RE: [PATCH v2 1/2] mfd: dt: tps6586x: Add power off control

nvpublic

On 08/24/2012 06:36 PM, Bill Huang wrote:

quoted

quoted

quoted

On Sun, Aug 19, 2012 at 06:07:55PM -0700, Bill Huang wrote:

quoted

Add DT property "ti,system-power-controller" telling whether or not
this pmic is in charge of controlling the system power, so the
power off routine can be hooked up to system call "pm_power_off".

...

quoted

quoted

quoted

I've seen the following while trying this patch applied on top of next-20120817:

[   40.581151] Power down.
[   41.583160] ------------[ cut here ]------------
[   41.587784] WARNING: at /home/thierry.reding/src/kernel/linux-ipmp.git/drivers/i2c/busses/i2c-
tegra.c:525 tegra_i2c_xfer+0x21c/0x29c()

...

quoted

quoted

Thanks Thierry, I can repro this on Tegra20 inconsistently and found,
if current cpu is not cpu0 when doing "machine_shutdown" (it will
call "smp_send_stop"), i2c controller will failed to do any transaction (looks like gic interrupt

will be disabled), I'll debug further to find out the root cause.

quoted

quoted

By the way, Tegra30 is good since it will always be cpu0 when doing
"machine_shutdown", I still don't know why it makes the difference
against Tegra20 since I'm not familiar with those cpu stuffs and what make it behave differently,

I'll study a bit, thanks.

quoted

I've sent the shutdown issue for discussion in ARM list: Shutdown problem in SMP system happened on

Tegra20.

quoted

The cause of the i2c timeout is pretty clear now and it is not
directly related to this patch, so is this patch series acceptable? Any thoughts or comment? Thanks.

I tend to agree; power off never worked without this patch, and sometimes does with the patch, due to
nothing wrong with this patch.

Bill, please do follow up on getting the underlying Tegra issue solved somehow though. IIRC, Joseph Lo
or Prashant has a patch which enabled the config option that Russell mentioned, so the fix may just be
to wait for that patch to get finalized, but please double-check that solves it.
Thanks!

As per the shutdown issue discussion, enabling CONFIG_PM_SLEEP_SMP is the only solution and I've confirmed that fix the issue, thanks.