On Thu, Jun 24, 2021 at 09:36:54PM +0100, Robin Murphy wrote:

On 2021-06-24 19:55, Catalin Marinas wrote:

quoted

On Thu, Jun 24, 2021 at 04:27:17PM +0000, Al Viro wrote:

quoted

On Thu, Jun 24, 2021 at 02:22:27PM +0100, Robin Murphy wrote:

quoted

FWIW I think the only way to make the kernel behaviour any more robust here
would be to make the whole uaccess API more expressive, such that rather
than simply saying "I only got this far" it could actually differentiate
between stopping due to a fault which may be recoverable and worth retrying,
and one which definitely isn't.

... and propagate that "more expressive" information through what, 3 or 4
levels in the call chain?

 From include/linux/uaccess.h:

  * If raw_copy_{to,from}_user(to, from, size) returns N, size - N bytes starting
  * at to must become equal to the bytes fetched from the corresponding area
  * starting at from.  All data past to + size - N must be left unmodified.
  *
  * If copying succeeds, the return value must be 0.  If some data cannot be
  * fetched, it is permitted to copy less than had been fetched; the only
  * hard requirement is that not storing anything at all (i.e. returning size)
  * should happen only when nothing could be copied.  In other words, you don't
  * have to squeeze as much as possible - it is allowed, but not necessary.

arm64 instances violate the aforementioned hard requirement.

After reading the above a few more times, I think I get it. The key
sentence is: not storing anything at all should happen only when nothing
could be copied. In the MTE case, something can still be copied.

quoted

Please, fix
it there; it's not hard.  All you need is an exception handler in .Ltiny15
that would fall back to (short) byte-by-byte copy if the faulting address
happened to be unaligned.  Or just do one-byte copy, not that it had been
considerably cheaper than a loop.  Will be cheaper than propagating that extra
information up the call chain, let alone paying for extra ->write_begin()
and ->write_end() for single byte in generic_perform_write().

Yeah, it's definitely fixable in the arch code. I misread the above
requirements and thought it could be fixed in the core code.

Quick hack, though I think in the actual exception handling path in .S
more sense (and it needs the copy_to_user for symmetry):

Hmm, if anything the asm version might be even more straightforward; I think
it's pretty much just this (untested):

That's what I thought but it was too late in the day to think in asm.

I don't think we over-read/write in the routine itself as this is based
on the user memcpy() which can't handle faults. And since we got a fault
before the end of the copy, we have at least one byte left in the
buffer (which may or may not trigger a fault).

I wonder whether we should skip the extra byte copy if something was
copied, i.e. start the exception handler with:

	cmp	dstin, dst
	b.ne	9997f

That said, the fall-back to bytewise copying may have some advantage. I
think we still have the issue where we copy some data to user but report
less (STP failing on the second 8-byte when the first had been already
written first 8). A byte copy loop would solve this, unless we pass the
fault address to the exception handler (I thought you had some patch for
this at some point).

-- 
Catalin