | |
1516 | } | |
1517 | default: | |
1518 | return 0; | |
1519 | } | |
1520 | } | |
bc06a678 | 1521 | |
a1e616b0 | 1522 | #define OBD_MD_FLID (0x00000001ULL) /* object ID */ |
d7e09d03 PT |
1523 | #define OBD_MD_FLATIME (0x00000002ULL) /* access time */ |
1524 | #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */ | |
1525 | #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */ | |
1526 | #define OBD_MD_FLSIZE (0x00000010ULL) /* size */ | |
1527 | #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */ | |
1528 | #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */ | |
1529 | #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */ | |
1530 | #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */ | |
1531 | #define OBD_MD_FLUID (0x00000200ULL) /* user ID */ | |
1532 | #define OBD_MD_FLGID (0x00000400ULL) /* group ID */ | |
1533 | #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */ | |
1534 | #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */ | |
1535 | #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */ | |
1536 | /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */ | |
1537 | #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */ | |
1538 | #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */ | |
1539 | #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */ | |
1540 | #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */ | |
1541 | #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */ | |
1542 | #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */ | |
1543 | /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */ | |
0a4bea92 | 1544 | /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */ |
d7e09d03 PT |
1545 | #define OBD_MD_FLGROUP (0x01000000ULL) /* group */ |
1546 | #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */ | |
1547 | #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */ | |
a1e616b0 OD |
1548 | /* ->mds if epoch opens or closes |
1549 | */ | |
d7e09d03 PT |
1550 | #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */ |
1551 | #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */ | |
1552 | #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */ | |
1553 | #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */ | |
1554 | #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */ | |
1555 | ||
a1e616b0 | 1556 | #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */ |
d7e09d03 | 1557 | #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */ |
a1e616b0 | 1558 | #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */ |
5ea17d6c | 1559 | #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */ |
d7e09d03 PT |
1560 | |
1561 | #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */ | |
1562 | #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */ | |
1563 | #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */ | |
a1e616b0 | 1564 | #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */ |
341f1f0a | 1565 | /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */ |
d7e09d03 PT |
1566 | #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */ |
1567 | #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */ | |
1568 | #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */ | |
1569 | #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */ | |
1570 | #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes | |
7fc1f831 AP |
1571 | * under lock; for xattr |
1572 | * requests means the | |
a1e616b0 OD |
1573 | * client holds the lock |
1574 | */ | |
d7e09d03 PT |
1575 | #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */ |
1576 | ||
341f1f0a FY |
1577 | /* OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) lfs lsetfacl, obsolete */ |
1578 | /* OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) lfs lgetfacl, obsolete */ | |
1579 | /* OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) lfs rsetfacl, obsolete */ | |
1580 | /* OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) lfs rgetfacl, obsolete */ | |
d7e09d03 PT |
1581 | |
1582 | #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */ | |
0ffaa9c8 HD |
1583 | #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent |
1584 | * executed | |
1585 | */ | |
d7e09d03 | 1586 | |
6e23ea98 | 1587 | #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */ |
1588 | ||
d7e09d03 PT |
1589 | #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \ |
1590 | OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \ | |
1591 | OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \ | |
1592 | OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \ | |
1593 | OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP) | |
1594 | ||
7fc1f831 AP |
1595 | #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS) |
1596 | ||
d7e09d03 | 1597 | /* don't forget obdo_fid which is way down at the bottom so it can |
a1e616b0 OD |
1598 | * come after the definition of llog_cookie |
1599 | */ | |
d7e09d03 PT |
1600 | |
1601 | enum hss_valid { | |
1602 | HSS_SETMASK = 0x01, | |
1603 | HSS_CLEARMASK = 0x02, | |
1604 | HSS_ARCHIVE_ID = 0x04, | |
1605 | }; | |
1606 | ||
1607 | struct hsm_state_set { | |
1608 | __u32 hss_valid; | |
1609 | __u32 hss_archive_id; | |
1610 | __u64 hss_setmask; | |
1611 | __u64 hss_clearmask; | |
1612 | }; | |
1613 | ||
d7e09d03 PT |
1614 | /* ost_body.data values for OST_BRW */ |
1615 | ||
a1e616b0 OD |
1616 | #define OBD_BRW_READ 0x01 |
1617 | #define OBD_BRW_WRITE 0x02 | |
1618 | #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE) | |
1619 | #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous | |
d7e09d03 | 1620 | * transfer and is not accounted in |
a1e616b0 OD |
1621 | * the grant. |
1622 | */ | |
1623 | #define OBD_BRW_CHECK 0x10 | |
d7e09d03 | 1624 | #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */ |
a1e616b0 OD |
1625 | #define OBD_BRW_GRANTED 0x40 /* the ost manages this */ |
1626 | #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */ | |
1627 | #define OBD_BRW_NOQUOTA 0x100 | |
1628 | #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */ | |
1629 | #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */ | |
d7e09d03 PT |
1630 | #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */ |
1631 | #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */ | |
1632 | #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */ | |
ad479287 PS |
1633 | #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server |
1634 | * that the client is running low on | |
1635 | * space for unstable pages; asking | |
1636 | * it to sync quickly | |
1637 | */ | |
d7e09d03 | 1638 | |
00c0a6ae | 1639 | #define OBD_OBJECT_EOF LUSTRE_EOF |
d7e09d03 PT |
1640 | |
1641 | #define OST_MIN_PRECREATE 32 | |
1642 | #define OST_MAX_PRECREATE 20000 | |
1643 | ||
1644 | struct obd_ioobj { | |
1645 | struct ost_id ioo_oid; /* object ID, if multi-obj BRW */ | |
1646 | __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4, | |
1647 | * now (PTLRPC_BULK_OPS_COUNT - 1) in | |
a1e616b0 OD |
1648 | * high 16 bits in 2.4 and later |
1649 | */ | |
d7e09d03 PT |
1650 | __u32 ioo_bufcnt; /* number of niobufs for this object */ |
1651 | }; | |
1652 | ||
5965de81 JX |
1653 | /* |
1654 | * NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in | |
1655 | * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS. | |
1656 | * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. | |
1657 | */ | |
d7e09d03 | 1658 | #define IOOBJ_MAX_BRW_BITS 16 |
d7e09d03 PT |
1659 | #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1) |
1660 | #define ioobj_max_brw_set(ioo, num) \ | |
1661 | do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0) | |
1662 | ||
d7e09d03 PT |
1663 | /* multiple of 8 bytes => can array */ |
1664 | struct niobuf_remote { | |
638814f6 JH |
1665 | __u64 rnb_offset; |
1666 | __u32 rnb_len; | |
1667 | __u32 rnb_flags; | |
d7e09d03 PT |
1668 | }; |
1669 | ||
d7e09d03 PT |
1670 | /* lock value block communicated between the filter and llite */ |
1671 | ||
1672 | /* OST_LVB_ERR_INIT is needed because the return code in rc is | |
a1e616b0 OD |
1673 | * negative, i.e. because ((MASK + rc) & MASK) != MASK. |
1674 | */ | |
d7e09d03 PT |
1675 | #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL |
1676 | #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL | |
1677 | #define OST_LVB_IS_ERR(blocks) \ | |
1678 | ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK) | |
1679 | #define OST_LVB_SET_ERR(blocks, rc) \ | |
1680 | do { blocks = OST_LVB_ERR_INIT + rc; } while (0) | |
1681 | #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT) | |
1682 | ||
1683 | struct ost_lvb_v1 { | |
1684 | __u64 lvb_size; | |
21aef7d9 OD |
1685 | __s64 lvb_mtime; |
1686 | __s64 lvb_atime; | |
1687 | __s64 lvb_ctime; | |
d7e09d03 PT |
1688 | __u64 lvb_blocks; |
1689 | }; | |
1690 | ||
d7e09d03 PT |
1691 | struct ost_lvb { |
1692 | __u64 lvb_size; | |
21aef7d9 OD |
1693 | __s64 lvb_mtime; |
1694 | __s64 lvb_atime; | |
1695 | __s64 lvb_ctime; | |
d7e09d03 PT |
1696 | __u64 lvb_blocks; |
1697 | __u32 lvb_mtime_ns; | |
1698 | __u32 lvb_atime_ns; | |
1699 | __u32 lvb_ctime_ns; | |
1700 | __u32 lvb_padding; | |
1701 | }; | |
1702 | ||
d7e09d03 PT |
1703 | /* |
1704 | * lquota data structures | |
1705 | */ | |
1706 | ||
03440c4e | 1707 | /* The lquota_id structure is a union of all the possible identifier types that |
d7e09d03 PT |
1708 | * can be used with quota, this includes: |
1709 | * - 64-bit user ID | |
1710 | * - 64-bit group ID | |
a1e616b0 OD |
1711 | * - a FID which can be used for per-directory quota in the future |
1712 | */ | |
d7e09d03 PT |
1713 | union lquota_id { |
1714 | struct lu_fid qid_fid; /* FID for per-directory quota */ | |
1715 | __u64 qid_uid; /* user identifier */ | |
1716 | __u64 qid_gid; /* group identifier */ | |
1717 | }; | |
1718 | ||
1719 | /* quotactl management */ | |
1720 | struct obd_quotactl { | |
1721 | __u32 qc_cmd; | |
1722 | __u32 qc_type; /* see Q_* flag below */ | |
1723 | __u32 qc_id; | |
1724 | __u32 qc_stat; | |
1725 | struct obd_dqinfo qc_dqinfo; | |
1726 | struct obd_dqblk qc_dqblk; | |
1727 | }; | |
1728 | ||
d7e09d03 PT |
1729 | #define Q_COPY(out, in, member) (out)->member = (in)->member |
1730 | ||
1731 | #define QCTL_COPY(out, in) \ | |
1732 | do { \ | |
1733 | Q_COPY(out, in, qc_cmd); \ | |
1734 | Q_COPY(out, in, qc_type); \ | |
1735 | Q_COPY(out, in, qc_id); \ | |
1736 | Q_COPY(out, in, qc_stat); \ | |
1737 | Q_COPY(out, in, qc_dqinfo); \ | |
1738 | Q_COPY(out, in, qc_dqblk); \ | |
1739 | } while (0) | |
1740 | ||
d7e09d03 PT |
1741 | /* Data structures associated with the quota locks */ |
1742 | ||
1743 | /* Glimpse descriptor used for the index & per-ID quota locks */ | |
1744 | struct ldlm_gl_lquota_desc { | |
1745 | union lquota_id gl_id; /* quota ID subject to the glimpse */ | |
1746 | __u64 gl_flags; /* see LQUOTA_FL* below */ | |
1747 | __u64 gl_ver; /* new index version */ | |
1748 | __u64 gl_hardlimit; /* new hardlimit or qunit value */ | |
1749 | __u64 gl_softlimit; /* new softlimit */ | |
1750 | __u64 gl_time; | |
1751 | __u64 gl_pad2; | |
1752 | }; | |
c9f6bb96 | 1753 | |
d7e09d03 PT |
1754 | /* quota glimpse flags */ |
1755 | #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */ | |
1756 | ||
1757 | /* LVB used with quota (global and per-ID) locks */ | |
1758 | struct lquota_lvb { | |
1759 | __u64 lvb_flags; /* see LQUOTA_FL* above */ | |
1760 | __u64 lvb_id_may_rel; /* space that might be released later */ | |
1761 | __u64 lvb_id_rel; /* space released by the slave for this ID */ | |
1762 | __u64 lvb_id_qunit; /* current qunit value */ | |
1763 | __u64 lvb_pad1; | |
1764 | }; | |
1765 | ||
d7e09d03 | 1766 | /* op codes */ |
115b4f9c | 1767 | enum quota_cmd { |
d7e09d03 PT |
1768 | QUOTA_DQACQ = 601, |
1769 | QUOTA_DQREL = 602, | |
1770 | QUOTA_LAST_OPC | |
115b4f9c | 1771 | }; |
d7e09d03 PT |
1772 | #define QUOTA_FIRST_OPC QUOTA_DQACQ |
1773 | ||
1774 | /* | |
1775 | * MDS REQ RECORDS | |
1776 | */ | |
1777 | ||
1778 | /* opcodes */ | |
303e4002 | 1779 | enum mds_cmd { |
d7e09d03 PT |
1780 | MDS_GETATTR = 33, |
1781 | MDS_GETATTR_NAME = 34, | |
1782 | MDS_CLOSE = 35, | |
1783 | MDS_REINT = 36, | |
1784 | MDS_READPAGE = 37, | |
1785 | MDS_CONNECT = 38, | |
1786 | MDS_DISCONNECT = 39, | |
1787 | MDS_GETSTATUS = 40, | |
1788 | MDS_STATFS = 41, | |
58c78cd2 JH |
1789 | MDS_PIN = 42, /* obsolete, never used in a release */ |
1790 | MDS_UNPIN = 43, /* obsolete, never used in a release */ | |
d7e09d03 | 1791 | MDS_SYNC = 44, |
a823acf5 | 1792 | MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */ |
d7e09d03 | 1793 | MDS_SET_INFO = 46, |
e57721e7 | 1794 | MDS_QUOTACHECK = 47, /* not used since 2.4 */ |
d7e09d03 PT |
1795 | MDS_QUOTACTL = 48, |
1796 | MDS_GETXATTR = 49, | |
1797 | MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */ | |
1798 | MDS_WRITEPAGE = 51, | |
58c78cd2 | 1799 | MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */ |
d7e09d03 PT |
1800 | MDS_GET_INFO = 53, |
1801 | MDS_HSM_STATE_GET = 54, | |
1802 | MDS_HSM_STATE_SET = 55, | |
1803 | MDS_HSM_ACTION = 56, | |
1804 | MDS_HSM_PROGRESS = 57, | |
1805 | MDS_HSM_REQUEST = 58, | |
1806 | MDS_HSM_CT_REGISTER = 59, | |
1807 | MDS_HSM_CT_UNREGISTER = 60, | |
1808 | MDS_SWAP_LAYOUTS = 61, | |
1809 | MDS_LAST_OPC | |
303e4002 | 1810 | }; |
d7e09d03 PT |
1811 | |
1812 | #define MDS_FIRST_OPC MDS_GETATTR | |
1813 | ||
d7e09d03 PT |
1814 | /* |
1815 | * Do not exceed 63 | |
1816 | */ | |
1817 | ||
07e2eb39 | 1818 | enum mdt_reint_cmd { |
d7e09d03 PT |
1819 | REINT_SETATTR = 1, |
1820 | REINT_CREATE = 2, | |
1821 | REINT_LINK = 3, | |
1822 | REINT_UNLINK = 4, | |
1823 | REINT_RENAME = 5, | |
1824 | REINT_OPEN = 6, | |
1825 | REINT_SETXATTR = 7, | |
1826 | REINT_RMENTRY = 8, | |
79496845 | 1827 | REINT_MIGRATE = 9, |
d7e09d03 | 1828 | REINT_MAX |
07e2eb39 | 1829 | }; |
d7e09d03 | 1830 | |
d7e09d03 PT |
1831 | /* the disposition of the intent outlines what was executed */ |
1832 | #define DISP_IT_EXECD 0x00000001 | |
1833 | #define DISP_LOOKUP_EXECD 0x00000002 | |
1834 | #define DISP_LOOKUP_NEG 0x00000004 | |
1835 | #define DISP_LOOKUP_POS 0x00000008 | |
1836 | #define DISP_OPEN_CREATE 0x00000010 | |
1837 | #define DISP_OPEN_OPEN 0x00000020 | |
f236f69b | 1838 | #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */ |
d7e09d03 PT |
1839 | #define DISP_ENQ_OPEN_REF 0x00800000 |
1840 | #define DISP_ENQ_CREATE_REF 0x01000000 | |
1841 | #define DISP_OPEN_LOCK 0x02000000 | |
d3a8a4e2 | 1842 | #define DISP_OPEN_LEASE 0x04000000 |
63d42578 | 1843 | #define DISP_OPEN_STRIPE 0x08000000 |
864d6a25 | 1844 | #define DISP_OPEN_DENY 0x10000000 |
d7e09d03 PT |
1845 | |
1846 | /* INODE LOCK PARTS */ | |
fe4c58af | 1847 | #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also |
1848 | * was used to protect permission (mode, | |
a1e616b0 OD |
1849 | * owner, group etc) before 2.4. |
1850 | */ | |
fe4c58af | 1851 | #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */ |
1852 | #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */ | |
1853 | #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */ | |
1854 | ||
1855 | /* The PERM bit is added int 2.4, and it is used to protect permission(mode, | |
1856 | * owner, group, acl etc), so to separate the permission from LOOKUP lock. | |
1857 | * Because for remote directories(in DNE), these locks will be granted by | |
1858 | * different MDTs(different ldlm namespace). | |
1859 | * | |
1860 | * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together. | |
1861 | * For Remote directory, the master MDT, where the remote directory is, will | |
1862 | * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is, | |
a1e616b0 OD |
1863 | * will grant LOOKUP_LOCK. |
1864 | */ | |
fe4c58af | 1865 | #define MDS_INODELOCK_PERM 0x000010 |
1866 | #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */ | |
d7e09d03 | 1867 | |
7fc1f831 | 1868 | #define MDS_INODELOCK_MAXSHIFT 5 |
d7e09d03 | 1869 | /* This FULL lock is useful to take on unlink sort of operations */ |
cd94f231 | 1870 | #define MDS_INODELOCK_FULL ((1 << (MDS_INODELOCK_MAXSHIFT + 1)) - 1) |
d7e09d03 | 1871 | |
d7e09d03 PT |
1872 | /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2], |
1873 | * but was moved into name[1] along with the OID to avoid consuming the | |
a1e616b0 OD |
1874 | * name[2,3] fields that need to be used for the quota id (also a FID). |
1875 | */ | |
d7e09d03 PT |
1876 | enum { |
1877 | LUSTRE_RES_ID_SEQ_OFF = 0, | |
1878 | LUSTRE_RES_ID_VER_OID_OFF = 1, | |
1879 | LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */ | |
1880 | LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2, | |
1881 | LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3, | |
1882 | LUSTRE_RES_ID_HSH_OFF = 3 | |
1883 | }; | |
1884 | ||
1885 | #define MDS_STATUS_CONN 1 | |
1886 | #define MDS_STATUS_LOV 2 | |
1887 | ||
d7e09d03 | 1888 | /* these should be identical to their EXT4_*_FL counterparts, they are |
a1e616b0 OD |
1889 | * redefined here only to avoid dragging in fs/ext4/ext4.h |
1890 | */ | |
d7e09d03 PT |
1891 | #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */ |
1892 | #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */ | |
1893 | #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */ | |
c90cfdba | 1894 | #define LUSTRE_NODUMP_FL 0x00000040 /* do not dump file */ |
d7e09d03 | 1895 | #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */ |
c90cfdba | 1896 | #define LUSTRE_INDEX_FL 0x00001000 /* hash-indexed directory */ |
d7e09d03 | 1897 | #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */ |
c90cfdba AD |
1898 | #define LUSTRE_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ |
1899 | #define LUSTRE_DIRECTIO_FL 0x00100000 /* Use direct i/o */ | |
1900 | #define LUSTRE_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */ | |
d7e09d03 PT |
1901 | |
1902 | /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values | |
1903 | * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire | |
1904 | * protocol equivalents of LDISKFS_*_FL values stored on disk, while | |
1905 | * the S_* flags are kernel-internal values that change between kernel | |
1906 | * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS. | |
a1e616b0 OD |
