If the kernel is configured with KASAN_EXTRA, the stack size is
increasted significantly due to enable this option will set
"-fstack-reuse" to "none" in GCC [1]. As the results, it could trigger
stack overrun quite often with 32k stack size compiled using GCC 8. For
example, this reproducer
https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/\
syscalls/madvise/madvise06.c
could trigger a "corrupted stack end detected inside scheduler" very
reliably with CONFIG_SCHED_STACK_END_CHECK enabled.
There are just too many functions that could have a large stack with
KASAN_EXTRA due to large local variables that have been called over and
over again without being able to reuse the stacks. Some noticiable ones
are,
size
7648 shrink_page_list
3584 xfs_rmap_convert
3312 migrate_page_move_mapping
3312 dev_ethtool
3200 migrate_misplaced_transhuge_page
3168 copy_process
There are other 49 functions are over 2k in size while compiling kernel
with "-Wframe-larger-than=" even with a related minimal config on this
machine. Hence, it is too much work to change Makefiles for each object
to compile without "-fsanitize-address-use-after-scope" individually.
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715#c23
Although there is a patch in GCC 9 to help the situation, GCC 9 probably
won't be released in a few months and then it probably take another
6-month to 1-year for all major distros to include it as a default.
Hence, the stack usage with KASAN_EXTRA can be revisited again in 2020
when GCC 9 is everywhere. Until then, this patch will help users avoid
stack overrun.
This has already been fixed for arm64 for the same reason via
6e8830674ea (arm64: kasan: Increase stack size for KASAN_EXTRA).
Signed-off-by: Qian Cai <[email protected]>
---
arch/x86/include/asm/page_64_types.h | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/arch/x86/include/asm/page_64_types.h b/arch/x86/include/asm/page_64_types.h
index 8f657286d599..0ce558a8150d 100644
--- a/arch/x86/include/asm/page_64_types.h
+++ b/arch/x86/include/asm/page_64_types.h
@@ -7,7 +7,11 @@
#endif
#ifdef CONFIG_KASAN
+#ifdef CONFIG_KASAN_EXTRA
+#define KASAN_STACK_ORDER 2
+#else
#define KASAN_STACK_ORDER 1
+#endif
#else
#define KASAN_STACK_ORDER 0
#endif
--
2.17.2 (Apple Git-113)
If the kernel is configured with KASAN_EXTRA, the stack size is
increasted significantly due to enable this option will set
"-fstack-reuse" to "none" in GCC [1]. As the results, it could trigger
stack overrun quite often with 32k stack size compiled using GCC 8. For
example, this reproducer
https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/\
syscalls/madvise/madvise06.c
could trigger a "corrupted stack end detected inside scheduler" very
reliably with CONFIG_SCHED_STACK_END_CHECK enabled.
There are just too many functions that could have a large stack with
KASAN_EXTRA due to large local variables that have been called over and
over again without being able to reuse the stacks. Some noticiable ones
are,
size
7648 shrink_page_list
3584 xfs_rmap_convert
3312 migrate_page_move_mapping
3312 dev_ethtool
3200 migrate_misplaced_transhuge_page
3168 copy_process
There are other 49 functions are over 2k in size while compiling kernel
with "-Wframe-larger-than=" even with a related minimal config on this
machine. Hence, it is too much work to change Makefiles for each object
to compile without "-fsanitize-address-use-after-scope" individually.
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715#c23
Although there is a patch in GCC 9 to help the situation, GCC 9 probably
won't be released in a few months and then it probably take another
6-month to 1-year for all major distros to include it as a default.
Hence, the stack usage with KASAN_EXTRA can be revisited again in 2020
when GCC 9 is everywhere. Until then, this patch will help users avoid
stack overrun.
This has already been fixed for arm64 for the same reason via
6e8830674ea (arm64: kasan: Increase stack size for KASAN_EXTRA).
Signed-off-by: Qian Cai <[email protected]>
---
arch/x86/include/asm/page_64_types.h | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/arch/x86/include/asm/page_64_types.h b/arch/x86/include/asm/page_64_types.h
index 8f657286d599..0ce558a8150d 100644
--- a/arch/x86/include/asm/page_64_types.h
+++ b/arch/x86/include/asm/page_64_types.h
@@ -7,7 +7,11 @@
#endif
#ifdef CONFIG_KASAN
+#ifdef CONFIG_KASAN_EXTRA
+#define KASAN_STACK_ORDER 2
+#else
#define KASAN_STACK_ORDER 1
+#endif
#else
#define KASAN_STACK_ORDER 0
#endif
--
2.17.2 (Apple Git-113)
On Wed, 9 Jan 2019 16:52:09 -0500 Qian Cai <[email protected]> wrote:
> If the kernel is configured with KASAN_EXTRA, the stack size is
> increasted significantly due to enable this option will set
> "-fstack-reuse" to "none" in GCC [1]. As the results, it could trigger
> stack overrun quite often with 32k stack size compiled using GCC 8. For
> example, this reproducer
>
> https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/\
> syscalls/madvise/madvise06.c
>
> could trigger a "corrupted stack end detected inside scheduler" very
> reliably with CONFIG_SCHED_STACK_END_CHECK enabled.
>
> There are just too many functions that could have a large stack with
> KASAN_EXTRA due to large local variables that have been called over and
> over again without being able to reuse the stacks. Some noticiable ones
> are,
>
> size
> 7648 shrink_page_list
> 3584 xfs_rmap_convert
> 3312 migrate_page_move_mapping
> 3312 dev_ethtool
> 3200 migrate_misplaced_transhuge_page
> 3168 copy_process
>
> There are other 49 functions are over 2k in size while compiling kernel
> with "-Wframe-larger-than=" even with a related minimal config on this
> machine. Hence, it is too much work to change Makefiles for each object
> to compile without "-fsanitize-address-use-after-scope" individually.
>
> [1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715#c23
>
> Although there is a patch in GCC 9 to help the situation, GCC 9 probably
> won't be released in a few months and then it probably take another
> 6-month to 1-year for all major distros to include it as a default.
> Hence, the stack usage with KASAN_EXTRA can be revisited again in 2020
> when GCC 9 is everywhere. Until then, this patch will help users avoid
> stack overrun.
>
> This has already been fixed for arm64 for the same reason via
> 6e8830674ea (arm64: kasan: Increase stack size for KASAN_EXTRA).
Sounds good.
> --- a/arch/x86/include/asm/page_64_types.h
> +++ b/arch/x86/include/asm/page_64_types.h
> @@ -7,7 +7,11 @@
> #endif
>
> #ifdef CONFIG_KASAN
> +#ifdef CONFIG_KASAN_EXTRA
> +#define KASAN_STACK_ORDER 2
> +#else
> #define KASAN_STACK_ORDER 1
> +#endif
> #else
> #define KASAN_STACK_ORDER 0
> #endif
I'm suspecting this logic could be performed in Kconfig, for all
architectures. Add a new always-defined CONFIG_KASAN_STACK_ORDER?
On 1/9/19 5:02 PM, Andrew Morton wrote:
>> --- a/arch/x86/include/asm/page_64_types.h
>> +++ b/arch/x86/include/asm/page_64_types.h
>> @@ -7,7 +7,11 @@
>> #endif
>>
>> #ifdef CONFIG_KASAN
>> +#ifdef CONFIG_KASAN_EXTRA
>> +#define KASAN_STACK_ORDER 2
>> +#else
>> #define KASAN_STACK_ORDER 1
>> +#endif
>> #else
>> #define KASAN_STACK_ORDER 0
>> #endif
>
> I'm suspecting this logic could be performed in Kconfig, for all
> architectures. Add a new always-defined CONFIG_KASAN_STACK_ORDER?
>
Sounds doable, but KASAN only support x86_64, arm64, s390 and xtensa so far. I
am not sure I care about the later two nor I have any way to test them if they
suffer the same problem.
I'll keep looking if more arches start to implement KASAN, and then might be a
good time to reduce all those code duplication.
On 1/10/19 12:52 AM, Qian Cai wrote:
> If the kernel is configured with KASAN_EXTRA, the stack size is
> increasted significantly due to enable this option will set
> "-fstack-reuse" to "none" in GCC [1]. As the results, it could trigger
> stack overrun quite often with 32k stack size compiled using GCC 8. For
> example, this reproducer
>
> https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/\
> syscalls/madvise/madvise06.c
>
> could trigger a "corrupted stack end detected inside scheduler" very
> reliably with CONFIG_SCHED_STACK_END_CHECK enabled.
>
> There are just too many functions that could have a large stack with
> KASAN_EXTRA due to large local variables that have been called over and
> over again without being able to reuse the stacks. Some noticiable ones
> are,
>
> size
> 7648 shrink_page_list
> 3584 xfs_rmap_convert
> 3312 migrate_page_move_mapping
> 3312 dev_ethtool
> 3200 migrate_misplaced_transhuge_page
> 3168 copy_process
>
> There are other 49 functions are over 2k in size while compiling kernel
> with "-Wframe-larger-than=" even with a related minimal config on this
> machine. Hence, it is too much work to change Makefiles for each object
> to compile without "-fsanitize-address-use-after-scope" individually.
>
> [1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715#c23
>
> Although there is a patch in GCC 9 to help the situation, GCC 9 probably
> won't be released in a few months and then it probably take another
> 6-month to 1-year for all major distros to include it as a default.
> Hence, the stack usage with KASAN_EXTRA can be revisited again in 2020
> when GCC 9 is everywhere. Until then, this patch will help users avoid
> stack overrun.
>
> This has already been fixed for arm64 for the same reason via
> 6e8830674ea (arm64: kasan: Increase stack size for KASAN_EXTRA).
>
> Signed-off-by: Qian Cai <[email protected]>
> ---
> arch/x86/include/asm/page_64_types.h | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/arch/x86/include/asm/page_64_types.h b/arch/x86/include/asm/page_64_types.h
> index 8f657286d599..0ce558a8150d 100644
> --- a/arch/x86/include/asm/page_64_types.h
> +++ b/arch/x86/include/asm/page_64_types.h
> @@ -7,7 +7,11 @@
> #endif
>
> #ifdef CONFIG_KASAN
> +#ifdef CONFIG_KASAN_EXTRA
> +#define KASAN_STACK_ORDER 2
So the kernel stack becomes 4-order page. That's above PAGE_ALLOC_COSTLY_ORDER, so people
will start seeing fork() failures with -ENOMEM due to high memory fragmentation. I don't think
we can afford such change.
Give that use-after-scope has proven to be almost useless for the kernel, I think we should just
remove it entirely.
> +#else
> #define KASAN_STACK_ORDER 1
> +#endif
> #else
> #define KASAN_STACK_ORDER 0
> #endif
>