Make the layout of kcov_remote_arg the same for 32-bit and 64-bit code.
This makes it more convenient to write userspace apps that can be compiled
into 32-bit or 64-bit binaries and still work with the same 64-bit kernel.
Also use proper __u32 types in uapi headers instead of unsigned ints.
Signed-off-by: Andrey Konovalov <[email protected]>
---
Hi Andrew,
We've noticed failures on 32 bit syzbot instances when the kcov patches
got merged into mainline. The reason is that the layout of kcov_remote_arg
depends on the alignment rules, which are different for 32/64 bit code.
We can deal with this issue in syzkaller [1], but I think it would be
cleander to get this fixed in the kernel.
I hope this patch is acceptable, since the change has just been merged and
is not included into a release kernel version. The patch breaks the newly
introduced kcov API for 32 bit apps.
Sorry for not testing it with 32 bit code earlier.
Thanks!
[1] https://github.com/google/syzkaller/commit/ba97c611a36b7729d489ebca5f97183c2ba7a90a
Documentation/dev-tools/kcov.rst | 7 ++++---
include/uapi/linux/kcov.h | 11 ++++++-----
2 files changed, 10 insertions(+), 8 deletions(-)
diff --git a/Documentation/dev-tools/kcov.rst b/Documentation/dev-tools/kcov.rst
index 36890b026e77..744df2bae1ed 100644
--- a/Documentation/dev-tools/kcov.rst
+++ b/Documentation/dev-tools/kcov.rst
@@ -251,9 +251,10 @@ selectively from different subsystems.
.. code-block:: c
struct kcov_remote_arg {
- unsigned trace_mode;
- unsigned area_size;
- unsigned num_handles;
+ uint32_t trace_mode;
+ uint32_t area_size;
+ uint32_t num_handles;
+ uint32_t reserved;
uint64_t common_handle;
uint64_t handles[0];
};
diff --git a/include/uapi/linux/kcov.h b/include/uapi/linux/kcov.h
index 409d3ad1e6e2..53267f9f1665 100644
--- a/include/uapi/linux/kcov.h
+++ b/include/uapi/linux/kcov.h
@@ -9,11 +9,12 @@
* and the comment before kcov_remote_start() for usage details.
*/
struct kcov_remote_arg {
- unsigned int trace_mode; /* KCOV_TRACE_PC or KCOV_TRACE_CMP */
- unsigned int area_size; /* Length of coverage buffer in words */
- unsigned int num_handles; /* Size of handles array */
- __u64 common_handle;
- __u64 handles[0];
+ __u32 trace_mode; /* KCOV_TRACE_PC or KCOV_TRACE_CMP */
+ __u32 area_size; /* Length of coverage buffer in words */
+ __u32 num_handles; /* Size of handles array */
+ __u32 reserved;
+ __u64 common_handle;
+ __u64 handles[0];
};
#define KCOV_REMOTE_MAX_HANDLES 0x100
--
2.24.0.393.g34dc348eaf-goog
On Fri, 6 Dec 2019 at 14:05, Andrey Konovalov <[email protected]> wrote:
>
> Make the layout of kcov_remote_arg the same for 32-bit and 64-bit code.
> This makes it more convenient to write userspace apps that can be compiled
> into 32-bit or 64-bit binaries and still work with the same 64-bit kernel.
> Also use proper __u32 types in uapi headers instead of unsigned ints.
>
> Signed-off-by: Andrey Konovalov <[email protected]>
> ---
>
> Hi Andrew,
>
> We've noticed failures on 32 bit syzbot instances when the kcov patches
> got merged into mainline. The reason is that the layout of kcov_remote_arg
> depends on the alignment rules, which are different for 32/64 bit code.
> We can deal with this issue in syzkaller [1], but I think it would be
> cleander to get this fixed in the kernel.
>
> I hope this patch is acceptable, since the change has just been merged and
> is not included into a release kernel version. The patch breaks the newly
> introduced kcov API for 32 bit apps.
>
> Sorry for not testing it with 32 bit code earlier.
>
> Thanks!
>
> [1] https://github.com/google/syzkaller/commit/ba97c611a36b7729d489ebca5f97183c2ba7a90a
>
> Documentation/dev-tools/kcov.rst | 7 ++++---
> include/uapi/linux/kcov.h | 11 ++++++-----
> 2 files changed, 10 insertions(+), 8 deletions(-)
>
> diff --git a/Documentation/dev-tools/kcov.rst b/Documentation/dev-tools/kcov.rst
> index 36890b026e77..744df2bae1ed 100644
> --- a/Documentation/dev-tools/kcov.rst
> +++ b/Documentation/dev-tools/kcov.rst
> @@ -251,9 +251,10 @@ selectively from different subsystems.
> .. code-block:: c
>
> struct kcov_remote_arg {
> - unsigned trace_mode;
> - unsigned area_size;
> - unsigned num_handles;
> + uint32_t trace_mode;
> + uint32_t area_size;
> + uint32_t num_handles;
> + uint32_t reserved;
> uint64_t common_handle;
> uint64_t handles[0];
> };
> diff --git a/include/uapi/linux/kcov.h b/include/uapi/linux/kcov.h
> index 409d3ad1e6e2..53267f9f1665 100644
> --- a/include/uapi/linux/kcov.h
> +++ b/include/uapi/linux/kcov.h
> @@ -9,11 +9,12 @@
> * and the comment before kcov_remote_start() for usage details.
> */
> struct kcov_remote_arg {
> - unsigned int trace_mode; /* KCOV_TRACE_PC or KCOV_TRACE_CMP */
> - unsigned int area_size; /* Length of coverage buffer in words */
> - unsigned int num_handles; /* Size of handles array */
> - __u64 common_handle;
> - __u64 handles[0];
> + __u32 trace_mode; /* KCOV_TRACE_PC or KCOV_TRACE_CMP */
> + __u32 area_size; /* Length of coverage buffer in words */
> + __u32 num_handles; /* Size of handles array */
> + __u32 reserved;
The kernel provides __aligned_u64 for this purpose (32/64 bit uapi
compatibility). Is adding an explicit 'reserved' better here? If so,
it'd be good to document.
Thanks,
-- Marco
On Fri, Dec 6, 2019 at 2:14 PM Marco Elver <[email protected]> wrote:
>
> On Fri, 6 Dec 2019 at 14:05, Andrey Konovalov <[email protected]> wrote:
> >
> > Make the layout of kcov_remote_arg the same for 32-bit and 64-bit code.
> > This makes it more convenient to write userspace apps that can be compiled
> > into 32-bit or 64-bit binaries and still work with the same 64-bit kernel.
> > Also use proper __u32 types in uapi headers instead of unsigned ints.
> >
> > Signed-off-by: Andrey Konovalov <[email protected]>
> > ---
> >
> > Hi Andrew,
> >
> > We've noticed failures on 32 bit syzbot instances when the kcov patches
> > got merged into mainline. The reason is that the layout of kcov_remote_arg
> > depends on the alignment rules, which are different for 32/64 bit code.
> > We can deal with this issue in syzkaller [1], but I think it would be
> > cleander to get this fixed in the kernel.
> >
> > I hope this patch is acceptable, since the change has just been merged and
> > is not included into a release kernel version. The patch breaks the newly
> > introduced kcov API for 32 bit apps.
> >
> > Sorry for not testing it with 32 bit code earlier.
> >
> > Thanks!
> >
> > [1] https://github.com/google/syzkaller/commit/ba97c611a36b7729d489ebca5f97183c2ba7a90a
> >
> > Documentation/dev-tools/kcov.rst | 7 ++++---
> > include/uapi/linux/kcov.h | 11 ++++++-----
> > 2 files changed, 10 insertions(+), 8 deletions(-)
> >
> > diff --git a/Documentation/dev-tools/kcov.rst b/Documentation/dev-tools/kcov.rst
> > index 36890b026e77..744df2bae1ed 100644
> > --- a/Documentation/dev-tools/kcov.rst
> > +++ b/Documentation/dev-tools/kcov.rst
> > @@ -251,9 +251,10 @@ selectively from different subsystems.
> > .. code-block:: c
> >
> > struct kcov_remote_arg {
> > - unsigned trace_mode;
> > - unsigned area_size;
> > - unsigned num_handles;
> > + uint32_t trace_mode;
> > + uint32_t area_size;
> > + uint32_t num_handles;
> > + uint32_t reserved;
> > uint64_t common_handle;
> > uint64_t handles[0];
> > };
> > diff --git a/include/uapi/linux/kcov.h b/include/uapi/linux/kcov.h
> > index 409d3ad1e6e2..53267f9f1665 100644
> > --- a/include/uapi/linux/kcov.h
> > +++ b/include/uapi/linux/kcov.h
> > @@ -9,11 +9,12 @@
> > * and the comment before kcov_remote_start() for usage details.
> > */
> > struct kcov_remote_arg {
> > - unsigned int trace_mode; /* KCOV_TRACE_PC or KCOV_TRACE_CMP */
> > - unsigned int area_size; /* Length of coverage buffer in words */
> > - unsigned int num_handles; /* Size of handles array */
> > - __u64 common_handle;
> > - __u64 handles[0];
> > + __u32 trace_mode; /* KCOV_TRACE_PC or KCOV_TRACE_CMP */
> > + __u32 area_size; /* Length of coverage buffer in words */
> > + __u32 num_handles; /* Size of handles array */
> > + __u32 reserved;
>
> The kernel provides __aligned_u64 for this purpose (32/64 bit uapi
> compatibility).
Oh yes, this is much better, sent v2, thank you!
> Is adding an explicit 'reserved' better here? If so,
> it'd be good to document.