On 2018-10-08 12:17 PM, Deucher, Alexander wrote:
-----Original Message----- From: Sasha Levin sashal@kernel.org Sent: Monday, October 8, 2018 10:49 AM To: Deucher, Alexander Alexander.Deucher@amd.com Cc: Greg Kroah-Hartman gregkh@linuxfoundation.org; linux- kernel@vger.kernel.org; stable@vger.kernel.org; Wentland, Harry Harry.Wentland@amd.com; Zhu, Rex Rex.Zhu@amd.com; Sasha Levin alexander.levin@microsoft.com Subject: Re: [PATCH 4.18 222/235] drm/amd/pp: Send khz clock values to DC for smu7/8
On Mon, Oct 08, 2018 at 02:33:56PM +0000, Deucher, Alexander wrote:
-----Original Message----- From: Greg Kroah-Hartman gregkh@linuxfoundation.org Sent: Monday, September 24, 2018 7:53 AM To: linux-kernel@vger.kernel.org Cc: Greg Kroah-Hartman gregkh@linuxfoundation.org; stable@vger.kernel.org; Wentland, Harry Harry.Wentland@amd.com; Deucher, Alexander Alexander.Deucher@amd.com; Zhu, Rex Rex.Zhu@amd.com; Sasha Levin alexander.levin@microsoft.com Subject: [PATCH 4.18 222/235] drm/amd/pp: Send khz clock values to DC for smu7/8
4.18-stable review patch. If anyone has any objections, please let me
know.
This regresses power usage on 4.18. Please revert. https://bugzilla.kernel.org/show_bug.cgi?id=201275
Hi Alex,
Thank you for the report.
I'm working on improving this process, I'd be very grateful if you could answer a few questions about this:
- Is the same breakage seen upstream? (if so, it should be reverted there as
well and we can grab the revert into -stable).
No regression in 4.19 or -next.
- Does the issue reported by this patch ("pipes seem to hang with a 4k DP
and 1080p HDMI display") exist in the 4.18 stable tree?
I don't think so, but I'm not 100% sure. Harry, Rex do you know if this is a general issue or was it just fall out from the changes to the interface?
It was intended to deal with fallout to smu7/8 caused by the change to move the kHz-to-10kHz-conversion from display to powerplay. I believe that conversion initially overlooked these blocks. Rex, please correct me if I'm wrong. You probably know more about this than I do.
So yes, I agree, it was never intended to be used without the change to drop the *10 in display code, which was done with this change:
commit 23ec3d1479fd79658cd52c47618d8ddd2f32550b Author: Rex Zhu Rex.Zhu@amd.com Date: Mon Jun 18 18:15:15 2018 +0800
drm/amd/pp: Convert clock unit to KHz as defined
Convert clock unit 10KHz to KHz as the data sturct defined. e.g. struct pp_clock_with_latency { uint32_t clocks_in_khz; uint32_t latency_in_us; }; Meanwhile revert the same conversion in display side.
Acked-by: Alex Deucher alexander.deucher@amd.com Acked-by: Harry Wentland harry.wentland@amd.com Signed-off-by: Rex Zhu Rex.Zhu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com
I don't see this change in the 4.18.y stable tree. I probably should've caught that with the initial 4.18-stable review patch.
Harry
- If not, could you briefly explain why?
We refactored the interface between the power and display components and this patch fixed up some of that fallout due to the differences in units used in each component.
The algorithm I use was very confident about this patch being stable material, and when I looked at it back then (and again now) I was very confident of the same. If I can understand where I was wrong I could improve my process.
There are some other dependent patches required that were not flagged in the patch itself. IIRC, they were a bit big for stable.
Alex