[PATCH] arm64: mm: Mark .rodata as RO

[PATCH] arm64: mm: Mark .rodata as RO

[Jeremy Linton]

Currently the .rodata section is actually still executable when DEBUG_RODATA
is enabled. This changes that so the .rodata is actually read only, no execute.

Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
---
 arch/arm64/kernel/vmlinux.lds.S |  5 +++--
 arch/arm64/mm/mmu.c             | 14 +++++++++++---
 2 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S
index ab4e436..2e2c053 100644
--- a/arch/arm64/kernel/vmlinux.lds.S
+++ b/arch/arm64/kernel/vmlinux.lds.S
@@ -114,8 +114,9 @@ SECTIONS
 		*(.got)			/* Global offset table		*/
 	}
 
-	RO_DATA(PAGE_SIZE)
-	EXCEPTION_TABLE(8)
+	ALIGN_DEBUG_RO_MIN(0)
+	RO_DATA(PAGE_SIZE)		/* everything from this point to */
+	EXCEPTION_TABLE(8)		/* _etext will be marked RO NX   */
 	NOTES
 
 	ALIGN_DEBUG_RO_MIN(PAGE_SIZE)
diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index ab69a99..a3f4112 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -453,10 +453,18 @@ static void __init map_mem(pgd_t *pgd)
 #ifdef CONFIG_DEBUG_RODATA
 void mark_rodata_ro(void)
 {
-	create_mapping_late(__pa(_stext), (unsigned long)_stext,
-				(unsigned long)_etext - (unsigned long)_stext,
-				PAGE_KERNEL_ROX);
+	unsigned long section_size;
 
+	section_size = (unsigned long)__start_rodata - (unsigned long)_stext;
+	create_mapping_late(__pa(_stext), (unsigned long)_stext,
+			    section_size, PAGE_KERNEL_ROX);
+	/*
+	 * mark .rodata as read only. Use _etext rather than __end_rodata to
+	 * cover NOTES and EXCEPTION_TABLE.
+	 */
+	section_size = (unsigned long)_etext - (unsigned long)__start_rodata;
+	create_mapping_late(__pa(__start_rodata), (unsigned long)__start_rodata,
+			    section_size, PAGE_KERNEL_RO);
 }
 #endif
 
-- 
2.4.3

[PATCH] arm64: mm: Mark .rodata as RO

[Mark Rutland]

On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
> Currently the .rodata section is actually still executable when DEBUG_RODATA
> is enabled. This changes that so the .rodata is actually read only, no execute.
> 
> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
> ---
>  arch/arm64/kernel/vmlinux.lds.S |  5 +++--
>  arch/arm64/mm/mmu.c             | 14 +++++++++++---
>  2 files changed, 14 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S
> index ab4e436..2e2c053 100644
> --- a/arch/arm64/kernel/vmlinux.lds.S
> +++ b/arch/arm64/kernel/vmlinux.lds.S
> @@ -114,8 +114,9 @@ SECTIONS
>  		*(.got)			/* Global offset table		*/
>  	}
>  
> -	RO_DATA(PAGE_SIZE)
> -	EXCEPTION_TABLE(8)
> +	ALIGN_DEBUG_RO_MIN(0)
> +	RO_DATA(PAGE_SIZE)		/* everything from this point to */
> +	EXCEPTION_TABLE(8)		/* _etext will be marked RO NX   */
>  	NOTES
>  
>  	ALIGN_DEBUG_RO_MIN(PAGE_SIZE)
> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> index ab69a99..a3f4112 100644
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -453,10 +453,18 @@ static void __init map_mem(pgd_t *pgd)
>  #ifdef CONFIG_DEBUG_RODATA
>  void mark_rodata_ro(void)
>  {
> -	create_mapping_late(__pa(_stext), (unsigned long)_stext,
> -				(unsigned long)_etext - (unsigned long)_stext,
> -				PAGE_KERNEL_ROX);
> +	unsigned long section_size;
>  
> +	section_size = (unsigned long)__start_rodata - (unsigned long)_stext;
> +	create_mapping_late(__pa(_stext), (unsigned long)_stext,
> +			    section_size, PAGE_KERNEL_ROX);
> +	/*
> +	 * mark .rodata as read only. Use _etext rather than __end_rodata to
> +	 * cover NOTES and EXCEPTION_TABLE.
> +	 */
> +	section_size = (unsigned long)_etext - (unsigned long)__start_rodata;
> +	create_mapping_late(__pa(__start_rodata), (unsigned long)__start_rodata,
> +			    section_size, PAGE_KERNEL_RO);
>  }

As you pointed out in the other thread, we'll also need to update
map_kernel to use equivalent chunks for .text and .rodata.

I think we can probably make .rodata RO from the outset in map_kernel,
too.

Could you please update mem_init to log .text and .rodata separately?

It looks like some core code makes assumptions about _etext, so I guess
that has to cover .rodata regardless.

Otherwise, looks good!

Thanks,
Mark.

[PATCH] arm64: mm: Mark .rodata as RO

[Kees Cook]

On Fri, Feb 12, 2016 at 10:25 AM, Mark Rutland <mark.rutland@arm.com> wrote:
> On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
>> Currently the .rodata section is actually still executable when DEBUG_RODATA
>> is enabled. This changes that so the .rodata is actually read only, no execute.
>>
>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>

Yikes, good catch. Is anyone running the lkdtm tests that check these things?

Reviewed-by: Kees Cook <keescook@chromium.org>

>> ---
>>  arch/arm64/kernel/vmlinux.lds.S |  5 +++--
>>  arch/arm64/mm/mmu.c             | 14 +++++++++++---
>>  2 files changed, 14 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S
>> index ab4e436..2e2c053 100644
>> --- a/arch/arm64/kernel/vmlinux.lds.S
>> +++ b/arch/arm64/kernel/vmlinux.lds.S
>> @@ -114,8 +114,9 @@ SECTIONS
>>               *(.got)                 /* Global offset table          */
>>       }
>>
>> -     RO_DATA(PAGE_SIZE)
>> -     EXCEPTION_TABLE(8)
>> +     ALIGN_DEBUG_RO_MIN(0)
>> +     RO_DATA(PAGE_SIZE)              /* everything from this point to */
>> +     EXCEPTION_TABLE(8)              /* _etext will be marked RO NX   */
>>       NOTES
>>
>>       ALIGN_DEBUG_RO_MIN(PAGE_SIZE)
>> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
>> index ab69a99..a3f4112 100644
>> --- a/arch/arm64/mm/mmu.c
>> +++ b/arch/arm64/mm/mmu.c
>> @@ -453,10 +453,18 @@ static void __init map_mem(pgd_t *pgd)
>>  #ifdef CONFIG_DEBUG_RODATA
>>  void mark_rodata_ro(void)
>>  {
>> -     create_mapping_late(__pa(_stext), (unsigned long)_stext,
>> -                             (unsigned long)_etext - (unsigned long)_stext,
>> -                             PAGE_KERNEL_ROX);
>> +     unsigned long section_size;
>>
>> +     section_size = (unsigned long)__start_rodata - (unsigned long)_stext;
>> +     create_mapping_late(__pa(_stext), (unsigned long)_stext,
>> +                         section_size, PAGE_KERNEL_ROX);
>> +     /*
>> +      * mark .rodata as read only. Use _etext rather than __end_rodata to
>> +      * cover NOTES and EXCEPTION_TABLE.
>> +      */
>> +     section_size = (unsigned long)_etext - (unsigned long)__start_rodata;
>> +     create_mapping_late(__pa(__start_rodata), (unsigned long)__start_rodata,
>> +                         section_size, PAGE_KERNEL_RO);
>>  }
>
> As you pointed out in the other thread, we'll also need to update
> map_kernel to use equivalent chunks for .text and .rodata.
>
> I think we can probably make .rodata RO from the outset in map_kernel,
> too.

