ioremap?

ioremap?

[Brian Murphy]

[-- Attachment #1: Type: text/plain, Size: 202 bytes --]

In three files/drivers I have had problems with physical addresses not 
being ioremapped.
I include my patch to fix things up and request comments. How come 
others don't have similar
problems?

/Brian

[-- Attachment #2: ioremap.diff --]
[-- Type: text/plain, Size: 1736 bytes --]

--- drivers/ide/ide-dma.c	2001/10/19 01:24:24	1.9
+++ drivers/ide/ide-dma.c	2002/05/29 11:49:14
@@ -741,7 +741,8 @@
 	if (hwif->mate && hwif->mate->dma_base) {
 		dma_base = hwif->mate->dma_base - (hwif->channel ? 0 : 8);
 	} else {
-		dma_base = pci_resource_start(dev, 4);
+		dma_base = ioremap(pci_resource_start(dev, 4), 
+				pci_resource_len(dev, 4));
 		if (!dma_base) {
 			printk("%s: dma_base is invalid (0x%04lx)\n", name, dma_base);
 			dma_base = 0;
--- drivers/ide/ide-pci.c	2001/11/19 13:53:59	1.20
+++ drivers/ide/ide-pci.c	2002/05/29 11:49:15
@@ -716,8 +716,10 @@
 		if (IDE_PCI_DEVID_EQ(d->devid, DEVID_HPT366) && (port) && (class_rev < 0x03))
 			return;
 		if ((dev->class >> 8) != PCI_CLASS_STORAGE_IDE || (dev->class & (port ? 4 : 1)) != 0) {
-			ctl  = dev->resource[(2*port)+1].start;
-			base = dev->resource[2*port].start;
+			ctl = ioremap(pci_resource_start(dev, (2*port)+1), 
+			              pci_resource_len(dev, (2*port)+1));
+			base = ioremap(pci_resource_start(dev, 2*port), 
+			              pci_resource_len(dev, 2*port));
 			if (!(ctl & PCI_BASE_ADDRESS_IO_MASK) ||
 			    !(base & PCI_BASE_ADDRESS_IO_MASK)) {
 				printk("%s: IO baseregs (BIOS) are reported as MEM, report to <andre@linux-ide.org>.\n", d->name);
--- drivers/net/pcnet32.c	2002/02/26 05:59:35	1.33.2.1
+++ drivers/net/pcnet32.c	2002/05/29 11:49:18
@@ -494,7 +494,8 @@
     }
     pci_set_master(pdev);
 
-    ioaddr = pci_resource_start (pdev, 0);
+    ioaddr = ioremap(pci_resource_start (pdev, 0),
+		    pci_resource_len (pdev, 0));
     printk(KERN_INFO "    ioaddr=%#08lx  resource_flags=%#08lx\n", ioaddr, pci_resource_flags (pdev, 0));
     if (!ioaddr) {
         printk (KERN_ERR "no PCI IO resources, aborting\n");

Re: ioremap?

[Bradley D. LaRonde]

Looks like those are for io (inb, outb, etc.) not mem (readb, writeb, etc.).
io addresses shouldn't be ioremapped.

Regards,
Brad

----- Original Message -----
From: "Brian Murphy" <brian@murphy.dk>
To: "linux-mips" <linux-mips@oss.sgi.com>
Sent: Wednesday, May 29, 2002 8:33 AM
Subject: ioremap?


> In three files/drivers I have had problems with physical addresses not
> being ioremapped.
> I include my patch to fix things up and request comments. How come
> others don't have similar
> problems?
>
> /Brian
>


----------------------------------------------------------------------------
----


> --- drivers/ide/ide-dma.c 2001/10/19 01:24:24 1.9
> +++ drivers/ide/ide-dma.c 2002/05/29 11:49:14
> @@ -741,7 +741,8 @@
>   if (hwif->mate && hwif->mate->dma_base) {
>   dma_base = hwif->mate->dma_base - (hwif->channel ? 0 : 8);
>   } else {
> - dma_base = pci_resource_start(dev, 4);
> + dma_base = ioremap(pci_resource_start(dev, 4),
> + pci_resource_len(dev, 4));
>   if (!dma_base) {
>   printk("%s: dma_base is invalid (0x%04lx)\n", name, dma_base);
>   dma_base = 0;
> --- drivers/ide/ide-pci.c 2001/11/19 13:53:59 1.20
> +++ drivers/ide/ide-pci.c 2002/05/29 11:49:15
> @@ -716,8 +716,10 @@
>   if (IDE_PCI_DEVID_EQ(d->devid, DEVID_HPT366) && (port) && (class_rev <
0x03))
>   return;
>   if ((dev->class >> 8) != PCI_CLASS_STORAGE_IDE || (dev->class & (port ?
4 : 1)) != 0) {
> - ctl  = dev->resource[(2*port)+1].start;
> - base = dev->resource[2*port].start;
> + ctl = ioremap(pci_resource_start(dev, (2*port)+1),
> +               pci_resource_len(dev, (2*port)+1));
> + base = ioremap(pci_resource_start(dev, 2*port),
> +               pci_resource_len(dev, 2*port));
>   if (!(ctl & PCI_BASE_ADDRESS_IO_MASK) ||
>       !(base & PCI_BASE_ADDRESS_IO_MASK)) {
>   printk("%s: IO baseregs (BIOS) are reported as MEM, report to
<andre@linux-ide.org>.\n", d->name);
> --- drivers/net/pcnet32.c 2002/02/26 05:59:35 1.33.2.1
> +++ drivers/net/pcnet32.c 2002/05/29 11:49:18
> @@ -494,7 +494,8 @@
>      }
>      pci_set_master(pdev);
>
> -    ioaddr = pci_resource_start (pdev, 0);
> +    ioaddr = ioremap(pci_resource_start (pdev, 0),
> +     pci_resource_len (pdev, 0));
>      printk(KERN_INFO "    ioaddr=%#08lx  resource_flags=%#08lx\n",
ioaddr, pci_resource_flags (pdev, 0));
>      if (!ioaddr) {
>          printk (KERN_ERR "no PCI IO resources, aborting\n");
>
>

Re: ioremap?

[Brian Murphy]

Bradley D. LaRonde wrote:

>Looks like those are for io (inb, outb, etc.) not mem (readb, writeb, etc.).
>io addresses shouldn't be ioremapped.
>
O.K.  I have now used set_io_port_base to set the base address to KSEG1.
What should be ioremapped then?

/Brian

Re: ioremap?

[Bradley D. LaRonde]

----- Original Message -----
From: "Brian Murphy" <brian@murphy.dk>
Cc: "linux-mips" <linux-mips@oss.sgi.com>
Sent: Wednesday, May 29, 2002 1:05 PM
Subject: Re: ioremap?


> Bradley D. LaRonde wrote:
>
> >Looks like those are for io (inb, outb, etc.) not mem (readb, writeb,
etc.).
> >io addresses shouldn't be ioremapped.
> >
> O.K.  I have now used set_io_port_base to set the base address to KSEG1.
> What should be ioremapped then?

Any address passed to readb, writeb, etc.  For example, a PCI mem resource
address.

Regards,
Brad