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 struct nfsd4_compoundres *resp;
173 dprintk("NFSD: nfsd4_open filename %.*s op_stateowner %p\n",
174 (int)open->op_fname.len, open->op_fname.data,
175 open->op_stateowner);
177 /* This check required by spec. */
178 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
183 /* check seqid for replay. set nfs4_owner */
184 resp = rqstp->rq_resp;
185 status = nfsd4_process_open1(&resp->cstate, open);
186 if (status == nfserr_replay_me) {
187 struct nfs4_replay *rp = &open->op_stateowner->so_replay;
188 fh_put(&cstate->current_fh);
189 fh_copy_shallow(&cstate->current_fh.fh_handle,
191 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
193 dprintk("nfsd4_open: replay failed"
194 " restoring previous filehandle\n");
196 status = nfserr_replay_me;
201 /* Openowner is now set, so sequence id will get bumped. Now we need
202 * these checks before we do any creates: */
203 status = nfserr_grace;
204 if (locks_in_grace() && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
206 status = nfserr_no_grace;
207 if (!locks_in_grace() && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
210 switch (open->op_claim_type) {
211 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
212 case NFS4_OPEN_CLAIM_NULL:
214 * (1) set CURRENT_FH to the file being opened,
215 * creating it if necessary, (2) set open->op_cinfo,
216 * (3) set open->op_truncate if the file is to be
217 * truncated after opening, (4) do permission checking.
219 status = do_open_lookup(rqstp, &cstate->current_fh,
224 case NFS4_OPEN_CLAIM_PREVIOUS:
225 open->op_stateowner->so_confirmed = 1;
227 * The CURRENT_FH is already set to the file being
228 * opened. (1) set open->op_cinfo, (2) set
229 * open->op_truncate if the file is to be truncated
230 * after opening, (3) do permission checking.
232 status = do_open_fhandle(rqstp, &cstate->current_fh,
237 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
238 open->op_stateowner->so_confirmed = 1;
239 dprintk("NFSD: unsupported OPEN claim type %d\n",
240 open->op_claim_type);
241 status = nfserr_notsupp;
244 dprintk("NFSD: Invalid OPEN claim type %d\n",
245 open->op_claim_type);
246 status = nfserr_inval;
250 * nfsd4_process_open2() does the actual opening of the file. If
251 * successful, it (1) truncates the file if open->op_truncate was
252 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
254 status = nfsd4_process_open2(rqstp, &cstate->current_fh, open);
256 if (open->op_stateowner) {
257 nfs4_get_stateowner(open->op_stateowner);
258 cstate->replay_owner = open->op_stateowner;
265 * filehandle-manipulating ops.
268 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
269 struct svc_fh **getfh)
271 if (!cstate->current_fh.fh_dentry)
272 return nfserr_nofilehandle;
274 *getfh = &cstate->current_fh;
279 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
280 struct nfsd4_putfh *putfh)
282 fh_put(&cstate->current_fh);
283 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
284 memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
286 return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
290 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
295 fh_put(&cstate->current_fh);
296 status = exp_pseudoroot(rqstp, &cstate->current_fh);
301 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
304 if (!cstate->save_fh.fh_dentry)
305 return nfserr_restorefh;
307 fh_dup2(&cstate->current_fh, &cstate->save_fh);
312 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
315 if (!cstate->current_fh.fh_dentry)
316 return nfserr_nofilehandle;
318 fh_dup2(&cstate->save_fh, &cstate->current_fh);
326 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
327 struct nfsd4_access *access)
329 if (access->ac_req_access & ~NFS3_ACCESS_FULL)
332 access->ac_resp_access = access->ac_req_access;
