On Thu, Aug 24, 2000 at 11:38:02PM +0200, almesber@lrc.di.epfl.ch wrote:
> BTW, this is a little off-topic, I kind of wonder what the right semantics
> for deadlock detection are with threads. POSIX 1003.1 1996-07-12 says:
>
> ... If the system detects that sleeping until a
> locked region is unlocked would cause a deadlock, the fcntl()
> function shall fail with an [EDEADLK] error."
>
> Okay, that's a little vague already. Now let's make it more interesting
> by adding threads:
>
> Process Thread Action (executed in this order)
> A 1 grab lock X
> B 1 grab lock Y
> A 2 try to get lock Y
> B 2 try to get lock X
>
> Solaris 2.5.1 thinks this is a deadlock. Linux 2.4.0-test1-ac16 thinks
> it isn't. I haven't found any definition of what "putting a process to
> sleep" means in POSIX.1.
I don't have POSIX in front of me, but I do not think
the system must detect deadlock. It is allowed to detect
deadlock, and in case it happens to do so EDEADLK is the
apprpriate error return.
Andries
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Aug 31 2000 - 21:00:15 EST