* Felipe Balbi [off-list ref] [141003 07:53]:

On Thu, Oct 02, 2014 at 04:59:30PM -0500, Felipe Balbi wrote:

quoted

On Thu, Oct 02, 2014 at 02:19:08PM -0700, Tony Lindgren wrote:

quoted

* Felipe Balbi [off-list ref] [141002 13:18]:

quoted

On Thu, Oct 02, 2014 at 12:52:38PM -0700, Tony Lindgren wrote:

quoted

Hmm here seems to be a link to similar issues from 2011:

http://e2e.ti.com/support/arm/sitara_arm/f/791/p/113593/628790.aspx

Looks like the issue can be potentially reproduced with:

# cyclictest -l100000000 -m -a0 -t1 -n -p99 -i200 -h200 -q

running here on am335x and am437x. On that same post, on person
mentions he reproduced on beagle bone.

OK I'll run it here too on my am37xx evm. Looks like Stanley was
running both cyclictest and hackbench the same time.

yeah I did that.

BTW, just got the following on BBB, AM437x SK is still running.

[ 3952.432262] Kernel panic - not syncing: Attempted to kill the idle task!752763
[ 3952.442403] CPU: 0 PID: 0 Comm: hackbench Not tainted 3.17.0-rc6-00456-gd8da063 #222
[ 3952.450517] [<c0017338>] (unwind_backtrace) from [<c0012fdc>] (show_stack+0x20/0x24)
[ 3952.458620] [<c0012fdc>] (show_stack) from [<c0647f0c>] (dump_stack+0x8c/0xa4)
[ 3952.466168] [<c0647f0c>] (dump_stack) from [<c0644da4>] (panic+0xa4/0x224)
[ 3952.473358] [<c0644da4>] (panic) from [<c004b510>] (do_exit+0x924/0x9d8)
[ 3952.480358] [<c004b510>] (do_exit) from [<c004c4d8>] (do_group_exit+0x50/0xc0)
[ 3952.487903] [<c004c4d8>] (do_group_exit) from [<c004c568>] (__wake_up_parent+0x0/0x30)
[ 3952.496179] [<c004c568>] (__wake_up_parent) from [<c000ed40>] (ret_fast_syscall+0x0/0x48)
[ 3952.504875] drm_kms_helper: panic occurred, switching back to text console
[ 3952.517844] ---[ end Kernel panic - not syncing: Attempted to kill the idle task!

quoted

And I'll also queue the following patch during the -rc cycle to
avoid apps segfaulting occasionally at random on omap3.

and maybe this will fix BBB :-) I'll add that locally. If I survive
until tomorrow, I'll add a Tested-by.

BBB died again with the same behavior as above, but I think it's
unrelated to this errata. Therefore:

Tested-by: Felipe Balbi <redacted>

BTW, I have revision r3p2:

CPU: ARMv7 Processor [413fc082] revision 2 (ARMv7), cr=10c5387d

And it still seems to need 430973.

Looks like my 37xx evm produced no errors overnight running cyclictest
and hackbench the same time.

Regards,

Tony