int main( )
{
uint8_t len = 1;
uint64_t loops = 50000;
scif_epd_t endpoint;
struct scif_portID portid;
int ret;
endpoint = scif_open( );
if( endpoint == SCIF_OPEN_FAILED )
{
printf("scif open failed\n");
return 1;
}
ret = scif_bind(endpoint, 23955);
if(ret==-1)
{
printf("scif_bind failed");
return 1;
}
portid.node = 0;
portid.port = 23968;
ret = scif_connect(endpoint, &portid);
for( int attempt = 0; ret == -1 && attempt < 10; ++attempt )
{
sleep(1);
ret = scif_connect(endpoint, &portid);
}
if (ret==-1)
{
printf("scif_connect failed\n");
return 1;
}
uint8_t* buffer = malloc(len * sizeof(uint8_t));
struct timespec start_latency, stop_latency;
clock_gettime(CLOCK_REALTIME, &start_latency);
for (uint64_t i = 0; i<loops; i++)
{
scif_recv(endpoint, (void*)buffer, len, SCIF_RECV_BLOCK);
scif_send(endpoint, &buffer[0], len, SCIF_SEND_BLOCK);
}
clock_gettime(CLOCK_REALTIME, & stop_latency);
double time_latency = (stop_latency.tv_sec - start_latency.tv_sec) + (stop_latency.tv_nsec - start_latency.tv_nsec) / NANOSECONDS;
double result = (time_latency/(2*loops));
printf("%1f\n", result);
free(buffer);
scif_close(endpoint);
}
static inline int mca_btl_scif_ep_connect_start_active (mca_btl_base_endpoint_t *ep) {
int rc = OPAL_SUCCESS;
BTL_VERBOSE(("initiaiting connection to remote peer %d with port: %u on local scif node: %u",
ep->peer_proc->proc_name.vpid, ep->port_id.port, ep->port_id.node));
opal_mutex_lock (&ep->lock);
do {
if (MCA_BTL_SCIF_EP_STATE_INIT != ep->state) {
/* the accept thread has already finished this connection */
rc = OPAL_SUCCESS;
break;
}
ep->state = MCA_BTL_SCIF_EP_STATE_CONNECTING;
ep->scif_epd = scif_open ();
if (OPAL_UNLIKELY(SCIF_OPEN_FAILED == ep->scif_epd)) {
BTL_VERBOSE(("error creating new scif endpoint"));
rc = OPAL_ERROR;
break;
}
rc = scif_connect (ep->scif_epd, &ep->port_id);
if (OPAL_UNLIKELY(-1 == rc)) {
/* the connection attempt failed. this could mean the peer is currently
* processing connections. we will to try again later. */
BTL_VERBOSE(("error connecting to scif peer. %d", errno));
rc = OPAL_ERR_RESOURCE_BUSY;
break;
}
rc = scif_send (ep->scif_epd, &OPAL_PROC_MY_NAME, sizeof (OPAL_PROC_MY_NAME), SCIF_SEND_BLOCK);
if (OPAL_UNLIKELY(-1 == rc)) {
BTL_VERBOSE(("error in scif_send"));
rc = OPAL_ERROR;
break;
}
/* build connection data */
rc = mca_btl_scif_ep_connect_finish (ep, false);
} while (0);
if (OPAL_SUCCESS != rc) {
scif_close (ep->scif_epd);
ep->scif_epd = -1;
ep->state = MCA_BTL_SCIF_EP_STATE_INIT;
}
opal_mutex_unlock (&ep->lock);
return rc;
}
scif_epd_t scif_obc(void)
{
scif_epd_t epd;
struct scif_portID portID;
int connection_port, request_port;
int err, tries = MAX_TRIES;
/* create "socket" and connect */
request_port = LOCAL_PORT;
portID.node = PEER_NODE;
portID.port = PEER_PORT;
if ((epd = scif_open()) == SCIF_OPEN_FAILED) {
fprintf(stderr, "= scif_open failed with error %d\n", (int) epd);
exit(EXIT_FAILURE);
}
if ((connection_port = scif_bind(epd, request_port)) < 0) {
fprintf(stderr, "= scif_bind failed with error %d\n",
connection_port);
exit(EXIT_FAILURE);
}
printf("= scif_bind to port %d success\n", connection_port);
__retry:
if ((err = scif_connect(epd, &portID)) < 0) {
if ((errno == ECONNREFUSED) && (tries > 0)) {
printf("= connection to node %d failed : trial %d\n",
