4 * Server-side procedures for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
10 * Andy Adamson <andros@umich.edu>
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its
22 * contributors may be used to endorse or promote products derived
23 * from this software without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
38 #include <linux/param.h>
39 #include <linux/major.h>
40 #include <linux/slab.h>
41 #include <linux/file.h>
43 #include <linux/sunrpc/svc.h>
44 #include <linux/nfsd/nfsd.h>
45 #include <linux/nfsd/cache.h>
46 #include <linux/nfs4.h>
47 #include <linux/nfsd/state.h>
48 #include <linux/nfsd/xdr4.h>
49 #include <linux/nfs4_acl.h>
50 #include <linux/sunrpc/gss_api.h>
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
55 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
60 cache_get(&src->fh_export->h);
65 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
69 if (open->op_truncate &&
70 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
73 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
74 accmode |= NFSD_MAY_READ;
75 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
76 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
77 if (open->op_share_deny & NFS4_SHARE_DENY_WRITE)
78 accmode |= NFSD_MAY_WRITE;
80 status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
86 do_open_lookup(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
92 fh_init(&resfh, NFS4_FHSIZE);
93 open->op_truncate = 0;
95 if (open->op_create) {
97 * Note: create modes (UNCHECKED,GUARDED...) are the same
100 status = nfsd_create_v3(rqstp, current_fh, open->op_fname.data,
101 open->op_fname.len, &open->op_iattr,
102 &resfh, open->op_createmode,
103 (u32 *)open->op_verf.data,
104 &open->op_truncate, &created);
107 * Following rfc 3530 14.2.16, use the returned bitmask
108 * to indicate which attributes we used to store the
111 if (open->op_createmode == NFS4_CREATE_EXCLUSIVE && status == 0)
112 open->op_bmval[1] = (FATTR4_WORD1_TIME_ACCESS |
113 FATTR4_WORD1_TIME_MODIFY);
115 status = nfsd_lookup(rqstp, current_fh,
116 open->op_fname.data, open->op_fname.len, &resfh);
117 fh_unlock(current_fh);
122 set_change_info(&open->op_cinfo, current_fh);
123 fh_dup2(current_fh, &resfh);
125 /* set reply cache */
126 fh_copy_shallow(&open->op_stateowner->so_replay.rp_openfh,
129 status = do_open_permission(rqstp, current_fh, open,
138 do_open_fhandle(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
142 /* Only reclaims from previously confirmed clients are valid */
143 if ((status = nfs4_check_open_reclaim(&open->op_clientid)))
146 /* We don't know the target directory, and therefore can not
147 * set the change info
150 memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
152 /* set replay cache */
153 fh_copy_shallow(&open->op_stateowner->so_replay.rp_openfh,
154 ¤t_fh->fh_handle);
156 open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
157 (open->op_iattr.ia_size == 0);
159 status = do_open_permission(rqstp, current_fh, open,
160 NFSD_MAY_OWNER_OVERRIDE);
167 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
168 struct nfsd4_open *open)
171 dprintk("NFSD: nfsd4_open filename %.*s op_stateowner %p\n",
172 (int)open->op_fname.len, open->op_fname.data,
173 open->op_stateowner);
175 /* This check required by spec. */
176 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
181 /* check seqid for replay. set nfs4_owner */
182 status = nfsd4_process_open1(open);
183 if (status == nfserr_replay_me) {
184 struct nfs4_replay *rp = &open->op_stateowner->so_replay;
185 fh_put(&cstate->current_fh);
186 fh_copy_shallow(&cstate->current_fh.fh_handle,
188 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
190 dprintk("nfsd4_open: replay failed"
191 " restoring previous filehandle\n");
193 status = nfserr_replay_me;
198 /* Openowner is now set, so sequence id will get bumped. Now we need
199 * these checks before we do any creates: */
200 status = nfserr_grace;
201 if (locks_in_grace() && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
203 status = nfserr_no_grace;
204 if (!locks_in_grace() && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
207 switch (open->op_claim_type) {
208 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
209 case NFS4_OPEN_CLAIM_NULL:
211 * (1) set CURRENT_FH to the file being opened,
212 * creating it if necessary, (2) set open->op_cinfo,
213 * (3) set open->op_truncate if the file is to be
214 * truncated after opening, (4) do permission checking.
216 status = do_open_lookup(rqstp, &cstate->current_fh,
221 case NFS4_OPEN_CLAIM_PREVIOUS:
222 open->op_stateowner->so_confirmed = 1;
224 * The CURRENT_FH is already set to the file being
225 * opened. (1) set open->op_cinfo, (2) set
226 * open->op_truncate if the file is to be truncated
227 * after opening, (3) do permission checking.
229 status = do_open_fhandle(rqstp, &cstate->current_fh,
234 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
235 open->op_stateowner->so_confirmed = 1;
236 dprintk("NFSD: unsupported OPEN claim type %d\n",
237 open->op_claim_type);
238 status = nfserr_notsupp;
241 dprintk("NFSD: Invalid OPEN claim type %d\n",
242 open->op_claim_type);
243 status = nfserr_inval;
247 * nfsd4_process_open2() does the actual opening of the file. If
248 * successful, it (1) truncates the file if open->op_truncate was
249 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
251 status = nfsd4_process_open2(rqstp, &cstate->current_fh, open);
253 if (open->op_stateowner) {
254 nfs4_get_stateowner(open->op_stateowner);
255 cstate->replay_owner = open->op_stateowner;
262 * filehandle-manipulating ops.
265 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
266 struct svc_fh **getfh)
268 if (!cstate->current_fh.fh_dentry)
269 return nfserr_nofilehandle;
271 *getfh = &cstate->current_fh;
276 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
277 struct nfsd4_putfh *putfh)
279 fh_put(&cstate->current_fh);
280 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
281 memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
283 return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
287 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
292 fh_put(&cstate->current_fh);
293 status = exp_pseudoroot(rqstp, &cstate->current_fh);
298 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
301 if (!cstate->save_fh.fh_dentry)
302 return nfserr_restorefh;
304 fh_dup2(&cstate->current_fh, &cstate->save_fh);
309 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
312 if (!cstate->current_fh.fh_dentry)
313 return nfserr_nofilehandle;
315 fh_dup2(&cstate->save_fh, &cstate->current_fh);
323 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
324 struct nfsd4_access *access)
326 if (access->ac_req_access & ~NFS3_ACCESS_FULL)
329 access->ac_resp_access = access->ac_req_access;
330 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
331 &access->ac_supported);
335 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
336 struct nfsd4_commit *commit)
340 u32 *p = (u32 *)commit->co_verf.data;
341 *p++ = nfssvc_boot.tv_sec;
342 *p++ = nfssvc_boot.tv_usec;
344 status = nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
346 if (status == nfserr_symlink)
347 status = nfserr_inval;
352 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
353 struct nfsd4_create *create)
359 fh_init(&resfh, NFS4_FHSIZE);
361 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR,
363 if (status == nfserr_symlink)
364 status = nfserr_notdir;
368 switch (create->cr_type) {
370 /* ugh! we have to null-terminate the linktext, or
371 * vfs_symlink() will choke. it is always safe to
372 * null-terminate by brute force, since at worst we
373 * will overwrite the first byte of the create namelen
374 * in the XDR buffer, which has already been extracted
377 create->cr_linkname[create->cr_linklen] = 0;
379 status = nfsd_symlink(rqstp, &cstate->current_fh,
380 create->cr_name, create->cr_namelen,
381 create->cr_linkname, create->cr_linklen,
382 &resfh, &create->cr_iattr);
386 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
387 if (MAJOR(rdev) != create->cr_specdata1 ||
388 MINOR(rdev) != create->cr_specdata2)
390 status = nfsd_create(rqstp, &cstate->current_fh,
391 create->cr_name, create->cr_namelen,
392 &create->cr_iattr, S_IFBLK, rdev, &resfh);
396 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
397 if (MAJOR(rdev) != create->cr_specdata1 ||
398 MINOR(rdev) != create->cr_specdata2)
400 status = nfsd_create(rqstp, &cstate->current_fh,
401 create->cr_name, create->cr_namelen,
402 &create->cr_iattr,S_IFCHR, rdev, &resfh);
406 status = nfsd_create(rqstp, &cstate->current_fh,
407 create->cr_name, create->cr_namelen,
408 &create->cr_iattr, S_IFSOCK, 0, &resfh);
412 status = nfsd_create(rqstp, &cstate->current_fh,
413 create->cr_name, create->cr_namelen,
414 &create->cr_iattr, S_IFIFO, 0, &resfh);
418 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
419 status = nfsd_create(rqstp, &cstate->current_fh,
420 create->cr_name, create->cr_namelen,
421 &create->cr_iattr, S_IFDIR, 0, &resfh);
425 status = nfserr_badtype;
429 fh_unlock(&cstate->current_fh);
430 set_change_info(&create->cr_cinfo, &cstate->current_fh);
431 fh_dup2(&cstate->current_fh, &resfh);
439 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
440 struct nfsd4_getattr *getattr)
444 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
448 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
451 getattr->ga_bmval[0] &= NFSD_SUPPORTED_ATTRS_WORD0;
452 getattr->ga_bmval[1] &= NFSD_SUPPORTED_ATTRS_WORD1;
454 getattr->ga_fhp = &cstate->current_fh;
459 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
460 struct nfsd4_link *link)
462 __be32 status = nfserr_nofilehandle;
464 if (!cstate->save_fh.fh_dentry)
466 status = nfsd_link(rqstp, &cstate->current_fh,
467 link->li_name, link->li_namelen, &cstate->save_fh);
469 set_change_info(&link->li_cinfo, &cstate->current_fh);
474 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
477 struct svc_fh tmp_fh;
480 fh_init(&tmp_fh, NFS4_FHSIZE);
481 ret = exp_pseudoroot(rqstp, &tmp_fh);
484 if (tmp_fh.fh_dentry == cstate->current_fh.fh_dentry) {
489 return nfsd_lookup(rqstp, &cstate->current_fh,
490 "..", 2, &cstate->current_fh);
494 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
495 struct nfsd4_lookup *lookup)
497 return nfsd_lookup(rqstp, &cstate->current_fh,
498 lookup->lo_name, lookup->lo_len,
499 &cstate->current_fh);
503 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
504 struct nfsd4_read *read)
508 /* no need to check permission - this will be done in nfsd_read() */
510 read->rd_filp = NULL;
511 if (read->rd_offset >= OFFSET_MAX)
516 if ((status = nfs4_preprocess_stateid_op(cstate, &read->rd_stateid,
517 RD_STATE, &read->rd_filp))) {
518 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
522 get_file(read->rd_filp);
526 read->rd_rqstp = rqstp;
527 read->rd_fhp = &cstate->current_fh;
532 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
533 struct nfsd4_readdir *readdir)
535 u64 cookie = readdir->rd_cookie;
536 static const nfs4_verifier zeroverf;
538 /* no need to check permission - this will be done in nfsd_readdir() */
540 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
543 readdir->rd_bmval[0] &= NFSD_SUPPORTED_ATTRS_WORD0;
544 readdir->rd_bmval[1] &= NFSD_SUPPORTED_ATTRS_WORD1;
546 if ((cookie > ~(u32)0) || (cookie == 1) || (cookie == 2) ||
547 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
548 return nfserr_bad_cookie;
550 readdir->rd_rqstp = rqstp;
551 readdir->rd_fhp = &cstate->current_fh;
556 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
557 struct nfsd4_readlink *readlink)
559 readlink->rl_rqstp = rqstp;
560 readlink->rl_fhp = &cstate->current_fh;
565 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
566 struct nfsd4_remove *remove)
570 if (locks_in_grace())
572 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
573 remove->rm_name, remove->rm_namelen);
574 if (status == nfserr_symlink)
575 return nfserr_notdir;
577 fh_unlock(&cstate->current_fh);
578 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
584 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
585 struct nfsd4_rename *rename)
587 __be32 status = nfserr_nofilehandle;
589 if (!cstate->save_fh.fh_dentry)
591 if (locks_in_grace() && !(cstate->save_fh.fh_export->ex_flags
592 & NFSEXP_NOSUBTREECHECK))
594 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
595 rename->rn_snamelen, &cstate->current_fh,
596 rename->rn_tname, rename->rn_tnamelen);
598 /* the underlying filesystem returns different error's than required
599 * by NFSv4. both save_fh and current_fh have been verified.. */
600 if (status == nfserr_isdir)
601 status = nfserr_exist;
602 else if ((status == nfserr_notdir) &&
603 (S_ISDIR(cstate->save_fh.fh_dentry->d_inode->i_mode) &&
604 S_ISDIR(cstate->current_fh.fh_dentry->d_inode->i_mode)))
605 status = nfserr_exist;
606 else if (status == nfserr_symlink)
607 status = nfserr_notdir;
610 set_change_info(&rename->rn_sinfo, &cstate->current_fh);
611 set_change_info(&rename->rn_tinfo, &cstate->save_fh);
617 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
618 struct nfsd4_secinfo *secinfo)
621 struct svc_export *exp;
622 struct dentry *dentry;
625 fh_init(&resfh, NFS4_FHSIZE);
626 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
627 secinfo->si_name, secinfo->si_namelen,
631 if (dentry->d_inode == NULL) {
635 secinfo->si_exp = exp;
641 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
642 struct nfsd4_setattr *setattr)
644 __be32 status = nfs_ok;
646 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
648 status = nfs4_preprocess_stateid_op(cstate,
649 &setattr->sa_stateid, WR_STATE, NULL);
652 dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
656 status = mnt_want_write(cstate->current_fh.fh_export->ex_path.mnt);
660 if (setattr->sa_acl != NULL)
661 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
665 status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
668 mnt_drop_write(cstate->current_fh.fh_export->ex_path.mnt);
673 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
674 struct nfsd4_write *write)
676 stateid_t *stateid = &write->wr_stateid;
677 struct file *filp = NULL;
679 __be32 status = nfs_ok;
682 /* no need to check permission - this will be done in nfsd_write() */
684 if (write->wr_offset >= OFFSET_MAX)
688 status = nfs4_preprocess_stateid_op(cstate, stateid, WR_STATE, &filp);
694 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
698 cnt = write->wr_buflen;
699 write->wr_how_written = write->wr_stable_how;
700 p = (u32 *)write->wr_verifier.data;
701 *p++ = nfssvc_boot.tv_sec;
702 *p++ = nfssvc_boot.tv_usec;
704 status = nfsd_write(rqstp, &cstate->current_fh, filp,
705 write->wr_offset, rqstp->rq_vec, write->wr_vlen,
706 &cnt, &write->wr_how_written);
710 write->wr_bytes_written = cnt;
712 if (status == nfserr_symlink)
713 status = nfserr_inval;
717 /* This routine never returns NFS_OK! If there are no other errors, it
718 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
719 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
720 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
723 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
724 struct nfsd4_verify *verify)
730 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
734 if ((verify->ve_bmval[0] & ~NFSD_SUPPORTED_ATTRS_WORD0)
735 || (verify->ve_bmval[1] & ~NFSD_SUPPORTED_ATTRS_WORD1))
736 return nfserr_attrnotsupp;
737 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
738 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
740 if (verify->ve_attrlen & 3)
744 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
746 count = 4 + (verify->ve_attrlen >> 2);
747 buf = kmalloc(count << 2, GFP_KERNEL);
749 return nfserr_resource;
751 status = nfsd4_encode_fattr(&cstate->current_fh,
752 cstate->current_fh.fh_export,
753 cstate->current_fh.fh_dentry, buf,
754 &count, verify->ve_bmval,
757 /* this means that nfsd4_encode_fattr() ran out of space */
758 if (status == nfserr_resource && count == 0)
759 status = nfserr_not_same;
764 status = nfserr_not_same;
765 if (ntohl(*p++) != verify->ve_attrlen)
767 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
768 status = nfserr_same;
776 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
777 struct nfsd4_verify *verify)
781 status = _nfsd4_verify(rqstp, cstate, verify);
782 return status == nfserr_not_same ? nfs_ok : status;
786 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
787 struct nfsd4_verify *verify)
791 status = _nfsd4_verify(rqstp, cstate, verify);
792 return status == nfserr_same ? nfs_ok : status;
799 nfsd4_proc_null(struct svc_rqst *rqstp, void *argp, void *resp)
804 static inline void nfsd4_increment_op_stats(u32 opnum)
806 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
807 nfsdstats.nfs4_opcount[opnum]++;
810 typedef __be32(*nfsd4op_func)(struct svc_rqst *, struct nfsd4_compound_state *,
812 enum nfsd4_op_flags {
813 ALLOWED_WITHOUT_FH = 1 << 0, /* No current filehandle required */
814 ALLOWED_ON_ABSENT_FS = 2 << 0, /* ops processed on absent fs */
815 ALLOWED_AS_FIRST_OP = 3 << 0, /* ops reqired first in compound */
818 struct nfsd4_operation {
819 nfsd4op_func op_func;
824 static struct nfsd4_operation nfsd4_ops[];
826 static const char *nfsd4_op_name(unsigned opnum);
829 * This is a replay of a compound for which no cache entry pages
830 * were used. Encode the sequence operation, and if cachethis is FALSE
831 * encode the uncache rep error on the next operation.
834 nfsd4_enc_uncached_replay(struct nfsd4_compoundargs *args,
835 struct nfsd4_compoundres *resp)
839 dprintk("--> %s resp->opcnt %d ce_cachethis %u \n", __func__,
840 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
842 /* Encode the replayed sequence operation */
843 BUG_ON(resp->opcnt != 1);
844 op = &args->ops[resp->opcnt - 1];
845 nfsd4_encode_operation(resp, op);
847 /*return nfserr_retry_uncached_rep in next operation. */
848 if (resp->cstate.slot->sl_cache_entry.ce_cachethis == 0) {
849 op = &args->ops[resp->opcnt++];
850 op->status = nfserr_retry_uncached_rep;
851 nfsd4_encode_operation(resp, op);
857 * Enforce NFSv4.1 COMPOUND ordering rules.
860 * - enforce NFS4ERR_NOT_ONLY_OP,
861 * - DESTROY_SESSION MUST be the final operation in the COMPOUND request.
863 static bool nfs41_op_ordering_ok(struct nfsd4_compoundargs *args)
865 if (args->minorversion && args->opcnt > 0) {
866 struct nfsd4_op *op = &args->ops[0];
867 return (op->status == nfserr_op_illegal) ||
868 (nfsd4_ops[op->opnum].op_flags & ALLOWED_AS_FIRST_OP);
877 nfsd4_proc_compound(struct svc_rqst *rqstp,
878 struct nfsd4_compoundargs *args,
879 struct nfsd4_compoundres *resp)
882 struct nfsd4_operation *opdesc;
883 struct nfsd4_compound_state *cstate = &resp->cstate;
887 resp->xbuf = &rqstp->rq_res;
888 resp->p = rqstp->rq_res.head[0].iov_base +
889 rqstp->rq_res.head[0].iov_len;
890 resp->tagp = resp->p;
891 /* reserve space for: taglen, tag, and opcnt */
892 resp->p += 2 + XDR_QUADLEN(args->taglen);
893 resp->end = rqstp->rq_res.head[0].iov_base + PAGE_SIZE;
894 resp->taglen = args->taglen;
895 resp->tag = args->tag;
898 resp->cstate.replay_owner = NULL;
899 fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
900 fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
901 /* Use the deferral mechanism only for NFSv4.0 compounds */
902 rqstp->rq_usedeferral = (args->minorversion == 0);
905 * According to RFC3010, this takes precedence over all other errors.
