int decode_DIS_ReqHdr(
struct tcp_chan *chan,
struct batch_request *preq,
int *proto_type,
int *proto_ver)
{
int rc;
/* We got the protocol type in process_pbs_server_port
We are just hard coding it here */
*proto_type = PBS_BATCH_PROT_TYPE;
*proto_ver = disrui(chan, &rc);
if (rc == 0)
{
preq->rq_type = disrui(chan, &rc);
}
if (rc != 0)
{
return(rc);
}
return(disrfst(chan, PBS_MAXUSER, preq->rq_user));
} /* END decode_DIS_ReqHdr() */
int decode_DIS_JobFile(
int sock,
struct batch_request *preq)
{
int rc;
size_t amt;
preq->rq_ind.rq_jobfile.rq_data = 0;
preq->rq_ind.rq_jobfile.rq_sequence = disrui(sock, &rc);
if (rc)
{
return(rc);
}
preq->rq_ind.rq_jobfile.rq_type = disrui(sock, &rc);
if (rc)
{
return(rc);
}
preq->rq_ind.rq_jobfile.rq_size = disrui(sock, &rc);
if (rc)
{
return(rc);
}
if ((rc = disrfst(sock, PBS_MAXSVRJOBID + 1, preq->rq_ind.rq_jobfile.rq_jobid)) != 0)
{
return(rc);
}
preq->rq_ind.rq_jobfile.rq_data = disrcs(sock, &amt, &rc);
if (((long)amt != preq->rq_ind.rq_jobfile.rq_size) && (rc == 0))
rc = DIS_EOD;
if (rc)
{
if (preq->rq_ind.rq_jobfile.rq_data)
free(preq->rq_ind.rq_jobfile.rq_data);
preq->rq_ind.rq_jobfile.rq_data = 0;
}
return(rc);
}
int decode_DIS_SchedLock(int sock, struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_lock = disrui(sock, &rc);
return rc;
}
/**
* @brief
* Decode PBS batch request to authenticate user
* @param [in] sock socket connection
* @param [in] preq PBS bath request
* @return in
* @retval 0 on success
* @retval > 0 on failure
*/
int
decode_DIS_AuthenResvPort(int sock, struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_authen_resvport.rq_port = disrui(sock, &rc);
return rc;
}
int
decode_DIS_ShutDown(int sock, struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_shutdown = disrui(sock, &rc);
return rc;
}
int decode_DIS_PowerState(
struct tcp_chan *chan,
struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_powerstate = disrui(chan,&rc);
return rc;
}
int decode_DIS_ShutDown(
struct tcp_chan *chan,
struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_shutdown = disrui(chan, &rc);
return rc;
}
int decode_DIS_Authen(
struct tcp_chan *chan,
struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_authen.rq_port = disrui(chan, &rc);
return(rc);
} /* END decode_DIS_Authen() */
int decode_DIS_ReqHdr(
int sock,
struct batch_request *preq,
int *proto_type,
int *proto_ver)
{
int rc;
*proto_type = disrui(sock, &rc);
if (rc != 0)
{
return(rc);
}
if (*proto_type != PBS_BATCH_PROT_TYPE)
{
return(DIS_PROTO);
}
*proto_ver = disrui(sock, &rc);
if (rc)
{
return(rc);
}
preq->rq_type = disrui(sock, &rc);
if (rc != 0)
{
return(rc);
}
return(disrfst(sock, PBS_MAXUSER + 1, preq->rq_user));
} /* END decode_DIS_ReqHdr() */
int
decode_DIS_JobCred(int sock, struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_jobcred.rq_data = 0;
preq->rq_ind.rq_jobcred.rq_type = disrui(sock, &rc);
if (rc) return rc;
preq->rq_ind.rq_jobcred.rq_data = disrcs(sock,
(size_t *)&preq->rq_ind.rq_jobcred.rq_size,
&rc);
return rc;
}
int
decode_DIS_ReqExtend(int sock, struct batch_request *preq)
{
int i;
int rc;
i = disrui(sock, &rc); /* indicates if an extension exists */
if (rc == 0) {
if (i != 0) {
preq->rq_extend = disrst(sock, &rc);
}
}
return (rc);
}
int
decode_DIS_AltAuthen(int sock, struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_authen.rq_port = disrui(sock, &rc);
if(rc != 0)
{
return(rc);
}
rc = disrfst(sock, PBS_MAXCREDENTIAL_LEN + 1, preq->rq_ind.rq_authen.rq_cred);
return rc;
}
int
decode_DIS_MessageJob(int sock, struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_message.rq_text = 0;
rc = disrfst(sock, PBS_MAXSVRJOBID+1, preq->rq_ind.rq_message.rq_jid);
if (rc) return rc;
preq->rq_ind.rq_message.rq_file = disrui(sock, &rc);
if (rc) return rc;
preq->rq_ind.rq_message.rq_text = disrst(sock, &rc);
return rc;
}
int
decode_DIS_TrackJob(int sock, struct batch_request *preq)
{
int rc;
rc = disrfst(sock, PBS_MAXSVRJOBID+1, preq->rq_ind.rq_track.rq_jid);
if (rc) return rc;
