On Wed, May 27, 2020 at 4:05 PM Ilia Mirkin <imirkin at alum.mit.edu> wrote:> > Isn't this already fixed by > > https://cgit.freedesktop.org/drm/drm/commit/?id=7dbbdd37f2ae7dd4175ba3f86f4335c463b18403Ok, I see that fixes the link error, but I when I created my fix, that did not seem like the correct solution because it reverts part of the original patch without reverting the rest of it. Unfortunately there was no changelog text in the first patch to explain why this is safe. Could you double-check if the behavior is still correct after the two patches? Arnd
On Thu, 28 May 2020 at 00:36, Arnd Bergmann <arnd at arndb.de> wrote:> > On Wed, May 27, 2020 at 4:05 PM Ilia Mirkin <imirkin at alum.mit.edu> wrote: > > > > Isn't this already fixed by > > > > https://cgit.freedesktop.org/drm/drm/commit/?id=7dbbdd37f2ae7dd4175ba3f86f4335c463b18403 > > Ok, I see that fixes the link error, but I when I created my fix, that did > not seem like the correct solution because it reverts part of the original > patch without reverting the rest of it. Unfortunately there was no > changelog text in the first patch to explain why this is safe.No it doesn't, I think you missed the pci in API name. The initial behaviour doesn't use the pci version of the API, the replacement did, and the fix used the drm wrapper around the pci one. So this patch isn't necessary now that I've fixed it the other way, Thanks, Dave.
On Thu, May 28, 2020 at 7:37 AM Dave Airlie <airlied at gmail.com> wrote:> > On Thu, 28 May 2020 at 00:36, Arnd Bergmann <arnd at arndb.de> wrote: > > > > On Wed, May 27, 2020 at 4:05 PM Ilia Mirkin <imirkin at alum.mit.edu> wrote: > > > > > > Isn't this already fixed by > > > > > > https://cgit.freedesktop.org/drm/drm/commit/?id=7dbbdd37f2ae7dd4175ba3f86f4335c463b18403 > > > > Ok, I see that fixes the link error, but I when I created my fix, that did > > not seem like the correct solution because it reverts part of the original > > patch without reverting the rest of it. Unfortunately there was no > > changelog text in the first patch to explain why this is safe. > > No it doesn't, I think you missed the pci in API name. > > The initial behaviour doesn't use the pci version of the API, the > replacement did, and the fix used the drm wrapper around the pci one. > > So this patch isn't necessary now that I've fixed it the other way,Ok, got it, thanks for explaining! Arnd