From: Thomas Tuttle <ttuttle@chromium.org>
To: linux-kernel@vger.kernel.org
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>, Johan Myreen <jem@iki.fi>
Subject: [PATCH] misc/serio_raw: Avoid circular locking.
Date: Wed, 5 Oct 2011 13:02:49 -0400 [thread overview]
Message-ID: <20111005170249.GA27129@google.com> (raw)
While debugging an unrelated bug in cdc-acm, PROVE_LOCKING found a
circular locking dependency between the misc and serio_raw:
<5>[ 4.311231] =======================================================
<5>[ 4.311242] [ INFO: possible circular locking dependency detected ]
<5>[ 4.311251] 3.0.3 #1
<5>[ 4.311257] -------------------------------------------------------
<5>[ 4.311264] X/394 is trying to acquire lock:
<5>[ 4.311273] (serio_raw_mutex){+.+.+.}, at: [<812be42c>] serio_raw_open+0x1b/0xb3
<5>[ 4.311301]
<5>[ 4.311303] but task is already holding lock:
<5>[ 4.311311] (misc_mtx){+.+.+.}, at: [<8120913d>] misc_open+0x25/0x137
<5>[ 4.311331]
<5>[ 4.311333] which lock already depends on the new lock.
<5>[ 4.311337]
<5>[ 4.311343]
<5>[ 4.311345] the existing dependency chain (in reverse order) is:
<5>[ 4.311353]
<5>[ 4.311355] -> #1 (misc_mtx){+.+.+.}:
<5>[ 4.311369] [<8105f177>] lock_acquire+0x8e/0xad
<5>[ 4.311383] [<8138b82d>] __mutex_lock_common+0x33/0x2a6
<5>[ 4.311399] [<8138bb92>] mutex_lock_nested+0x37/0x40
<5>[ 4.311412] [<812093af>] misc_register+0x21/0xee
<5>[ 4.311426] [<812be5cb>] serio_raw_connect+0x107/0x1a0
<5>[ 4.311440] [<812bbb46>] serio_connect_driver+0x25/0x35
<5>[ 4.311454] [<812bc6dc>] serio_rebind_driver+0x15a/0x1dd
<5>[ 4.311466] [<8125b1a0>] dev_attr_store+0x20/0x28
<5>[ 4.311479] [<810ffb7b>] sysfs_write_file+0xc5/0xf0
<5>[ 4.311491] [<810bd048>] vfs_write+0x8f/0x101
<5>[ 4.311501] [<810bd15b>] sys_write+0x40/0x65
<5>[ 4.311510] [<8138d697>] sysenter_do_call+0x12/0x36
<5>[ 4.311520]
<5>[ 4.311522] -> #0 (serio_raw_mutex){+.+.+.}:
<5>[ 4.311534] [<8105eb42>] __lock_acquire+0x92d/0xb92
<5>[ 4.311546] [<8105f177>] lock_acquire+0x8e/0xad
<5>[ 4.311558] [<8138b82d>] __mutex_lock_common+0x33/0x2a6
<5>[ 4.311572] [<8138bad7>] mutex_lock_interruptible_nested+0x37/0x3e
<5>[ 4.311586] [<812be42c>] serio_raw_open+0x1b/0xb3
<5>[ 4.311598] [<812091fb>] misc_open+0xe3/0x137
<5>[ 4.311612] [<810bf952>] chrdev_open+0x171/0x192
<5>[ 4.311625] [<810bb1fc>] __dentry_open+0x162/0x243
<5>[ 4.311635] [<810bb37b>] nameidata_to_filp+0x4a/0x58
<5>[ 4.311644] [<810c6d32>] do_last+0x3ea/0x4b6
<5>[ 4.311653] [<810c8053>] path_openat+0x9a/0x2cf
<5>[ 4.311662] [<810c833c>] do_filp_open+0x26/0x62
<5>[ 4.311671] [<810baf0f>] do_sys_open+0x61/0x10a
<5>[ 4.311682] [<810bb004>] sys_open+0x23/0x2b
<5>[ 4.311695] [<8138d697>] sysenter_do_call+0x12/0x36
<5>[ 4.311709]
<5>[ 4.311711] other info that might help us debug this:
<5>[ 4.311714]
<5>[ 4.311721] Possible unsafe locking scenario:
<5>[ 4.311724]
<5>[ 4.311730] CPU0 CPU1
<5>[ 4.311736] ---- ----
<5>[ 4.311741] lock(misc_mtx);
<5>[ 4.311752] lock(serio_raw_mutex);
<5>[ 4.311764] lock(misc_mtx);
<5>[ 4.311776] lock(serio_raw_mutex);
<5>[ 4.311786]
<5>[ 4.311788] *** DEADLOCK ***
<5>[ 4.311791]
<5>[ 4.311799] 1 lock held by X/394:
<5>[ 4.311805] #0: (misc_mtx){+.+.+.}, at: [<8120913d>] misc_open+0x25/0x137
<5>[ 4.311827]
<5>[ 4.311830] stack backtrace:
<5>[ 4.311839] Pid: 394, comm: X Not tainted 3.0.3 #1
<5>[ 4.311846] Call Trace:
<5>[ 4.311862] [<8105e17a>] print_circular_bug+0x18d/0x19a
<5>[ 4.311877] [<8105eb42>] __lock_acquire+0x92d/0xb92
<5>[ 4.311892] [<8105f177>] lock_acquire+0x8e/0xad
<5>[ 4.311906] [<812be42c>] ? serio_raw_open+0x1b/0xb3
