Message-ID: <CAHC9VhTkpSa665tesTEs8gBjaD3ahUMATGMXuGy+-unt7WL-UQ@mail.gmail.com>
Date: Wed, 20 Mar 2024 13:23:17 -0400
From: Paul Moore <paul@...l-moore.com>
To: Mike Snitzer <snitzer@...nel.org>
Cc: Fan Wu <wufan@...ux.microsoft.com>, corbet@....net, zohar@...ux.ibm.com, 
	jmorris@...ei.org, serge@...lyn.com, tytso@....edu, ebiggers@...nel.org, 
	axboe@...nel.dk, agk@...hat.com, eparis@...hat.com, linux-doc@...r.kernel.org, 
	linux-integrity@...r.kernel.org, linux-security-module@...r.kernel.org, 
	fsverity@...ts.linux.dev, linux-block@...r.kernel.org, 
	dm-devel@...ts.linux.dev, audit@...r.kernel.org, linux-kernel@...r.kernel.org, 
	Deven Bowers <deven.desai@...ux.microsoft.com>
Subject: Re: [PATCH RFC v15 14/21] dm verity: consume root hash digest and
 signature data via LSM hook

On Tue, Mar 19, 2024 at 10:19 PM Mike Snitzer <snitzer@...nel.org> wrote:
> On Tue, Mar 19 2024 at  7:00P -0400,
> Paul Moore <paul@...l-moore.com> wrote:
> > On Mar 15, 2024 Fan Wu <wufan@...ux.microsoft.com> wrote:
> > >
> > > dm-verity provides a strong guarantee of a block device's integrity. As
> > > a generic way to check the integrity of a block device, it provides
> > > those integrity guarantees to its higher layers, including the filesystem
> > > level.
> > >
> > > An LSM that control access to a resource on the system based on the
> > > available integrity claims can use this transitive property of
> > > dm-verity, by querying the underlying block_device of a particular
> > > file.
> > >
> > > The digest and signature information need to be stored in the block
> > > device to fulfill the next requirement of authorization via LSM policy.
> > > This will enable the LSM to perform revocation of devices that are still
> > > mounted, prohibiting execution of files that are no longer authorized
> > > by the LSM in question.
> > >
> > > This patch adds two security hook calls in dm-verity to save the
> > > dm-verity roothash and the roothash signature to the block device's
> > > LSM blobs. The hook calls are depended on CONFIG_IPE_PROP_DM_VERITY,
> > > which will be introduced in the next commit.
> > >
> > > Signed-off-by: Deven Bowers <deven.desai@...ux.microsoft.com>
> > > Signed-off-by: Fan Wu <wufan@...ux.microsoft.com>
> > > ---
> > > v2:
> > >   + No Changes
> > >
> > > v3:
> > >   + No changes
> > >
> > > v4:
> > >   + No changes
> > >
> > > v5:
> > >   + No changes
> > >
> > > v6:
> > >   + Fix an improper cleanup that can result in
> > >     a leak
> > >
> > > v7:
> > >   + Squash patch 08/12, 10/12 to [11/16]
> > >   + Use part0 for block_device, to retrieve the block_device, when
> > >     calling security_bdev_setsecurity
> > >
> > > v8:
> > >   + Undo squash of 08/12, 10/12 - separating drivers/md/ from
> > >     security/ & block/
> > >   + Use common-audit function for dmverity_signature.
> > >   + Change implementation for storing the dm-verity digest to use the
> > >     newly introduced dm_verity_digest structure introduced in patch
> > >     14/20.
> > >   + Create new structure, dm_verity_digest, containing digest algorithm,
> > >     size, and digest itself to pass to the LSM layer. V7 was missing the
> > >     algorithm.
> > >   + Create an associated public header containing this new structure and
> > >     the key values for the LSM hook, specific to dm-verity.
> > >   + Additional information added to commit, discussing the layering of
> > >     the changes and how the information passed will be used.
> > >
> > > v9:
