diff mbox series

[v1,1/2] certs: Remove panic() calls from blacklist_init()

Message ID 20220311174741.250424-2-mic@digikod.net (mailing list archive)
State New
Headers show
Series Remove panic() from keyring init calls | expand

Commit Message

Mickaël Salaün March 11, 2022, 5:47 p.m. UTC
From: Mickaël Salaün <mic@linux.microsoft.com>

Replace panic() calls from device_initcall(blacklist_init) with proper
error handling using -ENODEV.

Suggested-by: Jarkko Sakkinen <jarkko@kernel.org> [1]
Link: https://lore.kernel.org/r/Yik0C2t7G272YZ73@iki.fi [1]
Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Link: https://lore.kernel.org/r/20220311174741.250424-2-mic@digikod.net
---
 certs/blacklist.c | 27 +++++++++++++++++++++------
 1 file changed, 21 insertions(+), 6 deletions(-)

Comments

Paul Moore March 11, 2022, 10 p.m. UTC | #1
On Fri, Mar 11, 2022 at 12:47 PM Mickaël Salaün <mic@digikod.net> wrote:
> From: Mickaël Salaün <mic@linux.microsoft.com>
>
> Replace panic() calls from device_initcall(blacklist_init) with proper
> error handling using -ENODEV.
>
> Suggested-by: Jarkko Sakkinen <jarkko@kernel.org> [1]
> Link: https://lore.kernel.org/r/Yik0C2t7G272YZ73@iki.fi [1]
> Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
> Link: https://lore.kernel.org/r/20220311174741.250424-2-mic@digikod.net
> ---
>  certs/blacklist.c | 27 +++++++++++++++++++++------
>  1 file changed, 21 insertions(+), 6 deletions(-)

I'm not sure we can safely rely on a non-zero error code saving us in
the care of failure, can we?

The blacklist_init() function is registered as an initcall via
device_initcall() which I believe is either executed via
do_init_module() in the case of a dynamic module load, or via
do_initcalls() if built into the kernel.  In either case the result is
that the module/functionality doesn't load and the kernel continues on
executing.  While this could be acceptable for some non-critical
modules, if this particular module fails to load it defeats the
certificate/key based deny list for signed modules, yes?

I completely understand the strong desire to purge the kernel of
panic()s, BUG()s, and the like, but 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.

