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

Groups > linux.kernel > #1321434 > unrolled thread

[PATCHv2 0/2] Sanitization of buddy pages

Back to article view | Back to linux.kernel

Contents

  [PATCHv2 0/2] Sanitization of buddy pages Laura Abbott <labbott@fedoraproject.org> - 2016-01-29 03:40 +0100
    [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning Laura Abbott <labbott@fedoraproject.org> - 2016-01-29 03:40 +0100
      Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning "Rafael J. Wysocki" <rjw@rjwysocki.net> - 2016-01-29 05:00 +0100
      Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning Kees Cook <keescook@chromium.org> - 2016-01-29 05:50 +0100
        Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning Laura Abbott <labbott@redhat.com> - 2016-01-29 22:40 +0100
      Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning Pavel Machek <pavel@denx.de> - 2016-01-29 11:50 +0100
        Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning Laura Abbott <labbott@redhat.com> - 2016-01-29 22:40 +0100
          Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning Pavel Machek <pavel@denx.de> - 2016-01-30 16:40 +0100

#1321434 — [PATCHv2 0/2] Sanitization of buddy pages

Hi,

This is v2 of the series to add sanitization to the buddy allocator.
The standard sanitization blurb:

For those who aren't familiar with this, the goal of sanitization is to reduce
the severity of use after free and uninitialized data bugs. Memory is cleared
on free so any sensitive data is no longer available. Discussion of
sanitization was brough up in a thread about CVEs
(lkml.kernel.org/g/<20160119112812.GA10818@mwanda>)

Changes since v1:
- Squashed the refactor and adding the poisoning together. Having them separate
  didn't seem to give much extra benefit and lead to some churn as well.
- Corrected the order of poison vs. kernel_map in the alloc path
- zeroing can now be enabled with hibernation (enabling zero poisoning turns
  off hibernation)
- Added additional checks for skipping __GFP_ZERO. On SPARSEMEM systems the
  extended page flags are not initialized until after memory is freed to the
  buddy list which prevents the pages from being zeroed on first free via
  poisoning. This does also mean that any residual data that may be left in
  the pages from boot up will not be cleared which is a risk. I'm open to
  suggestions for fixing or it can be future work.
- A few spelling/checkpatch fixes.
- Addressed comments from Dave Hansen and Jianyu Zhan
- This series now depends on the change to allow debug_pagealloc_enabled
  to be used without !CONFIG_DEBUG_PAGEALLOC
  (http://article.gmane.org/gmane.linux.kernel.mm/145208)

Thanks,
Laura

Laura Abbott (2):
  mm/page_poison.c: Enable PAGE_POISONING as a separate option
  mm/page_poisoning.c: Allow for zero poisoning

 Documentation/kernel-parameters.txt |   5 +
 include/linux/mm.h                  |  15 +++
 include/linux/poison.h              |   4 +
 kernel/power/hibernate.c            |  17 ++++
 mm/Kconfig.debug                    |  36 ++++++-
 mm/Makefile                         |   2 +-
 mm/debug-pagealloc.c                | 137 ---------------------------
 mm/page_alloc.c                     |  13 ++-
 mm/page_ext.c                       |  10 +-
 mm/page_poison.c                    | 184 ++++++++++++++++++++++++++++++++++++
 10 files changed, 281 insertions(+), 142 deletions(-)
 delete mode 100644 mm/debug-pagealloc.c
 create mode 100644 mm/page_poison.c

-- 
2.5.0

[toc] | [next] | [standalone]

#1321442 — [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

By default, page poisoning uses a poison value (0xaa) on free. If this
is changed to 0, the page is not only sanitized but zeroing on alloc
with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
corruption from the poisoning is harder to detect. This feature also
cannot be used with hibernation since pages are not guaranteed to be
zeroed after hibernation.