portID.node, tries);
tries--;
sleep(1);
goto __retry;
}
fprintf(stderr, "= scif_connect failed with error %d\n", errno);
exit(EXIT_FAILURE);
}
printf("= conect to node %d success\n", portID.node);
return epd;
}
static int
mca_btl_scif_module_finalize (struct mca_btl_base_module_t *btl)
{
mca_btl_scif_module_t *scif_module = (mca_btl_scif_module_t *) btl;
unsigned int i;
OBJ_DESTRUCT(&mca_btl_scif_module.dma_frags);
OBJ_DESTRUCT(&mca_btl_scif_module.eager_frags);
mca_btl_scif_module.exiting = true;
/* close all open connections and release endpoints */
if (NULL != scif_module->endpoints) {
for (i = 0 ; i < scif_module->endpoint_count ; ++i) {
mca_btl_scif_ep_release (scif_module->endpoints + i);
}
free (scif_module->endpoints);
scif_module->endpoint_count = 0;
scif_module->endpoints = NULL;
}
/* close the listening endpoint */
if (mca_btl_scif_module.listening && -1 != mca_btl_scif_module.scif_fd) {
/* wake up the scif thread */
scif_epd_t tmpfd;
tmpfd = scif_open();
scif_connect (tmpfd, &mca_btl_scif_module.port_id);
pthread_join(mca_btl_scif_module.listen_thread, NULL);
scif_close(tmpfd);
scif_close (mca_btl_scif_module.scif_fd);
}
mca_btl_scif_module.scif_fd = -1;
return OPAL_SUCCESS;
}
void *
mic_monitor(void *arg)
{
struct mic_info *mic;
struct mpssd_info *mpssdi;
pthread_attr_t attr;
struct scif_portID sendID = {0, MPSSD_MONSEND};
struct scif_portID recvID;
scif_epd_t lep;
scif_epd_t recv_ep;
scif_epd_t send_ep;
unsigned int proto;
uint16_t send_port;
uint16_t remote_port = 0;
int err;
if ((lep = scif_open()) < 0) {
mpsslog(PINFO, "Cannot open mpssd monitor SCIF listen port: %s\n", strerror(errno));
pthread_exit((void *)1);
}
if (scif_bind(lep, MPSSD_MONRECV) < 0) {
mpsslog(PINFO, "Cannot bind to mpssd monitor SCIF PORT: %s\n", strerror(errno));
pthread_exit((void *)1);
}
if (scif_listen(lep, 16) < 0) {
mpsslog(PINFO, "Set Listen on mpssd monitor SCIF PORT fail: %s\n", strerror(errno));
pthread_exit((void *)1);
}
while (1) {
if (scif_accept(lep, &recvID, &recv_ep, SCIF_ACCEPT_SYNC)) {
if (errno != EINTR)
mpsslog(PINFO, "Wait for card connect failed: %s\n", strerror(errno));
sleep(1);
continue;
}
if ((mic = mpss_find_micid_inlist(miclist, recvID.node - 1)) == NULL) {
mpsslog(PINFO, "Cannot configure - node %d does not seem to exist\n",
recvID.node - 1);
scif_close(recv_ep);
continue;
}
mpssdi = (struct mpssd_info *)mic->data;
if ((send_ep = scif_open()) < 0) {
fprintf(logfp, "Failed to open SCIF: %s\n", strerror(errno));
scif_close(recv_ep);
pthread_exit((void *)1);
}
mpssdi->send_ep = send_ep;
if ((err = scif_recv(recv_ep, &proto, sizeof(proto), SCIF_RECV_BLOCK)) != sizeof(proto)) {
mpsslog(PINFO, "%s: MIC card mpssd daemon startup connection error %s\n",
mic->name, strerror(errno));
scif_close(recv_ep);
mpssdi->recv_ep = -1;
continue;
}
switch (proto) {
case MONITOR_START:
sendID.node = mic->id + 1;
while ((send_port = scif_connect(send_ep, &sendID)) < 0) {
fprintf(logfp, "Failed to connect to monitor thread on card: %s\n",
strerror(errno));
sleep(1);
}
// Over reliable connection, mpssd tells us which port number it uses
// to talk back to us. If this port matches actual recv_ep remote port
// then we know that recv_ep and send_ep reference the same client.
