AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add SSL mutual authentication support for rdci

Browse Source 
The RDC API is changed to pass the certificates to the gRPC.

Add the support to add all GPUs in the host to a group. Also before
add a GPU to a group, the RDC API will verify that GPU exists or not.

Add the support to fetch the temperature metrics.

Change-Id: I5857ef03fede233d16e8b2836be120f33172da93
This commit is contained in:
Bill(Shuzhou) Liu
2020-03-10 14:02:05 -04:00
committed by Chris Freehill
parent 023de40df7
commit 66e4e790c3
20 changed files with 232 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rdci/src/RdciDiscoverySubSystem.cc
+10 -11
View File
@@ -23,14 +23,14 @@ THE SOFTWARE.
#include <unistd.h>
#include "rdc_lib/rdc_common.h"
#include "rdc/rdc.h"
#include "RdcException.h"
#include "rdc_lib/RdcException.h"
#include "RdciDiscoverySubSystem.h"
namespace amd {
namespace rdc {
RdciDiscoverySubSystem::RdciDiscoverySubSystem() :show_help_(false) {
RdciDiscoverySubSystem::RdciDiscoverySubSystem() : show_help_(false) {
}
void RdciDiscoverySubSystem::parse_cmd_opts(int argc, char ** argv) {
@@ -38,13 +38,14 @@ void RdciDiscoverySubSystem::parse_cmd_opts(int argc, char ** argv) {
const struct option long_options[] = {
{"host", required_argument, nullptr, HOST_OPTIONS },
{"help", optional_argument, nullptr, 'h' },
{"unauth", optional_argument, nullptr, 'u' },
{ nullptr, 0 , nullptr, 0 }
};
int option_index = 0;
int opt = 0;
while ((opt = getopt_long(argc, argv, "h",
while ((opt = getopt_long(argc, argv, "hu",
long_options, &option_index)) != -1) {
switch (opt) {
case HOST_OPTIONS:
@@ -53,27 +54,25 @@ void RdciDiscoverySubSystem::parse_cmd_opts(int argc, char ** argv) {
case 'h':
show_help_ = true;
return;
case 'u':
use_auth_ = false;
break;
default:
show_help();
throw RdcException(RDC_ST_BAD_PARAMETER,
"Unknown command line options");
}
}
}
void RdciDiscoverySubSystem::show_help() const {
std::cout << " discovery -- Used to discover and identify GPUs "
<< "and their attributes.\n\n";
std::cout << "Usage\n";
std::cout << " rdci discovery [--host <IP/FQDN>:port]\n";
std::cout << " rdci discovery [--host <IP/FQDN>:port] [-u]\n";
std::cout << "\nFlags:\n";
std::cout << " --host <IP/FQDN>:port Connects to "
<< "specified IP or fully-qualified domain name.\n";
std::cout << " The port "
<< "must be specified.\n";
std::cout << " Default: localhost:50051\n";
std::cout << " -h --help Displays usage "
<< "information and exits.\n";
show_common_usage();
}
rdci/src/RdciSubSystem.cc
+76 -5
View File
@@ -20,14 +20,19 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
*/
#include "RdciSubSystem.h"
#include "RdcException.h"
#include "rdc_lib/RdcException.h"
#include "common/rdc_utils.h"
namespace amd {
namespace rdc {
RdciSubSystem::RdciSubSystem():
rdc_handle_(nullptr)
, ip_port_("localhost:50051") { // default host
, ip_port_("localhost:50051") // default host
, use_auth_(true)
, root_ca_("/etc/rdc/client/certs/rdc_cacert.pem")
, client_cert_("/etc/rdc/client/certs/rdc_client_cert.pem")
, client_key_("/etc/rdc/client/private/rdc_client_cert.key") {
rdc_status_t status = rdc_init(0);
if (status != RDC_ST_OK) {
throw RdcException(status, "RDC initialize fail");
@@ -35,12 +40,79 @@ RdciSubSystem::RdciSubSystem():
}
void RdciSubSystem::connect() {
rdc_status_t status = rdc_connect(ip_port_.c_str(), &rdc_handle_);
rdc_status_t status;
if (use_auth_) {
std::string ca_pem;
std::string client_cert_pem;
std::string client_key_pem;
if (!FileExists(root_ca_.c_str())) {
std::cout << "In order to use the SSL mutual authentication, the "
<< "root CA must be copied to " << root_ca_ << std::endl;
throw RdcException(RDC_ST_BAD_PARAMETER, "root CA not found");
}
int ret = ReadFile(root_ca_, &ca_pem);
if (ret) {
throw RdcException(RDC_ST_BAD_PARAMETER,
std::string("Fail to read root CA at") + root_ca_);
}
if (!FileExists(client_cert_.c_str())) {
std::cout << "In order to use the SSL mutual authentication, the "
<< "client certificate must be copied to "
<< client_cert_ << std::endl;
throw RdcException(RDC_ST_BAD_PARAMETER,
"client cert not found");
}
ret = ReadFile(client_cert_, &client_cert_pem);
if (ret) {
throw RdcException(RDC_ST_BAD_PARAMETER,
std::string("Fail to read client certificate at") + client_cert_);
}
if (!FileExists(client_key_.c_str())) {
std::cout << "In order to use the SSL mutual authentication, the "
<< "client private key must be copied to "
<< client_key_ << std::endl;
throw RdcException(RDC_ST_BAD_PARAMETER,
"client key not found");
}
ret = ReadFile(client_key_, &client_key_pem);
if (ret) {
throw RdcException(RDC_ST_BAD_PARAMETER,
std::string("Fail to read client key at ") + client_key_);
}
status = rdc_connect(ip_port_.c_str(), &rdc_handle_,
ca_pem.c_str(), client_cert_pem.c_str(), client_key_pem.c_str());
} else { // Not use the SSL mutual authentication
status = rdc_connect(ip_port_.c_str(), &rdc_handle_,
nullptr, nullptr, nullptr);
}
if (status != RDC_ST_OK) {
throw RdcException(status, "Fail to setup the connection");
throw RdcException(status,
"Fail to setup the connection. Please check all libraries in right folder");
}
}
void RdciSubSystem::show_common_usage() const {
std::cout << " --host <IP/FQDN>:port Connects to "
<< "specified IP or fully-qualified domain name.\n";
std::cout << " The port "
<< "must be specified.\n";
std::cout << " Default: localhost:50051\n";
std::cout << " -u --unauth Do not use the SSL mutual"
<< " authentication to encrypt the communication\n"
<< " Default: SSL mutual will be"
<< " used. You must copy the root CA to "
<< root_ca_ << "\n"
<< " Client certificate to "
<< client_cert_ << "\n"
<< " Client key to "
<< client_key_ << "\n";
std::cout << " -h --help Displays usage "
<< "information and exits.\n";
}
RdciSubSystem::~RdciSubSystem() {
if (rdc_handle_) {
@@ -51,6 +123,5 @@ RdciSubSystem::~RdciSubSystem() {
rdc_shutdown();
}
} // namespace rdc
} // namespace amd
rdci/src/rdci.cc
+1 -1
View File
@@ -24,7 +24,7 @@ THE SOFTWARE.
#include <string>
#include "rdc_lib/rdc_common.h"
#include "rdc/rdc.h"
#include "RdcException.h"
#include "rdc_lib/RdcException.h"
#include "RdciDiscoverySubSystem.h"