Re: [PATCH 5.4 49/52] powerpc/bpf: Fix BPF_MOD when imm == 1

[Christophe Leroy]

Le 11/10/2021 à 19:33, Naresh Kamboju a écrit :
> stable-rc 5.4 build failed due this patch.
>   - powerpc gcc-10-defconfig - FAILED
>   - powerpc gcc-11-defconfig - FAILED
>   - powerpc gcc-8-defconfig - FAILED
>   - powerpc gcc-9-defconfig - FAILED
>
>
> On Mon, 11 Oct 2021 at 19:28, Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > From: Naveen N. Rao <naveen.n.rao@xxxxxxxxxxxxxxxxxx>
> >
> > [ Upstream commit 8bbc9d822421d9ac8ff9ed26a3713c9afc69d6c8 ]
> >
> > Only ignore the operation if dividing by 1.
>
> <trim>
>
> In file included from arch/powerpc/net/bpf_jit64.h:11,
>                   from arch/powerpc/net/bpf_jit_comp64.c:19:
> arch/powerpc/net/bpf_jit_comp64.c: In function 'bpf_jit_build_body':
> arch/powerpc/net/bpf_jit_comp64.c:415:46: error: implicit declaration
> of function 'PPC_RAW_LI'; did you mean 'PPC_RLWIMI'?

PPC_RAW_LI() was added by commit 3a1812379163 ("powerpc/ppc-opcode: 
Consolidate powerpc instructions from bpf_jit.h")

Priori to that you have to use PPC_LI() instead, with the same arguments.

Christophe

> [-Werror=implicit-function-declaration]
>    415 |                                         EMIT(PPC_RAW_LI(dst_reg, 0));
>        |                                              ^~~~~~~~~~
> arch/powerpc/net/bpf_jit.h:32:34: note: in definition of macro 'PLANT_INSTR'
>     32 |         do { if (d) { (d)[idx] = instr; } idx++; } while (0)
>        |                                  ^~~~~
> arch/powerpc/net/bpf_jit_comp64.c:415:41: note: in expansion of macro 'EMIT'
>    415 |                                         EMIT(PPC_RAW_LI(dst_reg, 0));
>        |                                         ^~~~
> cc1: all warnings being treated as errors
>
> Reported-by: Linux Kernel Functional Testing <lkft@xxxxxxxxxx>
>
> --
> Linaro LKFT
> https://lkft.linaro.org
> ReplyReply to allForward