1907 | * See b=16526 for a full history. |
1908 | */ | |
d7e09d03 PT |
1909 | static inline int ll_ext_to_inode_flags(int flags) |
1910 | { | |
1911 | return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) | | |
1912 | ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) | | |
1913 | ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) | | |
d7e09d03 | 1914 | ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) | |
d7e09d03 PT |
1915 | ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0)); |
1916 | } | |
1917 | ||
1918 | static inline int ll_inode_to_ext_flags(int iflags) | |
1919 | { | |
1920 | return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) | | |
1921 | ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) | | |
1922 | ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) | | |
d7e09d03 | 1923 | ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) | |
d7e09d03 PT |
1924 | ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0)); |
1925 | } | |
1926 | ||
5ea17d6c JL |
1927 | /* 64 possible states */ |
1928 | enum md_transient_state { | |
1929 | MS_RESTORE = (1 << 0), /* restore is running */ | |
1930 | }; | |
1931 | ||
d7e09d03 | 1932 | struct mdt_body { |
2e1b5b8b JH |
1933 | struct lu_fid mbo_fid1; |
1934 | struct lu_fid mbo_fid2; | |
1935 | struct lustre_handle mbo_handle; | |
1936 | __u64 mbo_valid; | |
1937 | __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */ | |
1938 | __s64 mbo_mtime; | |
1939 | __s64 mbo_atime; | |
1940 | __s64 mbo_ctime; | |
1941 | __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */ | |
1942 | __u64 mbo_ioepoch; | |
1943 | __u64 mbo_t_state; /* transient file state defined in | |
1944 | * enum md_transient_state | |
1945 | * was "ino" until 2.4.0 | |
1946 | */ | |
1947 | __u32 mbo_fsuid; | |
1948 | __u32 mbo_fsgid; | |
1949 | __u32 mbo_capability; | |
1950 | __u32 mbo_mode; | |
1951 | __u32 mbo_uid; | |
1952 | __u32 mbo_gid; | |
c90cfdba | 1953 | __u32 mbo_flags; /* LUSTRE_*_FL file attributes */ |
2e1b5b8b JH |
1954 | __u32 mbo_rdev; |
1955 | __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */ | |
1956 | __u32 mbo_unused2; /* was "generation" until 2.4.0 */ | |
1957 | __u32 mbo_suppgid; | |
1958 | __u32 mbo_eadatasize; | |
1959 | __u32 mbo_aclsize; | |
1960 | __u32 mbo_max_mdsize; | |
0a4bea92 | 1961 | __u32 mbo_unused3; /* was max_cookiesize until 2.8 */ |
2e1b5b8b JH |
1962 | __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */ |
1963 | __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */ | |
1964 | __u32 mbo_padding_5; /* also fix lustre_swab_mdt_body */ | |
1965 | __u64 mbo_padding_6; | |
1966 | __u64 mbo_padding_7; | |
1967 | __u64 mbo_padding_8; | |
1968 | __u64 mbo_padding_9; | |
1969 | __u64 mbo_padding_10; | |
d7e09d03 PT |
1970 | }; /* 216 */ |
1971 | ||
d7e09d03 | 1972 | struct mdt_ioepoch { |
a823acf5 JH |
1973 | struct lustre_handle mio_handle; |
1974 | __u64 mio_unused1; /* was ioepoch */ | |
1975 | __u32 mio_unused2; /* was flags */ | |
1976 | __u32 mio_padding; | |
d7e09d03 PT |
1977 | }; |
1978 | ||
d7e09d03 PT |
1979 | /* permissions for md_perm.mp_perm */ |
1980 | enum { | |
1981 | CFS_SETUID_PERM = 0x01, | |
1982 | CFS_SETGID_PERM = 0x02, | |
1983 | CFS_SETGRP_PERM = 0x04, | |
d7e09d03 PT |
1984 | }; |
1985 | ||
d7e09d03 PT |
1986 | struct mdt_rec_setattr { |
1987 | __u32 sa_opcode; | |
1988 | __u32 sa_cap; | |
1989 | __u32 sa_fsuid; | |
1990 | __u32 sa_fsuid_h; | |
1991 | __u32 sa_fsgid; | |
1992 | __u32 sa_fsgid_h; | |
1993 | __u32 sa_suppgid; | |
1994 | __u32 sa_suppgid_h; | |
1995 | __u32 sa_padding_1; | |
1996 | __u32 sa_padding_1_h; | |
1997 | struct lu_fid sa_fid; | |
1998 | __u64 sa_valid; | |
1999 | __u32 sa_uid; | |
2000 | __u32 sa_gid; | |
2001 | __u64 sa_size; | |
2002 | __u64 sa_blocks; | |
21aef7d9 OD |
2003 | __s64 sa_mtime; |
2004 | __s64 sa_atime; | |
2005 | __s64 sa_ctime; | |
d7e09d03 PT |
2006 | __u32 sa_attr_flags; |
2007 | __u32 sa_mode; | |
2008 | __u32 sa_bias; /* some operation flags */ | |
2009 | __u32 sa_padding_3; | |
2010 | __u32 sa_padding_4; | |
2011 | __u32 sa_padding_5; | |
2012 | }; | |
2013 | ||
d7e09d03 PT |
2014 | /* |
2015 | * Attribute flags used in mdt_rec_setattr::sa_valid. | |
2016 | * The kernel's #defines for ATTR_* should not be used over the network | |
2017 | * since the client and MDS may run different kernels (see bug 13828) | |
2018 | * Therefore, we should only use MDS_ATTR_* attributes for sa_valid. | |
2019 | */ | |
a1e616b0 OD |
2020 | #define MDS_ATTR_MODE 0x1ULL /* = 1 */ |
2021 | #define MDS_ATTR_UID 0x2ULL /* = 2 */ | |
2022 | #define MDS_ATTR_GID 0x4ULL /* = 4 */ | |
2023 | #define MDS_ATTR_SIZE 0x8ULL /* = 8 */ | |
2024 | #define MDS_ATTR_ATIME 0x10ULL /* = 16 */ | |
2025 | #define MDS_ATTR_MTIME 0x20ULL /* = 32 */ | |
2026 | #define MDS_ATTR_CTIME 0x40ULL /* = 64 */ | |
d7e09d03 PT |
2027 | #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */ |
2028 | #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */ | |
2029 | #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */ | |
2030 | #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */ | |
2031 | #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */ | |
2032 | #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */ | |
2033 | #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */ | |
a1e616b0 OD |
2034 | #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, |
2035 | * ie O_TRUNC | |
2036 | */ | |
d7e09d03 PT |
2037 | #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */ |
2038 | ||
d7e09d03 PT |
2039 | #define MDS_FMODE_CLOSED 00000000 |
2040 | #define MDS_FMODE_EXEC 00000004 | |
a823acf5 JH |
2041 | /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */ |
2042 | /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */ | |
2043 | /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */ | |
d7e09d03 PT |
2044 | |
2045 | #define MDS_OPEN_CREATED 00000010 | |
2046 | #define MDS_OPEN_CROSS 00000020 | |
2047 | ||
2048 | #define MDS_OPEN_CREAT 00000100 | |
2049 | #define MDS_OPEN_EXCL 00000200 | |
2050 | #define MDS_OPEN_TRUNC 00001000 | |
2051 | #define MDS_OPEN_APPEND 00002000 | |
2052 | #define MDS_OPEN_SYNC 00010000 | |
2053 | #define MDS_OPEN_DIRECTORY 00200000 | |
2054 | ||
2055 | #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */ | |
2056 | #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */ | |
2057 | #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */ | |
2058 | #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file. | |
2059 | * We do not support JOIN FILE | |
2060 | * anymore, reserve this flags | |
2061 | * just for preventing such bit | |
a1e616b0 OD |
2062 | * to be reused. |
2063 | */ | |
d7e09d03 | 2064 | |
a1e616b0 | 2065 | #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */ |
d7e09d03 PT |
2066 | #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */ |
2067 | #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */ | |
2068 | #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */ | |
2069 | #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or | |
2070 | * hsm restore) */ | |
2071 | #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created | |
2072 | unlinked */ | |
d3a8a4e2 JX |
2073 | #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease |
2074 | * delegation, succeed if it's not | |
2075 | * being opened with conflict mode. | |
2076 | */ | |
48d23e61 | 2077 | #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */ |
d7e09d03 | 2078 | |
c1b66fcc LS |
2079 | #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \ |
2080 | MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \ | |
2081 | MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \ | |
2082 | MDS_OPEN_RELEASE) | |
2083 | ||
48d23e61 | 2084 | enum mds_op_bias { |
d7e09d03 PT |
2085 | MDS_CHECK_SPLIT = 1 << 0, |
2086 | MDS_CROSS_REF = 1 << 1, | |
2087 | MDS_VTX_BYPASS = 1 << 2, | |
2088 | MDS_PERM_BYPASS = 1 << 3, | |
a823acf5 | 2089 | /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */ |
d7e09d03 PT |
2090 | MDS_QUOTA_IGNORE = 1 << 5, |
2091 | MDS_CLOSE_CLEANUP = 1 << 6, | |
2092 | MDS_KEEP_ORPHAN = 1 << 7, | |
2093 | MDS_RECOV_OPEN = 1 << 8, | |
2094 | MDS_DATA_MODIFIED = 1 << 9, | |
2095 | MDS_CREATE_VOLATILE = 1 << 10, | |
2096 | MDS_OWNEROVERRIDE = 1 << 11, | |
48d23e61 | 2097 | MDS_HSM_RELEASE = 1 << 12, |
79496845 | 2098 | MDS_RENAME_MIGRATE = BIT(13), |
0ffaa9c8 | 2099 | MDS_CLOSE_LAYOUT_SWAP = BIT(14), |
d7e09d03 PT |
2100 | }; |
2101 | ||
2102 | /* instance of mdt_reint_rec */ | |
2103 | struct mdt_rec_create { | |
2104 | __u32 cr_opcode; | |
2105 | __u32 cr_cap; | |
2106 | __u32 cr_fsuid; | |
2107 | __u32 cr_fsuid_h; | |
2108 | __u32 cr_fsgid; | |
2109 | __u32 cr_fsgid_h; | |
2110 | __u32 cr_suppgid1; | |
2111 | __u32 cr_suppgid1_h; | |
2112 | __u32 cr_suppgid2; | |
2113 | __u32 cr_suppgid2_h; | |
2114 | struct lu_fid cr_fid1; | |
2115 | struct lu_fid cr_fid2; | |
2116 | struct lustre_handle cr_old_handle; /* handle in case of open replay */ | |
21aef7d9 | 2117 | __s64 cr_time; |
d7e09d03 PT |
2118 | __u64 cr_rdev; |
2119 | __u64 cr_ioepoch; | |
