I've plugged for testing a 3C996-T in a 32-bit slot on my Athlon system
lspci -v gives for the unused card:
00:09.0 Ethernet controller: BROADCOM Corporation NetXtreme BCM5700 Gigabit Ethernet (rev 12)
Subsystem: 3Com Corporation 3C996-T 1000BaseTX
Flags: bus master, 66Mhz, medium devsel, latency 64, IRQ 11
Memory at effe0000 (64-bit, non-prefetchable) [size=64K]
Capabilities: [40] #07 [0002]
Capabilities: [48] Power Management version 2
Capabilities: [50] Vital Product Data
Capabilities: [58] Message Signalled Interrupts: 64bit+ Queue=0/3 Enable-
If I access the card with the broadcom driver bcm5700 from 2.4.18
everything works fine (just brief testing). But using the Tigeon
v0.97 I cannot send any data to the network.
A modprobe of bcm5700 gives:
Broadcom Gigabit Ethernet Driver bcm5700 with Broadcom NIC Extension (NICE) ver. 2.2.8 (03/07/02)
AMD756: dev 14e4:1644, router pirq : 2 get irq : 11
PCI: Found IRQ 11 for device 00:09.0
Cannot get MAC addr from NVRAM. Using default.
VPD read failed
eth1: 3Com 3C996 10/100/1000 Server NIC found at mem effe0000, IRQ 11, node addr 001018686176
eth1: Broadcom BCM5401 Copper transceiver found
eth1: Scatter-gather ON, 64-bit DMA OFF, Tx Checksum ON, Rx Checksum ON
A modprobe of tg3 gives:
tg3.c:v0.97 (Mar 13, 2002)
AMD756: dev 14e4:1644, router pirq : 2 get irq : 11
PCI: Found IRQ 11 for device 00:09.0
eth1: Tigon3 [partno(BCM95700A6) rev 7102 PHY(5401)] (PCI:33MHz:32-bit) 10/100/1000BaseT Ethernet 00:04:76:3b:39:c7
And after ifconfig eth1 I get:
eth1: Link is up at 100 Mbps, full duplex.
eth1: Flow control is on for TX and on for RX.
But no packets go over the card - and sometimes the whole system
freezes. Is there some kind of problem with plugging the card into a
32-bit slot? The same card worked with no problems in a 64-bit slot..
Any further debugging that I can do?
Andreas
--
Andreas Jaeger
SuSE Labs [email protected]
private [email protected]
http://www.suse.de/~aj
It's an amd756 chipset bug. bcm5700 chooses to work around it in
their driver, when it really belongs as a generic PCI fixup in
the kernel.
"David S. Miller" <[email protected]> writes:
> It's an amd756 chipset bug. bcm5700 chooses to work around it in
> their driver, when it really belongs as a generic PCI fixup in
> the kernel.
What needs to be done for this? Can you point me to the PCI
workaround in the bcm driver?
I only found via grep in the sources a workaround for the AMD 762
northbridge but nothing directly for the 756.
Thanks,
Andreas
--
Andreas Jaeger
SuSE Labs [email protected]
private [email protected]
http://www.suse.de/~aj
David S. Miller wrote:
>It's an amd756 chipset bug. bcm5700 chooses to work around it in
>their driver, when it really belongs as a generic PCI fixup in
>the kernel.
>
bcm5700 works around AMD762 bug -- and that workaround should be in
stock 2.4.18 and 2.5.7 kernels now as a PCI quirk. I think something
else is going on here...
Jeff
> >It's an amd756 chipset bug. bcm5700 chooses to work around it in
> >their driver, when it really belongs as a generic PCI fixup in
> >the kernel.
>
> bcm5700 works around AMD762 bug -- and that workaround should be in
> stock 2.4.18 and 2.5.7 kernels now as a PCI quirk. I think something
> else is going on here...
Its not IMHO a chipset bug either. The documentation on the 762 is
quite explicit. It has a nice table and makes it absolutely clear which modes
are PCI 2.2 compliant.
Alan
> I've plugged for testing a 3C996-T in a 32-bit slot on my Athlon system
> lspci -v gives for the unused card:
What chipset ?
Alan
Alan Cox wrote:
>>I've plugged for testing a 3C996-T in a 32-bit slot on my Athlon system
>>lspci -v gives for the unused card:
>>
>
>What chipset ?
>
Andreas' full lspci attached...
Alan Cox <[email protected]> writes:
>> I've plugged for testing a 3C996-T in a 32-bit slot on my Athlon system
>> lspci -v gives for the unused card:
>
> What chipset ?
Gigabyte GA 7IXE,
Andreas
--
Andreas Jaeger
SuSE Labs [email protected]
private [email protected]
http://www.suse.de/~aj