[PATCH] amd_iommu: IO_PAGE_FAULTs on unity mapped regions during amd_iommu_init()

[PATCH] amd_iommu: IO_PAGE_FAULTs on unity mapped regions during amd_iommu_init()

[Shuah Khan]

When dma_ops are initialized the unity mappings are created. The
init_device_table_dma() function makes sure DMA from all devices is
blocked by default. This opens a short window in time where DMA to
unity mapped regions is blocked by the IOMMU. Make sure this does not
happen by initializing the device table after dma_ops.

Signed-off-by: Shuah Khan <shuah.khan@hp.com>
CC: stable@vger.kernel.org 3.4
---
 drivers/iommu/amd_iommu_init.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/amd_iommu_init.c b/drivers/iommu/amd_iommu_init.c
index ef0ae93..b573f80 100644
--- a/drivers/iommu/amd_iommu_init.c
+++ b/drivers/iommu/amd_iommu_init.c
@@ -1572,8 +1572,6 @@ int __init amd_iommu_init_hardware(void)
 	if (amd_iommu_pd_alloc_bitmap == NULL)
 		goto free;
 
-	/* init the device table */
-	init_device_table();
 
 	/*
 	 * let all alias entries point to itself
@@ -1655,6 +1653,7 @@ out:
  */
 static int __init amd_iommu_init(void)
 {
+	struct amd_iommu *iommu;
 	int ret = 0;
 
 	ret = amd_iommu_init_hardware();
@@ -1673,6 +1672,12 @@ static int __init amd_iommu_init(void)
 	if (ret)
 		goto free;
 
+	/* init the device table */
+	init_device_table();
+
+	for_each_iommu(iommu)
+		iommu_flush_all_caches(iommu);
+
 	amd_iommu_init_api();
 
 	x86_platform.iommu_shutdown = disable_iommus;
-- 
1.7.9.5

[PATCH] amd_iommu: IO_PAGE_FAULTs on unity mapped regions during amd_iommu_init()

[Shuah Khan]

When dma_ops are initialized the unity mappings are created. The
init_device_table_dma() function makes sure DMA from all devices is
blocked by default. This opens a short window in time where DMA to
unity mapped regions is blocked by the IOMMU. Make sure this does not
happen by initializing the device table after dma_ops.

Reference: http://www.gossamer-threads.com/lists/linux/kernel/1670769

Signed-off-by: Shuah Khan <shuah.khan@hp.com>
CC: stable@vger.kernel.org 3.0
---
 arch/x86/kernel/amd_iommu_init.c |   10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 33df6e8..d86aa3f 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -1363,6 +1363,7 @@ static struct syscore_ops amd_iommu_syscore_ops = {
  */
 static int __init amd_iommu_init(void)
 {
+	struct amd_iommu *iommu;
 	int i, ret = 0;
 
 	/*
@@ -1411,9 +1412,6 @@ static int __init amd_iommu_init(void)
 	if (amd_iommu_pd_alloc_bitmap == NULL)
 		goto free;
 
-	/* init the device table */
-	init_device_table();
-
 	/*
 	 * let all alias entries point to itself
 	 */
@@ -1463,6 +1461,12 @@ static int __init amd_iommu_init(void)
 	if (ret)
 		goto free_disable;
 
+	/* init the device table */
+	init_device_table();
+
+	for_each_iommu(iommu)
+		iommu_flush_all_caches(iommu);
+
 	amd_iommu_init_api();
 
 	amd_iommu_init_notifier();
-- 
1.7.9.5

Re: [PATCH] amd_iommu: IO_PAGE_FAULTs on unity mapped regions during amd_iommu_init()

[Ingo Molnar]

* Shuah Khan <shuah.khan-VXdhtT5mjnY@public.gmane.org> wrote:

> When dma_ops are initialized the unity mappings are created. The
> init_device_table_dma() function makes sure DMA from all devices is
> blocked by default. This opens a short window in time where DMA to
> unity mapped regions is blocked by the IOMMU. Make sure this does not
> happen by initializing the device table after dma_ops.
> 
> Reference: http://www.gossamer-threads.com/lists/linux/kernel/1670769
> 
> Signed-off-by: Shuah Khan <shuah.khan-VXdhtT5mjnY@public.gmane.org>
> CC: stable-u79uwXL29TY76Z2rM5mHXA@public.gmane.org 3.0
> ---
>  arch/x86/kernel/amd_iommu_init.c |   10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)

