Skip to contentSkip to conversation

hw/intc/xics: Add missing call to register vmstate_icp_server

PATCH In review Fabian Vogt 5 participants 5 messages Aug 14, 2025 – Aug 19, 2025

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hw/intc/xics: Add missing call to register vmstate_icp_server

An obsolete wrapper function with a workaround was removed entirely, without restoring the call it wrapped.

Without this, the guest is stuck after savevm/loadvm.

Fixes 24ee9229fe31 ("ppc/spapr: remove deprecated machine pseries-2.9")
---
 hw/intc/xics.c | 2 ++
 1 file changed, 2 insertions(+)
hw/intc/xics.c +2 -0
index d9a199e883..200710eb6c 100644
@@ -335,6 +335,8 @@ static void icp_realize(DeviceState *dev, Error **errp)
             return;
         }
     }
+
+    vmstate_register(NULL, icp->cs->cpu_index, &vmstate_icp_server, icp);
 }
 
 static void icp_unrealize(DeviceState *dev)
Signature · 1 line
2.50.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 16 lines · Fabian Vogt

An obsolete wrapper function with a workaround was removed entirely, without restoring the call it wrapped.

Without this, the guest is stuck after savevm/loadvm.

Fixes 24ee9229fe31 ("ppc/spapr: remove deprecated machine pseries-2.9")
---
  hw/intc/xics.c | 2 ++
  1 file changed, 2 insertions(+)
hw/intc/xics.c +0 -0
index d9a199e883..200710eb6c 100644

hw/intc/xics.c static void icp_realize(DeviceState *dev, Error **errp) · line 342

Quoted 8 lines
hw/intc/xics.c +2 -0
@@ -335,6 +335,8 @@ static void icp_realize(DeviceState *dev, Error **errp)
              return;
          }
      }
+
+    vmstate_register(NULL, icp->cs->cpu_index, &vmstate_icp_server, icp);
  }
 
  static void icp_unrealize(DeviceState *dev)
Acked-by on v1 1/1
#2 lore raw
In reply to Fabian Vogt

Fabian Vogt <[email protected]> writes:

+CC Peter

Quoted 5 lines

An obsolete wrapper function with a workaround was removed entirely, without restoring the call it wrapped.

Without this, the guest is stuck after savevm/loadvm.

I'd note that this only affects guests that have actually negotiated XICS, so guests using ic-mode=xive are safe.

With this patch, backward migration to QEMU versions >= 9.2 is broken. Which I can only assume is ok, since the hard break (vmstate gets rejected) is better than weird post-migration hangs.

If the PPC folks can spot a scenario where migration would provide reasonable results even without the emulated interrupt controller state being migrated, then we'll need a compat property for this.

Also, a 9.2 guest cannot move to a more recent QEMU version without a reboot. If this is an important scenario, then the new QEMU needs to provide some compatibility knob.

Otherwise:

#3 lore raw

+ Gautam, Glenn

Hi Fabian,

Quoted 4 lines · Fabiano Rosas

Fabian Vogt <[email protected]> writes:

+CC Peter

Quoted 8 lines
Quoted 5 lines

An obsolete wrapper function with a workaround was removed entirely, without restoring the call it wrapped.

Without this, the guest is stuck after savevm/loadvm.

I'd note that this only affects guests that have actually negotiated XICS, so guests using ic-mode=xive are safe.

Since this affected only guests using XICS, would it be a better idea to mention this explicitly in commit log as well (?

Quoted 5 lines

With this patch, backward migration to QEMU versions >= 9.2 is broken. Which I can only assume is ok, since the hard break (vmstate gets rejected) is better than weird post-migration hangs.

Quoted 7 lines

If the PPC folks can spot a scenario where migration would provide reasonable results even without the emulated interrupt controller state being migrated, then we'll need a compat property for this.

Also, a 9.2 guest cannot move to a more recent QEMU version without a reboot. If this is an important scenario, then the new QEMU needs to provide some compatibility knob.

Gautam, Glenn - thoughts on above comments?

Thanks for the patch.

regards,
Harsh

Quoted 5 lines

Otherwise:

#4 lore raw
Quoted 13 lines · Harsh Prateek Bora

+ Gautam, Glenn

Hi Fabian,

Quoted 7 lines · Fabiano Rosas

Fabian Vogt <[email protected]> writes:

+CC Peter

Quoted 3 lines

An obsolete wrapper function with a workaround was removed entirely, without restoring the call it wrapped.

Quoted 17 lines
Quoted 5 lines
Quoted 2 lines

Without this, the guest is stuck after savevm/loadvm.

I'd note that this only affects guests that have actually negotiated XICS, so guests using ic-mode=xive are safe.

Since this affected only guests using XICS, would it be a better idea to mention this explicitly in commit log as well (?

Quoted 8 lines

With this patch, backward migration to QEMU versions >= 9.2 is broken. Which I can only assume is ok, since the hard break (vmstate gets rejected) is better than weird post-migration hangs.

If the PPC folks can spot a scenario where migration would provide reasonable results even without the emulated interrupt controller state being migrated, then we'll need a compat property for this.

I don't know much about the XICS controller, but migration without restoring the interrupt controller state seems pretty risky to me. I wouldn't recommend it.

Glenn

Quoted 15 lines
Quoted 3 lines

Also, a 9.2 guest cannot move to a more recent QEMU version without a reboot. If this is an important scenario, then the new QEMU needs to provide some compatibility knob.

Gautam, Glenn - thoughts on above comments?

Thanks for the patch.

regards,
Harsh

Quoted 4 lines

Otherwise: