CFT update day 2

[René Ladan]

Op 29 mei 2016 9:16 p.m. schreef "René Ladan" <rene at freebsd.org>:
>
> On 29-05-16 18:37, Matthias Haas wrote:
> > Am 29.05.2016 um 16:51 schrieb René Ladan:
> >> On 23-05-16 10:12, Matthew Macy wrote:
> >>> The highlights for today are the following:
> >>>
> >>> Bug fixes:
> >>> - Will Andrews fixed attach for some laptops (such as the Carbon X1).
> >>>    The Carbon X1 has a quirky BIOS that doesn't allow the OS to
> >>>    enumerate the GPU's interrupt.
> >>> - Will Andrews identified a conditionally uninitialized return in
> >>>    idr_find that could lead to a panic in some cases.
> >>> - Fixed a panic in mtrr_del frequently seen when attach failed.
> >>> - Sleep/wakeups with interrupts are largely implemented correctly
> >>>    now. Previously a polling 10ms sleep was used. I'm still
> >>>    concerned that the code really needs to be level-triggered.
> >>>
> >>> Cleanups:
> >>> - Logging is now enabled for the first 10s after attach unless
> >>>    dev.drm.drm_debug_keep=1.
> >>> - Unimplemented warnings are off by default.
> >>>
> >> [...snip USB instructions...]
> >>> If using the github repo, make sure you're using the drm-next-4.6
> >>> branch.
> >>>
> >> I tested the latest github version on my laptop (an Acer Aspire
> >> E5-773G-78RN with an Intel HD 520 GPU, see [1]), some results:
> >>
> >> - xfce4 starts, no visual artifacts
> >> - XV is disabled but present according to xdpyinfo, i.e. mplayer
renders
> >> movies with black borders in full screen mode
> >> - glxgears gets up to 30 fps full screen (so something is not
> >> accelerated)
> >> - HDMI output works (when X is started after initially plugging in the
> >> cable), the TV image keeps getting updated if I close the lid
> >> - switching back and forth between X and the console works
> >> - stellarium works
> >>
> >> Maybe xf86-video-intel 2.21.15 is missing an PCI id?
> > It is indeed missing a few PCI ids, I created 2 patches that add those
> > missing ids, but that doesn't seem to be enough in my case (Iris 550).
> > You may try them anyway and see if they change anything for you, but I
> > can't give any support as I'm only a web developer and all this stuff is
> > not really my area of expertise.
>
> With a patched xf86-video-intel the screen freezes and switching back to
> the console does not work either. SSH login still works fine. Although
> Xorg looks frozen, Xorg.log shows that acceleration should work now, as
> shown in the attached Xorg.log diff (with timestamps removed). A kernel
> log from around the freeze is attached too.
>
> This is with the drm-next-4.6 branch at commit
> 1e9ceda8a2a5b5eb45b3cd692987edc8b410817f
>
A few more messages from before the"bad malloc flags":

<6>[drm] GMBUS [i915 gmbus dpb] timed out, falling back to bit banging on
pin 5
[drm:0xffffffff82f3687ds] *ERROR* Potential atomic update failure on pipe B
kernel: WARN_ON(!wm_changed)

> >> [1] https://wiki.freebsd.org/Laptops/Acer_Aspire_E5_773G_78RN
>
> Cheers,
> René
>