Steinar H. Gunderson fixed: Correctly clear sockets from the fd_set on in

several functions (write_tcp_data, read_tcp_data, read_udp_packets) so that
if it fails and the socket is closed the following code doesn't try to use
the file descriptor.
This commit is contained in:
Daniel Stenberg 2007-09-22 21:23:10 +00:00
parent 0f89a2e639
commit 0f4664d27f
2 changed files with 26 additions and 0 deletions

View File

@ -2,6 +2,11 @@
* September 22 2007 (Daniel Stenberg)
- Steinar H. Gunderson fixed: Correctly clear sockets from the fd_set on in
several functions (write_tcp_data, read_tcp_data, read_udp_packets) so that
if it fails and the socket is closed the following code doesn't try to use
the file descriptor.
- Steinar H. Gunderson modified c-ares to now also do to DNS retries even when
TCP is used since there are several edge cases where it still makes sense.

View File

@ -167,6 +167,13 @@ static void write_tcp_data(ares_channel channel,
continue;
}
/* If there's an error and we close this socket, then open
* another with the same fd to talk to another server, then we
* don't want to think that it was the new socket that was
* ready. This is not disastrous, but is likely to result in
* extra system calls and confusion. */
FD_CLR(server->tcp_socket, write_fds);
/* Count the number of send queue items. */
n = 0;
for (sendreq = server->qhead; sendreq; sendreq = sendreq->next)
@ -280,6 +287,13 @@ static void read_tcp_data(ares_channel channel, fd_set *read_fds,
continue;
}
/* If there's an error and we close this socket, then open
* another with the same fd to talk to another server, then we
* don't want to think that it was the new socket that was
* ready. This is not disastrous, but is likely to result in
* extra system calls and confusion. */
FD_CLR(server->tcp_socket, read_fds);
if (server->tcp_lenbuf_pos != 2)
{
/* We haven't yet read a length word, so read that (or
@ -370,6 +384,13 @@ static void read_udp_packets(ares_channel channel, fd_set *read_fds,
continue;
}
/* If there's an error and we close this socket, then open
* another with the same fd to talk to another server, then we
* don't want to think that it was the new socket that was
* ready. This is not disastrous, but is likely to result in
* extra system calls and confusion. */
FD_CLR(server->udp_socket, read_fds);
count = sread(server->udp_socket, buf, sizeof(buf));
if (count == -1 && try_again(SOCKERRNO))
continue;