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>
40 #include <linux/sunrpc/addr.h>
41 #include <linux/nfs_ssc.h>
47 #include "current_stateid.h"
53 static bool inter_copy_offload_enable;
54 module_param(inter_copy_offload_enable, bool, 0644);
55 MODULE_PARM_DESC(inter_copy_offload_enable,
56 "Enable inter server to server copy offload. Default: false");
58 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
59 static int nfsd4_ssc_umount_timeout = 900000; /* default to 15 mins */
60 module_param(nfsd4_ssc_umount_timeout, int, 0644);
61 MODULE_PARM_DESC(nfsd4_ssc_umount_timeout,
62 "idle msecs before unmount export from source server");
65 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
66 #include <linux/security.h>
69 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
71 struct inode *inode = d_inode(resfh->fh_dentry);
75 status = security_inode_setsecctx(resfh->fh_dentry,
76 label->data, label->len);
81 * XXX: We should really fail the whole open, but we may
82 * already have created a new file, so it may be too
83 * late. For now this seems the least of evils:
85 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
91 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
95 #define NFSDDBG_FACILITY NFSDDBG_PROC
97 static u32 nfsd_attrmask[] = {
98 NFSD_WRITEABLE_ATTRS_WORD0,
99 NFSD_WRITEABLE_ATTRS_WORD1,
100 NFSD_WRITEABLE_ATTRS_WORD2
103 static u32 nfsd41_ex_attrmask[] = {
104 NFSD_SUPPATTR_EXCLCREAT_WORD0,
105 NFSD_SUPPATTR_EXCLCREAT_WORD1,
106 NFSD_SUPPATTR_EXCLCREAT_WORD2
110 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
111 u32 *bmval, u32 *writable)
113 struct dentry *dentry = cstate->current_fh.fh_dentry;
114 struct svc_export *exp = cstate->current_fh.fh_export;
116 if (!nfsd_attrs_supported(cstate->minorversion, bmval))
117 return nfserr_attrnotsupp;
118 if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
119 return nfserr_attrnotsupp;
120 if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
121 !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
122 return nfserr_attrnotsupp;
123 if (writable && !bmval_is_subset(bmval, writable))
125 if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
126 (bmval[1] & FATTR4_WORD1_MODE))
132 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
133 struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
135 __be32 status = nfs_ok;
137 if (open->op_create == NFS4_OPEN_CREATE) {
138 if (open->op_createmode == NFS4_CREATE_UNCHECKED
139 || open->op_createmode == NFS4_CREATE_GUARDED)
140 status = check_attr_support(rqstp, cstate,
141 open->op_bmval, nfsd_attrmask);
142 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
143 status = check_attr_support(rqstp, cstate,
144 open->op_bmval, nfsd41_ex_attrmask);
151 is_create_with_attrs(struct nfsd4_open *open)
153 return open->op_create == NFS4_OPEN_CREATE
154 && (open->op_createmode == NFS4_CREATE_UNCHECKED
155 || open->op_createmode == NFS4_CREATE_GUARDED
156 || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
160 * if error occurs when setting the acl, just clear the acl bit
161 * in the returned attr bitmap.
164 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
165 struct nfs4_acl *acl, u32 *bmval)
169 status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
172 * We should probably fail the whole open at this point,
173 * but we've already created the file, so it's too late;
174 * So this seems the least of evils:
176 bmval[0] &= ~FATTR4_WORD0_ACL;
180 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
183 dget(src->fh_dentry);
185 exp_get(src->fh_export);
190 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
194 if (open->op_truncate &&
195 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
198 accmode |= NFSD_MAY_READ_IF_EXEC;
200 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
201 accmode |= NFSD_MAY_READ;
202 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
203 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
204 if (open->op_share_deny & NFS4_SHARE_DENY_READ)
205 accmode |= NFSD_MAY_WRITE;
207 status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
212 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
214 umode_t mode = d_inode(fh->fh_dentry)->i_mode;
221 * Using err_symlink as our catch-all case may look odd; but
222 * there's no other obvious error for this case in 4.0, and we
223 * happen to know that it will cause the linux v4 client to do
224 * the right thing on attempts to open something other than a
227 return nfserr_symlink;
230 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
232 if (nfsd4_has_session(cstate))
234 fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
239 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
241 struct svc_fh *current_fh = &cstate->current_fh;
245 *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
247 return nfserr_jukebox;
248 fh_init(*resfh, NFS4_FHSIZE);
249 open->op_truncate = false;
251 if (open->op_create) {
252 /* FIXME: check session persistence and pnfs flags.
253 * The nfsv4.1 spec requires the following semantics:
255 * Persistent | pNFS | Server REQUIRED | Client Allowed
256 * Reply Cache | server | |
257 * -------------+--------+-----------------+--------------------
258 * no | no | EXCLUSIVE4_1 | EXCLUSIVE4_1
260 * | | and EXCLUSIVE4 | or EXCLUSIVE4
262 * no | yes | EXCLUSIVE4_1 | EXCLUSIVE4_1
263 * yes | no | GUARDED4 | GUARDED4
264 * yes | yes | GUARDED4 | GUARDED4
268 * Note: create modes (UNCHECKED,GUARDED...) are the same
269 * in NFSv4 as in v3 except EXCLUSIVE4_1.
271 current->fs->umask = open->op_umask;
272 status = do_nfsd_create(rqstp, current_fh, open->op_fname,
273 open->op_fnamelen, &open->op_iattr,
274 *resfh, open->op_createmode,
275 (u32 *)open->op_verf.data,
276 &open->op_truncate, &open->op_created);
277 current->fs->umask = 0;
279 if (!status && open->op_label.len)
280 nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
283 * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
284 * use the returned bitmask to indicate which attributes
285 * we used to store the verifier:
287 if (nfsd_create_is_exclusive(open->op_createmode) && status == 0)
288 open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
289 FATTR4_WORD1_TIME_MODIFY);
292 * Note this may exit with the parent still locked.
293 * We will hold the lock until nfsd4_open's final
294 * lookup, to prevent renames or unlinks until we've had
295 * a chance to an acquire a delegation if appropriate.
297 status = nfsd_lookup(rqstp, current_fh,
298 open->op_fname, open->op_fnamelen, *resfh);
301 status = nfsd_check_obj_isreg(*resfh);
305 if (is_create_with_attrs(open) && open->op_acl != NULL)
306 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
308 nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
309 accmode = NFSD_MAY_NOP;
310 if (open->op_created ||
311 open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
312 accmode |= NFSD_MAY_OWNER_OVERRIDE;
313 status = do_open_permission(rqstp, *resfh, open, accmode);
314 set_change_info(&open->op_cinfo, current_fh);
320 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
322 struct svc_fh *current_fh = &cstate->current_fh;
326 /* We don't know the target directory, and therefore can not
327 * set the change info
330 memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
332 nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
334 open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
335 (open->op_iattr.ia_size == 0);
337 * In the delegation case, the client is telling us about an
338 * open that it *already* performed locally, some time ago. We
339 * should let it succeed now if possible.
341 * In the case of a CLAIM_FH open, on the other hand, the client
342 * may be counting on us to enforce permissions (the Linux 4.1
343 * client uses this for normal opens, for example).
345 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
346 accmode = NFSD_MAY_OWNER_OVERRIDE;
348 status = do_open_permission(rqstp, current_fh, open, accmode);
354 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
356 struct nfsd4_sessionid *sid =
357 (struct nfsd4_sessionid *)session->se_sessionid.data;
359 clid->cl_boot = sid->clientid.cl_boot;
360 clid->cl_id = sid->clientid.cl_id;
364 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
367 struct nfsd4_open *open = &u->open;
369 struct svc_fh *resfh = NULL;
370 struct net *net = SVC_NET(rqstp);
371 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
372 bool reclaim = false;
374 dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
375 (int)open->op_fnamelen, open->op_fname,
378 /* This check required by spec. */
379 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
382 open->op_created = false;
385 * Before RECLAIM_COMPLETE done, server should deny new lock
387 if (nfsd4_has_session(cstate) &&
388 !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
389 open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
392 if (nfsd4_has_session(cstate))
393 copy_clientid(&open->op_clientid, cstate->session);
395 /* check seqid for replay. set nfs4_owner */
396 status = nfsd4_process_open1(cstate, open, nn);
397 if (status == nfserr_replay_me) {
398 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
399 fh_put(&cstate->current_fh);
400 fh_copy_shallow(&cstate->current_fh.fh_handle,
402 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
404 dprintk("nfsd4_open: replay failed"
405 " restoring previous filehandle\n");
407 status = nfserr_replay_me;
411 if (open->op_xdr_error) {
412 status = open->op_xdr_error;
416 status = nfsd4_check_open_attributes(rqstp, cstate, open);
420 /* Openowner is now set, so sequence id will get bumped. Now we need
421 * these checks before we do any creates: */
422 status = nfserr_grace;
423 if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
425 status = nfserr_no_grace;
426 if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
429 switch (open->op_claim_type) {
430 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
431 case NFS4_OPEN_CLAIM_NULL:
432 status = do_open_lookup(rqstp, cstate, open, &resfh);
436 case NFS4_OPEN_CLAIM_PREVIOUS:
437 status = nfs4_check_open_reclaim(cstate->clp);
440 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
443 case NFS4_OPEN_CLAIM_FH:
444 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
445 status = do_open_fhandle(rqstp, cstate, open);
448 resfh = &cstate->current_fh;
450 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
451 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
452 dprintk("NFSD: unsupported OPEN claim type %d\n",
453 open->op_claim_type);
454 status = nfserr_notsupp;
457 dprintk("NFSD: Invalid OPEN claim type %d\n",
458 open->op_claim_type);
459 status = nfserr_inval;
463 * nfsd4_process_open2() does the actual opening of the file. If
464 * successful, it (1) truncates the file if open->op_truncate was
465 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
467 status = nfsd4_process_open2(rqstp, resfh, open);
468 WARN(status && open->op_created,
469 "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
470 be32_to_cpu(status));
471 if (reclaim && !status)
472 nn->somebody_reclaimed = true;
474 if (resfh && resfh != &cstate->current_fh) {
475 fh_dup2(&cstate->current_fh, resfh);
479 nfsd4_cleanup_open_state(cstate, open);
480 nfsd4_bump_seqid(cstate, status);
485 * OPEN is the only seqid-mutating operation whose decoding can fail
486 * with a seqid-mutating error (specifically, decoding of user names in
487 * the attributes). Therefore we have to do some processing to look up
488 * the stateowner so that we can bump the seqid.
