diff mbox series

cpufreq: check if policy is inactive before calling __cpufreq_get

Message ID 20190104114249.5355-1-sudeep.holla@arm.com
State New
Headers show
Series cpufreq: check if policy is inactive before calling __cpufreq_get | expand

Commit Message

Sudeep Holla Jan. 4, 2019, 11:42 a.m. UTC 
cpuinfo_cur_freq gets current CPU frequency as detected by hardware
while scaling_cur_freq last known CPU frequency. Some platforms may not
allow checking the CPU frequency of an offline CPU or the associated
resources may have been released via cpufreq_exit when the CPU gets
offlined. If we attempt to get current frequency from the hardware,
it may result in hang or crash.

For example on Juno, I see:

Unable to handle kernel NULL pointer dereference at virtual address 0000000000000188
[0000000000000188] pgd=0000000000000000
Internal error: Oops: 96000004 [#1] PREEMPT SMP
Modules linked in:
CPU: 5 PID: 4202 Comm: cat Not tainted 4.20.0-08251-ga0f2c0318a15-dirty #87
Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform
pstate: 40000005 (nZcv daif -PAN -UAO)
pc : scmi_cpufreq_get_rate+0x34/0xb0
lr : scmi_cpufreq_get_rate+0x34/0xb0
Call trace:
 scmi_cpufreq_get_rate+0x34/0xb0
 __cpufreq_get+0x34/0xc0
 show_cpuinfo_cur_freq+0x24/0x78
 show+0x40/0x60
 sysfs_kf_seq_show+0xc0/0x148
 kernfs_seq_show+0x44/0x50
 seq_read+0xd4/0x480
 kernfs_fop_read+0x15c/0x208
 __vfs_read+0x60/0x188
 vfs_read+0x94/0x150
 ksys_read+0x6c/0xd8
 __arm64_sys_read+0x24/0x30
 el0_svc_common+0x78/0x100
 el0_svc_handler+0x38/0x78
 el0_svc+0x8/0xc
---[ end trace 3d1024e58f77f6b2 ]---

Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

---
 drivers/cpufreq/cpufreq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Hi Viresh, Rafael,

I don't know the reason why we call __cpufreq_get here instead of
cpufreq_get which checks if the policy is active or not. I bumped into
this when I was testing Viresh's OPP fix. This patch can be improved
to avoid policy to cpu, back to policy reference in the path, but I
wanted to get the information I might be missing.

Regards,
Sudeep

-- 
2.17.1

Comments

Viresh Kumar Jan. 7, 2019, 6:02 a.m. UTC | #1 
On 04-01-19, 11:42, Sudeep Holla wrote:
> cpuinfo_cur_freq gets current CPU frequency as detected by hardware

> while scaling_cur_freq last known CPU frequency. Some platforms may not

> allow checking the CPU frequency of an offline CPU or the associated

> resources may have been released via cpufreq_exit when the CPU gets

> offlined. If we attempt to get current frequency from the hardware,

> it may result in hang or crash.

> 

> For example on Juno, I see:

> 

> Unable to handle kernel NULL pointer dereference at virtual address 0000000000000188

> [0000000000000188] pgd=0000000000000000

> Internal error: Oops: 96000004 [#1] PREEMPT SMP

> Modules linked in:

> CPU: 5 PID: 4202 Comm: cat Not tainted 4.20.0-08251-ga0f2c0318a15-dirty #87

> Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform

> pstate: 40000005 (nZcv daif -PAN -UAO)

> pc : scmi_cpufreq_get_rate+0x34/0xb0

> lr : scmi_cpufreq_get_rate+0x34/0xb0

> Call trace:

>  scmi_cpufreq_get_rate+0x34/0xb0

>  __cpufreq_get+0x34/0xc0

>  show_cpuinfo_cur_freq+0x24/0x78

>  show+0x40/0x60

>  sysfs_kf_seq_show+0xc0/0x148

>  kernfs_seq_show+0x44/0x50

>  seq_read+0xd4/0x480

>  kernfs_fop_read+0x15c/0x208

>  __vfs_read+0x60/0x188

>  vfs_read+0x94/0x150

>  ksys_read+0x6c/0xd8

>  __arm64_sys_read+0x24/0x30

>  el0_svc_common+0x78/0x100

>  el0_svc_handler+0x38/0x78

>  el0_svc+0x8/0xc

> ---[ end trace 3d1024e58f77f6b2 ]---

> 

> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>

> Cc: Viresh Kumar <viresh.kumar@linaro.org>

> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

> ---

>  drivers/cpufreq/cpufreq.c | 2 +-

>  1 file changed, 1 insertion(+), 1 deletion(-)

