2 * Server-side procedures for NFSv4.
4 * Copyright (c) 2002 The Regents of the University of Michigan.
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/file.h>
36 #include <linux/slab.h>
42 #include "current_stateid.h"
46 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
47 #include <linux/security.h>
50 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
52 struct inode *inode = resfh->fh_dentry->d_inode;
55 mutex_lock(&inode->i_mutex);
56 status = security_inode_setsecctx(resfh->fh_dentry,
57 label->data, label->len);
58 mutex_unlock(&inode->i_mutex);
62 * XXX: We should really fail the whole open, but we may
63 * already have created a new file, so it may be too
64 * late. For now this seems the least of evils:
66 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
72 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
76 #define NFSDDBG_FACILITY NFSDDBG_PROC
78 static u32 nfsd_attrmask[] = {
79 NFSD_WRITEABLE_ATTRS_WORD0,
80 NFSD_WRITEABLE_ATTRS_WORD1,
81 NFSD_WRITEABLE_ATTRS_WORD2
84 static u32 nfsd41_ex_attrmask[] = {
85 NFSD_SUPPATTR_EXCLCREAT_WORD0,
86 NFSD_SUPPATTR_EXCLCREAT_WORD1,
87 NFSD_SUPPATTR_EXCLCREAT_WORD2
91 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
92 u32 *bmval, u32 *writable)
94 struct dentry *dentry = cstate->current_fh.fh_dentry;
97 * Check about attributes are supported by the NFSv4 server or not.
98 * According to spec, unsupported attributes return ERR_ATTRNOTSUPP.
100 if ((bmval[0] & ~nfsd_suppattrs0(cstate->minorversion)) ||
101 (bmval[1] & ~nfsd_suppattrs1(cstate->minorversion)) ||
102 (bmval[2] & ~nfsd_suppattrs2(cstate->minorversion)))
103 return nfserr_attrnotsupp;
106 * Check FATTR4_WORD0_ACL can be supported
107 * in current environment or not.
109 if (bmval[0] & FATTR4_WORD0_ACL) {
110 if (!IS_POSIXACL(dentry->d_inode))
111 return nfserr_attrnotsupp;
115 * According to spec, read-only attributes return ERR_INVAL.
118 if ((bmval[0] & ~writable[0]) || (bmval[1] & ~writable[1]) ||
119 (bmval[2] & ~writable[2]))
127 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
128 struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
130 __be32 status = nfs_ok;
132 if (open->op_create == NFS4_OPEN_CREATE) {
133 if (open->op_createmode == NFS4_CREATE_UNCHECKED
134 || open->op_createmode == NFS4_CREATE_GUARDED)
135 status = check_attr_support(rqstp, cstate,
136 open->op_bmval, nfsd_attrmask);
137 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
138 status = check_attr_support(rqstp, cstate,
139 open->op_bmval, nfsd41_ex_attrmask);
146 is_create_with_attrs(struct nfsd4_open *open)
148 return open->op_create == NFS4_OPEN_CREATE
149 && (open->op_createmode == NFS4_CREATE_UNCHECKED
150 || open->op_createmode == NFS4_CREATE_GUARDED
151 || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
155 * if error occurs when setting the acl, just clear the acl bit
156 * in the returned attr bitmap.
159 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
160 struct nfs4_acl *acl, u32 *bmval)
164 status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
167 * We should probably fail the whole open at this point,
168 * but we've already created the file, so it's too late;
169 * So this seems the least of evils:
171 bmval[0] &= ~FATTR4_WORD0_ACL;
175 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
178 dget(src->fh_dentry);
180 cache_get(&src->fh_export->h);
185 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
189 if (open->op_truncate &&
190 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
193 accmode |= NFSD_MAY_READ_IF_EXEC;
195 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
196 accmode |= NFSD_MAY_READ;
197 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
198 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
199 if (open->op_share_deny & NFS4_SHARE_DENY_READ)
200 accmode |= NFSD_MAY_WRITE;
202 status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
207 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
209 umode_t mode = fh->fh_dentry->d_inode->i_mode;
216 * Using err_symlink as our catch-all case may look odd; but
217 * there's no other obvious error for this case in 4.0, and we
218 * happen to know that it will cause the linux v4 client to do
219 * the right thing on attempts to open something other than a
222 return nfserr_symlink;
225 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
227 if (nfsd4_has_session(cstate))
229 fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
234 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
236 struct svc_fh *current_fh = &cstate->current_fh;
240 *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
242 return nfserr_jukebox;
243 fh_init(*resfh, NFS4_FHSIZE);
244 open->op_truncate = 0;
246 if (open->op_create) {
247 /* FIXME: check session persistence and pnfs flags.
248 * The nfsv4.1 spec requires the following semantics:
250 * Persistent | pNFS | Server REQUIRED | Client Allowed
251 * Reply Cache | server | |
252 * -------------+--------+-----------------+--------------------
253 * no | no | EXCLUSIVE4_1 | EXCLUSIVE4_1
255 * | | and EXCLUSIVE4 | or EXCLUSIVE4
257 * no | yes | EXCLUSIVE4_1 | EXCLUSIVE4_1
258 * yes | no | GUARDED4 | GUARDED4
259 * yes | yes | GUARDED4 | GUARDED4
263 * Note: create modes (UNCHECKED,GUARDED...) are the same
264 * in NFSv4 as in v3 except EXCLUSIVE4_1.
266 status = do_nfsd_create(rqstp, current_fh, open->op_fname.data,
267 open->op_fname.len, &open->op_iattr,
268 *resfh, open->op_createmode,
269 (u32 *)open->op_verf.data,
270 &open->op_truncate, &open->op_created);
272 if (!status && open->op_label.len)
273 nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
276 * Following rfc 3530 14.2.16, use the returned bitmask
277 * to indicate which attributes we used to store the
280 if (open->op_createmode == NFS4_CREATE_EXCLUSIVE && status == 0)
281 open->op_bmval[1] = (FATTR4_WORD1_TIME_ACCESS |
282 FATTR4_WORD1_TIME_MODIFY);
285 * Note this may exit with the parent still locked.
286 * We will hold the lock until nfsd4_open's final
287 * lookup, to prevent renames or unlinks until we've had
288 * a chance to an acquire a delegation if appropriate.
290 status = nfsd_lookup(rqstp, current_fh,
291 open->op_fname.data, open->op_fname.len, *resfh);
294 status = nfsd_check_obj_isreg(*resfh);
298 if (is_create_with_attrs(open) && open->op_acl != NULL)
299 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
301 nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
302 accmode = NFSD_MAY_NOP;
303 if (open->op_created ||
304 open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
305 accmode |= NFSD_MAY_OWNER_OVERRIDE;
306 status = do_open_permission(rqstp, *resfh, open, accmode);
307 set_change_info(&open->op_cinfo, current_fh);
313 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
315 struct svc_fh *current_fh = &cstate->current_fh;
319 /* We don't know the target directory, and therefore can not
320 * set the change info
323 memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
325 nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
327 open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
328 (open->op_iattr.ia_size == 0);
330 * In the delegation case, the client is telling us about an
331 * open that it *already* performed locally, some time ago. We
332 * should let it succeed now if possible.
334 * In the case of a CLAIM_FH open, on the other hand, the client
335 * may be counting on us to enforce permissions (the Linux 4.1
336 * client uses this for normal opens, for example).
338 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
339 accmode = NFSD_MAY_OWNER_OVERRIDE;
341 status = do_open_permission(rqstp, current_fh, open, accmode);
347 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
349 struct nfsd4_sessionid *sid =
350 (struct nfsd4_sessionid *)session->se_sessionid.data;
352 clid->cl_boot = sid->clientid.cl_boot;
353 clid->cl_id = sid->clientid.cl_id;
357 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
358 struct nfsd4_open *open)
361 struct svc_fh *resfh = NULL;
362 struct nfsd4_compoundres *resp;
363 struct net *net = SVC_NET(rqstp);
364 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
366 dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
367 (int)open->op_fname.len, open->op_fname.data,
370 /* This check required by spec. */
371 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
374 open->op_created = 0;
377 * Before RECLAIM_COMPLETE done, server should deny new lock
379 if (nfsd4_has_session(cstate) &&
380 !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
381 &cstate->session->se_client->cl_flags) &&
382 open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
385 if (nfsd4_has_session(cstate))
386 copy_clientid(&open->op_clientid, cstate->session);
390 /* check seqid for replay. set nfs4_owner */
391 resp = rqstp->rq_resp;
392 status = nfsd4_process_open1(&resp->cstate, open, nn);
393 if (status == nfserr_replay_me) {
394 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
395 fh_put(&cstate->current_fh);
396 fh_copy_shallow(&cstate->current_fh.fh_handle,
398 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
400 dprintk("nfsd4_open: replay failed"
401 " restoring previous filehandle\n");
403 status = nfserr_replay_me;
407 if (open->op_xdr_error) {
408 status = open->op_xdr_error;
412 status = nfsd4_check_open_attributes(rqstp, cstate, open);
416 /* Openowner is now set, so sequence id will get bumped. Now we need
417 * these checks before we do any creates: */
418 status = nfserr_grace;
419 if (locks_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
421 status = nfserr_no_grace;
422 if (!locks_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
425 switch (open->op_claim_type) {
426 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
427 case NFS4_OPEN_CLAIM_NULL:
428 status = do_open_lookup(rqstp, cstate, open, &resfh);
432 case NFS4_OPEN_CLAIM_PREVIOUS:
433 status = nfs4_check_open_reclaim(&open->op_clientid,
434 cstate->minorversion,
438 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
439 case NFS4_OPEN_CLAIM_FH:
440 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
441 status = do_open_fhandle(rqstp, cstate, open);
444 resfh = &cstate->current_fh;
446 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
447 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
448 dprintk("NFSD: unsupported OPEN claim type %d\n",
449 open->op_claim_type);
450 status = nfserr_notsupp;
453 dprintk("NFSD: Invalid OPEN claim type %d\n",
454 open->op_claim_type);
455 status = nfserr_inval;
459 * nfsd4_process_open2() does the actual opening of the file. If
460 * successful, it (1) truncates the file if open->op_truncate was
461 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
463 status = nfsd4_process_open2(rqstp, resfh, open);
464 WARN_ON(status && open->op_created);
466 if (resfh && resfh != &cstate->current_fh) {
467 fh_dup2(&cstate->current_fh, resfh);
471 nfsd4_cleanup_open_state(open, status);
472 if (open->op_openowner && !nfsd4_has_session(cstate))
473 cstate->replay_owner = &open->op_openowner->oo_owner;
474 nfsd4_bump_seqid(cstate, status);
475 if (!cstate->replay_owner)
481 * OPEN is the only seqid-mutating operation whose decoding can fail
482 * with a seqid-mutating error (specifically, decoding of user names in
483 * the attributes). Therefore we have to do some processing to look up
484 * the stateowner so that we can bump the seqid.
486 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
488 struct nfsd4_open *open = (struct nfsd4_open *)&op->u;
490 if (!seqid_mutating_err(ntohl(op->status)))
492 if (nfsd4_has_session(cstate))
494 open->op_xdr_error = op->status;
495 return nfsd4_open(rqstp, cstate, open);
499 * filehandle-manipulating ops.
502 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
503 struct svc_fh **getfh)
505 if (!cstate->current_fh.fh_dentry)
506 return nfserr_nofilehandle;
508 *getfh = &cstate->current_fh;
513 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
514 struct nfsd4_putfh *putfh)
516 fh_put(&cstate->current_fh);
517 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
518 memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
520 return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
524 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
529 fh_put(&cstate->current_fh);
530 status = exp_pseudoroot(rqstp, &cstate->current_fh);
535 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
538 if (!cstate->save_fh.fh_dentry)
539 return nfserr_restorefh;
541 fh_dup2(&cstate->current_fh, &cstate->save_fh);
542 if (HAS_STATE_ID(cstate, SAVED_STATE_ID_FLAG)) {
543 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
544 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
550 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
553 if (!cstate->current_fh.fh_dentry)
554 return nfserr_nofilehandle;
556 fh_dup2(&cstate->save_fh, &cstate->current_fh);
557 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG)) {
558 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
559 SET_STATE_ID(cstate, SAVED_STATE_ID_FLAG);
568 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
569 struct nfsd4_access *access)
571 if (access->ac_req_access & ~NFS3_ACCESS_FULL)
574 access->ac_resp_access = access->ac_req_access;
575 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
576 &access->ac_supported);
579 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
582 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
584 verf[0] = (__be32)nn->nfssvc_boot.tv_sec;
585 verf[1] = (__be32)nn->nfssvc_boot.tv_usec;
586 memcpy(verifier->data, verf, sizeof(verifier->data));
590 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
591 struct nfsd4_commit *commit)
593 gen_boot_verifier(&commit->co_verf, SVC_NET(rqstp));
594 return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
599 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
600 struct nfsd4_create *create)
606 fh_init(&resfh, NFS4_FHSIZE);
608 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR,
613 status = check_attr_support(rqstp, cstate, create->cr_bmval,
618 switch (create->cr_type) {
620 status = nfsd_symlink(rqstp, &cstate->current_fh,
621 create->cr_name, create->cr_namelen,
622 create->cr_linkname, create->cr_linklen,
623 &resfh, &create->cr_iattr);
627 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
628 if (MAJOR(rdev) != create->cr_specdata1 ||
629 MINOR(rdev) != create->cr_specdata2)
631 status = nfsd_create(rqstp, &cstate->current_fh,
632 create->cr_name, create->cr_namelen,
633 &create->cr_iattr, S_IFBLK, rdev, &resfh);
637 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
638 if (MAJOR(rdev) != create->cr_specdata1 ||
639 MINOR(rdev) != create->cr_specdata2)
641 status = nfsd_create(rqstp, &cstate->current_fh,
642 create->cr_name, create->cr_namelen,
643 &create->cr_iattr,S_IFCHR, rdev, &resfh);
647 status = nfsd_create(rqstp, &cstate->current_fh,
648 create->cr_name, create->cr_namelen,
649 &create->cr_iattr, S_IFSOCK, 0, &resfh);
653 status = nfsd_create(rqstp, &cstate->current_fh,
654 create->cr_name, create->cr_namelen,
655 &create->cr_iattr, S_IFIFO, 0, &resfh);
659 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
660 status = nfsd_create(rqstp, &cstate->current_fh,
661 create->cr_name, create->cr_namelen,
662 &create->cr_iattr, S_IFDIR, 0, &resfh);
666 status = nfserr_badtype;
672 if (create->cr_label.len)
673 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
675 if (create->cr_acl != NULL)
676 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
679 fh_unlock(&cstate->current_fh);
680 set_change_info(&create->cr_cinfo, &cstate->current_fh);
681 fh_dup2(&cstate->current_fh, &resfh);
688 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
689 struct nfsd4_getattr *getattr)
693 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
697 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
700 getattr->ga_bmval[0] &= nfsd_suppattrs0(cstate->minorversion);
701 getattr->ga_bmval[1] &= nfsd_suppattrs1(cstate->minorversion);
702 getattr->ga_bmval[2] &= nfsd_suppattrs2(cstate->minorversion);
704 getattr->ga_fhp = &cstate->current_fh;
709 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
710 struct nfsd4_link *link)
712 __be32 status = nfserr_nofilehandle;
714 if (!cstate->save_fh.fh_dentry)
716 status = nfsd_link(rqstp, &cstate->current_fh,
717 link->li_name, link->li_namelen, &cstate->save_fh);
719 set_change_info(&link->li_cinfo, &cstate->current_fh);
723 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
725 struct svc_fh tmp_fh;
728 fh_init(&tmp_fh, NFS4_FHSIZE);
729 ret = exp_pseudoroot(rqstp, &tmp_fh);
732 if (tmp_fh.fh_dentry == fh->fh_dentry) {
737 return nfsd_lookup(rqstp, fh, "..", 2, fh);
741 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
744 return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
748 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
749 struct nfsd4_lookup *lookup)
751 return nfsd_lookup(rqstp, &cstate->current_fh,
752 lookup->lo_name, lookup->lo_len,
753 &cstate->current_fh);
757 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
758 struct nfsd4_read *read)
762 /* no need to check permission - this will be done in nfsd_read() */
764 read->rd_filp = NULL;
765 if (read->rd_offset >= OFFSET_MAX)
769 * If we do a zero copy read, then a client will see read data
770 * that reflects the state of the file *after* performing the
771 * following compound.
773 * To ensure proper ordering, we therefore turn off zero copy if
774 * the client wants us to do more in this compound:
776 if (!nfsd4_last_compound_op(rqstp))
777 rqstp->rq_splice_ok = false;
780 if ((status = nfs4_preprocess_stateid_op(SVC_NET(rqstp),
781 cstate, &read->rd_stateid,
782 RD_STATE, &read->rd_filp))) {
783 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
788 read->rd_rqstp = rqstp;
789 read->rd_fhp = &cstate->current_fh;
794 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
795 struct nfsd4_readdir *readdir)
797 u64 cookie = readdir->rd_cookie;
798 static const nfs4_verifier zeroverf;
800 /* no need to check permission - this will be done in nfsd_readdir() */
802 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
805 readdir->rd_bmval[0] &= nfsd_suppattrs0(cstate->minorversion);
806 readdir->rd_bmval[1] &= nfsd_suppattrs1(cstate->minorversion);
807 readdir->rd_bmval[2] &= nfsd_suppattrs2(cstate->minorversion);
809 if ((cookie == 1) || (cookie == 2) ||
810 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
811 return nfserr_bad_cookie;
813 readdir->rd_rqstp = rqstp;
814 readdir->rd_fhp = &cstate->current_fh;
819 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
820 struct nfsd4_readlink *readlink)
822 readlink->rl_rqstp = rqstp;
823 readlink->rl_fhp = &cstate->current_fh;
828 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
829 struct nfsd4_remove *remove)
833 if (locks_in_grace(SVC_NET(rqstp)))
835 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
836 remove->rm_name, remove->rm_namelen);
838 fh_unlock(&cstate->current_fh);
839 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
845 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
846 struct nfsd4_rename *rename)
