Newer ACPI revisions define the MADT GICC entry with Length = 82 bytes [1].
The current BAD_MADT_GICC_ENTRY() check rejects entries whose length does not
match the known values, which leads to:
  GICv3: No valid GICC entries exist.
as observed on the AmpereOne platform.

To fix this, import the logic from Linux commit 9eb1c92b47c7:
  The BAD_MADT_GICC_ENTRY check is a little too strict because
  it rejects MADT entries that don't match the currently known
  lengths. We should remove this restriction to avoid problems
  if the table length changes. Future code which might depend on
  additional fields should be written to validate those fields
  before using them, rather than trying to globally check
  known MADT version lengths.

  Link: https://lkml.kernel.org/r/20181012192937.3819951-1-jeremy.linton@arm.com
  Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
  [lorenzo.pieralisi@arm.com: added MADT macro comments]
  Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
  Acked-by: Sudeep Holla <sudeep.holla@arm.com>
  Cc: Will Deacon <will.deacon@arm.com>
  Cc: Catalin Marinas <catalin.marinas@arm.com>
  Cc: Al Stone <ahs3@redhat.com>
  Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
  Signed-off-by: Will Deacon <will.deacon@arm.com>

As ACPI_MADT_GICC_LENGTH is dropped, update the functions where it is
used. As we rewrite the MADT for hwdom, reuse the host GICC header length
instead of ACPI_MADT_GICC_LENGTH.

[1] https://uefi.org/specs/ACPI/6.6/05_ACPI_Software_Programming_Model.html#gic-cpu-interface-gicc-structure

Reported-By: Yann Dirson <yann.dirson@vates.tech>
Co-developed-by: Yann Sionneau <yann.sionneau@vates.tech>
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
---
I ran CI tests where it made sense for this patch, as I don’t see any CI job
that builds Xen with CONFIG_ACPI=y:
  https://gitlab.com/xen-project/people/olkur/xen/-/pipelines/2229673951

I also built Xen manually with CONFIG_ACPI=y enabled and tested it on the
AmpereOne platform.
---
Changes in v2:
 - Update the commit message:
   - Use more characters for commit ID.
   - Drop 'import from'.
 - Add Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>.
 - Make the local variables host_gicc const in  gic_get_hwdom_madt_size().
   (header variable isn't const as container_of() will discard 'const' qualifier
   and so compilation error will occur).
 - Return 0 instead of panic() in gic_get_hwdom_madt_size().
---
 xen/arch/arm/acpi/domain_build.c |  6 ++++++
 xen/arch/arm/gic-v2.c            |  3 ++-
 xen/arch/arm/gic-v3.c            |  3 ++-
 xen/arch/arm/gic.c               | 12 +++++++++++-
 xen/arch/arm/include/asm/acpi.h  | 21 +++++++++++++++------
 5 files changed, 36 insertions(+), 9 deletions(-)

diff --git a/xen/arch/arm/acpi/domain_build.c b/xen/arch/arm/acpi/domain_build.c
index 1c3555d814cc..959698d13ac3 100644
--- a/xen/arch/arm/acpi/domain_build.c
+++ b/xen/arch/arm/acpi/domain_build.c
@@ -458,6 +458,12 @@ static int __init estimate_acpi_efi_size(struct domain *d,
     acpi_size += ROUNDUP(sizeof(struct acpi_table_stao), 8);
 
     madt_size = gic_get_hwdom_madt_size(d);
+    if ( !madt_size )
+    {
+        printk("Unable to get hwdom MADT size\n");
+        return -EINVAL;
+    }
+
     acpi_size += ROUNDUP(madt_size, 8);
 
     addr = acpi_os_get_root_pointer();
diff --git a/xen/arch/arm/gic-v2.c b/xen/arch/arm/gic-v2.c
index b23e72a3d05d..aae6a7bf3076 100644
--- a/xen/arch/arm/gic-v2.c
+++ b/xen/arch/arm/gic-v2.c
@@ -1121,7 +1121,8 @@ static int gicv2_make_hwdom_madt(const struct domain *d, u32 offset)
     host_gicc = container_of(header, struct acpi_madt_generic_interrupt,
                              header);
 
-    size = ACPI_MADT_GICC_LENGTH;
+    size = host_gicc->header.length;
+
     /* Add Generic Interrupt */
     for ( i = 0; i < d->max_vcpus; i++ )
     {
diff --git a/xen/arch/arm/gic-v3.c b/xen/arch/arm/gic-v3.c
index bc07f97c16ab..75b89efad462 100644
--- a/xen/arch/arm/gic-v3.c
+++ b/xen/arch/arm/gic-v3.c
@@ -1672,7 +1672,8 @@ static int gicv3_make_hwdom_madt(const struct domain *d, u32 offset)
 
     host_gicc = container_of(header, struct acpi_madt_generic_interrupt,
                              header);
-    size = ACPI_MADT_GICC_LENGTH;
+    size = host_gicc->header.length;
+
     for ( i = 0; i < d->max_vcpus; i++ )
     {
         gicc = (struct acpi_madt_generic_interrupt *)(base_ptr + table_len);
diff --git a/xen/arch/arm/gic.c b/xen/arch/arm/gic.c
index ee75258fc3c3..e22feb46f5d4 100644
--- a/xen/arch/arm/gic.c
+++ b/xen/arch/arm/gic.c
@@ -418,8 +418,18 @@ unsigned long gic_get_hwdom_madt_size(const struct domain *d)
 {
     unsigned long madt_size;
 
+    struct acpi_subtable_header *header;
+    const struct acpi_madt_generic_interrupt *host_gicc;
+
+    header = acpi_table_get_entry_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT, 0);
+    if ( !header )
+        return 0;
+
+    host_gicc = container_of(header, struct acpi_madt_generic_interrupt,
+                             header);
+
     madt_size = sizeof(struct acpi_table_madt)
-                + ACPI_MADT_GICC_LENGTH * d->max_vcpus
+                + host_gicc->header.length * d->max_vcpus
                 + sizeof(struct acpi_madt_generic_distributor)
                 + gic_hw_ops->get_hwdom_extra_madt_size(d);
 
diff --git a/xen/arch/arm/include/asm/acpi.h b/xen/arch/arm/include/asm/acpi.h
index 13756dd341b4..30bc446d1f75 100644
--- a/xen/arch/arm/include/asm/acpi.h
+++ b/xen/arch/arm/include/asm/acpi.h
@@ -53,13 +53,22 @@ void acpi_smp_init_cpus(void);
  */
 paddr_t acpi_get_table_offset(struct membank tbl_add[], EFI_MEM_RES index);
 
-/* Macros for consistency checks of the GICC subtable of MADT */
-#define ACPI_MADT_GICC_LENGTH	\
-    (acpi_gbl_FADT.header.revision < 6 ? 76 : 80)
+/*
+ * MADT GICC minimum length refers to the MADT GICC structure table length as
+ * defined in the earliest ACPI version supported on arm64, ie ACPI 5.1.
+ *
+ * The efficiency_class member was added to the
+ * struct acpi_madt_generic_interrupt to represent the MADT GICC structure
+ * "Processor Power Efficiency Class" field, added in ACPI 6.0 whose offset
+ * is therefore used to delimit the MADT GICC structure minimum length
+ * appropriately.
+ */
+#define ACPI_MADT_GICC_MIN_LENGTH   ACPI_OFFSET( \
+    struct acpi_madt_generic_interrupt, efficiency_class)
 
-#define BAD_MADT_GICC_ENTRY(entry, end)						\
-    (!(entry) || (unsigned long)(entry) + sizeof(*(entry)) > (end) ||	\
-     (entry)->header.length != ACPI_MADT_GICC_LENGTH)
+#define BAD_MADT_GICC_ENTRY(entry, end) \
+    (!(entry) || (entry)->header.length < ACPI_MADT_GICC_MIN_LENGTH || \
+    (unsigned long)(entry) + (entry)->header.length > (end))
 
 #ifdef CONFIG_ACPI
 extern bool acpi_disabled;
-- 
2.52.0

On 07.01.2026 17:34, Oleksii Kurochko wrote:
> Newer ACPI revisions define the MADT GICC entry with Length = 82 bytes [1].
> The current BAD_MADT_GICC_ENTRY() check rejects entries whose length does not
> match the known values, which leads to:
>   GICv3: No valid GICC entries exist.
> as observed on the AmpereOne platform.
> 
> To fix this, import the logic from Linux commit 9eb1c92b47c7:
>   The BAD_MADT_GICC_ENTRY check is a little too strict because
>   it rejects MADT entries that don't match the currently known
>   lengths. We should remove this restriction to avoid problems
>   if the table length changes. Future code which might depend on
>   additional fields should be written to validate those fields
>   before using them, rather than trying to globally check
>   known MADT version lengths.
> 
>   Link: https://lkml.kernel.org/r/20181012192937.3819951-1-jeremy.linton@arm.com
>   Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>   [lorenzo.pieralisi@arm.com: added MADT macro comments]
>   Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>   Acked-by: Sudeep Holla <sudeep.holla@arm.com>
>   Cc: Will Deacon <will.deacon@arm.com>
>   Cc: Catalin Marinas <catalin.marinas@arm.com>
>   Cc: Al Stone <ahs3@redhat.com>
>   Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
>   Signed-off-by: Will Deacon <will.deacon@arm.com>
> 
> As ACPI_MADT_GICC_LENGTH is dropped, update the functions where it is
> used. As we rewrite the MADT for hwdom, reuse the host GICC header length
> instead of ACPI_MADT_GICC_LENGTH.
> 
> [1] https://uefi.org/specs/ACPI/6.6/05_ACPI_Software_Programming_Model.html#gic-cpu-interface-gicc-structure
> 
> Reported-By: Yann Dirson <yann.dirson@vates.tech>
> Co-developed-by: Yann Sionneau <yann.sionneau@vates.tech>
> Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
> Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
> ---
> I ran CI tests where it made sense for this patch, as I don’t see any CI job
> that builds Xen with CONFIG_ACPI=y:
>   https://gitlab.com/xen-project/people/olkur/xen/-/pipelines/2229673951
> 
> I also built Xen manually with CONFIG_ACPI=y enabled and tested it on the
> AmpereOne platform.
> ---
> Changes in v2:
>  - Update the commit message:
>    - Use more characters for commit ID.
>    - Drop 'import from'.
>  - Add Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>.

Was this a legitimate thing to do, considering ...

>  - Make the local variables host_gicc const in  gic_get_hwdom_madt_size().
>    (header variable isn't const as container_of() will discard 'const' qualifier
>    and so compilation error will occur).
>  - Return 0 instead of panic() in gic_get_hwdom_madt_size().

... all of these (plus leaving partly unaddressed a comment from Julien)?

> --- a/xen/arch/arm/gic.c
> +++ b/xen/arch/arm/gic.c
> @@ -418,8 +418,18 @@ unsigned long gic_get_hwdom_madt_size(const struct domain *d)
>  {
>      unsigned long madt_size;
>  
> +    struct acpi_subtable_header *header;

Why is there a blank line left between declarations? In unusual situations (very
many variables, for example) that may be okay, but otherwise the first blank line
generally wants to separate decls from statements.

Also Julien asked for this to be const. You claimed a compile error would occur
if you do, but afaict that's only because ...

> +    const struct acpi_madt_generic_interrupt *host_gicc;
> +
> +    header = acpi_table_get_entry_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT, 0);
> +    if ( !header )
> +        return 0;
> +
> +    host_gicc = container_of(header, struct acpi_madt_generic_interrupt,
> +                             header);

... you don't use const properly here as well.

Finally (possibly not for this patch, but mentioning since originally it was
pointed out as an option) the function imo wants to become __init anyway, for
(as said by Julien) its only caller being so.

