Date: Thu, 5 Mar 2009 10:04:36 -0500 From: John Baldwin <jhb@freebsd.org> To: Kip Macy <kmacy@freebsd.org> Cc: src-committers@freebsd.org, svn-src-user@freebsd.org Subject: Re: svn commit: r189374 - user/kmacy/HEAD_fast_net_merge/sys/netinet Message-ID: <200903051004.36626.jhb@freebsd.org> In-Reply-To: <3c1674c90903041520o145b0c42uc1dd2d41870432e6@mail.gmail.com> References: <200903042122.n24LMdO3053650@svn.freebsd.org> <200903041759.31535.jhb@freebsd.org> <3c1674c90903041520o145b0c42uc1dd2d41870432e6@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday 04 March 2009 6:20:55 pm Kip Macy wrote:
> Do we have any mechanism for handling sparse cpuids?
> If so I can do modulo the number of active cpus and then taking that
> value as an index. If not its a bit of an academic observation.
>=20
> -Kip
The rest of the kernel is careful to handle sparse cpuids already. You can=
=20
use CPU_ABSENT() to see if a CPU is present or not. You cound pre-populate=
a=20
table during boot if you wished by doing something like:
int active_cpus[MAXCPU];
sysinit_foo()
{
int i, cpu;
i =3D 0;
for (cpu =3D 0; cpu <=3D mp_maxid; cpu++) {
if (CPU_ABSENT(cpu)
continue;
active_cpus[i] =3D cpu;
i++;
}
KASSERT(i =3D=3D mp_ncpu, ("bad juju"));
}
Then you could use (active_cpus[flowid & mp_ncpu]) as a CPU. Eventually wh=
en=20
we support removing and adding CPUs at runtime (or disabling/enabling them =
at=20
runtime) there will be eventhandlers that you can hook to adjust your=20
active_cpus array.
=20
> On Wed, Mar 4, 2009 at 2:59 PM, John Baldwin <jhb@freebsd.org> wrote:
> > On Wednesday 04 March 2009 4:22:39 pm Kip Macy wrote:
> >> Author: kmacy
> >> Date: Wed Mar =A04 21:22:39 2009
> >> New Revision: 189374
> >> URL: http://svn.freebsd.org/changeset/base/189374
> >>
> >> Log:
> >> =A0 Use per-cpu callouts for tcp_timer
> >> =A0 186694, 187660
> >>
> >> =A0 186694:
> >> =A0 - convert tcp_timer_activate over to using
> >> =A0 =A0 per-cpu callouts
> >> =A0 - don't acquire the tcbinfo lock exclusively
> >> =A0 =A0 in tcp_timer_rexmt unless needed for tcp_drop
> >>
> >> =A0 187660:
> >> =A0 - mp_maxid may not be valid ensure that we
> >> =A0 =A0 re-schedule on cpuid less than or equal to
> >> =A0 =A0 the current one for tcp callouts
> >>
> >> Modified:
> >> =A0 user/kmacy/HEAD_fast_net_merge/sys/netinet/tcp_timer.c
> >>
> >> Modified: user/kmacy/HEAD_fast_net_merge/sys/netinet/tcp_timer.c
> >>
> >=20
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D
> >> --- user/kmacy/HEAD_fast_net_merge/sys/netinet/tcp_timer.c =A0 =A0Wed =
Mar =A04
> > 21:04:52 2009 =A0 (r189373)
> >> +++ user/kmacy/HEAD_fast_net_merge/sys/netinet/tcp_timer.c =A0 =A0Wed =
Mar =A04
> > 21:22:39 2009 =A0 (r189374)
> >> @@ -41,6 +41,7 @@ __FBSDID("$FreeBSD$");
> >> =A0#include <sys/mbuf.h>
> >> =A0#include <sys/mutex.h>
> >> =A0#include <sys/protosw.h>
> >> +#include <sys/smp.h>
> >> =A0#include <sys/socket.h>
> >> =A0#include <sys/socketvar.h>
> >> =A0#include <sys/sysctl.h>
> >> @@ -118,6 +119,8 @@ int =A0 =A0 =A0 tcp_maxpersistidle;
> >> =A0 =A0 =A0 /* max idle time in persist */
> >> =A0int =A0tcp_maxidle;
> >>
> >> +#define =A0 =A0 =A0INP_CPU(inp) =A0 =A0min(curcpu, ((inp)->inp_flowid=
% mp_maxid))
> >
> > This is not really safe. =A0CPU ID's may be sparse. =A0The only guarant=
ees you
> > have are that 0 is the boot CPU, and that all valid CPU IDs are in the=
=20
range
> > [0 .. mp_maxid] (inclusive). =A0Thus, you could have a system that only=
has
> > CPUs 0 and 3 and if you are on CPU 3 and flowid is 5, then this will=20
choose
> > min(3, 5 % 3) =3D=3D min(3, 2) =3D=3D 2 which is an invalid CPU.
> >
> > --
> > John Baldwin
> >
>=20
=2D-=20
John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200903051004.36626.jhb>