[PATCH] x86-64: pci-gart iommu sg chaining zeroes wrong sg.

[PATCH] x86-64: pci-gart iommu sg chaining zeroes wrong sg.

[FUJITA Tomonori]

This patch is for Jens' block tree (sg chaining branch).

I don't have the hardware but this looks like a bug.

---
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Subject: [PATCH] x86-64: pci-gart iommu sg chaining zeroes a wrong sg's dma_length

Needs to zero the end of the list.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
 arch/x86_64/kernel/pci-gart.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
index 27b7db4..a4151a7 100644
--- a/arch/x86_64/kernel/pci-gart.c
+++ b/arch/x86_64/kernel/pci-gart.c
@@ -425,9 +425,10 @@ int gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
 	if (dma_map_cont(start_sg, i - start, sgmap, pages, need) < 0)
 		goto error;
 	out++;
+	sgmap = sg_next(sgmap);
 	flush_gart();
 	if (out < nents)
-		ps->dma_length = 0;
+		sgmap->dma_length = 0;
 	return out;
 
 error:
-- 
1.5.2.4

Re: [PATCH] x86-64: pci-gart iommu sg chaining zeroes wrong sg.

[FUJITA Tomonori]

On Fri, 28 Sep 2007 01:38:27 +0900
FUJITA Tomonori <tomof@acm.org> wrote:

> This patch is for Jens' block tree (sg chaining branch).
> 
> I don't have the hardware but this looks like a bug.
> 
> ---
> From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> Subject: [PATCH] x86-64: pci-gart iommu sg chaining zeroes a wrong sg's dma_length
> 
> Needs to zero the end of the list.
> 
> Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> ---
>  arch/x86_64/kernel/pci-gart.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
> index 27b7db4..a4151a7 100644
> --- a/arch/x86_64/kernel/pci-gart.c
> +++ b/arch/x86_64/kernel/pci-gart.c
> @@ -425,9 +425,10 @@ int gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
>  	if (dma_map_cont(start_sg, i - start, sgmap, pages, need) < 0)
>  		goto error;
>  	out++;
> +	sgmap = sg_next(sgmap);
>  	flush_gart();
>  	if (out < nents)
> -		ps->dma_length = 0;
> +		sgmap->dma_length = 0;
>  	return out;

Sorry, it should be:


diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
index 27b7db4..cfcc84e 100644
--- a/arch/x86_64/kernel/pci-gart.c
+++ b/arch/x86_64/kernel/pci-gart.c
@@ -426,8 +426,10 @@ int gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
 		goto error;
 	out++;
 	flush_gart();
-	if (out < nents)
-		ps->dma_length = 0;
+	if (out < nents) {
+		sgmap = sg_next(sgmap);
+		sgmap->dma_length = 0;
+	}
 	return out;
 
 error:

Re: [PATCH] x86-64: pci-gart iommu sg chaining zeroes wrong sg.

[Benny Halevy]

On Sep 27, 2007, 18:46 +0200, FUJITA Tomonori <tomof@acm.org> wrote:
> On Fri, 28 Sep 2007 01:38:27 +0900
> FUJITA Tomonori <tomof@acm.org> wrote:
> 
>> This patch is for Jens' block tree (sg chaining branch).
>>
>> I don't have the hardware but this looks like a bug.
>>
>> ---
>> From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
>> Subject: [PATCH] x86-64: pci-gart iommu sg chaining zeroes a wrong sg's dma_length
>>
>> Needs to zero the end of the list.
>>
>> Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
>> ---
>>  arch/x86_64/kernel/pci-gart.c |    3 ++-
>>  1 files changed, 2 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
>> index 27b7db4..a4151a7 100644
>> --- a/arch/x86_64/kernel/pci-gart.c
>> +++ b/arch/x86_64/kernel/pci-gart.c
>> @@ -425,9 +425,10 @@ int gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
>>  	if (dma_map_cont(start_sg, i - start, sgmap, pages, need) < 0)
>>  		goto error;
>>  	out++;
>> +	sgmap = sg_next(sgmap);
>>  	flush_gart();
>>  	if (out < nents)
>> -		ps->dma_length = 0;
>> +		sgmap->dma_length = 0;
>>  	return out;
> 
> Sorry, it should be:
> 
> 
> diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
> index 27b7db4..cfcc84e 100644
> --- a/arch/x86_64/kernel/pci-gart.c
> +++ b/arch/x86_64/kernel/pci-gart.c
> @@ -426,8 +426,10 @@ int gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
>  		goto error;
>  	out++;
>  	flush_gart();
> -	if (out < nents)
> -		ps->dma_length = 0;
> +	if (out < nents) {
> +		sgmap = sg_next(sgmap);
> +		sgmap->dma_length = 0;
> +	}

looks correct to me.
ps points at the previous "scanned" sg entry while you want to zero out
dma_length at the entry immediately following the last entry mapped
(if (out < nents))

the original code before 62296749bd421904dace1e6b0fc3c4538aac7111 was:
-	if (out < nents) 
-		sg[out].dma_length = 0; 

Benny

>  	return out;
>  
>  error:
> -
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>