Re: [Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Dou Liyang]

Hi Igor,

At 09/04/2017 07:11 PM, Igor Mammedov wrote:
[...]

+if (mem_base <= HOLE_640K_START &&
+next_base > HOLE_640K_START) {
+mem_len -= next_base - HOLE_640K_START;
+if (mem_len > 0) {
+numamem = acpi_data_push(table_data, sizeof *numamem);
+build_srat_memory(numamem, mem_base, mem_len, i - 1,
+  MEM_AFFINITY_ENABLED);
+}
+
+/* Check for the rare case: 640K < RAM < 1M */
+if (next_base <= HOLE_640K_END) {
+next_base = HOLE_640K_END;

Is this assignment really necessary?



It is necessary, because we set mem_base to next_base before setting
next_base;

But, I can refine it:

MEM_AFFINITY_ENABLED);
  }

+mem_base = HOLE_640K_END;
  /* Check for the rare case: 640K < RAM < 1M */
  if (next_base <= HOLE_640K_END) {
-next_base = HOLE_640K_END;
  continue;
  }
-mem_base = HOLE_640K_END;
  mem_len = next_base - HOLE_640K_END;
  }

Is it?

I was wrong, so just leave it as it is now.



OK, I see.

Thanks,
dou.

Re: [Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Igor Mammedov]

On Mon, 4 Sep 2017 18:16:31 +0800
Dou Liyang  wrote:

> At 09/04/2017 05:39 PM, Igor Mammedov wrote:
> > On Thu, 31 Aug 2017 20:04:26 +0800
> > Dou Liyang  wrote:
> >  
> >> From: Eduardo Habkost 
> >>
> >> Currently, Using the fisrt node without memory on the machine makes
> >> QEMU unhappy. With this example command line:
> >>   ... \
> >>   -m 1024M,slots=4,maxmem=32G \
> >>   -numa node,nodeid=0 \
> >>   -numa node,mem=1024M,nodeid=1 \
> >>   -numa node,nodeid=2 \
> >>   -numa node,nodeid=3 \
> >> Guest reports "No NUMA configuration found" and the NUMA topology is
> >> wrong.
> >>
> >> This is because when QEMU builds ACPI SRAT, it regards node 0 as the
> >> default node to deal with the memory hole(640K-1M). this means the
> >> node0 must have some memory(>1M), but, actually it can have no
> >> memory.
> >>
> >> Fix this problem by  cut out the 640K hole in the same way the PCI
> >> 4G hole does. Also do some cleanup.
> >>
> >> Signed-off-by: Eduardo Habkost 
> >> Signed-off-by: Dou Liyang 
> >> ---
> >>  hw/i386/acpi-build.c | 30 +++---
> >>  1 file changed, 23 insertions(+), 7 deletions(-)
> >>
> >> diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> >> index 98dd424..48525a1 100644
> >> --- a/hw/i386/acpi-build.c
> >> +++ b/hw/i386/acpi-build.c
> >> @@ -2318,6 +2318,9 @@ build_tpm2(GArray *table_data, BIOSLinker *linker)
> >>   (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL, 
> >> NULL);
> >>  }
> >>
> >> +#define HOLE_640K_START  (640 * 1024)
> >> +#define HOLE_640K_END   (1024 * 1024)
> >> +
> >>  static void
> >>  build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
> >>  {
> >> @@ -2373,17 +2376,30 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
> >> MachineState *machine)
> >>  next_base = 0;
> >>  numa_start = table_data->len;
> >>
> >> -numamem = acpi_data_push(table_data, sizeof *numamem);
> >> -build_srat_memory(numamem, 0, 640 * 1024, 0, MEM_AFFINITY_ENABLED);
> >> -next_base = 1024 * 1024;
> >>  for (i = 1; i < pcms->numa_nodes + 1; ++i) {
> >>  mem_base = next_base;
> >>  mem_len = pcms->node_mem[i - 1];
> >> -if (i == 1) {
> >> -mem_len -= 1024 * 1024;
> >> -}
> >>  next_base = mem_base + mem_len;
> >>
> >> +/* Cut out the 640K hole */
> >> +if (mem_base <= HOLE_640K_START &&
> >> +next_base > HOLE_640K_START) {
> >> +mem_len -= next_base - HOLE_640K_START;
> >> +if (mem_len > 0) {
> >> +numamem = acpi_data_push(table_data, sizeof *numamem);
> >> +build_srat_memory(numamem, mem_base, mem_len, i - 1,
> >> +  MEM_AFFINITY_ENABLED);
> >> +}
> >> +
> >> +/* Check for the rare case: 640K < RAM < 1M */
> >> +if (next_base <= HOLE_640K_END) {
> >> +next_base = HOLE_640K_END;  
> > Is this assignment really necessary?
> >  
> 
> It is necessary, because we set mem_base to next_base before setting
> next_base;
> 
> But, I can refine it:
> 
> MEM_AFFINITY_ENABLED);
>   }
> 
> +mem_base = HOLE_640K_END;
>   /* Check for the rare case: 640K < RAM < 1M */
>   if (next_base <= HOLE_640K_END) {
> -next_base = HOLE_640K_END;
>   continue;
>   }
> -mem_base = HOLE_640K_END;
>   mem_len = next_base - HOLE_640K_END;
>   }
> 
> Is it?
I was wrong, so just leave it as it is now.

> 
> Thanks,
>   dou.
> 
> > it seems that next_base will be set at the start of the loop.
> >  
> 
> 
> >> +continue;
> >> +}
> >> +mem_base = HOLE_640K_END;
> >> +mem_len = next_base - HOLE_640K_END;
> >> +}
> >> +
> >>  /* Cut out the ACPI_PCI hole */
> >>  if (mem_base <= pcms->below_4g_mem_size &&
> >>  next_base > pcms->below_4g_mem_size) {
> >> @@ -2395,7 +2411,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
> >> MachineState *machine)
> >>  }
> >>  mem_base = 1ULL << 32;
> >>  mem_len = next_base - pcms->below_4g_mem_size;
> >> -next_base += (1ULL << 32) - pcms->below_4g_mem_size;
> >> +next_base = mem_base + mem_len;
> >>  }
> >>  numamem = acpi_data_push(table_data, sizeof *numamem);
> >>  build_srat_memory(numamem, mem_base, mem_len, i - 1,  
> >
> >
> >
> >  
> 
> 

Re: [Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Dou Liyang]

At 09/04/2017 05:39 PM, Igor Mammedov wrote:

On Thu, 31 Aug 2017 20:04:26 +0800
Dou Liyang  wrote:


From: Eduardo Habkost 

Currently, Using the fisrt node without memory on the machine makes
QEMU unhappy. With this example command line:
  ... \
  -m 1024M,slots=4,maxmem=32G \
  -numa node,nodeid=0 \
  -numa node,mem=1024M,nodeid=1 \
  -numa node,nodeid=2 \
  -numa node,nodeid=3 \
Guest reports "No NUMA configuration found" and the NUMA topology is
wrong.

This is because when QEMU builds ACPI SRAT, it regards node 0 as the
default node to deal with the memory hole(640K-1M). this means the
node0 must have some memory(>1M), but, actually it can have no
memory.

Fix this problem by  cut out the 640K hole in the same way the PCI
4G hole does. Also do some cleanup.

Signed-off-by: Eduardo Habkost 
Signed-off-by: Dou Liyang 
---
 hw/i386/acpi-build.c | 30 +++---
 1 file changed, 23 insertions(+), 7 deletions(-)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 98dd424..48525a1 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -2318,6 +2318,9 @@ build_tpm2(GArray *table_data, BIOSLinker *linker)
  (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL, NULL);
 }

+#define HOLE_640K_START  (640 * 1024)
+#define HOLE_640K_END   (1024 * 1024)
+
 static void
 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
 {
@@ -2373,17 +2376,30 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
MachineState *machine)
 next_base = 0;
 numa_start = table_data->len;

-numamem = acpi_data_push(table_data, sizeof *numamem);
-build_srat_memory(numamem, 0, 640 * 1024, 0, MEM_AFFINITY_ENABLED);
-next_base = 1024 * 1024;
 for (i = 1; i < pcms->numa_nodes + 1; ++i) {
 mem_base = next_base;
 mem_len = pcms->node_mem[i - 1];
-if (i == 1) {
-mem_len -= 1024 * 1024;
-}
 next_base = mem_base + mem_len;

+/* Cut out the 640K hole */
+if (mem_base <= HOLE_640K_START &&
+next_base > HOLE_640K_START) {
+mem_len -= next_base - HOLE_640K_START;
+if (mem_len > 0) {
+numamem = acpi_data_push(table_data, sizeof *numamem);
+build_srat_memory(numamem, mem_base, mem_len, i - 1,
+  MEM_AFFINITY_ENABLED);
+}
+
+/* Check for the rare case: 640K < RAM < 1M */
+if (next_base <= HOLE_640K_END) {
+next_base = HOLE_640K_END;

Is this assignment really necessary?



It is necessary, because we set mem_base to next_base before setting
next_base;

But, I can refine it:

   MEM_AFFINITY_ENABLED);
 }

+mem_base = HOLE_640K_END;
 /* Check for the rare case: 640K < RAM < 1M */
 if (next_base <= HOLE_640K_END) {
-next_base = HOLE_640K_END;
 continue;
 }
-mem_base = HOLE_640K_END;
 mem_len = next_base - HOLE_640K_END;
 }

Is it?

Thanks,
dou.


it seems that next_base will be set at the start of the loop.





+continue;
+}
+mem_base = HOLE_640K_END;
+mem_len = next_base - HOLE_640K_END;
+}
+
 /* Cut out the ACPI_PCI hole */
 if (mem_base <= pcms->below_4g_mem_size &&
 next_base > pcms->below_4g_mem_size) {
@@ -2395,7 +2411,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
MachineState *machine)
 }
 mem_base = 1ULL << 32;
 mem_len = next_base - pcms->below_4g_mem_size;
-next_base += (1ULL << 32) - pcms->below_4g_mem_size;
+next_base = mem_base + mem_len;
 }
 numamem = acpi_data_push(table_data, sizeof *numamem);
 build_srat_memory(numamem, mem_base, mem_len, i - 1,

Re: [Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Igor Mammedov]

On Thu, 31 Aug 2017 20:04:26 +0800
Dou Liyang  wrote:

> From: Eduardo Habkost 
> 
> Currently, Using the fisrt node without memory on the machine makes
> QEMU unhappy. With this example command line:
>   ... \
>   -m 1024M,slots=4,maxmem=32G \
>   -numa node,nodeid=0 \
>   -numa node,mem=1024M,nodeid=1 \
>   -numa node,nodeid=2 \
>   -numa node,nodeid=3 \
> Guest reports "No NUMA configuration found" and the NUMA topology is
> wrong.
> 
> This is because when QEMU builds ACPI SRAT, it regards node 0 as the
> default node to deal with the memory hole(640K-1M). this means the
> node0 must have some memory(>1M), but, actually it can have no
> memory.
> 
> Fix this problem by  cut out the 640K hole in the same way the PCI
> 4G hole does. Also do some cleanup.
> 
> Signed-off-by: Eduardo Habkost 
> Signed-off-by: Dou Liyang 
> ---
>  hw/i386/acpi-build.c | 30 +++---
>  1 file changed, 23 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> index 98dd424..48525a1 100644
> --- a/hw/i386/acpi-build.c
> +++ b/hw/i386/acpi-build.c
> @@ -2318,6 +2318,9 @@ build_tpm2(GArray *table_data, BIOSLinker *linker)
>   (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL, NULL);
>  }
>  
> +#define HOLE_640K_START  (640 * 1024)
> +#define HOLE_640K_END   (1024 * 1024)
> +
>  static void
>  build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
>  {
> @@ -2373,17 +2376,30 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
> MachineState *machine)
>  next_base = 0;
>  numa_start = table_data->len;
>  
> -numamem = acpi_data_push(table_data, sizeof *numamem);
> -build_srat_memory(numamem, 0, 640 * 1024, 0, MEM_AFFINITY_ENABLED);
> -next_base = 1024 * 1024;
>  for (i = 1; i < pcms->numa_nodes + 1; ++i) {
>  mem_base = next_base;
>  mem_len = pcms->node_mem[i - 1];
> -if (i == 1) {
> -mem_len -= 1024 * 1024;
> -}
>  next_base = mem_base + mem_len;
>  
> +/* Cut out the 640K hole */
> +if (mem_base <= HOLE_640K_START &&
> +next_base > HOLE_640K_START) {
> +mem_len -= next_base - HOLE_640K_START;
> +if (mem_len > 0) {
> +numamem = acpi_data_push(table_data, sizeof *numamem);
> +build_srat_memory(numamem, mem_base, mem_len, i - 1,
> +  MEM_AFFINITY_ENABLED);
> +}
> +
> +/* Check for the rare case: 640K < RAM < 1M */
> +if (next_base <= HOLE_640K_END) {
> +next_base = HOLE_640K_END;
Is this assignment really necessary?

