On Tue, 27 Nov 2012, Andre Oppermann wrote:

On 27.11.2012 23:35, Peter Wemm wrote:
Andre.. this breaks incoming connections. TCP is immediately reset and never even gets to the listener process. You need to back out of fix this urgently please.

I just found out and fixed it.  Sorry for the breakage.

I'd like to see a much more thorough use of "Reviewed by:" in socket and TCP-related commits -- this is very sensitive code, and a second pair of eyes is always valuable. Post-commit review is not a substitute. Looking back over similar changes in the socket code over the last two years, I see that almost all have reviewers, so I think it would be reasonable to consider it mandatory for these subsystems at this point. The good news is that we have lots of people with expertise in it.

Robert


--
Andre

On Tue, Nov 27, 2012 at 12:04 PM, Andre Oppermann <an...@freebsd.org> wrote:
Author: andre
Date: Tue Nov 27 20:04:52 2012
New Revision: 243627
URL: http://svnweb.freebsd.org/changeset/base/243627

Log:
   Fix a race on listen socket teardown where while draining the
   accept queues a new socket/connection may be added to the queue
   due to a race on the ACCEPT_LOCK.

   The submitted patch is slightly changed in comments, teardown
   and locking order and extended with KASSERT's.

   Submitted by: Vijay Singh <vijju.singh-at-gmail-dot-com>
   Found by:     His team.
   MFC after:    1 week

Modified:
   head/sys/kern/uipc_socket.c

Modified: head/sys/kern/uipc_socket.c
==============================================================================
--- head/sys/kern/uipc_socket.c Tue Nov 27 19:35:21 2012        (r243626)
+++ head/sys/kern/uipc_socket.c Tue Nov 27 20:04:52 2012        (r243627)
@@ -555,6 +555,16 @@ sonewconn(struct socket *head, int conns
         so->so_snd.sb_flags |= head->so_snd.sb_flags & SB_AUTOSIZE;
         so->so_state |= connstatus;
         ACCEPT_LOCK();
+       /*
+        * The accept socket may be tearing down but we just
+        * won a race on the ACCEPT_LOCK.
+        */
+       if (!(so->so_options & SO_ACCEPTCONN)) {
+               SOCK_LOCK(so);
+               so->so_head = NULL;
+ sofree(so); /* NB: returns ACCEPT_UNLOCK'ed. */
+               return (NULL);
+       }
         if (connstatus) {
                 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
                 so->so_qstate |= SQ_COMP;
@@ -780,9 +790,14 @@ soclose(struct socket *so)
  drop:
         if (so->so_proto->pr_usrreqs->pru_close != NULL)
                 (*so->so_proto->pr_usrreqs->pru_close)(so);
+       ACCEPT_LOCK();
         if (so->so_options & SO_ACCEPTCONN) {
                 struct socket *sp;
-               ACCEPT_LOCK();
+               /*
+                * Prevent new additions to the accept queues due
+                * to ACCEPT_LOCK races while we are draining them.
+                */
+               so->so_options &= ~SO_ACCEPTCONN;
                 while ((sp = TAILQ_FIRST(&so->so_incomp)) != NULL) {
                         TAILQ_REMOVE(&so->so_incomp, sp, so_list);
                         so->so_incqlen--;
@@ -801,13 +816,15 @@ drop:
                         soabort(sp);
                         ACCEPT_LOCK();
                 }
-               ACCEPT_UNLOCK();
+               KASSERT((TAILQ_EMPTY(&so->so_comp)),
+                   ("%s: so_comp populated", __func__));
+               KASSERT((TAILQ_EMPTY(&so->so_incomp)),
+                   ("%s: so_incomp populated", __func__));
         }
-       ACCEPT_LOCK();
         SOCK_LOCK(so);
         KASSERT((so->so_state & SS_NOFDREF) == 0, ("soclose: NOFDREF"));
         so->so_state |= SS_NOFDREF;
-       sorele(so);
+ sorele(so); /* NB: Returns with ACCEPT_UNLOCK(). */
         CURVNET_RESTORE();
         return (error);
  }





_______________________________________________
svn-src-head@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-head
To unsubscribe, send any mail to "svn-src-head-unsubscr...@freebsd.org"

Reply via email to