Re: Can we drop upstream Linux x32 support?

From: Andy Lutomirski
Date: Wed Dec 12 2018 - 13:03:47 EST

On Wed, Dec 12, 2018 at 8:52 AM Rich Felker <dalias@xxxxxxxx> wrote:
> On Wed, Dec 12, 2018 at 08:39:53AM -0800, Andy Lutomirski wrote:
> > > On Dec 11, 2018, at 6:33 PM, Thorsten Glaser <tg@xxxxxxxxx> wrote:
> > >
> > > Andy Lutomirski dixit:
> > >
> >
> > >
> > >> IMO the real right solution would be to push the whole problem to
> > >> userspace: get an ILP32 system working with almost or entirely LP64
> > >
> > > Is this a reflex of Linux kernel developers? ;-)
> > >
> > > I doubt that userspace is the right place for this, remember
> > > the recent glibc vs. syscalls debate. It would also need to
> > > multiply across various libcs.
> > >
> > >> How hard would it be to have __attribute__((ilp64)), with an optional
> > >> warning if any embedded structs are not ilp64? This plus a wrapper to
> > >
> > > You mean LP64. Impossible, because LP64 vs. ILP32 is not the only
> > > difference between amd64 and x32.
> >
> > I mean LP64. And I'm not suggesting that ILP32 is the only difference
> > between x32 and x86_64, nor am I suggesting that a technique like this
> > would implement x32 -- I'm suggesting it would implement something
> > better than x32.
> >
> > The kernel, as a practical matter, supports two ABIs on 64-bit builds:
> > LP64 and ILP32. ILP32 is what the kernel calls "compat". ("compat"
> > comes with other baggage -- it generally has a 32-bit signal frame,
> > syscall arguments are mostly limited to 32 bits, etc.) Allowing a
> > user program that runs in 64-bit mode to issue compat syscalls is not
> > really a big deal. x86_64 has allowed this forever using int $0x80 --
> > it's just slow. Adding a faster mechanism would be straightforward.
> > As I understand it, the arm64 ilp32 proposal involves using a genuine
> > ILP32 model for user code, so the syscalls will all (except for signal
> > handling) go through the compat path.
> >
> > x32 is not this at all. The kernel ABI part of x32 isn't ILP32. It's
> > IP32, 32-bit size_t, and *64-bit* long. The core kernel doesn't
> > really support this. The only good things I can think of about it are
> > that (a) it proves that somewhat odd ABIs are possible, at least in
> > principle, and (b) three users have come out of the woodwork to say
> > that they use it.
> >
> > I'm proposing another alternative. Given that x32 already proves that
> > the user bitness model doesn't have to match the kernel model (in x32,
> > user "long" is 32-bit but the kernel ABI "long" is 64-bit), I'm
> > proposing extending this to just make the kernel ABI be LP64. So
> > __kernel_size_t would be 64-bit and pointers in kernel data structures
> > would be 64-bit. In other words, most or all of the kernel ABI would
> > just match x86_64.
> >
> > As far as I can tell, the only thing that really needs unusual
> > toolchain features here is that C doesn't have an extra-wide pointer
> > type. The kernel headers would need a way to say "this pointer is
> > still logically a pointer, and user code may assume that it's 32 bits,
> > but it has 8-byte alignment."
> None of this works on the userspace/C side, nor should any attempt be
> made to make it work. Types fundamentally cannot have alignments
> larger than their size. If you want to make the alignment of some
> pointers 8, you have to make their size 8, and then you just have LP64
> again if you did it for all pointers.
> If on the other hand you tried to make just some pointers "wide
> pointers", you'd also be completely breaking the specified API
> contracts of standard interfaces. For example in struct iovec's
> iov_base, &foo->iov_base is no longer a valid pointer to an object of
> type void* that you can pass to interfaces expecting void**. Sloppy
> misunderstandings like what you're making now are exactly why x32 is
> already broken and buggy (&foo->tv_nsec already has wrong type for
> struct timespec foo).

I don't think it's quite that broken. For the struct iovec example,
we currently have:

struct iovec {
void *iov_base; /* Starting address */
size_t iov_len; /* Number of bytes to transfer */

we could have, instead: (pardon any whitespace damage)

struct iovec {
void *iov_base; /* Starting address */
uint32_t __pad0;
size_t iov_len; /* Number of bytes to transfer */
uint32_t __pad1;
} __attribute__((aligned(8));

or the same thing but where iov_len is uint64_t. A pointer to
iov_base still works exactly as expected. Something would need to be
done to ensure that the padding is all zeroed, which might be a real

No one wants to actually type all the macro gunk into the headers to
make this work, but this type of transformation is what I have in mind
when the compiler is asked to handle the headers. Or there could
potentially be a tool that automatically consumes the uapi headers and
spits out modified headers like this.

Realistically, I think a much better model would be to use true ILP32
code, where all the memory layouts in the uapi match i386.

> Unless it's a thin, "pure" library that doesn't need anything from
> libc, or needs sufficiently little that it could be satisfied by some
> shims, this would necessarily require having two libcs in the same
> process, which is not going to work.

That's a good point.