RE: [PATCH 1/2] perf: Add closing sibling events' file descriptors

From: David Laight
Date: Tue Aug 11 2020 - 17:06:30 EST


> On Tue, Aug 11, 2020 at 02:47:03PM +0000, David Laight wrote:
> > From: Andi Kleen
> > > On Mon, Aug 10, 2020 at 10:36:32PM +0200, Peter Zijlstra wrote:
> > > > On Mon, Aug 10, 2020 at 07:45:18AM -0700, Andi Kleen wrote:
> > > >
> > > > > Unfortunately we're kind of stuck with the old NFILE=1024 default
> > > > > even though it makes little sense on modern servers.
> > > >
> > > > Why can't that be changed? It seems to me all of userspace changes all
> > > > the time; heck that system-doofus thing flushed 20+ years of sysadmin
> > > > experience down the drain, just cause. Why can't we up a file limit?
> > >
> > > We could try, but I believe it's hard coded in various places outside
> > > the kernel.
> >
> > The place it really bites is select().
> > Although the kernel supports large bitmaps glibc doesn't.
> > The random bit overwrites are a PITA to debug.
>
> Good point.
>
> I remember I asked for a simple define to increase like glibc5 had, but
> they still didn't implement that.

Even windows lets you redefine FDSET_SIZE before including the header.
And their select() is much more like poll() - so the limit is the number
of sockets, not the highest socket number (which do happen to be small
integers).

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)