Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 27 Apr 2005 19:59:53 -0700
From:      Kris Kennaway <kris@obsecurity.org>
To:        Bruce Evans <bde@zeta.org.au>
Cc:        cvs-all@FreeBSD.org
Subject:   Re: cvs commit: src/sys/vm vm_init.c
Message-ID:  <20050428025953.GA50766@xor.obsecurity.org>
In-Reply-To: <20050428124206.C85412@delplex.bde.org>
References:  <200504280029.j3S0TOXp078389@repoman.freebsd.org> <20050428124206.C85412@delplex.bde.org>

next in thread | previous in thread | raw e-mail | index | archive | help

--0F1p//8PRICkK4MW
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Thu, Apr 28, 2005 at 12:52:06PM +1000, Bruce Evans wrote:
> On Thu, 28 Apr 2005, Kris Kennaway wrote:
>=20
> >kris        2005-04-28 00:29:24 UTC
> >
> > FreeBSD src repository
> >
> > Modified files:        (Branch: RELENG_5)
> >   sys/vm               vm_init.c
> > Log:
> > MFC:
> >
> >   Add the vm.exec_map_entries tunable and read-only sysctl, which contr=
ols
> >   the number of entries in exec_map (maximum number of simultaneous exe=
cs
> >   that can be handled by the kernel).  The default value of 16 is
> >   insufficient on heavily loaded machines (particularly SMP machines), =
and
> >   if it is exceeded then executing further processes will generate a=20
> >   SIGABRT.
> >
> >   This is a workaround until a better solution can be implemented.
>=20
> It seems to be insufficient on all reasonably fast machines.  Just
> fork-execing 16+ processes in the following benchmark seems to trigger it:
>=20
> %%%
> #!/bin/sh
>=20
> for i in 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
> do
> 	nice -$i sh -c "while :; do echo -n;done" &
> done
> top -o time
> %%%
>=20
> Lately, this has started failing with a strange "Abort trap" messages
> printed by the shell after I exit top.  The 3 processes with niceness
> 1, 2 and 3 tend to get killed more than others.

I didn't change the default value of 16 simultaneous execs to be
conservative and because of the proximity of 5.4, but on my SMP
package machines I typically need to increase it.  Perhaps the default
should be increased - this change was a minimal bandaid to at least
allow the problem to be worked around for now.  Alan has a patch that
addresses 2 of the 3 failure cases, so hopefully he'll be able to fix
the remaining one soon.

Kris

--0F1p//8PRICkK4MW
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (FreeBSD)

iD8DBQFCcFGoWry0BWjoQKURAokpAJ9gJxlh9ac2RideoKS6EeEOzuDYTgCfRGfV
wSMJC3sxNIk8ojAZI8kRGeg=
=WQeD
-----END PGP SIGNATURE-----

--0F1p//8PRICkK4MW--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050428025953.GA50766>