[v2,1/3] Revert "arm64: topology: divorce MC scheduling domain from core_siblings"

Message ID 1528209304-3280-1-git-send-email-sudeep.holla@arm.com
State New
Headers show
Series
  • [v2,1/3] Revert "arm64: topology: divorce MC scheduling domain from core_siblings"
Related show

Commit Message

Sudeep Holla June 5, 2018, 2:35 p.m.
This reverts commit 37c3ec2d810f87eac73822f76b30391a83bded19.

Currently on ARM64 platforms, we don't update the CPU topology masks
on each hotplug operation. However, the updates to cpu_coregroup_mask
done as part of ACPI PPTT support, in particular the commit being
reverted makes use of cpumask_of_node which returns the cpu_oneline_mask
instead of core_sibling as core_sibling masks are not updated for CPU
hotplug operations and the comparision to find NUMA in package or LLC
siblings fails.

This often leads to system hang or crash during CPU hotplug and system
suspend operation. This is mostly observed on HMP systems where the
CPU compute capacities are different and ends up in different scheduler
domains. Since cpumask_of_node is returned instead core_sibling, the
scheduler is confused with incorrect cpumasks(e.g. one CPU in two
different sched domains at the same time) on CPU hotplug.

The original commit is technically correct and since it depends on the
not yet supported feature, let's revert this for now. We can put it back
once we have the support for CPU topology masks update on hotplug merged.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

---
 arch/arm64/include/asm/topology.h |  2 --
 arch/arm64/kernel/topology.c      | 36 +-----------------------------------
 2 files changed, 1 insertion(+), 37 deletions(-)

v1->v2:
	- Updated commit log to describe the observations made as a
	  consequence of the issue as suggested by Geert's

-- 
2.7.4

--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Rafael J. Wysocki June 5, 2018, 3 p.m. | #1
On Tue, Jun 5, 2018 at 4:35 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:
> Though CONFIG_ACPI_PPTT is selected by platforms and nor user visible,

> it may be useful to support the build with CONFIG_ACPI_PPTT disabled.

>

> This patch adds the missing dummy/boiler plate implementation to fix

> the build.

>

> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>

> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

> ---

>  include/linux/acpi.h      | 15 +++++++++++++++

>  include/linux/cacheinfo.h |  2 +-

>  2 files changed, 16 insertions(+), 1 deletion(-)

>

> Hi Rafael,

>

> If you are fine with this, can you provide Ack, so that we route this

> through ARM64 tree where most of the ACPI PPTT support is present.

>

> Regards,

> Sudeep

>

> v1->v2:

>         - removed duplicate definition for acpi_find_last_cache_level

>

> diff --git a/include/linux/acpi.h b/include/linux/acpi.h

> index 8f2cdb0eca71..4b35a66383f9 100644

> --- a/include/linux/acpi.h

> +++ b/include/linux/acpi.h

> @@ -1299,8 +1299,23 @@ static inline int lpit_read_residency_count_address(u64 *address)

>  }

>  #endif

>

> +#ifdef CONFIG_ACPI_PPTT

>  int find_acpi_cpu_topology(unsigned int cpu, int level);

>  int find_acpi_cpu_topology_package(unsigned int cpu);

>  int find_acpi_cpu_cache_topology(unsigned int cpu, int level);

> +#else

> +static inline int find_acpi_cpu_topology(unsigned int cpu, int level)

> +{

> +       return -EINVAL;


Why -EINVAL?

> +}

> +static inline int find_acpi_cpu_topology_package(unsigned int cpu)

> +{

> +       return -EINVAL;

> +}

> +static inline int find_acpi_cpu_cache_topology(unsigned int cpu, int level)

> +{

> +       return -EINVAL;

> +}

> +#endif

>

>  #endif /*_LINUX_ACPI_H*/

> diff --git a/include/linux/cacheinfo.h b/include/linux/cacheinfo.h

> index 89397e30e269..70e19bc6cc9f 100644

> --- a/include/linux/cacheinfo.h

> +++ b/include/linux/cacheinfo.h

> @@ -98,7 +98,7 @@ struct cpu_cacheinfo *get_cpu_cacheinfo(unsigned int cpu);

>  int init_cache_level(unsigned int cpu);

>  int populate_cache_leaves(unsigned int cpu);

>  int cache_setup_acpi(unsigned int cpu);

> -#ifndef CONFIG_ACPI

> +#ifndef CONFIG_ACPI_PPTT

>  /*

>   * acpi_find_last_cache_level is only called on ACPI enabled

>   * platforms using the PPTT for topology. This means that if

> --

> 2.7.4

>

> --

> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in

> the body of a message to majordomo@vger.kernel.org

> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Rafael J. Wysocki June 5, 2018, 4:12 p.m. | #2
On Tue, Jun 5, 2018 at 5:33 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:
>

>

> On 05/06/18 16:00, Rafael J. Wysocki wrote:

>> On Tue, Jun 5, 2018 at 4:35 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:

>>> Though CONFIG_ACPI_PPTT is selected by platforms and nor user visible,

>>> it may be useful to support the build with CONFIG_ACPI_PPTT disabled.

>>>

>>> This patch adds the missing dummy/boiler plate implementation to fix

>>> the build.

>>>

>>> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>

>>> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

>>> ---

>>>  include/linux/acpi.h      | 15 +++++++++++++++

>>>  include/linux/cacheinfo.h |  2 +-

>>>  2 files changed, 16 insertions(+), 1 deletion(-)

>>>

>>> Hi Rafael,

>>>

>>> If you are fine with this, can you provide Ack, so that we route this

>>> through ARM64 tree where most of the ACPI PPTT support is present.

>>>

>>> Regards,

>>> Sudeep

>>>

>>> v1->v2:

>>>         - removed duplicate definition for acpi_find_last_cache_level

>>>

>>> diff --git a/include/linux/acpi.h b/include/linux/acpi.h

>>> index 8f2cdb0eca71..4b35a66383f9 100644

>>> --- a/include/linux/acpi.h

>>> +++ b/include/linux/acpi.h

>>> @@ -1299,8 +1299,23 @@ static inline int lpit_read_residency_count_address(u64 *address)

>>>  }

>>>  #endif

>>>

>>> +#ifdef CONFIG_ACPI_PPTT

>>>  int find_acpi_cpu_topology(unsigned int cpu, int level);

>>>  int find_acpi_cpu_topology_package(unsigned int cpu);

>>>  int find_acpi_cpu_cache_topology(unsigned int cpu, int level);

>>> +#else

>>> +static inline int find_acpi_cpu_topology(unsigned int cpu, int level)

>>> +{

>>> +       return -EINVAL;

>>

>> Why -EINVAL?

>>

>

> I am not sure either. I used to return -ENOTSUPP, but IIRC someone

> suggested to use it only for syscalls. Also I just based it on other

> existing functions in acpi.h

>

> I am open for any alternatives if you think that is better here.


It would be good to make it consistent with the error codes returned
by the functions when they are present.

Anyway, it's fine by me if that's consistent with the other acpi.h stubs.
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Sudeep Holla June 5, 2018, 4:18 p.m. | #3
On 05/06/18 17:12, Rafael J. Wysocki wrote:
> On Tue, Jun 5, 2018 at 5:33 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:

>>

>>

>> On 05/06/18 16:00, Rafael J. Wysocki wrote:

>>> On Tue, Jun 5, 2018 at 4:35 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:

>>>> Though CONFIG_ACPI_PPTT is selected by platforms and nor user visible,

>>>> it may be useful to support the build with CONFIG_ACPI_PPTT disabled.

>>>>

>>>> This patch adds the missing dummy/boiler plate implementation to fix

>>>> the build.

>>>>

>>>> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>

>>>> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

>>>> ---

>>>>  include/linux/acpi.h      | 15 +++++++++++++++

>>>>  include/linux/cacheinfo.h |  2 +-

>>>>  2 files changed, 16 insertions(+), 1 deletion(-)

>>>>

>>>> Hi Rafael,

>>>>

>>>> If you are fine with this, can you provide Ack, so that we route this

>>>> through ARM64 tree where most of the ACPI PPTT support is present.

