4.7.0-rc2

Robert John Small robert at robertsmall.org
Mon Jun 13 11:55:18 CEST 2016


Hullo there everyone,

In case nobody guessed I am the Robert that Paul keeps talking about. 
Hopefully he hasn't oversold me by too much.

I have been testing the 4.6.0 kernel and the 4.7.0-rc2-pm1 kernel. 
Hopefully some of my findings are interesting.

Configuration:
- EUFI 5.2.0
- 32bit Debian GNU/Linux
- All of the OS on the eMMC

Factors affecting stability:
- On both kernels loading the correct sound module seems to be related 
to (unexplained) lockups.
- On both kernels the wireless driver for the internal wireless seems 
to cause locks within the networking subsystem of the kernel
	- Tested with `ip a` which is pretty direct to the kernel, fails.
	- Tested trying to remove the module itself from the kernel, this 
fails as well.
	- Tested trying to restart NetworkManager, forcibly get a new DHCP 
address all locked up commands. Interestingly ping even got stuck once.
	- Once the WiFI driver has locked up even rebooting sometimes seems to 
need power button force.
	- Disabling the WiFi with NetworkManager (Gnome, just turned it off, 
not sure what that does on a hardware level, presume it is an rfkill) 
completely prevented lock ups.
	- If I disable the WiFi each time before putting the device away I can 
use it without any trouble at all.
- The 4.7.0-rc2-pm1 kernel seems slightly more stable but I'm not 100% 
sure.

Testing with CStates enabled (4.7.0-rc2-pm1):
- Idle drain of 2.2-2.7W vs 2.7-3.4W
- Active drain of 5.1.5.9W vs 5.6-6.9W
- Feels a little cooler when idling, YMMV.
- Ran two tests of just randomly opening, closing, leaving it for a bit 
and then coming back.
	- Test one lasted 6.5h until I needed to reboot for something else. No 
lock up at all.
	- Test two lasted 10h and locked up when I detached from charging. I 
also undocked and redocked it at one point to see if the keyboard still 
worked (it did). Not sure if the undock/dock may have been related to 
the lock up.
- Deliberately used the above method of turning WiFi off when not using 
the device.
- I may test this again further with the CStates enabled. 6.5h and 10h 
are both well over the margin where you'd be likely to conserve power 
by turning it off anyway.

If I were to make a wild accusation I would say that something common 
to sound and WiFi causes issues. ACPI? Certainly looks like it.

OK, that's everything I remember. Please excuse any failure to conform 
to mailing list etiquette I haven't used one in some years.

--Robert.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.tinc-vpn.org/pipermail/click/attachments/20160613/e7284fd8/attachment.html>


More information about the click mailing list