2 * Server-side procedures for NFSv4.
4 * Copyright (c) 2002 The Regents of the University of Michigan.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <andros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/fs_struct.h>
36 #include <linux/file.h>
37 #include <linux/falloc.h>
38 #include <linux/slab.h>
39 #include <linux/kthread.h>
45 #include "current_stateid.h"
51 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
52 #include <linux/security.h>
55 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
57 struct inode *inode = d_inode(resfh->fh_dentry);
61 status = security_inode_setsecctx(resfh->fh_dentry,
62 label->data, label->len);
67 * XXX: We should really fail the whole open, but we may
68 * already have created a new file, so it may be too
69 * late. For now this seems the least of evils:
71 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
77 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
81 #define NFSDDBG_FACILITY NFSDDBG_PROC
83 static u32 nfsd_attrmask[] = {
84 NFSD_WRITEABLE_ATTRS_WORD0,
85 NFSD_WRITEABLE_ATTRS_WORD1,
86 NFSD_WRITEABLE_ATTRS_WORD2
89 static u32 nfsd41_ex_attrmask[] = {
90 NFSD_SUPPATTR_EXCLCREAT_WORD0,
91 NFSD_SUPPATTR_EXCLCREAT_WORD1,
92 NFSD_SUPPATTR_EXCLCREAT_WORD2
96 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
97 u32 *bmval, u32 *writable)
99 struct dentry *dentry = cstate->current_fh.fh_dentry;
100 struct svc_export *exp = cstate->current_fh.fh_export;
102 if (!nfsd_attrs_supported(cstate->minorversion, bmval))
103 return nfserr_attrnotsupp;
104 if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
105 return nfserr_attrnotsupp;
106 if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
107 !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
108 return nfserr_attrnotsupp;
109 if (writable && !bmval_is_subset(bmval, writable))
111 if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
112 (bmval[1] & FATTR4_WORD1_MODE))
118 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
119 struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
121 __be32 status = nfs_ok;
123 if (open->op_create == NFS4_OPEN_CREATE) {
124 if (open->op_createmode == NFS4_CREATE_UNCHECKED
125 || open->op_createmode == NFS4_CREATE_GUARDED)
126 status = check_attr_support(rqstp, cstate,
127 open->op_bmval, nfsd_attrmask);
128 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
129 status = check_attr_support(rqstp, cstate,
130 open->op_bmval, nfsd41_ex_attrmask);
137 is_create_with_attrs(struct nfsd4_open *open)
139 return open->op_create == NFS4_OPEN_CREATE
140 && (open->op_createmode == NFS4_CREATE_UNCHECKED
141 || open->op_createmode == NFS4_CREATE_GUARDED
142 || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
146 * if error occurs when setting the acl, just clear the acl bit
147 * in the returned attr bitmap.
150 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
151 struct nfs4_acl *acl, u32 *bmval)
155 status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
158 * We should probably fail the whole open at this point,
159 * but we've already created the file, so it's too late;
160 * So this seems the least of evils:
162 bmval[0] &= ~FATTR4_WORD0_ACL;
166 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
169 dget(src->fh_dentry);
171 exp_get(src->fh_export);
176 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
180 if (open->op_truncate &&
181 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
184 accmode |= NFSD_MAY_READ_IF_EXEC;
186 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
187 accmode |= NFSD_MAY_READ;
188 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
189 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
190 if (open->op_share_deny & NFS4_SHARE_DENY_READ)
191 accmode |= NFSD_MAY_WRITE;
193 status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
198 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
200 umode_t mode = d_inode(fh->fh_dentry)->i_mode;
207 * Using err_symlink as our catch-all case may look odd; but
208 * there's no other obvious error for this case in 4.0, and we
209 * happen to know that it will cause the linux v4 client to do
210 * the right thing on attempts to open something other than a
213 return nfserr_symlink;
216 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
218 if (nfsd4_has_session(cstate))
220 fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
225 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
227 struct svc_fh *current_fh = &cstate->current_fh;
231 *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
233 return nfserr_jukebox;
234 fh_init(*resfh, NFS4_FHSIZE);
235 open->op_truncate = 0;
237 if (open->op_create) {
238 /* FIXME: check session persistence and pnfs flags.
239 * The nfsv4.1 spec requires the following semantics:
241 * Persistent | pNFS | Server REQUIRED | Client Allowed
242 * Reply Cache | server | |
243 * -------------+--------+-----------------+--------------------
244 * no | no | EXCLUSIVE4_1 | EXCLUSIVE4_1
246 * | | and EXCLUSIVE4 | or EXCLUSIVE4
248 * no | yes | EXCLUSIVE4_1 | EXCLUSIVE4_1
249 * yes | no | GUARDED4 | GUARDED4
250 * yes | yes | GUARDED4 | GUARDED4
254 * Note: create modes (UNCHECKED,GUARDED...) are the same
255 * in NFSv4 as in v3 except EXCLUSIVE4_1.
257 current->fs->umask = open->op_umask;
258 status = do_nfsd_create(rqstp, current_fh, open->op_fname.data,
259 open->op_fname.len, &open->op_iattr,
260 *resfh, open->op_createmode,
261 (u32 *)open->op_verf.data,
262 &open->op_truncate, &open->op_created);
263 current->fs->umask = 0;
265 if (!status && open->op_label.len)
266 nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
269 * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
270 * use the returned bitmask to indicate which attributes
271 * we used to store the verifier:
273 if (nfsd_create_is_exclusive(open->op_createmode) && status == 0)
274 open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
275 FATTR4_WORD1_TIME_MODIFY);
278 * Note this may exit with the parent still locked.
279 * We will hold the lock until nfsd4_open's final
280 * lookup, to prevent renames or unlinks until we've had
281 * a chance to an acquire a delegation if appropriate.
283 status = nfsd_lookup(rqstp, current_fh,
284 open->op_fname.data, open->op_fname.len, *resfh);
287 status = nfsd_check_obj_isreg(*resfh);
291 if (is_create_with_attrs(open) && open->op_acl != NULL)
292 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
294 nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
295 accmode = NFSD_MAY_NOP;
296 if (open->op_created ||
297 open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
298 accmode |= NFSD_MAY_OWNER_OVERRIDE;
299 status = do_open_permission(rqstp, *resfh, open, accmode);
300 set_change_info(&open->op_cinfo, current_fh);
306 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
308 struct svc_fh *current_fh = &cstate->current_fh;
312 /* We don't know the target directory, and therefore can not
313 * set the change info
316 memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
318 nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
320 open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
321 (open->op_iattr.ia_size == 0);
323 * In the delegation case, the client is telling us about an
324 * open that it *already* performed locally, some time ago. We
325 * should let it succeed now if possible.
327 * In the case of a CLAIM_FH open, on the other hand, the client
328 * may be counting on us to enforce permissions (the Linux 4.1
329 * client uses this for normal opens, for example).
