[musl] Re: [PATCHv3 00/24] ILP32 support in ARM64

[arnd@arndb.de (Arnd Bergmann)]

On Monday 16 February 2015 12:51:35 Rich Felker wrote:
> On Mon, Feb 16, 2015 at 06:20:18PM +0100, Arnd Bergmann wrote:
> > > Would it really be that hard to do:
> > > 
> > > 	if (ILP32_on_64_process) tv_nsec = (int)tv_nsec;
> > > 
> > > or similar? That's all that's needed.
> > > 
> > > > In some cases, there may also be a measurable performance penalty
> > > > in interpreting a user space data structure manually over copying
> > > > it (including the timespec values) in one chunk.
> > > 
> > > I don't think the above would be measurable.
> > 
> > It depends: Copying the structure first and then doing the conversion
> > in kernel space on the specific members as you do in the example
> > should indeed have a trivial performance impact. However, it is also
> > the hardest for driver writers to get right, and it's better not to
> > trust them with corner cases like this.
> > 
> > To make it more readable, we would probably introduce a helper function
> > that copies the timespec from user space memory to kernel space and
> > then does all the checks and conversions as required. However, doing
> > separate copies can (depending on the architecture) have a noticeable
> > impact. An example for this would be architectures that require setting
> > up a page table entry for the user space page in order to access the
> > data and then destroy it again afterwards, with the correct TLB flushes.
> > 
> > We can do something like this for the old-style compat handlers that
> > use 32-bit time_t, but I'd prefer not to have it in the fast path for
> > the native 64-bit time_t on 64-bit architectures.
> 
> I know this isn't the place to discuss large architectural kernel
> changes, but it would be really nice if the kernel had proper abstract
> knowledge, at syscall entry time, what regions of memory from
> userspace the syscall is going to need and a way of marshalling them
> all together as prep for enterring the code that implements the
> syscalls, and if conversion between different ABIs could take place
> mostly automatically at this layer. Perhaps this kind of thing is an
> idea that could be kept open for the future. I suspect the
> combinatorics of different legacy interfaces are going to continue
> getting worse, and it would be much nicer to have the support factored
> out of the actual syscall implementations.

Some subsystems (e.g. v4l) do this in their ioctls, and it sounds
like a nice idea, but as I count 17457 instances of copy_{to,from}_user
or {get,put}_user in the kernel, I believe we are basically stuck
with the current way.

> > > Generally I would think the kernel knows the model the process is
> > > using, but if not, all you need is separate ioctl numbers for
> > > userspace to use depending on which definition it's using.
> > 
> > I've checked now, and indeed the kernel knows for ilp32 x86 and arm, since
> > it uses a different ELF interpreter. I thought it might be running the
> > ilp32 binaries as ELF64, but it does not.
> 
> This would result in lots of problems like argv[], auxv[], envp[],
> etc. being in the wrong format.

Right. It depends a bit on the scope though: My impression is that
a lot of people who ask for ilp32 mode on arm64 are just interested
in getting to work one or two applications. If that was the only
goal, they could work around the problem (mostly) in user space, but
as it turns out, the kernel patch is doing the entire job of
implementing the new ABI at syscall level to the point where you
can (mostly) just recompile all of user space.

	Arnd