Skip to contentSkip to conversation

smbios: make memory device size configurable per Machine

PATCH Superseded Igor Mammedov 4 participants 9 messages Jul 11, 2024 – Feb 4, 2025

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 smbios: make memory device size configurable per Machine

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
 include/hw/boards.h |  4 ++++
 hw/arm/virt.c       |  1 +
 hw/core/machine.c   |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 hw/smbios/smbios.c  | 11 ++++++-----
 6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +4 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
  *    purposes only.
  *    Applies only to default memory backend, i.e., explicit memory backend
  *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"
  */
 struct MachineClass {
     /*< private >*/
@@ -304,6 +307,7 @@ struct MachineClass {
     const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
     int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
     ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;
 };
 
 /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
 static void virt_machine_9_0_options(MachineClass *mc)
 {
     virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
 }
 DEFINE_VIRT_MACHINE(9, 0)
hw/core/machine.c +1 -0
index bc38cad7f2..3cfdaec65d 100644
@@ -1004,6 +1004,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smbios_memory_device_size = 2047 * TiB;
 
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned
hw/i386/pc_piix.c +1 -0
index 9445b07b4f..d9e69243b4 100644
@@ -495,6 +495,7 @@ static void pc_i440fx_machine_9_0_options(MachineClass *m)
     pc_i440fx_machine_9_1_options(m);
     m->alias = NULL;
     m->is_default = false;
+    m->smbios_memory_device_size = 16 * GiB;
 
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
hw/i386/pc_q35.c +1 -0
index 71d3c6d122..9d108b194e 100644
@@ -374,6 +374,7 @@ static void pc_q35_machine_9_0_options(MachineClass *m)
     PCMachineClass *pcmc = PC_MACHINE_CLASS(m);
     pc_q35_machine_9_1_options(m);
     m->alias = NULL;
+    m->smbios_memory_device_size = 16 * GiB;
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
     pcmc->isa_bios_alias = false;
hw/smbios/smbios.c +6 -5
index 3b7703489d..a394514264 100644
@@ -1093,6 +1093,7 @@ static bool smbios_get_tables_ep(MachineState *ms,
                        Error **errp)
 {
     unsigned i, dimm_cnt, offset;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
     ERRP_GUARD();
 
     assert(ep_type == SMBIOS_ENTRY_POINT_TYPE_32 ||
@@ -1123,12 +1124,12 @@ static bool smbios_get_tables_ep(MachineState *ms,
     smbios_build_type_9_table(errp);
     smbios_build_type_11_table();
 
-#define MAX_DIMM_SZ (16 * GiB)
-#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? MAX_DIMM_SZ \
-                                        : ((current_machine->ram_size - 1) % MAX_DIMM_SZ) + 1)
+#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? mc->smbios_memory_device_size \
+    : ((current_machine->ram_size - 1) % mc->smbios_memory_device_size) + 1)
 
-    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size, MAX_DIMM_SZ) /
-               MAX_DIMM_SZ;
+    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size,
+                             mc->smbios_memory_device_size) /
+               mc->smbios_memory_device_size;
 
     /*
      * The offset determines if we need to keep additional space between
Signature · 1 line
2.43.0
#1 lore raw

Hi Igor,

Quoted 62 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
  include/hw/boards.h |  4 ++++
  hw/arm/virt.c       |  1 +
  hw/core/machine.c   |  1 +
  hw/i386/pc_piix.c   |  1 +
  hw/i386/pc_q35.c    |  1 +
  hw/smbios/smbios.c  | 11 ++++++-----
  6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +3 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
   *    purposes only.
   *    Applies only to default memory backend, i.e., explicit memory backend
   *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"
   */
  struct MachineClass {
      /*< private >*/

include/hw/boards.h struct MachineClass { · line 310

Quoted 4 lines
include/hw/boards.h +1 -0
@@ -304,6 +307,7 @@ struct MachineClass {
      const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
      int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
      ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;

Quick notes since I'm on holidays (not meant to block this patch):

- How will evolve this machine class property in the context of a heterogeneous machine (i.e. x86_64 cores and 1 riscv32 one)?

- Should this become a SmbiosProviderInterface later?

Quoted 15 lines
include/hw/boards.h +0 -0
  };
 
  /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
  static void virt_machine_9_0_options(MachineClass *mc)
  {
      virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
      compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
  }
  DEFINE_VIRT_MACHINE(9, 0)

[...]

#2 lore raw
Quoted 72 lines · On Thu, 11 Jul 2024 10:19:27 +0200 Philippe Mathieu-Daudé <[email protected]>…

Hi Igor,

Quoted 62 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
  include/hw/boards.h |  4 ++++
  hw/arm/virt.c       |  1 +
  hw/core/machine.c   |  1 +
  hw/i386/pc_piix.c   |  1 +
  hw/i386/pc_q35.c    |  1 +
  hw/smbios/smbios.c  | 11 ++++++-----
  6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +3 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
   *    purposes only.
   *    Applies only to default memory backend, i.e., explicit memory backend
   *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"
   */
  struct MachineClass {
      /*< private >*/

include/hw/boards.h struct MachineClass { · line 310

Quoted 4 lines
include/hw/boards.h +1 -0
@@ -304,6 +307,7 @@ struct MachineClass {
      const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
      int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
      ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;  

Quick notes since I'm on holidays (not meant to block this patch):

Quoted 2 lines

- How will evolve this machine class property in the context of a heterogeneous machine (i.e. x86_64 cores and 1 riscv32 one)?

I'm not aware of a SMBIOS spec (3.x) that cares about that heterogeneous setup yet. Are there anything in that area exists yet?

Quoted 1 line

- Should this become a SmbiosProviderInterface later?

if/when SMBIOS does get there (heterogeneous machines), introducing an interface might make a sense.

Quoted 19 lines
Quoted 16 lines
include/hw/boards.h +0 -0
  };
 
  /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
  static void virt_machine_9_0_options(MachineClass *mc)
  {
      virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
      compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
  }
  DEFINE_VIRT_MACHINE(9, 0)  

[...]

#3 lore raw
Quoted 75 lines · Igor Mammedov
Quoted 72 lines · On Thu, 11 Jul 2024 10:19:27 +0200 Philippe Mathieu-Daudé <[email protected]>…

Hi Igor,

Quoted 62 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
   include/hw/boards.h |  4 ++++
   hw/arm/virt.c       |  1 +
   hw/core/machine.c   |  1 +
   hw/i386/pc_piix.c   |  1 +
   hw/i386/pc_q35.c    |  1 +
   hw/smbios/smbios.c  | 11 ++++++-----
   6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +3 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
    *    purposes only.
    *    Applies only to default memory backend, i.e., explicit memory backend
    *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"
    */
   struct MachineClass {
       /*< private >*/

include/hw/boards.h struct MachineClass { · line 310

Quoted 4 lines
include/hw/boards.h +1 -0
@@ -304,6 +307,7 @@ struct MachineClass {
       const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
       int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
       ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;

Quick notes since I'm on holidays (not meant to block this patch):

Quoted 5 lines
Quoted 2 lines

- How will evolve this machine class property in the context of a heterogeneous machine (i.e. x86_64 cores and 1 riscv32 one)?

I'm not aware of a SMBIOS spec (3.x) that cares about that heterogeneous setup yet. Are there anything in that area exists yet?

Not yet.

Quoted 4 lines
Quoted 2 lines

- Should this become a SmbiosProviderInterface later?

if/when SMBIOS does get there (heterogeneous machines), introducing an interface might make a sense.

OK.

#4 lore raw
In reply to Igor Mammedov
Quoted 20 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

Quoted 10 lines

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

I was wondering what real hardware does, since the best way to avoid guest OS surprises is to align with real world behaviour. IIUC, there is usually one Type 17 structure per physical DIMM.

Most QEMU configs don't express DIMMs as a concept so in that case, we can presume 1 virtual DIMM, and thus having one type 17 structure is a match for physical hw practices.

What about when the QEMU config has used nvdimm, pc-dimm, or virtio-mem devices though ? It feels like the best practice would be to have a type 17 structure for each instance of one of those devices.

