| Message ID | 1367382644-30788-1-git-send-email-airlied@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, May 1, 2013 at 6:30 AM, Dave Airlie <airlied@gmail.com> wrote: > Since we ask the dmabuf owner to map the dma-buf into our device > address space, but for udl at present that is the CPU address space, > since we don't DMA directly from the mapped buffer. > > However if we don't set a dma mask on the usb device, the mapping > ends up using swiotlb on machines that have it enabled, which > is less than desireable. > > Signed-off-by: Dave Airlie <airlied@redhat.com> Fyi for everyone else who was not on irc when Dave&I discussed this: This really shouldn't be required and I think the real issue is that udl creates a dma_buf attachement (which is needed for device dma only), but only really wants to do cpu access through vmap/kmap. So not attached the device should be good enough. Cc'ing a few more lists for better fyi ;-) -Daniel > --- > drivers/gpu/drm/udl/udl_main.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/gpu/drm/udl/udl_main.c b/drivers/gpu/drm/udl/udl_main.c > index 0ce2d71..6770e1b 100644 > --- a/drivers/gpu/drm/udl/udl_main.c > +++ b/drivers/gpu/drm/udl/udl_main.c > @@ -293,6 +293,7 @@ int udl_driver_load(struct drm_device *dev, unsigned long flags) > udl->ddev = dev; > dev->dev_private = udl; > > + dma_set_mask(dev->dev, DMA_BIT_MASK(64)); > if (!udl_parse_vendor_descriptor(dev, dev->usbdev)) { > DRM_ERROR("firmware not recognized. Assume incompatible device\n"); > goto err; > -- > 1.8.2 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/dri-devel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
>> >> However if we don't set a dma mask on the usb device, the mapping >> ends up using swiotlb on machines that have it enabled, which >> is less than desireable. >> >> Signed-off-by: Dave Airlie <airlied@redhat.com> > > Fyi for everyone else who was not on irc when Dave&I discussed this: > This really shouldn't be required and I think the real issue is that > udl creates a dma_buf attachement (which is needed for device dma > only), but only really wants to do cpu access through vmap/kmap. So > not attached the device should be good enough. Cc'ing a few more lists > for better fyi ;-) Though I've looked at this a bit more, and since I want to be able to expose shared objects as proper GEM objects from the import side I really need that list of pages. So it looks like this won't really fly. Dave.
On Mon, May 06, 2013 at 10:35:35AM +1000, Dave Airlie wrote: > >> > >> However if we don't set a dma mask on the usb device, the mapping > >> ends up using swiotlb on machines that have it enabled, which > >> is less than desireable. > >> > >> Signed-off-by: Dave Airlie <airlied@redhat.com> > > > > Fyi for everyone else who was not on irc when Dave&I discussed this: > > This really shouldn't be required and I think the real issue is that > > udl creates a dma_buf attachement (which is needed for device dma > > only), but only really wants to do cpu access through vmap/kmap. So > > not attached the device should be good enough. Cc'ing a few more lists > > for better fyi ;-) > > Though I've looked at this a bit more, and since I want to be able to expose > shared objects as proper GEM objects from the import side I really > need that list of pages. Hm, what does "proper GEM object" mean in the context of udl? -Daniel
On Tue, May 7, 2013 at 1:59 AM, Daniel Vetter <daniel@ffwll.ch> wrote: > On Mon, May 06, 2013 at 10:35:35AM +1000, Dave Airlie wrote: >> >> >> >> However if we don't set a dma mask on the usb device, the mapping >> >> ends up using swiotlb on machines that have it enabled, which >> >> is less than desireable. >> >> >> >> Signed-off-by: Dave Airlie <airlied@redhat.com> >> > >> > Fyi for everyone else who was not on irc when Dave&I discussed this: >> > This really shouldn't be required and I think the real issue is that >> > udl creates a dma_buf attachement (which is needed for device dma >> > only), but only really wants to do cpu access through vmap/kmap. So >> > not attached the device should be good enough. Cc'ing a few more lists >> > for better fyi ;-) >> >> Though I've looked at this a bit more, and since I want to be able to expose >> shared objects as proper GEM objects from the import side I really >> need that list of pages. > > Hm, what does "proper GEM object" mean in the context of udl? One that appears the same as a GEM object created by userspace. i.e. mmap works. Dave.
On Mon, May 6, 2013 at 9:56 PM, Dave Airlie <airlied@gmail.com> wrote: > On Tue, May 7, 2013 at 1:59 AM, Daniel Vetter <daniel@ffwll.ch> wrote: >> On Mon, May 06, 2013 at 10:35:35AM +1000, Dave Airlie wrote: >>> >> >>> >> However if we don't set a dma mask on the usb device, the mapping >>> >> ends up using swiotlb on machines that have it enabled, which >>> >> is less than desireable. >>> >> >>> >> Signed-off-by: Dave Airlie <airlied@redhat.com> >>> > >>> > Fyi for everyone else who was not on irc when Dave&I discussed this: >>> > This really shouldn't be required and I think the real issue is that >>> > udl creates a dma_buf attachement (which is needed for device dma >>> > only), but only really wants to do cpu access through vmap/kmap. So >>> > not attached the device should be good enough. Cc'ing a few more lists >>> > for better fyi ;-) >>> >>> Though I've looked at this a bit more, and since I want to be able to expose >>> shared objects as proper GEM objects from the import side I really >>> need that list of pages. >> >> Hm, what does "proper GEM object" mean in the context of udl? > > One that appears the same as a GEM object created by userspace. i.e. mmap works. Oh, we have an mmap interface in the dma_buf thing for that, and iirc Rob Clark even bothered to implement the gem->dma_buf mmap forwarding somewhere. And iirc android's ion-on-dma_buf stuff is even using the mmap interface stuff. Now for prime "let's just ship this, dammit" prevailed for now. But I still think that hiding the backing storage a bit better (with the eventual goal of supporting eviction with Maarten's fence/ww_mutex madness) feels like a worthy long-term goal. Cheers, Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
On Mon, May 6, 2013 at 4:44 PM, Daniel Vetter <daniel@ffwll.ch> wrote: > On Mon, May 6, 2013 at 9:56 PM, Dave Airlie <airlied@gmail.com> wrote: >> On Tue, May 7, 2013 at 1:59 AM, Daniel Vetter <daniel@ffwll.ch> wrote: >>> On Mon, May 06, 2013 at 10:35:35AM +1000, Dave Airlie wrote: >>>> >> >>>> >> However if we don't set a dma mask on the usb device, the mapping >>>> >> ends up using swiotlb on machines that have it enabled, which >>>> >> is less than desireable. >>>> >> >>>> >> Signed-off-by: Dave Airlie <airlied@redhat.com> >>>> > >>>> > Fyi for everyone else who was not on irc when Dave&I discussed this: >>>> > This really shouldn't be required and I think the real issue is that >>>> > udl creates a dma_buf attachement (which is needed for device dma >>>> > only), but only really wants to do cpu access through vmap/kmap. So >>>> > not attached the device should be good enough. Cc'ing a few more lists >>>> > for better fyi ;-) >>>> >>>> Though I've looked at this a bit more, and since I want to be able to expose >>>> shared objects as proper GEM objects from the import side I really >>>> need that list of pages. >>> >>> Hm, what does "proper GEM object" mean in the context of udl? >> >> One that appears the same as a GEM object created by userspace. i.e. mmap works. > > Oh, we have an mmap interface in the dma_buf thing for that, and iirc > Rob Clark even bothered to implement the gem->dma_buf mmap forwarding > somewhere. And iirc android's ion-on-dma_buf stuff is even using the > mmap interface stuff. fwiw, what I did was dma_buf -> gem mmap fwding, ie. implement mmap for the dmabuf object by fwd'ing it to my normal gem mmap code. Might be the opposite of what you are looking for here. Although I suppose the reverse could work to, I hadn't really thought about it yet. BR, -R > Now for prime "let's just ship this, dammit" prevailed for now. But I > still think that hiding the backing storage a bit better (with the > eventual goal of supporting eviction with Maarten's fence/ww_mutex > madness) feels like a worthy long-term goal. > > Cheers, Daniel > -- > Daniel Vetter > Software Engineer, Intel Corporation > +41 (0) 79 365 57 48 - http://blog.ffwll.ch > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/dri-devel
>>> One that appears the same as a GEM object created by userspace. i.e. mmap works. >> >> Oh, we have an mmap interface in the dma_buf thing for that, and iirc >> Rob Clark even bothered to implement the gem->dma_buf mmap forwarding >> somewhere. And iirc android's ion-on-dma_buf stuff is even using the >> mmap interface stuff. > > fwiw, what I did was dma_buf -> gem mmap fwding, ie. implement mmap > for the dmabuf object by fwd'ing it to my normal gem mmap code. Might > be the opposite of what you are looking for here. Although I suppose > the reverse could work to, I hadn't really thought about it yet. > Yeah thats the opposite, I want to implement my GEM mmap ioctls using dma-buf mmaps :) Dave.
diff --git a/drivers/gpu/drm/udl/udl_main.c b/drivers/gpu/drm/udl/udl_main.c index 0ce2d71..6770e1b 100644 --- a/drivers/gpu/drm/udl/udl_main.c +++ b/drivers/gpu/drm/udl/udl_main.c @@ -293,6 +293,7 @@ int udl_driver_load(struct drm_device *dev, unsigned long flags) udl->ddev = dev; dev->dev_private = udl; + dma_set_mask(dev->dev, DMA_BIT_MASK(64)); if (!udl_parse_vendor_descriptor(dev, dev->usbdev)) { DRM_ERROR("firmware not recognized. Assume incompatible device\n"); goto err;
Since we ask the dmabuf owner to map the dma-buf into our device address space, but for udl at present that is the CPU address space, since we don't DMA directly from the mapped buffer. However if we don't set a dma mask on the usb device, the mapping ends up using swiotlb on machines that have it enabled, which is less than desireable. Signed-off-by: Dave Airlie <airlied@redhat.com> --- drivers/gpu/drm/udl/udl_main.c | 1 + 1 file changed, 1 insertion(+)