To avoid false positives, KMSAN needs to unpoison the data copied from
the userspace. To detect infoleaks - check the memory buffer passed to
copy_to_user().
Signed-off-by: Alexander Potapenko <[email protected]>
---
v2:
-- move implementation of kmsan_copy_to_user() here
Link: https://linux-review.googlesource.com/id/I43e93b9c02709e6be8d222342f1b044ac8bdbaaf
---
include/linux/instrumented.h | 5 ++++-
include/linux/kmsan-checks.h | 19 ++++++++++++++++++
mm/kmsan/hooks.c | 38 ++++++++++++++++++++++++++++++++++++
3 files changed, 61 insertions(+), 1 deletion(-)
diff --git a/include/linux/instrumented.h b/include/linux/instrumented.h
index ee8f7d17d34f5..c73c1b19e9227 100644
--- a/include/linux/instrumented.h
+++ b/include/linux/instrumented.h
@@ -2,7 +2,7 @@
/*
* This header provides generic wrappers for memory access instrumentation that
- * the compiler cannot emit for: KASAN, KCSAN.
+ * the compiler cannot emit for: KASAN, KCSAN, KMSAN.
*/
#ifndef _LINUX_INSTRUMENTED_H
#define _LINUX_INSTRUMENTED_H
@@ -10,6 +10,7 @@
#include <linux/compiler.h>
#include <linux/kasan-checks.h>
#include <linux/kcsan-checks.h>
+#include <linux/kmsan-checks.h>
#include <linux/types.h>
/**
@@ -117,6 +118,7 @@ instrument_copy_to_user(void __user *to, const void *from, unsigned long n)
{
kasan_check_read(from, n);
kcsan_check_read(from, n);
+ kmsan_copy_to_user(to, from, n, 0);
}
/**
@@ -151,6 +153,7 @@ static __always_inline void
instrument_copy_from_user_after(const void *to, const void __user *from,
unsigned long n, unsigned long left)
{
+ kmsan_unpoison_memory(to, n - left);
}
#endif /* _LINUX_INSTRUMENTED_H */
diff --git a/include/linux/kmsan-checks.h b/include/linux/kmsan-checks.h
index a6522a0c28df9..c4cae333deec5 100644
--- a/include/linux/kmsan-checks.h
+++ b/include/linux/kmsan-checks.h
@@ -46,6 +46,21 @@ void kmsan_unpoison_memory(const void *address, size_t size);
*/
void kmsan_check_memory(const void *address, size_t size);
+/**
+ * kmsan_copy_to_user() - Notify KMSAN about a data transfer to userspace.
+ * @to: destination address in the userspace.
+ * @from: source address in the kernel.
+ * @to_copy: number of bytes to copy.
+ * @left: number of bytes not copied.
+ *
+ * If this is a real userspace data transfer, KMSAN checks the bytes that were
+ * actually copied to ensure there was no information leak. If @to belongs to
+ * the kernel space (which is possible for compat syscalls), KMSAN just copies
+ * the metadata.
+ */
+void kmsan_copy_to_user(void __user *to, const void *from, size_t to_copy,
+ size_t left);
+
#else
static inline void kmsan_poison_memory(const void *address, size_t size,
@@ -58,6 +73,10 @@ static inline void kmsan_unpoison_memory(const void *address, size_t size)
static inline void kmsan_check_memory(const void *address, size_t size)
{
}
+static inline void kmsan_copy_to_user(void __user *to, const void *from,
+ size_t to_copy, size_t left)
+{
+}
#endif
diff --git a/mm/kmsan/hooks.c b/mm/kmsan/hooks.c
index 43a529569053d..1cdb4420977f1 100644
--- a/mm/kmsan/hooks.c
+++ b/mm/kmsan/hooks.c
@@ -212,6 +212,44 @@ void kmsan_iounmap_page_range(unsigned long start, unsigned long end)
}
EXPORT_SYMBOL(kmsan_iounmap_page_range);
+void kmsan_copy_to_user(void __user *to, const void *from, size_t to_copy,
+ size_t left)
+{
+ unsigned long ua_flags;
+
+ if (!kmsan_enabled || kmsan_in_runtime())
+ return;
+ /*
+ * At this point we've copied the memory already. It's hard to check it
+ * before copying, as the size of actually copied buffer is unknown.
+ */
+
+ /* copy_to_user() may copy zero bytes. No need to check. */
+ if (!to_copy)
+ return;
+ /* Or maybe copy_to_user() failed to copy anything. */
+ if (to_copy <= left)
+ return;
+
+ ua_flags = user_access_save();
+ if ((u64)to < TASK_SIZE) {
+ /* This is a user memory access, check it. */
+ kmsan_internal_check_memory((void *)from, to_copy - left, to,
+ REASON_COPY_TO_USER);
+ user_access_restore(ua_flags);
+ return;
+ }
+ /* Otherwise this is a kernel memory access. This happens when a compat
+ * syscall passes an argument allocated on the kernel stack to a real
+ * syscall.
