Re: [PATCH] Properly align memory allocations and temporary buffers

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2022-01-07 at 00:39:59, Jessica Clarke wrote:
> On 7 Jan 2022, at 00:31, brian m. carlson <sandals@xxxxxxxxxxxxxxxxxxxx> wrote:
> > If you want to get really language-lawyer-y about it, you can actually
> > argue that this is a compliant implementation of the C standard.
> > Integer types are permitted to have padding bits, and some combinations
> > of padding bits are allowed to be trap representations. Technically, in
> > our representation, the metadata bits are padding bits, because they do
> > not contribute to the precision like value bits. It is therefore the
> > case that the *value* of a uintptr_t still fits into a uintmax_t, but
> > the latter has no padding bits, and casting the latter to the former
> > yields a trap representation when further cast back to a pointer. This
> > may not the intent of the spec, and not how anyone thinks of it because
> > CHERI is the first implementation that pushes the boundary here, but
> > it’s technically legal under that interpretation. You may disagree with
> > the interpretation, and I don’t like to use it most of the time because
> > it’s complicated and involves yet more ill-defined parts of the spec
> > (e.g. it says arithmetic operations on valid values (they mean objects,
> > I assume, as the value only includes value bits, but the input could be
> > a trap representation on some implementations) never generate a trap
> > representation other than as part of an exceptional condition such as
> > an overflow, but nowhere defines what counts as an arithmetic
> > operation).
> 
> 
> So, no, C does not actually require what you say. It requires that void
> * -> uintptr_t -> void * give you a valid pointer. It requires that
> uintptr_t -> uintmax_t preserves the *value* of the uintptr_t, which we
> do, because the value is formed from only the value bits which
> contribute to the precision, which is 64 bits in this case, and
> uintmax_t is still 64-bit. It requires that uintmax_t -> uintptr_t,
> since uintptr_t’s precision is the same as uintmax_t’s, be always
> valid, which is is. But it does not require that that uintptr_t has the
> same representation as the original uintptr_t, which it does not for
> us. And therefore it does not require that casting that uintptr_t back
> to a void * yields a valid pointer. So if you want to really dig into
> the details of the standard, we are technically compliant, even if some
> might argue it’s not in the spirit of the standard.

Sure, implementations are allowed to have padding bits.  They're also
allowed, at the moment, to use signed-magnitude or ones' complement
integers, have CHAR_BIT greater than 8, have sizeof(char) ==
sizeof(short), not implement any of the customary sizes of intN_t or
uintN_t, not provide uintptr_t, and use middle-endian numbers.

However, if your ABI is only compliant in the face of those features
(especially when it could have been written in a way which would have
been compliant without the use of those features), it's intentionally
hostile to real-world developers, and I don't think we should support
it[0].  I'd be willing to revisit this if your ABI were defined in a
reasonable, sane way, where sizeof(uintmax_t) >= sizeof(uintptr_t),
without padding bits, where the alignment of pointers from malloc is
suitable for all types, and where the alignment of a type is no greater
than sizeof(type).

I'm not opposed to a small amount of finagling for this case, but I am
very much opposed to defining your C ABI in an intentionally difficult
way.  128-bit integers in 64-bit Linux were not originally part of the C
ABIs and if the ABI is ill defined now, that's a historical accident.
But this is a new ABI for a new architecture and it could have been
defined in a responsible way, but wasn't.

As an aside, I was actually going to point out that you could propose a
nice Rust or Go ABI with the status quo, but if your C ABI requires
padding bits, then you're probably going to have a hard time doing so,
since I don't believe those languages support padding bits and they need
to support the C ABI.

[0] For the record, I care strongly about portability, and I would not
accept a runtime having any of the qualities I mentioned in the first
paragraph.
-- 
brian m. carlson (he/him or they/them)
Toronto, Ontario, CA

Attachment: signature.asc
Description: PGP signature


[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux