SYS-OSS/FRET-qemu
4
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

hw/s390x/s390-virtio-ccw: Fix a record/replay deadlock

Browse Source 
Booting an s390x VM in record/replay mode hangs due to a deadlock
between rr_cpu_thread_fn() and s390_machine_reset(). The former needs
the record/replay mutex held by the latter, and the latter waits until
the former completes its run_on_cpu() request.

Fix by temporarily dropping the record/replay mutex, like it's done in
pause_all_vcpus().

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20250124112625.23050-1-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
This commit is contained in:
Ilya Leoshkevich 2025-01-24 12:25:48 +01:00 committed by Thomas Huth
parent b497b0376c
commit a4cda3f5df
1 changed files with 23 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
hw/s390x/s390-virtio-ccw.c
View File

@ -48,6 +48,7 @@
#include "kvm/kvm_s390x.h" #include "kvm/kvm_s390x.h"
#include "hw/virtio/virtio-md-pci.h" #include "hw/virtio/virtio-md-pci.h"
#include "hw/s390x/virtio-ccw-md.h" #include "hw/s390x/virtio-ccw-md.h"
#include "system/replay.h"
#include CONFIG_DEVICES #include CONFIG_DEVICES
static Error *pv_mig_blocker; static Error *pv_mig_blocker;
@ -454,6 +455,18 @@ static void s390_machine_reset(MachineState *machine, ResetType type)
CPUState *cs, *t; CPUState *cs, *t;
S390CPU *cpu; S390CPU *cpu;
/*
* Temporarily drop the record/replay mutex to let rr_cpu_thread_fn()
* process the run_on_cpu() requests below. This is safe, because at this
* point one of the following is true:
* - All CPU threads are not running, either because the machine is being
* initialized, or because the guest requested a reset using diag 308.
* There is no risk to desync the record/replay state.
* - A snapshot is about to be loaded. The record/replay state consistency
* is not important.
*/
replay_mutex_unlock();
/* get the reset parameters, reset them once done */ /* get the reset parameters, reset them once done */
s390_ipl_get_reset_request(&cs, &reset_type); s390_ipl_get_reset_request(&cs, &reset_type);
@ -533,7 +546,7 @@ static void s390_machine_reset(MachineState *machine, ResetType type)
* went wrong. * went wrong.
*/ */
s390_cpu_set_state(S390_CPU_STATE_OPERATING, cpu); s390_cpu_set_state(S390_CPU_STATE_OPERATING, cpu);
return; goto out_lock;
} }
run_on_cpu(cs, s390_do_cpu_load_normal, RUN_ON_CPU_NULL); run_on_cpu(cs, s390_do_cpu_load_normal, RUN_ON_CPU_NULL);
@ -546,6 +559,15 @@ static void s390_machine_reset(MachineState *machine, ResetType type)
run_on_cpu(t, s390_do_cpu_set_diag318, RUN_ON_CPU_HOST_ULONG(0)); run_on_cpu(t, s390_do_cpu_set_diag318, RUN_ON_CPU_HOST_ULONG(0));
} }
s390_ipl_clear_reset_request(); s390_ipl_clear_reset_request();
out_lock:
/*
* Re-take the record/replay mutex, temporarily dropping the BQL in order
* to satisfy the ordering requirements.
*/
bql_unlock();
replay_mutex_lock();
bql_lock();
} }
static void s390_machine_device_pre_plug(HotplugHandler *hotplug_dev, static void s390_machine_device_pre_plug(HotplugHandler *hotplug_dev,