Bypass taking ownership of the TPM2 if an srk_handle is specified.

This srk_handle must be usable with Null auth for the time being.

Signed-off-by: Jason Andryuk <jandryuk@gmail.com>
---
 docs/man/xen-vtpmmgr.7.pod |  7 +++++++
 stubdom/vtpmmgr/init.c     | 11 ++++++++++-
 2 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/docs/man/xen-vtpmmgr.7.pod b/docs/man/xen-vtpmmgr.7.pod
index 875dcce508..3286954568 100644
--- a/docs/man/xen-vtpmmgr.7.pod
+++ b/docs/man/xen-vtpmmgr.7.pod
@@ -92,6 +92,13 @@ Valid arguments:
 
 =over 4
 
+=item srk_handle=<HANDLE>
+
+Specify a srk_handle for TPM 2.0.  TPM 2.0 uses a key hierarchy, and
+this allow specifying the parent handle for vtpmmgr to create its own
+key under.  Using this option bypasses vtpmmgr trying to take ownership
+of the TPM.
+
 =item owner_auth=<AUTHSPEC>
 
 =item srk_auth=<AUTHSPEC>
diff --git a/stubdom/vtpmmgr/init.c b/stubdom/vtpmmgr/init.c
index 1506735051..c01d03e9f4 100644
--- a/stubdom/vtpmmgr/init.c
+++ b/stubdom/vtpmmgr/init.c
@@ -302,6 +302,11 @@ int parse_cmdline_opts(int argc, char** argv, struct Opts* opts)
             goto err_invalid;
          }
       }
+      else if(!strncmp(argv[i], "srk_handle:", 11)) {
+         if(sscanf(argv[i] + 11, "%x", &vtpm_globals.srk_handle) != 1) {
+            goto err_invalid;
+         }
+      }
       else if(!strncmp(argv[i], "tpmdriver=", 10)) {
          if(!strcmp(argv[i] + 10, "tpm_tis")) {
             opts->tpmdriver = TPMDRV_TPM_TIS;
@@ -586,7 +591,11 @@ TPM_RESULT vtpmmgr2_create(void)
 {
     TPM_RESULT status = TPM_SUCCESS;
 
-    TPMTRYRETURN(tpm2_take_ownership());
+    if ( vtpm_globals.srk_handle == 0 ) {
+        TPMTRYRETURN(tpm2_take_ownership());
+    } else {
+        tpm2_AuthArea_ctor(NULL, 0, &vtpm_globals.srk_auth_area);
+    }
 
    /* create SK */
     TPM2_Create_Params_out out;
-- 
2.30.2

Jason Andryuk, le mar. 04 mai 2021 08:48:37 -0400, a ecrit:
> Bypass taking ownership of the TPM2 if an srk_handle is specified.
> 
> This srk_handle must be usable with Null auth for the time being.
> 
> Signed-off-by: Jason Andryuk <jandryuk@gmail.com>
> ---
>  docs/man/xen-vtpmmgr.7.pod |  7 +++++++
>  stubdom/vtpmmgr/init.c     | 11 ++++++++++-
>  2 files changed, 17 insertions(+), 1 deletion(-)
> 
> diff --git a/docs/man/xen-vtpmmgr.7.pod b/docs/man/xen-vtpmmgr.7.pod
> index 875dcce508..3286954568 100644
> --- a/docs/man/xen-vtpmmgr.7.pod
> +++ b/docs/man/xen-vtpmmgr.7.pod
> @@ -92,6 +92,13 @@ Valid arguments:
>  
>  =over 4
>  
> +=item srk_handle=<HANDLE>

Is this actually srk_handle= or srk_handle: ?

The code tests for the latter. The problem seems to "exist" also for
owner_auth: and srk_auth: but both = and : work actually because strncmp
is told not to check for = and :

We'd better clean this up to avoid confusions.

Samuel

> +
> +Specify a srk_handle for TPM 2.0.  TPM 2.0 uses a key hierarchy, and
> +this allow specifying the parent handle for vtpmmgr to create its own
> +key under.  Using this option bypasses vtpmmgr trying to take ownership
> +of the TPM.
> +
>  =item owner_auth=<AUTHSPEC>
>  
>  =item srk_auth=<AUTHSPEC>
> diff --git a/stubdom/vtpmmgr/init.c b/stubdom/vtpmmgr/init.c
> index 1506735051..c01d03e9f4 100644
> --- a/stubdom/vtpmmgr/init.c
> +++ b/stubdom/vtpmmgr/init.c
> @@ -302,6 +302,11 @@ int parse_cmdline_opts(int argc, char** argv, struct Opts* opts)
>              goto err_invalid;
>           }
>        }
> +      else if(!strncmp(argv[i], "srk_handle:", 11)) {
> +         if(sscanf(argv[i] + 11, "%x", &vtpm_globals.srk_handle) != 1) {
> +            goto err_invalid;
> +         }
> +      }
>        else if(!strncmp(argv[i], "tpmdriver=", 10)) {
>           if(!strcmp(argv[i] + 10, "tpm_tis")) {
>              opts->tpmdriver = TPMDRV_TPM_TIS;
> @@ -586,7 +591,11 @@ TPM_RESULT vtpmmgr2_create(void)
>  {
>      TPM_RESULT status = TPM_SUCCESS;
>  
> -    TPMTRYRETURN(tpm2_take_ownership());
> +    if ( vtpm_globals.srk_handle == 0 ) {
> +        TPMTRYRETURN(tpm2_take_ownership());
> +    } else {
> +        tpm2_AuthArea_ctor(NULL, 0, &vtpm_globals.srk_auth_area);
> +    }
>  
>     /* create SK */
>      TPM2_Create_Params_out out;
> -- 
> 2.30.2
> 