Credit to Grsecurity/PaX team for inspiring this work

Signed-off-by: Laura Abbott <labbott@fedoraproject.org>
---
 include/linux/mm.h       |  2 ++
 include/linux/poison.h   |  4 ++++
 kernel/power/hibernate.c | 17 +++++++++++++++++
 mm/Kconfig.debug         | 14 ++++++++++++++
 mm/page_alloc.c          | 11 ++++++++++-
 mm/page_ext.c            | 10 ++++++++--
 mm/page_poison.c         |  7 +++++--
 7 files changed, 60 insertions(+), 5 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 966bf0e..59ce0dc 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2177,10 +2177,12 @@ extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
 #ifdef CONFIG_PAGE_POISONING
 extern bool page_poisoning_enabled(void);
 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
+extern bool page_is_poisoned(struct page *page);
 #else
 static inline bool page_poisoning_enabled(void) { return false; }
 static inline void kernel_poison_pages(struct page *page, int numpages,
 					int enable) { }
+static inline bool page_is_poisoned(struct page *page) { return false; }
 #endif
 
 #ifdef CONFIG_DEBUG_PAGEALLOC
diff --git a/include/linux/poison.h b/include/linux/poison.h
index 4a27153..51334ed 100644
--- a/include/linux/poison.h
+++ b/include/linux/poison.h
@@ -30,7 +30,11 @@
 #define TIMER_ENTRY_STATIC	((void *) 0x300 + POISON_POINTER_DELTA)
 
 /********** mm/debug-pagealloc.c **********/
+#ifdef CONFIG_PAGE_POISONING_ZERO
+#define PAGE_POISON 0x00
+#else
 #define PAGE_POISON 0xaa
+#endif
 
 /********** mm/page_alloc.c ************/
 
diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
index b7342a2..aa0f26b 100644
--- a/kernel/power/hibernate.c
+++ b/kernel/power/hibernate.c
@@ -1158,6 +1158,22 @@ static int __init kaslr_nohibernate_setup(char *str)
 	return nohibernate_setup(str);
 }
 
+static int __init page_poison_nohibernate_setup(char *str)
+{
+#ifdef CONFIG_PAGE_POISONING_ZERO
+	/*
+	 * The zeroing option for page poison skips the checks on alloc.
+	 * since hibernation doesn't save free pages there's no way to
+	 * guarantee the pages will still be zeroed.
+	 */
+	if (!strcmp(str, "on")) {
+		pr_info("Disabling hibernation due to page poisoning\n");
+		return nohibernate_setup(str);
+	}
+#endif
+	return 1;
+}
+
 __setup("noresume", noresume_setup);
 __setup("resume_offset=", resume_offset_setup);
 __setup("resume=", resume_setup);
@@ -1166,3 +1182,4 @@ __setup("resumewait", resumewait_setup);
 __setup("resumedelay=", resumedelay_setup);
 __setup("nohibernate", nohibernate_setup);
 __setup("kaslr", kaslr_nohibernate_setup);
+__setup("page_poison=", page_poison_nohibernate_setup);
diff --git a/mm/Kconfig.debug b/mm/Kconfig.debug
index 25c98ae..3d3b954 100644
--- a/mm/Kconfig.debug
+++ b/mm/Kconfig.debug
@@ -48,3 +48,17 @@ config PAGE_POISONING_NO_SANITY
 
 	   If you are only interested in sanitization, say Y. Otherwise
 	   say N.
+
+config PAGE_POISONING_ZERO
+	bool "Use zero for poisoning instead of random data"
+	depends on PAGE_POISONING
+	---help---
+	   Instead of using the existing poison value, fill the pages with
+	   zeros. This makes it harder to detect when errors are occurring
+	   due to sanitization but the zeroing at free means that it is
+	   no longer necessary to write zeros when GFP_ZERO is used on
+	   allocation.
+
+	   Enabling page poisoning with this option will disable hibernation
+
+	   If unsure, say N
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index cc4762a..59bd9dc 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1382,15 +1382,24 @@ static inline int check_new_page(struct page *page)
 	return 0;
 }
 
+static inline bool free_pages_prezeroed(bool poisoned)
+{
+	return IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) &&
+		page_poisoning_enabled() && poisoned;
+}
+
 static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
 								int alloc_flags)
 {
 	int i;
+	bool poisoned = true;
 
 	for (i = 0; i < (1 << order); i++) {
 		struct page *p = page + i;
 		if (unlikely(check_new_page(p)))
 			return 1;
+		if (poisoned)
+			poisoned &= page_is_poisoned(p);
 	}
 
 	set_page_private(page, 0);
@@ -1401,7 +1410,7 @@ static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
 	kernel_poison_pages(page, 1 << order, 1);
 	kasan_alloc_pages(page, order);
 
