8139too & APIC incompatibility (2.6.0-test6-mm1, 2.4.20)

8139too & APIC incompatibility (2.6.0-test6-mm1, 2.4.20)

[Andreas Schwarz]

When APIC is activated, the following messages appear in syslog from
time to time:                                                                   
                                                                                
kernel: NETDEV WATCHDOG: eth1: transmit timed out
kernel: eth1: Tx queue start entry 4  dirty entry 0.
kernel: eth1:  Tx descriptor 0 is 00002000. (queue head)
kernel: eth1:  Tx descriptor 1 is 00002000.
kernel: eth1:  Tx descriptor 2 is 00002000.
kernel: eth1:  Tx descriptor 3 is 00002000.
kernel: eth1: link up, 100Mbps, full-duplex, lpa 0x45E1

After this has happened the first time, the card fails to send or
receive any more packages.

The problem exists in both tested Kernel versions (2.6.0-test6-mm1 and
2.4.20).

lspci -v:

00:0f.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ (rev 10)
        Subsystem: Realtek Semiconductor Co., Ltd. RT8139
        Flags: bus master, medium devsel, latency 64, IRQ 5
        I/O ports at d400 [size=256]
        Memory at cfffde00 (32-bit, non-prefetchable) [size=256]
        Expansion ROM at cffe0000 [disabled] [size=64K]
        Capabilities: [50] Power Management version 2

Re: 8139too & APIC incompatibility (2.6.0-test6-mm1, 2.4.20)

[Jeff Garzik]

On Mon, Sep 29, 2003 at 03:47:29PM +0000, Andreas Schwarz wrote:
> When APIC is activated, the following messages appear in syslog from
> time to time:                                                                   
>                                                                                 
> kernel: NETDEV WATCHDOG: eth1: transmit timed out
> kernel: eth1: Tx queue start entry 4  dirty entry 0.
> kernel: eth1:  Tx descriptor 0 is 00002000. (queue head)
> kernel: eth1:  Tx descriptor 1 is 00002000.
> kernel: eth1:  Tx descriptor 2 is 00002000.
> kernel: eth1:  Tx descriptor 3 is 00002000.
> kernel: eth1: link up, 100Mbps, full-duplex, lpa 0x45E1
> 
> After this has happened the first time, the card fails to send or
> receive any more packages.

Yes.  Has nothing to do with 8139too, though.

This is one of 1001 similar symptoms of the same problem, "interrupt
routing bug(s)".

	Jeff

Interrupt routing (was Re: 8139too & APIC incompatibility (2.6.0-test6-mm1, 2.4.20))

[Jonathan McDowell]

On Mon, 29 Sep 2003 16:59:32, Jeff Garzik wrote:
> On Mon, Sep 29, 2003 at 03:47:29PM +0000, Andreas Schwarz wrote:
[snip tranmit timed out error]
> > After this has happened the first time, the card fails to send or
> > receive any more packages.
> Yes.  Has nothing to do with 8139too, though.
> 
> This is one of 1001 similar symptoms of the same problem, "interrupt
> routing bug(s)".

Bugs where? Does this count as a hardware issue or a software issue? I'm
seeing similar problems with DLink 580TX boards (sundance driver) and
both SiS 730 and Intel Camino chipsets, under 2.4.21 (and .22). I've
seen a suggestion that ACPI might help, but no definite answers. Equally
no mention of what hardware is a good choice if that's the root cause.

J.

-- 
                 /------------------------------------\
                 |   Zed's dead, baby. Zed's dead.    |
                 | http://www.blackcatnetworks.co.uk/ |
                 \------------------------------------/