907 status = nfserr_minor_vers_mismatch;
908 if (args->minorversion > NFSD_SUPPORTED_MINOR_VERSION)
911 if (!nfs41_op_ordering_ok(args)) {
913 op->status = nfserr_sequence_pos;
918 while (!status && resp->opcnt < args->opcnt) {
919 op = &args->ops[resp->opcnt++];
921 dprintk("nfsv4 compound op #%d/%d: %d (%s)\n",
922 resp->opcnt, args->opcnt, op->opnum,
923 nfsd4_op_name(op->opnum));
925 * The XDR decode routines may have pre-set op->status;
926 * for example, if there is a miscellaneous XDR error
927 * it will be set to nfserr_bad_xdr.
932 /* We must be able to encode a successful response to
933 * this operation, with enough room left over to encode a
934 * failed response to the next operation. If we don't
935 * have enough room, fail with ERR_RESOURCE.
937 slack_bytes = (char *)resp->end - (char *)resp->p;
938 if (slack_bytes < COMPOUND_SLACK_SPACE
939 + COMPOUND_ERR_SLACK_SPACE) {
940 BUG_ON(slack_bytes < COMPOUND_ERR_SLACK_SPACE);
941 op->status = nfserr_resource;
945 opdesc = &nfsd4_ops[op->opnum];
947 if (!cstate->current_fh.fh_dentry) {
948 if (!(opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
949 op->status = nfserr_nofilehandle;
952 } else if (cstate->current_fh.fh_export->ex_fslocs.migrated &&
953 !(opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
954 op->status = nfserr_moved;
959 op->status = opdesc->op_func(rqstp, cstate, &op->u);
961 BUG_ON(op->status == nfs_ok);
964 /* Only from SEQUENCE or CREATE_SESSION */
965 if (resp->cstate.status == nfserr_replay_cache) {
966 dprintk("%s NFS4.1 replay from cache\n", __func__);
967 if (nfsd4_not_cached(resp))
968 status = nfsd4_enc_uncached_replay(args, resp);
973 if (op->status == nfserr_replay_me) {
974 op->replay = &cstate->replay_owner->so_replay;
975 nfsd4_encode_replay(resp, op);
976 status = op->status = op->replay->rp_status;
978 nfsd4_encode_operation(resp, op);
982 dprintk("nfsv4 compound op %p opcnt %d #%d: %d: status %d\n",
983 args->ops, args->opcnt, resp->opcnt, op->opnum,
984 be32_to_cpu(status));
986 if (cstate->replay_owner) {
987 nfs4_put_stateowner(cstate->replay_owner);
988 cstate->replay_owner = NULL;
990 /* XXX Ugh, we need to get rid of this kind of special case: */
991 if (op->opnum == OP_READ && op->u.read.rd_filp)
992 fput(op->u.read.rd_filp);
994 nfsd4_increment_op_stats(op->opnum);
996 if (!rqstp->rq_usedeferral && status == nfserr_dropit) {
997 dprintk("%s Dropit - send NFS4ERR_DELAY\n", __func__);
998 status = nfserr_jukebox;
1001 resp->cstate.status = status;
1002 fh_put(&resp->cstate.current_fh);
1003 fh_put(&resp->cstate.save_fh);
1004 BUG_ON(resp->cstate.replay_owner);
1006 nfsd4_release_compoundargs(args);
1007 /* Reset deferral mechanism for RPC deferrals */
1008 rqstp->rq_usedeferral = 1;
1009 dprintk("nfsv4 compound returned %d\n", ntohl(status));
