Skip to contentSkip to conversation

acpi/nvdimm: support NVDIMM _LS{I,R,W} methods

PATCH RESEND In review Robert Hoo 3 participants 17 messages Apr 12, 2022 – May 19, 2022

lore
Share & tools
lore

Robert Hoo

author
#0 lore raw

cover acpi/nvdimm: support NVDIMM _LS{I,R,W} methods

The original NVDIMM _DSM functions (index 4~6) for label operations have been deprecated by new ACPI methods _LS{I,R,W}[1][2].

Patch 1 implements the new _LS{I,R,W} methods, on top of old _DSM implementation.

Patch 2 fixes some typo of logical and/or with bitwise and/or, though functionally they haven't causing trouble.

[1] https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/index.html, 6.5.10 NVDIMM Label Methods
[2] https://pmem.io/documents/IntelOptanePMem_DSM_Interface-V2.0.pdf, 3.10 Deprecated Functions

---
Resend for previous failed delivery to "[email protected]" due to 550-'Message headers fail syntax check'.

Robert Hoo (2):
acpi/nvdimm: Create _LS{I,R,W} method for NVDIMM device acpi/nvdimm: Fix aml_or() and aml_and() in if clause

 hw/acpi/nvdimm.c | 60 +++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 54 insertions(+), 6 deletions(-)

base-commit: 95a3fcc7487e5bef262e1f937ed8636986764c4e

Signature · 1 line
2.31.1
author
#1 lore raw

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

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 standard method _LSI, _LSR, _LSW respectively. Functions semantics are almost identical, so my implementation is to reuse existing _DSMs, just create _LS{I,R,W} interfaces and constructs parameters and call _DSMs.

Only child NVDIMM devices has these methods, rather Root device.

By this patch, new NVDIMM sub device in ACPI namespace will be like this:

Device (NV00)

{
	Name (_ADR, One)  // _ADR: Address
        Method (_LSI, 0, NotSerialized)  // _LSI: Label Storage Information
        {
             Return (NCAL (ToUUID ("4309ac30-0d11-11e4-9191-0800200c9a66"), 0x02, 0x04, Zero, One))
        }

        Method (_LSR, 2, Serialized)  // _LSR: Label Storage Read
        {
        	CreateDWordField (BUFF, Zero, DWD0)
                CreateDWordField (BUFF, 0x04, DWD1)
                Name (PKG1, Package (0x01)
                {
                    BUFF
                })
                DWD0 = Arg0
                DWD1 = Arg1
                Return (NCAL (ToUUID ("4309ac30-0d11-11e4-9191-0800200c9a66"), 0x02, 0x05, PKG1, One))
        }

        Method (_LSW, 3, Serialized)  // _LSW: Label Storage Write
        {
                CreateDWordField (BUFF, Zero, DWD0)
                CreateDWordField (BUFF, 0x04, DWD1)
                CreateField (BUFF, 0x40, 0x7FA0, FILD)
                Name (PKG1, Package (0x01)
                {
                    BUFF
                })
                DWD0 = Arg0
                DWD1 = Arg1
                FILD = Arg2
                Return (NCAL (ToUUID ("4309ac30-0d11-11e4-9191-0800200c9a66"), 0x02, 0x06, PKG1, One))
         }

         Method (_DSM, 4, NotSerialized)  // _DSM: Device-Specific Method
         {
                Return (NCAL (Arg0, Arg1, Arg2, Arg3, One))
         }
}
---
 hw/acpi/nvdimm.c | 56 ++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 52 insertions(+), 4 deletions(-)
hw/acpi/nvdimm.c +52 -4
index 0d43da19ea..7cc419401b 100644
@@ -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. */
+    if (in->revision != 0x1 && in->revision != 0x2) {
+        nvdimm_debug("Revision 0x%x is not supported, expect 0x1 or 0x2.\n",
+                     in->revision);
         nvdimm_dsm_no_payload(NVDIMM_DSM_RET_STATUS_UNSUPPORT, dsm_mem_addr);
         goto exit;
     }
@@ -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));
 
     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 */
+        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);
+
+        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"));
+        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_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))));
+        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))));
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }
Signature · 1 line
2.31.1
#2 lore raw

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

Quoted 3 lines · On Tue, 12 Apr 2022 14:57:52 +0800 Robert Hoo <[email protected]> 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.

Quoted 61 lines

standard method _LSI, _LSR, _LSW respectively. Functions semantics are almost identical, so my implementation is to reuse existing _DSMs, just create _LS{I,R,W} interfaces and constructs parameters and call _DSMs.

Only child NVDIMM devices has these methods, rather Root device.

By this patch, new NVDIMM sub device in ACPI namespace will be like this:

Device (NV00)

{
	Name (_ADR, One)  // _ADR: Address
        Method (_LSI, 0, NotSerialized)  // _LSI: Label Storage Information
        {
             Return (NCAL (ToUUID ("4309ac30-0d11-11e4-9191-0800200c9a66"), 0x02, 0x04, Zero, One))
        }

        Method (_LSR, 2, Serialized)  // _LSR: Label Storage Read
        {
        	CreateDWordField (BUFF, Zero, DWD0)
                CreateDWordField (BUFF, 0x04, DWD1)
                Name (PKG1, Package (0x01)
                {
                    BUFF
                })
                DWD0 = Arg0
                DWD1 = Arg1
                Return (NCAL (ToUUID ("4309ac30-0d11-11e4-9191-0800200c9a66"), 0x02, 0x05, PKG1, One))
        }

        Method (_LSW, 3, Serialized)  // _LSW: Label Storage Write
        {
                CreateDWordField (BUFF, Zero, DWD0)
                CreateDWordField (BUFF, 0x04, DWD1)
                CreateField (BUFF, 0x40, 0x7FA0, FILD)
                Name (PKG1, Package (0x01)
                {
                    BUFF
                })
                DWD0 = Arg0
                DWD1 = Arg1
                FILD = Arg2
                Return (NCAL (ToUUID ("4309ac30-0d11-11e4-9191-0800200c9a66"), 0x02, 0x06, PKG1, One))
         }

         Method (_DSM, 4, NotSerialized)  // _DSM: Device-Specific Method
         {
                Return (NCAL (Arg0, Arg1, Arg2, Arg3, One))
         }
}
---
 hw/acpi/nvdimm.c | 56 ++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 52 insertions(+), 4 deletions(-)
hw/acpi/nvdimm.c +0 -0
index 0d43da19ea..7cc419401b 100644

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr, uint64_t val, unsigned size) · line 851

Quoted 8 lines
hw/acpi/nvdimm.c +1 -4
@@ -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.

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr, uint64_t val, unsigned size) · line 854

Quoted 3 lines
hw/acpi/nvdimm.c +3 -0
+    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_').

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr, uint64_t val, unsigned size) · line 857

Quoted 3 lines
hw/acpi/nvdimm.c +0 -0
         nvdimm_dsm_no_payload(NVDIMM_DSM_RET_STATUS_UNSUPPORT, dsm_mem_addr);
         goto exit;
     }

this whole hunk should be a separate patch, properly documented

also I wonder if DSM

hw/acpi/nvdimm.c static void nvdimm_build_fit(Aml *dev) · line 1254

Quoted 9 lines
hw/acpi/nvdimm.c +5 -0
@@ -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?

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1272

Quoted 8 lines
hw/acpi/nvdimm.c +1 -0
     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

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1278

Quoted 6 lines
hw/acpi/nvdimm.c +6 -0
+        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

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1284

Quoted 5 lines
hw/acpi/nvdimm.c +5 -0
+        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

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1289

Quoted 5 lines
hw/acpi/nvdimm.c +5 -0
+        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

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.

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1293

Quoted 4 lines
hw/acpi/nvdimm.c +4 -0
+        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

Quoted 9 lines
hw/acpi/nvdimm.c +9 -0
+        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"));

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1312

Quoted 10 lines
hw/acpi/nvdimm.c +10 -0
+        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)

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.

Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }
author
#3 lore raw

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

Quoted 10 lines · Igor Mammedov
Quoted 6 lines · On Tue, 12 Apr 2022 14:57:52 +0800 Robert Hoo <[email protected]> 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

Quoted 3 lines

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.

Quoted 1 line

...

Quoted 8 lines
Quoted 8 lines
hw/acpi/nvdimm.c +0 -0
index 0d43da19ea..7cc419401b 100644
@@ -848,10 +848,10 @@ nvdimm_dsm_write(void *opaque, hwaddr addr,
 uint64_t val, unsigned size)
 
Quoted 13 lines

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr,

Quoted 10 lines
hw/acpi/nvdimm.c +1 -4
     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.

Quoted 10 lines
Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+    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.

Quoted 9 lines
Quoted 5 lines

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

Quoted 2 lines

also I wonder if DSM

It's not in SDM, but above-mentioned _DSM Interface spec by Intel.

Quoted 13 lines

hw/acpi/nvdimm.c static void nvdimm_build_fit(Aml *dev) · line 1255

Quoted 11 lines
hw/acpi/nvdimm.c +5 -0
@@ -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?

Quoted 10 lines
Quoted 10 lines
hw/acpi/nvdimm.c +0 -0
     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)));
 
Quoted 4 lines

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1272

Quoted 1 line
hw/acpi/nvdimm.c +1 -0
+        /* Build _LSI, _LSR, _LSW */

should be 1 comment per method with spec/ver and chapter where it's defined

OK

Quoted 10 lines

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1278

Quoted 8 lines
hw/acpi/nvdimm.c +6 -0
+        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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +3 -0
+        method = aml_method("_LSR", 2, AML_SERIALIZED);
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"), aml_int(0),

"DWD0"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"), aml_int(4),

"DWD1"));

Quoted 8 lines

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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 4 lines

perhaps use less magical names for fields, something like:
DOFF
TLEN
add appropriate comments

No problem.

Quoted 7 lines

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.

Quoted 8 lines
Quoted 6 lines
hw/acpi/nvdimm.c +2 -0
+        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.

Quoted 29 lines
Quoted 24 lines
hw/acpi/nvdimm.c +5 -0
+        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"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"), aml_int(4),

"DWD1"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_field(aml_name("BUFF"), aml_int(64),

aml_int(32672), "FILD"));

hw/acpi/nvdimm.c +4 -0
+        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")));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method, aml_return(aml_call5(NVDIMM_COMMON_DSM,
+                            aml_touuid("4309AC30-0D11-11E4-9191-
Quoted 4 lines

0800200C9A66"),
+ aml_int(2), aml_int(6),
aml_name("PKG1"),
+ aml_int(handle))));

Quoted 5 lines

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.

Quoted 13 lines

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.

Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }
#4 lore raw

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

Quoted 52 lines · On Fri, 29 Apr 2022 17:01:47 +0800 Robert Hoo <[email protected]> wrote:
Quoted 10 lines · Igor Mammedov
Quoted 6 lines · On Tue, 12 Apr 2022 14:57:52 +0800 Robert Hoo <[email protected]> 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

Quoted 3 lines

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.

Quoted 1 line

...

Quoted 8 lines
Quoted 8 lines
hw/acpi/nvdimm.c +0 -0
index 0d43da19ea..7cc419401b 100644
@@ -848,10 +848,10 @@ nvdimm_dsm_write(void *opaque, hwaddr addr,
 uint64_t val, unsigned size)
 
Quoted 13 lines

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr,

Quoted 10 lines
hw/acpi/nvdimm.c +1 -4
     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.

Quoted 6 lines
Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+    if (in->revision != 0x1 && in->revision != 0x2) {
+        nvdimm_debug("Revision 0x%x is not supported, expect 0x1

or 0x2.\n",
+ in->revision);

Quoted 6 lines
Quoted 4 lines

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.

just make conversion a separate patch

Quoted 111 lines
Quoted 9 lines
Quoted 5 lines

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

Quoted 2 lines

also I wonder if DSM

It's not in SDM, but above-mentioned _DSM Interface spec by Intel.

Quoted 13 lines

hw/acpi/nvdimm.c static void nvdimm_build_fit(Aml *dev) · line 1255

Quoted 11 lines
hw/acpi/nvdimm.c +5 -0
@@ -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?

Quoted 10 lines
Quoted 10 lines
hw/acpi/nvdimm.c +0 -0
     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)));
 
Quoted 4 lines

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1272

Quoted 1 line
hw/acpi/nvdimm.c +1 -0
+        /* Build _LSI, _LSR, _LSW */  

should be 1 comment per method with spec/ver and chapter where it's defined

OK

Quoted 10 lines

hw/acpi/nvdimm.c static void nvdimm_build_nvdimm_devices(Aml *root_dev, uint32_t ram_slots) · line 1278

Quoted 8 lines
hw/acpi/nvdimm.c +6 -0
+        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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +3 -0
+        method = aml_method("_LSR", 2, AML_SERIALIZED);
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"), aml_int(0),

"DWD0"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"), aml_int(4),

"DWD1"));

Quoted 8 lines

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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 4 lines

perhaps use less magical names for fields, something like:
DOFF
TLEN
add appropriate comments

No problem.

Quoted 14 lines

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.

Quoted 6 lines
hw/acpi/nvdimm.c +2 -0
+        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))));

Quoted 3 lines
Quoted 1 line

this shall return Package not a Buffer

Right, Going to re-implement these methods rather than wrapper NCAL.

wrapper is fine, you just need to repackage content of the Buffer into a Package

Quoted 29 lines
Quoted 29 lines
Quoted 28 lines
hw/acpi/nvdimm.c +5 -0
+        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"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"), aml_int(4),

"DWD1"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_field(aml_name("BUFF"), aml_int(64),

aml_int(32672), "FILD"));

hw/acpi/nvdimm.c +4 -0
+        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")));

hw/acpi/nvdimm.c +2 -0
+        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))));

Quoted 8 lines
Quoted 5 lines

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.

try to test it with Windows guest (it usually less tolerable to errors than Linux + it's own quirks that you need to carter to) Also it would e nice if you test and put results in cover letter not only for Linux but for Windows as well.

Quoted 14 lines
Quoted 13 lines

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.

Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }  
author
#5 lore raw

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

Quoted 67 lines · Igor Mammedov
Quoted 64 lines · On Fri, 29 Apr 2022 17:01:47 +0800 Robert Hoo <[email protected]> wrote:
Quoted 13 lines · Igor Mammedov
Quoted 9 lines · On Tue, 12 Apr 2022 14:57:52 +0800 Robert Hoo <[email protected]> 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

Quoted 4 lines

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.

Quoted 1 line

...

Quoted 9 lines
Quoted 9 lines
hw/acpi/nvdimm.c +0 -0
index 0d43da19ea..7cc419401b 100644
@@ -848,10 +848,10 @@ nvdimm_dsm_write(void *opaque, hwaddr
 addr,
 uint64_t val, unsigned size)
 
Quoted 16 lines

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr, uint64_t val, unsigned size) · line 851

Quoted 12 lines
hw/acpi/nvdimm.c +0 -2
     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",

hw/acpi/nvdimm.c +1 -1
-                     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.

Quoted 7 lines
Quoted 6 lines
hw/acpi/nvdimm.c +2 -0
+    if (in->revision != 0x1 && in->revision != 0x2) {
+        nvdimm_debug("Revision 0x%x is not supported, expect

0x1
or 0x2.\n",
+ in->revision);

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

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.

just make conversion a separate patch

Yeah, I supposed so too.

Quoted 130 lines
Quoted 130 lines
Quoted 10 lines
Quoted 6 lines

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

Quoted 2 lines

also I wonder if DSM

It's not in SDM, but above-mentioned _DSM Interface spec by Intel.

Quoted 14 lines

hw/acpi/nvdimm.c static void nvdimm_build_fit(Aml *dev) · line 1254

Quoted 12 lines
hw/acpi/nvdimm.c +0 -0
@@ -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;
hw/acpi/nvdimm.c +5 -0
+    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?

Quoted 11 lines
Quoted 11 lines
  
 
     for (slot = 0; slot < ram_slots; slot++) {
         uint32_t handle = nvdimm_slot_to_handle(slot);
hw/acpi/nvdimm.c +0 -0
@@ -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)));

Quoted 5 lines
Quoted 1 line

+ /* Build _LSI, _LSR, _LSW */

should be 1 comment per method with spec/ver and chapter where it's
defined

OK

Quoted 12 lines

hw/acpi/nvdimm.c

Quoted 11 lines
hw/acpi/nvdimm.c +2 -0
+        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),

hw/acpi/nvdimm.c +2 -0
+                            aml_int(handle))));
+        aml_append(nvdimm_dev, method);  
Quoted 1 line

_LSI should return Package

Right. See above.

Quoted 11 lines
Quoted 10 lines
hw/acpi/nvdimm.c +3 -0
+        method = aml_method("_LSR", 2, AML_SERIALIZED);
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(0),
"DWD0"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

Quoted 11 lines

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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 4 lines

perhaps use less magical names for fields, something like:
DOFF
TLEN
add appropriate comments

No problem.

Quoted 17 lines

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.

Quoted 8 lines

+ 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))));

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

this shall return Package not a Buffer

Right, Going to re-implement these methods rather than wrapper NCAL.

wrapper is fine, you just need to repackage content of the Buffer into a Package

I now prefer re-implementation, i.e. make _LS{I,R,W} their own functions, less NACL's burden and don't make it more complex, it's already not neat; and more point, I think by this we can save the 4K Buff at all.
Does this sound all right to you?

