libbinder: RPC use DEAD_OBJECT over -ECANCELED
When we shutdown a connection with FdTrigger, we return -ECANCELED.
However, the next call to that FD will return with -EPIPE
(DEAD_OBJECT). It's not really important: if your call happens to get
interrupted, you'll get -ECANCELED or if your call happened to be
scheduled slightly later, you'd get DEAD_OBJECT. Actually, we don't care
to distinguish between these two cases because existing clients of
libbinder only have and consider DEAD_OBJECT now. Whether the server is
shutting down right now, or you make the call a tiny bit later, we want
to return the exact same error code in order to be consistent w/ the
behavior of libbinder when it uses the binder driver.
This was causing a flake in a later CL (fixing the 'Die' race), when
rpcSend failing causes the server to shutdown.
Bug: 200167417
Test: binderRpcTest 'Callbacks' case repeated 100s of times
Change-Id: Id2977f05eb249691326955e6f2424d4e5e08b417
diff --git a/libs/binder/FdTrigger.cpp b/libs/binder/FdTrigger.cpp
index ecf13dc..49f83ff 100644
--- a/libs/binder/FdTrigger.cpp
+++ b/libs/binder/FdTrigger.cpp
@@ -53,7 +53,7 @@
continue;
}
if (pfd[1].revents & POLLHUP) {
- return -ECANCELED;
+ return DEAD_OBJECT;
}
return pfd[0].revents & event ? OK : DEAD_OBJECT;
}
diff --git a/libs/binder/RpcTransportTls.cpp b/libs/binder/RpcTransportTls.cpp
index 23088ad..79445d9 100644
--- a/libs/binder/RpcTransportTls.cpp
+++ b/libs/binder/RpcTransportTls.cpp
@@ -241,7 +241,7 @@
status_t handlePoll(int event, android::base::borrowed_fd fd, FdTrigger* fdTrigger,
const char* fnString) {
status_t ret = fdTrigger->triggerablePoll(fd, event);
- if (ret != OK && ret != DEAD_OBJECT && ret != -ECANCELED) {
+ if (ret != OK && ret != DEAD_OBJECT) {
ALOGE("triggerablePoll error while poll()-ing after %s(): %s", fnString,
statusToString(ret).c_str());
}
@@ -348,7 +348,7 @@
// Before doing any I/O, check trigger once. This ensures the trigger is checked at least
// once. The trigger is also checked via triggerablePoll() after every SSL_write().
- if (fdTrigger->isTriggered()) return -ECANCELED;
+ if (fdTrigger->isTriggered()) return DEAD_OBJECT;
while (buffer < end) {
size_t todo = std::min<size_t>(end - buffer, std::numeric_limits<int>::max());
@@ -379,7 +379,7 @@
// Before doing any I/O, check trigger once. This ensures the trigger is checked at least
// once. The trigger is also checked via triggerablePoll() after every SSL_write().
- if (fdTrigger->isTriggered()) return -ECANCELED;
+ if (fdTrigger->isTriggered()) return DEAD_OBJECT;
while (buffer < end) {
size_t todo = std::min<size_t>(end - buffer, std::numeric_limits<int>::max());
diff --git a/libs/binder/tests/binderRpcTest.cpp b/libs/binder/tests/binderRpcTest.cpp
index 6bcf102..cc1d2fa 100644
--- a/libs/binder/tests/binderRpcTest.cpp
+++ b/libs/binder/tests/binderRpcTest.cpp
@@ -1794,9 +1794,9 @@
}
status = serverTransport->interruptableWriteFully(fdTrigger, msg2.data(), msg2.size());
- if (status != -ECANCELED)
+ if (status != DEAD_OBJECT)
return AssertionFailure() << "When FdTrigger is shut down, interruptableWriteFully "
- "should return -ECANCELLED, but it is "
+ "should return DEAD_OBJECT, but it is "
<< statusToString(status);
return AssertionSuccess();
};
@@ -1830,7 +1830,7 @@
}
writeCv.notify_all();
// After this line, server thread unblocks and attempts to write the second message, but
- // shutdown is triggered, so write should failed with -ECANCELLED. See |serverPostConnect|.
+ // shutdown is triggered, so write should failed with DEAD_OBJECT. See |serverPostConnect|.
// On the client side, second read fails with DEAD_OBJECT
ASSERT_FALSE(client.readMessage(msg2));
}