>>>>

>>>> Regards,

>>>> Sudeep

>>>>

>>>> v1->v2:

>>>>         - removed duplicate definition for acpi_find_last_cache_level

>>>>

>>>> diff --git a/include/linux/acpi.h b/include/linux/acpi.h

>>>> index 8f2cdb0eca71..4b35a66383f9 100644

>>>> --- a/include/linux/acpi.h

>>>> +++ b/include/linux/acpi.h

>>>> @@ -1299,8 +1299,23 @@ static inline int lpit_read_residency_count_address(u64 *address)

>>>>  }

>>>>  #endif

>>>>

>>>> +#ifdef CONFIG_ACPI_PPTT

>>>>  int find_acpi_cpu_topology(unsigned int cpu, int level);

>>>>  int find_acpi_cpu_topology_package(unsigned int cpu);

>>>>  int find_acpi_cpu_cache_topology(unsigned int cpu, int level);

>>>> +#else

>>>> +static inline int find_acpi_cpu_topology(unsigned int cpu, int level)

>>>> +{

>>>> +       return -EINVAL;

>>>

>>> Why -EINVAL?

>>>

>>

>> I am not sure either. I used to return -ENOTSUPP, but IIRC someone

>> suggested to use it only for syscalls. Also I just based it on other

>> existing functions in acpi.h

>>

>> I am open for any alternatives if you think that is better here.

> 

> It would be good to make it consistent with the error codes returned

> by the functions when they are present.

> 

> Anyway, it's fine by me if that's consistent with the other acpi.h stubs.

> 


Thanks, indeed I copied it from existing stubs.

Can I take this as official Ack ?

-- 
Regards,
Sudeep
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Rafael J. Wysocki June 5, 2018, 4:20 p.m. | #4
On Tue, Jun 5, 2018 at 6:18 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:
>

>

> On 05/06/18 17:12, Rafael J. Wysocki wrote:

>> On Tue, Jun 5, 2018 at 5:33 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:

>>>

>>>

>>> On 05/06/18 16:00, Rafael J. Wysocki wrote:

>>>> On Tue, Jun 5, 2018 at 4:35 PM, Sudeep Holla <sudeep.holla@arm.com> wrote:

>>>>> Though CONFIG_ACPI_PPTT is selected by platforms and nor user visible,

>>>>> it may be useful to support the build with CONFIG_ACPI_PPTT disabled.

>>>>>

>>>>> This patch adds the missing dummy/boiler plate implementation to fix

>>>>> the build.

>>>>>

>>>>> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>

>>>>> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

>>>>> ---

>>>>>  include/linux/acpi.h      | 15 +++++++++++++++

>>>>>  include/linux/cacheinfo.h |  2 +-

>>>>>  2 files changed, 16 insertions(+), 1 deletion(-)

>>>>>

>>>>> Hi Rafael,

>>>>>

>>>>> If you are fine with this, can you provide Ack, so that we route this

>>>>> through ARM64 tree where most of the ACPI PPTT support is present.

>>>>>

>>>>> Regards,

>>>>> Sudeep

>>>>>

>>>>> v1->v2:

>>>>>         - removed duplicate definition for acpi_find_last_cache_level

>>>>>

>>>>> diff --git a/include/linux/acpi.h b/include/linux/acpi.h

>>>>> index 8f2cdb0eca71..4b35a66383f9 100644

>>>>> --- a/include/linux/acpi.h

>>>>> +++ b/include/linux/acpi.h

>>>>> @@ -1299,8 +1299,23 @@ static inline int lpit_read_residency_count_address(u64 *address)

>>>>>  }

>>>>>  #endif

>>>>>

>>>>> +#ifdef CONFIG_ACPI_PPTT

>>>>>  int find_acpi_cpu_topology(unsigned int cpu, int level);

>>>>>  int find_acpi_cpu_topology_package(unsigned int cpu);

>>>>>  int find_acpi_cpu_cache_topology(unsigned int cpu, int level);

>>>>> +#else

>>>>> +static inline int find_acpi_cpu_topology(unsigned int cpu, int level)

>>>>> +{

>>>>> +       return -EINVAL;

>>>>

>>>> Why -EINVAL?

>>>>

>>>

>>> I am not sure either. I used to return -ENOTSUPP, but IIRC someone

>>> suggested to use it only for syscalls. Also I just based it on other

>>> existing functions in acpi.h

>>>

>>> I am open for any alternatives if you think that is better here.

>>

>> It would be good to make it consistent with the error codes returned

>> by the functions when they are present.

