[PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[Benjamin Herrenschmidt]

The e1000 driver stores the content of the PCI resources into
unsigned long's before ioremapping. This breaks on 32 bits
platforms that support 64 bits MMIO resources such as ppc 44x.

This fixes it by removing those temporary variables and passing
directly the result of pci_resource_start/len to ioremap.

The side effect is that I removed the assignments to the netdev
fields mem_start, mem_end and base_addr, which are totally useless
for PCI devices.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
--

 drivers/net/e1000/e1000_main.c |   18 +++++-------------
 1 file changed, 5 insertions(+), 13 deletions(-)

Index: linux-work/drivers/net/e1000/e1000_main.c
===================================================================
--- linux-work.orig/drivers/net/e1000/e1000_main.c	2007-11-16 17:14:32.000000000 +1100
+++ linux-work/drivers/net/e1000/e1000_main.c	2007-11-16 18:32:07.000000000 +1100
@@ -861,8 +861,6 @@ e1000_probe(struct pci_dev *pdev,
 {
 	struct net_device *netdev;
 	struct e1000_adapter *adapter;
-	unsigned long mmio_start, mmio_len;
-	unsigned long flash_start, flash_len;
 
 	static int cards_found = 0;
 	static int global_quad_port_a = 0; /* global ksp3 port a indication */
@@ -905,11 +903,9 @@ e1000_probe(struct pci_dev *pdev,
 	adapter->hw.back = adapter;
 	adapter->msg_enable = (1 << debug) - 1;
 
-	mmio_start = pci_resource_start(pdev, BAR_0);
-	mmio_len = pci_resource_len(pdev, BAR_0);
-
 	err = -EIO;
-	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
+	adapter->hw.hw_addr = ioremap(pci_resource_start(pdev, BAR_0),
+				      pci_resource_len(pdev, BAR_0));
 	if (!adapter->hw.hw_addr)
 		goto err_ioremap;
 
@@ -944,10 +940,6 @@ e1000_probe(struct pci_dev *pdev,
 #endif
 	strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
 
-	netdev->mem_start = mmio_start;
-	netdev->mem_end = mmio_start + mmio_len;
-	netdev->base_addr = adapter->hw.io_base;
-
 	adapter->bd_number = cards_found;
 
 	/* setup the private structure */
@@ -960,9 +952,9 @@ e1000_probe(struct pci_dev *pdev,
 	 * because it depends on mac_type */
 	if ((adapter->hw.mac_type == e1000_ich8lan) &&
 	   (pci_resource_flags(pdev, 1) & IORESOURCE_MEM)) {
-		flash_start = pci_resource_start(pdev, 1);
-		flash_len = pci_resource_len(pdev, 1);
-		adapter->hw.flash_address = ioremap(flash_start, flash_len);
+		adapter->hw.flash_address =
+			ioremap(pci_resource_start(pdev, 1),
+				pci_resource_len(pdev, 1));
 		if (!adapter->hw.flash_address)
 			goto err_flashmap;
 	}

Re: [PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[Jeff Garzik]

Benjamin Herrenschmidt wrote:
> The e1000 driver stores the content of the PCI resources into
> unsigned long's before ioremapping. This breaks on 32 bits
> platforms that support 64 bits MMIO resources such as ppc 44x.
> 
> This fixes it by removing those temporary variables and passing
> directly the result of pci_resource_start/len to ioremap.
> 
> The side effect is that I removed the assignments to the netdev
> fields mem_start, mem_end and base_addr, which are totally useless
> for PCI devices.
> 
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> --
> 
>  drivers/net/e1000/e1000_main.c |   18 +++++-------------
>  1 file changed, 5 insertions(+), 13 deletions(-)

Looks good to me.  auke?

Re: [PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[Kok, Auke]

Jeff Garzik wrote:
> Benjamin Herrenschmidt wrote:
>> The e1000 driver stores the content of the PCI resources into
>> unsigned long's before ioremapping. This breaks on 32 bits
>> platforms that support 64 bits MMIO resources such as ppc 44x.
>>
>> This fixes it by removing those temporary variables and passing
>> directly the result of pci_resource_start/len to ioremap.
>>
>> The side effect is that I removed the assignments to the netdev
>> fields mem_start, mem_end and base_addr, which are totally useless
>> for PCI devices.
>>
>> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>> -- 
>>
>>  drivers/net/e1000/e1000_main.c |   18 +++++-------------
>>  1 file changed, 5 insertions(+), 13 deletions(-)
> 
> Looks good to me.  auke?

yes, please apply.

Acked-by: Auke Kok <auke-jan.h.kok@intel.com>



Auke

RE: [PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[Brandeburg, Jesse]

Benjamin Herrenschmidt wrote:
> The e1000 driver stores the content of the PCI resources into
> unsigned long's before ioremapping. This breaks on 32 bits
> platforms that support 64 bits MMIO resources such as ppc 44x.
>=20
> This fixes it by removing those temporary variables and passing
> directly the result of pci_resource_start/len to ioremap.

the patch is mostly fine, but looking through the tree, doesn't every
driver that has 64 bit capable resources have this problem?

in fact, skeleton module has this problem, if I'm not mistaken.  e1000
isn't the only one. :-)
=20
> The side effect is that I removed the assignments to the netdev
> fields mem_start, mem_end and base_addr, which are totally useless
> for PCI devices.

also, concerning this, ifconfig shows the output of these variables, I
don't think we want to blatantly remove them, as it is actually removing
information that is, if not used, at least displayed to user space.

eth6      Link encap:Ethernet  HWaddr 00:0E:01:0C:02:03
          inet addr:134.134.3.121  Bcast:134.134.3.255
Mask:255.255.255.0
          inet6 addr: fe80::20e:1ff:fe0c:203/64 Scope:Link
          UP BROADCAST NOTRAILERS RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:9022519 errors:0 dropped:0 overruns:0 frame:0
          TX packets:1303701 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:4044613454 (3857.2 Mb)  TX bytes:118365109 (112.8 Mb)
          Base address:0xf000 Memory:feac0000-feae0000
                                     ^^^^^^^^^^^^^^^^^

RE: [PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[Benjamin Herrenschmidt]

On Wed, 2007-11-28 at 10:47 -0800, Brandeburg, Jesse wrote:
> 
> > The side effect is that I removed the assignments to the netdev
> > fields mem_start, mem_end and base_addr, which are totally useless
> > for PCI devices.
> 
> also, concerning this, ifconfig shows the output of these variables, I
> don't think we want to blatantly remove them, as it is actually removing
> information that is, if not used, at least displayed to user space.

It's not the right place to have that information and there is no
room in the user space field to fit more than 32 bits on 32 bits
platforms so I'd rather remove it. It's really only for ISA drivers that
have user configurable addresses.

Ben.

Re: [PATCH] e1000: Fix for 32 bits platforms with 64 bits resources

[Jeff Garzik]

Benjamin Herrenschmidt wrote:
> The e1000 driver stores the content of the PCI resources into
> unsigned long's before ioremapping. This breaks on 32 bits
> platforms that support 64 bits MMIO resources such as ppc 44x.
> 
> This fixes it by removing those temporary variables and passing
> directly the result of pci_resource_start/len to ioremap.
> 
> The side effect is that I removed the assignments to the netdev
> fields mem_start, mem_end and base_addr, which are totally useless
> for PCI devices.
> 
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> --
> 
>  drivers/net/e1000/e1000_main.c |   18 +++++-------------
>  1 file changed, 5 insertions(+), 13 deletions(-)

applied