-	if (gfp_flags & __GFP_ZERO)
+	if (!free_pages_prezeroed(poisoned) && (gfp_flags & __GFP_ZERO))
 		for (i = 0; i < (1 << order); i++)
 			clear_highpage(page + i);
 
diff --git a/mm/page_ext.c b/mm/page_ext.c
index 292ca7b..2d864e6 100644
--- a/mm/page_ext.c
+++ b/mm/page_ext.c
@@ -106,12 +106,15 @@ struct page_ext *lookup_page_ext(struct page *page)
 	struct page_ext *base;
 
 	base = NODE_DATA(page_to_nid(page))->node_page_ext;
-#ifdef CONFIG_DEBUG_VM
+#if defined(CONFIG_DEBUG_VM) || defined(CONFIG_PAGE_POISONING)
 	/*
 	 * The sanity checks the page allocator does upon freeing a
 	 * page can reach here before the page_ext arrays are
 	 * allocated when feeding a range of pages to the allocator
 	 * for the first time during bootup or memory hotplug.
+	 *
+	 * This check is also necessary for ensuring page poisoning
+	 * works as expected when enabled
 	 */
 	if (unlikely(!base))
 		return NULL;
@@ -180,12 +183,15 @@ struct page_ext *lookup_page_ext(struct page *page)
 {
 	unsigned long pfn = page_to_pfn(page);
 	struct mem_section *section = __pfn_to_section(pfn);
-#ifdef CONFIG_DEBUG_VM
+#if defined(CONFIG_DEBUG_VM) || defined(CONFIG_PAGE_POISONING)
 	/*
 	 * The sanity checks the page allocator does upon freeing a
 	 * page can reach here before the page_ext arrays are
 	 * allocated when feeding a range of pages to the allocator
 	 * for the first time during bootup or memory hotplug.
+	 *
+	 * This check is also necessary for ensuring page poisoning
+	 * works as expected when enabled
 	 */
 	if (!section->page_ext)
 		return NULL;
diff --git a/mm/page_poison.c b/mm/page_poison.c
index 89d3bc7..479e7ea 100644
--- a/mm/page_poison.c
+++ b/mm/page_poison.c
@@ -71,11 +71,14 @@ static inline void clear_page_poison(struct page *page)
 	__clear_bit(PAGE_EXT_DEBUG_POISON, &page_ext->flags);
 }
 
-static inline bool page_poison(struct page *page)
+bool page_is_poisoned(struct page *page)
 {
 	struct page_ext *page_ext;
 
 	page_ext = lookup_page_ext(page);
+	if (!page_ext)
+		return false;
+
 	return test_bit(PAGE_EXT_DEBUG_POISON, &page_ext->flags);
 }
 
@@ -137,7 +140,7 @@ static void unpoison_page(struct page *page)
 {
 	void *addr;
 
-	if (!page_poison(page))
+	if (!page_is_poisoned(page))
 		return;
 
 	addr = kmap_atomic(page);
-- 
2.5.0

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

#1321471 — Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

On Thursday, January 28, 2016 06:38:19 PM Laura Abbott wrote:
> By default, page poisoning uses a poison value (0xaa) on free. If this
> is changed to 0, the page is not only sanitized but zeroing on alloc
> with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
> corruption from the poisoning is harder to detect. This feature also
> cannot be used with hibernation since pages are not guaranteed to be
> zeroed after hibernation.
> 
> Credit to Grsecurity/PaX team for inspiring this work
> 
> Signed-off-by: Laura Abbott <labbott@fedoraproject.org>

The hibernation disabling part is fine by me.

Please feel free to add an ACK from me to this if that helps.

Thanks,
Rafael

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

#1321484 — Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

On Thu, Jan 28, 2016 at 6:38 PM, Laura Abbott <labbott@fedoraproject.org> wrote:
> By default, page poisoning uses a poison value (0xaa) on free. If this
> is changed to 0, the page is not only sanitized but zeroing on alloc
> with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
> corruption from the poisoning is harder to detect. This feature also
> cannot be used with hibernation since pages are not guaranteed to be
> zeroed after hibernation.
>
> Credit to Grsecurity/PaX team for inspiring this work
>
> Signed-off-by: Laura Abbott <labbott@fedoraproject.org>
> ---
>  include/linux/mm.h       |  2 ++
>  include/linux/poison.h   |  4 ++++
>  kernel/power/hibernate.c | 17 +++++++++++++++++
>  mm/Kconfig.debug         | 14 ++++++++++++++
>  mm/page_alloc.c          | 11 ++++++++++-
>  mm/page_ext.c            | 10 ++++++++--
>  mm/page_poison.c         |  7 +++++--
>  7 files changed, 60 insertions(+), 5 deletions(-)
>
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 966bf0e..59ce0dc 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2177,10 +2177,12 @@ extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
>  #ifdef CONFIG_PAGE_POISONING
>  extern bool page_poisoning_enabled(void);
>  extern void kernel_poison_pages(struct page *page, int numpages, int enable);
> +extern bool page_is_poisoned(struct page *page);
>  #else
>  static inline bool page_poisoning_enabled(void) { return false; }
>  static inline void kernel_poison_pages(struct page *page, int numpages,
>                                         int enable) { }
> +static inline bool page_is_poisoned(struct page *page) { return false; }
>  #endif
>
>  #ifdef CONFIG_DEBUG_PAGEALLOC
> diff --git a/include/linux/poison.h b/include/linux/poison.h
> index 4a27153..51334ed 100644
> --- a/include/linux/poison.h
> +++ b/include/linux/poison.h
> @@ -30,7 +30,11 @@
>  #define TIMER_ENTRY_STATIC     ((void *) 0x300 + POISON_POINTER_DELTA)
>
>  /********** mm/debug-pagealloc.c **********/
> +#ifdef CONFIG_PAGE_POISONING_ZERO
> +#define PAGE_POISON 0x00
> +#else
>  #define PAGE_POISON 0xaa
> +#endif
>
>  /********** mm/page_alloc.c ************/
>
> diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
> index b7342a2..aa0f26b 100644
> --- a/kernel/power/hibernate.c
> +++ b/kernel/power/hibernate.c
> @@ -1158,6 +1158,22 @@ static int __init kaslr_nohibernate_setup(char *str)
>         return nohibernate_setup(str);
>  }
>
> +static int __init page_poison_nohibernate_setup(char *str)
> +{
> +#ifdef CONFIG_PAGE_POISONING_ZERO
> +       /*
> +        * The zeroing option for page poison skips the checks on alloc.
> +        * since hibernation doesn't save free pages there's no way to
> +        * guarantee the pages will still be zeroed.
> +        */
> +       if (!strcmp(str, "on")) {
> +               pr_info("Disabling hibernation due to page poisoning\n");
> +               return nohibernate_setup(str);
> +       }
> +#endif
> +       return 1;
> +}
> +
>  __setup("noresume", noresume_setup);
>  __setup("resume_offset=", resume_offset_setup);
>  __setup("resume=", resume_setup);
> @@ -1166,3 +1182,4 @@ __setup("resumewait", resumewait_setup);
>  __setup("resumedelay=", resumedelay_setup);
>  __setup("nohibernate", nohibernate_setup);
>  __setup("kaslr", kaslr_nohibernate_setup);
> +__setup("page_poison=", page_poison_nohibernate_setup);
> diff --git a/mm/Kconfig.debug b/mm/Kconfig.debug
> index 25c98ae..3d3b954 100644
> --- a/mm/Kconfig.debug
> +++ b/mm/Kconfig.debug
> @@ -48,3 +48,17 @@ config PAGE_POISONING_NO_SANITY
>
>            If you are only interested in sanitization, say Y. Otherwise
>            say N.
> +
> +config PAGE_POISONING_ZERO
> +       bool "Use zero for poisoning instead of random data"
> +       depends on PAGE_POISONING
> +       ---help---
> +          Instead of using the existing poison value, fill the pages with
> +          zeros. This makes it harder to detect when errors are occurring
> +          due to sanitization but the zeroing at free means that it is
> +          no longer necessary to write zeros when GFP_ZERO is used on
> +          allocation.

