Re: [Rstplib-users] Help: RSTP semaphores (again)

[Alex R. <ar...@mr...>]

On Tuesday, May 15, 2007 6:07 PM Pan Jieke asked:
> ... what's the utility of those semaphores, such as, =
STP_OUT_psos_init_semaphore,
> STP_OUT_psos_close_semaphore, etc.? I realized that they were just =
void functions
> for doing nothing, may be I misunderstood the idea. Can anyone tell me =
in which
> situation it should be used and how?=20

Hello,
RSTPLib is a library, that has to be linked into some process. The =
architecture of this process depends on system environment and/or your =
choice. If this process is not a "single thread" one, you must protect =
shared resources (data base) from simultaneous usage by different =
threads (race condition). Semaphores, mutexes, etc.  are dedicated to =
this purpose.
I tried to make RSTPLib system independent as far as possible. So the =
code contains calls to macros RSTP_CRITICAL_PATH_START and =
RSTP_CRITICAL_PATH_END in relevant places. The definitions of these =
macros, as I noted above, have to be system dependent. While RSTP =
simulator runs under Linux and is a single thread process, you may see, =
that in this case these macros are empty. In another environment you =
must decide - do you want to have non-empty semaphores or not? In our =
real bridges, running under pSOS, we must use semaphores. The functions =
STP_OUT_psos_*** serve to this purposes and are part of the environment. =
Our real routers run under Linux, and RSTP daemon (the same way as MSTP =
daemon) are single thread process.

Best regards, Alex