[PATCH v7 5/8] mm/util: Fix possible race condition in kstrdup()

[Linus Torvalds]

On Sat, 17 Aug 2024 at 01:48, Alejandro Colomar <alx at kernel.org> wrote:
>
> I would compact the above to:
>
>         len = strlen(s);
>         buf = kmalloc_track_caller(len + 1, gfp);
>         if (buf)
>                 strcpy(mempcpy(buf, s, len), "");

No, we're not doing this kind of horror.

If _FORTIFY_SOURCE has problems with a simple "memcpy and add NUL",
then _FORTIFY_SOURCE needs to be fixed.

We don't replace a "buf[len] = 0" with strcpy(,""). Yes, compilers may
simplify it, but dammit, it's an unreadable incomprehensible mess to
humans, and humans still matter a LOT more.

                Linus