# connect.c and errno

3 messages from 2006-06-28 to 2006-07-02. Participants: Morten Welinder, Petr Baudis, Junio C Hamano.
Thread: https://gitlist.dev/t/4697

## Morten Welinder, 2006-06-28 16:56

Subject: connect.c and errno
Message-ID: <118833cc0606280956s4081029ci5b3cd1fdf4b10c97@mail.gmail.com>
URL: https://gitlist.dev/e/118833cc0606280956s4081029ci5b3cd1fdf4b10c97%40mail.gmail.com

```
It looks like connect.c waits too long before it uses errno in both copies
of git_tcp_connect_sock.  Both close and freeaddrinfo can poke any
non-zero value in there.

M.

```

## Petr Baudis, 2006-07-01 21:56

Subject: [PATCH] Fix errno usage in connect.c
Message-ID: <20060701215626.GB29115@pasky.or.cz>
URL: https://gitlist.dev/e/20060701215626.GB29115%40pasky.or.cz
In-Reply-To: <118833cc0606280956s4081029ci5b3cd1fdf4b10c97@mail.gmail.com>

```
Dear diary, on Wed, Jun 28, 2006 at 06:56:12PM CEST, I got a letter
where Morten Welinder <mwelinder@gmail.com> said that...
> It looks like connect.c waits too long before it uses errno in both copies
> of git_tcp_connect_sock.  Both close and freeaddrinfo can poke any
> non-zero value in there.

Nice catch.

->8-

errno was used after it could've been modified by a subsequent library call.
Spotted by Morten Welinder.

Signed-off-by: Petr Baudis <pasky@suse.cz>
---

 connect.c |   18 ++++++++++++------
 1 files changed, 12 insertions(+), 6 deletions(-)

diff --git a/connect.c b/connect.c
index cb4656d..9a87bd9 100644
--- a/connect.c
+++ b/connect.c
@@ -328,7 +328,7 @@ #ifndef NO_IPV6
  */
 static int git_tcp_connect_sock(char *host)
 {
-	int sockfd = -1;
+	int sockfd = -1, saved_errno = 0;
 	char *colon, *end;
 	const char *port = STR(DEFAULT_GIT_PORT);
 	struct addrinfo hints, *ai0, *ai;
@@ -362,9 +362,12 @@ static int git_tcp_connect_sock(char *ho
 	for (ai0 = ai; ai; ai = ai->ai_next) {
 		sockfd = socket(ai->ai_family,
 				ai->ai_socktype, ai->ai_protocol);
-		if (sockfd < 0)
+		if (sockfd < 0) {
+			saved_errno = errno;
 			continue;
+		}
 		if (connect(sockfd, ai->ai_addr, ai->ai_addrlen) < 0) {
+			saved_errno = errno;
 			close(sockfd);
 			sockfd = -1;
 			continue;
@@ -375,7 +378,7 @@ static int git_tcp_connect_sock(char *ho
 	freeaddrinfo(ai0);
 
 	if (sockfd < 0)
-		die("unable to connect a socket (%s)", strerror(errno));
+		die("unable to connect a socket (%s)", strerror(saved_errno));
 
 	return sockfd;
 }
@@ -387,7 +390,7 @@ #else /* NO_IPV6 */
  */
 static int git_tcp_connect_sock(char *host)
 {
-	int sockfd = -1;
+	int sockfd = -1, saved_errno = 0;
 	char *colon, *end;
 	char *port = STR(DEFAULT_GIT_PORT), *ep;
 	struct hostent *he;
@@ -426,8 +429,10 @@ static int git_tcp_connect_sock(char *ho
 
 	for (ap = he->h_addr_list; *ap; ap++) {
 		sockfd = socket(he->h_addrtype, SOCK_STREAM, 0);
-		if (sockfd < 0)
+		if (sockfd < 0) {
+			saved_errno = errno;
 			continue;
+		}
 
 		memset(&sa, 0, sizeof sa);
 		sa.sin_family = he->h_addrtype;
@@ -435,6 +440,7 @@ static int git_tcp_connect_sock(char *ho
 		memcpy(&sa.sin_addr, *ap, he->h_length);
 
 		if (connect(sockfd, (struct sockaddr *)&sa, sizeof sa) < 0) {
+			saved_errno = errno;
 			close(sockfd);
 			sockfd = -1;
 			continue;
@@ -443,7 +449,7 @@ static int git_tcp_connect_sock(char *ho
 	}
 
 	if (sockfd < 0)
-		die("unable to connect a socket (%s)", strerror(errno));
+		die("unable to connect a socket (%s)", strerror(saved_errno));
 
 	return sockfd;
 }

-- 
				Petr "Pasky" Baudis
Stuff: http://pasky.or.cz/
Snow falling on Perl. White noise covering line noise.
Hides all the bugs too. -- J. Putnam

```

## Junio C Hamano, 2006-07-02 00:09

Subject: Re: [PATCH] Fix errno usage in connect.c
Message-ID: <7vd5codfzn.fsf@assigned-by-dhcp.cox.net>
URL: https://gitlist.dev/e/7vd5codfzn.fsf%40assigned-by-dhcp.cox.net
In-Reply-To: <20060701215626.GB29115@pasky.or.cz>

```
Petr Baudis <pasky@suse.cz> writes:

> Dear diary, on Wed, Jun 28, 2006 at 06:56:12PM CEST, I got a letter
> where Morten Welinder <mwelinder@gmail.com> said that...
>> It looks like connect.c waits too long before it uses errno in both copies
>> of git_tcp_connect_sock.  Both close and freeaddrinfo can poke any
>> non-zero value in there.
>
> Nice catch.
>
> ->8-
>
> errno was used after it could've been modified by a subsequent library call.
> Spotted by Morten Welinder.
>
> Signed-off-by: Petr Baudis <pasky@suse.cz>

Thanks.

```
