mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Rusty Russell <rusty@ozlabs.org>
To: David Howells <dhowells@redhat.com>
Cc: dhowells@redhat.com, keyrings@linux-nfs.org,
	linux-crypto@vger.kernel.org,
	linux-security-module@vger.kernel.org,
	linux-kernel@vger.kernel.org, dmitry.kasatkin@intel.com,
	zohar@linux.vnet.ibm.com, arjan.van.de.ven@intel.com,
	alan.cox@intel.com, Jon Masters <jcm@jonmasters.org>
Subject: Re: [PATCH 21/21] MODSIGN: Apply signature checking to modules on module load [ver #3]
Date: Mon, 12 Dec 2011 19:39:34 +1030	[thread overview]
Message-ID: <87ty56taup.fsf@rustcorp.com.au> (raw)
In-Reply-To: <26644.1323652900@redhat.com>

On Mon, 12 Dec 2011 01:21:40 +0000, David Howells <dhowells@redhat.com> wrote:
> Rusty Russell <rusty@ozlabs.org> wrote:
> 
> > I think you misunderstand, I'm talking about the modinfo command, not
> > the .modinfo section.
> 
> Sorry, yes.  But why do you need to enhance modinfo?

I was suggesting that you want it to print the signatures, or at least
indicate their existence.  Maybe check them too, but that might be a bit
too heavy for modinfo.

> > But I need to know exactly what these version-dependent mangling of
> > modules is.  Is it real?  Is it more than strip?  Is it so hard to fix
> > that it makes sense to add 450 lines of dense kernel code to allow
> > alteration of a module after signing?
> 
> The strip program (as far as I know that's the only binutil that we need worry
> about) rearranges and reorders the section, symbol and relocation tables when
> it discards stuff, and different versions of strip have done it
> differently.

OK, then you need to generate stripped modules as part of the build,
too.  It's a bit of a pain, sure, but hardly a showstopper.

> However, you said it should be fairly easy to jump over the ELF parcel to get
> to the signature.  How do you plan on doing that?

> I presume you would just parse sufficient of the ELF to find the
> theoretical ELF EOF and then look there for a whole string of
> signatures

You could do that.  But there's an easier way.  Took me longer to figure
out the damn crypto API than actually write the module part :(

Subject: module: simple signature support.

A signature contains a magic marker: it signs everything up to the
magic marker (ie. just append them):
	SUM=`md5sum drivers/block/loop.ko | cut -d\  -f1`; echo "@Module signature:$SUM" >> drivers/block/loop.ko

We can have false positives, but at worst that make us report EINVAL
(bad signature) instead of ENOENT (no signature).

diff --git a/kernel/module.c b/kernel/module.c
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -2374,6 +2374,98 @@ free_hdr:
 	return err;
 }
 
+/* CONFIG_MODULE_SIGN implies we don't trust modules: verify signature
+ * before we interpret (almost) anything. */
+#define MOD_SIGNATURE "@Module signature:"
+
+#include <linux/ctype.h>
+#include <crypto/hash.h>
+#include <crypto/md5.h>
+
+static int from_hex(char c)
+{
+	if (isdigit(c))
+		return c - '0';
+	if (isupper(c))
+		return c - 'A' + 10;
+	return c - 'a' + 10;
+}
+
+/* A signature signs everything before it. */
+static int try_signature(void *data, void *sig, unsigned long max_sig)
+{
+	unsigned long data_len = sig - data;
+
+	sig += strlen(MOD_SIGNATURE);
+	max_sig -= strlen(MOD_SIGNATURE);
+
+	/* Dummy: accept md5 as signature. */
+	{
+		struct crypto_api_blows {
+			struct shash_desc md5;
+			char morestuff[100];
+		} m;
+		u8 digest[MD5_DIGEST_SIZE], expected[MD5_DIGEST_SIZE];
+		char *s = sig;
+		int i;
+
+		/* Not a signature? */
+		if (max_sig < MD5_DIGEST_SIZE * 2) {
+			printk("Too close to end (%lu)\n", max_sig);
+			return -ENOENT;
+		}
+
+		for (i = 0; i < MD5_DIGEST_SIZE * 2; i += 2) {
+			/* Not a signature? */
+			if (!isxdigit(s[i]) || !isxdigit(s[i+1])) {
+				printk("Not hex digit (%c)\n", s[i]);
+				return -ENOENT;
+			}
+			digest[i/2] = (from_hex(s[i])<<4) | from_hex(s[i+1]);
+		}
+
+		m.md5.tfm = crypto_alloc_shash("md5", 0, 0);
+		if (IS_ERR(m.md5.tfm))
+			return PTR_ERR(m.md5.tfm);
+		m.md5.flags = 0;
+
+		crypto_shash_digest(&m.md5, data, data_len, expected);
+		crypto_free_shash(m.md5.tfm);
+		
+		if (memcmp(digest, expected, sizeof(digest)) != 0) {
+			printk("Mismatch: given %02x%02x%02x...,"
+			       " expect %02x%02x%02x...\n",
+			       digest[0], digest[1], digest[2],
+			       expected[0], expected[1], expected[2]);
+			return -EINVAL;
+		}
+		printk("Found valid signature!\n");
+		return 0;
+	}
+}
+
+/* -ENOENT if no signature found, -EINVAL if invalid, 0 if good. */
+static int find_and_check_signatures(struct load_info *info)
+{
+	void *p = info->hdr, *end = p + info->len;
+	const size_t sigsize = strlen(MOD_SIGNATURE);
+	int err = -ENOENT;
+
+	/* Poor man's memmem. len > sigsize */
+	while ((p = memchr(p, MOD_SIGNATURE[0], end - p))) {
+		if (p + sigsize > end)
+			break;
+
+		if (memcmp(p, MOD_SIGNATURE, sigsize) == 0) {
+			err = try_signature(info->hdr, p, end - p);
+			if (!err)
+				break;
+		}
+		p++;
+	}
+	return err;
+}
+
 static void free_copy(struct load_info *info)
 {
 	vfree(info->hdr);
@@ -2819,6 +2911,11 @@ static struct module *load_module(void _
 	if (err)
 		return ERR_PTR(err);
 
