Re: modutils (Re: cipe 1.2.0 & kernel 2.2.x)

Malcolm Beattie (mbeattie@sable.ox.ac.uk)
Thu, 25 Mar 1999 16:50:00 +0000 (GMT)

Y2K writes:
> On Thu, 25 Mar 1999, Malcolm Beattie wrote:
> > I wonder if it's worth having a closefdset() syscall used as follows:
> > fd_set most;
> > FD_ZERO(&most);
> > for (i=getdtablesize()-1; i>2; --i)
> > FD_SET(i, &most);
> > closefdset(&most);
[...]
> maybe a better solution would be a getopenfdset(fdset *set);
>
> so in bad pseudo c:
> long n;
> int bad=0;
> fdset all;
> getopenfdset(&all);
> FD_CLR(0,&all); FD_CLR(1,&all); FD_CLR(2,&all);
> while ( notempty(&all)&&(!bad) ) bad=close( firstfd(&all) );

Excellent suggestion. Moreover, we now no longer a new system call.
In pseudo C mixed with pseudo Perl

DIR *dir = opendir("/proc/$$/fd");
while (d = readdir(dir)) {
close(atoi(d->d_name)); /* and handle error */
}

Hurray, no new sys call needed. It scales nicely and may even be
faster on systems with only 256 descriptors since it replaces 253
close() syscalls with 1 open plus 1 getdents + n(open_files) closes.
One of the original messages said perl did the "lots of closes" thing
but it doesn't. It just sets close-on-exec on all descriptors above
the last "system" descriptor, changeable by the special $^F variable
or $SYSTEM_FD_MAX is you're "use English"-inclined. It's mainly things
like login, daemons and security-related code which close everything.
Getting them all to change, assuming I haven't missed anything and
assuming that the benefit outweighs the loss of portability (which it
might not), is going to be non-trivial though. But this is all now
off-topic for linux-kernel so I'll stop there.

--Malcolm

-- 
Malcolm Beattie <mbeattie@sable.ox.ac.uk>
Unix Systems Programmer
Oxford University Computing Services


-
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/