2120 | __u64 cr_padding_1; /* rr_blocks */ | |
2121 | __u32 cr_mode; | |
2122 | __u32 cr_bias; | |
2123 | /* use of helpers set/get_mrc_cr_flags() is needed to access | |
2124 | * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to | |
a1e616b0 OD |
2125 | * extend cr_flags size without breaking 1.8 compat |
2126 | */ | |
d7e09d03 PT |
2127 | __u32 cr_flags_l; /* for use with open, low 32 bits */ |
2128 | __u32 cr_flags_h; /* for use with open, high 32 bits */ | |
2129 | __u32 cr_umask; /* umask for create */ | |
2130 | __u32 cr_padding_4; /* rr_padding_4 */ | |
2131 | }; | |
2132 | ||
2133 | static inline void set_mrc_cr_flags(struct mdt_rec_create *mrc, __u64 flags) | |
2134 | { | |
2135 | mrc->cr_flags_l = (__u32)(flags & 0xFFFFFFFFUll); | |
2136 | mrc->cr_flags_h = (__u32)(flags >> 32); | |
2137 | } | |
2138 | ||
2139 | static inline __u64 get_mrc_cr_flags(struct mdt_rec_create *mrc) | |
2140 | { | |
2141 | return ((__u64)(mrc->cr_flags_l) | ((__u64)mrc->cr_flags_h << 32)); | |
2142 | } | |
2143 | ||
2144 | /* instance of mdt_reint_rec */ | |
2145 | struct mdt_rec_link { | |
2146 | __u32 lk_opcode; | |
2147 | __u32 lk_cap; | |
2148 | __u32 lk_fsuid; | |
2149 | __u32 lk_fsuid_h; | |
2150 | __u32 lk_fsgid; | |
2151 | __u32 lk_fsgid_h; | |
2152 | __u32 lk_suppgid1; | |
2153 | __u32 lk_suppgid1_h; | |
2154 | __u32 lk_suppgid2; | |
2155 | __u32 lk_suppgid2_h; | |
2156 | struct lu_fid lk_fid1; | |
2157 | struct lu_fid lk_fid2; | |
21aef7d9 | 2158 | __s64 lk_time; |
d7e09d03 PT |
2159 | __u64 lk_padding_1; /* rr_atime */ |
2160 | __u64 lk_padding_2; /* rr_ctime */ | |
2161 | __u64 lk_padding_3; /* rr_size */ | |
2162 | __u64 lk_padding_4; /* rr_blocks */ | |
2163 | __u32 lk_bias; | |
2164 | __u32 lk_padding_5; /* rr_mode */ | |
2165 | __u32 lk_padding_6; /* rr_flags */ | |
2166 | __u32 lk_padding_7; /* rr_padding_2 */ | |
2167 | __u32 lk_padding_8; /* rr_padding_3 */ | |
2168 | __u32 lk_padding_9; /* rr_padding_4 */ | |
2169 | }; | |
2170 | ||
2171 | /* instance of mdt_reint_rec */ | |
2172 | struct mdt_rec_unlink { | |
2173 | __u32 ul_opcode; | |
2174 | __u32 ul_cap; | |
2175 | __u32 ul_fsuid; | |
2176 | __u32 ul_fsuid_h; | |
2177 | __u32 ul_fsgid; | |
2178 | __u32 ul_fsgid_h; | |
2179 | __u32 ul_suppgid1; | |
2180 | __u32 ul_suppgid1_h; | |
2181 | __u32 ul_suppgid2; | |
2182 | __u32 ul_suppgid2_h; | |
2183 | struct lu_fid ul_fid1; | |
2184 | struct lu_fid ul_fid2; | |
21aef7d9 | 2185 | __s64 ul_time; |
d7e09d03 PT |
2186 | __u64 ul_padding_2; /* rr_atime */ |
2187 | __u64 ul_padding_3; /* rr_ctime */ | |
2188 | __u64 ul_padding_4; /* rr_size */ | |
2189 | __u64 ul_padding_5; /* rr_blocks */ | |
2190 | __u32 ul_bias; | |
2191 | __u32 ul_mode; | |
2192 | __u32 ul_padding_6; /* rr_flags */ | |
2193 | __u32 ul_padding_7; /* rr_padding_2 */ | |
2194 | __u32 ul_padding_8; /* rr_padding_3 */ | |
2195 | __u32 ul_padding_9; /* rr_padding_4 */ | |
2196 | }; | |
2197 | ||
2198 | /* instance of mdt_reint_rec */ | |
2199 | struct mdt_rec_rename { | |
2200 | __u32 rn_opcode; | |
2201 | __u32 rn_cap; | |
2202 | __u32 rn_fsuid; | |
2203 | __u32 rn_fsuid_h; | |
2204 | __u32 rn_fsgid; | |
2205 | __u32 rn_fsgid_h; | |
2206 | __u32 rn_suppgid1; | |
2207 | __u32 rn_suppgid1_h; | |
2208 | __u32 rn_suppgid2; | |
2209 | __u32 rn_suppgid2_h; | |
2210 | struct lu_fid rn_fid1; | |
2211 | struct lu_fid rn_fid2; | |
21aef7d9 | 2212 | __s64 rn_time; |
d7e09d03 PT |
2213 | __u64 rn_padding_1; /* rr_atime */ |
2214 | __u64 rn_padding_2; /* rr_ctime */ | |
2215 | __u64 rn_padding_3; /* rr_size */ | |
2216 | __u64 rn_padding_4; /* rr_blocks */ | |
2217 | __u32 rn_bias; /* some operation flags */ | |
2218 | __u32 rn_mode; /* cross-ref rename has mode */ | |
2219 | __u32 rn_padding_5; /* rr_flags */ | |
2220 | __u32 rn_padding_6; /* rr_padding_2 */ | |
2221 | __u32 rn_padding_7; /* rr_padding_3 */ | |
2222 | __u32 rn_padding_8; /* rr_padding_4 */ | |
2223 | }; | |
2224 | ||
2225 | /* instance of mdt_reint_rec */ | |
2226 | struct mdt_rec_setxattr { | |
2227 | __u32 sx_opcode; | |
2228 | __u32 sx_cap; | |
2229 | __u32 sx_fsuid; | |
2230 | __u32 sx_fsuid_h; | |
2231 | __u32 sx_fsgid; | |
2232 | __u32 sx_fsgid_h; | |
2233 | __u32 sx_suppgid1; | |
2234 | __u32 sx_suppgid1_h; | |
2235 | __u32 sx_suppgid2; | |
2236 | __u32 sx_suppgid2_h; | |
2237 | struct lu_fid sx_fid; | |
2238 | __u64 sx_padding_1; /* These three are rr_fid2 */ | |
2239 | __u32 sx_padding_2; | |
2240 | __u32 sx_padding_3; | |
2241 | __u64 sx_valid; | |
21aef7d9 | 2242 | __s64 sx_time; |
d7e09d03 PT |
2243 | __u64 sx_padding_5; /* rr_ctime */ |
2244 | __u64 sx_padding_6; /* rr_size */ | |
2245 | __u64 sx_padding_7; /* rr_blocks */ | |
2246 | __u32 sx_size; | |
2247 | __u32 sx_flags; | |
2248 | __u32 sx_padding_8; /* rr_flags */ | |
2249 | __u32 sx_padding_9; /* rr_padding_2 */ | |
2250 | __u32 sx_padding_10; /* rr_padding_3 */ | |
2251 | __u32 sx_padding_11; /* rr_padding_4 */ | |
2252 | }; | |
2253 | ||
2254 | /* | |
2255 | * mdt_rec_reint is the template for all mdt_reint_xxx structures. | |
2256 | * Do NOT change the size of various members, otherwise the value | |
2257 | * will be broken in lustre_swab_mdt_rec_reint(). | |
2258 | * | |
f16192ed | 2259 | * If you add new members in other mdt_reint_xxx structures and need to use the |
d7e09d03 PT |
2260 | * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also. |
2261 | */ | |
2262 | struct mdt_rec_reint { | |
2263 | __u32 rr_opcode; | |
2264 | __u32 rr_cap; | |
2265 | __u32 rr_fsuid; | |
2266 | __u32 rr_fsuid_h; | |
2267 | __u32 rr_fsgid; | |
2268 | __u32 rr_fsgid_h; | |
2269 | __u32 rr_suppgid1; | |
2270 | __u32 rr_suppgid1_h; | |
2271 | __u32 rr_suppgid2; | |
2272 | __u32 rr_suppgid2_h; | |
2273 | struct lu_fid rr_fid1; | |
2274 | struct lu_fid rr_fid2; | |
21aef7d9 OD |
2275 | __s64 rr_mtime; |
2276 | __s64 rr_atime; | |
2277 | __s64 rr_ctime; | |
d7e09d03 PT |
2278 | __u64 rr_size; |
2279 | __u64 rr_blocks; | |
2280 | __u32 rr_bias; | |
2281 | __u32 rr_mode; | |
2282 | __u32 rr_flags; | |
2283 | __u32 rr_flags_h; | |
2284 | __u32 rr_umask; | |
2285 | __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */ | |
2286 | }; | |
2287 | ||
b78c2b9b | 2288 | /* lmv structures */ |
d7e09d03 PT |
2289 | struct lmv_desc { |
2290 | __u32 ld_tgt_count; /* how many MDS's */ | |
2291 | __u32 ld_active_tgt_count; /* how many active */ | |
2292 | __u32 ld_default_stripe_count; /* how many objects are used */ | |
2de35386 | 2293 | __u32 ld_pattern; /* default hash pattern */ |
d7e09d03 PT |
2294 | __u64 ld_default_hash_size; |
2295 | __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */ | |
2296 | __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */ | |
2297 | __u32 ld_qos_maxage; /* in second */ | |
2298 | __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */ | |
2299 | __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */ | |
2300 | struct obd_uuid ld_uuid; | |
2301 | }; | |
2302 | ||
8f18c8a4 | 2303 | /* LMV layout EA, and it will be stored both in master and slave object */ |
2304 | struct lmv_mds_md_v1 { | |
2305 | __u32 lmv_magic; | |
2306 | __u32 lmv_stripe_count; | |
2307 | __u32 lmv_master_mdt_index; /* On master object, it is master | |
2308 | * MDT index, on slave object, it | |
2309 | * is stripe index of the slave obj | |
2310 | */ | |
2311 | __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate | |
2312 | * which hash function to be used, | |
2313 | * Note: only lower 16 bits is being | |
2314 | * used for now. Higher 16 bits will | |
2315 | * be used to mark the object status, | |
2316 | * for example migrating or dead. | |
2317 | */ | |
2318 | __u32 lmv_layout_version; /* Used for directory restriping */ | |
c8deb3cb FY |
2319 | __u32 lmv_padding1; |
2320 | __u64 lmv_padding2; | |
2321 | __u64 lmv_padding3; | |
aaf06e29 | 2322 | char lmv_pool_name[LOV_MAXPOOLNAME + 1];/* pool name */ |
8f18c8a4 | 2323 | struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */ |
2324 | }; | |
2de35386 | 2325 | |
8f18c8a4 | 2326 | #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */ |
2327 | #define LMV_MAGIC LMV_MAGIC_V1 | |
2328 | ||
2329 | /* #define LMV_USER_MAGIC 0x0CD30CD0 */ | |
2330 | #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */ | |
2331 | ||
2332 | /* | |
2333 | *Right now only the lower part(0-16bits) of lmv_hash_type is being used, | |
2334 | * and the higher part will be the flag to indicate the status of object, | |
2335 | * for example the object is being migrated. And the hash function | |
2336 | * might be interpreted differently with different flags. | |
2337 | */ | |
8f18c8a4 | 2338 | #define LMV_HASH_TYPE_MASK 0x0000ffff |
2339 | ||
2340 | #define LMV_HASH_FLAG_MIGRATION 0x80000000 | |
2341 | #define LMV_HASH_FLAG_DEAD 0x40000000 | |
2342 | ||
2de35386 | 2343 | /** |
2344 | * The FNV-1a hash algorithm is as follows: | |
2345 | * hash = FNV_offset_basis | |
2346 | * for each octet_of_data to be hashed | |
2347 | * hash = hash XOR octet_of_data | |
2348 | * hash = hash × FNV_prime | |
2349 | * return hash | |
2350 | * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash | |
2351 | * | |
2352 | * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source | |
2353 | * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL | |
2354 | **/ | |
2355 | #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL | |
2356 | #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL | |
2357 | static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size) | |
2358 | { | |
2359 | __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS; | |
2360 | const unsigned char *p = buf; | |
2361 | size_t i; | |
2362 | ||
2363 | for (i = 0; i < size; i++) { | |
2364 | hash ^= p[i]; | |
2365 | hash *= LUSTRE_FNV_1A_64_PRIME; | |
2366 | } | |
2367 | ||
2368 | return hash; | |
2369 | } | |
2370 | ||
8e9dfe8a | 2371 | union lmv_mds_md { |
2372 | __u32 lmv_magic; | |
2373 | struct lmv_mds_md_v1 lmv_md_v1; | |
2374 | struct lmv_user_md lmv_user_md; | |
2375 | }; | |
2376 | ||
2377 | static inline ssize_t lmv_mds_md_size(int stripe_count, unsigned int lmm_magic) | |
2378 | { | |
2379 | ssize_t len = -EINVAL; | |
2380 | ||
2381 | switch (lmm_magic) { | |
8f18c8a4 | 2382 | case LMV_MAGIC_V1: { |
8e9dfe8a | 2383 | struct lmv_mds_md_v1 *lmm1; |
2384 | ||
2385 | len = sizeof(*lmm1); | |
2386 | len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]); | |
2387 | break; } | |
2388 | default: | |
2389 | break; | |
2390 | } | |
2391 | return len; | |
2392 | } | |
2393 | ||
2394 | static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm) | |
2395 | { | |
2396 | switch (le32_to_cpu(lmm->lmv_magic)) { | |
2397 | case LMV_MAGIC_V1: | |
2398 | return le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count); | |
2399 | case LMV_USER_MAGIC: | |
2400 | return le32_to_cpu(lmm->lmv_user_md.lum_stripe_count); | |
2401 | default: | |
2402 | return -EINVAL; | |
2403 | } | |
2404 | } | |
2405 | ||
2406 | static inline int lmv_mds_md_stripe_count_set(union lmv_mds_md *lmm, | |
2407 | unsigned int stripe_count) | |
2408 | { | |
2409 | int rc = 0; | |
2410 | ||
2411 | switch (le32_to_cpu(lmm->lmv_magic)) { | |
2412 | case LMV_MAGIC_V1: | |
2413 | lmm->lmv_md_v1.lmv_stripe_count = cpu_to_le32(stripe_count); | |
2414 | break; | |
2415 | case LMV_USER_MAGIC: | |
2416 | lmm->lmv_user_md.lum_stripe_count = cpu_to_le32(stripe_count); | |
2417 | break; | |
2418 | default: | |
2419 | rc = -EINVAL; | |
2420 | break; | |
2421 | } | |
2422 | return rc; | |
2423 | } | |
2424 | ||
d7e09d03 | 2425 | enum fld_rpc_opc { |
b78c2b9b | 2426 | FLD_QUERY = 900, |
2427 | FLD_READ = 901, | |
d7e09d03 | 2428 | FLD_LAST_OPC, |
b78c2b9b | 2429 | FLD_FIRST_OPC = FLD_QUERY |
d7e09d03 PT |
2430 | }; |
2431 | ||
2432 | enum seq_rpc_opc { | |
2433 | SEQ_QUERY = 700, | |
2434 | SEQ_LAST_OPC, | |
2435 | SEQ_FIRST_OPC = SEQ_QUERY | |
2436 | }; | |
2437 | ||
2438 | enum seq_op { | |
2439 | SEQ_ALLOC_SUPER = 0, | |
2440 | SEQ_ALLOC_META = 1 | |
2441 | }; | |
2442 | ||
b78c2b9b | 2443 | enum fld_op { |
2444 | FLD_CREATE = 0, | |
2445 | FLD_DELETE = 1, | |
2446 | FLD_LOOKUP = 2, | |
2447 | }; | |
2448 | ||
d7e09d03 PT |
2449 | /* |
2450 | * LOV data structures | |
2451 | */ | |
2452 | ||
2453 | #define LOV_MAX_UUID_BUFFER_SIZE 8192 | |
2454 | /* The size of the buffer the lov/mdc reserves for the | |
2455 | * array of UUIDs returned by the MDS. With the current | |
a1e616b0 OD |
2456 | * protocol, this will limit the max number of OSTs per LOV |
2457 | */ | |
d7e09d03 PT |
2458 | |
2459 | #define LOV_DESC_MAGIC 0xB0CCDE5C | |
081b7265 JH |
2460 | #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */ |
2461 | #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS) | |
d7e09d03 PT |
2462 | |
2463 | /* LOV settings descriptor (should only contain static info) */ | |
2464 | struct lov_desc { | |
2465 | __u32 ld_tgt_count; /* how many OBD's */ | |
a1e616b0 OD |
2466 | __u32 ld_active_tgt_count; /* how many active */ |
2467 | __u32 ld_default_stripe_count; /* how many objects are used */ | |
2468 | __u32 ld_pattern; /* default PATTERN_RAID0 */ | |
2469 | __u64 ld_default_stripe_size; /* in bytes */ | |
2470 | __u64 ld_default_stripe_offset; /* in bytes */ | |
d7e09d03 | 2471 | __u32 ld_padding_0; /* unused */ |
a1e616b0 | 2472 | __u32 ld_qos_maxage; /* in second */ |
d7e09d03 PT |
2473 | __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */ |
2474 | __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */ | |
2475 | struct obd_uuid ld_uuid; | |
2476 | }; | |
2477 | ||
2478 | #define ld_magic ld_active_tgt_count /* for swabbing from llogs */ | |
2479 | ||
d7e09d03 PT |
2480 | /* |
2481 | * LDLM requests: | |
2482 | */ | |
2483 | /* opcodes -- MUST be distinct from OST/MDS opcodes */ | |
980b745f | 2484 | enum ldlm_cmd { |
d7e09d03 PT |
2485 | LDLM_ENQUEUE = 101, |
2486 | LDLM_CONVERT = 102, | |
2487 | LDLM_CANCEL = 103, | |
2488 | LDLM_BL_CALLBACK = 104, | |
2489 | LDLM_CP_CALLBACK = 105, | |
2490 | LDLM_GL_CALLBACK = 106, | |
2491 | LDLM_SET_INFO = 107, | |
2492 | LDLM_LAST_OPC | |
980b745f | 2493 | }; |
d7e09d03 PT |
2494 | #define LDLM_FIRST_OPC LDLM_ENQUEUE |
2495 | ||
2496 | #define RES_NAME_SIZE 4 | |
2497 | struct ldlm_res_id { | |
2498 | __u64 name[RES_NAME_SIZE]; | |
2499 | }; | |
2500 | ||
55f5a824 | 2501 | #define DLDLMRES "[%#llx:%#llx:%#llx].%llx" |
ce74f92d AD |
2502 | #define PLDLMRES(res) (res)->lr_name.name[0], (res)->lr_name.name[1], \ |
2503 | (res)->lr_name.name[2], (res)->lr_name.name[3] | |
2504 | ||
d8f183b3 JH |
2505 | static inline bool ldlm_res_eq(const struct ldlm_res_id *res0, |
2506 | const struct ldlm_res_id *res1) | |
d7e09d03 PT |
2507 | { |
2508 | return !memcmp(res0, res1, sizeof(*res0)); | |
2509 | } | |
2510 | ||
2511 | /* lock types */ | |
52ee0d20 | 2512 | enum ldlm_mode { |
d7e09d03 PT |
2513 | LCK_MINMODE = 0, |
2514 | LCK_EX = 1, | |
2515 | LCK_PW = 2, | |
2516 | LCK_PR = 4, | |
2517 | LCK_CW = 8, | |
2518 | LCK_CR = 16, | |
2519 | LCK_NL = 32, | |
2520 | LCK_GROUP = 64, | |
2521 | LCK_COS = 128, | |
2522 | LCK_MAXMODE | |
52ee0d20 | 2523 | }; |
d7e09d03 PT |
2524 | |
2525 | #define LCK_MODE_NUM 8 | |
2526 | ||
52ee0d20 | 2527 | enum ldlm_type { |
d7e09d03 PT |
2528 | LDLM_PLAIN = 10, |
2529 | LDLM_EXTENT = 11, | |
2530 | LDLM_FLOCK = 12, | |
2531 | LDLM_IBITS = 13, | |
2532 | LDLM_MAX_TYPE | |
52ee0d20 | 2533 | }; |
d7e09d03 PT |
2534 | |
2535 | #define LDLM_MIN_TYPE LDLM_PLAIN | |
2536 | ||
2537 | struct ldlm_extent { | |
2538 | __u64 start; | |
2539 | __u64 end; | |
2540 | __u64 gid; | |
2541 | }; | |
2542 | ||
ac8f0a5c JH |
2543 | static inline int ldlm_extent_overlap(const struct ldlm_extent *ex1, |
2544 | const struct ldlm_extent *ex2) | |
d7e09d03 PT |
2545 | { |
2546 | return (ex1->start <= ex2->end) && (ex2->start <= ex1->end); | |
2547 | } | |
2548 | ||
2549 | /* check if @ex1 contains @ex2 */ | |
ac8f0a5c JH |
2550 | static inline int ldlm_extent_contain(const struct ldlm_extent *ex1, |
2551 | const struct ldlm_extent *ex2) | |
d7e09d03 PT |
2552 | { |
2553 | return (ex1->start <= ex2->start) && (ex1->end >= ex2->end); | |
2554 | } | |
2555 | ||
2556 | struct ldlm_inodebits { | |
2557 | __u64 bits; | |
2558 | }; | |
2559 | ||
2560 | struct ldlm_flock_wire { | |
2561 | __u64 lfw_start; | |
2562 | __u64 lfw_end; | |
2563 | __u64 lfw_owner; | |
2564 | __u32 lfw_padding; | |
2565 | __u32 lfw_pid; | |
2566 | }; | |
2567 | ||
2568 | /* it's important that the fields of the ldlm_extent structure match | |
2569 | * the first fields of the ldlm_flock structure because there is only | |
2570 | * one ldlm_swab routine to process the ldlm_policy_data_t union. if | |
2571 | * this ever changes we will need to swab the union differently based | |
a1e616b0 OD |
2572 | * on the resource type. |
2573 | */ | |
d7e09d03 | 2574 | |
394a9726 | 2575 | union ldlm_wire_policy_data { |
d7e09d03 PT |
2576 | struct ldlm_extent l_extent; |
2577 | struct ldlm_flock_wire l_flock; | |
2578 | struct ldlm_inodebits l_inodebits; | |
394a9726 | 2579 | }; |
d7e09d03 | 2580 | |
d7e09d03 PT |
2581 | union ldlm_gl_desc { |
2582 | struct ldlm_gl_lquota_desc lquota_desc; | |
2583 | }; | |
2584 | ||
43eb3b33 BE |
2585 | enum ldlm_intent_flags { |
2586 | IT_OPEN = BIT(0), | |
2587 | IT_CREAT = BIT(1), | |
2588 | IT_OPEN_CREAT = BIT(1) | BIT(0), | |
2589 | IT_READDIR = BIT(2), | |
2590 | IT_GETATTR = BIT(3), | |
2591 | IT_LOOKUP = BIT(4), | |
2592 | IT_UNLINK = BIT(5), | |
2593 | IT_TRUNC = BIT(6), | |
2594 | IT_GETXATTR = BIT(7), | |
2595 | IT_EXEC = BIT(8), | |
2596 | IT_PIN = BIT(9), | |
2597 | IT_LAYOUT = BIT(10), | |
2598 | IT_QUOTA_DQACQ = BIT(11), | |
2599 | IT_QUOTA_CONN = BIT(12), | |
2600 | IT_SETXATTR = BIT(13), | |
2601 | }; | |
2602 | ||
d7e09d03 PT |
2603 | struct ldlm_intent { |
2604 | __u64 opc; | |
2605 | }; | |
2606 | ||
d7e09d03 | 2607 | struct ldlm_resource_desc { |
52ee0d20 | 2608 | enum ldlm_type lr_type; |
d7e09d03 PT |
2609 | __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */ |
2610 | struct ldlm_res_id lr_name; | |
2611 | }; | |
2612 | ||
d7e09d03 PT |
2613 | struct ldlm_lock_desc { |
2614 | struct ldlm_resource_desc l_resource; | |
52ee0d20 OD |
2615 | enum ldlm_mode l_req_mode; |
2616 | enum ldlm_mode l_granted_mode; | |
