Re: [3.5 regression] DRM: Massive (EDID-probing?) X startup delay on ATI Radeon RV770 (HD4870)

[Nix <nix@esperi.org.uk>]

On 6 Aug 2012, Alex Deucher verbalised:

> On Sat, Aug 4, 2012 at 12:13 PM, Nix <nix@esperi.org.uk> wrote:
>> Possibly-relevant info:
>>
>>  - Two DVI monitors, identical specs, one dual-head graphics card
>>    (so no VGA switcheroo or awesome-yet-terrifying PRIME madness needed)
>>
>>  - KMS, Xserver 1.12.3, driver 6.14.6-28 (trunk current as of today),
>>    Mesa 8.0.4, libdrm 2.4.37
>>
>> As of kernel 3.5 EDID probing of the older of my two monitors appears to
>> have subtly broken. The log shows that it appears to work -- KMS comes
>> up OK and I get a working console -- but then X stops during startup for
>> nearly a minute (with both monitors black) before coming back to life
>> again and EDID-probing the monitor a further six times for no obvious
>> reason. (Full log attached, and xorg.conf, for what little use it is.)

False alarm. Well, the massive number of EDID probes is consistent, but
that was present in earlier kernels too. The minute-long X start process
was what was unusual, and intermittent: even in 3.5, it starts in a
relatively 'normal' eight seconds normally. I suspect system load, or
i2c bus timeouts or something (it is a really crap bus, after all).
Anyway, it hasn't recurred, which makes it hard to track down further,
let alone bisect :(

I think we have to let this one rest. Dammit. I hate to let a bug slip
away.

-- 
NULL && (void)