/* $NetBSD: nfs_nfsdsocket.c,v 1.2 2016/12/13 21:50:32 pgoyette Exp $ */
/*-
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
/*
* Do an RPC. Basically, get the file handles translated to vnode pointers
* and then call the appropriate server routine. The server routines are
* split into groups, based on whether they use a file handle or file
* handle plus name or ...
* The NFS V4 Compound RPC is performed separately by nfsrvd_compound().
*/
APPLESTATIC void
nfsrvd_dorpc(struct nfsrv_descript *nd, int isdgram, u_char *tag, int taglen,
u_int32_t minorvers, NFSPROC_T *p)
{
int error = 0, lktype;
vnode_t vp;
mount_t mp = NULL;
struct nfsrvfh fh;
struct nfsexstuff nes;
/*
* Get a locked vnode for the first file handle
*/
if (!(nd->nd_flag & ND_NFSV4)) {
KASSERT(nd->nd_repstat == 0, ("nfsrvd_dorpc"));
/*
* For NFSv3, if the malloc/mget allocation is near limits,
* return NFSERR_DELAY.
*/
if ((nd->nd_flag & ND_NFSV3) && nfsrv_mallocmget_limit()) {
nd->nd_repstat = NFSERR_DELAY;
vp = NULL;
} else {
error = nfsrv_mtofh(nd, &fh);
if (error) {
if (error != EBADRPC)
printf("nfs dorpc err1=%d\n", error);
nd->nd_repstat = NFSERR_GARBAGE;
goto out;
}
if (nd->nd_procnum == NFSPROC_READ ||
nd->nd_procnum == NFSPROC_WRITE ||
nd->nd_procnum == NFSPROC_READDIR ||
nd->nd_procnum == NFSPROC_READDIRPLUS ||
nd->nd_procnum == NFSPROC_READLINK ||
nd->nd_procnum == NFSPROC_GETATTR ||
nd->nd_procnum == NFSPROC_ACCESS ||
nd->nd_procnum == NFSPROC_FSSTAT ||
nd->nd_procnum == NFSPROC_FSINFO)
lktype = LK_SHARED;
else
lktype = LK_EXCLUSIVE;
if (nd->nd_flag & ND_PUBLOOKUP)
nfsd_fhtovp(nd, &nfs_pubfh, lktype, &vp, &nes,
&mp, nfs_writerpc[nd->nd_procnum], p);
else
nfsd_fhtovp(nd, &fh, lktype, &vp, &nes,
&mp, nfs_writerpc[nd->nd_procnum], p);
if (nd->nd_repstat == NFSERR_PROGNOTV4)
goto out;
}
}
/*
* For V2 and 3, set the ND_SAVEREPLY flag for the recent request
* cache, as required.
* For V4, nfsrvd_compound() does this.
*/
if (!(nd->nd_flag & ND_NFSV4) && nfsrv_nonidempotent[nd->nd_procnum])
nd->nd_flag |= ND_SAVEREPLY;
nfsrvd_rephead(nd);
/*
* If nd_repstat is non-zero, just fill in the reply status
* to complete the RPC reply for V2. Otherwise, you must do
* the RPC.
*/
if (nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
*nd->nd_errp = nfsd_errmap(nd);
nfsrvd_statstart(nfsv3to4op[nd->nd_procnum], /*now*/ NULL);
nfsrvd_statend(nfsv3to4op[nd->nd_procnum], /*bytes*/ 0,
/*now*/ NULL, /*then*/ NULL);
if (mp != NULL && nfs_writerpc[nd->nd_procnum] != 0)
vn_finished_write(mp);
goto out;
}
/*
* Now the procedure can be performed. For V4, nfsrvd_compound()
* works through the sub-rpcs, otherwise just call the procedure.
* The procedures are in three groups with different arguments.
* The group is indicated by the value in nfs_retfh[].
*/
if (nd->nd_flag & ND_NFSV4) {
nfsrvd_compound(nd, isdgram, tag, taglen, minorvers, p);
} else {
struct bintime start_time;
/*
* Breaks down a compound RPC request and calls the server routines for
* the subprocedures.
* Some suboperations are performed directly here to simplify file handle<-->
* vnode pointer handling.
*/
static void
nfsrvd_compound(struct nfsrv_descript *nd, int isdgram, u_char *tag,
int taglen, u_int32_t minorvers, NFSPROC_T *p)
{
int i, lktype, op, op0 = 0, statsinprog = 0;
u_int32_t *tl;
struct nfsclient *clp, *nclp;
int numops, error = 0, igotlock;
u_int32_t retops = 0, *retopsp = NULL, *repp;
vnode_t vp, nvp, savevp;
struct nfsrvfh fh;
mount_t new_mp, temp_mp = NULL;
struct ucred *credanon;
struct nfsexstuff nes, vpnes, savevpnes;
fsid_t cur_fsid, save_fsid;
static u_int64_t compref = 0;
struct bintime start_time;
NFSVNO_EXINIT(&vpnes);
NFSVNO_EXINIT(&savevpnes);
/*
* Put the seq# of the current compound RPC in nfsrv_descript.
* (This is used by nfsrv_checkgetattr(), to see if the write
* delegation was created by the same compound RPC as the one
* with that Getattr in it.)
* Don't worry about the 64bit number wrapping around. It ain't
* gonna happen before this server gets shut down/rebooted.
*/
nd->nd_compref = compref++;
/*
* Check for and optionally get a lock on the root. This lock means that
* no nfsd will be fiddling with the V4 file system and state stuff. It
* is required when the V4 root is being changed, the stable storage
* restart file is being updated, or callbacks are being done.
* When any of the nfsd are processing an NFSv4 compound RPC, they must
* either hold a reference count (nfs_usecnt) or the lock. When
* nfsrv_unlock() is called to release the lock, it can optionally
* also get a reference count, which saves the need for a call to
* nfsrv_getref() after nfsrv_unlock().
*/
/*
* First, check to see if we need to wait for an update lock.
*/
igotlock = 0;
NFSLOCKV4ROOTMUTEX();
if (nfsrv_stablefirst.nsf_flags & NFSNSF_NEEDLOCK)
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
NFSV4ROOTLOCKMUTEXPTR, NULL);
else
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 0, NULL,
NFSV4ROOTLOCKMUTEXPTR, NULL);
NFSUNLOCKV4ROOTMUTEX();
if (igotlock) {
/*
* If I got the lock, I can update the stable storage file.
* Done when the grace period is over or a client has long
* since expired.
*/
nfsrv_stablefirst.nsf_flags &= ~NFSNSF_NEEDLOCK;
if ((nfsrv_stablefirst.nsf_flags &
(NFSNSF_GRACEOVER | NFSNSF_UPDATEDONE)) == NFSNSF_GRACEOVER)
nfsrv_updatestable(p);
/*
* If at least one client has long since expired, search
* the client list for them, write a REVOKE record on the
* stable storage file and then remove them from the client
* list.
*/
if (nfsrv_stablefirst.nsf_flags & NFSNSF_EXPIREDCLIENT) {
nfsrv_stablefirst.nsf_flags &= ~NFSNSF_EXPIREDCLIENT;
for (i = 0; i < nfsrv_clienthashsize; i++) {
LIST_FOREACH_SAFE(clp, &nfsclienthash[i], lc_hash,
nclp) {
if (clp->lc_flags & LCL_EXPIREIT) {
if (!LIST_EMPTY(&clp->lc_open) ||
!LIST_EMPTY(&clp->lc_deleg))
nfsrv_writestable(clp->lc_id,
clp->lc_idlen, NFSNST_REVOKE, p);
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
LIST_REMOVE(clp, lc_hash);
nfsrv_zapclient(clp, p);
}
}
}
}
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
} else {
/*
* If we didn't get the lock, we need to get a refcnt,
* which also checks for and waits for the lock.
*/
NFSLOCKV4ROOTMUTEX();
nfsv4_getref(&nfsv4rootfs_lock, NULL,
NFSV4ROOTLOCKMUTEXPTR, NULL);
NFSUNLOCKV4ROOTMUTEX();
}
/*
* If flagged, search for open owners that haven't had any opens
* for a long time.
