[PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Vikas Sajjan]

While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
connected with resolution 2560x1600, following error occured even with
IOMMU enabled:
[0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate buffer.
[0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0

This patch fixes the issue by adding a check for IOMMU.

Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
Signed-off-by: Arun Kumar <arun.kk@samsung.com>
---
 drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
index 8e60bd6..2a86666 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
@@ -16,6 +16,7 @@
 #include <drm/drm_crtc.h>
 #include <drm/drm_fb_helper.h>
 #include <drm/drm_crtc_helper.h>
+#include <drm/exynos_drm.h>
 
 #include "exynos_drm_drv.h"
 #include "exynos_drm_fb.h"
@@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct drm_fb_helper *helper,
 	struct platform_device *pdev = dev->platformdev;
 	unsigned long size;
 	int ret;
+	unsigned int flag;
 
 	DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
 			sizes->surface_width, sizes->surface_height,
@@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct drm_fb_helper *helper,
 	size = mode_cmd.pitches[0] * mode_cmd.height;
 
 	/* 0 means to allocate physically continuous memory */
-	exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
+	if (!is_drm_iommu_supported(dev))
+		flag = 0;
+	else
+		flag = EXYNOS_BO_NONCONTIG;
+
+	exynos_gem_obj = exynos_drm_gem_create(dev, flag, size);
 	if (IS_ERR(exynos_gem_obj)) {
 		ret = PTR_ERR(exynos_gem_obj);
 		goto err_release_framebuffer;
-- 
1.7.9.5

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Tomasz Figa]

Hi Vikas,

On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
> While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
> connected with resolution 2560x1600, following error occured even with
> IOMMU enabled:
> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
> buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
> 
> This patch fixes the issue by adding a check for IOMMU.
> 
> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
> ---
>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
> 100644
> --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> @@ -16,6 +16,7 @@
>  #include <drm/drm_crtc.h>
>  #include <drm/drm_fb_helper.h>
>  #include <drm/drm_crtc_helper.h>
> +#include <drm/exynos_drm.h>
> 
>  #include "exynos_drm_drv.h"
>  #include "exynos_drm_fb.h"
> @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
> drm_fb_helper *helper, struct platform_device *pdev = dev->platformdev;
>  	unsigned long size;
>  	int ret;
> +	unsigned int flag;
> 
>  	DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
>  			sizes->surface_width, sizes->surface_height,
> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
> drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
> 
>  	/* 0 means to allocate physically continuous memory */
> -	exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
> +	if (!is_drm_iommu_supported(dev))
> +		flag = 0;
> +	else
> +		flag = EXYNOS_BO_NONCONTIG;

While noncontig memory might be used for devices that support IOMMU, there 
should be no problem with using contig memory for them, so this seems more 
like masking the original problem rather than tracking it down.

Could you check why the allocation fails when requesting contiguous 
memory?

Best regards,
Tomasz

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Rob Clark]

On Thu, Aug 1, 2013 at 7:20 PM, Tomasz Figa <tomasz.figa@gmail.com> wrote:
> Hi Vikas,
>
> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
>> While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
>> connected with resolution 2560x1600, following error occured even with
>> IOMMU enabled:
>> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
>> buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
>>
>> This patch fixes the issue by adding a check for IOMMU.
>>
>> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
>> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
>> ---
>>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
>>  1 file changed, 8 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
>> 100644
>> --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> @@ -16,6 +16,7 @@
>>  #include <drm/drm_crtc.h>
>>  #include <drm/drm_fb_helper.h>
>>  #include <drm/drm_crtc_helper.h>
>> +#include <drm/exynos_drm.h>
>>
>>  #include "exynos_drm_drv.h"
>>  #include "exynos_drm_fb.h"
>> @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
>> drm_fb_helper *helper, struct platform_device *pdev = dev->platformdev;
>>       unsigned long size;
>>       int ret;
>> +     unsigned int flag;
>>
>>       DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
>>                       sizes->surface_width, sizes->surface_height,
>> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
>> drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
>>
>>       /* 0 means to allocate physically continuous memory */
>> -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
>> +     if (!is_drm_iommu_supported(dev))
>> +             flag = 0;
>> +     else
>> +             flag = EXYNOS_BO_NONCONTIG;
>
> While noncontig memory might be used for devices that support IOMMU, there
> should be no problem with using contig memory for them, so this seems more
> like masking the original problem rather than tracking it down.