848 __be32 status = nfserr_nofilehandle;
850 if (!cstate->save_fh.fh_dentry)
852 if (locks_in_grace(SVC_NET(rqstp)) &&
853 !(cstate->save_fh.fh_export->ex_flags & NFSEXP_NOSUBTREECHECK))
855 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
856 rename->rn_snamelen, &cstate->current_fh,
857 rename->rn_tname, rename->rn_tnamelen);
860 set_change_info(&rename->rn_sinfo, &cstate->current_fh);
861 set_change_info(&rename->rn_tinfo, &cstate->save_fh);
866 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
867 struct nfsd4_secinfo *secinfo)
870 struct svc_export *exp;
871 struct dentry *dentry;
874 fh_init(&resfh, NFS4_FHSIZE);
875 err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
878 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
879 secinfo->si_name, secinfo->si_namelen,
883 if (dentry->d_inode == NULL) {
887 secinfo->si_exp = exp;
889 if (cstate->minorversion)
890 /* See rfc 5661 section 2.6.3.1.1.8 */
891 fh_put(&cstate->current_fh);
896 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
897 struct nfsd4_secinfo_no_name *sin)
901 switch (sin->sin_style) {
902 case NFS4_SECINFO_STYLE4_CURRENT_FH:
904 case NFS4_SECINFO_STYLE4_PARENT:
905 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
912 exp_get(cstate->current_fh.fh_export);
913 sin->sin_exp = cstate->current_fh.fh_export;
914 fh_put(&cstate->current_fh);
919 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
920 struct nfsd4_setattr *setattr)
922 __be32 status = nfs_ok;
925 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
926 status = nfs4_preprocess_stateid_op(SVC_NET(rqstp), cstate,
927 &setattr->sa_stateid, WR_STATE, NULL);
929 dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
933 err = fh_want_write(&cstate->current_fh);
935 return nfserrno(err);
938 status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
943 if (setattr->sa_acl != NULL)
944 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
948 if (setattr->sa_label.len)
949 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
953 status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
956 fh_drop_write(&cstate->current_fh);
960 static int fill_in_write_vector(struct kvec *vec, struct nfsd4_write *write)
963 int buflen = write->wr_buflen;
965 vec[0].iov_base = write->wr_head.iov_base;
966 vec[0].iov_len = min_t(int, buflen, write->wr_head.iov_len);
967 buflen -= vec[0].iov_len;
970 vec[i].iov_base = page_address(write->wr_pagelist[i - 1]);
971 vec[i].iov_len = min_t(int, PAGE_SIZE, buflen);
972 buflen -= vec[i].iov_len;
979 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
980 struct nfsd4_write *write)
982 stateid_t *stateid = &write->wr_stateid;
983 struct file *filp = NULL;
984 __be32 status = nfs_ok;
988 /* no need to check permission - this will be done in nfsd_write() */
990 if (write->wr_offset >= OFFSET_MAX)
993 status = nfs4_preprocess_stateid_op(SVC_NET(rqstp),
994 cstate, stateid, WR_STATE, &filp);
996 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1000 cnt = write->wr_buflen;
1001 write->wr_how_written = write->wr_stable_how;
1002 gen_boot_verifier(&write->wr_verifier, SVC_NET(rqstp));
1004 nvecs = fill_in_write_vector(rqstp->rq_vec, write);
1005 WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1007 status = nfsd_write(rqstp, &cstate->current_fh, filp,
1008 write->wr_offset, rqstp->rq_vec, nvecs,
1009 &cnt, &write->wr_how_written);
1013 write->wr_bytes_written = cnt;
1018 /* This routine never returns NFS_OK! If there are no other errors, it
1019 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1020 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
1021 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1024 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1025 struct nfsd4_verify *verify)
1031 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1035 status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1039 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1040 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1041 return nfserr_inval;
1042 if (verify->ve_attrlen & 3)
1043 return nfserr_inval;
1046 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1048 count = 4 + (verify->ve_attrlen >> 2);
1049 buf = kmalloc(count << 2, GFP_KERNEL);
1051 return nfserr_jukebox;
1054 status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1055 cstate->current_fh.fh_export,
1056 cstate->current_fh.fh_dentry,
1060 * If nfsd4_encode_fattr() ran out of space, assume that's because
1061 * the attributes are longer (hence different) than those given:
1063 if (status == nfserr_resource)
1064 status = nfserr_not_same;
1069 p = buf + 1 + ntohl(buf[0]);
1070 status = nfserr_not_same;
1071 if (ntohl(*p++) != verify->ve_attrlen)
1073 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1074 status = nfserr_same;
1082 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1083 struct nfsd4_verify *verify)
1087 status = _nfsd4_verify(rqstp, cstate, verify);
1088 return status == nfserr_not_same ? nfs_ok : status;
1092 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1093 struct nfsd4_verify *verify)
1097 status = _nfsd4_verify(rqstp, cstate, verify);
1098 return status == nfserr_same ? nfs_ok : status;
1105 nfsd4_proc_null(struct svc_rqst *rqstp, void *argp, void *resp)
1110 static inline void nfsd4_increment_op_stats(u32 opnum)
1112 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
1113 nfsdstats.nfs4_opcount[opnum]++;
1116 typedef __be32(*nfsd4op_func)(struct svc_rqst *, struct nfsd4_compound_state *,
1118 typedef u32(*nfsd4op_rsize)(struct svc_rqst *, struct nfsd4_op *op);
1119 typedef void(*stateid_setter)(struct nfsd4_compound_state *, void *);
1120 typedef void(*stateid_getter)(struct nfsd4_compound_state *, void *);
1122 enum nfsd4_op_flags {
1123 ALLOWED_WITHOUT_FH = 1 << 0, /* No current filehandle required */
1124 ALLOWED_ON_ABSENT_FS = 1 << 1, /* ops processed on absent fs */
1125 ALLOWED_AS_FIRST_OP = 1 << 2, /* ops reqired first in compound */
1126 /* For rfc 5661 section 2.6.3.1.1: */
1127 OP_HANDLES_WRONGSEC = 1 << 3,
1128 OP_IS_PUTFH_LIKE = 1 << 4,
1130 * These are the ops whose result size we estimate before
1131 * encoding, to avoid performing an op then not being able to
1132 * respond or cache a response. This includes writes and setattrs
1133 * as well as the operations usually called "nonidempotent":
1135 OP_MODIFIES_SOMETHING = 1 << 5,
1137 * Cache compounds containing these ops in the xid-based drc:
1138 * We use the DRC for compounds containing non-idempotent
1139 * operations, *except* those that are 4.1-specific (since
1140 * sessions provide their own EOS), and except for stateful
1141 * operations other than setclientid and setclientid_confirm
1142 * (since sequence numbers provide EOS for open, lock, etc in
1145 OP_CACHEME = 1 << 6,
1147 * These are ops which clear current state id.
1149 OP_CLEAR_STATEID = 1 << 7,
1152 struct nfsd4_operation {
1153 nfsd4op_func op_func;
1156 /* Try to get response size before operation */
1157 nfsd4op_rsize op_rsize_bop;
1158 stateid_getter op_get_currentstateid;
1159 stateid_setter op_set_currentstateid;
1162 static struct nfsd4_operation nfsd4_ops[];
1164 static const char *nfsd4_op_name(unsigned opnum);
1167 * Enforce NFSv4.1 COMPOUND ordering rules:
1169 * Also note, enforced elsewhere:
1170 * - SEQUENCE other than as first op results in
1171 * NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
1172 * - BIND_CONN_TO_SESSION must be the only op in its compound.
