linux-kernel - Re: [PATCH 4/4] module: Add hook for security_kernel_post_read

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: <202007071951.605F38D43@keescook>
Date:   Tue, 7 Jul 2020 20:10:35 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Mimi Zohar <zohar@...ux.ibm.com>
Cc:     James Morris <jmorris@...ei.org>, Jessica Yu <jeyu@...nel.org>,
        Luis Chamberlain <mcgrof@...nel.org>,
        Scott Branden <scott.branden@...adcom.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Dmitry Kasatkin <dmitry.kasatkin@...il.com>,
        "Serge E. Hallyn" <serge@...lyn.com>,
        Casey Schaufler <casey@...aufler-ca.com>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Matthew Garrett <matthewgarrett@...gle.com>,
        David Howells <dhowells@...hat.com>,
        Mauro Carvalho Chehab <mchehab+huawei@...nel.org>,
        Randy Dunlap <rdunlap@...radead.org>,
        "Joel Fernandes (Google)" <joel@...lfernandes.org>,
        KP Singh <kpsingh@...gle.com>, Dave Olsthoorn <dave@...aar.me>,
        Hans de Goede <hdegoede@...hat.com>,
        Peter Jones <pjones@...hat.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Stephen Boyd <stephen.boyd@...aro.org>,
        Paul Moore <paul@...l-moore.com>, linux-kernel@...r.kernel.org,
        linux-fsdevel@...r.kernel.org, linux-integrity@...r.kernel.org,
        linux-security-module@...r.kernel.org
Subject: Re: [PATCH 4/4] module: Add hook for security_kernel_post_read_file()

On Tue, Jul 07, 2020 at 08:47:20PM -0400, Mimi Zohar wrote:
> On Tue, 2020-07-07 at 01:19 -0700, Kees Cook wrote:
> > Calls to security_kernel_load_data() should be paired with a call to
> > security_kernel_post_read_file() with a NULL file argument. Add the
> > missing call so the module contents are visible to the LSMs interested
> > in measuring the module content. (This also paves the way for moving
> > module signature checking out of the module core and into an LSM.)
> > 
> > Cc: Jessica Yu <jeyu@...nel.org>
> > Fixes: c77b8cdf745d ("module: replace the existing LSM hook in init_module")
> > Signed-off-by: Kees Cook <keescook@...omium.org>
> > ---
> >  kernel/module.c | 7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/kernel/module.c b/kernel/module.c
> > index 0c6573b98c36..af9679f8e5c6 100644
> > --- a/kernel/module.c
> > +++ b/kernel/module.c
> > @@ -2980,7 +2980,12 @@ static int copy_module_from_user(const void __user *umod, unsigned long len,
> >  		return -EFAULT;
> >  	}
> >  
> > -	return 0;
> > +	err = security_kernel_post_read_file(NULL, (char *)info->hdr,
> > +					     info->len, READING_MODULE);
> 
> There was a lot of push back on calling security_kernel_read_file()
> with a NULL file descriptor here.[1]  The result was defining a new
> security hook - security_kernel_load_data - and enumeration -
> LOADING_MODULE.  I would prefer calling the same pre and post security
> hook.
> 
> Mimi
> 
> [1] http://kernsec.org/pipermail/linux-security-module-archive/2018-May/007110.html

Ah yes, thanks for the pointer to the discussion.

I think we have four cases then, for differing LSM hooks:

- no "file", no contents
	e.g. init_module() before copying user buffer
	security_kernel_load_data()
- only a "file" available, no contents
	e.g. kernel_read_file() before actually reading anything
	security_kernel_read_file()
- "file" and contents
	e.g. kernel_read_file() after reading
	security_kernel_post_read_file()
- no "file" available, just the contents
	e.g. firmware platform fallback from EFI space (no "file")
	unimplemented!

If an LSM wants to be able to examine the contents of firmware, modules,
kexec, etc, it needs either a "file" or the full contents.

The "file" methods all pass through the kernel_read_file()-family. The
others happen via blobs coming from userspace or (more recently) the EFI
universe.

So, if a NULL file is unreasonable, we need, perhaps,
security_kernel_post_load_data()

?

-- 
Kees Cook