linux-kernel - eCryptFS possible circular locking

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 26 Jun 2008 17:59:04 +0400
From:	Cyrill Gorcunov <gorcunov@...il.com>
To:	Michael Halcrow <mhalcrow@...ibm.com>
Cc:	LKML <linux-kernel@...r.kernel.org>, Ingo Molnar <mingo@...e.hu>
Subject: eCryptFS possible circular locking

Hi Michael,

there is potential circularo locking in eCryptFS. Actually, I think Ingo
boots catched it. (Original report from Ingo was about a month ago).
Here is a log:

---

> [Ingo Molnar - Fri, May 09, 2008 at 11:44:23AM +0200]
> | 
> | * Ingo Molnar <mingo@...e.hu> wrote:
> | 
> | > x86.git testing found the following "possible circular locking" 
> | > lockdep warning, generated by the ecryptfs code:
> | 
> | the correct URL for the config is:
> | 
> |   http://redhat.com/~mingo/misc/config-Wed_May__7_16_17_31_CEST_2008.bad
> | 
> | this is reproducible, it happened a second time since i originally 
> | reported it.
> | 
> |     Ingo
>
> Hi Ingo, I've just sent patch, could you test it, please?
>
> http://lkml.org/lkml/2008/5/10/20

that patch is now upstream - but see below there's still a circular
dependency.

        Ingo

[   21.530026]
[   21.530026] =======================================================
[   21.530026] [ INFO: possible circular locking dependency detected ]
[   21.530026] 2.6.26-rc2-sched-devel.git #1205
[   21.530026] -------------------------------------------------------
[   21.530026] multipath.stati/1379 is trying to acquire lock:
[   21.530026]  (&ecryptfs_daemon_hash_mux){--..}, at: [<c020cca2>] ecryptfs_miscdev_open+0x22/0x140
[   21.530026]
[   21.530026] but task is already holding lock:
[   21.530026]  (misc_mtx){--..}, at: [<c02f4b92>] misc_open+0x22/0x120
[   21.530026]
[   21.530026] which lock already depends on the new lock.
[   21.530026]
[   21.530026]
[   21.530026] the existing dependency chain (in reverse order) is:
[   21.530026]
[   21.530026] -> #1 (misc_mtx){--..}:
[   21.530026]        [<c01438aa>] __lock_acquire+0xc3a/0x1100
[   21.530026]        [<c0143de6>] lock_acquire+0x76/0xa0
[   21.530026]        [<c05d032a>] mutex_lock_nested+0x7a/0x280
[   21.530026]        [<c02f4cb5>] misc_register+0x25/0x140
[   21.530026]        [<c020cb5c>] ecryptfs_init_ecryptfs_miscdev+0x2c/0x60
[   21.530026]        [<c020c5fe>] ecryptfs_init_messaging+0x1ee/0x280
[   21.530026]        [<c0837662>] ecryptfs_init+0xc2/0x1b0
[   21.530026]        [<c0824852>] kernel_init+0x132/0x250
[   21.530026]        [<c010365f>] kernel_thread_helper+0x7/0x10
[   21.530026]        [<ffffffff>] 0xffffffff
[   21.530026]
[   21.530026] -> #0 (&ecryptfs_daemon_hash_mux){--..}:
[   21.530026]        [<c01436cd>] __lock_acquire+0xa5d/0x1100
[   21.530026]        [<c0143de6>] lock_acquire+0x76/0xa0
[   21.530026]        [<c05d032a>] mutex_lock_nested+0x7a/0x280
[   21.530026]        [<c020cca2>] ecryptfs_miscdev_open+0x22/0x140
[   21.530026]        [<c02f4bf2>] misc_open+0x82/0x120
[   21.530026]        [<c0187c26>] chrdev_open+0x86/0x150
[   21.530026]        [<c0183289>] __dentry_open+0xa9/0x260
[   21.530026]        [<c0183487>] nameidata_to_filp+0x47/0x60
[   21.530026]        [<c018fda4>] do_filp_open+0x184/0x7a0
[   21.530026]        [<c01830aa>] do_sys_open+0x4a/0xb0
[   21.530026]        [<c018317e>] sys_open+0x2e/0x40
[   21.530026]        [<c010330e>] syscall_call+0x7/0xb
[   21.530026]        [<ffffffff>] 0xffffffff
[   21.530026]
[   21.530026] other info that might help us debug this:
[   21.530026]
[   21.530026] 1 lock held by multipath.stati/1379:
[   21.530026]  #0:  (misc_mtx){--..}, at: [<c02f4b92>] misc_open+0x22/0x120
[   21.530026]

---

Some analisys shows the following interesting things -

Call chain (if being compiled into the kernel)
----------------------------------------------

ecryptfs_init
	ecryptfs_init_messaging
		ecryptfs_init_ecryptfs_miscdev
			mutex_lock(&ecryptfs_daemon_hash_mux);
			misc_register(&ecryptfs_miscdev);
				mutex_lock(&misc_mtx);
				mutex_unlock(&misc_mtx);
1 --->
			mutex_unlock(&ecryptfs_daemon_hash_mux);

As only we have misc device registered it is valid to open it.
And that is happened by another proccess

chrdev_open
	misc_open
		mutex_lock(&misc_mtx);
			ecryptfs_miscdev_open
				mutex_lock(&ecryptfs_daemon_hash_mux);
			

If this happens at point (1) /ie chrdev_open called at this point/
we are getting to circular warning as on top. I'm not sure how to handle this
in more elegant way but I think we should defer all calls to misc
device until eCryptFS has been initialized. Could you take a look,
please?

		- Cyrill -
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/