| Message ID | 1424800730-32059-3-git-send-email-lorenzo.pieralisi@arm.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Delegated to: | Rafael Wysocki |
| Headers | show |
On 02/24/2015 06:58 PM, Lorenzo Pieralisi wrote: > The changes in commit: > > 381063133246 ("PM / sleep: Re-implement suspend-to-idle handling") > > let suspend-to-idle code bypass the cpuidle_select() function to > enter the deepest idle state. The sanity checks carried out in > cpuidle_select() are bypassed too and this can cause breakage > on systems that try to suspend-to-idle with no registered cpuidle > driver. > > This patch factors out a function cpuidle_device_disabled() that > is used to carry out sanity checks (ie CPUidle is disabled on the > cpu executing the code) in both cpuidle_select() and cpuidle_enter_freeze() > so that the checks are unified and carried out in both control paths. > > Cc: Rafael J. Wysocki <rjw@rjwysocki.net> > Cc: Daniel Lezcano <daniel.lezcano@linaro.org> > Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> > --- > drivers/cpuidle/cpuidle.c | 18 +++++++++++++----- > 1 file changed, 13 insertions(+), 5 deletions(-) > > diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c > index f47edc6c..344fe6c 100644 > --- a/drivers/cpuidle/cpuidle.c > +++ b/drivers/cpuidle/cpuidle.c > @@ -44,6 +44,12 @@ void disable_cpuidle(void) > off = 1; > } > > +static bool cpuidle_device_disabled(struct cpuidle_driver *drv, > + struct cpuidle_device *dev) > +{ > + return (off || !initialized || !drv || !dev || !dev->enabled); > +} This is getting a bit fuzzy IMO. What means disabled ? :) > /** > * cpuidle_play_dead - cpu off-lining > * > @@ -124,6 +130,11 @@ void cpuidle_enter_freeze(void) > struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev); > int index; I think this is exploding before because of dev == NULL in the line above. > + if (cpuidle_device_disabled(drv, dev)) { > + arch_cpu_idle(); > + return; > + } > + > /* > * Find the deepest state with ->enter_freeze present, which guarantees > * that interrupts won't be enabled when it exits and allows the tick to > @@ -202,11 +213,8 @@ int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv, > */ > int cpuidle_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) > { > - if (off || !initialized) > - return -ENODEV; > - > - if (!drv || !dev || !dev->enabled) > - return -EBUSY; > + if (cpuidle_device_disabled(drv, dev)) > + return -1; > > return cpuidle_curr_governor->select(drv, dev); > } >
On Wed, Feb 25, 2015 at 02:30:49PM +0000, Daniel Lezcano wrote: > On 02/24/2015 06:58 PM, Lorenzo Pieralisi wrote: > > The changes in commit: > > > > 381063133246 ("PM / sleep: Re-implement suspend-to-idle handling") > > > > let suspend-to-idle code bypass the cpuidle_select() function to > > enter the deepest idle state. The sanity checks carried out in > > cpuidle_select() are bypassed too and this can cause breakage > > on systems that try to suspend-to-idle with no registered cpuidle > > driver. > > > > This patch factors out a function cpuidle_device_disabled() that > > is used to carry out sanity checks (ie CPUidle is disabled on the > > cpu executing the code) in both cpuidle_select() and cpuidle_enter_freeze() > > so that the checks are unified and carried out in both control paths. > > > > Cc: Rafael J. Wysocki <rjw@rjwysocki.net> > > Cc: Daniel Lezcano <daniel.lezcano@linaro.org> > > Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> > > --- > > drivers/cpuidle/cpuidle.c | 18 +++++++++++++----- > > 1 file changed, 13 insertions(+), 5 deletions(-) > > > > diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c > > index f47edc6c..344fe6c 100644 > > --- a/drivers/cpuidle/cpuidle.c > > +++ b/drivers/cpuidle/cpuidle.c > > @@ -44,6 +44,12 @@ void disable_cpuidle(void) > > off = 1; > > } > > > > +static bool cpuidle_device_disabled(struct cpuidle_driver *drv, > > + struct cpuidle_device *dev) > > +{ > > + return (off || !initialized || !drv || !dev || !dev->enabled); > > +} > > This is getting a bit fuzzy IMO. What means disabled ? :) Well, that's just the current checks in cpuidle_select() (that by the way is supposed to return an index) merged together with a function name, to reuse the same checks in cpuidle_enter_freeze(). I have no problem leaving the checks as they are at the moment and replicate them in cpuidle_enter_freeze() but given your remark below, we should do something different in there. > > > /** > > * cpuidle_play_dead - cpu off-lining > > * > > @@ -124,6 +130,11 @@ void cpuidle_enter_freeze(void) > > struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev); > > int index; > > I think this is exploding before because of dev == NULL in the line above. Yes, good point so my attempt at consolidating the sanity checks above is not valid, but something has to be done regardless. Lorenzo > > + if (cpuidle_device_disabled(drv, dev)) { > > + arch_cpu_idle(); > > + return; > > + } > > + > > /* > > * Find the deepest state with ->enter_freeze present, which guarantees > > * that interrupts won't be enabled when it exits and allows the tick to > > @@ -202,11 +213,8 @@ int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv, > > */ > > int cpuidle_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) > > { > > - if (off || !initialized) > > - return -ENODEV; > > - > > - if (!drv || !dev || !dev->enabled) > > - return -EBUSY; > > + if (cpuidle_device_disabled(drv, dev)) > > + return -1; > > > > return cpuidle_curr_governor->select(drv, dev); > > } > > > > > -- > <http://www.linaro.org/> Linaro.org ? Open source software for ARM SoCs > > Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook | > <http://twitter.com/#!/linaroorg> Twitter | > <http://www.linaro.org/linaro-blog/> Blog > > -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, Feb 25, 2015 at 02:30:49PM +0000, Daniel Lezcano wrote: [...] > > diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c > > index f47edc6c..344fe6c 100644 > > --- a/drivers/cpuidle/cpuidle.c > > +++ b/drivers/cpuidle/cpuidle.c > > @@ -44,6 +44,12 @@ void disable_cpuidle(void) > > off = 1; > > } > > > > +static bool cpuidle_device_disabled(struct cpuidle_driver *drv, > > + struct cpuidle_device *dev) > > +{ > > + return (off || !initialized || !drv || !dev || !dev->enabled); > > +} > > This is getting a bit fuzzy IMO. What means disabled ? :) > > > /** > > * cpuidle_play_dead - cpu off-lining > > * > > @@ -124,6 +130,11 @@ void cpuidle_enter_freeze(void) > > struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev); > > int index; > > I think this is exploding before because of dev == NULL in the line above. Actually not, cpuidle_get_cpu_driver() checks the dev pointer, so we might end up with drv == NULL and dev == NULL, and the check I added still applies and it is effective I think. Lorenzo -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c index f47edc6c..344fe6c 100644 --- a/drivers/cpuidle/cpuidle.c +++ b/drivers/cpuidle/cpuidle.c @@ -44,6 +44,12 @@ void disable_cpuidle(void) off = 1; } +static bool cpuidle_device_disabled(struct cpuidle_driver *drv, + struct cpuidle_device *dev) +{ + return (off || !initialized || !drv || !dev || !dev->enabled); +} + /** * cpuidle_play_dead - cpu off-lining * @@ -124,6 +130,11 @@ void cpuidle_enter_freeze(void) struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev); int index; + if (cpuidle_device_disabled(drv, dev)) { + arch_cpu_idle(); + return; + } + /* * Find the deepest state with ->enter_freeze present, which guarantees * that interrupts won't be enabled when it exits and allows the tick to @@ -202,11 +213,8 @@ int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv, */ int cpuidle_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) { - if (off || !initialized) - return -ENODEV; - - if (!drv || !dev || !dev->enabled) - return -EBUSY; + if (cpuidle_device_disabled(drv, dev)) + return -1; return cpuidle_curr_governor->select(drv, dev); }
The changes in commit: 381063133246 ("PM / sleep: Re-implement suspend-to-idle handling") let suspend-to-idle code bypass the cpuidle_select() function to enter the deepest idle state. The sanity checks carried out in cpuidle_select() are bypassed too and this can cause breakage on systems that try to suspend-to-idle with no registered cpuidle driver. This patch factors out a function cpuidle_device_disabled() that is used to carry out sanity checks (ie CPUidle is disabled on the cpu executing the code) in both cpuidle_select() and cpuidle_enter_freeze() so that the checks are unified and carried out in both control paths. Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Daniel Lezcano <daniel.lezcano@linaro.org> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> --- drivers/cpuidle/cpuidle.c | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-)