Hi

Am 06.11.19 um 12:48 schrieb Noralf Trønnes:
> 
> 
> Den 06.11.2019 11.47, skrev Thomas Zimmermann:
>> Simply removes all the obsolete GEM code from udl. No functional
>> changes.
>>
>> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
>> ---
>>  drivers/gpu/drm/udl/Makefile     |   2 +-
>>  drivers/gpu/drm/udl/udl_dmabuf.c | 254 -------------------------------
>>  drivers/gpu/drm/udl/udl_drv.h    |  29 ----
>>  drivers/gpu/drm/udl/udl_gem.c    | 206 -------------------------
>>  4 files changed, 1 insertion(+), 490 deletions(-)
>>  delete mode 100644 drivers/gpu/drm/udl/udl_dmabuf.c
>>
> 
> <snip>
> 
>> -int udl_gem_vmap(struct udl_gem_object *obj)
>> -{
>> -	int page_count = obj->base.size / PAGE_SIZE;
>> -	int ret;
>> -
>> -	if (obj->base.import_attach) {
>> -		obj->vmapping = dma_buf_vmap(obj->base.import_attach->dmabuf);
>> -		if (!obj->vmapping)
>> -			return -ENOMEM;
>> -		return 0;
>> -	}
>> -
>> -	ret = udl_gem_get_pages(obj);
>> -	if (ret)
>> -		return ret;
>> -
>> -	obj->vmapping = vmap(obj->pages, page_count, 0, PAGE_KERNEL);
> 
> This differs from the shmem helper vmap:
> 
> 	shmem->vaddr = vmap(shmem->pages, obj->size >> PAGE_SHIFT,
> 			    VM_MAP, pgprot_writecombine(PAGE_KERNEL));
> 
> Boris added the WC in be7d9f05c53e:
> 
>  drm/gem_shmem: Use a writecombine mapping for ->vaddr
> 
>  Right now, the BO is mapped as a cached region when ->vmap() is called
>  and the underlying object is not a dmabuf.
>  Doing that makes cache management a bit more complicated (you'd need
>  to call dma_map/unmap_sg() on the ->sgt field everytime the BO is about
>  to be passed to the GPU/CPU), so let's map the BO with writecombine
>  attributes instead (as done in most drivers).
> 
> I don't know what the implications of this are, just pointing out a
> difference.

Thanks! Switching to SHMEM disables caching on these pages. The logic
around dma_map/unmap_sg() is the same in udl and generic prime functions
from what I can tell.

Actually, I've never seen any difference in display performance when
modifying these settings. (DisplayLink is always slow.) I'd like to
throw away the caching optimization and rather tell userspace to
allocate a shadow buffer if necessary.

Best regards
Thomas

> Noralf.
> 
>> -	if (!obj->vmapping)
>> -		return -ENOMEM;
>> -	return 0;
>> -}
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
> 

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Felix Imendörffer