333 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
334 &access->ac_supported);
338 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
339 struct nfsd4_commit *commit)
343 u32 *p = (u32 *)commit->co_verf.data;
344 *p++ = nfssvc_boot.tv_sec;
345 *p++ = nfssvc_boot.tv_usec;
347 status = nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
349 if (status == nfserr_symlink)
350 status = nfserr_inval;
355 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
356 struct nfsd4_create *create)
362 fh_init(&resfh, NFS4_FHSIZE);
364 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR,
366 if (status == nfserr_symlink)
367 status = nfserr_notdir;
371 switch (create->cr_type) {
373 /* ugh! we have to null-terminate the linktext, or
374 * vfs_symlink() will choke. it is always safe to
375 * null-terminate by brute force, since at worst we
376 * will overwrite the first byte of the create namelen
377 * in the XDR buffer, which has already been extracted
380 create->cr_linkname[create->cr_linklen] = 0;
382 status = nfsd_symlink(rqstp, &cstate->current_fh,
383 create->cr_name, create->cr_namelen,
384 create->cr_linkname, create->cr_linklen,
385 &resfh, &create->cr_iattr);
389 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
390 if (MAJOR(rdev) != create->cr_specdata1 ||
391 MINOR(rdev) != create->cr_specdata2)
393 status = nfsd_create(rqstp, &cstate->current_fh,
394 create->cr_name, create->cr_namelen,
395 &create->cr_iattr, S_IFBLK, rdev, &resfh);
399 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
400 if (MAJOR(rdev) != create->cr_specdata1 ||
401 MINOR(rdev) != create->cr_specdata2)
403 status = nfsd_create(rqstp, &cstate->current_fh,
404 create->cr_name, create->cr_namelen,
405 &create->cr_iattr,S_IFCHR, rdev, &resfh);
409 status = nfsd_create(rqstp, &cstate->current_fh,
410 create->cr_name, create->cr_namelen,
411 &create->cr_iattr, S_IFSOCK, 0, &resfh);
415 status = nfsd_create(rqstp, &cstate->current_fh,
416 create->cr_name, create->cr_namelen,
417 &create->cr_iattr, S_IFIFO, 0, &resfh);
421 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
422 status = nfsd_create(rqstp, &cstate->current_fh,
423 create->cr_name, create->cr_namelen,
424 &create->cr_iattr, S_IFDIR, 0, &resfh);
428 status = nfserr_badtype;
432 fh_unlock(&cstate->current_fh);
433 set_change_info(&create->cr_cinfo, &cstate->current_fh);
434 fh_dup2(&cstate->current_fh, &resfh);
442 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
443 struct nfsd4_getattr *getattr)
447 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
451 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
454 getattr->ga_bmval[0] &= NFSD_SUPPORTED_ATTRS_WORD0;
455 getattr->ga_bmval[1] &= NFSD_SUPPORTED_ATTRS_WORD1;
457 getattr->ga_fhp = &cstate->current_fh;
462 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
463 struct nfsd4_link *link)
465 __be32 status = nfserr_nofilehandle;
467 if (!cstate->save_fh.fh_dentry)
469 status = nfsd_link(rqstp, &cstate->current_fh,
470 link->li_name, link->li_namelen, &cstate->save_fh);
472 set_change_info(&link->li_cinfo, &cstate->current_fh);
477 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
480 struct svc_fh tmp_fh;
483 fh_init(&tmp_fh, NFS4_FHSIZE);
484 ret = exp_pseudoroot(rqstp, &tmp_fh);
487 if (tmp_fh.fh_dentry == cstate->current_fh.fh_dentry) {
492 return nfsd_lookup(rqstp, &cstate->current_fh,
493 "..", 2, &cstate->current_fh);
497 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
498 struct nfsd4_lookup *lookup)
500 return nfsd_lookup(rqstp, &cstate->current_fh,
501 lookup->lo_name, lookup->lo_len,
502 &cstate->current_fh);
506 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
507 struct nfsd4_read *read)
511 /* no need to check permission - this will be done in nfsd_read() */
513 read->rd_filp = NULL;
514 if (read->rd_offset >= OFFSET_MAX)
