freebsd-dev/sys/netsmb/smb_rq.c
2002-09-18 19:44:14 +00:00

755 lines
18 KiB
C

/*
* Copyright (c) 2000-2001, Boris Popov
* All rights reserved.
*
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Boris Popov.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/lock.h>
#include <sys/sysctl.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/mbuf.h>
#include <netsmb/smb.h>
#include <netsmb/smb_conn.h>
#include <netsmb/smb_rq.h>
#include <netsmb/smb_subr.h>
#include <netsmb/smb_tran.h>
MALLOC_DEFINE(M_SMBRQ, "SMBRQ", "SMB request");
MODULE_DEPEND(netsmb, libmchain, 1, 1, 1);
static int smb_rq_reply(struct smb_rq *rqp);
static int smb_rq_enqueue(struct smb_rq *rqp);
static int smb_rq_getenv(struct smb_connobj *layer,
struct smb_vc **vcpp, struct smb_share **sspp);
static int smb_rq_new(struct smb_rq *rqp, u_char cmd);
static int smb_t2_reply(struct smb_t2rq *t2p);
int
smb_rq_alloc(struct smb_connobj *layer, u_char cmd, struct smb_cred *scred,
struct smb_rq **rqpp)
{
struct smb_rq *rqp;
int error;
MALLOC(rqp, struct smb_rq *, sizeof(*rqp), M_SMBRQ, M_WAITOK);
if (rqp == NULL)
return ENOMEM;
error = smb_rq_init(rqp, layer, cmd, scred);
rqp->sr_flags |= SMBR_ALLOCED;
if (error) {
smb_rq_done(rqp);
return error;
}
*rqpp = rqp;
return 0;
}
static char tzero[12];
int
smb_rq_init(struct smb_rq *rqp, struct smb_connobj *layer, u_char cmd,
struct smb_cred *scred)
{
int error;
bzero(rqp, sizeof(*rqp));
smb_sl_init(&rqp->sr_slock, "srslock");
error = smb_rq_getenv(layer, &rqp->sr_vc, &rqp->sr_share);
if (error)
return error;
error = smb_vc_access(rqp->sr_vc, scred, SMBM_EXEC);
if (error)
return error;
if (rqp->sr_share) {
error = smb_share_access(rqp->sr_share, scred, SMBM_EXEC);
if (error)
return error;
}
rqp->sr_cred = scred;
rqp->sr_mid = smb_vc_nextmid(rqp->sr_vc);
return smb_rq_new(rqp, cmd);
}
static int
smb_rq_new(struct smb_rq *rqp, u_char cmd)
{
struct smb_vc *vcp = rqp->sr_vc;
struct mbchain *mbp = &rqp->sr_rq;
int error;
rqp->sr_sendcnt = 0;
mb_done(mbp);
md_done(&rqp->sr_rp);
error = mb_init(mbp);
if (error)
return error;
mb_put_mem(mbp, SMB_SIGNATURE, SMB_SIGLEN, MB_MSYSTEM);
mb_put_uint8(mbp, cmd);
mb_put_uint32le(mbp, 0); /* DosError */
mb_put_uint8(mbp, vcp->vc_hflags);
if (cmd == SMB_COM_TRANSACTION || cmd == SMB_COM_TRANSACTION_SECONDARY)
mb_put_uint16le(mbp, (vcp->vc_hflags2 & ~SMB_FLAGS2_UNICODE));
else
mb_put_uint16le(mbp, vcp->vc_hflags2);
mb_put_mem(mbp, tzero, 12, MB_MSYSTEM);
rqp->sr_rqtid = (u_int16_t*)mb_reserve(mbp, sizeof(u_int16_t));
mb_put_uint16le(mbp, 1 /*scred->sc_p->p_pid & 0xffff*/);
rqp->sr_rquid = (u_int16_t*)mb_reserve(mbp, sizeof(u_int16_t));
mb_put_uint16le(mbp, rqp->sr_mid);
return 0;
}
void
smb_rq_done(struct smb_rq *rqp)
{
mb_done(&rqp->sr_rq);
md_done(&rqp->sr_rp);
smb_sl_destroy(&rqp->sr_slock);
if (rqp->sr_flags & SMBR_ALLOCED)
free(rqp, M_SMBRQ);
}
/*
* Simple request-reply exchange
*/
int
smb_rq_simple(struct smb_rq *rqp)
{
struct smb_vc *vcp = rqp->sr_vc;
int error = EINVAL, i;
for (i = 0; i < SMB_MAXRCN; i++) {
rqp->sr_flags &= ~SMBR_RESTART;
rqp->sr_timo = vcp->vc_timo;
rqp->sr_state = SMBRQ_NOTSENT;
error = smb_rq_enqueue(rqp);
if (error)
return error;
error = smb_rq_reply(rqp);
if (error == 0)
break;
if ((rqp->sr_flags & (SMBR_RESTART | SMBR_NORESTART)) != SMBR_RESTART)
break;
}
return error;
}
static int
smb_rq_enqueue(struct smb_rq *rqp)
{
struct smb_share *ssp = rqp->sr_share;
int error;
if (ssp == NULL || rqp->sr_cred == &rqp->sr_vc->vc_iod->iod_scred) {
return smb_iod_addrq(rqp);
}
for (;;) {
SMBS_ST_LOCK(ssp);
if (ssp->ss_flags & SMBS_RECONNECTING) {
msleep(&ssp->ss_vcgenid, SMBS_ST_LOCKPTR(ssp),
PWAIT | PDROP, "90trcn", hz);
if (smb_proc_intr(rqp->sr_cred->scr_td->td_proc))
return EINTR;
continue;
}
if (smb_share_valid(ssp) || (ssp->ss_flags & SMBS_CONNECTED) == 0) {
SMBS_ST_UNLOCK(ssp);
} else {
SMBS_ST_UNLOCK(ssp);
error = smb_iod_request(rqp->sr_vc->vc_iod,
SMBIOD_EV_TREECONNECT | SMBIOD_EV_SYNC, ssp);
if (error)
return error;
}
error = smb_iod_addrq(rqp);
if (error != EXDEV)
break;
}
return error;
}
void
smb_rq_wstart(struct smb_rq *rqp)
{
rqp->sr_wcount = mb_reserve(&rqp->sr_rq, sizeof(u_int8_t));
rqp->sr_rq.mb_count = 0;
}
void
smb_rq_wend(struct smb_rq *rqp)
{
if (rqp->sr_wcount == NULL) {
SMBERROR("no wcount\n"); /* actually panic */
return;
}
if (rqp->sr_rq.mb_count & 1)
SMBERROR("odd word count\n");
*rqp->sr_wcount = rqp->sr_rq.mb_count / 2;
}
void
smb_rq_bstart(struct smb_rq *rqp)
{
rqp->sr_bcount = (u_short*)mb_reserve(&rqp->sr_rq, sizeof(u_short));
rqp->sr_rq.mb_count = 0;
}
void
smb_rq_bend(struct smb_rq *rqp)
{
int bcnt;
if (rqp->sr_bcount == NULL) {
SMBERROR("no bcount\n"); /* actually panic */
return;
}
bcnt = rqp->sr_rq.mb_count;
if (bcnt > 0xffff)
SMBERROR("byte count too large (%d)\n", bcnt);
*rqp->sr_bcount = htoles(bcnt);
}
int
smb_rq_intr(struct smb_rq *rqp)
{
struct proc *p = rqp->sr_cred->scr_td->td_proc;
if (rqp->sr_flags & SMBR_INTR)
return EINTR;
return smb_proc_intr(p);
}
int
smb_rq_getrequest(struct smb_rq *rqp, struct mbchain **mbpp)
{
*mbpp = &rqp->sr_rq;
return 0;
}
int
smb_rq_getreply(struct smb_rq *rqp, struct mdchain **mbpp)
{
*mbpp = &rqp->sr_rp;
return 0;
}
static int
smb_rq_getenv(struct smb_connobj *layer,
struct smb_vc **vcpp, struct smb_share **sspp)
{
struct smb_vc *vcp = NULL;
struct smb_share *ssp = NULL;
struct smb_connobj *cp;
int error = 0;
switch (layer->co_level) {
case SMBL_VC:
vcp = CPTOVC(layer);
if (layer->co_parent == NULL) {
SMBERROR("zombie VC %s\n", vcp->vc_srvname);
error = EINVAL;
break;
}
break;
case SMBL_SHARE:
ssp = CPTOSS(layer);
cp = layer->co_parent;
if (cp == NULL) {
SMBERROR("zombie share %s\n", ssp->ss_name);
error = EINVAL;
break;
}
error = smb_rq_getenv(cp, &vcp, NULL);
if (error)
break;
break;
default:
SMBERROR("invalid layer %d passed\n", layer->co_level);
error = EINVAL;
}
if (vcpp)
*vcpp = vcp;
if (sspp)
*sspp = ssp;
return error;
}
/*