1013 static struct nfsd4_operation nfsd4_ops[] = {
1015 .op_func = (nfsd4op_func)nfsd4_access,
1016 .op_name = "OP_ACCESS",
1019 .op_func = (nfsd4op_func)nfsd4_close,
1020 .op_name = "OP_CLOSE",
1023 .op_func = (nfsd4op_func)nfsd4_commit,
1024 .op_name = "OP_COMMIT",
1027 .op_func = (nfsd4op_func)nfsd4_create,
1028 .op_name = "OP_CREATE",
1030 [OP_DELEGRETURN] = {
1031 .op_func = (nfsd4op_func)nfsd4_delegreturn,
1032 .op_name = "OP_DELEGRETURN",
1035 .op_func = (nfsd4op_func)nfsd4_getattr,
1036 .op_flags = ALLOWED_ON_ABSENT_FS,
1037 .op_name = "OP_GETATTR",
1040 .op_func = (nfsd4op_func)nfsd4_getfh,
1041 .op_name = "OP_GETFH",
1044 .op_func = (nfsd4op_func)nfsd4_link,
1045 .op_name = "OP_LINK",
1048 .op_func = (nfsd4op_func)nfsd4_lock,
1049 .op_name = "OP_LOCK",
1052 .op_func = (nfsd4op_func)nfsd4_lockt,
1053 .op_name = "OP_LOCKT",
1056 .op_func = (nfsd4op_func)nfsd4_locku,
1057 .op_name = "OP_LOCKU",
1060 .op_func = (nfsd4op_func)nfsd4_lookup,
1061 .op_name = "OP_LOOKUP",
1064 .op_func = (nfsd4op_func)nfsd4_lookupp,
1065 .op_name = "OP_LOOKUPP",
1068 .op_func = (nfsd4op_func)nfsd4_nverify,
1069 .op_name = "OP_NVERIFY",
1072 .op_func = (nfsd4op_func)nfsd4_open,
1073 .op_name = "OP_OPEN",
1075 [OP_OPEN_CONFIRM] = {
1076 .op_func = (nfsd4op_func)nfsd4_open_confirm,
1077 .op_name = "OP_OPEN_CONFIRM",
1079 [OP_OPEN_DOWNGRADE] = {
1080 .op_func = (nfsd4op_func)nfsd4_open_downgrade,
1081 .op_name = "OP_OPEN_DOWNGRADE",
1084 .op_func = (nfsd4op_func)nfsd4_putfh,
1085 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1086 .op_name = "OP_PUTFH",
1089 .op_func = (nfsd4op_func)nfsd4_putrootfh,
1090 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1091 .op_name = "OP_PUTPUBFH",
1094 .op_func = (nfsd4op_func)nfsd4_putrootfh,
1095 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1096 .op_name = "OP_PUTROOTFH",
1099 .op_func = (nfsd4op_func)nfsd4_read,
1100 .op_name = "OP_READ",
1103 .op_func = (nfsd4op_func)nfsd4_readdir,
1104 .op_name = "OP_READDIR",
1107 .op_func = (nfsd4op_func)nfsd4_readlink,
1108 .op_name = "OP_READLINK",
1111 .op_func = (nfsd4op_func)nfsd4_remove,
1112 .op_name = "OP_REMOVE",
1115 .op_name = "OP_RENAME",
1116 .op_func = (nfsd4op_func)nfsd4_rename,
1119 .op_func = (nfsd4op_func)nfsd4_renew,
1120 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1121 .op_name = "OP_RENEW",
1124 .op_func = (nfsd4op_func)nfsd4_restorefh,
1125 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1126 .op_name = "OP_RESTOREFH",
1129 .op_func = (nfsd4op_func)nfsd4_savefh,
1130 .op_name = "OP_SAVEFH",
1133 .op_func = (nfsd4op_func)nfsd4_secinfo,
1134 .op_name = "OP_SECINFO",
1137 .op_func = (nfsd4op_func)nfsd4_setattr,
1138 .op_name = "OP_SETATTR",
1140 [OP_SETCLIENTID] = {
1141 .op_func = (nfsd4op_func)nfsd4_setclientid,