1173 * (Enforced in nfsd4_bind_conn_to_session().)
1174 * - DESTROY_SESSION must be the final operation in a compound, if
1175 * sessionid's in SEQUENCE and DESTROY_SESSION are the same.
1176 * (Enforced in nfsd4_destroy_session().)
1178 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
1180 struct nfsd4_op *op = &args->ops[0];
1182 /* These ordering requirements don't apply to NFSv4.0: */
1183 if (args->minorversion == 0)
1185 /* This is weird, but OK, not our problem: */
1186 if (args->opcnt == 0)
1188 if (op->status == nfserr_op_illegal)
1190 if (!(nfsd4_ops[op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
1191 return nfserr_op_not_in_session;
1192 if (op->opnum == OP_SEQUENCE)
1194 if (args->opcnt != 1)
1195 return nfserr_not_only_op;
1199 static inline struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
1201 return &nfsd4_ops[op->opnum];
1204 bool nfsd4_cache_this_op(struct nfsd4_op *op)
1206 if (op->opnum == OP_ILLEGAL)
1208 return OPDESC(op)->op_flags & OP_CACHEME;
1211 static bool need_wrongsec_check(struct svc_rqst *rqstp)
1213 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1214 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1215 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
1216 struct nfsd4_op *next = &argp->ops[resp->opcnt];
1217 struct nfsd4_operation *thisd;
1218 struct nfsd4_operation *nextd;
1220 thisd = OPDESC(this);
1222 * Most ops check wronsec on our own; only the putfh-like ops
1223 * have special rules.
1225 if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
1228 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
1229 * put-filehandle operation if we're not going to use the
1232 if (argp->opcnt == resp->opcnt)
1235 nextd = OPDESC(next);
1237 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
1238 * errors themselves as necessary; others should check for them
1241 return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
1244 static void svcxdr_init_encode(struct svc_rqst *rqstp,
1245 struct nfsd4_compoundres *resp)
1247 struct xdr_stream *xdr = &resp->xdr;
1248 struct xdr_buf *buf = &rqstp->rq_res;
1249 struct kvec *head = buf->head;
1253 xdr->p = head->iov_base + head->iov_len;
1254 xdr->end = head->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
1255 /* Tail and page_len should be zero at this point: */
1256 buf->len = buf->head[0].iov_len;
1257 xdr->scratch.iov_len = 0;
1258 xdr->page_ptr = buf->pages - 1;
1259 buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages)
1260 - rqstp->rq_auth_slack;
1267 nfsd4_proc_compound(struct svc_rqst *rqstp,
1268 struct nfsd4_compoundargs *args,
1269 struct nfsd4_compoundres *resp)
1271 struct nfsd4_op *op;
1272 struct nfsd4_operation *opdesc;
1273 struct nfsd4_compound_state *cstate = &resp->cstate;
1274 struct svc_fh *current_fh = &cstate->current_fh;
1275 struct svc_fh *save_fh = &cstate->save_fh;
1278 svcxdr_init_encode(rqstp, resp);
1279 resp->tagp = resp->xdr.p;
1280 /* reserve space for: taglen, tag, and opcnt */
1281 xdr_reserve_space(&resp->xdr, 8 + args->taglen);
1282 resp->taglen = args->taglen;
1283 resp->tag = args->tag;
1284 resp->rqstp = rqstp;
1285 cstate->minorversion = args->minorversion;
1286 fh_init(current_fh, NFS4_FHSIZE);
1287 fh_init(save_fh, NFS4_FHSIZE);
1289 * Don't use the deferral mechanism for NFSv4; compounds make it
1290 * too hard to avoid non-idempotency problems.
1292 rqstp->rq_usedeferral = 0;
1295 * According to RFC3010, this takes precedence over all other errors.
1297 status = nfserr_minor_vers_mismatch;
1298 if (nfsd_minorversion(args->minorversion, NFSD_TEST) <= 0)
1301 status = nfs41_check_op_ordering(args);
1304 op->status = status;
1308 while (!status && resp->opcnt < args->opcnt) {
1309 op = &args->ops[resp->opcnt++];
1311 dprintk("nfsv4 compound op #%d/%d: %d (%s)\n",
1312 resp->opcnt, args->opcnt, op->opnum,
1313 nfsd4_op_name(op->opnum));
1315 * The XDR decode routines may have pre-set op->status;
1316 * for example, if there is a miscellaneous XDR error
1317 * it will be set to nfserr_bad_xdr.