* Wait for reply on the request
*/
static int
smb_rq_reply(struct smb_rq *rqp)
{
struct mdchain *mdp = &rqp->sr_rp;
u_int32_t tdw;
u_int8_t tb;
int error, rperror = 0;
error = smb_iod_waitrq(rqp);
if (error)
return error;
error = md_get_uint32(mdp, &tdw);
if (error)
return error;
error = md_get_uint8(mdp, &tb);
if (rqp->sr_vc->vc_hflags2 & SMB_FLAGS2_ERR_STATUS) {
error = md_get_uint32le(mdp, &rqp->sr_error);
} else {
error = md_get_uint8(mdp, &rqp->sr_errclass);
error = md_get_uint8(mdp, &tb);
error = md_get_uint16le(mdp, &rqp->sr_serror);
if (!error)
rperror = smb_maperror(rqp->sr_errclass, rqp->sr_serror);
}
error = md_get_uint8(mdp, &rqp->sr_rpflags);
error = md_get_uint16le(mdp, &rqp->sr_rpflags2);
error = md_get_uint32(mdp, &tdw);
error = md_get_uint32(mdp, &tdw);
error = md_get_uint32(mdp, &tdw);
error = md_get_uint16le(mdp, &rqp->sr_rptid);
error = md_get_uint16le(mdp, &rqp->sr_rppid);
error = md_get_uint16le(mdp, &rqp->sr_rpuid);
error = md_get_uint16le(mdp, &rqp->sr_rpmid);
SMBSDEBUG("M:%04x, P:%04x, U:%04x, T:%04x, E: %d:%d\n",
rqp->sr_rpmid, rqp->sr_rppid, rqp->sr_rpuid, rqp->sr_rptid,
rqp->sr_errclass, rqp->sr_serror);
return error ? error : rperror;
}
#define ALIGN4(a) (((a) + 3) & ~3)
/*
* TRANS2 request implementation
*/
int
smb_t2_alloc(struct smb_connobj *layer, u_short setup, struct smb_cred *scred,
struct smb_t2rq **t2pp)
{
struct smb_t2rq *t2p;
int error;
MALLOC(t2p, struct smb_t2rq *, sizeof(*t2p), M_SMBRQ, M_WAITOK);
if (t2p == NULL)
return ENOMEM;
error = smb_t2_init(t2p, layer, setup, scred);
t2p->t2_flags |= SMBT2_ALLOCED;
if (error) {
smb_t2_done(t2p);
return error;
}
*t2pp = t2p;
return 0;
}
int
smb_t2_init(struct smb_t2rq *t2p, struct smb_connobj *source, u_short setup,
struct smb_cred *scred)
{
int error;
bzero(t2p, sizeof(*t2p));
t2p->t2_source = source;
t2p->t2_setupcount = 1;
t2p->t2_setupdata = t2p->t2_setup;
t2p->t2_setup[0] = setup;
t2p->t2_fid = 0xffff;
t2p->t2_cred = scred;
error = smb_rq_getenv(source, &t2p->t2_vc, NULL);
if (error)
return error;
return 0;
}
void
smb_t2_done(struct smb_t2rq *t2p)
{
mb_done(&t2p->t2_tparam);
mb_done(&t2p->t2_tdata);
md_done(&t2p->t2_rparam);
md_done(&t2p->t2_rdata);
if (t2p->t2_flags & SMBT2_ALLOCED)
free(t2p, M_SMBRQ);
}
static int
smb_t2_placedata(struct mbuf *mtop, u_int16_t offset, u_int16_t count,
struct mdchain *mdp)
{
struct mbuf *m, *m0;
int len;
m0 = m_split(mtop, offset, M_TRYWAIT);
if (m0 == NULL)
return EBADRPC;
len = m_length(m0, &m);
m->m_len -= len - count;
if (mdp->md_top == NULL) {
md_initm(mdp, m0);
} else
m_cat(mdp->md_top, m0);
return 0;
}
static int
smb_t2_reply(struct smb_t2rq *t2p)
{
struct mdchain *mdp;
struct smb_rq *rqp = t2p->t2_rq;
int error, totpgot, totdgot;
u_int16_t totpcount, totdcount, pcount, poff, doff, pdisp, ddisp;
u_int16_t tmp, bc, dcount;
u_int8_t wc;
error = smb_rq_reply(rqp);
if (error)
return error;
if ((t2p->t2_flags & SMBT2_ALLSENT) == 0) {
/*
* this is an interim response, ignore it.