Quoted 121 lines

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
 include/hw/boards.h |  4 ++++
 hw/arm/virt.c       |  1 +
 hw/core/machine.c   |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 hw/smbios/smbios.c  | 11 ++++++-----
 6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +4 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
  *    purposes only.
  *    Applies only to default memory backend, i.e., explicit memory backend
  *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"
  */
 struct MachineClass {
     /*< private >*/
@@ -304,6 +307,7 @@ struct MachineClass {
     const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
     int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
     ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;
 };
 
 /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
 static void virt_machine_9_0_options(MachineClass *mc)
 {
     virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
 }
 DEFINE_VIRT_MACHINE(9, 0)
hw/core/machine.c +1 -0
index bc38cad7f2..3cfdaec65d 100644
@@ -1004,6 +1004,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smbios_memory_device_size = 2047 * TiB;
 
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned
hw/i386/pc_piix.c +1 -0
index 9445b07b4f..d9e69243b4 100644
@@ -495,6 +495,7 @@ static void pc_i440fx_machine_9_0_options(MachineClass *m)
     pc_i440fx_machine_9_1_options(m);
     m->alias = NULL;
     m->is_default = false;
+    m->smbios_memory_device_size = 16 * GiB;
 
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
hw/i386/pc_q35.c +1 -0
index 71d3c6d122..9d108b194e 100644
@@ -374,6 +374,7 @@ static void pc_q35_machine_9_0_options(MachineClass *m)
     PCMachineClass *pcmc = PC_MACHINE_CLASS(m);
     pc_q35_machine_9_1_options(m);
     m->alias = NULL;
+    m->smbios_memory_device_size = 16 * GiB;
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
     pcmc->isa_bios_alias = false;
hw/smbios/smbios.c +6 -5
index 3b7703489d..a394514264 100644
@@ -1093,6 +1093,7 @@ static bool smbios_get_tables_ep(MachineState *ms,
                        Error **errp)
 {
     unsigned i, dimm_cnt, offset;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
     ERRP_GUARD();
 
     assert(ep_type == SMBIOS_ENTRY_POINT_TYPE_32 ||
@@ -1123,12 +1124,12 @@ static bool smbios_get_tables_ep(MachineState *ms,
     smbios_build_type_9_table(errp);
     smbios_build_type_11_table();
 
-#define MAX_DIMM_SZ (16 * GiB)
-#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? MAX_DIMM_SZ \
-                                        : ((current_machine->ram_size - 1) % MAX_DIMM_SZ) + 1)
+#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? mc->smbios_memory_device_size \
+    : ((current_machine->ram_size - 1) % mc->smbios_memory_device_size) + 1)
 
-    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size, MAX_DIMM_SZ) /
-               MAX_DIMM_SZ;
+    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size,
+                             mc->smbios_memory_device_size) /
+               mc->smbios_memory_device_size;
 
     /*
      * The offset determines if we need to keep additional space between
Signature · 1 line
2.43.0

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#5 lore raw
Quoted 42 lines · On Thu, 11 Jul 2024 09:43:46 +0100 Daniel P. Berrangé <[email protected]>…
Quoted 20 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

Quoted 10 lines

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

I was wondering what real hardware does, since the best way to avoid guest OS surprises is to align with real world behaviour. IIUC, there is usually one Type 17 structure per physical DIMM.

Most QEMU configs don't express DIMMs as a concept so in that case, we can presume 1 virtual DIMM, and thus having one type 17 structure is a match for physical hw practices.

Quoted 4 lines

What about when the QEMU config has used nvdimm, pc-dimm, or virtio-mem devices though ? It feels like the best practice would be to have a type 17 structure for each instance of one of those devices.

QEMU doesn't expose any memory beside initial one in SMBIOS. So from guest introspection pov when using only SMBIOS, those do not exists.

On tangent:
I think exposing those with hotplug in place makes
it messy especially with migration in mind (we would need to move smbios tables creation to reset time and enumerate all supported memory devices at that time to get somewhat reliable picture, which would reflect machine config _only_ at boot time).

Also it would help to model initial RAM as DIMM(s) device to avoid faking RAM entry, and do it consistently with DIMM devices.

(but yeah, nobody asked for anything like that so far).

