Linux powerpc new system call instruction and ABI

[Nicholas Piggin]

Excerpts from Dmitry V. Levin's message of May 19, 2021 11:26 pm:
> On Wed, May 19, 2021 at 08:59:05PM +1000, Nicholas Piggin wrote:
>> Excerpts from Dmitry V. Levin's message of May 19, 2021 8:24 pm:
>> > On Wed, May 19, 2021 at 12:50:24PM +1000, Nicholas Piggin wrote:
>> > [...]
>> >> With this patch, I think the ptrace ABI should mostly be fixed. I think 
>> >> a problem remains with applications that look at system call return 
>> >> registers directly and have powerpc specific error cases. Those probably
>> >> will just need to be updated unfortunately. Michael thought it might be
>> >> possible to return an indication via ptrace somehow that the syscall is
>> >> using a new ABI, so such apps can be updated to test for it. I don't 
>> >> know how that would be done.
>> > 
>> > Is there any sane way for these applications to handle the scv case?
>> > How can they tell that the scv semantics is being used for the given
>> > syscall invocation?  Can this information be obtained e.g. from struct
>> > pt_regs?
>> 
>> Not that I know of. Michael suggested there might be a way to add 
>> something. ptrace_syscall_info has some pad bytes, could
>> we use one for flags bits and set a bit for "new system call ABI"?
> 
> PTRACE_GET_SYSCALL_INFO is an architecture-agnostic API, it hides all
> architecture-specific details behind struct ptrace_syscall_info which has
> the same meaning on all architectures.  ptrace_syscall_info.exit contains
> both rval and is_error fields to support every architecture regardless of
> its syscall ABI.
> 
> ptrace_syscall_info.exit is extensible, but every architecture would have
> to define a method of telling whether the system call follows the "new
> system call ABI" conventions to export this bit of information.

It's already architecture speicfic if you look at registers of syscall 
exit state so I don't see a problem with a flag that ppc can use for
ABI.

> 
> This essentially means implementing something like
> static inline long syscall_get_error_abi(struct task_struct *task, struct pt_regs *regs)
> for every architecture, and using it along with syscall_get_error
> in ptrace_get_syscall_info_exit to initialize the new field in
> ptrace_syscall_info.exit structure.

Yes this could work. Other architectures can just use a generic
implementation if they don't define their own so that's easy. And
in userspace they can continue to ignore the flag.

> 
>> As a more hacky thing you could make a syscall with -1 and see how
>> the error looks, and then assume all syscalls will be the same.
> 
> This would be very unreliable because sc and scv are allowed to intermingle,
> so every syscall invocation can follow any of these two error handling
> conventions.
> 
>> Or... is it possible at syscall entry to peek the address of
>> the instruction which caused the call and see if that was a
>> scv instruction? That would be about as reliable as possible
>> without having that new flag bit.
> 
> No other architecture requires peeking into tracee memory just to find out
> the syscall ABI.  This would make powerpc the most ugly architecture for
> ptracing.
> 
> I wonder why can't this information be just exported to the tracer via
> struct pt_regs?

It might be able to, I don't see why that would be superior though.

Where could you put it... I guess it could go in the trap field in a 
high bit. But could that break things that just test for syscall 
trap number (and don't care about register ABI)? I'm not sure.

Thanks,
Nick