+	/* Before we trust it, carefully check signatures. */
+	err = find_and_check_signatures(&info);
+	if (err)
+		goto free_copy;
+
 	/* Figure out module layout, and allocate all the memory. */
 	mod = layout_and_allocate(&info);
 	if (IS_ERR(mod)) {



  reply	other threads:[~2011-12-12  9:11 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-12-02 18:42 [RFC][PATCH 00/21] Crypto keys and module signing " David Howells
2011-12-02 18:42 ` [PATCH 01/21] MPILIB: Export some more symbols " David Howells
2011-12-02 18:42 ` [PATCH 02/21] MPILIB: Add a missing ENOMEM check " David Howells
2011-12-02 18:43 ` [PATCH 03/21] KEYS: Permit key_serial() to be called with a const key pointer " David Howells
2011-12-02 18:43 ` [PATCH 04/21] KEYS: Move the key config into security/keys/Kconfig " David Howells
2011-12-02 18:43 ` [PATCH 05/21] KEYS: Announce key type (un)registration " David Howells
2011-12-02 18:43 ` [PATCH 06/21] KEYS: Reorganise keys Makefile " David Howells
2011-12-02 18:43 ` [PATCH 07/21] KEYS: Create a key type that can be used for general cryptographic operations " David Howells
2012-01-16 12:53   ` Mimi Zohar
2012-01-17 15:32   ` David Howells
2012-01-18 10:56     ` Kasatkin, Dmitry
2011-12-02 18:44 ` [PATCH 08/21] KEYS: Add signature verification facility " David Howells
2012-01-18 11:20   ` Kasatkin, Dmitry
2012-01-18 12:26   ` David Howells
2012-01-18 13:26     ` Kasatkin, Dmitry
2012-01-18 15:13     ` David Howells
2012-01-18 15:20       ` Kasatkin, Dmitry
2012-01-18 19:59       ` David Howells
2012-01-20  1:52         ` Herbert Xu
2011-12-02 18:44 ` [PATCH 09/21] KEYS: Asymmetric public-key algorithm crypto key subtype " David Howells
2011-12-02 18:44 ` [PATCH 10/21] KEYS: DSA signature verification algorithm " David Howells
2011-12-02 18:44 ` [PATCH 11/21] KEYS: RSA " David Howells
2011-12-02 18:44 ` [PATCH 12/21] PGPLIB: PGP definitions (RFC 4880) " David Howells
2011-12-02 18:45 ` [PATCH 13/21] PGPLIB: Basic packet parser " David Howells
2011-12-02 18:45 ` [PATCH 14/21] PGPLIB: Signature " David Howells
2011-12-02 18:45 ` [PATCH 15/21] KEYS: PGP data " David Howells
2011-12-02 18:45 ` [PATCH 16/21] KEYS: PGP-based public key signature verification " David Howells
2012-01-18 11:36   ` Kasatkin, Dmitry
2012-01-18 12:49   ` David Howells
2012-01-18 13:34     ` Kasatkin, Dmitry
2011-12-02 18:46 ` [PATCH 17/21] KEYS: PGP format signature parser " David Howells
2011-12-02 18:46 ` [PATCH 18/21] KEYS: Provide a function to load keys from a PGP keyring blob " David Howells
2011-12-02 18:46 ` [PATCH 19/21] MODSIGN: Add indications of module ELF types " David Howells
2011-12-02 18:46 ` [PATCH 20/21] MODSIGN: Module ELF verifier " David Howells
2011-12-02 18:46 ` [PATCH 21/21] MODSIGN: Apply signature checking to modules on module load " David Howells
2011-12-09 11:18   ` Rusty Russell
2011-12-09 18:43   ` David Howells
2011-12-10  7:01     ` Rusty Russell
2011-12-10 18:37       ` Arjan van de Ven
2011-12-11  4:59         ` Rusty Russell
2011-12-10 14:08     ` David Howells
2011-12-11  4:57       ` Rusty Russell
2011-12-12  1:21       ` David Howells
2011-12-12  9:09         ` Rusty Russell [this message]
2011-12-12 16:11         ` David Howells
2011-12-13  2:15           ` Rusty Russell
2011-12-15  0:14           ` David Howells
2011-12-16  0:41             ` Rusty Russell
2012-01-08 22:02 ` [RFC][PATCH 00/21] Crypto keys and module signing " Mimi Zohar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87ty56taup.fsf@rustcorp.com.au \
    --to=rusty@ozlabs.org \
    --cc=alan.cox@intel.com \
    --cc=arjan.van.de.ven@intel.com \
    --cc=dhowells@redhat.com \
    --cc=dmitry.kasatkin@intel.com \
    --cc=jcm@jonmasters.org \
    --cc=keyrings@linux-nfs.org \
    --cc=linux-crypto@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-security-module@vger.kernel.org \
    --cc=zohar@linux.vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

Powered by JetHome