[parisc-linux] Recent CVS kernel: boot failure on 715/50 & inb() undefined

[parisc-linux] Recent CVS kernel: boot failure on 715/50 & inb() undefined

[Daniel Engstrom]

Since about yesterday CVS kernels seems to stop booting after the
traps_init message.

Hardware: 715/50, serial console. I can remeber this happening before 
but it was 'fixed' by disabling support for PCI chipsets.

Also: I had to tweak the kernel a bit so it would link, because inb()
and outb() was undefined (referenced in driverd/cher/mem.c and
drivers/char/serial.c). 

I could see that the code in mem.c had some '#if !defined(__mc68000__)'
around them, perhaps there should be a CONFIG_PORT_IO #define to test for
which could be undefined on machines such as the 715 without a wax driver.

Or maybe inb() and outb() should always be defined?

/Daniel  

-- 

Re: [parisc-linux] Recent CVS kernel: boot failure on 715/50 & inb() undefined

[Grant Grundler]

Daniel Engstrom wrote:

> Or maybe inb() and outb() should always be defined?

inb/outb need to be defined for the type of I/O bus being supported.

PArisc doesn't have inb/outb instructions like x86.
PCI bus support is platform specific - ie different depending on HBA.
GSC bus support requires faking "I/O port space".
EISA is similar to PCI in this regard.

But HP does have systems with GSC, EISA, and PCI in the same machine.
Right now, only PCI defines a dynamic interface to support all three
at the same time. This "design" definitely needs more work for GSC/EISA
bus support if you (or anyone else) is interested.

grant

Grant Grundler
Unix Systems Enablement Lab
+1.408.447.7253

Re: [parisc-linux] Recent CVS kernel: boot failure on 715/50 & inb() undefined

[Matthew Wilcox]

On Tue, Dec 05, 2000 at 11:18:24PM +0100, Daniel Engstrom wrote:
> Also: I had to tweak the kernel a bit so it would link, because inb()
> and outb() was undefined (referenced in driverd/cher/mem.c and
> drivers/char/serial.c). 

OK, this is Alan Cox's fault :-)  He insists that inb et al should not
have dummy defines, but the drivers should be fixed instead.  please send
your .config.

> I could see that the code in mem.c had some '#if !defined(__mc68000__)'
> around them, perhaps there should be a CONFIG_PORT_IO #define to test for
> which could be undefined on machines such as the 715 without a wax driver.

possibly, it's difficult to say wihout knowing exactly what's going wrong.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] Recent CVS kernel: boot failure on 715/50 & inb() undefined

[Alan Cox]

> OK, this is Alan Cox's fault :-)  He insists that inb et al should not
> have dummy defines, but the drivers should be fixed instead.  please send
> your .config.

I am allowed to be wrong here. I would like to take a look at the specific
drivers if you can tell me which before you leave.

Re: [parisc-linux] Recent CVS kernel: boot failure on 715/50 & inb() undefined

[Alan Cox]

> OK, this is Alan Cox's fault :-)  He insists that inb et al should not
> have dummy defines, but the drivers should be fixed instead.  please send
> your .config.

Oh wait ignore my other comment. I realised the context this one was meant.
You want it to fail to build drivers for things you cannot have, not 
mysteriously fail in some driver thats mostly mmio but has an inb/inw you
didnt notice..

> 
>