Skip to contentSkip to conversation

PM: hibernate: add sanity check on power_kobj

PATCH In review Abel Wu 2 participants 4 messages Feb 1, 2021 – Feb 2, 2021

Subsystems (2)
lore
Share & tools
lore

Abel Wu

author
#0 lore raw

PATCH 1/1 PM: hibernate: add sanity check on power_kobj

The @power_kobj is initialized in pm_init() which is the same initcall level as pm_disk_init(). Although this dependency is guaranteed based on the current initcall serial execution model, it would still be better do a cost-less sanity check to avoid oops once the dependency is broken.

---
 kernel/power/hibernate.c | 3 +++
 1 file changed, 3 insertions(+)
kernel/power/hibernate.c +3 -0
index da0b41914177..060089cc261d 100644
@@ -1262,6 +1262,9 @@ static const struct attribute_group attr_group = {
 
 static int __init pm_disk_init(void)
 {
+	if (!power_kobj)
+		return -EINVAL;
+
 	return sysfs_create_group(power_kobj, &attr_group);
 }
 
Signature · 1 line
2.27.0
#1 lore raw
Quoted 5 lines · Abel Wu

The @power_kobj is initialized in pm_init() which is the same initcall level as pm_disk_init(). Although this dependency is guaranteed based on the current initcall serial execution model, it would still be better do a cost-less sanity check to avoid oops once the dependency is broken.

I don't believe this is good idea. If the dependency is ever broken, this will make failure more subtle and harder to debug.

Quoted 21 lines
---
 kernel/power/hibernate.c | 3 +++
 1 file changed, 3 insertions(+)
kernel/power/hibernate.c +3 -0
index da0b41914177..060089cc261d 100644
@@ -1262,6 +1262,9 @@ static const struct attribute_group attr_group = {
 
 static int __init pm_disk_init(void)
 {
+	if (!power_kobj)
+		return -EINVAL;
+
 	return sysfs_create_group(power_kobj, &attr_group);
 }
 
Signature · 1 line
2.27.0
Signature · 1 line
http://www.livejournal.com/~pavelmachek
author
#2 lore raw
Quoted 13 lines · Pavel Machek
Quoted 5 lines · Abel Wu

The @power_kobj is initialized in pm_init() which is the same initcall level as pm_disk_init(). Although this dependency is guaranteed based on the current initcall serial execution model, it would still be better do a cost-less sanity check to avoid oops once the dependency is broken.

I don't believe this is good idea. If the dependency is ever broken, this will make failure more subtle and harder to debug.

Thanks for reviewing. I think the cmdline parameter initcall_debug will help in this case.
Actually we are trying to make initcalls being called asynchronously to reduce boot time which is crucial to our cloud-native business. And we resolve this kind of dependencies by retrying failed initcalls.

Best regards,
Abel

Quoted 25 lines
Quoted 22 lines

---
kernel/power/hibernate.c | 3 +++

1 file changed, 3 insertions(+)
kernel/power/hibernate.c +3 -0
index da0b41914177..060089cc261d 100644
@@ -1262,6 +1262,9 @@ static const struct attribute_group attr_group = {
 
 static int __init pm_disk_init(void)
 {
+	if (!power_kobj)
+		return -EINVAL;
+
 	return sysfs_create_group(power_kobj, &attr_group);
 }
 
Signature · 1 line
2.27.0
Signature · 1 line
http://www.livejournal.com/~pavelmachek
#3 lore raw
Quoted 11 lines · Abel Wu
Quoted 11 lines · Pavel Machek
Quoted 5 lines · Abel Wu

The @power_kobj is initialized in pm_init() which is the same initcall level as pm_disk_init(). Although this dependency is guaranteed based on the current initcall serial execution model, it would still be better do a cost-less sanity check to avoid oops once the dependency is broken.

Quoted 8 lines
Quoted 2 lines

I don't believe this is good idea. If the dependency is ever broken, this will make failure more subtle and harder to debug.

Thanks for reviewing. I think the cmdline parameter initcall_debug will help in this case.
Actually we are trying to make initcalls being called asynchronously to reduce boot time which is crucial to our cloud-native business. And we resolve this kind of dependencies by retrying failed initcalls.

And this patch is okay if that gets mainlined, but not before.

Best regards,
Pavel

Signature · 1 line
http://www.livejournal.com/~pavelmachek