> 

> Hi Viresh, Rafael,

> 

> I don't know the reason why we call __cpufreq_get here instead of

> cpufreq_get which checks if the policy is active or not.


We stopped removing the policy sysfs files, when the last CPU of the policy
exits, sometime back and probably missed this combination.

> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c

> index 6f23ebb395f1..cfc502385855 100644

> --- a/drivers/cpufreq/cpufreq.c

> +++ b/drivers/cpufreq/cpufreq.c

> @@ -712,7 +712,7 @@ store_one(scaling_max_freq, max);

>  static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,

>  					char *buf)

>  {

> -	unsigned int cur_freq = __cpufreq_get(policy);

> +	unsigned int cur_freq = cpufreq_get(policy->cpu);

>  

>  	if (cur_freq)

>  		return sprintf(buf, "%u\n", cur_freq);


I will fix it a bit differently:

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index 81de54cce637..8b1591804ec8 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -1541,17 +1541,16 @@ static unsigned int __cpufreq_get(struct cpufreq_policy *policy)
 {
 	unsigned int ret_freq = 0;
 
-	if (!cpufreq_driver->get)
+	if (unlikely(policy_is_inactive(policy)) || !cpufreq_driver->get)
 		return ret_freq;
 
 	ret_freq = cpufreq_driver->get(policy->cpu);
 
 	/*
-	 * Updating inactive policies is invalid, so avoid doing that.  Also
-	 * if fast frequency switching is used with the given policy, the check
+	 * If fast frequency switching is used with the given policy, the check
 	 * against policy->cur is pointless, so skip it in that case too.
 	 */
-	if (unlikely(policy_is_inactive(policy)) || policy->fast_switch_enabled)
+	if (policy->fast_switch_enabled)
 		return ret_freq;
 
 	if (ret_freq && policy->cur &&
@@ -1580,10 +1579,7 @@ unsigned int cpufreq_get(unsigned int cpu)
 
 	if (policy) {
 		down_read(&policy->rwsem);
-
-		if (!policy_is_inactive(policy))
-			ret_freq = __cpufreq_get(policy);
-
+		ret_freq = __cpufreq_get(policy);
 		up_read(&policy->rwsem);
 
 		cpufreq_cpu_put(policy);

-- 
viresh
Sudeep Holla Jan. 7, 2019, 9:40 a.m. UTC | #2 
On Mon, Jan 07, 2019 at 11:32:41AM +0530, Viresh Kumar wrote:
> On 04-01-19, 11:42, Sudeep Holla wrote:

> > cpuinfo_cur_freq gets current CPU frequency as detected by hardware

> > while scaling_cur_freq last known CPU frequency. Some platforms may not

> > allow checking the CPU frequency of an offline CPU or the associated

> > resources may have been released via cpufreq_exit when the CPU gets

> > offlined. If we attempt to get current frequency from the hardware,

> > it may result in hang or crash.

> >

> > For example on Juno, I see:

> >

> > Unable to handle kernel NULL pointer dereference at virtual address 0000000000000188

> > [0000000000000188] pgd=0000000000000000

> > Internal error: Oops: 96000004 [#1] PREEMPT SMP

> > Modules linked in:

> > CPU: 5 PID: 4202 Comm: cat Not tainted 4.20.0-08251-ga0f2c0318a15-dirty #87

> > Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform

> > pstate: 40000005 (nZcv daif -PAN -UAO)

> > pc : scmi_cpufreq_get_rate+0x34/0xb0

> > lr : scmi_cpufreq_get_rate+0x34/0xb0

> > Call trace:

> >  scmi_cpufreq_get_rate+0x34/0xb0

> >  __cpufreq_get+0x34/0xc0

> >  show_cpuinfo_cur_freq+0x24/0x78

> >  show+0x40/0x60

> >  sysfs_kf_seq_show+0xc0/0x148

> >  kernfs_seq_show+0x44/0x50

> >  seq_read+0xd4/0x480

> >  kernfs_fop_read+0x15c/0x208

> >  __vfs_read+0x60/0x188

> >  vfs_read+0x94/0x150

> >  ksys_read+0x6c/0xd8

> >  __arm64_sys_read+0x24/0x30

> >  el0_svc_common+0x78/0x100

> >  el0_svc_handler+0x38/0x78

> >  el0_svc+0x8/0xc

> > ---[ end trace 3d1024e58f77f6b2 ]---

> >

> > Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>

> > Cc: Viresh Kumar <viresh.kumar@linaro.org>

> > Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

> > ---

> >  drivers/cpufreq/cpufreq.c | 2 +-

> >  1 file changed, 1 insertion(+), 1 deletion(-)

> >

> > Hi Viresh, Rafael,

> >

> > I don't know the reason why we call __cpufreq_get here instead of

> > cpufreq_get which checks if the policy is active or not.

>

> We stopped removing the policy sysfs files, when the last CPU of the policy

> exits, sometime back and probably missed this combination.

>


Ah OK, I did a quick search as I did remember something similar but
didn't get the answer for this.

> > diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c

> > index 6f23ebb395f1..cfc502385855 100644

> > --- a/drivers/cpufreq/cpufreq.c

> > +++ b/drivers/cpufreq/cpufreq.c

> > @@ -712,7 +712,7 @@ store_one(scaling_max_freq, max);

> >  static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,

> >  					char *buf)

> >  {

> > -	unsigned int cur_freq = __cpufreq_get(policy);

> > +	unsigned int cur_freq = cpufreq_get(policy->cpu);

> >

> >  	if (cur_freq)

> >  		return sprintf(buf, "%u\n", cur_freq);

>

> I will fix it a bit differently:

>

> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c

> index 81de54cce637..8b1591804ec8 100644

> --- a/drivers/cpufreq/cpufreq.c

> +++ b/drivers/cpufreq/cpufreq.c

> @@ -1541,17 +1541,16 @@ static unsigned int __cpufreq_get(struct cpufreq_policy *policy)

>  {

>  	unsigned int ret_freq = 0;

>

> -	if (!cpufreq_driver->get)

> +	if (unlikely(policy_is_inactive(policy)) || !cpufreq_driver->get)

>  		return ret_freq;

>

>  	ret_freq = cpufreq_driver->get(policy->cpu);

>

>  	/*

> -	 * Updating inactive policies is invalid, so avoid doing that.  Also

> -	 * if fast frequency switching is used with the given policy, the check

> +	 * If fast frequency switching is used with the given policy, the check

>  	 * against policy->cur is pointless, so skip it in that case too.

>  	 */

> -	if (unlikely(policy_is_inactive(policy)) || policy->fast_switch_enabled)

> +	if (policy->fast_switch_enabled)

>  		return ret_freq;

>

>  	if (ret_freq && policy->cur &&

> @@ -1580,10 +1579,7 @@ unsigned int cpufreq_get(unsigned int cpu)

>

>  	if (policy) {

>  		down_read(&policy->rwsem);

> -

> -		if (!policy_is_inactive(policy))

> -			ret_freq = __cpufreq_get(policy);

> -

> +		ret_freq = __cpufreq_get(policy);


This should work. Do you want to post this or want me to do ? Let me know.

--
Regards,
Sudeep
Viresh Kumar Jan. 7, 2019, 10:59 a.m. UTC | #3 
On 07-01-19, 09:40, Sudeep Holla wrote:
> This should work. Do you want to post this or want me to do ? Let me know.


You can send it.

-- 
viresh
diff mbox series

Patch

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index 6f23ebb395f1..cfc502385855 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -712,7 +712,7 @@  store_one(scaling_max_freq, max);
 static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,
 					char *buf)
 {
-	unsigned int cur_freq = __cpufreq_get(policy);
+	unsigned int cur_freq = cpufreq_get(policy->cpu);
 
 	if (cur_freq)
 		return sprintf(buf, "%u\n", cur_freq);