freebsd-dev/sys/netsmb/smb_conn.c

873 lines
20 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.
*
* $FreeBSD$
*/
/*
* Connection engine.
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/lock.h>
#include <sys/sysctl.h>
#include <sys/socketvar.h>
#include <sys/iconv.h>
#include <netsmb/smb.h>
#include <netsmb/smb_subr.h>
#include <netsmb/smb_conn.h>
#include <netsmb/smb_tran.h>
#include <netsmb/smb_trantcp.h>
static struct smb_connobj smb_vclist;
static int smb_vcnext = 1; /* next unique id for VC */
SYSCTL_NODE(_net, OID_AUTO, smb, CTLFLAG_RW, NULL, "SMB protocol");
MALLOC_DEFINE(M_SMBCONN, "SMB conn", "SMB connection");
static void smb_co_init(struct smb_connobj *cp, int level, char *objname,
struct thread *td);
static void smb_co_done(struct smb_connobj *cp);
static int smb_co_lockstatus(struct smb_connobj *cp, struct thread *td);
static int smb_vc_disconnect(struct smb_vc *vcp);
static void smb_vc_free(struct smb_connobj *cp);
static void smb_vc_gone(struct smb_connobj *cp, struct smb_cred *scred);
static smb_co_free_t smb_share_free;
static smb_co_gone_t smb_share_gone;
static int smb_sysctl_treedump(SYSCTL_HANDLER_ARGS);
SYSCTL_PROC(_net_smb, OID_AUTO, treedump, CTLFLAG_RD | CTLTYPE_OPAQUE,
NULL, 0, smb_sysctl_treedump, "S,treedump", "Requester tree");
int
smb_sm_init(void)
{
smb_co_init(&smb_vclist, SMBL_SM, "smbsm", curthread);
smb_co_unlock(&smb_vclist, 0, curthread);
return 0;
}
int
smb_sm_done(void)
{
/* XXX: hold the mutex */
if (smb_vclist.co_usecount > 1) {
SMBERROR("%d connections still active\n", smb_vclist.co_usecount - 1);
return EBUSY;
}
smb_co_done(&smb_vclist);
return 0;
}
static int
smb_sm_lockvclist(int flags, struct thread *td)
{
return smb_co_lock(&smb_vclist, flags | LK_CANRECURSE, td);
}
static void
smb_sm_unlockvclist(struct thread *td)
{
smb_co_unlock(&smb_vclist, LK_RELEASE, td);
}
static int
smb_sm_lookupint(struct smb_vcspec *vcspec, struct smb_sharespec *shspec,
struct smb_cred *scred, struct smb_vc **vcpp)
{
struct thread *td = scred->scr_td;
struct smb_vc *vcp;
int exact = 1;
int error;
vcspec->shspec = shspec;
error = ENOENT;
SMBCO_FOREACH((struct smb_connobj*)vcp, &smb_vclist) {
error = smb_vc_lock(vcp, LK_EXCLUSIVE, td);
if (error)
continue;
itry {
if ((vcp->obj.co_flags & SMBV_PRIVATE) ||
!CONNADDREQ(vcp->vc_paddr, vcspec->sap) ||
strcmp(vcp->vc_username, vcspec->username) != 0)
ithrow(1);
if (vcspec->owner != SMBM_ANY_OWNER) {
if (vcp->vc_uid != vcspec->owner)
ithrow(1);
} else
exact = 0;
if (vcspec->group != SMBM_ANY_GROUP) {
if (vcp->vc_grp != vcspec->group)
ithrow(1);
} else
exact = 0;
if (vcspec->mode & SMBM_EXACT) {
if (!exact ||
(vcspec->mode & SMBM_MASK) != vcp->vc_mode)
ithrow(1);
}
if (smb_vc_access(vcp, scred, vcspec->mode) != 0)
ithrow(1);
vcspec->ssp = NULL;
if (shspec)
ithrow(smb_vc_lookupshare(vcp, shspec, scred, &vcspec->ssp));
error = 0;
break;
} icatch(error) {
smb_vc_unlock(vcp, 0, td);
} ifinally {
} iendtry;
if (error == 0)
break;
}
if (vcp) {
smb_vc_ref(vcp);
*vcpp = vcp;
}
return error;
}
int
smb_sm_lookup(struct smb_vcspec *vcspec, struct smb_sharespec *shspec,
struct smb_cred *scred, struct smb_vc **vcpp)
{