// We also know that send_ep, references mpssd on mic, as port we
// connect to on that endpoint requires privliges to listen on.
if (scif_recv(send_ep, &remote_port, sizeof(remote_port), SCIF_RECV_BLOCK) < 0) {
mpsslog(PINFO, "%s: MIC card mpssd daemon handshake error %s\n",
mic->name, strerror(errno));
scif_close(send_ep);
scif_close(recv_ep);
continue; // go back to next iteration of while(1), we cannot break the while loop because hosts mpssd can connect with multiple mic cards
}
if (remote_port != recvID.port || sendID.node != recvID.node) {
mpsslog(PINFO, "%s: Failed to authenticate connection with mic mpssd\n",
mic->name);
scif_close(send_ep);
scif_close(recv_ep);
continue; // go back to next iteration of while(1), we cannot break the while loop because hosts mpssd can connect with multiple mic cards
}
// Similarily, provide info for the client, so that he can also verify
// that both connections send_ep & recv_ep belong to us.
if (scif_send(recv_ep, &send_port, sizeof(send_port), SCIF_SEND_BLOCK) < 0) {
mpsslog(PINFO, "%s: MIC card mpssd daemon handshake error %s\n",
mic->name, strerror(errno));
scif_close(send_ep);
scif_close(recv_ep);
continue; // go back to next iteration of while(1), we cannot break the while loop because hosts mpssd can connect with multiple mic cards
}
mpssdi->recv_ep = recv_ep;
pthread_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
pthread_create(&mpssdi->monitor_pth, &attr, monitor, mic);
proto = MONITOR_START_ACK;
scif_send(send_ep, &proto, sizeof(proto), SCIF_RECV_BLOCK);
mpsslog(PINFO, "%s: Monitor connection established\n", mic->name);
break;
}
}
}
int MPID_nem_scif_vc_init(MPIDI_VC_t * vc)
{
int mpi_errno = MPI_SUCCESS;
MPIDI_CH3I_VC *vc_ch = &vc->ch;
MPID_nem_scif_vc_area *vc_scif = VC_SCIF(vc);
int ret;
size_t s;
scifconn_t *sc;
off_t offset;
MPIDI_STATE_DECL(MPID_STATE_MPID_NEM_SCIF_VC_INIT);
MPIDI_FUNC_ENTER(MPID_STATE_MPID_NEM_SCIF_VC_INIT);
vc->sendNoncontig_fn = MPID_nem_scif_SendNoncontig;
vc_ch->iStartContigMsg = MPID_nem_scif_iStartContigMsg;
vc_ch->iSendContig = MPID_nem_scif_iSendContig;
vc_ch->next = NULL;
vc_ch->prev = NULL;
ASSIGN_SC_TO_VC(vc_scif, NULL);
vc_scif->send_queue.head = vc_scif->send_queue.tail = NULL;
vc_scif->sc = sc = &MPID_nem_scif_conns[vc->pg_rank];
vc_scif->terminate = 0;
sc->vc = vc;
/* do the connection */
if (vc->pg_rank < MPID_nem_scif_myrank) {
sc->fd = scif_open();
MPIU_ERR_CHKANDJUMP1(sc->fd == -1, mpi_errno, MPI_ERR_OTHER,
"**scif_open", "**scif_open %s", MPIU_Strerror(errno));
mpi_errno = get_addr(vc, &sc->addr);
if (mpi_errno)
MPIU_ERR_POP(mpi_errno);
ret = scif_connect(sc->fd, &sc->addr);
MPIU_ERR_CHKANDJUMP1(ret == -1, mpi_errno, MPI_ERR_OTHER,
"**scif_connect", "**scif_connect %s", MPIU_Strerror(errno));
}
else {
ret = scif_accept(listen_fd, &sc->addr, &sc->fd, SCIF_ACCEPT_SYNC);
MPIU_ERR_CHKANDJUMP1(ret, mpi_errno, MPI_ERR_OTHER,
"**scif_accept", "**scif_accept %s", MPIU_Strerror(errno));
}
MPIDI_CHANGE_VC_STATE(vc, ACTIVE);
ret = MPID_nem_scif_init_shmsend(&sc->csend, sc->fd, vc->pg_rank);
MPIU_ERR_CHKANDJUMP1(ret, mpi_errno, MPI_ERR_OTHER,
