Added cgid capture to stirling with testing

src/shared/metadata/cgroup_path_resolver.cc

@@ -24,7 +24,10 @@
 
 #include <absl/strings/str_replace.h>
 
+#include <fstream>
 #include "src/common/fs/fs_wrapper.h"
+#include "src/common/system/config.h"
+#include "src/common/system/proc_pid_path.h"
 #include "src/shared/metadata/cgroup_path_resolver.h"
 
 DEFINE_bool(force_cgroup2_mode, true, "Flag to force assume cgroup2 fs for testing purposes");
@@ -76,6 +79,35 @@ StatusOr<std::string> FindSelfCGroupProcs(std::string_view base_path) {
   return error::NotFound("Could not find self as a template.");
 }
 
+StatusOr<uint64_t> FindCgroupIDFromPID(pid_t pid) {
+  auto cgroup_name = px::system::ProcPidPath(pid, "cgroup");
+  std::ifstream proc_cgroup(cgroup_name);
+
+  // get rid of the id at the front
+  uint64_t blank;
+  proc_cgroup >> blank;
+  std::string line;
+  std::getline(proc_cgroup, line);
+
+  if (proc_cgroup.fail() || proc_cgroup.bad()) {
+    std::string failure_string =
+        absl::StrCat("procfs cgroup file getline failed with", "fail: ", proc_cgroup.fail(),
+                     "bad : ", proc_cgroup.bad(), "eof: ", proc_cgroup.eof());
+    return error::NotFound(failure_string);
+  }
+
+  std::string cgroup_path_name =
+      absl::StrCat(px::system::Config::GetInstance().sysfs_path().string(), "/cgroup/",
+                   line.substr(2, line.size()));
+
+  // lstat what we are pointing to
+  struct stat stat_buf;
+  if (stat(cgroup_path_name.c_str(), &stat_buf)) {
+    return error::NotFound(absl::StrCat(cgroup_path_name, " sysfs cgroup stat failed."));
+  }
+  return stat_buf.st_ino;
+}
+
 StatusOr<CGroupTemplateSpec> CreateCGroupTemplateSpecFromPath(std::string_view path) {
   // Pattern match for a pod ID.
   // Examples:

src/shared/metadata/cgroup_path_resolver.h

@@ -90,6 +90,11 @@ StatusOr<std::string> CGroupBasePath(std::string_view sysfs_path);
  */
 StatusOr<std::string> FindSelfCGroupProcs(std::string_view base_path);
 
+/**
+ * Finds the cgroup id corresponding to the pid
+ **/
+StatusOr<uint64_t> FindCgroupIDFromPID(pid_t pid);
+
 /**
  * Given a path to a sample cgroup.procs file for a K8s pod,
  * this function produces a templated spec from which paths for other pods can be generated.

src/stirling/core/canonical_types.h

@@ -40,11 +40,19 @@ constexpr DataElement kTime = {
 constexpr std::string_view kUPIDColName = "upid";
 constexpr DataElement kUPID = {
     kUPIDColName,
-    "An opaque numeric ID that globally identify a running process inside the cluster.",
+    "An opaque numeric ID that globally identifies a running process inside the cluster.",
     types::DataType::UINT128,
     types::SemanticType::ST_UPID,
     types::PatternType::GENERAL};
 
+constexpr std::string_view kCGIDColName = "cgid";
+constexpr DataElement kCGID = {
+    kCGIDColName,
+    "The cgroup id that the process was in when this probe was triggered.",
+    types::DataType::INT64,
+    types::SemanticType::ST_NONE,
+    types::PatternType::GENERAL};
+
 // clang-format on
 
 }  // namespace canonical_data_elements

src/stirling/source_connectors/socket_tracer/amqp_table.h

@@ -31,6 +31,7 @@ namespace stirling {
 static constexpr DataElement kAMQPElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -84,6 +85,7 @@ DEFINE_PRINT_TABLE(AMQP)
 
 constexpr int kAMQPTimeIdx = kAMQPTable.ColIndex("time_");
 constexpr int kAMQPUUIDIdx = kAMQPTable.ColIndex("upid");
+constexpr int kAMQPCGIDIdx = kAMQPTable.ColIndex("cgid");
 constexpr int kAMQPFrameTypeIdx = kAMQPTable.ColIndex("frame_type");
 constexpr int kAMQPChannelIdx = kAMQPTable.ColIndex("channel");
 constexpr int kAMQPLatencyIdx = kAMQPTable.ColIndex("latency");

src/stirling/source_connectors/socket_tracer/bcc_bpf/socket_trace.c

@@ -136,6 +136,8 @@ static __inline void init_conn_id(uint32_t tgid, int32_t fd, struct conn_id_t* c
   conn_id->upid.start_time_ticks = get_tgid_start_time();
   conn_id->fd = fd;
   conn_id->tsid = bpf_ktime_get_ns();
+  // Until we have proper cgids tracing
+  conn_id->cgid = bpf_get_current_cgroup_id();
 }
 
 static __inline void init_conn_info(uint32_t tgid, int32_t fd, struct conn_info_t* conn_info) {

src/stirling/source_connectors/socket_tracer/bcc_bpf_intf/common.h

@@ -79,6 +79,8 @@ struct conn_id_t {
   int32_t fd;
   // Unique id of the conn_id (timestamp).
   uint64_t tsid;
+  // cgid of the connection
+  uint64_t cgid;
 };
 
 // Specifies the corresponding indexes of the entries of a per-cpu array.

src/stirling/source_connectors/socket_tracer/cass_table.h

@@ -37,6 +37,7 @@ static const std::map<int64_t, std::string_view> kCQLRespOpDecoder =
 static constexpr DataElement kCQLElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -71,6 +72,7 @@ DEFINE_PRINT_TABLE(CQL)
 
 static constexpr int kCQLTraceRoleIdx = kCQLTable.ColIndex("trace_role");
 static constexpr int kCQLUPIDIdx = kCQLTable.ColIndex("upid");
+static constexpr int kCQLCGIDIdx = kCQLTable.ColIndex("cgid");
 static constexpr int kCQLReqOp = kCQLTable.ColIndex("req_op");
 static constexpr int kCQLReqBody = kCQLTable.ColIndex("req_body");
 static constexpr int kCQLRespOp = kCQLTable.ColIndex("resp_op");

src/stirling/source_connectors/socket_tracer/conn_stats.cc

@@ -76,6 +76,7 @@ absl::flat_hash_map<ConnStats::AggKey, ConnStats::Stats>& ConnStats::UpdateStats
     stats.conn_close += conn_close;
     stats.bytes_recv += bytes_recv;
     stats.bytes_sent += bytes_sent;
+    stats.cgid_curr = tracker->conn_id().cgid;
 
     stats.last_update = update_counter_;
   }

src/stirling/source_connectors/socket_tracer/conn_stats.h

@@ -75,6 +75,7 @@ class ConnStats {
     uint64_t conn_close = 0;
     uint64_t bytes_sent = 0;
     uint64_t bytes_recv = 0;
+    uint64_t cgid_curr = 0;
 
     // Keep track of whether this stats object has ever been previously reported.
     // Used to determine whether it should be reported in the future, after the connection

src/stirling/source_connectors/socket_tracer/conn_stats_table.h

@@ -29,6 +29,7 @@ namespace stirling {
 constexpr DataElement kConnStatsElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -68,6 +69,7 @@ namespace conn_stats_idx {
 
 constexpr int kTime = kConnStatsTable.ColIndex("time_");
 constexpr int kUPID = kConnStatsTable.ColIndex("upid");
+constexpr int kCGID = kConnStatsTable.ColIndex("cgid");
 constexpr int kRemoteAddr = kConnStatsTable.ColIndex("remote_addr");
 constexpr int kRemotePort = kConnStatsTable.ColIndex("remote_port");
 constexpr int kAddrFamily = kConnStatsTable.ColIndex("addr_family");

src/stirling/source_connectors/socket_tracer/conn_stats_test.cc

@@ -89,6 +89,7 @@ TEST_F(ConnStatsTest, Basic) {
       .upid = {.pid = 12345, .start_time_ticks = 1000},
       .fd = 3,
       .tsid = 111110,
+      .cgid = UINT64_MAX,
   };
 
