For annoying architectural reasons, it's very difficult to support inline
instrumentation on powerpc64.
Add a Kconfig flag to allow an arch to disable inline. (It's a bit
annoying to be 'backwards', but I'm not aware of any way to have
an arch force a symbol to be 'n', rather than 'y'.)
We also disable stack instrumentation in this case as it does things that
are functionally equivalent to inline instrumentation, namely adding
code that touches the shadow directly without going through a C helper.
Signed-off-by: Daniel Axtens <[email protected]>
---
lib/Kconfig.kasan | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/lib/Kconfig.kasan b/lib/Kconfig.kasan
index cffc2ebbf185..935814f332a7 100644
--- a/lib/Kconfig.kasan
+++ b/lib/Kconfig.kasan
@@ -12,6 +12,15 @@ config HAVE_ARCH_KASAN_HW_TAGS
config HAVE_ARCH_KASAN_VMALLOC
bool
+# Sometimes an architecture might not be able to support inline instrumentation
+# but might be able to support outline instrumentation. This option allows an
+# arch to prevent inline and stack instrumentation from being enabled.
+# ppc64 turns on virtual memory late in boot, after calling into generic code
+# like the device-tree parser, so it uses this in conjuntion with a hook in
+# outline mode to avoid invalid access early in boot.
+config ARCH_DISABLE_KASAN_INLINE
+ bool
+
config CC_HAS_KASAN_GENERIC
def_bool $(cc-option, -fsanitize=kernel-address)
@@ -130,6 +139,7 @@ config KASAN_OUTLINE
config KASAN_INLINE
bool "Inline instrumentation"
+ depends on !ARCH_DISABLE_KASAN_INLINE
help
Compiler directly inserts code checking shadow memory before
memory accesses. This is faster than outline (in some workloads
@@ -141,6 +151,7 @@ endchoice
config KASAN_STACK
bool "Enable stack instrumentation (unsafe)" if CC_IS_CLANG && !COMPILE_TEST
depends on KASAN_GENERIC || KASAN_SW_TAGS
+ depends on !ARCH_DISABLE_KASAN_INLINE
default y if CC_IS_GCC
help
The LLVM stack address sanitizer has a know problem that
@@ -154,6 +165,9 @@ config KASAN_STACK
but clang users can still enable it for builds without
CONFIG_COMPILE_TEST. On gcc it is assumed to always be safe
to use and enabled by default.
+ If the architecture disables inline instrumentation, this is
+ also disabled as it adds inline-style instrumentation that
+ is run unconditionally.
config KASAN_SW_TAGS_IDENTIFY
bool "Enable memory corruption identification"
--
2.27.0
On Tue, 15 Jun 2021 at 03:47, Daniel Axtens <[email protected]> wrote:
>
> For annoying architectural reasons, it's very difficult to support inline
> instrumentation on powerpc64.
>
> Add a Kconfig flag to allow an arch to disable inline. (It's a bit
> annoying to be 'backwards', but I'm not aware of any way to have
> an arch force a symbol to be 'n', rather than 'y'.)
>
> We also disable stack instrumentation in this case as it does things that
> are functionally equivalent to inline instrumentation, namely adding
> code that touches the shadow directly without going through a C helper.
>
> Signed-off-by: Daniel Axtens <[email protected]>
> ---
> lib/Kconfig.kasan | 14 ++++++++++++++
> 1 file changed, 14 insertions(+)
>
> diff --git a/lib/Kconfig.kasan b/lib/Kconfig.kasan
> index cffc2ebbf185..935814f332a7 100644
> --- a/lib/Kconfig.kasan
> +++ b/lib/Kconfig.kasan
> @@ -12,6 +12,15 @@ config HAVE_ARCH_KASAN_HW_TAGS
> config HAVE_ARCH_KASAN_VMALLOC
> bool
>
> +# Sometimes an architecture might not be able to support inline instrumentation
> +# but might be able to support outline instrumentation. This option allows an
> +# arch to prevent inline and stack instrumentation from being enabled.
This comment could be moved into 'help' of this new config option.
> +# ppc64 turns on virtual memory late in boot, after calling into generic code
> +# like the device-tree parser, so it uses this in conjuntion with a hook in
> +# outline mode to avoid invalid access early in boot.
I think the ppc64-related comment isn't necessary and can be moved to
arch/ppc64 somewhere, if there isn't one already.
> +config ARCH_DISABLE_KASAN_INLINE
> + bool
> +
> config CC_HAS_KASAN_GENERIC
> def_bool $(cc-option, -fsanitize=kernel-address)
>
> @@ -130,6 +139,7 @@ config KASAN_OUTLINE
>
> config KASAN_INLINE
> bool "Inline instrumentation"
> + depends on !ARCH_DISABLE_KASAN_INLINE
> help
> Compiler directly inserts code checking shadow memory before
> memory accesses. This is faster than outline (in some workloads
> @@ -141,6 +151,7 @@ endchoice
> config KASAN_STACK
> bool "Enable stack instrumentation (unsafe)" if CC_IS_CLANG && !COMPILE_TEST
> depends on KASAN_GENERIC || KASAN_SW_TAGS
> + depends on !ARCH_DISABLE_KASAN_INLINE
> default y if CC_IS_GCC
> help
> The LLVM stack address sanitizer has a know problem that
> @@ -154,6 +165,9 @@ config KASAN_STACK
> but clang users can still enable it for builds without
> CONFIG_COMPILE_TEST. On gcc it is assumed to always be safe
> to use and enabled by default.
> + If the architecture disables inline instrumentation, this is
> + also disabled as it adds inline-style instrumentation that
> + is run unconditionally.
>
> config KASAN_SW_TAGS_IDENTIFY
> bool "Enable memory corruption identification"
> --
> 2.27.0
>
> --
> You received this message because you are subscribed to the Google Groups "kasan-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
> To view this discussion on the web visit https://groups.google.com/d/msgid/kasan-dev/20210615014705.2234866-2-dja%40axtens.net.
Hi Marco,
@@ -12,6 +12,15 @@ config HAVE_ARCH_KASAN_HW_TAGS
>> config HAVE_ARCH_KASAN_VMALLOC
>> bool
>>
>> +# Sometimes an architecture might not be able to support inline instrumentation
>> +# but might be able to support outline instrumentation. This option allows an
>> +# arch to prevent inline and stack instrumentation from being enabled.
>
> This comment could be moved into 'help' of this new config option.
It could. I did wonder if that made sense given that this is not a user
selectable option so I'm not sure if the help will ever be visible, but
I see that we do this sort of thing in Kconfig.kcsan and Kconfig.kgdb.
I've changed it over.
>> +# ppc64 turns on virtual memory late in boot, after calling into generic code
>> +# like the device-tree parser, so it uses this in conjuntion with a hook in
>> +# outline mode to avoid invalid access early in boot.
>
> I think the ppc64-related comment isn't necessary and can be moved to
> arch/ppc64 somewhere, if there isn't one already.
Fair enough. I'll pull it out of this file and look for a good place to
put the information in arch/powerpc in a later patch/series.
Kind regards,
Daniel