Re: [PATCH 1/2] x86/mm/pat: Change pat_disable() to emulate PAT table

From: Borislav Petkov <bp@alien8.de>
To: Toshi Kani <toshi.kani@hpe.com>
Cc: "mingo@kernel.org" <mingo@kernel.org>,
	"hpa@zytor.com" <hpa@zytor.com>,
	"tglx@linutronix.de" <tglx@linutronix.de>,
	"mcgrof@suse.com" <mcgrof@suse.com>,
	"jgross@suse.com" <jgross@suse.com>,
	"paul.gortmaker@windriver.com" <paul.gortmaker@windriver.com>,
	"x86@kernel.org" <x86@kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 1/2] x86/mm/pat: Change pat_disable() to emulate PAT table
Date: Tue, 15 Mar 2016 12:00:05 +0100	[thread overview]
Message-ID: <20160315110005.GB4559@pd.tnic> (raw)
In-Reply-To: <1457991443.6393.290.camel@hpe.com>

On Mon, Mar 14, 2016 at 03:37:23PM -0600, Toshi Kani wrote:
> Your patch is a simplified version of mine.  So, yes, it fixes the Paul's
> issue, but it does not address other issues that my patchset also
> addressed.  In specific, I think your patch has the following issues.

You couldnt've structured your reply better: remember how I split a
convoluted patch of yours already? A patch which was trying to do a
bunch of things in one go.

The situation here is the same. You need to do *one* *logical*
*non-trivial* thing in a patch. If there's something else that needs to
be done, add it in a *separate* patch which explains why that new change
is needed.

> - pat_disable() is now callable from other modules. So, it needs to check
> with boot_cpu_done. We cannot disable PAT once it is initialized.

That should be a separate patch which explains *why* the change is being
done.

> - mtrr_bp_init() needs to check with mtrr_enabled() when it
> calls mtrr_pat_setup_bp(). Otherwise, PAT is left initialized on BSP only
> when MTRR is disabled by its MSR. In your patch, mtrr_bp_init() calls
> pat_setup() again, but it does not help since boot_cpu_done is set.

The code which you carved out from get_mtrr_state() didn't check
mtrr_enabled() before. That needs to be another patch *again* with
explanations.

> - When PAT is disabled in CPU feature, pat_bsp_init() calls pat_disable()
> and returns. However, it does not initialize a PAT table by calling
> pat_init_cache_modes().

Yet another patch.

> - When CONFIG_MTRR is unset, it does not call pat_setup().

Aaaand... can you guess what I'm going to say here?

I hope it is coming across as I intend it: please use my hunk to do a
single fix and then prepare all those changes above in separate patches
with explanations:

"Problem is A. We need to do B. I'm doing it/I'm doing C because."

Ok?

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.