Quoted 45 lines
Quoted 41 lines
Quoted 33 lines
Quoted 24 lines
hw/acpi/nvdimm.c +5 -0
+        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"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_field(aml_name("BUFF"), aml_int(64),

aml_int(32672), "FILD"));

hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 8 lines

+ 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))));

Quoted 7 lines

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)

Quoted 3 lines

I observed guest kernel ACPI component complaining this; just warning,
no harm. I'll re-implement it.

Quoted 6 lines

try to test it with Windows guest (it usually less tolerable to errors
than Linux + it's own quirks that you need to carter to) Also it would e nice if you test and put results in cover letter not only for Linux but for Windows as well.

I'll try to, but have no Windows resource at hand, I'll ask around if any test resource to cover that.

Quoted 16 lines
Quoted 14 lines
Quoted 14 lines

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.

Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }  
#6 lore raw

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

Quoted 213 lines · On Thu, 05 May 2022 11:07:53 +0800 Robert Hoo <[email protected]> wrote:
Quoted 67 lines · Igor Mammedov
Quoted 64 lines · On Fri, 29 Apr 2022 17:01:47 +0800 Robert Hoo <[email protected]> wrote:
Quoted 13 lines · Igor Mammedov
Quoted 9 lines · On Tue, 12 Apr 2022 14:57:52 +0800 Robert Hoo <[email protected]> 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

Quoted 4 lines

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.

Quoted 1 line

...

Quoted 9 lines
Quoted 9 lines
hw/acpi/nvdimm.c +0 -0
index 0d43da19ea..7cc419401b 100644
@@ -848,10 +848,10 @@ nvdimm_dsm_write(void *opaque, hwaddr
 addr,
 uint64_t val, unsigned size)
 
Quoted 16 lines

hw/acpi/nvdimm.c nvdimm_dsm_write(void *opaque, hwaddr addr, uint64_t val, unsigned size) · line 851

Quoted 12 lines
hw/acpi/nvdimm.c +0 -2
     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",

hw/acpi/nvdimm.c +1 -1
-                     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.

Quoted 7 lines
Quoted 6 lines
hw/acpi/nvdimm.c +2 -0
+    if (in->revision != 0x1 && in->revision != 0x2) {
+        nvdimm_debug("Revision 0x%x is not supported, expect

0x1
or 0x2.\n",
+ in->revision);

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

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.

just make conversion a separate patch

Yeah, I supposed so too.

Quoted 135 lines
Quoted 134 lines
Quoted 10 lines
Quoted 6 lines

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

Quoted 2 lines

also I wonder if DSM

It's not in SDM, but above-mentioned _DSM Interface spec by Intel.

Quoted 14 lines

hw/acpi/nvdimm.c static void nvdimm_build_fit(Aml *dev) · line 1254

Quoted 12 lines
hw/acpi/nvdimm.c +0 -0
@@ -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;
hw/acpi/nvdimm.c +5 -0
+    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?

Quoted 11 lines
Quoted 11 lines
    
 
     for (slot = 0; slot < ram_slots; slot++) {
         uint32_t handle = nvdimm_slot_to_handle(slot);
hw/acpi/nvdimm.c +0 -0
@@ -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)));

Quoted 5 lines
Quoted 1 line

+ /* Build _LSI, _LSR, _LSW */

should be 1 comment per method with spec/ver and chapter where it's
defined

OK

Quoted 12 lines

hw/acpi/nvdimm.c

Quoted 11 lines
hw/acpi/nvdimm.c +2 -0
+        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),

hw/acpi/nvdimm.c +2 -0
+                            aml_int(handle))));
+        aml_append(nvdimm_dev, method);    
Quoted 1 line

_LSI should return Package

Right. See above.

Quoted 11 lines
Quoted 10 lines
hw/acpi/nvdimm.c +3 -0
+        method = aml_method("_LSR", 2, AML_SERIALIZED);
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(0),
"DWD0"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

Quoted 11 lines

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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 4 lines

perhaps use less magical names for fields, something like:
DOFF
TLEN
add appropriate comments

No problem.

Quoted 17 lines

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.

Quoted 8 lines

+ 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))));

Quoted 1 line

this shall return Package not a Buffer

Right, Going to re-implement these methods rather than wrapper NCAL.

Quoted 8 lines
Quoted 3 lines

wrapper is fine, you just need to repackage content of the Buffer into a Package

I now prefer re-implementation, i.e. make _LS{I,R,W} their own functions, less NACL's burden and don't make it more complex, it's already not neat; and more point, I think by this we can save the 4K Buff at all.
Does this sound all right to you?

On one hand what you propose will be bit simpler (but not mach) than repacking (and only on AML guest side), however it will add to host side an extra interface/ABI that we will have to maintain, also it won't save space as buffer will still be there for legacy interface.

So unless we have to add new host/guest ABI, I'd prefer reusing existing one and complicate only new _LS{I,R,W} AML without touching NACL or host side.

Quoted 71 lines
Quoted 46 lines
Quoted 42 lines
Quoted 34 lines
Quoted 25 lines
hw/acpi/nvdimm.c +5 -0
+        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"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

hw/acpi/nvdimm.c +2 -0
+        aml_append(method,
+            aml_create_field(aml_name("BUFF"), aml_int(64),

aml_int(32672), "FILD"));

hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 8 lines

+ 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))));

Quoted 7 lines

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)

Quoted 3 lines

I observed guest kernel ACPI component complaining this; just warning,
no harm. I'll re-implement it.

Quoted 6 lines

try to test it with Windows guest (it usually less tolerable to errors
than Linux + it's own quirks that you need to carter to) Also it would e nice if you test and put results in cover letter not only for Linux but for Windows as well.

