Re: [PATCH 1/2] acpi/nvdimm: Create _LS{I,R,W} method for NVDIMM device

[Robert Hoo]

On Wed, 2022-04-27 at 16:34 +0200, Igor Mammedov wrote:
> On Tue, 12 Apr 2022 14:57:52 +0800
> Robert Hoo <robert.hu@linux.intel.com> wrote:
> 
> > Since ACPI 6.2, previous NVDIMM/_DSM funcions "Get Namespace Label
> > Data
> > Size (function index 4)", "Get Namespace Label Data (function index
> > 5)",
> > "Set Namespace Label Data (function index 6)" has been deprecated
> > by ACPI
> 
> where it's said that old way was deprecated, should be mentioned here
> including
> pointer to spec where it came into effect.

OK. https://pmem.io/documents/IntelOptanePMem_DSM_Interface-V2.0.pdf,
3.10 Deprecated Functions.
I put it in cover letter. Will also mention it here.
> 
...
> > 
> > diff --git a/hw/acpi/nvdimm.c b/hw/acpi/nvdimm.c
> > index 0d43da19ea..7cc419401b 100644
> > --- a/hw/acpi/nvdimm.c
> > +++ b/hw/acpi/nvdimm.c
> > @@ -848,10 +848,10 @@ nvdimm_dsm_write(void *opaque, hwaddr addr,
> > uint64_t val, unsigned size)
> >  
> >      nvdimm_debug("Revision 0x%x Handler 0x%x Function 0x%x.\n",
> > in->revision,
> >                   in->handle, in->function);
> > -
> > -    if (in->revision != 0x1 /* Currently we only support DSM Spec
> > Rev1. */) {
> > -        nvdimm_debug("Revision 0x%x is not supported, expect
> > 0x%x.\n",
> > -                     in->revision, 0x1);
> > +    /* Currently we only support DSM Spec Rev1 and Rev2. */
> 
> where does revision 2 come from? It would be better to add a pointer
> to relevant spec.

https://pmem.io/documents/IntelOptanePMem_DSM_Interface-V2.0.pdf,
Section 3 "_DSM Interface for the NVDIMM Device", table 3-A and 3-B.

I'll add this in comments in next version.
> 
> > +    if (in->revision != 0x1 && in->revision != 0x2) {
> > +        nvdimm_debug("Revision 0x%x is not supported, expect 0x1
> > or 0x2.\n",
> > +                     in->revision);
> 
> since you are touching nvdimm_debug(), please replace it with
> tracing,
> see docs/devel/tracing.rst and any commit that adds tracing calls
> (functions starting with 'trace_').

OK I'll have a try.
> 
> >          nvdimm_dsm_no_payload(NVDIMM_DSM_RET_STATUS_UNSUPPORT,
> > dsm_mem_addr);
> >          goto exit;
> >      }
> 
> 
> this whole hunk should be a separate patch, properly documented
> 
OK
> 
> also I wonder if DSM

It's not in SDM, but above-mentioned _DSM Interface spec by Intel.
> 
> > @@ -1247,6 +1247,11 @@ static void nvdimm_build_fit(Aml *dev)
> >  static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t
> > ram_slots)
> >  {
> >      uint32_t slot;
> > +    Aml *method, *pkg, *buff;
> > +
> > +    /* Build common shared buffer for params pass in/out */
> > +    buff = aml_buffer(4096, NULL);
> > +    aml_append(root_dev, aml_name_decl("BUFF", buff));
> 
> is there a reason to use global variable instead of LocalX?

Local in root_dev but global to its sub devices? I think it is doable.

But given your below comments on return param _LS{I,R,W}, I now think,
in v2, I'm not going to reuse existing "NCAL" method, but implement
_LS{I,R,W} their own, stringently follow interface spec. Then, no buff
required at all. How do you like this?
> 
> >  
> >      for (slot = 0; slot < ram_slots; slot++) {
> >          uint32_t handle = nvdimm_slot_to_handle(slot);
> > @@ -1264,6 +1269,49 @@ static void nvdimm_build_nvdimm_devices(Aml
> > *root_dev, uint32_t ram_slots)
> >           */
> >          aml_append(nvdimm_dev, aml_name_decl("_ADR",
> > aml_int(handle)));
> >  
> > +        /* Build _LSI, _LSR, _LSW */
> 
> should be 1 comment per method with spec/ver and chapter where it's
> defined

OK
> 
> > +        method = aml_method("_LSI", 0, AML_NOTSERIALIZED);
> > +        aml_append(method, aml_return(aml_call5(NVDIMM_COMMON_DSM,
> > +                            aml_touuid("4309AC30-0D11-11E4-9191-
> > 0800200C9A66"),
> > +                            aml_int(2), aml_int(4), aml_int(0),
> > +                            aml_int(handle))));
> > +        aml_append(nvdimm_dev, method);
> 
> _LSI should return Package

Right. See above.
> 
> > +        method = aml_method("_LSR", 2, AML_SERIALIZED);
> > +        aml_append(method,
> > +            aml_create_dword_field(aml_name("BUFF"), aml_int(0),
> > "DWD0"));
> > +        aml_append(method,
> > +            aml_create_dword_field(aml_name("BUFF"), aml_int(4),
> > "DWD1"));
> 
> theoretically aml_create_dword_field() takes TermArg as source
> buffer,
> so it doesn't have to be a global named buffer.
> Try keep buffer in LocalX variable and check if it works in earliest
> Windows version that supports NVDIMMs. If it does then drop BUFF and
> use
> Local variable, if not then that fact should be mentioned in commit
> message/patch

Thanks Igor. I'm new to asl grammar, I'll take your advice.

> 
> > +        pkg = aml_package(1);
> > +        aml_append(pkg, aml_name("BUFF"));
> > +        aml_append(method, aml_name_decl("PKG1", pkg));
> > +        aml_append(method, aml_store(aml_arg(0),
> > aml_name("DWD0")));
> > +        aml_append(method, aml_store(aml_arg(1),
> > aml_name("DWD1")));
> 
> perhaps use less magical names for fields, something like:
>   DOFF
>   TLEN
> add appropriate comments

No problem.
> 
> Also I'd prepare/fill in buffer first and only then declare
> initialize
> Package + don't use named object for Package if it can be done with
> help
> of Local variables.
> 
> > +        aml_append(method, aml_return(aml_call5(NVDIMM_COMMON_DSM,
> > +                            aml_touuid("4309AC30-0D11-11E4-9191-
> > 0800200C9A66"),
> > +                            aml_int(2), aml_int(5),
> > aml_name("PKG1"),
> > +                            aml_int(handle))));
> 
> this shall return Package not a Buffer

Right, Going to re-implement these methods rather than wrapper NCAL.
> 
> > +        aml_append(nvdimm_dev, method);
> > +
> > +        method = aml_method("_LSW", 3, AML_SERIALIZED);
> > +        aml_append(method,
> > +            aml_create_dword_field(aml_name("BUFF"), aml_int(0),
> > "DWD0"));
> > +        aml_append(method,
> > +            aml_create_dword_field(aml_name("BUFF"), aml_int(4),
> > "DWD1"));
> > +        aml_append(method,
> > +            aml_create_field(aml_name("BUFF"), aml_int(64),
> > aml_int(32672), "FILD"));
> > +        pkg = aml_package(1);
> > +        aml_append(pkg, aml_name("BUFF"));
> > +        aml_append(method, aml_name_decl("PKG1", pkg));
> > +        aml_append(method, aml_store(aml_arg(0),
> > aml_name("DWD0")));
> > +        aml_append(method, aml_store(aml_arg(1),
> > aml_name("DWD1")));
> > +        aml_append(method, aml_store(aml_arg(2),
> > aml_name("FILD")));
> > +        aml_append(method, aml_return(aml_call5(NVDIMM_COMMON_DSM,
> > +                            aml_touuid("4309AC30-0D11-11E4-9191-
> > 0800200C9A66"),
> > +                            aml_int(2), aml_int(6),
> > aml_name("PKG1"),
> > +                            aml_int(handle))));
> 
> should return Integer not Buffer, it looks like implicit conversion
> will take care of it,
> but it would be better to explicitly convert it to Integer even if
> it's only for the sake
> of documenting expected return value (or add a comment)

I observed guest kernel ACPI component complaining this; just warning,
no harm. I'll re-implement it.
> 
> Also returned value in case of error NVDIMM_DSM_RET_STATUS_INVALID,
> in NVDIMM and ACPI spec differ. So fix the spec or remap returned
> value.
> 
> 
> > +        aml_append(nvdimm_dev, method);
> > +
> >          nvdimm_build_device_dsm(nvdimm_dev, handle);
> >          aml_append(root_dev, nvdimm_dev);
> >      }
> 
>