25 replies to this topic

[goddeex]

Hello,

I just want my own kernel, to be able to have USB On The Go (if that's possible !). And to understand much.

I build the .config after copying the one in the kernel directory on arch/arm/configs/msm7627a-perf_defconfig

After that, make does its jobs and gcc compiles some c files, until I get this error:

make[1]: ***  Aucune règle pour fabriquer la cible « /kernel/include/oem/oem_rapi_events.h », nécessaire pour « arch/arm/mach-msm/proc_comm_version_checksum.h ». Arrêt.

(no rule to make target), (needed for)

The problem is that I don't have oem_rapi_events.h !

Do you know where can I find it ?

Thx in advance

Xav

[vache]

I never found it too, to bypass this error have a look @ https://github.com/V...ch-msm/Makefile

[goddeex]

What a buggy kernel source !
Thks for your Makefile, this way it starts to compile

Had to edit "gcc-wrapper.py" to be able to compile (disabling halt on unallowed warnings)
Had to disable CONFIG_DEBUG_FS
Had to disable CONFIG_MSM_SMD

Had to comment some code for cpu check.
Had to comment some include for MTD compiles however.

Now seams that HW3D doesn't compile. See you next.

Also, got the linux kernel from kernel.org, compiles for ARM without problem, but doesn't support our processor so I'm still working on the buggy kernel.

[vache]

All path to oem/*.h are wrong, you have to edit them.
There's also some wrong hardcoded path that have to be edited.

There's no need to comment line(s). (gcc-wrapper excluded)

[goddeex]

Indeed, all is starting to go wrong when I get:

arch/arm/mach-msm/proc_comm.c:193: error: 'SMEM_PC_OEM_CHECK_VERSION' undeclared (first use in this function)

After that, lots of fixes are needed.
Also a problem with boot_mode.h

Also I replace #include oem_smem_struct.h with a comment line. I don't know where to find this file.

[goddeex]

Hey, what a mess, it is in /include/oem/oem_smem_struct.h !

Alright will try tomorrow !

[vache]

That's what i said, you have to change path to <oem/*.h>.
There is around 10 path to fix.

[lupohirp]

 goddeex, on 04 December 2012 - 09:10 AM, said:

What a buggy kernel source !
Thks for your Makefile, this way it starts to compile

Had to edit "gcc-wrapper.py" to be able to compile (disabling halt on unallowed warnings)
Had to disable CONFIG_DEBUG_FS
Had to disable CONFIG_MSM_SMD

Had to comment some code for cpu check.
Had to comment some include for MTD compiles however.

Now seams that HW3D doesn't compile. See you next.

Also, got the linux kernel from kernel.org, compiles for ARM without problem, but doesn't support our processor so I'm still working on the buggy kernel.

DON'T disable msm_smd....we have a mtd device here....without that sdcard could be not work
although a request for vache....which .ko modules are needed for glow to boot???


EDIT: GREAT NEWS!!! I HAVE FINALLY A BOOTING KERNEL!!!!...ok now i have problem with sleep and with wifi that doesn't power on in other words when i press lockscreen phone doesn't power up anymore. seems a .ko modules....i've seen that dimension of my modules are too big compared to vache's custom_kernel.zip....vache did you stripped the modules??if yes...how did you that??

Edited by lupohirp, 04 December 2012 - 10:45 PM.

[vache]

You don't need any module to boot, and i didn't stripped them. Did you had a look at system and kernel logs ?

[goddeex]

Hello, I finally managed to boot my own linux kernel, although I didn't have sim card support and wifi !
I have put the path to the .h files in the LINUXINCLUDE section of Makefile.

Now trying to play with menuconfig to get it work.

How do you access kernel logs, please ?

[lupohirp]

 goddeex, on 05 December 2012 - 09:30 AM, said:

Hello, I finally managed to boot my own linux kernel, although I didn't have sim card support and wifi !
I have put the path to the .h files in the LINUXINCLUDE section of Makefile.

Now trying to play with menuconfig to get it work.

How do you access kernel logs, please ?

you can use dmesg from adb

 vache, on 04 December 2012 - 10:56 PM, said:

You don't need any module to boot, and i didn't stripped them. Did you had a look at system and kernel logs ?

kernel logs stopped when i press the on/off button....for wifi it says that system can't load module.....i had similar problems with p500 when i was workin' at cm9 and i remembered that i compiled kernel with android ndk toolchain instead of linaro (also i had big sizes of module with ndk)....later i will try to recompile with linaro and codesourcery

[goddeex]

I can't access to my phone using usb,  so can't do "adb shell", so I thing something is missing in my .config file.

I also activated USB OTG in the config file, but an error appears at compile time. Put some coments in the C file, that compiles, but i'm unsure about what I've done. Result: no USB OTG.

[vache]

 lupohirp, on 05 December 2012 - 09:35 AM, said:

kernel logs stopped when i press the on/off button....for wifi it says that system can't load module.....i had similar problems with p500 when i was workin' at cm9 and i remembered that i compiled kernel with android ndk toolchain instead of linaro (also i had big sizes of module with ndk)....later i will try to recompile with linaro and codesourcery

Try to add "CFLAGS_MODULE  = -fno-pic" in makefile then rebuild modules.
Or you can try to disable CONFIG_MODVERSIONS.

 goddeex, on 05 December 2012 - 12:10 PM, said:

I can't access to my phone using usb,  so can't do "adb shell", so I thing something is missing in my .config file.

I also activated USB OTG in the config file, but an error appears at compile time. Put some coments in the C file, that compiles, but i'm unsure about what I've done. Result: no USB OTG.

Using the provided msm7627a-perf_defconfig is enought to have a working kernel.

Edited by vache, 05 December 2012 - 01:37 PM.

[goddeex]

Yes, with this default config I have a working kernel, but the wifi and the sim card are not enabled. So I can't even make a call.

From what I have seen, it seams that if you try to modify .config your going to have problem because the arm code is somehow beta version.

[lupohirp]

ok i've resolved wifi problems....but when screen goes off phone is dead and doesn't worked anymore...must pull out battery for start another time...any suggestions??


EDIT: NO MATTER I HAVE RESOLVED THIS ERROR FINALLY!!!!WE HAVE A FULL BOOTING KERNEL NOW!!!!BUT WHAT A CRAP THESE SOURCES!!! even lg's sources was better than these....i will upload all my code on github....BUT A HUGE THANKS GOES TO VACHE

Edited by lupohirp, 05 December 2012 - 10:47 PM.

[goddeex]

Hello, I have found that some files are "read only" in the kernel, this is why I had problems when modifying .config. I did "chmod a+w -R *" and it seams to solve the automatic "make oldconfig" problem asking for every parameter when launching make.

However, hub.c doesn't compile when activating USB_OTG.

[vache]

Could you post the error you get ?

[goddeex]

Yes I get this error:

drivers/usb/core/hub.c: In function 'usb_enumerate_device_otg':
drivers/usb/core/hub.c:1814: error: 'struct usb_otg_descriptor' has no member named 'bcdOTG'

And, it's true, I searched everywhere in the kernel source, I found that there is no member named 'bcdOTG' in the struct. So I think it doesn't come from a dependency error.

There are comments "FIXME" everywhere in hub.c
The new generic linux kernel 3.6.9 has new code for this function in hub.c

The problem comes from this line in the function that enumerates usb OTG devices:

/* a_hnp_support is not required for devices
* compliant to revision 2.0 or subsequent
* versions.
*/
if (le16_to_cpu(desc->bcdOTG) >= 0x0200)
    goto out;

I put a comment on the if and goto out lines, that compiles but I get no usb OTG !
Wonder what is a_hnp_support ?

[vache]

I found out that commenting this is a good way for compiling, keeping USB OTG support.

I think that the issue is not here.

[lupohirp]

my sources are up

https://github.com/l...uid_Glow_Kernel