pci_set_method() and bochs

[Andrew Kohlsmith]

I am debugging my bochs-makes-the-bios-jump-off-to-never-never-land problem 
and have just realized that the linuxbios_c.map file shows me the RAM 
addresses (bliss!).

I've been tracking this back and have traced it down to lib/newpci.c, line 
274, where a far call to a null pointer seems to occur because struct pci_ops 
*conf is never assigned a value.

pci_set_method() assigns it a value, but under bochs neither PCI direct method 
1 nor 2 works.  I suppose I could hack around it but I'm wondering if anyone 
has done anything a little more slick than that.  I'm reluctant to just flash 
one of these boards because if she don't boot, I have to unsolder a TQFP, buy 
an adaptor to reflash it, (probably buy a TQFP socket too while I'm at it), 
and try again.  :-)

Regards,
Andrew

[Gregg C Levine]

Hello from Gregg C Levine
Well, I can confirm that it works here. Andrew, check to see if you
remembered to install the PCI utilities for your distribution. I
neglected to do so at first, that caused it to pull the same fault that
you reported, then after installing it, it ran to completion, it now
wants me to attach a kernel to boot. Any suggestions? Such as what
kernel version you are using? I chose arbitrarily the 2.4.17 version,
and added the usual patch series for the version. Also, for what you
sent me privately can you confirm what date you chose for the CVS drop?
Basically your image runs to completion, with the message that it can't
boot an elf image. 
-------------------
Gregg C Levine [e-mail]
------------------------------------------------------------
"The Force will be with you...Always." Obi-Wan Kenobi
"Use the Force, Luke."  Obi-Wan Kenobi
(This company dedicates this E-Mail to General Obi-Wan Kenobi )
(This company dedicates this E-Mail to Master Yoda )
> -----Original Message-----
> From: [e-mail] [mailto:linuxbios-
> [e-mail]] On Behalf Of Andrew Kohlsmith
> Sent: Monday, November 18, 2002 8:51 PM
> To: LinuxBIOS Mailing List
> Subject: pci_set_method() and bochs
>
> I am debugging my bochs-makes-the-bios-jump-off-to-never-never-land

problem
> and have just realized that the linuxbios_c.map file shows me the RAM
> addresses (bliss!).
>
> I've been tracking this back and have traced it down to lib/newpci.c,

line
> 274, where a far call to a null pointer seems to occur because struct

pci_ops
> *conf is never assigned a value.
>
> pci_set_method() assigns it a value, but under bochs neither PCI

direct method
> 1 nor 2 works.  I suppose I could hack around it but I'm wondering if

anyone
> has done anything a little more slick than that.  I'm reluctant to

just flash
> one of these boards because if she don't boot, I have to unsolder a

TQFP, buy
> an adaptor to reflash it, (probably buy a TQFP socket too while I'm at

it),
> and try again.  :-)
>
> Regards,
> Andrew
> _______________________________________________
> Linuxbios mailing list
> [e-mail]
> http://www.clustermatic.org/mailman/listinfo/linuxbios

[Andrew Kohlsmith]

> Well, I can confirm that it works here. Andrew, check to see if you
> remembered to install the PCI utilities for your distribution. I

_which_ PCI utilities?  I do have lspci and friends, but I'm not sure what 
other utilities you are referring to.
> neglected to do so at first, that caused it to pull the same fault that
> you reported, then after installing it, it ran to completion, it now
> wants me to attach a kernel to boot. Any suggestions? Such as what

That's where I'm at.  I set a breakpoint at pci_set_method to override its 
setting conf to NULL, then I set a breakpoint at elfboot() and try to see 
where it's dying.

It seems that elfboot searches 0xa0000 --> ???? for an ELF header.  In my case 
it never seems to find it.  :-(

BTW: printk_spew() seems to be a null (i.e. empty) function/macro.  The 
messages from it never appear.
> kernel version you are using? I chose arbitrarily the 2.4.17 version,
> and added the usual patch series for the version. Also, for what you
> sent me privately can you confirm what date you chose for the CVS drop?
> Basically your image runs to completion, with the message that it can't
> boot an elf image.

I am using 2.4.19 with the sis patch, although I don't think that's necessary.  
I'm not going ot be booting a kernel since I don't have enough memory for 
that (I only have 256k FLASH, although as Ron said, I may be able to just 
swap the chip out if the address lines are there).  I put an etherboot elf 
image at 0xc0000, but I think what needs to happen is configure linuxbios to 
boot from IDE instead of do it through etherboot (I have a CompactFlash card 
located at ide0 master (hda)).

I pulled linuxbios from CVS 4pm 10 Nov 2002.

Regards,
Andrew

[Ronald G. Minnich]

and, when you get this done, I'll take a HOWTO :-)

ron

[Ronald G. Minnich]

We can initialize that to something at compile time I suppose. Pick one 
:-)

ron

[Andrew Kohlsmith]

> We can initialize that to something at compile time I suppose. Pick one

It is initialized to method 1, but it's zapped my pci_check_direct.

How is this? (attached)

Regards,
Andrew