May be worth noting the security benefit in this help text.

> +
> +          Enabling page poisoning with this option will disable hibernation

This isn't obvious to me. It looks like you need to both use
CONFIG_PAGE_POISOING_ZERO and put "page_poison=on" on the command line
to enable it?

-Kees

> +
> +          If unsure, say N
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index cc4762a..59bd9dc 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -1382,15 +1382,24 @@ static inline int check_new_page(struct page *page)
>         return 0;
>  }
>
> +static inline bool free_pages_prezeroed(bool poisoned)
> +{
> +       return IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) &&
> +               page_poisoning_enabled() && poisoned;
> +}
> +
>  static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
>                                                                 int alloc_flags)
>  {
>         int i;
> +       bool poisoned = true;
>
>         for (i = 0; i < (1 << order); i++) {
>                 struct page *p = page + i;
>                 if (unlikely(check_new_page(p)))
>                         return 1;
> +               if (poisoned)
> +                       poisoned &= page_is_poisoned(p);
>         }
>
>         set_page_private(page, 0);
> @@ -1401,7 +1410,7 @@ static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
>         kernel_poison_pages(page, 1 << order, 1);
>         kasan_alloc_pages(page, order);
>
> -       if (gfp_flags & __GFP_ZERO)
> +       if (!free_pages_prezeroed(poisoned) && (gfp_flags & __GFP_ZERO))
>                 for (i = 0; i < (1 << order); i++)
>                         clear_highpage(page + i);
>
> diff --git a/mm/page_ext.c b/mm/page_ext.c
> index 292ca7b..2d864e6 100644
> --- a/mm/page_ext.c
> +++ b/mm/page_ext.c
> @@ -106,12 +106,15 @@ struct page_ext *lookup_page_ext(struct page *page)
>         struct page_ext *base;
>
>         base = NODE_DATA(page_to_nid(page))->node_page_ext;
> -#ifdef CONFIG_DEBUG_VM
> +#if defined(CONFIG_DEBUG_VM) || defined(CONFIG_PAGE_POISONING)
>         /*
>          * The sanity checks the page allocator does upon freeing a
>          * page can reach here before the page_ext arrays are
>          * allocated when feeding a range of pages to the allocator
>          * for the first time during bootup or memory hotplug.
> +        *
> +        * This check is also necessary for ensuring page poisoning
> +        * works as expected when enabled
>          */
>         if (unlikely(!base))
>                 return NULL;
> @@ -180,12 +183,15 @@ struct page_ext *lookup_page_ext(struct page *page)
>  {
>         unsigned long pfn = page_to_pfn(page);
>         struct mem_section *section = __pfn_to_section(pfn);
> -#ifdef CONFIG_DEBUG_VM
> +#if defined(CONFIG_DEBUG_VM) || defined(CONFIG_PAGE_POISONING)
>         /*
>          * The sanity checks the page allocator does upon freeing a
>          * page can reach here before the page_ext arrays are
>          * allocated when feeding a range of pages to the allocator
>          * for the first time during bootup or memory hotplug.
> +        *
> +        * This check is also necessary for ensuring page poisoning
> +        * works as expected when enabled
>          */
>         if (!section->page_ext)
>                 return NULL;
> diff --git a/mm/page_poison.c b/mm/page_poison.c
> index 89d3bc7..479e7ea 100644
> --- a/mm/page_poison.c
> +++ b/mm/page_poison.c
> @@ -71,11 +71,14 @@ static inline void clear_page_poison(struct page *page)
>         __clear_bit(PAGE_EXT_DEBUG_POISON, &page_ext->flags);
>  }
>
> -static inline bool page_poison(struct page *page)
> +bool page_is_poisoned(struct page *page)
>  {
>         struct page_ext *page_ext;
>
>         page_ext = lookup_page_ext(page);
> +       if (!page_ext)
> +               return false;
> +
>         return test_bit(PAGE_EXT_DEBUG_POISON, &page_ext->flags);
>  }
>
> @@ -137,7 +140,7 @@ static void unpoison_page(struct page *page)
>  {
>         void *addr;
>
> -       if (!page_poison(page))
> +       if (!page_is_poisoned(page))
>                 return;
>
>         addr = kmap_atomic(page);
> --
> 2.5.0
>



