Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 24 Dec 2018 01:22:57 -0800
From:   Dmitry Torokhov <dmitry.torokhov@gmail.com>
To:     Greg KH <gregkh@linuxfoundation.org>
Cc:     Marcus Meissner <meissner@suse.de>, linux-kernel@vger.kernel.org
Subject: Re: FYI: Userland breakage caused by udev bind commit
Message-ID: <20181224092257.GB122208@dtor-ws>
References: <20181223164954.hib4lbchftspidsd@suse.de>
 <20181224091229.GA26796@kroah.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20181224091229.GA26796@kroah.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Dec 24, 2018 at 10:12:29AM +0100, Greg KH wrote:
> On Sun, Dec 23, 2018 at 05:49:54PM +0100, Marcus Meissner wrote:
> > Hi,
> > 
> > I am the maintainer of libmtp and libgphoto2
> > 
> > Some months ago I was made aware of this bug:
> > 	https://bugs.kde.org/show_bug.cgi?id=387454
> > 
> > This was fallout identified to come from this kernel commit:
> > 
> > 	commit 1455cf8dbfd06aa7651dcfccbadb7a093944ca65
> > 	Author: Dmitry Torokhov <dmitry.torokhov@gmail.com>
> > 	Date:   Wed Jul 19 17:24:30 2017 -0700
> > 
> > If distributions would be using libmtp and libgphoto2 udev rules
> > that just triggered on "add" events, and not the new "bind" events,
> > the missing "attribute tagging" of the "bind" events would confused the
> > KDE Solid device detection and make the devices no longer detected.
> > 
> > This did not affect distributions that rely on the newer "hwdb"
> > device detection method.
> > 
> > I have released fixed libmtp and libgphoto2 versions in November, so
> > this is under control, but wanted to bring this up as a "kernel caused
> > userland breakage".
> 
> This is complex, sorry.  When this first commit was merged, we did get
> some reports of problems, so we reverted it.  Dmitry worked through the
> issues and then we added it back again.
> 
> That was back in July of 2017, and since then, we had not heard of any
> problems that happened until this month, a very long time.
> 
> So I really don't understand the root problem here, all of the distros
> that have been shipping kernels with this code for over a year didn't
> seem to have any issues.  My systems never had any issues, and so I
> can't figure out what suddenly changed to cause problems.
> 
> Was it the fact that we all are using distros that use hwdb?  Who does
> _not_ use hwdb these days?  Heck, I would have expected Debian to report
> problems as they are the ones that are known to use old userspace code
> with kernel developers using new kernels.
> 
> So what changed to cause the problem recently?

I think this is new systemd that had my patch to handle bind/unbind
instead of ignoring them is catching up in distros. So:

- old systemd with old kernels OK
- old systemd with new kernels OK
- new systemd with old kernels OK
- new systemd with new kernels - NOT OK - losing tags on bind

Systemd folks merged my patch to disable bind/unbind again until we
teach it to no longer flush entire device state on each new uevent and
it should be well now.

Thanks.

-- 
Dmitry