preq->rq_ind.rq_track.rq_hopcount = disrui(sock, &rc);
if (rc) return rc;
rc = disrfst(sock, PBS_MAXDEST+1, preq->rq_ind.rq_track.rq_location);
if (rc) return rc;
preq->rq_ind.rq_track.rq_state[0] = disruc(sock, &rc);
return rc;
}
int
decode_DIS_UserCred(int sock, struct batch_request *preq)
{
int rc;
rc = disrfst(sock, PBS_MAXUSER+1, preq->rq_ind.rq_usercred.rq_user);
if (rc) return rc;
preq->rq_ind.rq_usercred.rq_type = disrui(sock, &rc);
if (rc) return rc;
preq->rq_ind.rq_usercred.rq_data = 0;
preq->rq_ind.rq_usercred.rq_data = disrcs(sock,
(size_t *)&preq->rq_ind.rq_usercred.rq_size,
&rc);
return rc;
}
int decode_DIS_AltAuthen(
struct tcp_chan *chan,
struct batch_request *preq)
{
int rc;
preq->rq_ind.rq_authen.rq_port = disrui(chan, &rc);
if (rc != 0)
{
return(rc);
}
rc = disrfst(chan, PBS_MAXCREDENTIAL_LEN, preq->rq_ind.rq_authen.rq_cred);
return rc;
} /* END decode_DIS_AltAuthen() */
/**
* @brief
* vn_decode_DIS - read verison 3 or 4 vnode definition information from
* Mom.
* @par Functionality:
* The V4 over-the-wire representation of a placement set list (vnl_t) is
* a superset of V3. V4 adds the ability to specify the type of an
* attribute/resource (and reserves a place in the protocol for flags).
* The V3 over-the-wire representation of a placement set list (vnl_t) is
*
* version unsigned integer the version of the following
* information
*
* Version PS_DIS_V3 consists of
*
* vnl_modtime signed long this OTW format could be
* problematic: the Open Group
* Base Specifications Issue 6
* says that time_t ``shall be
* integer or real-floating''
*
* vnl_used unsigned integer number of entries in the vnal_t
* array to follow
*
*
* There follows, for each element of the vnal_t array,
*
* vnal_id string
*
* vnal_used unsigned integer number of entries in the vna_t
* array to follow
*
* vna_name string name of resource
* vna_val string value of resource
* Following added in V4
* vna_type int type of attribute/resource
* vna_flag int flag of attribute/resource (-h)
*
*
* @param[in] fd - file (socket) descriptor from which to read
* @param[out] rcp - pointer to int into which to return the error value,
* either DIS_SUCCESS or some DIS_* error.
*
* @return vnl_t *
* @retval pointer to decoded vnode information which has been malloc-ed.
* @retval NULL on error, see rcp value
*
* @par Side Effects: None
*
* @par MT-safe: yes
*
*/
vnl_t *
vn_decode_DIS(int fd, int *rcp)
{
unsigned int vers;
vers = disrui(fd, rcp);
if (*rcp != DIS_SUCCESS)
return ((vnl_t *) NULL);
switch (vers) {
case PS_DIS_V3:
return (vn_decode_DIS_V3(fd, rcp));
case PS_DIS_V4:
return (vn_decode_DIS_V4(fd, rcp));
default:
*rcp = DIS_PROTO;
return ((vnl_t *) NULL);
}
}
int dis_request_read(
struct tcp_chan *chan,
struct batch_request *request) /* server internal structure */
{
int proto_type;
int proto_ver;
int rc; /* return code */
char log_buf[LOCAL_LOG_BUF_SIZE];
#ifdef PBS_MOM
/* NYI: talk to Ken about this. This is necessary due to the changes to
* decode_DIS_ReqHdr */
proto_type = disrui(chan, &rc);
#endif
/* Decode the Request Header, that will tell the request type */
if ((rc = decode_DIS_ReqHdr(chan, request, &proto_type, &proto_ver)))
{
if (rc == DIS_EOF)
{
return(EOF);
}
sprintf(log_buf, "req header bad, dis error %d (%s), type=%s",
rc,
dis_emsg[rc],
reqtype_to_txt(request->rq_type));
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
return(PBSE_DISPROTO);
}
if (proto_ver != PBS_BATCH_PROT_VER)
{
sprintf(log_buf, "conflicting version numbers, %d detected, %d expected",
proto_ver,
PBS_BATCH_PROT_VER);
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
return(PBSE_DISPROTO);
}
if ((request->rq_type < 0) || (request->rq_type >= PBS_BATCH_CEILING))
{
sprintf(log_buf, "invalid request type: %d", request->rq_type);
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
return(PBSE_DISPROTO);
}
/* Decode the Request Body based on the type */
if (LOGLEVEL >= 5)
{
sprintf(log_buf, "decoding command %s from %s",
