linux-kernel - stracpy

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 20 Aug 2019 17:58:27 -0700
From:   Joe Perches <joe@...ches.com>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Stephen Rothwell <sfr@...b.auug.org.au>,
        Julia Lawall <julia.lawall@...6.fr>,
        "Gustavo A. R. Silva" <gustavo@...eddedor.com>,
        LKML <linux-kernel@...r.kernel.org>,
        clang-built-linux@...glegroups.com,
        Linux Next Mailing List <linux-next@...r.kernel.org>
Subject: stracpy

On Tue, 2019-08-20 at 17:43 -0700, Linus Torvalds wrote:
> On Tue, Aug 20, 2019 at 5:20 PM Joe Perches <joe@...ches.com> wrote:
> > Umm, btw: have you actually looked at stracpy?
> 
> Yes, Joe, I have.
> 
> What part of "there are now so many of them that no human being can
> keep track of them" didn't you see as a problem?

Well, the actual post-conversion uses to stracpy make the old ones
(strcpy/strlcpy/strncpy) exceptional uses that can be analyzed
quite a bit more easily.

btw: I really don't care what any convenience macro is named.

Most all of the strlcpy and strscpy uses actually _do_ copy to
a char array and strscpy is a simple interface that is somewhat
frequently misused.

> How many broken string functions are we going to do, adding yet
> another one when you notice that the _last_ one wasn't great?
>
> We never seem to remove the broken ones. We just add yet another one,
> and have a never-ending jumble of random letters.

<shrug> Intermediate problems.

> I would seriously suggest doing something like
>  and 
>    copy_string( dst, dstsize, src, srcsize, FLAGS );
> 
> where FLAGS migth be "pad" or whatever. Make it return the size of the
> resulting string, because while it can be convenient to pass 'dst" on,
> it's not useful.

That's simply not convenient for pointers.

Auditing the kernel for those unsized uses is a
large scope problem.  Even anything that uses
PAGE_SIZE sized allocations does sprintf
instead of snprintf.  Any show_<foo> for instance.

> And then maybe just add the helper macro that turns an array into a
> "pointer, size" combination, rather than yet another letter jumble.

Good luck with that when an unsized char pointer is the thing
passed to a function.

There are _way_ too many of those already in the kernel.
Simple strcpy is already used > 2000 times.