-- 
Kees Cook
Chrome OS & Brillo Security

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

#1322112 — Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

On 01/28/2016 08:46 PM, Kees Cook wrote:
> On Thu, Jan 28, 2016 at 6:38 PM, Laura Abbott <labbott@fedoraproject.org> wrote:
>> By default, page poisoning uses a poison value (0xaa) on free. If this
>> is changed to 0, the page is not only sanitized but zeroing on alloc
>> with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
>> corruption from the poisoning is harder to detect. This feature also
>> cannot be used with hibernation since pages are not guaranteed to be
>> zeroed after hibernation.
>>
>> Credit to Grsecurity/PaX team for inspiring this work
>>
>> Signed-off-by: Laura Abbott <labbott@fedoraproject.org>
>> ---
>>   include/linux/mm.h       |  2 ++
>>   include/linux/poison.h   |  4 ++++
>>   kernel/power/hibernate.c | 17 +++++++++++++++++
>>   mm/Kconfig.debug         | 14 ++++++++++++++
>>   mm/page_alloc.c          | 11 ++++++++++-
>>   mm/page_ext.c            | 10 ++++++++--
>>   mm/page_poison.c         |  7 +++++--
>>   7 files changed, 60 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>> index 966bf0e..59ce0dc 100644
>> --- a/include/linux/mm.h
>> +++ b/include/linux/mm.h
>> @@ -2177,10 +2177,12 @@ extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
>>   #ifdef CONFIG_PAGE_POISONING
>>   extern bool page_poisoning_enabled(void);
>>   extern void kernel_poison_pages(struct page *page, int numpages, int enable);
>> +extern bool page_is_poisoned(struct page *page);
>>   #else
>>   static inline bool page_poisoning_enabled(void) { return false; }
>>   static inline void kernel_poison_pages(struct page *page, int numpages,
>>                                          int enable) { }
>> +static inline bool page_is_poisoned(struct page *page) { return false; }
>>   #endif
>>
>>   #ifdef CONFIG_DEBUG_PAGEALLOC
>> diff --git a/include/linux/poison.h b/include/linux/poison.h
>> index 4a27153..51334ed 100644
>> --- a/include/linux/poison.h
>> +++ b/include/linux/poison.h
>> @@ -30,7 +30,11 @@
>>   #define TIMER_ENTRY_STATIC     ((void *) 0x300 + POISON_POINTER_DELTA)
>>
>>   /********** mm/debug-pagealloc.c **********/
>> +#ifdef CONFIG_PAGE_POISONING_ZERO
>> +#define PAGE_POISON 0x00
>> +#else
>>   #define PAGE_POISON 0xaa
>> +#endif
>>
>>   /********** mm/page_alloc.c ************/
>>
>> diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
>> index b7342a2..aa0f26b 100644
>> --- a/kernel/power/hibernate.c
>> +++ b/kernel/power/hibernate.c
>> @@ -1158,6 +1158,22 @@ static int __init kaslr_nohibernate_setup(char *str)
>>          return nohibernate_setup(str);
>>   }
>>
>> +static int __init page_poison_nohibernate_setup(char *str)
>> +{
>> +#ifdef CONFIG_PAGE_POISONING_ZERO
>> +       /*
>> +        * The zeroing option for page poison skips the checks on alloc.
>> +        * since hibernation doesn't save free pages there's no way to
>> +        * guarantee the pages will still be zeroed.
>> +        */
>> +       if (!strcmp(str, "on")) {
>> +               pr_info("Disabling hibernation due to page poisoning\n");
>> +               return nohibernate_setup(str);
>> +       }
>> +#endif
>> +       return 1;
>> +}
>> +
>>   __setup("noresume", noresume_setup);
>>   __setup("resume_offset=", resume_offset_setup);
>>   __setup("resume=", resume_setup);
>> @@ -1166,3 +1182,4 @@ __setup("resumewait", resumewait_setup);
>>   __setup("resumedelay=", resumedelay_setup);
>>   __setup("nohibernate", nohibernate_setup);
>>   __setup("kaslr", kaslr_nohibernate_setup);
>> +__setup("page_poison=", page_poison_nohibernate_setup);
>> diff --git a/mm/Kconfig.debug b/mm/Kconfig.debug
>> index 25c98ae..3d3b954 100644
>> --- a/mm/Kconfig.debug
>> +++ b/mm/Kconfig.debug
>> @@ -48,3 +48,17 @@ config PAGE_POISONING_NO_SANITY
>>
>>             If you are only interested in sanitization, say Y. Otherwise
>>             say N.
>> +
>> +config PAGE_POISONING_ZERO
>> +       bool "Use zero for poisoning instead of random data"
>> +       depends on PAGE_POISONING
>> +       ---help---
>> +          Instead of using the existing poison value, fill the pages with
>> +          zeros. This makes it harder to detect when errors are occurring
>> +          due to sanitization but the zeroing at free means that it is
>> +          no longer necessary to write zeros when GFP_ZERO is used on
>> +          allocation.
>
> May be worth noting the security benefit in this help text.
>

This is supposed to build on the existing page poisoning which mentions the
security bit. I think this text needs to be clarified how this works.
  
>> +
>> +          Enabling page poisoning with this option will disable hibernation
>
> This isn't obvious to me. It looks like you need to both use
> CONFIG_PAGE_POISOING_ZERO and put "page_poison=on" on the command line
> to enable it?

Yeah, this isn't really clear. I'll make it more obvious this is an extension
of page poisoning so page poisoning must be enabled first.

>
> -Kees
>

Thanks,
Laura
  
>> +
>> +          If unsure, say N
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index cc4762a..59bd9dc 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -1382,15 +1382,24 @@ static inline int check_new_page(struct page *page)
>>          return 0;
>>   }
>>
>> +static inline bool free_pages_prezeroed(bool poisoned)
>> +{
>> +       return IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) &&
>> +               page_poisoning_enabled() && poisoned;
>> +}
>> +
>>   static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
>>                                                                  int alloc_flags)
>>   {
>>          int i;
>> +       bool poisoned = true;
>>
>>          for (i = 0; i < (1 << order); i++) {
>>                  struct page *p = page + i;
>>                  if (unlikely(check_new_page(p)))
>>                          return 1;
>> +               if (poisoned)
>> +                       poisoned &= page_is_poisoned(p);
>>          }
>>
>>          set_page_private(page, 0);
>> @@ -1401,7 +1410,7 @@ static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
>>          kernel_poison_pages(page, 1 << order, 1);
>>          kasan_alloc_pages(page, order);
>>
>> -       if (gfp_flags & __GFP_ZERO)
>> +       if (!free_pages_prezeroed(poisoned) && (gfp_flags & __GFP_ZERO))
>>                  for (i = 0; i < (1 << order); i++)
>>                          clear_highpage(page + i);
>>
>> diff --git a/mm/page_ext.c b/mm/page_ext.c
>> index 292ca7b..2d864e6 100644
>> --- a/mm/page_ext.c
>> +++ b/mm/page_ext.c
>> @@ -106,12 +106,15 @@ struct page_ext *lookup_page_ext(struct page *page)
>>          struct page_ext *base;
>>
>>          base = NODE_DATA(page_to_nid(page))->node_page_ext;
>> -#ifdef CONFIG_DEBUG_VM
>> +#if defined(CONFIG_DEBUG_VM) || defined(CONFIG_PAGE_POISONING)
>>          /*
>>           * The sanity checks the page allocator does upon freeing a
>>           * page can reach here before the page_ext arrays are
>>           * allocated when feeding a range of pages to the allocator
>>           * for the first time during bootup or memory hotplug.
>> +        *
>> +        * This check is also necessary for ensuring page poisoning
>> +        * works as expected when enabled
>>           */
>>          if (unlikely(!base))
>>                  return NULL;
>> @@ -180,12 +183,15 @@ struct page_ext *lookup_page_ext(struct page *page)
>>   {
>>          unsigned long pfn = page_to_pfn(page);
>>          struct mem_section *section = __pfn_to_section(pfn);
>> -#ifdef CONFIG_DEBUG_VM
>> +#if defined(CONFIG_DEBUG_VM) || defined(CONFIG_PAGE_POISONING)
>>          /*
>>           * The sanity checks the page allocator does upon freeing a
>>           * page can reach here before the page_ext arrays are
>>           * allocated when feeding a range of pages to the allocator
>>           * for the first time during bootup or memory hotplug.
>> +        *
>> +        * This check is also necessary for ensuring page poisoning
>> +        * works as expected when enabled
>>           */
>>          if (!section->page_ext)
>>                  return NULL;
>> diff --git a/mm/page_poison.c b/mm/page_poison.c
>> index 89d3bc7..479e7ea 100644
>> --- a/mm/page_poison.c
>> +++ b/mm/page_poison.c
>> @@ -71,11 +71,14 @@ static inline void clear_page_poison(struct page *page)
>>          __clear_bit(PAGE_EXT_DEBUG_POISON, &page_ext->flags);
>>   }
>>
>> -static inline bool page_poison(struct page *page)
>> +bool page_is_poisoned(struct page *page)
>>   {
>>          struct page_ext *page_ext;
>>
>>          page_ext = lookup_page_ext(page);
>> +       if (!page_ext)
>> +               return false;
>> +
>>          return test_bit(PAGE_EXT_DEBUG_POISON, &page_ext->flags);
>>   }
>>
>> @@ -137,7 +140,7 @@ static void unpoison_page(struct page *page)
>>   {
>>          void *addr;
>>
>> -       if (!page_poison(page))
>> +       if (!page_is_poisoned(page))
>>                  return;
>>
>>          addr = kmap_atomic(page);
>> --
>> 2.5.0
>>
>
>
>

