thr3ads.net - Xen devel - [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU [Nov 2009]

If this information is useful, please help other people find it:
Share via:

Jan Beulich

2009-Nov-19 09:50 UTC

[Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

Replacing virt_to_phys() by virt_to_bus(), and adding code to ensure
contiguity as required by the firmware.

As usual, written and tested on 2.6.27.37 and made apply to the 2.6.18
tree without further testing.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Tested-by: Douglas Warzecha <Douglas_Warzecha@Dell.com>

--- sle11-2009-10-16.orig/drivers/firmware/dell_rbu.c	2009-02-02
09:22:26.000000000 +0100
+++ sle11-2009-10-16/drivers/firmware/dell_rbu.c	2009-10-26 16:34:16.000000000
+0100
@@ -169,9 +169,28 @@ static int create_packet(void *data, siz
 			spin_lock(&rbu_data.lock);
 			goto out_alloc_packet_array;
 		}
+#ifdef CONFIG_XEN
+		if (ordernum && xen_create_contiguous_region(
+			(unsigned long)packet_data_temp_buf, ordernum, 0)) {
+			free_pages((unsigned long)packet_data_temp_buf,
+				   ordernum);
+			printk(KERN_WARNING
+				"dell_rbu:%s: failed to adjust new "
+				"packet\n", __func__);
+			retval = -ENOMEM;
+			spin_lock(&rbu_data.lock);
+			goto out_alloc_packet_array;
+		}
+#endif
 
-		if ((unsigned long)virt_to_phys(packet_data_temp_buf)
+		if ((unsigned long)virt_to_bus(packet_data_temp_buf)
 				< allocation_floor) {
+#ifdef CONFIG_XEN
+			if (ordernum)
+				xen_destroy_contiguous_region(
+					(unsigned long)packet_data_temp_buf,
+					ordernum);
+#endif
 			pr_debug("packet 0x%lx below floor at 0x%lx.\n",
 					(unsigned long)virt_to_phys(
 						packet_data_temp_buf),
@@ -185,7 +204,7 @@ static int create_packet(void *data, siz
 	newpacket->data = packet_data_temp_buf;
 
 	pr_debug("create_packet: newpacket at physical addr %lx\n",
-		(unsigned long)virt_to_phys(newpacket->data));
+		(unsigned long)virt_to_bus(newpacket->data));
 
 	/* packets may not have fixed size */
 	newpacket->length = length;
@@ -204,7 +223,7 @@ out_alloc_packet_array:
 	/* always free packet array */
 	for (;idx>0;idx--) {
 		pr_debug("freeing unused packet below floor 0x%lx.\n",
-			(unsigned long)virt_to_phys(
+			(unsigned long)virt_to_bus(
 				invalid_addr_packet_array[idx-1]));
 		free_pages((unsigned long)invalid_addr_packet_array[idx-1],
 			ordernum);
@@ -348,6 +367,13 @@ static void packet_empty_list(void)
 		 * to make sure there are no stale RBU packets left in memory
 		 */
 		memset(newpacket->data, 0, rbu_data.packetsize);
+#ifdef CONFIG_XEN
+		if (newpacket->ordernum)
+			xen_destroy_contiguous_region(
+				(unsigned long)newpacket->data,
+				newpacket->ordernum);
+#endif
+
 		free_pages((unsigned long) newpacket->data,
 			newpacket->ordernum);
 		kfree(newpacket);
@@ -402,7 +428,9 @@ static int img_update_realloc(unsigned l
 {
 	unsigned char *image_update_buffer = NULL;
 	unsigned long rc;
+#ifndef CONFIG_XEN
 	unsigned long img_buf_phys_addr;
+#endif
 	int ordernum;
 	int dma_alloc = 0;
 
@@ -433,15 +461,19 @@ static int img_update_realloc(unsigned l
 
 	spin_unlock(&rbu_data.lock);
 
