UME S3 sleep is not stable

[Yu Ping]

Description:
while in UME beta 4 testing, we found the S3 sleep is not stable. After entering S3 state, the wake-up is not stable. Sometimes, the screen turns to white, but could not recover to the orignal state when sleeping.

Reproduce Steps:
(1)use "pm-suspend --quirk-vbestate-restore" to make the system go to S3 state
(2)wait for the system go to S3 state
(3)press power button to wake up the system

Expected result:
system could be easy to go to S3 state and easy to recover from S3 state.

[CheonwoeiNg (cheon-woei-ng)]

It look like this issue is specific to Ubuntu-UME.

Test result from 7/23/08:

Moblin:
D1 B7 driver on Poulsbo D1 – Pass (unable to reproduce the failure, LVDS display is fine after resume from S3)

UME:
D1 B7 driver on Poulsbo D1 –
Fail (LVDS showing blank white screen after resume from S3)

[CheonwoeiNg (cheon-woei-ng)]

Test result from 8/7/2008

Tested on Ubuntu PV
Problem reproduced:-
-Without LVDS the display OK.
-If LVDS display is connected, the display will show a blank white screen

Tested on Moblin Beta 4
-Without LVDS the display OK.
-If LVDS display is connected, the display will show a distorted white screen and then able to enter the normal mode (desktop was shown).

This issue is still specific to UME.

[Loïc Minier (lool)]

I checked with the USG team working on the ODM device and they moved away of pm-suspend because:
a) it wasn't working properly
b) it was very slow

I'll attach the replacement sleep script they use.

[Loïc Minier (lool)]

For Ubuntu MID, we should check whether the bug still happens in intrepid and if it does add a quirks file to fix whatever pm-suspend thinks it needs to do/not to do: perhaps a VBIOS repost or the lack of it could help for instance.

[CheonwoeiNg (cheon-woei-ng)]

Just tested the S3 sleep script on non-ODM specific Hardy based Ubuntu-UME PV image. With both LCD and LVDS connected, the display would not resume.

[Loïc Minier (lool)]

"pm-suspend" and "pm-suspend --quirk-vbestate-restore" both allow to suspend and resume fine; the display comes back.

This is likely a result of fixing bug #264485.

[CheonwoeiNg (cheon-woei-ng)]

Still see issue.

[CheonwoeiNg (cheon-woei-ng)]

Loic,

Have you perform test with both LCD and LVDS?

The original issue is this:

-Without LVDS the display OK.
-If LVDS display is connected, the display will show a blank white screen

[Loïc Minier (lool)]

I have both a LCD and the LVDS connected, but didn't look at the LVDS upon resume.

[CheonwoeiNg (cheon-woei-ng)]

Observed Intrepid Desktop (without Intel driver) S3 resumed issue on Menlow D2. When pressing power button to resume, monitor is not awake.

*Note: My setup is D2 with USB keyboard and mouse. Tried CRT monitor and LCD monitor.

FYI, found a temporary workaround that would make S3 resume works on my setup. Note that workaround does not work on D1.

Steps for workaround:
To suspend, use the menu: System-> Shut Down->Suspend
To resume, press the power button. Wait for about 6 seconds and press Ctrl-Alt-F7. Wait ~20 seconds. Press Ctrl-Alt-F8 and wait for terminal text to appear. After that, press Ctrl-Alt-F7 and I will get the log in screen to log back in.

[CheonwoeiNg (cheon-woei-ng)]

Per Canonical business team, S3 resume issue in Hardy is a known issue in Xserver. The symptom is that Hyper Threading (HT) and Virture Console ( Ctrl + Alt + F# ) created a race condition from S3 resume. The workaround is to delay Virtual Console from S3 resume.

I wonder if this is the same issue for Intrepid. Please consider this workaround if S3 issues still exist in Jaunty.

[Loïc Minier (lool)]

I've pushed the pm-utils change to disable HT on suspend and restore on resume to the hardy ubuntu-mobile ppa.

This is a workaround, not an actual fix.

It's pending QA and a subsequent push to production.

If you experience the /same/ bug under intrepid, please check with a different video driver (e.g. vesa), and note here your linux kernel version and hardware (jax10 or CB I guess).

I fear it's too late to push the workaround to disable ht in pm-utils in intrepid; it would be a risk for all users of intrepid and it's actually a workaround, not a fix; it only affects uncommon hardware -- unless it's affecting all ht hardware, but I didn't see widespread reports of this.

[CheonwoeiNg (cheon-woei-ng)]

FYI, the Intrepid S3 bug I mentioned earlier is confirmed under VESA driver, not Intel's driver.

The kernel version is 2.6.27-7-generic.

The S3 bug under VESA was confirmed on three different setups:
1. CB D1 board with USB Keyboard and USB Mouse with LCD monitor.
2. CB D2 board with USB Keyboard and USB Mouse with LCD monitor.
3. CB D2 board with USB Kayboard and USB Mouse with CRT monitor.

Also, although I do not see widespread reporting on S3, doing a quick search I found some bugs including #298927(new) and #34155 (old one and confirmed on Intrepid too).

[CheonwoeiNg (cheon-woei-ng)]

Using the new (untested per Loic) pm-util from PPA, the S3 issue still exists on Intrepid Desktop (2.6.27-7-generic kernel with VESA driver) on CB D2 board with USB Kayboard and USB Mouse with CRT monitor.

Tested two different ways (using the menu and also pm-suspend in terminal) to put system to S3. Both ways failed to resume after pressing power button.

However, using workaround mentioned earlier, I was able to get system to resume.