That file does not exist anymore, it died 2.5+ years ago:

  403f81d8ee53 iommu/amd: Move missing parts to drivers/iommu

Thanks,

	Ingo

Re: [PATCH] amd_iommu: IO_PAGE_FAULTs on unity mapped regions during amd_iommu_init()

[Joerg Roedel]

On Thu, Feb 07, 2013 at 11:49:05AM +0100, Ingo Molnar wrote:
> 
> * Shuah Khan <shuah.khan-VXdhtT5mjnY@public.gmane.org> wrote:
> >  arch/x86/kernel/amd_iommu_init.c |   10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> That file does not exist anymore, it died 2.5+ years ago:
> 
>   403f81d8ee53 iommu/amd: Move missing parts to drivers/iommu

Right, the file was moved for Linux v3.1 AFAIR. But this patch is a
stable-packport for v3.0 of a patch I sent yersterday :)


	Joerg

[PATCH 3.4] iommu/amd: Initialize device table after dma_ops

[Shuah Khan]

When dma_ops are initialized the unity mappings are created. The
init_device_table_dma() function makes sure DMA from all devices is
blocked by default. This opens a short window in time where DMA to
unity mapped regions is blocked by the IOMMU. Make sure this does not
happen by initializing the device table after dma_ops.

Back-port upstream commit: f528d980c17b8714aedc918ba86e058af914d66b

Signed-off-by: Joerg Roedel <joro@8bytes.org>
Signed-off-by: Shuah Khan <shuah.khan@hp.com>
CC: stable@vger.kernel.org 3.4
---
 drivers/iommu/amd_iommu_init.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/amd_iommu_init.c b/drivers/iommu/amd_iommu_init.c
index ef0ae93..b573f80 100644
--- a/drivers/iommu/amd_iommu_init.c
+++ b/drivers/iommu/amd_iommu_init.c
@@ -1572,8 +1572,6 @@ int __init amd_iommu_init_hardware(void)
 	if (amd_iommu_pd_alloc_bitmap == NULL)
 		goto free;
 
-	/* init the device table */
-	init_device_table();
 
 	/*
 	 * let all alias entries point to itself
@@ -1655,6 +1653,7 @@ out:
  */
 static int __init amd_iommu_init(void)
 {
+	struct amd_iommu *iommu;
 	int ret = 0;
 
 	ret = amd_iommu_init_hardware();
@@ -1673,6 +1672,12 @@ static int __init amd_iommu_init(void)
 	if (ret)
 		goto free;
 
+	/* init the device table */
+	init_device_table();
+
+	for_each_iommu(iommu)
+		iommu_flush_all_caches(iommu);
+
 	amd_iommu_init_api();
 
 	x86_platform.iommu_shutdown = disable_iommus;
-- 
1.7.9.5

Re: [PATCH 3.4] iommu/amd: Initialize device table after dma_ops

[Shuah Khan]

On Tue, 2013-02-26 at 16:49 -0700, Shuah Khan wrote:
> When dma_ops are initialized the unity mappings are created. The
> init_device_table_dma() function makes sure DMA from all devices is
> blocked by default. This opens a short window in time where DMA to
> unity mapped regions is blocked by the IOMMU. Make sure this does not
> happen by initializing the device table after dma_ops.
> 
> Back-port upstream commit: f528d980c17b8714aedc918ba86e058af914d66b
> 
> Signed-off-by: Joerg Roedel <joro@8bytes.org>
> Signed-off-by: Shuah Khan <shuah.khan@hp.com>
> CC: stable@vger.kernel.org 3.4

Please consider this for 3.5 as well.