it is probably a good idea to not require contig memory when it is not
needed for performance or functionality (and if it is only
performance, then fallback gracefully to non-contig).. but yeah, would
be good to know if this is masking another issue all the same

BR,
-R

> Could you check why the allocation fails when requesting contiguous
> memory?
>
> Best regards,
> Tomasz
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-media" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Vikas Sajjan]

Hi Tomasz,


On 2 August 2013 04:50, Tomasz Figa <tomasz.figa@gmail.com> wrote:
>
> Hi Vikas,
>
> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
> > While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
> > connected with resolution 2560x1600, following error occured even with
> > IOMMU enabled:
> > [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
> > buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
> >
> > This patch fixes the issue by adding a check for IOMMU.
> >
> > Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
> > Signed-off-by: Arun Kumar <arun.kk@samsung.com>
> > ---
> >  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
> >  1 file changed, 8 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> > b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
> > 100644
> > --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> > +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> > @@ -16,6 +16,7 @@
> >  #include <drm/drm_crtc.h>
> >  #include <drm/drm_fb_helper.h>
> >  #include <drm/drm_crtc_helper.h>
> > +#include <drm/exynos_drm.h>
> >
> >  #include "exynos_drm_drv.h"
> >  #include "exynos_drm_fb.h"
> > @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
> > drm_fb_helper *helper, struct platform_device *pdev = dev->platformdev;
> >       unsigned long size;
> >       int ret;
> > +     unsigned int flag;
> >
> >       DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
> >                       sizes->surface_width, sizes->surface_height,
> > @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
> > drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
> >
> >       /* 0 means to allocate physically continuous memory */
> > -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
> > +     if (!is_drm_iommu_supported(dev))
> > +             flag = 0;
> > +     else
> > +             flag = EXYNOS_BO_NONCONTIG;
>
> While noncontig memory might be used for devices that support IOMMU, there
> should be no problem with using contig memory for them, so this seems more
> like masking the original problem rather than tracking it down.
>
> Could you check why the allocation fails when requesting contiguous
> memory?
>

It is failing if i am giving bigger resolution like 2560x1600, but if
i try for 1280x780 resolution, it   works fine.
Looks like arm_dma_alloc() is NOT able to allocate bigger buffer of
size 2560 * 1600 * 4.
Hence i used flag = EXYNOS_BO_NONCONTIG;


>
> Best regards,
> Tomasz
>



-- 
Thanks and Regards
 Vikas Sajjan

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Sachin Kamat]

Hi Vikas,

On 2 August 2013 09:23, Vikas Sajjan <vikas.sajjan@linaro.org> wrote:
> Hi Tomasz,
>
>
> On 2 August 2013 04:50, Tomasz Figa <tomasz.figa@gmail.com> wrote:
>>
>> Hi Vikas,
>>
>> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
>> > While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
>> > connected with resolution 2560x1600, following error occured even with
>> > IOMMU enabled:
>> > [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
>> > buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
>> >
>> > This patch fixes the issue by adding a check for IOMMU.
>> >
>> > Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
>> > Signed-off-by: Arun Kumar <arun.kk@samsung.com>
>> > ---
>> >  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
>> >  1 file changed, 8 insertions(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> > b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
>> > 100644
>> > --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> > +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> > @@ -16,6 +16,7 @@
>> >  #include <drm/drm_crtc.h>
>> >  #include <drm/drm_fb_helper.h>
>> >  #include <drm/drm_crtc_helper.h>
>> > +#include <drm/exynos_drm.h>
>> >
>> >  #include "exynos_drm_drv.h"
>> >  #include "exynos_drm_fb.h"
>> > @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
>> > drm_fb_helper *helper, struct platform_device *pdev = dev->platformdev;
>> >       unsigned long size;
>> >       int ret;
>> > +     unsigned int flag;
>> >
>> >       DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
>> >                       sizes->surface_width, sizes->surface_height,
>> > @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
>> > drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
>> >
>> >       /* 0 means to allocate physically continuous memory */
>> > -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
>> > +     if (!is_drm_iommu_supported(dev))
>> > +             flag = 0;
>> > +     else
>> > +             flag = EXYNOS_BO_NONCONTIG;
>>
>> While noncontig memory might be used for devices that support IOMMU, there
>> should be no problem with using contig memory for them, so this seems more
>> like masking the original problem rather than tracking it down.
>>
>> Could you check why the allocation fails when requesting contiguous
>> memory?
>>
>
> It is failing if i am giving bigger resolution like 2560x1600, but if
> i try for 1280x780 resolution, it   works fine.
> Looks like arm_dma_alloc() is NOT able to allocate bigger buffer of
> size 2560 * 1600 * 4.

