return read error on non EAGAIN errno

Summary: Tested over weekend with D2571554. There was no EXC_RESOURCE/CPU crash when read errrors were returned on ENOTCONN. EAGAIN seems innocent as the spin detector was disabled on testing group.

Patch folly. I can add #ifdef __APPLE__ around the errno checking code but this should be good practice for non apple code as well.

I will remove debugging code in fbios master in another diff.

Reviewed By: djwatson

Differential Revision: D2580819

fb-gh-sync-id: 9162a3deba01af8b07cd2b336d7da3da040c67a9

diff --git a/folly/io/async/AsyncSSLSocket.cpp b/folly/io/async/AsyncSSLSocket.cpp
index 9ca8e8e51518ceb007780deb5047cae77fb11120..e34df5641d6510bf7ccf32f56b01eb8d128a2dc1 100644 (file)
--- a/folly/io/async/AsyncSSLSocket.cpp
+++ b/folly/io/async/AsyncSSLSocket.cpp
@@ -1164,7 +1164,11 @@ AsyncSSLSocket::performRead(void** buf, size_t* buflen, size_t* offset) {
     int error = SSL_get_error(ssl_, bytes);
     if (error == SSL_ERROR_WANT_READ) {
       // The caller will register for read event if not already.
-      return READ_BLOCKING;
+      if (errno == EWOULDBLOCK || errno == EAGAIN) {
+        return READ_BLOCKING;
+      } else {
+        return READ_ERROR;
+      }
     } else if (error == SSL_ERROR_WANT_WRITE) {
       // TODO: Even though we are attempting to read data, SSL_read() may
       // need to write data if renegotiation is being performed.  We currently