3.12-rc4: kernel testing time...

RobertCNelson · October 11, 2013, 12:09am

So here is the deal; 3.8 is old, and 3.12-rc4 is newer/better/etc...

Functionally, it seems ready in my small 5 board sample size. 3 of
which have been running 24/7 @ 100% since rc3..

Anywho, we need community testing. (Here's looking at you!)

Kernel Branch:
https://github.com/beagleboard/linux/tree/3.12

Kernel src Script:
https://github.com/beagleboard/kernel/tree/3.12

Regression/testing:
Are we missing any kernel config settings from 3.8 that you used?
Does your Cape still work?

Other thoughts...
Some users will build/install these from src, however anyone
interested in a quick "test-me.sh*" script which will install the new
kernel on a daily/weekly basis?

* script currently doesn't exist, but after a few beers it might...

Regards,

Li926744 · October 11, 2013, 1:06am

awesome

William_Hermans · October 11, 2013, 1:14am

I’d be willing to test Robert,

However, I’d want to netboot it. So, I’d need some way to achieve that. If the 3.12-rcx instructions on your eewiki cover that, then I should be able to deal with that.

Now if there were a way I could use your prebuilt test images . . . that’d be even better. But I have not yet taken a look at them.

RC4 has working USB ?

William_Hermans · October 11, 2013, 1:16am

err sorry, i meant if your 3.12-rcx instructions covers what you’re talking about here.

Context → BBB

RobertCNelson · October 11, 2013, 1:38am

Correct, we just fixed it today.. (it was a config issue the whole
time, and usb had been working since rc1..)
https://github.com/beagleboard/kernel/commit/e1ec2a113936a779cb78df286bd66d931bfe9ea5

With my "netinstall" you can use the [--use-beta-kernel] just not all
netinstall targets are fully built yet.. It's 'bone5' for the tag..

Regards,

Charles_Steinkuehler · October 11, 2013, 1:43am

Regression/testing:
Are we missing any kernel config settings from 3.8 that you used?
Does your Cape still work?

Does it come with Xenomai support?!?

And when can we expect SGX acceleration?

I'm just beginning to test SGX acceleration with the 3.8 kernel (I found
some magic sauce by talking to Jason Kridner who hooked me up with
Prabindh Sundareson). I haven't had a chance to test anything yet (just
got back from a business trip), but I've managed to compile drivers
without anything that looks like a fatal error.

Other thoughts...
Some users will build/install these from src, however anyone
interested in a quick "test-me.sh*" script which will install the new
kernel on a daily/weekly basis?

* script currently doesn't exist, but after a few beers it might...

If you want, I'll send you some of the local 8-bit brew:

http://www.tallgrassbeer.com/beers/8-bit

...it's the least I can do for all your contributions to the community!

jkridner · October 11, 2013, 3:06am

So here is the deal; 3.8 is old, and 3.12-rc4 is newer/better/etc...

Functionally, it seems ready in my small 5 board sample size. 3 of
which have been running 24/7 @ 100% since rc3..

Anywho, we need community testing. (Here's looking at you!)

Kernel Branch:
GitHub - beagleboard/linux at 3.12

I probably need to add the built M3 firmware and copy the defconfig
over from the other tree. Thoughts?

I'm seeing an error in the am335x-boneblack.dts where emmc_pins isn't
defined. Anybody have a definition for that?

The kernel also seems to be missing the reset control line which is
critical for disabling the eMMC.

RobertCNelson · October 11, 2013, 12:49pm

Regression/testing:
Are we missing any kernel config settings from 3.8 that you used?
Does your Cape still work?

Does it come with Xenomai support?!?

I sense, we have a new maintainer for github.com/beagleboard/kernel

If we are going to add the Xenomai patchset, i'd feel more comfortable
if we had a developer who is using/testing/etc..

Part of the goal of that repo, is just have enough for everyone to use
and base off it (for Fedora, arch, etc..)

And when can we expect SGX acceleration?

I can't answer that. I'm a user too like you guys, and i hate NDA's
with a passion and closed "stuff"..

