Re: Drivers situation

[Geert U. <ge...@li...>]

On Tue, 4 Jul 2000 fh...@at... wrote:
> Actually my last name is "Heitkamp".   I have yet to get the driver to
> pass the cache test.  I am not sure what is wrong, but Ken seems to think
> the CPU cache is getting in the way.  I have let the pieces that reserve
> the I/O space alone.  The driver uses check_region/request_region to
> reserve I/O space for the kernel, but that doesn't seem to allow the SCSI
> chip to access memory properly.  What I plan to do next is use the bits
> out of the 53C7xx driver to see if that makes any difference, unless
> someone has a better idea. 

{check,request}_region() just mark that that region is in use by a driver, it
doesn't allow access itself.

Since you're speaking about request_region(), this sounds like the chip uses
PCI I/O space? And to access PCI I/O space you have to know where PCI (ISA) I/O
space is mapped (isa_io_base). AFAIK, it doesn't exist on APUS.

Now forget the previous paragraph, since the 53c7xx is not for PCI! AFAIK it's
meant to be connected directly to the CPU bus. Hence you should not use
request_region() at all, but request_mem_region(). And to make everything I
said above even more useless: you don't need request_mem_region() to get
anything to work neither (but you do want request_mem_region() later).

> Yes the code is written in a completely different style from that of the
> 53c7xx driver but it appears to be a close cousin of the  driver used in
> the sym53c8xx driver, which is used for a number of  cards for the PC. 
> Alas they are all ISA or PCI cards.  I am trying to  find someone who can
> tell me the major differences between

Not only on the PC. I have a plain sym53c875 PCI card in my CHRP box, which
works fine (17 MB/s on my Quantum Viking II).

> accessing cards on the Amiga vs. the PC.  If the driver ever does get to

On PC, the sym53c8xx driver can use both PCI I/O space and PCI memory space,
configurable through a CONFIG_* option. On some platforms only one of them
work fine.

On Amiga, you should just read/write to the correct memory addresses.

> work, I would eventually like to see it combined with the sym53c8xx.c in
> order to take advantage of the active development. If that's all possible
> of course.

Should be possible, since 53c7xx and 53c8xx are close cousins.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- Linux/{m68k~Amiga,PPC~CHRP} -- ge...@li...

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds