Direct gtt range is used in the page fault scene in current driver, instead of stolen page. So no need to keep relative process.
--- drivers/gpu/drm/gma500/gem.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/gma500/gem.c b/drivers/gpu/drm/gma500/gem.c index 6d1cb6b..53cb6704 100644 --- a/drivers/gpu/drm/gma500/gem.c +++ b/drivers/gpu/drm/gma500/gem.c @@ -201,10 +201,7 @@ int psb_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf) >> PAGE_SHIFT;
/* CPU view of the page, don't go via the GART for CPU writes */ - if (r->stolen) - pfn = (dev_priv->stolen_base + r->offset) >> PAGE_SHIFT; - else - pfn = page_to_pfn(r->pages[page_offset]); + pfn = page_to_pfn(r->pages[page_offset]); ret = vm_insert_pfn(vma, (unsigned long)vmf->virtual_address, pfn);
fail: -- 2.1.0
On Tue, Aug 30, 2016 at 7:10 AM, jiang.biao2@zte.com.cn wrote:
Direct gtt range is used in the page fault scene in current driver, instead of stolen page. So no need to keep relative process.
Hi
Are you saying that we don't use stolen memory? Afaik stolen memory should be accessed through the stolen range so we do need this.
-Patrik
drivers/gpu/drm/gma500/gem.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/gma500/gem.c b/drivers/gpu/drm/gma500/gem.c index 6d1cb6b..53cb6704 100644 --- a/drivers/gpu/drm/gma500/gem.c +++ b/drivers/gpu/drm/gma500/gem.c @@ -201,10 +201,7 @@ int psb_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf) >> PAGE_SHIFT;
/* CPU view of the page, don't go via the GART for CPU writes */
if (r->stolen)pfn = (dev_priv->stolen_base + r->offset) >> PAGE_SHIFT;elsepfn = page_to_pfn(r->pages[page_offset]);
pfn = page_to_pfn(r->pages[page_offset]); ret = vm_insert_pfn(vma, (unsigned long)vmf->virtual_address, pfn);fail:
2.1.0
Patrik Jakobsson patrik.r.jakobsson@gmail.com wrote on 2016/08/30 18:21:08:
Patrik Jakobsson patrik.r.jakobsson@gmail.com 2016/08/30 18:21
From jiang.biao2@zte.com.cn, cc dri-devel dri-devel@lists.freedesktop.org Re: [PATCH] drm/gma500: remove the process of stolen page in page fault handler.
On Tue, Aug 30, 2016 at 7:10 AM, jiang.biao2@zte.com.cn wrote:
Direct gtt range is used in the page fault scene in current driver, instead of stolen page. So no need to keep relative process.
Hi
Are you saying that we don't use stolen memory? Afaik stolen memory should be accessed through the stolen range so we do need this.
-Patrik
As far as I can see, the stolen memory is only used by fbdev driver in gma500, but the fbdev driver maps the stloen memory directly in psbfb_vm_fault, not using psb_gem_fault to map the stolen memory. The only scenario using the psb_gem_fault is the gtt range created by psb_gem_create, which alloc the gtt range without stolen memory backed.
If I missed something, pls enlighten me. Thanks a lot.
dri-devel@lists.freedesktop.org
-
jiang.biao2＠zte.com.cn -
Patrik Jakobsson