[v1,1/1] certs: Explain the rational to call panic()

Commit Message

Mickaël Salaün March 21, 2022, 5:45 p.m. UTC

From: Mickaël Salaün <[e-mail]>

The blacklist_init() function calls panic() for memory allocation
errors.  This change documents the reason why we don't return -ENODEV.

Suggested-by: Paul Moore <[e-mail]> [1]
Requested-by: Jarkko Sakkinen <[e-mail]> [1]
Link: https://lore.kernel.org/r/[e-mail] [1]
Signed-off-by: Mickaël Salaün <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
---
 certs/blacklist.c | 8 ++++++++
 1 file changed, 8 insertions(+)

Comments

Paul Moore March 21, 2022, 6:23 p.m. UTC | #1

On Mon, Mar 21, 2022 at 1:45 PM Mickaël Salaün <[e-mail]> wrote:
>
> From: Mickaël Salaün <[e-mail]>
>
> The blacklist_init() function calls panic() for memory allocation
> errors.  This change documents the reason why we don't return -ENODEV.
>
> Suggested-by: Paul Moore <[e-mail]> [1]
> Requested-by: Jarkko Sakkinen <[e-mail]> [1]
> Link: https://lore.kernel.org/r/[e-mail] [1]
> Signed-off-by: Mickaël Salaün <[e-mail]>
> Link: https://lore.kernel.org/r/[e-mail]
> ---
>  certs/blacklist.c | 8 ++++++++
>  1 file changed, 8 insertions(+)

I would suggest changing the second sentence as shown below, but
otherwise it looks good to me.

Reviewed-by: Paul Moore <[e-mail]>

> diff --git a/certs/blacklist.c b/certs/blacklist.c
> index 486ce0dd8e9c..ac26bcf9b9a5 100644
> --- a/certs/blacklist.c
> +++ b/certs/blacklist.c
> @@ -307,6 +307,14 @@ static int restrict_link_for_blacklist(struct key *dest_keyring,
>
>  /*
>   * Initialise the blacklist
> + *
> + * The blacklist_init() function is registered as an initcall via
> + * device_initcall().  As a result the functionality doesn't load and the

"As a result if the blacklist_init() function fails for any reason the
kernel continues to execute."

> + * kernel continues on executing.  While cleanly returning -ENODEV could be
> + * acceptable for some non-critical kernel parts, if the blacklist keyring
> + * fails to load it defeats the certificate/key based deny list for signed
> + * modules.  If a critical piece of security functionality that users expect to
> + * be present fails to initialize, panic()ing is likely the right thing to do.
>   */
>  static int __init blacklist_init(void)
>  {

--
paul-moore.com

Jarkko Sakkinen March 21, 2022, 11:53 p.m. UTC | #2

On Mon, Mar 21, 2022 at 02:23:54PM -0400, Paul Moore wrote:
> On Mon, Mar 21, 2022 at 1:45 PM Mickaël Salaün <[e-mail]> wrote:
> >
> > From: Mickaël Salaün <[e-mail]>
> >
> > The blacklist_init() function calls panic() for memory allocation
> > errors.  This change documents the reason why we don't return -ENODEV.
> >
> > Suggested-by: Paul Moore <[e-mail]> [1]
> > Requested-by: Jarkko Sakkinen <[e-mail]> [1]
> > Link: https://lore.kernel.org/r/[e-mail] [1]
> > Signed-off-by: Mickaël Salaün <[e-mail]>
> > Link: https://lore.kernel.org/r/[e-mail]
> > ---
> >  certs/blacklist.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> 
> I would suggest changing the second sentence as shown below, but
> otherwise it looks good to me.
> 
> Reviewed-by: Paul Moore <[e-mail]>

Reviewed-by: Jarkko Sakkinen <[e-mail]>

Mickaël, I think since your patch set was not huge in the first place, I'm
considering making it part of rc2 pull request while I normally try to
avoid any features after rc1. It's anyway throughly tested, and generally
has been around for a *long time*. I've even tested it myself a few times.

Just trying to be responsible as a maintainer and if something does not
feel right, I don't  try to pretend that "I get it", if you know what 
I mean. This fully clarifies "not getting it" part :-)

Thanks!