I'm just beginning to test SGX acceleration with the 3.8 kernel (I found
some magic sauce by talking to Jason Kridner who hooked me up with
Prabindh Sundareson). I haven't had a chance to test anything yet (just
got back from a business trip), but I've managed to compile drivers
without anything that looks like a fatal error.

Other thoughts...
Some users will build/install these from src, however anyone
interested in a quick "test-me.sh*" script which will install the new
kernel on a daily/weekly basis?

* script currently doesn't exist, but after a few beers it might...

If you want, I'll send you some of the local 8-bit brew:

http://www.tallgrassbeer.com/beers/8-bit

...it's the least I can do for all your contributions to the community!

Tall grass makes awesome beer.. PS, i never short of beer at my place..

Regards,

RobertCNelson · October 11, 2013, 12:51pm

So here is the deal; 3.8 is old, and 3.12-rc4 is newer/better/etc...

Functionally, it seems ready in my small 5 board sample size. 3 of
which have been running 24/7 @ 100% since rc3..

Anywho, we need community testing. (Here's looking at you!)

Kernel Branch:
GitHub - beagleboard/linux at 3.12

I probably need to add the built M3 firmware and copy the defconfig
over from the other tree. Thoughts?

It should be part of the build already..

github.com

beagleboard/kernel/blob/3.12/patches/general-fixes/0001-add-PM-firmware.patch

From ec313f7c43675adece510e57dc737cf495afa4a2 Mon Sep 17 00:00:00 2001
From: Koen Kooi <koen@dominion.thruhere.net>
Date: Thu, 6 Jun 2013 17:09:38 +0200
Subject: [PATCH 1/3] add PM firmware

Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
---
 firmware/am335x-pm-firmware.bin | Bin 0 -> 10964 bytes
 1 file changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 firmware/am335x-pm-firmware.bin

diff --git a/firmware/am335x-pm-firmware.bin b/firmware/am335x-pm-firmware.bin
new file mode 100644
index 0000000000000000000000000000000000000000..1e86780666ed2cad030adb969ecf3697501b67bb
GIT binary patch
literal 10964
zcmds7eQ;aVmA}$UV!IB+jU6hage2=Mm?#h@Znt*Kc4doI+l59UVM8Z;j9eF14#O75
zv#ew7QW0n|TZWQuhJxwrgbZ|P%QWDo(-M=hGbDe&veZsmXp&`68anKHVCGq<maJ##
z?eE+Vz58T2A@E;4bKmv(xW9YuIrrUr?tNl))$FqdV-KJ_g0dIoDU@eW%Kv@>-b<4I
zGU}Qt#(GhInsW40w0}*uU9o_%FQDv3*^8pv^ApfVQBI*uq0}s7Y&A+7%1tOeD7`3;

This file has been truncated. show original

but is it "new" enough, probally should pull from the *.bin repo again..

I'm seeing an error in the am335x-boneblack.dts where emmc_pins isn't
defined. Anybody have a definition for that?

The kernel also seems to be missing the reset control line which is
critical for disabling the eMMC.

I didn't notice those issues, I'll take a look today at it..

Regards,

jkridner · October 11, 2013, 1:53pm

So here is the deal; 3.8 is old, and 3.12-rc4 is newer/better/etc...

Functionally, it seems ready in my small 5 board sample size. 3 of
which have been running 24/7 @ 100% since rc3..

Anywho, we need community testing. (Here's looking at you!)

Kernel Branch:
GitHub - beagleboard/linux at 3.12

I probably need to add the built M3 firmware and copy the defconfig
over from the other tree. Thoughts?

It should be part of the build already..
kernel/patches/general-fixes/0001-add-PM-firmware.patch at 3.12 · beagleboard/kernel · GitHub

Sorry, I did notice it is in there. I won't update it until I have
things otherwise working.

but is it "new" enough, probally should pull from the *.bin repo again..

I'm seeing an error in the am335x-boneblack.dts where emmc_pins isn't
defined. Anybody have a definition for that?

The kernel also seems to be missing the reset control line which is
critical for disabling the eMMC.

I didn't notice those issues, I'll take a look today at it..

