Hi Kevin,

On Wed, Oct 1, 2014 at 10:57 PM, Kevin Hilman [off-list ref] wrote:

On Sun, Sep 21, 2014 at 11:30 PM, Sudhir Sreedharan
[off-list ref] wrote:

quoted

For console devices having UART_CAP_SLEEP capability, the uart_pm_state has
to be initialized to UART_PM_STATE_ON. Otherwise the LCR regiser values
are reinitialized when uart_change_pm is called from uart_configure_port.

Signed-off-by: Sudhir Sreedharan <redacted>

Multiple boot failures on ARM[1] were bisected down to this patch.

How was this patch tested, and on which platforms?

This patch was tested on x86-64(haswell) board, which uses ST16650V2
uart(which has UART_CAP_SLEEP).
While serial driver gets initialized, console port LCR register is
getting reinitalized to 0.
Then boot logs will be seen as garbage characters.

I will re-check why this failed on the boards/archs you mentioned.

Thanks,
Sudhir