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  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:	Mon, 16 Jun 2014 11:03:02 +0800
From:	Allen Yu <alleny@...dia.com>
To:	Alan Stern <stern@...land.harvard.edu>
CC:	"Rafael J. Wysocki" <rjw@...ysocki.net>,
	Pavel Machek <pavel@....cz>, "Len Brown" <len.brown@...el.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	"linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH 1/1] PM / Runtime: let rpm_resume fail if rpm disabled
 and device suspended.

On Sat, 14 Jun 2014, Alan Stern wrote:

> > dev->power.is_suspended is set after core suspends device during system
> suspend.
> > This flag mostly means device is not operational (all I/O been
> > quiesced, no more data read or write acceptible, etc.), hence it's
> > dangerous to access hardware if device is suspended even though runtime
> PM status is RPM_ACTIVE.
> >
> > In turn, we should allow device to be accessed in case device is *not*
> > suspended and runtime PM status is RPM_ACTIVE, even if runtime PM
> > disabled. This corner case can happen to a device in a generic PM
> > domain if the domain is not powered off while preparing for a system-wide
> power transition.
> 
> I don't understand.  Even if the PM domain isn't powered off, the device's
> is_suspended flag will still be set by __device_suspend().

Yes, is_suspended flag will be set by __device_suspend(). But runtime PM can be disabled in pm_genpd_prepare():

914 static int pm_genpd_prepare(struct device *dev){
...
956         /*
957          * The PM domain must be in the GPD_STATE_ACTIVE state at this point,
958          * so pm_genpd_poweron() will return immediately, but if the device
959          * is suspended (e.g. it's been stopped by genpd_stop_dev()), we need
960          * to make it operational.
961          */
962         pm_runtime_resume(dev);
963         __pm_runtime_disable(dev, false);
...
978 }

And there is a gap between pm_genpd_prepare() and __device_suspend(), during which is_suspended flag is *not* yet set but runtime PM is disabled and device status is RPM_ACTIVE. 

> 
> >  In this case, runtime PM status will
> > be set to RPM_ACTIVE and then runtime PM is disabled. After that,
> > device driver may call pm_runtime_get_sync() and rpm_resume() should
> > return 1, because the device is still active as long as not been suspended.
> 
> Isn't that what the existing code does already?  Your patch seems to change
> it so that it _doesn't_ behave the way you want.
> 

The existing code return 1 for case is_suspended flag is set.  That means __device_suspend() has been called and device is not in operational state. Whereas the case I mentioned above is before device is suspended.
It's dangerous to access device if it's in suspended state, so I propose only allowing access to a device if it's not suspended (i.e. value of "is_suspneded" flag is false).

> > Signed-off-by: Allen Yu <alleny@...dia.com>
> > ---
> >  drivers/base/power/runtime.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/base/power/runtime.c
> > b/drivers/base/power/runtime.c index 67c7938..39885f1 100644
> > --- a/drivers/base/power/runtime.c
> > +++ b/drivers/base/power/runtime.c
> > @@ -608,7 +608,7 @@ static int rpm_resume(struct device *dev, int
> rpmflags)
> >   repeat:
> >  	if (dev->power.runtime_error)
> >  		retval = -EINVAL;
> > -	else if (dev->power.disable_depth == 1 && dev-
> >power.is_suspended
> > +	else if (dev->power.disable_depth == 1 && !dev-
> >power.is_suspended
> >  	    && dev->power.runtime_status == RPM_ACTIVE)
> >  		retval = 1;
> 
> 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/

Powered by blists - more mailing lists