> diff --git a/certs/blacklist.c b/certs/blacklist.c
> index 486ce0dd8e9c..ea7a77f156da 100644
> --- a/certs/blacklist.c
> +++ b/certs/blacklist.c
> @@ -313,12 +313,16 @@ static int __init blacklist_init(void)
>         const char *const *bl;
>         struct key_restriction *restriction;
>
> -       if (register_key_type(&key_type_blacklist) < 0)
> -               panic("Can't allocate system blacklist key type\n");
> +       if (register_key_type(&key_type_blacklist) < 0) {
> +               pr_err("Can't allocate system blacklist key type\n");
> +               return -ENODEV;
> +       }
>
>         restriction = kzalloc(sizeof(*restriction), GFP_KERNEL);
> -       if (!restriction)
> -               panic("Can't allocate blacklist keyring restriction\n");
> +       if (!restriction) {
> +               pr_err("Can't allocate blacklist keyring restriction\n");
> +               goto err_restriction;
> +       }
>         restriction->check = restrict_link_for_blacklist;
>
>         blacklist_keyring =
> @@ -333,13 +337,24 @@ static int __init blacklist_init(void)
>                               , KEY_ALLOC_NOT_IN_QUOTA |
>                               KEY_ALLOC_SET_KEEP,
>                               restriction, NULL);
> -       if (IS_ERR(blacklist_keyring))
> -               panic("Can't allocate system blacklist keyring\n");
> +       if (IS_ERR(blacklist_keyring)) {
> +               pr_err("Can't allocate system blacklist keyring\n");
> +               goto err_keyring;
> +       }
>
>         for (bl = blacklist_hashes; *bl; bl++)
>                 if (mark_raw_hash_blacklisted(*bl) < 0)
>                         pr_err("- blacklisting failed\n");
>         return 0;
> +
> +
> +err_keyring:
> +       kfree(restriction);
> +
> +err_restriction:
> +       unregister_key_type(&key_type_blacklist);
> +
> +       return -ENODEV;
>  }
Jarkko Sakkinen March 20, 2022, 9:04 p.m. UTC | #2
On Fri, Mar 11, 2022 at 05:00:32PM -0500, Paul Moore wrote:
> On Fri, Mar 11, 2022 at 12:47 PM Mickaël Salaün <mic@digikod.net> wrote:
> > From: Mickaël Salaün <mic@linux.microsoft.com>
> >
> > Replace panic() calls from device_initcall(blacklist_init) with proper
> > error handling using -ENODEV.
> >
> > Suggested-by: Jarkko Sakkinen <jarkko@kernel.org> [1]
> > Link: https://lore.kernel.org/r/Yik0C2t7G272YZ73@iki.fi [1]
> > Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
> > Link: https://lore.kernel.org/r/20220311174741.250424-2-mic@digikod.net
> > ---
> >  certs/blacklist.c | 27 +++++++++++++++++++++------
> >  1 file changed, 21 insertions(+), 6 deletions(-)
> 
> I'm not sure we can safely rely on a non-zero error code saving us in
> the care of failure, can we?
> 
> The blacklist_init() function is registered as an initcall via
> device_initcall() which I believe is either executed via
> do_init_module() in the case of a dynamic module load, or via
> do_initcalls() if built into the kernel.  In either case the result is
> that the module/functionality doesn't load and the kernel continues on
> executing.  While this could be acceptable for some non-critical
> modules, if this particular module fails to load it defeats the
> certificate/key based deny list for signed modules, yes?
> 
> I completely understand the strong desire to purge the kernel of
> panic()s, BUG()s, and the like, but 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.

OK, I get this. 

What this function should have is this information documented in
the header. Otherwise, this is just confusing.

BR, Jarkko
diff mbox series

Patch

diff --git a/certs/blacklist.c b/certs/blacklist.c
index 486ce0dd8e9c..ea7a77f156da 100644
--- a/certs/blacklist.c
+++ b/certs/blacklist.c
@@ -313,12 +313,16 @@  static int __init blacklist_init(void)
 	const char *const *bl;
 	struct key_restriction *restriction;
 
-	if (register_key_type(&key_type_blacklist) < 0)
-		panic("Can't allocate system blacklist key type\n");
+	if (register_key_type(&key_type_blacklist) < 0) {
+		pr_err("Can't allocate system blacklist key type\n");
+		return -ENODEV;
+	}
 
 	restriction = kzalloc(sizeof(*restriction), GFP_KERNEL);
-	if (!restriction)
-		panic("Can't allocate blacklist keyring restriction\n");
+	if (!restriction) {
+		pr_err("Can't allocate blacklist keyring restriction\n");
+		goto err_restriction;
+	}
 	restriction->check = restrict_link_for_blacklist;
 
 	blacklist_keyring =
@@ -333,13 +337,24 @@  static int __init blacklist_init(void)
 			      , KEY_ALLOC_NOT_IN_QUOTA |
 			      KEY_ALLOC_SET_KEEP,
 			      restriction, NULL);
-	if (IS_ERR(blacklist_keyring))
-		panic("Can't allocate system blacklist keyring\n");
+	if (IS_ERR(blacklist_keyring)) {
+		pr_err("Can't allocate system blacklist keyring\n");
+		goto err_keyring;
+	}
 
 	for (bl = blacklist_hashes; *bl; bl++)
 		if (mark_raw_hash_blacklisted(*bl) < 0)
 			pr_err("- blacklisting failed\n");
 	return 0;
+
+
+err_keyring:
+	kfree(restriction);
+
+err_restriction:
+	unregister_key_type(&key_type_blacklist);
+
+	return -ENODEV;
 }
 
 /*