[PATCH v5 0/2] x86_64: fix KASan shadow region page tables

[PATCH v5 0/2] x86_64: fix KASan shadow region page tables

[Alexander Popov]

This patch series:
  - fixes kernel halt caused by incorrect physical addresses
      in KASan shadow region page tables in case of non-zero phys_base;
  - consolidates early KASan initialization.

Changes from v2:
  - move KASan shadow region page tables to BSS;
  - use __PAGE_KERNEL flags for describing kasan_zero_page in kasan_zero_pte.

Changes from v3:
  - improve commit message.

Changes from v4:
  - add Andrey's patch which removes faulty clear_page(init_level4_pgt);
  - call kasan_map_early_shadow() for early_level4_pgt and init_level4_pgt
      in kasan_early_init().

Alexander Popov (1):
  x86_64: fix KASan shadow region page tables

Andrey Ryabinin (1):
  x86_64: remove not needed clear_page for init_level4_page

 arch/x86/include/asm/kasan.h |  8 ++------
 arch/x86/kernel/head64.c     | 12 ++++++------
 arch/x86/kernel/head_64.S    | 29 -----------------------------
 arch/x86/mm/kasan_init_64.c  | 36 ++++++++++++++++++++++++++++++++++--
 4 files changed, 42 insertions(+), 43 deletions(-)

-- 
1.9.1

[PATCH v5 1/2] x86_64: remove not needed clear_page for init_level4_page

[Alexander Popov]

From: Andrey Ryabinin <a.ryabinin@samsung.com>

Commit 8170e6bed465 ("x86, 64bit: Use a #PF handler to materialize
early mappings on demand") introduced clear_page(init_level4_pgt);
call in x86_64_start_kernel(). However, this clear_page is useless
because init_level4_page already filled with zeroes in head_64.S

Commit message in 8170e6bed465 says that this clear_page() was
dropped in v7, but it accidentally reappeared in later versions
of that patchset.

Signed-off-by: Andrey Ryabinin <a.ryabinin@samsung.com>
---
 arch/x86/kernel/head64.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index 5a46681..6a6eefd 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -177,7 +177,6 @@ asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
 	 */
 	load_ucode_bsp();
 
-	clear_page(init_level4_pgt);
 	/* set init_level4_pgt kernel high mapping*/
 	init_level4_pgt[511] = early_level4_pgt[511];
 
-- 
1.9.1

[PATCH v5 2/2] x86_64: fix KASan shadow region page tables

[Alexander Popov]

Physical addresses in KASan shadow region page tables need fixup similarly
to the other page tables. Current code doesn't do it which causes
kernel halt if phys_base is not zero.
So let's initialize KASan shadow region page tables in kasan_early_init()
using __pa_nodebug() which considers phys_base.

Signed-off-by: Alexander Popov <alpopov@ptsecurity.com>
---
 arch/x86/include/asm/kasan.h |  8 ++------
 arch/x86/kernel/head64.c     | 11 ++++++-----
 arch/x86/kernel/head_64.S    | 29 -----------------------------
 arch/x86/mm/kasan_init_64.c  | 36 ++++++++++++++++++++++++++++++++++--
 4 files changed, 42 insertions(+), 42 deletions(-)

diff --git a/arch/x86/include/asm/kasan.h b/arch/x86/include/asm/kasan.h
index 8b22422..74a2a8d 100644
--- a/arch/x86/include/asm/kasan.h
+++ b/arch/x86/include/asm/kasan.h
@@ -14,15 +14,11 @@
 
 #ifndef __ASSEMBLY__
 
-extern pte_t kasan_zero_pte[];
-extern pte_t kasan_zero_pmd[];
-extern pte_t kasan_zero_pud[];
-
 #ifdef CONFIG_KASAN
-void __init kasan_map_early_shadow(pgd_t *pgd);
+void __init kasan_early_init(void);
 void __init kasan_init(void);
 #else
-static inline void kasan_map_early_shadow(pgd_t *pgd) { }
+static inline void kasan_early_init(void) { }
 static inline void kasan_init(void) { }
 #endif
 
diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index 6a6eefd..74da193 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -161,11 +161,14 @@ asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
 	/* Kill off the identity-map trampoline */
 	reset_early_page_tables();
 
-	kasan_map_early_shadow(early_level4_pgt);
-
-	/* clear bss before set_intr_gate with early_idt_handler */
+	/*
+	 * Clear bss before kasan_early_init and set_intr_gate
+	 * with early_idt_handler
+	 */
 	clear_bss();
 
+	kasan_early_init();
+
 	for (i = 0; i < NUM_EXCEPTION_VECTORS; i++)
 		set_intr_gate(i, early_idt_handler_array[i]);
 	load_idt((const struct desc_ptr *)&idt_descr);
@@ -180,8 +183,6 @@ asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
 	/* set init_level4_pgt kernel high mapping*/
 	init_level4_pgt[511] = early_level4_pgt[511];
 
-	kasan_map_early_shadow(init_level4_pgt);
-
 	x86_64_start_reservations(real_mode_data);
 }
 
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index df7e780..7e5da2c 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -516,38 +516,9 @@ ENTRY(phys_base)
 	/* This must match the first entry in level2_kernel_pgt */
 	.quad   0x0000000000000000
 
-#ifdef CONFIG_KASAN
-#define FILL(VAL, COUNT)				\
-	.rept (COUNT) ;					\
-	.quad	(VAL) ;					\
-	.endr
-
-NEXT_PAGE(kasan_zero_pte)
-	FILL(kasan_zero_page - __START_KERNEL_map + _KERNPG_TABLE, 512)
-NEXT_PAGE(kasan_zero_pmd)
-	FILL(kasan_zero_pte - __START_KERNEL_map + _KERNPG_TABLE, 512)
-NEXT_PAGE(kasan_zero_pud)
-	FILL(kasan_zero_pmd - __START_KERNEL_map + _KERNPG_TABLE, 512)
-
-#undef FILL
-#endif
-
-
 #include "../../x86/xen/xen-head.S"
 	
 	__PAGE_ALIGNED_BSS
 NEXT_PAGE(empty_zero_page)
 	.skip PAGE_SIZE
 
-#ifdef CONFIG_KASAN
-/*
- * This page used as early shadow. We don't use empty_zero_page
- * at early stages, stack instrumentation could write some garbage
- * to this page.
- * Latter we reuse it as zero shadow for large ranges of memory
- * that allowed to access, but not instrumented by kasan
- * (vmalloc/vmemmap ...).
- */
-NEXT_PAGE(kasan_zero_page)
-	.skip PAGE_SIZE
-#endif
diff --git a/arch/x86/mm/kasan_init_64.c b/arch/x86/mm/kasan_init_64.c
index 4860906..0e4a05f 100644
--- a/arch/x86/mm/kasan_init_64.c
+++ b/arch/x86/mm/kasan_init_64.c
@@ -11,7 +11,19 @@
 extern pgd_t early_level4_pgt[PTRS_PER_PGD];
 extern struct range pfn_mapped[E820_X_MAX];
 
-extern unsigned char kasan_zero_page[PAGE_SIZE];
+static pud_t kasan_zero_pud[PTRS_PER_PUD] __page_aligned_bss;
+static pmd_t kasan_zero_pmd[PTRS_PER_PMD] __page_aligned_bss;
+static pte_t kasan_zero_pte[PTRS_PER_PTE] __page_aligned_bss;
+
+/*
+ * This page used as early shadow. We don't use empty_zero_page
+ * at early stages, stack instrumentation could write some garbage
+ * to this page.
+ * Latter we reuse it as zero shadow for large ranges of memory
+ * that allowed to access, but not instrumented by kasan
+ * (vmalloc/vmemmap ...).
+ */
+static unsigned char kasan_zero_page[PAGE_SIZE] __page_aligned_bss;
 
 static int __init map_range(struct range *range)
 {
@@ -36,7 +48,7 @@ static void __init clear_pgds(unsigned long start,
 		pgd_clear(pgd_offset_k(start));
 }
 