I'll try to, but have no Windows resource at hand, I'll ask around if any test resource to cover that.

Quoted 16 lines
Quoted 14 lines
Quoted 14 lines

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.

Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }    
author
#7 lore raw

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

On Thu, 2022-05-05 at 10:50 +0200, Igor Mammedov wrote:
...

Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 8 lines
Quoted 8 lines
(quoted diff) +0 -0
@@ -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;

Quoted 8 lines

hw/acpi/nvdimm.c static void nvdimm_build_fit(Aml *dev) · line 1254

Quoted 5 lines
(quoted diff) +2 -0
+
+    /* Build common shared buffer for params pass in/out

*/

hw/acpi/nvdimm.c +2 -0
+    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?

Quoted 1 line

...

Quoted 93 lines
Quoted 86 lines
Quoted 77 lines
Quoted 77 lines
Quoted 13 lines

Review

Quoted 12 lines
(quoted diff) +2 -0
+        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),

(quoted diff) +2 -0
+                            aml_int(handle))));
+        aml_append(nvdimm_dev, method);    
Quoted 1 line

_LSI should return Package

Right. See above.

Quoted 11 lines
Quoted 10 lines
hw/acpi/nvdimm.c +3 -0
+        method = aml_method("_LSR", 2, AML_SERIALIZED);
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(0),
"DWD0"));

(quoted diff) +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

Quoted 13 lines

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.

Quoted 9 lines
Quoted 8 lines
hw/acpi/nvdimm.c +4 -0
+        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")));

Quoted 4 lines

perhaps use less magical names for fields, something like:
DOFF
TLEN
add appropriate comments

No problem.

Quoted 19 lines

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.

Quoted 9 lines

+ 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))));

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

this shall return Package not a Buffer

Right, Going to re-implement these methods rather than wrapper NCAL.

wrapper is fine, you just need to repackage content of the Buffer into a Package

Quoted 6 lines

I now prefer re-implementation, i.e. make _LS{I,R,W} their own functions, less NACL's burden and don't make it more complex, it's already not neat; and more point, I think by this we can save the 4K
Buff at all.
Does this sound all right to you?

Quoted 4 lines

On one hand what you propose will be bit simpler (but not mach) than repacking (and only on AML guest side), however it will add to host side an extra interface/ABI that we will have to maintain, also it won't save space as buffer will still be there for legacy interface.

No, sorry, I didn't explain it clear.
No extra interface/ABI but these 3 must _LS{I,R,W} nvdimm-sub-device methods. Of course, I'm going to extract 'SystemIO' and 'SystemMemory' operation regions out of NACL to be globally available.

The buffer (BUFF in above patch) will be gone. It is added by my this patch, its mere use is to covert param of _LS{I,R,W} into those of NACL. If I implemented each _LS{I,R,W} on their own, rather than wrap the multi-purpose NACL, no buffer needed, at least I now assume so. And, why declare the 4K buffer global to sub-nvdimm? I now recall that it is because if not each sub-nvdimm device would contain a 4K buff, which will make this SSDT enormously large.

Quoted 4 lines

So unless we have to add new host/guest ABI, I'd prefer reusing existing one and complicate only new _LS{I,R,W} AML without touching NACL or host side.

As mentioned above, I assume no new host/guest ABI, just extract 'SystemIO' and 'SystemMemory' operation regions to a higher level scope.

Quoted 73 lines
Quoted 73 lines
Quoted 50 lines
Quoted 46 lines
Quoted 37 lines
Quoted 28 lines
hw/acpi/nvdimm.c +5 -0
+        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"));

(quoted diff) +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

(quoted diff) +2 -0
+        aml_append(method,
+            aml_create_field(aml_name("BUFF"),

aml_int(64),
aml_int(32672), "FILD"));

hw/acpi/nvdimm.c +4 -0
+        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,

Quoted 8 lines

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))));

Quoted 8 lines

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)

Quoted 3 lines

I observed guest kernel ACPI component complaining this; just warning,
no harm. I'll re-implement it.

Quoted 6 lines

try to test it with Windows guest (it usually less tolerable to errors
than Linux + it's own quirks that you need to carter to) Also it would e nice if you test and put results in cover letter not only for Linux but for Windows as well.

I'll try to, but have no Windows resource at hand, I'll ask around if
any test resource to cover that.

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines

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.

Quoted 5 lines
hw/acpi/nvdimm.c +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }    
#8 lore raw

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

Quoted 135 lines · On Thu, 05 May 2022 21:26:59 +0800 Robert Hoo <[email protected]> wrote:

On Thu, 2022-05-05 at 10:50 +0200, Igor Mammedov wrote:
...

Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 8 lines
Quoted 8 lines
(quoted diff) +0 -0
@@ -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;

Quoted 8 lines

Review

Quoted 5 lines
(quoted diff) +2 -0
+
+    /* Build common shared buffer for params pass in/out

*/

(quoted diff) +2 -0
+    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?

Quoted 1 line

...

Quoted 93 lines
Quoted 86 lines
Quoted 77 lines
Quoted 77 lines
Quoted 13 lines

Review

Quoted 12 lines
(quoted diff) +2 -0
+        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),

(quoted diff) +2 -0
+                            aml_int(handle))));
+        aml_append(nvdimm_dev, method);      
Quoted 1 line

_LSI should return Package

Right. See above.

Quoted 11 lines
Quoted 10 lines
(quoted diff) +3 -0
+        method = aml_method("_LSR", 2, AML_SERIALIZED);
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(0),
"DWD0"));

(quoted diff) +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

Quoted 13 lines

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.

Quoted 9 lines
Quoted 8 lines
(quoted diff) +4 -0
+        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")));

Quoted 4 lines

perhaps use less magical names for fields, something like:
DOFF
TLEN
add appropriate comments

No problem.

Quoted 19 lines

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.

Quoted 9 lines

+ 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))));

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

this shall return Package not a Buffer

Right, Going to re-implement these methods rather than wrapper NCAL.

wrapper is fine, you just need to repackage content of the Buffer into a Package

Quoted 6 lines

I now prefer re-implementation, i.e. make _LS{I,R,W} their own functions, less NACL's burden and don't make it more complex, it's already not neat; and more point, I think by this we can save the 4K
Buff at all.
Does this sound all right to you?

Quoted 4 lines

On one hand what you propose will be bit simpler (but not mach) than repacking (and only on AML guest side), however it will add to host side an extra interface/ABI that we will have to maintain, also it won't save space as buffer will still be there for legacy interface.

No, sorry, I didn't explain it clear.
No extra interface/ABI but these 3 must _LS{I,R,W} nvdimm-sub-device methods. Of course, I'm going to extract 'SystemIO' and 'SystemMemory' operation regions out of NACL to be globally available.

Quoted 7 lines

The buffer (BUFF in above patch) will be gone. It is added by my this patch, its mere use is to covert param of _LS{I,R,W} into those of NACL. If I implemented each _LS{I,R,W} on their own, rather than wrap the multi-purpose NACL, no buffer needed, at least I now assume so. And, why declare the 4K buffer global to sub-nvdimm? I now recall that it is because if not each sub-nvdimm device would contain a 4K buff, which will make this SSDT enormously large.

ok, lets see how it will look like when you are done.

Quoted 82 lines
Quoted 4 lines

So unless we have to add new host/guest ABI, I'd prefer reusing existing one and complicate only new _LS{I,R,W} AML without touching NACL or host side.

As mentioned above, I assume no new host/guest ABI, just extract 'SystemIO' and 'SystemMemory' operation regions to a higher level scope.

Quoted 73 lines
Quoted 73 lines
Quoted 50 lines
Quoted 46 lines
Quoted 37 lines
Quoted 28 lines
(quoted diff) +5 -0
+        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"));

(quoted diff) +2 -0
+        aml_append(method,
+            aml_create_dword_field(aml_name("BUFF"),

aml_int(4),
"DWD1"));

(quoted diff) +2 -0
+        aml_append(method,
+            aml_create_field(aml_name("BUFF"),

aml_int(64),
aml_int(32672), "FILD"));

(quoted diff) +4 -0
+        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,

Quoted 8 lines

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))));

Quoted 8 lines

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)

Quoted 3 lines

I observed guest kernel ACPI component complaining this; just warning,
no harm. I'll re-implement it.

Quoted 6 lines

try to test it with Windows guest (it usually less tolerable to errors
than Linux + it's own quirks that you need to carter to) Also it would e nice if you test and put results in cover letter not only for Linux but for Windows as well.

I'll try to, but have no Windows resource at hand, I'll ask around if
any test resource to cover that.

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines

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.

Quoted 5 lines
(quoted diff) +2 -0
+        aml_append(nvdimm_dev, method);
+
         nvdimm_build_device_dsm(nvdimm_dev, handle);
         aml_append(root_dev, nvdimm_dev);
     }      
author
#9 lore raw

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

Quoted 21 lines · Igor Mammedov
Quoted 9 lines

No, sorry, I didn't explain it clear.
No extra interface/ABI but these 3 must _LS{I,R,W} nvdimm-sub- device
methods. Of course, I'm going to extract 'SystemIO' and 'SystemMemory'
operation regions out of NACL to be globally available.

Quoted 11 lines

The buffer (BUFF in above patch) will be gone. It is added by my this
patch, its mere use is to covert param of _LS{I,R,W} into those of NACL. If I implemented each _LS{I,R,W} on their own, rather than wrap
the multi-purpose NACL, no buffer needed, at least I now assume so. And, why declare the 4K buffer global to sub-nvdimm? I now recall that
it is because if not each sub-nvdimm device would contain a 4K buff,
which will make this SSDT enormously large.

Quoted 1 line

ok, lets see how it will look like when you are done.

In ASL, can we define package with Arg in? e.g.

Name (PKG1, Package ()

            {
                Arg0,
                Arg1,
                Arg2
            })

But it cannot pass compilation. Any approach to achieve this? if so, we can still use simpler wrap scheme like v1 and save the 4K buffer.

Quoted 10 lines
Quoted 10 lines
Quoted 5 lines

So unless we have to add new host/guest ABI, I'd prefer reusing existing one and complicate only new _LS{I,R,W} AML without touching NACL or host side.

As mentioned above, I assume no new host/guest ABI, just extract 'SystemIO' and 'SystemMemory' operation regions to a higher level scope.

Quoted 1 line
#10 lore raw

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

Quoted 24 lines · On Wed, 18 May 2022 08:20:56 +0800 Robert Hoo <[email protected]> wrote:
Quoted 21 lines · Igor Mammedov
Quoted 9 lines

No, sorry, I didn't explain it clear.
No extra interface/ABI but these 3 must _LS{I,R,W} nvdimm-sub- device
methods. Of course, I'm going to extract 'SystemIO' and 'SystemMemory'
operation regions out of NACL to be globally available.

Quoted 11 lines

The buffer (BUFF in above patch) will be gone. It is added by my this
patch, its mere use is to covert param of _LS{I,R,W} into those of NACL. If I implemented each _LS{I,R,W} on their own, rather than wrap
the multi-purpose NACL, no buffer needed, at least I now assume so. And, why declare the 4K buffer global to sub-nvdimm? I now recall that
it is because if not each sub-nvdimm device would contain a 4K buff,
which will make this SSDT enormously large.

Quoted 1 line

ok, lets see how it will look like when you are done.

Quoted 8 lines

In ASL, can we define package with Arg in? e.g.

Name (PKG1, Package ()

            {
                Arg0,
                Arg1,
                Arg2
            })

Looking at the spec it doesn't seem to be a valid construct. see "DefPackage :=" and "PackageElement :=" definitions.

However you can try to play with RefOf to turn ArgX into reference (mind 'read' rules fro ArgTerm).

Quoted 16 lines

But it cannot pass compilation. Any approach to achieve this? if so, we can still use simpler wrap scheme like v1 and save the 4K buffer.

Quoted 10 lines
Quoted 10 lines
Quoted 5 lines

So unless we have to add new host/guest ABI, I'd prefer reusing existing one and complicate only new _LS{I,R,W} AML without touching NACL or host side.

As mentioned above, I assume no new host/guest ABI, just extract 'SystemIO' and 'SystemMemory' operation regions to a higher level scope.

Quoted 1 line
author
#11 lore raw

PATCH 2/2 acpi/nvdimm: Fix aml_or() and aml_and() in if clause

In reply to Robert Hoo

It should be some typo originally, where in If condition, using bitwise and/or, rather than logical and/or.

The resulting change in AML code:

If (((Local6 == Zero) | (Arg0 != Local0)))
==>
If (((Local6 == Zero) || (Arg0 != Local0)))

If (((ObjectType (Arg3) == 0x04) & (SizeOf (Arg3) == One))) ==>
If (((ObjectType (Arg3) == 0x04) && (SizeOf (Arg3) == One)))

Fixes 90623ebf603 ("nvdimm acpi: check UUID")
Fixes 4568c948066 ("nvdimm acpi: save arg3 of _DSM method")
---
 hw/acpi/nvdimm.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
hw/acpi/nvdimm.c +3 -4
index 7cc419401b..2cd26bb9e9 100644
@@ -1040,7 +1040,7 @@ static void nvdimm_build_common_dsm(Aml *dev,
 
     uuid_invalid = aml_lnot(aml_equal(uuid, expected_uuid));
 
-    unsupport = aml_if(aml_or(unpatched, uuid_invalid, NULL));
+    unsupport = aml_if(aml_lor(unpatched, uuid_invalid));
 
     /*
      * function 0 is called to inquire what functions are supported by
@@ -1072,10 +1072,9 @@ static void nvdimm_build_common_dsm(Aml *dev,
      * in the DSM Spec.
      */
     pckg = aml_arg(3);
-    ifctx = aml_if(aml_and(aml_equal(aml_object_type(pckg),
+    ifctx = aml_if(aml_land(aml_equal(aml_object_type(pckg),
                    aml_int(4 /* Package */)) /* It is a Package? */,
-                   aml_equal(aml_sizeof(pckg), aml_int(1)) /* 1 element? */,
-                   NULL));
+                   aml_equal(aml_sizeof(pckg), aml_int(1)) /* 1 element? */));
 
     pckg_index = aml_local(2);
     pckg_buf = aml_local(3);
Signature · 1 line
2.31.1
Reviewed-by on v1 2/2
#12 lore raw

Re: [PATCH 2/2] acpi/nvdimm: Fix aml_or() and aml_and() in if clause

Quoted 9 lines · On Tue, 12 Apr 2022 14:57:53 +0800 Robert Hoo <[email protected]> wrote:

It should be some typo originally, where in If condition, using bitwise and/or, rather than logical and/or.

The resulting change in AML code:

If (((Local6 == Zero) | (Arg0 != Local0)))
==>
If (((Local6 == Zero) || (Arg0 != Local0)))

Quoted 8 lines

If (((ObjectType (Arg3) == 0x04) & (SizeOf (Arg3) == One))) ==>
If (((ObjectType (Arg3) == 0x04) && (SizeOf (Arg3) == One)))

Fixes 90623ebf603 ("nvdimm acpi: check UUID")
Fixes 4568c948066 ("nvdimm acpi: save arg3 of _DSM method")
Quoted 30 lines
---
 hw/acpi/nvdimm.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
hw/acpi/nvdimm.c +3 -4
index 7cc419401b..2cd26bb9e9 100644
@@ -1040,7 +1040,7 @@ static void nvdimm_build_common_dsm(Aml *dev,
 
     uuid_invalid = aml_lnot(aml_equal(uuid, expected_uuid));
 
-    unsupport = aml_if(aml_or(unpatched, uuid_invalid, NULL));
+    unsupport = aml_if(aml_lor(unpatched, uuid_invalid));
 
     /*
      * function 0 is called to inquire what functions are supported by
@@ -1072,10 +1072,9 @@ static void nvdimm_build_common_dsm(Aml *dev,
      * in the DSM Spec.
      */
     pckg = aml_arg(3);
-    ifctx = aml_if(aml_and(aml_equal(aml_object_type(pckg),
+    ifctx = aml_if(aml_land(aml_equal(aml_object_type(pckg),
                    aml_int(4 /* Package */)) /* It is a Package? */,
-                   aml_equal(aml_sizeof(pckg), aml_int(1)) /* 1 element? */,
-                   NULL));
+                   aml_equal(aml_sizeof(pckg), aml_int(1)) /* 1 element? */));
 
     pckg_index = aml_local(2);
     pckg_buf = aml_local(3);
#13 lore raw

Re: [PATCH 2/2] acpi/nvdimm: Fix aml_or() and aml_and() in if clause

In reply to Robert Hoo
Quoted 9 lines · Robert Hoo

It should be some typo originally, where in If condition, using bitwise and/or, rather than logical and/or.

