Hi,
On Tue, Apr 10, 2018 at 11:34 AM, Karthik Ramasubramanian
<kramasub@xxxxxxxxxxxxxx> wrote:
> Hi Bjorn,
>
> Thanks for pulling it into your tree and helping with the verification.
>
> On 4/9/2018 5:46 PM, Bjorn Andersson wrote:
>> On Mon 09 Apr 12:38 PDT 2018, Karthikeyan Ramasubramanian wrote:
>>
>>> Add early console support in Qualcomm Technologies Inc., GENI based
>>> UART controller.
>>>
>>
>> Hi Karthikeyan,
>>
>> I pulled this into my SDM845 tree and added "earlycon" to my command
>> line, the result is a working console up until the serial driver is
>> probed at which point the console stops.
>>
>>
>> I'm expecting that this should hand over to the normal console as it is
>> probed, unless I specify keep_bootcon in which case I expect all log
>> entries to be printed both through the earlycon and the normal console
>> (i.e. show up twice on the uart).
> I suspect the issue is something to do with clocks. I will look into it
> further and fix in the next submission.
>
> This is my current hypothesis: When the driver gets probed, it registers
> the uart port which in turn invokes the console setup. The console setup
> enables the "se" clock (geni_resources_on()->clk_prepare_enable()).
> Enabling the "se" clock is disturbing the rate that the clock is already
> on. Early console write is simultaneously going on while the console is
> being set up and this clock disturbance is causing the halt.
Or it could be a simple bug in this code. Specifically, your "custom
implementation instead of readl_poll_atomic" is wrong. You don't
ensure that "timeout_us" is evenly divisible by 10, yet you do:
while (timeout_us) {
...;
udelay(10);
timeout_us -= 10;
}
One way to fix this is with this before the loop:
timeout_us = DIV_ROUND_UP(timeout_us, 10) * 10;
BTW: are you planning to spin this patch series anytime soon?
-Doug
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
