On 05/03/2026 13.10, Igor Mammedov wrote:
On Wed, 4 Mar 2026 18:20:27 +0100
Thomas Huth <[email protected]> wrote:

On 03/03/2026 15.00, Igor Mammedov wrote:
From: Philippe Mathieu-Daudé <[email protected]>

v2:
     - do not check for SMI features if hotplug happens when
       SMI is not enabled. (matters for qtest and possibly seabios)
     - removing property also removes default ICH9_LPC_SMI_F_BROADCAST_BIT
       put default back in place only set it initfn() instead

The ICH9_LPC_SMI_F_BROADCAST_BIT feature bit was only set
in the pc_compat_2_8[] array, via the 'x-smi-broadcast=off'
property. We removed all machines using that array, lets remove
that property and all the code around it.

Signed-off-by: Philippe Mathieu-Daudé <[email protected]>
Signed-off-by: Igor Mammedov <[email protected]>
---
   hw/acpi/ich9.c    |  4 ++--
   hw/isa/lpc_ich9.c | 23 ++++-------------------
   2 files changed, 6 insertions(+), 21 deletions(-)

diff --git a/hw/acpi/ich9.c b/hw/acpi/ich9.c
index bbb1bd60a2..87afe86bcc 100644
--- a/hw/acpi/ich9.c
+++ b/hw/acpi/ich9.c
@@ -432,7 +432,7 @@ void ich9_pm_device_pre_plug_cb(HotplugHandler 
*hotplug_dev, DeviceState *dev,
       if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
           uint64_t negotiated = lpc->smi_negotiated_features;
- if (negotiated & BIT_ULL(ICH9_LPC_SMI_F_BROADCAST_BIT) &&
+        if (lpc->pm.smi_en & ICH9_PMIO_SMI_EN_APMC_EN &&
               !(negotiated & BIT_ULL(ICH9_LPC_SMI_F_CPU_HOTPLUG_BIT))) {
               error_setg(errp, "cpu hotplug with SMI wasn't enabled by 
firmware");
               error_append_hint(errp, "update machine type to newer than 5.1 "
@@ -476,7 +476,7 @@ void ich9_pm_device_unplug_request_cb(HotplugHandler 
*hotplug_dev,
       } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
           uint64_t negotiated = lpc->smi_negotiated_features;
- if (negotiated & BIT_ULL(ICH9_LPC_SMI_F_BROADCAST_BIT) &&
+        if (lpc->pm.smi_en & ICH9_PMIO_SMI_EN_APMC_EN &&
               !(negotiated & BIT_ULL(ICH9_LPC_SMI_F_CPU_HOT_UNPLUG_BIT))) {
               error_setg(errp, "cpu hot-unplug with SMI wasn't enabled "
                                "by firmware");
diff --git a/hw/isa/lpc_ich9.c b/hw/isa/lpc_ich9.c
index 51dc680029..04169ffa24 100644
--- a/hw/isa/lpc_ich9.c
+++ b/hw/isa/lpc_ich9.c
@@ -404,15 +404,6 @@ static void smi_features_ok_callback(void *opaque)
       guest_cpu_hotplug_features = guest_features &
                                    (BIT_ULL(ICH9_LPC_SMI_F_CPU_HOTPLUG_BIT) |
                                     
BIT_ULL(ICH9_LPC_SMI_F_CPU_HOT_UNPLUG_BIT));
-    if (!(guest_features & BIT_ULL(ICH9_LPC_SMI_F_BROADCAST_BIT)) &&
-        guest_cpu_hotplug_features) {
-        /*
-         * cpu hot-[un]plug with SMI requires SMI broadcast,
-         * leave @features_ok at zero
-         */
-        return;
-    }
-
       if (guest_cpu_hotplug_features ==
           BIT_ULL(ICH9_LPC_SMI_F_CPU_HOT_UNPLUG_BIT)) {
           /* cpu hot-unplug is unsupported without cpu-hotplug */
@@ -474,14 +465,9 @@ static void ich9_apm_ctrl_changed(uint32_t val, void *arg)
/* SMI_EN = PMBASE + 30. SMI control and enable register */
       if (lpc->pm.smi_en & ICH9_PMIO_SMI_EN_APMC_EN) {
-        if (lpc->smi_negotiated_features &
-            (UINT64_C(1) << ICH9_LPC_SMI_F_BROADCAST_BIT)) {
-            CPUState *cs;
-            CPU_FOREACH(cs) {
-                cpu_interrupt(cs, CPU_INTERRUPT_SMI);
-            }
-        } else {
-            cpu_interrupt(current_cpu, CPU_INTERRUPT_SMI);
+        CPUState *cs;
+        CPU_FOREACH(cs) {
+            cpu_interrupt(cs, CPU_INTERRUPT_SMI);
           }
       }
   }
@@ -685,6 +671,7 @@ static void ich9_lpc_initfn(Object *obj)
static const uint8_t acpi_enable_cmd = ICH9_APM_ACPI_ENABLE;
       static const uint8_t acpi_disable_cmd = ICH9_APM_ACPI_DISABLE;
+    lpc->smi_host_features = BIT_ULL(ICH9_LPC_SMI_F_BROADCAST_BIT);

Shouldn't that rather be a "|=" instead of the "=" so that we don't lose the
other bits?

Indeed it should be '|=',
can you fix it up on merge or should I better send fixed up version?

I tried to fix it, but I still see an issue with the bios-tables-test qtest during "make check" - it does not finish anymore and is finally killed by the timeout. Does "make check-qtest" work for you?

 Thomas


Reply via email to