[parisc-linux] [Q] on EISA support

[parisc-linux] [Q] on EISA support

[Christian Suder]

Hello,

I finally got the C110 running a couple of days ago (thanks to the "magic" 
FW-SCSI command) and things are well so far. I'm trying to add a network 
card needed for the planned use of the machine but so far no luck.

I'm not an expert on the HP HW, but I do remember the (short) EISA days on 
PCs where a .cfg file was needed to teach the PC the specifics of the cards 
(IO ports, interrupts etc.). How does this work with the HP machines ? I 
tried an SMC Ultra 32 and an AC3200 with no luck. One hurdle here is that 
there is a function call in both drivers (as well as other drivers that use 
similar controller chips) that the compiler barks about. (Something like 
isa_eth_copy_and_chksm - don't have the exact at hand right now). I found a 
way around it but the cards are not detected at all. Note these cards use a 
memory window - is that possible with the current EISA support ? Is it 
possible to use generic EISA cards at all or do they have to be supported 
by the boot rom in some way ?

Tried a NE2000 clone too, which was detected but with the wrong interrupt. 
That might be because it's a "hard" configured plug&pray card. PCs have no 
problems detecting the settings correctly, though.

Thanks,
	Christian

Re: [parisc-linux] [Q] on EISA support

[Helge Deller]

On Wednesday 26 December 2001 23:13, Christian Suder wrote:
> Tried a NE2000 clone too, which was detected but with the wrong interrupt.
> That might be because it's a "hard" configured plug&pray card. PCs have no
> problems detecting the settings correctly, though.

Hi Christian,

Did you tried the "eisa_irq_edge=3,4,5,7,9,10,11,14,15" palo parameter ?
In case your NE2000 card is an ISA card you need to set the EISA irq's to
edge mode which is done with the above command.
Maybe we should think about adding a /proc-interface here ?

Greetings,
Helge 

Re: [parisc-linux] [Q] on EISA support

[Christian Suder]

On Wed, 26 Dec 2001 23:30:16 +0100
"Helge Deller" <deller@gmx.de> wrote:

> On Wednesday 26 December 2001 23:13, Christian Suder wrote:
> > Tried a NE2000 clone too, which was detected but with the wrong interrupt.
> > That might be because it's a "hard" configured plug&pray card. PCs have no
> > problems detecting the settings correctly, though.
> 
> Hi Christian,
> 
> Did you tried the "eisa_irq_edge=3,4,5,7,9,10,11,14,15" palo parameter ?
> In case your NE2000 card is an ISA card you need to set the EISA irq's to
> edge mode which is done with the above command.
> Maybe we should think about adding a /proc-interface here ?
> 

Thanks Helge,

that "basically" fixed it, i could use the card with 2.4.16-pa23. However as X has problems with this kernel (X server starts but graphics are garbled). I compiled 2.4.16-pa25, and this is what I get: 

Dec 26 20:45:04 c110-linux kernel: Wax EISA Adapter found at 0xfc000000
Dec 26 20:45:04 c110-linux kernel: Warning : device (11, 0x1e, 0x0, 0x90) NOT cl
aimed by EISA Bus Adapter

and then : 
Dec 26 20:45:04 c110-linux kernel: PCI or EISA Host Bus Adapter 0 not registered
. in8(0x300) returning -1
Dec 26 20:45:04 c110-linux kernel: 
Dec 26 20:45:04 c110-linux kernel: PCI or EISA Host Bus Adapter 0 not registered
. in8(0x280) returning -1
Dec 26 20:45:04 c110-linux kernel:
Dec 26 20:45:04 c110-linux kernel: PCI or EISA Host Bus Adapter 0 not registered
. in8(0x320) returning -1

The EISA support should be compiled in, at least I selected it. X is still garbled so no improvement in that area. 

BTW is there a compilation of all the "secret" boot parameters ? It's the second time now that it fixes a problem for me....


Danke nochmal,

	Christian 

Re: [parisc-linux] [Q] on EISA support

[Helge Deller]

On Thursday 27 December 2001 06:25, Christian Suder wrote:
> On Wed, 26 Dec 2001 23:30:16 +0100
>
> "Helge Deller" <deller@gmx.de> wrote:
> > On Wednesday 26 December 2001 23:13, Christian Suder wrote:
> > > Tried a NE2000 clone too, which was detected but with the wrong
> > > interrupt. That might be because it's a "hard" configured plug&pray
> > > card. PCs have no problems detecting the settings correctly, though.
> >
> > Hi Christian,
> >
> > Did you tried the "eisa_irq_edge=3,4,5,7,9,10,11,14,15" palo parameter ?
> > In case your NE2000 card is an ISA card you need to set the EISA irq's to
> > edge mode which is done with the above command.
> > Maybe we should think about adding a /proc-interface here ?
>
> Thanks Helge,
>
> that "basically" fixed it, i could use the card with 2.4.16-pa23. However
> as X has problems with this kernel (X server starts but graphics are
> garbled). I compiled 2.4.16-pa25, and this is what I get:
>
> Dec 26 20:45:04 c110-linux kernel: Wax EISA Adapter found at 0xfc000000
> Dec 26 20:45:04 c110-linux kernel: Warning : device (11, 0x1e, 0x0, 0x90)
> NOT cl aimed by EISA Bus Adapter
>
> and then :
> Dec 26 20:45:04 c110-linux kernel: PCI or EISA Host Bus Adapter 0 not
> registered . in8(0x300) returning -1
> Dec 26 20:45:04 c110-linux kernel:
> Dec 26 20:45:04 c110-linux kernel: PCI or EISA Host Bus Adapter 0 not
> registered . in8(0x280) returning -1
> Dec 26 20:45:04 c110-linux kernel:
> Dec 26 20:45:04 c110-linux kernel: PCI or EISA Host Bus Adapter 0 not
> registered . in8(0x320) returning -1
>
> The EISA support should be compiled in, at least I selected it. X is still
> garbled so no improvement in that area.
>
> BTW is there a compilation of all the "secret" boot parameters ? It's the
> second time now that it fixes a problem for me....

Hi Christian,

try the default config (remove your .config file and then run "make oldconfig").
It just booted here for me with -pa27, so I assume you may have forgotten the
CONFIG_GSC_WAX=y configuration entry. 
Regarding the "secret" boot parameters I think the eisa_irq_edge and some
SCSI variants are the only one we have. I will add them later today to the FAQ.

Greetings,
Helge

Re: [parisc-linux] [Q] on EISA support

[Jochen Friedrich]

Hi Christian,

> Note these cards use a memory window - is that possible with the
> current EISA support ?

There seems to be an issue with memory mapped windows currently. I have
similar problems with an token ring card which is detected correctly
(detection only uses port IO), but the memory window is only partially
seen by the CPU.

Cheers,
Jochen