reqtype_to_txt(request->rq_type),
request->rq_user);
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
}
switch (request->rq_type)
{
case PBS_BATCH_Disconnect:
return(PBSE_SOCKET_CLOSE); /* set EOF return */
/*NOTREACHED*/
break;
case PBS_BATCH_QueueJob:
CLEAR_HEAD(request->rq_ind.rq_queuejob.rq_attr);
rc = decode_DIS_QueueJob(chan, request);
break;
case PBS_BATCH_JobCred:
rc = decode_DIS_JobCred(chan, request);
break;
case PBS_BATCH_jobscript:
case PBS_BATCH_MvJobFile:
rc = decode_DIS_JobFile(chan, request);
break;
case PBS_BATCH_RdytoCommit:
case PBS_BATCH_Commit:
case PBS_BATCH_Rerun:
rc = decode_DIS_JobId(chan, request->rq_ind.rq_commit);
break;
case PBS_BATCH_DeleteJob:
case PBS_BATCH_HoldJob:
case PBS_BATCH_CheckpointJob:
case PBS_BATCH_ModifyJob:
case PBS_BATCH_AsyModifyJob:
rc = decode_DIS_Manage(chan, request);
break;
case PBS_BATCH_MessJob:
rc = decode_DIS_MessageJob(chan, request);
break;
case PBS_BATCH_Shutdown:
rc = decode_DIS_ShutDown(chan, request);
break;
case PBS_BATCH_SignalJob:
rc = decode_DIS_SignalJob(chan, request);
break;
case PBS_BATCH_StatusJob:
rc = decode_DIS_Status(chan, request);
break;
case PBS_BATCH_GpuCtrl:
rc = decode_DIS_GpuCtrl(chan, request);
break;
#ifndef PBS_MOM
case PBS_BATCH_LocateJob:
rc = decode_DIS_JobId(chan, request->rq_ind.rq_locate);
break;
case PBS_BATCH_Manager:
case PBS_BATCH_ReleaseJob:
rc = decode_DIS_Manage(chan, request);
break;
case PBS_BATCH_MoveJob:
case PBS_BATCH_OrderJob:
rc = decode_DIS_MoveJob(chan, request);
break;
case PBS_BATCH_RunJob:
case PBS_BATCH_AsyrunJob:
case PBS_BATCH_StageIn:
rc = decode_DIS_RunJob(chan, request);
break;
case PBS_BATCH_SelectJobs:
case PBS_BATCH_SelStat:
CLEAR_HEAD(request->rq_ind.rq_select);
rc = decode_DIS_svrattrl(chan, &request->rq_ind.rq_select);
break;
case PBS_BATCH_StatusNode:
case PBS_BATCH_StatusQue:
case PBS_BATCH_StatusSvr:
/* DIAGTODO: add PBS_BATCH_StatusDiag */
rc = decode_DIS_Status(chan, request);
break;
case PBS_BATCH_TrackJob:
rc = decode_DIS_TrackJob(chan, request);
break;
case PBS_BATCH_Rescq:
case PBS_BATCH_ReserveResc:
case PBS_BATCH_ReleaseResc:
rc = decode_DIS_Rescl(chan, request);
break;
case PBS_BATCH_RegistDep:
rc = decode_DIS_Register(chan, request);
break;
case PBS_BATCH_AuthenUser:
rc = decode_DIS_Authen(chan, request);
break;
case PBS_BATCH_AltAuthenUser:
rc = decode_DIS_AltAuthen(chan, request);
break;
case PBS_BATCH_JobObit:
rc = decode_DIS_JobObit(chan, request);
break;
#else /* PBS_MOM */
/* pbs_mom services */
case PBS_BATCH_DeleteReservation:
/* NO-OP: this one is just a header and an extension string */
break;
case PBS_BATCH_ReturnFiles:
rc = decode_DIS_ReturnFiles(chan, request);
break;
case PBS_BATCH_CopyFiles:
case PBS_BATCH_DelFiles:
rc = decode_DIS_CopyFiles(chan, request);
break;
#endif /* PBS_MOM */
default:
sprintf(log_buf, "%s: %d from %s",
pbse_to_txt(PBSE_NOSUP),
request->rq_type,
request->rq_user);
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
rc = PBSE_UNKREQ;
break;
} /* END switch (request->rq_type) */
if (rc == DIS_SUCCESS)
{
/* Decode the Request Extension, if present */
rc = decode_DIS_ReqExtend(chan, request);
if (rc != 0)
{
sprintf(log_buf, "req extension bad, dis error %d (%s), type=%s",
rc,
dis_emsg[rc],
reqtype_to_txt(request->rq_type));
log_event(PBSEVENT_DEBUG,PBS_EVENTCLASS_REQUEST,"?",log_buf);
rc = PBSE_DISPROTO;
}
}
else if (rc != PBSE_UNKREQ)
{
sprintf(log_buf, "req body bad, dis error %d (%s), type=%s",
rc,
dis_emsg[rc],
reqtype_to_txt(request->rq_type));
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_REQUEST, "?", log_buf);
rc = PBSE_DISPROTO;
}
return(rc);
} /* END dis_request_read() */
/*************************************************
* svr_is_request
*
* Return: svr_is_request always returns a non-zero value
* and it must call close_conn to close the connection
* before returning. PBSE_SOCKET_CLOSE is the code
* for a successful return. But which ever retun
* code is iused it must terminate the while loop
* in start_process_pbs_server_port.
