Bug #982

bge0 watchdog timeout

Added by jspringe over 6 years ago. Updated 3 months ago.

Status:ClosedStart date:
Priority:NormalDue date:
Assignee:sepherosa% Done:

0%

Category:Driver
Target version:3.8.0

Description

Hi there,
after playing around with dragonfly for quite some weeks I decided
it is time to set uo a production environment and I remembered an old IBM eServer 325
that I could use for this purpose.
Installing and compiling a SMP kernel worked flawlessly however there is one major bug:
The machine has two bge cards bge0 and bge1 the latter works fine but the first
always gives watchdog timeouts.
This happens regardless of wheather I am running a UP or SMP kernel.
Turnig off APIC_IO while using an SMP kernel also made no difference.
Any ideas to why this might be happening ?
Machines are down for maintenance and I have not yet grabbed a full dmesg,
sorry for that.
Tobias

History

#1 Updated by sepherosa over 6 years ago

On Thu, Mar 27, 2008 at 6:48 PM, Jost Tobias Springenberg
<> wrote:
> Hi there,
> after playing around with dragonfly for quite some weeks I decided
> it is time to set uo a production environment and I remembered an old IBM eServer 325
> that I could use for this purpose.
> Installing and compiling a SMP kernel worked flawlessly however there is one major bug:
> The machine has two bge cards bge0 and bge1 the latter works fine but the first

Mmm, are two bge same?
Have you tried polling?
try setting hw.bge0.tx_coal_ticks to 400 without polling enabled.
Default value _may_ be too big for certain bge chips.

Best Regards,
sephe

#2 Updated by jspringe over 6 years ago

Little bit of a delay here ...
I will try your suggestions and reply back mid next week.

On Thu, 27 Mar 2008 19:15:39 +0800
"Sepherosa Ziehau" <> wrote:

#3 Updated by sepherosa about 6 years ago

hand over to myself

#4 Updated by adamk about 5 years ago

I recently installed 2.2.1 on a Dell Precision 490. It has one bge0 nic. It
works fine with the GENERIC UP kernel. When I rebooted into an SMP kernel that
I built, though, I started getting the watchdog timeout errors. I found this
report and set hw.bge0.tx_coal_ticks to 400 in sysctl.conf, but that didn't
change anything.

Anything else I can test? The NIC in question:

bge0@pci11:0:0: class=0x020000 card=0x01c11028 chip=0x160014e4 rev=0x02 hdr=0x00
vendor = 'Broadcom Corporation'
device = 'BCM5752 NetXtreme BCM5752 Gigabit Ethernet PCI Express'
class = network
subclass = ethernet

#5 Updated by sepherosa about 5 years ago

On Wed, Jul 8, 2009 at 3:48 AM, Adam Kirchhoff (via DragonFly issue
tracker)<> wrote:
>
> Adam Kirchhoff <> added the comment:
>
> I recently installed 2.2.1 on a Dell Precision 490. It has one bge0 nic.  It
> works fine with the GENERIC UP kernel.  When I rebooted into an SMP kernel that
> I built, though, I started getting the watchdog timeout errors.  I found this
> report and set hw.bge0.tx_coal_ticks to 400 in sysctl.conf, but that didn't
> change anything.
>
> Anything else I can test?  The NIC in question:
>
> bge0@pci11:0:0: class=0x020000 card=0x01c11028 chip=0x160014e4 rev=0x02 hdr=0x00
>    vendor   = 'Broadcom Corporation'
>    device   = 'BCM5752 NetXtreme BCM5752 Gigabit Ethernet PCI Express'
>    class    = network
>    subclass = ethernet

Did you have APIC_IO in your kernel configure?
If "yes", then try removing APIC_IO and try again.

Best Regards,
sephe

#6 Updated by adamk about 5 years ago

Yes, rebuilding the SMP kernel without APIC_IO seems to have done it. Thanks.

Adam

#7 Updated by tuxillo 6 months ago

  • Description updated (diff)
  • Category set to Driver
  • Status changed from New to Feedback
  • Target version set to 3.8.0

Sephe,

After all the work you did on the SMP front and in the bge(4) driver, do you think this is still relevant?

Cheers,
Antonio Huete

#8 Updated by tuxillo 3 months ago

  • Status changed from Feedback to Closed
  • Assignee set to sepherosa

Hi,

Been using bge(4) for a long time in my previous NAS w/o any issues.
Also there has been a ton of work to the bge(4) driver since this ticket was opened.

So please open a new one if you run into problems again. Closing this one.

Best regards,
Antonio Huete

Also available in: Atom PDF