*/
SMBRQ_SLOCK(rqp);
md_next_record(&rqp->sr_rp);
SMBRQ_SUNLOCK(rqp);
return 0;
}
/*
* Now we have to get all subsequent responses. The CIFS specification
* says that they can be disordered which is weird.
* TODO: timo
*/
totpgot = totdgot = 0;
totpcount = totdcount = 0xffff;
mdp = &rqp->sr_rp;
for (;;) {
m_dumpm(mdp->md_top);
if ((error = md_get_uint8(mdp, &wc)) != 0)
break;
if (wc < 10) {
error = ENOENT;
break;
}
if ((error = md_get_uint16le(mdp, &tmp)) != 0)
break;
if (totpcount > tmp)
totpcount = tmp;
md_get_uint16le(mdp, &tmp);
if (totdcount > tmp)
totdcount = tmp;
if ((error = md_get_uint16le(mdp, &tmp)) != 0 || /* reserved */
(error = md_get_uint16le(mdp, &pcount)) != 0 ||
(error = md_get_uint16le(mdp, &poff)) != 0 ||
(error = md_get_uint16le(mdp, &pdisp)) != 0)
break;
if (pcount != 0 && pdisp != totpgot) {
SMBERROR("Can't handle disordered parameters %d:%d\n",
pdisp, totpgot);
error = EINVAL;
break;
}
if ((error = md_get_uint16le(mdp, &dcount)) != 0 ||
(error = md_get_uint16le(mdp, &doff)) != 0 ||
(error = md_get_uint16le(mdp, &ddisp)) != 0)
break;
if (dcount != 0 && ddisp != totdgot) {
SMBERROR("Can't handle disordered data\n");
error = EINVAL;
break;
}
md_get_uint8(mdp, &wc);
md_get_uint8(mdp, NULL);
tmp = wc;
while (tmp--)
md_get_uint16(mdp, NULL);
if ((error = md_get_uint16le(mdp, &bc)) != 0)
break;
/* tmp = SMB_HDRLEN + 1 + 10 * 2 + 2 * wc + 2;*/
if (dcount) {
error = smb_t2_placedata(mdp->md_top, doff, dcount,
&t2p->t2_rdata);
if (error)
break;
}
if (pcount) {
error = smb_t2_placedata(mdp->md_top, poff, pcount,
&t2p->t2_rparam);
if (error)
break;
}
totpgot += pcount;
totdgot += dcount;
if (totpgot >= totpcount && totdgot >= totdcount) {
error = 0;
t2p->t2_flags |= SMBT2_ALLRECV;
break;
}
/*
* We're done with this reply, look for the next one.