331 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
332 accmode = NFSD_MAY_OWNER_OVERRIDE;
334 status = do_open_permission(rqstp, current_fh, open, accmode);
340 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
342 struct nfsd4_sessionid *sid =
343 (struct nfsd4_sessionid *)session->se_sessionid.data;
345 clid->cl_boot = sid->clientid.cl_boot;
346 clid->cl_id = sid->clientid.cl_id;
350 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
353 struct nfsd4_open *open = &u->open;
355 struct svc_fh *resfh = NULL;
356 struct net *net = SVC_NET(rqstp);
357 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
358 bool reclaim = false;
360 dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
361 (int)open->op_fname.len, open->op_fname.data,
364 /* This check required by spec. */
365 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
368 open->op_created = 0;
371 * Before RECLAIM_COMPLETE done, server should deny new lock
373 if (nfsd4_has_session(cstate) &&
374 !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
375 &cstate->session->se_client->cl_flags) &&
376 open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
379 if (nfsd4_has_session(cstate))
380 copy_clientid(&open->op_clientid, cstate->session);
382 /* check seqid for replay. set nfs4_owner */
383 status = nfsd4_process_open1(cstate, open, nn);
384 if (status == nfserr_replay_me) {
385 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
386 fh_put(&cstate->current_fh);
387 fh_copy_shallow(&cstate->current_fh.fh_handle,
389 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
391 dprintk("nfsd4_open: replay failed"
392 " restoring previous filehandle\n");
394 status = nfserr_replay_me;
398 if (open->op_xdr_error) {
399 status = open->op_xdr_error;
403 status = nfsd4_check_open_attributes(rqstp, cstate, open);
407 /* Openowner is now set, so sequence id will get bumped. Now we need
408 * these checks before we do any creates: */
409 status = nfserr_grace;
410 if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
412 status = nfserr_no_grace;
413 if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
416 switch (open->op_claim_type) {
417 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
418 case NFS4_OPEN_CLAIM_NULL:
419 status = do_open_lookup(rqstp, cstate, open, &resfh);
423 case NFS4_OPEN_CLAIM_PREVIOUS:
424 status = nfs4_check_open_reclaim(&open->op_clientid,
428 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
431 case NFS4_OPEN_CLAIM_FH:
432 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
433 status = do_open_fhandle(rqstp, cstate, open);
436 resfh = &cstate->current_fh;
438 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
439 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
440 dprintk("NFSD: unsupported OPEN claim type %d\n",
441 open->op_claim_type);
442 status = nfserr_notsupp;
445 dprintk("NFSD: Invalid OPEN claim type %d\n",
446 open->op_claim_type);
447 status = nfserr_inval;
451 * nfsd4_process_open2() does the actual opening of the file. If
452 * successful, it (1) truncates the file if open->op_truncate was
453 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
455 status = nfsd4_process_open2(rqstp, resfh, open);
456 WARN(status && open->op_created,
457 "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
458 be32_to_cpu(status));
459 if (reclaim && !status)
460 nn->somebody_reclaimed = true;
462 if (resfh && resfh != &cstate->current_fh) {
463 fh_dup2(&cstate->current_fh, resfh);
467 nfsd4_cleanup_open_state(cstate, open);
468 nfsd4_bump_seqid(cstate, status);
473 * OPEN is the only seqid-mutating operation whose decoding can fail
474 * with a seqid-mutating error (specifically, decoding of user names in
475 * the attributes). Therefore we have to do some processing to look up
476 * the stateowner so that we can bump the seqid.
478 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
480 struct nfsd4_open *open = &op->u.open;
482 if (!seqid_mutating_err(ntohl(op->status)))
484 if (nfsd4_has_session(cstate))
486 open->op_xdr_error = op->status;
487 return nfsd4_open(rqstp, cstate, &op->u);
491 * filehandle-manipulating ops.
494 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
497 u->getfh = &cstate->current_fh;
502 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
505 struct nfsd4_putfh *putfh = &u->putfh;
507 fh_put(&cstate->current_fh);
508 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
509 memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
511 return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
515 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
520 fh_put(&cstate->current_fh);
521 status = exp_pseudoroot(rqstp, &cstate->current_fh);
526 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
529 if (!cstate->save_fh.fh_dentry)
530 return nfserr_restorefh;
532 fh_dup2(&cstate->current_fh, &cstate->save_fh);
533 if (HAS_STATE_ID(cstate, SAVED_STATE_ID_FLAG)) {
534 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
535 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
541 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
544 fh_dup2(&cstate->save_fh, &cstate->current_fh);
545 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG)) {
546 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
547 SET_STATE_ID(cstate, SAVED_STATE_ID_FLAG);
556 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
559 struct nfsd4_access *access = &u->access;
561 if (access->ac_req_access & ~NFS3_ACCESS_FULL)
564 access->ac_resp_access = access->ac_req_access;
565 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
566 &access->ac_supported);
569 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
571 __be32 *verf = (__be32 *)verifier->data;
573 BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
575 nfsd_copy_boot_verifier(verf, net_generic(net, nfsd_net_id));
579 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
582 struct nfsd4_commit *commit = &u->commit;
584 gen_boot_verifier(&commit->co_verf, SVC_NET(rqstp));
585 return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
590 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
593 struct nfsd4_create *create = &u->create;
598 fh_init(&resfh, NFS4_FHSIZE);
600 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
604 status = check_attr_support(rqstp, cstate, create->cr_bmval,
609 current->fs->umask = create->cr_umask;
610 switch (create->cr_type) {
612 status = nfsd_symlink(rqstp, &cstate->current_fh,
613 create->cr_name, create->cr_namelen,
614 create->cr_data, &resfh);
618 status = nfserr_inval;
619 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
620 if (MAJOR(rdev) != create->cr_specdata1 ||
621 MINOR(rdev) != create->cr_specdata2)
623 status = nfsd_create(rqstp, &cstate->current_fh,
624 create->cr_name, create->cr_namelen,
625 &create->cr_iattr, S_IFBLK, rdev, &resfh);
629 status = nfserr_inval;
630 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
631 if (MAJOR(rdev) != create->cr_specdata1 ||
632 MINOR(rdev) != create->cr_specdata2)
634 status = nfsd_create(rqstp, &cstate->current_fh,
635 create->cr_name, create->cr_namelen,
636 &create->cr_iattr,S_IFCHR, rdev, &resfh);
640 status = nfsd_create(rqstp, &cstate->current_fh,
641 create->cr_name, create->cr_namelen,
642 &create->cr_iattr, S_IFSOCK, 0, &resfh);
646 status = nfsd_create(rqstp, &cstate->current_fh,
647 create->cr_name, create->cr_namelen,
648 &create->cr_iattr, S_IFIFO, 0, &resfh);
652 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
653 status = nfsd_create(rqstp, &cstate->current_fh,
654 create->cr_name, create->cr_namelen,
655 &create->cr_iattr, S_IFDIR, 0, &resfh);
659 status = nfserr_badtype;
665 if (create->cr_label.len)
666 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
668 if (create->cr_acl != NULL)
669 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
672 fh_unlock(&cstate->current_fh);
673 set_change_info(&create->cr_cinfo, &cstate->current_fh);
674 fh_dup2(&cstate->current_fh, &resfh);
678 current->fs->umask = 0;
683 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
686 struct nfsd4_getattr *getattr = &u->getattr;
689 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
693 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
696 getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
697 getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
698 getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
700 getattr->ga_fhp = &cstate->current_fh;
705 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
708 struct nfsd4_link *link = &u->link;
711 status = nfsd_link(rqstp, &cstate->current_fh,
712 link->li_name, link->li_namelen, &cstate->save_fh);
714 set_change_info(&link->li_cinfo, &cstate->current_fh);
718 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
720 struct svc_fh tmp_fh;
723 fh_init(&tmp_fh, NFS4_FHSIZE);
724 ret = exp_pseudoroot(rqstp, &tmp_fh);
727 if (tmp_fh.fh_dentry == fh->fh_dentry) {
732 return nfsd_lookup(rqstp, fh, "..", 2, fh);
736 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
739 return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
743 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
746 return nfsd_lookup(rqstp, &cstate->current_fh,
747 u->lookup.lo_name, u->lookup.lo_len,
748 &cstate->current_fh);
752 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
755 struct nfsd4_read *read = &u->read;
759 if (read->rd_offset >= OFFSET_MAX)
762 trace_nfsd_read_start(rqstp, &cstate->current_fh,
763 read->rd_offset, read->rd_length);
766 * If we do a zero copy read, then a client will see read data
767 * that reflects the state of the file *after* performing the
768 * following compound.
770 * To ensure proper ordering, we therefore turn off zero copy if
771 * the client wants us to do more in this compound:
773 if (!nfsd4_last_compound_op(rqstp))
774 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
777 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
778 &read->rd_stateid, RD_STATE,
781 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
786 read->rd_rqstp = rqstp;
787 read->rd_fhp = &cstate->current_fh;
793 nfsd4_read_release(union nfsd4_op_u *u)
796 nfsd_file_put(u->read.rd_nf);
797 trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
798 u->read.rd_offset, u->read.rd_length);
802 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
805 struct nfsd4_readdir *readdir = &u->readdir;
806 u64 cookie = readdir->rd_cookie;
807 static const nfs4_verifier zeroverf;
809 /* no need to check permission - this will be done in nfsd_readdir() */
811 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
814 readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
815 readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
816 readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
818 if ((cookie == 1) || (cookie == 2) ||
819 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
820 return nfserr_bad_cookie;
822 readdir->rd_rqstp = rqstp;
823 readdir->rd_fhp = &cstate->current_fh;
828 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
831 u->readlink.rl_rqstp = rqstp;
832 u->readlink.rl_fhp = &cstate->current_fh;
837 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
840 struct nfsd4_remove *remove = &u->remove;
843 if (opens_in_grace(SVC_NET(rqstp)))
845 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
846 remove->rm_name, remove->rm_namelen);
848 fh_unlock(&cstate->current_fh);
849 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
855 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
858 struct nfsd4_rename *rename = &u->rename;
861 if (opens_in_grace(SVC_NET(rqstp)))
863 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
864 rename->rn_snamelen, &cstate->current_fh,
865 rename->rn_tname, rename->rn_tnamelen);
868 set_change_info(&rename->rn_sinfo, &cstate->current_fh);
869 set_change_info(&rename->rn_tinfo, &cstate->save_fh);
874 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
877 struct nfsd4_secinfo *secinfo = &u->secinfo;
878 struct svc_export *exp;
879 struct dentry *dentry;
882 err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
885 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
886 secinfo->si_name, secinfo->si_namelen,
890 fh_unlock(&cstate->current_fh);
891 if (d_really_is_negative(dentry)) {
895 secinfo->si_exp = exp;
897 if (cstate->minorversion)
898 /* See rfc 5661 section 2.6.3.1.1.8 */
899 fh_put(&cstate->current_fh);
904 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
909 switch (u->secinfo_no_name.sin_style) {
910 case NFS4_SECINFO_STYLE4_CURRENT_FH:
912 case NFS4_SECINFO_STYLE4_PARENT:
913 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
921 u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
922 fh_put(&cstate->current_fh);
927 nfsd4_secinfo_release(union nfsd4_op_u *u)
929 if (u->secinfo.si_exp)
930 exp_put(u->secinfo.si_exp);
934 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
936 if (u->secinfo_no_name.sin_exp)
937 exp_put(u->secinfo_no_name.sin_exp);
941 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
944 struct nfsd4_setattr *setattr = &u->setattr;
945 __be32 status = nfs_ok;
948 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
949 status = nfs4_preprocess_stateid_op(rqstp, cstate,
950 &cstate->current_fh, &setattr->sa_stateid,
953 dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
957 err = fh_want_write(&cstate->current_fh);
959 return nfserrno(err);
962 status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
967 if (setattr->sa_acl != NULL)
968 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
972 if (setattr->sa_label.len)
973 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
977 status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
980 fh_drop_write(&cstate->current_fh);
985 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
988 struct nfsd4_write *write = &u->write;
989 stateid_t *stateid = &write->wr_stateid;
990 struct nfsd_file *nf = NULL;
991 __be32 status = nfs_ok;
995 if (write->wr_offset >= OFFSET_MAX)
998 cnt = write->wr_buflen;
999 trace_nfsd_write_start(rqstp, &cstate->current_fh,
1000 write->wr_offset, cnt);
1001 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1002 stateid, WR_STATE, &nf);
1004 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1008 write->wr_how_written = write->wr_stable_how;
1009 gen_boot_verifier(&write->wr_verifier, SVC_NET(rqstp));
1011 nvecs = svc_fill_write_vector(rqstp, write->wr_pagelist,
1012 &write->wr_head, write->wr_buflen);
1013 WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1015 status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf->nf_file,
1016 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1017 write->wr_how_written);
1020 write->wr_bytes_written = cnt;
1021 trace_nfsd_write_done(rqstp, &cstate->current_fh,
1022 write->wr_offset, cnt);
1027 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1028 stateid_t *src_stateid, struct nfsd_file **src,
1029 stateid_t *dst_stateid, struct nfsd_file **dst)
1033 if (!cstate->save_fh.fh_dentry)
1034 return nfserr_nofilehandle;
1036 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1037 src_stateid, RD_STATE, src);
1039 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1043 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1044 dst_stateid, WR_STATE, dst);
1046 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1050 /* fix up for NFS-specific error code */
1051 if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1052 !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1053 status = nfserr_wrong_type;
1060 nfsd_file_put(*dst);
1062 nfsd_file_put(*src);
1067 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1068 union nfsd4_op_u *u)
1070 struct nfsd4_clone *clone = &u->clone;
1071 struct nfsd_file *src, *dst;
1074 status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1075 &clone->cl_dst_stateid, &dst);
1079 status = nfsd4_clone_file_range(src->nf_file, clone->cl_src_pos,
1080 dst->nf_file, clone->cl_dst_pos, clone->cl_count);
1088 void nfs4_put_copy(struct nfsd4_copy *copy)
1090 if (!refcount_dec_and_test(©->refcount))
1096 check_and_set_stop_copy(struct nfsd4_copy *copy)
1100 spin_lock(©->cp_clp->async_lock);
1101 value = copy->stopped;
1103 copy->stopped = true;
1104 spin_unlock(©->cp_clp->async_lock);
1108 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1110 /* only 1 thread should stop the copy */
1111 if (!check_and_set_stop_copy(copy))
1112 kthread_stop(copy->copy_task);
1113 nfs4_put_copy(copy);
1116 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1118 struct nfsd4_copy *copy = NULL;
1120 spin_lock(&clp->async_lock);
1121 if (!list_empty(&clp->async_copies)) {
1122 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1124 refcount_inc(©->refcount);
1126 spin_unlock(&clp->async_lock);
1130 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1132 struct nfsd4_copy *copy;
1134 while ((copy = nfsd4_get_copy(clp)) != NULL)
1135 nfsd4_stop_copy(copy);
1138 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1140 struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1142 nfs4_put_copy(copy);
1145 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1146 struct rpc_task *task)
1151 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1152 .release = nfsd4_cb_offload_release,
1153 .done = nfsd4_cb_offload_done
1156 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1158 copy->cp_res.wr_stable_how = NFS_UNSTABLE;
1159 copy->cp_synchronous = sync;
1160 gen_boot_verifier(©->cp_res.wr_verifier, copy->cp_clp->net);
1163 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1165 ssize_t bytes_copied = 0;
1166 size_t bytes_total = copy->cp_count;
1167 u64 src_pos = copy->cp_src_pos;
1168 u64 dst_pos = copy->cp_dst_pos;
1171 if (kthread_should_stop())
1173 bytes_copied = nfsd_copy_file_range(copy->nf_src->nf_file,
1174 src_pos, copy->nf_dst->nf_file, dst_pos,
1176 if (bytes_copied <= 0)
1178 bytes_total -= bytes_copied;
1179 copy->cp_res.wr_bytes_written += bytes_copied;
1180 src_pos += bytes_copied;
1181 dst_pos += bytes_copied;
1182 } while (bytes_total > 0 && !copy->cp_synchronous);
1183 return bytes_copied;
1186 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1191 bytes = _nfsd_copy_file_range(copy);
1192 /* for async copy, we ignore the error, client can always retry
1195 if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1196 status = nfserrno(bytes);
1198 nfsd4_init_copy_res(copy, sync);
1202 nfsd_file_put(copy->nf_src);
1203 nfsd_file_put(copy->nf_dst);
1207 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1209 dst->cp_src_pos = src->cp_src_pos;
1210 dst->cp_dst_pos = src->cp_dst_pos;
1211 dst->cp_count = src->cp_count;
1212 dst->cp_synchronous = src->cp_synchronous;
1213 memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1214 memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1215 dst->cp_clp = src->cp_clp;
1216 dst->nf_dst = nfsd_file_get(src->nf_dst);
1217 dst->nf_src = nfsd_file_get(src->nf_src);
