[PATCH] parisc: Suspend lockup detectors before system halt

[PATCH] parisc: Suspend lockup detectors before system halt

[Helge Deller]

Some machines can't power off the machine, so disable the lockup detectors to
avoid this watchdog BUG to show up every few seconds:
watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [systemd-shutdow:1]

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # 4.9+

diff --git a/arch/parisc/kernel/process.c b/arch/parisc/kernel/process.c
index b64d7d2..a45a67d 100644
--- a/arch/parisc/kernel/process.c
+++ b/arch/parisc/kernel/process.c
@@ -53,6 +53,7 @@
 #include <linux/uaccess.h>
 #include <linux/rcupdate.h>
 #include <linux/random.h>
+#include <linux/nmi.h>
 
 #include <asm/io.h>
 #include <asm/asm-offsets.h>
@@ -145,6 +146,7 @@ void machine_power_off(void)
 
 	/* prevent soft lockup/stalled CPU messages for endless loop. */
 	rcu_sysrq_start();
+	lockup_detector_suspend();
 	for (;;);
 }
 

Re: [PATCH] parisc: Suspend lockup detectors before system halt

[Rolf Eike Beer]

[-- Attachment #1: Type: text/plain, Size: 364 bytes --]

Helge Deller wrote:
> Some machines can't power off the machine, so disable the lockup detectors
> to avoid this watchdog BUG to show up every few seconds:
> watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [systemd-shutdow:1]

I don't think HPPA is the only architecture that has this problem, so isn't 
there a generic solution? Or: is it time to add one?

Eike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: [PATCH] parisc: Suspend lockup detectors before system halt

[Helge Deller]

On 25.07.2017 21:47, Rolf Eike Beer wrote:
> Helge Deller wrote:
>> Some machines can't power off the machine, so disable the lockup detectors
>> to avoid this watchdog BUG to show up every few seconds:
>> watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [systemd-shutdow:1]
> 
> I don't think HPPA is the only architecture that has this problem,

Maybe.
I haven't heard of others yet, probably because most other architectures
do power off themselves, so they won't see this problem.

> so isn't there a generic solution? Or: is it time to add one?

Sure, you are welcome to test, post and discuss patches upstream if you like.
It's not a very critical issue, so I can wait with pushing my patch.

Helge

Re: [PATCH] parisc: Suspend lockup detectors before system halt

[Helge Deller]

On 25.07.2017 21:55, Helge Deller wrote:
> On 25.07.2017 21:47, Rolf Eike Beer wrote:
>> Helge Deller wrote:
>>> Some machines can't power off the machine, so disable the lockup detectors
>>> to avoid this watchdog BUG to show up every few seconds:
>>> watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [systemd-shutdow:1]
>>
>> I don't think HPPA is the only architecture that has this problem,
> 
> Maybe.
> I haven't heard of others yet, probably because most other architectures
> do power off themselves, so they won't see this problem.
> 
>> so isn't there a generic solution? Or: is it time to add one?
> 
> Sure, you are welcome to test, post and discuss patches upstream if you like.

By the way, I still see another lockup once (workqueue lockup) even after this patch.
In general it seems most lockup detectors don't unregister themselves in the
poweroff/shutdown phase (which might be intentional):  

[  123.664431] BUG: workqueue lockup - pool cpus=0 node=0 flags=0x0 nice=0 stuck for 36s!
[  123.857429] Showing busy workqueues and worker pools:
[  123.917919] workqueue events: flags=0x0
[  123.963830]   pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=1/256                 
[  123.967817]     pending: vmstat_shepherd                                         
[  124.082949] workqueue events_freezable: flags=0x4                                
[  124.139364]   pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=1/256                 
[  124.143350]     pending: pci_pme_list_scan                                       
[  124.260448] workqueue events_power_efficient: flags=0x80
[  124.324059]   pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=1/256
[  124.328045]     pending: neigh_periodic_work
[  124.447237] workqueue mm_percpu_wq: flags=0x8
[  124.499488]   pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=1/256
[  124.510543]     pending: vmstat_update
[  124.616432] workqueue mpt_poll_1: flags=0x8
[  124.666595]   pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=1/256
[  124.673079]     pending: mpt_fault_reset_work

Helge

Re: [PATCH] parisc: Suspend lockup detectors before system halt

[Rolf Eike Beer]

[-- Attachment #1: Type: text/plain, Size: 917 bytes --]

Helge Deller wrote:
> On 25.07.2017 21:47, Rolf Eike Beer wrote:
> > Helge Deller wrote:
> >> Some machines can't power off the machine, so disable the lockup
> >> detectors
> >> to avoid this watchdog BUG to show up every few seconds:
> >> watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [systemd-shutdow:1]
> > 
> > I don't think HPPA is the only architecture that has this problem,
> 
> Maybe.
> I haven't heard of others yet, probably because most other architectures
> do power off themselves, so they won't see this problem.
> 
> > so isn't there a generic solution? Or: is it time to add one?
> 
> Sure, you are welcome to test, post and discuss patches upstream if you
> like. It's not a very critical issue, so I can wait with pushing my patch.

I wont be able to do anything like that this week (maybe longer), so please go 
ahead with your patch to get rid of the annoying warning. Cleanups later.

Eike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 181 bytes --]