Re: MPC8250 + PCMCIA + airo_cs card + IO Port mapping proble

On Thu, 2003-05-22 at 11:35, Jerry Van Baren wrote:

Kernel memory space starts at 0xC0000000.  You need to set your I/O well
above this (above the kernel executable).  It is illegal to have I/O below
the kernel (its a memory management issue).  That is why the common IMMR
setting is 0xF0000000.  I would suggest you try setting your PCI window at
0xF8000000 or something like that.

gvb

Ok, thanks for the pointer. I've scooted my PCI IO space up in memory to
where I had room.

This is how my IO is now setup:

    pci_init_resource(&hose->io_resource,
                       0xd0000000,
                       0xdfffffff,
                       IORESOURCE_IO,
                       "PCI host bridge");

    pci_init_resource (&hose->mem_resources[0],
                        0x80000000,
                        0x803fffff,
                        IORESOURCE_MEM,
                        "PCI host bridge");

     hose->io_space.start = 0xd0000000;
     hose->io_space.end = 0xdfffffff;
     hose->mem_space.start = 0x7ff00000;
     hose->mem_space.end = 0x803fffff;
     hose->io_base_virt = (void *)0;

my _IO_BASE is now 0xd0000000 and I do

io_block_mapping(0xd0000000, 0xd0000000, 0x1000000, _PAGE_IO);

early when the board is getting setup.


I've also setup PCIBR0 and PCIBR1 in the boot loader to have windows of
0x80000000 to 0xbfffffff and 0xd0000000 to 0xdfffffff.


Now when I try to write to the PCMCIA card, I don't see anything go out
on the PCMCIA bus (write enable and chipselect don't change).  A printk
in the out_le16() routine shows that the address seems to be correct
(0xd00000102), but when I try to read from the card I get a Machine
Check Exception that is "Transfer ack error". I suspect that my IO
mapping is still not correct.

I know that the board is wired up correctly because the I am able to
read the card's CIS through io memory. Why aren't my IO Ports working?


-- James

At 12:29 PM 5/22/2003 -0400, you wrote:

quoted

On Wed, 2003-05-21 at 21:33, Dan Malek wrote:

quoted

Jim Rowe wrote:

quoted

quoted

From what I can see - and please correct me if I'm wrong, the way that

Linux translates the Port IO to the memory mapped IO is through the outb
and outw macros, which add an offset of _IO_BASE. How do I determine the
correct value of _IO_BASE?

The value of _IO_BASE depends upon how you configure and map the host PCI
bridge on the 8250.

Here is how my PCI bridge gets mapped:

    pci_init_resource(&hose->io_resource,
                      0xb0000000,
                      0xb00fffff,
                      IORESOURCE_IO,
                      "PCI host bridge");

    pci_init_resource (&hose->mem_resources[0],
                       0x80000000,
                       0x803fffff,
                       IORESOURCE_MEM,
                       "PCI host bridge");

    hose->io_space.start = 0xb0000000;
    hose->io_space.end = 0xb00fffff;
    hose->mem_space.start = 0x7ff00000;
    hose->mem_space.end = 0x803fffff;
    hose->io_base_virt = (void *)0;

quoted

quoted

... I looked in m8260_setup.c and found

The 8260 doesn't have a PCI bridge, so these mappings aren't appropriate
for your board.  The 0xf0000000 is needed to get to the IMMR space, but
those accesses don't (shouldn't) be using in/out macros.  The 0xe0000000
space is generally used for flash prom, external devices, local sdram
and so on.

I am basing the 8250 startup from the m8260_setup.c file.

quoted

There are a variety of configuration options for the PCI bridge and all of
the other I/O in your system and the mapping using either BATs or page
tables.  It depends upon the address ranges of all of the I/O in your
system and their physical space allocation.

It looks like my _IO_BASE should be 0xb0000000. I've added a line in
m8260_setup.c to do

        io_block_mapping(0xb0000000, 0xb0000000, 0x000fffff, _PAGE_IO);

however, I get an Oops when I try to write with an outw(). From my debug
statements I see the address I'm trying to write to is

16-bit write: addr:b0000102 val:00000000

which looks correct since it is adding the 0xb0000000 offset.

Does it look like I have my mapping setup properly, and if so why am I
still getting Oopses?

quoted

      -- Dan

--
Jim Rowe
Advanced CounterMeasure Systems
Phone: (916)669-4304
Email: jrowe@acmsystems.com

**********************************************************************
This e-mail and any files transmitted with it may be confidential and
may be legally privileged or otherwise exempt from disclosure under
applicable law. This e-mail and its files are intended solely for
the individual or entity to whom they are addressed and their content
is the property of Smiths Aerospace.  If you are not the intended
recipient, please do not read, copy, use or disclose this communication.
If you have received this e-mail in error please notify the e-mail
administrator at postmaster@smiths-aerospace.com and then delete this
e-mail, its files and any copies.

This footnote also confirms that this e-mail message has been scanned
for the presence of known computer viruses.

Smiths addresses are changing!  The new addresses are of the form
firstname.lastname@smiths-aerospace.com.  Please update your address
books!  Please begin using the new form immediately.
***********************************************************************

--
Jim Rowe
Advanced CounterMeasure Systems
Phone: (916)669-4304
Email: jrowe@acmsystems.com


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Jim Rowe wrote:

....It would be even better if you are
using an MPC82xx chip with an integrated PCI controller.

Take a look at the 8245 with the integrated MPC10x PCI bridge.
Basically, the Sandpoint port.  The 8250 is a more simple
implementation than this.  The 405GP/Walnut would also be
a good example, just don't get mired in the details that it
doesn't have BATs :-)


	-- Dan


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

If anyone has a 16-bit pcmcia card working with a Cardbus to PCI bridge
it would be very helpful if I could see your /proc/ioports and
/proc/iomem with the card working. It would be even better if you are
using an MPC82xx chip with an integrated PCI controller.

Thanks,
-- Jim


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/