KEYS: Fix keyring quota misaccounting on key replacement and unlink
authorDavid Howells <dhowells@redhat.com>
Wed, 30 Oct 2013 11:15:24 +0000 (11:15 +0000)
committerDavid Howells <dhowells@redhat.com>
Wed, 30 Oct 2013 11:15:24 +0000 (11:15 +0000)
If a key is displaced from a keyring by a matching one, then four more bytes
of quota are allocated to the keyring - despite the fact that the keyring does
not change in size.

Further, when a key is unlinked from a keyring, the four bytes of quota
allocated the link isn't recovered and returned to the user's pool.

The first can be tested by repeating:

keyctl add big_key a fred @s
cat /proc/key-users

(Don't put it in a shell loop otherwise the garbage collector won't have time
to clear the displaced keys, thus affecting the result).

This was causing the kerberos keyring to run out of room fairly quickly.

The second can be tested by:

cat /proc/key-users
a=`keyctl add user a a @s`
cat /proc/key-users
keyctl unlink $a
sleep 1 # Give RCU a chance to delete the key
cat /proc/key-users

assuming no system activity that otherwise adds/removes keys, the amount of
key data allocated should go up (say 40/20000 -> 47/20000) and then return to
the original value at the end.

Reported-by: Stephen Gallagher <sgallagh@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
security/keys/keyring.c

index 8c05ebd7203d3d25577e818ec91d7fb067fafc93..d80311e571c345a16142fb4fbe2255e234d6a6db 100644 (file)
@@ -1063,12 +1063,6 @@ int __key_link_begin(struct key *keyring,
        if (index_key->type == &key_type_keyring)
                down_write(&keyring_serialise_link_sem);
 
-       /* check that we aren't going to overrun the user's quota */
-       ret = key_payload_reserve(keyring,
-                                 keyring->datalen + KEYQUOTA_LINK_BYTES);
-       if (ret < 0)
-               goto error_sem;
-
        /* Create an edit script that will insert/replace the key in the
         * keyring tree.
         */
@@ -1078,17 +1072,25 @@ int __key_link_begin(struct key *keyring,
                                  NULL);
        if (IS_ERR(edit)) {
                ret = PTR_ERR(edit);
-               goto error_quota;
+               goto error_sem;
+       }
+
+       /* If we're not replacing a link in-place then we're going to need some
+        * extra quota.
+        */
+       if (!edit->dead_leaf) {
+               ret = key_payload_reserve(keyring,
+                                         keyring->datalen + KEYQUOTA_LINK_BYTES);
+               if (ret < 0)
+                       goto error_cancel;
        }
 
        *_edit = edit;
        kleave(" = 0");
        return 0;
 
-error_quota:
-       /* undo the quota changes */
-       key_payload_reserve(keyring,
-                           keyring->datalen - KEYQUOTA_LINK_BYTES);
+error_cancel:
+       assoc_array_cancel_edit(edit);
 error_sem:
        if (index_key->type == &key_type_keyring)
                up_write(&keyring_serialise_link_sem);
@@ -1146,7 +1148,7 @@ void __key_link_end(struct key *keyring,
        if (index_key->type == &key_type_keyring)
                up_write(&keyring_serialise_link_sem);
 
-       if (edit) {
+       if (edit && !edit->dead_leaf) {
                key_payload_reserve(keyring,
                                    keyring->datalen - KEYQUOTA_LINK_BYTES);
                assoc_array_cancel_edit(edit);
@@ -1243,6 +1245,7 @@ int key_unlink(struct key *keyring, struct key *key)
                goto error;
 
        assoc_array_apply_edit(edit);
+       key_payload_reserve(keyring, keyring->datalen - KEYQUOTA_LINK_BYTES);
        ret = 0;
 
 error: