[Prism54-devel] [patch 0/6] [prism54 svn trunk] various fixes for prism54 svn

[Bob Beers]

On 5/18/05, Jean-Baptiste Note <jean-baptiste.note at wanadoo.fr> wrote:
> 
> Hello Bob,
> 
Hi,

> Let's go further along the debugging !
> 
Yes, let's.

 
> Can you do in islpci_mgt.h ,
> 
> #define VERBOSE                                 0x01 | SHOW_QUEUE_INDEXES
> 
Certainly.

> It'll fill the logs rather rapidly, but if you can trigger the problem
> fast, it should be okay.
> 
I will try.  I did another test overnight of the same setup I described
 before and got almost to 14 hours, but then hard lockup, and no log
 info.  I'll put in this new VERBOSE and try to trigger the failure more
 quickly (much more quickly would be nice).
BTW, my other setup with the PCI XG-901 is still up and running.
 (but I do get some occasional errors in the debug  log about my wired
ethernet interfaces:

May 19 07:16:11 rm_bob_2610 kernel: eth0: Transmit timeout, status 0c
0005 c07f media 18.
May 19 07:16:11 rm_bob_2610 kernel: eth0: Tx queue start entry 1185 
dirty entry 1181.
May 19 07:16:11 rm_bob_2610 kernel: eth0:  Tx descriptor 0 is 9008a06a.
May 19 07:16:11 rm_bob_2610 kernel: eth0:  Tx descriptor 1 is
9008a06a. (queue head)
May 19 07:16:11 rm_bob_2610 kernel: eth0:  Tx descriptor 2 is 9008a06a.
May 19 07:16:11 rm_bob_2610 kernel: eth0:  Tx descriptor 3 is 9008a06a.

 )

> What interests me is the very beginning of the logs, and around the time
> when the problem occurs. Aim is to know if the mgmt timeout appears
> because of a real problem in the mgmt queue device-wise, or (more
> likely) because we somehow screw up the mgmt queue in the driver. For
> instance, we may miss a mgmt frame...
> 
Ok, I'll let you know.

> Thanks for your patience !
>
Thank you.

-Bob