1142 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1143 .op_name = "OP_SETCLIENTID",
1145 [OP_SETCLIENTID_CONFIRM] = {
1146 .op_func = (nfsd4op_func)nfsd4_setclientid_confirm,
1147 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1148 .op_name = "OP_SETCLIENTID_CONFIRM",
1151 .op_func = (nfsd4op_func)nfsd4_verify,
1152 .op_name = "OP_VERIFY",
1155 .op_func = (nfsd4op_func)nfsd4_write,
1156 .op_name = "OP_WRITE",
1158 [OP_RELEASE_LOCKOWNER] = {
1159 .op_func = (nfsd4op_func)nfsd4_release_lockowner,
1160 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1161 .op_name = "OP_RELEASE_LOCKOWNER",
1164 /* NFSv4.1 operations */
1165 [OP_EXCHANGE_ID] = {
1166 .op_func = (nfsd4op_func)nfsd4_exchange_id,
1167 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1168 .op_name = "OP_EXCHANGE_ID",
1170 [OP_CREATE_SESSION] = {
1171 .op_func = (nfsd4op_func)nfsd4_create_session,
1172 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1173 .op_name = "OP_CREATE_SESSION",
1175 [OP_DESTROY_SESSION] = {
1176 .op_func = (nfsd4op_func)nfsd4_destroy_session,
1177 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1178 .op_name = "OP_DESTROY_SESSION",
1181 .op_func = (nfsd4op_func)nfsd4_sequence,
1182 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1183 .op_name = "OP_SEQUENCE",
1187 static const char *nfsd4_op_name(unsigned opnum)
1189 if (opnum < ARRAY_SIZE(nfsd4_ops))
1190 return nfsd4_ops[opnum].op_name;
1191 return "unknown_operation";
1194 #define nfs4svc_decode_voidargs NULL
1195 #define nfs4svc_release_void NULL
1196 #define nfsd4_voidres nfsd4_voidargs
1197 #define nfs4svc_release_compound NULL
1198 struct nfsd4_voidargs { int dummy; };
1200 #define PROC(name, argt, rest, relt, cache, respsize) \
1201 { (svc_procfunc) nfsd4_proc_##name, \
1202 (kxdrproc_t) nfs4svc_decode_##argt##args, \
1203 (kxdrproc_t) nfs4svc_encode_##rest##res, \
1204 (kxdrproc_t) nfs4svc_release_##relt, \
1205 sizeof(struct nfsd4_##argt##args), \
1206 sizeof(struct nfsd4_##rest##res), \
1213 * TODO: At the present time, the NFSv4 server does not do XID caching
1214 * of requests. Implementing XID caching would not be a serious problem,
1215 * although it would require a mild change in interfaces since one
1216 * doesn't know whether an NFSv4 request is idempotent until after the
1217 * XDR decode. However, XID caching totally confuses pynfs (Peter
1218 * Astrand's regression testsuite for NFSv4 servers), which reuses
1219 * XID's liberally, so I've left it unimplemented until pynfs generates
1222 static struct svc_procedure nfsd_procedures4[2] = {
1223 PROC(null, void, void, void, RC_NOCACHE, 1),
1224 PROC(compound, compound, compound, compound, RC_NOCACHE, NFSD_BUFSIZE/4)
1227 struct svc_version nfsd_version4 = {
1230 .vs_proc = nfsd_procedures4,
1231 .vs_dispatch = nfsd_dispatch,
1232 .vs_xdrsize = NFS4_SVC_XDRSIZE,