Date: Wed, 21 Sep 2022 22:15:21 -0700 From: Kevin Oberman <rkoberman@gmail.com> To: Paul Procacci <pprocacci@gmail.com> Cc: "freebsd-questions@freebsd.org" <questions@freebsd.org> Subject: Re: Disabling CPUs Message-ID: <CAN6yY1sqwFQ1EkaV5PBNU7uOMuZoi90=FBvLjVaOuXpmwM00ow@mail.gmail.com> In-Reply-To: <CAFbbPuijZ9YmRkS=0JdHNNj34Zv8Qoh78nbnP2JbHCyotD-fiA@mail.gmail.com> References: <CAN6yY1s=ds_=yWsEWtVfoXchD5XXqU5uPw6bdFNW%2BdaU6t491A@mail.gmail.com> <CAFbbPuijZ9YmRkS=0JdHNNj34Zv8Qoh78nbnP2JbHCyotD-fiA@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--0000000000003f06b105e93d27a7
Content-Type: text/plain; charset="UTF-8"
On Wed, Sep 21, 2022 at 7:26 PM Paul Procacci <pprocacci@gmail.com> wrote:
> (Resending with list included)
>
> Hey Keith,
>
> Taken from smp(4):
>
> FreeBSD allows specific CPUs on a multi-processor system to be
> disabled.
> This can be done using the hint.lapic.X.disabled tunable, where X is
> the
> APIC ID of a CPU. Setting this tunable to 1 will result in the
> corresponding CPU being disabled.
>
> smp(4) further describes how to detect CPU topologies and whatnot.
> Specifically the following sysctl seems useful: kern.sched.topology_spec
>
> Though I'm not entirely sure, I'd personally start with the above.
>
> ~Paul
>
> On Wed, Sep 21, 2022 at 10:06 PM Kevin Oberman <rkoberman@gmail.com>
> wrote:
>
>> I am looking to disable all 8 E-cores on my Alder Lake system to prevent
>> repeated crashes. The man page has an example of this:
>> Modify the cpuset all threads are in by default to contain only the first
>> 4 CPUs, leaving the rest idle:
>> cpuset -l 0-3 -s 1
>> I did this, but in subsequent port build, all 12 "CPUs" were running at
>> 100%. Am I missing something? Maybe use -p 1" instead of "-s 1".
>>
>> I also found suggestions to use "hint.lapic.N.disabled", but teh lines
>> that were supposed to be in dmesg and the messages log were not present. I
>> am baffled, but really need to do something to stop the crashes currently
>> impacting Alder Lake systems.
>> --
>> Kevin Oberman, Part time kid herder and retired Network Engineer
>> E-mail: rkoberman@gmail.com
>> PGP Fingerprint: D03FB98AFA78E3B78C1694B318AB39EF1B055683
>>
>
>
> --
> __________________
>
> :(){ :|:& };:
>
Nothing looks useful in kern.sched. I am wondering if SCHED_4BSD is
impacting this. My knowledge of the schedulers is minimal, but 4BSD
provided much better responsiveness than ULE.I 've been running 4BSD on my
systems (excluding servers) for over a decade... Probably much longer. I
may try building a kernel with ULE and see what effect that has.
# sysctl kern.sched
kern.sched.runq_fuzz: 1
kern.sched.ipiwakeup.useloop: 0
kern.sched.ipiwakeup.usemask: 1
kern.sched.ipiwakeup.delivered: 26954287
kern.sched.ipiwakeup.requested: 26899799
kern.sched.ipiwakeup.enabled: 1
kern.sched.slice: 12
kern.sched.quantum: 94488
kern.sched.name: 4BSD
kern.sched.preemption: 1
kern.sched.cpusetsize: 32
-
Kevin Oberman, Part time kid herder and retired Network Engineer
E-mail: rkoberman@gmail.com
PGP Fingerprint: D03FB98AFA78E3B78C1694B318AB39EF1B055683
--0000000000003f06b105e93d27a7
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
<div dir=3D"ltr"><div dir=3D"ltr"><div class=3D"gmail_default" style=3D"fon=
t-family:tahoma,sans-serif;font-size:small">On Wed, Sep 21, 2022 at 7:26 PM=
Paul Procacci <<a href=3D"mailto:pprocacci@gmail.com">pprocacci@gmail.c=
om</a>> wrote:<br></div></div><div class=3D"gmail_quote"><blockquote cla=
ss=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid =
rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">
<div>(Resending with list included)<br><br>Hey Keith,</div><div><br></div><=
div>Taken from smp(4):<br><br></div><div>=C2=A0 =C2=A0=C2=A0 FreeBSD allows=
specific CPUs on a multi-processor system to be disabled.<br>=C2=A0 =C2=A0=
=C2=A0This can be done using the hint.lapic.X.disabled tunable, where X is=
the<br>=C2=A0 =C2=A0 =C2=A0APIC ID of a CPU.=C2=A0 Setting this tunable to=
1 will result in the<br>=C2=A0 =C2=A0 =C2=A0corresponding CPU being disabl=
ed.<br><br></div><div>smp(4) further describes how to detect CPU topologies=
and whatnot.<br></div><div>Specifically the following sysctl seems useful:=
kern.sched.topology_spec<br></div><div><br></div><div>Though I'm not e=
ntirely sure, I'd personally start with the above.<font color=3D"#88888=
8"><br><br></font></div><font color=3D"#888888"><div>~Paul</div></font>
</div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">=
On Wed, Sep 21, 2022 at 10:06 PM Kevin Oberman <<a href=3D"mailto:rkober=
man@gmail.com" target=3D"_blank">rkoberman@gmail.com</a>> wrote:<br></di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div s=
tyle=3D"font-family:tahoma,sans-serif;font-size:small">I am looking to disa=
ble all 8 E-cores on my Alder Lake system to prevent repeated crashes. The =
man page has an example of this:</div><div style=3D"font-family:tahoma,sans=
-serif;font-size:small">Modify the cpuset all threads are in by default to =
contain only the first 4 CPUs, leaving the rest idle:<br>=C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0cpuset -l 0-3 -s 1</div><div style=3D"font-family:t=
ahoma,sans-serif;font-size:small">I did this, but in subsequent port build,=
all 12 "CPUs" were running at 100%. Am I missing something? Mayb=
e use -p 1" instead of "-s 1".</div><div style=3D"font-famil=
y:tahoma,sans-serif;font-size:small"><br></div><div style=3D"font-family:ta=
homa,sans-serif;font-size:small">I also found suggestions to use "hint=
.lapic.N.disabled", but teh lines that were supposed to be in dmesg an=
d the messages log were not present. I am baffled, but really need to do so=
mething to stop the crashes currently impacting Alder Lake systems.<br></di=
v>-- <br><div dir=3D"ltr"><div dir=3D"ltr"><div><div dir=3D"ltr"><div><div =
dir=3D"ltr"><div><div dir=3D"ltr">Kevin Oberman, Part time kid herder and r=
etired Network Engineer<br>E-mail: <a href=3D"mailto:rkoberman@gmail.com" t=
arget=3D"_blank">rkoberman@gmail.com</a><br></div><div>PGP Fingerprint: D03=
FB98AFA78E3B78C1694B318AB39EF1B055683</div></div></div></div></div></div></=
div></div></div>
</blockquote></div><br clear=3D"all"><br>-- <br><div dir=3D"ltr">__________=
________<br><br>:(){ :|:& };:</div>
</blockquote></div><br clear=3D"all"><br><div><br></div><div><span class=3D=
"gmail_default" style=3D"font-family:tahoma,sans-serif;font-size:small">Not=
hing looks useful in kern.sched. I am wondering if SCHED_4BSD is impacting =
this. My knowledge of the schedulers is minimal, but 4BSD provided much bet=
ter responsiveness than ULE.I 've been running 4BSD on my systems (excl=
uding servers) for over a decade... Probably much longer. I may try buildin=
g a kernel with ULE and see what effect that has.<br></span></div><div><spa=
n class=3D"gmail_default" style=3D"font-family:tahoma,sans-serif;font-size:=
small"><br></span></div><div><span class=3D"gmail_default" style=3D"font-fa=
mily:tahoma,sans-serif;font-size:small"></span># sysctl kern.sched</div>ker=
n.sched.runq_fuzz: 1<br>kern.sched.ipiwakeup.useloop: 0<br>kern.sched.ipiwa=
keup.usemask: 1<br>kern.sched.ipiwakeup.delivered: 26954287<br>kern.sched.i=
piwakeup.requested: 26899799<br>kern.sched.ipiwakeup.enabled: 1<br>kern.sch=
ed.slice: 12<br>kern.sched.quantum: 94488<br><a href=3D"http://kern.sched.n=
ame">kern.sched.name</a>: 4BSD<br>kern.sched.preemption: 1<br>kern.sched.cp=
usetsize: 32<br>- <br><div dir=3D"ltr" class=3D"gmail_signature"><div dir=
=3D"ltr"><div><div dir=3D"ltr"><div><div dir=3D"ltr"><div><div dir=3D"ltr">=
Kevin Oberman, Part time kid herder and retired Network Engineer<br>E-mail:=
<a href=3D"mailto:rkoberman@gmail.com" target=3D"_blank">rkoberman@gmail.c=
om</a><br></div><div>PGP Fingerprint: D03FB98AFA78E3B78C1694B318AB39EF1B055=
683</div></div></div></div></div></div></div></div></div>
--0000000000003f06b105e93d27a7--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAN6yY1sqwFQ1EkaV5PBNU7uOMuZoi90=FBvLjVaOuXpmwM00ow>