linux-kernel - Re: [BUG] Device unbound in a resumed state

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]

Date:	Fri, 11 Mar 2016 13:46:12 -0500 (EST)
From:	Alan Stern <stern@...land.harvard.edu>
To:	Krzysztof Kozlowski <k.kozlowski@...sung.com>
cc:	Ulf Hansson <ulf.hansson@...aro.org>,
	Kevin Hilman <khilman@...aro.org>,
	"Rafael J. Wysocki" <rjw@...ysocki.net>,
	Len Brown <len.brown@...el.com>, Pavel Machek <pavel@....cz>,
	"linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
	linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: [BUG] Device unbound in a resumed state

On Fri, 11 Mar 2016, Krzysztof Kozlowski wrote:

> Hi,
> 
> 
> Could be related (the same?) with [0].
> 
> I have a driver (hwrng/exynos-rng) which in probe does:
> pm_runtime_set_autosuspend_delay(&pdev->dev, EXYNOS_AUTOSUSPEND_DELAY);
> pm_runtime_use_autosuspend(&pdev->dev);
> pm_runtime_enable(&pdev->dev);
> 
> and in remove:
> pm_runtime_disable(&pdev->dev)

But not pm_runtime_dont_use_autosuspend()?

Why disable runtime PM if you want the runtime-PM methods to put the 
device into a low-power state?

> Just before unbinding in __device_release_driver() the device is resumed
> but unfortunately not suspended later. I mean the
> __device_release_driver()->pm_runtime_put_sync() does not trigger
> runtime suspend.

Because autosuspend is still in use at this point.

> This leads to leaving the device in active state (e.g. clocks enabled).
> 
> It does not happen after removal of autosuspend. Also runtime suspend
> happens after very fast unbind-bind.

Overall it sounds like the system is behaving the way it is supposed 
to.

But maybe we should make pm_runtime_use_autosuspend() call
pm_runtime_mark_last_busy(), to avoid the unbind - bind - immediate
autosuspend behavior.

Alan Stern