Re: MLD problems (again) [PATCH]

Takashi,
Great! I actually built it for 2.4.24 myself and was about to ask for
more information, since it worked for me.

                        +-DLS


Takashi Hibi [off-list ref] on 01/12/2004 11:59:41 PM

To:    David Stevens/Beaverton/IBM@IBMUS
cc:    netdev@oss.sgi.com, davem@redhat.com
Subject:    Re: MLD problems (again) [PATCH]



David,

Good news.
I found a mistake which I made when I built the kernel.
After fixing it, the problem is solved and the Linux box can join
multicast group at once.
The patch is OK.

Thank you.

Takashi Hibi

David,

The patch is very simple and I have no problem to apply it to 2.4 line

kernel.

I already applied this patch, but it didn't solve the problem.

Regards,
Takashi Hibi

quoted

Takashi,
      The patch I sent was for 2.6.1, but it should also apply to
2.4 line kernels and fix the problem. If you have problems getting
it to apply in 2.4, let me know what version and I can provide you
a patch specific for that kernel.

                              +-DLS


Takashi Hibi [off-list ref] on 01/12/2004 03:04:48 AM

To:    David Stevens/Beaverton/IBM@IBMUS
cc:    netdev@oss.sgi.com, davem@redhat.com
Subject:    Re: MLD problems (again) [PATCH]



David,

Thank you for the patch.
But at least it doesn't solve the problem for 2.4 line kernel.
I will also test with 2.6.1 later.

Regards,
Takashi Hibi

quoted

Takashi,
      I believe the patch below will fix the problem you
had with MCAST_JOIN_SOURCE_GROUP not sending a
report. There was a typo in the source filter switching that did
two deletes, rather than an delete and an add.

Dave,
      Although IGMPv3 didn't have any problems, this patch
also re-arranges the order of the filter changes. I think it's cleaner
to add the new one first and then delete the old one, rather than
having a small window with no filter set. So, this is a bug fix for MLD
and a code clean-up for IGMPv3.
      This bug and patch should also apply to the 2.4 line.

                        +-DLS

[included in-line for viewing and as an attachment for unmangled
      whitespace]

--- linux-2.6.1/net/ipv6/mcast.c    2004-01-08 22:59:56.000000000 -0800
+++ linux-2.6.1F1/net/ipv6/mcast.c  2004-01-11 21:06:05.000000000 -0800

@@ -372,9 +372,9 @@
                  goto done;
      } else if (pmc->sfmode != omode) {
            /* allow mode switches for empty-set filters */
+           ip6_mc_add_src(idev, group, omode, 0, 0, 0);
            ip6_mc_del_src(idev, group, pmc->sfmode, 0, 0, 0);
            pmc->sfmode = omode;
-           ip6_mc_del_src(idev, group, pmc->sfmode, 0, 0, 0);
      }

      psl = pmc->sflist;
--- linux-2.6.1/net/ipv4/igmp.c     2004-01-08 23:00:12.000000000 -0800
+++ linux-2.6.1F1/net/ipv4/igmp.c   2004-01-11 21:27:41.000000000 -0800

@@ -1749,11 +1749,10 @@
                  goto done;
      } else if (pmc->sfmode != omode) {
            /* allow mode switches for empty-set filters */
+           ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, 0,

0);

quoted

quoted

            ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode,

0,

quoted

quoted

                  0, 0);
            pmc->sfmode = omode;
-           ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode,

0,

quoted

quoted

-                 0, 0);
      }

      psl = pmc->sflist;

(See attached file: 2.6.1MLD.patch)