You may need to increase the zoneorder (from current default 11 to say
13) if you need large physically contiguous pages.
However I am not sure if it is recommended.

-- 
With warm regards,
Sachin

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Sachin Kamat]

Hi Vikas,

On 1 August 2013 16:49, Vikas Sajjan <vikas.sajjan@linaro.org> wrote:
> While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
> connected with resolution 2560x1600, following error occured even with
> IOMMU enabled:
> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate buffer.
> [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
>
> This patch fixes the issue by adding a check for IOMMU.
>
> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
> ---
>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> index 8e60bd6..2a86666 100644
> --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> @@ -16,6 +16,7 @@
>  #include <drm/drm_crtc.h>
>  #include <drm/drm_fb_helper.h>
>  #include <drm/drm_crtc_helper.h>
> +#include <drm/exynos_drm.h>
>
>  #include "exynos_drm_drv.h"
>  #include "exynos_drm_fb.h"
> @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct drm_fb_helper *helper,
>         struct platform_device *pdev = dev->platformdev;
>         unsigned long size;
>         int ret;
> +       unsigned int flag;
>
>         DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
>                         sizes->surface_width, sizes->surface_height,
> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct drm_fb_helper *helper,
>         size = mode_cmd.pitches[0] * mode_cmd.height;
>
>         /* 0 means to allocate physically continuous memory */

This comment is now wrongly placed. Please use EXYNOS_BO_CONTIG instead of 0
and get rid of this comment altogether.

> -       exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
> +       if (!is_drm_iommu_supported(dev))
> +               flag = 0;
> +       else
> +               flag = EXYNOS_BO_NONCONTIG;
> +
> +       exynos_gem_obj = exynos_drm_gem_create(dev, flag, size);
>         if (IS_ERR(exynos_gem_obj)) {
>                 ret = PTR_ERR(exynos_gem_obj);
>                 goto err_release_framebuffer;

-- 
With warm regards,
Sachin

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Vikas Sajjan]

Hi Rob,

On 2 August 2013 06:03, Rob Clark <robdclark@gmail.com> wrote:
> On Thu, Aug 1, 2013 at 7:20 PM, Tomasz Figa <tomasz.figa@gmail.com> wrote:
>> Hi Vikas,
>>
>> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
>>> While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
>>> connected with resolution 2560x1600, following error occured even with
>>> IOMMU enabled:
>>> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
>>> buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
>>>
>>> This patch fixes the issue by adding a check for IOMMU.
>>>
>>> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
>>> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
>>> ---
>>>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
>>>  1 file changed, 8 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>>> b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
>>> 100644
>>> --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>>> +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>>> @@ -16,6 +16,7 @@
>>>  #include <drm/drm_crtc.h>
>>>  #include <drm/drm_fb_helper.h>
>>>  #include <drm/drm_crtc_helper.h>
>>> +#include <drm/exynos_drm.h>
>>>
>>>  #include "exynos_drm_drv.h"
>>>  #include "exynos_drm_fb.h"
>>> @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
>>> drm_fb_helper *helper, struct platform_device *pdev = dev->platformdev;
>>>       unsigned long size;
>>>       int ret;
>>> +     unsigned int flag;
>>>
>>>       DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
>>>                       sizes->surface_width, sizes->surface_height,
>>> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
>>> drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
>>>
>>>       /* 0 means to allocate physically continuous memory */
>>> -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
>>> +     if (!is_drm_iommu_supported(dev))
>>> +             flag = 0;
>>> +     else
>>> +             flag = EXYNOS_BO_NONCONTIG;
>>
>> While noncontig memory might be used for devices that support IOMMU, there
>> should be no problem with using contig memory for them, so this seems more
>> like masking the original problem rather than tracking it down.
>
> it is probably a good idea to not require contig memory when it is not
> needed for performance or functionality (and if it is only
> performance, then fallback gracefully to non-contig).. but yeah, would
> be good to know if this is masking another issue all the same
>

