[meta-freescale] linux-imx/linux-wandboard 3.10.17 kernel and HDMI display

Mon Mar 24 09:44:12 PDT 2014

Hello guys,

I can confirm this problem, I've seen it for ages on many kernels.

Right now the standard TechNexion supplied kernel (
http://git.wandboard.org/git/kernel_imx.git/log/?h=jb4.3_1.1.0-ga__wandboard)
has somewhat same problems:
1) When
setting videomode=display0=dev=lcd0,mode=CUSTOM_DISPLAY,if=RGB24,bpp=24
display1=dev=hdmi0,mode=1280x720MR-16 at 60,if=RGB24,bpp=32 or similar - board
locks up with RCU stall/hard lockup when HDMI is connected.
2) When setting videomode=display0=dev=hdmi0,mode=1280x720 at 60,if=RGB24,bpp=32
board works only when HDMI cable is connected. If HDMI cable is
disconnected - system locks up.

Everything locks up in here, 1 of 10 tries I had a luck getting backtraces:
[   62.169747] INFO: rcu_sched_state detected stall on CPU 0 (t=6000
jiffies)
[   62.191716] INFO: rcu_sched_state detected stalls on CPUs/tasks: { 0}
(detected by 2, t=6002 jiffies)
< ... >
[   64.162699] CPU1: stopping
[   64.165420] [<8004e160>] (unwind_backtrace+0x0/0xdc) from [<800422c8>]
(do_IPI+0xd4/0x19c)
[   64.173695] [<800422c8>] (do_IPI+0xd4/0x19c) from [<8004718c>]
(__irq_svc+0x4c/0xe4)
[   64.181441] Exception stack(0xe206ff98 to 0xe206ffe0)
[   64.186495] ff80:
00000000 00000001
[   64.194680] ffa0: f40dc008 00000000 80606b90 e206e028 e206e000 8085ed3c
1000406a 412fc09a
[   64.202864] ffc0: 00000000 00000000 00000001 e206ffe0 800536d8 80048238
80000013 ffffffff
[   64.211050] [<8004718c>] (__irq_svc+0x4c/0xe4) from [<80048238>]
(default_idle+0x24/0x28)
[   64.219235] [<80048238>] (default_idle+0x24/0x28) from [<800484bc>]
(cpu_idle+0xb4/0x118)
[   64.227419] [<800484bc>] (cpu_idle+0xb4/0x118) from [<105f9c94>]
(0x105f9c94)
[   64.234557] CPU0: stopping
[   64.237275] [<8004e160>] (unwind_backtrace+0x0/0xdc) from [<800422c8>]
(do_IPI+0xd4/0x19c)
[   64.245550] [<800422c8>] (do_IPI+0xd4/0x19c) from [<8004718c>]
(__irq_svc+0x4c/0xe4)
[   64.253295] Exception stack(0xe24c5d28 to 0xe24c5d70)
[   64.258352] 5d20:                   00000000 00000000 00000000 808a0860
00000202 80817cc0
[   64.266535] 5d40: 80040ae8 e24c4030 0000000a 7f7fbfbc 808a0860 00000000
00000000 e24c5d70
[   64.274717] 5d60: 80080590 80080020 20000113 ffffffff
[   64.279780] [<8004718c>] (__irq_svc+0x4c/0xe4) from [<80080020>]
(__do_softirq+0x70/0x15c)
[   64.288054] [<80080020>] (__do_softirq+0x70/0x15c) from [<80080590>]
(irq_exit+0x4c/0xa4)
[   64.296240] [<80080590>] (irq_exit+0x4c/0xa4) from [<800421d0>]
(do_local_timer+0x5c/0x80)
[   64.304514] [<800421d0>] (do_local_timer+0x5c/0x80) from [<8004718c>]
(__irq_svc+0x4c/0xe4)
[   64.312868] Exception stack(0xe24c5dd0 to 0xe24c5e18)
[   64.317923] 5dc0:                                     e2013800 e223c3c8
e24c5e18 00000000
[   64.326107] 5de0: e2013800 e223c3c8 00000000 e24c5f28 e2017738 00000038
8028075c 00000000
[   64.334289] 5e00: 00000000 e24c5e18 8027f4b8 8007aab0 60000013 ffffffff
[   64.340912] [<8004718c>] (__irq_svc+0x4c/0xe4) from [<8007aab0>]
(console_lock+0x20/0x60)
[   64.349099] [<8007aab0>] (console_lock+0x20/0x60) from [<8027f4b8>]
(mxc_hdmi_notify_fb+0x3c/0x74)
[   64.358068] [<8027f4b8>] (mxc_hdmi_notify_fb+0x3c/0x74) from
[<80280748>] (mxc_hdmi_set_mode+0xf4/0x108)
[   64.367556] [<80280748>] (mxc_hdmi_set_mode+0xf4/0x108) from
[<80280944>] (hotplug_worker+0x1e8/0x340)
[   64.376870] [<80280944>] (hotplug_worker+0x1e8/0x340) from [<8008e6d4>]
(process_one_work+0x1f8/0x318)
[   64.386187] [<8008e6d4>] (process_one_work+0x1f8/0x318) from
[<8008ee00>] (worker_thread+0x1b0/0x2d0)
[   64.395416] [<8008ee00>] (worker_thread+0x1b0/0x2d0) from [<80092df8>]
(kthread+0x7c/0x84)
[   64.403690] [<80092df8>] (kthread+0x7c/0x84) from [<80048154>]
(kernel_thread_exit+0x0/0x8)

