[jas0npc]

ok mate will do that now

[jas0npc]

do i need to delete the 3.6.7 folder?

[popcornmix]

do i need to delete the 3.6.7 folder?

You can - but not essential.
If it still reboots, I'd be interested to know if:
sudo rpi-update 788cddf57e1c965_0f34b2cdea8c74410626805a9
works. (Latest 3.6.10 firmware, but without my cma mod)

[jas0npc]

Yep ok i will let you know mate do i just use rpi-update to grab it?

---

I just had boot loop again, so will reinstall raspbmc, and try again, whats the best way to update libs, modules?

[popcornmix]

Did you run:
sudo rpi-update 788cddf57e1c965_0f34b2cdea8c74410626805a9
(minus the _)
To update the modules (after getting the kernel_nocma.img):
cd ~
wget http://dl.dropbox.com/u/3669512/temp/ker...odules.tgz
cd /
sudo tar xf kernel_nocma.modules.tgz
will probably work...

[jas0npc]

I must be totally thick copied the above perfect but still getting the relax boot loop

[Sam.Nazarko]

just a heads up I am bundling /opt/vc with XBMC now. Probably not a good idea but it's been quite successful as of late.

---

Make sure you don't get vchiq incompatible library by copying to /opt/xbmc-bcm/xbmc-bin/lib/xbmc/system

[jas0npc]

Fresh day, Fresh Mind, lets try again, Will post logs update, soon

[jas0npc]

Ok started again this morning, fresh install of raspbmc used rpi-updater to update fw to 3.6.10

Code:

https://dl.dropbox.com/u/94360623/xbmc%20log%20testing/putty.log%20install%203.6.10%20fw

now stuck in relax loop, so did the above to update libs modules and still in relax loop

Code:

https://dl.dropbox.com/u/94360623/xbmc%20log%20testing/putty.log%20install%20update%20libs%20modules

[rbej]

Use OpenElec Beta 5.

Copy advancedsettings.xml from usr/share/xbmc/system to storage/.xbmc/userdata and change <cachemembuffersize>5242880</cachemembuffersize> in advancedsettings.xml.

Try set "2000000". If this dont help try other values cache memory buffer size.

Im now tested small buffer 2621440.

[jas0npc]

Hi mate, Is this a fix or a workaround, Have you tried the fw and libs and modules the popcornmix posted above? as this happens in all xbmc distros on the pi.

[rbej]

I found this solution on Raspbmc forum.

nope only the 256 version

Scratch my previous... it seems it still reboot doing full 1080p streaming off navi x.. not sure why though :-( ... a good test if using navi x is finding nemo 1080p in BMs HD movies-- Tests to follow...

So the tests..

t i set the cache to 30971520
i paused a video and let it buffer, and watched the memory usage in putty with the "top" command once the memory got to 65% which was about 27-28mb buffered, the pi restarts

i then set the cache to 20971520 and played the same 1080p clip ( not finding nemo)
This time the buffer got to 23.32 , with memory showing 62% in use... and it didn't restart :-)

However on that finding Nemo one, restarts instatly sometimes.. so im guessing that uses more memory as its a proper 1080p so for me 20971520 is right on the edge if its a full 1080p clip, the Finding nemo one is a proper 1080p stream, the resolution is actually 1920 x 1080 with a high bitrate...
i managed to get it to play , and did the same test , using 20971520... i let it buffer to max which was 21.89 and the mem usage was 66% so im guessing thts on the edge of restart....

just fyi

So iv put my cache mem down to 2000000 and have so far had no issues.

[jas0npc]

I tried every version of OE on thestateofme and only found this problem started when fw 3.6.3 was used up untill now, yesterday i did a fresh install of raspbmc tried "the campain" in hd caused xbmc task to be killed, downgraded fw to 3.2 played perfect, did a fresh install of latest OE same thing, downgraded fw to 3.2 played perfect

[rbej]

How downgrade Kernel in OpenElec Beta 5 to 3.2??.

[popcornmix]

Use OpenElec Beta 5.

Copy advancedsettings.xml from usr/share/xbmc/system to storage/.xbmc/userdata and change <cachemembuffersize>5242880</cachemembuffersize> in advancedsettings.xml.

Try set "2000000". If this dont help try other values cache memory buffer size.

Im now tested small buffer 2621440.

Can I ask people to try this. From my testing, I'm not convinced the newer kernel has a leak. The 3.2 kernel does start with 80M free on ARM at boot.
The 3.6 kernel starts with 76M free on boot. The 3.6 kernel with CMA disabled has 78M free on boot.

I think possibly the extra 4M is producing the difference between Campaign playing or not playing.

I think the key thing is that when playing streams from internet, the cachemembuffersize of 5242880 requires an extra allocation of ~16M which is very tight on a 256M board.
Lowering it to 2000000, means it only allocates 6M.