Re: RTNL: assertion failed at net/core/dev.c (3961)

[Eric Dumazet <eric.dumazet@gmail.com>]

Meelis Roos a écrit :
> I'm running 2.6.31-rc4 on an i386 pc and got these two similar 
> but different in cause assertion failures  on boot. Seems to be 
> repeatable. I saw similar messages on a kernel around -rc3 but it had 
> SIT troubles so I did not test more earlier rc-s.
> 
> This seems to be VLAN+IPv6 related, so some more info on the 
> configurartion:
> 
> eth0 is plain ethernet, internal LAN, IPv6 address autoconfigured from 
> external 6to4 address.
> 
> eth1 is ethernet, external interface. Dynamic IPv4, 6to4 tunnel is 
> created automatically from it. Only link-local IPv6 on this 
> interface. This is untagged, but it also carries VLAN 4 tagged.
> 
> eth1.4 is the VLAN on eth1 that carries iptv. Dynamic IPv4, only 
> link-local IPv6.
> 
> eth2 is plain internal LAN, like eth0 - static IPv4, IPv6 from 6to4.
> 
> tun6to4 is sti tunnel for 6to4, cofigured from eth1's IP.
> 
> eth0 and eth1 are 8139too, eth2 is uli526x.
> 
> RTNL: assertion failed at net/core/dev.c (3961)
> Pid: 0, comm: swapper Not tainted 2.6.31-rc4 #327
> Call Trace:
>  [<c127fd2b>] ? printk+0x1d/0x32
>  [<c120c43c>] dev_unicast_sync+0x3b/0xfe
>  [<f8a0b19d>] vlan_dev_set_rx_mode+0x2f/0x45 [8021q]
>  [<c120bab5>] __dev_set_rx_mode+0x88/0x9e
>  [<c1211c48>] dev_mc_add+0x52/0x7c
>  [<f890d5d1>] igmp6_group_added+0x5e/0x160 [ipv6]
>  [<f8900923>] ? fib6_clean_node+0x0/0x9f [ipv6]
>  [<f890d9e6>] ipv6_dev_mc_inc+0x2b9/0x2dc [ipv6]
>  [<f88f6745>] addrconf_join_solict+0x43/0x56 [ipv6]
>  [<f88f197d>] ipv6_dev_ac_inc+0x15c/0x1ab [ipv6]
>  [<f88f5f55>] addrconf_join_anycast+0x82/0x9b [ipv6]
>  [<f88f75b4>] __ipv6_ifa_notify+0xd6/0x153 [ipv6]
>  [<f88f8584>] ? addrconf_dad_timer+0x0/0x120 [ipv6]
>  [<f88f7662>] ipv6_ifa_notify+0x31/0x4c [ipv6]
>  [<f88f76a3>] addrconf_dad_completed+0x26/0xb7 [ipv6]
>  [<f88f8607>] addrconf_dad_timer+0x83/0x120 [ipv6]
>  [<f89fd942>] ? garp_join_timer+0x0/0x79 [garp]
>  [<f89fd9a5>] ? garp_join_timer+0x63/0x79 [garp]
>  [<c102dbff>] run_timer_softirq+0x152/0x1ca
>  [<c1029b9b>] __do_softirq+0x69/0xe6
>  [<c1029b32>] ? __do_softirq+0x0/0xe6
>  <IRQ>  [<c10298a2>] ? irq_exit+0x36/0x7b
>  [<c1012a71>] ? smp_apic_timer_interrupt+0x70/0x8c
>  [<c10032a1>] ? apic_timer_interrupt+0x31/0x38
>  [<c1008b67>] ? default_idle+0x36/0x5c
>  [<c1001cbb>] ? cpu_idle+0x72/0x9f
>  [<c1278d08>] ? rest_init+0x74/0x87
>  [<c13c9af9>] ? start_kernel+0x278/0x28e
>  [<c13c9312>] ? i386_start_kernel+0x40/0x56
> ADDRCONF(NETDEV_UP): eth2: link is not ready
> powernow-k8: Found 1 AMD Athlon(tm) 64 Processor 3000+ processors (1 cpu 
> cores) (version 2.20.00)
> powernow-k8:    0 : fid 0xa (1800 MHz), vid 0x6
> powernow-k8:    1 : fid 0x2 (1000 MHz), vid 0x12
> Marking TSC unstable due to cpufreq changes
> RTNL: assertion failed at net/core/dev.c (3961)
> Pid: 2744, comm: avahi-daemon Not tainted 2.6.31-rc4 #327
> Call Trace:
>  [<c127fd2b>] ? printk+0x1d/0x32
>  [<c120c43c>] dev_unicast_sync+0x3b/0xfe
>  [<f8a0b19d>] vlan_dev_set_rx_mode+0x2f/0x45 [8021q]
>  [<c120bab5>] __dev_set_rx_mode+0x88/0x9e
>  [<c1211c48>] dev_mc_add+0x52/0x7c
>  [<f890d5d1>] igmp6_group_added+0x5e/0x160 [ipv6]
>  [<c1202560>] ? sock_alloc_send_pskb+0x9c/0x267
>  [<f890d9e6>] ipv6_dev_mc_inc+0x2b9/0x2dc [ipv6]
>  [<f890eb15>] ipv6_sock_mc_join+0x15d/0x1ce [ipv6]
>  [<f8902565>] do_ipv6_setsockopt+0x8a7/0xc2e [ipv6]
>  [<c11ff715>] ? sock_recvmsg+0xdf/0x108
>  [<c11ff877>] ? sock_sendmsg+0xd8/0x100
>  [<c10384fe>] ? autoremove_wake_function+0x0/0x54
>  [<c108eb47>] ? file_update_time+0x9e/0xd0
>  [<c11ffba2>] ? sys_sendto+0x10d/0x146
>  [<f890293f>] ipv6_setsockopt+0x53/0xb0 [ipv6]
>  [<f8905951>] udpv6_setsockopt+0x4c/0x67 [ipv6]
>  [<c1201262>] sock_common_setsockopt+0x22/0x38
>  [<c11ff150>] sys_setsockopt+0x79/0xa8
>  [<c1200cd1>] sys_socketcall+0x129/0x18a
>  [<c1002bbb>] sysenter_do_call+0x12/0x26
> 
> 

Thanks for your detailed report Meelis

Seems because of following commit.

So assumption that dev_unicast_sync() is always called with RTNL is currently
false.


commit ccffad25b5136958d4769ed6de5e87992dd9c65c
Author: Jiri Pirko <jpirko@redhat.com>
Date:   Fri May 22 23:22:17 2009 +0000

    net: convert unicast addr list

    This patch converts unicast address list to standard list_head using
    previously introduced struct netdev_hw_addr. It also relaxes the
    locking. Original spinlock (still used for multicast addresses) is not
    needed and is no longer used for a protection of this list. All
    reading and writing takes place under rtnl (with no changes).

    I also removed a possibility to specify the length of the address
    while adding or deleting unicast address. It's always dev->addr_len.

    The convertion touched especially e1000 and ixgbe codes when the
    change is not so trivial.

    Signed-off-by: Jiri Pirko <jpirko@redhat.com>