-- Shuah
> ---
>  drivers/iommu/amd_iommu_init.c |    9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/iommu/amd_iommu_init.c b/drivers/iommu/amd_iommu_init.c
> index ef0ae93..b573f80 100644
> --- a/drivers/iommu/amd_iommu_init.c
> +++ b/drivers/iommu/amd_iommu_init.c
> @@ -1572,8 +1572,6 @@ int __init amd_iommu_init_hardware(void)
>  	if (amd_iommu_pd_alloc_bitmap == NULL)
>  		goto free;
>  
> -	/* init the device table */
> -	init_device_table();
>  
>  	/*
>  	 * let all alias entries point to itself
> @@ -1655,6 +1653,7 @@ out:
>   */
>  static int __init amd_iommu_init(void)
>  {
> +	struct amd_iommu *iommu;
>  	int ret = 0;
>  
>  	ret = amd_iommu_init_hardware();
> @@ -1673,6 +1672,12 @@ static int __init amd_iommu_init(void)
>  	if (ret)
>  		goto free;
>  
> +	/* init the device table */
> +	init_device_table();
> +
> +	for_each_iommu(iommu)
> +		iommu_flush_all_caches(iommu);
> +
>  	amd_iommu_init_api();
>  
>  	x86_platform.iommu_shutdown = disable_iommus;

Re: [PATCH 3.4] iommu/amd: Initialize device table after dma_ops

[Shuah Khan]

On Thu, 2013-02-28 at 12:52 -0700, Shuah Khan wrote:
> On Tue, 2013-02-26 at 16:49 -0700, Shuah Khan wrote:
> > When dma_ops are initialized the unity mappings are created. The
> > init_device_table_dma() function makes sure DMA from all devices is
> > blocked by default. This opens a short window in time where DMA to
> > unity mapped regions is blocked by the IOMMU. Make sure this does not
> > happen by initializing the device table after dma_ops.
> > 
> > Back-port upstream commit: f528d980c17b8714aedc918ba86e058af914d66b
> > 
> > Signed-off-by: Joerg Roedel <joro@8bytes.org>
> > Signed-off-by: Shuah Khan <shuah.khan@hp.com>
> > CC: stable@vger.kernel.org 3.4
> 
> Please consider this for 3.5 as well.
> 
> -- Shuah

Please consider this for 3.5. Patch applies and fix works on 3.5.7. My
previous reply to Herton Ronaldo Krzesinski bounced. Including Luis
Henriques on this one.

-- Shuah

Re: [PATCH 3.4] iommu/amd: Initialize device table after dma_ops

[Luis Henriques]

Hi Shuah,

On Thu, Feb 28, 2013 at 12:52:00PM -0700, Shuah Khan wrote:
> On Tue, 2013-02-26 at 16:49 -0700, Shuah Khan wrote:
> > When dma_ops are initialized the unity mappings are created. The
> > init_device_table_dma() function makes sure DMA from all devices is
> > blocked by default. This opens a short window in time where DMA to
> > unity mapped regions is blocked by the IOMMU. Make sure this does not
> > happen by initializing the device table after dma_ops.
> > 
> > Back-port upstream commit: f528d980c17b8714aedc918ba86e058af914d66b
> > 
> > Signed-off-by: Joerg Roedel <joro@8bytes.org>
> > Signed-off-by: Shuah Khan <shuah.khan@hp.com>
> > CC: stable@vger.kernel.org 3.4
> 
> Please consider this for 3.5 as well.

Thank you, I will queue this for the 3.5 kernel.

Cheers,
--
Luis

> 
> -- Shuah
> > ---
> >  drivers/iommu/amd_iommu_init.c |    9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/iommu/amd_iommu_init.c b/drivers/iommu/amd_iommu_init.c
> > index ef0ae93..b573f80 100644
> > --- a/drivers/iommu/amd_iommu_init.c
> > +++ b/drivers/iommu/amd_iommu_init.c
> > @@ -1572,8 +1572,6 @@ int __init amd_iommu_init_hardware(void)
> >  	if (amd_iommu_pd_alloc_bitmap == NULL)
> >  		goto free;
> >  
> > -	/* init the device table */
> > -	init_device_table();
> >  
> >  	/*
> >  	 * let all alias entries point to itself
> > @@ -1655,6 +1653,7 @@ out:
> >   */
> >  static int __init amd_iommu_init(void)
> >  {
> > +	struct amd_iommu *iommu;
> >  	int ret = 0;
> >  
> >  	ret = amd_iommu_init_hardware();
> > @@ -1673,6 +1672,12 @@ static int __init amd_iommu_init(void)
> >  	if (ret)
> >  		goto free;
> >  
> > +	/* init the device table */
> > +	init_device_table();
> > +
> > +	for_each_iommu(iommu)
> > +		iommu_flush_all_caches(iommu);
> > +
> >  	amd_iommu_init_api();
> >  
> >  	x86_platform.iommu_shutdown = disable_iommus;
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe stable" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html