394a9726 | 2617 | union ldlm_wire_policy_data l_policy_data; |
d7e09d03 PT |
2618 | }; |
2619 | ||
d7e09d03 PT |
2620 | #define LDLM_LOCKREQ_HANDLES 2 |
2621 | #define LDLM_ENQUEUE_CANCEL_OFF 1 | |
2622 | ||
2623 | struct ldlm_request { | |
2624 | __u32 lock_flags; | |
2625 | __u32 lock_count; | |
2626 | struct ldlm_lock_desc lock_desc; | |
2627 | struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES]; | |
2628 | }; | |
2629 | ||
d7e09d03 | 2630 | /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available. |
a1e616b0 OD |
2631 | * Otherwise, 2 are available. |
2632 | */ | |
1d8cb70c | 2633 | #define ldlm_request_bufsize(count, type) \ |
d7e09d03 PT |
2634 | ({ \ |
2635 | int _avail = LDLM_LOCKREQ_HANDLES; \ | |
2636 | _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \ | |
2637 | sizeof(struct ldlm_request) + \ | |
2638 | (count > _avail ? count - _avail : 0) * \ | |
2639 | sizeof(struct lustre_handle); \ | |
2640 | }) | |
2641 | ||
2642 | struct ldlm_reply { | |
2643 | __u32 lock_flags; | |
2644 | __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */ | |
2645 | struct ldlm_lock_desc lock_desc; | |
2646 | struct lustre_handle lock_handle; | |
2647 | __u64 lock_policy_res1; | |
2648 | __u64 lock_policy_res2; | |
2649 | }; | |
2650 | ||
d7e09d03 PT |
2651 | #define ldlm_flags_to_wire(flags) ((__u32)(flags)) |
2652 | #define ldlm_flags_from_wire(flags) ((__u64)(flags)) | |
2653 | ||
2654 | /* | |
2655 | * Opcodes for mountconf (mgs and mgc) | |
2656 | */ | |
66423bff | 2657 | enum mgs_cmd { |
d7e09d03 PT |
2658 | MGS_CONNECT = 250, |
2659 | MGS_DISCONNECT, | |
2660 | MGS_EXCEPTION, /* node died, etc. */ | |
2661 | MGS_TARGET_REG, /* whenever target starts up */ | |
2662 | MGS_TARGET_DEL, | |
2663 | MGS_SET_INFO, | |
2664 | MGS_CONFIG_READ, | |
2665 | MGS_LAST_OPC | |
66423bff | 2666 | }; |
d7e09d03 PT |
2667 | #define MGS_FIRST_OPC MGS_CONNECT |
2668 | ||
2669 | #define MGS_PARAM_MAXLEN 1024 | |
2670 | #define KEY_SET_INFO "set_info" | |
2671 | ||
2672 | struct mgs_send_param { | |
2673 | char mgs_param[MGS_PARAM_MAXLEN]; | |
2674 | }; | |
2675 | ||
2676 | /* We pass this info to the MGS so it can write config logs */ | |
2677 | #define MTI_NAME_MAXLEN 64 | |
2678 | #define MTI_PARAM_MAXLEN 4096 | |
2679 | #define MTI_NIDS_MAX 32 | |
2680 | struct mgs_target_info { | |
2681 | __u32 mti_lustre_ver; | |
2682 | __u32 mti_stripe_index; | |
2683 | __u32 mti_config_ver; | |
2684 | __u32 mti_flags; | |
2685 | __u32 mti_nid_count; | |
2686 | __u32 mti_instance; /* Running instance of target */ | |
2687 | char mti_fsname[MTI_NAME_MAXLEN]; | |
2688 | char mti_svname[MTI_NAME_MAXLEN]; | |
2689 | char mti_uuid[sizeof(struct obd_uuid)]; | |
2690 | __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/ | |
2691 | char mti_params[MTI_PARAM_MAXLEN]; | |
2692 | }; | |
8150a97f | 2693 | |
d7e09d03 PT |
2694 | struct mgs_nidtbl_entry { |
2695 | __u64 mne_version; /* table version of this entry */ | |
2696 | __u32 mne_instance; /* target instance # */ | |
2697 | __u32 mne_index; /* target index */ | |
2698 | __u32 mne_length; /* length of this entry - by bytes */ | |
2699 | __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */ | |
2700 | __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */ | |
2701 | __u8 mne_nid_size; /* size of each NID, by bytes */ | |
2702 | __u8 mne_nid_count; /* # of NIDs in buffer */ | |
2703 | union { | |
2704 | lnet_nid_t nids[0]; /* variable size buffer for NIDs. */ | |
2705 | } u; | |
2706 | }; | |
8150a97f | 2707 | |
d7e09d03 PT |
2708 | struct mgs_config_body { |
2709 | char mcb_name[MTI_NAME_MAXLEN]; /* logname */ | |
2710 | __u64 mcb_offset; /* next index of config log to request */ | |
2711 | __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */ | |
2712 | __u8 mcb_reserved; | |
2713 | __u8 mcb_bits; /* bits unit size of config log */ | |
2714 | __u32 mcb_units; /* # of units for bulk transfer */ | |
2715 | }; | |
8150a97f | 2716 | |
d7e09d03 PT |
2717 | struct mgs_config_res { |
2718 | __u64 mcr_offset; /* index of last config log */ | |
2719 | __u64 mcr_size; /* size of the log */ | |
2720 | }; | |
8150a97f | 2721 | |
d7e09d03 PT |
2722 | /* Config marker flags (in config log) */ |
2723 | #define CM_START 0x01 | |
2724 | #define CM_END 0x02 | |
2725 | #define CM_SKIP 0x04 | |
2726 | #define CM_UPGRADE146 0x08 | |
2727 | #define CM_EXCLUDE 0x10 | |
2728 | #define CM_START_SKIP (CM_START | CM_SKIP) | |
2729 | ||
2730 | struct cfg_marker { | |
2731 | __u32 cm_step; /* aka config version */ | |
2732 | __u32 cm_flags; | |
2733 | __u32 cm_vers; /* lustre release version number */ | |
2734 | __u32 cm_padding; /* 64 bit align */ | |
21aef7d9 OD |
2735 | __s64 cm_createtime; /*when this record was first created */ |
2736 | __s64 cm_canceltime; /*when this record is no longer valid*/ | |
d7e09d03 PT |
2737 | char cm_tgtname[MTI_NAME_MAXLEN]; |
2738 | char cm_comment[MTI_NAME_MAXLEN]; | |
2739 | }; | |
2740 | ||
d7e09d03 PT |
2741 | /* |
2742 | * Opcodes for multiple servers. | |
2743 | */ | |
2744 | ||
deadbe9a | 2745 | enum obd_cmd { |
d7e09d03 PT |
2746 | OBD_PING = 400, |
2747 | OBD_LOG_CANCEL, | |
e57721e7 | 2748 | OBD_QC_CALLBACK, /* not used since 2.4 */ |
d7e09d03 PT |
2749 | OBD_IDX_READ, |
2750 | OBD_LAST_OPC | |
deadbe9a | 2751 | }; |
d7e09d03 PT |
2752 | #define OBD_FIRST_OPC OBD_PING |
2753 | ||
b14b3ba5 MP |
2754 | /** |
2755 | * llog contexts indices. | |
2756 | * | |
2757 | * There is compatibility problem with indexes below, they are not | |
2758 | * continuous and must keep their numbers for compatibility needs. | |
2759 | * See LU-5218 for details. | |
2760 | */ | |
2761 | enum llog_ctxt_id { | |
2762 | LLOG_CONFIG_ORIG_CTXT = 0, | |
2763 | LLOG_CONFIG_REPL_CTXT = 1, | |
2764 | LLOG_MDS_OST_ORIG_CTXT = 2, | |
2765 | LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */ | |
2766 | LLOG_SIZE_ORIG_CTXT = 4, | |
2767 | LLOG_SIZE_REPL_CTXT = 5, | |
2768 | LLOG_TEST_ORIG_CTXT = 8, | |
2769 | LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */ | |
2770 | LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */ | |
2771 | LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */ | |
2772 | /* for multiple changelog consumers */ | |
2773 | LLOG_CHANGELOG_USER_ORIG_CTXT = 14, | |
2774 | LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */ | |
2775 | LLOG_MAX_CTXTS | |
2776 | }; | |
d7e09d03 PT |
2777 | |
2778 | /** Identifier for a single log object */ | |
2779 | struct llog_logid { | |
2780 | struct ost_id lgl_oi; | |
2781 | __u32 lgl_ogen; | |
a6e19b4d | 2782 | } __packed; |
d7e09d03 PT |
2783 | |
2784 | /** Records written to the CATALOGS list */ | |
2785 | #define CATLIST "CATALOGS" | |
2786 | struct llog_catid { | |
2787 | struct llog_logid lci_logid; | |
2788 | __u32 lci_padding1; | |
2789 | __u32 lci_padding2; | |
2790 | __u32 lci_padding3; | |
a6e19b4d | 2791 | } __packed; |
d7e09d03 PT |
2792 | |
2793 | /* Log data record types - there is no specific reason that these need to | |
2794 | * be related to the RPC opcodes, but no reason not to (may be handy later?) | |
2795 | */ | |
2796 | #define LLOG_OP_MAGIC 0x10600000 | |
2797 | #define LLOG_OP_MASK 0xfff00000 | |
2798 | ||
bfea4767 | 2799 | enum llog_op_type { |
d7e09d03 PT |
2800 | LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000, |
2801 | OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00, | |
2802 | /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */ | |
2803 | MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | | |
2804 | REINT_UNLINK, /* obsolete after 2.5.0 */ | |
2805 | MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) | | |
2806 | REINT_UNLINK, | |
2807 | /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */ | |
2808 | MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) | | |
2809 | REINT_SETATTR, | |
2810 | OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000, | |
2811 | /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */ | |
2812 | LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000, | |
2813 | /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */ | |
2814 | CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000, | |
2815 | CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000, | |
99a92265 | 2816 | HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000, |
d7e09d03 PT |
2817 | LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539, |
2818 | LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b, | |
bfea4767 | 2819 | }; |
d7e09d03 PT |
2820 | |
2821 | #define LLOG_REC_HDR_NEEDS_SWABBING(r) \ | |
2822 | (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC)) | |
2823 | ||
2824 | /** Log record header - stored in little endian order. | |
2825 | * Each record must start with this struct, end with a llog_rec_tail, | |
2826 | * and be a multiple of 256 bits in size. | |
2827 | */ | |
2828 | struct llog_rec_hdr { | |
2829 | __u32 lrh_len; | |
2830 | __u32 lrh_index; | |
2831 | __u32 lrh_type; | |
2832 | __u32 lrh_id; | |
2833 | }; | |
2834 | ||
2835 | struct llog_rec_tail { | |
2836 | __u32 lrt_len; | |
2837 | __u32 lrt_index; | |
