All of lore.kernel.org
 help / color / mirror / Atom feed
From: Roland Dreier <rdreier@cisco.com>
To: tyhicks@linux.vnet.ibm.com, kirkland@canonical.com,
	ecryptfs-devel@lists.launchpad.net
Cc: linux-kernel@vger.kernel.org
Subject: lockdep reported AB-BA problem in ecryptfs
Date: Wed, 01 Jul 2009 10:06:33 -0700	[thread overview]
Message-ID: <adahbxwgwo6.fsf@cisco.com> (raw)

I have an Ubuntu 9.10 system, using ecryptfs for my home directory,
running a 2.6.31-rc1+git kernel with lockdep enabled, and I got the
report below on login.  This looks like a valid AB-BA issue; there are
two paths through the code, first:

  ecryptfs_new_file_context() ->
    ecryptfs_copy_mount_wide_sigs_to_inode_sigs() ->

	mutex_lock(&mount_crypt_stat->global_auth_tok_list_mutex);

	  -> ecryptfs_add_keysig() ->

		mutex_lock(&crypt_stat->keysig_list_mutex);

so global_auth_tok_list_mutex is taken before keysig_list_mutex.

and second:

  ecryptfs_generate_key_packet_set() ->

	mutex_lock(&crypt_stat->keysig_list_mutex);

	  -> ecryptfs_find_global_auth_tok_for_sig() ->

		mutex_lock(&mount_crypt_stat->global_auth_tok_list_mutex);

so keysig_list_mutex is taken before global_auth_tok_list_mutex here.

I'm not sure if this could actually deadlock in practice (ie can we have
threads racing on these two paths?) but it seems risky.

I'm not an expert on ecryptfs locking bug the simplest fix I see for
this particular issue is to move taking keysig_list_mutex to before
global_auth_tok_list_mutex in ecryptfs_copy_mount_wide_sigs_to_inode_sigs(),
since ecryptfs_add_keysig() has only one caller and so we can rely on
that caller to take the needed lock.  I can send a patch for this if
this sounds good.


here's the full report:

[   24.787210] =======================================================
[   24.787213] [ INFO: possible circular locking dependency detected ]
[   24.787216] 2.6.31-2-generic #14~rbd2
[   24.787217] -------------------------------------------------------
[   24.787219] gdm/2640 is trying to acquire lock:
[   24.787221]  (&mount_crypt_stat->global_auth_tok_list_mutex){+.+.+.}, at: [<ffffffff8121591e>] ecryptfs_find_global_auth_tok_for_sig+0x2e/0x90
[   24.787231] 
[   24.787231] but task is already holding lock:
[   24.787233]  (&crypt_stat->keysig_list_mutex){+.+.+.}, at: [<ffffffff81217728>] ecryptfs_generate_key_packet_set+0x58/0x2b0
[   24.787239] 
[   24.787239] which lock already depends on the new lock.
[   24.787240] 
[   24.787241] 
[   24.787242] the existing dependency chain (in reverse order) is:
[   24.787244] 
[   24.787244] -> #1 (&crypt_stat->keysig_list_mutex){+.+.+.}:
[   24.787248]        [<ffffffff8108c897>] check_prev_add+0x2a7/0x370
[   24.787253]        [<ffffffff8108cfc1>] validate_chain+0x661/0x750
[   24.787256]        [<ffffffff8108d2e7>] __lock_acquire+0x237/0x430
[   24.787259]        [<ffffffff8108d585>] lock_acquire+0xa5/0x150
[   24.787263]        [<ffffffff815526cd>] __mutex_lock_common+0x4d/0x3d0
[   24.787267]        [<ffffffff81552b56>] mutex_lock_nested+0x46/0x60
[   24.787270]        [<ffffffff8121526a>] ecryptfs_add_keysig+0x5a/0xb0
[   24.787273]        [<ffffffff81213299>] ecryptfs_copy_mount_wide_sigs_to_inode_sigs+0x59/0xb0
[   24.787276]        [<ffffffff81214b06>] ecryptfs_new_file_context+0xa6/0x1a0
[   24.787280]        [<ffffffff8120e42a>] ecryptfs_initialize_file+0x4a/0x140
[   24.787283]        [<ffffffff8120e54d>] ecryptfs_create+0x2d/0x60
[   24.787286]        [<ffffffff8113a7d4>] vfs_create+0xb4/0xe0
[   24.787290]        [<ffffffff8113a8c4>] __open_namei_create+0xc4/0x110
[   24.787293]        [<ffffffff8113d1c1>] do_filp_open+0xa01/0xae0
[   24.787296]        [<ffffffff8112d8d9>] do_sys_open+0x69/0x140
[   24.787300]        [<ffffffff8112d9f0>] sys_open+0x20/0x30
[   24.787303]        [<ffffffff81013132>] system_call_fastpath+0x16/0x1b
[   24.787308]        [<ffffffffffffffff>] 0xffffffffffffffff
[   24.787326] 
[   24.787326] -> #0 (&mount_crypt_stat->global_auth_tok_list_mutex){+.+.+.}:
[   24.787330]        [<ffffffff8108c675>] check_prev_add+0x85/0x370
[   24.787333]        [<ffffffff8108cfc1>] validate_chain+0x661/0x750
[   24.787337]        [<ffffffff8108d2e7>] __lock_acquire+0x237/0x430
[   24.787340]        [<ffffffff8108d585>] lock_acquire+0xa5/0x150
[   24.787343]        [<ffffffff815526cd>] __mutex_lock_common+0x4d/0x3d0
[   24.787346]        [<ffffffff81552b56>] mutex_lock_nested+0x46/0x60
[   24.787349]        [<ffffffff8121591e>] ecryptfs_find_global_auth_tok_for_sig+0x2e/0x90
[   24.787352]        [<ffffffff812177d5>] ecryptfs_generate_key_packet_set+0x105/0x2b0
[   24.787356]        [<ffffffff81212f49>] ecryptfs_write_headers_virt+0xc9/0x120
[   24.787359]        [<ffffffff8121306d>] ecryptfs_write_metadata+0xcd/0x200
[   24.787362]        [<ffffffff8120e44b>] ecryptfs_initialize_file+0x6b/0x140
[   24.787365]        [<ffffffff8120e54d>] ecryptfs_create+0x2d/0x60
[   24.787368]        [<ffffffff8113a7d4>] vfs_create+0xb4/0xe0
[   24.787371]        [<ffffffff8113a8c4>] __open_namei_create+0xc4/0x110
[   24.787374]        [<ffffffff8113d1c1>] do_filp_open+0xa01/0xae0
[   24.787377]        [<ffffffff8112d8d9>] do_sys_open+0x69/0x140
[   24.787380]        [<ffffffff8112d9f0>] sys_open+0x20/0x30
[   24.787383]        [<ffffffff81013132>] system_call_fastpath+0x16/0x1b
[   24.787386]        [<ffffffffffffffff>] 0xffffffffffffffff
[   24.787390] 
[   24.787390] other info that might help us debug this:
[   24.787391] 
[   24.787393] 2 locks held by gdm/2640:
[   24.787394]  #0:  (&sb->s_type->i_mutex_key#11){+.+.+.}, at: [<ffffffff8113cb8b>] do_filp_open+0x3cb/0xae0
[   24.787401]  #1:  (&crypt_stat->keysig_list_mutex){+.+.+.}, at: [<ffffffff81217728>] ecryptfs_generate_key_packet_set+0x58/0x2b0
[   24.787407] 
[   24.787407] stack backtrace:
[   24.787410] Pid: 2640, comm: gdm Tainted: G         C 2.6.31-2-generic #14~rbd2
[   24.787412] Call Trace:
[   24.787416]  [<ffffffff8108b988>] print_circular_bug_tail+0xa8/0xf0
[   24.787419]  [<ffffffff8108c675>] check_prev_add+0x85/0x370
[   24.787423]  [<ffffffff81094912>] ? __module_text_address+0x12/0x60
[   24.787426]  [<ffffffff8108cfc1>] validate_chain+0x661/0x750
[   24.787429]  [<ffffffff81017275>] ? print_context_stack+0x85/0x140
[   24.787433]  [<ffffffff81089c68>] ? find_usage_backwards+0x38/0x160
[   24.787436]  [<ffffffff8108d2e7>] __lock_acquire+0x237/0x430
[   24.787439]  [<ffffffff8108d585>] lock_acquire+0xa5/0x150
[   24.787442]  [<ffffffff8121591e>] ? ecryptfs_find_global_auth_tok_for_sig+0x2e/0x90
[   24.787445]  [<ffffffff8108b0b0>] ? check_usage_backwards+0x0/0xb0
[   24.787448]  [<ffffffff815526cd>] __mutex_lock_common+0x4d/0x3d0
[   24.787451]  [<ffffffff8121591e>] ? ecryptfs_find_global_auth_tok_for_sig+0x2e/0x90
[   24.787454]  [<ffffffff8121591e>] ? ecryptfs_find_global_auth_tok_for_sig+0x2e/0x90
[   24.787458]  [<ffffffff8108c02c>] ? mark_held_locks+0x6c/0xa0
[   24.787461]  [<ffffffff81125b0d>] ? kmem_cache_alloc+0xfd/0x1a0
[   24.787464]  [<ffffffff8108c34d>] ? trace_hardirqs_on_caller+0x14d/0x190
[   24.787467]  [<ffffffff81552b56>] mutex_lock_nested+0x46/0x60
[   24.787470]  [<ffffffff8121591e>] ecryptfs_find_global_auth_tok_for_sig+0x2e/0x90
[   24.787473]  [<ffffffff812177d5>] ecryptfs_generate_key_packet_set+0x105/0x2b0
[   24.787476]  [<ffffffff81212f49>] ecryptfs_write_headers_virt+0xc9/0x120
[   24.787479]  [<ffffffff8121306d>] ecryptfs_write_metadata+0xcd/0x200
[   24.787482]  [<ffffffff81210240>] ? ecryptfs_init_persistent_file+0x60/0xe0
[   24.787485]  [<ffffffff8120e44b>] ecryptfs_initialize_file+0x6b/0x140
[   24.787488]  [<ffffffff8120e54d>] ecryptfs_create+0x2d/0x60
[   24.787491]  [<ffffffff8113a7d4>] vfs_create+0xb4/0xe0
[   24.787494]  [<ffffffff8113a8c4>] __open_namei_create+0xc4/0x110
[   24.787496]  [<ffffffff8113d1c1>] do_filp_open+0xa01/0xae0
[   24.787502]  [<ffffffff8129a93e>] ? _raw_spin_unlock+0x5e/0xb0
[   24.787505]  [<ffffffff8155410b>] ? _spin_unlock+0x2b/0x40
[   24.787508]  [<ffffffff81139e9b>] ? getname+0x3b/0x240
[   24.787511]  [<ffffffff81148a5a>] ? alloc_fd+0xfa/0x140
[   24.787515]  [<ffffffff8112d8d9>] do_sys_open+0x69/0x140
[   24.787517]  [<ffffffff81553b8f>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[   24.787520]  [<ffffffff8112d9f0>] sys_open+0x20/0x30
[   24.787523]  [<ffffffff81013132>] system_call_fastpath+0x16/0x1b

             reply	other threads:[~2009-07-01 17:06 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-07-01 17:06 Roland Dreier [this message]
2009-07-01 22:48 ` [PATCH] eCryptfs: Fix lockdep-reported AB-BA mutex issue Roland Dreier
2009-07-02 19:20   ` Tyler Hicks
2009-07-04 22:04     ` Roland Dreier
2009-07-01 23:08 ` Another lockdep issue reported with ecryptfs Roland Dreier
2009-07-02  0:31   ` Roland Dreier
2009-07-02  0:41   ` Yet another " Roland Dreier
2009-07-02  3:58     ` Roland Dreier

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=adahbxwgwo6.fsf@cisco.com \
    --to=rdreier@cisco.com \
    --cc=ecryptfs-devel@lists.launchpad.net \
    --cc=kirkland@canonical.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=tyhicks@linux.vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.