struct thread *td = scred->scr_td;
struct smb_vc *vcp;
struct smb_share *ssp = NULL;
int error;
*vcpp = vcp = NULL;
error = smb_sm_lockvclist(LK_EXCLUSIVE, td);
if (error)
return error;
error = smb_sm_lookupint(vcspec, shspec, scred, vcpp);
if (error == 0 || (vcspec->flags & SMBV_CREATE) == 0) {
smb_sm_unlockvclist(td);
return error;
}
error = smb_sm_lookupint(vcspec, NULL, scred, &vcp);
if (error) {
error = smb_vc_create(vcspec, scred, &vcp);
if (error)
goto out;
error = smb_vc_connect(vcp, scred);
if (error)
goto out;
}
if (shspec == NULL)
goto out;
error = smb_share_create(vcp, shspec, scred, &ssp);
if (error)
goto out;
error = smb_smb_treeconnect(ssp, scred);
if (error == 0)
vcspec->ssp = ssp;
else
smb_share_put(ssp, scred);
out:
smb_sm_unlockvclist(td);
if (error == 0)
*vcpp = vcp;
else if (vcp)
smb_vc_put(vcp, scred);
return error;
}
/*
* Common code for connection object
*/
static void
smb_co_init(struct smb_connobj *cp, int level, char *objname, struct thread *td)
{
SLIST_INIT(&cp->co_children);
smb_sl_init(&cp->co_interlock, objname);
lockinit(&cp->co_lock, PZERO, objname, 0, 0);
cp->co_level = level;
cp->co_usecount = 1;
KASSERT(smb_co_lock(cp, LK_EXCLUSIVE, td) == 0, ("smb_co_init: lock failed"));
}
static void
smb_co_done(struct smb_connobj *cp)
{
smb_sl_destroy(&cp->co_interlock);
lockdestroy(&cp->co_lock);
}
static void
smb_co_gone(struct smb_connobj *cp, struct smb_cred *scred)
{
struct smb_connobj *parent;
if (cp->co_gone)
cp->co_gone(cp, scred);
parent = cp->co_parent;
if (parent) {
smb_co_lock(parent, LK_EXCLUSIVE, scred->scr_td);
SLIST_REMOVE(&parent->co_children, cp, smb_connobj, co_next);
smb_co_put(parent, scred);
}
if (cp->co_free)
cp->co_free(cp);
}
void
smb_co_ref(struct smb_connobj *cp)
{
SMB_CO_LOCK(cp);
cp->co_usecount++;
SMB_CO_UNLOCK(cp);
}
void
smb_co_rele(struct smb_connobj *cp, struct smb_cred *scred)
{
struct thread *td = scred->scr_td;
SMB_CO_LOCK(cp);
if (cp->co_usecount > 1) {
cp->co_usecount--;
SMB_CO_UNLOCK(cp);
return;
}
if (cp->co_usecount == 0) {
SMBERROR("negative use_count for object %d", cp->co_level);
SMB_CO_UNLOCK(cp);
return;
}
cp->co_usecount--;
cp->co_flags |= SMBO_GONE;
lockmgr(&cp->co_lock, LK_DRAIN | LK_INTERLOCK, &cp->co_interlock, td);
smb_co_gone(cp, scred);
}
int
smb_co_get(struct smb_connobj *cp, int flags, struct smb_cred *scred)
{
int error;
if ((flags & LK_INTERLOCK) == 0)
SMB_CO_LOCK(cp);
cp->co_usecount++;
error = smb_co_lock(cp, flags | LK_INTERLOCK, scred->scr_td);
if (error) {
SMB_CO_LOCK(cp);
cp->co_usecount--;
SMB_CO_UNLOCK(cp);
return error;
}
return 0;
}
void
smb_co_put(struct smb_connobj *cp, struct smb_cred *scred)
{
struct thread *td = scred->scr_td;
int flags;
flags = LK_RELEASE;
SMB_CO_LOCK(cp);
if (cp->co_usecount > 1) {
cp->co_usecount--;
} else if (cp->co_usecount == 1) {
cp->co_usecount--;
cp->co_flags |= SMBO_GONE;
flags = LK_DRAIN;
} else {
SMBERROR("negative usecount");
}
lockmgr(&cp->co_lock, LK_RELEASE | LK_INTERLOCK, &cp->co_interlock, td);
if ((cp->co_flags & SMBO_GONE) == 0)
return;
lockmgr(&cp->co_lock, LK_DRAIN, NULL, td);
smb_co_gone(cp, scred);
}
int
smb_co_lockstatus(struct smb_connobj *cp, struct thread *td)
{
return lockstatus(&cp->co_lock, td);
}
int
smb_co_lock(struct smb_connobj *cp, int flags, struct thread *td)
{
if (cp->co_flags & SMBO_GONE)
return EINVAL;
if ((flags & LK_TYPE_MASK) == 0)
flags |= LK_EXCLUSIVE;
if (smb_co_lockstatus(cp, td) == LK_EXCLUSIVE &&
(flags & LK_CANRECURSE) == 0) {
SMBERROR("recursive lock for object %d\n", cp->co_level);
return 0;
}
return lockmgr(&cp->co_lock, flags, &cp->co_interlock, td);
}
void
smb_co_unlock(struct smb_connobj *cp, int flags, struct thread *td)
{
(void)lockmgr(&cp->co_lock, flags | LK_RELEASE, &cp->co_interlock, td);
}
static void
smb_co_addchild(struct smb_connobj *parent, struct smb_connobj *child)
{
KASSERT(smb_co_lockstatus(parent, curthread) == LK_EXCLUSIVE, ("smb_co_addchild: parent not locked"));
KASSERT(smb_co_lockstatus(child, curthread) == LK_EXCLUSIVE, ("smb_co_addchild: child not locked"));
smb_co_ref(parent);
SLIST_INSERT_HEAD(&parent->co_children, child, co_next);
child->co_parent = parent;
}
/*
* Session implementation
*/
int
smb_vc_create(struct smb_vcspec *vcspec,
struct smb_cred *scred, struct smb_vc **vcpp)
{
struct smb_vc *vcp;
struct thread *td = scred->scr_td;
struct ucred *cred = scred->scr_cred;
uid_t uid = vcspec->owner;
gid_t gid = vcspec->group;
uid_t realuid = cred->cr_uid;
char *domain = vcspec->domain;
int error, isroot;
isroot = smb_suser(cred) == 0;
/*
* Only superuser can create VCs with different uid and gid
*/
if (uid != SMBM_ANY_OWNER && uid != realuid && !isroot)
return EPERM;
if (gid != SMBM_ANY_GROUP && !groupmember(gid, cred) && !isroot)
return EPERM;
vcp = smb_zmalloc(sizeof(*vcp), M_SMBCONN, M_WAITOK);
smb_co_init(VCTOCP(vcp), SMBL_VC, "smb_vc", td);
vcp->obj.co_free = smb_vc_free;
vcp->obj.co_gone = smb_vc_gone;
vcp->vc_number = smb_vcnext++;
vcp->vc_timo = SMB_DEFRQTIMO;
vcp->vc_smbuid = SMB_UID_UNKNOWN;
vcp->vc_mode = vcspec->rights & SMBM_MASK;
vcp->obj.co_flags = vcspec->flags & (SMBV_PRIVATE | SMBV_SINGLESHARE);
vcp->vc_tdesc = &smb_tran_nbtcp_desc;
if (uid == SMBM_ANY_OWNER)
uid = realuid;
if (gid == SMBM_ANY_GROUP)
gid = cred->cr_groups[0];
vcp->vc_uid = uid;
vcp->vc_grp = gid;
smb_sl_init(&vcp->vc_stlock, "vcstlock");
error = 0;
itry {
vcp->vc_paddr = dup_sockaddr(vcspec->sap, 1);
ierror(vcp->vc_paddr == NULL, ENOMEM);
vcp->vc_laddr = dup_sockaddr(vcspec->lap, 1);
ierror(vcp->vc_laddr == NULL, ENOMEM);
ierror((vcp->vc_pass = smb_strdup(vcspec->pass)) == NULL, ENOMEM);
vcp->vc_domain = smb_strdup((domain && domain[0]) ? domain : "NODOMAIN");
ierror(vcp->vc_domain == NULL, ENOMEM);
ierror((vcp->vc_srvname = smb_strdup(vcspec->srvname)) == NULL, ENOMEM);
ierror((vcp->vc_username = smb_strdup(vcspec->username)) == NULL, ENOMEM);
ithrow(iconv_open("tolower", vcspec->localcs, &vcp->vc_tolower));
ithrow(iconv_open("toupper", vcspec->localcs, &vcp->vc_toupper));
if (vcspec->servercs[0]) {
ithrow(iconv_open(vcspec->servercs, vcspec->localcs,
&vcp->vc_toserver));
ithrow(iconv_open(vcspec->localcs, vcspec->servercs,
&vcp->vc_tolocal));
}
ithrow(smb_iod_create(vcp));
*vcpp = vcp;
smb_co_addchild(&smb_vclist, VCTOCP(vcp));
} icatch(error) {
smb_vc_put(vcp, scred);
} ifinally {
} iendtry;
return error;
}