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

#1321649 — Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

Hi!

> By default, page poisoning uses a poison value (0xaa) on free. If this
> is changed to 0, the page is not only sanitized but zeroing on alloc
> with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
> corruption from the poisoning is harder to detect. This feature also
> cannot be used with hibernation since pages are not guaranteed to be
> zeroed after hibernation.

So... this makes kernel harder to debug for performance advantage...?
If so.. how big is the performance advantage?
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

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

#1322110 — Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

On 01/29/2016 02:45 AM, Pavel Machek wrote:
> Hi!
>
>> By default, page poisoning uses a poison value (0xaa) on free. If this
>> is changed to 0, the page is not only sanitized but zeroing on alloc
>> with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
>> corruption from the poisoning is harder to detect. This feature also
>> cannot be used with hibernation since pages are not guaranteed to be
>> zeroed after hibernation.
>
> So... this makes kernel harder to debug for performance advantage...?
> If so.. how big is the performance advantage?
> 									Pavel
>

The performance advantage really depends on the benchmark you are running.
It was pointed out this may help some unknown amount with merging pages
in VMs since the pages are now identical and can be merged. The debugging
is also only slightly more difficult. With the non-zero poisoning value
it's easier to see that a crash was caused by triggering the poison vs.
just some random NULL pointer.

As as been pointed out, this help text could use some updating so I'll
clarify this more.

Thanks,
Laura

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

#1322394 — Re: [PATCHv2 2/2] mm/page_poisoning.c: Allow for zero poisoning

Hi!

> >>By default, page poisoning uses a poison value (0xaa) on free. If this
> >>is changed to 0, the page is not only sanitized but zeroing on alloc
> >>with __GFP_ZERO can be skipped as well. The tradeoff is that detecting
> >>corruption from the poisoning is harder to detect. This feature also
> >>cannot be used with hibernation since pages are not guaranteed to be
> >>zeroed after hibernation.
> >
> >So... this makes kernel harder to debug for performance advantage...?
> >If so.. how big is the performance advantage?

> 
> The performance advantage really depends on the benchmark you are
> running.

You are trying to improve performance, so you should publish at least
one benchmark where it helps.

Alternatively, quote kernel build times with and without the
patch.

If it speeds kernel compile twice, I guess I may even help with
hibernation support. If it makes kernel compile faster by .00000034%
(or slows it down), we should probably simply ignore this patch.

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web