RE: [RFC v1] arm/virt: Add memory hot remove support

[Shameerali Kolothum Thodi <shameerali.kolothum.thodi@huawei.com>]

Hi Eric,

> -----Original Message-----
> From: Auger Eric [mailto:eric.auger@redhat.com]
> Sent: 26 March 2020 11:01
> To: Shameerali Kolothum Thodi <shameerali.kolothum.thodi@huawei.com>;
> qemu-devel@nongnu.org; qemu-arm@nongnu.org
> Cc: imammedo@redhat.com; peter.maydell@linaro.org; mst@redhat.com;
> xuwei (O) <xuwei5@huawei.com>; Zengtao (B) <prime.zeng@hisilicon.com>;
> Linuxarm <linuxarm@huawei.com>; Anshuman Khandual
> <anshuman.khandual@arm.com>
> Subject: Re: [RFC v1] arm/virt: Add memory hot remove support
> 
> Hi Shameer,
> 
> On 3/18/20 1:37 PM, Shameer Kolothum wrote:
> > This adds support for memory hot remove on arm/virt that
> > uses acpi ged device.
> 
> I gave this a try and it works fine if the PCDIMM slot was initially
> hotplugged:
> (QEMU) object-add qom-type=memory-backend-ram id=mem1
> props.size=4294967296
> {"return": {}}
> (QEMU) device_add driver=pc-dimm  id=pcdimm1 memdev=mem1
> (QEMU) device_del id=pcdimm1
> {"return": {}}
> 
> on guest I can see:
> [   82.466321] Offlined Pages 262144
> [   82.541712] Offlined Pages 262144
> [   82.589236] Offlined Pages 262144
> [   82.969166] Offlined Pages 262144
> 
> However I noticed that if qemu is launched directly with
> 
> -m 16G,maxmem=32G,slots=2 \
> -object memory-backend-ram,id=mem1,size=4G \
> -device pc-dimm,memdev=mem1,id=dimm1,driver=pc-dimm -device
> 
> and then in the qmp shell:
> (QEMU) device_del id=dimm1
> 
> the hot-unplug fails in guest:
> 
> [   78.897407] Offlined Pages 262144
> [   79.260811] Offlined Pages 262144
> [   79.308105] Offlined Pages 262144
> [   79.333675] page:fffffe00137d1f40 refcount:1 mapcount:0
> mapping:ffff0004ea9f20b1 index:0xaaab11c6e
> [   79.335927] anon flags: 0x17ffff8000080024(uptodate|active|swapbacked)
> [   79.337571] raw: 17ffff8000080024 dead000000000100
> dead000000000122
> ffff0004ea9f20b1
> [   79.339502] raw: 0000000aaab11c6e 0000000000000000 00000001ffffffff
> ffff0004fd4e3000
> [   79.341701] page dumped because: unmovable page
> [   79.342887] page->mem_cgroup:ffff0004fd4e3000
> [   79.354729] page:fffffe00137d1f40 refcount:1 mapcount:0
> mapping:ffff0004ea9f20b1 index:0xaaab11c6e
> [   79.357012] anon flags: 0x17ffff8000080024(uptodate|active|swapbacked)
> [   79.358658] raw: 17ffff8000080024 dead000000000100
> dead000000000122
> ffff0004ea9f20b1
> [   79.360611] raw: 0000000aaab11c6e 0000000000000000 00000001ffffffff
> ffff0004fd4e3000
> [   79.362560] page dumped because: unmovable page
> [   79.363742] page->mem_cgroup:ffff0004fd4e3000
> [   79.368636] memory memory20: Offline failed.
> 
> I did not expect this. The PCDIMM slot in that case does not seem to be
> interpreted as a hot-unpluggable one (?). I added Anshuman in cc.

Could you please try adding "movable_node" to qemu guest kernel command line params.
This will prevent any kernel allocation from hotplugable memory nodes which I think is
causing the behavior you are seeing.