"**scif_init_shmsend", "**scif_init_shmsend %s",
MPIU_Strerror(errno));
/* Exchange offsets */
s = scif_send(sc->fd, &sc->csend.offset, sizeof(off_t), SCIF_SEND_BLOCK);
MPIU_ERR_CHKANDJUMP1(s != sizeof(off_t), mpi_errno, MPI_ERR_OTHER,
"**scif_send", "**scif_send %s", MPIU_Strerror(errno));
s = scif_recv(sc->fd, &offset, sizeof(off_t), SCIF_RECV_BLOCK);
MPIU_ERR_CHKANDJUMP1(s != sizeof(off_t), mpi_errno, MPI_ERR_OTHER,
"**scif_recv", "**scif_recv %s", MPIU_Strerror(errno));
ret = MPID_nem_scif_init_shmrecv(&sc->crecv, sc->fd, offset, vc->pg_rank);
MPIU_ERR_CHKANDJUMP1(ret, mpi_errno, MPI_ERR_OTHER,
"**scif_init_shmrecv", "**scif_init_shmrecv %s",
MPIU_Strerror(errno));
MPIDI_FUNC_EXIT(MPID_STATE_MPID_NEM_SCIF_VC_INIT);
fn_exit:
return mpi_errno;
fn_fail:
goto fn_exit;
}
int main( )
{
size_t len = 536870912;
int align = 4096;
scif_epd_t endpoint;
struct scif_portID portid;
int ret;
uint8_t *in_key = malloc(16 * sizeof(uint8_t));
struct crypto_tfm *tfm
= malloc(
sizeof(struct crypto_tfm) +
sizeof(struct crypto_aes_ctx)
);
struct crypto_aes_ctx *ctx
= crypto_tfm_ctx(tfm);
ctx->key_length = AES_KEYSIZE_256;
crypto_aes_set_key(tfm, in_key, AES_KEYSIZE_256);
endpoint = scif_open( );
if( endpoint == SCIF_OPEN_FAILED )
{
printf("scif open failed\n");
return 1;
}
ret = scif_bind(endpoint, 23955);
if(ret==-1)
{
printf("scif_bind failed");
return 1;
}
portid.node = 0;
portid.port = 23968;
ret = scif_connect(endpoint, &portid);
for( int attempt = 0; ret == -1 && attempt < 10; ++attempt )
{
sleep(1);
ret = scif_connect(endpoint, &portid);
}
if (ret==-1)
{
printf("scif_connect failed\n");
return 1;
}
void *ptr;
ret = posix_memalign((void**)&ptr, align, len);
if (ret)
{
printf("Allocating memory failed\n");
return 1;
}
memset(ptr, 0, len);
if( SCIF_REGISTER_FAILED == scif_register(endpoint, ptr, len, (long)ptr, SCIF_PROT_READ | SCIF_PROT_WRITE, SCIF_MAP_FIXED ) )
{
printf("scif_register of ptr failed due to: %s\n", strerror(errno));
return 1;
}
void *tempbuffer;
ret = posix_memalign((void**)&tempbuffer, align, len);
if (ret)
{
printf("Allocating tempbuffer failed\n");
return 1;
}
if( SCIF_REGISTER_FAILED == scif_register(endpoint, tempbuffer, len, (long)tempbuffer, SCIF_PROT_READ | SCIF_PROT_WRITE, SCIF_MAP_FIXED ) )
{
printf("scif_register of temp failed due to: %s\n", strerror(errno));
return 1;
}
void *outbuffer;
ret = posix_memalign((void**)&outbuffer, align, len);
if (ret)
{
printf("Allocating outbuffer failed %s\n", strerror(errno));
return 1;
}
if( SCIF_REGISTER_FAILED == scif_register(endpoint, outbuffer, len, (long)outbuffer, SCIF_PROT_READ | SCIF_PROT_WRITE, SCIF_MAP_FIXED ) )
{
printf("scif_register of outbuffer failed due to: %s\n", strerror(errno));
return 1;
}
void *remote_ptr;
void *return_ptr;
ret = scif_recv(endpoint, &remote_ptr, sizeof(void*), SCIF_RECV_BLOCK);
if (ret==-1)
{
printf("scif_recv failed due to: %s\n", strerror(errno));
return 1;
}
ret = scif_recv(endpoint, &return_ptr, sizeof(void*), SCIF_RECV_BLOCK);
if (ret==-1)
{
printf("scif_recv failed due to: %s\n", strerror(errno));
return 1;
}
struct timespec start_enc, stop_enc;
clock_gettime(CLOCK_REALTIME, &start_enc);
if (scif_readfrom(endpoint, (long)ptr, len, (long)remote_ptr, SCIF_RMA_SYNC))
{
printf("scif_readfrom failed due to: %s\n", strerror(errno));
return 1;
}
#pragma omp parallel for
for (int k = 0; k<len; k+=16)
{ aes_encrypt(tfm, (uint8_t*)&tempbuffer[k], (uint8_t*)&ptr[k]); }
if (scif_writeto(endpoint, (long)tempbuffer, len, (long)return_ptr, SCIF_RMA_SYNC))
{
printf("scif_writeto failed due to: %s\n", strerror(errno));
return 1;
}
clock_gettime(CLOCK_REALTIME, &stop_enc);
double time_enc = (stop_enc.tv_sec - start_enc.tv_sec) + ( stop_enc.tv_nsec - start_enc.tv_nsec) / NANOSECONDS;
double result0 = len/time_enc/1048576;
printf("%1f,", result0);
struct timespec start_for, stop_for;
clock_gettime(CLOCK_REALTIME, &start_for);
if (scif_readfrom(endpoint, (long)ptr, len, (long)remote_ptr, SCIF_RMA_SYNC))
{
printf("scif_readfrom failed due to: %s\n", strerror(errno));
return 1;
}
#pragma omp parallel for
for (int k=0; k<len; k+=16)
{ aes_decrypt(tfm, (uint8_t*)&outbuffer[k], (uint8_t*)&tempbuffer[k]); }
if (scif_writeto(endpoint, (long)outbuffer, len, (long)return_ptr, SCIF_RMA_SYNC))
{
printf("scif_writeto failed due to: %s\n", strerror(errno));
return 1;
}
clock_gettime(CLOCK_REALTIME, &stop_for);
double time_for = (stop_for.tv_sec - start_for.tv_sec) + ( stop_for.tv_nsec - start_for.tv_nsec) / NANOSECONDS;
double result = 536870912/time_for/1048576;
printf("%1f \n", result);
ret = scif_send(endpoint, &ptr, sizeof(long), SCIF_SEND_BLOCK);
if (ret==-1)
{
printf("scif_send failed due to: %s\n", strerror(errno));
return 1;
}
ret = scif_unregister(endpoint, (off_t)ptr, len );
if(ret==-1 && errno!=ENOTCONN )
{
printf("scif_unregister failed %s\n", strerror(errno));
return 1;
}
scif_close(endpoint);
}
int pm_scif_init(void)
{
int err = 1;
int retry = 0;
FUNCTION_ENTRY;
PM_DB("pm_scif insmoded \n");
#ifdef PM_SCIF_IOCTL
if ((err = spm_dev_init())) {
PM_DB(" spm_dev_init failed\n");
goto done;
}
#endif
atomic_set(&epinuse,0);
pm_scif = kzalloc(sizeof(mic_pm_scif), GFP_KERNEL);
if (!pm_scif) {
err = -ENOMEM;
goto end_con;
}
pm_scif_register(&micpmscif);
if ((pm_scif->ep = scif_open()) == NULL) {
PM_DB(" scif_open failed\n");
goto end_con;
}
if ((pm_scif->lport = scif_bind(pm_scif->ep, 0)) < 0) {
PM_DB(" scif_bind failed\n");
goto end_con;
}
PM_DB(" scif_bind successfull. Local port number = %d, ep = \n",
pm_scif->lport);
dump_ep(pm_scif->ep, __func__,__LINE__);
pm_scif->rport_id.node = 0;
pm_scif->rport_id.port = SCIF_PM_PORT_0;
while ((err = scif_connect(pm_scif->ep, &pm_scif->rport_id)) != 0) {
PM_DB(" scif_connect failed with err = %d ep %p\n",err,
pm_scif->ep);
msleep(1000);
if (retry++ > PM_SCIF_RETRY_COUNT)
goto end_con;
}
pm_scif->pm_recvq = create_singlethread_workqueue("pm_recvq");
INIT_WORK(&pm_scif->pm_recv, pm_recv_from_host);
queue_work(pm_scif->pm_recvq, &pm_scif->pm_recv);
pm_scif->con_state = PM_CONNECTED;
err = 0;
#ifdef PM_SCIF_IOCTL
done:
#endif
return err;
end_con:
pm_scif_exit();
FUNCTION_EXIT;
return err;
}