+#ifndef CONFIG_XEN
 	ordernum = get_order(size);
 	image_update_buffer  		(unsigned char *) __get_free_pages(GFP_KERNEL,
ordernum);
 
 	img_buf_phys_addr -		(unsigned long) virt_to_phys(image_update_buffer);
+		(unsigned long) virt_to_bus(image_update_buffer);
 
 	if (img_buf_phys_addr > BIOS_SCAN_LIMIT) {
 		free_pages((unsigned long) image_update_buffer, ordernum);
+#else
+	{
+#endif
 		ordernum = -1;
 		image_update_buffer = dma_alloc_coherent(NULL, size,
 			&dell_rbu_dmaaddr, GFP_KERNEL);
@@ -706,6 +738,12 @@ static struct bin_attribute rbu_packet_s
 static int __init dcdrbu_init(void)
 {
 	int rc = 0;
+
+#ifdef CONFIG_XEN
+	if (!is_initial_xendomain())
+		return -ENODEV;
+#endif
+
 	spin_lock_init(&rbu_data.lock);
 
 	init_packet_head();





_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Jan Beulich

2009-Nov-19 10:01 UTC

head link

Re: [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

>>> "Jan Beulich" <JBeulich@novell.com> 19.11.09 10:50
>>>
>Replacing virt_to_phys() by virt_to_bus(), and adding code to ensure
>contiguity as required by the firmware.
While similar address translation changes are also needed in dcdbas,
doing them is not sufficient for the driver to work under Xen. It also
requires that the issuing on SMIs happens on physical CPU#0, which
presently we can arrange for only by using domctl interfaces. I''m not
certain however that I could, not being there is any interest in this in
the first place, reliably backport to 2.6.18 the hack we use for this in
our forward ported trees. Of course, some mechanism will certainly be
needed in the pv-ops Dom0 case once that reaches the point where
it can be used in products...

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Konrad Rzeszutek Wilk

2009-Nov-19 17:22 UTC

head link

Re: [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

On Thu, Nov 19, 2009 at 09:50:35AM +0000, Jan Beulich
wrote:> Replacing virt_to_phys() by virt_to_bus(), and adding code to ensure
> contiguity as required by the firmware.
> 
> As usual, written and tested on 2.6.27.37 and made apply to the 2.6.18
> tree without further testing.
> 
> Signed-off-by: Jan Beulich <jbeulich@novell.com>
> Tested-by: Douglas Warzecha <Douglas_Warzecha@Dell.com>
> 
> --- sle11-2009-10-16.orig/drivers/firmware/dell_rbu.c	2009-02-02
09:22:26.000000000 +0100
> +++ sle11-2009-10-16/drivers/firmware/dell_rbu.c	2009-10-26
16:34:16.000000000 +0100
> @@ -169,9 +169,28 @@ static int create_packet(void *data, siz
>  			spin_lock(&rbu_data.lock);
>  			goto out_alloc_packet_array;
>  		}
> +#ifdef CONFIG_XEN
> +		if (ordernum && xen_create_contiguous_region(
> +			(unsigned long)packet_data_temp_buf, ordernum, 0)) {
> +			free_pages((unsigned long)packet_data_temp_buf,
> +				   ordernum);
> +			printk(KERN_WARNING
> +				"dell_rbu:%s: failed to adjust new "
> +				"packet\n", __func__);
> +			retval = -ENOMEM;
> +			spin_lock(&rbu_data.lock);
> +			goto out_alloc_packet_array;
> +		}
> +#endif
I was under the impression we did not want to create any of the #ifdef
CONFIG_XEN
at all. As a matter of fact, to actually eliminate or minimize the level of
them.

Is there no other way to do this?

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Konrad Rzeszutek Wilk

2009-Nov-19 17:35 UTC

head link

Re: [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

On Thu, Nov 19, 2009 at 12:22:45PM -0500, Konrad Rzeszutek Wilk
wrote:> On Thu, Nov 19, 2009 at 09:50:35AM +0000, Jan Beulich wrote:
> > Replacing virt_to_phys() by virt_to_bus(), and adding code to ensure
> > contiguity as required by the firmware.
> > 
> > As usual, written and tested on 2.6.27.37 and made apply to the 2.6.18
> > tree without further testing.
Duh! I need my coffee. Obviously this patch is not for the pv-ops kernel -
please
do ignore my comments.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Xen devel - Nov 2009 - [PATCH] linux: force proper address translation in Dell RBU

[Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

Re: [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

Re: [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU

Re: [Xen-devel] [PATCH] linux: force proper address translation in Dell RBU