FreeTDS Development Group

["ZIGLIO, Frediano, VF-IT" <Frediano.Ziglio AT vodafone.com>]

> 
> Dear list,
> 
> I use this great library (libtds) to communicate with MSSQL 
> server on an
> embedded system. It's rather important that this system never stops
> sending data for a longer period of time. But as far as I 
> could observe
> this occurs every time the server becomes unreachable. In both the
> stable release and the CVS version a timeout can be set for this case,
> but in the stable version this triggers an assertion and in the CVS

Mmm... assertion == DoS here... not that fine...

> version as far as I understand the code it leads to an infinite loop.
> For the latter case there is a rather simple fix which makes the
> goodread function abort instead of waiting forever. I only insert data
> into the database but this has worked reliably now for the 
> last 4 days.
> 
> best regards
> Omar Siam
> 

Well, this code has been revise a lot of time... I still don't
understand why with correct timeouts you should have an infinite loop...
probably dblib say continue or sending cancel connection close are not
detected...

Let's look at the code a bit deeper.

291         global_start = start = tds->query_start_time ?
tds->query_start_time : time(NULL);
292
293         while (buflen > 0) {
294
295                 int len;
296                 time_t now = time(NULL);
297
298                 if (IS_TDSDEAD(tds))
299                         return -1;
300
301                 FD_ZERO(&rfds);
302                 FD_SET(tds->s, &rfds);
303
304                 timeout = 0;
305                 if (tds->query_timeout > 0) {
306                         timeout = tds->query_timeout - (now -
start);
307                         if (timeout < 1)
308                                 timeout = 1;
309                 }
310                 tv.tv_sec = timeout;
311
312                 if ((len = select(tds->s + 1, &rfds, NULL, NULL,
timeout > 0 ? &tv : NULL)) > 0) {
313                         len = 0;
314                         if (FD_ISSET(tds->s, &rfds)) {
315 #ifndef MSG_NOSIGNAL
316                                 len = READSOCKET(tds->s, buf + got,
buflen);
317 #else
318                                 len = recv(tds->s, buf + got,
buflen, MSG_NOSIGNAL);
319 #endif
320                         }
321                 }
322
323                 if (len < 0) {

Here you suggest len <= 0 however len == 0 means mainly a timeout in
select. It seems that if connection is closed select in this case just
timeouts. I would try to change line 312 in

   if ((len = select(tds->s + 1, &rfds, NULL, &rfds, timeout > 0 ? &tv :
NULL)) > 0) {

This should detect socket error and next read should set correctly len
and sock_errno. Can you try this ??
If this works as I expect some other selects in net.c should be changed
too...

324                         switch(sock_errno) {
325                         case TDSSOCK_EINTR:
326                                 tdsdump_log(TDS_DBG_NETWORK, "socket
read interrupted\n");
327                         case EAGAIN:
328                         case TDSSOCK_EINPROGRESS:
329                                 break;
330                         default:
331                                 return -1;
332                         }
333                         len = 0;
334                 }
335
336                 buflen -= len;
337                 got += len;
338
339                 now = time(NULL);
340                 if (tds->query_timeout > 0 && now - start >=
tds->query_timeout) {
341
342                         int timeout_action = TDS_INT_CONTINUE;
343
344                         tdsdump_log(TDS_DBG_NETWORK, "exceeded query
timeout: %d\n", tds->query_timeout);
345
346                         if (tds->query_timeout_func &&
tds->query_timeout)
347                                 timeout_action =
(*tds->query_timeout_func) (tds->query_timeout_param, now -
global_start);
348
349                         switch (timeout_action) {
350                         case TDS_INT_EXIT:
351                                 exit(EXIT_FAILURE);
352                                 break;
353                         case TDS_INT_CANCEL:
354                                 tds_send_cancel(tds);
355                                 break;
356
357                         case TDS_INT_CONTINUE:
358                                 start = now;
359                         default:
360                                 break;
361                         }
362                 }
363
364                 if (unfinished && got)
365                         return got;
366         }
367         return got;
368 }

freddy77