*/
if (nfsrv_stablefirst.nsf_flags & NFSNSF_NOOPENS) {
nfsrv_throwawayopens(p);
}
/* If taglen < 0, there was a parsing error in nfsd_getminorvers(). */
if (taglen < 0) {
error = EBADRPC;
goto nfsmout;
}
(void) nfsm_strtom(nd, tag, taglen);
NFSM_BUILD(retopsp, u_int32_t *, NFSX_UNSIGNED);
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (minorvers != NFSV4_MINORVERSION && minorvers != NFSV41_MINORVERSION)
nd->nd_repstat = NFSERR_MINORVERMISMATCH;
if (nd->nd_repstat)
numops = 0;
else
numops = fxdr_unsigned(int, *tl);
/*
* Loop around doing the sub ops.
* vp - is an unlocked vnode pointer for the CFH
* savevp - is an unlocked vnode pointer for the SAVEDFH
* (at some future date, it might turn out to be more appropriate
* to keep the file handles instead of vnode pointers?)
* savevpnes and vpnes - are the export flags for the above.
*/
for (i = 0; i < numops; i++) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
NFSM_BUILD(repp, u_int32_t *, 2 * NFSX_UNSIGNED);
*repp = *tl;
op = fxdr_unsigned(int, *tl);
NFSD_DEBUG(4, "op=%d\n", op);
/*
* Check for a referral on the current FH and, if so, return
* NFSERR_MOVED for all ops that allow it, except Getattr.
*/
if (vp != NULL && op != NFSV4OP_GETATTR &&
nfsv4root_getreferral(vp, NULL, 0) != NULL &&
nfsrv_errmoved(op)) {
nd->nd_repstat = NFSERR_MOVED;
*repp = nfsd_errmap(nd);
retops++;
break;
}
/*
* For NFSv4.1, check for a Sequence Operation being first
* or one of the other allowed operations by itself.
*/
if ((nd->nd_flag & ND_NFSV41) != 0) {
if (i != 0 && op == NFSV4OP_SEQUENCE)
nd->nd_repstat = NFSERR_SEQUENCEPOS;
else if (i == 0 && op != NFSV4OP_SEQUENCE &&
op != NFSV4OP_EXCHANGEID &&
op != NFSV4OP_CREATESESSION &&
op != NFSV4OP_BINDCONNTOSESS &&
op != NFSV4OP_DESTROYCLIENTID &&
op != NFSV4OP_DESTROYSESSION)
nd->nd_repstat = NFSERR_OPNOTINSESS;
else if (i != 0 && op0 != NFSV4OP_SEQUENCE)
nd->nd_repstat = NFSERR_NOTONLYOP;
if (nd->nd_repstat != 0) {
*repp = nfsd_errmap(nd);
retops++;
break;
}
}
nd->nd_procnum = op;
/*
* If over flood level, reply NFSERR_RESOURCE, if at the first
* Op. (Since a client recovery from NFSERR_RESOURCE can get
* really nasty for certain Op sequences, I'll play it safe
* and only return the error at the beginning.) The cache
* will still function over flood level, but uses lots of
* mbufs.)
* If nfsrv_mallocmget_limit() returns True, the system is near
* to its limit for memory that malloc()/mget() can allocate.
*/
if (i == 0 && (nd->nd_rp == NULL ||
nd->nd_rp->rc_refcnt == 0) &&
(nfsrv_mallocmget_limit() ||
nfsrc_tcpsavedreplies > nfsrc_floodlevel)) {
if (nfsrc_tcpsavedreplies > nfsrc_floodlevel)
printf("nfsd server cache flooded, try "
"increasing vfs.nfsd.tcphighwater\n");
nd->nd_repstat = NFSERR_RESOURCE;
*repp = nfsd_errmap(nd);
if (op == NFSV4OP_SETATTR) {
/*
* Setattr replies require a bitmap.
* even for errors like these.
*/
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = 0;
}
retops++;
break;
}
if (nfsv4_opflag[op].savereply)
nd->nd_flag |= ND_SAVEREPLY;
switch (op) {
case NFSV4OP_PUTFH:
error = nfsrv_mtofh(nd, &fh);
if (error)
goto nfsmout;
if (!nd->nd_repstat)
nfsd_fhtovp(nd, &fh, LK_SHARED, &nvp, &nes,
NULL, 0, p);
/* For now, allow this for non-export FHs */
if (!nd->nd_repstat) {
if (vp)
vrele(vp);
vp = nvp;
cur_fsid = vp->v_mount->mnt_stat.f_fsid;
NFSVOPUNLOCK(vp, 0);
vpnes = nes;
}
break;
case NFSV4OP_PUTPUBFH:
if (nfs_pubfhset)
nfsd_fhtovp(nd, &nfs_pubfh, LK_SHARED, &nvp,
&nes, NULL, 0, p);
else
nd->nd_repstat = NFSERR_NOFILEHANDLE;
if (!nd->nd_repstat) {
if (vp)
vrele(vp);
vp = nvp;
cur_fsid = vp->v_mount->mnt_stat.f_fsid;
NFSVOPUNLOCK(vp, 0);
vpnes = nes;
}
break;
case NFSV4OP_PUTROOTFH:
if (nfs_rootfhset) {
nfsd_fhtovp(nd, &nfs_rootfh, LK_SHARED, &nvp,
&nes, NULL, 0, p);
if (!nd->nd_repstat) {
if (vp)
vrele(vp);
vp = nvp;
cur_fsid = vp->v_mount->mnt_stat.f_fsid;
NFSVOPUNLOCK(vp, 0);
vpnes = nes;
}
} else
nd->nd_repstat = NFSERR_NOFILEHANDLE;
break;
case NFSV4OP_SAVEFH:
if (vp && NFSVNO_EXPORTED(&vpnes)) {
nd->nd_repstat = 0;
/* If vp == savevp, a no-op */
if (vp != savevp) {
if (savevp)
vrele(savevp);
VREF(vp);
savevp = vp;
savevpnes = vpnes;
save_fsid = cur_fsid;
}
} else {
nd->nd_repstat = NFSERR_NOFILEHANDLE;
}
break;
case NFSV4OP_RESTOREFH:
if (savevp) {
nd->nd_repstat = 0;
/* If vp == savevp, a no-op */
if (vp != savevp) {
VREF(savevp);
vrele(vp);
vp = savevp;
vpnes = savevpnes;
cur_fsid = save_fsid;
}
} else {
nd->nd_repstat = NFSERR_RESTOREFH;
}
break;
default:
/*
* Allow a Lookup, Getattr, GetFH, Secinfo on an
* non-exported directory if
* nfs_rootfhset. Do I need to allow any other Ops?
* (You can only have a non-exported vpnes if
* nfs_rootfhset is true. See nfsd_fhtovp())
* Allow AUTH_SYS to be used for file systems
* exported GSS only for certain Ops, to allow
* clients to do mounts more easily.
*/
if (nfsv4_opflag[op].needscfh && vp) {
if (!NFSVNO_EXPORTED(&vpnes) &&
op != NFSV4OP_LOOKUP &&
op != NFSV4OP_GETATTR &&
op != NFSV4OP_GETFH &&
op != NFSV4OP_ACCESS &&
op != NFSV4OP_READLINK &&
op != NFSV4OP_SECINFO)
nd->nd_repstat = NFSERR_NOFILEHANDLE;
else if (nfsvno_testexp(nd, &vpnes) &&
op != NFSV4OP_LOOKUP &&
op != NFSV4OP_GETFH &&
op != NFSV4OP_GETATTR &&
op != NFSV4OP_SECINFO)
nd->nd_repstat = NFSERR_WRONGSEC;
if (nd->nd_repstat) {
if (op == NFSV4OP_SETATTR) {
/*
* Setattr reply requires a bitmap
* even for errors like these.
*/
NFSM_BUILD(tl, u_int32_t *,
NFSX_UNSIGNED);
*tl = 0;
}
break;
}
}
if (nfsv4_opflag[op].retfh == 1) {
if (!vp) {
nd->nd_repstat = NFSERR_NOFILEHANDLE;
break;
}
VREF(vp);
if (nfsv4_opflag[op].modifyfs)
vn_start_write(vp, &temp_mp, V_WAIT);
error = (*(nfsrv4_ops1[op]))(nd, isdgram, vp,
&nvp, (fhandle_t *)fh.nfsrvfh_data, p, &vpnes);
if (!error && !nd->nd_repstat) {
if (op == NFSV4OP_LOOKUP || op == NFSV4OP_LOOKUPP) {
new_mp = nvp->v_mount;
if (cur_fsid.val[0] !=
new_mp->mnt_stat.f_fsid.val[0] ||
cur_fsid.val[1] !=
new_mp->mnt_stat.f_fsid.val[1]) {
/* crossed a server mount point */
nd->nd_repstat = nfsvno_checkexp(new_mp,
nd->nd_nam, &nes, &credanon);
if (!nd->nd_repstat)
nd->nd_repstat = nfsd_excred(nd,
&nes, credanon);
if (credanon != NULL)
crfree(credanon);
if (!nd->nd_repstat) {
vpnes = nes;
cur_fsid = new_mp->mnt_stat.f_fsid;
}
}
/* Lookup ops return a locked vnode */
NFSVOPUNLOCK(nvp, 0);
}
if (!nd->nd_repstat) {
vrele(vp);
vp = nvp;
} else
vrele(nvp);
}
if (nfsv4_opflag[op].modifyfs)
vn_finished_write(temp_mp);
} else if (nfsv4_opflag[op].retfh == 2) {
if (vp == NULL || savevp == NULL) {
nd->nd_repstat = NFSERR_NOFILEHANDLE;
break;
} else if (cur_fsid.val[0] != save_fsid.val[0] ||
cur_fsid.val[1] != save_fsid.val[1]) {
nd->nd_repstat = NFSERR_XDEV;
break;
}
if (nfsv4_opflag[op].modifyfs)
vn_start_write(savevp, &temp_mp, V_WAIT);
if (NFSVOPLOCK(savevp, LK_EXCLUSIVE) == 0) {
VREF(vp);
VREF(savevp);
error = (*(nfsrv4_ops2[op]))(nd, isdgram,
savevp, vp, p, &savevpnes, &vpnes);
} else
nd->nd_repstat = NFSERR_PERM;
if (nfsv4_opflag[op].modifyfs)
vn_finished_write(temp_mp);
} else {
if (nfsv4_opflag[op].retfh != 0)
panic("nfsrvd_compound");
if (nfsv4_opflag[op].needscfh) {
if (vp != NULL) {
lktype = nfsv4_opflag[op].lktype;
if (nfsv4_opflag[op].modifyfs) {
vn_start_write(vp, &temp_mp,
V_WAIT);
if (op == NFSV4OP_WRITE &&
MNT_SHARED_WRITES(temp_mp))
lktype = LK_SHARED;
}
if (NFSVOPLOCK(vp, lktype) == 0)
VREF(vp);
else
nd->nd_repstat = NFSERR_PERM;
} else {
nd->nd_repstat = NFSERR_NOFILEHANDLE;
if (op == NFSV4OP_SETATTR) {
/*
* Setattr reply requires a
* bitmap even for errors like
* these.
*/
NFSM_BUILD(tl, u_int32_t *,
NFSX_UNSIGNED);
*tl = 0;
}
break;
}
if (nd->nd_repstat == 0)
error = (*(nfsrv4_ops0[op]))(nd,
isdgram, vp, p, &vpnes);
if (nfsv4_opflag[op].modifyfs)
vn_finished_write(temp_mp);
} else {
error = (*(nfsrv4_ops0[op]))(nd, isdgram,
NULL, p, &vpnes);
}
}
}
if (error) {
if (error == EBADRPC || error == NFSERR_BADXDR) {
nd->nd_repstat = NFSERR_BADXDR;
} else {
nd->nd_repstat = error;
printf("nfsv4 comperr0=%d\n", error);
}
error = 0;
}