The resulting change in AML code:

If (((Local6 == Zero) | (Arg0 != Local0)))
==>
If (((Local6 == Zero) || (Arg0 != Local0)))

Quoted 8 lines

If (((ObjectType (Arg3) == 0x04) & (SizeOf (Arg3) == One))) ==>
If (((ObjectType (Arg3) == 0x04) && (SizeOf (Arg3) == One)))

Fixes 90623ebf603 ("nvdimm acpi: check UUID")
Fixes 4568c948066 ("nvdimm acpi: save arg3 of _DSM method")

This changes existing AML, you need to do the dance
with updating bios test tables, see header of ./tests/qtest/bios-tables-test.c

Quoted 32 lines
---
 hw/acpi/nvdimm.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
hw/acpi/nvdimm.c +3 -4
index 7cc419401b..2cd26bb9e9 100644
@@ -1040,7 +1040,7 @@ static void nvdimm_build_common_dsm(Aml *dev,
 
     uuid_invalid = aml_lnot(aml_equal(uuid, expected_uuid));
 
-    unsupport = aml_if(aml_or(unpatched, uuid_invalid, NULL));
+    unsupport = aml_if(aml_lor(unpatched, uuid_invalid));
 
     /*
      * function 0 is called to inquire what functions are supported by
@@ -1072,10 +1072,9 @@ static void nvdimm_build_common_dsm(Aml *dev,
      * in the DSM Spec.
      */
     pckg = aml_arg(3);
-    ifctx = aml_if(aml_and(aml_equal(aml_object_type(pckg),
+    ifctx = aml_if(aml_land(aml_equal(aml_object_type(pckg),
                    aml_int(4 /* Package */)) /* It is a Package? */,
-                   aml_equal(aml_sizeof(pckg), aml_int(1)) /* 1 element? */,
-                   NULL));
+                   aml_equal(aml_sizeof(pckg), aml_int(1)) /* 1 element? */));
 
     pckg_index = aml_local(2);
     pckg_buf = aml_local(3);
Signature · 1 line
2.31.1
author
#14 lore raw
In reply to Robert Hoo

Ping...

Quoted 29 lines · Robert Hoo

The original NVDIMM _DSM functions (index 4~6) for label operations have
been deprecated by new ACPI methods _LS{I,R,W}[1][2].

Patch 1 implements the new _LS{I,R,W} methods, on top of old _DSM implementation.

Patch 2 fixes some typo of logical and/or with bitwise and/or, though functionally they haven't causing trouble.

[1] https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/index.html, 6.5.10 NVDIMM Label Methods
[2] https://pmem.io/documents/IntelOptanePMem_DSM_Interface-V2.0.pdf,
3.10 Deprecated Functions

---
Resend for previous failed delivery to "[email protected]" due to 550-'Message headers fail syntax check'.

Robert Hoo (2):
acpi/nvdimm: Create _LS{I,R,W} method for NVDIMM device acpi/nvdimm: Fix aml_or() and aml_and() in if clause

 hw/acpi/nvdimm.c | 60 +++++++++++++++++++++++++++++++++++++++++++---
--
 1 file changed, 54 insertions(+), 6 deletions(-)

base-commit: 95a3fcc7487e5bef262e1f937ed8636986764c4e

#15 lore raw
In reply to Robert Hoo
Quoted 8 lines · On Tue, 12 Apr 2022 14:57:51 +0800 Robert Hoo <[email protected]> wrote:

The original NVDIMM _DSM functions (index 4~6) for label operations have been deprecated by new ACPI methods _LS{I,R,W}[1][2].

Patch 1 implements the new _LS{I,R,W} methods, on top of old _DSM implementation.

Patch 2 fixes some typo of logical and/or with bitwise and/or, though functionally they haven't causing trouble.

generic requirement for ACPI patches,
the should pass bios-tables-test (part of 'make check')

for that you need to update testcase expected data,
see tests/qtest/bios-tables-test.c for the process
also see https://www.mail-archive.com/[email protected]/msg875304.html for example

Quoted 17 lines

[1] https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/index.html, 6.5.10 NVDIMM Label Methods
[2] https://pmem.io/documents/IntelOptanePMem_DSM_Interface-V2.0.pdf, 3.10 Deprecated Functions

---
Resend for previous failed delivery to "[email protected]" due to 550-'Message headers fail syntax check'.

Robert Hoo (2):
acpi/nvdimm: Create _LS{I,R,W} method for NVDIMM device acpi/nvdimm: Fix aml_or() and aml_and() in if clause

 hw/acpi/nvdimm.c | 60 +++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 54 insertions(+), 6 deletions(-)

base-commit: 95a3fcc7487e5bef262e1f937ed8636986764c4e

author
#16 lore raw
Quoted 14 lines · Igor Mammedov
Quoted 10 lines · On Tue, 12 Apr 2022 14:57:51 +0800 Robert Hoo <[email protected]> wrote:

The original NVDIMM _DSM functions (index 4~6) for label operations have
been deprecated by new ACPI methods _LS{I,R,W}[1][2].

Patch 1 implements the new _LS{I,R,W} methods, on top of old _DSM implementation.

Patch 2 fixes some typo of logical and/or with bitwise and/or, though
functionally they haven't causing trouble.

Quoted 9 lines

generic requirement for ACPI patches,
the should pass bios-tables-test (part of 'make check')

for that you need to update testcase expected data,
see tests/qtest/bios-tables-test.c for the process
also see
https://www.mail-archive.com/[email protected]/msg875304.html for example

Got it. Thanks Igor.

Quoted 24 lines
Quoted 22 lines

[1] https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/index.html,
6.5.10 NVDIMM Label Methods
[2]
https://pmem.io/documents/IntelOptanePMem_DSM_Interface-V2.0.pdf,
3.10 Deprecated Functions

---
Resend for previous failed delivery to "[email protected]" due to
550-'Message headers fail syntax check'.

Robert Hoo (2):
acpi/nvdimm: Create _LS{I,R,W} method for NVDIMM device acpi/nvdimm: Fix aml_or() and aml_and() in if clause

 hw/acpi/nvdimm.c | 60 +++++++++++++++++++++++++++++++++++++++++++-
----
 1 file changed, 54 insertions(+), 6 deletions(-)

base-commit: 95a3fcc7487e5bef262e1f937ed8636986764c4e