Re: [PATCH] libselinux: android: fix lax service context lookup

[Stephen Smalley]

On 09/28/2016 12:35 PM, Janis Danisevskis wrote:
> 
> 
> On Wed, Sep 28, 2016 at 5:17 PM, Stephen Smalley  > wrote:
> 
> On 09/28/2016 12:04 PM, Janis Danisevskis wrote:
> > We use the same lookup function for service contexts
> > that we use for property contexts. However, property
> > contexts are namespace based and only compare the
> > prefix. This may lead to service associations with
> > a wrong label.
> >
> > This patch introduces a stricter lookup function for
> > services contexts. Now the service name must match
> > the key of the service label exactly.
> >
> > Signed-off-by: Janis Danisevskis  >
> > ---
> >  libselinux/include/selinux/label.h  |  2 ++
> >  libselinux/src/label.c  |  1 +
> >  libselinux/src/label_android_property.c | 50 
> +
> >  libselinux/src/label_internal.h |  3 ++
> >  4 files changed, 56 insertions(+)
> 
> Normally each backend would go into its own file, so service would get
> its own.  Alternatively, since we are unlikely to ever support one
> without the other, perhaps label_android_property.c should be renamed to
> label_android.c and contain all of the Android-unique backends.
> 
> I was thinking along the same lines, was just eager to get home... I'll send
> a refactored patch by tomorrow.

Also, please update libselinux/utils/selabel_lookup.c and
selabel_digest.c to know about your new backend.
>  
> 
> >
> > diff --git a/libselinux/include/selinux/label.h 
> b/libselinux/include/selinux/label.h
> > index f0b1e10..277287e 100644
> > --- a/libselinux/include/selinux/label.h
> > +++ b/libselinux/include/selinux/label.h
> > @@ -34,6 +34,8 @@ struct selabel_handle;
> >  #define SELABEL_CTX_DB   3
> >  /* Android property service contexts */
> >  #define SELABEL_CTX_ANDROID_PROP 4
> > +/* Android service contexts */
> > +#define SELABEL_CTX_ANDROID_SERVICE 5
> >
> >  /*
> >   * Available options
> > diff --git a/libselinux/src/label.c b/libselinux/src/label.c
> > index 96a4ff1..eb0e766 100644
> > --- a/libselinux/src/label.c
> > +++ b/libselinux/src/label.c
> > @@ -45,6 +45,7 @@ static selabel_initfunc initfuncs[] = {
> >   CONFIG_X_BACKEND(selabel_x_init),
> >   CONFIG_DB_BACKEND(selabel_db_init),
> >   _property_init,
> > + _service_init,
> 
> Wondering if we should support selective enablement of the property and
> service backends too, similar to what William introduced for media, x,
> and db so that he could disable them on Android (in our case, so we can
> disable property and service backends on Linux distributions).
> 
> 
> Oh, that is what these wrappers are for :-) . Sure.
>  
> 
> 
> >  };
> >
> >  static void selabel_subs_fini(struct selabel_sub *ptr)
> > diff --git a/libselinux/src/label_android_property.c
> b/libselinux/src/label_android_property.c
> > index 290b438..69d6afd 100644
> > --- a/libselinux/src/label_android_property.c
> > +++ b/libselinux/src/label_android_property.c
> > @@ -279,6 +279,38 @@ finish:
> >   return ret;
> >  }
> >
> > +static struct selabel_lookup_rec *service_lookup(struct
> selabel_handle *rec,
> > + const char *key, int __attribute__((unused)) type)
> > +{
> > + struct saved_data *data = (struct saved_data *)rec->data;
> > + spec_t *spec_arr = data->spec_arr;
> > + unsigned int i;
> > + struct selabel_lookup_rec *ret = NULL;
> > +
> > + if (!data->nspec) {
> > + errno = ENOENT;
> > + goto finish;
> > + }
> > +
> > + for (i = 0; i < data->nspec; i++) {
> > + if (strcmp(spec_arr[i].property_key, key) == 0)
> > + break;
> > + if (strcmp(spec_arr[i].property_key, "*") == 0)
> > + break;
> > + }
> > +
> > + if (i >= data->nspec) {
> > + /* No matching specification. */
> > + errno = ENOENT;
> > + goto finish;
> > + }
> > +
> > + ret = _arr[i].lr;
> > +
> > +finish:
> > + return ret;
> > +}
> > +
> >  static void stats(struct selabel_handle __attribute__((unused)) *rec)
> >  {
> >   selinux_log(SELINUX_WARNING, "'stats' functionality not
> implemented.\n");
> > @@ -302,3 +334,21 @@ int selabel_property_init(struct
> selabel_handle *rec,
> >
> >   return init(rec, opts, nopts);
> >  }
> > +
> > +int selabel_service_init(struct selabel_handle *rec,
> > + const struct selinux_opt *opts, unsigned nopts)
> > +{
> > + struct 

Re: [PATCH] libselinux: android: fix lax service context lookup

[Stephen Smalley]

On 09/28/2016 12:43 PM, William Roberts wrote:
> On Wed, Sep 28, 2016 at 12:42 PM, Stephen Smalley  wrote:
>> On 09/28/2016 12:25 PM, William Roberts wrote:
>>> On Wed, Sep 28, 2016 at 12:17 PM, Stephen Smalley  
>>> wrote:
 On 09/28/2016 12:04 PM, Janis Danisevskis wrote:
> We use the same lookup function for service contexts
> that we use for property contexts. However, property
> contexts are namespace based and only compare the
> prefix. This may lead to service associations with
> a wrong label.
>
> This patch introduces a stricter lookup function for
> services contexts. Now the service name must match
> the key of the service label exactly.
>
> Signed-off-by: Janis Danisevskis 
> ---
>  libselinux/include/selinux/label.h  |  2 ++
>  libselinux/src/label.c  |  1 +
>  libselinux/src/label_android_property.c | 50 
> +
>  libselinux/src/label_internal.h |  3 ++
>  4 files changed, 56 insertions(+)

 Normally each backend would go into its own file, so service would get
 its own.  Alternatively, since we are unlikely to ever support one
 without the other, perhaps label_android_property.c should be renamed to
 label_android.c and contain all of the Android-unique backends.

>
> diff --git a/libselinux/include/selinux/label.h 
> b/libselinux/include/selinux/label.h
> index f0b1e10..277287e 100644
> --- a/libselinux/include/selinux/label.h
> +++ b/libselinux/include/selinux/label.h
> @@ -34,6 +34,8 @@ struct selabel_handle;
>  #define SELABEL_CTX_DB   3
>  /* Android property service contexts */
>  #define SELABEL_CTX_ANDROID_PROP 4
> +/* Android service contexts */
> +#define SELABEL_CTX_ANDROID_SERVICE 5
>
>  /*
>   * Available options
> diff --git a/libselinux/src/label.c b/libselinux/src/label.c
> index 96a4ff1..eb0e766 100644
> --- a/libselinux/src/label.c
> +++ b/libselinux/src/label.c
> @@ -45,6 +45,7 @@ static selabel_initfunc initfuncs[] = {
>   CONFIG_X_BACKEND(selabel_x_init),
>   CONFIG_DB_BACKEND(selabel_db_init),
>   _property_init,
> + _service_init,

 Wondering if we should support selective enablement of the property and
 service backends too, similar to what William introduced for media, x,
 and db so that he could disable them on Android (in our case, so we can
 disable property and service backends on Linux distributions).
>>>
>>> I was wondering that too, im for it. If ANDROID_HOST patch is applied, we
>>> should just set the defaults to strip them out and only on ANDROID_HOST=y
>>> add them in.
>>>
>>> We'd also need to coordinate with the AOSP patches, but I can
>>> routinely update those
>>> based on whats going on.
>>
>> I could be wrong, but I think we only need the property and service
>> backends on the target, not on the build host.  The file backend is
>> needed on the build host to label the filesystem images when they are
>> created.  We are likely only building the property backend on the host
>> because we don't allow conditionally excluding it presently.
> 
> checkfc I thought uses them for checking property and service backends?

Ah, you're right.  Never mind...

___
Seandroid-list mailing list
Seandroid-list@tycho.nsa.gov
To unsubscribe, send email to seandroid-list-le...@tycho.nsa.gov.
To get help, send an email containing "help" to 
seandroid-list-requ...@tycho.nsa.gov.

Re: [PATCH] libselinux: android: fix lax service context lookup

[William Roberts]

On Wed, Sep 28, 2016 at 12:42 PM, Stephen Smalley  wrote:
> On 09/28/2016 12:25 PM, William Roberts wrote:
>> On Wed, Sep 28, 2016 at 12:17 PM, Stephen Smalley  wrote:
>>> On 09/28/2016 12:04 PM, Janis Danisevskis wrote:
 We use the same lookup function for service contexts
 that we use for property contexts. However, property
 contexts are namespace based and only compare the
 prefix. This may lead to service associations with
 a wrong label.

 This patch introduces a stricter lookup function for
 services contexts. Now the service name must match
 the key of the service label exactly.

 Signed-off-by: Janis Danisevskis 
 ---
  libselinux/include/selinux/label.h  |  2 ++
  libselinux/src/label.c  |  1 +
  libselinux/src/label_android_property.c | 50 
 +
  libselinux/src/label_internal.h |  3 ++
  4 files changed, 56 insertions(+)
>>>
>>> Normally each backend would go into its own file, so service would get
>>> its own.  Alternatively, since we are unlikely to ever support one
>>> without the other, perhaps label_android_property.c should be renamed to
>>> label_android.c and contain all of the Android-unique backends.
>>>

 diff --git a/libselinux/include/selinux/label.h 
 b/libselinux/include/selinux/label.h
 index f0b1e10..277287e 100644
 --- a/libselinux/include/selinux/label.h
 +++ b/libselinux/include/selinux/label.h
 @@ -34,6 +34,8 @@ struct selabel_handle;
  #define SELABEL_CTX_DB   3
  /* Android property service contexts */
  #define SELABEL_CTX_ANDROID_PROP 4
 +/* Android service contexts */
 +#define SELABEL_CTX_ANDROID_SERVICE 5

  /*
   * Available options
 diff --git a/libselinux/src/label.c b/libselinux/src/label.c
 index 96a4ff1..eb0e766 100644
 --- a/libselinux/src/label.c
 +++ b/libselinux/src/label.c
 @@ -45,6 +45,7 @@ static selabel_initfunc initfuncs[] = {
   CONFIG_X_BACKEND(selabel_x_init),
   CONFIG_DB_BACKEND(selabel_db_init),
   _property_init,
 + _service_init,
>>>
>>> Wondering if we should support selective enablement of the property and
>>> service backends too, similar to what William introduced for media, x,
>>> and db so that he could disable them on Android (in our case, so we can
>>> disable property and service backends on Linux distributions).
>>
>> I was wondering that too, im for it. If ANDROID_HOST patch is applied, we
>> should just set the defaults to strip them out and only on ANDROID_HOST=y
>> add them in.
>>
>> We'd also need to coordinate with the AOSP patches, but I can
>> routinely update those
>> based on whats going on.
>
> I could be wrong, but I think we only need the property and service
> backends on the target, not on the build host.  The file backend is
> needed on the build host to label the filesystem images when they are
> created.  We are likely only building the property backend on the host
> because we don't allow conditionally excluding it presently.

checkfc I thought uses them for checking property and service backends?

>
> OTOH, being able to look things up on the build host could be useful
> from a development POV, e.g. if you were to add utils/selabel_lookup to
> the build host targets and kept the property and service backends in the
> host libselinux, then one could check what would match a given key.



-- 
Respectfully,

William C Roberts
___
Seandroid-list mailing list
Seandroid-list@tycho.nsa.gov
To unsubscribe, send email to seandroid-list-le...@tycho.nsa.gov.
To get help, send an email containing "help" to 
seandroid-list-requ...@tycho.nsa.gov.

Re: [PATCH] libselinux: android: fix lax service context lookup

[Janis Danisevskis]

On Wed, Sep 28, 2016 at 5:17 PM, Stephen Smalley  wrote:

> On 09/28/2016 12:04 PM, Janis Danisevskis wrote:
> > We use the same lookup function for service contexts
> > that we use for property contexts. However, property
> > contexts are namespace based and only compare the
> > prefix. This may lead to service associations with
> > a wrong label.
> >
> > This patch introduces a stricter lookup function for
> > services contexts. Now the service name must match
> > the key of the service label exactly.
> >
> > Signed-off-by: Janis Danisevskis 
> > ---
> >  libselinux/include/selinux/label.h  |  2 ++
> >  libselinux/src/label.c  |  1 +
> >  libselinux/src/label_android_property.c | 50
> +
> >  libselinux/src/label_internal.h |  3 ++
> >  4 files changed, 56 insertions(+)
>
> Normally each backend would go into its own file, so service would get
> its own.  Alternatively, since we are unlikely to ever support one
> without the other, perhaps label_android_property.c should be renamed to
> label_android.c and contain all of the Android-unique backends.
>
> I was thinking along the same lines, was just eager to get home... I'll
send
a refactored patch by tomorrow.


> >
> > diff --git a/libselinux/include/selinux/label.h
> b/libselinux/include/selinux/label.h
> > index f0b1e10..277287e 100644
> > --- a/libselinux/include/selinux/label.h
> > +++ b/libselinux/include/selinux/label.h
> > @@ -34,6 +34,8 @@ struct selabel_handle;
> >  #define SELABEL_CTX_DB   3
> >  /* Android property service contexts */
> >  #define SELABEL_CTX_ANDROID_PROP 4
> > +/* Android service contexts */
> > +#define SELABEL_CTX_ANDROID_SERVICE 5
> >
> >  /*
> >   * Available options
> > diff --git a/libselinux/src/label.c b/libselinux/src/label.c
> > index 96a4ff1..eb0e766 100644
> > --- a/libselinux/src/label.c
> > +++ b/libselinux/src/label.c
> > @@ -45,6 +45,7 @@ static selabel_initfunc initfuncs[] = {
> >   CONFIG_X_BACKEND(selabel_x_init),
> >   CONFIG_DB_BACKEND(selabel_db_init),
> >   _property_init,
> > + _service_init,
>
> Wondering if we should support selective enablement of the property and
> service backends too, similar to what William introduced for media, x,
> and db so that he could disable them on Android (in our case, so we can
> disable property and service backends on Linux distributions).
>

Oh, that is what these wrappers are for :-) . Sure.


>
> >  };
> >
> >  static void selabel_subs_fini(struct selabel_sub *ptr)
> > diff --git a/libselinux/src/label_android_property.c
> b/libselinux/src/label_android_property.c
> > index 290b438..69d6afd 100644
> > --- a/libselinux/src/label_android_property.c
> > +++ b/libselinux/src/label_android_property.c
> > @@ -279,6 +279,38 @@ finish:
> >   return ret;
> >  }
> >
> > +static struct selabel_lookup_rec *service_lookup(struct selabel_handle
> *rec,
> > + const char *key, int __attribute__((unused)) type)
> > +{
> > + struct saved_data *data = (struct saved_data *)rec->data;
> > + spec_t *spec_arr = data->spec_arr;
> > + unsigned int i;
> > + struct selabel_lookup_rec *ret = NULL;
> > +
> > + if (!data->nspec) {
> > + errno = ENOENT;
> > + goto finish;
> > + }
> > +
> > + for (i = 0; i < data->nspec; i++) {
> > + if (strcmp(spec_arr[i].property_key, key) == 0)
> > + break;
> > + if (strcmp(spec_arr[i].property_key, "*") == 0)
> > + break;
> > + }
> > +
> > + if (i >= data->nspec) {
> > + /* No matching specification. */
> > + errno = ENOENT;
> > + goto finish;
> > + }
> > +
> > + ret = _arr[i].lr;
> > +
> > +finish:
> > + return ret;
> > +}
> > +
> >  static void stats(struct selabel_handle __attribute__((unused)) *rec)
> >  {
> >   selinux_log(SELINUX_WARNING, "'stats' functionality not
> implemented.\n");
> > @@ -302,3 +334,21 @@ int selabel_property_init(struct selabel_handle
> *rec,
> >
> >   return init(rec, opts, nopts);
> >  }
> > +
> > +int selabel_service_init(struct selabel_handle *rec,
> > + const struct selinux_opt *opts, unsigned nopts)
> > +{
> > + struct saved_data *data;
> > +
> > + data = (struct saved_data *)malloc(sizeof(*data));
> > + if (!data)
> > + return -1;
> > + memset(data, 0, sizeof(*data));
> > +
> > + rec->data = data;
> > + rec->func_close = 
> > + rec->func_stats = 
> > + rec->func_lookup = _lookup;
> > +
> > + return init(rec, opts, nopts);
> > +}
> > diff --git a/libselinux/src/label_internal.h b/libselinux/src/label_
> internal.h
> > index 7c55531..6a9481a 100644
> > --- a/libselinux/src/label_internal.h
> > +++ b/libselinux/src/label_internal.h
> > @@ -39,6 +39,9 @@ int selabel_db_init(struct selabel_handle *rec,
> >  int selabel_property_init(struct 

Re: [PATCH] libselinux: android: fix lax service context lookup

[Stephen Smalley]

On 09/28/2016 12:25 PM, William Roberts wrote:
> On Wed, Sep 28, 2016 at 12:17 PM, Stephen Smalley  wrote:
>> On 09/28/2016 12:04 PM, Janis Danisevskis wrote:
>>> We use the same lookup function for service contexts
>>> that we use for property contexts. However, property
>>> contexts are namespace based and only compare the
>>> prefix. This may lead to service associations with
>>> a wrong label.
>>>
>>> This patch introduces a stricter lookup function for
>>> services contexts. Now the service name must match
>>> the key of the service label exactly.
>>>
>>> Signed-off-by: Janis Danisevskis 
>>> ---
>>>  libselinux/include/selinux/label.h  |  2 ++
>>>  libselinux/src/label.c  |  1 +
>>>  libselinux/src/label_android_property.c | 50 
>>> +
>>>  libselinux/src/label_internal.h |  3 ++
>>>  4 files changed, 56 insertions(+)
>>
>> Normally each backend would go into its own file, so service would get
>> its own.  Alternatively, since we are unlikely to ever support one
>> without the other, perhaps label_android_property.c should be renamed to
>> label_android.c and contain all of the Android-unique backends.
>>
>>>
>>> diff --git a/libselinux/include/selinux/label.h 
>>> b/libselinux/include/selinux/label.h
>>> index f0b1e10..277287e 100644
>>> --- a/libselinux/include/selinux/label.h
>>> +++ b/libselinux/include/selinux/label.h
>>> @@ -34,6 +34,8 @@ struct selabel_handle;
>>>  #define SELABEL_CTX_DB   3
>>>  /* Android property service contexts */
>>>  #define SELABEL_CTX_ANDROID_PROP 4
>>> +/* Android service contexts */
>>> +#define SELABEL_CTX_ANDROID_SERVICE 5
>>>
>>>  /*
>>>   * Available options
>>> diff --git a/libselinux/src/label.c b/libselinux/src/label.c
>>> index 96a4ff1..eb0e766 100644
>>> --- a/libselinux/src/label.c
>>> +++ b/libselinux/src/label.c
>>> @@ -45,6 +45,7 @@ static selabel_initfunc initfuncs[] = {
>>>   CONFIG_X_BACKEND(selabel_x_init),
>>>   CONFIG_DB_BACKEND(selabel_db_init),
>>>   _property_init,
>>> + _service_init,
>>
>> Wondering if we should support selective enablement of the property and
>> service backends too, similar to what William introduced for media, x,
>> and db so that he could disable them on Android (in our case, so we can
>> disable property and service backends on Linux distributions).
> 
> I was wondering that too, im for it. If ANDROID_HOST patch is applied, we
> should just set the defaults to strip them out and only on ANDROID_HOST=y
> add them in.
> 
> We'd also need to coordinate with the AOSP patches, but I can
> routinely update those
> based on whats going on.

I could be wrong, but I think we only need the property and service
backends on the target, not on the build host.  The file backend is
needed on the build host to label the filesystem images when they are
created.  We are likely only building the property backend on the host
because we don't allow conditionally excluding it presently.

OTOH, being able to look things up on the build host could be useful
from a development POV, e.g. if you were to add utils/selabel_lookup to
the build host targets and kept the property and service backends in the
host libselinux, then one could check what would match a given key.
___
Seandroid-list mailing list
Seandroid-list@tycho.nsa.gov
To unsubscribe, send email to seandroid-list-le...@tycho.nsa.gov.
To get help, send an email containing "help" to 
seandroid-list-requ...@tycho.nsa.gov.