[Intel-gfx] broken eDP device types

[Jon Pry]

Hi,

   I'm trying to work out some bugs on Asus T100TA which is Baytrail-T
platform. Current code in use is 3.15_rc4. In general I have been
having problems with the backlight control. For some reason the kernel
is detecting the panel as a VGA device and intel_crt.c does not load
intel_backlight, so I hacked something together real quick and ended
up getting something that actual changes PWM registers, but this had
no effect on actual screen brightness.

   Without any real solid theory as to why PWM is not doing anything.
I started wondering why exactly the kernel thinks the panel is VGA
since it is kind of unlikely the panel is analog especially
considering Baytrail-T does not have any analog interfaces. So I
nabbed the VBT to take a look.

Whole thing is here if your interested http://pastebin.com/crht1nDU

Pretty sure the relevant portion is:

        EFP device info:
                Device type: 0x1400 (unknown)
                Port: 0x15 (unknown)
                DDC pin: 0x04
                Dock port: 0x00 (N/A)
                HDMI compatible? No
                Info: HDMI certified
                Aux channel: 0x00
                Dongle detect: 0x00

The VBT does include tables for LVDS/eDP operation of the panel. It
seems just the device type is fubar. So my questions are, why is the
type screwed up? What would windows driver do upon seeing that, and
what is the best way to override it in the kernel?

The practical impact here I think will be during sleep. As taking down
the eDP link is unlikely to work so long as driver thinks it is a CRT.

Also I notice that the backlight block contains these values:

        I2C slave addr: 0x58
        I2C command: 0xaa

Which are not used in the linux driver. Is this something the windows
driver actually does? Any plans to implement i2c backlight control in
i915?

Regards,

Jon Pry
jonpry at gmail.com