490 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
492 struct nfsd4_open *open = &op->u.open;
494 if (!seqid_mutating_err(ntohl(op->status)))
496 if (nfsd4_has_session(cstate))
498 open->op_xdr_error = op->status;
499 return nfsd4_open(rqstp, cstate, &op->u);
503 * filehandle-manipulating ops.
506 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
509 u->getfh = &cstate->current_fh;
514 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
517 struct nfsd4_putfh *putfh = &u->putfh;
520 fh_put(&cstate->current_fh);
521 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
522 memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
524 ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
525 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
526 if (ret == nfserr_stale && putfh->no_verify) {
527 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
535 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
540 fh_put(&cstate->current_fh);
541 status = exp_pseudoroot(rqstp, &cstate->current_fh);
546 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
549 if (!cstate->save_fh.fh_dentry)
550 return nfserr_restorefh;
552 fh_dup2(&cstate->current_fh, &cstate->save_fh);
553 if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
554 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
555 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
561 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
564 fh_dup2(&cstate->save_fh, &cstate->current_fh);
565 if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
566 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
567 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
576 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
579 struct nfsd4_access *access = &u->access;
582 access_full = NFS3_ACCESS_FULL;
583 if (cstate->minorversion >= 2)
584 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
587 if (access->ac_req_access & ~access_full)
590 access->ac_resp_access = access->ac_req_access;
591 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
592 &access->ac_supported);
595 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
597 __be32 *verf = (__be32 *)verifier->data;
599 BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
601 nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
605 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
608 struct nfsd4_commit *commit = &u->commit;
610 return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
612 (__be32 *)commit->co_verf.data);
616 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
619 struct nfsd4_create *create = &u->create;
624 fh_init(&resfh, NFS4_FHSIZE);
626 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
630 status = check_attr_support(rqstp, cstate, create->cr_bmval,
635 current->fs->umask = create->cr_umask;
636 switch (create->cr_type) {
638 status = nfsd_symlink(rqstp, &cstate->current_fh,
639 create->cr_name, create->cr_namelen,
640 create->cr_data, &resfh);
644 status = nfserr_inval;
645 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
646 if (MAJOR(rdev) != create->cr_specdata1 ||
647 MINOR(rdev) != create->cr_specdata2)
649 status = nfsd_create(rqstp, &cstate->current_fh,
650 create->cr_name, create->cr_namelen,
651 &create->cr_iattr, S_IFBLK, rdev, &resfh);
655 status = nfserr_inval;
656 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
657 if (MAJOR(rdev) != create->cr_specdata1 ||
658 MINOR(rdev) != create->cr_specdata2)
660 status = nfsd_create(rqstp, &cstate->current_fh,
661 create->cr_name, create->cr_namelen,
662 &create->cr_iattr,S_IFCHR, rdev, &resfh);
666 status = nfsd_create(rqstp, &cstate->current_fh,
667 create->cr_name, create->cr_namelen,
668 &create->cr_iattr, S_IFSOCK, 0, &resfh);
672 status = nfsd_create(rqstp, &cstate->current_fh,
673 create->cr_name, create->cr_namelen,
674 &create->cr_iattr, S_IFIFO, 0, &resfh);
678 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
679 status = nfsd_create(rqstp, &cstate->current_fh,
680 create->cr_name, create->cr_namelen,
681 &create->cr_iattr, S_IFDIR, 0, &resfh);
685 status = nfserr_badtype;
691 if (create->cr_label.len)
692 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
694 if (create->cr_acl != NULL)
695 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
698 fh_unlock(&cstate->current_fh);
699 set_change_info(&create->cr_cinfo, &cstate->current_fh);
700 fh_dup2(&cstate->current_fh, &resfh);
704 current->fs->umask = 0;
709 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
712 struct nfsd4_getattr *getattr = &u->getattr;
715 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
719 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
722 getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
723 getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
724 getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
726 getattr->ga_fhp = &cstate->current_fh;
731 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
734 struct nfsd4_link *link = &u->link;
737 status = nfsd_link(rqstp, &cstate->current_fh,
738 link->li_name, link->li_namelen, &cstate->save_fh);
740 set_change_info(&link->li_cinfo, &cstate->current_fh);
744 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
746 struct svc_fh tmp_fh;
749 fh_init(&tmp_fh, NFS4_FHSIZE);
750 ret = exp_pseudoroot(rqstp, &tmp_fh);
753 if (tmp_fh.fh_dentry == fh->fh_dentry) {
758 return nfsd_lookup(rqstp, fh, "..", 2, fh);
762 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
765 return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
769 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
772 return nfsd_lookup(rqstp, &cstate->current_fh,
773 u->lookup.lo_name, u->lookup.lo_len,
774 &cstate->current_fh);
778 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
781 struct nfsd4_read *read = &u->read;
786 trace_nfsd_read_start(rqstp, &cstate->current_fh,
787 read->rd_offset, read->rd_length);
789 read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
790 if (read->rd_offset > (u64)OFFSET_MAX)
791 read->rd_offset = (u64)OFFSET_MAX;
792 if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
793 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
796 * If we do a zero copy read, then a client will see read data
797 * that reflects the state of the file *after* performing the
798 * following compound.
800 * To ensure proper ordering, we therefore turn off zero copy if
801 * the client wants us to do more in this compound:
803 if (!nfsd4_last_compound_op(rqstp))
804 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
807 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
808 &read->rd_stateid, RD_STATE,
811 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
816 read->rd_rqstp = rqstp;
817 read->rd_fhp = &cstate->current_fh;
823 nfsd4_read_release(union nfsd4_op_u *u)
826 nfsd_file_put(u->read.rd_nf);
827 trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
828 u->read.rd_offset, u->read.rd_length);
832 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
835 struct nfsd4_readdir *readdir = &u->readdir;
836 u64 cookie = readdir->rd_cookie;
837 static const nfs4_verifier zeroverf;
839 /* no need to check permission - this will be done in nfsd_readdir() */
841 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
844 readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
845 readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
846 readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
848 if ((cookie == 1) || (cookie == 2) ||
849 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
850 return nfserr_bad_cookie;
852 readdir->rd_rqstp = rqstp;
853 readdir->rd_fhp = &cstate->current_fh;
858 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
861 u->readlink.rl_rqstp = rqstp;
862 u->readlink.rl_fhp = &cstate->current_fh;
867 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
870 struct nfsd4_remove *remove = &u->remove;
873 if (opens_in_grace(SVC_NET(rqstp)))
875 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
876 remove->rm_name, remove->rm_namelen);
878 fh_unlock(&cstate->current_fh);
879 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
885 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
888 struct nfsd4_rename *rename = &u->rename;
891 if (opens_in_grace(SVC_NET(rqstp)))
893 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
894 rename->rn_snamelen, &cstate->current_fh,
895 rename->rn_tname, rename->rn_tnamelen);
898 set_change_info(&rename->rn_sinfo, &cstate->current_fh);
899 set_change_info(&rename->rn_tinfo, &cstate->save_fh);
904 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
907 struct nfsd4_secinfo *secinfo = &u->secinfo;
908 struct svc_export *exp;
909 struct dentry *dentry;
912 err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
915 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
916 secinfo->si_name, secinfo->si_namelen,
920 fh_unlock(&cstate->current_fh);
921 if (d_really_is_negative(dentry)) {
925 secinfo->si_exp = exp;
927 if (cstate->minorversion)
928 /* See rfc 5661 section 2.6.3.1.1.8 */
929 fh_put(&cstate->current_fh);
934 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
939 switch (u->secinfo_no_name.sin_style) {
940 case NFS4_SECINFO_STYLE4_CURRENT_FH:
942 case NFS4_SECINFO_STYLE4_PARENT:
943 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
951 u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
952 fh_put(&cstate->current_fh);
957 nfsd4_secinfo_release(union nfsd4_op_u *u)
959 if (u->secinfo.si_exp)
960 exp_put(u->secinfo.si_exp);
964 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
966 if (u->secinfo_no_name.sin_exp)
967 exp_put(u->secinfo_no_name.sin_exp);
971 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
974 struct nfsd4_setattr *setattr = &u->setattr;
975 __be32 status = nfs_ok;
978 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
979 status = nfs4_preprocess_stateid_op(rqstp, cstate,
980 &cstate->current_fh, &setattr->sa_stateid,
981 WR_STATE, NULL, NULL);
983 dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
987 err = fh_want_write(&cstate->current_fh);
989 return nfserrno(err);
992 status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
997 if (setattr->sa_acl != NULL)
998 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
1002 if (setattr->sa_label.len)
1003 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
1004 &setattr->sa_label);
1007 status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
1010 fh_drop_write(&cstate->current_fh);
1015 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1016 union nfsd4_op_u *u)
1018 struct nfsd4_write *write = &u->write;
1019 stateid_t *stateid = &write->wr_stateid;
1020 struct nfsd_file *nf = NULL;
1021 __be32 status = nfs_ok;
1025 if (write->wr_offset > (u64)OFFSET_MAX ||
1026 write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1029 cnt = write->wr_buflen;
1030 trace_nfsd_write_start(rqstp, &cstate->current_fh,
1031 write->wr_offset, cnt);
1032 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1033 stateid, WR_STATE, &nf, NULL);
1035 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1039 write->wr_how_written = write->wr_stable_how;
1041 nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1042 WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1044 status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1045 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1046 write->wr_how_written,
1047 (__be32 *)write->wr_verifier.data);
1050 write->wr_bytes_written = cnt;
1051 trace_nfsd_write_done(rqstp, &cstate->current_fh,
1052 write->wr_offset, cnt);
1057 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1058 stateid_t *src_stateid, struct nfsd_file **src,
1059 stateid_t *dst_stateid, struct nfsd_file **dst)
1063 if (!cstate->save_fh.fh_dentry)
1064 return nfserr_nofilehandle;
1066 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1067 src_stateid, RD_STATE, src, NULL);
1069 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1073 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1074 dst_stateid, WR_STATE, dst, NULL);
1076 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1080 /* fix up for NFS-specific error code */
1081 if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1082 !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1083 status = nfserr_wrong_type;
1090 nfsd_file_put(*dst);
1092 nfsd_file_put(*src);
1097 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1098 union nfsd4_op_u *u)
1100 struct nfsd4_clone *clone = &u->clone;
1101 struct nfsd_file *src, *dst;
1104 status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1105 &clone->cl_dst_stateid, &dst);
1109 status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1110 dst, clone->cl_dst_pos, clone->cl_count,
1111 EX_ISSYNC(cstate->current_fh.fh_export));
1119 void nfs4_put_copy(struct nfsd4_copy *copy)
1121 if (!refcount_dec_and_test(©->refcount))
1127 check_and_set_stop_copy(struct nfsd4_copy *copy)
1131 spin_lock(©->cp_clp->async_lock);
1132 value = copy->stopped;
1134 copy->stopped = true;
1135 spin_unlock(©->cp_clp->async_lock);
1139 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1141 /* only 1 thread should stop the copy */
1142 if (!check_and_set_stop_copy(copy))
1143 kthread_stop(copy->copy_task);
1144 nfs4_put_copy(copy);
1147 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1149 struct nfsd4_copy *copy = NULL;
1151 spin_lock(&clp->async_lock);
1152 if (!list_empty(&clp->async_copies)) {
1153 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1155 refcount_inc(©->refcount);
1157 spin_unlock(&clp->async_lock);
1161 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1163 struct nfsd4_copy *copy;
1165 while ((copy = nfsd4_get_copy(clp)) != NULL)
1166 nfsd4_stop_copy(copy);
1168 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1170 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1171 struct nfs_fh *src_fh,
1172 nfs4_stateid *stateid);
1173 extern void nfs42_ssc_close(struct file *filep);
1175 extern void nfs_sb_deactive(struct super_block *sb);
1177 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1180 * setup a work entry in the ssc delayed unmount list.
