Re: [Qemu-ppc] [RFC PATCH v0 3/3] spapr: Memory hot-unplug support

[Michael Roth]

Quoting Bharata B Rao (2015-08-26 04:57:51)
> On Mon, Aug 24, 2015 at 09:39:31PM -0500, Michael Roth wrote:
> > Quoting Bharata B Rao (2015-08-19 01:56:11)
> > > Add support to hot remove pc-dimm memory devices.
> > > 
> > > Signed-off-by: Bharata B Rao <address@hidden>
> > > ---
> > >  hw/ppc/spapr.c         | 114 
> > > ++++++++++++++++++++++++++++++++++++++++++++++++-
> > >  hw/ppc/spapr_drc.c     |  21 +++++++++
> > >  include/hw/ppc/spapr.h |   2 +
> > >  3 files changed, 136 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > > index 06d000d..441012d 100644
> > > --- a/hw/ppc/spapr.c
> > > +++ b/hw/ppc/spapr.c
> > > @@ -2110,6 +2110,109 @@ out:
> > >      error_propagate(errp, local_err);
> > >  }
> > > 
> > > +typedef struct sPAPRDIMMState {
> > > +    uint32_t nr_lmbs;
> > > +} sPAPRDIMMState;
> > > +
> > > +/*
> > > + * Called from spapr_drc.c: set_isolation_state().
> > > + *
> > > + * If the drc is being marked as ISOLATED, ensure that the corresponding
> > > + * LMB is part of the DIMM device which is being deleted.
> > > + */
> > > +int spapr_lmb_in_removable_dimm(sPAPRDRConnector *drc,
> > > +                                sPAPRDRIsolationState state)
> > > +{
> > > +    DeviceState *dev = drc->dev;
> > > +    PCDIMMDevice *dimm = PC_DIMM(dev);
> > > +
> > > +    if (state != SPAPR_DR_ISOLATION_STATE_ISOLATED) {
> > > +        return 0;
> > > +    }
> > > +
> > > +    if (!dimm->delete_pending) {
> > > +        return -1;
> > > +    }
> > > +
> > > +    return 0;
> > > +}
> > > +
> > > +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));
> > > +}
> > 
> > In the current code the unplug() and request_unplug() are mutually
> > exclusive. Are the plans on making the unplug() do something in the
> > prescence of request_unplug()? If so, I'd imagine it would've be a
> > forced removal, except maybe as a fallback if the request is determined
> > to fail somehow?
> 
> Like x86 memory hotremoval, our model too fits into async type of removal
> where we first send removal notification to guest in ->unplug_request() and
> when the guest indeed removes the memory, we cleanup the pc-dimm device
> in ->unplug().
> 
> Since we implement both ->unplug() and ->unplug_request(), and given that
> the removal works like above, I don't see why we would ever end up doing a
> forced removal from ->unplug().

Ah, sorry, misunderstanding on my part: as far as *qdev* lifecycle is
concerned, unplug() never gets called directly if unplug_request() is
defined, so it seemed the 2 were mutually exclusive. But in the case
of memory x86 unplug, unplug_request()'s implementation is such that
the unplug() handler gets called explicitly via guest acknowledgement's
callback, which is the same as what you've modelled here.

It's a little wierd that the unplug() callback is needlessly exposed
outside of the async unplug_request() implementation, but it does
allow for future code where maybe a --force option could be introduced
for certain situations where a guest isn't cooperative.

> 
> > > diff --git a/hw/ppc/spapr_drc.c b/hw/ppc/spapr_drc.c
> > > index 8cbcf4d..b9d7c71 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"
> > > @@ -63,9 +64,29 @@ static int set_isolation_state(sPAPRDRConnector *drc,
> > >                                 sPAPRDRIsolationState state)
> > >  {
> > >      sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
> > > +    int ret;
> > > 
> > >      DPRINTFN("drc: %x, set_isolation_state: %x", get_index(drc), state);
> > > 
> > > +    /*
> > > +     * Fail any requests 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.
> > > +     *
> > > +     * TODO: Calling out from spapr_drc.c like this doesn't look good.
> > > +     */
> > > +    if (drc->type == SPAPR_DR_CONNECTOR_TYPE_LMB) {
> > > +        ret = spapr_lmb_in_removable_dimm(drc, state);
> > > +        if (ret) {
> > > +            return RTAS_OUT_HW_ERROR;
> > > +        }
> > > +    }
> 
> I am not sure if this call out is the right way to do this. Do you have
> suggestions here ?

Since drc->awaiting_release is only set in cases where the corresponding
dimm is pending delete, I think maybe we can rely on
drc->awaiting_release instead and avoid the call out.

> 
> Regards,
> Bharata.
>