"Can't obtain network semaphore"

[Till Straumann]

Ian Caddy wrote:
> Till Straumann wrote:
> 
>> Ian Caddy wrote:
>>
>>> What would happen if the network semaphore object id was corrupted.  
>>> It wouldn't even get to the section of code to get the mutex, but 
>>> would return on invalid ID error when it tried to get the object.
>>>
>>> Steve did not mention what the actual return code from the call to 
>>> the semaphore get, does it correspond to the internal return code of 4?
>>
>>
>>
>> Yes, Steve did report earlier a return code of 4
>>
> 
> If the actual return code from the actual call:
> 
>     sc = rtems_semaphore_obtain

You're off track. The ID isn't used but the networking code
uses the core mutex directly. Read libnetworking/rtems/rtems_glue.c
(RTEMS_FAST_MUTEX defined)

T.

> 
> was 4 then (at least on 4.5.0) indicates a:
> 
>   RTEMS_INVALID_ID               =  4, /* invalid object id */
> 
> In this case it would not go into any of the mutex code.
> 
> 
> 
>>> This could have been returned from another earlier function call.
>>
>>
>>
>> No. _CORE_mutex_Seize() sets this to 0 on entry (in 'trylock').
>>
>> T.
>>
> 
> But if the ID is invalid, it will not enter the seize code, therefore 
> the Wait.return_code will not get initialised or even changed from what 
> ever it was set to before the call, in other words from the last call 
> that modified it.
> 
> The actual return code from the function is different to the 
> Wait.return_code as it is translated at the end of semobtain.c:
> 
>         return _Semaphore_Translate_core_mutex_return_code(
>                   _Thread_Executing->Wait.return_code );
> 
> 
> The acual rtems_status_code of 4 is not possible through this mechanism.
> 
> Ian Caddy
>