Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1703297 > unrolled thread

[PATCH] arm64: avoid overflow in VA_START and PAGE_OFFSET

Started byNick Desaulniers <ndesaulniers@google.com>
First post2017-08-03 19:20 +0200
Last post2017-08-03 20:10 +0200
Articles 4 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH] arm64: avoid overflow in VA_START and PAGE_OFFSET Nick Desaulniers <ndesaulniers@google.com> - 2017-08-03 19:20 +0200
    Re: [PATCH] arm64: avoid overflow in VA_START and PAGE_OFFSET Ard Biesheuvel <ard.biesheuvel@linaro.org> - 2017-08-03 19:30 +0200
      Re: [PATCH] arm64: avoid overflow in VA_START and PAGE_OFFSET Nick Desaulniers <ndesaulniers@google.com> - 2017-08-03 20:00 +0200
        [PATCH v2] arm64: avoid overflow in VA_START and PAGE_OFFSET Nick Desaulniers <ndesaulniers@google.com> - 2017-08-03 20:10 +0200

#1703297 — [PATCH] arm64: avoid overflow in VA_START and PAGE_OFFSET

FromNick Desaulniers <ndesaulniers@google.com>
Date2017-08-03 19:20 +0200
Subject[PATCH] arm64: avoid overflow in VA_START and PAGE_OFFSET
Message-ID<uarWX-37e-41@gated-at.bofh.it>
The bitmask used to define these values produces overflow, as seen by
this compiler warning:

arch/arm64/kernel/head.S:47:8: warning:
      integer overflow in preprocessor expression
  #elif (PAGE_OFFSET & 0x1fffff) != 0
         ^~~~~~~~~~~
arch/arm64/include/asm/memory.h:52:46: note:
      expanded from macro 'PAGE_OFFSET'
  #define PAGE_OFFSET             (UL(0xffffffffffffffff) << (VA_BITS -
1))
                                      ~~~~~~~~~~~~~~~~~~  ^

It would be preferrable to use GENMASK_ULL() instead, but it's not set
up to be used from assembly (the UL() macro token pastes UL suffixes
when not included in assembly sources).

Suggested-by: Yury Norov <ynorov@caviumnetworks.com>
Suggested-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
---
 arch/arm64/include/asm/memory.h | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/memory.h b/arch/arm64/include/asm/memory.h
index 32f82723338a..dde717a31dee 100644
--- a/arch/arm64/include/asm/memory.h
+++ b/arch/arm64/include/asm/memory.h
@@ -64,8 +64,9 @@
  * TASK_UNMAPPED_BASE - the lower boundary of the mmap VM area.
  */
 #define VA_BITS			(CONFIG_ARM64_VA_BITS)
-#define VA_START		(UL(0xffffffffffffffff) << VA_BITS)
-#define PAGE_OFFSET		(UL(0xffffffffffffffff) << (VA_BITS - 1))
+#define VA_START		((UL(0xffffffffffffffff) >> VA_BITS) << VA_BITS)
+#define PAGE_OFFSET		((UL(0xffffffffffffffff) >> (VA_BITS - 1)) \
+	<< (VA_BITS - 1))
 #define KIMAGE_VADDR		(MODULES_END)
 #define MODULES_END		(MODULES_VADDR + MODULES_VSIZE)
 #define MODULES_VADDR		(VA_START + KASAN_SHADOW_SIZE)
-- 
2.14.0.rc1.383.gd1ce394fe2-goog

[toc] | [next] | [standalone]

#1703303

FromArd Biesheuvel <ard.biesheuvel@linaro.org>
Date2017-08-03 19:30 +0200
Message-ID<uas6C-3bf-35@gated-at.bofh.it>
In reply to#1703297
On 3 August 2017 at 18:11, Nick Desaulniers <ndesaulniers@google.com> wrote:
> The bitmask used to define these values produces overflow, as seen by
> this compiler warning:
>
> arch/arm64/kernel/head.S:47:8: warning:
>       integer overflow in preprocessor expression
>   #elif (PAGE_OFFSET & 0x1fffff) != 0
>          ^~~~~~~~~~~
> arch/arm64/include/asm/memory.h:52:46: note:
>       expanded from macro 'PAGE_OFFSET'
>   #define PAGE_OFFSET             (UL(0xffffffffffffffff) << (VA_BITS -
> 1))
>                                       ~~~~~~~~~~~~~~~~~~  ^
>
> It would be preferrable to use GENMASK_ULL() instead, but it's not set
> up to be used from assembly (the UL() macro token pastes UL suffixes
> when not included in assembly sources).
>
> Suggested-by: Yury Norov <ynorov@caviumnetworks.com>
> Suggested-by: Matthias Kaehlcke <mka@chromium.org>
> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
> ---
>  arch/arm64/include/asm/memory.h | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/include/asm/memory.h b/arch/arm64/include/asm/memory.h
> index 32f82723338a..dde717a31dee 100644
> --- a/arch/arm64/include/asm/memory.h
> +++ b/arch/arm64/include/asm/memory.h
> @@ -64,8 +64,9 @@
>   * TASK_UNMAPPED_BASE - the lower boundary of the mmap VM area.
>   */
>  #define VA_BITS                        (CONFIG_ARM64_VA_BITS)
> -#define VA_START               (UL(0xffffffffffffffff) << VA_BITS)
> -#define PAGE_OFFSET            (UL(0xffffffffffffffff) << (VA_BITS - 1))
> +#define VA_START               ((UL(0xffffffffffffffff) >> VA_BITS) << VA_BITS)
> +#define PAGE_OFFSET            ((UL(0xffffffffffffffff) >> (VA_BITS - 1)) \
> +       << (VA_BITS - 1))
>  #define KIMAGE_VADDR           (MODULES_END)
>  #define MODULES_END            (MODULES_VADDR + MODULES_VSIZE)
>  #define MODULES_VADDR          (VA_START + KASAN_SHADOW_SIZE)
> --
> 2.14.0.rc1.383.gd1ce394fe2-goog
>

Would

#define VA_START (UL(0xffffffffffffffff) - (1 << VA_BITS) + 1)

also work?

[toc] | [prev] | [next] | [standalone]

#1703316

FromNick Desaulniers <ndesaulniers@google.com>
Date2017-08-03 20:00 +0200
Message-ID<uaszE-3m4-19@gated-at.bofh.it>
In reply to#1703303
> Would
>
> #define VA_START (UL(0xffffffffffffffff) - (1 << VA_BITS) + 1)
>
> also work?

I think you'd have to wrap the 1 in a UL(), ex:

#define VA_START (UL(0xffffffffffffffff) - (UL(1) << VA_BITS) + 1)

Otherwise IIUC a integral literal (`1`) is treated as an int, which on
arm64 is LP64 making it 32b, where most configs set VA_BITS
to larger than 32b.  Shifting by more than the width is undefined
behavior. And without it, I get compile errors.

I'll send v2 with your suggestion, thanks.
-- 
Thanks,
~Nick Desaulniers

[toc] | [prev] | [next] | [standalone]

#1703321 — [PATCH v2] arm64: avoid overflow in VA_START and PAGE_OFFSET

FromNick Desaulniers <ndesaulniers@google.com>
Date2017-08-03 20:10 +0200
Subject[PATCH v2] arm64: avoid overflow in VA_START and PAGE_OFFSET
Message-ID<uasJj-3EH-7@gated-at.bofh.it>
In reply to#1703316
The bitmask used to define these values produces overflow, as seen by
this compiler warning:

arch/arm64/kernel/head.S:47:8: warning:
      integer overflow in preprocessor expression
  #elif (PAGE_OFFSET & 0x1fffff) != 0
         ^~~~~~~~~~~
arch/arm64/include/asm/memory.h:52:46: note:
      expanded from macro 'PAGE_OFFSET'
  #define PAGE_OFFSET             (UL(0xffffffffffffffff) << (VA_BITS -
1))
                                      ~~~~~~~~~~~~~~~~~~  ^

It would be preferrable to use GENMASK_ULL() instead, but it's not set
up to be used from assembly (the UL() macro token pastes UL suffixes
when not included in assembly sources).

Suggested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Suggested-by: Yury Norov <ynorov@caviumnetworks.com>
Suggested-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
---
 arch/arm64/include/asm/memory.h | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/memory.h b/arch/arm64/include/asm/memory.h
index 32f82723338a..ef39dcb9ca6a 100644
--- a/arch/arm64/include/asm/memory.h
+++ b/arch/arm64/include/asm/memory.h
@@ -64,8 +64,10 @@
  * TASK_UNMAPPED_BASE - the lower boundary of the mmap VM area.
  */
 #define VA_BITS			(CONFIG_ARM64_VA_BITS)
-#define VA_START		(UL(0xffffffffffffffff) << VA_BITS)
-#define PAGE_OFFSET		(UL(0xffffffffffffffff) << (VA_BITS - 1))
+#define VA_START		(UL(0xffffffffffffffff) - \
+	(UL(1) << VA_BITS) + 1)
+#define PAGE_OFFSET		(UL(0xffffffffffffffff) - \
+	(UL(1) << (VA_BITS - 1)) + 1)
 #define KIMAGE_VADDR		(MODULES_END)
 #define MODULES_END		(MODULES_VADDR + MODULES_VSIZE)
 #define MODULES_VADDR		(VA_START + KASAN_SHADOW_SIZE)
-- 
2.14.0.rc1.383.gd1ce394fe2-goog

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web