Re: intel 82571EB gigabit fails to see link on 2.6.20-rc5 in-tree e1000 driver (regression)

[Jesse Brandeburg]

On 1/16/07, Allen Parker <parker@xxxxxxxxxxx> wrote:
> Allen Parker wrote:
> > I have a PCI-E pro/1000 MT Quad Port adapter, which works quite well
> > under 2.6.19.2 but fails to see link under 2.6.20-rc5. Earlier today I
> > reported this to e1000-devel@xxxxxxxxxxxx, but thought I should get the
> > word out in case someone else is testing this kernel on this nic chipset.

Upon some further investigation with Allen, I got this info, and it
appears that his system is not freeing MSI irq's correctly.

from our discussion:
> Allen wrote:
> > Jesse Brandeburg wrote:
> > I believe you may have an interrupt delivery problem, due to kernel
> > changes.  When you don't get interrupts delivered correctly, you may
> > not achieve link up.
> >
> > you can try disabling CONFIG_PCI_MSI in the kernel, as another
> > trouble shooting step.  Also, what model motherboard/machine are you
> > using, and please check to make sure your BIOS is up to date.
>
> CONFIG_PCI_MSI seems to fix it as well, however, i'm not sure about
> possible performance implications with leaving it turned off.
>
> Tyan S2927G2NR w/ 2x Opteron 2210 bios rev 1.02 iirc

Also, here is an excerpt of his his proc/interrupts
Allen Parker wrote:
      CPU0       CPU1       CPU2       CPU3
498:      1          8        129       8723   PCI-MSI-edge      eth3
499:      0         42        122      17135   PCI-MSI-edge      eth2
500:      0         19        275    1418326   PCI-MSI-edge      eth1
501:      0          0          0          0   PCI-MSI-edge      eth1
502:      0        414         26      25227   PCI-MSI-edge      eth1
503:      0         37        264    1418695   PCI-MSI-edge      eth0
504:      0          0          0        139   PCI-MSI-edge      eth0
505:      0          6        714      22806   PCI-MSI-edge      eth0
NMI:    275        207        225        262
LOC:3531851    3531829    3531805    3531778
ERR:      0

The disturbing bit is that on this Hypertransport to PCIe system, his
interrupts appear to be getting registered multiple times on different
IRQ numbers.

help?
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/