Quoted 123 lines
Quoted 120 lines

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
 include/hw/boards.h |  4 ++++
 hw/arm/virt.c       |  1 +
 hw/core/machine.c   |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 hw/smbios/smbios.c  | 11 ++++++-----
 6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +4 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
  *    purposes only.
  *    Applies only to default memory backend, i.e., explicit memory backend
  *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"
  */
 struct MachineClass {
     /*< private >*/
@@ -304,6 +307,7 @@ struct MachineClass {
     const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
     int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
     ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;
 };
 
 /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
 static void virt_machine_9_0_options(MachineClass *mc)
 {
     virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
 }
 DEFINE_VIRT_MACHINE(9, 0)
hw/core/machine.c +1 -0
index bc38cad7f2..3cfdaec65d 100644
@@ -1004,6 +1004,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smbios_memory_device_size = 2047 * TiB;
 
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned
hw/i386/pc_piix.c +1 -0
index 9445b07b4f..d9e69243b4 100644
@@ -495,6 +495,7 @@ static void pc_i440fx_machine_9_0_options(MachineClass *m)
     pc_i440fx_machine_9_1_options(m);
     m->alias = NULL;
     m->is_default = false;
+    m->smbios_memory_device_size = 16 * GiB;
 
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
hw/i386/pc_q35.c +1 -0
index 71d3c6d122..9d108b194e 100644
@@ -374,6 +374,7 @@ static void pc_q35_machine_9_0_options(MachineClass *m)
     PCMachineClass *pcmc = PC_MACHINE_CLASS(m);
     pc_q35_machine_9_1_options(m);
     m->alias = NULL;
+    m->smbios_memory_device_size = 16 * GiB;
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
     pcmc->isa_bios_alias = false;
hw/smbios/smbios.c +6 -5
index 3b7703489d..a394514264 100644
@@ -1093,6 +1093,7 @@ static bool smbios_get_tables_ep(MachineState *ms,
                        Error **errp)
 {
     unsigned i, dimm_cnt, offset;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
     ERRP_GUARD();
 
     assert(ep_type == SMBIOS_ENTRY_POINT_TYPE_32 ||
@@ -1123,12 +1124,12 @@ static bool smbios_get_tables_ep(MachineState *ms,
     smbios_build_type_9_table(errp);
     smbios_build_type_11_table();
 
-#define MAX_DIMM_SZ (16 * GiB)
-#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? MAX_DIMM_SZ \
-                                        : ((current_machine->ram_size - 1) % MAX_DIMM_SZ) + 1)
+#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? mc->smbios_memory_device_size \
+    : ((current_machine->ram_size - 1) % mc->smbios_memory_device_size) + 1)
 
-    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size, MAX_DIMM_SZ) /
-               MAX_DIMM_SZ;
+    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size,
+                             mc->smbios_memory_device_size) /
+               mc->smbios_memory_device_size;
 
     /*
      * The offset determines if we need to keep additional space between
Signature · 1 line
2.43.0

With regards,
Daniel

#6 lore raw
In reply to Igor Mammedov
Quoted 52 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
 include/hw/boards.h |  4 ++++
 hw/arm/virt.c       |  1 +
 hw/core/machine.c   |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 hw/smbios/smbios.c  | 11 ++++++-----
 6 files changed, 14 insertions(+), 5 deletions(-)
include/hw/boards.h +0 -0
index ef6f18f2c1..48ff6d8b93 100644

include/hw/boards.h typedef struct { · line 242

Quoted 6 lines
include/hw/boards.h +3 -0
@@ -237,6 +237,9 @@ typedef struct {
  *    purposes only.
  *    Applies only to default memory backend, i.e., explicit memory backend
  *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"

Maybe it would be better to just make this a boolean, and put the spec related logic in smbios.c ? WDYT?

Quoted 28 lines
include/hw/boards.h +1 -0
  */
 struct MachineClass {
     /*< private >*/
@@ -304,6 +307,7 @@ struct MachineClass {
     const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
     int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
     ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;
 };
 
 /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
 static void virt_machine_9_0_options(MachineClass *mc)
 {
     virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
 }
 DEFINE_VIRT_MACHINE(9, 0)
hw/core/machine.c +0 -0
index bc38cad7f2..3cfdaec65d 100644

hw/core/machine.c static void machine_class_init(ObjectClass *oc, void *data) · line 1010

Quoted 7 lines
hw/core/machine.c +1 -0
@@ -1004,6 +1004,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smbios_memory_device_size = 2047 * TiB;
 
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned

All these values really should be documented.
And I feel

Quoted 56 lines
hw/i386/pc_piix.c +1 -0
index 9445b07b4f..d9e69243b4 100644
@@ -495,6 +495,7 @@ static void pc_i440fx_machine_9_0_options(MachineClass *m)
     pc_i440fx_machine_9_1_options(m);
     m->alias = NULL;
     m->is_default = false;
+    m->smbios_memory_device_size = 16 * GiB;
 
