Re: [PATCH 14/15] gatchat: added command completion by timeout

[Denis Kenzior <denkenz@gmail.com>]

Hi Максим,

On 10/27/23 04:45, Любимов Максим wrote:
> On 18/10/2023 в 21:21 -0500, Denis Kenzior wrote:
>> So I think you're going to need to give a bit more context for how
>> this is
>> expected to work.  There are commands (on older modems) that can take
>> minutes to
>> complete.  This is why multiplexers were invented.  Now this is
>> probably quite
>> rare, but I imagine it can still happen.
> 
> This functionality was added because some modems sometimes do not
> respond to the command. Therefore, there is an endless wait for a
> response and blocking of other functions with a message that the action

Ok, but which modems?  Do you have an OFONO_AT_DEBUG log trace you can share?

Anyhow, at the very least you need to make this an opt-in feature.  As I pointed 
out, opting in *all* modem drivers into this functionality will break some of them.

> is still in progress. It was experimentally verified that if the
> command is repeated, the modem can immediately respond to it and there
> will be no blocking. What do you suggest to do if the modem does not
> respond to the command either after a minute, or after an hour, never?

Call the vendor and have them fix their firmware? :)  But seriously, how many 
commands are we talking about?  If it is one, then handle it specifically for 
that command.  If it is a dozen, create a utility function.

I'm not convinced adding this functionality directly to GAtChat is the right 
approach.  There is no such thing as 'AT command timeout' in V.250 specification 
or any other standard.  AT command cancellation as a concept exists, but it only 
works sometimes and only for some commands, and isn't what you're doing here 
from what I can tell.

What you're describing is vendor specific behavior and has to be treated as such.

> 
>> Also, what happens if the command times out?  You generate a
>> synthetic CME 100
>> error, but what then?
> 
> The function that sent the command must take action based on its
> operating logic, for example, repeat the command, or fail.
> 

Okay, so where is this code?

Regards,
-Denis