Bug 109245 - Commit 516a49cc19467e298d08a404f73a6e311f4548d1 causes increase in boot time on Sandy Bridge
Summary: Commit 516a49cc19467e298d08a404f73a6e311f4548d1 causes increase in boot time ...
Status: NEEDINFO
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-01-07 17:53 UTC by Daniel Kamil Kozar
Modified: 2019-01-08 21:56 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments
dmesg log with drm.debug=0xe (33.87 KB, application/x-bzip)
2019-01-07 17:53 UTC, Daniel Kamil Kozar
no flags Details
Output of intel_reg dump without i915 loaded (16.53 KB, text/plain)
2019-01-08 17:45 UTC, Daniel Kamil Kozar
no flags Details
[PATCH] drm/i915: Try to sanitize bogus DPLL state left over by broken SNB BIOSen (4.94 KB, patch)
2019-01-08 20:56 UTC, Ville Syrjala
no flags Details | Splinter Review

Note You need to log in before you can comment on or make changes to this bug.
Description Daniel Kamil Kozar 2019-01-07 17:53:36 UTC
Created attachment 143003 [details]
dmesg log with drm.debug=0xe

As outlined in https://lkml.org/lkml/2019/1/4/611 , I experienced a significant increase in boot time after moving from kernel 4.19.12 to 4.20. The hardware is a Asus K53SV with a i7-2630QM (Sandy Bridge) with an external display connected via HDMI.

Bisecting the commits revealed 516a49cc19467e298d08a404f73a6e311f4548d1 as the cause.

The requested dmesg log with drm.debug=0xe is attached to this ticket.
Comment 1 Ville Syrjala 2019-01-08 13:26:14 UTC
The hardware seems to be in a rather weird state when the driver loads. The PLL is on but apparently misconfigured, DP-C is on but not attached to the right pipe, etc.

Please reboot without loading i915, then run 'intel_reg dump > regdump' (intel_reg comes from igt-gpu-tools), and attach the results here.
Comment 2 Daniel Kamil Kozar 2019-01-08 17:45:39 UTC
Created attachment 143017 [details]
Output of intel_reg dump without i915 loaded

Here's the register dump that was taken without i915 loaded, with the monitor attached (but no signal being output), the laptop display turned off, and kernel running in runlevel 1.

If it helps, the monitor in question is a Dell U2515H whose native resolution is 2560x1440, which is probably reported in the EDID. The maximum resolution supported by this chip is supposedly 1080p, though I have no problems with running my desktop in 1440p@44 with a custom modeline specified in Xorg.conf.

The warning messages visible in the kernel output (attached earlier) also appeared on the old kernel version, and didn't seem to cause any trouble besides being there.
Comment 3 Ville Syrjala 2019-01-08 20:56:39 UTC
Created attachment 143020 [details] [review]
[PATCH] drm/i915: Try to sanitize bogus DPLL state left over by  broken SNB BIOSen

I think this ought to work around the problem. It's not ideal though since we still get a bunch of WARNs.

As mentioned in the patch I think this same problem was seen on some other SNB machine years ago, but it got fixed by a BIOS update. Sadly looks like your machine doesn't have that option (unless I missed something on the ASUS website).

Please test and let me know whether this helps.
Comment 4 Daniel Kamil Kozar 2019-01-08 21:56:03 UTC
Works great for me! The boot time, and the messages, are back to what they were in 4.19.12.
I applied the patch on top of 5.0-rc1 and only had to change IS_GEN(dev_priv, 6) to IS_GEN6(dev_priv) which I hope does the same thing.
Unfortunately, you're right about the BIOS : I'm running the latest revision, which has not been updated since 2011.

Thanks for getting this sorted out so quickly.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.