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
hw/i386/pc_q35.c +1 -0
index 71d3c6d122..9d108b194e 100644
@@ -374,6 +374,7 @@ static void pc_q35_machine_9_0_options(MachineClass *m)
     PCMachineClass *pcmc = PC_MACHINE_CLASS(m);
     pc_q35_machine_9_1_options(m);
     m->alias = NULL;
+    m->smbios_memory_device_size = 16 * GiB;
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
     pcmc->isa_bios_alias = false;
hw/smbios/smbios.c +6 -5
index 3b7703489d..a394514264 100644
@@ -1093,6 +1093,7 @@ static bool smbios_get_tables_ep(MachineState *ms,
                        Error **errp)
 {
     unsigned i, dimm_cnt, offset;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
     ERRP_GUARD();
 
     assert(ep_type == SMBIOS_ENTRY_POINT_TYPE_32 ||
@@ -1123,12 +1124,12 @@ static bool smbios_get_tables_ep(MachineState *ms,
     smbios_build_type_9_table(errp);
     smbios_build_type_11_table();
 
-#define MAX_DIMM_SZ (16 * GiB)
-#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? MAX_DIMM_SZ \
-                                        : ((current_machine->ram_size - 1) % MAX_DIMM_SZ) + 1)
+#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? mc->smbios_memory_device_size \
+    : ((current_machine->ram_size - 1) % mc->smbios_memory_device_size) + 1)
 
-    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size, MAX_DIMM_SZ) /
-               MAX_DIMM_SZ;
+    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size,
+                             mc->smbios_memory_device_size) /
+               mc->smbios_memory_device_size;
 
     /*
      * The offset determines if we need to keep additional space between
Signature · 1 line
2.43.0
#7 lore raw
Quoted 48 lines · On Thu, 11 Jul 2024 07:13:27 -0400 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 47 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
 include/hw/boards.h |  4 ++++
 hw/arm/virt.c       |  1 +
 hw/core/machine.c   |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 hw/smbios/smbios.c  | 11 ++++++-----
 6 files changed, 14 insertions(+), 5 deletions(-)
Quoted 15 lines

include/hw/boards.h typedef struct { · line 242

Quoted 11 lines
include/hw/boards.h +3 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
  *    purposes only.
  *    Applies only to default memory backend, i.e., explicit memory backend
  *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"  

Maybe it would be better to just make this a boolean, and put the spec related logic in smbios.c ? WDYT?

Using bool here, seems awkward to me,
i.e. not clear semantics and compat handling would be complicated as well.

And if we have to expose it someday to users, it would be logical to make it machine property. Given it's used not only by x86, having it as value here lets each machine to customize if necessary using well established pattern (incl. compat machinery)

Quoted 33 lines
Quoted 33 lines
include/hw/boards.h +1 -0
  */
 struct MachineClass {
     /*< private >*/
@@ -304,6 +307,7 @@ struct MachineClass {
     const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
     int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
     ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;
 };
 
 /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
 static void virt_machine_9_0_options(MachineClass *mc)
 {
     virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
 }
 DEFINE_VIRT_MACHINE(9, 0)
hw/core/machine.c +1 -0
index bc38cad7f2..3cfdaec65d 100644
@@ -1004,6 +1004,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smbios_memory_device_size = 2047 * TiB;
 
Quoted 6 lines

hw/core/machine.c static void machine_class_init(ObjectClass *oc, void *data) · line 1010

Quoted 2 lines
hw/core/machine.c +0 -0
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned  

All these values really should be documented.

It's in commit message, but right I'll document value here on respin so it would be easier for reader to see where it comes from.

Quoted 1 line

And I feel

???

