how to hard hang the kernel from userspace?

how to hard hang the kernel from userspace?

[Daniel Walker]

Hi,


I was able to formulate this command,

echo -1 > /proc/sys/kernel/sched_rt_runtime_us
chrt -f 99 yes &> /dev/null

and it works sometimes on UP, but it doesn't work on SMP. I also tried 
adding in "taskset" to pin multiple version of the "yes" command to each 
cpu in the system, but that just resulted in lots of rcu warnings.


The goal is to make the watchdog trigger by starving out the watchdog 
petting process. Anyone have any ideas? I'd rather not make a kernel module.


Thanks,

Daniel

how to hard hang the kernel from userspace?

[valdis.kletnieks at vt.edu]

On Wed, 06 Dec 2017 16:32:26 -0800, Daniel Walker said:

> and it works sometimes on UP, but it doesn't work on SMP. I also tried

Thinking about *why* that hangs a UP should explain why it doesn't work
on SMP.

> adding in "taskset" to pin multiple version of the "yes" command to each
> cpu in the system, but that just resulted in lots of rcu warnings.

That's.... unexpected.

> The goal is to make the watchdog trigger by starving out the watchdog
> petting process. Anyone have any ideas? I'd rather not make a kernel module.

What are you trying to achieve by doing that?

Have you tried just pointing a 'kill -9' at the watchdog monitor process?  Are you using
in-kernel watchdogd, or one in userspace?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 486 bytes
Desc: not available
Url : http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20171207/ec60fe15/attachment.bin 

how to hard hang the kernel from userspace?

[Daniel Walker]

On 12/07/2017 07:49 AM, valdis.kletnieks at vt.edu wrote:
> On Wed, 06 Dec 2017 16:32:26 -0800, Daniel Walker said:
>
>> and it works sometimes on UP, but it doesn't work on SMP. I also tried
> Thinking about *why* that hangs a UP should explain why it doesn't work
> on SMP.
>
>> adding in "taskset" to pin multiple version of the "yes" command to each
>> cpu in the system, but that just resulted in lots of rcu warnings.
> That's.... unexpected.

May have to do with not being able to start them all instantly enough, 
something like that.

>> The goal is to make the watchdog trigger by starving out the watchdog
>> petting process. Anyone have any ideas? I'd rather not make a kernel module.
> What are you trying to achieve by doing that?
>
> Have you tried just pointing a 'kill -9' at the watchdog monitor process?  Are you using
> in-kernel watchdogd, or one in userspace?


So typically you have a process which updates the watchdog periodically 
which I call a petting process. As I said the purpose is to trigger the 
watchdog by starving the petting process (i.e. kick process) , then the 
watchdog reboots the system.


Daniel

how to hard hang the kernel from userspace?

[valdis.kletnieks at vt.edu]

On Thu, 07 Dec 2017 15:12:15 -0800, Daniel Walker said:

> So typically you have a process which updates the watchdog periodically
> which I call a petting process. As I said the purpose is to trigger the
> watchdog by starving the petting process (i.e. kick process) , then the
> watchdog reboots the system.

If you're doing the updating from userspace, a kill -9 will get the job done.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 486 bytes
Desc: not available
Url : http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20171207/482d6acd/attachment.bin 

how to hard hang the kernel from userspace?

[Daniel Walker]

On 12/07/2017 06:40 PM, valdis.kletnieks at vt.edu wrote:
> On Thu, 07 Dec 2017 15:12:15 -0800, Daniel Walker said:
>
>> So typically you have a process which updates the watchdog periodically
>> which I call a petting process. As I said the purpose is to trigger the
>> watchdog by starving the petting process (i.e. kick process) , then the
>> watchdog reboots the system.
> If you're doing the updating from userspace, a kill -9 will get the job done.
>

True, but that wasn't what I had in mind.. I was more interesting in 
simulating an actual lockup.


Daniel