Re: [PATCH v2] tools lib bpf: Support libbpf on PowerPC

From: Naveen N. Rao
Date: Mon Dec 14 2015 - 10:44:22 EST


On 2015/12/14 11:57AM, Wang Nan wrote:
> Support basic PowerPC compiling.
>
> Checks BPF syscall number, turn off libbpf building on platform doesn't
> support sys_bpf instead of blocking compiling.
>
> Reported-by: Naveen N. Rao <naveen.n.rao@xxxxxxxxxxxxxxxxxx>
> Signed-off-by: Wang Nan <wangnan0@xxxxxxxxxx>
> Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> Cc: Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx>
> Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
> Cc: Sukadev Bhattiprolu <sukadev@xxxxxxxxxxxxxxxxxx>
> ---
>
> v1 -> v2: __powerpc64__ -> __powerpc__. Code is from
> ./tools/perf/perf-sys.h, but not tested yet.
>
> Naveen, please help me try this patch on PPC.

Thanks, this works. However...

>
> Thank you.
>
> ---
> tools/build/feature/test-bpf.c | 23 ++++++++++++++++++++++-
> tools/lib/bpf/bpf.c | 6 ++++--
> 2 files changed, 26 insertions(+), 3 deletions(-)
>
> diff --git a/tools/build/feature/test-bpf.c b/tools/build/feature/test-bpf.c
> index 062bac8..19497f7 100644
> --- a/tools/build/feature/test-bpf.c
> +++ b/tools/build/feature/test-bpf.c
> @@ -1,9 +1,26 @@
> +#include <asm/unistd.h>
> #include <linux/bpf.h>
> +#include <unistd.h>
> +
> +#ifndef __NR_bpf
> +# if defined(__i386__)
> +# define __NR_bpf 357
> +# elif defined(__x86_64__)
> +# define __NR_bpf 321
> +# elif defined(__aarch64__)
> +# define __NR_bpf 280
> +# elif defined(__powerpc__)
> +# define __NR_bpf 361

I think we should drop __aarch64__ and __powerpc__ here since this
builds fine on ppc without these hunks.

> +# else
> +# error __NR_bpf not defined. libbpf does not support your arch.
> +# endif
> +#endif
>
> int main(void)
> {
> union bpf_attr attr;
>
> + /* Check fields in attr */
> attr.prog_type = BPF_PROG_TYPE_KPROBE;
> attr.insn_cnt = 0;
> attr.insns = 0;
> @@ -14,5 +31,9 @@ int main(void)
> attr.kern_version = 0;
>
> attr = attr;
> - return 0;
> + /*
> + * Test existance of __NR_bpf and BPF_PROG_LOAD.

Nit... 'for existence'.

> + * This call should fail if we run the testcase.
> + */
> + return syscall(__NR_bpf, BPF_PROG_LOAD, attr, sizeof(attr));
> }
> diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c
> index 5bdc6ea..fd25c58 100644
> --- a/tools/lib/bpf/bpf.c
> +++ b/tools/lib/bpf/bpf.c
> @@ -14,8 +14,8 @@
> #include "bpf.h"
>
> /*
> - * When building perf, unistd.h is override. Define __NR_bpf is
> - * required to be defined.
> + * When building perf, unistd.h is overrided. __NR_bpf is
> + * required to be defined explicitly.
> */
> #ifndef __NR_bpf
> # if defined(__i386__)
> @@ -24,6 +24,8 @@
> # define __NR_bpf 321
> # elif defined(__aarch64__)
> # define __NR_bpf 280
> +# elif defined(__powerpc__)
> +# define __NR_bpf 361

And, I think we should drop this hunk, but include the patch I sent.
That ensures that the proper headers from the kernel source tree are
included, so there won't be a need to explicitly define __NR_bpf for
each architecture.


Regards,
Naveen

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/