KEYS: Fix handling of stored error in a negatively instantiated user key
authorDavid Howells <dhowells@redhat.com>
Tue, 24 Nov 2015 21:36:31 +0000 (21:36 +0000)
committerJames Morris <james.l.morris@oracle.com>
Wed, 25 Nov 2015 03:19:47 +0000 (14:19 +1100)
commit096fe9eaea40a17e125569f9e657e34cdb6d73bd
tree1fcbf56162ce88d2174ac9add5802fa634797ab1
parent6ffeba9607343f15303a399bc402a538800d89d9
KEYS: Fix handling of stored error in a negatively instantiated user key

If a user key gets negatively instantiated, an error code is cached in the
payload area.  A negatively instantiated key may be then be positively
instantiated by updating it with valid data.  However, the ->update key
type method must be aware that the error code may be there.

The following may be used to trigger the bug in the user key type:

    keyctl request2 user user "" @u
    keyctl add user user "a" @u

which manifests itself as:

BUG: unable to handle kernel paging request at 00000000ffffff8a
IP: [<ffffffff810a376f>] __call_rcu.constprop.76+0x1f/0x280 kernel/rcu/tree.c:3046
PGD 7cc30067 PUD 0
Oops: 0002 [#1] SMP
Modules linked in:
CPU: 3 PID: 2644 Comm: a.out Not tainted 4.3.0+ #49
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
task: ffff88003ddea700 ti: ffff88003dd88000 task.ti: ffff88003dd88000
RIP: 0010:[<ffffffff810a376f>]  [<ffffffff810a376f>] __call_rcu.constprop.76+0x1f/0x280
 [<ffffffff810a376f>] __call_rcu.constprop.76+0x1f/0x280 kernel/rcu/tree.c:3046
RSP: 0018:ffff88003dd8bdb0  EFLAGS: 00010246
RAX: 00000000ffffff82 RBX: 0000000000000000 RCX: 0000000000000001
RDX: ffffffff81e3fe40 RSI: 0000000000000000 RDI: 00000000ffffff82
RBP: ffff88003dd8bde0 R08: ffff88007d2d2da0 R09: 0000000000000000
R10: 0000000000000000 R11: ffff88003e8073c0 R12: 00000000ffffff82
R13: ffff88003dd8be68 R14: ffff88007d027600 R15: ffff88003ddea700
FS:  0000000000b92880(0063) GS:ffff88007fd00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00000000ffffff8a CR3: 000000007cc5f000 CR4: 00000000000006e0
Stack:
 ffff88003dd8bdf0 ffffffff81160a8a 0000000000000000 00000000ffffff82
 ffff88003dd8be68 ffff88007d027600 ffff88003dd8bdf0 ffffffff810a39e5
 ffff88003dd8be20 ffffffff812a31ab ffff88007d027600 ffff88007d027620
Call Trace:
 [<ffffffff810a39e5>] kfree_call_rcu+0x15/0x20 kernel/rcu/tree.c:3136
 [<ffffffff812a31ab>] user_update+0x8b/0xb0 security/keys/user_defined.c:129
 [<     inline     >] __key_update security/keys/key.c:730
 [<ffffffff8129e5c1>] key_create_or_update+0x291/0x440 security/keys/key.c:908
 [<     inline     >] SYSC_add_key security/keys/keyctl.c:125
 [<ffffffff8129fc21>] SyS_add_key+0x101/0x1e0 security/keys/keyctl.c:60
 [<ffffffff8185f617>] entry_SYSCALL_64_fastpath+0x12/0x6a arch/x86/entry/entry_64.S:185

Note the error code (-ENOKEY) in EDX.

A similar bug can be tripped by:

    keyctl request2 trusted user "" @u
    keyctl add trusted user "a" @u

This should also affect encrypted keys - but that has to be correctly
parameterised or it will fail with EINVAL before getting to the bit that
will crashes.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
Signed-off-by: James Morris <james.l.morris@oracle.com>
security/keys/encrypted-keys/encrypted.c
security/keys/trusted.c
security/keys/user_defined.c