Re: [PATCH v2] ppc/spapr: Add support for implement support for H_SCM_HEALTH

[Vaibhav Jain <vaibhav@linux.ibm.com>]

Hi Greg,

Thanks for looking into this patch.

Greg Kurz <groug@kaod.org> writes:

> On Thu, 1 Apr 2021 13:26:11 +1100
> David Gibson <david@gibson.dropbear.id.au> wrote:
>
>> On Thu, Apr 01, 2021 at 06:35:19AM +0530, Vaibhav Jain wrote:
>> > Add support for H_SCM_HEALTH hcall described at [1] for spapr
>> > nvdimms. This enables guest to detect the 'unarmed' status of a
>> > specific spapr nvdimm identified by its DRC and if its unarmed, mark
>> > the region backed by the nvdimm as read-only.
>> > 
>> > The patch adds h_scm_health() to handle the H_SCM_HEALTH hcall which
>> > returns two 64-bit bitmaps (health bitmap, health bitmap mask) derived
>> > from 'struct nvdimm->unarmed' member.
>> > 
>> > Linux kernel side changes to enable handling of 'unarmed' nvdimms for
>> > ppc64 are proposed at [2].
>> > 
>> > References:
>> > [1] "Hypercall Op-codes (hcalls)"
>> >     https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/powerpc/papr_hcalls.rst#n220
>> > [2] "powerpc/papr_scm: Mark nvdimm as unarmed if needed during probe"
>> >     https://lore.kernel.org/linux-nvdimm/20210329113103.476760-1-vaibhav@linux.ibm.com/
>> > 
>> > Signed-off-by: Vaibhav Jain <vaibhav@linux.ibm.com>
>> 
>> As well as the handful of comments below, this will definitely need to
>> wait for ppc-6.1 at this point.
>> 
>> > ---
>> > Changelog
>> > 
>> > v2:
>> > * Added a check for drc->dev to ensure that the dimm is plugged in
>> >   when servicing H_SCM_HEALTH. [ Shiva ]
>> > * Instead of accessing the 'nvdimm->unarmed' member directly use the
>> >   object_property_get_bool accessor to fetch it. [ Shiva ]
>> > * Update the usage of PAPR_PMEM_UNARMED* macros [ Greg ]
>> > * Updated patch description reference#1 to point appropriate section
>> >   in the documentation. [ Greg ]
>> > ---
>> >  hw/ppc/spapr_nvdimm.c  | 38 ++++++++++++++++++++++++++++++++++++++
>> >  include/hw/ppc/spapr.h |  3 ++-
>> >  2 files changed, 40 insertions(+), 1 deletion(-)
>> > 
>> > diff --git a/hw/ppc/spapr_nvdimm.c b/hw/ppc/spapr_nvdimm.c
>> > index b46c36917c..34096e4718 100644
>> > --- a/hw/ppc/spapr_nvdimm.c
>> > +++ b/hw/ppc/spapr_nvdimm.c
>> > @@ -31,6 +31,13 @@
>> >  #include "qemu/range.h"
>> >  #include "hw/ppc/spapr_numa.h"
>> >  
>> > +/* DIMM health bitmap bitmap indicators. Taken from kernel's papr_scm.c */
>> > +/* SCM device is unable to persist memory contents */
>> > +#define PAPR_PMEM_UNARMED (1ULL << (63 - 0))
>> 
>> You can use PPC_BIT() for more clarity here.
>> 
>
> I had already suggested PPC_BIT(0) but since this macro was copied
> from the kernel source, I've let Vaibhav decide whether to use
> PPC_BIT() or keep the macro and comment it comes from the kernel.
>
> I agree I prefer PPC_BIT(0) :-)
>
Have switched to PPC_BIT in v3
>> > +/* Bits status indicators for health bitmap indicating unarmed dimm */
>> > +#define PAPR_PMEM_UNARMED_MASK (PAPR_PMEM_UNARMED)
>> 
>> I'm not sure why you want two equal #defines here.
>> 
>
> Especially, this define doesn't make sense for the hypervisor IMHO.
>
> It is _just_ the mask of bits for the "unarmed" state in the kernel.
>
Have gotten rid of this define in v3. We can revisit this In future when
support for more bits is added.

>> > +
>> >  bool spapr_nvdimm_validate(HotplugHandler *hotplug_dev, NVDIMMDevice *nvdimm,
>> >                             uint64_t size, Error **errp)
>> >  {
>> > @@ -467,6 +474,36 @@ static target_ulong h_scm_unbind_all(PowerPCCPU *cpu, SpaprMachineState *spapr,
>> >      return H_SUCCESS;
>> >  }
>> >  
>> > +static target_ulong h_scm_health(PowerPCCPU *cpu, SpaprMachineState *spapr,
>> > +                                 target_ulong opcode, target_ulong *args)
>> > +{
>> > +    uint32_t drc_index = args[0];
>> > +    SpaprDrc *drc = spapr_drc_by_index(drc_index);
>> > +    NVDIMMDevice *nvdimm;
>> > +
>> > +    if (drc && spapr_drc_type(drc) != SPAPR_DR_CONNECTOR_TYPE_PMEM) {
>> 
>> This will fail badly if !drc (given index is way out of bounds).  I'm
>> pretty sure you want
>> 	if (!drc || spapr_drc_type(drc) != SPAPR_DR_CONNECTOR_TYPE_PMEM) {
>> 
>> 
>> > +        return H_PARAMETER;
>> > +    }
>> > +
>> > +    /* Ensure that the dimm is plugged in */
>> > +    if (!drc->dev) {
>> > +        return H_HARDWARE;
>> 
>> H_HARDWARE doesn't seem right - it's the guest that has chosen to
>> attempt this on an unplugged LMB, not the (virtual) hardware's fault.
>> 
>
> Yes. As already suggested, simply do the same as in other hcall
> implementations in this file, e.g. from h_scm_bind_mem() :
>
>     if (!drc || !drc->dev ||
>         spapr_drc_type(drc) != SPAPR_DR_CONNECTOR_TYPE_PMEM) {
>         return H_PARAMETER;
>     }
>
Yes,  have used the same construct in the fixed v3

>> > +    }
>> > +
>> > +    nvdimm = NVDIMM(drc->dev);
>> > +
>> > +    args[0] = 0;
>> > +    /* Check if the nvdimm is unarmed and send its status via health bitmaps */
>
> Not sure this comment is super useful.
>
Have reworked this function in v3.

>> > +    if (object_property_get_bool(OBJECT(nvdimm), NVDIMM_UNARMED_PROP, NULL)) {
>> > +        args[0] |= PAPR_PMEM_UNARMED;
>> > +    }
>> > +
>> > +    /* Update the health bitmap with the applicable mask */
>> > +    args[1] = PAPR_PMEM_UNARMED_MASK;
>
> I still think this is a misuse of PAPR_PMEM_UNARMED_MASK... The
> meaning of args[1] is "health-bit-valid-bitmap indicate which
> bits in health-bitmap are valid" according to the documentation.
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/powerpc/papr_hcalls.rst#n228
>
> Without any further detail, I tend to consider this as a hint
> to the guest on the bits supported by the hypervisor. Since
> we can only set PAPR_PMEM_UNARMED, for now, args[1] should
> be set to just that bit PAPR_PMEM_UNARMED. Other bits can
> be added later if QEMU supports more of them.
>
Agree and addressed in v3.

>> > +
>> > +    return H_SUCCESS;
>> > +}
>> > +
>> >  static void spapr_scm_register_types(void)
>> >  {
>> >      /* qemu/scm specific hcalls */
>> > @@ -475,6 +512,7 @@ static void spapr_scm_register_types(void)
>> >      spapr_register_hypercall(H_SCM_BIND_MEM, h_scm_bind_mem);
>> >      spapr_register_hypercall(H_SCM_UNBIND_MEM, h_scm_unbind_mem);
>> >      spapr_register_hypercall(H_SCM_UNBIND_ALL, h_scm_unbind_all);
>> > +    spapr_register_hypercall(H_SCM_HEALTH, h_scm_health);
>> >  }
>> >  
>> >  type_init(spapr_scm_register_types)
>> > diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
>> > index 47cebaf3ac..6e1eafb05d 100644
>> > --- a/include/hw/ppc/spapr.h
>> > +++ b/include/hw/ppc/spapr.h
>> > @@ -538,8 +538,9 @@ struct SpaprMachineState {
>> >  #define H_SCM_BIND_MEM          0x3EC
>> >  #define H_SCM_UNBIND_MEM        0x3F0
>> >  #define H_SCM_UNBIND_ALL        0x3FC
>> > +#define H_SCM_HEALTH            0x400
>> >  
>> > -#define MAX_HCALL_OPCODE        H_SCM_UNBIND_ALL
>> > +#define MAX_HCALL_OPCODE        H_SCM_HEALTH
>> >  
>> >  /* The hcalls above are standardized in PAPR and implemented by pHyp
>> >   * as well.
>> 
>

-- 
Cheers
~ Vaibhav