From mboxrd@z Thu Jan 1 00:00:00 1970 From: linux@armlinux.org.uk (Russell King - ARM Linux) Date: Fri, 4 Nov 2016 12:21:07 +0000 Subject: [PATCH] bus: vexpress-config: fix device reference leak In-Reply-To: <1478259746.17152.126.camel@arm.com> References: <1477997017-29103-1-git-send-email-johan@kernel.org> <1478259746.17152.126.camel@arm.com> Message-ID: <20161104122107.GA1041@n2100.armlinux.org.uk> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Fri, Nov 04, 2016 at 11:42:26AM +0000, Pawel Moll wrote: > On Tue, 2016-11-01 at 11:43 +0100, Johan Hovold wrote: > > Make sure to drop the reference to the parent device taken by > > class_find_device() after populating the bus. > > > > Fixes: 3b9334ac835b ("mfd: vexpress: Convert custom func API to > > regmap") > > Signed-off-by: Johan Hovold > > You're right. May I ask how did you figure it out? The get_device() > happening in class_find_device() is a bit obscure, It's not obscure at all - all the functions that find a device do so under a lock to ensure that the device does not go away, and they take a reference count on the device before returning the pointer for exactly the same reason. If they didn't do that, the find function could locate a struct device while another thread is deleting the struct device, and it would then return a stale pointer - and dereferencing that pointer would then be a use-after-free bug. So not obscure, but rather fundamentally necessary. -- RMK's Patch system: http://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up according to speedtest.net.