[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170109221754.GA47453@linux.intel.com>
Date: Mon, 9 Jan 2017 14:17:55 -0800
From: Fenghua Yu <fenghua.yu@...el.com>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: Fenghua Yu <fenghua.yu@...el.com>, Shaohua Li <shli@...com>,
LKML <linux-kernel@...r.kernel.org>,
Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH] x86/intel_rdt: reinitialize cbm for new group allocation
On Mon, Jan 09, 2017 at 11:03:59PM +0100, Thomas Gleixner wrote:
> On Mon, 9 Jan 2017, Fenghua Yu wrote:
> > On Fri, Jan 06, 2017 at 04:05:19PM -0800, Shaohua Li wrote:
> > But since you come here now, I would think reseting the CBM in
> > closid_free() is better.
>
> No. closid_free() is the wrong place. closid_alloc/free() merily deal with
> the bitmap and nothing else.
>
> If you really want to do that, then this needs a seperate function called
> from rmdir.
>
> > The reason is user can see "right" max_cbm even through rdmsr after
> > rmdir, ie no gap for cbm values between rmdir and the next mkdir.
>
> And the value pf this is?
Not very useful value, but user may get a bit confused by a non max_cbm if
a closid is not used after rdmsr. A couple of users actually asked this.
Reseting to initial max_cbm may reduce questions from users.
>
> The closid is not usable after the rmdir, so it's really completely
> uninteresting when the user can read the old value from that configuration
> MSR.
Agree with you. The cbm value is uninteresting after rmdir.
>
> When the closid is reused then it hardly matters either what's in those cbm
> values (the old or max_cbm).
>
> Thanks,
>
> tglx
Powered by blists - more mailing lists