Message-ID: <634aa365-1f51-8684-24ae-3b68aba1e12a@redhat.com>
Date:   Mon, 23 Jan 2023 10:28:23 +0100
From:   David Hildenbrand <david@...hat.com>
To:     Rick Edgecombe <rick.p.edgecombe@...el.com>, x86@...nel.org,
        "H . Peter Anvin" <hpa@...or.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, linux-kernel@...r.kernel.org,
        linux-doc@...r.kernel.org, linux-mm@...ck.org,
        linux-arch@...r.kernel.org, linux-api@...r.kernel.org,
        Arnd Bergmann <arnd@...db.de>,
        Andy Lutomirski <luto@...nel.org>,
        Balbir Singh <bsingharora@...il.com>,
        Borislav Petkov <bp@...en8.de>,
        Cyrill Gorcunov <gorcunov@...il.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Eugene Syromiatnikov <esyr@...hat.com>,
        Florian Weimer <fweimer@...hat.com>,
        "H . J . Lu" <hjl.tools@...il.com>, Jann Horn <jannh@...gle.com>,
        Jonathan Corbet <corbet@....net>,
        Kees Cook <keescook@...omium.org>,
        Mike Kravetz <mike.kravetz@...cle.com>,
        Nadav Amit <nadav.amit@...il.com>,
        Oleg Nesterov <oleg@...hat.com>, Pavel Machek <pavel@....cz>,
        Peter Zijlstra <peterz@...radead.org>,
        Randy Dunlap <rdunlap@...radead.org>,
        Weijiang Yang <weijiang.yang@...el.com>,
        "Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>,
        John Allen <john.allen@....com>, kcc@...gle.com,
        eranian@...gle.com, rppt@...nel.org, jamorris@...ux.microsoft.com,
        dethoma@...rosoft.com, akpm@...ux-foundation.org,
        Andrew.Cooper3@...rix.com, christina.schimpe@...el.com
Cc:     Yu-cheng Yu <yu-cheng.yu@...el.com>
Subject: Re: [PATCH v5 10/39] x86/mm: Introduce _PAGE_COW

On 19.01.23 22:22, Rick Edgecombe wrote:
> Some OSes have a greater dependence on software available bits in PTEs than
> Linux. That left the hardware architects looking for a way to represent a
> new memory type (shadow stack) within the existing bits. They chose to
> repurpose a lightly-used state: Write=0,Dirty=1. So in order to support
> shadow stack memory, Linux should avoid creating memory with this PTE bit
> combination unless it intends for it to be shadow stack.
> 
> The reason it's lightly used is that Dirty=1 is normally set by HW
> _before_ a write. A write with a Write=0 PTE would typically only generate
> a fault, not set Dirty=1. Hardware can (rarely) both set Dirty=1 *and*
> generate the fault, resulting in a Write=0,Dirty=1 PTE. Hardware which
> supports shadow stacks will no longer exhibit this oddity.
> 
> So that leaves Write=0,Dirty=1 PTEs created in software. To achieve this,
> in places where Linux normally creates Write=0,Dirty=1, it can use the
> software-defined _PAGE_COW in place of the hardware _PAGE_DIRTY. In other
> words, whenever Linux needs to create Write=0,Dirty=1, it instead creates
> Write=0,Cow=1 except for shadow stack, which is Write=0,Dirty=1.
> Further differentiated by VMA flags, these PTE bit combinations would be
> set as follows for various types of memory:
> 
> (Write=0,Cow=1,Dirty=0):
>   - A modified, copy-on-write (COW) page. Previously when a typical
>     anonymous writable mapping was made COW via fork(), the kernel would
>     mark it Write=0,Dirty=1. Now it will instead use the Cow bit. This
>     happens in copy_present_pte().
>   - A R/O page that has been COW'ed. The user page is in a R/O VMA,
>     and get_user_pages(FOLL_FORCE) needs a writable copy. The page fault
>     handler creates a copy of the page and sets the new copy's PTE as
>     Write=0 and Cow=1.
>   - A shared shadow stack PTE. When a shadow stack page is being shared
>     among processes (this happens at fork()), its PTE is made Dirty=0, so
>     the next shadow stack access causes a fault, and the page is
>     duplicated and Dirty=1 is set again. This is the COW equivalent for
>     shadow stack pages, even though it's copy-on-access rather than
>     copy-on-write.
> 
> (Write=0,Cow=0,Dirty=1):
>   - A shadow stack PTE.
>   - A Cow PTE created when a processor without shadow stack support set
>     Dirty=1.
> 
> There are six bits left available to software in the 64-bit PTE after
> consuming a bit for _PAGE_COW. No space is consumed in 32-bit kernels
> because shadow stacks are not enabled there.
> 
> Implement only the infrastructure for _PAGE_COW. Changes to start
> creating _PAGE_COW PTEs will follow once other pieces are in place.
> 
> Tested-by: Pengfei Xu <pengfei.xu@...el.com>
> Tested-by: John Allen <john.allen@....com>
> Co-developed-by: Yu-cheng Yu <yu-cheng.yu@...el.com>
> Signed-off-by: Yu-cheng Yu <yu-cheng.yu@...el.com>
> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@...el.com>
> ---
> 
> v5:
>   - Fix log, comments and whitespace (Boris)
>   - Remove capitalization on shadow stack (Boris)
> 
> v4:
>   - Teach pte_flags_need_flush() about _PAGE_COW bit
>   - Break apart patch for better bisectability
> 
> v3:
>   - Add comment around _PAGE_TABLE in response to comment
>     from (Andrew Cooper)
>   - Check for PSE in pmd_shstk (Andrew Cooper)
>   - Get to the point quicker in commit log (Andrew Cooper)
>   - Clarify and reorder commit log for why the PTE bit examples have
>     multiple entries. Apply same changes for comment. (peterz)
>   - Fix comment that implied dirty bit for COW was a specific x86 thing
>     (peterz)
>   - Fix swapping of Write/Dirty (PeterZ)
> 
> v2:
>   - Update commit log with comments (Dave Hansen)
>   - Add comments in code to explain pte modification code better (Dave)
>   - Clarify info on the meaning of various Write,Cow,Dirty combinations
> 
>   arch/x86/include/asm/pgtable.h       | 78 ++++++++++++++++++++++++++++
>   arch/x86/include/asm/pgtable_types.h | 59 +++++++++++++++++++--
>   arch/x86/include/asm/tlbflush.h      |  3 +-
>   3 files changed, 134 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
> index b39f16c0d507..6d2f612c04b5 100644
> --- a/arch/x86/include/asm/pgtable.h
> +++ b/arch/x86/include/asm/pgtable.h
> @@ -301,6 +301,44 @@ static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
>   	return native_make_pte(v & ~clear);
>   }
>   
> +/*
> + * Normally COW memory can result in Dirty=1,Write=0 PTEs. But in the case
> + * of X86_FEATURE_USER_SHSTK, the software COW bit is used, since the
> + * Dirty=1,Write=0 will result in the memory being treated as shadow stack
> + * by the HW. So when creating COW memory, a software bit is used
> + * _PAGE_BIT_COW. The following functions pte_mkcow() and pte_clear_cow()
> + * take a PTE marked conventionally COW (Dirty=1) and transition it to the
> + * shadow stack compatible version of COW (Cow=1).
> + */

TBH, I find that all highly confusing.

Dirty=1,Write=0 does not indicate a COW page reliably. You could have 
both, false negatives and false positives.

False negative: fork() on a clean anon page.

False positives: wrpotect() of a dirty anon page.


I wonder if it really has to be that complicated: what you really want 
to achieve is to disallow "Dirty=1,Write=0" if it's not a shadow stack 
page, correct?

-- 
Thanks,

David / dhildenb

Powered by blists - more mailing lists