Skip to contentSkip to conversation

kobject (ffff88003ffbb4b8): tried to init an initialized object, something is seriously wrong.

DISCUSSION Konrad Rzeszutek Wilk 2 participants 7 messages Feb 3, 2012 – Feb 8, 2012 Driver core

lore
Share & tools
lore
#0 lore raw

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

[   34.984105] Disabling CPU3 present 
[   34.984801] arch_unregister_cpu: CPU 3

sh-4.1#
sh-4.1#
sh-4.1# [ 42.297542] Contemplating CPU 0 which should be online

[   42.298447] Enabling CPU0 present online
[   42.299229] Contemplating CPU 1 which should be online
[   42.300001] Enabling CPU1 present online
[   42.300783] Contemplating CPU 2 which should be online
[   42.301558] Enabling CPU2 present online
[   42.302387] Contemplating CPU 3 which should be online
[   42.303198] Enabling CPU3  
[   42.303895] arch_register_cpu: CPU 3
[   42.304188] kobject (ffff88003ffbb4b8): tried to init an initialized object, something is seriously wrong.
[   42.304188] Pid: 30, comm: xenwatch Tainted: G           O 3.3.0-rc2-00008-g4b64969 #1
[   42.304188] Call Trace:
[   42.304188]  [<ffffffff812ce4d9>] kobject_init+0x89/0xa0
[   42.304188]  [<ffffffff813ca917>] device_initialize+0x27/0x90
[   42.304188]  [<ffffffff813cbbc1>] device_register+0x11/0x20
[   42.304188]  [<ffffffff815f2210>] register_cpu+0x52/0xb6
[   42.304188]  [<ffffffff815d66c0>] arch_register_cpu+0x60/0x70
[   42.304188]  [<ffffffff8136b095>] handle_vcpu_hotplug_event+0x175/0x180
[   42.304188]  [<ffffffff81367140>] xenwatch_thread+0xb0/0x180
[   42.304188]  [<ffffffff810a7ec0>] ? wake_up_bit+0x40/0x40
[   42.304188]  [<ffffffff81367090>] ? split+0xf0/0xf0
[   42.304188]  [<ffffffff810a77e6>] kthread+0x96/0xa0
[   42.304188]  [<ffffffff815fec24>] kernel_thread_helper+0x4/0x10
[   42.304188]  [<ffffffff815f6b40>] ? retint_restore_args+0x5/0x6
[   42.304188]  [<ffffffff815fec20>] ? gs_change+0x13/0x13
maintainer
#1 lore raw
Quoted 11 lines · Konrad Rzeszutek Wilk

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

No, it's related to the previous issue, let me knock up a patch to fix this as well...

thanks,

greg k-h

maintainer
#2 lore raw
Quoted 10 lines · Greg KH
Quoted 9 lines · Konrad Rzeszutek Wilk

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

Quoted 5 lines
Quoted 2 lines

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

No, it's related to the previous issue, let me knock up a patch to fix this as well...

Does this patch solve the problem for you?

thanks,

greg k-h

drivers/base/cpu.c +1 -0
index 23f2c4c..4dabf50 100644
@@ -240,6 +240,7 @@ int __cpuinit register_cpu(struct cpu *cpu, int num)
 	int error;
 
 	cpu->node_id = cpu_to_node(num);
+	memset(&cpu->dev, 0x00, sizeof(struct device));
 	cpu->dev.id = num;
 	cpu->dev.bus = &cpu_subsys;
 	cpu->dev.release = cpu_device_release;
#3 lore raw
Quoted 11 lines · Greg KH
Quoted 10 lines · Greg KH
Quoted 9 lines · Konrad Rzeszutek Wilk

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

No, it's related to the previous issue, let me knock up a patch to fix this as well...

Does this patch solve the problem for you?

Yeah, no warnings. But neither is the CPU coming back up :-(

Thought that might be related to something else - let me double check with 3.2 whether there are some race-conditions that I might be hitting.

Quoted 17 lines

thanks,

greg k-h

drivers/base/cpu.c +1 -0
index 23f2c4c..4dabf50 100644
@@ -240,6 +240,7 @@ int __cpuinit register_cpu(struct cpu *cpu, int num)
 	int error;
 
 	cpu->node_id = cpu_to_node(num);
+	memset(&cpu->dev, 0x00, sizeof(struct device));
 	cpu->dev.id = num;
 	cpu->dev.bus = &cpu_subsys;
 	cpu->dev.release = cpu_device_release;
maintainer
#4 lore raw
Quoted 15 lines · Konrad Rzeszutek Wilk
Quoted 14 lines · Greg KH
Quoted 13 lines · Greg KH
Quoted 11 lines · Konrad Rzeszutek Wilk

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

No, it's related to the previous issue, let me knock up a patch to fix this as well...

Does this patch solve the problem for you?

Yeah, no warnings. But neither is the CPU coming back up :-(

Very wierd.

Quoted 2 lines

Thought that might be related to something else - let me double check with 3.2 whether there are some race-conditions that I might be hitting.

Ok, I'll wait to queue this up until you get back to me.

greg k-h

#5 lore raw
Quoted 16 lines · Greg KH
Quoted 15 lines · Konrad Rzeszutek Wilk
Quoted 14 lines · Greg KH
Quoted 13 lines · Greg KH
Quoted 11 lines · Konrad Rzeszutek Wilk

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

No, it's related to the previous issue, let me knock up a patch to fix this as well...

Does this patch solve the problem for you?

Yeah, no warnings. But neither is the CPU coming back up :-(

Very wierd.

Found the culprit and have a patch for it. But it does not affect baremetal and it was in the drivers/xen/cpu_hotplug.c code - so

Quoted 5 lines
Quoted 3 lines

Thought that might be related to something else - let me double check with 3.2 whether there are some race-conditions that I might be hitting.

Ok, I'll wait to queue this up until you get back to me.

.. you can go ahead and queue it up with my Tested-by tag. Thanks!

maintainer
#6 lore raw
Quoted 30 lines · Konrad Rzeszutek Wilk
Quoted 16 lines · Greg KH
Quoted 15 lines · Konrad Rzeszutek Wilk
Quoted 14 lines · Greg KH
Quoted 13 lines · Greg KH
Quoted 11 lines · Konrad Rzeszutek Wilk

Hey Greg,

With or without your previous patch (the one that added an cpu_device_release function) I am getting this in 3.3-rc2. I did not get this in 3.2.

I added some extra printks to make sure that the handle_vcpu_hotplug_event code actually did call 'arch_unregister_cpu' which is certainly looks to be doing.

I hadn't done a bisection yet, but was wondering if this is related to the previous issues with the sysdev. Or kobject getting more observant.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

No, it's related to the previous issue, let me knock up a patch to fix this as well...

Does this patch solve the problem for you?

Yeah, no warnings. But neither is the CPU coming back up :-(

Very wierd.

Found the culprit and have a patch for it. But it does not affect baremetal and it was in the drivers/xen/cpu_hotplug.c code - so

Quoted 1 line
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

Thought that might be related to something else - let me double check with 3.2 whether there are some race-conditions that I might be hitting.

Ok, I'll wait to queue this up until you get back to me.

.. you can go ahead and queue it up with my Tested-by tag. Thanks!

Great, will do.