1320 if (op->opnum == OP_OPEN)
1321 op->status = nfsd4_open_omfg(rqstp, cstate, op);
1325 opdesc = OPDESC(op);
1327 if (!current_fh->fh_dentry) {
1328 if (!(opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
1329 op->status = nfserr_nofilehandle;
1332 } else if (current_fh->fh_export->ex_fslocs.migrated &&
1333 !(opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
1334 op->status = nfserr_moved;
1338 fh_clear_wcc(current_fh);
1340 /* If op is non-idempotent */
1341 if (opdesc->op_flags & OP_MODIFIES_SOMETHING) {
1343 * Don't execute this op if we couldn't encode a
1346 u32 plen = opdesc->op_rsize_bop(rqstp, op);
1348 * Plus if there's another operation, make sure
1349 * we'll have space to at least encode an error:
1351 if (resp->opcnt < args->opcnt)
1352 plen += COMPOUND_ERR_SLACK_SPACE;
1353 op->status = nfsd4_check_resp_size(resp, plen);
1359 if (opdesc->op_get_currentstateid)
1360 opdesc->op_get_currentstateid(cstate, &op->u);
1361 op->status = opdesc->op_func(rqstp, cstate, &op->u);
1364 if (opdesc->op_set_currentstateid)
1365 opdesc->op_set_currentstateid(cstate, &op->u);
1367 if (opdesc->op_flags & OP_CLEAR_STATEID)
1368 clear_current_stateid(cstate);
1370 if (need_wrongsec_check(rqstp))
1371 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
1375 /* Only from SEQUENCE */
1376 if (cstate->status == nfserr_replay_cache) {
1377 dprintk("%s NFS4.1 replay from cache\n", __func__);
1378 status = op->status;
1381 if (op->status == nfserr_replay_me) {
1382 op->replay = &cstate->replay_owner->so_replay;
1383 nfsd4_encode_replay(&resp->xdr, op);
1384 status = op->status = op->replay->rp_status;
1386 nfsd4_encode_operation(resp, op);
1387 status = op->status;
1390 dprintk("nfsv4 compound op %p opcnt %d #%d: %d: status %d\n",
1391 args->ops, args->opcnt, resp->opcnt, op->opnum,
1392 be32_to_cpu(status));
1394 if (cstate->replay_owner) {
1395 nfs4_unlock_state();
1396 cstate->replay_owner = NULL;
1398 /* XXX Ugh, we need to get rid of this kind of special case: */
1399 if (op->opnum == OP_READ && op->u.read.rd_filp)
1400 fput(op->u.read.rd_filp);
1402 nfsd4_increment_op_stats(op->opnum);
1405 cstate->status = status;
1408 BUG_ON(cstate->replay_owner);
1410 /* Reset deferral mechanism for RPC deferrals */
1411 rqstp->rq_usedeferral = 1;
1412 dprintk("nfsv4 compound returned %d\n", ntohl(status));
1416 #define op_encode_hdr_size (2)
1417 #define op_encode_stateid_maxsz (XDR_QUADLEN(NFS4_STATEID_SIZE))
1418 #define op_encode_verifier_maxsz (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
1419 #define op_encode_change_info_maxsz (5)
1420 #define nfs4_fattr_bitmap_maxsz (4)
1422 /* We'll fall back on returning no lockowner if run out of space: */
1423 #define op_encode_lockowner_maxsz (0)
1424 #define op_encode_lock_denied_maxsz (8 + op_encode_lockowner_maxsz)
1426 #define nfs4_owner_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
1428 #define op_encode_ace_maxsz (3 + nfs4_owner_maxsz)
1429 #define op_encode_delegation_maxsz (1 + op_encode_stateid_maxsz + 1 + \
1430 op_encode_ace_maxsz)
1432 #define op_encode_channel_attrs_maxsz (6 + 1 + 1)
1434 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1436 return (op_encode_hdr_size) * sizeof(__be32);
1439 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1441 return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
1444 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1446 return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
1449 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1451 return (op_encode_hdr_size + op_encode_change_info_maxsz
1452 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
1456 * Note since this is an idempotent operation we won't insist on failing
1457 * the op prematurely if the estimate is too large. We may turn off splice
1458 * reads unnecessarily.
1460 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
1461 struct nfsd4_op *op)
1463 u32 *bmap = op->u.getattr.ga_bmval;
1464 u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
1467 if (bmap0 & FATTR4_WORD0_ACL)
1468 return svc_max_payload(rqstp);
1469 if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
1470 return svc_max_payload(rqstp);
1472 if (bmap1 & FATTR4_WORD1_OWNER) {
1473 ret += IDMAP_NAMESZ + 4;
1474 bmap1 &= ~FATTR4_WORD1_OWNER;
1476 if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
1477 ret += IDMAP_NAMESZ + 4;
1478 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
1480 if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
1481 ret += NFS4_FHSIZE + 4;
1482 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
1484 if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
1485 ret += NFSD4_MAX_SEC_LABEL_LEN + 12;
1486 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
1489 * Largest of remaining attributes are 16 bytes (e.g.,
1490 * supported_attributes)
1492 ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
1493 /* bitmask, length */
1498 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1500 return (op_encode_hdr_size + op_encode_change_info_maxsz)
1504 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1506 return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
1510 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1512 return (op_encode_hdr_size + op_encode_stateid_maxsz
1513 + op_encode_change_info_maxsz + 1
1514 + nfs4_fattr_bitmap_maxsz
1515 + op_encode_delegation_maxsz) * sizeof(__be32);
1518 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1520 u32 maxcount = 0, rlen = 0;
1522 maxcount = svc_max_payload(rqstp);
1523 rlen = op->u.read.rd_length;
1525 if (rlen > maxcount)
1528 return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
1531 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1533 u32 maxcount = svc_max_payload(rqstp);
1534 u32 rlen = op->u.readdir.rd_maxcount;
1536 if (rlen > maxcount)
1539 return (op_encode_hdr_size + op_encode_verifier_maxsz +
1540 XDR_QUADLEN(rlen)) * sizeof(__be32);
1543 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1545 return (op_encode_hdr_size + op_encode_change_info_maxsz)
1549 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1551 return (op_encode_hdr_size + op_encode_change_info_maxsz
1552 + op_encode_change_info_maxsz) * sizeof(__be32);
1555 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
1556 struct nfsd4_op *op)
