linux-kernel - Re: [PATCH] mm: export copy_to_kernel

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]

Message-ID: <CANpmjNPvBnov-EFk1PNO4GEOF7XLG7S1bYYjg9i4Ej=ZzaA6ag@mail.gmail.com>
Date: Wed, 23 Oct 2024 09:02:23 +0200
From: Marco Elver <elver@...gle.com>
To: Christoph Hellwig <hch@...radead.org>
Cc: Arnd Bergmann <arnd@...nel.org>, Andrew Morton <akpm@...ux-foundation.org>, 
	Sabyrzhan Tasbolatov <snovitoll@...il.com>, Andrey Konovalov <andreyknvl@...il.com>, 
	Arnd Bergmann <arnd@...db.de>, linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm: export copy_to_kernel_nofault

On Wed, 23 Oct 2024 at 08:59, Christoph Hellwig <hch@...radead.org> wrote:
>
> On Wed, Oct 23, 2024 at 08:57:02AM +0200, Marco Elver wrote:
> > On Wed, 23 Oct 2024 at 08:53, Christoph Hellwig <hch@...radead.org> wrote:
> > >
> > > On Fri, Oct 18, 2024 at 03:11:09PM +0000, Arnd Bergmann wrote:
> > > > From: Arnd Bergmann <arnd@...db.de>
> > > >
> > > > This symbol is now used on the kasan test module, so it needs to be
> > > > exported.
> > > >
> > > > ERROR: modpost: "copy_to_kernel_nofault" [mm/kasan/kasan_test.ko] undefined!
> > >
> > > Meh, it would be great not to export internal helpers just because
> > > someone wants to test them.  Please just mark that test built-in only
> > > instead.
> >
> > We have EXPORT_SYMBOL_IF_KUNIT. See include/kunit/visibility.h -
> > that's more appropriate, and also adjust kasan_test.c to do
> > MODULE_IMPORT_NS(EXPORTED_FOR_KUNIT_TESTING).
>
> Thats a little better, but at least in this case I still think it is
> a very bad idea.  copy_to_kernel_nofault is a perfect vector for
> exploit code to probe writing to kernel address without causing faults
> so it really should never ever be exported.

Another alternative is to just #ifndef MODULE the offending test case,
so it's only available if built-in. No need to just make the whole
test built-in only. I know there are users of this particular test
that rely on it being a module.