1182 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1183 struct nfsd4_ssc_umount_item **retwork, struct vfsmount **ss_mnt)
1185 struct nfsd4_ssc_umount_item *ni = NULL;
1186 struct nfsd4_ssc_umount_item *work = NULL;
1187 struct nfsd4_ssc_umount_item *tmp;
1192 work = kzalloc(sizeof(*work), GFP_KERNEL);
1194 spin_lock(&nn->nfsd_ssc_lock);
1195 list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1196 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1199 if (ni->nsui_busy) {
1200 /* wait - and try again */
1201 prepare_to_wait(&nn->nfsd_ssc_waitq, &wait,
1202 TASK_INTERRUPTIBLE);
1203 spin_unlock(&nn->nfsd_ssc_lock);
1205 /* allow 20secs for mount/unmount for now - revisit */
1206 if (signal_pending(current) ||
1207 (schedule_timeout(20*HZ) == 0)) {
1209 return nfserr_eagain;
1211 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1214 *ss_mnt = ni->nsui_vfsmount;
1215 refcount_inc(&ni->nsui_refcnt);
1216 spin_unlock(&nn->nfsd_ssc_lock);
1219 /* return vfsmount in ss_mnt */
1223 strncpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr));
1224 refcount_set(&work->nsui_refcnt, 2);
1225 work->nsui_busy = true;
1226 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1229 spin_unlock(&nn->nfsd_ssc_lock);
1233 static void nfsd4_ssc_update_dul_work(struct nfsd_net *nn,
1234 struct nfsd4_ssc_umount_item *work, struct vfsmount *ss_mnt)
1236 /* set nsui_vfsmount, clear busy flag and wakeup waiters */
1237 spin_lock(&nn->nfsd_ssc_lock);
1238 work->nsui_vfsmount = ss_mnt;
1239 work->nsui_busy = false;
1240 wake_up_all(&nn->nfsd_ssc_waitq);
1241 spin_unlock(&nn->nfsd_ssc_lock);
1244 static void nfsd4_ssc_cancel_dul_work(struct nfsd_net *nn,
1245 struct nfsd4_ssc_umount_item *work)
1247 spin_lock(&nn->nfsd_ssc_lock);
1248 list_del(&work->nsui_list);
1249 wake_up_all(&nn->nfsd_ssc_waitq);
1250 spin_unlock(&nn->nfsd_ssc_lock);
1255 * Support one copy source server for now.
1258 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1259 struct vfsmount **mount)
1261 struct file_system_type *type;
1262 struct vfsmount *ss_mnt;
1263 struct nfs42_netaddr *naddr;
1264 struct sockaddr_storage tmp_addr;
1265 size_t tmp_addrlen, match_netid_len = 3;
1266 char *startsep = "", *endsep = "", *match_netid = "tcp";
1267 char *ipaddr, *dev_name, *raw_data;
1269 __be32 status = nfserr_inval;
1270 struct nfsd4_ssc_umount_item *work = NULL;
1271 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1273 naddr = &nss->u.nl4_addr;
1274 tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1276 (struct sockaddr *)&tmp_addr,
1278 if (tmp_addrlen == 0)
1281 if (tmp_addr.ss_family == AF_INET6) {
1284 match_netid = "tcp6";
1285 match_netid_len = 4;
1288 if (naddr->netid_len != match_netid_len ||
1289 strncmp(naddr->netid, match_netid, naddr->netid_len))
1292 /* Construct the raw data for the vfs_kern_mount call */
1293 len = RPC_MAX_ADDRBUFLEN + 1;
1294 ipaddr = kzalloc(len, GFP_KERNEL);
1298 rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1300 /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1302 raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1303 raw_data = kzalloc(raw_len, GFP_KERNEL);
1305 goto out_free_ipaddr;
1307 snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1309 status = nfserr_nodev;
1310 type = get_fs_type("nfs");
1312 goto out_free_rawdata;
1314 /* Set the server:<export> for the vfs_kern_mount call */
1315 dev_name = kzalloc(len + 5, GFP_KERNEL);
1317 goto out_free_rawdata;
1318 snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1320 status = nfsd4_ssc_setup_dul(nn, ipaddr, &work, &ss_mnt);
1322 goto out_free_devname;
1326 /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1327 ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1328 module_put(type->owner);
1329 if (IS_ERR(ss_mnt)) {
1330 status = nfserr_nodev;
1332 nfsd4_ssc_cancel_dul_work(nn, work);
1333 goto out_free_devname;
1336 nfsd4_ssc_update_dul_work(nn, work, ss_mnt);
1352 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1354 nfs_do_sb_deactive(ss_mnt->mnt_sb);
1359 * Verify COPY destination stateid.
1361 * Connect to the source server with NFSv4.1.
1362 * Create the source struct file for nfsd_copy_range.
1363 * Called with COPY cstate:
1364 * SAVED_FH: source filehandle
1365 * CURRENT_FH: destination filehandle
1368 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1369 struct nfsd4_compound_state *cstate,
1370 struct nfsd4_copy *copy, struct vfsmount **mount)
1372 struct svc_fh *s_fh = NULL;
1373 stateid_t *s_stid = ©->cp_src_stateid;
1374 __be32 status = nfserr_inval;
1376 /* Verify the destination stateid and set dst struct file*/
1377 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1378 ©->cp_dst_stateid,
1379 WR_STATE, ©->nf_dst, NULL);
1383 status = nfsd4_interssc_connect(©->cp_src, rqstp, mount);
1387 s_fh = &cstate->save_fh;
1389 copy->c_fh.size = s_fh->fh_handle.fh_size;
1390 memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1391 copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1392 memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1393 sizeof(stateid_opaque_t));
1401 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1402 struct nfsd_file *dst)
1406 struct nfsd4_ssc_umount_item *tmp;
1407 struct nfsd4_ssc_umount_item *ni = NULL;
1408 struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1410 nfs42_ssc_close(src->nf_file);
1418 spin_lock(&nn->nfsd_ssc_lock);
1419 timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1420 list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1421 if (ni->nsui_vfsmount->mnt_sb == ss_mnt->mnt_sb) {
1422 list_del(&ni->nsui_list);
1424 * vfsmount can be shared by multiple exports,
1425 * decrement refcnt. If the count drops to 1 it
1426 * will be unmounted when nsui_expire expires.
