On Thu, 2002-08-08 at 13:56, Hubertus Franke wrote:
> (a) Vanilla version which breaks down after 22K pids and
> really sucks above 32.5K
> (b) Bill Irwin's, which keeps track of which PID is free and
> which one is not ?
> (c) Andrea's patch which searches in the bitmap when we are
> running out
> (d) Paul's patch, which I believe was based on one of my earlier
> submission (03/02) that essentially switches between (a)+(c) at
> the break even point.
> I feel (c) or (d) is a better solution over (a) and (b)... Open for
> discussion.
> I have a test program that does the random pid deletion and pid allocation
> in user space. All what's required is to copy the get_pid() code from the
> kernel into there... Can make that available ...
>
> I don't know what Paul has done to the patch since then ....
It's pretty much the same, with a couple of small changes. After
forward porting your patch and the one from AA's patches (hch says that
one was actually done by Ihno Krumreich, thanks for the info), I added
the fix from (c) for the race, and moved the check for if(flags &
CLONE_IDLETASK) outside of get_pid into copy_process() since there's no
need to call get_pid and return just to find out that we need to use 0.
-Paul Larson
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Aug 15 2002 - 22:00:17 EST