519 if ((status = nfs4_preprocess_stateid_op(cstate, &read->rd_stateid,
520 RD_STATE, &read->rd_filp))) {
521 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
525 get_file(read->rd_filp);
529 read->rd_rqstp = rqstp;
530 read->rd_fhp = &cstate->current_fh;
535 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
536 struct nfsd4_readdir *readdir)
538 u64 cookie = readdir->rd_cookie;
539 static const nfs4_verifier zeroverf;
541 /* no need to check permission - this will be done in nfsd_readdir() */
543 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
546 readdir->rd_bmval[0] &= NFSD_SUPPORTED_ATTRS_WORD0;
547 readdir->rd_bmval[1] &= NFSD_SUPPORTED_ATTRS_WORD1;
549 if ((cookie > ~(u32)0) || (cookie == 1) || (cookie == 2) ||
550 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
551 return nfserr_bad_cookie;
553 readdir->rd_rqstp = rqstp;
554 readdir->rd_fhp = &cstate->current_fh;
559 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
560 struct nfsd4_readlink *readlink)
562 readlink->rl_rqstp = rqstp;
563 readlink->rl_fhp = &cstate->current_fh;
568 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
569 struct nfsd4_remove *remove)
573 if (locks_in_grace())
575 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
576 remove->rm_name, remove->rm_namelen);
577 if (status == nfserr_symlink)
578 return nfserr_notdir;
580 fh_unlock(&cstate->current_fh);
581 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
587 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
588 struct nfsd4_rename *rename)
590 __be32 status = nfserr_nofilehandle;
592 if (!cstate->save_fh.fh_dentry)
594 if (locks_in_grace() && !(cstate->save_fh.fh_export->ex_flags
595 & NFSEXP_NOSUBTREECHECK))
597 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
598 rename->rn_snamelen, &cstate->current_fh,
599 rename->rn_tname, rename->rn_tnamelen);
601 /* the underlying filesystem returns different error's than required
602 * by NFSv4. both save_fh and current_fh have been verified.. */
603 if (status == nfserr_isdir)
604 status = nfserr_exist;
605 else if ((status == nfserr_notdir) &&
606 (S_ISDIR(cstate->save_fh.fh_dentry->d_inode->i_mode) &&
607 S_ISDIR(cstate->current_fh.fh_dentry->d_inode->i_mode)))
608 status = nfserr_exist;
609 else if (status == nfserr_symlink)
610 status = nfserr_notdir;
613 set_change_info(&rename->rn_sinfo, &cstate->current_fh);
614 set_change_info(&rename->rn_tinfo, &cstate->save_fh);
620 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
621 struct nfsd4_secinfo *secinfo)
624 struct svc_export *exp;
625 struct dentry *dentry;
628 fh_init(&resfh, NFS4_FHSIZE);
629 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
630 secinfo->si_name, secinfo->si_namelen,
634 if (dentry->d_inode == NULL) {
638 secinfo->si_exp = exp;
644 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
645 struct nfsd4_setattr *setattr)
647 __be32 status = nfs_ok;
649 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
651 status = nfs4_preprocess_stateid_op(cstate,
652 &setattr->sa_stateid, WR_STATE, NULL);
655 dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
659 status = mnt_want_write(cstate->current_fh.fh_export->ex_path.mnt);
663 if (setattr->sa_acl != NULL)
664 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
668 status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
671 mnt_drop_write(cstate->current_fh.fh_export->ex_path.mnt);
676 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
677 struct nfsd4_write *write)
679 stateid_t *stateid = &write->wr_stateid;
680 struct file *filp = NULL;
682 __be32 status = nfs_ok;
685 /* no need to check permission - this will be done in nfsd_write() */
687 if (write->wr_offset >= OFFSET_MAX)
691 status = nfs4_preprocess_stateid_op(cstate, stateid, WR_STATE, &filp);
697 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