BR, Jarkko

Mickaël Salaün March 22, 2022, 10:53 a.m. UTC | #3

On 21/03/2022 19:23, Paul Moore wrote:
> On Mon, Mar 21, 2022 at 1:45 PM Mickaël Salaün <[e-mail]> wrote:
>>
>> From: Mickaël Salaün <[e-mail]>
>>
>> The blacklist_init() function calls panic() for memory allocation
>> errors.  This change documents the reason why we don't return -ENODEV.
>>
>> Suggested-by: Paul Moore <[e-mail]> [1]
>> Requested-by: Jarkko Sakkinen <[e-mail]> [1]
>> Link: https://lore.kernel.org/r/[e-mail] [1]
>> Signed-off-by: Mickaël Salaün <[e-mail]>
>> Link: https://lore.kernel.org/r/[e-mail]
>> ---
>>   certs/blacklist.c | 8 ++++++++
>>   1 file changed, 8 insertions(+)
> 
> I would suggest changing the second sentence as shown below, but
> otherwise it looks good to me.
> 
> Reviewed-by: Paul Moore <[e-mail]>
> 
>> diff --git a/certs/blacklist.c b/certs/blacklist.c
>> index 486ce0dd8e9c..ac26bcf9b9a5 100644
>> --- a/certs/blacklist.c
>> +++ b/certs/blacklist.c
>> @@ -307,6 +307,14 @@ static int restrict_link_for_blacklist(struct key *dest_keyring,
>>
>>   /*
>>    * Initialise the blacklist
>> + *
>> + * The blacklist_init() function is registered as an initcall via
>> + * device_initcall().  As a result the functionality doesn't load and the
> 
> "As a result if the blacklist_init() function fails for any reason the
> kernel continues to execute."

Thanks, I'll fix that.

> 
>> + * kernel continues on executing.  While cleanly returning -ENODEV could be
>> + * acceptable for some non-critical kernel parts, if the blacklist keyring
>> + * fails to load it defeats the certificate/key based deny list for signed
>> + * modules.  If a critical piece of security functionality that users expect to
>> + * be present fails to initialize, panic()ing is likely the right thing to do.
>>    */
>>   static int __init blacklist_init(void)
>>   {
> 
> --
> paul-moore.com

Mickaël Salaün March 22, 2022, 10:54 a.m. UTC | #4

On 22/03/2022 00:53, Jarkko Sakkinen wrote:
> On Mon, Mar 21, 2022 at 02:23:54PM -0400, Paul Moore wrote:
>> On Mon, Mar 21, 2022 at 1:45 PM Mickaël Salaün <[e-mail]> wrote:
>>>
>>> From: Mickaël Salaün <[e-mail]>
>>>
>>> The blacklist_init() function calls panic() for memory allocation
>>> errors.  This change documents the reason why we don't return -ENODEV.
>>>
>>> Suggested-by: Paul Moore <[e-mail]> [1]
>>> Requested-by: Jarkko Sakkinen <[e-mail]> [1]
>>> Link: https://lore.kernel.org/r/[e-mail] [1]
>>> Signed-off-by: Mickaël Salaün <[e-mail]>
>>> Link: https://lore.kernel.org/r/[e-mail]
>>> ---
>>>   certs/blacklist.c | 8 ++++++++
>>>   1 file changed, 8 insertions(+)
>>
>> I would suggest changing the second sentence as shown below, but
>> otherwise it looks good to me.
>>
>> Reviewed-by: Paul Moore <[e-mail]>
> 
> Reviewed-by: Jarkko Sakkinen <[e-mail]>
> 
> Mickaël, I think since your patch set was not huge in the first place, I'm
> considering making it part of rc2 pull request while I normally try to
> avoid any features after rc1. It's anyway throughly tested, and generally
> has been around for a *long time*. I've even tested it myself a few times.
> 
> Just trying to be responsible as a maintainer and if something does not
> feel right, I don't  try to pretend that "I get it", if you know what
> I mean. This fully clarifies "not getting it" part :-)
> 
> Thanks!

Thanks Jarkko, I get it. ;)

Patch

diff --git a/certs/blacklist.c b/certs/blacklist.c
index 486ce0dd8e9c..ac26bcf9b9a5 100644
--- a/certs/blacklist.c
+++ b/certs/blacklist.c
@@ -307,6 +307,14 @@  static int restrict_link_for_blacklist(struct key *dest_keyring,
 
 /*
  * Initialise the blacklist
+ *
+ * The blacklist_init() function is registered as an initcall via
+ * device_initcall().  As a result the functionality doesn't load and the
+ * kernel continues on executing.  While cleanly returning -ENODEV could be
+ * acceptable for some non-critical kernel parts, if the blacklist keyring
+ * fails to load it defeats the certificate/key based deny list for signed
+ * modules.  If a critical piece of security functionality that users expect to
+ * be present fails to initialize, panic()ing is likely the right thing to do.
  */
 static int __init blacklist_init(void)
 {