There's a benefit to marking it late in that .rodata can live in the
same (upcoming) section as .data..ro_after_init, which is marked ro
after mark_rodata_ro() to help constify more things.

> Could you please update mem_init to log .text and .rodata separately?
>
> It looks like some core code makes assumptions about _etext, so I guess
> that has to cover .rodata regardless.
>
> Otherwise, looks good!
>
> Thanks,
> Mark.

-Kees

-- 
Kees Cook
Chrome OS & Brillo Security

[kernel-hardening] Re: [PATCH] arm64: mm: Mark .rodata as RO

[Kees Cook]

On Fri, Feb 12, 2016 at 10:25 AM, Mark Rutland <mark.rutland@arm.com> wrote:
> On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
>> Currently the .rodata section is actually still executable when DEBUG_RODATA
>> is enabled. This changes that so the .rodata is actually read only, no execute.
>>
>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>

Yikes, good catch. Is anyone running the lkdtm tests that check these things?

Reviewed-by: Kees Cook <keescook@chromium.org>

>> ---
>>  arch/arm64/kernel/vmlinux.lds.S |  5 +++--
>>  arch/arm64/mm/mmu.c             | 14 +++++++++++---
>>  2 files changed, 14 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S
>> index ab4e436..2e2c053 100644
>> --- a/arch/arm64/kernel/vmlinux.lds.S
>> +++ b/arch/arm64/kernel/vmlinux.lds.S
>> @@ -114,8 +114,9 @@ SECTIONS
>>               *(.got)                 /* Global offset table          */
>>       }
>>
>> -     RO_DATA(PAGE_SIZE)
>> -     EXCEPTION_TABLE(8)
>> +     ALIGN_DEBUG_RO_MIN(0)
>> +     RO_DATA(PAGE_SIZE)              /* everything from this point to */
>> +     EXCEPTION_TABLE(8)              /* _etext will be marked RO NX   */
>>       NOTES
>>
>>       ALIGN_DEBUG_RO_MIN(PAGE_SIZE)
>> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
>> index ab69a99..a3f4112 100644
>> --- a/arch/arm64/mm/mmu.c
>> +++ b/arch/arm64/mm/mmu.c
>> @@ -453,10 +453,18 @@ static void __init map_mem(pgd_t *pgd)
>>  #ifdef CONFIG_DEBUG_RODATA
>>  void mark_rodata_ro(void)
>>  {
>> -     create_mapping_late(__pa(_stext), (unsigned long)_stext,
>> -                             (unsigned long)_etext - (unsigned long)_stext,
>> -                             PAGE_KERNEL_ROX);
>> +     unsigned long section_size;
>>
>> +     section_size = (unsigned long)__start_rodata - (unsigned long)_stext;
>> +     create_mapping_late(__pa(_stext), (unsigned long)_stext,
>> +                         section_size, PAGE_KERNEL_ROX);
>> +     /*
>> +      * mark .rodata as read only. Use _etext rather than __end_rodata to
>> +      * cover NOTES and EXCEPTION_TABLE.
>> +      */
>> +     section_size = (unsigned long)_etext - (unsigned long)__start_rodata;
>> +     create_mapping_late(__pa(__start_rodata), (unsigned long)__start_rodata,
>> +                         section_size, PAGE_KERNEL_RO);
>>  }
>
> As you pointed out in the other thread, we'll also need to update
> map_kernel to use equivalent chunks for .text and .rodata.
>
> I think we can probably make .rodata RO from the outset in map_kernel,
> too.

There's a benefit to marking it late in that .rodata can live in the
same (upcoming) section as .data..ro_after_init, which is marked ro
after mark_rodata_ro() to help constify more things.

> Could you please update mem_init to log .text and .rodata separately?
>
> It looks like some core code makes assumptions about _etext, so I guess
> that has to cover .rodata regardless.
>
> Otherwise, looks good!
>
> Thanks,
> Mark.

-Kees

-- 
Kees Cook
Chrome OS & Brillo Security

[PATCH] arm64: mm: Mark .rodata as RO

[Laura Abbott]

On 2/16/16 10:10 AM, Kees Cook wrote:
> On Fri, Feb 12, 2016 at 10:25 AM, Mark Rutland <mark.rutland@arm.com> wrote:
>> On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
>>> Currently the .rodata section is actually still executable when DEBUG_RODATA
>>> is enabled. This changes that so the .rodata is actually read only, no execute.
>>>
>>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>
> Yikes, good catch. Is anyone running the lkdtm tests that check these things?
>

I don't think the current lkdtm test would have caught this since the exec
test is using rw data and not ro data. That test could be expanded though
to include a rodata buffer as well.
  
Thanks,
Laura

[kernel-hardening] Re: [PATCH] arm64: mm: Mark .rodata as RO

[Laura Abbott]

On 2/16/16 10:10 AM, Kees Cook wrote:
> On Fri, Feb 12, 2016 at 10:25 AM, Mark Rutland <mark.rutland@arm.com> wrote:
>> On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
>>> Currently the .rodata section is actually still executable when DEBUG_RODATA
>>> is enabled. This changes that so the .rodata is actually read only, no execute.
>>>
>>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>
> Yikes, good catch. Is anyone running the lkdtm tests that check these things?
>

I don't think the current lkdtm test would have caught this since the exec
test is using rw data and not ro data. That test could be expanded though
to include a rodata buffer as well.
  
Thanks,
Laura

[PATCH] arm64: mm: Mark .rodata as RO

[Kees Cook]

On Tue, Feb 16, 2016 at 10:48 AM, Laura Abbott <laura@labbott.name> wrote:
> On 2/16/16 10:10 AM, Kees Cook wrote:
>>
>> On Fri, Feb 12, 2016 at 10:25 AM, Mark Rutland <mark.rutland@arm.com>
>> wrote:
>>>
>>> On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
>>>>
>>>> Currently the .rodata section is actually still executable when
>>>> DEBUG_RODATA
>>>> is enabled. This changes that so the .rodata is actually read only, no
>>>> execute.
>>>>
>>>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>>
>>
>> Yikes, good catch. Is anyone running the lkdtm tests that check these
>> things?
>>
>
> I don't think the current lkdtm test would have caught this since the exec
> test is using rw data and not ro data. That test could be expanded though
> to include a rodata buffer as well.

Oh, yeah, excellent point. I'll send a patch.

-Kees

-- 
Kees Cook
Chrome OS & Brillo Security

[kernel-hardening] Re: [PATCH] arm64: mm: Mark .rodata as RO

[Kees Cook]

On Tue, Feb 16, 2016 at 10:48 AM, Laura Abbott <laura@labbott.name> wrote:
> On 2/16/16 10:10 AM, Kees Cook wrote:
>>
>> On Fri, Feb 12, 2016 at 10:25 AM, Mark Rutland <mark.rutland@arm.com>
>> wrote:
>>>
>>> On Fri, Feb 12, 2016 at 10:13:19AM -0600, Jeremy Linton wrote:
>>>>
>>>> Currently the .rodata section is actually still executable when
>>>> DEBUG_RODATA
>>>> is enabled. This changes that so the .rodata is actually read only, no
>>>> execute.
>>>>
>>>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>>
>>
>> Yikes, good catch. Is anyone running the lkdtm tests that check these
>> things?
>>
>
> I don't think the current lkdtm test would have caught this since the exec
> test is using rw data and not ro data. That test could be expanded though
> to include a rodata buffer as well.

Oh, yeah, excellent point. I'll send a patch.

-Kees

-- 
Kees Cook
Chrome OS & Brillo Security