1218 memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1221 static void cleanup_async_copy(struct nfsd4_copy *copy)
1223 nfs4_free_cp_state(copy);
1224 nfsd_file_put(copy->nf_dst);
1225 nfsd_file_put(copy->nf_src);
1226 spin_lock(©->cp_clp->async_lock);
1227 list_del(©->copies);
1228 spin_unlock(©->cp_clp->async_lock);
1229 nfs4_put_copy(copy);
1232 static int nfsd4_do_async_copy(void *data)
1234 struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1235 struct nfsd4_copy *cb_copy;
1237 copy->nfserr = nfsd4_do_copy(copy, 0);
1238 cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1241 memcpy(&cb_copy->cp_res, ©->cp_res, sizeof(copy->cp_res));
1242 cb_copy->cp_clp = copy->cp_clp;
1243 cb_copy->nfserr = copy->nfserr;
1244 memcpy(&cb_copy->fh, ©->fh, sizeof(copy->fh));
1245 nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1246 &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1247 nfsd4_run_cb(&cb_copy->cp_cb);
1249 cleanup_async_copy(copy);
1254 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1255 union nfsd4_op_u *u)
1257 struct nfsd4_copy *copy = &u->copy;
1259 struct nfsd4_copy *async_copy = NULL;
1261 status = nfsd4_verify_copy(rqstp, cstate, ©->cp_src_stateid,
1262 ©->nf_src, ©->cp_dst_stateid,
1267 copy->cp_clp = cstate->clp;
1268 memcpy(©->fh, &cstate->current_fh.fh_handle,
1269 sizeof(struct knfsd_fh));
1270 if (!copy->cp_synchronous) {
1271 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1273 status = nfserrno(-ENOMEM);
1274 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1277 if (!nfs4_init_cp_state(nn, copy)) {
1281 refcount_set(&async_copy->refcount, 1);
1282 memcpy(©->cp_res.cb_stateid, ©->cp_stateid,
1283 sizeof(copy->cp_stateid));
1284 dup_copy_fields(copy, async_copy);
1285 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1286 async_copy, "%s", "copy thread");
1287 if (IS_ERR(async_copy->copy_task))
1289 spin_lock(&async_copy->cp_clp->async_lock);
1290 list_add(&async_copy->copies,
1291 &async_copy->cp_clp->async_copies);
1292 spin_unlock(&async_copy->cp_clp->async_lock);
1293 wake_up_process(async_copy->copy_task);
1296 status = nfsd4_do_copy(copy, 1);
1300 cleanup_async_copy(async_copy);
1305 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1307 struct nfsd4_copy *copy;
1309 spin_lock(&clp->async_lock);
1310 list_for_each_entry(copy, &clp->async_copies, copies) {
1311 if (memcmp(©->cp_stateid, stateid, NFS4_STATEID_SIZE))
1313 refcount_inc(©->refcount);
1314 spin_unlock(&clp->async_lock);
1317 spin_unlock(&clp->async_lock);
1322 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1323 struct nfsd4_compound_state *cstate,
1324 union nfsd4_op_u *u)
1326 struct nfsd4_offload_status *os = &u->offload_status;
1328 struct nfsd4_copy *copy;
1329 struct nfs4_client *clp = cstate->clp;
1331 copy = find_async_copy(clp, &os->stateid);
1333 nfsd4_stop_copy(copy);
1335 status = nfserr_bad_stateid;
1341 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1342 struct nfsd4_fallocate *fallocate, int flags)
1345 struct nfsd_file *nf;
1347 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1348 &fallocate->falloc_stateid,
1350 if (status != nfs_ok) {
1351 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1355 status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1356 fallocate->falloc_offset,
1357 fallocate->falloc_length,
1363 nfsd4_offload_status(struct svc_rqst *rqstp,
1364 struct nfsd4_compound_state *cstate,
1365 union nfsd4_op_u *u)
1367 struct nfsd4_offload_status *os = &u->offload_status;
1369 struct nfsd4_copy *copy;
1370 struct nfs4_client *clp = cstate->clp;
1372 copy = find_async_copy(clp, &os->stateid);
1374 os->count = copy->cp_res.wr_bytes_written;
1375 nfs4_put_copy(copy);
1377 status = nfserr_bad_stateid;
1383 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1384 union nfsd4_op_u *u)
1386 return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1390 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1391 union nfsd4_op_u *u)
1393 return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1394 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1398 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1399 union nfsd4_op_u *u)
1401 struct nfsd4_seek *seek = &u->seek;
1404 struct nfsd_file *nf;
1406 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1407 &seek->seek_stateid,
1410 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1414 switch (seek->seek_whence) {
1415 case NFS4_CONTENT_DATA:
1418 case NFS4_CONTENT_HOLE:
1422 status = nfserr_union_notsupp;
1427 * Note: This call does change file->f_pos, but nothing in NFSD
1428 * should ever file->f_pos.
1430 seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
1431 if (seek->seek_pos < 0)
1432 status = nfserrno(seek->seek_pos);
1433 else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
1434 seek->seek_eof = true;
1441 /* This routine never returns NFS_OK! If there are no other errors, it
1442 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1443 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
1444 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1447 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1448 struct nfsd4_verify *verify)
1454 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1458 status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1462 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1463 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1464 return nfserr_inval;
1465 if (verify->ve_attrlen & 3)
1466 return nfserr_inval;
1469 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1471 count = 4 + (verify->ve_attrlen >> 2);
1472 buf = kmalloc(count << 2, GFP_KERNEL);
1474 return nfserr_jukebox;
1477 status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1478 cstate->current_fh.fh_export,
1479 cstate->current_fh.fh_dentry,
1483 * If nfsd4_encode_fattr() ran out of space, assume that's because
1484 * the attributes are longer (hence different) than those given:
1486 if (status == nfserr_resource)
1487 status = nfserr_not_same;
1492 p = buf + 1 + ntohl(buf[0]);
1493 status = nfserr_not_same;
1494 if (ntohl(*p++) != verify->ve_attrlen)
1496 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1497 status = nfserr_same;
1505 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1506 union nfsd4_op_u *u)
1510 status = _nfsd4_verify(rqstp, cstate, &u->verify);
1511 return status == nfserr_not_same ? nfs_ok : status;
1515 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1516 union nfsd4_op_u *u)
1520 status = _nfsd4_verify(rqstp, cstate, &u->nverify);
1521 return status == nfserr_same ? nfs_ok : status;
1524 #ifdef CONFIG_NFSD_PNFS
1525 static const struct nfsd4_layout_ops *
1526 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
1528 if (!exp->ex_layout_types) {
1529 dprintk("%s: export does not support pNFS\n", __func__);
1533 if (layout_type >= LAYOUT_TYPE_MAX ||
1534 !(exp->ex_layout_types & (1 << layout_type))) {
1535 dprintk("%s: layout type %d not supported\n",
1536 __func__, layout_type);
1540 return nfsd4_layout_ops[layout_type];
1544 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
1545 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1547 struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
1548 const struct nfsd4_layout_ops *ops;
1549 struct nfsd4_deviceid_map *map;
1550 struct svc_export *exp;
1553 dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
1555 gdp->gd_layout_type,
1556 gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
1559 map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
1561 dprintk("%s: couldn't find device ID to export mapping!\n",
1563 return nfserr_noent;
1566 exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
1568 dprintk("%s: could not find device id\n", __func__);
1569 return nfserr_noent;
1572 nfserr = nfserr_layoutunavailable;
1573 ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
1578 if (gdp->gd_maxcount != 0) {
1579 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
1580 rqstp, cstate->session->se_client, gdp);
1583 gdp->gd_notify_types &= ops->notify_types;
1590 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
1592 kfree(u->getdeviceinfo.gd_device);
1596 nfsd4_layoutget(struct svc_rqst *rqstp,
1597 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1599 struct nfsd4_layoutget *lgp = &u->layoutget;
1600 struct svc_fh *current_fh = &cstate->current_fh;
1601 const struct nfsd4_layout_ops *ops;
1602 struct nfs4_layout_stateid *ls;
1604 int accmode = NFSD_MAY_READ_IF_EXEC;
1606 switch (lgp->lg_seg.iomode) {
1608 accmode |= NFSD_MAY_READ;
1611 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
1614 dprintk("%s: invalid iomode %d\n",
1615 __func__, lgp->lg_seg.iomode);
1616 nfserr = nfserr_badiomode;
1620 nfserr = fh_verify(rqstp, current_fh, 0, accmode);
1624 nfserr = nfserr_layoutunavailable;
1625 ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
1630 * Verify minlength and range as per RFC5661:
1631 * o If loga_length is less than loga_minlength,
1632 * the metadata server MUST return NFS4ERR_INVAL.
1633 * o If the sum of loga_offset and loga_minlength exceeds
1634 * NFS4_UINT64_MAX, and loga_minlength is not
1635 * NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
1636 * o If the sum of loga_offset and loga_length exceeds
1637 * NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
1638 * the error NFS4ERR_INVAL MUST result.