Thanks,
Shameer


> Thanks
> 
> Eric
> 
> 
> 
> >
> > Signed-off-by: Shameer Kolothum <shameerali.kolothum.thodi@huawei.com>
> > ---
> >  -RFC because linux kernel support for mem hot remove is just queued
> >   for 5.7[1].
> >  -Tested with guest kernel 5.6-rc5 + [1]
> >
> > 1. https://patchwork.kernel.org/cover/11419301/
> > ---
> >  hw/acpi/generic_event_device.c | 28 +++++++++++++++++
> >  hw/arm/virt.c                  | 56
> ++++++++++++++++++++++++++++++++--
> >  2 files changed, 82 insertions(+), 2 deletions(-)
> >
> > diff --git a/hw/acpi/generic_event_device.c
> b/hw/acpi/generic_event_device.c
> > index 021ed2bf23..3e28c110fa 100644
> > --- a/hw/acpi/generic_event_device.c
> > +++ b/hw/acpi/generic_event_device.c
> > @@ -182,6 +182,32 @@ static void
> acpi_ged_device_plug_cb(HotplugHandler *hotplug_dev,
> >      }
> >  }
> >
> > +static void acpi_ged_unplug_request_cb(HotplugHandler *hotplug_dev,
> > +                                       DeviceState *dev, Error
> **errp)
> > +{
> > +    AcpiGedState *s = ACPI_GED(hotplug_dev);
> > +
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
> > +        acpi_memory_unplug_request_cb(hotplug_dev,
> &s->memhp_state, dev, errp);
> > +    } else {
> > +        error_setg(errp, "acpi: device unplug request for unsupported
> device"
> > +                   " type: %s", object_get_typename(OBJECT(dev)));
> > +    }
> > +}
> > +
> > +static void acpi_ged_unplug_cb(HotplugHandler *hotplug_dev,
> > +                               DeviceState *dev, Error **errp)
> > +{
> > +    AcpiGedState *s = ACPI_GED(hotplug_dev);
> > +
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
> > +        acpi_memory_unplug_cb(&s->memhp_state, dev, errp);
> > +    } else {
> > +        error_setg(errp, "acpi: device unplug for unsupported device"
> > +                   " type: %s", object_get_typename(OBJECT(dev)));
> > +    }
> > +}
> > +
> >  static void acpi_ged_send_event(AcpiDeviceIf *adev, AcpiEventStatusBits
> ev)
> >  {
> >      AcpiGedState *s = ACPI_GED(adev);
> > @@ -286,6 +312,8 @@ static void acpi_ged_class_init(ObjectClass *class,
> void *data)
> >      dc->vmsd = &vmstate_acpi_ged;
> >
> >      hc->plug = acpi_ged_device_plug_cb;
> > +    hc->unplug_request = acpi_ged_unplug_request_cb;
> > +    hc->unplug = acpi_ged_unplug_cb;
> >
> >      adevc->send_event = acpi_ged_send_event;
> >  }
> > diff --git a/hw/arm/virt.c b/hw/arm/virt.c
> > index 94f93dda54..91974e4e80 100644
> > --- a/hw/arm/virt.c
> > +++ b/hw/arm/virt.c
> > @@ -2096,11 +2096,62 @@ static void
> virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
> >      }
> >  }
> >
> > +static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
> > +                                     DeviceState *dev, Error
> **errp)
> > +{
> > +    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
> > +    Error *local_err = NULL;
> > +
> > +    if (!vms->acpi_dev) {
> > +        error_setg(errp,
> > +                   "memory hotplug is not enabled: missing acpi-ged
> device");
> > +        goto out;
> > +    }
> > +
> > +    hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev),
> dev,
> > +                                   &local_err);
> > +out:
> > +    error_propagate(errp, local_err);
> > +}
> > +
> > +static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
> > +                             DeviceState *dev, Error **errp)
> > +{
> > +    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
> > +    Error *local_err = NULL;
> > +
> > +    hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev,
> &local_err);
> > +    if (local_err) {
> > +        goto out;
> > +    }
> > +
> > +    pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
> > +    object_property_set_bool(OBJECT(dev), false, "realized", NULL);
> > +
> > + out:
> > +    error_propagate(errp, local_err);
> > +}
> > +
> >  static void virt_machine_device_unplug_request_cb(HotplugHandler
> *hotplug_dev,
> >                                            DeviceState *dev, Error
> **errp)
> >  {
> > -    error_setg(errp, "device unplug request for unsupported device"
> > -               " type: %s", object_get_typename(OBJECT(dev)));
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
> > +        virt_dimm_unplug_request(hotplug_dev, dev, errp);
> > +    } else {
> > +        error_setg(errp, "device unplug request for unsupported device"
> > +                   " type: %s", object_get_typename(OBJECT(dev)));
> > +    }
> > +}
> > +
> > +static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
> > +                                          DeviceState *dev, Error
> **errp)
> > +{
> > +    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
> > +        virt_dimm_unplug(hotplug_dev, dev, errp);
> > +    } else {
> > +        error_setg(errp, "virt: device unplug for unsupported device"
> > +                   " type: %s", object_get_typename(OBJECT(dev)));
> > +    }
> >  }
> >
> >  static HotplugHandler *virt_machine_get_hotplug_handler(MachineState
> *machine,
> > @@ -2181,6 +2232,7 @@ static void virt_machine_class_init(ObjectClass
> *oc, void *data)
> >      hc->pre_plug = virt_machine_device_pre_plug_cb;
> >      hc->plug = virt_machine_device_plug_cb;
> >      hc->unplug_request = virt_machine_device_unplug_request_cb;
> > +    hc->unplug = virt_machine_device_unplug_cb;
> >      mc->numa_mem_supported = true;
> >      mc->auto_enable_numa_with_memhp = true;
> >      mc->default_ram_id = "mach-virt.ram";
> >