I've sent this to TechNexion (Tapani, Edward and other guys) 4 days ago,
but it looks like they seem to not care. I hope they are just busy and will
notice this at some point.

I still did not find time to test 3.10.y on wandboard, but I guess it will
have same problems, Jacob, is this what you see running JTAG trace?

Also, 3.0.35 kernels from freescale had this check enabled:
http://git.freescale.com/git/cgit.cgi/imx/linux-2.6-imx.git/tree/arch/arm/mach-mx6/board-mx6q_sabrelite.c?h=imx_3.0.35_4.1.0#n883
 Does 3.10.y have this? HDMI core is different on different tapeouts and
I'm not sure if wandboard has all boards with same CPU revision, especially
Solo/DL.

Anyway, I'm ready to test anything on my HDMI setup that is known to lockup
the board.

Best Regards,
Dmitriy Beykun

2014-03-24 20:16 GMT+04:00 John Weber <rjohnweber at gmail.com>:

> Hi Jacob,
>
>
> On 3/24/14, 10:10 AM, Otavio Salvador wrote:
>
>> Hello Kroon,
>>
>> On Mon, Mar 17, 2014 at 6:06 AM, Jacob Kroon <jacob.kroon at gmail.com>
>> wrote:
>>
>>> Hi,
>>>
>>> Just some test reporting, I finally got the 3.10.17 linux-imx kernel from
>>> meta-fsl-arm working with my Sony TV, but I had to revert commit
>>>
>>> ce224e3c4fc3c1b00a94641c28724be2adcd78be
>>> "ENGR00288972 HDMI: unmask overflow interrupter when HDMI phy power on"
>>>
>>> otherwise the kernel would lockup once I plugged in the HDMI-cable.
>>> This was on a Wandboard Solo. linux-wandboard 3.10.17 kernel from
>>> meta-fsl-arm-extra showed the same problem, I didn't test if reverting
>>> the
>>> same commit would solve the problem but I'm guessing so.
>>>
>> While there are some issues with the 3.10.17 kernel for Wandboard with
> HDMI, I'm not aware of this one.  Then again, I don't have a Sony HDMI
> monitor.  I've been working with an Acer 22" monitor with HDMI input.  I
> suspect that most of the problems I've seen are related to the lack of
> working EDID on Wandboard.
>
>  Did you find out why this was  need?
>>
>> I am adding John and Trevor in Cc so they can comment on this.
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/meta-freescale/attachments/20140324/be1c9d46/attachment-0001.html>