4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2015, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #define DEBUG_SUBSYSTEM S_MDC
38 #include <linux/user_namespace.h>
39 #ifdef HAVE_UIDGID_HEADER
40 # include <linux/uidgid.h>
42 #include <lustre_net.h>
43 #include <lustre/lustre_idl.h>
44 #include <obd_class.h>
46 #include <cl_object.h>
47 #include "mdc_internal.h"
49 static void set_mrc_cr_flags(struct mdt_rec_create *mrc, __u64 flags)
51 mrc->cr_flags_l = (__u32)(flags & 0xFFFFFFFFUll);
52 mrc->cr_flags_h = (__u32)(flags >> 32);
55 static void __mdc_pack_body(struct mdt_body *b, __u32 suppgid)
59 b->mbo_suppgid = suppgid;
60 b->mbo_uid = from_kuid(&init_user_ns, current_uid());
61 b->mbo_gid = from_kgid(&init_user_ns, current_gid());
62 b->mbo_fsuid = from_kuid(&init_user_ns, current_fsuid());
63 b->mbo_fsgid = from_kgid(&init_user_ns, current_fsgid());
64 b->mbo_capability = cfs_curproc_cap_pack();
67 void mdc_swap_layouts_pack(struct ptlrpc_request *req,
68 struct md_op_data *op_data)
70 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
73 __mdc_pack_body(b, op_data->op_suppgids[0]);
74 b->mbo_fid1 = op_data->op_fid1;
75 b->mbo_fid2 = op_data->op_fid2;
76 b->mbo_valid |= OBD_MD_FLID;
79 void mdc_pack_body(struct ptlrpc_request *req, const struct lu_fid *fid,
80 u64 valid, size_t ea_size, u32 suppgid, u32 flags)
82 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
86 b->mbo_eadatasize = ea_size;
88 __mdc_pack_body(b, suppgid);
91 b->mbo_valid |= OBD_MD_FLID;
96 * Pack a name (path component) into a request
98 * \param[in] req request
99 * \param[in] field request field (usually RMF_NAME)
100 * \param[in] name path component
101 * \param[in] name_len length of path component
103 * \a field must be present in \a req and of size \a name_len + 1.
105 * \a name must be '\0' terminated of length \a name_len and represent
106 * a single path component (not contain '/').
108 static void mdc_pack_name(struct ptlrpc_request *req,
109 const struct req_msg_field *field,
110 const char *name, size_t name_len)
116 buf = req_capsule_client_get(&req->rq_pill, field);
117 buf_size = req_capsule_get_size(&req->rq_pill, field, RCL_CLIENT);
119 LASSERT(name != NULL && name_len != 0 &&
120 buf != NULL && buf_size == name_len + 1);
122 cpy_len = strlcpy(buf, name, buf_size);
124 LASSERT(cpy_len == name_len && lu_name_is_valid_2(buf, cpy_len));
127 void mdc_file_secctx_pack(struct ptlrpc_request *req, const char *secctx_name,
128 const void *secctx, size_t secctx_size)
133 if (secctx_name == NULL)
136 buf = req_capsule_client_get(&req->rq_pill, &RMF_FILE_SECCTX_NAME);
137 buf_size = req_capsule_get_size(&req->rq_pill, &RMF_FILE_SECCTX_NAME,
140 LASSERT(buf_size == strlen(secctx_name) + 1);
141 memcpy(buf, secctx_name, buf_size);
143 buf = req_capsule_client_get(&req->rq_pill, &RMF_FILE_SECCTX);
144 buf_size = req_capsule_get_size(&req->rq_pill, &RMF_FILE_SECCTX,
147 LASSERT(buf_size == secctx_size);
148 memcpy(buf, secctx, buf_size);
151 void mdc_readdir_pack(struct ptlrpc_request *req, __u64 pgoff, size_t size,
152 const struct lu_fid *fid)
154 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
157 b->mbo_valid |= OBD_MD_FLID;
158 b->mbo_size = pgoff; /* !! */
159 b->mbo_nlink = size; /* !! */
160 __mdc_pack_body(b, -1);
161 b->mbo_mode = LUDA_FID | LUDA_TYPE;
164 /* packing of MDS records */
165 void mdc_create_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
166 const void *data, size_t datalen, umode_t mode,
167 uid_t uid, gid_t gid, cfs_cap_t cap_effective, __u64 rdev)
169 struct mdt_rec_create *rec;
173 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
174 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
177 rec->cr_opcode = REINT_CREATE;
180 rec->cr_cap = cap_effective;
181 rec->cr_fid1 = op_data->op_fid1;
182 rec->cr_fid2 = op_data->op_fid2;
185 rec->cr_time = op_data->op_mod_time;
186 rec->cr_suppgid1 = op_data->op_suppgids[0];
187 rec->cr_suppgid2 = op_data->op_suppgids[1];
189 if (op_data->op_bias & MDS_CREATE_VOLATILE)
190 flags |= MDS_OPEN_VOLATILE;
191 set_mrc_cr_flags(rec, flags);
192 rec->cr_bias = op_data->op_bias;
193 rec->cr_umask = current_umask();
195 mdc_pack_name(req, &RMF_NAME, op_data->op_name, op_data->op_namelen);
197 tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
198 memcpy(tmp, data, datalen);
201 mdc_file_secctx_pack(req, op_data->op_file_secctx_name,
202 op_data->op_file_secctx,
203 op_data->op_file_secctx_size);
206 static inline __u64 mds_pack_open_flags(__u64 flags)
208 __u64 cr_flags = (flags & (FMODE_READ | FMODE_WRITE |
209 MDS_OPEN_FL_INTERNAL));
212 cr_flags |= MDS_OPEN_CREAT;
214 cr_flags |= MDS_OPEN_EXCL;
216 cr_flags |= MDS_OPEN_TRUNC;
217 if (flags & O_APPEND)
218 cr_flags |= MDS_OPEN_APPEND;
220 cr_flags |= MDS_OPEN_SYNC;
221 if (flags & O_DIRECTORY)
222 cr_flags |= MDS_OPEN_DIRECTORY;
224 if (flags & FMODE_EXEC)
225 cr_flags |= MDS_FMODE_EXEC;
227 if (cl_is_lov_delay_create(flags))
228 cr_flags |= MDS_OPEN_DELAY_CREATE;
230 if (flags & O_NONBLOCK)
231 cr_flags |= MDS_OPEN_NORESTORE;
236 /* packing of MDS records */
237 void mdc_open_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
238 umode_t mode, __u64 rdev, __u64 flags, const void *lmm,
241 struct mdt_rec_create *rec;
245 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
246 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
248 /* XXX do something about time, uid, gid */
249 rec->cr_opcode = REINT_OPEN;
250 rec->cr_fsuid = from_kuid(&init_user_ns, current_fsuid());
251 rec->cr_fsgid = from_kgid(&init_user_ns, current_fsgid());
252 rec->cr_cap = cfs_curproc_cap_pack();
254 cr_flags = mds_pack_open_flags(flags);
256 rec->cr_umask = current_umask();
257 if (op_data != NULL) {
258 rec->cr_fid1 = op_data->op_fid1;
259 rec->cr_fid2 = op_data->op_fid2;
260 rec->cr_time = op_data->op_mod_time;
261 rec->cr_suppgid1 = op_data->op_suppgids[0];
262 rec->cr_suppgid2 = op_data->op_suppgids[1];
263 rec->cr_bias = op_data->op_bias;
264 rec->cr_old_handle = op_data->op_handle;
266 if (op_data->op_name) {
267 mdc_pack_name(req, &RMF_NAME, op_data->op_name,
268 op_data->op_namelen);
270 if (op_data->op_bias & MDS_CREATE_VOLATILE)
271 cr_flags |= MDS_OPEN_VOLATILE;
274 mdc_file_secctx_pack(req, op_data->op_file_secctx_name,
275 op_data->op_file_secctx,
276 op_data->op_file_secctx_size);
280 cr_flags |= MDS_OPEN_HAS_EA;
281 tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
282 memcpy(tmp, lmm, lmmlen);
284 set_mrc_cr_flags(rec, cr_flags);
287 static inline __u64 attr_pack(unsigned int ia_valid) {
290 if (ia_valid & ATTR_MODE)
291 sa_valid |= MDS_ATTR_MODE;
292 if (ia_valid & ATTR_UID)
293 sa_valid |= MDS_ATTR_UID;
294 if (ia_valid & ATTR_GID)
295 sa_valid |= MDS_ATTR_GID;
296 if (ia_valid & ATTR_SIZE)
297 sa_valid |= MDS_ATTR_SIZE;
298 if (ia_valid & ATTR_ATIME)
299 sa_valid |= MDS_ATTR_ATIME;
300 if (ia_valid & ATTR_MTIME)
301 sa_valid |= MDS_ATTR_MTIME;
302 if (ia_valid & ATTR_CTIME)
303 sa_valid |= MDS_ATTR_CTIME;
304 if (ia_valid & ATTR_ATIME_SET)
305 sa_valid |= MDS_ATTR_ATIME_SET;
306 if (ia_valid & ATTR_MTIME_SET)
307 sa_valid |= MDS_ATTR_MTIME_SET;
308 if (ia_valid & ATTR_FORCE)
309 sa_valid |= MDS_ATTR_FORCE;
310 if (ia_valid & ATTR_ATTR_FLAG)
311 sa_valid |= MDS_ATTR_ATTR_FLAG;
312 if (ia_valid & ATTR_KILL_SUID)
313 sa_valid |= MDS_ATTR_KILL_SUID;
314 if (ia_valid & ATTR_KILL_SGID)
315 sa_valid |= MDS_ATTR_KILL_SGID;
316 if (ia_valid & ATTR_CTIME_SET)
317 sa_valid |= MDS_ATTR_CTIME_SET;
318 if (ia_valid & ATTR_FROM_OPEN)
319 sa_valid |= MDS_ATTR_FROM_OPEN;
320 if (ia_valid & ATTR_BLOCKS)
321 sa_valid |= MDS_ATTR_BLOCKS;
322 if (ia_valid & MDS_OPEN_OWNEROVERRIDE)
323 /* NFSD hack (see bug 5781) */
324 sa_valid |= MDS_OPEN_OWNEROVERRIDE;
328 static void mdc_setattr_pack_rec(struct mdt_rec_setattr *rec,
329 struct md_op_data *op_data)
331 rec->sa_opcode = REINT_SETATTR;
332 rec->sa_fsuid = from_kuid(&init_user_ns, current_fsuid());
333 rec->sa_fsgid = from_kgid(&init_user_ns, current_fsgid());
334 rec->sa_cap = cfs_curproc_cap_pack();
335 rec->sa_suppgid = -1;
337 rec->sa_fid = op_data->op_fid1;
338 rec->sa_valid = attr_pack(op_data->op_attr.ia_valid);
339 rec->sa_mode = op_data->op_attr.ia_mode;
340 rec->sa_uid = from_kuid(&init_user_ns, op_data->op_attr.ia_uid);
341 rec->sa_gid = from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
342 rec->sa_size = op_data->op_attr.ia_size;
343 rec->sa_blocks = op_data->op_attr_blocks;
344 rec->sa_atime = LTIME_S(op_data->op_attr.ia_atime);
345 rec->sa_mtime = LTIME_S(op_data->op_attr.ia_mtime);
346 rec->sa_ctime = LTIME_S(op_data->op_attr.ia_ctime);
347 rec->sa_attr_flags = op_data->op_attr_flags;
348 if ((op_data->op_attr.ia_valid & ATTR_GID) &&
349 in_group_p(op_data->op_attr.ia_gid))
351 from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
353 rec->sa_suppgid = op_data->op_suppgids[0];
355 rec->sa_bias = op_data->op_bias;
358 static void mdc_ioepoch_pack(struct mdt_ioepoch *epoch,
359 struct md_op_data *op_data)
361 epoch->mio_handle = op_data->op_handle;
362 epoch->mio_unused1 = 0;
363 epoch->mio_unused2 = 0;
364 epoch->mio_padding = 0;
367 void mdc_setattr_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
368 void *ea, size_t ealen)
370 struct mdt_rec_setattr *rec;
371 struct lov_user_md *lum = NULL;
373 CLASSERT(sizeof(struct mdt_rec_reint) ==sizeof(struct mdt_rec_setattr));
374 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
375 mdc_setattr_pack_rec(rec, op_data);
380 lum = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
381 if (ea == NULL) { /* Remove LOV EA */
382 lum->lmm_magic = cpu_to_le32(LOV_USER_MAGIC_V1);
383 lum->lmm_stripe_size = 0;
384 lum->lmm_stripe_count = 0;
385 lum->lmm_stripe_offset = (typeof(lum->lmm_stripe_offset))(-1);
387 memcpy(lum, ea, ealen);
391 void mdc_unlink_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
393 struct mdt_rec_unlink *rec;
395 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_unlink));
396 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
397 LASSERT(rec != NULL);
399 rec->ul_opcode = op_data->op_cli_flags & CLI_RM_ENTRY ?
400 REINT_RMENTRY : REINT_UNLINK;
401 rec->ul_fsuid = op_data->op_fsuid;
402 rec->ul_fsgid = op_data->op_fsgid;
403 rec->ul_cap = op_data->op_cap;
404 rec->ul_mode = op_data->op_mode;
405 rec->ul_suppgid1= op_data->op_suppgids[0];
406 rec->ul_suppgid2= -1;
407 rec->ul_fid1 = op_data->op_fid1;
408 rec->ul_fid2 = op_data->op_fid2;
409 rec->ul_time = op_data->op_mod_time;
410 rec->ul_bias = op_data->op_bias;
412 mdc_pack_name(req, &RMF_NAME, op_data->op_name, op_data->op_namelen);
415 void mdc_link_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
417 struct mdt_rec_link *rec;
419 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_link));
420 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
421 LASSERT (rec != NULL);
423 rec->lk_opcode = REINT_LINK;
424 rec->lk_fsuid = op_data->op_fsuid;//current->fsuid;
425 rec->lk_fsgid = op_data->op_fsgid;//current->fsgid;
426 rec->lk_cap = op_data->op_cap;//current->cap_effective;
427 rec->lk_suppgid1 = op_data->op_suppgids[0];
428 rec->lk_suppgid2 = op_data->op_suppgids[1];
429 rec->lk_fid1 = op_data->op_fid1;
430 rec->lk_fid2 = op_data->op_fid2;
431 rec->lk_time = op_data->op_mod_time;
432 rec->lk_bias = op_data->op_bias;
434 mdc_pack_name(req, &RMF_NAME, op_data->op_name, op_data->op_namelen);
437 static void mdc_intent_close_pack(struct ptlrpc_request *req,
438 struct md_op_data *op_data)
440 struct close_data *data;
441 struct ldlm_lock *lock;
442 enum mds_op_bias bias = op_data->op_bias;
444 if (!(bias & (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP |
445 MDS_RENAME_MIGRATE)))
448 data = req_capsule_client_get(&req->rq_pill, &RMF_CLOSE_DATA);
449 LASSERT(data != NULL);
451 lock = ldlm_handle2lock(&op_data->op_lease_handle);
453 data->cd_handle = lock->l_remote_handle;
456 ldlm_cli_cancel(&op_data->op_lease_handle, LCF_LOCAL);
458 data->cd_data_version = op_data->op_data_version;
459 data->cd_fid = op_data->op_fid2;
462 void mdc_rename_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
463 const char *old, size_t oldlen,
464 const char *new, size_t newlen)
466 struct mdt_rec_rename *rec;
468 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_rename));
469 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
471 /* XXX do something about time, uid, gid */
472 rec->rn_opcode = op_data->op_cli_flags & CLI_MIGRATE ?
473 REINT_MIGRATE : REINT_RENAME;
474 rec->rn_fsuid = op_data->op_fsuid;
475 rec->rn_fsgid = op_data->op_fsgid;
476 rec->rn_cap = op_data->op_cap;
477 rec->rn_suppgid1 = op_data->op_suppgids[0];
478 rec->rn_suppgid2 = op_data->op_suppgids[1];
479 rec->rn_fid1 = op_data->op_fid1;
480 rec->rn_fid2 = op_data->op_fid2;
481 rec->rn_time = op_data->op_mod_time;
482 rec->rn_mode = op_data->op_mode;
483 rec->rn_bias = op_data->op_bias;
485 mdc_pack_name(req, &RMF_NAME, old, oldlen);
488 mdc_pack_name(req, &RMF_SYMTGT, new, newlen);
490 if (op_data->op_cli_flags & CLI_MIGRATE &&
491 op_data->op_bias & MDS_RENAME_MIGRATE) {
492 struct mdt_ioepoch *epoch;
494 mdc_intent_close_pack(req, op_data);
495 epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
496 mdc_ioepoch_pack(epoch, op_data);
500 void mdc_getattr_pack(struct ptlrpc_request *req, __u64 valid, __u32 flags,
501 struct md_op_data *op_data, size_t ea_size)
503 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
506 b->mbo_valid = valid;
507 if (op_data->op_bias & MDS_CHECK_SPLIT)
508 b->mbo_valid |= OBD_MD_FLCKSPLIT;
509 if (op_data->op_bias & MDS_CROSS_REF)
510 b->mbo_valid |= OBD_MD_FLCROSSREF;
511 b->mbo_eadatasize = ea_size;
512 b->mbo_flags = flags;
513 __mdc_pack_body(b, op_data->op_suppgids[0]);
515 b->mbo_fid1 = op_data->op_fid1;
516 b->mbo_fid2 = op_data->op_fid2;
517 b->mbo_valid |= OBD_MD_FLID;
519 if (op_data->op_name != NULL)
520 mdc_pack_name(req, &RMF_NAME, op_data->op_name,
521 op_data->op_namelen);
524 void mdc_close_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
526 struct mdt_ioepoch *epoch;
527 struct mdt_rec_setattr *rec;
529 epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
530 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
532 mdc_setattr_pack_rec(rec, op_data);
534 * The client will zero out local timestamps when losing the IBITS lock
535 * so any new RPC timestamps will update the client inode's timestamps.
536 * There was a defect on the server side which allowed the atime to be
537 * overwritten by a zeroed-out atime packed into the close RPC.
539 * Proactively clear the MDS_ATTR_ATIME flag in the RPC in this case
540 * to avoid zeroing the atime on old unpatched servers. See LU-8041.
542 if (rec->sa_atime == 0)
543 rec->sa_valid &= ~MDS_ATTR_ATIME;
545 mdc_ioepoch_pack(epoch, op_data);
546 mdc_intent_close_pack(req, op_data);