At Fri, 12 Jun 2015 16:33:32 +0200, David Henningsson wrote:
On 2015-06-12 16:14, Takashi Iwai wrote:
At Fri, 12 Jun 2015 08:37:38 +0200, Takashi Iwai wrote:
At Fri, 12 Jun 2015 08:17:12 +0200, David Henningsson wrote:
If BIOS disables the power well, then we would get the "audio register I/O error" in the boot phase, there will be no codecs detected, and thus no broken sound card showing up.
The only issue is if the i915 driver disables the power well, but the case that the i915 should be functioning enough to turn off the power well at some point, and at the same time so broken that we can't bind to it, seems extremely unlikely to me.
Either the i915 driver works, or it does not. In both cases, continuing probing is the more useful option.
I don't see much usefulness there but rather simplicity.
The binding with i915 won't happen at the later stage, thus the device is just a placeholder. Any other goodies by keeping the dead device?
Fair enough. Then I'm going to queue the patch below in addition to for-linus branch.
We'll have to agree to disagree on that, then.
One point that hasn't been mentioned is the behavior change from the previous kernels. The previous kernels abort probing, and it was OK since we supposed only HSW/BDW before 4.1. Now, with continuing the probe, user sees the new device that hasn't been there, and yet the device is dead and merely a placeholder.
Is this convincing enough for you? Or would you still like to make the broken HDMI/DP device enabled for HSW/BDW?
I guess the question is whether HDMI/DP can actually work without the i915 driver, e g when running with nomodeset. If there is absolutely no possibility whatsoever (e g maybe BIOS could set up the device so that it works?), then it does not matter much, I suppose.
It can't work because we don't get proper ELD thus cannot setup HDMI/DP CA bits.
Takashi