Re: [PATCH v2 3/3] io: Add FORTIFY_SOURCE check for fcntl arguments

[Sergey Bugaev <bugaevc@gmail.com>]

On Mon, May 29, 2023 at 11:14 PM Adhemerval Zanella Netto
<adhemerval.zanella@linaro.org> wrote:
> Now that you brought Rust black_box, we already something similar on
> benchtests: DO_NOT_OPTIMIZE_OUT.

I see, thanks -- but that also seems different in that it doesn't mark
the value as the output of the asm block, so the compiler could see
that it's returned unmodified. So it wouldn't be suitable here.

Rust's black_box is / was [0] instead implemented as

llvm_asm!("" : : "r"(&mut dummy) : "memory" : "volatile");

i.e. it marks the mutable reference to the value as inline asm's
input, not the value itself.

[0]: nowadays it's a compiler intrinsic that still gets codegenned to
a similar asm block when using the LLVM codegen backend

> The LFS names are not considered a namespace pollution, so I think that's
> why it always provided (just check tst-fortify.c LFS name usage, like
> pread64).

They may be always provided when building tst-fortify.c, but they're
certainly not always provided when building user code (not even
provided by default):

$ gcc use-fcntl64.c -Werror
use-fcntl64.c: In function ‘main’:
use-fcntl64.c:6:3: error: implicit declaration of function ‘fcntl64’;
did you mean ‘fcntl’? [-Werror=implicit-function-declaration]
    6 |   fcntl64 (0, F_GETFD);
      |   ^~~~~~~
      |   fcntl
cc1: all warnings being treated as errors
$ gcc use-fcntl64.c -D _LARGEFILE64_SOURCE -Werror
(builds successfully)

So this has to be supported, and ideally tested too.

Sergey