Re: Accelerometers jerky with new stable kernels?
Andy Green wrote: > Somebody in the thread at some point said: > | Using this reverting list [1] to get a woking stable kernel (image [2]). > | With itthe accelerometers work well also after suspend/resume (that > | works here, since it seems needed to the > | fix-glamo-mci-slow-clock-until-first-bulk.patch as reported in the > | kernel ML). > | > | I'm using my moko with OpenDoom in these days and all is completely > usable! > > It's great you found some suspend stability, but on 2.6.24 / stable > there is zero doubt it is purely random chance build by build. I would > guess that's why you can 'tune' stability by just removing patches that > affect the resume race by timing, even though the patches themselves > have no direct impact on suspend otherwise. I found I could make or > break a 2.6.24 kernel for suspend merely by adding printk()s around. Ah, ok... I knew that many of these patches were not causing direct troubles, but since I found that starting from a particular commit I was building non working (from the suspend/resume point of view) kernels, I decided to take this commit as my base and then adding only the patches (i.e. removing the others) that were more important for the standard usage. > Real solution is coming closer in stable-tracking for these problems > hopefully, although right now accel comms is completely broken in there. > ~ But it should be straight in next couple of weeks kind of timeframe. Nice to read... I'll give it a try as soon as I can, and I think it will be my default kernel as soon as the fixes you mentioned will be there! :) -- Treviño's World - Life and Linux http://www.3v1n0.net/ ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Somebody in the thread at some point said: | Timo Jyrinki wrote: |> 2008/11/6 Michael Zanetti <[EMAIL PROTECTED]>: |>> What do you mean with "working fine now"? Are they working as they were always |>> or are they now really working (no crashes on suspend or after some minutes of |>> use)? |> "Working fine" as in accelerometer using applications (I have |> AccelGame and Doom) work, unlike with the patch (jerkiness). I don't |> use suspend with my own kernels since resume is always broken - |> probably because I only flash the self-compiled kernel and don't know |> how I could easily also copy the compiled modules into use on Neo. | | Using this reverting list [1] to get a woking stable kernel (image [2]). | With itthe accelerometers work well also after suspend/resume (that | works here, since it seems needed to the | fix-glamo-mci-slow-clock-until-first-bulk.patch as reported in the | kernel ML). | | I'm using my moko with OpenDoom in these days and all is completely usable! It's great you found some suspend stability, but on 2.6.24 / stable there is zero doubt it is purely random chance build by build. I would guess that's why you can 'tune' stability by just removing patches that affect the resume race by timing, even though the patches themselves have no direct impact on suspend otherwise. I found I could make or break a 2.6.24 kernel for suspend merely by adding printk()s around. Real solution is coming closer in stable-tracking for these problems hopefully, although right now accel comms is completely broken in there. ~ But it should be straight in next couple of weeks kind of timeframe. - -Andy -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (GNU/Linux) Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org iEYEARECAAYFAkkUAsIACgkQOjLpvpq7dMp1dwCdE5T8mmGHzDCg5qFD3bTBlZ4K +0IAn2Y4lP/CSx26OduO+KY6RJ9aifa7 =7RO2 -END PGP SIGNATURE- ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
Timo Jyrinki wrote: > 2008/11/6 Michael Zanetti <[EMAIL PROTECTED]>: >> What do you mean with "working fine now"? Are they working as they were >> always >> or are they now really working (no crashes on suspend or after some minutes >> of >> use)? > > "Working fine" as in accelerometer using applications (I have > AccelGame and Doom) work, unlike with the patch (jerkiness). I don't > use suspend with my own kernels since resume is always broken - > probably because I only flash the self-compiled kernel and don't know > how I could easily also copy the compiled modules into use on Neo. Using this reverting list [1] to get a woking stable kernel (image [2]). With itthe accelerometers work well also after suspend/resume (that works here, since it seems needed to the fix-glamo-mci-slow-clock-until-first-bulk.patch as reported in the kernel ML). I'm using my moko with OpenDoom in these days and all is completely usable! [1] git-revert list to stable. Not completely studied: git-revert --no-edit 933f18caaf8f96d1094f064312d416a2fe0b4337 git-revert --no-edit ca57dbb5effd245e2619f4b730f8247f7279f954 git-revert --no-edit d3c59382c7e8429e2c86eb92cc2518dce26c4a69 git-revert --no-edit 5e2b5ddaf95985744e0e0c28ba6dac83c093fd15 git-revert --no-edit 3318a7cb3165a33046e122294e1878e3bd6555d8 git-revert --no-edit 35d910f40ce2c7c959edfbaa671f88e8da5fa6b1 git-revert --no-edit 1c6f5a92c816db43128444606c26507ebe6e5a43 git-revert --no-edit 73a16ab06c18ae6da52178c8997f097363ca0a17 git-revert --no-edit f43814abe0f52033d5915881016408bd1da29d2b git-revert --no-edit f5b973489beb1a1239dfad53e3ad6e36ff7ee958 git-revert --no-edit 2a6788eaaf0185f8db56353777e8f3f5dab1d665 git-revert --no-edit bc2caff9cdef8a16297a2cb196306f88fa5ea2c2 git-revert --no-edit 968c41d0c32099d78927849a71e2ef3143cc05e7 git-revert --no-edit 9e7ba57b104e9293f746342b7450b10d5fa0c4cd git-revert --no-edit f6b1f49300a24c329f5d1049031f711c6268be95 git-revert --no-edit 853fcd2329b427788677d11431e6165d108f935b git-revert --no-edit ca19d156400f817960efe0d14680324b2ea34171 [2] http://downloads.tuxfamily.org/3v1deb/openmoko/uImage-gta02-stable-3v1n0-git-mixture-2.bin -- Treviño's World - Life and Linux http://www.3v1n0.net/ ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
2008/11/6 Timo Jyrinki <[EMAIL PROTECTED]>: > -Timo (waiting for the evening or weekend to play more) Ok, lied. Doing echo 10 > /sys/devices/platform/lis302dl.1/threshold echo 10 > /sys/devices/platform/lis302dl.2/threshold fixes the problem on the current testing problem. cat /sys/devices/platform/lis302dl.1/threshold still says 0, and I don't know what's the default value either, but doing those two echo commands after reboot makes it work. So possibly some default could be set according to those... -Timo ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
2008/11/6 Timo Jyrinki <[EMAIL PROTECTED]>: > I installed/flashed back the kernel from testing. It might be the > threshold should just be lower. Unfortunately currently it seems that > /sys/devices/platform/lis302dl.1/sample_rate is fixed to 100. Ie. if I > do echo 10 > > /sys/devices/platform/lis302dl.1/sample_rate, it's immediately back to > 100 if I cat it. So I really can't test if the lower value would help. Uh, clearly too busy right now. Anyway the _threshold_ seems to be likewise fixed to 0 currently, ie. cannot be changed. The sample_rate was talked about in the commit notes. -Timo (waiting for the evening or weekend to play more) ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
2008/11/6 Michael Zanetti <[EMAIL PROTECTED]>: > What do you mean with "working fine now"? Are they working as they were always > or are they now really working (no crashes on suspend or after some minutes of > use)? "Working fine" as in accelerometer using applications (I have AccelGame and Doom) work, unlike with the patch (jerkiness). I don't use suspend with my own kernels since resume is always broken - probably because I only flash the self-compiled kernel and don't know how I could easily also copy the compiled modules into use on Neo. 2008/11/6 Alastair Johnson <[EMAIL PROTECTED]>: > Does it need reverting, or do you just need to echo a smaller value to > threshold when you want to use the accels for game control? I installed/flashed back the kernel from testing. It might be the threshold should just be lower. Unfortunately currently it seems that /sys/devices/platform/lis302dl.1/sample_rate is fixed to 100. Ie. if I do echo 10 > /sys/devices/platform/lis302dl.1/sample_rate, it's immediately back to 100 if I cat it. So I really can't test if the lower value would help. Reverting is probably pointless as such, as most probably the whole rework on the accelerometer side has it meanings (like the suggested suspend problems), but I just haven't found another way to get accelerometers working fluently without reverting the patch in question. -Timo ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
