state terminology

state terminology

[Jan Engelhardt]

Hi,


If this example kernel functions is called, it will generate an oops, kill 
the current process, and subsequent tries to call it will hang the process 
trying to do so.

        static struct semaphore x;
        void f(void) {
            int *p = NULL;
            down(&x);
            *p++;
            up(&x);
        }

Is there a specific term (other than "hang") associated with this 
situation? It's not a "dead-lock", because there is no other process 
(anymore) which could potentially up the semaphore.


Jan Engelhardt
-- 
| Alphagate Systems, http://alphagate.hopto.org/
| jengelh's site, http://jengelh.hopto.org/

Re: state terminology

[Ingo Oeser]

[-- Attachment #1: Type: text/plain, Size: 970 bytes --]

Hi Jan,

On Saturday 14 January 2006 22:34, you wrote:
> Is there a specific term (other than "hang") associated with this 
> situation? It's not a "dead-lock", because there is no other process 
> (anymore) which could potentially up the semaphore.

This is a simple "resource leak" (or "semaphore leak" in this case).

Explanation follows:

The resource semaphore is not usable by anyone anymore 
and is still around.

Its pretty much the same as a memory leak. There is no one, who
could free the memory anymore.

The reasons for the resource not being usable anymore is
not significant for a resource leak.

Also insignificant is the fact that the amount of semaphores
are just limited by available memory. If you repeat starting threads 
doing the semaphore leak game from your example, you'll run out
of memory and thus out of semaphores. This is another sign of leakage.

Do the above explanations sound ok?


Regards

Ingo Oeser


[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: state terminology

[Paul Dickson]

On Sun, 15 Jan 2006 10:58:48 +0100, Ingo Oeser wrote:

> Hi Jan,
> 
> On Saturday 14 January 2006 22:34, you wrote:
> > Is there a specific term (other than "hang") associated with this 
> > situation? It's not a "dead-lock", because there is no other process 
> > (anymore) which could potentially up the semaphore.
> 
> This is a simple "resource leak" (or "semaphore leak" in this case).
> 
> Explanation follows:
> 
> The resource semaphore is not usable by anyone anymore 
> and is still around.
> 
> Its pretty much the same as a memory leak. There is no one, who
> could free the memory anymore.
> 
> The reasons for the resource not being usable anymore is
> not significant for a resource leak.
> 
> Also insignificant is the fact that the amount of semaphores
> are just limited by available memory. If you repeat starting threads 
> doing the semaphore leak game from your example, you'll run out
> of memory and thus out of semaphores. This is another sign of leakage.
> 
> Do the above explanations sound ok?

But it's the functionality rather than the resource that's being lost.
So I wouldn't consider it to be a leak.

How about "locked-out" or "lock-out"?  It's akin to a locked room, with
the keys left inside.

	-Paul

Re: state terminology

[Jan Engelhardt]

>> > Is there a specific term (other than "hang") associated with this 
>> > situation? It's not a "dead-lock", because there is no other process 
>> > (anymore) which could potentially up the semaphore.
>> 
>> This is a simple "resource leak" (or "semaphore leak" in this case).
>> 
>How about "locked-out" or "lock-out"?  It's akin to a locked room, with
>the keys left inside.

Yeah, comes closer. I also thought about cul-de-sac. (Which does not allow to
proceed further, but let's some waiting process abort.)


Jan Engelhardt
--