On Tue, Jan 14, 2020 at 3:57 PM Mimi Zohar [off-list ref] wrote:

On Tue, 2020-01-14 at 14:58 +0100, Dmitry Vyukov wrote:

quoted

On Tue, Jan 14, 2020 at 2:56 PM syzbot
[off-list ref] wrote:

quoted

Hello,

syzbot found the following crash on:

HEAD commit:    1b851f98 Add linux-next specific files for 20200114
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=12bcbb25e00000
kernel config:  https://syzkaller.appspot.com/x/.config?x=3e7d9cf7ebfa08ad
dashboard link: https://syzkaller.appspot.com/bug?extid=a4a503d7f37292ae1664
compiler:       gcc (GCC) 9.0.0 20181231 (experimental)

Unfortunately, I don't have any reproducer for this crash yet.

IMPORTANT: if you fix the bug, please add the following tag to the commit:
Reported-by: syzbot+a4a503d7f37292ae1664@syzkaller.appspotmail.com

+Lakshmi, you seem to have submitted a number of changes to this file recently.

This completely breaks linux-next testing for us, every kernel crashes
a few minutes after boot.

2020/01/14 14:45:00 vm-26: crash: inconsistent lock state in
ima_process_queued_keys

Yikes!  Are you running with an IMA policy?

I don't know.

 I assume this is being
caused by commit 8f5d2d06f217 ("IMA: Defined timer to free queued
keys".  Does reverting it prevent this from happening?

The following seems to help, but don't know if it's the right fix or not.

diff --git a/security/integrity/ima/ima_asymmetric_keys.c

b/security/integrity/ima/ima_asymmetric_keys.c
index 61e478f9e8199..49d559501fe62 100644

--- a/security/integrity/ima/ima_asymmetric_keys.c
+++ b/security/integrity/ima/ima_asymmetric_keys.c

@@ -103,17 +103,18 @@ static bool ima_queue_key(struct key *keyring,

const void *payload,
 {
        bool queued = false;
        struct ima_key_entry *entry;
+       unsigned long flags;

        entry = ima_alloc_key_entry(keyring, payload, payload_len);
        if (!entry)
                return false;

-       spin_lock(&ima_keys_lock);
+       spin_lock_irqsave(&ima_keys_lock, flags);
        if (!ima_process_keys) {
                list_add_tail(&entry->list, &ima_keys);
                queued = true;
        }
-       spin_unlock(&ima_keys_lock);
+       spin_unlock_irqrestore(&ima_keys_lock, flags);

        if (!queued)
                ima_free_key_entry(entry);