Re: [RFC] string: Allow 2-argument strscpy()

[Kees Cook]

On Mon, Jan 29, 2024 at 09:55:25PM +0000, Justin Stitt wrote:
> Hi,
> 
> On Mon, Jan 29, 2024 at 12:29:04PM -0800, Kees Cook wrote:
> > Using sizeof(dst) is the overwhelmingly common case for strscpy().
> > Instead of requiring this everywhere, allow a 2-argument version to be
> > used that will use the sizeof() internally.
> 
> Yeah, this is definitely the case. I have a ton of patches replacing
> strncpy with strscpy [1] and many of them match the pattern of:
> | strscpy(dest, src, sizeof(dest))
> 
> BTW, this hack for function overloading is insane. Never really looked into
> it before.

It very much is. :P Hence the RFC nature of this patch. I don't think we
any any other API in the kernel that does this (though there are plenty
of wild macro wrappers to do similar tricks, like the syscall wrappers).

> > Cc: Rasmus Villemoes <linux@xxxxxxxxxxxxxxxxxx>
> > Cc: Andy Shevchenko <andy@xxxxxxxxxx>
> > Cc: linux-hardening@xxxxxxxxxxxxxxx
> > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > ---
> > What do people think of this idea? It's usually very redundant to
> > include the 3rd argument, so this might improve readability (and
> > perhaps make things more robust by avoiding mistakes when the
> > destination name changes).
> 
> I like this, though, should you include documentation changes/additions?

Yeah, though I'm not sure how to do this -- kerndoc expects a fixed
number of arguments. :P Maybe I can just do something like add
"optional" to @size:

 * strscpy - Copy a C-string into a sized buffer
 * @p: Where to copy the string to
 * @q: Where to copy the string from
 * @size: Size of destination buffer (optional)

> 
> Reviewed-by: Justin Stitt <justinstitt@xxxxxxxxxx>
> 
> > ---
> >  include/linux/fortify-string.h | 4 ++--
> >  include/linux/string.h         | 9 ++++++++-
> >  lib/string.c                   | 4 ++--
> >  3 files changed, 12 insertions(+), 5 deletions(-)
> >
> > diff --git a/include/linux/fortify-string.h b/include/linux/fortify-string.h
> > index 89a6888f2f9e..56be4d4a5dea 100644
> > --- a/include/linux/fortify-string.h
> > +++ b/include/linux/fortify-string.h
> > @@ -215,7 +215,7 @@ __kernel_size_t __fortify_strlen(const char * const POS p)
> >  }
> >
> >  /* Defined after fortified strnlen() to reuse it. */
> > -extern ssize_t __real_strscpy(char *, const char *, size_t) __RENAME(strscpy);
> > +extern ssize_t __real_strscpy(char *, const char *, size_t) __RENAME(sized_strscpy);
> >  /**
> >   * strscpy - Copy a C-string into a sized buffer
> >   *
> > @@ -234,7 +234,7 @@ extern ssize_t __real_strscpy(char *, const char *, size_t) __RENAME(strscpy);
> >   * Returns the number of characters copied in @p (not including the
> >   * trailing %NUL) or -E2BIG if @size is 0 or the copy of @q was truncated.
> >   */
> > -__FORTIFY_INLINE ssize_t strscpy(char * const POS p, const char * const POS q, size_t size)
> > +__FORTIFY_INLINE ssize_t sized_strscpy(char * const POS p, const char * const POS q, size_t size)
> >  {
> >  	/* Use string size rather than possible enclosing struct size. */
> >  	const size_t p_size = __member_size(p);
> > diff --git a/include/linux/string.h b/include/linux/string.h
> > index ab148d8dbfc1..0bb1c8d05f18 100644
> > --- a/include/linux/string.h
> > +++ b/include/linux/string.h
> > @@ -67,9 +67,16 @@ extern char * strcpy(char *,const char *);
> >  extern char * strncpy(char *,const char *, __kernel_size_t);
> >  #endif
> >  #ifndef __HAVE_ARCH_STRSCPY
> > -ssize_t strscpy(char *, const char *, size_t);
> > +ssize_t sized_strscpy(char *, const char *, size_t);
> >  #endif
> >
> > +#define __strscpy0(dst, src, ...)	sized_strscpy(dst, src, sizeof(dst))

In thinking about this slightly longer, I realize that the size may be
better as: sizeof(dst) + __must_be_array(dst)

otherwise a "char *" will be allowed as a dst for the 2-arg method, and
will get a 1 byte size. :)

> > +
> > +#define __strscpy1(dst, src, size)	sized_strscpy(dst, src, size)
> > +

And I should probably relocate the kern-doc to here...

> > +#define strscpy(dst, src, ...)	\
> > +	CONCATENATE(__strscpy, COUNT_ARGS(__VA_ARGS__))(dst, src, __VA_ARGS__)
> > +
> >  /* Wraps calls to strscpy()/memset(), no arch specific code required */
> >  ssize_t strscpy_pad(char *dest, const char *src, size_t count);
> >
> > diff --git a/lib/string.c b/lib/string.c
> > index 6891d15ce991..2869895a1180 100644
> > --- a/lib/string.c
> > +++ b/lib/string.c
> > @@ -104,7 +104,7 @@ EXPORT_SYMBOL(strncpy);
> >  #endif
> >
> >  #ifndef __HAVE_ARCH_STRSCPY
> > -ssize_t strscpy(char *dest, const char *src, size_t count)
> > +ssize_t sized_strscpy(char *dest, const char *src, size_t count)
> >  {
> >  	const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS;
> >  	size_t max = count;
> > @@ -170,7 +170,7 @@ ssize_t strscpy(char *dest, const char *src, size_t count)
> >
> >  	return -E2BIG;
> >  }
> > -EXPORT_SYMBOL(strscpy);
> > +EXPORT_SYMBOL(sized_strscpy);
> >  #endif
> >
> >  /**
> > --
> > 2.34.1
> >
> 
> [1]: https://lore.kernel.org/all/?q=f%3A%22justinstitt%40google.com%22+AND+b%3Astrscpy+AND+NOT+s%3A%22Re%22
> 
> Thanks
> Justin

-- 
Kees Cook