qla1280 mem-mapped I/O fix

qla1280 mem-mapped I/O fix

[David Mosberger]

With the fix in the second hunk, I don't see any reason not to turn on
MEMORY_MAPPED_IO in qla1280.  It seems to work fine on my machine
with this controller (ia64 Big Sur).

	--david

diff -Nru a/drivers/scsi/qla1280.c b/drivers/scsi/qla1280.c
--- a/drivers/scsi/qla1280.c	Sat May 10 01:47:43 2003
+++ b/drivers/scsi/qla1280.c	Sat May 10 01:47:43 2003
@@ -284,7 +284,7 @@
 #define  QL1280_TARGET_MODE_SUPPORT    0	/* Target mode support */
 #define  QL1280_LUN_SUPPORT            0
 #define  WATCHDOGTIMER                 0
-#define  MEMORY_MAPPED_IO              0
+#define  MEMORY_MAPPED_IO              1
 #define  DEBUG_QLA1280_INTR            0
 #define  USE_NVRAM_DEFAULTS	       0
 #define  DEBUG_PRINT_NVRAM             0
@@ -2634,7 +2634,7 @@
 	/*
 	 * Get memory mapped I/O address.
 	 */
-	pci_read_config_word (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
+	pci_read_config_dword (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
 	mmapbase &= PCI_BASE_ADDRESS_MEM_MASK;
 
 	/*

Re: qla1280 mem-mapped I/O fix

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 319 bytes --]


> @@ -2634,7 +2634,7 @@
>  	/*
>  	 * Get memory mapped I/O address.
>  	 */
> -	pci_read_config_word (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
> +	pci_read_config_dword (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
>  	mmapbase &= PCI_BASE_ADDRESS_MEM_MASK;
>  
>  	
shouldn't this be pci_resource_start() ?

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: qla1280 mem-mapped I/O fix

[David Mosberger]

>>>>> On 10 May 2003 12:15:08 +0200, Arjan van de Ven <arjanv@redhat.com> said:

  >> @@ -2634,7 +2634,7 @@
  >> /*
  >> * Get memory mapped I/O address.
  >> */
  >> -	pci_read_config_word (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
  >> +	pci_read_config_dword (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
  >> mmapbase &= PCI_BASE_ADDRESS_MEM_MASK;

  Arjan> shouldn't this be pci_resource_start() ?

Probably should be.  I wanted a minimal fix, because if you start
cleaning up qla1280, it won't stop there... ;-(

	--david

Re: qla1280 mem-mapped I/O fix

[David S. Miller]

On Sat, 2003-05-10 at 02:51, David Mosberger wrote:
> With the fix in the second hunk, I don't see any reason not to turn on
> MEMORY_MAPPED_IO in qla1280.  It seems to work fine on my machine
> with this controller (ia64 Big Sur).

David, you absolute MAY NOT pass this:

> -	pci_read_config_word (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
> +	pci_read_config_dword (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);

into ioremap() which is exactly what this driver is doing.
One must use the PCI device struct resource values.

-- 
David S. Miller <davem@redhat.com>

Re: qla1280 mem-mapped I/O fix

[David Mosberger]

>>>>> On 10 May 2003 04:30:11 -0700, "David S. Miller" <davem@redhat.com> said:

  DaveM> On Sat, 2003-05-10 at 02:51, David Mosberger wrote:
  >> With the fix in the second hunk, I don't see any reason not to turn on
  >> MEMORY_MAPPED_IO in qla1280.  It seems to work fine on my machine
  >> with this controller (ia64 Big Sur).

  DaveM> David, you absolute MAY NOT pass this:

Me?  It's the driver that's doing it! ;-)

	--david

Re: qla1280 mem-mapped I/O fix

[David S. Miller]

   From: David Mosberger <davidm@napali.hpl.hp.com>
   Date: Sat, 10 May 2003 04:34:42 -0700

   >>>>> On 10 May 2003 04:30:11 -0700, "David S. Miller" <davem@redhat.com> said:
   
     DaveM> David, you absolute MAY NOT pass this:
   
   Me?  It's the driver that's doing it! ;-)
   
It won't do it until your changes :-)

Re: qla1280 mem-mapped I/O fix

[Jes Sorensen]

>>>>> "Arjan" == Arjan van de Ven <arjanv@redhat.com> writes:

>> @@ -2634,7 +2634,7 @@ /* * Get memory mapped I/O address.  */ -
>> pci_read_config_word (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase); +
>> pci_read_config_dword (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
>> mmapbase &= PCI_BASE_ADDRESS_MEM_MASK;
>> 
>> 
Arjan> shouldn't this be pci_resource_start() ?

Yep,

The existing code is a nightmare, I am working on cleaning this up so
we can get rid of all the I/O ports crap.

Jes

Re: qla1280 mem-mapped I/O fix

[Andrey Panin]

On 135, 05 15, 2003 at 11:52:45AM -0400, Jes Sorensen wrote:
> >>>>> "Arjan" == Arjan van de Ven <arjanv@redhat.com> writes:
> 
> >> @@ -2634,7 +2634,7 @@ /* * Get memory mapped I/O address.  */ -
> >> pci_read_config_word (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase); +
> >> pci_read_config_dword (ha->pdev, PCI_BASE_ADDRESS_1, &mmapbase);
> >> mmapbase &= PCI_BASE_ADDRESS_MEM_MASK;
> >> 
> >> 
> Arjan> shouldn't this be pci_resource_start() ?
> 
> Yep,
> 
> The existing code is a nightmare, I am working on cleaning this up so
> we can get rid of all the I/O ports crap.

Do you want to convert this driver to use memmaped io only ?

That's bad news for me and visws subarch.

-- 
Andrey Panin		| Embedded systems software developer
pazke@orbita1.ru	| PGP key: wwwkeys.pgp.net