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

Groups > linux.kernel > #1342521 > unrolled thread

[RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option

Started byKees Cook <keescook@chromium.org>
First post2016-02-25 00:40 +0100
Last post2016-02-26 23:30 +0100
Articles 5 — 3 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

  [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option Kees Cook <keescook@chromium.org> - 2016-02-25 00:40 +0100
    Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a  separate option Jianyu Zhan <nasa4836@gmail.com> - 2016-02-26 04:00 +0100
      Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a  separate option Laura Abbott <labbott@redhat.com> - 2016-02-26 05:50 +0100
        Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a  separate option Jianyu Zhan <nasa4836@gmail.com> - 2016-02-26 06:40 +0100
          Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a  separate option Laura Abbott <labbott@redhat.com> - 2016-02-26 23:30 +0100

#1342521 — [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option

FromKees Cook <keescook@chromium.org>
Date2016-02-25 00:40 +0100
Subject[RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option
Message-ID<r5Rce-3uS-5@gated-at.bofh.it>
From: Laura Abbott <labbott@fedoraproject.org>

Page poisoning is currently setup as a feature if architectures don't
have architecture debug page_alloc to allow unmapping of pages. It has
uses apart from that though. Clearing of the pages on free provides
an increase in security as it helps to limit the risk of information
leaks. Allow page poisoning to be enabled as a separate option
independent of any other debug feature. Because of how hiberanation
is implemented, the checks on alloc cannot occur if hibernation is
enabled. This option can also be set on !HIBERNATION as well.

Credit to Grsecurity/PaX team for inspiring this work

Signed-off-by: Laura Abbott <labbott@fedoraproject.org>
[rebased by Kees Cook <keescook@chromium.org>]
Tested-by: Kees Cook <keescook@chromium.org>
---
 include/linux/mm.h |  7 +++----
 mm/Kconfig.debug   | 22 +++++++++++++++++++++-
 mm/Makefile        |  4 ----
 mm/page_alloc.c    |  2 ++
 mm/page_poison.c   | 29 +++++++++++++++++++++++++----
 5 files changed, 51 insertions(+), 13 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index ea5de9d3e00b..6cdd8d91e5ef 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2199,13 +2199,12 @@ extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
 
 
 #ifdef CONFIG_PAGE_POISONING
-extern void poison_pages(struct page *page, int n);
-extern void unpoison_pages(struct page *page, int n);
 extern bool page_poisoning_enabled(void);
+extern void kernel_poison_pages(struct page *page, int numpages, int enable);
 #else
-static inline void poison_pages(struct page *page, int n) { }
-static inline void unpoison_pages(struct page *page, int n) { }
 static inline bool page_poisoning_enabled(void) { return false; }
+static inline void kernel_poison_pages(struct page *page, int numpages,
+				       int enable) { }
 #endif
 
 #ifdef CONFIG_DEBUG_PAGEALLOC
diff --git a/mm/Kconfig.debug b/mm/Kconfig.debug
index a0c136af9c91..ddf71d7cb6ba 100644
--- a/mm/Kconfig.debug
+++ b/mm/Kconfig.debug
@@ -41,4 +41,24 @@ config DEBUG_PAGEALLOC_ENABLE_DEFAULT
 	  can be overridden by debug_pagealloc=off|on.
 
 config PAGE_POISONING
-	bool
+	bool "Poison pages after freeing"
+	select PAGE_EXTENSION
+	select PAGE_POISONING_NO_SANITY if HIBERNATION
+	---help---
+	  Fill the pages with poison patterns after free_pages() and verify
+	  the patterns before alloc_pages. The filling of the memory helps
+	  reduce the risk of information leaks from freed data. This does
+	  have a potential performance impact.
+
+	  If unsure, say N
+
+config PAGE_POISONING_NO_SANITY
+	depends on PAGE_POISONING
+	bool "Only poison, don't sanity check"
+	---help---
+	   Skip the sanity checking on alloc, only fill the pages with
+	   poison on free. This reduces some of the overhead of the
+	   poisoning feature.
+
+	   If you are only interested in sanitization, say Y. Otherwise
+	   say N.
diff --git a/mm/Makefile b/mm/Makefile
index fb1a7948c107..ec59c071b4f9 100644
--- a/mm/Makefile
+++ b/mm/Makefile
@@ -13,7 +13,6 @@ KCOV_INSTRUMENT_slob.o := n
 KCOV_INSTRUMENT_slab.o := n
 KCOV_INSTRUMENT_slub.o := n
 KCOV_INSTRUMENT_page_alloc.o := n
-KCOV_INSTRUMENT_debug-pagealloc.o := n
 KCOV_INSTRUMENT_kmemleak.o := n
 KCOV_INSTRUMENT_kmemcheck.o := n
 KCOV_INSTRUMENT_memcontrol.o := n
@@ -63,9 +62,6 @@ obj-$(CONFIG_SPARSEMEM_VMEMMAP) += sparse-vmemmap.o
 obj-$(CONFIG_SLOB) += slob.o
 obj-$(CONFIG_MMU_NOTIFIER) += mmu_notifier.o
 obj-$(CONFIG_KSM) += ksm.o
-ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
-	obj-$(CONFIG_DEBUG_PAGEALLOC) += debug-pagealloc.o
-endif
 obj-$(CONFIG_PAGE_POISONING) += page_poison.o
 obj-$(CONFIG_SLAB) += slab.o
 obj-$(CONFIG_SLUB) += slub.o
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index a34c359d8e81..0bdb3cfd83b5 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1026,6 +1026,7 @@ static bool free_pages_prepare(struct page *page, unsigned int order)
 					   PAGE_SIZE << order);
 	}
 	arch_free_page(page, order);
+	kernel_poison_pages(page, 1 << order, 0);
 	kernel_map_pages(page, 1 << order, 0);
 
 	return true;
@@ -1497,6 +1498,7 @@ static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
 
 	arch_alloc_page(page, order);
 	kernel_map_pages(page, 1 << order, 1);
+	kernel_poison_pages(page, 1 << order, 1);
 	kasan_alloc_pages(page, order);
 
 	if (gfp_flags & __GFP_ZERO)
diff --git a/mm/page_poison.c b/mm/page_poison.c
index 92ead727b8f0..884a6f854432 100644
--- a/mm/page_poison.c
+++ b/mm/page_poison.c
@@ -80,7 +80,7 @@ static void poison_page(struct page *page)
 	kunmap_atomic(addr);
 }
 
-void poison_pages(struct page *page, int n)
+static void poison_pages(struct page *page, int n)
 {
 	int i;
 
@@ -101,6 +101,9 @@ static void check_poison_mem(unsigned char *mem, size_t bytes)
 	unsigned char *start;
 	unsigned char *end;
 
+	if (IS_ENABLED(CONFIG_PAGE_POISONING_NO_SANITY))
+		return;
+
 	start = memchr_inv(mem, PAGE_POISON, bytes);
 	if (!start)
 		return;
@@ -113,9 +116,9 @@ static void check_poison_mem(unsigned char *mem, size_t bytes)
 	if (!__ratelimit(&ratelimit))
 		return;
 	else if (start == end && single_bit_flip(*start, PAGE_POISON))
-		printk(KERN_ERR "pagealloc: single bit error\n");
+		pr_err("pagealloc: single bit error\n");
 	else
-		printk(KERN_ERR "pagealloc: memory corruption\n");
+		pr_err("pagealloc: memory corruption\n");
 
 	print_hex_dump(KERN_ERR, "", DUMP_PREFIX_ADDRESS, 16, 1, start,
 			end - start + 1, 1);
@@ -135,10 +138,28 @@ static void unpoison_page(struct page *page)
 	kunmap_atomic(addr);
 }
 
-void unpoison_pages(struct page *page, int n)
+static void unpoison_pages(struct page *page, int n)
 {
 	int i;
 
 	for (i = 0; i < n; i++)
 		unpoison_page(page + i);
 }
+
+void kernel_poison_pages(struct page *page, int numpages, int enable)
+{
+	if (!page_poisoning_enabled())
+		return;
+
+	if (enable)
+		unpoison_pages(page, numpages);
+	else
+		poison_pages(page, numpages);
+}
+
+#ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
+void __kernel_map_pages(struct page *page, int numpages, int enable)
+{
+	/* This function does nothing, all work is done via poison pages */
+}
+#endif
-- 
2.6.3

[toc] | [next] | [standalone]

#1343756 — Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option

FromJianyu Zhan <nasa4836@gmail.com>
Date2016-02-26 04:00 +0100
SubjectRe: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option
Message-ID<r6gNk-4Yk-5@gated-at.bofh.it>
In reply to#1342521
On Thu, Feb 25, 2016 at 7:35 AM, Kees Cook <keescook@chromium.org> wrote:
>  config PAGE_POISONING
> -       bool
> +       bool "Poison pages after freeing"
> +       select PAGE_EXTENSION
> +       select PAGE_POISONING_NO_SANITY if HIBERNATION
> +       ---help---
> +         Fill the pages with poison patterns after free_pages() and verify
> +         the patterns before alloc_pages. The filling of the memory helps
> +         reduce the risk of information leaks from freed data. This does
> +         have a potential performance impact.
> +
> +         If unsure, say N
> +

I would suggest that you add some wording in the help text to clarify
that what "poisoning"
means here is not the same as that in "HWPoison".

The previous one is pattern padding, while the latter one is just
nomenclature borrowed from
Intel for memory failure.

