This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: recv and errno during a connection reset/closed by peer


Peter,

FYI - I tried setting the SO_LINGER and the SO_RCVTIMEO on Linux and on
Cygwin (see code below).  It makes sense to me that if the timeout has been
exceeded then the recv(..., MSG_PEEK) should do something.  On Cygwin it
doesn't do anything.

On Linux it DOES time out, but it returns -1 and errno=EAGAIN.  This is an
interesting choice for errno (perhaps ECONNABORTED, ESHUTDOWN, ECOMM, ETIME,
ETIMEDOUT, EHOSTUNREACH, or ENOLINK would have been better choices), but at
least it is consistent.  In the case below, every 10.1 seconds I get -1 and
errno=EAGAIN from my MSG_PEEK, so it is useable.

I may get even better results now by combining the SO_RCVTIMEO with
SO_KEEPALIVE now as Juris had suggested.

The spec is kinda vague about the response to a MSG_PEEK when the connection
is lost, but I think that you could write a book on this subject with all
the possible combinations.

Pete

#include <errno.h>
#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <string.h>

#define MAX_CHARS 18
#define MAX_BUF (MAX_CHARS+2) // +2 is for the CRLF

void show_status(int fd);
void err(char * pre);
int handle_it(char * buf);

int main(int argc, char **argv)
{
   int lfd=-1, afd=-1;
   char buf[MAX_BUF];

   int vlen=0, temp=0;
   socklen_t vslen=0;
   struct linger vlinger;
   struct timeval vtime;

   lfd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
   if (-1 == lfd)
      printf("Error at socket(): %i\n", errno);

   sockaddr_in svc;
   svc.sin_family = AF_INET;
   svc.sin_addr.s_addr = inet_addr("192.168.1.244");
   svc.sin_port = htons(27015);

   if (-1 == bind(lfd, (struct sockaddr *) &svc, sizeof(svc)))
      printf("bind() failed.\n");

   if (-1 == listen(lfd, 1 ))
      printf("Error listening on socket.\n");

   printf("Waiting for client to connect...\n");
   while( -1 == afd ){
      afd = accept( lfd, NULL, NULL );
   }
   printf("Client connected.\n");

   // set receive timeout
   vslen = sizeof(struct timeval);
   temp = getsockopt(afd, SOL_SOCKET, SO_RCVTIMEO, &vtime, &vslen);
   if(-1 == temp){
     err("Error at getsockopt(SO_RCVTIMEO)");
   }
   else{
     vtime.tv_sec=10;
     vtime.tv_usec=5;
     vlen = sizeof(struct timeval);
     temp = setsockopt(afd, SOL_SOCKET, SO_RCVTIMEO, &vtime, vlen);
     if(-1 == temp){
       err("Error at setsockopt(SO_RCVTIMEO)");
     }
   }

   int ret_val = 0;
   int done = 0;
   do{
     ret_val = recv(afd, buf, MAX_BUF, MSG_PEEK);
     if(0 > ret_val)
       err(" >>> ERRNO");
     else if(ret_val > 0){
        ret_val = recv(afd, buf, MAX_BUF, 0);
       buf[ret_val]='\0';
       printf("(%i) \t> %s", ret_val, buf);
     }

     usleep(250);
     done = handle_it(buf);

   }while(!done);

   printf("Completed\n");

   return 0;
}

-----Original Message-----
From: Peter A. Castro [mailto:doctor@fruitbat.org]
Sent: Monday, March 28, 2005 7:02 PM
To: Peter Stephens
Cc: cygwin@cygwin.com
Subject: RE: recv and errno during a connection reset/closed by peer

On Mon, 28 Mar 2005, Peter Stephens wrote:

> Brian
>
> Are you saying that there is no way to distinguish a dropped 
> connection from a MSG_PEEK with no data to retrieve?

As someone who's seen this behaviour on several platforms, it can happen.
I've had to deal with this little annoyance in other products by having a
retry counter loop.  So many consecutive recv()s of 0 length constitues a
closed connection.  Something like this might work here as well?

> Pete
>
> P.S.  I use this standard: http://www.unix.org/unix03.html .  Is this 
> incorrect?

Brians' link is to the Single Unix Spec V2, so it's not all that old.

> Peter A. Stephens
> ptfoof@sbcglobal.net
>
> -----Original Message-----
> From: Brian Ford [mailto:ford@vss.fsi.com]
> Sent: Monday, March 28, 2005 11:37 AM
> To: Peter Stephens
> Cc: cygwin@cygwin.com
> Subject: RE: recv and errno during a connection reset/closed by peer
>
> On Fri, 25 Mar 2005, Peter Stephens wrote:
>
>> I boiled this down to nothing(see below).  I must be missing 
>> something basic.
>
> Yup.
>
>> I tried the suggestions made so far and it never gets to:
>>
>> 	printf(" >>> ERRNO %i\n", errno);
>>
>> I would expect that on a disconnect (I use putty in telnet or raw
>> mode) it would return -1 whether it is doing MSG_PEEK or an actual 
>> retrieval.  No luck.
>
> Nope (http://www.opengroup.org/onlinepubs/7990989775/xns/recv.html):
>
> "If no messages are available to be received and the peer has 
> performed an orderly shutdown, recv() returns 0."
>
> I think that's what you're missing.
>
> --
> Brian Ford
> Senior Realtime Software Engineer
> VITAL - Visual Simulation Systems
> FlightSafety International
> the best safety device in any aircraft is a well-trained pilot...
>
>
> --
> Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
> Problem reports:       http://cygwin.com/problems.html
> Documentation:         http://cygwin.com/docs.html
> FAQ:                   http://cygwin.com/faq/
>

--
Peter A. Castro <doctor@fruitbat.org> or <Peter.Castro@oracle.com>
 	"Cats are just autistic Dogs" -- Dr. Tony Attwood


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]