. Im Root-Verzeichnís habe ich folgende Auswahlmöglichkeiten:
.
slohstraeter schrieb:
Die Kernel sind seit langer Zeit in den Rom's integriert, du hast den Froyo Kernel doch nicht etwas geflasht, oder? 
.
Blechdose schrieb:Good evening
Had some time to make another update. Short changelog;
* changed the behaviour of interactive in terms of scaling up to max_freq instead of hispeed_freq
* patched in google's sync driver for android
* backported the cgroups timer slack controller
* some little fixes, check out github
* (and of course all previous changes in case you missed it.)
Yep, the last change is quite interesting. It also requires a modified ramdisk, so this update contains a modified ramdisk by me.
I attached both CM10 and CM10.1 zips. The one for CM10.1 should also work for CM10.2, but keep in mind i have only tested
it with CM10.1.
The following is confusing and can cause headaches. Be warned.
So you are probably asking yourself "what about ion?". Well ion is tough. Everything looks, it compiles, it boots up, it
registers the ion_device properly. The funny part is the initialization; on our 3.0 branch we use module_init(ion_init) to
load the ion driver like a module. On 2.6.32 we have to use subsys_initcallion_init) otherwise it wont even boot and we
have no clue why. Now the subsyscall has some special thing; its order of execution and because of that the ion_device
can't allocate its memory and can't use anything.
So here is what i assume; because of the very early init the ion_device tries to allocate memory thats not ready yet and
therefore fails. With module_init(ion_device) this is probably no problem because it gets way later initialized. Now we have some
other devices that gets memory allocated that early, namely our gpu or the be more precisely vram. In fact it gets allocated
right before the ion stuff is happening. Also i assume the problem is not related to our driver/gpu/ion driver instead i think
the problem is in our platform code namely in arch/arm/mach-omap2.
To put even more questionmarks in your heads; the current nexus 4 kernel head uses ion too. But they init it with subsys_initcall.
As i said, pretty tough.