-void __init kasan_map_early_shadow(pgd_t *pgd)
+static void __init kasan_map_early_shadow(pgd_t *pgd)
 {
 	int i;
 	unsigned long start = KASAN_SHADOW_START;
@@ -166,6 +178,26 @@ static struct notifier_block kasan_die_notifier = {
 };
 #endif
 
+void __init kasan_early_init(void)
+{
+	int i;
+	pteval_t pte_val = __pa_nodebug(kasan_zero_page) | __PAGE_KERNEL;
+	pmdval_t pmd_val = __pa_nodebug(kasan_zero_pte) | _KERNPG_TABLE;
+	pudval_t pud_val = __pa_nodebug(kasan_zero_pmd) | _KERNPG_TABLE;
+
+	for (i = 0; i < PTRS_PER_PTE; i++)
+		kasan_zero_pte[i] = __pte(pte_val);
+
+	for (i = 0; i < PTRS_PER_PMD; i++)
+		kasan_zero_pmd[i] = __pmd(pmd_val);
+
+	for (i = 0; i < PTRS_PER_PUD; i++)
+		kasan_zero_pud[i] = __pud(pud_val);
+
+	kasan_map_early_shadow(early_level4_pgt);
+	kasan_map_early_shadow(init_level4_pgt);
+}
+
 void __init kasan_init(void)
 {
 	int i;
-- 
1.9.1

Re: [PATCH v5 2/2] x86_64: fix KASan shadow region page tables

[Andrey Ryabinin]

2015-06-09 12:42 GMT+03:00 Alexander Popov <alpopov@ptsecurity.com>:
> Physical addresses in KASan shadow region page tables need fixup similarly
> to the other page tables. Current code doesn't do it which causes
> kernel halt if phys_base is not zero.
> So let's initialize KASan shadow region page tables in kasan_early_init()
> using __pa_nodebug() which considers phys_base.
>
> Signed-off-by: Alexander Popov <alpopov@ptsecurity.com>

Acked-by: Andrey Ryabinin <a.ryabinin@samsung.com>

Re: [PATCH v5 1/2] x86_64: remove not needed clear_page for init_level4_page

[Borislav Petkov]

On Tue, Jun 09, 2015 at 12:42:00PM +0300, Alexander Popov wrote:
> From: Andrey Ryabinin <a.ryabinin@samsung.com>
> 
> Commit 8170e6bed465 ("x86, 64bit: Use a #PF handler to materialize
> early mappings on demand") introduced clear_page(init_level4_pgt);
> call in x86_64_start_kernel(). However, this clear_page is useless
> because init_level4_page already filled with zeroes in head_64.S

I really doubt that, see below:

> Commit message in 8170e6bed465 says that this clear_page() was
> dropped in v7, but it accidentally reappeared in later versions
> of that patchset.
> 
> Signed-off-by: Andrey Ryabinin <a.ryabinin@samsung.com>
> ---
>  arch/x86/kernel/head64.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
> index 5a46681..6a6eefd 100644
> --- a/arch/x86/kernel/head64.c
> +++ b/arch/x86/kernel/head64.c
> @@ -177,7 +177,6 @@ asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
>  	 */
>  	load_ucode_bsp();
>  
> -	clear_page(init_level4_pgt);
>  	/* set init_level4_pgt kernel high mapping*/
>  	init_level4_pgt[511] = early_level4_pgt[511];

vmlinux has:

ffffffff81a00000 <init_level4_pgt>:
ffffffff81a00000:       63 10                   movslq (%rax),%edx
ffffffff81a00002:       a0 01 00 00 00 00 00    movabs 0x1,%al
ffffffff81a00009:       00 00 
        ...
ffffffff81a0087f:       00 63 10                add    %ah,0x10(%rbx)
ffffffff81a00882:       a0 01 00 00 00 00 00    movabs 0x1,%al
ffffffff81a00889:       00 00 
        ...
ffffffff81a00ff7:       00 67 30                add    %ah,0x30(%rdi)
ffffffff81a00ffa:       a0 01 00 00 00 00 63    movabs 0xa020630000000001,%al
ffffffff81a01001:       20 a0 

ffffffff81a01000 <level3_ident_pgt>:
ffffffff81a01000:       63 20                   movslq (%rax),%esp
ffffffff81a01002:       a0 01 00 00 00 00 00    movabs 0x1,%al
ffffffff81a01009:       00 00 
        ...

