Panda Boot hang w/ linaro 2.6.39 kernel

Config for sha 54dab05c329c5fb6121d89fddf9b82e0fed35efb from linaro-2.6.39 tree that can be used to reproduce the hang.

Interestingly it seems the second config specific hang (after "regulator_init_complete: VUSB: incomplete constraints, leaving on") seems to be linked to early printk.

Avik, since the PMU patch appears to make a difference, could you have
a look and confirm if you get the same problem?

With the PMU patch I could boot to shell prompt with the config file attached in comment #1 (see attached log). Though sometimes I do get hang at "regulator_init_complete: VUSB: incomplete constraints, leaving on"

Yea, the second hang seems to be more sporadic.

Hrm. Maybe its a gcc versioning issue? My gcc is 4.4.0. Maybe that's too old? Or maybe there's an actual board difference?

Updating my crosscompiler to 4.6.0 to verify its a gcc issue or not.

Reproduced the same thing w/ gcc 4.6.0. Maybe something specific to the board then?

John, do you think this is likely related to #803142?

I don't think its directly related (as that bug deals with 2.6.38 based tree) , but bug #803142 does point out that uboot differences need to be taken into account as well.

FYI, I've been using "U-Boot 2011.03 (Apr 20 2011 - 07:37:43)" and my Pandaboard is Rev A1.

I see this with this build:

https://android-build.linaro.org/builds/~linaro-android/panda-11.06-release/#build=3

after I have replaced the uImage with jstultz's tip:

git://git.linaro.org/people/jstultz/android

commit 666bc7df26c1510701b3edb6d7e07b484b0081fe
Merge: 9072700 54dab05
Author: John Stultz <email address hidden>
Date: Fri Jun 24 15:59:23 2011 -0700

    Merge branch 'upstream/linaro.39' into linaro-android.39

    Conflicts:
        drivers/mmc/card/block.c
        drivers/mmc/core/core.c
        drivers/mmc/core/mmc_ops.c
        drivers/mmc/core/quirks.c
        include/linux/mmc/card.h
        include/linux/mmc/core.h
        include/linux/mmc/host.h

Zach seems to be seeing the same issue with "U-Boot 2011.06 (Jun 29 2011 - 07:59:26)"
http://pastebin.ubuntu.com/635883/

Info from #pandaboard

nhg> pfefferz: sounds like the function in the display driver that reads the resolution info from your monitor is failing...see some similar queries: http://www.spinics.net/lists/linux-omap/msg48636.html
<pfefferz> hey cool
<pfefferz> thanks

Avik: I realize you don't see the issue, and that makes it really hard to debug. Could you maybe send out a debug patch to me that would help narrow down where the issue is?

Otherwise, I might push pretty hard to get the patch reverted until the hang can be figured out. And I'd be happy to test patches from you so we could later get it back in.

Its just these sorts of breakage are way too common in the ARM tree these days (see the linux-3.0 ehci boot hang on panda), and its holding back our ability to make solid releases. So I think we really have to be diligent about quickly reverting code that breaks users if a fix isn't immediately found.

Yeah...we're tracking tip and if its broken than people won't be able to develop. Going forward we won't be able to take commits that cause tip not to boot.

John, I don't have any debug patch as such. The commit 4fefbe94a77c3d3b5e75d386628c36298c72c57f was taken from http://lists.infradead.org/pipermail/linux-arm-kernel/2011-March/045283.html written by Ming Lei.

If we don't have any steps to resolve this, I'm going to go ahead and push to revert.

The tilt-tracking-android tree has some kind of boot race in it at the moment with similar symptoms ~50% of the boots. It doesn't have this patch in.

It also hangs after reporting UART latency stuff.

I can also change the probability of seeing it with loglevel=8 and earlyprintk, so it seems like some kind of race.

3.0 doesn't have any of the CTI stuff in from 4fefbe94a77c3d3b5e75d386628c36298c72c57f so if these issues have a common cause, I am not sure that is it.

Nico has applied the revert. Upstream (3.0) doesn't seem to be affected, so marking this fixed.