Jan

On 1/8/26 9:41 AM, Jan Beulich wrote:
> On 07.01.2026 17:34, Oleksii Kurochko wrote:
>> Newer ACPI revisions define the MADT GICC entry with Length = 82 bytes [1].
>> The current BAD_MADT_GICC_ENTRY() check rejects entries whose length does not
>> match the known values, which leads to:
>>    GICv3: No valid GICC entries exist.
>> as observed on the AmpereOne platform.
>>
>> To fix this, import the logic from Linux commit 9eb1c92b47c7:
>>    The BAD_MADT_GICC_ENTRY check is a little too strict because
>>    it rejects MADT entries that don't match the currently known
>>    lengths. We should remove this restriction to avoid problems
>>    if the table length changes. Future code which might depend on
>>    additional fields should be written to validate those fields
>>    before using them, rather than trying to globally check
>>    known MADT version lengths.
>>
>>    Link: https://lkml.kernel.org/r/20181012192937.3819951-1-jeremy.linton@arm.com
>>    Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>>    [lorenzo.pieralisi@arm.com: added MADT macro comments]
>>    Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>>    Acked-by: Sudeep Holla <sudeep.holla@arm.com>
>>    Cc: Will Deacon <will.deacon@arm.com>
>>    Cc: Catalin Marinas <catalin.marinas@arm.com>
>>    Cc: Al Stone <ahs3@redhat.com>
>>    Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
>>    Signed-off-by: Will Deacon <will.deacon@arm.com>
>>
>> As ACPI_MADT_GICC_LENGTH is dropped, update the functions where it is
>> used. As we rewrite the MADT for hwdom, reuse the host GICC header length
>> instead of ACPI_MADT_GICC_LENGTH.
>>
>> [1] https://uefi.org/specs/ACPI/6.6/05_ACPI_Software_Programming_Model.html#gic-cpu-interface-gicc-structure
>>
>> Reported-By: Yann Dirson <yann.dirson@vates.tech>
>> Co-developed-by: Yann Sionneau <yann.sionneau@vates.tech>
>> Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
>> Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
>> ---
>> I ran CI tests where it made sense for this patch, as I don’t see any CI job
>> that builds Xen with CONFIG_ACPI=y:
>>    https://gitlab.com/xen-project/people/olkur/xen/-/pipelines/2229673951
>>
>> I also built Xen manually with CONFIG_ACPI=y enabled and tested it on the
>> AmpereOne platform.
>> ---
>> Changes in v2:
>>   - Update the commit message:
>>     - Use more characters for commit ID.
>>     - Drop 'import from'.
>>   - Add Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>.
> Was this a legitimate thing to do, considering ...
>
>>   - Make the local variables host_gicc const in  gic_get_hwdom_madt_size().
>>     (header variable isn't const as container_of() will discard 'const' qualifier
>>     and so compilation error will occur).
>>   - Return 0 instead of panic() in gic_get_hwdom_madt_size().
> ... all of these (plus leaving partly unaddressed a comment from Julien)?

Probably you are right, it was to early. I will drop it and ask Stefano to re-review.

>
>> --- a/xen/arch/arm/gic.c
>> +++ b/xen/arch/arm/gic.c
>> @@ -418,8 +418,18 @@ unsigned long gic_get_hwdom_madt_size(const struct domain *d)
>>   {
>>       unsigned long madt_size;
>>   
>> +    struct acpi_subtable_header *header;
> Why is there a blank line left between declarations? In unusual situations (very
> many variables, for example) that may be okay, but otherwise the first blank line
> generally wants to separate decls from statements.

It was visually better for me to have header and host_gicc variable in a "separate
block". I can drop a blank line.

>
> Also Julien asked for this to be const. You claimed a compile error would occur
> if you do, but afaict that's only because ...
>
>> +    const struct acpi_madt_generic_interrupt *host_gicc;
>> +
>> +    header = acpi_table_get_entry_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT, 0);
>> +    if ( !header )
>> +        return 0;
>> +
>> +    host_gicc = container_of(header, struct acpi_madt_generic_interrupt,
>> +                             header);
> ... you don't use const properly here as well.

Oh, right, I missed to add const for the second argument of container_of().

>
> Finally (possibly not for this patch, but mentioning since originally it was
> pointed out as an option) the function imo wants to become __init anyway, for
> (as said by Julien) its only caller being so.

I didn't put __init here as you mentioned it was pointed out as an option. Anyway
I think I will put __init in the next patch version.

Thanks.

~ Oleksii