1640 nfserr = nfserr_inval;
1641 if (lgp->lg_seg.length < lgp->lg_minlength ||
1642 (lgp->lg_minlength != NFS4_MAX_UINT64 &&
1643 lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
1644 (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
1645 lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
1647 if (lgp->lg_seg.length == 0)
1650 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
1651 true, lgp->lg_layout_type, &ls);
1653 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
1657 nfserr = nfserr_recallconflict;
1658 if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
1661 nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
1666 nfserr = nfsd4_insert_layout(lgp, ls);
1669 mutex_unlock(&ls->ls_mutex);
1670 nfs4_put_stid(&ls->ls_stid);
1676 nfsd4_layoutget_release(union nfsd4_op_u *u)
1678 kfree(u->layoutget.lg_content);
1682 nfsd4_layoutcommit(struct svc_rqst *rqstp,
1683 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1685 struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
1686 const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
1687 struct svc_fh *current_fh = &cstate->current_fh;
1688 const struct nfsd4_layout_ops *ops;
1689 loff_t new_size = lcp->lc_last_wr + 1;
1690 struct inode *inode;
1691 struct nfs4_layout_stateid *ls;
1694 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
1698 nfserr = nfserr_layoutunavailable;
1699 ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
1702 inode = d_inode(current_fh->fh_dentry);
1704 nfserr = nfserr_inval;
1705 if (new_size <= seg->offset) {
1706 dprintk("pnfsd: last write before layout segment\n");
1709 if (new_size > seg->offset + seg->length) {
1710 dprintk("pnfsd: last write beyond layout segment\n");
1713 if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
1714 dprintk("pnfsd: layoutcommit beyond EOF\n");
1718 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
1719 false, lcp->lc_layout_type,
1722 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
1723 /* fixup error code as per RFC5661 */
1724 if (nfserr == nfserr_bad_stateid)
1725 nfserr = nfserr_badlayout;
1729 /* LAYOUTCOMMIT does not require any serialization */
1730 mutex_unlock(&ls->ls_mutex);
1732 if (new_size > i_size_read(inode)) {
1733 lcp->lc_size_chg = 1;
1734 lcp->lc_newsize = new_size;
1736 lcp->lc_size_chg = 0;
1739 nfserr = ops->proc_layoutcommit(inode, lcp);
1740 nfs4_put_stid(&ls->ls_stid);
1746 nfsd4_layoutreturn(struct svc_rqst *rqstp,
1747 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1749 struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
1750 struct svc_fh *current_fh = &cstate->current_fh;
1753 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
1757 nfserr = nfserr_layoutunavailable;
1758 if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
1761 switch (lrp->lr_seg.iomode) {
1767 dprintk("%s: invalid iomode %d\n", __func__,
1768 lrp->lr_seg.iomode);
1769 nfserr = nfserr_inval;
1773 switch (lrp->lr_return_type) {
1775 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
1779 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
1782 dprintk("%s: invalid return_type %d\n", __func__,
1783 lrp->lr_return_type);
1784 nfserr = nfserr_inval;
1790 #endif /* CONFIG_NFSD_PNFS */
1796 nfsd4_proc_null(struct svc_rqst *rqstp)
1801 static inline void nfsd4_increment_op_stats(u32 opnum)
1803 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
1804 nfsdstats.nfs4_opcount[opnum]++;
1807 static const struct nfsd4_operation nfsd4_ops[];
1809 static const char *nfsd4_op_name(unsigned opnum);
1812 * Enforce NFSv4.1 COMPOUND ordering rules:
1814 * Also note, enforced elsewhere:
1815 * - SEQUENCE other than as first op results in
1816 * NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
1817 * - BIND_CONN_TO_SESSION must be the only op in its compound.
1818 * (Enforced in nfsd4_bind_conn_to_session().)
1819 * - DESTROY_SESSION must be the final operation in a compound, if
1820 * sessionid's in SEQUENCE and DESTROY_SESSION are the same.
1821 * (Enforced in nfsd4_destroy_session().)
1823 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
1825 struct nfsd4_op *first_op = &args->ops[0];
1827 /* These ordering requirements don't apply to NFSv4.0: */
1828 if (args->minorversion == 0)
1830 /* This is weird, but OK, not our problem: */
1831 if (args->opcnt == 0)
1833 if (first_op->status == nfserr_op_illegal)
1835 if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
1836 return nfserr_op_not_in_session;
1837 if (first_op->opnum == OP_SEQUENCE)
1840 * So first_op is something allowed outside a session, like
1841 * EXCHANGE_ID; but then it has to be the only op in the
1844 if (args->opcnt != 1)
1845 return nfserr_not_only_op;
1849 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
1851 return &nfsd4_ops[op->opnum];
1854 bool nfsd4_cache_this_op(struct nfsd4_op *op)
1856 if (op->opnum == OP_ILLEGAL)
1858 return OPDESC(op)->op_flags & OP_CACHEME;
1861 static bool need_wrongsec_check(struct svc_rqst *rqstp)
1863 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1864 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1865 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
1866 struct nfsd4_op *next = &argp->ops[resp->opcnt];
1867 const struct nfsd4_operation *thisd = OPDESC(this);
1868 const struct nfsd4_operation *nextd;
1871 * Most ops check wronsec on our own; only the putfh-like ops
1872 * have special rules.
1874 if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
1877 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
1878 * put-filehandle operation if we're not going to use the
1881 if (argp->opcnt == resp->opcnt)
1883 if (next->opnum == OP_ILLEGAL)
1885 nextd = OPDESC(next);
1887 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
1888 * errors themselves as necessary; others should check for them
1891 return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
1894 static void svcxdr_init_encode(struct svc_rqst *rqstp,
1895 struct nfsd4_compoundres *resp)
1897 struct xdr_stream *xdr = &resp->xdr;
1898 struct xdr_buf *buf = &rqstp->rq_res;
1899 struct kvec *head = buf->head;
1903 xdr->p = head->iov_base + head->iov_len;
1904 xdr->end = head->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
1905 /* Tail and page_len should be zero at this point: */
1906 buf->len = buf->head[0].iov_len;
1907 xdr->scratch.iov_len = 0;
1908 xdr->page_ptr = buf->pages - 1;
1909 buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages)
1910 - rqstp->rq_auth_slack;
1917 nfsd4_proc_compound(struct svc_rqst *rqstp)
1919 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1920 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1921 struct nfsd4_op *op;
1922 struct nfsd4_compound_state *cstate = &resp->cstate;
1923 struct svc_fh *current_fh = &cstate->current_fh;
1924 struct svc_fh *save_fh = &cstate->save_fh;
1925 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1928 svcxdr_init_encode(rqstp, resp);
1929 resp->tagp = resp->xdr.p;
1930 /* reserve space for: taglen, tag, and opcnt */
1931 xdr_reserve_space(&resp->xdr, 8 + args->taglen);
1932 resp->taglen = args->taglen;
1933 resp->tag = args->tag;
1934 resp->rqstp = rqstp;
1935 cstate->minorversion = args->minorversion;
1936 fh_init(current_fh, NFS4_FHSIZE);
1937 fh_init(save_fh, NFS4_FHSIZE);
1939 * Don't use the deferral mechanism for NFSv4; compounds make it
1940 * too hard to avoid non-idempotency problems.
1942 clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1945 * According to RFC3010, this takes precedence over all other errors.
1947 status = nfserr_minor_vers_mismatch;
1948 if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
1950 status = nfserr_resource;
1951 if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
1954 status = nfs41_check_op_ordering(args);
1957 op->status = status;
1962 trace_nfsd_compound(rqstp, args->opcnt);
1963 while (!status && resp->opcnt < args->opcnt) {
1964 op = &args->ops[resp->opcnt++];
1967 * The XDR decode routines may have pre-set op->status;
1968 * for example, if there is a miscellaneous XDR error
1969 * it will be set to nfserr_bad_xdr.
1972 if (op->opnum == OP_OPEN)
1973 op->status = nfsd4_open_omfg(rqstp, cstate, op);
1977 if (!current_fh->fh_dentry) {
1978 if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
1979 op->status = nfserr_nofilehandle;
1982 } else if (current_fh->fh_export->ex_fslocs.migrated &&
1983 !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
1984 op->status = nfserr_moved;
1988 fh_clear_wcc(current_fh);
1990 /* If op is non-idempotent */
1991 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
1993 * Don't execute this op if we couldn't encode a
1996 u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
1998 * Plus if there's another operation, make sure
1999 * we'll have space to at least encode an error:
2001 if (resp->opcnt < args->opcnt)
2002 plen += COMPOUND_ERR_SLACK_SPACE;
2003 op->status = nfsd4_check_resp_size(resp, plen);
2009 if (op->opdesc->op_get_currentstateid)
2010 op->opdesc->op_get_currentstateid(cstate, &op->u);
2011 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2013 /* Only from SEQUENCE */
2014 if (cstate->status == nfserr_replay_cache) {
2015 dprintk("%s NFS4.1 replay from cache\n", __func__);
2016 status = op->status;
2020 if (op->opdesc->op_set_currentstateid)
2021 op->opdesc->op_set_currentstateid(cstate, &op->u);
2023 if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2024 clear_current_stateid(cstate);
2026 if (need_wrongsec_check(rqstp))
2027 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2030 if (op->status == nfserr_replay_me) {
2031 op->replay = &cstate->replay_owner->so_replay;
2032 nfsd4_encode_replay(&resp->xdr, op);
2033 status = op->status = op->replay->rp_status;
2035 nfsd4_encode_operation(resp, op);
2036 status = op->status;
2039 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2040 nfsd4_op_name(op->opnum));
2042 nfsd4_cstate_clear_replay(cstate);
2043 nfsd4_increment_op_stats(op->opnum);
2046 cstate->status = status;
2049 BUG_ON(cstate->replay_owner);
2051 /* Reset deferral mechanism for RPC deferrals */
2052 set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2053 dprintk("nfsv4 compound returned %d\n", ntohl(status));
2057 #define op_encode_hdr_size (2)
2058 #define op_encode_stateid_maxsz (XDR_QUADLEN(NFS4_STATEID_SIZE))
2059 #define op_encode_verifier_maxsz (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2060 #define op_encode_change_info_maxsz (5)
2061 #define nfs4_fattr_bitmap_maxsz (4)
2063 /* We'll fall back on returning no lockowner if run out of space: */
2064 #define op_encode_lockowner_maxsz (0)
2065 #define op_encode_lock_denied_maxsz (8 + op_encode_lockowner_maxsz)
2067 #define nfs4_owner_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2069 #define op_encode_ace_maxsz (3 + nfs4_owner_maxsz)
2070 #define op_encode_delegation_maxsz (1 + op_encode_stateid_maxsz + 1 + \
2071 op_encode_ace_maxsz)
2073 #define op_encode_channel_attrs_maxsz (6 + 1 + 1)
2075 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2077 return (op_encode_hdr_size) * sizeof(__be32);
2080 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2082 return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2085 static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2087 /* ac_supported, ac_resp_access */
2088 return (op_encode_hdr_size + 2)* sizeof(__be32);
2091 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2093 return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2096 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2098 return (op_encode_hdr_size + op_encode_change_info_maxsz
2099 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2103 * Note since this is an idempotent operation we won't insist on failing
2104 * the op prematurely if the estimate is too large. We may turn off splice
2105 * reads unnecessarily.
2107 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2108 struct nfsd4_op *op)
2110 u32 *bmap = op->u.getattr.ga_bmval;
2111 u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2114 if (bmap0 & FATTR4_WORD0_ACL)
2115 return svc_max_payload(rqstp);
2116 if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2117 return svc_max_payload(rqstp);
2119 if (bmap1 & FATTR4_WORD1_OWNER) {
2120 ret += IDMAP_NAMESZ + 4;
2121 bmap1 &= ~FATTR4_WORD1_OWNER;
2123 if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2124 ret += IDMAP_NAMESZ + 4;
2125 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2127 if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2128 ret += NFS4_FHSIZE + 4;
2129 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2131 if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2132 ret += NFS4_MAXLABELLEN + 12;
2133 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2136 * Largest of remaining attributes are 16 bytes (e.g.,
2137 * supported_attributes)
2139 ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2140 /* bitmask, length */
2145 static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2147 return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2150 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2152 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2156 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2158 return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2162 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2164 return (op_encode_hdr_size + op_encode_stateid_maxsz
2165 + op_encode_change_info_maxsz + 1
2166 + nfs4_fattr_bitmap_maxsz
2167 + op_encode_delegation_maxsz) * sizeof(__be32);
2170 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2172 u32 maxcount = 0, rlen = 0;
2174 maxcount = svc_max_payload(rqstp);
2175 rlen = min(op->u.read.rd_length, maxcount);
2177 return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2180 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2182 u32 maxcount = 0, rlen = 0;
2184 maxcount = svc_max_payload(rqstp);
2185 rlen = min(op->u.readdir.rd_maxcount, maxcount);
2187 return (op_encode_hdr_size + op_encode_verifier_maxsz +
2188 XDR_QUADLEN(rlen)) * sizeof(__be32);
2191 static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2193 return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2196 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2198 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2202 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2204 return (op_encode_hdr_size + op_encode_change_info_maxsz
2205 + op_encode_change_info_maxsz) * sizeof(__be32);
2208 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2209 struct nfsd4_op *op)
2211 return (op_encode_hdr_size
2212 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2215 static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2217 return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2221 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2223 return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2226 static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2228 return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2229 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2232 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2234 return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2238 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2240 return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2243 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2245 return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2246 1 + 1 + /* eir_flags, spr_how */\
2247 4 + /* spo_must_enforce & _allow with bitmap */\
2248 2 + /*eir_server_owner.so_minor_id */\
2249 /* eir_server_owner.so_major_id<> */\
2250 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2251 /* eir_server_scope<> */\
2252 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2253 1 + /* eir_server_impl_id array length */\
2254 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2257 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2259 return (op_encode_hdr_size + \
2260 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2261 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2264 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2266 return (op_encode_hdr_size + \
2267 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2268 2 + /* csr_sequence, csr_flags */\
2269 op_encode_channel_attrs_maxsz + \
2270 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2273 static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2275 return (op_encode_hdr_size +
2276 1 /* wr_callback */ +
2277 op_encode_stateid_maxsz /* wr_callback */ +
2279 1 /* wr_committed */ +
2280 op_encode_verifier_maxsz +
2281 1 /* cr_consecutive */ +
2282 1 /* cr_synchronous */) * sizeof(__be32);
2285 static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
2286 struct nfsd4_op *op)
2288 return (op_encode_hdr_size +
2290 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
2293 #ifdef CONFIG_NFSD_PNFS
2294 static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2296 u32 maxcount = 0, rlen = 0;
2298 maxcount = svc_max_payload(rqstp);
2299 rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
2301 return (op_encode_hdr_size +
2302 1 /* gd_layout_type*/ +
2304 2 /* gd_notify_types */) * sizeof(__be32);
2308 * At this stage we don't really know what layout driver will handle the request,
2309 * so we need to define an arbitrary upper bound here.
2311 #define MAX_LAYOUT_SIZE 128
2312 static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2314 return (op_encode_hdr_size +
2315 1 /* logr_return_on_close */ +
2316 op_encode_stateid_maxsz +
2317 1 /* nr of layouts */ +
2318 MAX_LAYOUT_SIZE) * sizeof(__be32);
2321 static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2323 return (op_encode_hdr_size +
2324 1 /* locr_newsize */ +
2325 2 /* ns_size */) * sizeof(__be32);
2328 static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2330 return (op_encode_hdr_size +
2331 1 /* lrs_stateid */ +
2332 op_encode_stateid_maxsz) * sizeof(__be32);
2334 #endif /* CONFIG_NFSD_PNFS */
2337 static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2339 return (op_encode_hdr_size + 3) * sizeof(__be32);
2342 static const struct nfsd4_operation nfsd4_ops[] = {
2344 .op_func = nfsd4_access,
2345 .op_name = "OP_ACCESS",
2346 .op_rsize_bop = nfsd4_access_rsize,
2349 .op_func = nfsd4_close,
2350 .op_flags = OP_MODIFIES_SOMETHING,
2351 .op_name = "OP_CLOSE",
2352 .op_rsize_bop = nfsd4_status_stateid_rsize,
2353 .op_get_currentstateid = nfsd4_get_closestateid,
2354 .op_set_currentstateid = nfsd4_set_closestateid,
2357 .op_func = nfsd4_commit,
2358 .op_flags = OP_MODIFIES_SOMETHING,
2359 .op_name = "OP_COMMIT",
2360 .op_rsize_bop = nfsd4_commit_rsize,
2363 .op_func = nfsd4_create,
2364 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
2365 .op_name = "OP_CREATE",
2366 .op_rsize_bop = nfsd4_create_rsize,
2368 [OP_DELEGRETURN] = {
2369 .op_func = nfsd4_delegreturn,
2370 .op_flags = OP_MODIFIES_SOMETHING,
2371 .op_name = "OP_DELEGRETURN",
2372 .op_rsize_bop = nfsd4_only_status_rsize,
2373 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
2376 .op_func = nfsd4_getattr,
2377 .op_flags = ALLOWED_ON_ABSENT_FS,
2378 .op_rsize_bop = nfsd4_getattr_rsize,
2379 .op_name = "OP_GETATTR",
2382 .op_func = nfsd4_getfh,
2383 .op_name = "OP_GETFH",
2384 .op_rsize_bop = nfsd4_getfh_rsize,
2387 .op_func = nfsd4_link,
2388 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
2390 .op_name = "OP_LINK",
2391 .op_rsize_bop = nfsd4_link_rsize,
2394 .op_func = nfsd4_lock,
2395 .op_flags = OP_MODIFIES_SOMETHING |
2396 OP_NONTRIVIAL_ERROR_ENCODE,
2397 .op_name = "OP_LOCK",
2398 .op_rsize_bop = nfsd4_lock_rsize,
2399 .op_set_currentstateid = nfsd4_set_lockstateid,
2402 .op_func = nfsd4_lockt,
2403 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
2404 .op_name = "OP_LOCKT",
2405 .op_rsize_bop = nfsd4_lock_rsize,
2408 .op_func = nfsd4_locku,
2409 .op_flags = OP_MODIFIES_SOMETHING,
2410 .op_name = "OP_LOCKU",
2411 .op_rsize_bop = nfsd4_status_stateid_rsize,
2412 .op_get_currentstateid = nfsd4_get_lockustateid,
2415 .op_func = nfsd4_lookup,
2416 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2417 .op_name = "OP_LOOKUP",
2418 .op_rsize_bop = nfsd4_only_status_rsize,
2421 .op_func = nfsd4_lookupp,
2422 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2423 .op_name = "OP_LOOKUPP",
2424 .op_rsize_bop = nfsd4_only_status_rsize,
2427 .op_func = nfsd4_nverify,
2428 .op_name = "OP_NVERIFY",
2429 .op_rsize_bop = nfsd4_only_status_rsize,
2432 .op_func = nfsd4_open,
2433 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2434 .op_name = "OP_OPEN",
2435 .op_rsize_bop = nfsd4_open_rsize,
2436 .op_set_currentstateid = nfsd4_set_openstateid,
2438 [OP_OPEN_CONFIRM] = {
2439 .op_func = nfsd4_open_confirm,
2440 .op_flags = OP_MODIFIES_SOMETHING,
2441 .op_name = "OP_OPEN_CONFIRM",
2442 .op_rsize_bop = nfsd4_status_stateid_rsize,
2444 [OP_OPEN_DOWNGRADE] = {
2445 .op_func = nfsd4_open_downgrade,
2446 .op_flags = OP_MODIFIES_SOMETHING,
2447 .op_name = "OP_OPEN_DOWNGRADE",
2448 .op_rsize_bop = nfsd4_status_stateid_rsize,
2449 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
2450 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
2453 .op_func = nfsd4_putfh,
2454 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2455 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2456 .op_name = "OP_PUTFH",
2457 .op_rsize_bop = nfsd4_only_status_rsize,
2460 .op_func = nfsd4_putrootfh,
2461 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2462 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2463 .op_name = "OP_PUTPUBFH",
2464 .op_rsize_bop = nfsd4_only_status_rsize,
2467 .op_func = nfsd4_putrootfh,
2468 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2469 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2470 .op_name = "OP_PUTROOTFH",
2471 .op_rsize_bop = nfsd4_only_status_rsize,
2474 .op_func = nfsd4_read,
2475 .op_release = nfsd4_read_release,
2476 .op_name = "OP_READ",
2477 .op_rsize_bop = nfsd4_read_rsize,
2478 .op_get_currentstateid = nfsd4_get_readstateid,
2481 .op_func = nfsd4_readdir,
2482 .op_name = "OP_READDIR",
2483 .op_rsize_bop = nfsd4_readdir_rsize,
2486 .op_func = nfsd4_readlink,
2487 .op_name = "OP_READLINK",
2488 .op_rsize_bop = nfsd4_readlink_rsize,
2491 .op_func = nfsd4_remove,
2492 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2493 .op_name = "OP_REMOVE",
2494 .op_rsize_bop = nfsd4_remove_rsize,
2497 .op_func = nfsd4_rename,
2498 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2499 .op_name = "OP_RENAME",
2500 .op_rsize_bop = nfsd4_rename_rsize,
2503 .op_func = nfsd4_renew,
2504 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2505 | OP_MODIFIES_SOMETHING,
2506 .op_name = "OP_RENEW",
2507 .op_rsize_bop = nfsd4_only_status_rsize,
2511 .op_func = nfsd4_restorefh,
2512 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2513 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
2514 .op_name = "OP_RESTOREFH",
2515 .op_rsize_bop = nfsd4_only_status_rsize,
2518 .op_func = nfsd4_savefh,
2519 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2520 .op_name = "OP_SAVEFH",
2521 .op_rsize_bop = nfsd4_only_status_rsize,
2524 .op_func = nfsd4_secinfo,
2525 .op_release = nfsd4_secinfo_release,
2526 .op_flags = OP_HANDLES_WRONGSEC,
2527 .op_name = "OP_SECINFO",
2528 .op_rsize_bop = nfsd4_secinfo_rsize,
2531 .op_func = nfsd4_setattr,
2532 .op_name = "OP_SETATTR",
2533 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
2534 | OP_NONTRIVIAL_ERROR_ENCODE,
2535 .op_rsize_bop = nfsd4_setattr_rsize,
2536 .op_get_currentstateid = nfsd4_get_setattrstateid,
2538 [OP_SETCLIENTID] = {
2539 .op_func = nfsd4_setclientid,
2540 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2541 | OP_MODIFIES_SOMETHING | OP_CACHEME
2542 | OP_NONTRIVIAL_ERROR_ENCODE,
2543 .op_name = "OP_SETCLIENTID",
2544 .op_rsize_bop = nfsd4_setclientid_rsize,
2546 [OP_SETCLIENTID_CONFIRM] = {
2547 .op_func = nfsd4_setclientid_confirm,
2548 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2549 | OP_MODIFIES_SOMETHING | OP_CACHEME,
2550 .op_name = "OP_SETCLIENTID_CONFIRM",
2551 .op_rsize_bop = nfsd4_only_status_rsize,
2554 .op_func = nfsd4_verify,
2555 .op_name = "OP_VERIFY",
2556 .op_rsize_bop = nfsd4_only_status_rsize,
2559 .op_func = nfsd4_write,
2560 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2561 .op_name = "OP_WRITE",
2562 .op_rsize_bop = nfsd4_write_rsize,
2563 .op_get_currentstateid = nfsd4_get_writestateid,
2565 [OP_RELEASE_LOCKOWNER] = {
2566 .op_func = nfsd4_release_lockowner,
2567 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2568 | OP_MODIFIES_SOMETHING,
2569 .op_name = "OP_RELEASE_LOCKOWNER",
2570 .op_rsize_bop = nfsd4_only_status_rsize,
2573 /* NFSv4.1 operations */
2574 [OP_EXCHANGE_ID] = {
2575 .op_func = nfsd4_exchange_id,
2576 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2577 | OP_MODIFIES_SOMETHING,
2578 .op_name = "OP_EXCHANGE_ID",
2579 .op_rsize_bop = nfsd4_exchange_id_rsize,
2581 [OP_BACKCHANNEL_CTL] = {
2582 .op_func = nfsd4_backchannel_ctl,
2583 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2584 .op_name = "OP_BACKCHANNEL_CTL",
2585 .op_rsize_bop = nfsd4_only_status_rsize,
2587 [OP_BIND_CONN_TO_SESSION] = {
2588 .op_func = nfsd4_bind_conn_to_session,
2589 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2590 | OP_MODIFIES_SOMETHING,
2591 .op_name = "OP_BIND_CONN_TO_SESSION",
2592 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
2594 [OP_CREATE_SESSION] = {
2595 .op_func = nfsd4_create_session,
2596 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2597 | OP_MODIFIES_SOMETHING,
2598 .op_name = "OP_CREATE_SESSION",
2599 .op_rsize_bop = nfsd4_create_session_rsize,
2601 [OP_DESTROY_SESSION] = {
2602 .op_func = nfsd4_destroy_session,
2603 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2604 | OP_MODIFIES_SOMETHING,
2605 .op_name = "OP_DESTROY_SESSION",
2606 .op_rsize_bop = nfsd4_only_status_rsize,
2609 .op_func = nfsd4_sequence,
2610 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
2611 .op_name = "OP_SEQUENCE",
2612 .op_rsize_bop = nfsd4_sequence_rsize,
2614 [OP_DESTROY_CLIENTID] = {
2615 .op_func = nfsd4_destroy_clientid,
2616 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2617 | OP_MODIFIES_SOMETHING,
2618 .op_name = "OP_DESTROY_CLIENTID",
2619 .op_rsize_bop = nfsd4_only_status_rsize,
2621 [OP_RECLAIM_COMPLETE] = {
2622 .op_func = nfsd4_reclaim_complete,
2623 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2624 .op_name = "OP_RECLAIM_COMPLETE",
2625 .op_rsize_bop = nfsd4_only_status_rsize,
2627 [OP_SECINFO_NO_NAME] = {
2628 .op_func = nfsd4_secinfo_no_name,
2629 .op_release = nfsd4_secinfo_no_name_release,
2630 .op_flags = OP_HANDLES_WRONGSEC,
2631 .op_name = "OP_SECINFO_NO_NAME",
2632 .op_rsize_bop = nfsd4_secinfo_rsize,
2634 [OP_TEST_STATEID] = {
2635 .op_func = nfsd4_test_stateid,
2636 .op_flags = ALLOWED_WITHOUT_FH,
2637 .op_name = "OP_TEST_STATEID",
2638 .op_rsize_bop = nfsd4_test_stateid_rsize,
2640 [OP_FREE_STATEID] = {
2641 .op_func = nfsd4_free_stateid,
2642 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2643 .op_name = "OP_FREE_STATEID",
2644 .op_get_currentstateid = nfsd4_get_freestateid,
2645 .op_rsize_bop = nfsd4_only_status_rsize,
2647 #ifdef CONFIG_NFSD_PNFS
2648 [OP_GETDEVICEINFO] = {
2649 .op_func = nfsd4_getdeviceinfo,
2650 .op_release = nfsd4_getdeviceinfo_release,
2651 .op_flags = ALLOWED_WITHOUT_FH,
2652 .op_name = "OP_GETDEVICEINFO",
2653 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
2656 .op_func = nfsd4_layoutget,
2657 .op_release = nfsd4_layoutget_release,
2658 .op_flags = OP_MODIFIES_SOMETHING,
2659 .op_name = "OP_LAYOUTGET",
2660 .op_rsize_bop = nfsd4_layoutget_rsize,
2662 [OP_LAYOUTCOMMIT] = {
2663 .op_func = nfsd4_layoutcommit,
2664 .op_flags = OP_MODIFIES_SOMETHING,
2665 .op_name = "OP_LAYOUTCOMMIT",
2666 .op_rsize_bop = nfsd4_layoutcommit_rsize,
2668 [OP_LAYOUTRETURN] = {
2669 .op_func = nfsd4_layoutreturn,
2670 .op_flags = OP_MODIFIES_SOMETHING,
2671 .op_name = "OP_LAYOUTRETURN",
2672 .op_rsize_bop = nfsd4_layoutreturn_rsize,
2674 #endif /* CONFIG_NFSD_PNFS */
2676 /* NFSv4.2 operations */
2678 .op_func = nfsd4_allocate,
2679 .op_flags = OP_MODIFIES_SOMETHING,
2680 .op_name = "OP_ALLOCATE",
2681 .op_rsize_bop = nfsd4_only_status_rsize,
2684 .op_func = nfsd4_deallocate,
2685 .op_flags = OP_MODIFIES_SOMETHING,
2686 .op_name = "OP_DEALLOCATE",
2687 .op_rsize_bop = nfsd4_only_status_rsize,
2690 .op_func = nfsd4_clone,
2691 .op_flags = OP_MODIFIES_SOMETHING,
2692 .op_name = "OP_CLONE",
2693 .op_rsize_bop = nfsd4_only_status_rsize,
2696 .op_func = nfsd4_copy,
2697 .op_flags = OP_MODIFIES_SOMETHING,
2698 .op_name = "OP_COPY",
2699 .op_rsize_bop = nfsd4_copy_rsize,
2702 .op_func = nfsd4_seek,
2703 .op_name = "OP_SEEK",
2704 .op_rsize_bop = nfsd4_seek_rsize,
2706 [OP_OFFLOAD_STATUS] = {
2707 .op_func = nfsd4_offload_status,
2708 .op_name = "OP_OFFLOAD_STATUS",
2709 .op_rsize_bop = nfsd4_offload_status_rsize,
2711 [OP_OFFLOAD_CANCEL] = {
2712 .op_func = nfsd4_offload_cancel,
2713 .op_flags = OP_MODIFIES_SOMETHING,
2714 .op_name = "OP_OFFLOAD_CANCEL",
2715 .op_rsize_bop = nfsd4_only_status_rsize,
2720 * nfsd4_spo_must_allow - Determine if the compound op contains an
2721 * operation that is allowed to be sent with machine credentials
2723 * @rqstp: a pointer to the struct svc_rqst
2725 * Checks to see if the compound contains a spo_must_allow op
2726 * and confirms that it was sent with the proper machine creds.
2729 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
2731 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2732 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2733 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2734 struct nfsd4_compound_state *cstate = &resp->cstate;
2735 struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
2738 if (!cstate->minorversion)
2741 if (cstate->spo_must_allowed == true)
2744 opiter = resp->opcnt;
2745 while (opiter < argp->opcnt) {
2746 this = &argp->ops[opiter++];
2747 if (test_bit(this->opnum, allow->u.longs) &&
2748 cstate->clp->cl_mach_cred &&
2749 nfsd4_mach_creds_match(cstate->clp, rqstp)) {
2750 cstate->spo_must_allowed = true;
2754 cstate->spo_must_allowed = false;
2758 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
2760 if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
2761 return op_encode_hdr_size * sizeof(__be32);
2763 BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
2764 return OPDESC(op)->op_rsize_bop(rqstp, op);
2767 void warn_on_nonidempotent_op(struct nfsd4_op *op)
2769 if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
2770 pr_err("unable to encode reply to nonidempotent op %d (%s)\n",
2771 op->opnum, nfsd4_op_name(op->opnum));
2776 static const char *nfsd4_op_name(unsigned opnum)
2778 if (opnum < ARRAY_SIZE(nfsd4_ops))
2779 return nfsd4_ops[opnum].op_name;
2780 return "unknown_operation";
2783 #define nfsd4_voidres nfsd4_voidargs
2784 struct nfsd4_voidargs { int dummy; };
2786 static const struct svc_procedure nfsd_procedures4[2] = {
2788 .pc_func = nfsd4_proc_null,
2789 .pc_encode = nfs4svc_encode_voidres,
2790 .pc_argsize = sizeof(struct nfsd4_voidargs),
2791 .pc_ressize = sizeof(struct nfsd4_voidres),
2792 .pc_cachetype = RC_NOCACHE,
2795 [NFSPROC4_COMPOUND] = {
2796 .pc_func = nfsd4_proc_compound,
2797 .pc_decode = nfs4svc_decode_compoundargs,
2798 .pc_encode = nfs4svc_encode_compoundres,
2799 .pc_argsize = sizeof(struct nfsd4_compoundargs),
2800 .pc_ressize = sizeof(struct nfsd4_compoundres),
2801 .pc_release = nfsd4_release_compoundargs,
2802 .pc_cachetype = RC_NOCACHE,
2803 .pc_xdrressize = NFSD_BUFSIZE/4,
2807 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
2808 const struct svc_version nfsd_version4 = {
2811 .vs_proc = nfsd_procedures4,
2812 .vs_count = nfsd_count3,
2813 .vs_dispatch = nfsd_dispatch,
2814 .vs_xdrsize = NFS4_SVC_XDRSIZE,
2815 .vs_rpcb_optnl = true,
2816 .vs_need_cong_ctrl = true,