I’m working on developing TDMA on GR using RedHat Enterprise Linux WS
2.6.
I intend to use semaphores or mutex for controlling access to shared
memory
when events like packet arrivals and timeouts occur.
However, I understand based on the man pages that a number of the calls
for
semaphores and mutex may cause a deadlock when the calling thread
happens to
be handling an asynchronous signal.
First of all, what are asynchronous signals in this context? If they
refer
to events whose time occurance is unpredictable, wouldn’t such signals
apply
to a really large number of cases, including packet arrivals?
The man pages say that sem_post is the only POSIX call that handles
asynchronous signals fine. What use is this if I can’t use sem_wait with
asynchronous signals?
What is your experience with semaphores and mutex?
Jeremy
On Thu, Oct 26, 2006 at 10:43:08AM +0800, Jeremy Chew wrote:
I’m working on developing TDMA on GR using RedHat Enterprise Linux WS 2.6.
I intend to use semaphores or mutex for controlling access to shared memory
when events like packet arrivals and timeouts occur.
OK
However, I understand based on the man pages that a number of the calls for
semaphores and mutex may cause a deadlock when the calling thread happens
to be handling an asynchronous signal.
OK
First of all, what are asynchronous signals in this context?
They mean “signals” as defined by the kernel system calls
signal(2), kill(2), sigaction(2), etc. Try “man sigaction”
If they refer to events whose time occurance is unpredictable,
wouldn’t such signals apply to a really large number of cases,
including packet arrivals?
No.
The man pages say that sem_post is the only POSIX call that handles
asynchronous signals fine. What use is this if I can’t use sem_wait with
asynchronous signals?
What is your experience with semaphores and mutex?
More years than I like to think about 
You may want to spend some time with a good book on Unix or Linux
programming. Sorry, I don’t have a specific recommendation.
Eric