Re: [PATCH v6 03/25] drm/gem-dma: Compute dumb-buffer sizes with drm_mode_size_dumb()

[<Ludovic.Desroches@xxxxxxxxxxxxx>]

On 11/26/25 08:40, Thomas Zimmermann wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know 
> the content is safe
> 
> Hi
> 
> Am 26.11.25 um 07:24 schrieb Ludovic.Desroches@xxxxxxxxxxxxx:
>> On 11/25/25 16:03, Thomas Zimmermann wrote:
>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know
>>> the content is safe
>>>
>>> Hi
>>>
>>> Am 25.11.25 um 15:39 schrieb Ludovic.Desroches@xxxxxxxxxxxxx:
>>>> On 8/21/25 10:17, Thomas Zimmermann wrote:
>>>>> Call drm_mode_size_dumb() to compute dumb-buffer scanline pitch and
>>>>> buffer size. Align the pitch to a multiple of 8.
>>> I missed a chance to explain it here. :/
>> I was wondering if it was 8-bits or 8-bytes.
>>
>>>>> Push the current calculation into the only direct caller imx. Imx's
>>>>> hardware requires the framebuffer width to be aligned to 8. The
>>>>> driver's current approach is actually incorrect, as it only guarantees
>>>>> this implicitly and requires bpp to be a multiple of 8 already. A
>>>>> later commit will fix this problem by aligning the scanline pitch
>>>>> such that an aligned width still fits into each scanline's memory.
>>>>>
>>>>> A number of other drivers are build on top of gem-dma helpers and
>>>>> implement their own dumb-buffer allocation. These drivers invoke
>>>>> drm_gem_dma_dumb_create_internal(), which is not affected by this
>>>>> commit.
>>>>>
>>>>> v5:
>>>>> - avoid reset of arguments (Tomi)
>>>>>
>>>>> Signed-off-by: Thomas Zimmermann <tzimmermann@xxxxxxx>
>>>>> Reviewed-by: Tomi Valkeinen <tomi.valkeinen@xxxxxxxxxxxxxxxx>
>>>>> ---
>>>>>     drivers/gpu/drm/drm_gem_dma_helper.c     | 7 +++++--
>>>>>     drivers/gpu/drm/imx/ipuv3/imx-drm-core.c | 4 +++-
>>>>>     2 files changed, 8 insertions(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/drm_gem_dma_helper.c b/drivers/gpu/drm/
>>>>> drm_gem_dma_helper.c
>>>>> index 4f0320df858f..ab1a70b1d6f1 100644
>>>>> --- a/drivers/gpu/drm/drm_gem_dma_helper.c
>>>>> +++ b/drivers/gpu/drm/drm_gem_dma_helper.c
>>>>> @@ -20,6 +20,7 @@
>>>>>     #include <drm/drm.h>
>>>>>     #include <drm/drm_device.h>
>>>>>     #include <drm/drm_drv.h>
>>>>> +#include <drm/drm_dumb_buffers.h>
>>>>>     #include <drm/drm_gem_dma_helper.h>
>>>>>     #include <drm/drm_vma_manager.h>
>>>>>
>>>>> @@ -304,9 +305,11 @@ int drm_gem_dma_dumb_create(struct drm_file
>>>>> *file_priv,
>>>>>                           struct drm_mode_create_dumb *args)
>>>>>     {
>>>>>       struct drm_gem_dma_object *dma_obj;
>>>>> +    int ret;
>>>>>
>>>>> -    args->pitch = DIV_ROUND_UP(args->width * args->bpp, 8);
>>>>> -    args->size = args->pitch * args->height;
>>>>> +    ret = drm_mode_size_dumb(drm, args, SZ_8, 0);
>>>>> +    if (ret)
>>>>> +            return ret;
>>>> Hi,
>>>>
>>>> Was it intentional for this change to alter the pitch?
>>> Most hardware does not need the pitch to be of a certain alignment. But
>>> these buffers are possibly shared with other hardware, which sometimes
>>> needs alignment to certain values. Using SZ_8 improves compatible with
>>> that hardware.
>>>
>>>> The alignment requirement has been updated—from 8-bit alignment to
>>>> 64-bit alignment. Since the pitch is expressed in bytes, we should pass
>>>> SZ_1 instead of SZ_8 for hw_patch_align.
>>>>
>>>> For example, for an 850×480 framebuffer at 16 bpp, the pitch should be
>>>> 1700 bytes. With the new alignment, the pitch becomes 1704 bytes.
>>> Many display modes have an 8-byte alignment in their width. 850 pixels
>>> is somewhat of an exception.
>> Indeed, but overlay-type planes are often used to display windows that
>> are not the same size as the screen, and therefore potentially not on
>> 8-bytes aligned.
>>
>>>> Please let me know if you’d like me to submit a fix.
>>> Do you see a bug with your hardware? Unless this creates a real problem,
>>> I'd like to keep it as it is now.
>> Yes, that’s how I noticed this change. I have an application that
>> requests a framebuffer larger than my screen, in this case 850x480 in
>> RGB565 for an 800x480 display. So I have a pitch of 1700 bytes. Now it’s
>> using a pitch of 1704, which shifts my lines.
> 
> Honestly, your application is broken. The ioctl returns the pitch value
> to user space and that's what the program should use. Some hardware has
> hard constraints on the pitch, so you will sooner or later run into the
> same problem again.

Yes, I’m currently trying to find where the pitch information is being 
lost, given that we do correctly retrieve the value provided by the 
kernel after the ioctl.

> 
> Anyway, feel free to submit a patch to call drm_mode_size_dumb() without
> SZ_8. The best value would be 0, so that the function picks a default by
> itself. Since you're at it, maybe also look at patches 4 and 5 of this
> series. They have similar code for other memory managers.

I have no strong opinion about this 8-bytes alignment if it was 
intended. This could potentially break other applications that had been 
working until now, even if it does indeed reveal an issue on the 
application side. And it can result in larger buffers without it 
necessarily being needed.

I'll prepare patches and we'll see where that leads.

Regards,
Ludovic

> 
> Best regards
> Thomas
> 
>>
>> Regards,
>> Ludovic
>>
>>> Best regards
>>> Thomas
>>>
>>>> Regards,
>>>> Ludovic
>>>>
>>>>
>>>>>       dma_obj = drm_gem_dma_create_with_handle(file_priv, drm, args-
>>>>>> size,
>>>>>                                                &args->handle);
>>>>> diff --git a/drivers/gpu/drm/imx/ipuv3/imx-drm-core.c b/drivers/gpu/
>>>>> drm/imx/ipuv3/imx-drm-core.c
>>>>> index ec5fd9a01f1e..af4a30311e18 100644
>>>>> --- a/drivers/gpu/drm/imx/ipuv3/imx-drm-core.c
>>>>> +++ b/drivers/gpu/drm/imx/ipuv3/imx-drm-core.c
>>>>> @@ -145,8 +145,10 @@ static int imx_drm_dumb_create(struct drm_file
>>>>> *file_priv,
>>>>>       int ret;
>>>>>
>>>>>       args->width = ALIGN(width, 8);
>>>>> +    args->pitch = DIV_ROUND_UP(args->width * args->bpp, 8);
>>>>> +    args->size = args->pitch * args->height;
>>>>>
>>>>> -    ret = drm_gem_dma_dumb_create(file_priv, drm, args);
>>>>> +    ret = drm_gem_dma_dumb_create_internal(file_priv, drm, args);
>>>>>       if (ret)
>>>>>               return ret;
>>>>>
>>> -- 
>>> -- 
>>> Thomas Zimmermann
>>> Graphics Driver Developer
>>> SUSE Software Solutions Germany GmbH
>>> Frankenstr. 146, 90461 Nürnberg, Germany, www.suse.com
>>> GF: Jochen Jaser, Andrew McDonald, Werner Knoblich, (HRB 36809, AG
>>> Nürnberg)
>>>
>>>
>>
> 
> -- 
> -- 
> Thomas Zimmermann
> Graphics Driver Developer
> SUSE Software Solutions Germany GmbH
> Frankenstr. 146, 90461 Nürnberg, Germany, www.suse.com
> GF: Jochen Jaser, Andrew McDonald, Werner Knoblich, (HRB 36809, AG 
> Nürnberg)
> 
>