static void
smb_vc_free(struct smb_connobj *cp)
{
struct smb_vc *vcp = CPTOVC(cp);
if (vcp->vc_iod)
smb_iod_destroy(vcp->vc_iod);
SMB_STRFREE(vcp->vc_username);
SMB_STRFREE(vcp->vc_srvname);
SMB_STRFREE(vcp->vc_pass);
SMB_STRFREE(vcp->vc_domain);
if (vcp->vc_paddr)
free(vcp->vc_paddr, M_SONAME);
if (vcp->vc_laddr)
free(vcp->vc_laddr, M_SONAME);
if (vcp->vc_tolower)
iconv_close(vcp->vc_tolower);
if (vcp->vc_toupper)
iconv_close(vcp->vc_toupper);
if (vcp->vc_tolocal)
iconv_close(vcp->vc_tolocal);
if (vcp->vc_toserver)
iconv_close(vcp->vc_toserver);
smb_co_done(VCTOCP(vcp));
smb_sl_destroy(&vcp->vc_stlock);
free(vcp, M_SMBCONN);
}
/*
* Called when use count of VC dropped to zero.
* VC should be locked on enter with LK_DRAIN.
*/
static void
smb_vc_gone(struct smb_connobj *cp, struct smb_cred *scred)
{
struct smb_vc *vcp = CPTOVC(cp);
smb_vc_disconnect(vcp);
}
void
smb_vc_ref(struct smb_vc *vcp)
{
smb_co_ref(VCTOCP(vcp));
}
void
smb_vc_rele(struct smb_vc *vcp, struct smb_cred *scred)
{
smb_co_rele(VCTOCP(vcp), scred);
}
int
smb_vc_get(struct smb_vc *vcp, int flags, struct smb_cred *scred)
{
return smb_co_get(VCTOCP(vcp), flags, scred);
}
void
smb_vc_put(struct smb_vc *vcp, struct smb_cred *scred)
{
smb_co_put(VCTOCP(vcp), scred);
}
int
smb_vc_lock(struct smb_vc *vcp, int flags, struct thread *td)
{
return smb_co_lock(VCTOCP(vcp), flags, td);
}
void
smb_vc_unlock(struct smb_vc *vcp, int flags, struct thread *td)
{
smb_co_unlock(VCTOCP(vcp), flags, td);
}
int
smb_vc_access(struct smb_vc *vcp, struct smb_cred *scred, mode_t mode)
{
struct ucred *cred = scred->scr_cred;
if (smb_suser(cred) == 0 || cred->cr_uid == vcp->vc_uid)
return 0;
mode >>= 3;
if (!groupmember(vcp->vc_grp, cred))
mode >>= 3;
return (vcp->vc_mode & mode) == mode ? 0 : EACCES;
}
static int
smb_vc_cmpshare(struct smb_share *ssp, struct smb_sharespec *dp)
{
int exact = 1;
if (strcmp(ssp->ss_name, dp->name) != 0)
return 1;
if (dp->owner != SMBM_ANY_OWNER) {
if (ssp->ss_uid != dp->owner)
return 1;
} else
exact = 0;
if (dp->group != SMBM_ANY_GROUP) {
if (ssp->ss_grp != dp->group)
return 1;
} else
exact = 0;
if (dp->mode & SMBM_EXACT) {
if (!exact)
return 1;
return (dp->mode & SMBM_MASK) == ssp->ss_mode ? 0 : 1;
}
if (smb_share_access(ssp, dp->scred, dp->mode) != 0)
return 1;
return 0;
}
/*
* Lookup share in the given VC. Share referenced and locked on return.
* VC expected to be locked on entry and will be left locked on exit.
*/
int
smb_vc_lookupshare(struct smb_vc *vcp, struct smb_sharespec *dp,
struct smb_cred *scred, struct smb_share **sspp)
{
struct thread *td = scred->scr_td;
struct smb_share *ssp = NULL;
int error;
*sspp = NULL;
dp->scred = scred;
SMBCO_FOREACH((struct smb_connobj*)ssp, VCTOCP(vcp)) {
error = smb_share_lock(ssp, LK_EXCLUSIVE, td);
if (error)
continue;
if (smb_vc_cmpshare(ssp, dp) == 0)
break;
smb_share_unlock(ssp, 0, td);
}
if (ssp) {
smb_share_ref(ssp);
*sspp = ssp;
error = 0;
} else
error = ENOENT;
return error;
}
int
smb_vc_connect(struct smb_vc *vcp, struct smb_cred *scred)
{
return smb_iod_request(vcp->vc_iod, SMBIOD_EV_CONNECT | SMBIOD_EV_SYNC, NULL);
}
/*
* Destroy VC to server, invalidate shares linked with it.
* Transport should be locked on entry.
*/
int
smb_vc_disconnect(struct smb_vc *vcp)
{
smb_iod_request(vcp->vc_iod, SMBIOD_EV_DISCONNECT | SMBIOD_EV_SYNC, NULL);
return 0;
}
static char smb_emptypass[] = "";
const char *
smb_vc_getpass(struct smb_vc *vcp)
{
if (vcp->vc_pass)
return vcp->vc_pass;
return smb_emptypass;
}
static int
smb_vc_getinfo(struct smb_vc *vcp, struct smb_vc_info *vip)
{
bzero(vip, sizeof(struct smb_vc_info));
vip->itype = SMB_INFO_VC;
vip->usecount = vcp->obj.co_usecount;
vip->uid = vcp->vc_uid;
vip->gid = vcp->vc_grp;
vip->mode = vcp->vc_mode;
vip->flags = vcp->obj.co_flags;
vip->sopt = vcp->vc_sopt;
vip->iodstate = vcp->vc_iod->iod_state;
bzero(&vip->sopt.sv_skey, sizeof(vip->sopt.sv_skey));
snprintf(vip->srvname, sizeof(vip->srvname), "%s", vcp->vc_srvname);
snprintf(vip->vcname, sizeof(vip->vcname), "%s", vcp->vc_username);
return 0;
}
u_short
smb_vc_nextmid(struct smb_vc *vcp)
{
u_short r;
SMB_CO_LOCK(&vcp->obj);
r = vcp->vc_mid++;
SMB_CO_UNLOCK(&vcp->obj);
return r;
}
/*
* Share implementation
*/
/*
* Allocate share structure and attach it to the given VC
* Connection expected to be locked on entry. Share will be returned
* in locked state.
*/
int
smb_share_create(struct smb_vc *vcp, struct smb_sharespec *shspec,
struct smb_cred *scred, struct smb_share **sspp)
{
struct smb_share *ssp;
struct thread *td = scred->scr_td;
struct ucred *cred = scred->scr_cred;
uid_t realuid = cred->cr_uid;
uid_t uid = shspec->owner;
gid_t gid = shspec->group;
int error, isroot;
isroot = smb_suser(cred) == 0;
/*
* Only superuser can create shares with different uid and gid
*/
if (uid != SMBM_ANY_OWNER && uid != realuid && !isroot)
return EPERM;
if (gid != SMBM_ANY_GROUP && !groupmember(gid, cred) && !isroot)
return EPERM;
error = smb_vc_lookupshare(vcp, shspec, scred, &ssp);
if (!error) {
smb_share_put(ssp, scred);
return EEXIST;
}
if (uid == SMBM_ANY_OWNER)
uid = realuid;
if (gid == SMBM_ANY_GROUP)
gid = cred->cr_groups[0];
ssp = smb_zmalloc(sizeof(*ssp), M_SMBCONN, M_WAITOK);
smb_co_init(SSTOCP(ssp), SMBL_SHARE, "smbss", td);
ssp->obj.co_free = smb_share_free;
ssp->obj.co_gone = smb_share_gone;
smb_sl_init(&ssp->ss_stlock, "ssstlock");
ssp->ss_name = smb_strdup(shspec->name);
if (shspec->pass && shspec->pass[0])
ssp->ss_pass = smb_strdup(shspec->pass);
ssp->ss_type = shspec->stype;
ssp->ss_tid = SMB_TID_UNKNOWN;
ssp->ss_uid = uid;
ssp->ss_grp = gid;
ssp->ss_mode = shspec->rights & SMBM_MASK;
smb_co_addchild(VCTOCP(vcp), SSTOCP(ssp));
*sspp = ssp;
return 0;
}
static void
smb_share_free(struct smb_connobj *cp)
{
struct smb_share *ssp = CPTOSS(cp);
SMB_STRFREE(ssp->ss_name);
SMB_STRFREE(ssp->ss_pass);
smb_sl_destroy(&ssp->ss_stlock);
smb_co_done(SSTOCP(ssp));
free(ssp, M_SMBCONN);
}
static void
smb_share_gone(struct smb_connobj *cp, struct smb_cred *scred)
{
struct smb_share *ssp = CPTOSS(cp);
smb_smb_treedisconnect(ssp, scred);
}
void
smb_share_ref(struct smb_share *ssp)
{
smb_co_ref(SSTOCP(ssp));
}
void
smb_share_rele(struct smb_share *ssp, struct smb_cred *scred)
{
smb_co_rele(SSTOCP(ssp), scred);
}
int
smb_share_get(struct smb_share *ssp, int flags, struct smb_cred *scred)
{
return smb_co_get(SSTOCP(ssp), flags, scred);
}
void
smb_share_put(struct smb_share *ssp, struct smb_cred *scred)
{
smb_co_put(SSTOCP(ssp), scred);
}
int
smb_share_lock(struct smb_share *ssp, int flags, struct thread *td)
{
return smb_co_lock(SSTOCP(ssp), flags, td);
}
void
smb_share_unlock(struct smb_share *ssp, int flags, struct thread *td)
{
smb_co_unlock(SSTOCP(ssp), flags, td);
}
int
smb_share_access(struct smb_share *ssp, struct smb_cred *scred, mode_t mode)
{
struct ucred *cred = scred->scr_cred;
if (smb_suser(cred) == 0 || cred->cr_uid == ssp->ss_uid)
return 0;
mode >>= 3;
if (!groupmember(ssp->ss_grp, cred))
mode >>= 3;
return (ssp->ss_mode & mode) == mode ? 0 : EACCES;
}
void
smb_share_invalidate(struct smb_share *ssp)
{
ssp->ss_tid = SMB_TID_UNKNOWN;
}
int
smb_share_valid(struct smb_share *ssp)
{
return ssp->ss_tid != SMB_TID_UNKNOWN &&
ssp->ss_vcgenid == SSTOVC(ssp)->vc_genid;
}
const char*
smb_share_getpass(struct smb_share *ssp)
{
struct smb_vc *vcp;
if (ssp->ss_pass)
return ssp->ss_pass;
vcp = SSTOVC(ssp);
if (vcp->vc_pass)
return vcp->vc_pass;
return smb_emptypass;
}
static int
smb_share_getinfo(struct smb_share *ssp, struct smb_share_info *sip)
{
bzero(sip, sizeof(struct smb_share_info));
sip->itype = SMB_INFO_SHARE;
sip->usecount = ssp->obj.co_usecount;
sip->tid = ssp->ss_tid;
sip->type= ssp->ss_type;
sip->uid = ssp->ss_uid;
sip->gid = ssp->ss_grp;
sip->mode= ssp->ss_mode;
sip->flags = ssp->obj.co_flags;
snprintf(sip->sname, sizeof(sip->sname), "%s", ssp->ss_name);
return 0;
}
/*
* Dump an entire tree into sysctl call
*/
static int
smb_sysctl_treedump(SYSCTL_HANDLER_ARGS)
{
struct thread *td = req->td;
struct smb_cred scred;
struct smb_vc *vcp;
struct smb_share *ssp;
struct smb_vc_info vci;
struct smb_share_info ssi;
int error, itype;
smb_makescred(&scred, td, td->td_ucred);
error = smb_sm_lockvclist(LK_SHARED, td);
if (error)
return error;
SMBCO_FOREACH((struct smb_connobj*)vcp, &smb_vclist) {
error = smb_vc_lock(vcp, LK_SHARED, td);
if (error)
continue;
smb_vc_getinfo(vcp, &vci);
error = SYSCTL_OUT(req, &vci, sizeof(struct smb_vc_info));
if (error) {
smb_vc_unlock(vcp, 0, td);
break;
}
SMBCO_FOREACH((struct smb_connobj*)ssp, VCTOCP(vcp)) {
error = smb_share_lock(ssp, LK_SHARED, td);
if (error) {
error = 0;
continue;
}
smb_share_getinfo(ssp, &ssi);
smb_share_unlock(ssp, 0, td);
error = SYSCTL_OUT(req, &ssi, sizeof(struct smb_share_info));
if (error)
break;
}
smb_vc_unlock(vcp, 0, td);
if (error)
break;
}
if (!error) {
itype = SMB_INFO_NONE;
error = SYSCTL_OUT(req, &itype, sizeof(itype));
}
smb_sm_unlockvclist(td);
return error;
}