Quoted 60 lines
Quoted 59 lines
hw/i386/pc_piix.c +1 -0
index 9445b07b4f..d9e69243b4 100644
@@ -495,6 +495,7 @@ static void pc_i440fx_machine_9_0_options(MachineClass *m)
     pc_i440fx_machine_9_1_options(m);
     m->alias = NULL;
     m->is_default = false;
+    m->smbios_memory_device_size = 16 * GiB;
 
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
hw/i386/pc_q35.c +1 -0
index 71d3c6d122..9d108b194e 100644
@@ -374,6 +374,7 @@ static void pc_q35_machine_9_0_options(MachineClass *m)
     PCMachineClass *pcmc = PC_MACHINE_CLASS(m);
     pc_q35_machine_9_1_options(m);
     m->alias = NULL;
+    m->smbios_memory_device_size = 16 * GiB;
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
     pcmc->isa_bios_alias = false;
hw/smbios/smbios.c +6 -5
index 3b7703489d..a394514264 100644
@@ -1093,6 +1093,7 @@ static bool smbios_get_tables_ep(MachineState *ms,
                        Error **errp)
 {
     unsigned i, dimm_cnt, offset;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
     ERRP_GUARD();
 
     assert(ep_type == SMBIOS_ENTRY_POINT_TYPE_32 ||
@@ -1123,12 +1124,12 @@ static bool smbios_get_tables_ep(MachineState *ms,
     smbios_build_type_9_table(errp);
     smbios_build_type_11_table();
 
-#define MAX_DIMM_SZ (16 * GiB)
-#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? MAX_DIMM_SZ \
-                                        : ((current_machine->ram_size - 1) % MAX_DIMM_SZ) + 1)
+#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? mc->smbios_memory_device_size \
+    : ((current_machine->ram_size - 1) % mc->smbios_memory_device_size) + 1)
 
-    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size, MAX_DIMM_SZ) /
-               MAX_DIMM_SZ;
+    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size,
+                             mc->smbios_memory_device_size) /
+               mc->smbios_memory_device_size;
 
     /*
      * The offset determines if we need to keep additional space between
Signature · 1 line
2.43.0  
#8 lore raw
Quoted 117 lines · Igor Mammedov
Quoted 48 lines · On Thu, 11 Jul 2024 07:13:27 -0400 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 47 lines · Igor Mammedov

Currently SMBIOS maximum memory device chunk is capped at 16Gb, which is fine for the most cases (QEMU uses it to describe initial RAM (type 17 SMBIOS table entries)).
However when starting guest with terabytes of RAM this leads to too many memory device structures, which eventually upsets linux kernel as it reserves only 64K for these entries and when that border is crossed out it runs out of reserved memory.

Instead of partitioning initial RAM on 16Gb chunks, use maximum possible chunk size that SMBIOS spec allows[1]. Which lets encode RAM in Mb units in uint32_t-1 field (upto 2047Tb). As result initial RAM will generate only one type 17 structure until host/guest reach ability to use more RAM in the future.

Compat changes:
We can't unconditionally change chunk size as it will break QEMU<->guest ABI (and migration). Thus introduce a new machine class field that would let older versioned machines to use 16Gb chunks while new machine type could use maximum possible chunk size.

While it might seem to be risky to rise max entry size this much (much beyond of what current physical RAM modules support), I'd not expect it causing much issues, modulo uncovering bugs in software running within guest. And those should be fixed on guest side to handle SMBIOS spec properly, especially if guest is expected to support so huge RAM configs.
In worst case, QEMU can reduce chunk size later if we would care enough about introducing a workaround for some 'unfixable' guest OS, either by fixing up the next machine type or giving users a CLI option to customize it.

1) SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size

PS:
* tested on 8Tb host with RHEL6 guest, which seems to parse type 17 SMBIOS table entries correctly (according to 'dmidecode').

---
 include/hw/boards.h |  4 ++++
 hw/arm/virt.c       |  1 +
 hw/core/machine.c   |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 hw/smbios/smbios.c  | 11 ++++++-----
 6 files changed, 14 insertions(+), 5 deletions(-)
Quoted 15 lines

include/hw/boards.h typedef struct { · line 242

Quoted 11 lines
include/hw/boards.h +3 -0
index ef6f18f2c1..48ff6d8b93 100644
@@ -237,6 +237,9 @@ typedef struct {
  *    purposes only.
  *    Applies only to default memory backend, i.e., explicit memory backend
  *    wasn't used.
+ * @smbios_memory_device_size:
+ *    Default size of memory device,
+ *    SMBIOS 3.1.0 "7.18 Memory Device (Type 17)"  

Maybe it would be better to just make this a boolean, and put the spec related logic in smbios.c ? WDYT?

Using bool here, seems awkward to me,
i.e. not clear semantics and compat handling would be complicated as well.