Whats happening with CONTIG flag and with IOMMU,  is

 __iommu_alloc_buffer() ---> dma_alloc_from_contiguous() and in this
function it fails at
this condition check  if (pageno >= cma->count)

So I tried increasing the CONFIG_CMA_SIZE_MBYTES to 24,  this check
succeeds and it works well without my patch.

But what about the case where CONFIG_CMA is disabled , yet i want
bigger memory for a device.
 I think using IOMMU we can achieve this.

 correct me, if i am wrong.


> BR,
> -R
>
>> Could you check why the allocation fails when requesting contiguous
>> memory?
>>
>> Best regards,
>> Tomasz
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-media" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html



-- 
Thanks and Regards
 Vikas Sajjan

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Tomasz Figa]

Hi Vikas,

On Friday 02 of August 2013 12:08:52 Vikas Sajjan wrote:
> Hi Rob,
> 
> On 2 August 2013 06:03, Rob Clark <robdclark@gmail.com> wrote:
> > On Thu, Aug 1, 2013 at 7:20 PM, Tomasz Figa <tomasz.figa@gmail.com> 
wrote:
> >> Hi Vikas,
> >> 
> >> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
> >>> While trying to get boot-logo up on exynos5420 SMDK which has eDP
> >>> panel
> >>> connected with resolution 2560x1600, following error occured even
> >>> with
> >>> IOMMU enabled:
> >>> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
> >>> buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor
> >>> 0
> >>> 
> >>> This patch fixes the issue by adding a check for IOMMU.
> >>> 
> >>> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
> >>> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
> >>> ---
> >>> 
> >>>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
> >>>  1 file changed, 8 insertions(+), 1 deletion(-)
[snip]
> >>> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
> >>> drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
> >>> 
> >>>       /* 0 means to allocate physically continuous memory */
> >>> 
> >>> -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
> >>> +     if (!is_drm_iommu_supported(dev))
> >>> +             flag = 0;
> >>> +     else
> >>> +             flag = EXYNOS_BO_NONCONTIG;
> >> 
> >> While noncontig memory might be used for devices that support IOMMU,
> >> there should be no problem with using contig memory for them, so
> >> this seems more like masking the original problem rather than
> >> tracking it down.> 
> > it is probably a good idea to not require contig memory when it is not
> > needed for performance or functionality (and if it is only
> > performance, then fallback gracefully to non-contig).. but yeah, would
> > be good to know if this is masking another issue all the same
> 
> Whats happening with CONTIG flag and with IOMMU,  is
> 
>  __iommu_alloc_buffer() ---> dma_alloc_from_contiguous() and in this
> function it fails at
> this condition check  if (pageno >= cma->count)
> 
> So I tried increasing the CONFIG_CMA_SIZE_MBYTES to 24,  this check
> succeeds and it works well without my patch.
> 
> But what about the case where CONFIG_CMA is disabled , yet i want
> bigger memory for a device.
>  I think using IOMMU we can achieve this.
>
>  correct me, if i am wrong.

This is probably fine. I'm not sure about performance aspects of using 
noncontig memory as framebuffer, though. This needs to be checked and if 
there is some performance penalty, I would make noncontig allocation a 
fallback case, if contig fails, as Rob has suggested.

Also I think you should adjust the commit message to say that non-
contiguous memory can be used when IOMMU is supported, so there is no need 
to force contiguous allocations, since this is not a bug fix, but rather a 
feature this patch is adding.

Best regards,
Tomasz
 
> > BR,
> > -R
> > 
> >> Could you check why the allocation fails when requesting contiguous
> >> memory?
> >> 
> >> Best regards,
> >> Tomasz
> >> 
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe
> >> linux-media" in the body of a message to majordomo@vger.kernel.org
> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Inki Dae]

[-- Attachment #1.1: Type: text/plain, Size: 4616 bytes --]

2013/8/2 Vikas Sajjan <vikas.sajjan@linaro.org>

> Hi Rob,
>
> On 2 August 2013 06:03, Rob Clark <robdclark@gmail.com> wrote:
> > On Thu, Aug 1, 2013 at 7:20 PM, Tomasz Figa <tomasz.figa@gmail.com>
> wrote:
> >> Hi Vikas,
> >>
> >> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
> >>> While trying to get boot-logo up on exynos5420 SMDK which has eDP panel
> >>> connected with resolution 2560x1600, following error occured even with
> >>> IOMMU enabled:
> >>> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
> >>> buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
> >>>
> >>> This patch fixes the issue by adding a check for IOMMU.
> >>>
> >>> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
> >>> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
> >>> ---
> >>>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
> >>>  1 file changed, 8 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> >>> b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
> >>> 100644
> >>> --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> >>> +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
> >>> @@ -16,6 +16,7 @@
> >>>  #include <drm/drm_crtc.h>
> >>>  #include <drm/drm_fb_helper.h>
> >>>  #include <drm/drm_crtc_helper.h>
> >>> +#include <drm/exynos_drm.h>
> >>>
> >>>  #include "exynos_drm_drv.h"
> >>>  #include "exynos_drm_fb.h"
> >>> @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
> >>> drm_fb_helper *helper, struct platform_device *pdev = dev->platformdev;
> >>>       unsigned long size;
> >>>       int ret;
> >>> +     unsigned int flag;
> >>>
> >>>       DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
> >>>                       sizes->surface_width, sizes->surface_height,
> >>> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
> >>> drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
> >>>
> >>>       /* 0 means to allocate physically continuous memory */
> >>> -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
> >>> +     if (!is_drm_iommu_supported(dev))
> >>> +             flag = 0;
> >>> +     else
> >>> +             flag = EXYNOS_BO_NONCONTIG;
> >>
> >> While noncontig memory might be used for devices that support IOMMU,
> there
> >> should be no problem with using contig memory for them, so this seems
> more
> >> like masking the original problem rather than tracking it down.
> >
> > it is probably a good idea to not require contig memory when it is not
> > needed for performance or functionality (and if it is only
> > performance, then fallback gracefully to non-contig).. but yeah, would
> > be good to know if this is masking another issue all the same
> >
>
> Whats happening with CONTIG flag and with IOMMU,  is
>
>  __iommu_alloc_buffer() ---> dma_alloc_from_contiguous() and in this
> function it fails at
> this condition check  if (pageno >= cma->count)
>
> So I tried increasing the CONFIG_CMA_SIZE_MBYTES to 24,  this check
> succeeds and it works well without my patch.
>
> But what about the case where CONFIG_CMA is disabled , yet i want
> bigger memory for a device.
>  I think using IOMMU we can achieve this.
>
>  correct me, if i am wrong.
>
>
I'm on summer vacation so I'm afraid that I cannot test and look into it
but I guess you guy didn't declare CMA region for Exynos drm. And in this
case, the size of CMA declared region is 16MB as default. That is why works
well after increasing default size, CONFIG_CMA_SIZE_MBYTES, to 24MB. And I
mentioned long time ago, we are required to use physically contiguous
memory in case that bootloader uses physically contiguous memory for its
own framebuffer, and kernel wants to share the bootloader's framebuffer
region to resolve flickering issue while booted; that is required for
product. And one question, is there any reason that you guy should use
non-contiguous memory for framebuffer with iommu?

Thanks,
Inki Dae


>
> > BR,
> > -R
> >
> >> Could you check why the allocation fails when requesting contiguous
> >> memory?
> >>
> >> Best regards,
> >> Tomasz
> >>
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe linux-media"
> in
> >> the body of a message to majordomo@vger.kernel.org
> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>
>
> --
> Thanks and Regards
>  Vikas Sajjan
> --
> To unsubscribe from this list: send the line "unsubscribe linux-media" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

[-- Attachment #1.2: Type: text/html, Size: 6714 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Vikas Sajjan]

Hi Inki Dae,

On 2 August 2013 12:58, Inki Dae <inki.dae@samsung.com> wrote:
>
>
> 2013/8/2 Vikas Sajjan <vikas.sajjan@linaro.org>
>>
>> Hi Rob,
>>
>> On 2 August 2013 06:03, Rob Clark <robdclark@gmail.com> wrote:
>> > On Thu, Aug 1, 2013 at 7:20 PM, Tomasz Figa <tomasz.figa@gmail.com>
>> > wrote:
>> >> Hi Vikas,
>> >>
>> >> On Thursday 01 of August 2013 16:49:32 Vikas Sajjan wrote:
>> >>> While trying to get boot-logo up on exynos5420 SMDK which has eDP
>> >>> panel
>> >>> connected with resolution 2560x1600, following error occured even with
>> >>> IOMMU enabled:
>> >>> [0.880000] [drm:lowlevel_buffer_allocate] *ERROR* failed to allocate
>> >>> buffer. [0.890000] [drm] Initialized exynos 1.0.0 20110530 on minor 0
>> >>>
>> >>> This patch fixes the issue by adding a check for IOMMU.
>> >>>
>> >>> Signed-off-by: Vikas Sajjan <vikas.sajjan@linaro.org>
>> >>> Signed-off-by: Arun Kumar <arun.kk@samsung.com>
>> >>> ---
>> >>>  drivers/gpu/drm/exynos/exynos_drm_fbdev.c |    9 ++++++++-
>> >>>  1 file changed, 8 insertions(+), 1 deletion(-)
>> >>>
>> >>> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> >>> b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c index 8e60bd6..2a86666
>> >>> 100644
>> >>> --- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> >>> +++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
>> >>> @@ -16,6 +16,7 @@
>> >>>  #include <drm/drm_crtc.h>
>> >>>  #include <drm/drm_fb_helper.h>
>> >>>  #include <drm/drm_crtc_helper.h>
>> >>> +#include <drm/exynos_drm.h>
>> >>>
>> >>>  #include "exynos_drm_drv.h"
>> >>>  #include "exynos_drm_fb.h"
>> >>> @@ -143,6 +144,7 @@ static int exynos_drm_fbdev_create(struct
>> >>> drm_fb_helper *helper, struct platform_device *pdev =
>> >>> dev->platformdev;
>> >>>       unsigned long size;
>> >>>       int ret;
>> >>> +     unsigned int flag;
>> >>>
>> >>>       DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
>> >>>                       sizes->surface_width, sizes->surface_height,
>> >>> @@ -166,7 +168,12 @@ static int exynos_drm_fbdev_create(struct
>> >>> drm_fb_helper *helper, size = mode_cmd.pitches[0] * mode_cmd.height;
>> >>>
>> >>>       /* 0 means to allocate physically continuous memory */
>> >>> -     exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
>> >>> +     if (!is_drm_iommu_supported(dev))
>> >>> +             flag = 0;
>> >>> +     else
>> >>> +             flag = EXYNOS_BO_NONCONTIG;
>> >>
>> >> While noncontig memory might be used for devices that support IOMMU,
>> >> there
>> >> should be no problem with using contig memory for them, so this seems
>> >> more
>> >> like masking the original problem rather than tracking it down.
>> >
>> > it is probably a good idea to not require contig memory when it is not
>> > needed for performance or functionality (and if it is only
>> > performance, then fallback gracefully to non-contig).. but yeah, would
>> > be good to know if this is masking another issue all the same
>> >
>>
>> Whats happening with CONTIG flag and with IOMMU,  is
>>
>>  __iommu_alloc_buffer() ---> dma_alloc_from_contiguous() and in this
>> function it fails at
>> this condition check  if (pageno >= cma->count)
>>
>> So I tried increasing the CONFIG_CMA_SIZE_MBYTES to 24,  this check
>> succeeds and it works well without my patch.
>>
>> But what about the case where CONFIG_CMA is disabled , yet i want
>> bigger memory for a device.
>>  I think using IOMMU we can achieve this.
>>
>>  correct me, if i am wrong.
>>
>
> I'm on summer vacation so I'm afraid that I cannot test and look into it but
> I guess you guy didn't declare CMA region for Exynos drm. And in this case,
> the size of CMA declared region is 16MB as default. That is why works well
> after increasing default size, CONFIG_CMA_SIZE_MBYTES, to 24MB. And I
> mentioned long time ago, we are required to use physically contiguous memory
> in case that bootloader uses physically contiguous memory for its own
> framebuffer, and kernel wants to share the bootloader's framebuffer region
> to resolve flickering issue while booted; that is required for product. And
> one question, is there any reason that you guy should use non-contiguous
> memory for framebuffer with iommu?
>