On Tue, May 4, 2021 at 9:13 AM Samuel Thibault
<samuel.thibault@ens-lyon.org> wrote:
>
> Jason Andryuk, le mar. 04 mai 2021 08:48:37 -0400, a ecrit:
> > Bypass taking ownership of the TPM2 if an srk_handle is specified.
> >
> > This srk_handle must be usable with Null auth for the time being.
> >
> > Signed-off-by: Jason Andryuk <jandryuk@gmail.com>
> > ---
> >  docs/man/xen-vtpmmgr.7.pod |  7 +++++++
> >  stubdom/vtpmmgr/init.c     | 11 ++++++++++-
> >  2 files changed, 17 insertions(+), 1 deletion(-)
> >
> > diff --git a/docs/man/xen-vtpmmgr.7.pod b/docs/man/xen-vtpmmgr.7.pod
> > index 875dcce508..3286954568 100644
> > --- a/docs/man/xen-vtpmmgr.7.pod
> > +++ b/docs/man/xen-vtpmmgr.7.pod
> > @@ -92,6 +92,13 @@ Valid arguments:
> >
> >  =over 4
> >
> > +=item srk_handle=<HANDLE>
>
> Is this actually srk_handle= or srk_handle: ?

Whoops.  It's srk_handle: .  I just copy and pasted here.

> The code tests for the latter. The problem seems to "exist" also for
> owner_auth: and srk_auth: but both = and : work actually because strncmp
> is told not to check for = and :

owner_auth & srk_auth don't check :, but then they don't skip : or =
when passing the string to parse_auth_string.  So they can't work
properly?

srk_handle: does check for that entire string.

> We'd better clean this up to avoid confusions.

Right, so what do we want?  I'm leaning toward standardizing on =
since the tpm.*= options look to parse properly.  Given : doesn't seem
like it could work, we don't need to attempt to maintain backwards
compatibility.

Thanks for the review.

-Jason

Jason Andryuk, le mar. 04 mai 2021 13:04:47 -0400, a ecrit:
> owner_auth & srk_auth don't check :, but then they don't skip : or =
> when passing the string to parse_auth_string.  So they can't work
> properly?

They happen to "work" just because there is no other parameter prefixed
the same.

> > We'd better clean this up to avoid confusions.
> 
> Right, so what do we want?  I'm leaning toward standardizing on =
> since the tpm.*= options look to parse properly.

I'd say so too. Also because that's what is apparently documented.

Samuel

On Tue, May 4, 2021 at 1:07 PM Samuel Thibault
<samuel.thibault@ens-lyon.org> wrote:
>
> Jason Andryuk, le mar. 04 mai 2021 13:04:47 -0400, a ecrit:
> > owner_auth & srk_auth don't check :, but then they don't skip : or =
> > when passing the string to parse_auth_string.  So they can't work
> > properly?
>
> They happen to "work" just because there is no other parameter prefixed
> the same.

parse_auth_string fails on the ":".

Just tested "owner_auth:well-known"
ERROR[VTPM]: Invalid auth string :well-known
ERROR[VTPM]: Invalid Option owner_auth:well-known
ERROR[VTPM]: Command line parsing failed! exiting..

> > > We'd better clean this up to avoid confusions.
> >
> > Right, so what do we want?  I'm leaning toward standardizing on =
> > since the tpm.*= options look to parse properly.
>
> I'd say so too. Also because that's what is apparently documented.

Ok, thanks.

Regards,
Jason

Jason Andryuk, le mar. 04 mai 2021 13:27:36 -0400, a ecrit:
> On Tue, May 4, 2021 at 1:07 PM Samuel Thibault
> <samuel.thibault@ens-lyon.org> wrote:
> >
> > Jason Andryuk, le mar. 04 mai 2021 13:04:47 -0400, a ecrit:
> > > owner_auth & srk_auth don't check :, but then they don't skip : or =
> > > when passing the string to parse_auth_string.  So they can't work
> > > properly?
> >
> > They happen to "work" just because there is no other parameter prefixed
> > the same.
> 
> parse_auth_string fails on the ":".
> 
> Just tested "owner_auth:well-known"

owner_auth happens to have the proper size, but srk_auth doesn't.

Samuel