*************************************************/
int svr_is_request(
struct tcp_chan *chan,
int version)
{
int command = 0;
int ret = DIS_SUCCESS;
int i;
int err;
char nodename[PBS_MAXHOSTNAME];
int perm = ATR_DFLAG_MGRD | ATR_DFLAG_MGWR;
unsigned long ipaddr;
unsigned short mom_port;
unsigned short rm_port;
unsigned long tmpaddr;
struct sockaddr_in *addr = NULL;
struct sockaddr s_addr;
unsigned int len = sizeof(s_addr);
struct pbsnode *node = NULL;
char *node_name = NULL;
char log_buf[LOCAL_LOG_BUF_SIZE+1];
command = disrsi(chan, &ret);
if (ret != DIS_SUCCESS)
goto err;
if (LOGLEVEL >= 4)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"message received from sock %d (version %d)",
chan->sock,
version);
log_event(PBSEVENT_ADMIN,PBS_EVENTCLASS_SERVER,__func__,log_buf);
}
if (getpeername(chan->sock, &s_addr, &len) != 0)
{
close_conn(chan->sock, FALSE);
log_err(errno,__func__, (char *)"Cannot get socket name using getpeername\n");
return(PBSE_SOCKET_CLOSE);
}
addr = (struct sockaddr_in *)&s_addr;
if (version != IS_PROTOCOL_VER)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "protocol version %d unknown from %s",
version,
netaddr(addr));
log_err(-1, __func__, log_buf);
close_conn(chan->sock, FALSE);
return PBSE_SOCKET_DATA;
}
/* check that machine is known */
mom_port = disrsi(chan, &ret);
rm_port = disrsi(chan, &ret);
if (LOGLEVEL >= 3)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"message received from addr %s: mom_port %d - rm_port %d",
netaddr(addr),
mom_port,
rm_port);
log_event(PBSEVENT_ADMIN,PBS_EVENTCLASS_SERVER,__func__,log_buf);
}
ipaddr = ntohl(addr->sin_addr.s_addr);
if ((node = AVL_find(ipaddr, mom_port, ipaddrs)) != NULL)
{
lock_node(node, __func__, "AVL_find", LOGLEVEL);
} /* END if AVL_find != NULL) */
else if (allow_any_mom)
{
char *name = get_cached_nameinfo(addr);
if (name != NULL)
snprintf(nodename, sizeof(nodename), "%s", name);
else if (getnameinfo(&s_addr, len, nodename, sizeof(nodename)-1, NULL, 0, 0) != 0)
{
tmpaddr = ntohl(addr->sin_addr.s_addr);
sprintf(nodename, "0x%lX", tmpaddr);
}
else
insert_addr_name_info(nodename, NULL, addr);
err = create_partial_pbs_node(nodename, ipaddr, perm);
if (err == PBSE_NONE)
{
node = AVL_find(ipaddr, 0, ipaddrs);
lock_node(node, __func__, "no error", LOGLEVEL);
}
}
if (node == NULL)
{
/* node not listed in trusted ipaddrs list */
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"bad attempt to connect from %s (address not trusted - check entry in server_priv/nodes)",
netaddr(addr));
if (LOGLEVEL >= 2)
{
log_record(PBSEVENT_SCHED, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
}
else
{
log_err(-1, __func__, log_buf);
}
close_conn(chan->sock, FALSE);
return PBSE_SOCKET_CLOSE;
}
if (LOGLEVEL >= 3)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"message %s (%d) received from mom on host %s (%s) (sock %d)",
PBSServerCmds2[command],
command,
node->nd_name,
netaddr(addr),
chan->sock);
log_event(PBSEVENT_ADMIN,PBS_EVENTCLASS_SERVER,__func__,log_buf);
}
switch (command)
{
case IS_NULL: /* a ping from server */
DBPRT(("%s: IS_NULL\n", __func__))
break;
case IS_UPDATE:
DBPRT(("%s: IS_UPDATE\n", __func__))
i = disrui(chan, &ret);
if (ret != DIS_SUCCESS)
{
if (LOGLEVEL >= 1)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"IS_UPDATE error %d on node %s\n", ret, node->nd_name);
log_err(ret, __func__, log_buf);
}
goto err;
}
DBPRT(("%s: IS_UPDATE %s 0x%x\n", __func__, node->nd_name, i))
update_node_state(node, i);
if ((node->nd_state & INUSE_DOWN) != 0)
{
node->nd_mom_reported_down = TRUE;
}
break;
case IS_STATUS:
if (LOGLEVEL >= 2)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"IS_STATUS received from %s", node->nd_name);
log_event(PBSEVENT_ADMIN, PBS_EVENTCLASS_SERVER, __func__, log_buf);
}
if ((node_name = strdup(node->nd_name)) == NULL)
goto err;
unlock_node(node, __func__, "before is_stat_get", LOGLEVEL);
ret = is_stat_get(node_name, chan);
node = find_nodebyname(node_name);
if (ret == SEND_HELLO)
{
struct hello_info *hi = (struct hello_info *)calloc(1, sizeof(struct hello_info));
write_tcp_reply(chan, IS_PROTOCOL, IS_PROTOCOL_VER, IS_STATUS, DIS_SUCCESS);
hi->name = strdup(node_name);
enqueue_threadpool_request(send_hierarchy_threadtask, hi);
ret = DIS_SUCCESS;
}
else
write_tcp_reply(chan,IS_PROTOCOL,IS_PROTOCOL_VER,IS_STATUS,ret);
if(node != NULL)
node->nd_stream = -1;
if (ret != DIS_SUCCESS)
{
if (LOGLEVEL >= 1)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"IS_STATUS error %d on node %s", ret, node_name);
log_err(ret, __func__, log_buf);
}
free(node_name);
goto err;
}
free(node_name);
break;
default:
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"unknown command %d sent from %s",
command,
node->nd_name);
log_err(-1, __func__, log_buf);
goto err;
break;
} /* END switch (command) */
/* must be closed because mom opens and closes this connection each time */
close_conn(chan->sock, FALSE);
if(node != NULL)
unlock_node(node, __func__, "close", LOGLEVEL);
return PBSE_SOCKET_CLOSE;
err:
/* a DIS write error has occurred */
if (node != NULL)
{
if (LOGLEVEL >= 1)
{
DBPRT(("%s: error processing node %s\n",
__func__,
node->nd_name))
}
sprintf(log_buf, "%s from %s(%s)",
dis_emsg[ret],
node->nd_name,
netaddr(addr));
unlock_node(node, __func__, "err", LOGLEVEL);
}
else
{
int
decode_DIS_svrattrl(int sock, pbs_list_head *phead)
{
int i;
unsigned int hasresc;
size_t ls;
unsigned int data_len;
unsigned int numattr;
svrattrl *psvrat;
int rc;
size_t tsize;
numattr = disrui(sock, &rc); /* number of attributes in set */
if (rc) return rc;
for (i=0; i<numattr; ++i) {
data_len = disrui(sock, &rc); /* here it is used */
if (rc) return rc;
tsize = sizeof(svrattrl) + data_len;
if ((psvrat = (svrattrl *)malloc(tsize)) == 0)
return DIS_NOMALLOC;
CLEAR_LINK(psvrat->al_link);
psvrat->al_sister = (svrattrl *)0;
psvrat->al_atopl.next = 0;
psvrat->al_tsize = tsize;
psvrat->al_name = (char *)psvrat + sizeof(svrattrl);
psvrat->al_resc = 0;
psvrat->al_value = 0;
psvrat->al_nameln = 0;
psvrat->al_rescln = 0;
psvrat->al_valln = 0;
psvrat->al_flags = 0;
psvrat->al_refct = 1;
if ((rc = disrfcs(sock, &ls, data_len, psvrat->al_name)) != 0)
break;
*(psvrat->al_name + ls++) = '\0';
psvrat->al_nameln = (int)ls;
data_len -= ls;
hasresc = disrui(sock, &rc);
if (rc) break;
if (hasresc) {
psvrat->al_resc = psvrat->al_name + ls;
rc = disrfcs(sock, &ls, data_len, psvrat->al_resc);
if (rc)
break;
*(psvrat->al_resc + ls++) = '\0';
psvrat->al_rescln = (int)ls;
data_len -= ls;
}
psvrat->al_value = psvrat->al_name + psvrat->al_nameln +
psvrat->al_rescln;
if ((rc = disrfcs(sock, &ls, data_len, psvrat->al_value)) != 0)
break;
*(psvrat->al_value + ls++) = '\0';
psvrat->al_valln = (int)ls;
psvrat->al_op = (enum batch_op)disrui(sock, &rc);
if (rc) break;
append_link(phead, &psvrat->al_link, psvrat);
}
if (rc) {
(void)free(psvrat);
}
return (rc);
}
/**
* @brief
* vn_decode_DIS_V3 - decode version 3 vnode information from Mom
*
* @par Functionality:
* See vn_decode_DIS() above, This is called from there to decode
* V3 information.
*
* @param[in] fd - socket descriptor from which to read
* @param[out] rcp - pointer to place to return error code if error.
*
* @return vnl_t *
* @retval pointer to decoded vnode information which has been malloc-ed.
* @retval NULL on error, see rcp value
*
* @par Side Effects: None
*
* @par MT-safe: yes
*
*/
static vnl_t *
vn_decode_DIS_V3(int fd, int *rcp)
{
unsigned int i, j;
unsigned int size;
time_t t;
vnl_t *vnlp;
if ((vnlp = malloc(sizeof(vnl_t))) == NULL) {
*rcp = DIS_NOMALLOC;
return ((vnl_t *) NULL);
}
t = (time_t) disrsl(fd, rcp);
if (*rcp != DIS_SUCCESS)
return ((vnl_t *) NULL);
else
vnlp->vnl_modtime = t;
size = disrui(fd, rcp);
if (*rcp != DIS_SUCCESS)
return ((vnl_t *) NULL);
else
vnlp->vnl_nelem = vnlp->vnl_used = size;
if ((vnlp->vnl_list = calloc(vnlp->vnl_nelem,
sizeof(vnal_t))) == NULL) {
free(vnlp);
*rcp = DIS_NOMALLOC;
return ((vnl_t *) NULL);
}
for (i = 0; i < vnlp->vnl_used; i++) {
vnal_t *curreslist = VNL_NODENUM(vnlp, i);
/*
* In case an error occurs and we need to free
* whatever's been allocated so far, we use the
* vnal_cur entry to record the number of vnal_t
* entries to free.
*/
vnlp->vnl_cur = i;
curreslist->vnal_id = disrst(fd, rcp);
if (*rcp != DIS_SUCCESS)
return (free_and_return(vnlp));
size = disrui(fd, rcp);
if (*rcp != DIS_SUCCESS)
return (free_and_return(vnlp));
else
curreslist->vnal_nelem = curreslist->vnal_used = size;
if ((curreslist->vnal_list = calloc(curreslist->vnal_nelem,
sizeof(vna_t))) == NULL)
return (free_and_return(vnlp));
for (j = 0; j < size; j++) {
vna_t *curres = VNAL_NODENUM(curreslist, j);
/*
* In case an error occurs and we need to free
* whatever's been allocated so far, we use the
* vnal_cur entry to record the number of vna_t
* entries to free.
*/
curreslist->vnal_cur = j;
curres->vna_name = disrst(fd, rcp);
if (*rcp != DIS_SUCCESS)
return (free_and_return(vnlp));
curres->vna_val = disrst(fd, rcp);
if (*rcp != DIS_SUCCESS)
return (free_and_return(vnlp));
}
}
*rcp = DIS_SUCCESS;
return (vnlp);
}
int
decode_DIS_attropl(int sock, struct attropl **ppatt)
{
int hasresc;
unsigned int i;
unsigned int name_len;
unsigned int numpat;
struct attropl *pat = NULL;
struct attropl *patprior = NULL;
int rc;
numpat = disrui(sock, &rc);
if (rc) return rc;
for (i = 0; i < numpat; ++i)
{
name_len = disrui(sock, &rc); /* name_len is unused here */
if (rc) break;
pat = malloc(sizeof(struct attropl));
if (pat == 0)
return DIS_NOMALLOC;
pat->next = (struct attropl *)0;
pat->name = (char *)0;
pat->resource = (char *)0;
pat->value = (char *)0;
pat->name = disrst(sock, &rc);
if (rc) break;
hasresc = disrui(sock, &rc);
if (rc) break;
if (hasresc)
{
pat->resource = disrst(sock, &rc);
if (rc) break;
}
pat->value = disrst(sock, &rc);
if (rc) break;
pat->op = (enum batch_op)disrui(sock, &rc);
if (rc) break;
if (i == 0)
{
/* first one, link to passing in pointer */
*ppatt = pat;
}
else
{
patprior->next = pat;
}
patprior = pat;
}
if (rc)
PBS_free_aopl(pat);
return rc;
}
int decode_DIS_svrattrl(
struct tcp_chan *chan,
tlist_head *phead)
{
unsigned int i;
unsigned int hasresc;
size_t ls;
unsigned int data_len;
unsigned int numattr;
svrattrl *psvrat = NULL;
int rc;
size_t tsize;
numattr = disrui(chan, &rc); /* number of attributes in set */
if (rc) return rc;
for (i = 0; i < numattr; ++i)
{
data_len = disrui(chan, &rc); /* here it is used */
if (data_len == 0)
data_len = sizeof(char);
if (rc)
return(rc);
tsize = sizeof(svrattrl) + data_len + 2; /* 2 more for nulls in name & resc*/
if ((psvrat = (svrattrl *)calloc(1, tsize)) == 0)
return DIS_NOMALLOC;
CLEAR_LINK(psvrat->al_link);
psvrat->al_atopl.next = 0;
psvrat->al_tsize = tsize;
psvrat->al_name = (char *)psvrat + sizeof(svrattrl);
psvrat->al_resc = 0;
psvrat->al_value = 0;
psvrat->al_nameln = 0;
psvrat->al_rescln = 0;
psvrat->al_valln = 0;
psvrat->al_flags = 0;
if ((rc = disrfcs(chan, &ls, data_len, psvrat->al_name)))
break;
*(psvrat->al_name + ls++) = '\0';
psvrat->al_nameln = (int)ls;
data_len -= ls;
hasresc = disrui(chan, &rc);
if (rc)
break;
if (hasresc)
{
psvrat->al_resc = psvrat->al_name + ls;
if ((rc = disrfcs(chan, &ls, data_len, psvrat->al_resc)))
break;
*(psvrat->al_resc + ls++) = '\0';
psvrat->al_rescln = (int)ls;
data_len -= ls;
}
psvrat->al_value = psvrat->al_name + psvrat->al_nameln +
psvrat->al_rescln;
if ((rc = disrfcs(chan, &ls, data_len, psvrat->al_value)))
break;
*(psvrat->al_value + ls++) = '\0';
psvrat->al_valln = (int)ls;
psvrat->al_op = (enum batch_op)disrui(chan, &rc);
if (rc) break;
append_link(phead, &psvrat->al_link, psvrat);
}
if (rc)
{
(void)free(psvrat);
}
return (rc);
}
/*************************************************
* svr_is_request
*
* Return: svr_is_request always returns a non-zero value
* and it must call close_conn to close the connection
* before returning. PBSE_SOCKET_CLOSE is the code
* for a successful return. But which ever retun
* code is iused it must terminate the while loop
* in start_process_pbs_server_port.