yeah, we could not allocate CMA region for FIMD, because the function
dma_declare_contiguous() needs "dev" as the first argument and we have
access to "dev" node only if it is NON-DT way of probing like the way
it is done in arch/arm/mach-davinci/devices-da8xx.c
But now, since the probing is through DT way, there is NO way ( Let me
know if something is newly added ) to call dma_declare_contiguous()
and reserve CMA region .

we don't have any specific requirement for NON_CONTIG or CONTIG
memory, but only requirement was to allocate a bigger memory like (
2560 * 1600 * 4 ) for FB.

But as Rob suggested, we should have fall-back case if CONTIG memory
allocation fails, as below

diff --git a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
index df43fa9..15de626 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_fbdev.c
@@ -144,7 +144,6 @@ static int exynos_drm_fbdev_create(struct
drm_fb_helper *helper,
        struct platform_device *pdev = dev->platformdev;
        unsigned long size;
        int ret;

        DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
                        sizes->surface_width, sizes->surface_height,
@@ -167,16 +166,14 @@ static int exynos_drm_fbdev_create(struct
drm_fb_helper *helper,

        size = mode_cmd.pitches[0] * mode_cmd.height;

-       /* 0 means to allocate physically continuous memory */
-       exynos_gem_obj = exynos_drm_gem_create(dev, 0, size);
+       exynos_gem_obj = exynos_drm_gem_create(dev, EXYNOS_BO_CONTIG, size);
        if (IS_ERR(exynos_gem_obj)) {
-               ret = PTR_ERR(exynos_gem_obj);
-               goto err_release_framebuffer;
+               if(is_drm_iommu_supported(dev))
+                       exynos_gem_obj = exynos_drm_gem_create(dev,
EXYNOS_BO_NONCONTIG, size);
+               if (IS_ERR(exynos_gem_obj)) {
+                       ret = PTR_ERR(exynos_gem_obj);
+                       goto err_release_framebuffer;
+               }
+               dev_warn("\n exynos_gem_obj for FB is allocated with
non physically continuous       +               memory \n");
        }



> Thanks,
> Inki Dae
>
>>
>>
>> > BR,
>> > -R
>> >
>> >> Could you check why the allocation fails when requesting contiguous
>> >> memory?
>> >>
>> >> Best regards,
>> >> Tomasz
>> >>
>> >> --
>> >> To unsubscribe from this list: send the line "unsubscribe linux-media"
>> >> in
>> >> the body of a message to majordomo@vger.kernel.org
>> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>>
>>
>> --
>> Thanks and Regards
>>  Vikas Sajjan
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-media" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>



-- 
Thanks and Regards
 Vikas Sajjan

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Sylwester Nawrocki]

Hi Vikas,

On 08/02/2013 12:10 PM, Vikas Sajjan wrote:
> yeah, we could not allocate CMA region for FIMD, because the function
> dma_declare_contiguous() needs "dev" as the first argument and we have
> access to "dev" node only if it is NON-DT way of probing like the way
> it is done in arch/arm/mach-davinci/devices-da8xx.c
> But now, since the probing is through DT way, there is NO way ( Let me
> know if something is newly added ) to call dma_declare_contiguous()
> and reserve CMA region .

See this patch series [1]. We have have been using this kind of bindings
for assigning physically contiguous memory regions to the Exynos 
multimedia devices, instead of what's currently in mainline where same
physical addresses are repeated in dts for various boards without much
thought. And where custom device specific parsing code is required at 
arch side.

$ git grep mfc\-[lr] arch/arm/boot/dts

arch/arm/boot/dts/exynos4210-origen.dts:     samsung,mfc-r = <0x43000000 0x800000>;
arch/arm/boot/dts/exynos4210-origen.dts:     samsung,mfc-l = <0x51000000 0x800000>;
arch/arm/boot/dts/exynos4210-smdkv310.dts:   samsung,mfc-r = <0x43000000 0x800000>;
arch/arm/boot/dts/exynos4210-smdkv310.dts:   samsung,mfc-l = <0x51000000 0x800000>;
arch/arm/boot/dts/exynos4412-origen.dts:     samsung,mfc-r = <0x43000000 0x800000>;
arch/arm/boot/dts/exynos4412-origen.dts:     samsung,mfc-l = <0x51000000 0x800000>;
arch/arm/boot/dts/exynos4412-smdk4412.dts:   samsung,mfc-r = <0x43000000 0x800000>;
arch/arm/boot/dts/exynos4412-smdk4412.dts:   samsung,mfc-l = <0x51000000 0x800000>;
arch/arm/boot/dts/exynos5250-arndale.dts:    samsung,mfc-r = <0x43000000 0x800000>;
arch/arm/boot/dts/exynos5250-arndale.dts:    samsung,mfc-l = <0x51000000 0x800000>;
arch/arm/boot/dts/exynos5250-smdk5250.dts:   samsung,mfc-r = <0x43000000 0x800000>;
arch/arm/boot/dts/exynos5250-smdk5250.dts:   samsung,mfc-l = <0x51000000 0x800000>;


[1] http://www.spinics.net/lists/arm-kernel/msg263130.html

Regards,
Sylwester

Re: [PATCH] drm/exynos: Add check for IOMMU while passing physically continous memory flag

[Vikas Sajjan]

Hi Sylwester,

On 2 August 2013 16:28, Sylwester Nawrocki <s.nawrocki@samsung.com> wrote:
> Hi Vikas,
>
> On 08/02/2013 12:10 PM, Vikas Sajjan wrote:
>> yeah, we could not allocate CMA region for FIMD, because the function
>> dma_declare_contiguous() needs "dev" as the first argument and we have
>> access to "dev" node only if it is NON-DT way of probing like the way
>> it is done in arch/arm/mach-davinci/devices-da8xx.c
>> But now, since the probing is through DT way, there is NO way ( Let me
>> know if something is newly added ) to call dma_declare_contiguous()
>> and reserve CMA region .
>
> See this patch series [1]. We have have been using this kind of bindings
> for assigning physically contiguous memory regions to the Exynos
> multimedia devices, instead of what's currently in mainline where same
> physical addresses are repeated in dts for various boards without much
> thought. And where custom device specific parsing code is required at
> arch side.
>
> $ git grep mfc\-[lr] arch/arm/boot/dts
>
> arch/arm/boot/dts/exynos4210-origen.dts:     samsung,mfc-r = <0x43000000 0x800000>;
> arch/arm/boot/dts/exynos4210-origen.dts:     samsung,mfc-l = <0x51000000 0x800000>;
> arch/arm/boot/dts/exynos4210-smdkv310.dts:   samsung,mfc-r = <0x43000000 0x800000>;
> arch/arm/boot/dts/exynos4210-smdkv310.dts:   samsung,mfc-l = <0x51000000 0x800000>;
> arch/arm/boot/dts/exynos4412-origen.dts:     samsung,mfc-r = <0x43000000 0x800000>;
> arch/arm/boot/dts/exynos4412-origen.dts:     samsung,mfc-l = <0x51000000 0x800000>;
> arch/arm/boot/dts/exynos4412-smdk4412.dts:   samsung,mfc-r = <0x43000000 0x800000>;
> arch/arm/boot/dts/exynos4412-smdk4412.dts:   samsung,mfc-l = <0x51000000 0x800000>;
> arch/arm/boot/dts/exynos5250-arndale.dts:    samsung,mfc-r = <0x43000000 0x800000>;
> arch/arm/boot/dts/exynos5250-arndale.dts:    samsung,mfc-l = <0x51000000 0x800000>;
> arch/arm/boot/dts/exynos5250-smdk5250.dts:   samsung,mfc-r = <0x43000000 0x800000>;
> arch/arm/boot/dts/exynos5250-smdk5250.dts:   samsung,mfc-l = <0x51000000 0x800000>;
>
>
> [1] http://www.spinics.net/lists/arm-kernel/msg263130.html
>


Thanks, its good that now we have a new way to reserve CMA region.


> Regards,
> Sylwester



-- 
Thanks and Regards
 Vikas Sajjan