Re: [Lksctp-developers] BUG: recv() from rebooted hosts before accept()

[Vlad Y. <vla...@hp...>]

Nathan Straz wrote:
> On Dec 18 11:35, Vlad Yasevich wrote:
>> Nathan Straz wrote:
>>> Okay, so because I reuse the src port on the rebooted host, I get an
>>> association restart and the connection as far as the server is concerned
>>> is restored.  Is that behavior specific to SCTP?
>> Yes.
>>
>>>> If you test doesn't expect this, you need to randomize your ports, but
>>>> even then this situation may happen anyway.
>>> I'm not sure how to set the source port of a connection so I think I'll
>>> teach my program how to handle the association restart.  Is there a way
>>> to detect the association restart on the client or server side?
>> To set the source port of the connection, bind to an explicit port.  You can
>> either call random() and bind to a that random port or apply the kernel patch
>> and that will try to randomize ports picked by the kernel.
> 
> Ah, good to know.  I decided to just handle the reconnect since my
> protocol includes a join message.  Everything was working great until I
> hit a panic on two nodes.  I filed a bug on it in the Red Hat Bugzilla.
> 
> http://bugzilla.redhat.com/show_bug.cgi?id=426234
> 
> I don't know if this backtrace will ring any bells.

See if this fixes your issue:

http://git.kernel.org/?p=linux/kernel/git/davem/net-2.6.git;a=commitdiff_plain;h=f26f7c480555812ca7c4037e0a50fa54afe2cb4a

-vlad

> 
> BUG: unable to handle kernel NULL pointer dereference at virtual address 00000024
>  printing eip:
> f8e12ffb
> *pde = 55add067
> Oops: 0000 [#1]
> SMP 
> last sysfs file: /devices/pci0000:00/0000:00:00.0/irq
> Modules linked in: sctp gfs(U) lock_dlm gfs2 dlm configfs autofs4 hidp rfcomm
> l2cap bluetooth sunrpc ipv6 dm_multipath video sbs backlight i2c_ec button
> battery asus_acpi ac lp intel_rng floppy e7xxx_edac e1000 edac_mc i2c_i801
> ide_cd parport_pc pcspkr i2c_core parport cdrom sg dm_snapshot dm_zero dm_mirror
> dm_mod qla2xxx scsi_transport_fc ata_piix libata sd_mod scsi_mod ext3 jbd
> ehci_hcd ohci_hcd uhci_hcd
> CPU:    0
> EIP:    0060:[<f8e12ffb>]    Not tainted VLI
> EFLAGS: 00010286   (2.6.18-53.el5 #1) 
> EIP is at sctp_get_port+0x12/0x44 [sctp]
> eax: 00000000   ebx: ef935b80   ecx: f8e22ae0   edx: ef935b80
> esi: ef935b80   edi: 00000000   ebp: 00000010   esp: ee797ea0
> ds: 007b   es: 007b   ss: 0068
> Process d_doio (pid: 9256, ti=ee797000 task=e4db8aa0 task.ti=ee797000)
> Stack: c05a5a4a f133e500 00000040 f7bb7d80 ee797ebc c05a4ad2 ef935b80 ef935b80 
>        ef935b80 ee797eec c05e6d1f f8e17bc0 c05e75f1 f8e17bc0 f5a8eb00 ee797eec 
>        00000010 c05a46d8 00000002 2b810002 62590f0a 00000000 00000000 00000000 
> Call Trace:
>  [<c05a5a4a>] lock_sock+0x8e/0x96
>  [<c05a4ad2>] sys_recvfrom+0x101/0x137
>  [<c05e6d1f>] inet_autobind+0x1c/0x51
>  [<c05e75f1>] inet_dgram_connect+0x30/0x4e
>  [<c05a46d8>] sys_connect+0x7d/0xa9
>  [<c0492369>] inotify_d_instantiate+0x3c/0x5f
>  [<c0483308>] d_rehash+0x1c/0x2b
>  [<c05a3e90>] sock_attach_fd+0x6c/0xcc
>  [<c06058c7>] _spin_lock_bh+0x8/0x18
>  [<c05a5a4a>] lock_sock+0x8e/0x96
>  [<c05a5a4a>] lock_sock+0x8e/0x96
>  [<c06058c7>] _spin_lock_bh+0x8/0x18
>  [<c05a4ec1>] sys_socketcall+0x8c/0x19e
>  [<c0407ef7>] do_syscall_trace+0xab/0xb1
>  [<c0404eff>] syscall_call+0x7/0xb
>  =======================
> Code: ff ff bd f4 ff ff ff e9 5f fb ff ff 81 c4 24 02 00 00 89 e8 5b 5e 5f 5d c3
> 57 89 d7 56 53 89 c3 83 ec 1c 8b 80 f4 01 00 00 89 da <8b> 48 24 89 e0 ff 51 30
> 0f b7 d7 89 d0 c1 e8 08 c1 e2 08 09 c2 
> EIP: [<f8e12ffb>] sctp_get_port+0x12/0x44 [sctp] SS:ESP 0068:ee797ea0
>  <0>Kernel panic - not syncing: Fatal exception
> 
> Nate Straz
>