On Wed, Sep 16, 2020 at 04:32:43PM +0000, Wei Liu wrote:
> On Tue, Sep 15, 2020 at 12:27:16PM +0200, Vitaly Kuznetsov wrote:
> > Wei Liu <wei.liu@xxxxxxxxxx> writes:
> [...]
> > >
> > > +void __init hv_get_partition_id(void)
> > > +{
> > > + struct hv_get_partition_id *output_page;
> > > + int status;
> > > + unsigned long flags;
> > > +
> > > + local_irq_save(flags);
> > > + output_page = *this_cpu_ptr(hyperv_pcpu_output_arg);
> > > + status = hv_do_hypercall(HVCALL_GET_PARTITION_ID, NULL, output_page) &
> > > + HV_HYPERCALL_RESULT_MASK;
> >
> > Nit: in this case status is 'u16', we can define it as such (instead of
> > signed int).
>
> Fixed.
>
> >
> > > + if (status != HV_STATUS_SUCCESS)
> > > + pr_err("Failed to get partition ID: %d\n", status);
> > > + else
> > > + hv_current_partition_id = output_page->partition_id;
> > > + local_irq_restore(flags);
> > > +
> > > + /* No point in proceeding if this failed */
> > > + BUG_ON(status != HV_STATUS_SUCCESS);
> > > +}
> > > +
> > > /*
> > > * This function is to be invoked early in the boot sequence after the
> > > * hypervisor has been detected.
> > > @@ -440,6 +463,9 @@ void __init hyperv_init(void)
> > >
> > > register_syscore_ops(&hv_syscore_ops);
> > >
> > > + if (hv_root_partition)
> > > + hv_get_partition_id();
> >
> > According to TLFS, partition ID is available when AccessPartitionId
> > privilege is granted. I'd suggest we check that instead of
> > hv_root_partition (and we can set hv_current_partition_id to something
> > like U64_MAX so we know it wasn't acuired). So the BUG_ON condition will
> > move here:
> >
> > hv_get_partition_id();
> > BUG_ON(hv_root_partition && hv_current_partition_id == U64_MAX);
> >
>
> Good point. I will reorganize this a bit.
Actually, our current code never stashed the feature mask that contains
that privilege anywhere. Getting access to it will require a few more
extra patches -- I would really like to rename those fields (features,
misc_features) inside ms_hyperv to something more appropriate.
We will gate it wit hv_root_partition anyway, since we a child VM may
not have the privilege, making an unconditional BUG_ON fatal.
All in all, the current code is not too bad. I intend to keep the
current structure for my RFC v2. I will see if I can find some time to
rework the feature mask extraction code and get that upstreamed first.
Wei.
>
> Wei.
