[parisc-linux] EISA stuff now in CVS

[parisc-linux] EISA stuff now in CVS

[Helge Deller]

Hi all,
I've just committed Daniel Engstrom's <5116@telia.com> EISA patches
to CVS head. Get yourself 2.4.9-pa79 and try yourself :-)

I don't have any EISA cards, so I tested it here with an
old ISA NE2000 NIC clone in my 715/64:

dmesg:
ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
Last modified Nov 1, 2000 by Paul Gortmaker
NE*000 ethercard probe at 0x340: 08 00 00 50 11 13
eth0: NE2000 found at 0x340, using IRQ 10.

pa64:~# ping www.kde.org
PING www.kde.org (213.203.58.36): 56 data bytes
64 bytes from 213.203.58.36: icmp_seq=0 ttl=237 time=149.7 ms
64 bytes from 213.203.58.36: icmp_seq=1 ttl=237 time=108.7 ms

--- www.kde.org ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss
round-trip min/avg/max = 108.7/129.2/149.7 ms

pa64:~# cat /proc/interrupts
 IRQ      count         Region    ISR
  2:          0            EISA  cascade
 10:       1645            EISA  NE2000
 32:      93357       PA-CPU-00  timer
 33:       8865       PA-CPU-00  lasi
 39:       1647       PA-CPU-00  wax
 69:          1            Lasi  keyboard, psaux
 86:       7852            Lasi  lasi710
 88:          0            Lasi  parport0
 90:       1008            Lasi  serial
117:       1645             Wax  EISA
125:          0             Wax  wax_hil_nmi
126:          2             Wax  hil

Greetings,
Helge

Re: [parisc-linux] EISA stuff now in CVS

[Jochen Friedrich]

Hi Helge,

> I don't have any EISA cards, so I tested it here with an
> old ISA NE2000 NIC clone in my 715/64:
> 
> dmesg:
> ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
> Last modified Nov 1, 2000 by Paul Gortmaker
> NE*000 ethercard probe at 0x340: 08 00 00 50 11 13
> eth0: NE2000 found at 0x340, using IRQ 10.
> 
> pa64:~# ping www.kde.org
> PING www.kde.org (213.203.58.36): 56 data bytes
> 64 bytes from 213.203.58.36: icmp_seq=0 ttl=237 time=149.7 ms

Pretty strange you're getting any interrupts, as with the current patch 
the EISA PIC isn't initialized, at all...

Index: gsc/eisa.c
===================================================================
RCS file: /var/cvs/linux/drivers/gsc/eisa.c,v
retrieving revision 1.12
diff -u -r1.12 eisa.c
--- gsc/eisa.c	2001/11/08 00:12:40	1.12
+++ gsc/eisa.c	2001/11/09 21:07:19
@@ -336,8 +336,6 @@
 	}
 	pcibios_register_hba(&eisa_dev.hba);
 
-	init_eisa_pic();
-	
 	result = request_irq(dev->irq, eisa_irq, SA_SHIRQ, "EISA", NULL);
 	if (result) {
 		printk(KERN_ERR "EISA: request_irq failed!\n");
@@ -352,6 +350,7 @@
 	irq_region[0] = &eisa_irq_region;
 	
 	EISA_bus = 1;
+	init_eisa_pic();
 	
 	return 0;
 }

Re: [parisc-linux] EISA stuff now in CVS

[Helge Deller]

On Friday 09 November 2001 22:08, Jochen Friedrich wrote:
> Pretty strange you're getting any interrupts, as with the current patch
> the EISA PIC isn't initialized, at all...

Hi Jochen,
Daniel's newest patch is in CVS only. Update your tree or download the whole
kernel source from ftp://parisc-linux.org/cvs

Greetings,
Helge

Re: [parisc-linux] EISA stuff now in CVS

[Jochen Friedrich]

Hi Helge,

> Hi Jochen,
> Daniel's newest patch is in CVS only. Update your tree or download the whole
> kernel source from ftp://parisc-linux.org/cvs

That's what i did and checked. 

However in the CVS version, init_eisa_pic() is called before EISA_bus is 
set to 1, so all the initialisation isn't really written to the chip...

That's what my patch fixes.

Cheers,
Jochen

Re: [parisc-linux] EISA stuff now in CVS

[Helge Deller]

On Friday 09 November 2001 23:47, Jochen Friedrich wrote:
> Hi Helge,
>
> > Hi Jochen,
> > Daniel's newest patch is in CVS only. Update your tree or download the
> > whole kernel source from ftp://parisc-linux.org/cvs
>
> That's what i did and checked.
>
> However in the CVS version, init_eisa_pic() is called before EISA_bus is
> set to 1, so all the initialisation isn't really written to the chip...
>
> That's what my patch fixes.

Ooops....
Yes, I mixed that up. I'll check and fix that.
Now I'm wondering myself how it could have worked here. Maybe I just was lucky 
and the PDC in the 715/64 already did the proper initialization itself ?

Helge

Re: [parisc-linux] EISA stuff now in CVS

[Grant Grundler]

Helge Deller wrote:
> Now I'm wondering myself how it could have worked here. Maybe I just
> was lucky 
> and the PDC in the 715/64 already did the proper initialization itself ?

lucky most likely. PDC/IODC don't  use interrupts.
I've been told all firmware IO is polled.

grant

Re: [parisc-linux] EISA stuff now in CVS

[Tom]

> > Helge:
> > I don't have any EISA cards, so I tested it here with an
> > old ISA NE2000 NIC clone in my 715/64:
> >
> > dmesg:
> > ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
> > Last modified Nov 1, 2000 by Paul Gortmaker
> > NE*000 ethercard probe at 0x340: 08 00 00 50 11 13
> > eth0: NE2000 found at 0x340, using IRQ 10.
>
> On Fri, 9 Nov 2001, Jochen Friedrich wrote:
> Pretty strange you're getting any interrupts, as with the current patch
> the EISA PIC isn't initialized, at all...

Not that strange, I'm thinking. Lots of ISA cards have the IO and IRQ
hard-coded on them, either via jumper or logically. I'd guess this NIC is
advertising it's presence and resources used, so the system is picking up
on it. Impressive that it's working, though...

BTW, how stands the 'apt-get update' fixes? Are we safe to go, or do we
need another day or two for things to spread about?

Re: [parisc-linux] EISA stuff now in CVS

[Helge Deller]

> > On Fri, 9 Nov 2001, Jochen Friedrich wrote:
> > Pretty strange you're getting any interrupts, as with the current patch
> > the EISA PIC isn't initialized, at all...
>
> Not that strange, I'm thinking. Lots of ISA cards have the IO and IRQ
> hard-coded on them, either via jumper or logically. I'd guess this NIC is
> advertising it's presence and resources used, so the system is picking up
> on it. Impressive that it's working, though...

It's even better :-)
This card uses the ne2.c driver and this driver probes all adresses 0x300, 0x320, 0x340, ...
and then uses Linux's irq autodetection code. Nothing hardcoded here.

Helge