[PATCH] x86/acpi: keep x86_cpu_to_acpiid mapping valid on cpu hotplug

[PATCH] x86/acpi: keep x86_cpu_to_acpiid mapping valid on cpu hotplug

[Vitaly Kuznetsov]

We may or may not have all possible CPUs in MADT on boot but in any case
we're overwriting x86_cpu_to_acpiid mapping with U32_MAX when
acpi_register_lapic() is called again on the CPU hotplug path:
acpi_processor_hotadd_init() -> acpi_map_cpu() -> acpi_register_lapic().

As we have the required acpi_id information in acpi_processor_hotadd_init()
propagate it to acpi_map_cpu() to always keep x86_cpu_to_acpiid mapping
valid.

Reported-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
---
 arch/ia64/kernel/acpi.c       | 3 ++-
 arch/x86/kernel/acpi/boot.c   | 5 +++--
 drivers/acpi/acpi_processor.c | 4 ++--
 include/linux/acpi.h          | 3 ++-
 4 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/arch/ia64/kernel/acpi.c b/arch/ia64/kernel/acpi.c
index 9273e03..7508c30 100644
--- a/arch/ia64/kernel/acpi.c
+++ b/arch/ia64/kernel/acpi.c
@@ -887,7 +887,8 @@ static int _acpi_map_lsapic(acpi_handle handle, int physid, int *pcpu)
 }
 
 /* wrapper to silence section mismatch warning */
-int __ref acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, int *pcpu)
+int __ref acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, u32 acpi_id,
+		       int *pcpu)
 {
 	return _acpi_map_lsapic(handle, physid, pcpu);
 }
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 64422f8..04bc5f3 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -723,11 +723,12 @@ int acpi_map_cpu2node(acpi_handle handle, int cpu, int physid)
 	return 0;
 }
 
-int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, int *pcpu)
+int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, u32 acpi_id,
+		 int *pcpu)
 {
 	int cpu;
 
-	cpu = acpi_register_lapic(physid, U32_MAX, ACPI_MADT_ENABLED);
+	cpu = acpi_register_lapic(physid, acpi_id, ACPI_MADT_ENABLED);
 	if (cpu < 0) {
 		pr_info(PREFIX "Unable to map lapic to logical cpu number\n");
 		return cpu;
diff --git a/drivers/acpi/acpi_processor.c b/drivers/acpi/acpi_processor.c
index 3de3b6b..4467a80 100644
--- a/drivers/acpi/acpi_processor.c
+++ b/drivers/acpi/acpi_processor.c
@@ -165,7 +165,7 @@ static int acpi_processor_errata(void)
 
 #ifdef CONFIG_ACPI_HOTPLUG_CPU
 int __weak acpi_map_cpu(acpi_handle handle,
-		phys_cpuid_t physid, int *pcpu)
+		phys_cpuid_t physid, u32 acpi_id, int *pcpu)
 {
 	return -ENODEV;
 }
@@ -203,7 +203,7 @@ static int acpi_processor_hotadd_init(struct acpi_processor *pr)
 	cpu_maps_update_begin();
 	cpu_hotplug_begin();
 
-	ret = acpi_map_cpu(pr->handle, pr->phys_id, &pr->id);
+	ret = acpi_map_cpu(pr->handle, pr->phys_id, pr->acpi_id, &pr->id);
 	if (ret)
 		goto out;
 
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index 5b36974..6ab47e9 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -291,7 +291,8 @@ bool acpi_processor_validate_proc_id(int proc_id);
 
 #ifdef CONFIG_ACPI_HOTPLUG_CPU
 /* Arch dependent functions for cpu hotplug support */
-int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, int *pcpu);
+int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, u32 acpi_id,
+		 int *pcpu);
 int acpi_unmap_cpu(int cpu);
 int acpi_map_cpu2node(acpi_handle handle, int cpu, int physid);
 #endif /* CONFIG_ACPI_HOTPLUG_CPU */
-- 
2.9.3

Re: [PATCH] x86/acpi: keep x86_cpu_to_acpiid mapping valid on cpu hotplug

[Rafael J. Wysocki]

On Mon, Feb 6, 2017 at 6:01 PM, Vitaly Kuznetsov <vkuznets@redhat.com> wrote:
> We may or may not have all possible CPUs in MADT on boot but in any case
> we're overwriting x86_cpu_to_acpiid mapping with U32_MAX when
> acpi_register_lapic() is called again on the CPU hotplug path:
> acpi_processor_hotadd_init() -> acpi_map_cpu() -> acpi_register_lapic().
>
> As we have the required acpi_id information in acpi_processor_hotadd_init()
> propagate it to acpi_map_cpu() to always keep x86_cpu_to_acpiid mapping
> valid.
>
> Reported-by: Andrew Jones <drjones@redhat.com>
> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>

Is the bug report available and if so, do you have a pointer to it?

