On 18 April 2015 at 02:39, Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@linaro.org> wrote:
On 04/17/2015 10:01 AM, Daniel Thompson wrote:
> On 17/04/15 12:25, Jorge Ramirez-Ortiz wrote:
>> On 04/17/2015 04:37 AM, Daniel Thompson wrote:
>>> On 17/04/15 02:58, Jorge Ramirez-Ortiz wrote:
>>>> On 04/16/2015 08:58 PM, Scott Bambrough wrote:
>>>>> BT works, I get this on 258, 260 -- hadn't seen it before.
>>>> it means that BT is taken longer to initialize - ie: user space (UIM) is
>>>> taking
>>>> longer to run measured from the time the btwilink.ko is modprobed by udev.
>>> Forgive my interfering but I'm not sure all this comes from userspace
>>> (although the cause could be in userspace).
>> please everyone more than welcome to interfere.
>>> I had a look at the ti driver code earlier in the week and first four messages
>>> looked to me like a failure to get the TI device into firmware loading mode.
>> yes - well kind of.
>> when the driver is loaded, st_kim_start waits for the line discipline to be
>> installed.
>> [0]https://github.com/96boards/linux/blob/hikey/drivers/misc/ti-st/st_kim.c#L461
>> the completion is waiting for is usermode opening the tty[1] and the completion
>> being executed (from st_tty_open) [2]
>> [1] https://github.com/96boards/uim/blob/master/uim.c#L289
>> [2]
>> https://github.com/96boards/linux/blob/hikey/drivers/misc/ti-st/st_core.c#L702
>> once opened, the line discipline is installed and the firmarware version present
>> in the device is read at dhe default settings (115kbps)
>> The driver reads "TIInit_11.8.32.bts" - meaning the that device is accessible
>> and healthy.
>> After this has happened, usermode then takes control (UIM) and changes the baud
>> rate command in the firmware to the value presented by the device tree to the
>> user via sysfs.
>> https://github.com/96boards/uim/blob/master/uim.c#L356
>> this is done by sending 0xFF36 opcode (see here):
>> https://github.com/96boards/uim/blob/master/uim.c#L356
>> and I think this is pretty much what is happening.
> Thanks for the detailed reply.
> As you probably guessed I could see the tiny details but not the bigger
> picture. Your e-mail is very clear.

I had to review the BT stack a few months ago so I kind of remembered it a little.
btw I am not saying there is not a problem somewhere: just that I dont think it
is BT related and that the messages look OK.

However the fact that UIM is running late is probably not a good sign (since the
UIM hasnt changed in a while it must be something else)

User space starts up much slower, that's one regression in build #258. And it has been fixed in build #260, see for details.

if the problem continues to happen I can spend some time tracing it.

Dev mailing list