Re: [qemu-s390x] current s390-next crashes with ERROR:cpus.c:1767:qemu_mutex_lock_iothread: assertion failed: (!qemu_mutex_iothread_locked())

[David Hildenbrand]

On 29.06.2018 16:47, Christian Borntraeger wrote:
> Running the QEMU I/O test now results in errors like
> ERROR:cpus.c:1767:qemu_mutex_lock_iothread: assertion failed: 
> (!qemu_mutex_iothread_locked())
> in several cases (e.g. 068)
> 
> 
> +**
> +ERROR:/home/cborntra/REPOS/qemu/cpus.c:1767:qemu_mutex_lock_iothread: 
> assertion failed: (!qemu_mutex_iothread_locked())
> +./common.rc: line 51: 48580 Aborted                 (core dumped) ( if [ -n 
> "${QEMU_NEED_PID}" ]; then
> +    echo $BASHPID > "${QEMU_TEST_DIR}/qemu-${_QEMU_HANDLE}.pid";
> +fi; exec "$QEMU_PROG" $QEMU_OPTIONS "$@" )
>  QEMU X.Y.Z monitor - type 'help' for more information
>  (qemu) loadvm 0
>  (qemu) loadvm 0
> -(qemu) quit
>  
>  === Saving and reloading a VM state to/from a qcow2 image (-object 
> iothread,id=iothread0 -set device.hba0.iothread=iothread0) ===
>  
> @@ -17,8 +21,11 @@
>  QEMU X.Y.Z monitor - type 'help' for more information
>  (qemu) savevm 0
>  (qemu) quit
> +**
> +ERROR:/home/cborntra/REPOS/qemu/cpus.c:1767:qemu_mutex_lock_iothread: 
> assertion failed: (!qemu_mutex_iothread_locked())
> +./common.rc: line 51: 48624 Aborted                 (core dumped) ( if [ -n 
> "${QEMU_NEED_PID}" ]; then
> +    echo $BASHPID > "${QEMU_TEST_DIR}/qemu-${_QEMU_HANDLE}.pid";
> +fi; exec "$QEMU_PROG" $QEMU_OPTIONS "$@" )
>  QEMU X.Y.Z monitor - type 'help' for more information
>  (qemu) loadvm 0
> -(qemu) loadvm 0
> -(qemu) quit
> -*** done
> +(qemu) loa*** done
> Failures: 068
> Failed 1 of 1 tests
> 
> Bisect points to 
> 
> commit 6a29769cb54f69d346fa12f922f35021c386b596
> Author: David Hildenbrand <address@hidden>
> Date:   Wed Jun 27 15:44:07 2018 +0200
> 
>     s390x/tcg: SET CLOCK COMPARATOR can clear CKC interrupt
> 
> 
> and reverting that helps.
> 

The real problem might be

commit 00abe0c7711db097931972793e4c1671400b8c27
Author: David Hildenbrand <address@hidden>
Date:   Tue Feb 13 12:23:53 2018 +0100

    s390x/tcg: rearm the CKC timer during migration

It is probably best to make tcg_s390_tod_updated() not go via the
helper. So we avoid locking issues.

I'll send a fixup.

Thanks for testing.

-- 

Thanks,

David / dhildenb