it seems that next_base will be set at the start of the loop.

> +continue;
> +}
> +mem_base = HOLE_640K_END;
> +mem_len = next_base - HOLE_640K_END;
> +}
> +
>  /* Cut out the ACPI_PCI hole */
>  if (mem_base <= pcms->below_4g_mem_size &&
>  next_base > pcms->below_4g_mem_size) {
> @@ -2395,7 +2411,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
> MachineState *machine)
>  }
>  mem_base = 1ULL << 32;
>  mem_len = next_base - pcms->below_4g_mem_size;
> -next_base += (1ULL << 32) - pcms->below_4g_mem_size;
> +next_base = mem_base + mem_len;
>  }
>  numamem = acpi_data_push(table_data, sizeof *numamem);
>  build_srat_memory(numamem, mem_base, mem_len, i - 1,

Re: [Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Dou Liyang]

Hi, Eduardo

At 09/01/2017 05:36 AM, Eduardo Habkost wrote:

On Thu, Aug 31, 2017 at 08:04:26PM +0800, Dou Liyang wrote:

From: Eduardo Habkost 

Currently, Using the fisrt node without memory on the machine makes
QEMU unhappy. With this example command line:
  ... \
  -m 1024M,slots=4,maxmem=32G \
  -numa node,nodeid=0 \
  -numa node,mem=1024M,nodeid=1 \
  -numa node,nodeid=2 \
  -numa node,nodeid=3 \
Guest reports "No NUMA configuration found" and the NUMA topology is
wrong.

This is because when QEMU builds ACPI SRAT, it regards node 0 as the
default node to deal with the memory hole(640K-1M). this means the
node0 must have some memory(>1M), but, actually it can have no
memory.

Fix this problem by  cut out the 640K hole in the same way the PCI
4G hole does. Also do some cleanup.

Signed-off-by: Eduardo Habkost 
Signed-off-by: Dou Liyang 
---
 hw/i386/acpi-build.c | 30 +++---
 1 file changed, 23 insertions(+), 7 deletions(-)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 98dd424..48525a1 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -2318,6 +2318,9 @@ build_tpm2(GArray *table_data, BIOSLinker *linker)
  (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL, NULL);
 }

+#define HOLE_640K_START  (640 * 1024)
+#define HOLE_640K_END   (1024 * 1024)
+
 static void
 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
 {
@@ -2373,17 +2376,30 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
MachineState *machine)
 next_base = 0;
 numa_start = table_data->len;

-numamem = acpi_data_push(table_data, sizeof *numamem);
-build_srat_memory(numamem, 0, 640 * 1024, 0, MEM_AFFINITY_ENABLED);
-next_base = 1024 * 1024;
 for (i = 1; i < pcms->numa_nodes + 1; ++i) {
 mem_base = next_base;
 mem_len = pcms->node_mem[i - 1];
-if (i == 1) {
-mem_len -= 1024 * 1024;
-}
 next_base = mem_base + mem_len;