1428 refcount_dec(&ni->nsui_refcnt);
1429 ni->nsui_expire = jiffies + timeout;
1430 list_add_tail(&ni->nsui_list, &nn->nfsd_ssc_mount_list);
1435 spin_unlock(&nn->nfsd_ssc_lock);
1442 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1445 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1446 struct nfsd4_compound_state *cstate,
1447 struct nfsd4_copy *copy,
1448 struct vfsmount **mount)
1451 return nfserr_inval;
1455 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1456 struct nfsd_file *dst)
1461 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1465 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1466 struct nfs_fh *src_fh,
1467 nfs4_stateid *stateid)
1471 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1474 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1475 struct nfsd4_compound_state *cstate,
1476 struct nfsd4_copy *copy)
1478 return nfsd4_verify_copy(rqstp, cstate, ©->cp_src_stateid,
1479 ©->nf_src, ©->cp_dst_stateid,
1484 nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
1490 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1492 struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1494 nfs4_put_copy(copy);
1497 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1498 struct rpc_task *task)
1503 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1504 .release = nfsd4_cb_offload_release,
1505 .done = nfsd4_cb_offload_done
1508 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1510 copy->cp_res.wr_stable_how =
1511 copy->committed ? NFS_FILE_SYNC : NFS_UNSTABLE;
1512 copy->cp_synchronous = sync;
1513 gen_boot_verifier(©->cp_res.wr_verifier, copy->cp_clp->net);
1516 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1518 struct file *dst = copy->nf_dst->nf_file;
1519 struct file *src = copy->nf_src->nf_file;
1521 ssize_t bytes_copied = 0;
1522 u64 bytes_total = copy->cp_count;
1523 u64 src_pos = copy->cp_src_pos;
1524 u64 dst_pos = copy->cp_dst_pos;
1527 /* See RFC 7862 p.67: */
1528 if (bytes_total == 0)
1529 bytes_total = ULLONG_MAX;
1531 if (kthread_should_stop())
1533 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1535 if (bytes_copied <= 0)
1537 bytes_total -= bytes_copied;
1538 copy->cp_res.wr_bytes_written += bytes_copied;
1539 src_pos += bytes_copied;
1540 dst_pos += bytes_copied;
1541 } while (bytes_total > 0 && !copy->cp_synchronous);
1542 /* for a non-zero asynchronous copy do a commit of data */
1543 if (!copy->cp_synchronous && copy->cp_res.wr_bytes_written > 0) {
1544 since = READ_ONCE(dst->f_wb_err);
1545 status = vfs_fsync_range(dst, copy->cp_dst_pos,
1546 copy->cp_res.wr_bytes_written, 0);
1548 status = filemap_check_wb_err(dst->f_mapping, since);
1550 copy->committed = true;
1552 return bytes_copied;
1555 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1560 bytes = _nfsd_copy_file_range(copy);
1561 /* for async copy, we ignore the error, client can always retry
1564 if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1565 status = nfserrno(bytes);
1567 nfsd4_init_copy_res(copy, sync);
1571 if (!copy->cp_intra) /* Inter server SSC */
1572 nfsd4_cleanup_inter_ssc(copy->ss_mnt, copy->nf_src,
1575 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1580 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1582 dst->cp_src_pos = src->cp_src_pos;
1583 dst->cp_dst_pos = src->cp_dst_pos;
1584 dst->cp_count = src->cp_count;
1585 dst->cp_synchronous = src->cp_synchronous;
1586 memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1587 memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1588 dst->cp_clp = src->cp_clp;
1589 dst->nf_dst = nfsd_file_get(src->nf_dst);
1590 dst->cp_intra = src->cp_intra;
1591 if (src->cp_intra) /* for inter, file_src doesn't exist yet */
1592 dst->nf_src = nfsd_file_get(src->nf_src);
1594 memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1595 memcpy(&dst->cp_src, &src->cp_src, sizeof(struct nl4_server));
1596 memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1597 memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1598 dst->ss_mnt = src->ss_mnt;
1601 static void cleanup_async_copy(struct nfsd4_copy *copy)
1603 nfs4_free_copy_state(copy);
1604 nfsd_file_put(copy->nf_dst);
1606 nfsd_file_put(copy->nf_src);
1607 spin_lock(©->cp_clp->async_lock);
1608 list_del(©->copies);
1609 spin_unlock(©->cp_clp->async_lock);
1610 nfs4_put_copy(copy);
1613 static int nfsd4_do_async_copy(void *data)
1615 struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1616 struct nfsd4_copy *cb_copy;
1618 if (!copy->cp_intra) { /* Inter server SSC */
1619 copy->nf_src = kzalloc(sizeof(struct nfsd_file), GFP_KERNEL);
1620 if (!copy->nf_src) {
1621 copy->nfserr = nfserr_serverfault;
1622 nfsd4_interssc_disconnect(copy->ss_mnt);
1625 copy->nf_src->nf_file = nfs42_ssc_open(copy->ss_mnt, ©->c_fh,
1627 if (IS_ERR(copy->nf_src->nf_file)) {
1628 copy->nfserr = nfserr_offload_denied;
1629 nfsd4_interssc_disconnect(copy->ss_mnt);
1634 copy->nfserr = nfsd4_do_copy(copy, 0);
1636 cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1639 refcount_set(&cb_copy->refcount, 1);
1640 memcpy(&cb_copy->cp_res, ©->cp_res, sizeof(copy->cp_res));
1641 cb_copy->cp_clp = copy->cp_clp;
1642 cb_copy->nfserr = copy->nfserr;
1643 memcpy(&cb_copy->fh, ©->fh, sizeof(copy->fh));
1644 nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1645 &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1646 trace_nfsd_cb_offload(copy->cp_clp, ©->cp_res.cb_stateid,
1647 ©->fh, copy->cp_count, copy->nfserr);
1648 nfsd4_run_cb(&cb_copy->cp_cb);
1650 if (!copy->cp_intra)
1651 kfree(copy->nf_src);
1652 cleanup_async_copy(copy);
1657 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1658 union nfsd4_op_u *u)
1660 struct nfsd4_copy *copy = &u->copy;
1662 struct nfsd4_copy *async_copy = NULL;
1664 if (!copy->cp_intra) { /* Inter server SSC */
1665 if (!inter_copy_offload_enable || copy->cp_synchronous) {
1666 status = nfserr_notsupp;
1669 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy,
1672 return nfserr_offload_denied;
1674 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1679 copy->cp_clp = cstate->clp;
1680 memcpy(©->fh, &cstate->current_fh.fh_handle,
1681 sizeof(struct knfsd_fh));
1682 if (!copy->cp_synchronous) {
1683 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1685 status = nfserrno(-ENOMEM);
1686 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1689 if (!nfs4_init_copy_state(nn, copy))
1691 refcount_set(&async_copy->refcount, 1);
1692 memcpy(©->cp_res.cb_stateid, ©->cp_stateid.stid,
1693 sizeof(copy->cp_res.cb_stateid));
1694 dup_copy_fields(copy, async_copy);
1695 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1696 async_copy, "%s", "copy thread");
1697 if (IS_ERR(async_copy->copy_task))
1699 spin_lock(&async_copy->cp_clp->async_lock);
1700 list_add(&async_copy->copies,
1701 &async_copy->cp_clp->async_copies);
1702 spin_unlock(&async_copy->cp_clp->async_lock);
1703 wake_up_process(async_copy->copy_task);
1706 status = nfsd4_do_copy(copy, 1);
1712 cleanup_async_copy(async_copy);
1713 status = nfserrno(-ENOMEM);
1714 if (!copy->cp_intra)
1715 nfsd4_interssc_disconnect(copy->ss_mnt);
1720 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1722 struct nfsd4_copy *copy;
1724 spin_lock(&clp->async_lock);
1725 list_for_each_entry(copy, &clp->async_copies, copies) {
1726 if (memcmp(©->cp_stateid.stid, stateid, NFS4_STATEID_SIZE))
1728 refcount_inc(©->refcount);
1729 spin_unlock(&clp->async_lock);
1732 spin_unlock(&clp->async_lock);
1737 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1738 struct nfsd4_compound_state *cstate,
1739 union nfsd4_op_u *u)
1741 struct nfsd4_offload_status *os = &u->offload_status;
1742 struct nfsd4_copy *copy;
1743 struct nfs4_client *clp = cstate->clp;
1745 copy = find_async_copy(clp, &os->stateid);
1747 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1749 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1751 nfsd4_stop_copy(copy);
1757 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1758 union nfsd4_op_u *u)
1760 struct nfsd4_copy_notify *cn = &u->copy_notify;
1762 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1763 struct nfs4_stid *stid;
1764 struct nfs4_cpntf_state *cps;
1765 struct nfs4_client *clp = cstate->clp;
1767 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1768 &cn->cpn_src_stateid, RD_STATE, NULL,
1773 cn->cpn_sec = nn->nfsd4_lease;
1776 status = nfserrno(-ENOMEM);
1777 cps = nfs4_alloc_init_cpntf_state(nn, stid);
1780 memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.stid, sizeof(stateid_t));
1781 memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1782 memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1784 /* For now, only return one server address in cpn_src, the
1785 * address used by the client to connect to this server.
1787 cn->cpn_src.nl4_type = NL4_NETADDR;
1788 status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1789 &cn->cpn_src.u.nl4_addr);
1790 WARN_ON_ONCE(status);
1792 nfs4_put_cpntf_state(nn, cps);
1796 nfs4_put_stid(stid);
1801 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1802 struct nfsd4_fallocate *fallocate, int flags)
1805 struct nfsd_file *nf;
1807 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1808 &fallocate->falloc_stateid,
1809 WR_STATE, &nf, NULL);
1810 if (status != nfs_ok) {
1811 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1815 status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1816 fallocate->falloc_offset,
1817 fallocate->falloc_length,
1823 nfsd4_offload_status(struct svc_rqst *rqstp,
1824 struct nfsd4_compound_state *cstate,
1825 union nfsd4_op_u *u)
1827 struct nfsd4_offload_status *os = &u->offload_status;
1829 struct nfsd4_copy *copy;
1830 struct nfs4_client *clp = cstate->clp;
1832 copy = find_async_copy(clp, &os->stateid);
1834 os->count = copy->cp_res.wr_bytes_written;
1835 nfs4_put_copy(copy);
1837 status = nfserr_bad_stateid;
1843 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1844 union nfsd4_op_u *u)
1846 return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1850 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1851 union nfsd4_op_u *u)
1853 return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1854 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1858 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1859 union nfsd4_op_u *u)
1861 struct nfsd4_seek *seek = &u->seek;
1864 struct nfsd_file *nf;
1866 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1867 &seek->seek_stateid,
1868 RD_STATE, &nf, NULL);
1870 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1874 switch (seek->seek_whence) {
1875 case NFS4_CONTENT_DATA:
1878 case NFS4_CONTENT_HOLE:
1882 status = nfserr_union_notsupp;
1887 * Note: This call does change file->f_pos, but nothing in NFSD
1888 * should ever file->f_pos.
