Confusing issue of Android m5_rc14

I'm working on and off porting m5_rc14 to Zaurus. Yet I haven't seen anything on LCD. Just black out. In the emulator, fb0 the frame buffer device is in /dev/graphics instead of /dev like Zaurus Angstrom. Crossing fingers, I copied it to graphics directory.

# mkdir /dev/graphics
# cp /dev/fb0 /dev/graphics/fb0
# chmod 666 /dev/graphics/fb0

No luck. Still blank screen. Though fuser command and strace says that runtime and dalvikvm opens successfully /dev/graphics/fb0. Weird.

When I try to see Android log by logcat command, it says an error

/dev/log: Not a directory

Well, it is not a directory but a file in Zaurus. In the emulator, it is a directory. So I checked the source code of the logger and found

#define LOGGER_LOG_RADIO	"log_radio"	/* radio-related messages */
#define LOGGER_LOG_EVENTS	"log_events"	/* system/hardware events */
#define LOGGER_LOG_MAIN		"log_main"	/* everything else */

in include/linux/logger.h. This is really weird. logcat expects /dev/log as directory and logger driver creates log_xxx in /dev. Is there any glitches between m5_rc14 kernel and m5_rc14 Android?

Comments

android-sdk_m5-rc14 on Real Target "armadillo-500"
http://cafe.naver.com/ArticleRead.nhn?articleid=859&sc=e0d0301f0f46299f1c&clubid=14355692#

Posted by mung at 02/20/2008 11:36

Thanks, mung. Hmmm... What's the difference...

Posted by androidzaurus at 02/20/2008 12:36

> # cp /dev/fb0 /dev/graphics/fb0
This is the most stupid thing I have ever seen :-)

You should have created a symlink, like this:
# ln -s /dev/fb0 /dev/graphics/fb0
So just do a "rm /dev/graphics/fb0" and then create the symlink as I stated.

Now for the logs: just remove the current /var/log file and create a a brand new /dev/log directory. The angstrom kernel won't be able to log anything but Android will.

My 2 cents...

Posted by Olivier at 02/20/2008 18:21

if you want to copy a device, you need to do

cp -prd /dev/XXX .

otherwise you're reading the file represented by the file

Posted by Paul at 02/22/2008 07:55

p.s. better to use mknod to create the device entry!

Posted by Paul at 02/22/2008 07:57

Thanks, guys!

Sometime I forget the very basic thing...

Posted by androidzaurus at 02/22/2008 08:35

> No luck. Still blank screen.

Did you solve this already?

Have similar issue on an other device (TI OMAP, ARM9 based):

http://marc.info/?l=linux-omap&m=120430992719131&w=2

Posted by Dirk at 03/01/2008 05:17

Dirk,

Cortez implemented double buffer and pan function in pxafb.c and m5 is finally up.
http://www.omegamoon.com/blog/index.php?entry=entry080228-220020

Posted by androidzaurus at 03/01/2008 11:13

Many thanks for the hint! Are the changes available somewhere (modified pxafb.c or patch)?

All I could find yet is

http://www.oesf.org/forum/index.php?showtopic=25517&st=15

"More info later..."

Posted by Dirk at 03/01/2008 15:32

Dirk,

No patch so far I've found. I was thinking to write a code myself, but cortez has already done it. So I'm just looking forward to it.

Posted by androidzaurus at 03/01/2008 16:31

For the archives: For pxafb see

http://androidzaurus.seesaa.net/article/87973048.html

Posted by Dirk at 03/02/2008 17:27

Yeah, I've done it ;-)

But it's really primitive. Needs some brush up.

Posted by androidzaurus at 03/02/2008 18:03

バイアグラ販売

Posted by バイアグラ販売 at 01/17/2013 22:13