[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20180822205707.GA16015@u40b0340c692b58f6553c.ant.amazon.com>
Date: Wed, 22 Aug 2018 13:57:07 -0700
From: Eduardo Valentin <eduval@...zon.com>
To: Andi Kleen <ak@...ux.intel.com>
CC: Eduardo Valentin <eduval@...zon.com>,
Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...hat.com>,
"H. Peter Anvin" <hpa@...or.com>, <x86@...nel.org>,
Peter Zijlstra <peterz@...radead.org>,
"Kan Liang" <kan.liang@...ux.intel.com>,
Dan Carpenter <dan.carpenter@...cle.com>,
Jia Zhang <qianyue.zj@...baba-inc.com>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/1] perf/x86/intel: make error messages less confusing
Hey,
On Tue, Aug 21, 2018 at 04:59:08PM -0700, Andi Kleen wrote:
> On Tue, Aug 21, 2018 at 04:05:22PM -0700, Eduardo Valentin wrote:
> > On Tue, Aug 21, 2018 at 03:09:37PM -0700, Andi Kleen wrote:
> > > On Tue, Aug 21, 2018 at 02:15:28PM -0700, Eduardo Valentin wrote:
> > > > On a system with X86_FEATURE_ARCH_PERFMON disabled
> > > > and with a model not known by family PMU drivers,
> > > > user gets a kernel message log like the following:
> > > > [ 0.100114] Performance Events: unsupported p6 CPU model 85 no PMU driver, software events only.
> > > >
> > > > The "unsupported .. CPU" part may be confusing for some
> > > > users. Rewording the messages on the failure path to:
> > > > [ 0.667154] Performance Events: unknown p6 PMU on CPU model 85: !X86_FEATURE_ARCH_PERFMON: no PMU driver, software events only.
> > >
> > > Are you sure users even know what ARCH_PERFMON is?
> > >
> > > Maybe it is confusing (why exactly?), but it doesn't seem to me that your
> > > new message is any better.
> >
> > Yeah, the part that says "unsupported CPU" is the confusing part,
>
> That makes sense.
>
> > I get people thinking that the specific reported CPU model is not
> > supported by the kernel :-)
> >
> > >
> > > If you refer to VMs not exposing the PMU perhaps that should be explicitely mentioned.
> > >
> > > Of course the real fix is to always expose the PMU, not improve the error messages...
> >
> > I agree that best is simply to enable PMU. But it does not hurt to improve the error messaging, does it?
> >
> > Any suggestions there, given that the initial attempt seams to make it even worse :-)
>
> Perhaps just say
>
> "CPU does not support PMU"
>
Ok.
> which is really what the problem is here.
>
Fair enough.
> The other option would be to move this message after the big model switch,
> but would need to be very careful that it doesn't have any unintended
> side effects.
I will simply remove those messages with CPU model details and on
the failure path build the message to look like:
[ 0.666785] Performance Events: CPU does not support PMU: no PMU driver, software events only.
>
> -Andi
>
--
All the best,
Eduardo Valentin
Powered by blists - more mailing lists