Received-SPF: pass (google.com: domain of linux-kernel+bounces-187680-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Precedence: bulk
MIME-Version: 1.0
References: <20240403234054.2020347-1-debug@rivosinc.com> <20240403234054.2020347-14-debug@rivosinc.com>
 <276fa17b-cd62-433d-b0ec-fa98c65a46ca@ghiti.fr> <ZkJOs6ENmDHFsq/U@debug.ba.rivosinc.com>
In-Reply-To: <ZkJOs6ENmDHFsq/U@debug.ba.rivosinc.com>
From: Alexandre Ghiti <alexghiti@rivosinc.com>
Date: Thu, 23 May 2024 16:59:30 +0200
Message-ID: <CAHVXubhS3CJ87DxC+9+8z6CiWDV1bQ8nK+iOZUDvMiT7vszFLA@mail.gmail.com>
Subject: Re: [PATCH v3 13/29] riscv mmu: write protect and shadow stack
To: Deepak Gupta <debug@rivosinc.com>
Cc: Alexandre Ghiti <alex@ghiti.fr>, paul.walmsley@sifive.com, rick.p.edgecombe@intel.com, 
	broonie@kernel.org, Szabolcs.Nagy@arm.com, kito.cheng@sifive.com, 
	keescook@chromium.org, ajones@ventanamicro.com, conor.dooley@microchip.com, 
	cleger@rivosinc.com, atishp@atishpatra.org, bjorn@rivosinc.com, 
	samuel.holland@sifive.com, conor@kernel.org, linux-doc@vger.kernel.org, 
	linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org, 
	devicetree@vger.kernel.org, linux-mm@kvack.org, linux-arch@vger.kernel.org, 
	linux-kselftest@vger.kernel.org, corbet@lwn.net, palmer@dabbelt.com, 
	aou@eecs.berkeley.edu, robh+dt@kernel.org, krzysztof.kozlowski+dt@linaro.org, 
	oleg@redhat.com, akpm@linux-foundation.org, arnd@arndb.de, 
	ebiederm@xmission.com, Liam.Howlett@oracle.com, vbabka@suse.cz, 
	lstoakes@gmail.com, shuah@kernel.org, brauner@kernel.org, 
	andy.chiu@sifive.com, jerry.shih@sifive.com, hankuan.chen@sifive.com, 
	greentime.hu@sifive.com, evan@rivosinc.com, xiao.w.wang@intel.com, 
	charlie@rivosinc.com, apatel@ventanamicro.com, mchitale@ventanamicro.com, 
	dbarboza@ventanamicro.com, sameo@rivosinc.com, shikemeng@huaweicloud.com, 
	willy@infradead.org, vincent.chen@sifive.com, guoren@kernel.org, 
	samitolvanen@google.com, songshuaishuai@tinylab.org, gerg@kernel.org, 
	heiko@sntech.de, bhe@redhat.com, jeeheng.sia@starfivetech.com, 
	cyy@cyyself.name, maskray@google.com, ancientmodern4@gmail.com, 
	mathis.salmen@matsal.de, cuiyunhui@bytedance.com, bgray@linux.ibm.com, 
	mpe@ellerman.id.au, baruch@tkos.co.il, alx@kernel.org, david@redhat.com, 
	catalin.marinas@arm.com, revest@chromium.org, josh@joshtriplett.org, 
	shr@devkernel.io, deller@gmx.de, omosnace@redhat.com, ojeda@kernel.org, 
	jhubbard@nvidia.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hi Deepak,

