Re: kernel/6043: panic: rtfree 2

[Insan Praja SW]

On Wed, 25 Feb 2009 03:12:09 +0700, Claudio Jeker clau...@openbsd.org  
wrote:



On Tue, Feb 24, 2009 at 08:37:22AM +0700, Insan Praja SW wrote:

On Tue, 24 Feb 2009 06:20:05 +0700, Claudio Jeker clau...@openbsd.org
wrote:

Could you please run with the following diff and report if you still  
hit

the panics. It seems some drivers (maybe even em) fail to call the link
state update function without the needed splsoftnet protection.
Using splnet in rt_if_track may be to protective but we can reconsider
that if the panic is gone.

thanks


Hi Claudio,
I will test this diff today and I will get back to you with result.
Thanks,



I got other negative feedback (in other words the panic still happens).


Well, same here.
panic: rtfree 2
Stopped atDebugger+0x4: leave

ddb trace
Debugger (d2ee1f58,d2ab7100,4,e087fc0c,d2ee1f00) at Debugger+0x4
panic (d06f6a96,d2ee1f78,d2ab7100,0,0) at panic+0x55
rtfree (e087fc0c,30,0,d2ab7100) at rtfree+0xbd
route_output (d899d100,d8871c70,0,de84cd50,d0202de9) at route_output+0x341
raw_usrreq (d8871c70,9,d899d100,0,0) at raw_usrreq+0xf2
route_usrreq (d8871c70,9,d899d100,0,0) at route_usrreq+0xe8
sosend (d8871c70,0,de84cea8,d899d100,0,0) at sosend+0x357
soo_write (d8a00ec8,d8a00ee4,de84cea8,d8a1e000) at soo_write+0x1c
dofilewritev (d89f8c44,6,d8a00ec8,cfbd44905) at dofilewritev+0xeb
sys_writev (d89f8c44,de84cf68,de84cf58,cfbd4490,d89f8c44) at  
sys_writev+0x47

syscall () at syscall+0x24e
--- syscall (number 121) ---
0x3623d9d


So
as we're right before release the best thing is to turn the featrue of  
for

now. So that we don't ship broken code in 4.5.

Could you test this diff instead?

 Ok.
Thanks,

Insan

--
insandotpraja(at)gmaildotcom

Re: kernel/6043: panic: rtfree 2

[Insan Praja SW]

Hi Claudio and Misc@,
On Wed, 25 Feb 2009 03:12:09 +0700, Claudio Jeker clau...@openbsd.org  
wrote:



On Tue, Feb 24, 2009 at 08:37:22AM +0700, Insan Praja SW wrote:

On Tue, 24 Feb 2009 06:20:05 +0700, Claudio Jeker clau...@openbsd.org
wrote:

Could you please run with the following diff and report if you still  
hit

the panics. It seems some drivers (maybe even em) fail to call the link
state update function without the needed splsoftnet protection.
Using splnet in rt_if_track may be to protective but we can reconsider
that if the panic is gone.

thanks


Hi Claudio,
I will test this diff today and I will get back to you with result.
Thanks,



I got other negative feedback (in other words the panic still happens).  
So
as we're right before release the best thing is to turn the featrue of  
for

now. So that we don't ship broken code in 4.5.

Could you test this diff instead?


Yeah I did.. and it works.. So it disabled rt_if_track? how 'bout sending  
half of the data to nirvana things on a multipath routing situation? is  
it affected?

Thanks.

--
insandotpraja(at)gmaildotcom