701 cnt = write->wr_buflen;
702 write->wr_how_written = write->wr_stable_how;
703 p = (u32 *)write->wr_verifier.data;
704 *p++ = nfssvc_boot.tv_sec;
705 *p++ = nfssvc_boot.tv_usec;
707 status = nfsd_write(rqstp, &cstate->current_fh, filp,
708 write->wr_offset, rqstp->rq_vec, write->wr_vlen,
709 &cnt, &write->wr_how_written);
713 write->wr_bytes_written = cnt;
715 if (status == nfserr_symlink)
716 status = nfserr_inval;
720 /* This routine never returns NFS_OK! If there are no other errors, it
721 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
722 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
723 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
726 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
727 struct nfsd4_verify *verify)
733 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
737 if ((verify->ve_bmval[0] & ~NFSD_SUPPORTED_ATTRS_WORD0)
738 || (verify->ve_bmval[1] & ~NFSD_SUPPORTED_ATTRS_WORD1))
739 return nfserr_attrnotsupp;
740 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
741 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
743 if (verify->ve_attrlen & 3)
747 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
749 count = 4 + (verify->ve_attrlen >> 2);
750 buf = kmalloc(count << 2, GFP_KERNEL);
752 return nfserr_resource;
754 status = nfsd4_encode_fattr(&cstate->current_fh,
755 cstate->current_fh.fh_export,
756 cstate->current_fh.fh_dentry, buf,
757 &count, verify->ve_bmval,
760 /* this means that nfsd4_encode_fattr() ran out of space */
761 if (status == nfserr_resource && count == 0)
762 status = nfserr_not_same;
767 status = nfserr_not_same;
768 if (ntohl(*p++) != verify->ve_attrlen)
770 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
771 status = nfserr_same;
779 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
780 struct nfsd4_verify *verify)
784 status = _nfsd4_verify(rqstp, cstate, verify);
785 return status == nfserr_not_same ? nfs_ok : status;
789 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
790 struct nfsd4_verify *verify)
794 status = _nfsd4_verify(rqstp, cstate, verify);
795 return status == nfserr_same ? nfs_ok : status;
802 nfsd4_proc_null(struct svc_rqst *rqstp, void *argp, void *resp)
807 static inline void nfsd4_increment_op_stats(u32 opnum)
809 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
810 nfsdstats.nfs4_opcount[opnum]++;
813 typedef __be32(*nfsd4op_func)(struct svc_rqst *, struct nfsd4_compound_state *,
815 enum nfsd4_op_flags {
816 ALLOWED_WITHOUT_FH = 1 << 0, /* No current filehandle required */
817 ALLOWED_ON_ABSENT_FS = 2 << 0, /* ops processed on absent fs */
818 ALLOWED_AS_FIRST_OP = 3 << 0, /* ops reqired first in compound */
821 struct nfsd4_operation {
822 nfsd4op_func op_func;
827 static struct nfsd4_operation nfsd4_ops[];
829 static const char *nfsd4_op_name(unsigned opnum);
832 * This is a replay of a compound for which no cache entry pages
833 * were used. Encode the sequence operation, and if cachethis is FALSE
834 * encode the uncache rep error on the next operation.
837 nfsd4_enc_uncached_replay(struct nfsd4_compoundargs *args,
838 struct nfsd4_compoundres *resp)
842 dprintk("--> %s resp->opcnt %d ce_cachethis %u \n", __func__,
843 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
845 /* Encode the replayed sequence operation */
846 BUG_ON(resp->opcnt != 1);
847 op = &args->ops[resp->opcnt - 1];
848 nfsd4_encode_operation(resp, op);
850 /*return nfserr_retry_uncached_rep in next operation. */
851 if (resp->cstate.slot->sl_cache_entry.ce_cachethis == 0) {
852 op = &args->ops[resp->opcnt++];
853 op->status = nfserr_retry_uncached_rep;
854 nfsd4_encode_operation(resp, op);
860 * Enforce NFSv4.1 COMPOUND ordering rules.
863 * - enforce NFS4ERR_NOT_ONLY_OP,
864 * - DESTROY_SESSION MUST be the final operation in the COMPOUND request.
866 static bool nfs41_op_ordering_ok(struct nfsd4_compoundargs *args)
868 if (args->minorversion && args->opcnt > 0) {
869 struct nfsd4_op *op = &args->ops[0];
870 return (op->status == nfserr_op_illegal) ||
871 (nfsd4_ops[op->opnum].op_flags & ALLOWED_AS_FIRST_OP);
880 nfsd4_proc_compound(struct svc_rqst *rqstp,
881 struct nfsd4_compoundargs *args,
882 struct nfsd4_compoundres *resp)
885 struct nfsd4_operation *opdesc;
886 struct nfsd4_compound_state *cstate = &resp->cstate;
890 resp->xbuf = &rqstp->rq_res;
891 resp->p = rqstp->rq_res.head[0].iov_base +
892 rqstp->rq_res.head[0].iov_len;
893 resp->tagp = resp->p;
894 /* reserve space for: taglen, tag, and opcnt */
895 resp->p += 2 + XDR_QUADLEN(args->taglen);
896 resp->end = rqstp->rq_res.head[0].iov_base + PAGE_SIZE;
897 resp->taglen = args->taglen;
898 resp->tag = args->tag;
901 resp->cstate.replay_owner = NULL;
902 fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
903 fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
904 /* Use the deferral mechanism only for NFSv4.0 compounds */
905 rqstp->rq_usedeferral = (args->minorversion == 0);
908 * According to RFC3010, this takes precedence over all other errors.
910 status = nfserr_minor_vers_mismatch;
911 if (args->minorversion > NFSD_SUPPORTED_MINOR_VERSION)
914 if (!nfs41_op_ordering_ok(args)) {
916 op->status = nfserr_sequence_pos;
921 while (!status && resp->opcnt < args->opcnt) {
922 op = &args->ops[resp->opcnt++];
924 dprintk("nfsv4 compound op #%d/%d: %d (%s)\n",
925 resp->opcnt, args->opcnt, op->opnum,
926 nfsd4_op_name(op->opnum));
928 * The XDR decode routines may have pre-set op->status;
929 * for example, if there is a miscellaneous XDR error
930 * it will be set to nfserr_bad_xdr.
935 /* We must be able to encode a successful response to
936 * this operation, with enough room left over to encode a
937 * failed response to the next operation. If we don't
938 * have enough room, fail with ERR_RESOURCE.
940 slack_bytes = (char *)resp->end - (char *)resp->p;
941 if (slack_bytes < COMPOUND_SLACK_SPACE
942 + COMPOUND_ERR_SLACK_SPACE) {
943 BUG_ON(slack_bytes < COMPOUND_ERR_SLACK_SPACE);
944 op->status = nfserr_resource;
948 opdesc = &nfsd4_ops[op->opnum];
950 if (!cstate->current_fh.fh_dentry) {
951 if (!(opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
952 op->status = nfserr_nofilehandle;
955 } else if (cstate->current_fh.fh_export->ex_fslocs.migrated &&
956 !(opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
957 op->status = nfserr_moved;
962 op->status = opdesc->op_func(rqstp, cstate, &op->u);
964 BUG_ON(op->status == nfs_ok);
967 /* Only from SEQUENCE or CREATE_SESSION */
968 if (resp->cstate.status == nfserr_replay_cache) {
969 dprintk("%s NFS4.1 replay from cache\n", __func__);
970 if (nfsd4_not_cached(resp))
971 status = nfsd4_enc_uncached_replay(args, resp);
976 if (op->status == nfserr_replay_me) {
977 op->replay = &cstate->replay_owner->so_replay;
978 nfsd4_encode_replay(resp, op);
979 status = op->status = op->replay->rp_status;
981 nfsd4_encode_operation(resp, op);
985 dprintk("nfsv4 compound op %p opcnt %d #%d: %d: status %d\n",
986 args->ops, args->opcnt, resp->opcnt, op->opnum,
987 be32_to_cpu(status));
