Re: Idle Loop

Hi Adam,

On 7/26/24 12:03 PM, Adam Pigg wrote:
> Hi Dennis
> 
> Wondered if I could pick your brain about a new issue I have.
> 
> In the patches I apply ontop of stock Ofono for Sailfish, there is a plugin
> called the Slot Manager.
> 
> On 2.7 all is fine.
> 
> On 2.9 not so.
> 
> The code in the slot manager, after its creation waits for 5 iterations of the
> event loop before disallowing any more drivers to register, and it seems that
> in 2.9, this is much more likely to happen, though its a bit inconsistent.

Why would it be doing that?

> 
> I wondered if it could be related to https://git.kernel.org/pub/scm/network/
> ofono/ofono.git/commit/?id=063856a2ba9308e1776a3f906bd2ed9a9f760744 and what a
> correct solution might look like (I understand this is my issue as the code
> isnt upstream!)

This commit makes drivers register on probe such that no main loop iterations 
are needed between probe() and register().

I assume we are talking about atom drivers here (everything in drivers/*).  Atom 
drivers when probed (probe() is invoked) start their initialization process. 
This process can take as much time as these drivers need.  Putting a cap of the 
number of iterations doesn't make any sense.

> 
> Potentially the check for 5 iterations of the loop isnt even necessary (and
> seems very arbitrary!) so Ill test without it.
> 
> Is there a reason that the gobi plugin/qmimodem driver combination now takes
> an indeterminate amount of event loop time to appear?

See above.  gobi / qmimodem now perform more steps after probe() and before 
register().

> 
> Any thoughts welcome,
> Kind Regards
> 
> Adam
> 
> 

Regards,
-Denis