> +config PAGE_POISONING_NO_SANITY
> +       depends on PAGE_POISONING
> +       bool "Only poison, don't sanity check"
> +       ---help---
> +          Skip the sanity checking on alloc, only fill the pages with
> +          poison on free. This reduces some of the overhead of the
> +          poisoning feature.
> +
> +          If you are only interested in sanitization, say Y. Otherwise
> +          say N.
> diff --git a/mm/Makefile b/mm/Makefile
> index fb1a7948c107..ec59c071b4f9 100644
> --- a/mm/Makefile
> +++ b/mm/Makefile
> @@ -13,7 +13,6 @@ KCOV_INSTRUMENT_slob.o := n
>  KCOV_INSTRUMENT_slab.o := n
>  KCOV_INSTRUMENT_slub.o := n
>  KCOV_INSTRUMENT_page_alloc.o := n
> -KCOV_INSTRUMENT_debug-pagealloc.o := n
>  KCOV_INSTRUMENT_kmemleak.o := n
>  KCOV_INSTRUMENT_kmemcheck.o := n
>  KCOV_INSTRUMENT_memcontrol.o := n
> @@ -63,9 +62,6 @@ obj-$(CONFIG_SPARSEMEM_VMEMMAP) += sparse-vmemmap.o
>  obj-$(CONFIG_SLOB) += slob.o
>  obj-$(CONFIG_MMU_NOTIFIER) += mmu_notifier.o
>  obj-$(CONFIG_KSM) += ksm.o
> -ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
> -       obj-$(CONFIG_DEBUG_PAGEALLOC) += debug-pagealloc.o
> -endif
>  obj-$(CONFIG_PAGE_POISONING) += page_poison.o
>  obj-$(CONFIG_SLAB) += slab.o
>  obj-$(CONFIG_SLUB) += slub.o
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index a34c359d8e81..0bdb3cfd83b5 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -1026,6 +1026,7 @@ static bool free_pages_prepare(struct page *page, unsigned int order)
>                                            PAGE_SIZE << order);
>         }
>         arch_free_page(page, order);
> +       kernel_poison_pages(page, 1 << order, 0);
>         kernel_map_pages(page, 1 << order, 0);
>
>         return true;
> @@ -1497,6 +1498,7 @@ static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
>
>         arch_alloc_page(page, order);
>         kernel_map_pages(page, 1 << order, 1);
> +       kernel_poison_pages(page, 1 << order, 1);
>         kasan_alloc_pages(page, order);
>
>         if (gfp_flags & __GFP_ZERO)
> diff --git a/mm/page_poison.c b/mm/page_poison.c
> index 92ead727b8f0..884a6f854432 100644
> --- a/mm/page_poison.c
> +++ b/mm/page_poison.c
> @@ -80,7 +80,7 @@ static void poison_page(struct page *page)
>         kunmap_atomic(addr);
>  }
>
> -void poison_pages(struct page *page, int n)
> +static void poison_pages(struct page *page, int n)
>  {
>         int i;
>
> @@ -101,6 +101,9 @@ static void check_poison_mem(unsigned char *mem, size_t bytes)
>         unsigned char *start;
>         unsigned char *end;
>
> +       if (IS_ENABLED(CONFIG_PAGE_POISONING_NO_SANITY))
> +               return;
> +
>         start = memchr_inv(mem, PAGE_POISON, bytes);
>         if (!start)
>                 return;
> @@ -113,9 +116,9 @@ static void check_poison_mem(unsigned char *mem, size_t bytes)
>         if (!__ratelimit(&ratelimit))
>                 return;
>         else if (start == end && single_bit_flip(*start, PAGE_POISON))
> -               printk(KERN_ERR "pagealloc: single bit error\n");
> +               pr_err("pagealloc: single bit error\n");
>         else
> -               printk(KERN_ERR "pagealloc: memory corruption\n");
> +               pr_err("pagealloc: memory corruption\n");
>
>         print_hex_dump(KERN_ERR, "", DUMP_PREFIX_ADDRESS, 16, 1, start,
>                         end - start + 1, 1);
> @@ -135,10 +138,28 @@ static void unpoison_page(struct page *page)
>         kunmap_atomic(addr);
>  }
>
> -void unpoison_pages(struct page *page, int n)
> +static void unpoison_pages(struct page *page, int n)
>  {
>         int i;
>
>         for (i = 0; i < n; i++)
>                 unpoison_page(page + i);
>  }
> +
> +void kernel_poison_pages(struct page *page, int numpages, int enable)
> +{
> +       if (!page_poisoning_enabled())
> +               return;
> +
> +       if (enable)
> +               unpoison_pages(page, numpages);
> +       else
> +               poison_pages(page, numpages);
> +}
> +
> +#ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
> +void __kernel_map_pages(struct page *page, int numpages, int enable)
> +{
> +       /* This function does nothing, all work is done via poison pages */
> +}
> +#endif

IMHO,  kernel_map_pages is originally incorporated for debugging page
allocation.
And latter for archs that do not support arch-specific page poisoning,
a software poisoning
method was used.