2838 | }; | |
2839 | ||
2840 | /* Where data follow just after header */ | |
2841 | #define REC_DATA(ptr) \ | |
2842 | ((void *)((char *)ptr + sizeof(struct llog_rec_hdr))) | |
2843 | ||
2844 | #define REC_DATA_LEN(rec) \ | |
2845 | (rec->lrh_len - sizeof(struct llog_rec_hdr) - \ | |
2846 | sizeof(struct llog_rec_tail)) | |
2847 | ||
2848 | struct llog_logid_rec { | |
2849 | struct llog_rec_hdr lid_hdr; | |
2850 | struct llog_logid lid_id; | |
2851 | __u32 lid_padding1; | |
2852 | __u64 lid_padding2; | |
2853 | __u64 lid_padding3; | |
2854 | struct llog_rec_tail lid_tail; | |
a6e19b4d | 2855 | } __packed; |
d7e09d03 PT |
2856 | |
2857 | struct llog_unlink_rec { | |
2858 | struct llog_rec_hdr lur_hdr; | |
21aef7d9 OD |
2859 | __u64 lur_oid; |
2860 | __u32 lur_oseq; | |
2861 | __u32 lur_count; | |
d7e09d03 | 2862 | struct llog_rec_tail lur_tail; |
a6e19b4d | 2863 | } __packed; |
d7e09d03 PT |
2864 | |
2865 | struct llog_unlink64_rec { | |
2866 | struct llog_rec_hdr lur_hdr; | |
2867 | struct lu_fid lur_fid; | |
21aef7d9 | 2868 | __u32 lur_count; /* to destroy the lost precreated */ |
d7e09d03 PT |
2869 | __u32 lur_padding1; |
2870 | __u64 lur_padding2; | |
2871 | __u64 lur_padding3; | |
2872 | struct llog_rec_tail lur_tail; | |
a6e19b4d | 2873 | } __packed; |
d7e09d03 PT |
2874 | |
2875 | struct llog_setattr64_rec { | |
2876 | struct llog_rec_hdr lsr_hdr; | |
2877 | struct ost_id lsr_oi; | |
2878 | __u32 lsr_uid; | |
2879 | __u32 lsr_uid_h; | |
2880 | __u32 lsr_gid; | |
2881 | __u32 lsr_gid_h; | |
f7aafa7c | 2882 | __u64 lsr_valid; |
d7e09d03 | 2883 | struct llog_rec_tail lsr_tail; |
a6e19b4d | 2884 | } __packed; |
d7e09d03 PT |
2885 | |
2886 | struct llog_size_change_rec { | |
2887 | struct llog_rec_hdr lsc_hdr; | |
2888 | struct ll_fid lsc_fid; | |
2889 | __u32 lsc_ioepoch; | |
2890 | __u32 lsc_padding1; | |
2891 | __u64 lsc_padding2; | |
2892 | __u64 lsc_padding3; | |
2893 | struct llog_rec_tail lsc_tail; | |
a6e19b4d | 2894 | } __packed; |
d7e09d03 | 2895 | |
d7e09d03 PT |
2896 | /* changelog llog name, needed by client replicators */ |
2897 | #define CHANGELOG_CATALOG "changelog_catalog" | |
2898 | ||
2899 | struct changelog_setinfo { | |
2900 | __u64 cs_recno; | |
2901 | __u32 cs_id; | |
a6e19b4d | 2902 | } __packed; |
d7e09d03 PT |
2903 | |
2904 | /** changelog record */ | |
2905 | struct llog_changelog_rec { | |
c9fe1f7f HD |
2906 | struct llog_rec_hdr cr_hdr; |
2907 | struct changelog_rec cr; /**< Variable length field */ | |
2908 | struct llog_rec_tail cr_do_not_use; /**< for_sizezof_only */ | |
a6e19b4d | 2909 | } __packed; |
d7e09d03 | 2910 | |
d7e09d03 PT |
2911 | struct llog_changelog_user_rec { |
2912 | struct llog_rec_hdr cur_hdr; | |
2913 | __u32 cur_id; | |
2914 | __u32 cur_padding; | |
2915 | __u64 cur_endrec; | |
2916 | struct llog_rec_tail cur_tail; | |
a6e19b4d | 2917 | } __packed; |
d7e09d03 | 2918 | |
99a92265 | 2919 | enum agent_req_status { |
2920 | ARS_WAITING, | |
2921 | ARS_STARTED, | |
2922 | ARS_FAILED, | |
2923 | ARS_CANCELED, | |
2924 | ARS_SUCCEED, | |
2925 | }; | |
2926 | ||
ac8f0a5c | 2927 | static inline const char *agent_req_status2name(const enum agent_req_status ars) |
99a92265 | 2928 | { |
2929 | switch (ars) { | |
2930 | case ARS_WAITING: | |
2931 | return "WAITING"; | |
2932 | case ARS_STARTED: | |
2933 | return "STARTED"; | |
2934 | case ARS_FAILED: | |
2935 | return "FAILED"; | |
2936 | case ARS_CANCELED: | |
2937 | return "CANCELED"; | |
2938 | case ARS_SUCCEED: | |
2939 | return "SUCCEED"; | |
2940 | default: | |
2941 | return "UNKNOWN"; | |
2942 | } | |
2943 | } | |
2944 | ||
2945 | static inline bool agent_req_in_final_state(enum agent_req_status ars) | |
2946 | { | |
2947 | return ((ars == ARS_SUCCEED) || (ars == ARS_FAILED) || | |
2948 | (ars == ARS_CANCELED)); | |
2949 | } | |
2950 | ||
2951 | struct llog_agent_req_rec { | |
2952 | struct llog_rec_hdr arr_hdr; /**< record header */ | |
2953 | __u32 arr_status; /**< status of the request */ | |
2954 | /* must match enum | |
a1e616b0 OD |
2955 | * agent_req_status |
2956 | */ | |
99a92265 | 2957 | __u32 arr_archive_id; /**< backend archive number */ |
2958 | __u64 arr_flags; /**< req flags */ | |
a1e616b0 | 2959 | __u64 arr_compound_id;/**< compound cookie */ |
99a92265 | 2960 | __u64 arr_req_create; /**< req. creation time */ |
2961 | __u64 arr_req_change; /**< req. status change time */ | |
2962 | struct hsm_action_item arr_hai; /**< req. to the agent */ | |
a1e616b0 | 2963 | struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */ |
a6e19b4d | 2964 | } __packed; |
99a92265 | 2965 | |
d7e09d03 PT |
2966 | /* Old llog gen for compatibility */ |
2967 | struct llog_gen { | |
2968 | __u64 mnt_cnt; | |
2969 | __u64 conn_cnt; | |
a6e19b4d | 2970 | } __packed; |
d7e09d03 PT |
2971 | |
2972 | struct llog_gen_rec { | |
2973 | struct llog_rec_hdr lgr_hdr; | |
2974 | struct llog_gen lgr_gen; | |
2975 | __u64 padding1; | |
2976 | __u64 padding2; | |
2977 | __u64 padding3; | |
2978 | struct llog_rec_tail lgr_tail; | |
2979 | }; | |
2980 | ||
d7e09d03 PT |
2981 | /* flags for the logs */ |
2982 | enum llog_flag { | |
2983 | LLOG_F_ZAP_WHEN_EMPTY = 0x1, | |
2984 | LLOG_F_IS_CAT = 0x2, | |
2985 | LLOG_F_IS_PLAIN = 0x4, | |
c9fe1f7f | 2986 | LLOG_F_EXT_JOBID = BIT(3), |
b5367061 | 2987 | LLOG_F_IS_FIXSIZE = BIT(4), |
c9fe1f7f | 2988 | |
b5367061 | 2989 | /* |
2990 | * Note: Flags covered by LLOG_F_EXT_MASK will be inherited from | |
2991 | * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here, | |
2992 | * because the catlog record is usually fixed size, but its plain | |
2993 | * log record can be variable | |
2994 | */ | |
c9fe1f7f | 2995 | LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID, |
d7e09d03 PT |
2996 | }; |
2997 | ||
11c647ca AD |
2998 | /* On-disk header structure of each log object, stored in little endian order */ |
2999 | #define LLOG_MIN_CHUNK_SIZE 8192 | |
3000 | #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + | |
3001 | * sizeof(llh_tail) - sizeof(llh_bitmap) | |
3002 | */ | |
3003 | #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE) | |
3004 | #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */ | |
3005 | ||
3006 | /* flags for the logs */ | |
d7e09d03 PT |
3007 | struct llog_log_hdr { |
3008 | struct llog_rec_hdr llh_hdr; | |
21aef7d9 | 3009 | __s64 llh_timestamp; |
d7e09d03 PT |
3010 | __u32 llh_count; |
3011 | __u32 llh_bitmap_offset; | |
3012 | __u32 llh_size; | |
3013 | __u32 llh_flags; | |
3014 | __u32 llh_cat_idx; | |
3015 | /* for a catalog the first plain slot is next to it */ | |
3016 | struct obd_uuid llh_tgtuuid; | |
cd94f231 | 3017 | __u32 llh_reserved[LLOG_HEADER_SIZE / sizeof(__u32) - 23]; |
11c647ca AD |
3018 | /* These fields must always be at the end of the llog_log_hdr. |
3019 | * Note: llh_bitmap size is variable because llog chunk size could be | |
3020 | * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192 | |
3021 | * bytes, and the real size is stored in llh_hdr.lrh_len, which means | |
3022 | * llh_tail should only be referred by LLOG_HDR_TAIL(). | |
3023 | * But this structure is also used by client/server llog interface | |
3024 | * (see llog_client.c), it will be kept in its original way to avoid | |
3025 | * compatibility issue. | |
3026 | */ | |
cd94f231 | 3027 | __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)]; |
d7e09d03 | 3028 | struct llog_rec_tail llh_tail; |
a6e19b4d | 3029 | } __packed; |
d7e09d03 | 3030 | |
11c647ca AD |
3031 | #undef LLOG_HEADER_SIZE |
3032 | #undef LLOG_BITMAP_BYTES | |
3033 | ||
3034 | #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \ | |
3035 | llh->llh_bitmap_offset - \ | |
3036 | sizeof(llh->llh_tail)) * 8) | |
3037 | #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \ | |
3038 | (llh)->llh_bitmap_offset) | |
3039 | #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \ | |
3040 | llh->llh_hdr.lrh_len - \ | |
3041 | sizeof(llh->llh_tail))) | |
d7e09d03 | 3042 | |
a1e616b0 OD |
3043 | /** log cookies are used to reference a specific log file and a record |
3044 | * therein | |
3045 | */ | |
d7e09d03 PT |
3046 | struct llog_cookie { |
3047 | struct llog_logid lgc_lgl; | |
3048 | __u32 lgc_subsys; | |
3049 | __u32 lgc_index; | |
3050 | __u32 lgc_padding; | |
a6e19b4d | 3051 | } __packed; |
d7e09d03 PT |
3052 | |
3053 | /** llog protocol */ | |
3054 | enum llogd_rpc_ops { | |
3055 | LLOG_ORIGIN_HANDLE_CREATE = 501, | |
3056 | LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502, | |
3057 | LLOG_ORIGIN_HANDLE_READ_HEADER = 503, | |
3058 | LLOG_ORIGIN_HANDLE_WRITE_REC = 504, | |
3059 | LLOG_ORIGIN_HANDLE_CLOSE = 505, | |
a1e616b0 | 3060 | LLOG_ORIGIN_CONNECT = 506, |
d7e09d03 PT |
3061 | LLOG_CATINFO = 507, /* deprecated */ |
3062 | LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508, | |
3063 | LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/ | |
3064 | LLOG_LAST_OPC, | |
3065 | LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE | |
3066 | }; | |
3067 | ||
3068 | struct llogd_body { | |
3069 | struct llog_logid lgd_logid; | |
3070 | __u32 lgd_ctxt_idx; | |