1558 return NFS4_MAX_SESSIONID_LEN + 20;
1561 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1563 return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
1566 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1568 return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
1572 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1574 return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
1577 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1579 return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
1580 1 + 1 + /* eir_flags, spr_how */\
1581 4 + /* spo_must_enforce & _allow with bitmap */\
1582 2 + /*eir_server_owner.so_minor_id */\
1583 /* eir_server_owner.so_major_id<> */\
1584 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
1585 /* eir_server_scope<> */\
1586 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
1587 1 + /* eir_server_impl_id array length */\
1588 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
1591 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1593 return (op_encode_hdr_size + \
1594 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
1595 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
1598 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1600 return (op_encode_hdr_size + \
1601 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
1602 2 + /* csr_sequence, csr_flags */\
1603 op_encode_channel_attrs_maxsz + \
1604 op_encode_channel_attrs_maxsz) * sizeof(__be32);
1607 static struct nfsd4_operation nfsd4_ops[] = {
1609 .op_func = (nfsd4op_func)nfsd4_access,
1610 .op_name = "OP_ACCESS",
1613 .op_func = (nfsd4op_func)nfsd4_close,
1614 .op_flags = OP_MODIFIES_SOMETHING,
1615 .op_name = "OP_CLOSE",
1616 .op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1617 .op_get_currentstateid = (stateid_getter)nfsd4_get_closestateid,
1618 .op_set_currentstateid = (stateid_setter)nfsd4_set_closestateid,
1621 .op_func = (nfsd4op_func)nfsd4_commit,
1622 .op_flags = OP_MODIFIES_SOMETHING,
1623 .op_name = "OP_COMMIT",
1624 .op_rsize_bop = (nfsd4op_rsize)nfsd4_commit_rsize,
1627 .op_func = (nfsd4op_func)nfsd4_create,
1628 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
1629 .op_name = "OP_CREATE",
1630 .op_rsize_bop = (nfsd4op_rsize)nfsd4_create_rsize,
1632 [OP_DELEGRETURN] = {
1633 .op_func = (nfsd4op_func)nfsd4_delegreturn,
1634 .op_flags = OP_MODIFIES_SOMETHING,
1635 .op_name = "OP_DELEGRETURN",
1636 .op_rsize_bop = nfsd4_only_status_rsize,
1637 .op_get_currentstateid = (stateid_getter)nfsd4_get_delegreturnstateid,
1640 .op_func = (nfsd4op_func)nfsd4_getattr,
1641 .op_flags = ALLOWED_ON_ABSENT_FS,
1642 .op_rsize_bop = nfsd4_getattr_rsize,
1643 .op_name = "OP_GETATTR",
1646 .op_func = (nfsd4op_func)nfsd4_getfh,
1647 .op_name = "OP_GETFH",
1650 .op_func = (nfsd4op_func)nfsd4_link,
1651 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
1653 .op_name = "OP_LINK",
1654 .op_rsize_bop = (nfsd4op_rsize)nfsd4_link_rsize,
1657 .op_func = (nfsd4op_func)nfsd4_lock,
1658 .op_flags = OP_MODIFIES_SOMETHING,
1659 .op_name = "OP_LOCK",
1660 .op_rsize_bop = (nfsd4op_rsize)nfsd4_lock_rsize,
1661 .op_set_currentstateid = (stateid_setter)nfsd4_set_lockstateid,
1664 .op_func = (nfsd4op_func)nfsd4_lockt,
1665 .op_name = "OP_LOCKT",
1668 .op_func = (nfsd4op_func)nfsd4_locku,
1669 .op_flags = OP_MODIFIES_SOMETHING,
1670 .op_name = "OP_LOCKU",
1671 .op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1672 .op_get_currentstateid = (stateid_getter)nfsd4_get_lockustateid,
1675 .op_func = (nfsd4op_func)nfsd4_lookup,
1676 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
1677 .op_name = "OP_LOOKUP",
1680 .op_func = (nfsd4op_func)nfsd4_lookupp,
1681 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
1682 .op_name = "OP_LOOKUPP",
1685 .op_func = (nfsd4op_func)nfsd4_nverify,
1686 .op_name = "OP_NVERIFY",
1689 .op_func = (nfsd4op_func)nfsd4_open,
1690 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
1691 .op_name = "OP_OPEN",
1692 .op_rsize_bop = (nfsd4op_rsize)nfsd4_open_rsize,
1693 .op_set_currentstateid = (stateid_setter)nfsd4_set_openstateid,
1695 [OP_OPEN_CONFIRM] = {
1696 .op_func = (nfsd4op_func)nfsd4_open_confirm,
1697 .op_flags = OP_MODIFIES_SOMETHING,
1698 .op_name = "OP_OPEN_CONFIRM",
1699 .op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1701 [OP_OPEN_DOWNGRADE] = {
1702 .op_func = (nfsd4op_func)nfsd4_open_downgrade,
1703 .op_flags = OP_MODIFIES_SOMETHING,
1704 .op_name = "OP_OPEN_DOWNGRADE",
1705 .op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1706 .op_get_currentstateid = (stateid_getter)nfsd4_get_opendowngradestateid,
1707 .op_set_currentstateid = (stateid_setter)nfsd4_set_opendowngradestateid,
1710 .op_func = (nfsd4op_func)nfsd4_putfh,
1711 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1712 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
1713 .op_name = "OP_PUTFH",
1714 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1717 .op_func = (nfsd4op_func)nfsd4_putrootfh,
1718 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1719 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
1720 .op_name = "OP_PUTPUBFH",
1721 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1724 .op_func = (nfsd4op_func)nfsd4_putrootfh,
1725 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1726 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
1727 .op_name = "OP_PUTROOTFH",
1728 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1731 .op_func = (nfsd4op_func)nfsd4_read,
1732 .op_name = "OP_READ",
1733 .op_rsize_bop = (nfsd4op_rsize)nfsd4_read_rsize,
1734 .op_get_currentstateid = (stateid_getter)nfsd4_get_readstateid,
1737 .op_func = (nfsd4op_func)nfsd4_readdir,
1738 .op_name = "OP_READDIR",
1739 .op_rsize_bop = (nfsd4op_rsize)nfsd4_readdir_rsize,
1742 .op_func = (nfsd4op_func)nfsd4_readlink,
1743 .op_name = "OP_READLINK",
1746 .op_func = (nfsd4op_func)nfsd4_remove,
1747 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1748 .op_name = "OP_REMOVE",
1749 .op_rsize_bop = (nfsd4op_rsize)nfsd4_remove_rsize,
1752 .op_func = (nfsd4op_func)nfsd4_rename,
1753 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1754 .op_name = "OP_RENAME",
1755 .op_rsize_bop = (nfsd4op_rsize)nfsd4_rename_rsize,
1758 .op_func = (nfsd4op_func)nfsd4_renew,
1759 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1760 | OP_MODIFIES_SOMETHING,
1761 .op_name = "OP_RENEW",
1762 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1766 .op_func = (nfsd4op_func)nfsd4_restorefh,
1767 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1768 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
1769 .op_name = "OP_RESTOREFH",
1770 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1773 .op_func = (nfsd4op_func)nfsd4_savefh,
1774 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
1775 .op_name = "OP_SAVEFH",
1776 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1779 .op_func = (nfsd4op_func)nfsd4_secinfo,
1780 .op_flags = OP_HANDLES_WRONGSEC,
1781 .op_name = "OP_SECINFO",
1784 .op_func = (nfsd4op_func)nfsd4_setattr,
1785 .op_name = "OP_SETATTR",
1786 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1787 .op_rsize_bop = (nfsd4op_rsize)nfsd4_setattr_rsize,
1788 .op_get_currentstateid = (stateid_getter)nfsd4_get_setattrstateid,
1790 [OP_SETCLIENTID] = {
1791 .op_func = (nfsd4op_func)nfsd4_setclientid,
1792 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1793 | OP_MODIFIES_SOMETHING | OP_CACHEME,
1794 .op_name = "OP_SETCLIENTID",
1795 .op_rsize_bop = (nfsd4op_rsize)nfsd4_setclientid_rsize,
1797 [OP_SETCLIENTID_CONFIRM] = {
1798 .op_func = (nfsd4op_func)nfsd4_setclientid_confirm,
1799 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1800 | OP_MODIFIES_SOMETHING | OP_CACHEME,
1801 .op_name = "OP_SETCLIENTID_CONFIRM",
1802 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1805 .op_func = (nfsd4op_func)nfsd4_verify,
1806 .op_name = "OP_VERIFY",
1809 .op_func = (nfsd4op_func)nfsd4_write,
1810 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1811 .op_name = "OP_WRITE",
1812 .op_rsize_bop = (nfsd4op_rsize)nfsd4_write_rsize,
1813 .op_get_currentstateid = (stateid_getter)nfsd4_get_writestateid,
1815 [OP_RELEASE_LOCKOWNER] = {
1816 .op_func = (nfsd4op_func)nfsd4_release_lockowner,
1817 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1818 | OP_MODIFIES_SOMETHING,
1819 .op_name = "OP_RELEASE_LOCKOWNER",
1820 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1823 /* NFSv4.1 operations */
1824 [OP_EXCHANGE_ID] = {
1825 .op_func = (nfsd4op_func)nfsd4_exchange_id,
1826 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1827 | OP_MODIFIES_SOMETHING,
1828 .op_name = "OP_EXCHANGE_ID",
1829 .op_rsize_bop = (nfsd4op_rsize)nfsd4_exchange_id_rsize,
1831 [OP_BACKCHANNEL_CTL] = {
1832 .op_func = (nfsd4op_func)nfsd4_backchannel_ctl,
1833 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
1834 .op_name = "OP_BACKCHANNEL_CTL",
1835 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1837 [OP_BIND_CONN_TO_SESSION] = {
1838 .op_func = (nfsd4op_func)nfsd4_bind_conn_to_session,
1839 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1840 | OP_MODIFIES_SOMETHING,
1841 .op_name = "OP_BIND_CONN_TO_SESSION",
1842 .op_rsize_bop = (nfsd4op_rsize)nfsd4_bind_conn_to_session_rsize,
1844 [OP_CREATE_SESSION] = {
1845 .op_func = (nfsd4op_func)nfsd4_create_session,
1846 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1847 | OP_MODIFIES_SOMETHING,
1848 .op_name = "OP_CREATE_SESSION",
1849 .op_rsize_bop = (nfsd4op_rsize)nfsd4_create_session_rsize,
1851 [OP_DESTROY_SESSION] = {
1852 .op_func = (nfsd4op_func)nfsd4_destroy_session,
1853 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1854 | OP_MODIFIES_SOMETHING,
1855 .op_name = "OP_DESTROY_SESSION",
1856 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1859 .op_func = (nfsd4op_func)nfsd4_sequence,
1860 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1861 .op_name = "OP_SEQUENCE",
1863 [OP_DESTROY_CLIENTID] = {
1864 .op_func = (nfsd4op_func)nfsd4_destroy_clientid,
1865 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1866 | OP_MODIFIES_SOMETHING,
1867 .op_name = "OP_DESTROY_CLIENTID",
1868 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1870 [OP_RECLAIM_COMPLETE] = {
1871 .op_func = (nfsd4op_func)nfsd4_reclaim_complete,
1872 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
1873 .op_name = "OP_RECLAIM_COMPLETE",
1874 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1876 [OP_SECINFO_NO_NAME] = {
1877 .op_func = (nfsd4op_func)nfsd4_secinfo_no_name,
1878 .op_flags = OP_HANDLES_WRONGSEC,
1879 .op_name = "OP_SECINFO_NO_NAME",
1881 [OP_TEST_STATEID] = {
1882 .op_func = (nfsd4op_func)nfsd4_test_stateid,
1883 .op_flags = ALLOWED_WITHOUT_FH,
1884 .op_name = "OP_TEST_STATEID",
1886 [OP_FREE_STATEID] = {
1887 .op_func = (nfsd4op_func)nfsd4_free_stateid,
1888 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
1889 .op_name = "OP_FREE_STATEID",
1890 .op_get_currentstateid = (stateid_getter)nfsd4_get_freestateid,
1891 .op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1895 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
1897 struct nfsd4_operation *opdesc;
1898 nfsd4op_rsize estimator;
1900 if (op->opnum == OP_ILLEGAL)
1901 return op_encode_hdr_size * sizeof(__be32);
1902 opdesc = OPDESC(op);
1903 estimator = opdesc->op_rsize_bop;
1904 return estimator ? estimator(rqstp, op) : PAGE_SIZE;
1907 void warn_on_nonidempotent_op(struct nfsd4_op *op)
1909 if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
1910 pr_err("unable to encode reply to nonidempotent op %d (%s)\n",
1911 op->opnum, nfsd4_op_name(op->opnum));
1916 static const char *nfsd4_op_name(unsigned opnum)
1918 if (opnum < ARRAY_SIZE(nfsd4_ops))
1919 return nfsd4_ops[opnum].op_name;
1920 return "unknown_operation";
1923 #define nfsd4_voidres nfsd4_voidargs
1924 struct nfsd4_voidargs { int dummy; };
1926 static struct svc_procedure nfsd_procedures4[2] = {
1928 .pc_func = (svc_procfunc) nfsd4_proc_null,
1929 .pc_encode = (kxdrproc_t) nfs4svc_encode_voidres,
1930 .pc_argsize = sizeof(struct nfsd4_voidargs),
1931 .pc_ressize = sizeof(struct nfsd4_voidres),
1932 .pc_cachetype = RC_NOCACHE,
1935 [NFSPROC4_COMPOUND] = {
1936 .pc_func = (svc_procfunc) nfsd4_proc_compound,
1937 .pc_decode = (kxdrproc_t) nfs4svc_decode_compoundargs,
1938 .pc_encode = (kxdrproc_t) nfs4svc_encode_compoundres,
1939 .pc_argsize = sizeof(struct nfsd4_compoundargs),
1940 .pc_ressize = sizeof(struct nfsd4_compoundres),
1941 .pc_release = nfsd4_release_compoundargs,
1942 .pc_cachetype = RC_NOCACHE,
1943 .pc_xdrressize = NFSD_BUFSIZE/4,
1947 struct svc_version nfsd_version4 = {
1950 .vs_proc = nfsd_procedures4,
1951 .vs_dispatch = nfsd_dispatch,
1952 .vs_xdrsize = NFS4_SVC_XDRSIZE,