suspend/resume test

[Jeremy Chang]

On Fri, Dec 05, 2008 at 01:34:43PM +0000, Andy Green wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Somebody in the thread at some point said:
> | Hi, lists:
> |     Recently, I am trying to test suspend/resume auto recursively with a
> | script that I wrote as here,
> | http://svn.openmoko.org/developers/jeremy/winker/madwinker.
> |
> | This script uses wkalrm that's by Werner to wakeup the device. you can
> | set a parameter to decide how many times you want to test. It's
> | basically a while loop that with "wkalrm +5s && echo mem >
> /sys/power/state" inside.
> |
> |     For now, I run it with the GTA02 daily kernel stable image, the
> | result is very exciting, The device I tested is with sim card inside,
> | and I test it with bluetooth and gps on, I tested it with 1000/2000/4000
> | times test by running the script. The result is that all the tests were
> | completed, and the device is still alive, can make a phone call.
> |
> |     So, It looks suspend/resume is very stable in stable branch.
>
> Well it's pleasing to hear it but I think we find it's only true build
> by build for the 2.6.24 / stable branch.  We know from experience if we
> add some printks or change driver timing trouble will come.  These
> problems are meant to be properly solved in andy-tracking.
>
> Also, we know that different trouble can come according to the wake
> source, so we had a lot of problems that only came if the wake source
> was GSM unit for example.
>
> |     For andy-tracking, I just tried using it to test, but now I met a
> | problem from the beginning, the device can't be waked up by the rtc. I
> | don't know what's the cause yet.  Any clue for this?
>
> It matters when you last pulled andy-tracking... before this patch
>
> http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=5430bdb92ec379dd0cd4d52d80180efc0b79e92b
>
> last night it was completely broken for resume for a day or so, but it
> should be fine now.  About rtc wakeup specifically I never tried it...
> if it's not the problem can you check dmesg looking for "wake enabled
> for irq 53"?
>
> |     Hi, Andy, do you have any plan about how I can do this kind of
> test for
> | andy-tracking image? For gta02, when will be likely andy-tracking the next
> | stable/testing image? I am a newbie here.
>
> Lately git is updating several times a day :-)  If you are using binary
> images from http://people.openmoko.org/andy I am updating it right now
> and will keep doing so every day or so for the next while.
>
> If RTC remains busted even with newer build let us know and we can look
> more into it, however I know Balaji has tested it on andy-tracking
> recently too so I am hoping this was just a broken tree when you tried it.

    Hi, Andy, I just tried the latest, rtc wakeup still has no response
here. Though I did see the "wake enabled for irq 53" in dmesg message.
even pressing power button, no response either.

    But if I make the device suspend by "apm -s" or pressing power
button, Then pressing power key has response but still can't resume
thoroughly either, instead, it comes a blank gray screen, with sometimes 
LED keeping flashing,sometimes not, and it can respond the ping but ssh 
can't work.

    I will go to study pmu code, and datasheet this weekend as a start
to know more about suspend/resume.

>
> | Later I and Matt have a plan to improve this script, like testing the
> | device in various and random different situation, recording the current
> | value to check if the suspend really works. Even it could be used on
> | manufacture line in factory.
>
> For sure, but if you go down that path make sure to talk to Christopher
> about his new scipt / test system.
        Roger. Thanks !
BR.
Jeremy
>
> In fact suspend / resume has been very scary thing, after a lot of
> effort it is beaten into some kind of shape now and these automated
> tests can provide very nice regression detection, so it is good to hear
> you work on it.
>
> - -Andy
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.9 (GNU/Linux)
> Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org
>
> iEYEARECAAYFAkk5LfMACgkQOjLpvpq7dMrNswCeL0ixngntCXWsRS/NGaGMd7ZP
> 1uEAn0oCa8QeF+NCw54iC89hUFl463ea
> =tLXL
> -----END PGP SIGNATURE-----
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
Url : http://lists.openmoko.org/pipermail/openmoko-kernel/attachments/20081206/4c333a01/attachment.pgp