   // The basic conn_stats_event template.
@@ -165,6 +166,7 @@ TEST_F(ConnStatsTest, ServerSide) {
       .upid = {.pid = 12345, .start_time_ticks = 1000},
       .fd = 3,
       .tsid = 10000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn0_stats_event;
@@ -193,6 +195,7 @@ TEST_F(ConnStatsTest, ServerSide) {
       .upid = {.pid = 12345, .start_time_ticks = 1000},
       .fd = 4,
       .tsid = 20000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn1_stats_event;
@@ -221,6 +224,7 @@ TEST_F(ConnStatsTest, ServerSide) {
       .upid = {.pid = 12345, .start_time_ticks = 1000},
       .fd = 5,
       .tsid = 30000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn2_stats_event;
@@ -257,6 +261,7 @@ TEST_F(ConnStatsTest, ClientSide) {
       .upid = {.pid = 11111, .start_time_ticks = 1000},
       .fd = 3,
       .tsid = 10000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn0_stats_event;
@@ -285,6 +290,7 @@ TEST_F(ConnStatsTest, ClientSide) {
       .upid = {.pid = 11111, .start_time_ticks = 1000},
       .fd = 4,
       .tsid = 20000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn1_stats_event;
@@ -313,6 +319,7 @@ TEST_F(ConnStatsTest, ClientSide) {
       .upid = {.pid = 11111, .start_time_ticks = 1000},
       .fd = 5,
       .tsid = 30000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn2_stats_event;
@@ -344,6 +351,7 @@ TEST_F(ConnStatsTest, NoEventsIfNoRemoteAddr) {
       .upid = {.pid = 11111, .start_time_ticks = 1000},
       .fd = 3,
       .tsid = 10000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn_stats_event;
@@ -371,6 +379,7 @@ TEST_F(ConnStatsTest, DisabledConnTracker) {
       .upid = {.pid = 11111, .start_time_ticks = 1000},
       .fd = 3,
       .tsid = 10000,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn_stats_event;

src/stirling/source_connectors/socket_tracer/conn_tracker_http2_test.cc

@@ -47,7 +47,7 @@ class ConnTrackerHTTP2Test : public ::testing::Test {
   }
 
   const conn_id_t kConnID = {
-      .upid = {{.pid = 123}, .start_time_ticks = 11000000}, .fd = 3, .tsid = 21};
+      .upid = {{.pid = 123}, .start_time_ticks = 11000000}, .fd = 3, .tsid = 21, .cgid = 33};
 
   ConnTracker tracker_;
   testing::MockClock mock_clock_;
@@ -193,7 +193,7 @@ TEST_F(ConnTrackerHTTP2Test, MixedHeadersAndData) {
 TEST_F(ConnTrackerHTTP2Test, ZeroFD) {
   ConnTracker tracker;
   const conn_id_t kConnID = {
-      .upid = {{.pid = 123}, .start_time_ticks = 11000000}, .fd = 0, .tsid = 21};
+      .upid = {{.pid = 123}, .start_time_ticks = 11000000}, .fd = 0, .tsid = 21, .cgid = 33};
   tracker.SetConnID(kConnID);
 
   const int kStreamID = 7;

src/stirling/source_connectors/socket_tracer/conn_tracker_test.cc

@@ -769,6 +769,7 @@ TEST_F(ConnTrackerTest, ConnStats) {
       .upid = {.pid = 12345, .start_time_ticks = 1000},
       .fd = 3,
       .tsid = 111110,
+      .cgid = UINT64_MAX,
   };
 
   struct conn_stats_event_t conn_stats_event;

src/stirling/source_connectors/socket_tracer/conn_trackers_manager_test.cc

@@ -84,7 +84,7 @@ TEST_F(ConnTrackersManagerTest, Fuzz) {
       int32_t fd = 1;
       uint64_t tsid = tsid_dist(rng_);
 
-      struct conn_id_t conn_id = {{{pid}, 0}, fd, tsid};
+      struct conn_id_t conn_id = {{{pid}, 0}, fd, tsid, UINT64_MAX};
       TrackerEvent(conn_id, protocol.value());
     } else if (x < 0.95) {
       int death_countdown = death_countdown_dist(rng_);

src/stirling/source_connectors/socket_tracer/dns_table.h

@@ -32,6 +32,7 @@ namespace stirling {
 static constexpr DataElement kDNSElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -63,6 +64,7 @@ static constexpr auto kDNSTable =
 DEFINE_PRINT_TABLE(DNS)
 
 static constexpr int kDNSUPIDIdx = kDNSTable.ColIndex("upid");
+static constexpr int kDNSCGIDIdx = kDNSTable.ColIndex("cgid");
 static constexpr int kDNSReqHdrIdx = kDNSTable.ColIndex("req_header");
 static constexpr int kDNSReqBodyIdx = kDNSTable.ColIndex("req_body");
 static constexpr int kDNSRespHdrIdx = kDNSTable.ColIndex("resp_header");

src/stirling/source_connectors/socket_tracer/http2_messages_table.h

@@ -27,6 +27,7 @@ namespace stirling {
 constexpr DataElement kHTTP2MessagesElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -58,6 +59,7 @@ DEFINE_PRINT_TABLE(HTTP2Messages)
 
 constexpr int kHTTP2MessagesTimeIdx = kHTTP2MessagesTable.ColIndex("time_");
 constexpr int kHTTP2MessagesUPIDIdx = kHTTP2MessagesTable.ColIndex("upid");
+constexpr int kHTTP2MessagesCGIDIdx = kHTTP2MessagesTable.ColIndex("cgid");
 constexpr int kHTTP2MessagesRemoteAddrIdx = kHTTP2MessagesTable.ColIndex("remote_addr");
 constexpr int kHTTP2MessagesRemotePortIdx = kHTTP2MessagesTable.ColIndex("remote_port");
 constexpr int kHTTP2MessagesTraceRoleIdx = kHTTP2MessagesTable.ColIndex("trace_role");

src/stirling/source_connectors/socket_tracer/http_table.h

@@ -41,6 +41,7 @@ static const std::map<int64_t, std::string_view> kHTTPContentTypeDecoder =
 constexpr DataElement kHTTPElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -110,6 +111,7 @@ DEFINE_PRINT_TABLE(HTTP)
 
 constexpr int kHTTPTimeIdx = kHTTPTable.ColIndex("time_");
 constexpr int kHTTPUPIDIdx = kHTTPTable.ColIndex("upid");
+constexpr int kHTTPCGIDIdx = kHTTPTable.ColIndex("cgid");
 constexpr int kHTTPRemoteAddrIdx = kHTTPTable.ColIndex("remote_addr");
 constexpr int kHTTPRemotePortIdx = kHTTPTable.ColIndex("remote_port");
 constexpr int kHTTPTraceRoleIdx = kHTTPTable.ColIndex("trace_role");

src/stirling/source_connectors/socket_tracer/kafka_table.h

@@ -35,6 +35,7 @@ static const std::map<int64_t, std::string_view> kKafkaAPIKeyDecoder =
 static constexpr DataElement kKafkaElements[] = {
       canonical_data_elements::kTime,
       canonical_data_elements::kUPID,
+      canonical_data_elements::kCGID,
       canonical_data_elements::kRemoteAddr,
       canonical_data_elements::kRemotePort,
       canonical_data_elements::kTraceRole,
@@ -68,6 +69,7 @@ DEFINE_PRINT_TABLE(Kafka)
 
