Re: Is _PAGE_PROTNONE set only for user mappings?

From: Hyeonggon Yoo <42.hyeyoo@gmail.com>
To: Dave Hansen <dave.hansen@intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>,
	Tom Lendacky <thomas.lendacky@amd.com>,
	Rick Edgecombe <rick.p.edgecombe@intel.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Andy Lutomirski <luto@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
	"Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
	Tianyu Lan <Tianyu.Lan@microsoft.com>,
	"Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org, vbabka@suse.cz,
	akpm@linux-foundation.org, willy@infradead.org
Subject: Re: Is _PAGE_PROTNONE set only for user mappings?
Date: Sun, 29 May 2022 10:32:37 +0000	[thread overview]
Message-ID: <YpNLxXvGFQpyPjRa@n2.us-central1-a.c.spheric-algebra-350919.internal> (raw)
In-Reply-To: <3f2f7c09-ddf3-6052-9860-8554a4ff2798@intel.com>

On Mon, May 16, 2022 at 07:04:32AM -0700, Dave Hansen wrote:
> On 5/12/22 22:33, Hyeonggon Yoo wrote:
> > Thanks Mel, and IIUC nor does do_kern_addr_fault() in arch/x86/mm/fault.c
> > expect _PAGE_PROTNONE instead of _PAGE_GLOBAL. I want to make it clear
> > in the code that _PAGE_PROTNONE is only used for user mappings.
> > 
> > How does below look?
> > 
> > diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
> > index 40497a9020c6..f8d02b91a90c 100644
> > --- a/arch/x86/include/asm/pgtable_types.h
> > +++ b/arch/x86/include/asm/pgtable_types.h
> > @@ -35,7 +35,10 @@
> >  #define _PAGE_BIT_DEVMAP	_PAGE_BIT_SOFTW4
> >  
> >  /* If _PAGE_BIT_PRESENT is clear, we use these: */
> > -/* - if the user mapped it with PROT_NONE; pte_present gives true */
> > +/*
> > + * if the user mapped it with PROT_NONE; pte_present gives true
> > + * this is only used for user mappings (with _PAGE_BIT_USER)
> > + */
> >  #define _PAGE_BIT_PROTNONE	_PAGE_BIT_GLOBAL
> >  
> >  #define _PAGE_PRESENT	(_AT(pteval_t, 1) << _PAGE_BIT_PRESENT)
> > @@ -115,7 +118,8 @@
> >  #define _PAGE_DEVMAP	(_AT(pteval_t, 0))
> >  #endif
> >  
> > -#define _PAGE_PROTNONE	(_AT(pteval_t, 1) << _PAGE_BIT_PROTNONE)
> > +#define _PAGE_PROTNONE	((_AT(pteval_t, 1) << _PAGE_BIT_USER) | \
> > +			 (_AT(pteval_t, 1) << _PAGE_BIT_PROTNONE))
> >  
> >  /*
> >   * Set of bits not changed in pte_modify.  The pte's
> 
> I don't like the idea of _PAGE_BIT_USER being so implicit.  It is
> something kernel users should know explicitly that they are messing with.
> 
> I was thinking of something more along the lines of taking the
> set_memory.c code and ensuring that it never sets (or even observes)
> _PAGE_BIT_GLOBAL on a _PAGE_USER mapping.  There was also a question of
> if set_memory.c is ever used on userspace mappings.  It would be good to
> validate whether it's possible in-tree today and if not, enforce that
> _PAGE_USER PTEs should never even be observed with set_memory.c.

Writing code I'm a bit confused:
commit d1440b23c922d8 ("x86/mm: Factor out pageattr
_PAGE_GLOBAL setting") says:

	"This unconditional setting of _PAGE_GLOBAL is a problem when we have
	PTI and non-PTI and we want some areas to have _PAGE_GLOBAL and some
	not." 

Is this this sentence not valid anymore in PTI,
and just unconditionally setting _PAGE_GLOBAL would be okay in kernel
side regardless of PTI?

I'm wondering it because previously I thought "Let's not clear
_PAGE_GLOBAL in set_memory.c for kernel mappings and make pmd/pte_present() not
confuse when _PAGE_USER is not set"

But you don't like it as it's a bit implicit.
Then I wonder - how do we know when to set _PAGE_GLOBAL again?

> The arch/x86/mm/dump_pagetables.c code is also a reasonable place to put
> assumptions about the page tables since it walks *everything* when asked.