Timo Jyrinki wrote: > 2008/11/6 Andy Green <[EMAIL PROTECTED]>: >> | During the weekend I tried one self-compiled kernel with all the >> | "lis302dl" starting patches from >> | http://git.openmoko.org/?p=kernel.git;a=shortlog;h=stable reverted, >> | but it didn't help at all. Any idea which patch would be causing the >> | problem? >> >> How about this one: >> >> http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=35d910f40ce2c7c959edfbaa671f88e8da5fa6b1 > > Right! Reverting also that one fixed the problem, accelerometers work > fine now again. So apparently that patch would need some more work or > reverting. Does it need reverting, or do you just need to echo a smaller value to threshold when you want to use the accels for game control? ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
Am Donnerstag 06 November 2008 11:00:02 schrieb Timo Jyrinki: > Right! Reverting also that one fixed the problem, accelerometers work > fine now again. So apparently that patch would need some more work or > reverting. > What do you mean with "working fine now"? Are they working as they were always or are they now really working (no crashes on suspend or after some minutes of use)? ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
2008/11/6 Andy Green <[EMAIL PROTECTED]>: > | During the weekend I tried one self-compiled kernel with all the > | "lis302dl" starting patches from > | http://git.openmoko.org/?p=kernel.git;a=shortlog;h=stable reverted, > | but it didn't help at all. Any idea which patch would be causing the > | problem? > > How about this one: > > http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=35d910f40ce2c7c959edfbaa671f88e8da5fa6b1 Right! Reverting also that one fixed the problem, accelerometers work fine now again. So apparently that patch would need some more work or reverting. -Timo ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Re: Accelerometers jerky with new stable kernels?
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Somebody in the thread at some point said: | Hi, | | I get huge "stalls" with accelerometers if using the current kernel in | testing or self-compiled stable branch kernel. Eg. in accelgame or | opendoom, the fps is ca. 1-3, sometimes a few frame rapidly in | succession and then 1 second of stalling. Anyone else experiencing the | problem, possibly with a cure? | | During the weekend I tried one self-compiled kernel with all the | "lis302dl" starting patches from | http://git.openmoko.org/?p=kernel.git;a=shortlog;h=stable reverted, | but it didn't help at all. Any idea which patch would be causing the | problem? | | Okay, while writing this I noticed I didn't revert | "fix-lid302dl-bitbang-all-the-way-baby.patch", so it might be it, but | I'd like to hear from others and also if it's a known problem to be | fixed. | | -Timo, wants the accelerometers back (without losing other goodies in | recent stable kernel) How about this one: http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=35d910f40ce2c7c959edfbaa671f88e8da5fa6b1 - -Andy -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (GNU/Linux) Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org iEYEARECAAYFAkkSuygACgkQOjLpvpq7dMoQUgCggZ1GSNemdnhdV1BnegngF+3U a5YAniuPHsER0WLhv2k6HSiKlMEhw8sn =zn5m -END PGP SIGNATURE- ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support
Accelerometers jerky with new stable kernels?
Hi, I get huge "stalls" with accelerometers if using the current kernel in testing or self-compiled stable branch kernel. Eg. in accelgame or opendoom, the fps is ca. 1-3, sometimes a few frame rapidly in succession and then 1 second of stalling. Anyone else experiencing the problem, possibly with a cure? During the weekend I tried one self-compiled kernel with all the "lis302dl" starting patches from http://git.openmoko.org/?p=kernel.git;a=shortlog;h=stable reverted, but it didn't help at all. Any idea which patch would be causing the problem? Okay, while writing this I noticed I didn't revert "fix-lid302dl-bitbang-all-the-way-baby.patch", so it might be it, but I'd like to hear from others and also if it's a known problem to be fixed. -Timo, wants the accelerometers back (without losing other goodies in recent stable kernel) ___ support mailing list support@lists.openmoko.org https://lists.openmoko.org/mailman/listinfo/support