2.6.17-rc5-mm2 problem ?

2.6.17-rc5-mm2 problem ?

[Badari Pulavarty]

Hi Andrew,

Have you seen this on 2.6.17-rc5-mm2 before ? Could be due to my
patchset (fileop cleanups), but doesn't seem like it ..

Thanks,
Badari

Unable to handle kernel NULL pointer dereference at 0000000000000008
RIP:
 [<ffffffff80448d5f>] skb_dequeue+0x2c/0x50
PGD 18024d067 PUD 19d2ab067 PMD 0
Oops: 0002 [1] SMP
last sysfs file: /class/net/sit0/address
CPU 2
Modules linked in: usbserial parport_pc lp parport ipv6 ohci_hcd
i2c_amd756 i2c_core usbcore dm_mod
Pid: 4438, comm: syslogd Not tainted 2.6.17-rc5-mm2 #2
RIP: 0010:[<ffffffff80448d5f>]  [<ffffffff80448d5f>] skb_dequeue
+0x2c/0x50
RSP: 0018:ffff81019d157b68  EFLAGS: 00010046
RAX: 0000000000000000 RBX: ffff81019f7f61d8 RCX: ffff81019d157c54
RDX: 0000000000000000 RSI: 0000000000000246 RDI: ffff81019f7f61ec
RBP: ffff8101bdff05c0 R08: 0000000000000000 R09: 0000000000000000
R10: ffff81019d157bd8 R11: 0000000000000028 R12: ffff81019f7f61ec
R13: ffff81019d157cb8 R14: ffff81019d157c54 R15: 0000000000000000
FS:  00002b2f389756e0(0000) GS:ffff8101a00ca640(0000)
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 0000000000000008 CR3: 000000019d950000 CR4: 00000000000006e0
Process syslogd (pid: 4438, threadinfo ffff81019d156000, task
ffff81019d41d090)
Stack: 0000000000000000 ffff81019f7f6140 ffff81019f7f61d8
ffffffff8044b673
       ffff81019f7f6370 7fffffffffffffff ffff81019d156000
ffff81019d157bd8
       ffff81019d41d090 ffffffff80240137
Call Trace:
 [<ffffffff8044b673>] skb_recv_datagram+0x93/0x1f8
 [<ffffffff80240137>] debug_mutex_add_waiter+0x50/0x65
 [<ffffffff804bd59b>] __mutex_lock_slowpath+0x1d4/0x218
 [<ffffffff804a3cc6>] unix_dgram_recvmsg+0x6e/0x1e5
 [<ffffffff80240137>] debug_mutex_add_waiter+0x50/0x65
 [<ffffffff80444e4b>] sock_recvmsg+0x117/0x13b
 [<ffffffff80253c7e>] generic_file_aio_write+0x72/0xc7
 [<ffffffff80253c0c>] generic_file_aio_write+0x0/0xc7
 [<ffffffff80276b88>] do_sync_readv_writev+0xc8/0x10f
 [<ffffffff8023d005>] autoremove_wake_function+0x0/0x2e
 [<ffffffff80445d5b>] sys_recvfrom+0xd3/0x12f
 [<ffffffff804c0e28>] kprobe_flush_task+0x1e/0x53
 [<ffffffff802091e6>] system_call+0x7e/0x83

Re: 2.6.17-rc5-mm2 problem ?

[Andrew Morton]

On Mon, 05 Jun 2006 09:37:40 -0700
Badari Pulavarty <pbadari@us.ibm.com> wrote:

> Have you seen this on 2.6.17-rc5-mm2 before ? Could be due to my
> patchset (fileop cleanups), but doesn't seem like it ..
> 
> Thanks,
> Badari
> 
> Unable to handle kernel NULL pointer dereference at 0000000000000008
> RIP:
>  [<ffffffff80448d5f>] skb_dequeue+0x2c/0x50

It could be the LLC bug.  That thing caused untold amounts of grief.

--- devel/net/llc/llc_input.c~git-net-llc-fix	2006-06-02 11:55:38.000000000 -0700
+++ devel-akpm/net/llc/llc_input.c	2006-06-02 11:55:38.000000000 -0700
@@ -176,7 +176,6 @@ int llc_rcv(struct sk_buff *skb, struct 
  		struct sk_buff *cskb = skb_clone(skb, GFP_ATOMIC);
  		if (cskb)
  			rcv(cskb, dev, pt, orig_dev);
-		rcv(skb, dev, pt, orig_dev);
 	}
 	dest = llc_pdu_type(skb);
 	if (unlikely(!dest || !llc_type_handlers[dest - 1]))
_

Re: 2.6.17-rc5-mm2 problem ?

[Badari Pulavarty]

On Mon, 2006-06-05 at 11:41 -0700, Andrew Morton wrote:
> On Mon, 05 Jun 2006 09:37:40 -0700
> Badari Pulavarty <pbadari@us.ibm.com> wrote:
> 
> > Have you seen this on 2.6.17-rc5-mm2 before ? Could be due to my
> > patchset (fileop cleanups), but doesn't seem like it ..
> > 
> > Thanks,
> > Badari
> > 
> > Unable to handle kernel NULL pointer dereference at 0000000000000008
> > RIP:
> >  [<ffffffff80448d5f>] skb_dequeue+0x2c/0x50
> 
> It could be the LLC bug.  That thing caused untold amounts of grief.
> 
> --- devel/net/llc/llc_input.c~git-net-llc-fix	2006-06-02 11:55:38.000000000 -0700
> +++ devel-akpm/net/llc/llc_input.c	2006-06-02 11:55:38.000000000 -0700
> @@ -176,7 +176,6 @@ int llc_rcv(struct sk_buff *skb, struct 
>   		struct sk_buff *cskb = skb_clone(skb, GFP_ATOMIC);
>   		if (cskb)
>   			rcv(cskb, dev, pt, orig_dev);
> -		rcv(skb, dev, pt, orig_dev);
>  	}
>  	dest = llc_pdu_type(skb);
>  	if (unlikely(!dest || !llc_type_handlers[dest - 1]))
> _

Thanks. 2.6.17-rc5-mm3 is behaving much better, haven't seen these
issues on -mm3 with my patchset. I don't want to be the reason for
breaking -mm tree with fileop changes (I already used up my free
credits) ;)

Thanks,
Badari