On Mon, May 13, 2024 at 7:32=E2=80=AFPM Deepak Gupta <debug@rivosinc.com> w=
rote:
>
> On Sun, May 12, 2024 at 06:31:24PM +0200, Alexandre Ghiti wrote:
> >On 04/04/2024 01:35, Deepak Gupta wrote:
> >>`fork` implements copy on write (COW) by making pages readonly in child
> >>and parent both.
> >>
> >>ptep_set_wrprotect and pte_wrprotect clears _PAGE_WRITE in PTE.
> >>Assumption is that page is readable and on fault copy on write happens.
> >>
> >>To implement COW on such pages,
> >
> >
> >I guess you mean "shadow stack pages" here.
>
> Yes I meant shadow stack pages. Will fix the message.
>
> >
> >
> >>  clearing up W bit makes them XWR =3D 000.
> >>This will result in wrong PTE setting which says no perms but V=3D1 and=
 PFN
> >>field pointing to final page. Instead desired behavior is to turn it in=
to
> >>a readable page, take an access (load/store) fault on sspush/sspop
> >>(shadow stack) and then perform COW on such pages.
> >>This way regular reads
> >>would still be allowed and not lead to COW maintaining current behavior
> >>of COW on non-shadow stack but writeable memory.
> >>
> >>On the other hand it doesn't interfere with existing COW for read-write
> >>memory. Assumption is always that _PAGE_READ must have been set and thu=
s
> >>setting _PAGE_READ is harmless.
> >>
> >>Signed-off-by: Deepak Gupta <debug@rivosinc.com>
> >>---
> >>  arch/riscv/include/asm/pgtable.h | 12 ++++++++++--
> >>  1 file changed, 10 insertions(+), 2 deletions(-)
> >>
> >>diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/=
pgtable.h
> >>index 9b837239d3e8..7a1c2a98d272 100644
> >>--- a/arch/riscv/include/asm/pgtable.h
> >>+++ b/arch/riscv/include/asm/pgtable.h
> >>@@ -398,7 +398,7 @@ static inline int pte_special(pte_t pte)
> >>  static inline pte_t pte_wrprotect(pte_t pte)
> >>  {
> >>-     return __pte(pte_val(pte) & ~(_PAGE_WRITE));
> >>+     return __pte((pte_val(pte) & ~(_PAGE_WRITE)) | (_PAGE_READ));
> >>  }
> >>  /* static inline pte_t pte_mkread(pte_t pte) */
> >>@@ -581,7 +581,15 @@ static inline pte_t ptep_get_and_clear(struct mm_s=
truct *mm,
> >>  static inline void ptep_set_wrprotect(struct mm_struct *mm,
> >>                                    unsigned long address, pte_t *ptep)
> >>  {
> >>-     atomic_long_and(~(unsigned long)_PAGE_WRITE, (atomic_long_t *)pte=
p);
> >>+     volatile pte_t read_pte =3D *ptep;

Sorry I missed this ^. You need to use ptep_get() to get the value of
a pte. And why do you need the volatile here?

> >>+     /*
> >>+      * ptep_set_wrprotect can be called for shadow stack ranges too.
> >>+      * shadow stack memory is XWR =3D 010 and thus clearing _PAGE_WRI=
TE will lead to
> >>+      * encoding 000b which is wrong encoding with V =3D 1. This shoul=
d lead to page fault
> >>+      * but we dont want this wrong configuration to be set in page ta=
bles.
> >>+      */
> >>+     atomic_long_set((atomic_long_t *)ptep,
> >>+                     ((pte_val(read_pte) & ~(unsigned long)_PAGE_WRITE=
) | _PAGE_READ));
> >>  }
> >>  #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
> >
> >
> >Doesn't making the shadow stack page readable allow "normal" loads to
> >access the page? If it does, isn't that an issue (security-wise)?
>
> When shadow stack permissions are there (i.e. R=3D0, W=3D1, X=3D0), then =
also shadow stack is
> readable through "normal" loads. So nothing changes when it converts into=
 a readonly page
> from page permissions perspective.
>
> Security-wise it's not a concern because from threat modeling perspective=
, if attacker had
> read-write primitives (via some bug in program) available to read and wri=
te address space
> of process/task; then they would have availiblity of return addresses on =
normal stack. It's
> the write primitive that is concerning and to be protected against. And t=
hat's why shadow stack
> is not writeable using "normal" stores.
>
> >

Thanks for the explanation!

With the use of ptep_get(), you can add:

Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>

Thanks,

Alex