>>

>> Anyway, it's fine by me if that's consistent with the other acpi.h stubs.

>>

>

> Thanks, indeed I copied it from existing stubs.

>

> Can I take this as official Ack ?


Yes, please.
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/arch/arm64/include/asm/topology.h b/arch/arm64/include/asm/topology.h
index df48212f767b..6b10459e6905 100644
--- a/arch/arm64/include/asm/topology.h
+++ b/arch/arm64/include/asm/topology.h
@@ -8,10 +8,8 @@  struct cpu_topology {
 	int thread_id;
 	int core_id;
 	int package_id;
-	int llc_id;
 	cpumask_t thread_sibling;
 	cpumask_t core_sibling;
-	cpumask_t llc_siblings;
 };
 
 extern struct cpu_topology cpu_topology[NR_CPUS];
diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
index 7415c166281f..047d98e68502 100644
--- a/arch/arm64/kernel/topology.c
+++ b/arch/arm64/kernel/topology.c
@@ -13,7 +13,6 @@ 
 
 #include <linux/acpi.h>
 #include <linux/arch_topology.h>
-#include <linux/cacheinfo.h>
 #include <linux/cpu.h>
 #include <linux/cpumask.h>
 #include <linux/init.h>
@@ -215,19 +214,7 @@  EXPORT_SYMBOL_GPL(cpu_topology);
 
 const struct cpumask *cpu_coregroup_mask(int cpu)
 {
-	const cpumask_t *core_mask = cpumask_of_node(cpu_to_node(cpu));
-
-	/* Find the smaller of NUMA, core or LLC siblings */
-	if (cpumask_subset(&cpu_topology[cpu].core_sibling, core_mask)) {
-		/* not numa in package, lets use the package siblings */
-		core_mask = &cpu_topology[cpu].core_sibling;
-	}
-	if (cpu_topology[cpu].llc_id != -1) {
-		if (cpumask_subset(&cpu_topology[cpu].llc_siblings, core_mask))
-			core_mask = &cpu_topology[cpu].llc_siblings;
-	}
-
-	return core_mask;
+	return &cpu_topology[cpu].core_sibling;
 }
 
 static void update_siblings_masks(unsigned int cpuid)
@@ -239,9 +226,6 @@  static void update_siblings_masks(unsigned int cpuid)
 	for_each_possible_cpu(cpu) {
 		cpu_topo = &cpu_topology[cpu];
 
-		if (cpuid_topo->llc_id == cpu_topo->llc_id)
-			cpumask_set_cpu(cpu, &cpuid_topo->llc_siblings);
-
 		if (cpuid_topo->package_id != cpu_topo->package_id)
 			continue;
 
@@ -307,10 +291,6 @@  static void __init reset_cpu_topology(void)
 		cpu_topo->core_id = 0;
 		cpu_topo->package_id = -1;
 
-		cpu_topo->llc_id = -1;
-		cpumask_clear(&cpu_topo->llc_siblings);
-		cpumask_set_cpu(cpu, &cpu_topo->llc_siblings);
-
 		cpumask_clear(&cpu_topo->core_sibling);
 		cpumask_set_cpu(cpu, &cpu_topo->core_sibling);
 		cpumask_clear(&cpu_topo->thread_sibling);
@@ -331,8 +311,6 @@  static int __init parse_acpi_topology(void)
 	is_threaded = read_cpuid_mpidr() & MPIDR_MT_BITMASK;
 
 	for_each_possible_cpu(cpu) {
-		int i, cache_id;
-
 		topology_id = find_acpi_cpu_topology(cpu, 0);
 		if (topology_id < 0)
 			return topology_id;
@@ -347,18 +325,6 @@  static int __init parse_acpi_topology(void)
 		}
 		topology_id = find_acpi_cpu_topology_package(cpu);
 		cpu_topology[cpu].package_id = topology_id;
-
-		i = acpi_find_last_cache_level(cpu);
-
-		if (i > 0) {
-			/*
-			 * this is the only part of cpu_topology that has
-			 * a direct relationship with the cache topology
-			 */
-			cache_id = find_acpi_cpu_cache_topology(cpu, i);
-			if (cache_id > 0)
-				cpu_topology[cpu].llc_id = cache_id;
-		}
 	}
 
 	return 0;