1890 seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
1891 if (seek->seek_pos < 0)
1892 status = nfserrno(seek->seek_pos);
1893 else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
1894 seek->seek_eof = true;
1901 /* This routine never returns NFS_OK! If there are no other errors, it
1902 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1903 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
1904 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1907 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1908 struct nfsd4_verify *verify)
1914 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1918 status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1922 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1923 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1924 return nfserr_inval;
1925 if (verify->ve_attrlen & 3)
1926 return nfserr_inval;
1929 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1931 count = 4 + (verify->ve_attrlen >> 2);
1932 buf = kmalloc(count << 2, GFP_KERNEL);
1934 return nfserr_jukebox;
1937 status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1938 cstate->current_fh.fh_export,
1939 cstate->current_fh.fh_dentry,
1943 * If nfsd4_encode_fattr() ran out of space, assume that's because
1944 * the attributes are longer (hence different) than those given:
1946 if (status == nfserr_resource)
1947 status = nfserr_not_same;
1952 p = buf + 1 + ntohl(buf[0]);
1953 status = nfserr_not_same;
1954 if (ntohl(*p++) != verify->ve_attrlen)
1956 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1957 status = nfserr_same;
1965 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1966 union nfsd4_op_u *u)
1970 status = _nfsd4_verify(rqstp, cstate, &u->verify);
1971 return status == nfserr_not_same ? nfs_ok : status;
1975 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1976 union nfsd4_op_u *u)
1980 status = _nfsd4_verify(rqstp, cstate, &u->nverify);
1981 return status == nfserr_same ? nfs_ok : status;
1984 #ifdef CONFIG_NFSD_PNFS
1985 static const struct nfsd4_layout_ops *
1986 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
1988 if (!exp->ex_layout_types) {
1989 dprintk("%s: export does not support pNFS\n", __func__);
1993 if (layout_type >= LAYOUT_TYPE_MAX ||
1994 !(exp->ex_layout_types & (1 << layout_type))) {
1995 dprintk("%s: layout type %d not supported\n",
1996 __func__, layout_type);
2000 return nfsd4_layout_ops[layout_type];
2004 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2005 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2007 struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2008 const struct nfsd4_layout_ops *ops;
2009 struct nfsd4_deviceid_map *map;
2010 struct svc_export *exp;
2013 dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2015 gdp->gd_layout_type,
2016 gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2019 map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2021 dprintk("%s: couldn't find device ID to export mapping!\n",
2023 return nfserr_noent;
2026 exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2028 dprintk("%s: could not find device id\n", __func__);
2029 return nfserr_noent;
2032 nfserr = nfserr_layoutunavailable;
2033 ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2038 if (gdp->gd_maxcount != 0) {
2039 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2040 rqstp, cstate->clp, gdp);
2043 gdp->gd_notify_types &= ops->notify_types;
2050 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2052 kfree(u->getdeviceinfo.gd_device);
2056 nfsd4_layoutget(struct svc_rqst *rqstp,
2057 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2059 struct nfsd4_layoutget *lgp = &u->layoutget;
2060 struct svc_fh *current_fh = &cstate->current_fh;
2061 const struct nfsd4_layout_ops *ops;
2062 struct nfs4_layout_stateid *ls;
2064 int accmode = NFSD_MAY_READ_IF_EXEC;
2066 switch (lgp->lg_seg.iomode) {
2068 accmode |= NFSD_MAY_READ;
2071 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2074 dprintk("%s: invalid iomode %d\n",
2075 __func__, lgp->lg_seg.iomode);
2076 nfserr = nfserr_badiomode;
2080 nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2084 nfserr = nfserr_layoutunavailable;
2085 ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2090 * Verify minlength and range as per RFC5661:
2091 * o If loga_length is less than loga_minlength,
2092 * the metadata server MUST return NFS4ERR_INVAL.
2093 * o If the sum of loga_offset and loga_minlength exceeds
2094 * NFS4_UINT64_MAX, and loga_minlength is not
2095 * NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2096 * o If the sum of loga_offset and loga_length exceeds
2097 * NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2098 * the error NFS4ERR_INVAL MUST result.
2100 nfserr = nfserr_inval;
2101 if (lgp->lg_seg.length < lgp->lg_minlength ||
2102 (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2103 lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2104 (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2105 lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2107 if (lgp->lg_seg.length == 0)
2110 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2111 true, lgp->lg_layout_type, &ls);
2113 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2117 nfserr = nfserr_recallconflict;
2118 if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2121 nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2126 nfserr = nfsd4_insert_layout(lgp, ls);
2129 mutex_unlock(&ls->ls_mutex);
2130 nfs4_put_stid(&ls->ls_stid);
2136 nfsd4_layoutget_release(union nfsd4_op_u *u)
2138 kfree(u->layoutget.lg_content);
2142 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2143 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2145 struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2146 const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2147 struct svc_fh *current_fh = &cstate->current_fh;
2148 const struct nfsd4_layout_ops *ops;
2149 loff_t new_size = lcp->lc_last_wr + 1;
2150 struct inode *inode;
2151 struct nfs4_layout_stateid *ls;
2154 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2158 nfserr = nfserr_layoutunavailable;
2159 ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2162 inode = d_inode(current_fh->fh_dentry);
2164 nfserr = nfserr_inval;
2165 if (new_size <= seg->offset) {
2166 dprintk("pnfsd: last write before layout segment\n");
2169 if (new_size > seg->offset + seg->length) {
2170 dprintk("pnfsd: last write beyond layout segment\n");
2173 if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2174 dprintk("pnfsd: layoutcommit beyond EOF\n");
2178 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2179 false, lcp->lc_layout_type,
2182 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2183 /* fixup error code as per RFC5661 */
2184 if (nfserr == nfserr_bad_stateid)
2185 nfserr = nfserr_badlayout;
2189 /* LAYOUTCOMMIT does not require any serialization */
2190 mutex_unlock(&ls->ls_mutex);
2192 if (new_size > i_size_read(inode)) {
2193 lcp->lc_size_chg = 1;
2194 lcp->lc_newsize = new_size;
2196 lcp->lc_size_chg = 0;
2199 nfserr = ops->proc_layoutcommit(inode, lcp);
2200 nfs4_put_stid(&ls->ls_stid);
2206 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2207 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2209 struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2210 struct svc_fh *current_fh = &cstate->current_fh;
2213 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2217 nfserr = nfserr_layoutunavailable;
2218 if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2221 switch (lrp->lr_seg.iomode) {
2227 dprintk("%s: invalid iomode %d\n", __func__,
2228 lrp->lr_seg.iomode);
2229 nfserr = nfserr_inval;
2233 switch (lrp->lr_return_type) {
2235 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2239 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2242 dprintk("%s: invalid return_type %d\n", __func__,
2243 lrp->lr_return_type);
2244 nfserr = nfserr_inval;
2250 #endif /* CONFIG_NFSD_PNFS */
2253 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2254 union nfsd4_op_u *u)
2256 struct nfsd4_getxattr *getxattr = &u->getxattr;
2258 return nfsd_getxattr(rqstp, &cstate->current_fh,
2259 getxattr->getxa_name, &getxattr->getxa_buf,
2260 &getxattr->getxa_len);
2264 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2265 union nfsd4_op_u *u)
2267 struct nfsd4_setxattr *setxattr = &u->setxattr;
2270 if (opens_in_grace(SVC_NET(rqstp)))
2271 return nfserr_grace;
2273 ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2274 setxattr->setxa_buf, setxattr->setxa_len,
2275 setxattr->setxa_flags);
2278 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2284 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2285 union nfsd4_op_u *u)
2288 * Get the entire list, then copy out only the user attributes
2289 * in the encode function.
2291 return nfsd_listxattr(rqstp, &cstate->current_fh,
2292 &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2296 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2297 union nfsd4_op_u *u)
2299 struct nfsd4_removexattr *removexattr = &u->removexattr;
2302 if (opens_in_grace(SVC_NET(rqstp)))
2303 return nfserr_grace;
2305 ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2306 removexattr->rmxa_name);
2309 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2318 nfsd4_proc_null(struct svc_rqst *rqstp)
2323 static inline void nfsd4_increment_op_stats(u32 opnum)
2325 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2326 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2329 static const struct nfsd4_operation nfsd4_ops[];
2331 static const char *nfsd4_op_name(unsigned opnum);
2334 * Enforce NFSv4.1 COMPOUND ordering rules:
2336 * Also note, enforced elsewhere:
2337 * - SEQUENCE other than as first op results in
2338 * NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2339 * - BIND_CONN_TO_SESSION must be the only op in its compound.
2340 * (Enforced in nfsd4_bind_conn_to_session().)
2341 * - DESTROY_SESSION must be the final operation in a compound, if
2342 * sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2343 * (Enforced in nfsd4_destroy_session().)
2345 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2347 struct nfsd4_op *first_op = &args->ops[0];
2349 /* These ordering requirements don't apply to NFSv4.0: */
2350 if (args->minorversion == 0)
2352 /* This is weird, but OK, not our problem: */
2353 if (args->opcnt == 0)
2355 if (first_op->status == nfserr_op_illegal)
2357 if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2358 return nfserr_op_not_in_session;
2359 if (first_op->opnum == OP_SEQUENCE)
2362 * So first_op is something allowed outside a session, like
2363 * EXCHANGE_ID; but then it has to be the only op in the
2366 if (args->opcnt != 1)
2367 return nfserr_not_only_op;
2371 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2373 return &nfsd4_ops[op->opnum];
2376 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2378 if (op->opnum == OP_ILLEGAL)
2380 return OPDESC(op)->op_flags & OP_CACHEME;
2383 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2385 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2386 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2387 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2388 struct nfsd4_op *next = &argp->ops[resp->opcnt];
2389 const struct nfsd4_operation *thisd = OPDESC(this);
2390 const struct nfsd4_operation *nextd;
2393 * Most ops check wronsec on our own; only the putfh-like ops
2394 * have special rules.
