I've been using the 2.6.17 series with "all-generic-ide irqpoll" on a
Gigabyte 965P-S3 motherboard [using the ICH8 chipset].
It worked semi decent [other than random interrupt 177 hehehe].
However, this morning I took the same .config and applied it to 2.6.18
with some nasties. It compiled fine, the disks work now (my SATA
drives also show up as /dev/sd* finally) but now the sky2 device
disappeared.
lspci can still see it, it's a Marvell "Uknown Device 4364, rev 12"
sitting off the PCI-E bus.
Odd...
Tom
Tom St Denis wrote:
> I've been using the 2.6.17 series with "all-generic-ide irqpoll" on a
> Gigabyte 965P-S3 motherboard [using the ICH8 chipset].
>
> It worked semi decent [other than random interrupt 177 hehehe].
>
> However, this morning I took the same .config and applied it to 2.6.18
> with some nasties. It compiled fine, the disks work now (my SATA
> drives also show up as /dev/sd* finally) but now the sky2 device
> disappeared.
>
> lspci can still see it, it's a Marvell "Uknown Device 4364, rev 12"
> sitting off the PCI-E bus.
This device was not listed in the 2.6.17 sky2 driver and is not included
in the 2.6.18 version either. Either you modified your 2.6.17 kernel or
your distro did.
You can simply add the device ID to the list in 2.6.18. Patches to do
this are already on their way to mainline for 2.6.19.
Daniel
On 9/22/06, Daniel Drake <[email protected]> wrote:
> This device was not listed in the 2.6.17 sky2 driver and is not included
> in the 2.6.18 version either. Either you modified your 2.6.17 kernel or
> your distro did.
It is detected with the latest 2.6.17 kernel automatically [during
udev] and works fine.
> You can simply add the device ID to the list in 2.6.18. Patches to do
> this are already on their way to mainline for 2.6.19.
This won't be fixed as part of 2.6.18.x? Looking at the source for
the latest gentoo-sources in the 2.6.17 stream I don't see ID 4364 in
the source for sky2.c [sky2_id_table[]]. So why is it detected and
working there but not in 2.6.18?
I'll try adding the device to the table and see what happens.
Hint: ICH8 was "fixed" between 2.6.17 and 2.6.18 and all of a sudden
my gige stops working which is attached off the ICH8.
Tom
On 9/22/06, Tom St Denis <[email protected]> wrote:
> This won't be fixed as part of 2.6.18.x? Looking at the source for
> the latest gentoo-sources in the 2.6.17 stream I don't see ID 4364 in
> the source for sky2.c [sky2_id_table[]]. So why is it detected and
> working there but not in 2.6.18?
>
> I'll try adding the device to the table and see what happens.
Adding the line to the table works now. [yipee, I get proper SATA and
net now woot!]
May I request that someone insert that one line to drivers/net/sky2.c
before 2.6.19?
--- sky2.old 2006-09-22 21:34:36.000000000 +0000
+++ sky2.c 2006-09-22 21:28:03.000000000 +0000
@@ -121,6 +121,7 @@
{ PCI_DEVICE(PCI_VENDOR_ID_MARVELL, 0x4361) },
{ PCI_DEVICE(PCI_VENDOR_ID_MARVELL, 0x4362) },
{ PCI_DEVICE(PCI_VENDOR_ID_MARVELL, 0x4363) },
+ { PCI_DEVICE(PCI_VENDOR_ID_MARVELL, 0x4364) },
{ 0 }
};
[yeah I know not a normal patch ... I'm also not a kernel developer...]
I'm still confused on how it works in 2.6.17 without the ID?
Thanks,
Tom
Tom St Denis wrote:
> This won't be fixed as part of 2.6.18.x?
Probably not.
> So why is it detected and
> working there but not in 2.6.18?
It wasn't detected under 2.6.17. Either your kernel is modified, or you
were using the vendor sk98lin driver or something like that. If you have
2.6.17 still bootable you could boot it and check the dmesg output to
make sense of things.
Daniel
On 9/18/06, Daniel Drake <[email protected]> wrote:
> Tom St Denis wrote:
> > This won't be fixed as part of 2.6.18.x?
>
> Probably not.
It's a one line fix!!! :-/
> It wasn't detected under 2.6.17. Either your kernel is modified, or you
> were using the vendor sk98lin driver or something like that. If you have
> 2.6.17 still bootable you could boot it and check the dmesg output to
> make sense of things.
No, i get sky2 printk after udev kicks in. It was a 2.6.17-gentoo-r8
kernel which has devices 4364 through 4368 [just checked].
This means my patch is incomplete, and the fix Gentoo made against
2.6.17 didn't make it into 2.6.18. ARRG.
Tom
Tom St Denis wrote:
> No, i get sky2 printk after udev kicks in. It was a 2.6.17-gentoo-r8
> kernel which has devices 4364 through 4368 [just checked].
So your 2.6.17 kernel *is* modified :)
> This means my patch is incomplete, and the fix Gentoo made against
> 2.6.17 didn't make it into 2.6.18. ARRG.
I maintain the Gentoo kernel and yes we do include sky2 changes,
sometimes before they make it to Linus for whatever reason, but always
only if they are in the netdev tree.
If you had switched from gentoo-sources-2.6.17 to gentoo-sources-2.6.18
you would not have noticed any loss of functionality since the patches
are obviously carried across. The difference appears to be that you went
from gentoo-2.6.17 to vanilla-2.6.18...
Daniel