989 if (cstate->replay_owner) {
990 nfs4_put_stateowner(cstate->replay_owner);
991 cstate->replay_owner = NULL;
993 /* XXX Ugh, we need to get rid of this kind of special case: */
994 if (op->opnum == OP_READ && op->u.read.rd_filp)
995 fput(op->u.read.rd_filp);
997 nfsd4_increment_op_stats(op->opnum);
999 if (!rqstp->rq_usedeferral && status == nfserr_dropit) {
1000 dprintk("%s Dropit - send NFS4ERR_DELAY\n", __func__);
1001 status = nfserr_jukebox;
1004 resp->cstate.status = status;
1005 fh_put(&resp->cstate.current_fh);
1006 fh_put(&resp->cstate.save_fh);
1007 BUG_ON(resp->cstate.replay_owner);
1009 nfsd4_release_compoundargs(args);
1010 /* Reset deferral mechanism for RPC deferrals */
1011 rqstp->rq_usedeferral = 1;
1012 dprintk("nfsv4 compound returned %d\n", ntohl(status));
1016 static struct nfsd4_operation nfsd4_ops[] = {
1018 .op_func = (nfsd4op_func)nfsd4_access,
1019 .op_name = "OP_ACCESS",
1022 .op_func = (nfsd4op_func)nfsd4_close,
1023 .op_name = "OP_CLOSE",
1026 .op_func = (nfsd4op_func)nfsd4_commit,
1027 .op_name = "OP_COMMIT",
1030 .op_func = (nfsd4op_func)nfsd4_create,
1031 .op_name = "OP_CREATE",
1033 [OP_DELEGRETURN] = {
1034 .op_func = (nfsd4op_func)nfsd4_delegreturn,
1035 .op_name = "OP_DELEGRETURN",
1038 .op_func = (nfsd4op_func)nfsd4_getattr,
1039 .op_flags = ALLOWED_ON_ABSENT_FS,
1040 .op_name = "OP_GETATTR",
1043 .op_func = (nfsd4op_func)nfsd4_getfh,
1044 .op_name = "OP_GETFH",
1047 .op_func = (nfsd4op_func)nfsd4_link,
1048 .op_name = "OP_LINK",
1051 .op_func = (nfsd4op_func)nfsd4_lock,
1052 .op_name = "OP_LOCK",
1055 .op_func = (nfsd4op_func)nfsd4_lockt,
1056 .op_name = "OP_LOCKT",
1059 .op_func = (nfsd4op_func)nfsd4_locku,
1060 .op_name = "OP_LOCKU",
1063 .op_func = (nfsd4op_func)nfsd4_lookup,
1064 .op_name = "OP_LOOKUP",
1067 .op_func = (nfsd4op_func)nfsd4_lookupp,
1068 .op_name = "OP_LOOKUPP",
1071 .op_func = (nfsd4op_func)nfsd4_nverify,
1072 .op_name = "OP_NVERIFY",
1075 .op_func = (nfsd4op_func)nfsd4_open,
1076 .op_name = "OP_OPEN",
1078 [OP_OPEN_CONFIRM] = {
1079 .op_func = (nfsd4op_func)nfsd4_open_confirm,
1080 .op_name = "OP_OPEN_CONFIRM",
1082 [OP_OPEN_DOWNGRADE] = {
1083 .op_func = (nfsd4op_func)nfsd4_open_downgrade,
1084 .op_name = "OP_OPEN_DOWNGRADE",
1087 .op_func = (nfsd4op_func)nfsd4_putfh,
1088 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1089 .op_name = "OP_PUTFH",
1092 .op_func = (nfsd4op_func)nfsd4_putrootfh,
1093 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1094 .op_name = "OP_PUTPUBFH",
1097 .op_func = (nfsd4op_func)nfsd4_putrootfh,
1098 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1099 .op_name = "OP_PUTROOTFH",
1102 .op_func = (nfsd4op_func)nfsd4_read,
1103 .op_name = "OP_READ",
1106 .op_func = (nfsd4op_func)nfsd4_readdir,
1107 .op_name = "OP_READDIR",
1110 .op_func = (nfsd4op_func)nfsd4_readlink,
1111 .op_name = "OP_READLINK",
1114 .op_func = (nfsd4op_func)nfsd4_remove,
1115 .op_name = "OP_REMOVE",
1118 .op_name = "OP_RENAME",
1119 .op_func = (nfsd4op_func)nfsd4_rename,
1122 .op_func = (nfsd4op_func)nfsd4_renew,
1123 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1124 .op_name = "OP_RENEW",
1127 .op_func = (nfsd4op_func)nfsd4_restorefh,
