linux-kernel - Re: [PATCH] AB3100 regulator support v1

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <Pine.LNX.4.44L0.0908301217410.10608-100000@netrider.rowland.org>
Date:	Sun, 30 Aug 2009 12:27:32 -0400 (EDT)
From:	Alan Stern <stern@...land.harvard.edu>
To:	Mark Brown <broonie@...nsource.wolfsonmicro.com>
cc:	Linus Walleij <linus.ml.walleij@...il.com>,
	<linux-kernel@...r.kernel.org>,
	Linus Walleij <linus.walleij@...ricsson.com>,
	Liam Girdwood <lrg@...mlogic.co.uk>,
	Samuel Ortiz <sameo@...ux.intel.com>,
	Russell King <linux@....linux.org.uk>,
	<linux-arm-kernel@...ts.arm.linux.org.uk>
Subject: Re: [PATCH] AB3100 regulator support v1

On Sun, 30 Aug 2009, Mark Brown wrote:

> On Sun, Aug 30, 2009 at 01:03:38AM +0200, Linus Walleij wrote:
> 
> [While setting up a child platform device from a platform driver
> probe...]
> 
> > >> +     /* This would seem logical but makes everything break... */
> > >> +     /* pdev->dev.parent = parent; */
> 
> > > Err...  it does?  In what way?
> 
> > The sub-platform devices are added, but when I add a platform driver for
> > them, the probing does not commence, because they hang in
> > driver base/dd.c: __driver_attach() trying to take the parent semaphore,
> > i.e. this line:
> 
> > 	if (dev->parent)	/* Needed for USB */
> > 		down(&dev->parent->sem);
> 
> > In this case the parent is ab3100-regulators, the platform device for
> > the set of regulators.
> 
> On the face of it (and without having actually looked at a running
> system or anything yet) I'm rather surprised that platform based MFD
> drivers aren't running into this issue more often.  CCing in Alan who
> made the change.

I'm not at all familiar with the detailed issues involved here.  
Perhaps because of this, I don't see why there's any reason for
deadlocking.  __driver_attach() is invoked when a new driver is
registered; to avoid problems all you have to do is make sure you
aren't holding any device locks when you register a driver.

In general, drivers are registered as part of a module initialization
routine, which is quite separate from device probing.  Hence the code
doing the registration doesn't hold any locks.

Is the problem that the probe routine for a platform device is 
registering some child devices and then registering a driver for them?  
If that's the case, the problem can be solved by doing thing in the 
opposite order: Register the driver _before_ registering the devices it 
is supposed to manage.  Then __driver_attach() won't run -- 
device_attach() will instead -- so the problem won't arise.

Alan Stern

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/