RPC Binder: Retry Trusty connections several times
Sometimes the restarting of the Trusty service races
with tipc_connect, causing the latter to return -ENOTCONN
since the service has not come up yet. This puts the
connection attempts in a loop and retry them up to 5 times,
sleeping for a short duration between attempts.
Should make binderRpcToTrustyTest a lot less flaky.
Bug: 230135749
Test: run binderRpcToTrustyTest several times
Change-Id: I2c4e8dfe929535dbb77b5af3c677e5d27f00f50e
diff --git a/libs/binder/tests/binderRpcTest.cpp b/libs/binder/tests/binderRpcTest.cpp
index 8d1def1..504b3ce 100644
--- a/libs/binder/tests/binderRpcTest.cpp
+++ b/libs/binder/tests/binderRpcTest.cpp
@@ -382,9 +382,16 @@
status = session->setupPreconnectedClient({}, [=]() {
#ifdef BINDER_RPC_TO_TRUSTY_TEST
auto port = trustyIpcPort(serverVersion);
- int tipcFd = tipc_connect(kTrustyIpcDevice, port.c_str());
- return tipcFd >= 0 ? android::base::unique_fd(tipcFd)
- : android::base::unique_fd();
+ for (size_t i = 0; i < 5; i++) {
+ // Try to connect several times,
+ // in case the service is slow to start
+ int tipcFd = tipc_connect(kTrustyIpcDevice, port.c_str());
+ if (tipcFd >= 0) {
+ return android::base::unique_fd(tipcFd);
+ }
+ usleep(50000);
+ }
+ return android::base::unique_fd();
#else
LOG_ALWAYS_FATAL("Tried to connect to Trusty outside of vendor");
return android::base::unique_fd();