3071 | __u32 lgd_llh_flags; | |
3072 | __u32 lgd_index; | |
3073 | __u32 lgd_saved_index; | |
3074 | __u32 lgd_len; | |
3075 | __u64 lgd_cur_offset; | |
a6e19b4d | 3076 | } __packed; |
d7e09d03 PT |
3077 | |
3078 | struct llogd_conn_body { | |
3079 | struct llog_gen lgdc_gen; | |
3080 | struct llog_logid lgdc_logid; | |
3081 | __u32 lgdc_ctxt_idx; | |
a6e19b4d | 3082 | } __packed; |
d7e09d03 PT |
3083 | |
3084 | /* Note: 64-bit types are 64-bit aligned in structure */ | |
3085 | struct obdo { | |
21aef7d9 OD |
3086 | __u64 o_valid; /* hot fields in this obdo */ |
3087 | struct ost_id o_oi; | |
3088 | __u64 o_parent_seq; | |
3089 | __u64 o_size; /* o_size-o_blocks == ost_lvb */ | |
3090 | __s64 o_mtime; | |
3091 | __s64 o_atime; | |
3092 | __s64 o_ctime; | |
3093 | __u64 o_blocks; /* brw: cli sent cached bytes */ | |
3094 | __u64 o_grant; | |
d7e09d03 PT |
3095 | |
3096 | /* 32-bit fields start here: keep an even number of them via padding */ | |
21aef7d9 OD |
3097 | __u32 o_blksize; /* optimal IO blocksize */ |
3098 | __u32 o_mode; /* brw: cli sent cache remain */ | |
3099 | __u32 o_uid; | |
3100 | __u32 o_gid; | |
3101 | __u32 o_flags; | |
3102 | __u32 o_nlink; /* brw: checksum */ | |
3103 | __u32 o_parent_oid; | |
3104 | __u32 o_misc; /* brw: o_dropped */ | |
d7e09d03 PT |
3105 | |
3106 | __u64 o_ioepoch; /* epoch in ost writes */ | |
3107 | __u32 o_stripe_idx; /* holds stripe idx */ | |
3108 | __u32 o_parent_ver; | |
a1e616b0 OD |
3109 | struct lustre_handle o_handle; /* brw: lock handle to prolong locks |
3110 | */ | |
0a4bea92 JH |
3111 | struct llog_cookie o_lcookie; /* destroy: unlink cookie from MDS, |
3112 | * obsolete in 2.8, reused in OSP | |
a1e616b0 | 3113 | */ |
d7e09d03 PT |
3114 | __u32 o_uid_h; |
3115 | __u32 o_gid_h; | |
3116 | ||
3117 | __u64 o_data_version; /* getattr: sum of iversion for | |
3118 | * each stripe. | |
3119 | * brw: grant space consumed on | |
a1e616b0 OD |
3120 | * the client for the write |
3121 | */ | |
d7e09d03 PT |
3122 | __u64 o_padding_4; |
3123 | __u64 o_padding_5; | |
3124 | __u64 o_padding_6; | |
3125 | }; | |
3126 | ||
3127 | #define o_dirty o_blocks | |
3128 | #define o_undirty o_mode | |
3129 | #define o_dropped o_misc | |
3130 | #define o_cksum o_nlink | |
3131 | #define o_grant_used o_data_version | |
3132 | ||
d7e09d03 PT |
3133 | /* request structure for OST's */ |
3134 | struct ost_body { | |
3135 | struct obdo oa; | |
3136 | }; | |
3137 | ||
3138 | /* Key for FIEMAP to be used in get_info calls */ | |
3139 | struct ll_fiemap_info_key { | |
cbd4d4a8 BJ |
3140 | char lfik_name[8]; |
3141 | struct obdo lfik_oa; | |
3142 | struct fiemap lfik_fiemap; | |
d7e09d03 PT |
3143 | }; |
3144 | ||
d7e09d03 PT |
3145 | /* Functions for dumping PTLRPC fields */ |
3146 | void dump_rniobuf(struct niobuf_remote *rnb); | |
3147 | void dump_ioo(struct obd_ioobj *nb); | |
d7e09d03 PT |
3148 | void dump_ost_body(struct ost_body *ob); |
3149 | void dump_rcs(__u32 *rc); | |
3150 | ||
d7e09d03 | 3151 | /* security opcodes */ |
9e04cbfc | 3152 | enum sec_cmd { |
d7e09d03 PT |
3153 | SEC_CTX_INIT = 801, |
3154 | SEC_CTX_INIT_CONT = 802, | |
3155 | SEC_CTX_FINI = 803, | |
3156 | SEC_LAST_OPC, | |
3157 | SEC_FIRST_OPC = SEC_CTX_INIT | |
9e04cbfc | 3158 | }; |
d7e09d03 PT |
3159 | |
3160 | /* | |
3161 | * capa related definitions | |
3162 | */ | |
3163 | #define CAPA_HMAC_MAX_LEN 64 | |
3164 | #define CAPA_HMAC_KEY_MAX_LEN 56 | |
3165 | ||
3166 | /* NB take care when changing the sequence of elements this struct, | |
a1e616b0 OD |
3167 | * because the offset info is used in find_capa() |
3168 | */ | |
d7e09d03 PT |
3169 | struct lustre_capa { |
3170 | struct lu_fid lc_fid; /** fid */ | |
3171 | __u64 lc_opc; /** operations allowed */ | |
3172 | __u64 lc_uid; /** file owner */ | |
3173 | __u64 lc_gid; /** file group */ | |
3174 | __u32 lc_flags; /** HMAC algorithm & flags */ | |
3175 | __u32 lc_keyid; /** key# used for the capability */ | |
3176 | __u32 lc_timeout; /** capa timeout value (sec) */ | |
a6c8746d | 3177 | /* FIXME: y2038 time_t overflow: */ |
d7e09d03 PT |
3178 | __u32 lc_expiry; /** expiry time (sec) */ |
3179 | __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */ | |
a6e19b4d | 3180 | } __packed; |
d7e09d03 | 3181 | |
d7e09d03 PT |
3182 | /** lustre_capa::lc_opc */ |
3183 | enum { | |
cd94f231 OD |
3184 | CAPA_OPC_BODY_WRITE = 1 << 0, /**< write object data */ |
3185 | CAPA_OPC_BODY_READ = 1 << 1, /**< read object data */ | |
3186 | CAPA_OPC_INDEX_LOOKUP = 1 << 2, /**< lookup object fid */ | |
3187 | CAPA_OPC_INDEX_INSERT = 1 << 3, /**< insert object fid */ | |
3188 | CAPA_OPC_INDEX_DELETE = 1 << 4, /**< delete object fid */ | |
3189 | CAPA_OPC_OSS_WRITE = 1 << 5, /**< write oss object data */ | |
3190 | CAPA_OPC_OSS_READ = 1 << 6, /**< read oss object data */ | |
3191 | CAPA_OPC_OSS_TRUNC = 1 << 7, /**< truncate oss object */ | |
3192 | CAPA_OPC_OSS_DESTROY = 1 << 8, /**< destroy oss object */ | |
3193 | CAPA_OPC_META_WRITE = 1 << 9, /**< write object meta data */ | |
3194 | CAPA_OPC_META_READ = 1 << 10, /**< read object meta data */ | |
d7e09d03 PT |
3195 | }; |
3196 | ||
3197 | #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE) | |
3198 | #define CAPA_OPC_MDS_ONLY \ | |
3199 | (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \ | |
3200 | CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE) | |
3201 | #define CAPA_OPC_OSS_ONLY \ | |
3202 | (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \ | |
3203 | CAPA_OPC_OSS_DESTROY) | |
3204 | #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY | |
3205 | #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY) | |
3206 | ||
d7e09d03 PT |
3207 | struct lustre_capa_key { |
3208 | __u64 lk_seq; /**< mds# */ | |
3209 | __u32 lk_keyid; /**< key# */ | |
3210 | __u32 lk_padding; | |
3211 | __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */ | |
a6e19b4d | 3212 | } __packed; |
d7e09d03 | 3213 | |
d7e09d03 PT |
3214 | /** The link ea holds 1 \a link_ea_entry for each hardlink */ |
3215 | #define LINK_EA_MAGIC 0x11EAF1DFUL | |
3216 | struct link_ea_header { | |
3217 | __u32 leh_magic; | |
3218 | __u32 leh_reccount; | |
3219 | __u64 leh_len; /* total size */ | |
3220 | /* future use */ | |
3221 | __u32 padding1; | |
3222 | __u32 padding2; | |
3223 | }; | |
3224 | ||
3225 | /** Hardlink data is name and parent fid. | |
3226 | * Stored in this crazy struct for maximum packing and endian-neutrality | |
3227 | */ | |
3228 | struct link_ea_entry { | |
3229 | /** __u16 stored big-endian, unaligned */ | |
3230 | unsigned char lee_reclen[2]; | |
3231 | unsigned char lee_parent_fid[sizeof(struct lu_fid)]; | |
3232 | char lee_name[0]; | |
a6e19b4d | 3233 | } __packed; |
d7e09d03 PT |
3234 | |
3235 | /** fid2path request/reply structure */ | |
3236 | struct getinfo_fid2path { | |
3237 | struct lu_fid gf_fid; | |
3238 | __u64 gf_recno; | |
3239 | __u32 gf_linkno; | |
3240 | __u32 gf_pathlen; | |
3241 | char gf_path[0]; | |
a6e19b4d | 3242 | } __packed; |
d7e09d03 | 3243 | |
a6d879fd HD |
3244 | /** path2parent request/reply structures */ |
3245 | struct getparent { | |
3246 | struct lu_fid gp_fid; /**< parent FID */ | |
3247 | __u32 gp_linkno; /**< hardlink number */ | |
3248 | __u32 gp_name_size; /**< size of the name field */ | |
3249 | char gp_name[0]; /**< zero-terminated link name */ | |
3250 | } __packed; | |
3251 | ||
d7e09d03 PT |
3252 | enum { |
3253 | LAYOUT_INTENT_ACCESS = 0, | |
3254 | LAYOUT_INTENT_READ = 1, | |
3255 | LAYOUT_INTENT_WRITE = 2, | |
3256 | LAYOUT_INTENT_GLIMPSE = 3, | |
3257 | LAYOUT_INTENT_TRUNC = 4, | |
3258 | LAYOUT_INTENT_RELEASE = 5, | |
3259 | LAYOUT_INTENT_RESTORE = 6 | |
3260 | }; | |
3261 | ||
3262 | /* enqueue layout lock with intent */ | |
3263 | struct layout_intent { | |
3264 | __u32 li_opc; /* intent operation for enqueue, read, write etc */ | |
3265 | __u32 li_flags; | |
3266 | __u64 li_start; | |
3267 | __u64 li_end; | |
3268 | }; | |
3269 | ||
d7e09d03 PT |
3270 | /** |
3271 | * On the wire version of hsm_progress structure. | |
3272 | * | |
3273 | * Contains the userspace hsm_progress and some internal fields. | |
3274 | */ | |
3275 | struct hsm_progress_kernel { | |
3276 | /* Field taken from struct hsm_progress */ | |
d8f6bc9a | 3277 | struct lu_fid hpk_fid; |
d7e09d03 PT |
3278 | __u64 hpk_cookie; |
3279 | struct hsm_extent hpk_extent; | |
3280 | __u16 hpk_flags; | |
3281 | __u16 hpk_errval; /* positive val */ | |
3282 | __u32 hpk_padding1; | |
3283 | /* Additional fields */ | |
3284 | __u64 hpk_data_version; | |
3285 | __u64 hpk_padding2; | |
a6e19b4d | 3286 | } __packed; |
d7e09d03 | 3287 | |
d7e09d03 PT |
3288 | /** layout swap request structure |
3289 | * fid1 and fid2 are in mdt_body | |
3290 | */ | |
3291 | struct mdc_swap_layouts { | |
3292 | __u64 msl_flags; | |
3293 | } __packed; | |
3294 | ||
48d23e61 JX |
3295 | struct close_data { |
3296 | struct lustre_handle cd_handle; | |
3297 | struct lu_fid cd_fid; | |
3298 | __u64 cd_data_version; | |
3299 | __u64 cd_reserved[8]; | |
3300 | }; | |
3301 | ||
d7e09d03 PT |
3302 | #endif |
3303 | /** @} lustreidl */ |