 constexpr int kKafkaTimeIdx = kKafkaTable.ColIndex("time_");
 constexpr int kKafkaUPIDIdx = kKafkaTable.ColIndex("upid");
+constexpr int kKafkaCGIDIdx = kKafkaTable.ColIndex("cgid");
 constexpr int kKafkaReqCmdIdx = kKafkaTable.ColIndex("req_cmd");
 constexpr int kKafkaClientIDIdx = kKafkaTable.ColIndex("client_id");
 constexpr int kKafkaReqBodyIdx = kKafkaTable.ColIndex("req_body");

src/stirling/source_connectors/socket_tracer/mux_table.h

@@ -32,6 +32,7 @@ namespace stirling {
 static constexpr DataElement kMuxElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -56,6 +57,7 @@ static constexpr auto kMuxTable =
 DEFINE_PRINT_TABLE(Mux)
 
 static constexpr int kMuxUPIDIdx = kMuxTable.ColIndex("upid");
+static constexpr int kMuxCGIDIdx = kMuxTable.ColIndex("cgid");
 static constexpr int kMuxReqTypeIdx = kMuxTable.ColIndex("req_type");
 
 }  // namespace stirling

src/stirling/source_connectors/socket_tracer/mysql_table.h

@@ -37,6 +37,7 @@ static const std::map<int64_t, std::string_view> kMySQLRespStatusDecoder =
 static constexpr DataElement kMySQLElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -71,6 +72,7 @@ DEFINE_PRINT_TABLE(MySQL)
 
 constexpr int kMySQLTimeIdx = kMySQLTable.ColIndex("time_");
 constexpr int kMySQLUPIDIdx = kMySQLTable.ColIndex("upid");
+constexpr int kMySQLCGIDIdx = kMySQLTable.ColIndex("cgid");
 constexpr int kMySQLReqCmdIdx = kMySQLTable.ColIndex("req_cmd");
 constexpr int kMySQLReqBodyIdx = kMySQLTable.ColIndex("req_body");
 constexpr int kMySQLRespStatusIdx = kMySQLTable.ColIndex("resp_status");

src/stirling/source_connectors/socket_tracer/nats_table.h

@@ -29,6 +29,7 @@ namespace stirling {
 constexpr DataElement kNATSElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -53,6 +54,7 @@ namespace nats_idx {
 
 constexpr int kTime = kNATSTable.ColIndex("time_");
 constexpr int kUPID = kNATSTable.ColIndex("upid");
+constexpr int kCGID = kNATSTable.ColIndex("cgid");
 constexpr int kRemoteAddr = kNATSTable.ColIndex("remote_addr");
 constexpr int kRemotePort = kNATSTable.ColIndex("remote_port");
 constexpr int kRemoteRole = kNATSTable.ColIndex("trace_role");

src/stirling/source_connectors/socket_tracer/pgsql_table.h

@@ -29,6 +29,7 @@ namespace stirling {
 static constexpr DataElement kPGSQLElements[] = {
         canonical_data_elements::kTime,
         canonical_data_elements::kUPID,
+        canonical_data_elements::kCGID,
         canonical_data_elements::kRemoteAddr,
         canonical_data_elements::kRemotePort,
         canonical_data_elements::kTraceRole,
@@ -56,6 +57,7 @@ static constexpr auto kPGSQLTable = DataTableSchema(
 DEFINE_PRINT_TABLE(PGSQL)
 
 constexpr int kPGSQLUPIDIdx = kPGSQLTable.ColIndex("upid");
+constexpr int kPGSQLCGIDIdx = kPGSQLTable.ColIndex("cgid");
 constexpr int kPGSQLReqIdx = kPGSQLTable.ColIndex("req");
 constexpr int kPGSQLRespIdx = kPGSQLTable.ColIndex("resp");
 constexpr int kPGSQLReqCmdIdx = kPGSQLTable.ColIndex("req_cmd");