*************************************************/
void *svr_is_request(
void *v)
{
int command = 0;
int ret = DIS_SUCCESS;
int i;
int err;
char nodename[PBS_MAXHOSTNAME];
int perm = ATR_DFLAG_MGRD | ATR_DFLAG_MGWR;
unsigned long ipaddr;
unsigned short mom_port;
unsigned short rm_port;
unsigned long tmpaddr;
struct sockaddr_in addr;
struct pbsnode *node = NULL;
char log_buf[LOCAL_LOG_BUF_SIZE+1];
char msg_buf[80];
char tmp[80];
int version;
struct tcp_chan *chan;
long *args;
is_request_info *isr = (is_request_info *)v;
if (isr == NULL)
return(NULL);
chan = isr->chan;
args = isr->args;
version = disrsi(chan, &ret);
if (ret != DIS_SUCCESS)
{
log_err(-1, __func__, "Cannot read version - skipping this request.\n");
close_conn(chan->sock, FALSE);
DIS_tcp_cleanup(chan);
return(NULL);
}
command = disrsi(chan, &ret);
if (ret != DIS_SUCCESS)
{
snprintf(log_buf, sizeof(log_buf), "could not read command: %d", ret);
log_err(-1, __func__, log_buf);
close_conn(chan->sock, FALSE);
DIS_tcp_cleanup(chan);
return(NULL);
}
if (LOGLEVEL >= 4)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"message received from sock %d (version %d)",
chan->sock,
version);
log_event(PBSEVENT_ADMIN, PBS_EVENTCLASS_SERVER, __func__, log_buf);
}
/* Just a note to let us know we only do IPv4 for now */
addr.sin_family = AF_INET;
memcpy(&addr.sin_addr, (void *)&args[1], sizeof(struct in_addr));
addr.sin_port = args[2];
if (version != IS_PROTOCOL_VER)
{
netaddr_long(args[1], tmp);
sprintf(msg_buf, "%s:%ld", tmp, args[2]);
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "protocol version %d unknown from %s",
version,
msg_buf);
log_err(-1, __func__, log_buf);
close_conn(chan->sock, FALSE);
DIS_tcp_cleanup(chan);
return(NULL);
}
/* check that machine is known */
mom_port = disrsi(chan, &ret);
rm_port = disrsi(chan, &ret);
if (LOGLEVEL >= 3)
{
netaddr_long(args[1], tmp);
sprintf(msg_buf, "%s:%ld", tmp, args[2]);
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"message received from addr %s: mom_port %d - rm_port %d",
msg_buf,
mom_port,
rm_port);
log_event(PBSEVENT_ADMIN,PBS_EVENTCLASS_SERVER,__func__,log_buf);
}
ipaddr = args[1];
if ((node = AVL_find(ipaddr, mom_port, ipaddrs)) != NULL)
{
node->lock_node(__func__, "AVL_find", LOGLEVEL);
} /* END if AVL_find != NULL) */
else if (allow_any_mom)
{
const char *name = get_cached_nameinfo(&addr);
if (name != NULL)
snprintf(nodename, sizeof(nodename), "%s", name);
else if (getnameinfo((struct sockaddr *)&addr, sizeof(addr), nodename, sizeof(nodename)-1, NULL, 0, 0) != 0)
{
tmpaddr = ntohl(addr.sin_addr.s_addr);
sprintf(nodename, "0x%lX", tmpaddr);
}
else
insert_addr_name_info(NULL, nodename);
err = create_partial_pbs_node(nodename, ipaddr, perm);
if (err == PBSE_NONE)
{
node = AVL_find(ipaddr, 0, ipaddrs);
node->lock_node(__func__, "no error", LOGLEVEL);
}
}
if (node == NULL)
{
/* node not listed in trusted ipaddrs list */
netaddr_long(args[1], tmp);
sprintf(msg_buf, "%s:%ld", tmp, args[2]);
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"bad attempt to connect from %s (address not trusted - check entry in server_priv/nodes)",
msg_buf);
if (LOGLEVEL >= 2)
{
log_record(PBSEVENT_SCHED, PBS_EVENTCLASS_REQUEST, __func__, log_buf);
}
else
{
log_err(-1, __func__, log_buf);
}
close_conn(chan->sock, FALSE);
DIS_tcp_cleanup(chan);
return(NULL);
}
if (LOGLEVEL >= 3)
{
netaddr_long(args[1], tmp);
sprintf(msg_buf, "%s:%ld", tmp, args[2]);
