[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200430095913.GA3996@zn.tnic>
Date: Thu, 30 Apr 2020 11:59:13 +0200
From: Borislav Petkov <bp@...en8.de>
To: Reinette Chatre <reinette.chatre@...el.com>
Cc: tglx@...utronix.de, fenghua.yu@...el.com, tony.luck@...el.com,
kuo-lang.tseng@...el.com, mingo@...hat.com, babu.moger@....com,
hpa@...or.com, x86@...nel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/2] x86/resctrl: Support CPUID enumeration of MBM
counter width
On Wed, Apr 29, 2020 at 11:42:03AM -0700, Reinette Chatre wrote:
> This would essentially be resubmitting [1] though. Do you expect that
> this change would receive a different reception at this time?
Right, Thomas and I talked it over a bit last night. So the proper
thing to do is to read all that information *once* and put it in
boot_cpu_data. Because that information is replicated the same over
CPUID on each core. If there's per-CPU stuff, then it should remain
per-CPU but looking at how the RDT code uses boot_cpu_data, I'd say this
is global info.
So, it should be parsed once on the BSP during boot and put into
boot_cpu_data. And then silly stuff like x86_init_cache_qos() should go
away too.
If this info is needed on Intel only, then it should be parsed in
cpu/intel.c, in a ->c_bsp_init helper and if it is needed on AMD too,
then a function which does this should be called by the respective
c_bsp_init helper.
Then all its users can continue reading it out of boot_cpu_data and
future RDT hw info can be added there.
Makes sense?
Thx.
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
Powered by blists - more mailing lists