2396 if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2399 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2400 * put-filehandle operation if we're not going to use the
2403 if (argp->opcnt == resp->opcnt)
2405 if (next->opnum == OP_ILLEGAL)
2407 nextd = OPDESC(next);
2409 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2410 * errors themselves as necessary; others should check for them
2413 return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2416 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2418 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2420 struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2421 struct nfsd4_copy *copy;
2422 struct nfsd4_putfh *putfh;
2425 /* traverse all operation and if it's a COPY compound, mark the
2426 * source filehandle to skip verification
2428 for (i = 0; i < args->opcnt; i++) {
2430 if (op->opnum == OP_PUTFH)
2432 else if (op->opnum == OP_SAVEFH)
2433 saved_op = current_op;
2434 else if (op->opnum == OP_RESTOREFH)
2435 current_op = saved_op;
2436 else if (op->opnum == OP_COPY) {
2437 copy = (struct nfsd4_copy *)&op->u;
2439 op->status = nfserr_nofilehandle;
2442 putfh = (struct nfsd4_putfh *)&saved_op->u;
2443 if (!copy->cp_intra)
2444 putfh->no_verify = true;
2450 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2459 nfsd4_proc_compound(struct svc_rqst *rqstp)
2461 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2462 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2463 struct nfsd4_op *op;
2464 struct nfsd4_compound_state *cstate = &resp->cstate;
2465 struct svc_fh *current_fh = &cstate->current_fh;
2466 struct svc_fh *save_fh = &cstate->save_fh;
2467 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2470 resp->xdr = &rqstp->rq_res_stream;
2471 resp->statusp = resp->xdr->p;
2473 /* reserve space for: NFS status code */
2474 xdr_reserve_space(resp->xdr, XDR_UNIT);
2476 /* reserve space for: taglen, tag, and opcnt */
2477 xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2478 resp->taglen = args->taglen;
2479 resp->tag = args->tag;
2480 resp->rqstp = rqstp;
2481 cstate->minorversion = args->minorversion;
2482 fh_init(current_fh, NFS4_FHSIZE);
2483 fh_init(save_fh, NFS4_FHSIZE);
2485 * Don't use the deferral mechanism for NFSv4; compounds make it
2486 * too hard to avoid non-idempotency problems.
2488 clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2491 * According to RFC3010, this takes precedence over all other errors.
2493 status = nfserr_minor_vers_mismatch;
2494 if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2496 status = nfserr_resource;
2497 if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
2500 status = nfs41_check_op_ordering(args);
2503 op->status = status;
2507 check_if_stalefh_allowed(args);
2509 rqstp->rq_lease_breaker = (void **)&cstate->clp;
2511 trace_nfsd_compound(rqstp, args->opcnt);
2512 while (!status && resp->opcnt < args->opcnt) {
2513 op = &args->ops[resp->opcnt++];
2516 * The XDR decode routines may have pre-set op->status;
2517 * for example, if there is a miscellaneous XDR error
2518 * it will be set to nfserr_bad_xdr.
2521 if (op->opnum == OP_OPEN)
2522 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2525 if (!current_fh->fh_dentry &&
2526 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2527 if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2528 op->status = nfserr_nofilehandle;
2531 } else if (current_fh->fh_export &&
2532 current_fh->fh_export->ex_fslocs.migrated &&
2533 !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2534 op->status = nfserr_moved;
2538 fh_clear_pre_post_attrs(current_fh);
2540 /* If op is non-idempotent */
2541 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2543 * Don't execute this op if we couldn't encode a
2546 u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2548 * Plus if there's another operation, make sure
2549 * we'll have space to at least encode an error:
2551 if (resp->opcnt < args->opcnt)
2552 plen += COMPOUND_ERR_SLACK_SPACE;
2553 op->status = nfsd4_check_resp_size(resp, plen);
2559 if (op->opdesc->op_get_currentstateid)
2560 op->opdesc->op_get_currentstateid(cstate, &op->u);
2561 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2563 /* Only from SEQUENCE */
2564 if (cstate->status == nfserr_replay_cache) {
2565 dprintk("%s NFS4.1 replay from cache\n", __func__);
2566 status = op->status;
2570 if (op->opdesc->op_set_currentstateid)
2571 op->opdesc->op_set_currentstateid(cstate, &op->u);
2573 if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2574 clear_current_stateid(cstate);
2576 if (current_fh->fh_export &&
2577 need_wrongsec_check(rqstp))
2578 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2581 if (op->status == nfserr_replay_me) {
2582 op->replay = &cstate->replay_owner->so_replay;
2583 nfsd4_encode_replay(resp->xdr, op);
2584 status = op->status = op->replay->rp_status;
2586 nfsd4_encode_operation(resp, op);
2587 status = op->status;
2590 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2591 nfsd4_op_name(op->opnum));
2593 nfsd4_cstate_clear_replay(cstate);
2594 nfsd4_increment_op_stats(op->opnum);
2599 BUG_ON(cstate->replay_owner);
2601 cstate->status = status;
2602 /* Reset deferral mechanism for RPC deferrals */
2603 set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2607 #define op_encode_hdr_size (2)
2608 #define op_encode_stateid_maxsz (XDR_QUADLEN(NFS4_STATEID_SIZE))
2609 #define op_encode_verifier_maxsz (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2610 #define op_encode_change_info_maxsz (5)
2611 #define nfs4_fattr_bitmap_maxsz (4)
2613 /* We'll fall back on returning no lockowner if run out of space: */
2614 #define op_encode_lockowner_maxsz (0)
2615 #define op_encode_lock_denied_maxsz (8 + op_encode_lockowner_maxsz)
2617 #define nfs4_owner_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2619 #define op_encode_ace_maxsz (3 + nfs4_owner_maxsz)
2620 #define op_encode_delegation_maxsz (1 + op_encode_stateid_maxsz + 1 + \
2621 op_encode_ace_maxsz)
2623 #define op_encode_channel_attrs_maxsz (6 + 1 + 1)
2625 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2627 return (op_encode_hdr_size) * sizeof(__be32);
2630 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2632 return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2635 static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2637 /* ac_supported, ac_resp_access */
2638 return (op_encode_hdr_size + 2)* sizeof(__be32);
2641 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2643 return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2646 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2648 return (op_encode_hdr_size + op_encode_change_info_maxsz
2649 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2653 * Note since this is an idempotent operation we won't insist on failing
2654 * the op prematurely if the estimate is too large. We may turn off splice
2655 * reads unnecessarily.
2657 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2658 struct nfsd4_op *op)
2660 u32 *bmap = op->u.getattr.ga_bmval;
2661 u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2664 if (bmap0 & FATTR4_WORD0_ACL)
2665 return svc_max_payload(rqstp);
2666 if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2667 return svc_max_payload(rqstp);
2669 if (bmap1 & FATTR4_WORD1_OWNER) {
2670 ret += IDMAP_NAMESZ + 4;
2671 bmap1 &= ~FATTR4_WORD1_OWNER;
2673 if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2674 ret += IDMAP_NAMESZ + 4;
2675 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2677 if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2678 ret += NFS4_FHSIZE + 4;
2679 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2681 if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2682 ret += NFS4_MAXLABELLEN + 12;
2683 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2686 * Largest of remaining attributes are 16 bytes (e.g.,
2687 * supported_attributes)
2689 ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2690 /* bitmask, length */
2695 static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2697 return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2700 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2702 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2706 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2708 return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2712 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2714 return (op_encode_hdr_size + op_encode_stateid_maxsz
2715 + op_encode_change_info_maxsz + 1
2716 + nfs4_fattr_bitmap_maxsz
2717 + op_encode_delegation_maxsz) * sizeof(__be32);
2720 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2722 u32 maxcount = 0, rlen = 0;
2724 maxcount = svc_max_payload(rqstp);
2725 rlen = min(op->u.read.rd_length, maxcount);
2727 return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2730 static inline u32 nfsd4_read_plus_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2732 u32 maxcount = svc_max_payload(rqstp);
2733 u32 rlen = min(op->u.read.rd_length, maxcount);
2735 * If we detect that the file changed during hole encoding, then we
2736 * recover by encoding the remaining reply as data. This means we need
2737 * to set aside enough room to encode two data segments.
2739 u32 seg_len = 2 * (1 + 2 + 1);
2741 return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2744 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2746 u32 maxcount = 0, rlen = 0;
2748 maxcount = svc_max_payload(rqstp);
2749 rlen = min(op->u.readdir.rd_maxcount, maxcount);
2751 return (op_encode_hdr_size + op_encode_verifier_maxsz +
2752 XDR_QUADLEN(rlen)) * sizeof(__be32);
2755 static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2757 return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2760 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2762 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2766 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2768 return (op_encode_hdr_size + op_encode_change_info_maxsz
2769 + op_encode_change_info_maxsz) * sizeof(__be32);
2772 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2773 struct nfsd4_op *op)
2775 return (op_encode_hdr_size
2776 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2779 static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2781 return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2785 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2787 return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2790 static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2792 return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2793 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2796 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2798 return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2802 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2804 return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2807 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2809 return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2810 1 + 1 + /* eir_flags, spr_how */\
2811 4 + /* spo_must_enforce & _allow with bitmap */\
2812 2 + /*eir_server_owner.so_minor_id */\
2813 /* eir_server_owner.so_major_id<> */\
2814 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2815 /* eir_server_scope<> */\
2816 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2817 1 + /* eir_server_impl_id array length */\
2818 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2821 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2823 return (op_encode_hdr_size + \
2824 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2825 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2828 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2830 return (op_encode_hdr_size + \
2831 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2832 2 + /* csr_sequence, csr_flags */\
2833 op_encode_channel_attrs_maxsz + \
2834 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2837 static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2839 return (op_encode_hdr_size +
2840 1 /* wr_callback */ +
2841 op_encode_stateid_maxsz /* wr_callback */ +
2843 1 /* wr_committed */ +
2844 op_encode_verifier_maxsz +
2845 1 /* cr_consecutive */ +
2846 1 /* cr_synchronous */) * sizeof(__be32);
2849 static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
2850 struct nfsd4_op *op)
2852 return (op_encode_hdr_size +
2854 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
2857 static inline u32 nfsd4_copy_notify_rsize(struct svc_rqst *rqstp,
2858 struct nfsd4_op *op)
2860 return (op_encode_hdr_size +
2861 3 /* cnr_lease_time */ +
2862 1 /* We support one cnr_source_server */ +
2863 1 /* cnr_stateid seq */ +
2864 op_encode_stateid_maxsz /* cnr_stateid */ +
2865 1 /* num cnr_source_server*/ +
2868 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
2872 #ifdef CONFIG_NFSD_PNFS
2873 static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2875 u32 maxcount = 0, rlen = 0;
2877 maxcount = svc_max_payload(rqstp);
2878 rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
2880 return (op_encode_hdr_size +
2881 1 /* gd_layout_type*/ +
2883 2 /* gd_notify_types */) * sizeof(__be32);
2887 * At this stage we don't really know what layout driver will handle the request,
2888 * so we need to define an arbitrary upper bound here.
2890 #define MAX_LAYOUT_SIZE 128
2891 static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2893 return (op_encode_hdr_size +
2894 1 /* logr_return_on_close */ +
2895 op_encode_stateid_maxsz +
2896 1 /* nr of layouts */ +
2897 MAX_LAYOUT_SIZE) * sizeof(__be32);
2900 static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2902 return (op_encode_hdr_size +
2903 1 /* locr_newsize */ +
2904 2 /* ns_size */) * sizeof(__be32);
2907 static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2909 return (op_encode_hdr_size +
2910 1 /* lrs_stateid */ +
2911 op_encode_stateid_maxsz) * sizeof(__be32);
2913 #endif /* CONFIG_NFSD_PNFS */
2916 static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2918 return (op_encode_hdr_size + 3) * sizeof(__be32);
2921 static inline u32 nfsd4_getxattr_rsize(struct svc_rqst *rqstp,
2922 struct nfsd4_op *op)
2926 maxcount = svc_max_payload(rqstp);
2927 rlen = min_t(u32, XATTR_SIZE_MAX, maxcount);
2929 return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2932 static inline u32 nfsd4_setxattr_rsize(struct svc_rqst *rqstp,
2933 struct nfsd4_op *op)
2935 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2938 static inline u32 nfsd4_listxattrs_rsize(struct svc_rqst *rqstp,
2939 struct nfsd4_op *op)
2943 maxcount = svc_max_payload(rqstp);
2944 rlen = min(op->u.listxattrs.lsxa_maxcount, maxcount);
2946 return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2949 static inline u32 nfsd4_removexattr_rsize(struct svc_rqst *rqstp,
2950 struct nfsd4_op *op)
2952 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2957 static const struct nfsd4_operation nfsd4_ops[] = {
2959 .op_func = nfsd4_access,
2960 .op_name = "OP_ACCESS",
2961 .op_rsize_bop = nfsd4_access_rsize,
2964 .op_func = nfsd4_close,
2965 .op_flags = OP_MODIFIES_SOMETHING,
2966 .op_name = "OP_CLOSE",
2967 .op_rsize_bop = nfsd4_status_stateid_rsize,
2968 .op_get_currentstateid = nfsd4_get_closestateid,
2969 .op_set_currentstateid = nfsd4_set_closestateid,
2972 .op_func = nfsd4_commit,
2973 .op_flags = OP_MODIFIES_SOMETHING,
2974 .op_name = "OP_COMMIT",
2975 .op_rsize_bop = nfsd4_commit_rsize,
2978 .op_func = nfsd4_create,
2979 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
2980 .op_name = "OP_CREATE",
2981 .op_rsize_bop = nfsd4_create_rsize,
2983 [OP_DELEGRETURN] = {
2984 .op_func = nfsd4_delegreturn,
2985 .op_flags = OP_MODIFIES_SOMETHING,
2986 .op_name = "OP_DELEGRETURN",
2987 .op_rsize_bop = nfsd4_only_status_rsize,
2988 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
2991 .op_func = nfsd4_getattr,
2992 .op_flags = ALLOWED_ON_ABSENT_FS,
2993 .op_rsize_bop = nfsd4_getattr_rsize,
2994 .op_name = "OP_GETATTR",
2997 .op_func = nfsd4_getfh,
2998 .op_name = "OP_GETFH",
2999 .op_rsize_bop = nfsd4_getfh_rsize,
3002 .op_func = nfsd4_link,
3003 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3005 .op_name = "OP_LINK",
3006 .op_rsize_bop = nfsd4_link_rsize,
3009 .op_func = nfsd4_lock,
3010 .op_flags = OP_MODIFIES_SOMETHING |
3011 OP_NONTRIVIAL_ERROR_ENCODE,
3012 .op_name = "OP_LOCK",
3013 .op_rsize_bop = nfsd4_lock_rsize,
3014 .op_set_currentstateid = nfsd4_set_lockstateid,
3017 .op_func = nfsd4_lockt,
3018 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3019 .op_name = "OP_LOCKT",
3020 .op_rsize_bop = nfsd4_lock_rsize,
3023 .op_func = nfsd4_locku,
3024 .op_flags = OP_MODIFIES_SOMETHING,
3025 .op_name = "OP_LOCKU",
3026 .op_rsize_bop = nfsd4_status_stateid_rsize,
3027 .op_get_currentstateid = nfsd4_get_lockustateid,
3030 .op_func = nfsd4_lookup,
3031 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3032 .op_name = "OP_LOOKUP",
3033 .op_rsize_bop = nfsd4_only_status_rsize,
3036 .op_func = nfsd4_lookupp,
3037 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3038 .op_name = "OP_LOOKUPP",
3039 .op_rsize_bop = nfsd4_only_status_rsize,
3042 .op_func = nfsd4_nverify,
3043 .op_name = "OP_NVERIFY",
3044 .op_rsize_bop = nfsd4_only_status_rsize,
3047 .op_func = nfsd4_open,
3048 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3049 .op_name = "OP_OPEN",
3050 .op_rsize_bop = nfsd4_open_rsize,
3051 .op_set_currentstateid = nfsd4_set_openstateid,
3053 [OP_OPEN_CONFIRM] = {
3054 .op_func = nfsd4_open_confirm,
3055 .op_flags = OP_MODIFIES_SOMETHING,
3056 .op_name = "OP_OPEN_CONFIRM",
3057 .op_rsize_bop = nfsd4_status_stateid_rsize,
3059 [OP_OPEN_DOWNGRADE] = {
3060 .op_func = nfsd4_open_downgrade,
3061 .op_flags = OP_MODIFIES_SOMETHING,
3062 .op_name = "OP_OPEN_DOWNGRADE",
3063 .op_rsize_bop = nfsd4_status_stateid_rsize,
3064 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3065 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3068 .op_func = nfsd4_putfh,
3069 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3070 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3071 .op_name = "OP_PUTFH",
3072 .op_rsize_bop = nfsd4_only_status_rsize,
3075 .op_func = nfsd4_putrootfh,
3076 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3077 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3078 .op_name = "OP_PUTPUBFH",
3079 .op_rsize_bop = nfsd4_only_status_rsize,
3082 .op_func = nfsd4_putrootfh,
3083 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3084 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3085 .op_name = "OP_PUTROOTFH",
3086 .op_rsize_bop = nfsd4_only_status_rsize,
3089 .op_func = nfsd4_read,
3090 .op_release = nfsd4_read_release,
3091 .op_name = "OP_READ",
3092 .op_rsize_bop = nfsd4_read_rsize,
3093 .op_get_currentstateid = nfsd4_get_readstateid,
3096 .op_func = nfsd4_readdir,
3097 .op_name = "OP_READDIR",
3098 .op_rsize_bop = nfsd4_readdir_rsize,
3101 .op_func = nfsd4_readlink,
3102 .op_name = "OP_READLINK",
3103 .op_rsize_bop = nfsd4_readlink_rsize,
3106 .op_func = nfsd4_remove,
3107 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3108 .op_name = "OP_REMOVE",
3109 .op_rsize_bop = nfsd4_remove_rsize,
3112 .op_func = nfsd4_rename,
3113 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3114 .op_name = "OP_RENAME",
3115 .op_rsize_bop = nfsd4_rename_rsize,
3118 .op_func = nfsd4_renew,
3119 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3120 | OP_MODIFIES_SOMETHING,
3121 .op_name = "OP_RENEW",
3122 .op_rsize_bop = nfsd4_only_status_rsize,
3126 .op_func = nfsd4_restorefh,
3127 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3128 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3129 .op_name = "OP_RESTOREFH",
3130 .op_rsize_bop = nfsd4_only_status_rsize,
3133 .op_func = nfsd4_savefh,
3134 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3135 .op_name = "OP_SAVEFH",
3136 .op_rsize_bop = nfsd4_only_status_rsize,
3139 .op_func = nfsd4_secinfo,
3140 .op_release = nfsd4_secinfo_release,
3141 .op_flags = OP_HANDLES_WRONGSEC,
3142 .op_name = "OP_SECINFO",
3143 .op_rsize_bop = nfsd4_secinfo_rsize,
3146 .op_func = nfsd4_setattr,
3147 .op_name = "OP_SETATTR",
3148 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3149 | OP_NONTRIVIAL_ERROR_ENCODE,
3150 .op_rsize_bop = nfsd4_setattr_rsize,
3151 .op_get_currentstateid = nfsd4_get_setattrstateid,
3153 [OP_SETCLIENTID] = {
3154 .op_func = nfsd4_setclientid,
3155 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3156 | OP_MODIFIES_SOMETHING | OP_CACHEME
3157 | OP_NONTRIVIAL_ERROR_ENCODE,
3158 .op_name = "OP_SETCLIENTID",
3159 .op_rsize_bop = nfsd4_setclientid_rsize,
3161 [OP_SETCLIENTID_CONFIRM] = {
3162 .op_func = nfsd4_setclientid_confirm,
3163 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3164 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3165 .op_name = "OP_SETCLIENTID_CONFIRM",
3166 .op_rsize_bop = nfsd4_only_status_rsize,
3169 .op_func = nfsd4_verify,
3170 .op_name = "OP_VERIFY",
3171 .op_rsize_bop = nfsd4_only_status_rsize,
3174 .op_func = nfsd4_write,
3175 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3176 .op_name = "OP_WRITE",
3177 .op_rsize_bop = nfsd4_write_rsize,
3178 .op_get_currentstateid = nfsd4_get_writestateid,
3180 [OP_RELEASE_LOCKOWNER] = {
3181 .op_func = nfsd4_release_lockowner,
3182 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3183 | OP_MODIFIES_SOMETHING,
3184 .op_name = "OP_RELEASE_LOCKOWNER",
3185 .op_rsize_bop = nfsd4_only_status_rsize,
