Not all compilers have __builtin_bswap16() and __builtin_bswap32(), thus not all compilers are able to compile the following code (bpf_htons):
(__builtin_constant_p(x) ? \ ___constant_swab16(x) : __builtin_bswap16(x))
That's why, for instance, bpf_htons() doesn't work on GCC < 4.8:
error: implicit declaration of function '__builtin_bswap16'
We can use __builtin_bswap16() only if compiler has this built-in, that is, only if __HAVE_BUILTIN_BSWAP16__ is defined. Standard UAPI __swab16()/__swab32() take care of that, and, additionally, handle __builtin_constant_p() cases as well (if compiler doesn't provide builtin bswap with constants folding):
#ifdef __HAVE_BUILTIN_BSWAP16__ #define __swab16(x) (__u16)__builtin_bswap16((__u16)(x)) #else #define __swab16(x) \ (__builtin_constant_p((__u16)(x)) ? \ ___constant_swab16(x) : \ __fswab16(x)) #endif
So we can tweak selftests/bpf/bpf_endian.h and use UAPI __swab16()/__swab32().
Signed-off-by: Sergey Senozhatsky sergey.senozhatsky@gmail.com --- tools/testing/selftests/bpf/bpf_endian.h | 37 +++++------------------- 1 file changed, 8 insertions(+), 29 deletions(-)
diff --git a/tools/testing/selftests/bpf/bpf_endian.h b/tools/testing/selftests/bpf/bpf_endian.h index b25595ea4a78..ba06222963d5 100644 --- a/tools/testing/selftests/bpf/bpf_endian.h +++ b/tools/testing/selftests/bpf/bpf_endian.h @@ -20,38 +20,17 @@ * use different targets. */ #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ -# define __bpf_ntohs(x) __builtin_bswap16(x) -# define __bpf_htons(x) __builtin_bswap16(x) -# define __bpf_constant_ntohs(x) ___constant_swab16(x) -# define __bpf_constant_htons(x) ___constant_swab16(x) -# define __bpf_ntohl(x) __builtin_bswap32(x) -# define __bpf_htonl(x) __builtin_bswap32(x) -# define __bpf_constant_ntohl(x) ___constant_swab32(x) -# define __bpf_constant_htonl(x) ___constant_swab32(x) +# define bpf_ntohs(x) __swab16(x) +# define bpf_htons(x) __swab16(x) +# define bpf_ntohl(x) __swab32(x) +# define bpf_htonl(x) __swab32(x) #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ -# define __bpf_ntohs(x) (x) -# define __bpf_htons(x) (x) -# define __bpf_constant_ntohs(x) (x) -# define __bpf_constant_htons(x) (x) -# define __bpf_ntohl(x) (x) -# define __bpf_htonl(x) (x) -# define __bpf_constant_ntohl(x) (x) -# define __bpf_constant_htonl(x) (x) +# define bpf_ntohs(x) (x) +# define bpf_htons(x) (x) +# define bpf_ntohl(x) (x) +# define bpf_htonl(x) (x) #else # error "Fix your compiler's __BYTE_ORDER__?!" #endif
-#define bpf_htons(x) \ - (__builtin_constant_p(x) ? \ - __bpf_constant_htons(x) : __bpf_htons(x)) -#define bpf_ntohs(x) \ - (__builtin_constant_p(x) ? \ - __bpf_constant_ntohs(x) : __bpf_ntohs(x)) -#define bpf_htonl(x) \ - (__builtin_constant_p(x) ? \ - __bpf_constant_htonl(x) : __bpf_htonl(x)) -#define bpf_ntohl(x) \ - (__builtin_constant_p(x) ? \ - __bpf_constant_ntohl(x) : __bpf_ntohl(x)) - #endif /* __BPF_ENDIAN__ */
On 03/19, Sergey Senozhatsky wrote:
Not all compilers have __builtin_bswap16() and __builtin_bswap32(), thus not all compilers are able to compile the following code (bpf_htons):
(__builtin_constant_p(x) ? \ ___constant_swab16(x) : __builtin_bswap16(x))
That's why, for instance, bpf_htons() doesn't work on GCC < 4.8:
error: implicit declaration of function '__builtin_bswap16'
We can use __builtin_bswap16() only if compiler has this built-in, that is, only if __HAVE_BUILTIN_BSWAP16__ is defined. Standard UAPI __swab16()/__swab32() take care of that, and, additionally, handle __builtin_constant_p() cases as well (if compiler doesn't provide builtin bswap with constants folding):
#ifdef __HAVE_BUILTIN_BSWAP16__ #define __swab16(x) (__u16)__builtin_bswap16((__u16)(x)) #else #define __swab16(x) \ (__builtin_constant_p((__u16)(x)) ? \ ___constant_swab16(x) : \ __fswab16(x)) #endif
So we can tweak selftests/bpf/bpf_endian.h and use UAPI __swab16()/__swab32().
Signed-off-by: Sergey Senozhatsky sergey.senozhatsky@gmail.com
tools/testing/selftests/bpf/bpf_endian.h | 37 +++++------------------- 1 file changed, 8 insertions(+), 29 deletions(-)
diff --git a/tools/testing/selftests/bpf/bpf_endian.h b/tools/testing/selftests/bpf/bpf_endian.h index b25595ea4a78..ba06222963d5 100644 --- a/tools/testing/selftests/bpf/bpf_endian.h +++ b/tools/testing/selftests/bpf/bpf_endian.h @@ -20,38 +20,17 @@
- use different targets.
*/ #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ -# define __bpf_ntohs(x) __builtin_bswap16(x) -# define __bpf_htons(x) __builtin_bswap16(x) -# define __bpf_constant_ntohs(x) ___constant_swab16(x) -# define __bpf_constant_htons(x) ___constant_swab16(x)
This breaks the build until your next patch is applied (in other words, breaks bisection). Can we do it in three steps? Convert to swab (without breaking existing tests), convert the tests, remove unused __bpf_xyz defines?
Could you also send it as a series (git format-patch --thread)? Those patches depend on each other. And pls use [PATCH bpf-next] ... subj.
-# define __bpf_ntohl(x) __builtin_bswap32(x) -# define __bpf_htonl(x) __builtin_bswap32(x) -# define __bpf_constant_ntohl(x) ___constant_swab32(x) -# define __bpf_constant_htonl(x) ___constant_swab32(x) +# define bpf_ntohs(x) __swab16(x) +# define bpf_htons(x) __swab16(x) +# define bpf_ntohl(x) __swab32(x) +# define bpf_htonl(x) __swab32(x) #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ -# define __bpf_ntohs(x) (x) -# define __bpf_htons(x) (x) -# define __bpf_constant_ntohs(x) (x) -# define __bpf_constant_htons(x) (x) -# define __bpf_ntohl(x) (x) -# define __bpf_htonl(x) (x) -# define __bpf_constant_ntohl(x) (x) -# define __bpf_constant_htonl(x) (x) +# define bpf_ntohs(x) (x) +# define bpf_htons(x) (x) +# define bpf_ntohl(x) (x) +# define bpf_htonl(x) (x) #else # error "Fix your compiler's __BYTE_ORDER__?!" #endif -#define bpf_htons(x) \
- (__builtin_constant_p(x) ? \
__bpf_constant_htons(x) : __bpf_htons(x))
-#define bpf_ntohs(x) \
- (__builtin_constant_p(x) ? \
__bpf_constant_ntohs(x) : __bpf_ntohs(x))
-#define bpf_htonl(x) \
- (__builtin_constant_p(x) ? \
__bpf_constant_htonl(x) : __bpf_htonl(x))
-#define bpf_ntohl(x) \
- (__builtin_constant_p(x) ? \
__bpf_constant_ntohl(x) : __bpf_ntohl(x))
#endif /* __BPF_ENDIAN__ */
2.21.0
On (03/19/19 09:01), Stanislav Fomichev wrote:
diff --git a/tools/testing/selftests/bpf/bpf_endian.h b/tools/testing/selftests/bpf/bpf_endian.h index b25595ea4a78..ba06222963d5 100644 --- a/tools/testing/selftests/bpf/bpf_endian.h +++ b/tools/testing/selftests/bpf/bpf_endian.h @@ -20,38 +20,17 @@
- use different targets.
*/ #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ -# define __bpf_ntohs(x) __builtin_bswap16(x) -# define __bpf_htons(x) __builtin_bswap16(x) -# define __bpf_constant_ntohs(x) ___constant_swab16(x) -# define __bpf_constant_htons(x) ___constant_swab16(x)
This breaks the build until your next patch is applied (in other words, breaks bisection). Can we do it in three steps?
Bummer! I thought about applying the second patch (flow_dissector.c) first and then the first one (bpf/bpf_endian.h).
Convert to swab (without breaking existing tests), convert the tests, remove unused __bpf_xyz defines?
OK.
Could you also send it as a series (git format-patch --thread)? Those patches depend on each other. And pls use [PATCH bpf-next] ... subj.
Right. I figured out that I also need to patch flow_dissector.c after I sent out bpf/bpf_endian.h patch.
Sorry about that, will fix.
-ss
linux-kselftest-mirror@lists.linaro.org