Booting a kvm quest and halting it before the clear_page:

---
        /*
         * Load microcode early on BSP.
         */
        load_ucode_bsp();

//      clear_page(init_level4_pgt);

        while(1)
                cpu_relax();

	/* set init_level4_pgt kernel high mapping*/
	init_level4_pgt[511] = early_level4_pgt[511];
---

and dumping the memory at ffffffff81a00000 gives:

---
00000000  63 10 a0 01 00 00 00 00  00 00 00 00 00 00 00 00  |c...............|
00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000880  63 10 a0 01 00 00 00 00  00 00 00 00 00 00 00 00  |c...............|
00000890  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000ff0  00 00 00 00 00 00 00 00  67 30 a0 01 00 00 00 00  |........g0......|
00001000
---

These are basically the PTE entries it gets for the CONFIG_XEN case
which we clear because we're on baremetal.

Now my hunch is that those entries get overwritten later but I wouldn't
want to debug any strange bugs from leftovers in init_level4_pgt so
having the clear_page() is actually a good thing.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH v5 1/2] x86_64: remove not needed clear_page for init_level4_page

[Borislav Petkov]

On Tue, Jun 16, 2015 at 01:16:32PM +0200, Borislav Petkov wrote:
> Now my hunch is that those entries get overwritten later but I wouldn't
> want to debug any strange bugs from leftovers in init_level4_pgt so
> having the clear_page() is actually a good thing.

So I guess we can do that:

#ifndef CONFIG_KASAN
        clear_page(init_level4_pgt);
#endif

Hmm...

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH v5 1/2] x86_64: remove not needed clear_page for init_level4_page

[Andrey Ryabinin]

On 06/16/2015 02:34 PM, Borislav Petkov wrote:
> On Tue, Jun 16, 2015 at 01:16:32PM +0200, Borislav Petkov wrote:
>> Now my hunch is that those entries get overwritten later but I wouldn't
>> want to debug any strange bugs from leftovers in init_level4_pgt so
>> having the clear_page() is actually a good thing.
> 
> So I guess we can do that:
> 
> #ifndef CONFIG_KASAN
>         clear_page(init_level4_pgt);
> #endif
> 

Ugh..

Can't we just move clear_page(init_level4_pgt) higher? Before or right after clear_bss() (iow before kasan_early_init()).

Re: [PATCH v5 1/2] x86_64: remove not needed clear_page for init_level4_page

[Borislav Petkov]

On Tue, Jun 16, 2015 at 02:45:06PM +0300, Andrey Ryabinin wrote:
> Can't we just move clear_page(init_level4_pgt) higher? Before or right
> after clear_bss() (iow before kasan_early_init()).

That sounds much better. And I don't see anything depending on
init_level4_pgt before we clear it. And it boots here if I move it
before

	kasan_map_early_shadow(early_level4_pgt);

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH v5 1/2] x86_64: remove not needed clear_page for init_level4_page

[Alexander Popov]

On 16.06.2015 18:46, Borislav Petkov wrote:
> On Tue, Jun 16, 2015 at 02:45:06PM +0300, Andrey Ryabinin wrote:
>> Can't we just move clear_page(init_level4_pgt) higher? Before or right
>> after clear_bss() (iow before kasan_early_init()).
> 
> That sounds much better. And I don't see anything depending on
> init_level4_pgt before we clear it. And it boots here if I move it
> before
> 
> 	kasan_map_early_shadow(early_level4_pgt);

Thanks, Borislav and Andrey.
I'll return with version 6.

Best regards,
Alexander