maemo.org - Talk - Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n810"?

maemo.org - Talk (https://talk.maemo.org/index.php)

- General (https://talk.maemo.org/forumdisplay.php?f=7)

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

anyone else have trouble with the built-in media player slipping after intstalling this? I've been having a small amount of stutter but there are a few other things that I could blame this on as well ;-)

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

Quote:

Originally Posted by brontide (Post 210260)

anyone else have trouble with the built-in media player slipping after intstalling this?

try running as root

Code:

renice 0 `pidof mmcqd`

If that helps add this line somewhere to /usr/sbin/osso-mmc-mount.sh. It needs to be run each time card is inserted.

Quote:

Originally Posted by brontide (Post 210260)

I've been having a small amount of stutter but there are a few other things that I could blame this on as well ;-)

Yes, there are more things to blame so maybe the tip above won't help you.

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

FYI, that seemed to do the trick for me. What is mmcqd and what was it's original priority?

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

Quote:

Originally Posted by brontide (Post 210467)

What is mmcqd and what was it's original priority?

MMC workqueue kernel thread. Sometime around 2.6.18 they changed API for starting kernel threads. Once MMC code started to use new API, priority of mmcqd 'automagically' raised from 0 to -5. I'm not sure if it is bug or feature. -5 is above most important system daemons including osso-media-server. IMO lowering it back to 0 should not do any harm, there should be no time critical code that needs higher priority in there.

New mmcqd thread is started for each slot when card is inserted and it stops when card is removed. I don't think the stutter is related to 24 vs 48MHz bus speed but never verified this theory. It happens more often when booting from card due to higher MMC I/O traffic overall.

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

Hmmm...wirdness.

I was looking over everything and I noticed that /media/mmcX didn't have noatime. So I did a mount -o remount,noatime and the problem went away, but now canola is flipping out. It works *significantly* faster displaying images until the images start to corrupt and then it just barfs. I have to kill it off with a kill -9 on the python-launcher.

It's the only application that I'm having trouble with at this point. Funny enough just setting noatime and without the renice 0 `pidof mmcqd` media player doesn't skip anymore either.

I'm going to try with python-launcher disabled.

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

Two observations

1) Locking the max CPU speed to 330,000 fixes the problem
2) Going back to the non-highspeed kernel fixes the problem

I think I may have actually come across the corruption that is warned about. Thankfully I was only doing reading and not writing.

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

Quote:

Originally Posted by brontide (Post 210571)

Hmmm...wirdness.

I was looking over everything and I noticed that /media/mmcX didn't have noatime. So I did a mount -o remount,noatime and the problem went away

wow, noatime makes sense with vfat, never thought of it (there is no access time field in fat filesystem). Did some google search and indeed

Quote:

Alert readers may be thinking: “Wait, there's no atime on FAT filesystems!” True. There's only a single timestamp field for each file or directory, but Linux deals with this by updating that single time value on any occasion it ordinarily would update atime, mtime or ctime. So, disabling atime still reduces the frequency of erase/write operations, even on FAT.

If this really makes difference I guess it should be reported as maemo bug. metalayer-crawler scans everything and also file open dialog scans a lot files, both may cause big atime updates ==> unneeded writes, audio stutter and cpu burning in mmcqd.

Quote:

Originally Posted by brontide (Post 210571)

, but now canola is flipping out. It works *significantly* faster displaying images until the images start to corrupt and then it just barfs. I have to kill it off with a kill -9 on the python-launcher.

See kernel log via dmesg, are there some mmc related errors? If not then silent mmc corruption is unlikely. 48MHz mode is indeed unsupported by TI but my guess is that if it doesn't work under some conditions you would see device hangs and ton of errors in kernel log, not silent corruption of data.

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

So.. anyone know why rotation won't work on my tablet after installing the .deb? Or do I still have to flash it with an image?

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

You have to install the xomap-xserver too. I have a post with all the files you need but it's easier installing emjayes's sliderotate, running the script to install the kernel, rebooting and reinstalling my kernel with rotation and mmc speed from this thread.

Re: Why there is lack of interest "Highspeed enabled kernel for the Nokia n800 and n8

Quote:

Originally Posted by fanoush (Post 210676)

I'll reapply the kernel and monitor dmesg and post any results. Between high-speed and noatime the difference was significant over stock. It's also possible that the problems were cause by running out of free space on /.

I have already posted this as a bug report and it's a 1 *WORD* fix that i have included with the bug report.

https://bugs.maemo.org/show_bug.cgi?id=3543