Re: kern/49410: -current: panic: "KERNEL_LOCKED_P()" failed: file ".../if_loop.c", line 221

[Ryota Ozaki <ozaki-r%netbsd.org@localhost>]

On Sun, Nov 23, 2014 at 12:35 AM,  <kardel%netbsd.org@localhost> wrote:
>>Number:         49410
>>Category:       kern
>>Synopsis:       -current MROUTING enabled kernel panics in if_loop.c:221 with KERNEL_LOCKED_P assertion
>>Confidential:   no
>>Severity:       serious
>>Priority:       high
>>Responsible:    kern-bug-people
>>State:          open
>>Class:          sw-bug
>>Submitter-Id:   net
>>Arrival-Date:   Sat Nov 22 15:35:00 +0000 2014
>>Originator:     Charlie Root
>>Release:        NetBSD 7.99.1
>>Organization:
>>Environment:
> System: NetBSD gateway.kardel.name 7.99.1 NetBSD 7.99.1 (GATEWAY) #21: Sat Nov 22 13:49:32 CET 2014 kardel%pip.kardel.name@localhost:/fs/raid1a/src/NetBSD/cur/src/obj.amd64/sys/arch/amd64/compile/GATEWAY amd64
> Architecture: x86_64
> Machine: amd64
>>Description:
>         A -current kernel with multicast support panics due to locking error.
>
> panic: kernel diagnostic assertion "KERNEL_LOCKED_P()" failed: file "/fs/raid1a/src/NetBSD/cur/src/sys/net/if_loop.c", line 221
> cpu1: Begin traceback...
> vpanic() at netbsd:vpanic+0x13c
> kern_assert() at netbsd:kern_assert+0x4f
> looutput() at netbsd:looutput+0x39e
> ip_output() at netbsd:ip_output+0x8b0
> igmp_sendpkt.isra.1() at netbsd:igmp_sendpkt.isra.1+0x117
> igmp_fasttimo() at netbsd:igmp_fasttimo+0xb3
> pffasttimo() at netbsd:pffasttimo+0x31
> callout_softclock() at netbsd:callout_softclock+0x1d0
> softint_dispatch() at netbsd:softint_dispatch+0xd3
>>How-To-Repeat:
>         pick -current kernel with MROUTING enabled and run with
>         multicast services like mrouted and multiple interfaces
>         and watch it panic while trying to reach multi user.
>>Fix:
>         With MROUTING disabled the kernel will survive the startup
>         but that is not the solution.

Thank you for the report. A possible fix is the below patch.
I'm still checking whether the place I put KERNEL_LOCK is appropriate
or not, anyway it should fix the assertion failure.

Regards,
  ozaki-r

diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c
index a92c61b..4d31253 100644
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -1713,5 +1713,7 @@ ip_mloopback(struct ifnet *ifp, struct mbuf *m,
const struct sockaddr_in *dst)

  ip->ip_sum = 0;
  ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
+ KERNEL_LOCK(1, NULL);
  (void)looutput(ifp, copym, sintocsa(dst), NULL);
+ KERNEL_UNLOCK_ONE(NULL);
 }