3188 /* NFSv4.1 operations */
3189 [OP_EXCHANGE_ID] = {
3190 .op_func = nfsd4_exchange_id,
3191 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3192 | OP_MODIFIES_SOMETHING,
3193 .op_name = "OP_EXCHANGE_ID",
3194 .op_rsize_bop = nfsd4_exchange_id_rsize,
3196 [OP_BACKCHANNEL_CTL] = {
3197 .op_func = nfsd4_backchannel_ctl,
3198 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3199 .op_name = "OP_BACKCHANNEL_CTL",
3200 .op_rsize_bop = nfsd4_only_status_rsize,
3202 [OP_BIND_CONN_TO_SESSION] = {
3203 .op_func = nfsd4_bind_conn_to_session,
3204 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3205 | OP_MODIFIES_SOMETHING,
3206 .op_name = "OP_BIND_CONN_TO_SESSION",
3207 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3209 [OP_CREATE_SESSION] = {
3210 .op_func = nfsd4_create_session,
3211 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3212 | OP_MODIFIES_SOMETHING,
3213 .op_name = "OP_CREATE_SESSION",
3214 .op_rsize_bop = nfsd4_create_session_rsize,
3216 [OP_DESTROY_SESSION] = {
3217 .op_func = nfsd4_destroy_session,
3218 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3219 | OP_MODIFIES_SOMETHING,
3220 .op_name = "OP_DESTROY_SESSION",
3221 .op_rsize_bop = nfsd4_only_status_rsize,
3224 .op_func = nfsd4_sequence,
3225 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3226 .op_name = "OP_SEQUENCE",
3227 .op_rsize_bop = nfsd4_sequence_rsize,
3229 [OP_DESTROY_CLIENTID] = {
3230 .op_func = nfsd4_destroy_clientid,
3231 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3232 | OP_MODIFIES_SOMETHING,
3233 .op_name = "OP_DESTROY_CLIENTID",
3234 .op_rsize_bop = nfsd4_only_status_rsize,
3236 [OP_RECLAIM_COMPLETE] = {
3237 .op_func = nfsd4_reclaim_complete,
3238 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3239 .op_name = "OP_RECLAIM_COMPLETE",
3240 .op_rsize_bop = nfsd4_only_status_rsize,
3242 [OP_SECINFO_NO_NAME] = {
3243 .op_func = nfsd4_secinfo_no_name,
3244 .op_release = nfsd4_secinfo_no_name_release,
3245 .op_flags = OP_HANDLES_WRONGSEC,
3246 .op_name = "OP_SECINFO_NO_NAME",
3247 .op_rsize_bop = nfsd4_secinfo_rsize,
3249 [OP_TEST_STATEID] = {
3250 .op_func = nfsd4_test_stateid,
3251 .op_flags = ALLOWED_WITHOUT_FH,
3252 .op_name = "OP_TEST_STATEID",
3253 .op_rsize_bop = nfsd4_test_stateid_rsize,
3255 [OP_FREE_STATEID] = {
3256 .op_func = nfsd4_free_stateid,
3257 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3258 .op_name = "OP_FREE_STATEID",
3259 .op_get_currentstateid = nfsd4_get_freestateid,
3260 .op_rsize_bop = nfsd4_only_status_rsize,
3262 #ifdef CONFIG_NFSD_PNFS
3263 [OP_GETDEVICEINFO] = {
3264 .op_func = nfsd4_getdeviceinfo,
3265 .op_release = nfsd4_getdeviceinfo_release,
3266 .op_flags = ALLOWED_WITHOUT_FH,
3267 .op_name = "OP_GETDEVICEINFO",
3268 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3271 .op_func = nfsd4_layoutget,
3272 .op_release = nfsd4_layoutget_release,
3273 .op_flags = OP_MODIFIES_SOMETHING,
3274 .op_name = "OP_LAYOUTGET",
3275 .op_rsize_bop = nfsd4_layoutget_rsize,
3277 [OP_LAYOUTCOMMIT] = {
3278 .op_func = nfsd4_layoutcommit,
3279 .op_flags = OP_MODIFIES_SOMETHING,
3280 .op_name = "OP_LAYOUTCOMMIT",
3281 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3283 [OP_LAYOUTRETURN] = {
3284 .op_func = nfsd4_layoutreturn,
3285 .op_flags = OP_MODIFIES_SOMETHING,
3286 .op_name = "OP_LAYOUTRETURN",
3287 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3289 #endif /* CONFIG_NFSD_PNFS */
3291 /* NFSv4.2 operations */
3293 .op_func = nfsd4_allocate,
3294 .op_flags = OP_MODIFIES_SOMETHING,
3295 .op_name = "OP_ALLOCATE",
3296 .op_rsize_bop = nfsd4_only_status_rsize,
3299 .op_func = nfsd4_deallocate,
3300 .op_flags = OP_MODIFIES_SOMETHING,
3301 .op_name = "OP_DEALLOCATE",
3302 .op_rsize_bop = nfsd4_only_status_rsize,
3305 .op_func = nfsd4_clone,
3306 .op_flags = OP_MODIFIES_SOMETHING,
3307 .op_name = "OP_CLONE",
3308 .op_rsize_bop = nfsd4_only_status_rsize,
3311 .op_func = nfsd4_copy,
3312 .op_flags = OP_MODIFIES_SOMETHING,
3313 .op_name = "OP_COPY",
3314 .op_rsize_bop = nfsd4_copy_rsize,
3317 .op_func = nfsd4_read,
3318 .op_release = nfsd4_read_release,
3319 .op_name = "OP_READ_PLUS",
3320 .op_rsize_bop = nfsd4_read_plus_rsize,
3321 .op_get_currentstateid = nfsd4_get_readstateid,
3324 .op_func = nfsd4_seek,
3325 .op_name = "OP_SEEK",
3326 .op_rsize_bop = nfsd4_seek_rsize,
3328 [OP_OFFLOAD_STATUS] = {
3329 .op_func = nfsd4_offload_status,
3330 .op_name = "OP_OFFLOAD_STATUS",
3331 .op_rsize_bop = nfsd4_offload_status_rsize,
3333 [OP_OFFLOAD_CANCEL] = {
3334 .op_func = nfsd4_offload_cancel,
3335 .op_flags = OP_MODIFIES_SOMETHING,
3336 .op_name = "OP_OFFLOAD_CANCEL",
3337 .op_rsize_bop = nfsd4_only_status_rsize,
3339 [OP_COPY_NOTIFY] = {
3340 .op_func = nfsd4_copy_notify,
3341 .op_flags = OP_MODIFIES_SOMETHING,
3342 .op_name = "OP_COPY_NOTIFY",
3343 .op_rsize_bop = nfsd4_copy_notify_rsize,
3346 .op_func = nfsd4_getxattr,
3347 .op_name = "OP_GETXATTR",
3348 .op_rsize_bop = nfsd4_getxattr_rsize,
3351 .op_func = nfsd4_setxattr,
3352 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3353 .op_name = "OP_SETXATTR",
3354 .op_rsize_bop = nfsd4_setxattr_rsize,
3357 .op_func = nfsd4_listxattrs,
3358 .op_name = "OP_LISTXATTRS",
3359 .op_rsize_bop = nfsd4_listxattrs_rsize,
3361 [OP_REMOVEXATTR] = {
3362 .op_func = nfsd4_removexattr,
3363 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3364 .op_name = "OP_REMOVEXATTR",
3365 .op_rsize_bop = nfsd4_removexattr_rsize,
3370 * nfsd4_spo_must_allow - Determine if the compound op contains an
3371 * operation that is allowed to be sent with machine credentials
3373 * @rqstp: a pointer to the struct svc_rqst
3375 * Checks to see if the compound contains a spo_must_allow op
3376 * and confirms that it was sent with the proper machine creds.
3379 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3381 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3382 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3383 struct nfsd4_op *this;
3384 struct nfsd4_compound_state *cstate = &resp->cstate;
3385 struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3388 if (!cstate->minorversion)
3391 if (cstate->spo_must_allowed)
3394 opiter = resp->opcnt;
3395 while (opiter < argp->opcnt) {
3396 this = &argp->ops[opiter++];
3397 if (test_bit(this->opnum, allow->u.longs) &&
3398 cstate->clp->cl_mach_cred &&
3399 nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3400 cstate->spo_must_allowed = true;
3404 cstate->spo_must_allowed = false;
3408 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3410 if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3411 return op_encode_hdr_size * sizeof(__be32);
3413 BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3414 return OPDESC(op)->op_rsize_bop(rqstp, op);
3417 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3419 if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3420 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3421 op->opnum, nfsd4_op_name(op->opnum));
3426 static const char *nfsd4_op_name(unsigned opnum)
3428 if (opnum < ARRAY_SIZE(nfsd4_ops))
3429 return nfsd4_ops[opnum].op_name;
3430 return "unknown_operation";
3433 static const struct svc_procedure nfsd_procedures4[2] = {
3435 .pc_func = nfsd4_proc_null,
3436 .pc_decode = nfssvc_decode_voidarg,
3437 .pc_encode = nfssvc_encode_voidres,
3438 .pc_argsize = sizeof(struct nfsd_voidargs),
3439 .pc_ressize = sizeof(struct nfsd_voidres),
3440 .pc_cachetype = RC_NOCACHE,
3444 [NFSPROC4_COMPOUND] = {
3445 .pc_func = nfsd4_proc_compound,
3446 .pc_decode = nfs4svc_decode_compoundargs,
3447 .pc_encode = nfs4svc_encode_compoundres,
3448 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3449 .pc_ressize = sizeof(struct nfsd4_compoundres),
3450 .pc_release = nfsd4_release_compoundargs,
3451 .pc_cachetype = RC_NOCACHE,
3452 .pc_xdrressize = NFSD_BUFSIZE/4,
3453 .pc_name = "COMPOUND",
3457 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
3458 const struct svc_version nfsd_version4 = {
3461 .vs_proc = nfsd_procedures4,
3462 .vs_count = nfsd_count3,
3463 .vs_dispatch = nfsd_dispatch,
3464 .vs_xdrsize = NFS4_SVC_XDRSIZE,
3465 .vs_rpcb_optnl = true,
3466 .vs_need_cong_ctrl = true,