So I think it is not appropriate to use two interfaces in the alloc/free hooks.

The kernel_poison_pages actually should be an implementation detail
and should be hided
in the kernel_map_pages interface.


Thanks,
Jianyu Zhan

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

#1343779 — Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option

FromLaura Abbott <labbott@redhat.com>
Date2016-02-26 05:50 +0100
SubjectRe: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option
Message-ID<r6ivM-6n1-3@gated-at.bofh.it>
In reply to#1343756
On 02/25/2016 06:53 PM, Jianyu Zhan wrote:
> On Thu, Feb 25, 2016 at 7:35 AM, Kees Cook <keescook@chromium.org> wrote:
>>   config PAGE_POISONING
>> -       bool
>> +       bool "Poison pages after freeing"
>> +       select PAGE_EXTENSION
>> +       select PAGE_POISONING_NO_SANITY if HIBERNATION
>> +       ---help---
>> +         Fill the pages with poison patterns after free_pages() and verify
>> +         the patterns before alloc_pages. The filling of the memory helps
>> +         reduce the risk of information leaks from freed data. This does
>> +         have a potential performance impact.
>> +
>> +         If unsure, say N
>> +
>
> I would suggest that you add some wording in the help text to clarify
> that what "poisoning"
> means here is not the same as that in "HWPoison".
>
> The previous one is pattern padding, while the latter one is just
> nomenclature borrowed from
> Intel for memory failure.
>

Do you have some suggestion on wording here? I'm not sure what else to
say besides poison patterns to differentiate from hardware poison.
  
>> +config PAGE_POISONING_NO_SANITY
>> +       depends on PAGE_POISONING
>> +       bool "Only poison, don't sanity check"
>> +       ---help---
>> +          Skip the sanity checking on alloc, only fill the pages with
>> +          poison on free. This reduces some of the overhead of the
>> +          poisoning feature.
>> +
>> +          If you are only interested in sanitization, say Y. Otherwise
>> +          say N.
>> diff --git a/mm/Makefile b/mm/Makefile
>> index fb1a7948c107..ec59c071b4f9 100644
>> --- a/mm/Makefile
>> +++ b/mm/Makefile
>> @@ -13,7 +13,6 @@ KCOV_INSTRUMENT_slob.o := n
>>   KCOV_INSTRUMENT_slab.o := n
>>   KCOV_INSTRUMENT_slub.o := n
>>   KCOV_INSTRUMENT_page_alloc.o := n
>> -KCOV_INSTRUMENT_debug-pagealloc.o := n
>>   KCOV_INSTRUMENT_kmemleak.o := n
>>   KCOV_INSTRUMENT_kmemcheck.o := n
>>   KCOV_INSTRUMENT_memcontrol.o := n
>> @@ -63,9 +62,6 @@ obj-$(CONFIG_SPARSEMEM_VMEMMAP) += sparse-vmemmap.o
>>   obj-$(CONFIG_SLOB) += slob.o
>>   obj-$(CONFIG_MMU_NOTIFIER) += mmu_notifier.o
>>   obj-$(CONFIG_KSM) += ksm.o
>> -ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
>> -       obj-$(CONFIG_DEBUG_PAGEALLOC) += debug-pagealloc.o
>> -endif
>>   obj-$(CONFIG_PAGE_POISONING) += page_poison.o
>>   obj-$(CONFIG_SLAB) += slab.o
>>   obj-$(CONFIG_SLUB) += slub.o
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index a34c359d8e81..0bdb3cfd83b5 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -1026,6 +1026,7 @@ static bool free_pages_prepare(struct page *page, unsigned int order)
>>                                             PAGE_SIZE << order);
>>          }
>>          arch_free_page(page, order);
>> +       kernel_poison_pages(page, 1 << order, 0);
>>          kernel_map_pages(page, 1 << order, 0);
>>
>>          return true;
>> @@ -1497,6 +1498,7 @@ static int prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
>>
>>          arch_alloc_page(page, order);
>>          kernel_map_pages(page, 1 << order, 1);
>> +       kernel_poison_pages(page, 1 << order, 1);
>>          kasan_alloc_pages(page, order);
>>
>>          if (gfp_flags & __GFP_ZERO)
>> diff --git a/mm/page_poison.c b/mm/page_poison.c
>> index 92ead727b8f0..884a6f854432 100644
>> --- a/mm/page_poison.c
>> +++ b/mm/page_poison.c
>> @@ -80,7 +80,7 @@ static void poison_page(struct page *page)
>>          kunmap_atomic(addr);
>>   }
>>
>> -void poison_pages(struct page *page, int n)
>> +static void poison_pages(struct page *page, int n)
>>   {
>>          int i;
>>
>> @@ -101,6 +101,9 @@ static void check_poison_mem(unsigned char *mem, size_t bytes)
>>          unsigned char *start;
>>          unsigned char *end;
>>
>> +       if (IS_ENABLED(CONFIG_PAGE_POISONING_NO_SANITY))
>> +               return;
>> +
>>          start = memchr_inv(mem, PAGE_POISON, bytes);
>>          if (!start)
>>                  return;
>> @@ -113,9 +116,9 @@ static void check_poison_mem(unsigned char *mem, size_t bytes)
>>          if (!__ratelimit(&ratelimit))
>>                  return;
>>          else if (start == end && single_bit_flip(*start, PAGE_POISON))
>> -               printk(KERN_ERR "pagealloc: single bit error\n");
>> +               pr_err("pagealloc: single bit error\n");
>>          else
>> -               printk(KERN_ERR "pagealloc: memory corruption\n");
>> +               pr_err("pagealloc: memory corruption\n");
>>
>>          print_hex_dump(KERN_ERR, "", DUMP_PREFIX_ADDRESS, 16, 1, start,
>>                          end - start + 1, 1);
>> @@ -135,10 +138,28 @@ static void unpoison_page(struct page *page)
>>          kunmap_atomic(addr);
>>   }
>>
>> -void unpoison_pages(struct page *page, int n)
>> +static void unpoison_pages(struct page *page, int n)
>>   {
>>          int i;
>>
>>          for (i = 0; i < n; i++)
>>                  unpoison_page(page + i);
>>   }
>> +
>> +void kernel_poison_pages(struct page *page, int numpages, int enable)
>> +{
>> +       if (!page_poisoning_enabled())
>> +               return;
>> +
>> +       if (enable)
>> +               unpoison_pages(page, numpages);
>> +       else
>> +               poison_pages(page, numpages);
>> +}
>> +
>> +#ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
>> +void __kernel_map_pages(struct page *page, int numpages, int enable)
>> +{
>> +       /* This function does nothing, all work is done via poison pages */
>> +}
>> +#endif
>
> IMHO,  kernel_map_pages is originally incorporated for debugging page
> allocation.
> And latter for archs that do not support arch-specific page poisoning,
> a software poisoning
> method was used.
>
> So I think it is not appropriate to use two interfaces in the alloc/free hooks.
>
> The kernel_poison_pages actually should be an implementation detail
> and should be hided
> in the kernel_map_pages interface.
>

We want to have the poisoning independent of anything that kernel_map_pages
does. It was originally added for software poisoning for arches that
didn't have the full ARCH_SUPPORTS_DEBUG_PAGEALLOC support but there's
nothing that specifically ties it to mapping. It's beneficial even when
we aren't mapping/unmapping the pages so putting it in kernel_map_pages
would defeat what we're trying to accomplish here.
  
>
> Thanks,
> Jianyu Zhan
>

Thanks,
Laura

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

#1343789 — Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option

FromJianyu Zhan <nasa4836@gmail.com>
Date2016-02-26 06:40 +0100
SubjectRe: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option
Message-ID<r6ji9-6W8-7@gated-at.bofh.it>
In reply to#1343779
On Fri, Feb 26, 2016 at 12:45 PM, Laura Abbott <labbott@redhat.com> wrote:
> Do you have some suggestion on wording here? I'm not sure what else to
> say besides poison patterns to differentiate from hardware poison.
>


Is the below wording OK?


config PAGE_POISONING
        bool
        bool "Poison pages after freeing"
        select PAGE_EXTENSION
        select PAGE_POISONING_NO_SANITY if HIBERNATION
        ---help---
             Fill the pages with poison patterns after free_pages() and verify
             the patterns before alloc_pages. The filling of the memory helps
             reduce the risk of information leaks from freed data. This does
             have a potential performance impact.

             Note that "poison" here is not the same thing as that in "HWPoison"
             for CONFIG_MEMORY_FAILURE, in which "poison" is just a nomenclature
             borrowed from Intel , for the processor support for
"poisoned" memory, an
             adaptive method for flagging and recovering from memory errors

>
>>>
>>> +config PAGE_POISONING_NO_SANITY
>>> +       depends on PAGE_POISONING
>>> +       bool "Only poison, don't sanity check"
>>> +       ---help---
>>> +          Skip the sanity checking on alloc, only fill the pages with
>>> +          poison on free. This reduces some of the overhead of the
>>> +          poisoning feature.
>>> +
>>> +          If you are only interested in sanitization, say Y. Otherwise
>>> +          say N.
>>> diff --git a/mm/Makefile b/mm/Makefile
>>> index fb1a7948c107..ec59c071b4f9 100644
>>> --- a/mm/Makefile
>>> +++ b/mm/Makefile
>>> @@ -13,7 +13,6 @@ KCOV_INSTRUMENT_slob.o := n
>>>   KCOV_INSTRUMENT_slab.o := n
>>>   KCOV_INSTRUMENT_slub.o := n
>>>   KCOV_INSTRUMENT_page_alloc.o := n
>>> -KCOV_INSTRUMENT_debug-pagealloc.o := n
>>>   KCOV_INSTRUMENT_kmemleak.o := n
>>>   KCOV_INSTRUMENT_kmemcheck.o := n
>>>   KCOV_INSTRUMENT_memcontrol.o := n
>>> @@ -63,9 +62,6 @@ obj-$(CONFIG_SPARSEMEM_VMEMMAP) += sparse-vmemmap.o
>>>   obj-$(CONFIG_SLOB) += slob.o
>>>   obj-$(CONFIG_MMU_NOTIFIER) += mmu_notifier.o
>>>   obj-$(CONFIG_KSM) += ksm.o
>>> -ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
>>> -       obj-$(CONFIG_DEBUG_PAGEALLOC) += debug-pagealloc.o
>>> -endif
>>>   obj-$(CONFIG_PAGE_POISONING) += page_poison.o
>>>   obj-$(CONFIG_SLAB) += slab.o
>>>   obj-$(CONFIG_SLUB) += slub.o
>>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>>> index a34c359d8e81..0bdb3cfd83b5 100644
>>> --- a/mm/page_alloc.c
>>> +++ b/mm/page_alloc.c
>>> @@ -1026,6 +1026,7 @@ static bool free_pages_prepare(struct page *page,
>>> unsigned int order)
>>>                                             PAGE_SIZE << order);
>>>          }
>>>          arch_free_page(page, order);
>>> +       kernel_poison_pages(page, 1 << order, 0);
>>>          kernel_map_pages(page, 1 << order, 0);
>>>
>>>          return true;
>>> @@ -1497,6 +1498,7 @@ static int prep_new_page(struct page *page,
>>> unsigned int order, gfp_t gfp_flags,
>>>
>>>          arch_alloc_page(page, order);
>>>          kernel_map_pages(page, 1 << order, 1);
>>> +       kernel_poison_pages(page, 1 << order, 1);
>>>          kasan_alloc_pages(page, order);
>>>
>>>          if (gfp_flags & __GFP_ZERO)
>>> diff --git a/mm/page_poison.c b/mm/page_poison.c
>>> index 92ead727b8f0..884a6f854432 100644
>>> --- a/mm/page_poison.c
>>> +++ b/mm/page_poison.c
>>> @@ -80,7 +80,7 @@ static void poison_page(struct page *page)
>>>          kunmap_atomic(addr);
>>>   }
>>>
>>> -void poison_pages(struct page *page, int n)
>>> +static void poison_pages(struct page *page, int n)
>>>   {
>>>          int i;
>>>
>>> @@ -101,6 +101,9 @@ static void check_poison_mem(unsigned char *mem,
>>> size_t bytes)
>>>          unsigned char *start;
>>>          unsigned char *end;
>>>
>>> +       if (IS_ENABLED(CONFIG_PAGE_POISONING_NO_SANITY))
>>> +               return;
>>> +
>>>          start = memchr_inv(mem, PAGE_POISON, bytes);
>>>          if (!start)
>>>                  return;
>>> @@ -113,9 +116,9 @@ static void check_poison_mem(unsigned char *mem,
>>> size_t bytes)
>>>          if (!__ratelimit(&ratelimit))
>>>                  return;
>>>          else if (start == end && single_bit_flip(*start, PAGE_POISON))
>>> -               printk(KERN_ERR "pagealloc: single bit error\n");
>>> +               pr_err("pagealloc: single bit error\n");
>>>          else
>>> -               printk(KERN_ERR "pagealloc: memory corruption\n");
>>> +               pr_err("pagealloc: memory corruption\n");
>>>
>>>          print_hex_dump(KERN_ERR, "", DUMP_PREFIX_ADDRESS, 16, 1, start,
>>>                          end - start + 1, 1);
>>> @@ -135,10 +138,28 @@ static void unpoison_page(struct page *page)
>>>          kunmap_atomic(addr);
>>>   }
>>>
>>> -void unpoison_pages(struct page *page, int n)
>>> +static void unpoison_pages(struct page *page, int n)
>>>   {
>>>          int i;
>>>
>>>          for (i = 0; i < n; i++)
>>>                  unpoison_page(page + i);
>>>   }
>>> +
>>> +void kernel_poison_pages(struct page *page, int numpages, int enable)
>>> +{
>>> +       if (!page_poisoning_enabled())
>>> +               return;
>>> +
>>> +       if (enable)
>>> +               unpoison_pages(page, numpages);
>>> +       else
>>> +               poison_pages(page, numpages);
>>> +}
>>> +
>>> +#ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
>>> +void __kernel_map_pages(struct page *page, int numpages, int enable)
>>> +{
>>> +       /* This function does nothing, all work is done via poison pages
>>> */
>>> +}
>>> +#endif
>>
>>
>> IMHO,  kernel_map_pages is originally incorporated for debugging page
>> allocation.
>> And latter for archs that do not support arch-specific page poisoning,
>> a software poisoning
>> method was used.
>>
>> So I think it is not appropriate to use two interfaces in the alloc/free
>> hooks.
>>
>> The kernel_poison_pages actually should be an implementation detail
>> and should be hided
>> in the kernel_map_pages interface.
>>
>
> We want to have the poisoning independent of anything that kernel_map_pages
> does. It was originally added for software poisoning for arches that
> didn't have the full ARCH_SUPPORTS_DEBUG_PAGEALLOC support but there's
> nothing that specifically ties it to mapping. It's beneficial even when
> we aren't mapping/unmapping the pages so putting it in kernel_map_pages
> would defeat what we're trying to accomplish here.
>

Ok, fair enough. If so,  I suggest you add this clarification into the
code, or as least, in
the changelog.


Thanks,
Jianyu Zhan

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

#1344774 — Re: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option

FromLaura Abbott <labbott@redhat.com>
Date2016-02-26 23:30 +0100
SubjectRe: [RFC][PATCH v3 1/2] mm/page_poison.c: Enable PAGE_POISONING as a separate option
Message-ID<r6z3B-1uq-57@gated-at.bofh.it>
In reply to#1343789
On 02/25/2016 09:34 PM, Jianyu Zhan wrote:
> On Fri, Feb 26, 2016 at 12:45 PM, Laura Abbott <labbott@redhat.com> wrote:
>> Do you have some suggestion on wording here? I'm not sure what else to
>> say besides poison patterns to differentiate from hardware poison.
>>
>
>
> Is the below wording OK?
>
>
> config PAGE_POISONING
>          bool
>          bool "Poison pages after freeing"
>          select PAGE_EXTENSION
>          select PAGE_POISONING_NO_SANITY if HIBERNATION
>          ---help---
>               Fill the pages with poison patterns after free_pages() and verify
>               the patterns before alloc_pages. The filling of the memory helps
>               reduce the risk of information leaks from freed data. This does
>               have a potential performance impact.
>
>               Note that "poison" here is not the same thing as that in "HWPoison"
>               for CONFIG_MEMORY_FAILURE, in which "poison" is just a nomenclature
>               borrowed from Intel , for the processor support for
> "poisoned" memory, an
>               adaptive method for flagging and recovering from memory errors
>

Okay, I see what you are getting at here. This sounds okay.
  
>>
>>>>
>>>> +config PAGE_POISONING_NO_SANITY
>>>> +       depends on PAGE_POISONING
>>>> +       bool "Only poison, don't sanity check"
>>>> +       ---help---
>>>> +          Skip the sanity checking on alloc, only fill the pages with
>>>> +          poison on free. This reduces some of the overhead of the
>>>> +          poisoning feature.
>>>> +
>>>> +          If you are only interested in sanitization, say Y. Otherwise
>>>> +          say N.
>>>> diff --git a/mm/Makefile b/mm/Makefile
>>>> index fb1a7948c107..ec59c071b4f9 100644
>>>> --- a/mm/Makefile
>>>> +++ b/mm/Makefile
>>>> @@ -13,7 +13,6 @@ KCOV_INSTRUMENT_slob.o := n
>>>>    KCOV_INSTRUMENT_slab.o := n
>>>>    KCOV_INSTRUMENT_slub.o := n
>>>>    KCOV_INSTRUMENT_page_alloc.o := n
>>>> -KCOV_INSTRUMENT_debug-pagealloc.o := n
>>>>    KCOV_INSTRUMENT_kmemleak.o := n
>>>>    KCOV_INSTRUMENT_kmemcheck.o := n
>>>>    KCOV_INSTRUMENT_memcontrol.o := n
>>>> @@ -63,9 +62,6 @@ obj-$(CONFIG_SPARSEMEM_VMEMMAP) += sparse-vmemmap.o
>>>>    obj-$(CONFIG_SLOB) += slob.o
>>>>    obj-$(CONFIG_MMU_NOTIFIER) += mmu_notifier.o
>>>>    obj-$(CONFIG_KSM) += ksm.o
>>>> -ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
>>>> -       obj-$(CONFIG_DEBUG_PAGEALLOC) += debug-pagealloc.o
>>>> -endif
>>>>    obj-$(CONFIG_PAGE_POISONING) += page_poison.o
>>>>    obj-$(CONFIG_SLAB) += slab.o
>>>>    obj-$(CONFIG_SLUB) += slub.o
>>>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>>>> index a34c359d8e81..0bdb3cfd83b5 100644
>>>> --- a/mm/page_alloc.c
>>>> +++ b/mm/page_alloc.c
>>>> @@ -1026,6 +1026,7 @@ static bool free_pages_prepare(struct page *page,
>>>> unsigned int order)
>>>>                                              PAGE_SIZE << order);
>>>>           }
>>>>           arch_free_page(page, order);
>>>> +       kernel_poison_pages(page, 1 << order, 0);
>>>>           kernel_map_pages(page, 1 << order, 0);
>>>>
>>>>           return true;
>>>> @@ -1497,6 +1498,7 @@ static int prep_new_page(struct page *page,
>>>> unsigned int order, gfp_t gfp_flags,
>>>>
>>>>           arch_alloc_page(page, order);
>>>>           kernel_map_pages(page, 1 << order, 1);
>>>> +       kernel_poison_pages(page, 1 << order, 1);
>>>>           kasan_alloc_pages(page, order);
>>>>
>>>>           if (gfp_flags & __GFP_ZERO)
>>>> diff --git a/mm/page_poison.c b/mm/page_poison.c
>>>> index 92ead727b8f0..884a6f854432 100644
>>>> --- a/mm/page_poison.c
>>>> +++ b/mm/page_poison.c
>>>> @@ -80,7 +80,7 @@ static void poison_page(struct page *page)
>>>>           kunmap_atomic(addr);
>>>>    }
>>>>
>>>> -void poison_pages(struct page *page, int n)
>>>> +static void poison_pages(struct page *page, int n)
>>>>    {
>>>>           int i;
>>>>
>>>> @@ -101,6 +101,9 @@ static void check_poison_mem(unsigned char *mem,
>>>> size_t bytes)
>>>>           unsigned char *start;
>>>>           unsigned char *end;
>>>>
>>>> +       if (IS_ENABLED(CONFIG_PAGE_POISONING_NO_SANITY))
>>>> +               return;
>>>> +
>>>>           start = memchr_inv(mem, PAGE_POISON, bytes);
>>>>           if (!start)
>>>>                   return;
>>>> @@ -113,9 +116,9 @@ static void check_poison_mem(unsigned char *mem,
>>>> size_t bytes)
>>>>           if (!__ratelimit(&ratelimit))
>>>>                   return;
>>>>           else if (start == end && single_bit_flip(*start, PAGE_POISON))
>>>> -               printk(KERN_ERR "pagealloc: single bit error\n");
>>>> +               pr_err("pagealloc: single bit error\n");
>>>>           else
>>>> -               printk(KERN_ERR "pagealloc: memory corruption\n");
>>>> +               pr_err("pagealloc: memory corruption\n");
>>>>
>>>>           print_hex_dump(KERN_ERR, "", DUMP_PREFIX_ADDRESS, 16, 1, start,
>>>>                           end - start + 1, 1);
>>>> @@ -135,10 +138,28 @@ static void unpoison_page(struct page *page)
>>>>           kunmap_atomic(addr);
>>>>    }
>>>>
>>>> -void unpoison_pages(struct page *page, int n)
>>>> +static void unpoison_pages(struct page *page, int n)
>>>>    {
>>>>           int i;
>>>>
>>>>           for (i = 0; i < n; i++)
>>>>                   unpoison_page(page + i);
>>>>    }
>>>> +
>>>> +void kernel_poison_pages(struct page *page, int numpages, int enable)
>>>> +{
>>>> +       if (!page_poisoning_enabled())
>>>> +               return;
>>>> +
>>>> +       if (enable)
>>>> +               unpoison_pages(page, numpages);
>>>> +       else
>>>> +               poison_pages(page, numpages);
>>>> +}
>>>> +
>>>> +#ifndef CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC
>>>> +void __kernel_map_pages(struct page *page, int numpages, int enable)
>>>> +{
>>>> +       /* This function does nothing, all work is done via poison pages
>>>> */
>>>> +}
>>>> +#endif
>>>
>>>
>>> IMHO,  kernel_map_pages is originally incorporated for debugging page
>>> allocation.
>>> And latter for archs that do not support arch-specific page poisoning,
>>> a software poisoning
>>> method was used.
>>>
>>> So I think it is not appropriate to use two interfaces in the alloc/free
>>> hooks.
>>>
>>> The kernel_poison_pages actually should be an implementation detail
>>> and should be hided
>>> in the kernel_map_pages interface.
>>>
>>
>> We want to have the poisoning independent of anything that kernel_map_pages
>> does. It was originally added for software poisoning for arches that
>> didn't have the full ARCH_SUPPORTS_DEBUG_PAGEALLOC support but there's
>> nothing that specifically ties it to mapping. It's beneficial even when
>> we aren't mapping/unmapping the pages so putting it in kernel_map_pages
>> would defeat what we're trying to accomplish here.
>>
>
> Ok, fair enough. If so,  I suggest you add this clarification into the
> code, or as least, in
> the changelog.

Sounds fine.

>
>
> Thanks,
> Jianyu Zhan
>

Thanks,
Laura

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web