FS#50694 - [linux] Kernel 4.7 causes boot time increase due to ACPI EC
Attached to Project:
Arch Linux
Opened by John Henderson (jwhendy) - Wednesday, 07 September 2016, 14:07 GMT
Last edited by Jan de Groot (JGC) - Wednesday, 04 October 2017, 03:28 GMT
Opened by John Henderson (jwhendy) - Wednesday, 07 September 2016, 14:07 GMT
Last edited by Jan de Groot (JGC) - Wednesday, 04 October 2017, 03:28 GMT
|
Details
Description:
After upgrading to kernel 4.7, my boot time went from ~13sec to 24sec. I posted around and didn't get much response, so I created a kernel bug report: - https://bugzilla.kernel.org/show_bug.cgi?id=153511 Here was Lv Zheng's summary of the issue: ### This is in fact not an issue. Just because _REG (which consumes time on this platform) is executed more times than the previous version. However it really can be tuned not to execute so many times by checking if 2 EC devices are equivelent. ### Additional info: * package: core/linux (at 4.7.2-1 as of 2016-09-06) LTS vs. 4.7.0 (which is what I was on at the time): ## kernel 4.4.16 $ systemd-analyze Startup finished in 4.683s (kernel) + 7.625s (initrd) + 1.084s (userspace) = 13.393s ### kernel 4.7.0 $ systemd-analyze Startup finished in 10.746s (kernel) + 10.518s (initrd) + 4.060s (userspace) = 25.325s Bug report: https://bugzilla.kernel.org/show_bug.cgi?id=153511 Steps to reproduce: - boot system as usual on 4.7 vs. 4.6 or LTS. The bug report resulted in a patch that has me back to just over 14sec, so I considered the issue solved. They say the fix won't be implemented until 4.9, so I thought I'd pass along the info here in the event we'd want to include this in the Arch patchset. I'm not sure how those decisions get made, but it'd be awesome not to have to compile my own with every update! This is the patch that worked on 4.7.2 compiled via ABS: https://bugzilla.kernel.org/attachment.cgi?id=232071 |
This task depends upon