To:     linux-kernel@vger.rutgers.edu
From:   torvalds@transmeta.com (Linus Torvalds)
Subject: Re: Linux 2.3.39 has 32bit uid. What about 32bit pid?
Date:   16 Jan 2000 11:09:27 -0800

In article <UTC200001152131.WAA07579.aeb@arend.cwi.nl>,
 <Andries.Brouwer@cwi.nl> wrote:
>
>    No, its worse sadly -- think fcntl(fd,F_SETOWN,&pid) sort of thing.
>    We will need a new F_SETOWN or esle the SIGIO stuff will break.
>
>But where is the problem?
>Let me repeat: pid_t has *always* been 32-bit.
>In libc5. In libc4. In Linux 0.01.

Indeed.

In fact, I think that in Linux 0.01 it actually _used_ all bits.

I remember how I was chasing a strange bug in bash, which turned out to
be because _bash_ used "short" somewhere to store a pid.  And this was
long long before Linux development and me became arrogant enough to say
"oh, bash is broken, tell Chet to fix it", so what I ended up doing was
to just make sure that yes, "pid_t" was still 32-bit, but we only ever
selected pids in the 15-bit range. 

We could start using 32-bit pid's any day, but for (a) /proc and (b) I'm
not sure what makes sense in a cluster. Do we want to have the high bits
be just high bits, or do we want them to be cluster machine ID, or do we
want them to be thread-ID related?

So the dynamic range of pids stays at 15 bits until it's clear what the
right thing is. The type stays at 32 bits..

		Linus


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/