> > >   + No changes
> > >
> > > v10:
> > >   + No changes
> > >
> > > v11:
> > >   + Add an optional field to save signature
> > >   + Move the security hook call to the new finalize hook
> > >
> > > v12:
> > >   + No changes
> > >
> > > v13:
> > >   + No changes
> > >
> > > v14:
> > >   + Correct code format
> > >   + Remove unnecessary header and switch to dm_disk()
> > >
> > > v15:
> > >   + Refactor security_bdev_setsecurity() to security_bdev_setintegrity()
> > >   + Remove unnecessary headers
> > > ---
> > >  drivers/md/dm-verity-target.c | 73 +++++++++++++++++++++++++++++++++++
> > >  drivers/md/dm-verity.h        |  6 +++
> > >  include/linux/dm-verity.h     | 12 ++++++
> > >  include/linux/security.h      |  2 +
> > >  4 files changed, 93 insertions(+)
> > >  create mode 100644 include/linux/dm-verity.h
> > >
> > > diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
> > > index bb5da66da4c1..e94cc6a755d5 100644
> > > --- a/drivers/md/dm-verity-target.c
> > > +++ b/drivers/md/dm-verity-target.c
> > > @@ -22,6 +22,8 @@
> > >  #include <linux/scatterlist.h>
> > >  #include <linux/string.h>
> > >  #include <linux/jump_label.h>
> > > +#include <linux/security.h>
> > > +#include <linux/dm-verity.h>
> > >
> > >  #define DM_MSG_PREFIX                      "verity"
> > >
> > > @@ -1017,6 +1019,38 @@ static void verity_io_hints(struct dm_target *ti, struct queue_limits *limits)
> > >     blk_limits_io_min(limits, limits->logical_block_size);
> > >  }
> > >
> > > +#ifdef CONFIG_IPE_PROP_DM_VERITY
> > > +
> > > +static int verity_init_sig(struct dm_verity *v, const void *sig,
> > > +                      size_t sig_size)
> > > +{
> > > +   v->sig_size = sig_size;
> > > +   v->root_digest_sig = kmalloc(v->sig_size, GFP_KERNEL);
> > > +   if (!v->root_digest)
> > > +           return -ENOMEM;
> >
> > Either you meant to copy @sig into @v->root_digest_sig and forgot to
> > add the code for that, or we don't need to include @sig as a parameter
> > to this function.  I'm guessing it is the former as it wouldn't make
> > sense to even have dm_verity::root_digest_sig if we weren't stashing
> > it here.
> >
> > I'd also suggest looking at kmemdup() instead of a kmalloc()/memcpy()
> > combo.
> >
> > > +   return 0;
> > > +}
> > > +
> > > +static void verity_free_sig(struct dm_verity *v)
> > > +{
> > > +   kfree(v->root_digest_sig);
> > > +}
> > > +#else
> > > +
> > > +static inline int verity_init_sig(struct dm_verity *v, const void *sig,
> > > +                             size_t sig_size)
> > > +{
> > > +   return 0;
> > > +}
> > > +
> > > +static inline void verity_free_sig(struct dm_verity *v)
> > > +{
> > > +}
> > > +
> > > +#endif /* CONFIG_IPE_PROP_DM_VERITY */
> >
> > It's been a while since I looked at this patch in the patchset, so
> > maybe I'm missing something, but in general we don't want CONFIG_XXX
> > checks in the kernel, outside of security/, that are specific to a
> > particular LSM (what happens when multiple LSMs need this?).  Please
> > use CONFIG_SECURITY instead.
> >
> > >  static void verity_dtr(struct dm_target *ti)
> > >  {
> > >     struct dm_verity *v = ti->private;
> > > @@ -1035,6 +1069,7 @@ static void verity_dtr(struct dm_target *ti)
> > >     kfree(v->salt);
> > >     kfree(v->root_digest);
> > >     kfree(v->zero_digest);
> > > +   verity_free_sig(v);
> > >
> > >     if (v->tfm)
> > >             crypto_free_ahash(v->tfm);
> > > @@ -1434,6 +1469,13 @@ static int verity_ctr(struct dm_target *ti, unsigned int argc, char **argv)
> > >             ti->error = "Root hash verification failed";
> > >             goto bad;
> > >     }
> > > +
> > > +   r = verity_init_sig(v, verify_args.sig, verify_args.sig_size);
> > > +   if (r < 0) {
> > > +           ti->error = "Cannot allocate root digest signature";
> > > +           goto bad;
> > > +   }
> > > +
> > >     v->hash_per_block_bits =
> > >             __fls((1 << v->hash_dev_block_bits) / v->digest_size);
> > >
> > > @@ -1584,6 +1626,34 @@ int dm_verity_get_root_digest(struct dm_target *ti, u8 **root_digest, unsigned i
> > >     return 0;
> > >  }
> > >
> > > +#ifdef CONFIG_IPE_PROP_DM_VERITY
> > > +
> > > +static int verity_finalize(struct dm_target *ti)
> > > +{
> > > +   struct block_device *bdev;
> > > +   struct dm_verity_digest root_digest;
> > > +   struct dm_verity *v;
> > > +   int r;
> > > +
> > > +   v = ti->private;
> > > +   bdev = dm_disk(dm_table_get_md(ti->table))->part0;
> > > +   root_digest.digest = v->root_digest;
> > > +   root_digest.digest_len = v->digest_size;
> > > +   root_digest.alg = v->alg_name;
> > > +
> > > +   r = security_bdev_setintegrity(bdev, LSM_INTGR_DMV_ROOTHASH, &root_digest,
> > > +                                  sizeof(root_digest));
> > > +   if (r)
> > > +           return r;
> > > +
> > > +   return security_bdev_setintegrity(bdev,
> > > +                                     LSM_INTGR_DMV_SIG,
> > > +                                     v->root_digest_sig,
> > > +                                     v->sig_size);
> >
> > What happens if the second call fails, should we clear the
> > LSM_INTGR_DMV_ROOTHASH state in the LSM?
> >
> > > +}
> > > +
> > > +#endif /* CONFIG_IPE_PROP_DM_VERITY */
> >
> > See my comments about CONFIG_SECURITY above.  In fact, I would suggest
> > moving this up into that part of the file so you only need one #ifdef
> > block relating to CONFIG_SECURITY.
> >
> > I would also recommend making a dummy function so we can get rid of
> > the conditional compilation in @verity_target below.  For example:
> >
> >   #ifdef CONFIG_SECURITY
> >   static int verity_finalize(struct dm_target *ti)
> >   {
> >     /* real implementation */
> >   }
> >   #else
> >   static int verity_finalize(struct dm_target *ti)
> >   {
> >     return 0;
> >   }
> >   #endif /* CONFIG_SECURITY */
> >
> > >  static struct target_type verity_target = {
> > >     .name           = "verity",
> > >     .features       = DM_TARGET_SINGLETON | DM_TARGET_IMMUTABLE,
> > > @@ -1596,6 +1666,9 @@ static struct target_type verity_target = {
> > >     .prepare_ioctl  = verity_prepare_ioctl,
> > >     .iterate_devices = verity_iterate_devices,
> > >     .io_hints       = verity_io_hints,
> > > +#ifdef CONFIG_IPE_PROP_DM_VERITY
> > > +   .finalize       = verity_finalize,
> > > +#endif /* CONFIG_IPE_PROP_DM_VERITY */
> > >  };
> > >  module_dm(verity);
> >
> > If you create a dummy verity_finalize() function like above you can
> > get rid of the #ifdef checks.
>
> Think it is better to leave it as-is, to avoid calling the .finalize
> hook if it isn't actually needed.

Fair enough, my personal preference is to minimize Kconfig conditional
code flow changes such as this, but I understand your point of view
and device-mapper is your code after all.

I believe the other issues still need to be addressed, but other than
that are you generally okay with the new "finalize" hook approach?

-- 
paul-moore.com

Powered by blists - more mailing lists