Clarification re: outstanding IPC flaws

Clarification re: outstanding IPC flaws

[Tom Burns]

Hi List,

  Reading current kernel git ipc/sem.c I see the following comment:

--
 * - The previous code had two flaws:
...
 *   2) It did not wake up all zero waiting processes. We try to do
 *      better but only get the semops right which only wait for zero or
 *      increase. If there are decrement operations in the operations
 *      array we do the same as before.

--

I have some questions about this comment.

Most importantly - does this mean that if I have multiple processes
queued inside semop() calls with sem_val = -1 (also, if it matters,
SEM_UNDO flag is set), I can expect to eventually see none of the
queued processes woken up when whoever has the semaphore releases it?

If that's the case, is there any semop best-use example I should be
following for multiple multi-threaded processes sharing multiple
semaphores?

Currently I have some globally reachable semaphores doing semop calls
with SEM_UNDO.  They usually work fine with multiple processes waiting
on each other to lock the semaphore, but under certain circumstances a
semaphore will be released and none of the processes waiting on the
semaphore will be woken up.

Thanks in advance,
Tom Burns
Software Developer

Re: Clarification re: outstanding IPC flaws

[Tom Burns]

Sorry for the double post.

On Nov 26, 2007 12:01 PM, Tom Burns <tom.i.burns@gmail.com> wrote:
> I have some questions about this comment.
>
> Most importantly - does this mean that if I have multiple processes
> queued inside semop() calls with sem_val = -1 (also, if it matters,
> SEM_UNDO flag is set),

This should read "sem_op = -1" as opposed to "sem_val = -1".

Cheers,
Tom Burns