+ * Don't check anything, just copy the shadow of the copied bytes.
+ */
+ kmsan_internal_memmove_metadata((void *)to, (void *)from,
+ to_copy - left);
+ user_access_restore(ua_flags);
+}
+EXPORT_SYMBOL(kmsan_copy_to_user);
+
/* Functions from kmsan-checks.h follow. */
void kmsan_poison_memory(const void *address, size_t size, gfp_t flags)
{
--
2.37.0.rc0.161.g10f37bed90-goog
On Fri, 1 Jul 2022 at 16:24, Alexander Potapenko <[email protected]> wrote:
>
> To avoid false positives, KMSAN needs to unpoison the data copied from
> the userspace. To detect infoleaks - check the memory buffer passed to
> copy_to_user().
>
> Signed-off-by: Alexander Potapenko <[email protected]>
Reviewed-by: Marco Elver <[email protected]>
With the code simplification below.
[...]
> --- a/mm/kmsan/hooks.c
> +++ b/mm/kmsan/hooks.c
> @@ -212,6 +212,44 @@ void kmsan_iounmap_page_range(unsigned long start, unsigned long end)
> }
> EXPORT_SYMBOL(kmsan_iounmap_page_range);
>
> +void kmsan_copy_to_user(void __user *to, const void *from, size_t to_copy,
> + size_t left)
> +{
> + unsigned long ua_flags;
> +
> + if (!kmsan_enabled || kmsan_in_runtime())
> + return;
> + /*
> + * At this point we've copied the memory already. It's hard to check it
> + * before copying, as the size of actually copied buffer is unknown.
> + */
> +
> + /* copy_to_user() may copy zero bytes. No need to check. */
> + if (!to_copy)
> + return;
> + /* Or maybe copy_to_user() failed to copy anything. */
> + if (to_copy <= left)
> + return;
> +
> + ua_flags = user_access_save();
> + if ((u64)to < TASK_SIZE) {
> + /* This is a user memory access, check it. */
> + kmsan_internal_check_memory((void *)from, to_copy - left, to,
> + REASON_COPY_TO_USER);
This could just do "} else {" and the stuff below, and would result in
simpler code with no explicit "return" and no duplicated
user_access_restore().
> + user_access_restore(ua_flags);
> + return;
> + }
> + /* Otherwise this is a kernel memory access. This happens when a compat
> + * syscall passes an argument allocated on the kernel stack to a real
> + * syscall.
> + * Don't check anything, just copy the shadow of the copied bytes.
> + */
> + kmsan_internal_memmove_metadata((void *)to, (void *)from,
> + to_copy - left);
> + user_access_restore(ua_flags);
> +}
> +EXPORT_SYMBOL(kmsan_copy_to_user);
On Tue, Jul 12, 2022 at 3:52 PM Marco Elver <[email protected]> wrote:
>
> On Fri, 1 Jul 2022 at 16:24, Alexander Potapenko <[email protected]> wrote:
> >
> > To avoid false positives, KMSAN needs to unpoison the data copied from
> > the userspace. To detect infoleaks - check the memory buffer passed to
> > copy_to_user().
> >
> > Signed-off-by: Alexander Potapenko <[email protected]>
>
> Reviewed-by: Marco Elver <[email protected]>
>
> With the code simplification below.
>
> [...]
> > --- a/mm/kmsan/hooks.c
> > +++ b/mm/kmsan/hooks.c
> > @@ -212,6 +212,44 @@ void kmsan_iounmap_page_range(unsigned long start, unsigned long end)
> > }
> > EXPORT_SYMBOL(kmsan_iounmap_page_range);
> >
> > +void kmsan_copy_to_user(void __user *to, const void *from, size_t to_copy,
> > + size_t left)
> > +{
> > + unsigned long ua_flags;
> > +
> > + if (!kmsan_enabled || kmsan_in_runtime())
> > + return;
> > + /*
> > + * At this point we've copied the memory already. It's hard to check it
> > + * before copying, as the size of actually copied buffer is unknown.
> > + */
> > +
> > + /* copy_to_user() may copy zero bytes. No need to check. */
> > + if (!to_copy)
> > + return;
> > + /* Or maybe copy_to_user() failed to copy anything. */
> > + if (to_copy <= left)
> > + return;
> > +
> > + ua_flags = user_access_save();
> > + if ((u64)to < TASK_SIZE) {
> > + /* This is a user memory access, check it. */
> > + kmsan_internal_check_memory((void *)from, to_copy - left, to,
> > + REASON_COPY_TO_USER);
>
> This could just do "} else {" and the stuff below, and would result in
> simpler code with no explicit "return" and no duplicated
> user_access_restore().
Sounds good, will do.
--
Alexander Potapenko
Software Engineer
Google Germany GmbH
Erika-Mann-Straße, 33
80636 München
Geschäftsführer: Paul Manicle, Liana Sebastian
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg