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.
Signed-off-by: Abel Wu <[email protected]>
---
kernel/power/hibernate.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
index da0b41914177..060089cc261d 100644
--- a/kernel/power/hibernate.c
+++ b/kernel/power/hibernate.c
@@ -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);
}
--
2.27.0
On Mon 2021-02-01 02:50:41, Abel Wu wrote:
> 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.
> Signed-off-by: Abel Wu <[email protected]>
> ---
> kernel/power/hibernate.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
> index da0b41914177..060089cc261d 100644
> --- a/kernel/power/hibernate.c
> +++ b/kernel/power/hibernate.c
> @@ -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);
> }
>
> --
> 2.27.0
--
http://www.livejournal.com/~pavelmachek
> On Feb 1, 2021, at 6:52 PM, Pavel Machek <[email protected]> wrote:
>
> On Mon 2021-02-01 02:50:41, Abel Wu wrote:
>> 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
>
>> Signed-off-by: Abel Wu <[email protected]>
>> ---
>> kernel/power/hibernate.c | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
>> index da0b41914177..060089cc261d 100644
>> --- a/kernel/power/hibernate.c
>> +++ b/kernel/power/hibernate.c
>> @@ -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);
>> }
>>
>> --
>> 2.27.0
>
> --
> http://www.livejournal.com/~pavelmachek
On Tue 2021-02-02 09:59:11, Abel Wu wrote:
>
>
> > On Feb 1, 2021, at 6:52 PM, Pavel Machek <[email protected]> wrote:
> >
> > On Mon 2021-02-01 02:50:41, Abel Wu wrote:
> >> 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.
And this patch is okay if that gets mainlined, but not before.
Best regards,
Pavel
--
http://www.livejournal.com/~pavelmachek