linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Ira Weiny <ira.weiny@intel.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	Andy Lutomirski <luto@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Fenghua Yu <fenghua.yu@intel.com>,
	x86@kernel.org, linux-kernel@vger.kernel.org,
	Andrew Morton <akpm@linux-foundation.org>,
	linux-doc@vger.kernel.org, linux-nvdimm@lists.01.org,
	linux-mm@kvack.org, linux-kselftest@vger.kernel.org,
	Dan Williams <dan.j.williams@intel.com>,
	Greg KH <gregkh@linuxfoundation.org>
Subject: Re: [PATCH V3 04/10] x86/pks: Preserve the PKRS MSR on context switch
Date: Thu, 17 Dec 2020 20:05:13 -0800	[thread overview]
Message-ID: <20201218040513.GB2506510@iweiny-DESK2.sc.intel.com> (raw)
In-Reply-To: <871rfoscz4.fsf@nanos.tec.linutronix.de>

On Thu, Dec 17, 2020 at 03:50:55PM +0100, Thomas Gleixner wrote:
> On Fri, Nov 06 2020 at 15:29, ira weiny wrote:
> > --- a/arch/x86/kernel/process.c
> > +++ b/arch/x86/kernel/process.c
> > @@ -43,6 +43,7 @@
> >  #include <asm/io_bitmap.h>
> >  #include <asm/proto.h>
> >  #include <asm/frame.h>
> > +#include <asm/pkeys_common.h>
> >  
> >  #include "process.h"
> >  
> > @@ -187,6 +188,27 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
> >  	return ret;
> >  }
> >  
> > +#ifdef CONFIG_ARCH_HAS_SUPERVISOR_PKEYS
> > +DECLARE_PER_CPU(u32, pkrs_cache);
> > +static inline void pks_init_task(struct task_struct *tsk)
> 
> First of all. I asked several times now not to glue stuff onto a
> function without a newline inbetween. It's unreadable.

Fixed.

> 
> But what's worse is that the declaration of pkrs_cache which is global
> is in a C file and not in a header. And pkrs_cache is not even used in
> this file. So what?

OK, this was just a complete rebase/refactor mess up on my part.  The
global'ness is not required until we need a global update of the pkrs which was
not part of this series.

I've removed it from this patch.  And cleaned it up in patch 6/10 as well.  And
cleaned it up in the global pkrs patch which you found in my git tree.

> 
> > +{
> > +	/* New tasks get the most restrictive PKRS value */
> > +	tsk->thread.saved_pkrs = INIT_PKRS_VALUE;
> > +}
> > +static inline void pks_sched_in(void)
> 
> Newline between functions. It's fine for stubs, but not for a real implementation.

Again my apologies.

Fixed.

> 
> > diff --git a/arch/x86/mm/pkeys.c b/arch/x86/mm/pkeys.c
> > index d1dfe743e79f..76a62419c446 100644
> > --- a/arch/x86/mm/pkeys.c
> > +++ b/arch/x86/mm/pkeys.c
> > @@ -231,3 +231,34 @@ u32 update_pkey_val(u32 pk_reg, int pkey, unsigned int flags)
> >  
> >  	return pk_reg;
> >  }
> > +
> > +DEFINE_PER_CPU(u32, pkrs_cache);
> 
> Again, why is this global?

In this patch it does not need to be.  I've changed it to static.

> 
> > +void write_pkrs(u32 new_pkrs)
> > +{
> > +	u32 *pkrs;
> > +
> > +	if (!static_cpu_has(X86_FEATURE_PKS))
> > +		return;
> > +
> > +	pkrs = get_cpu_ptr(&pkrs_cache);
> 
> So this is called from various places including schedule and also from
> the low level entry/exit code. Why do we need to have an extra
> preempt_disable/enable() there via get/put_cpu_ptr()?
> 
> Just because performance in those code paths does not matter?

Honestly I don't recall the full history at this point.  The
preempt_disable/enable() is required when this is called from
pks_update_protection()  AKA when a user is trying to update the protections of
their key.  What I do remember is that this was originally not preempt safe and we
had a comment to that effect in the early patches.[1]

Somewhere along the line the preempt discussion lead us to make write_pkrs()
'self contained' with the preemption protection here.  I just did not think
about any performance issues.  It is safe to call preempt_disable() from a
preempt disabled region, correct?  I seem to recall asking that and the answer
was 'yes'.

I will audit the calls again and adjust the preemption disable as needed.

[1] https://lore.kernel.org/lkml/20200717072056.73134-5-ira.weiny@intel.com/#t

> 
> > +	if (*pkrs != new_pkrs) {
> > +		*pkrs = new_pkrs;
> > +		wrmsrl(MSR_IA32_PKRS, new_pkrs);
> > +	}
> > +	put_cpu_ptr(pkrs);
> 
> Now back to the context switch:
> 
> > @@ -644,6 +668,8 @@ void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p)
> >
> >	 if ((tifp ^ tifn) & _TIF_SLD)
> >		 switch_to_sld(tifn);
> > +
> > +	pks_sched_in();
> >  }
> 
> How is this supposed to work? 
> 
> switch_to() {
>    ....
>    switch_to_extra() {
>       ....
>       if (unlikely(next_tif & _TIF_WORK_CTXSW_NEXT ||
> 	           prev_tif & _TIF_WORK_CTXSW_PREV))
> 	   __switch_to_xtra(prev, next);
> 
> I.e. __switch_to_xtra() is only invoked when the above condition is
> true, which is not guaranteed at all.

I did not know that.  I completely missunderstood what __switch_to_xtra()
meant.  I thought it was arch specific 'extra' stuff so it seemed reasonable to
me.

Also, our test seemed to work.  I'm still investigating what may be wrong.

> 
> While I have to admit that I dropped the ball on the update for the
> entry patch, I'm not too sorry about it anymore when looking at this.
> 
> Are you still sure that this is ready for merging?

Nope...

Thanks for the review,
Ira

> 
> Thanks,
> 
>         tglx

  parent reply	other threads:[~2020-12-18  4:06 UTC|newest]

Thread overview: 45+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-11-06 23:28 [PATCH V3 00/10] PKS: Add Protection Keys Supervisor (PKS) support V3 ira.weiny
2020-11-06 23:28 ` [PATCH V3 01/10] x86/pkeys: Create pkeys_common.h ira.weiny
2020-11-06 23:29 ` [PATCH V3 02/10] x86/fpu: Refactor arch_set_user_pkey_access() for PKS support ira.weiny
2020-11-06 23:29 ` [PATCH V3 03/10] x86/pks: Add PKS defines and Kconfig options ira.weiny
2020-11-06 23:29 ` [PATCH V3 04/10] x86/pks: Preserve the PKRS MSR on context switch ira.weiny
2020-12-17 14:50   ` Thomas Gleixner
2020-12-17 22:43     ` Thomas Gleixner
2020-12-18 13:57       ` Thomas Gleixner
2020-12-18 19:20         ` Dan Williams
2020-12-18 21:06           ` Thomas Gleixner
2020-12-18 21:58             ` Dan Williams
2020-12-18 22:44               ` Thomas Gleixner
2020-12-18 19:42         ` Ira Weiny
2020-12-18 20:10           ` Dave Hansen
2020-12-18 21:30           ` Thomas Gleixner
2020-12-18  4:05     ` Ira Weiny [this message]
2020-12-17 20:41   ` [NEEDS-REVIEW] " Dave Hansen
2020-12-18  4:10     ` Ira Weiny
2020-12-18 15:33       ` Dave Hansen
2020-11-06 23:29 ` [PATCH V3 05/10] x86/entry: Pass irqentry_state_t by reference ira.weiny
2020-11-15 18:58   ` Thomas Gleixner
2020-11-16 18:49     ` Ira Weiny
2020-11-16 20:36       ` Thomas Gleixner
2020-11-24  6:09   ` [PATCH V3.1] entry: " ira.weiny
2020-12-11 22:14     ` Andy Lutomirski
2020-12-16  1:32       ` Ira Weiny
2020-12-16  2:09         ` Andy Lutomirski
2020-12-17  0:38           ` Ira Weiny
2020-12-17 13:07       ` Thomas Gleixner
2020-12-17 13:19         ` Peter Zijlstra
2020-12-17 15:35           ` Andy Lutomirski
2020-12-17 16:58     ` Thomas Gleixner
2020-11-06 23:29 ` [PATCH V3 06/10] x86/entry: Preserve PKRS MSR across exceptions ira.weiny
2020-12-17 15:28   ` Thomas Gleixner
2020-11-06 23:29 ` [PATCH V3 07/10] x86/fault: Report the PKRS state on fault ira.weiny
2020-11-06 23:29 ` [PATCH V3 08/10] x86/pks: Add PKS kernel API ira.weiny
2020-12-23 20:39   ` Randy Dunlap
2020-11-06 23:29 ` [PATCH V3 09/10] x86/pks: Enable Protection Keys Supervisor (PKS) ira.weiny
2020-11-06 23:29 ` [PATCH V3 10/10] x86/pks: Add PKS test code ira.weiny
2020-12-17 20:55   ` Dave Hansen
2020-12-18  4:05     ` Ira Weiny
2020-12-18 16:59       ` Dan Williams
2020-12-07 22:14 ` [PATCH V3 00/10] PKS: Add Protection Keys Supervisor (PKS) support V3 Ira Weiny
2020-12-08 15:55   ` Thomas Gleixner
2020-12-08 17:22     ` Ira Weiny

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201218040513.GB2506510@iweiny-DESK2.sc.intel.com \
    --to=ira.weiny@intel.com \
    --cc=akpm@linux-foundation.org \
    --cc=bp@alien8.de \
    --cc=dan.j.williams@intel.com \
    --cc=dave.hansen@linux.intel.com \
    --cc=fenghua.yu@intel.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux-nvdimm@lists.01.org \
    --cc=luto@kernel.org \
    --cc=mingo@redhat.com \
    --cc=peterz@infradead.org \
    --cc=tglx@linutronix.de \
    --cc=x86@kernel.org \
    --subject='Re: [PATCH V3 04/10] x86/pks: Preserve the PKRS MSR on context switch' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).