*/
SMBRQ_SLOCK(rqp);
md_next_record(&rqp->sr_rp);
SMBRQ_SUNLOCK(rqp);
error = smb_rq_reply(rqp);
if (error)
break;
}
return error;
}
/*
* Perform a full round of TRANS2 request
*/
static int
smb_t2_request_int(struct smb_t2rq *t2p)
{
struct smb_vc *vcp = t2p->t2_vc;
struct smb_cred *scred = t2p->t2_cred;
struct mbchain *mbp;
struct mdchain *mdp, mbparam, mbdata;
struct mbuf *m;
struct smb_rq *rqp;
int totpcount, leftpcount, totdcount, leftdcount, len, txmax, i;
int error, doff, poff, txdcount, txpcount, nmlen;
m = t2p->t2_tparam.mb_top;
if (m) {
md_initm(&mbparam, m); /* do not free it! */
totpcount = m_fixhdr(m);
if (totpcount > 0xffff) /* maxvalue for u_short */
return EINVAL;
} else
totpcount = 0;
m = t2p->t2_tdata.mb_top;
if (m) {
md_initm(&mbdata, m); /* do not free it! */
totdcount = m_fixhdr(m);
if (totdcount > 0xffff)
return EINVAL;
} else
totdcount = 0;
leftdcount = totdcount;
leftpcount = totpcount;
txmax = vcp->vc_txmax;
error = smb_rq_alloc(t2p->t2_source, t2p->t_name ?
SMB_COM_TRANSACTION : SMB_COM_TRANSACTION2, scred, &rqp);
if (error)
return error;
rqp->sr_flags |= SMBR_MULTIPACKET;
t2p->t2_rq = rqp;
mbp = &rqp->sr_rq;
smb_rq_wstart(rqp);
mb_put_uint16le(mbp, totpcount);
mb_put_uint16le(mbp, totdcount);
mb_put_uint16le(mbp, t2p->t2_maxpcount);
mb_put_uint16le(mbp, t2p->t2_maxdcount);
mb_put_uint8(mbp, t2p->t2_maxscount);
mb_put_uint8(mbp, 0); /* reserved */
mb_put_uint16le(mbp, 0); /* flags */
mb_put_uint32le(mbp, 0); /* Timeout */
mb_put_uint16le(mbp, 0); /* reserved 2 */
len = mb_fixhdr(mbp);
/*
* now we have known packet size as
* ALIGN4(len + 5 * 2 + setupcount * 2 + 2 + strlen(name) + 1),
* and need to decide which parts should go into the first request
*/
nmlen = t2p->t_name ? strlen(t2p->t_name) : 0;
len = ALIGN4(len + 5 * 2 + t2p->t2_setupcount * 2 + 2 + nmlen + 1);
if (len + leftpcount > txmax) {
txpcount = min(leftpcount, txmax - len);
poff = len;
txdcount = 0;
doff = 0;
} else {
txpcount = leftpcount;
poff = txpcount ? len : 0;
len = ALIGN4(len + txpcount);
txdcount = min(leftdcount, txmax - len);
doff = txdcount ? len : 0;
}
leftpcount -= txpcount;
leftdcount -= txdcount;
mb_put_uint16le(mbp, txpcount);
mb_put_uint16le(mbp, poff);
mb_put_uint16le(mbp, txdcount);
mb_put_uint16le(mbp, doff);
mb_put_uint8(mbp, t2p->t2_setupcount);
mb_put_uint8(mbp, 0);
for (i = 0; i < t2p->t2_setupcount; i++)
mb_put_uint16le(mbp, t2p->t2_setupdata[i]);
smb_rq_wend(rqp);
smb_rq_bstart(rqp);
/* TDUNICODE */
if (t2p->t_name)
mb_put_mem(mbp, t2p->t_name, nmlen, MB_MSYSTEM);
mb_put_uint8(mbp, 0); /* terminating zero */
len = mb_fixhdr(mbp);
if (txpcount) {
mb_put_mem(mbp, NULL, ALIGN4(len) - len, MB_MZERO);
error = md_get_mbuf(&mbparam, txpcount, &m);
SMBSDEBUG("%d:%d:%d\n", error, txpcount, txmax);
if (error)
goto freerq;
mb_put_mbuf(mbp, m);
}
len = mb_fixhdr(mbp);
if (txdcount) {
mb_put_mem(mbp, NULL, ALIGN4(len) - len, MB_MZERO);
error = md_get_mbuf(&mbdata, txdcount, &m);
if (error)
goto freerq;
mb_put_mbuf(mbp, m);
}
smb_rq_bend(rqp); /* incredible, but thats it... */
error = smb_rq_enqueue(rqp);
if (error)
goto freerq;
if (leftpcount == 0 && leftdcount == 0)
t2p->t2_flags |= SMBT2_ALLSENT;
error = smb_t2_reply(t2p);
if (error)
goto bad;
while (leftpcount || leftdcount) {
error = smb_rq_new(rqp, t2p->t_name ?
SMB_COM_TRANSACTION_SECONDARY : SMB_COM_TRANSACTION2_SECONDARY);
if (error)
goto bad;
mbp = &rqp->sr_rq;
smb_rq_wstart(rqp);
mb_put_uint16le(mbp, totpcount);
mb_put_uint16le(mbp, totdcount);
len = mb_fixhdr(mbp);
/*
* now we have known packet size as
* ALIGN4(len + 7 * 2 + 2) for T2 request, and -2 for T one,
* and need to decide which parts should go into request
*/
len = ALIGN4(len + 6 * 2 + 2);
if (t2p->t_name == NULL)
len += 2;
if (len + leftpcount > txmax) {
txpcount = min(leftpcount, txmax - len);
poff = len;
txdcount = 0;
doff = 0;
} else {
txpcount = leftpcount;
poff = txpcount ? len : 0;
len = ALIGN4(len + txpcount);
txdcount = min(leftdcount, txmax - len);
doff = txdcount ? len : 0;
}
mb_put_uint16le(mbp, txpcount);
mb_put_uint16le(mbp, poff);
mb_put_uint16le(mbp, totpcount - leftpcount);
mb_put_uint16le(mbp, txdcount);
mb_put_uint16le(mbp, doff);
mb_put_uint16le(mbp, totdcount - leftdcount);
leftpcount -= txpcount;
leftdcount -= txdcount;
if (t2p->t_name == NULL)
mb_put_uint16le(mbp, t2p->t2_fid);
smb_rq_wend(rqp);
smb_rq_bstart(rqp);
mb_put_uint8(mbp, 0); /* name */
len = mb_fixhdr(mbp);
if (txpcount) {
mb_put_mem(mbp, NULL, ALIGN4(len) - len, MB_MZERO);
error = md_get_mbuf(&mbparam, txpcount, &m);
if (error)
goto bad;
mb_put_mbuf(mbp, m);
}
len = mb_fixhdr(mbp);
if (txdcount) {
mb_put_mem(mbp, NULL, ALIGN4(len) - len, MB_MZERO);
error = md_get_mbuf(&mbdata, txdcount, &m);
if (error)
goto bad;
mb_put_mbuf(mbp, m);
}
smb_rq_bend(rqp);
rqp->sr_state = SMBRQ_NOTSENT;
error = smb_iod_request(vcp->vc_iod, SMBIOD_EV_NEWRQ, NULL);
if (error)
goto bad;
} /* while left params or data */
t2p->t2_flags |= SMBT2_ALLSENT;
mdp = &t2p->t2_rdata;
if (mdp->md_top) {
m_fixhdr(mdp->md_top);
md_initm(mdp, mdp->md_top);
}
mdp = &t2p->t2_rparam;
if (mdp->md_top) {
m_fixhdr(mdp->md_top);
md_initm(mdp, mdp->md_top);
}
bad:
smb_iod_removerq(rqp);
freerq:
smb_rq_done(rqp);
if (error) {
if (rqp->sr_flags & SMBR_RESTART)
t2p->t2_flags |= SMBT2_RESTART;
md_done(&t2p->t2_rparam);
md_done(&t2p->t2_rdata);
}
return error;
}
int
smb_t2_request(struct smb_t2rq *t2p)
{
int error = EINVAL, i;
for (i = 0; i < SMB_MAXRCN; i++) {
t2p->t2_flags &= ~SMBR_RESTART;
error = smb_t2_request_int(t2p);
if (error == 0)
break;
if ((t2p->t2_flags & (SMBT2_RESTART | SMBT2_NORESTART)) != SMBT2_RESTART)
break;
}
return error;
}