Intel Workaround For Graphics Driver Regression: "The Platform Problem Going Crazy"

[phoronix]

Phoronix: Intel Workaround For Graphics Driver Regression: "The Platform Problem Going Crazy"

Sent out over the weekend was a patch series for the Intel Linux kernel graphics driver entitled "Time, where did it go?" This set of 42 patches aims to provide incremental improvements to the driver to offset a performance regression in Linux 5.7 that Intel hasn't been able to track down. This this increased complication of the driver to offset the regression is now under the microscope...

http://www.phoronix.com/scan.php?page=news_item&px=Intel-Graphics-WA-For-Regress

[gnarlin]

Call me crazy, but isn't this exactly what kernel bisection is for?

[ms178]

I am a bit surprised that this wasn't caught and fixed earlier as Intel puts much effort into CI testing their stack.

[Widefox]

Yup, with today's (early) state of 5.9, I've now got a kernel warning for i915 :

[ 3.892417] ------------[ cut here ]------------
[ 3.892421] i915 0000:00:02.0: drm_WARN_ON_ONCE(level >= n_entries)
[ 3.892595] WARNING: CPU: 0 PID: 113 at drivers/gpu/drm/i915/display/intel_ddi.c:1065 intel_ddi_pre_enable+0xd37/0xdb0 [i915]
[ 3.892596] Modules linked in: mmc_block i915 rtsx_pci_sdmmc mmc_core cec i2c_algo_bit drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops e1000e drm crc32c_intel serio_raw ptp rtsx_pci pps_core wmi$
[ 3.892619] CPU: 0 PID: 113 Comm: kworker/u16:3 Not tainted 5.9.0rc0.2user+ #174

Update - the bug is when connecting an external monitor. It's a regression sometime since 5.3.11

[cl333r]

This is a hard NAK for this sort of refactoring

What's NAK?

[Widefox]

What's NAK?

en.wikipedia.org/w/index.php?title=NAK "negative-acknowledgement (NAK or NACK) signal is sent to reject a previously received message or to indicate some kind of error"

[Mitch]

What's NAK?

I think a NAK is the opposite of approval for changes. I imagine he thinks a hurried and messy refactor was done for intel's GFX as a knee jerk reaction to a performance regression. In effect, he's saying no to any present or future refactoring of this type.

[programmerjake]

What's NAK?

negative/no acknowledge -- basically saying it isn't accepted.

[Linuxhippy]

Quite interesting - my desktop performance regressed quite a lot when upgrading from Fedora 31 -> 32.
However, with the update changing basically everything from kernel to mesa to userspace/xorg/compositor version, I was unable to track the culprit down.

However, having an old 1st gen Intel HD graphics (arrandale) chip, I guess nobody would care a lot finding the root cause of the issue.
Funny, other regressions also seem to pop up for newer platforms.

[intelfx]

I don’t understand. Each of these patches is an improvement in itself, regardless of the regression, right? Each of these patches has a commit message that makes sense. And what happens is the DRM maintainer basically says “no, the kernel will not get these performance improvements, never ever”.

WTF?