[Bug 107583] Linux: REGRESSION: drm: cmdline EDID override mechanisms broken since 4.15

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=107583

--- Comment #18 from Jani Nikula <jani.nikula at intel.com> ---
The problem here and in the AST case [1] is that drm_helper_probe_detect()
correctly reports connected, but the DDC probe in drm_get_edid() fails. Here
the connected status comes from hotplug detect, in AST it's always connected.

Using the DDC probe even with override/firmware EDID is useful in cases where
DDC works, but the EDID data is corrupt.

The idea was that one would use video=VGA-1:e to force the connector on when
DDC does not work at all. In the AST case where the connector status is always
connected, there is no functional regression, but it does require adding
another parameter. The same with all cases where detection is based solely on
DDC probe in general, and you have to use connector force anyway.

In this bug, the hotplug detect works, and the connected status from that takes
precedence over DDC. Which means that in this bug, the requirement to use
video=VGA-1:e regresses the hotplug functionality.

The override/firmware EDID mechanism was moved to the low level to
transparently fix a plethora of bugs where, confusingly, both the real EDID and
the override/firmware EDID were used. For example, it was impossible to
override audio details.

I think there's a conflicting set of valid requirements here, and the only way
to preserve functionality for all cases is to add another level of connector
force probe, which skips DDC probe in case of override/firmware EDID.

[1]
http://mid.mail-archive.com/alpine.DEB.2.20.1905262211270.24390@whs-18.cs.helsinki.fi

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20190528/c8af3f47/attachment.html>