[Qemu-devel] [RFC PATCH v1] spapr: Memory hot-unplug support

[Bharata B Rao]

Add support to hot remove pc-dimm memory devices.

TODO: In response to memory hot removal operation on a DIMM device,
guest kernel might refuse to offline a few LMBs that are part of that device.
In such cases, we will have a DIMM device that has some LMBs online and some
LMBs offline. To avoid this situation, drmgr could be enhanced to support
a command line option that results in removal of all the LMBs or none.

Signed-off-by: Bharata B Rao <address@hidden>
---
Changes in v1:
- Got rid of the patch that introduced a field in PCDIMMDevice to track
  DIMM marked for removal since we can track that using within DRC
  object.
- Removed the patch that added return value to rtas_set_indicator()
  since the required changes are already pushed by Michael Roth.

v0:

 hw/ppc/spapr.c     | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++++-
 hw/ppc/spapr_drc.c | 18 +++++++++++
 2 files changed, 107 insertions(+), 1 deletion(-)

diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index e1202ce..f5b1ac2 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -2174,6 +2174,85 @@ out:
     error_propagate(errp, local_err);
 }
 
+typedef struct sPAPRDIMMState {
+    uint32_t nr_lmbs;
+} sPAPRDIMMState;
+
+static void spapr_lmb_release(DeviceState *dev, void *opaque)
+{
+    sPAPRDIMMState *ds = (sPAPRDIMMState *)opaque;
+    HotplugHandler *hotplug_ctrl = NULL;
+    Error *local_err = NULL;
+
+    if (--ds->nr_lmbs) {
+        return;
+    }
+
+    g_free(ds);
+
+    /*
+     * Now that all the LMBs have been removed by the guest, call the
+     * pc-dimm unplug handler to cleanup up the pc-dimm device.
+     */
+    hotplug_ctrl = qdev_get_hotplug_handler(dev);
+    hotplug_handler_unplug(hotplug_ctrl, dev, &local_err);
+}
+
+static void spapr_del_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
+                           Error **errp)
+{
+    sPAPRDRConnector *drc;
+    sPAPRDRConnectorClass *drck;
+    uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
+    Error *local_err = NULL;
+    int i;
+    sPAPRDIMMState *ds = g_malloc0(sizeof(sPAPRDIMMState));
+
+    ds->nr_lmbs = nr_lmbs;
+    for (i = 0; i < nr_lmbs; i++) {
+        drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
+                addr/SPAPR_MEMORY_BLOCK_SIZE);
+        g_assert(drc);
+
+        drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
+        drck->detach(drc, dev, spapr_lmb_release, ds, &local_err);
+        addr += SPAPR_MEMORY_BLOCK_SIZE;
+    }
+    spapr_hotplug_req_remove_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
+}
+
+static void spapr_memory_unplug(HotplugHandler *hotplug_dev, DeviceState *dev,
+                                Error **errp)
+{
+    sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
+    PCDIMMDevice *dimm = PC_DIMM(dev);
+    PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
+    MemoryRegion *mr = ddc->get_memory_region(dimm);
+
+    pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
+    object_unparent(OBJECT(dev));
+}
+
+static void spapr_memory_unplug_request(HotplugHandler *hotplug_dev,
+                                        DeviceState *dev, Error **errp)
+{
+    Error *local_err = NULL;
+    PCDIMMDevice *dimm = PC_DIMM(dev);
+    PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
+    MemoryRegion *mr = ddc->get_memory_region(dimm);
+    uint64_t size = memory_region_size(mr);
+    uint64_t addr;
+
+    addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, 
&local_err);
+    if (local_err) {
+        goto out;
+    }
+
+    spapr_del_lmbs(dev, addr, size, &local_err);
+out:
+    error_propagate(errp, local_err);
+}
+
 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
                                       DeviceState *dev, Error **errp)
 {
@@ -2221,7 +2300,15 @@ static void spapr_machine_device_unplug(HotplugHandler 
*hotplug_dev,
                                       DeviceState *dev, Error **errp)
 {
     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
-        error_setg(errp, "Memory hot unplug not supported by sPAPR");
+        spapr_memory_unplug(hotplug_dev, dev, errp);
+    }
+}
+
+static void spapr_machine_device_unplug_request(HotplugHandler *hotplug_dev,
+                                                DeviceState *dev, Error **errp)
+{
+    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
+        spapr_memory_unplug_request(hotplug_dev, dev, errp);
     }
 }
 
@@ -2263,6 +2350,7 @@ static void spapr_machine_class_init(ObjectClass *oc, 
void *data)
     hc->plug = spapr_machine_device_plug;
     hc->unplug = spapr_machine_device_unplug;
     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
+    hc->unplug_request = spapr_machine_device_unplug_request;
 
     smc->dr_lmb_enabled = false;
     fwc->get_dev_path = spapr_get_fw_dev_path;
diff --git a/hw/ppc/spapr_drc.c b/hw/ppc/spapr_drc.c
index 5d6ea7c..59b6ea9 100644
--- a/hw/ppc/spapr_drc.c
+++ b/hw/ppc/spapr_drc.c
@@ -11,6 +11,7 @@
  */
 
 #include "hw/ppc/spapr_drc.h"
+#include "hw/ppc/spapr.h"
 #include "qom/object.h"
 #include "hw/qdev.h"
 #include "qapi/visitor.h"
@@ -77,6 +78,23 @@ static uint32_t set_isolation_state(sPAPRDRConnector *drc,
         }
     }
 
+    /*
+     * Fail any request to ISOLATE the LMB DRC if this LMB doesn't
+     * belong to a DIMM device that is marked for removal.
+     *
+     * Currently the guest userspace tool drmgr that drives the memory
+     * hotplug/unplug will just try to remove a set of 'removable' LMBs
+     * in response to a hot unplug request that is based on drc-count.
+     * If the LMB being removed doesn't belong to a DIMM device that is
+     * actually being unplugged, fail the isolation request here.
+     */
+    if (drc->type == SPAPR_DR_CONNECTOR_TYPE_LMB) {
+        if ((state == SPAPR_DR_ISOLATION_STATE_ISOLATED) &&
+             !drc->awaiting_release) {
+            return RTAS_OUT_HW_ERROR;
+        }
+    }
+
     drc->isolation_state = state;
 
     if (drc->isolation_state == SPAPR_DR_ISOLATION_STATE_ISOLATED) {
-- 
2.1.0