From mboxrd@z Thu Jan 1 00:00:00 1970
From: Philipp Zabel
Subject: Re: [PATCH v14 00/11] mux controller abstraction and iio/i2c muxes
Date: Mon, 24 Apr 2017 12:52:59 +0200
Message-ID: <1493031179.2446.9.camel@pengutronix.de>
References: <1493022995-16917-1-git-send-email-peda@axentia.se>
Mime-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
Return-path:
In-Reply-To: <1493022995-16917-1-git-send-email-peda@axentia.se>
Sender: linux-kernel-owner@vger.kernel.org
To: Peter Rosin
Cc: Jonathan Cameron , linux-kernel@vger.kernel.org, Greg Kroah-Hartman , Wolfram Sang , Rob Herring , Mark Rutland , Hartmut Knaack , Lars-Peter Clausen , Peter Meerwald-Stadler , Jonathan Corbet , linux-i2c@vger.kernel.org, devicetree@vger.kernel.org, linux-iio@vger.kernel.org, linux-doc@vger.kernel.org, Andrew Morton , Colin Ian King , Paul Gortmaker , kernel@pengutronix.de
List-Id: devicetree@vger.kernel.org
On Mon, 2017-04-24 at 10:36 +0200, Peter Rosin wrote:
> Hi!
>
> The big change since v13 is that the mux state is now locked with a mutex
> instead of an rwsem. Other that that, it is mostly restructuring and doc
> changes. There are a few other "real" changes as well, but those changes
> feel kind of minor. I guess what I'm trying to say is that although the
> list of changes for v14 is longish, it's still basically the same as last
> time.
I have hooked this up to the video-multiplexer and now I trigger
the lockdep debug_check_no_locks_held error message when selecting the
mux input from userspace:
$ media-ctl --links "'imx6-mipi-csi2':1->'ipu1_csi0_mux':0[1]"
[ 66.258368]
[ 66.259919] =====================================
[ 66.265369] [ BUG: media-ctl/258 still has locks held! ]
[ 66.270810] 4.11.0-rc8-20170424-1+ #1305 Not tainted
[ 66.275863] -------------------------------------
[ 66.282158] 1 lock held by media-ctl/258:
[ 66.286464] #0: (&mux->lock){+.+.+.}, at: [<8074a6c0>] mux_control_select+0x24/0x50
[ 66.294424]
[ 66.294424] stack backtrace:
[ 66.298980] CPU: 0 PID: 258 Comm: media-ctl Not tainted 4.11.0-rc8+ #1305
[ 66.306771] Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
[ 66.313334] Backtrace:
[ 66.315858] [<8010e5a4>] (dump_backtrace) from [<8010e8ac>] (show_stack+0x20/0x24)
[ 66.323473] r7:80e518d0 r6:80e518d0 r5:600d0013 r4:00000000
[ 66.329190] [<8010e88c>] (show_stack) from [<80496cf4>] (dump_stack+0xb0/0xdc)
[ 66.336470] [<80496c44>] (dump_stack) from [<8017e404>] (debug_check_no_locks_held+0xb8/0xbc)
[ 66.345043] r9:ae8566b8 r8:ad88dc84 r7:ad88df58 r6:ad88dc84 r5:ad88df58 r4:ae856400
[ 66.352837] [<8017e34c>] (debug_check_no_locks_held) from [<8012b258>] (do_exit+0x79c/0xcc8)
[ 66.361321] [<8012aabc>] (do_exit) from [<8012d25c>] (do_group_exit+0x4c/0xcc)
[ 66.368581] r7:000000f8
[ 66.371161] [<8012d210>] (do_group_exit) from [<8012d2fc>] (__wake_up_parent+0x0/0x30)
[ 66.379120] r7:000000f8 r6:76f71798 r5:00000000 r4:00000001
[ 66.384837] [<8012d2dc>] (SyS_exit_group) from [<80109380>] (ret_fast_syscall+0x0/0x1c)
That correctly warns that the media-ctl process caused the mux->lock to
be locked and still held when the process exited. Do we need a usage
counter based mechanism for muxes that are (indirectly) controlled from
userspace?
regards
Philipp