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"message %s (%d) received from mom on host %s (%s) (sock %d)",
PBSServerCmds2[command],
command,
node->get_name(),
msg_buf,
chan->sock);
log_event(PBSEVENT_ADMIN,PBS_EVENTCLASS_SERVER,__func__,log_buf);
}
mutex_mgr node_mutex(&node->nd_mutex, true);
switch (command)
{
case IS_NULL: /* a ping from server */
DBPRT(("%s: IS_NULL\n", __func__))
break;
case IS_UPDATE:
DBPRT(("%s: IS_UPDATE\n", __func__))
i = disrui(chan, &ret);
if (ret != DIS_SUCCESS)
{
if (LOGLEVEL >= 1)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"IS_UPDATE error %d on node %s\n", ret, node->get_name());
log_err(ret, __func__, log_buf);
}
goto err;
}
DBPRT(("%s: IS_UPDATE %s 0x%x\n", __func__, node->get_name(), i))
update_node_state(node, i);
if ((node->nd_state & INUSE_DOWN) != 0)
{
node->nd_mom_reported_down = TRUE;
}
break;
case IS_STATUS:
{
std::string node_name = node->get_name();
if (LOGLEVEL >= 2)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"IS_STATUS received from %s", node->get_name());
log_event(PBSEVENT_ADMIN, PBS_EVENTCLASS_SERVER, __func__, log_buf);
}
node_mutex.unlock();
ret = is_stat_get(node_name.c_str(), chan);
node = find_nodebyname(node_name.c_str());
if (node != NULL)
{
node->nd_stream = -1;
node_mutex.mark_as_locked();
if (ret == SEND_HELLO)
{
//struct hello_info *hi = new hello_info(node->nd_id);
write_tcp_reply(chan, IS_PROTOCOL, IS_PROTOCOL_VER, IS_STATUS, DIS_SUCCESS);
hierarchy_handler.sendHierarchyToANode(node);
ret = DIS_SUCCESS;
}
else
write_tcp_reply(chan,IS_PROTOCOL,IS_PROTOCOL_VER,IS_STATUS,ret);
}
if (ret != DIS_SUCCESS)
{
if (LOGLEVEL >= 1)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"IS_STATUS error %d on node %s", ret, node_name.c_str());
log_err(ret, __func__, log_buf);
}
goto err;
}
break;
}
default:
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"unknown command %d sent from %s",
command,
node->get_name());
log_err(-1, __func__, log_buf);
goto err;
break;
} /* END switch (command) */
/* must be closed because mom opens and closes this connection each time */
close_conn(chan->sock, FALSE);
DIS_tcp_cleanup(chan);
return(NULL);
err:
/* a DIS write error has occurred */
if (node != NULL)
{
if (LOGLEVEL >= 1)
{
DBPRT(("%s: error processing node %s\n",
__func__,
node->get_name()))
}
netaddr_long(args[1], tmp);
sprintf(msg_buf, "%s:%ld", tmp, args[2]);
sprintf(log_buf, "%s from %s(%s)",
dis_emsg[ret],
node->get_name(),
msg_buf);
}
else
{
int
decode_DIS_attrl(int sock, struct attrl **ppatt)
{
int hasresc;
int i;
unsigned int numpat;
struct attrl *pat = 0;
struct attrl *patprior = 0;
int rc;
numpat = disrui(sock, &rc);
if (rc) return rc;
for (i=0; i < numpat; ++i) {
(void) disrui(sock, &rc);
if (rc) break;
pat = malloc(sizeof(struct attrl));
if (pat == 0)
return DIS_NOMALLOC;
pat->next = (struct attrl *)0;
pat->name = (char *)0;
pat->resource = (char *)0;
pat->value = (char *)0;
pat->name = disrst(sock, &rc);
if (rc) break;
hasresc = disrui(sock, &rc);
if (rc) break;
if (hasresc) {
pat->resource = disrst(sock, &rc);
if (rc) break;
}
pat->value = disrst(sock, &rc);
if (rc) break;
#ifdef NAS /* localmod 005 */
pat->op = (enum batch_op) disrui(sock, &rc);
#else
pat->op = disrui(sock, &rc);
#endif /* localmod 005 */
if (rc) break;
if (i == 0) {
/* first one, link to passing in pointer */
*ppatt = pat;
} else {
patprior->next = pat;
}
patprior = pat;
}
if (rc)
PBS_free_aopl((struct attropl *)pat);
return rc;
}