+/* Cut out the 640K hole */
+if (mem_base <= HOLE_640K_START &&
+next_base > HOLE_640K_START) {
+mem_len -= next_base - HOLE_640K_START;
+if (mem_len > 0) {
+numamem = acpi_data_push(table_data, sizeof *numamem);
+build_srat_memory(numamem, mem_base, mem_len, i - 1,
+  MEM_AFFINITY_ENABLED);
+}
+
+/* Check for the rare case: 640K < RAM < 1M */
+if (next_base <= HOLE_640K_END) {
+next_base = HOLE_640K_END;
+continue;
+}
+mem_base = HOLE_640K_END;
+mem_len = next_base - HOLE_640K_END;
+}
+
 /* Cut out the ACPI_PCI hole */
 if (mem_base <= pcms->below_4g_mem_size &&
 next_base > pcms->below_4g_mem_size) {
@@ -2395,7 +2411,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
MachineState *machine)
 }
 mem_base = 1ULL << 32;
 mem_len = next_base - pcms->below_4g_mem_size;
-next_base += (1ULL << 32) - pcms->below_4g_mem_size;
+next_base = mem_base + mem_len;


Is this extra change intentional?



Yes, it is, Just for readability. :-)


I find the code more readable with it, but it should go in a
separate patch because it is unrelated to the bug fix.



Indeed, I will split it out.

Thanks,
dou.

Re: [Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Eduardo Habkost]

On Thu, Aug 31, 2017 at 08:04:26PM +0800, Dou Liyang wrote:
> From: Eduardo Habkost 
> 
> Currently, Using the fisrt node without memory on the machine makes
> QEMU unhappy. With this example command line:
>   ... \
>   -m 1024M,slots=4,maxmem=32G \
>   -numa node,nodeid=0 \
>   -numa node,mem=1024M,nodeid=1 \
>   -numa node,nodeid=2 \
>   -numa node,nodeid=3 \
> Guest reports "No NUMA configuration found" and the NUMA topology is
> wrong.
> 
> This is because when QEMU builds ACPI SRAT, it regards node 0 as the
> default node to deal with the memory hole(640K-1M). this means the
> node0 must have some memory(>1M), but, actually it can have no
> memory.
> 
> Fix this problem by  cut out the 640K hole in the same way the PCI
> 4G hole does. Also do some cleanup.
> 
> Signed-off-by: Eduardo Habkost 
> Signed-off-by: Dou Liyang 
> ---
>  hw/i386/acpi-build.c | 30 +++---
>  1 file changed, 23 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
> index 98dd424..48525a1 100644
> --- a/hw/i386/acpi-build.c
> +++ b/hw/i386/acpi-build.c
> @@ -2318,6 +2318,9 @@ build_tpm2(GArray *table_data, BIOSLinker *linker)
>   (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL, NULL);
>  }
>  
> +#define HOLE_640K_START  (640 * 1024)
> +#define HOLE_640K_END   (1024 * 1024)
> +
>  static void
>  build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
>  {
> @@ -2373,17 +2376,30 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
> MachineState *machine)
>  next_base = 0;
>  numa_start = table_data->len;
>  
> -numamem = acpi_data_push(table_data, sizeof *numamem);
> -build_srat_memory(numamem, 0, 640 * 1024, 0, MEM_AFFINITY_ENABLED);
> -next_base = 1024 * 1024;
>  for (i = 1; i < pcms->numa_nodes + 1; ++i) {
>  mem_base = next_base;
>  mem_len = pcms->node_mem[i - 1];
> -if (i == 1) {
> -mem_len -= 1024 * 1024;
> -}
>  next_base = mem_base + mem_len;
>  
> +/* Cut out the 640K hole */
> +if (mem_base <= HOLE_640K_START &&
> +next_base > HOLE_640K_START) {
> +mem_len -= next_base - HOLE_640K_START;
> +if (mem_len > 0) {
> +numamem = acpi_data_push(table_data, sizeof *numamem);
> +build_srat_memory(numamem, mem_base, mem_len, i - 1,
> +  MEM_AFFINITY_ENABLED);
> +}
> +
> +/* Check for the rare case: 640K < RAM < 1M */
> +if (next_base <= HOLE_640K_END) {
> +next_base = HOLE_640K_END;
> +continue;
> +}
> +mem_base = HOLE_640K_END;
> +mem_len = next_base - HOLE_640K_END;
> +}
> +
>  /* Cut out the ACPI_PCI hole */
>  if (mem_base <= pcms->below_4g_mem_size &&
>  next_base > pcms->below_4g_mem_size) {
> @@ -2395,7 +2411,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
> MachineState *machine)
>  }
>  mem_base = 1ULL << 32;
>  mem_len = next_base - pcms->below_4g_mem_size;
> -next_base += (1ULL << 32) - pcms->below_4g_mem_size;
> +next_base = mem_base + mem_len;

Is this extra change intentional?

I find the code more readable with it, but it should go in a
separate patch because it is unrelated to the bug fix.

-- 
Eduardo

[Qemu-devel] [PATCH v5 1/3] hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

[Dou Liyang]

From: Eduardo Habkost 

Currently, Using the fisrt node without memory on the machine makes
QEMU unhappy. With this example command line:
  ... \
  -m 1024M,slots=4,maxmem=32G \
  -numa node,nodeid=0 \
  -numa node,mem=1024M,nodeid=1 \
  -numa node,nodeid=2 \
  -numa node,nodeid=3 \
Guest reports "No NUMA configuration found" and the NUMA topology is
wrong.

This is because when QEMU builds ACPI SRAT, it regards node 0 as the
default node to deal with the memory hole(640K-1M). this means the
node0 must have some memory(>1M), but, actually it can have no
memory.

Fix this problem by  cut out the 640K hole in the same way the PCI
4G hole does. Also do some cleanup.

Signed-off-by: Eduardo Habkost 
Signed-off-by: Dou Liyang 
---
 hw/i386/acpi-build.c | 30 +++---
 1 file changed, 23 insertions(+), 7 deletions(-)

diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index 98dd424..48525a1 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -2318,6 +2318,9 @@ build_tpm2(GArray *table_data, BIOSLinker *linker)
  (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL, NULL);
 }
 
+#define HOLE_640K_START  (640 * 1024)
+#define HOLE_640K_END   (1024 * 1024)
+
 static void
 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
 {
@@ -2373,17 +2376,30 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
MachineState *machine)
 next_base = 0;
 numa_start = table_data->len;
 
-numamem = acpi_data_push(table_data, sizeof *numamem);
-build_srat_memory(numamem, 0, 640 * 1024, 0, MEM_AFFINITY_ENABLED);
-next_base = 1024 * 1024;
 for (i = 1; i < pcms->numa_nodes + 1; ++i) {
 mem_base = next_base;
 mem_len = pcms->node_mem[i - 1];
-if (i == 1) {
-mem_len -= 1024 * 1024;
-}
 next_base = mem_base + mem_len;
 
+/* Cut out the 640K hole */
+if (mem_base <= HOLE_640K_START &&
+next_base > HOLE_640K_START) {
+mem_len -= next_base - HOLE_640K_START;
+if (mem_len > 0) {
+numamem = acpi_data_push(table_data, sizeof *numamem);
+build_srat_memory(numamem, mem_base, mem_len, i - 1,
+  MEM_AFFINITY_ENABLED);
+}
+
+/* Check for the rare case: 640K < RAM < 1M */
+if (next_base <= HOLE_640K_END) {
+next_base = HOLE_640K_END;
+continue;
+}
+mem_base = HOLE_640K_END;
+mem_len = next_base - HOLE_640K_END;
+}
+
 /* Cut out the ACPI_PCI hole */
 if (mem_base <= pcms->below_4g_mem_size &&
 next_base > pcms->below_4g_mem_size) {
@@ -2395,7 +2411,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, 
MachineState *machine)
 }
 mem_base = 1ULL << 32;
 mem_len = next_base - pcms->below_4g_mem_size;
-next_base += (1ULL << 32) - pcms->below_4g_mem_size;
+next_base = mem_base + mem_len;
 }
 numamem = acpi_data_push(table_data, sizeof *numamem);
 build_srat_memory(numamem, mem_base, mem_len, i - 1,
-- 
2.5.5