Currently the macro that inserts entries into the SPU syscall table doesn't actually use the "nr" (syscall number) parameter.
This does work, but it relies on the exact right number of syscall entries being emitted in order for the syscal numbers to line up with the array entries. If for example we had two entries with the same syscall number we wouldn't get an error, it would just cause all subsequent syscalls to be off by one in the spu_syscall_table. So instead change the macro to assign to the specific entry of the array, meaning any numbering overlap will be caught by the compiler. Signed-off-by: Michael Ellerman <m...@ellerman.id.au> Acked-by: Arnd Bergmann <a...@arndb.de> --- arch/powerpc/platforms/cell/spu_callbacks.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) v2: Add Arnd's ack. diff --git a/arch/powerpc/platforms/cell/spu_callbacks.c b/arch/powerpc/platforms/cell/spu_callbacks.c index 2f03671ac8db..8f94acdd6d6d 100644 --- a/arch/powerpc/platforms/cell/spu_callbacks.c +++ b/arch/powerpc/platforms/cell/spu_callbacks.c @@ -34,7 +34,7 @@ */ void *spu_syscall_table[] = { -#define __SYSCALL(nr, entry, nargs) entry, +#define __SYSCALL(nr, entry, nargs) [nr] = entry, #include <asm/syscall_table_spu.h> #undef __SYSCALL }; -- 2.20.1