[PATCH 1/5] drm/vmwgfx: unbind in vmw_ttm_unpopulate

Daniel Vetter daniel at ffwll.ch
Thu Aug 26 12:50:16 UTC 2021


On Thu, Aug 26, 2021 at 12:11:04PM +0200, Christian König wrote:
> 
> 
> Am 26.08.21 um 10:49 schrieb Daniel Vetter:
> > On Mon, Aug 23, 2021 at 01:15:20PM +0200, Thomas Hellström wrote:
> > > On Mon, 2021-08-23 at 13:05 +0200, Christian König wrote:
> > > > Adding Thomas on CC as well.
> > > > 
> > > > Just a gentle ping. I think the patch set makes sense now.
> > > > 
> > > > Regards,
> > > > Christian.
> > > > 
> > > > Am 28.07.21 um 15:05 schrieb Christian König:
> > > > > Doing this in vmw_ttm_destroy() is to late.
> > > > > 
> > > > > It turned out that this is not a good idea at all because it leaves
> > > > > pointers
> > > > > to freed up system memory pages in the GART tables of the drivers.
> > > > > 
> > > > > Signed-off-by: Christian König <christian.koenig at amd.com>
> > > > > ---
> > > > >    drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 9 +++------
> > > > >    1 file changed, 3 insertions(+), 6 deletions(-)
> > > > > 
> > > Reviewed-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>
> > For next time around I think recording a bit more of the discussions and
> > git history in these would be really good. At least I'd like to get more
> > people ramped up on ttm work, and for that to work out things need to be a
> > bit more accessible ... The above commit message is pretty much useless if
> > you ever hit it in a git blame, if you haven't been involved in any of
> > these discussions.
> 
> I've pushed it with a link tag back to the patches in patchwork, but I
> should probably include a link tag to the older versions as well for
> completeness.

Imo references to the commits that landed which broke stuff and full
quotes of the relevant discussions in the other thread. Just adding the
links still means you have to painstakingly stitch the story together
yourself, but at least it would be somewhat possible.

Really I think stuff like this should be documented in kerneldoc for these
callbacks, and we just got to start doing that. Waiting until magically
someone else fixes the ttm kerneldoc wont happen :-)
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch


More information about the dri-devel mailing list