1128 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1129 .op_name = "OP_RESTOREFH",
1132 .op_func = (nfsd4op_func)nfsd4_savefh,
1133 .op_name = "OP_SAVEFH",
1136 .op_func = (nfsd4op_func)nfsd4_secinfo,
1137 .op_name = "OP_SECINFO",
1140 .op_func = (nfsd4op_func)nfsd4_setattr,
1141 .op_name = "OP_SETATTR",
1143 [OP_SETCLIENTID] = {
1144 .op_func = (nfsd4op_func)nfsd4_setclientid,
1145 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1146 .op_name = "OP_SETCLIENTID",
1148 [OP_SETCLIENTID_CONFIRM] = {
1149 .op_func = (nfsd4op_func)nfsd4_setclientid_confirm,
1150 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1151 .op_name = "OP_SETCLIENTID_CONFIRM",
1154 .op_func = (nfsd4op_func)nfsd4_verify,
1155 .op_name = "OP_VERIFY",
1158 .op_func = (nfsd4op_func)nfsd4_write,
1159 .op_name = "OP_WRITE",
1161 [OP_RELEASE_LOCKOWNER] = {
1162 .op_func = (nfsd4op_func)nfsd4_release_lockowner,
1163 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS,
1164 .op_name = "OP_RELEASE_LOCKOWNER",
1167 /* NFSv4.1 operations */
1168 [OP_EXCHANGE_ID] = {
1169 .op_func = (nfsd4op_func)nfsd4_exchange_id,
1170 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1171 .op_name = "OP_EXCHANGE_ID",
1173 [OP_CREATE_SESSION] = {
1174 .op_func = (nfsd4op_func)nfsd4_create_session,
1175 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1176 .op_name = "OP_CREATE_SESSION",
1178 [OP_DESTROY_SESSION] = {
1179 .op_func = (nfsd4op_func)nfsd4_destroy_session,
1180 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1181 .op_name = "OP_DESTROY_SESSION",
1184 .op_func = (nfsd4op_func)nfsd4_sequence,
1185 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1186 .op_name = "OP_SEQUENCE",
1190 static const char *nfsd4_op_name(unsigned opnum)
1192 if (opnum < ARRAY_SIZE(nfsd4_ops))
1193 return nfsd4_ops[opnum].op_name;
1194 return "unknown_operation";
1197 #define nfs4svc_decode_voidargs NULL
1198 #define nfs4svc_release_void NULL
1199 #define nfsd4_voidres nfsd4_voidargs
1200 #define nfs4svc_release_compound NULL
1201 struct nfsd4_voidargs { int dummy; };
1203 #define PROC(name, argt, rest, relt, cache, respsize) \
1204 { (svc_procfunc) nfsd4_proc_##name, \
1205 (kxdrproc_t) nfs4svc_decode_##argt##args, \
1206 (kxdrproc_t) nfs4svc_encode_##rest##res, \
1207 (kxdrproc_t) nfs4svc_release_##relt, \
1208 sizeof(struct nfsd4_##argt##args), \
1209 sizeof(struct nfsd4_##rest##res), \
1216 * TODO: At the present time, the NFSv4 server does not do XID caching
1217 * of requests. Implementing XID caching would not be a serious problem,
1218 * although it would require a mild change in interfaces since one
1219 * doesn't know whether an NFSv4 request is idempotent until after the
1220 * XDR decode. However, XID caching totally confuses pynfs (Peter
1221 * Astrand's regression testsuite for NFSv4 servers), which reuses
1222 * XID's liberally, so I've left it unimplemented until pynfs generates
1225 static struct svc_procedure nfsd_procedures4[2] = {
1226 PROC(null, void, void, void, RC_NOCACHE, 1),
1227 PROC(compound, compound, compound, compound, RC_NOCACHE, NFSD_BUFSIZE/4)
1230 struct svc_version nfsd_version4 = {
1233 .vs_proc = nfsd_procedures4,
1234 .vs_dispatch = nfsd_dispatch,
1235 .vs_xdrsize = NFS4_SVC_XDRSIZE,