On Fri, May 27, 2016 at 08:03:57AM +0200, Heiko Carstens wrote: > > > > The cost is pretty trivial though. See kernel/compat_wrapper.o: > > > > COMPAT_SYSCALL_WRAP2(creat, const char __user *, pathname, umode_t, mode); > > > > 0: a9bf7bfd stp x29, x30, [sp,#-16]! > > > > 4: 910003fd mov x29, sp > > > > 8: 2a0003e0 mov w0, w0 > > > > c: 94000000 bl 0 <sys_creat> > > > > 10: a8c17bfd ldp x29, x30, [sp],#16 > > > > 14: d65f03c0 ret > > > > > > I would say the above could be more expensive than 8 movs (16 bytes to > > > write, read, a branch and a ret). You can also add the I-cache locality, > > > having wrappers for each syscalls instead of a single place for zeroing > > > the upper half (where no other wrapper is necessary). > > > > > > Can we trick the compiler into doing a tail call optimisation. This > > > could have simply been: > > > > > > COMPAT_SYSCALL_WRAP2(creat, ...): > > > mov w0, w0 > > > b <sys_creat> > > > > What you talk about was in my initial version. But Heiko insisted on having all > > wrappers together. > > http://www.spinics.net/lists/linux-s390/msg11593.html > > > > Grep your email for discussion. > > I think Catalin's question was more about why there is even a stack frame > generated. It looks like it is not necessary. I did ask this too a couple > of months ago, when we discussed this. Indeed, I was questioning the need for prologue/epilogue, not the use of COMPAT_SYSCALL_WRAPx(). Maybe something like __naked would do. > > > > > Cost wise, this seems like it all cancels out in the end, but what > > > > > do I know? > > > > > > > > I think you know something, and I also think Heiko and other s390 guys > > > > know something as well. So I'd like to listen their arguments here. > > If it comes to 64 bit arguments for compat system calls: s390 also has an > x32-like ABI extension which allows user space to use full 64 bit > registers. As far as I know hardly anybody ever made use of that. > > However even if that would be widely used, to me it wouldn't make sense to > add new compat system calls which allow 64 bit arguments, simply because > something like > > c = (u32)a | (u64)b << 32; > > can be done with a single 1-cycle instruction. It's just not worth the > extra effort to maintain additional system call variants. If we split 64-bit arguments in two, we can go a step further and avoid most of the COMPAT_SYSCALL_WRAPx annotations in favour of a common upper-half zeroing of the argument registers on ILP32 syscall entry. There would be a few exceptions where we need to re-build 64-bit arguments on sign-extend. -- Catalin -- To unsubscribe from this list: send the line "unsubscribe linux-arch" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
