Message-ID: <1432312047.1428.34.camel@misato.fc.hp.com>
Subject: Re: [PATCH v9 9/10] x86, mm, pat: Refactor !pat_enabled handling
From: Toshi Kani <toshi.kani@hp.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: hpa@zytor.com, mingo@redhat.com, akpm@linux-foundation.org, arnd@arndb.de,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org, x86@kernel.org,
        linux-nvdimm@ml01.01.org, jgross@suse.com, stefan.bader@canonical.com,
        luto@amacapital.net, hmh@hmh.eng.br, yigal@plexistor.com,
        konrad.wilk@oracle.com, Elliott@hp.com, mcgrof@suse.com, hch@lst.de
Date: Fri, 22 May 2015 10:27:27 -0600
In-Reply-To: <alpine.DEB.2.11.1505220958050.5457@nanos>
References: <1431551151-19124-1-git-send-email-toshi.kani@hp.com>
	 <1431551151-19124-10-git-send-email-toshi.kani@hp.com>
	 <alpine.DEB.2.11.1505220958050.5457@nanos>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3055
Lines: 126

On Fri, 2015-05-22 at 10:34 +0200, Thomas Gleixner wrote:
> On Wed, 13 May 2015, Toshi Kani wrote:
> 
> > --- a/arch/x86/mm/pat.c
> > +++ b/arch/x86/mm/pat.c
> > @@ -182,7 +182,11 @@ void pat_init_cache_modes(void)
> >  	char pat_msg[33];
> >  	u64 pat;
> >  
> > -	rdmsrl(MSR_IA32_CR_PAT, pat);
> > +	if (pat_enabled)
> > +		rdmsrl(MSR_IA32_CR_PAT, pat);
> > +	else
> > +		pat = boot_pat_state;
> 
> boot_pat_state is 0 if pat is disabled, but this boot_pat_state multi
> purpose usage is really horrible. We do 5 things at once with it and
> of course all of it completely undocumented.

boot_pat_state is set even if pat is disabled so that this case can be
handled in the same framework.

	:
  if (!pat_enabled) {
	/*
	 * No PAT. Emulate the PAT table that corresponds to the two
	:
	pat = PAT(0, WB) | PAT(1, WT) | PAT(2, UC_MINUS) | PAT(3, UC) |
	      PAT(4, WB) | PAT(5, WT) | PAT(6, UC_MINUS) | PAT(7, UC);
	if (!boot_pat_state)
		boot_pat_state = pat;
	:

That said, yes, I agree that the use of boot_pat_state is overloaded.

>   	pat_msg[32] = 0;
> >  	for (i = 7; i >= 0; i--) {
> >  		cache = pat_get_cache_mode((pat >> (i * 8)) & 7,
> > @@ -200,28 +204,58 @@ void pat_init(void)
> >  	bool boot_cpu = !boot_pat_state;
> 
> The crap starts here and this really wants to be distangled.

Agreed.

> void pat_init(void)
> {
> 	static bool boot_done;
> 
> 	if (!boot_done) {
> 	   	if (!cpu_has_pat)
>   			pat_disable("PAT not supported by CPU.");
> 
> 		if (pat_enabled) {
> 		   	rdmsrl(MSR_IA32_CR_PAT, boot_pat_state);
> 			if (!boot_pat_state)
> 				pat_disable("PAT read returns always zero, disabled.");
> 		}
> 	} else if (!cpu_has_pat && pat_enabled) {
> 		/*
> 		 * If this happens we are on a secondary CPU, but
> 		 * switched to PAT on the boot CPU. We have no way to
> 		 * undo PAT.
> 		 */
> 		pr_err("PAT enabled but not supported by secondary CPU\n");
> 		BUG();
> 	}
> 
> 	
> 	if (!pat_enabled) {
> 	   .....
> 	} else {
> 	   .....	
> 	}
> 
> 	if (!boot_done) {
> 	    ....
> 	    boot_done = true;	
> 	}
> }
> 
> And this cleanup wants to be done as a seperate patch before you do
> this other stuff.

Yes, this looks much better!  Will add a patch for this clean up.

> > @@ -275,16 +309,8 @@ void pat_init(void)
> >  		      PAT(4, WB) | PAT(5, WC) | PAT(6, UC_MINUS) | PAT(7, WT);
> >  	}
> >  
> > -	/* Boot CPU check */
> > -	if (!boot_pat_state) {
> > -		rdmsrl(MSR_IA32_CR_PAT, boot_pat_state);
> > -		if (!boot_pat_state) {
> > -			pat_disable("PAT read returns always zero, disabled.");
> > -			return;
> > -		}
> > -	}
> > -
> > -	wrmsrl(MSR_IA32_CR_PAT, pat);
> > +	if (pat_enabled)
> > +		wrmsrl(MSR_IA32_CR_PAT, pat);
> 
> Sigh.

Yeah...

> 
> 	if (!pat_enabled) {
> 	   ....
> 	} else {
> 	   ....
> 	}
> 	
> +	if (pat_enabled)
> 
> Thanks,

Thanks a lot!
-Toshi

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/