I'm back to playing with this today in hopes of enabling more people
to do some testing, but it seems we still have quite a few out-of-tree
patches.

I don't see this one from Koen in the patches:
https://lkml.org/lkml/2013/9/12/427

That should fix emmc pins. The tree is missing just about everything
in firmware/capes, so those'll need to be applied as well.

Charles_Steinkuehler · October 11, 2013, 3:11pm

Regression/testing:
Are we missing any kernel config settings from 3.8 that you used?
Does your Cape still work?

If it wasn't obvious, my initial comments were meant to be
toung-in-cheek. Great work on 3.12, but I need enough out-of-tree code
I'll be stuck playing catch-up on 3.8 for a while.

Does it come with Xenomai support?!?

I sense, we have a new maintainer for github.com/beagleboard/kernel

I'm sure I don't know what you mean...

< me: runs and hides >

If we are going to add the Xenomai patchset, i'd feel more comfortable
if we had a developer who is using/testing/etc..

Part of the goal of that repo, is just have enough for everyone to use
and base off it (for Fedora, arch, etc..)

I don't think the xenomai patches should be in the defalut Beagle*
kernels. I was meaning more that xenomai just recently began working on
3.8 kernels for the ARM architecture, and work is ongoing to get 3.10
kernels working on x86. I suspect it will be a while before ARM 3.12
kernels have an easy-to-apply xenomai patch. :-/

And when can we expect SGX acceleration?

I can't answer that. I'm a user too like you guys, and i hate NDA's
with a passion and closed "stuff"..

I'm hopeing I can get the SGX drivers working on 3.8. I'm back home and
can test now, if only I could find some time...

I did hear rumors here about TI jumping from 3.2 to 3.12 kernel support
for the SGX drivers, so if that actually happens, I can probably manage
porting the xenomai stuff. At least it's open source!

RobertCNelson · October 11, 2013, 3:29pm

Good point, i'm re-basing the patch set right now so we have all the
changes from Benoit Cousson's tree for v3.13-rcX

Regards,

David1 · October 11, 2013, 4:00pm

Excuse my ignorance, but could you point me to an "idiots guide" for download/building this kernel. The README.md still mentions the 3.8 kernel. Is it simply a matter of RTFM plus changing 3.8 to 3.12?

TIA,

Dave.

RobertCNelson · October 11, 2013, 7:29pm

David,

I just pushed an update to fix the readme file.

We can probably simplify a few more things in it.

Does Angstrom use zImage or uImage for the bbb in the default flasher image?

Regards,

David1 · October 11, 2013, 7:47pm

Excuse my ignorance, but could you point me to an "idiots guide" for
download/building this kernel. The README.md still mentions the 3.8 kernel.
Is it simply a matter of RTFM plus changing 3.8 to 3.12?

David,

I just pushed an update to fix the readme file.

Thanks. I will try that

We can probably simplify a few more things in it.

Does Angstrom use zImage or uImage for the bbb in the default flasher image?

Mine uses uImage

HTH,

Dave.

RobertCNelson · October 11, 2013, 7:51pm

ah... okay because uImage gets interesting in v3.12

make ARCH=arm LOADADDR=0x80008000 uImage

Regards,

Robert_P_J_Day · October 11, 2013, 8:15pm

is there any actual *technical* reason to continue using uImage now
that u-boot for BBB has switched to using zImage by default?

rday

RobertCNelson · October 11, 2013, 8:21pm

ah... okay because uImage gets interesting in v3.12

make ARCH=arm LOADADDR=0x80008000 uImage

is there any actual *technical* reason to continue using uImage now
that u-boot for BBB has switched to using zImage by default?

Really none, unless the user doesn't want to update their
bootloader/uEnv.txt just to test..

Regards,

David1 · October 11, 2013, 8:30pm

Still confused. If I try to clone using:

git clone https://github.com/beagleboard/kernel.git

I get the README.md:
dlambert@development ~/src/kernel $ cat README.md
kernel

RobertCNelson · October 11, 2013, 8:48pm

git clone https://github.com/beagleboard/kernel.git
cd kernel
git checkout origin/3.12 -b 3.12

Regards,