And if we have to expose it someday to users, it would be logical to make it machine property. Given it's used not only by x86, having it as value here lets each machine to customize if necessary using well established pattern (incl. compat machinery)

Quoted 33 lines
Quoted 33 lines
include/hw/boards.h +1 -0
  */
 struct MachineClass {
     /*< private >*/
@@ -304,6 +307,7 @@ struct MachineClass {
     const CPUArchIdList *(*possible_cpu_arch_ids)(MachineState *machine);
     int64_t (*get_default_cpu_node_id)(const MachineState *ms, int idx);
     ram_addr_t (*fixup_ram_size)(ram_addr_t size);
+    uint64_t smbios_memory_device_size;
 };
 
 /**
hw/arm/virt.c +1 -0
index b0c68d66a3..719e83e6a1 100644
@@ -3308,6 +3308,7 @@ DEFINE_VIRT_MACHINE_AS_LATEST(9, 1)
 static void virt_machine_9_0_options(MachineClass *mc)
 {
     virt_machine_9_1_options(mc);
+    mc->smbios_memory_device_size = 16 * GiB;
     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
 }
 DEFINE_VIRT_MACHINE(9, 0)
hw/core/machine.c +1 -0
index bc38cad7f2..3cfdaec65d 100644
@@ -1004,6 +1004,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smbios_memory_device_size = 2047 * TiB;
 
Quoted 6 lines

hw/core/machine.c static void machine_class_init(ObjectClass *oc, void *data) · line 1010

Quoted 2 lines
hw/core/machine.c +0 -0
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned  

All these values really should be documented.

Quoted 6 lines

It's in commit message, but right I'll document value here on respin so it would be easier for reader to see where it comes from.

Quoted 1 line

And I feel

???

sorry. what I said above. can we find some place where it is not awkward to quote spec?

Quoted 60 lines
Quoted 60 lines
Quoted 59 lines
hw/i386/pc_piix.c +1 -0
index 9445b07b4f..d9e69243b4 100644
@@ -495,6 +495,7 @@ static void pc_i440fx_machine_9_0_options(MachineClass *m)
     pc_i440fx_machine_9_1_options(m);
     m->alias = NULL;
     m->is_default = false;
+    m->smbios_memory_device_size = 16 * GiB;
 
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
hw/i386/pc_q35.c +1 -0
index 71d3c6d122..9d108b194e 100644
@@ -374,6 +374,7 @@ static void pc_q35_machine_9_0_options(MachineClass *m)
     PCMachineClass *pcmc = PC_MACHINE_CLASS(m);
     pc_q35_machine_9_1_options(m);
     m->alias = NULL;
+    m->smbios_memory_device_size = 16 * GiB;
     compat_props_add(m->compat_props, hw_compat_9_0, hw_compat_9_0_len);
     compat_props_add(m->compat_props, pc_compat_9_0, pc_compat_9_0_len);
     pcmc->isa_bios_alias = false;
hw/smbios/smbios.c +6 -5
index 3b7703489d..a394514264 100644
@@ -1093,6 +1093,7 @@ static bool smbios_get_tables_ep(MachineState *ms,
                        Error **errp)
 {
     unsigned i, dimm_cnt, offset;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
     ERRP_GUARD();
 
     assert(ep_type == SMBIOS_ENTRY_POINT_TYPE_32 ||
@@ -1123,12 +1124,12 @@ static bool smbios_get_tables_ep(MachineState *ms,
     smbios_build_type_9_table(errp);
     smbios_build_type_11_table();
 
-#define MAX_DIMM_SZ (16 * GiB)
-#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? MAX_DIMM_SZ \
-                                        : ((current_machine->ram_size - 1) % MAX_DIMM_SZ) + 1)
+#define GET_DIMM_SZ ((i < dimm_cnt - 1) ? mc->smbios_memory_device_size \
+    : ((current_machine->ram_size - 1) % mc->smbios_memory_device_size) + 1)
 
-    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size, MAX_DIMM_SZ) /
-               MAX_DIMM_SZ;
+    dimm_cnt = QEMU_ALIGN_UP(current_machine->ram_size,
+                             mc->smbios_memory_device_size) /
+               mc->smbios_memory_device_size;
 
     /*
      * The offset determines if we need to keep additional space between
Signature · 1 line
2.43.0