[PATCH] Bug in sn_pci_map_sg causes MCA

[PATCH] Bug in sn_pci_map_sg causes MCA

[Alan Mayer]

If sg->dma_address is set, we try to do a __pa() on a dma_address,
then, later, create a dma_addresss from a munged dma_address.  When
this bogus dma_address is used by the card, it results in MCAs.

		--ajm



--- /usr/tmp/TmpDir.5614-0/linux/arch/ia64/sn/io/machvec/pci_dma.c_1.5  Mon Nov  3 14:32:30 2003
+++ linux/arch/ia64/sn/io/machvec/pci_dma.c     Mon Nov  3 13:16:20 2003
@@ -279,8 +279,7 @@
         * scatterlist.
         */
        for (i = 0; i < nents; i++, sg++) {
-               phys_addr = __pa(sg->dma_address ? sg->dma_address :
-                       (unsigned long)page_address(sg->page) + sg->offset);
+               phys_addr = __pa((unsigned long)page_address(sg->page) + sg->offset);
 
                /*
                 * Handle the most common case: 64 bit cards.  This

Re: [PATCH] Bug in sn_pci_map_sg causes MCA

[Jesse Barnes]

David, this fixes a critical bug for us.  Can you push it to Linus or
should we send it straight to Andrew?

Thanks,
Jesse

On Mon, Nov 03, 2003 at 02:36:29PM -0600, Alan Mayer wrote:
> If sg->dma_address is set, we try to do a __pa() on a dma_address,
> then, later, create a dma_addresss from a munged dma_address.  When
> this bogus dma_address is used by the card, it results in MCAs.
> 
> 		--ajm
> 
> 
> 
> --- /usr/tmp/TmpDir.5614-0/linux/arch/ia64/sn/io/machvec/pci_dma.c_1.5  Mon Nov  3 14:32:30 2003
> +++ linux/arch/ia64/sn/io/machvec/pci_dma.c     Mon Nov  3 13:16:20 2003
> @@ -279,8 +279,7 @@
>          * scatterlist.
>          */
>         for (i = 0; i < nents; i++, sg++) {
> -               phys_addr = __pa(sg->dma_address ? sg->dma_address :
> -                       (unsigned long)page_address(sg->page) + sg->offset);
> +               phys_addr = __pa((unsigned long)page_address(sg->page) + sg->offset);
>  
>                 /*
>                  * Handle the most common case: 64 bit cards.  This
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] Bug in sn_pci_map_sg causes MCA

[David Mosberger]

>>>>> On Mon, 3 Nov 2003 12:47:55 -0800, jbarnes@sgi.com (Jesse Barnes) said:

  Jesse> David, this fixes a critical bug for us.  Can you push it to
  Jesse> Linus or should we send it straight to Andrew?

The patch doesn't apply for me, otherwise, I'd have been happy to take
it (since it's a critical bug fix).  I think it fails to apply because
TABs got replaced by blanks.

	--david

Re: [PATCH] Bug in sn_pci_map_sg causes MCA

[Jesse Barnes]

On Mon, Nov 03, 2003 at 01:14:00PM -0800, David Mosberger wrote:
> >>>>> On Mon, 3 Nov 2003 12:47:55 -0800, jbarnes@sgi.com (Jesse Barnes) said:
> 
>   Jesse> David, this fixes a critical bug for us.  Can you push it to
>   Jesse> Linus or should we send it straight to Andrew?
> 
> The patch doesn't apply for me, otherwise, I'd have been happy to take
> it (since it's a critical bug fix).  I think it fails to apply because
> TABs got replaced by blanks.

Sorry about that, please try this one.

Jesse

--- /usr/tmp/TmpDir.16109-0/linux/arch/ia64/sn/io/machvec/pci_dma.c_1.5	Mon Nov  3 13:00:46 2003
+++ linux/arch/ia64/sn/io/machvec/pci_dma.c	Mon Nov  3 13:00:27 2003
@@ -279,8 +279,7 @@
 	 * scatterlist.
 	 */
 	for (i = 0; i < nents; i++, sg++) {
-		phys_addr = __pa(sg->dma_address ? sg->dma_address :
-			(unsigned long)page_address(sg->page) + sg->offset);
+		phys_addr = __pa((unsigned long)page_address(sg->page) + sg->offset);
 
 		/*
 		 * Handle the most common case: 64 bit cards.  This

Re: [PATCH] Bug in sn_pci_map_sg causes MCA

[Alan Mayer]

Sorry, here's one that should actually work:


--- /home/poppy15/ajm/work/davidm.bk/arch/ia64/sn/io/machvec/pci_dma.c	Mon Nov  3 11:49:00 2003
+++ arch/ia64/sn/io/machvec/pci_dma.c	Mon Nov  3 13:16:20 2003
@@ -279,8 +279,7 @@
 	 * scatterlist.
 	 */
 	for (i = 0; i < nents; i++, sg++) {
-		phys_addr = __pa(sg->dma_address ? sg->dma_address :
-			(unsigned long)page_address(sg->page) + sg->offset);
+		phys_addr = __pa((unsigned long)page_address(sg->page) + sg->offset);
 
 		/*
 		 * Handle the most common case: 64 bit cards.  This


		--ajm


On Mon, 3 Nov 2003, David Mosberger wrote:

> >>>>> On Mon, 3 Nov 2003 12:47:55 -0800, jbarnes@sgi.com (Jesse Barnes) said:
> 
>   Jesse> David, this fixes a critical bug for us.  Can you push it to
>   Jesse> Linus or should we send it straight to Andrew?
> 
> The patch doesn't apply for me, otherwise, I'd have been happy to take
> it (since it's a critical bug fix).  I think it fails to apply because
> TABs got replaced by blanks.
> 
> 	--david
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

But he can't be a man,
'cause he doesn't smoke
the same cigarettes as me.
--
Alan J. Mayer
SGI
ajm@sgi.com
WORK: 651-683-3131
HOME: 651-407-0134
--

Re: [PATCH] Bug in sn_pci_map_sg causes MCA

[David Mosberger]

>>>>> On Mon, 3 Nov 2003 16:07:53 -0600 (CST), Alan Mayer <ajm@sgi.com> said:

  Alan> Sorry, here's one that should actually work:

Well, no it doesn't apply with patch -p1.  But it's OK this time, I fixed
it myself.

	--david