On Sun, 2 Jun 2024 at 10:53, Eric W. Biederman [off-list ref] wrote:

The read may race with a write that is changing the location
of '\0'.  Especially if the new value is shorter than
the old value.

It *shouldn't* happen.

So 'strscpy()' itself is written to be NUL-safe, in that if it ever
copies a NUL character, it will stop. Admittedly the byte loop at the
end might technically need a READ_ONCE() for that to eb strictly true
in theory, but in practice it already is.

And even if the new string is shorter, the comm[] array will always
have a NUL terminator _somewhere_, in how the last byte is never
non-NUL.

Now, the only real issue is if something writes *to* the  comm[] array
without following the rules properly - like writing a non-NULL
character to the end of the array before then filling it in with NUL
again.

But that would be a bug on the comm[] writing side, I feel, not a bug
on the reader side.

               Linus