<5>[ 4.311920] [<8138b82d>] __mutex_lock_common+0x33/0x2a6
<5>[ 4.311935] [<812be42c>] ? serio_raw_open+0x1b/0xb3
<5>[ 4.311949] [<8138bad7>] mutex_lock_interruptible_nested+0x37/0x3e
<5>[ 4.311962] [<812be42c>] ? serio_raw_open+0x1b/0xb3
<5>[ 4.311975] [<812be42c>] serio_raw_open+0x1b/0xb3
<5>[ 4.311989] [<812091fb>] misc_open+0xe3/0x137
<5>[ 4.312004] [<810bf952>] chrdev_open+0x171/0x192
<5>[ 4.312017] [<810bb1fc>] __dentry_open+0x162/0x243
<5>[ 4.312019] [<810bb37b>] nameidata_to_filp+0x4a/0x58
<5>[ 4.312019] [<810bf7e1>] ? cdev_put+0x1f/0x1f
<5>[ 4.312019] [<810c6d32>] do_last+0x3ea/0x4b6
<5>[ 4.312019] [<810c8053>] path_openat+0x9a/0x2cf
<5>[ 4.312019] [<8105ee2d>] ? lock_release_non_nested+0x86/0x1d8
<5>[ 4.312019] [<810c833c>] do_filp_open+0x26/0x62
<5>[ 4.312019] [<8138cc4f>] ? _raw_spin_unlock+0x22/0x25
<5>[ 4.312019] [<810d149d>] ? alloc_fd+0xc2/0xcf
<5>[ 4.312019] [<810baf0f>] do_sys_open+0x61/0x10a
<5>[ 4.312019] [<810bd0a6>] ? vfs_write+0xed/0x101
<5>[ 4.312019] [<810bb004>] sys_open+0x23/0x2b
<5>[ 4.312019] [<8138d697>] sysenter_do_call+0x12/0x36
This patch rearranges serio_raw_connect so that it takes the mutex once
to allocate a device number and open the serio device, drops it to call
misc_register, then takes it again to add the serio_raw device to the
list.
BUG=chromium-os:21113
TEST=Adhoc: PROVE_LOCKING no longer finds circular locking dependency
Change-Id: I414c5a3d606c28c81e42dcf989e5c55918eb4ea3
Signed-off-by: Thomas Tuttle <ttuttle@chromium.org>
---
I think this works, but I'm not familiar enough with the misc and serio
subsystems to know if this is completely correct, nor if it is the best
way to break the circular dependency. Feedback would be appreciated.
drivers/input/serio/serio_raw.c | 14 +++++++++-----
1 files changed, 9 insertions(+), 5 deletions(-)
diff --git a/drivers/input/serio/serio_raw.c b/drivers/input/serio/serio_raw.c
index b7ba459..d5757e8 100644
--- a/drivers/input/serio/serio_raw.c
+++ b/drivers/input/serio/serio_raw.c
@@ -294,13 +294,13 @@ static int serio_raw_connect(struct serio *serio, struct serio_driver *drv)
if (err)
goto out_free;
- list_add_tail(&serio_raw->node, &serio_raw_list);
-
serio_raw->dev.minor = PSMOUSE_MINOR;
serio_raw->dev.name = serio_raw->name;
serio_raw->dev.parent = &serio->dev;
serio_raw->dev.fops = &serio_raw_fops;
+ mutex_unlock(&serio_raw_mutex);
+
err = misc_register(&serio_raw->dev);
if (err) {
serio_raw->dev.minor = MISC_DYNAMIC_MINOR;
@@ -313,18 +313,22 @@ static int serio_raw_connect(struct serio *serio, struct serio_driver *drv)
goto out_close;
}
+ mutex_lock(&serio_raw_mutex);
+ list_add_tail(&serio_raw->node, &serio_raw_list);
+ mutex_unlock(&serio_raw_mutex);
+
printk(KERN_INFO "serio_raw: raw access enabled on %s (%s, minor %d)\n",
serio->phys, serio_raw->name, serio_raw->dev.minor);
- goto out;
+ return 0;
out_close:
+ mutex_lock(&serio_raw_mutex);
serio_close(serio);
- list_del_init(&serio_raw->node);
out_free:
serio_set_drvdata(serio, NULL);
+ mutex_unlock(&serio_raw_mutex);
kfree(serio_raw);
out:
- mutex_unlock(&serio_raw_mutex);
return err;
}
--
1.7.3.1
next reply other threads:[~2011-10-05 17:02 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-10-05 17:02 Thomas Tuttle [this message]
2011-10-06 5:07 ` [PATCH] misc/serio_raw: Avoid circular locking Dmitry Torokhov
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=20111005170249.GA27129@google.com \
--to=ttuttle@chromium.org \
--cc=dmitry.torokhov@gmail.com \
--cc=jem@iki.fi \
--cc=linux-kernel@vger.kernel.org \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox