mainboard/advantech/pcm-5823/config.example - added options for rtl8139 and eepro100 mainboard/advantech/pcm-5823/mainboard.c - assigning irqs for ethernet ports and added option for rtl8139 and eepro100 southbridge/nsc/cs5530/southbridge.c - usb uses INTC instead of INTA
Ron,would you add the eepro portion in mainboard/advantech/pcm-5823/mainboard.c, since I don't have pcm-5823 with eepro board. Thanks.
-Andrew
On Mon, 11 Nov 2002, Andrew Ip wrote:
mainboard/advantech/pcm-5823/config.example
- added options for rtl8139 and eepro100
mainboard/advantech/pcm-5823/mainboard.c
- assigning irqs for ethernet ports and added option for rtl8139 and eepro100
why did you need to do this? linux was doing this work just fine either way?
ron
On Mon, 11 Nov 2002, Ronald G. Minnich wrote:
mainboard/advantech/pcm-5823/config.example
- added options for rtl8139 and eepro100
mainboard/advantech/pcm-5823/mainboard.c
- assigning irqs for ethernet ports and added option for rtl8139 and eepro100
why did you need to do this? linux was doing this work just fine either way?
browsing cvs at the sourceforge I don't see these changes anyway. What's going on ...
ron
Ron,
why did you need to do this? linux was doing this work just fine either way?
Mine doesn't behave like this. ethernet doesn't get irq without those code.
-Andrew
Andrew Ip aip@cwlinux.com writes:
Ron,
why did you need to do this? linux was doing this work just fine either way?
Mine doesn't behave like this. ethernet doesn't get irq without those code.
Does someone want to write some generic irq assignment code, that we can report in the pci registers? I think in general this is something we want to do even if linux let's us get away with it most of the time. Other OS's that are less hotplug friendly won't and it shouldn't be that much code to write.
Eric
Does someone want to write some generic irq assignment code, that we can report in the pci registers? I think in general this is something we want to do even if linux let's us get away with it most of the time. Other OS's that are less hotplug friendly won't and it shouldn't be that much code to write.
I wouldn't mind to take it, 'coz I have sooo many irq problem lately.
-Andrew
On 11 Nov 2002, Eric W. Biederman wrote:
Does someone want to write some generic irq assignment code, that we can report in the pci registers? I think in general this is something we want to do even if linux let's us get away with it most of the time. Other OS's that are less hotplug friendly won't and it shouldn't be that much code to write.
It would be useful, but .... Andrew's problem is something else. It is almost useful in this case to have the error propagate to Linux since it is revealing a problem.
I wonder if Adam Sulmicki has done IRQ initialization support since they support many OSes.
Also, generic code won't solve the special-cases we are seeing. But it is still worth having. It would make our Plan 9 work much easier, since Plan 9's IRQ code doesn't work correctly in several cases.
ron
I wonder if Adam Sulmicki has done IRQ initialization support since they support many OSes.
For now it just "standard" PIC classic programming and that's it
It is fine as far as goes booting of linux/openbsd/lilo/gurb. It seems it even works for Win98.
Although I suepct It is not enough to get WinXP and Win2k to boot.
Coincidentally just yesterday we were looking at how to get the Int15/EAH=B1xx services implemented. Basically the info needs to be copied over from LinuxBIOS. It is that just someone just got sit down and implement that.
obtw : in reference to my lats email "warm boot" means shotcutting reset pins on the motherboard. So it is not a "power cycle", nor just jumping to start of LinuxBIOS.
Looks like I have located the problem. The reason of the ethernet can't get irq is because pirq routing table is not written properly. If option CONFIG_COMPRESS is set to 0, then everything back to normal. compression may affect pirq routing table area. Also, 2.4.17 and 2.4.19 also have different behavior. On 2.4.19, ping doesn't work even it has an irq. However, if writing an irq to PCI_INTERRUPT_LINE in mainboard.c, then ping works. 2.4.17 works either way.
-Andrew
On Tue, Nov 12, 2002 at 07:54:29AM -0700, Ronald G. Minnich wrote:
On 11 Nov 2002, Eric W. Biederman wrote:
Does someone want to write some generic irq assignment code, that we can report in the pci registers? I think in general this is something we want to do even if linux let's us get away with it most of the time. Other OS's that are less hotplug friendly won't and it shouldn't be that much code to write.
It would be useful, but .... Andrew's problem is something else. It is almost useful in this case to have the error propagate to Linux since it is revealing a problem.
I wonder if Adam Sulmicki has done IRQ initialization support since they support many OSes.
Also, generic code won't solve the special-cases we are seeing. But it is still worth having. It would make our Plan 9 work much easier, since Plan 9's IRQ code doesn't work correctly in several cases.
ron
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
On Wed, 13 Nov 2002, Andrew Ip wrote:
Looks like I have located the problem. The reason of the ethernet can't get irq is because pirq routing table is not written properly. If option CONFIG_COMPRESS is set to 0, then everything back to normal. compression may affect pirq routing table area. Also, 2.4.17 and 2.4.19 also have different behavior. On 2.4.19, ping doesn't work even it has an irq. However, if writing an irq to PCI_INTERRUPT_LINE in mainboard.c, then ping works. 2.4.17 works either way.
2.4.19 is broken. I posted the fix here a few days ago and I can send it to you.
I'm glad compression is an option :-)
interesting problem and thanks for the fix. Here is an interesting case where having linuxbios do less (letting Linux parse the IRQ table) revealed a bug in linuxbios.
ron
-
Adam Sulmicki -
Andrew Ip -
ebiederman＠lnxi.com -
Ronald G. Minnich