commit de6e50d00eacf8895c33351572381948aa5079a3
parent 2b041aea4e05b11b47e37a38e8294de36ea8db44
Author: Christian Grothoff <christian@grothoff.org>
Date: Sun, 15 May 2016 10:20:04 +0000
handle EMFILE, ENFILE, ENOBUF on accept
Diffstat:
4 files changed, 50 insertions(+), 9 deletions(-)
diff --git a/ChangeLog b/ChangeLog
@@ -1,3 +1,7 @@
+Sun May 15 12:17:25 CEST 2016
+ Fix handling system or process resource limit exhaustion upon
+ accept(). -CG/CP
+
Thu May 12 08:42:19 CEST 2016
Fix handling of partial writes in MHD_USE_EPOLL_LINUX_ONLY; only
consider sockets returning EAGAIN as unready. -CG/CP
diff --git a/src/include/microhttpd.h b/src/include/microhttpd.h
@@ -130,7 +130,7 @@ typedef intptr_t ssize_t;
* Current version of the library.
* 0x01093001 = 1.9.30-1.
*/
-#define MHD_VERSION 0x00094901
+#define MHD_VERSION 0x00094902
/**
* MHD-internal return code for "YES".
diff --git a/src/microhttpd/daemon.c b/src/microhttpd/daemon.c
@@ -2085,6 +2085,28 @@ MHD_accept_connection (struct MHD_Daemon *daemon)
MHD_PANIC ("close failed\n");
/* just in case */
}
+ if ( (EMFILE == err) ||
+ (ENFILE == err) ||
+ (ENOMEM == err) ||
+ (ENOBUFS == err) )
+ {
+ /* system/process out of resources */
+ if (0 == daemon->connections)
+ {
+ /* Not setting 'at_limit' flag, as there is no way it
+ would ever be cleared. Instead trying to produce
+ bit fat ugly warning. */
+ MHD_DLOG (daemon,
+ "Hit process or system resource limit at FIRST connection. This is really bad as there is no sane way to proceed. Will try busy waiting for system resources to become magically available.\n");
+ }
+ else
+ {
+ daemon->at_limit = MHD_YES;
+ MHD_DLOG (daemon,
+ "Hit process or system resource limit at %u connections, temporarily suspending accept(). Consider setting a lower MHD_OPTION_CONNECTION_LIMIT.\n",
+ (unsigned int) daemon->connections);
+ }
+ }
return MHD_NO;
}
#if !defined(USE_ACCEPT4)
@@ -2142,6 +2164,7 @@ MHD_cleanup_connections (struct MHD_Daemon *daemon)
gnutls_deinit (pos->tls_session);
#endif
daemon->connections--;
+ daemon->at_limit = MHD_NO;
if (NULL != daemon->notify_connection)
daemon->notify_connection (daemon->notify_connection_cls,
pos,
@@ -2420,9 +2443,10 @@ MHD_select (struct MHD_Daemon *daemon,
we do not miss the shutdown, so only do this
optimization if we have a shutdown signaling
pipe. */
- if ( (MHD_INVALID_SOCKET != daemon->socket_fd) &&
- (daemon->connections == daemon->connection_limit) &&
- (0 != (daemon->options & MHD_USE_PIPE_FOR_SHUTDOWN)) )
+ if ( ( (MHD_INVALID_SOCKET != daemon->socket_fd) &&
+ (daemon->connections == daemon->connection_limit) &&
+ (0 != (daemon->options & MHD_USE_PIPE_FOR_SHUTDOWN)) ) ||
+ (MHD_YES == daemon->at_limit) )
FD_CLR (daemon->socket_fd, &rs);
}
else
@@ -2558,7 +2582,8 @@ MHD_poll_all (struct MHD_Daemon *daemon,
poll_server = 0;
poll_listen = -1;
if ( (MHD_INVALID_SOCKET != daemon->socket_fd) &&
- (daemon->connections < daemon->connection_limit) )
+ (daemon->connections < daemon->connection_limit) &&
+ (MHD_NO == daemon->at_limit) )
{
/* only listen if we are not at the connection limit */
p[poll_server].fd = daemon->socket_fd;
@@ -2794,7 +2819,8 @@ MHD_epoll (struct MHD_Daemon *daemon,
return MHD_NO;
if ( (MHD_INVALID_SOCKET != daemon->socket_fd) &&
(daemon->connections < daemon->connection_limit) &&
- (MHD_NO == daemon->listen_socket_in_epoll) )
+ (MHD_NO == daemon->listen_socket_in_epoll) &&
+ (MHD_NO == daemon->at_limit) )
{
event.events = EPOLLIN;
event.data.ptr = daemon;
@@ -2812,8 +2838,9 @@ MHD_epoll (struct MHD_Daemon *daemon,
}
daemon->listen_socket_in_epoll = MHD_YES;
}
- if ( (MHD_YES == daemon->listen_socket_in_epoll) &&
- (daemon->connections == daemon->connection_limit) )
+ if ( ( (MHD_YES == daemon->listen_socket_in_epoll) &&
+ (daemon->connections == daemon->connection_limit) ) ||
+ (MHD_YES == daemon->at_limit) )
{
/* we're at the connection limit, disable listen socket
for event loop for now */
@@ -2912,7 +2939,8 @@ MHD_epoll (struct MHD_Daemon *daemon,
series_length = 0;
while ( (MHD_YES == MHD_accept_connection (daemon)) &&
(daemon->connections < daemon->connection_limit) &&
- (series_length < 128) )
+ (series_length < 128) &&
+ (MHD_NO == daemon->at_limit) )
series_length++;
}
}
diff --git a/src/microhttpd/internal.h b/src/microhttpd/internal.h
@@ -1164,6 +1164,15 @@ struct MHD_Daemon
*/
int shutdown;
+ /**
+ * Did we hit some system or process-wide resource limit while
+ * trying to accept() the last time? If so, we don't accept new
+ * connections until we close an existing one. This effectively
+ * temporarily lowers the "connection_limit" to the current
+ * number of connections.
+ */
+ int at_limit;
+
/*
* Do we need to process resuming connections?
*/