> ---
>  arch/ia64/kernel/acpi.c       | 3 ++-
>  arch/x86/kernel/acpi/boot.c   | 5 +++--
>  drivers/acpi/acpi_processor.c | 4 ++--
>  include/linux/acpi.h          | 3 ++-
>  4 files changed, 9 insertions(+), 6 deletions(-)
>
> diff --git a/arch/ia64/kernel/acpi.c b/arch/ia64/kernel/acpi.c
> index 9273e03..7508c30 100644
> --- a/arch/ia64/kernel/acpi.c
> +++ b/arch/ia64/kernel/acpi.c
> @@ -887,7 +887,8 @@ static int _acpi_map_lsapic(acpi_handle handle, int physid, int *pcpu)
>  }
>
>  /* wrapper to silence section mismatch warning */
> -int __ref acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, int *pcpu)
> +int __ref acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, u32 acpi_id,
> +                      int *pcpu)
>  {
>         return _acpi_map_lsapic(handle, physid, pcpu);
>  }
> diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
> index 64422f8..04bc5f3 100644
> --- a/arch/x86/kernel/acpi/boot.c
> +++ b/arch/x86/kernel/acpi/boot.c
> @@ -723,11 +723,12 @@ int acpi_map_cpu2node(acpi_handle handle, int cpu, int physid)
>         return 0;
>  }
>
> -int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, int *pcpu)
> +int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, u32 acpi_id,
> +                int *pcpu)
>  {
>         int cpu;
>
> -       cpu = acpi_register_lapic(physid, U32_MAX, ACPI_MADT_ENABLED);
> +       cpu = acpi_register_lapic(physid, acpi_id, ACPI_MADT_ENABLED);
>         if (cpu < 0) {
>                 pr_info(PREFIX "Unable to map lapic to logical cpu number\n");
>                 return cpu;
> diff --git a/drivers/acpi/acpi_processor.c b/drivers/acpi/acpi_processor.c
> index 3de3b6b..4467a80 100644
> --- a/drivers/acpi/acpi_processor.c
> +++ b/drivers/acpi/acpi_processor.c
> @@ -165,7 +165,7 @@ static int acpi_processor_errata(void)
>
>  #ifdef CONFIG_ACPI_HOTPLUG_CPU
>  int __weak acpi_map_cpu(acpi_handle handle,
> -               phys_cpuid_t physid, int *pcpu)
> +               phys_cpuid_t physid, u32 acpi_id, int *pcpu)
>  {
>         return -ENODEV;
>  }
> @@ -203,7 +203,7 @@ static int acpi_processor_hotadd_init(struct acpi_processor *pr)
>         cpu_maps_update_begin();
>         cpu_hotplug_begin();
>
> -       ret = acpi_map_cpu(pr->handle, pr->phys_id, &pr->id);
> +       ret = acpi_map_cpu(pr->handle, pr->phys_id, pr->acpi_id, &pr->id);
>         if (ret)
>                 goto out;
>
> diff --git a/include/linux/acpi.h b/include/linux/acpi.h
> index 5b36974..6ab47e9 100644
> --- a/include/linux/acpi.h
> +++ b/include/linux/acpi.h
> @@ -291,7 +291,8 @@ bool acpi_processor_validate_proc_id(int proc_id);
>
>  #ifdef CONFIG_ACPI_HOTPLUG_CPU
>  /* Arch dependent functions for cpu hotplug support */
> -int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, int *pcpu);
> +int acpi_map_cpu(acpi_handle handle, phys_cpuid_t physid, u32 acpi_id,
> +                int *pcpu);
>  int acpi_unmap_cpu(int cpu);
>  int acpi_map_cpu2node(acpi_handle handle, int cpu, int physid);
>  #endif /* CONFIG_ACPI_HOTPLUG_CPU */
> --

Thanks,
Rafael

Re: [PATCH] x86/acpi: keep x86_cpu_to_acpiid mapping valid on cpu hotplug

[Vitaly Kuznetsov]

"Rafael J. Wysocki" <rafael@kernel.org> writes:

> On Mon, Feb 6, 2017 at 6:01 PM, Vitaly Kuznetsov <vkuznets@redhat.com> wrote:
>> We may or may not have all possible CPUs in MADT on boot but in any case
>> we're overwriting x86_cpu_to_acpiid mapping with U32_MAX when
>> acpi_register_lapic() is called again on the CPU hotplug path:
>> acpi_processor_hotadd_init() -> acpi_map_cpu() -> acpi_register_lapic().
>>
>> As we have the required acpi_id information in acpi_processor_hotadd_init()
>> propagate it to acpi_map_cpu() to always keep x86_cpu_to_acpiid mapping
>> valid.
>>
>> Reported-by: Andrew Jones <drjones@redhat.com>
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>
> Is the bug report available and if so, do you have a pointer to it?
>

There was no 'real' bug, the issue was found by code inspection.
Potentially, the following scenario is currently broken:
- boot Xen HVM guest and do kdump on a secondary CPU
- with kdump kernel running try to hot-plug additional CPUs.
we'll end up with U32_MAX in x86_cpu_to_acpiid for hot-plugged CPUs so
we'll be assuming direct Linux<->Xen mapping which won't be the case and
all hypercalls referring to these CPUs will end up affecting some other
CPUs.

-- 
  Vitaly