From mboxrd@z Thu Jan  1 00:00:00 1970
From: Mika Westerberg <mika.westerberg@linux.intel.com>
Subject: Re: [PATCH v4 14/38] platform/x86: intel_scu_ipc: Introduce new SCU
 IPC API
Date: Wed, 22 Jan 2020 17:04:04 +0200
Message-ID: <20200122150404.GZ2665@lahna.fi.intel.com>
References: <20200121160114.60007-1-mika.westerberg@linux.intel.com>
 <20200121160114.60007-15-mika.westerberg@linux.intel.com>
 <20200122134359.GE4963@kroah.com>
 <20200122144048.GV2665@lahna.fi.intel.com>
 <20200122144939.GA59473@kroah.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Return-path: <linux-kernel-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <20200122144939.GA59473@kroah.com>
Sender: linux-kernel-owner@vger.kernel.org
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>, Darren Hart <dvhart@infradead.org>, Lee Jones <lee.jones@linaro.org>, Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>, "H . Peter Anvin" <hpa@zytor.com>, x86@kernel.org, Zha Qipeng <qipeng.zha@intel.com>, "David E . Box" <david.e.box@linux.intel.com>, Guenter Roeck <linux@roeck-us.net>, Heikki Krogerus <heikki.krogerus@linux.intel.com>, Wim Van Sebroeck <wim@linux-watchdog.org>, Mark Brown <broonie@kernel.org>, platform-driver-x86@vger.kernel.org, linux-kernel@vger.kernel.org
List-Id: platform-driver-x86.vger.kernel.org

On Wed, Jan 22, 2020 at 03:49:39PM +0100, Greg Kroah-Hartman wrote:
> On Wed, Jan 22, 2020 at 04:40:48PM +0200, Mika Westerberg wrote:
> > On Wed, Jan 22, 2020 at 02:43:59PM +0100, Greg Kroah-Hartman wrote:
> > > > +	if (!scu->dev)
> > > > +		goto err_unlock;
> > > > +	if (!try_module_get(scu->dev->driver->owner))
> > > > +		goto err_unlock;
> > > > +	mutex_unlock(&ipclock);
> > > > +	return scu;
> > > 
> > > NO REFERENCE COUNT INCREMENT???
> > 
> > You mean increment the scu->dev reference count? I kind of thought that
> > the try_module_get() should make sure the thing stays there as long as
> > the caller has not called _put() but now when I think about it bit more
> > we would need to do device_get() here as well.
> 
> module reference counts handle _code_ while structure reference counts
> handle _data_.
> 
> You should almost never need to worry about module reference counts if
> your code is structured properly, only handle the reference counts on
> the pointers you throw around.
> 
> The fact that you are even calling try_module_get() is a huge flag that
> something is wrong here.

Thanks for the explanation! I'll keep this in mind and make sure the
next version has reference counting done accordingly.