If the the Processor ID valid is not set for a Physical Processor Package
node, then the node table offset is used as a substitute. As such, we
may get info like this from sysfs:
root@(none)$ pwd
/sys/devices/system/cpu/cpu0/topology
root@(none)$ more physical_package_id
56
Inform the user of this in the bootlog, as it is much less than ideal, and
they can remedy this in their FW.
This topic was originally discussed in:
https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
Signed-off-by: John Garry <[email protected]>
diff --git a/drivers/acpi/pptt.c b/drivers/acpi/pptt.c
index 4ae93350b70d..b4ed3c818e00 100644
--- a/drivers/acpi/pptt.c
+++ b/drivers/acpi/pptt.c
@@ -515,6 +515,8 @@ static int topology_get_acpi_cpu_tag(struct acpi_table_header *table,
if (level == 0 ||
cpu_node->flags & ACPI_PPTT_ACPI_PROCESSOR_ID_VALID)
return cpu_node->acpi_processor_id;
+ if (level == PPTT_ABORT_PACKAGE)
+ pr_notice_once("Physical package node Processor ID valid not set, will use table offset as substitute\n");
return ACPI_PTR_DIFF(cpu_node, table);
}
pr_warn_once("PPTT table found, but unable to locate core %d (%d)\n",
--
2.16.4
(I see 2 copies of this patch, replying on the latest)
On Thu, Apr 02, 2020 at 08:22:25PM +0800, John Garry wrote:
> If the the Processor ID valid is not set for a Physical Processor Package
> node, then the node table offset is used as a substitute. As such, we
> may get info like this from sysfs:
>
> root@(none)$ pwd
> /sys/devices/system/cpu/cpu0/topology
> root@(none)$ more physical_package_id
> 56
>
> Inform the user of this in the bootlog, as it is much less than ideal, and
> they can remedy this in their FW.
>
> This topic was originally discussed in:
> https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
>
Acked-by: Sudeep Holla <[email protected]>
--
Regards,
Sudeep
On 2020/4/2 20:22, John Garry wrote:
> If the the Processor ID valid is not set for a Physical Processor Package
> node, then the node table offset is used as a substitute. As such, we
> may get info like this from sysfs:
>
> root@(none)$ pwd
> /sys/devices/system/cpu/cpu0/topology
> root@(none)$ more physical_package_id
> 56
>
> Inform the user of this in the bootlog, as it is much less than ideal, and
> they can remedy this in their FW.
>
> This topic was originally discussed in:
> https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
>
> Signed-off-by: John Garry <[email protected]>
>
> diff --git a/drivers/acpi/pptt.c b/drivers/acpi/pptt.c
> index 4ae93350b70d..b4ed3c818e00 100644
> --- a/drivers/acpi/pptt.c
> +++ b/drivers/acpi/pptt.c
> @@ -515,6 +515,8 @@ static int topology_get_acpi_cpu_tag(struct acpi_table_header *table,
> if (level == 0 ||
> cpu_node->flags & ACPI_PPTT_ACPI_PROCESSOR_ID_VALID)
> return cpu_node->acpi_processor_id;
> + if (level == PPTT_ABORT_PACKAGE)
> + pr_notice_once("Physical package node Processor ID valid not set, will use table offset as substitute\n");
> return ACPI_PTR_DIFF(cpu_node, table);
> }
> pr_warn_once("PPTT table found, but unable to locate core %d (%d)\n",
Looks good to me,
Reviewed-by: Hanjun Guo <[email protected]>
On 02/04/2020 14:08, Sudeep Holla wrote:
> (I see 2 copies of this patch, replying on the latest)
I sent 2x copies accidentally. WFH....
>
> On Thu, Apr 02, 2020 at 08:22:25PM +0800, John Garry wrote:
>> If the the Processor ID valid is not set for a Physical Processor Package
>> node, then the node table offset is used as a substitute. As such, we
>> may get info like this from sysfs:
>>
>> root@(none)$ pwd
>> /sys/devices/system/cpu/cpu0/topology
>> root@(none)$ more physical_package_id
>> 56
>>
>> Inform the user of this in the bootlog, as it is much less than ideal, and
>> they can remedy this in their FW.
>>
>> This topic was originally discussed in:
>> https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
>>
>
> Acked-by: Sudeep Holla <[email protected]>
thanks
>
> --
> Regards,
> Sudeep
> .
>
Hi,
On 4/2/20 7:22 AM, John Garry wrote:
> If the the Processor ID valid is not set for a Physical Processor Package
> node, then the node table offset is used as a substitute. As such, we
> may get info like this from sysfs:
>
> root@(none)$ pwd
> /sys/devices/system/cpu/cpu0/topology
> root@(none)$ more physical_package_id
> 56
>
> Inform the user of this in the bootlog, as it is much less than ideal, and
> they can remedy this in their FW.
>
> This topic was originally discussed in:
> https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
>
> Signed-off-by: John Garry <[email protected]>
>
> diff --git a/drivers/acpi/pptt.c b/drivers/acpi/pptt.c
> index 4ae93350b70d..b4ed3c818e00 100644
> --- a/drivers/acpi/pptt.c
> +++ b/drivers/acpi/pptt.c
> @@ -515,6 +515,8 @@ static int topology_get_acpi_cpu_tag(struct acpi_table_header *table,
> if (level == 0 ||
> cpu_node->flags & ACPI_PPTT_ACPI_PROCESSOR_ID_VALID)
> return cpu_node->acpi_processor_id;
> + if (level == PPTT_ABORT_PACKAGE)
> + pr_notice_once("Physical package node Processor ID valid not set, will use table offset as substitute\n");
What happens in the find_acpi_cpu_topology_hetro_id() case, if the last
IDENTICAL node isn't a socket/etc. Are we expecting to warn of a missing
processor container there as well?
> return ACPI_PTR_DIFF(cpu_node, table);
> }
> pr_warn_once("PPTT table found, but unable to locate core %d (%d)\n",
>
On 02/04/2020 14:55, Jeremy Linton wrote:
> Hi,
>
> On 4/2/20 7:22 AM, John Garry wrote:
>> If the the Processor ID valid is not set for a Physical Processor Package
>> node, then the node table offset is used as a substitute. As such, we
>> may get info like this from sysfs:
>>
>> root@(none)$ pwd
>> /sys/devices/system/cpu/cpu0/topology
>> root@(none)$ more physical_package_id
>> 56
>>
>> Inform the user of this in the bootlog, as it is much less than ideal,
>> and
>> they can remedy this in their FW.
>>
>> This topic was originally discussed in:
>> https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
>>
>>
>> Signed-off-by: John Garry <[email protected]>
>>
>> diff --git a/drivers/acpi/pptt.c b/drivers/acpi/pptt.c
>> index 4ae93350b70d..b4ed3c818e00 100644
>> --- a/drivers/acpi/pptt.c
>> +++ b/drivers/acpi/pptt.c
>> @@ -515,6 +515,8 @@ static int topology_get_acpi_cpu_tag(struct
>> acpi_table_header *table,
>> if (level == 0 ||
>> cpu_node->flags & ACPI_PPTT_ACPI_PROCESSOR_ID_VALID)
>> return cpu_node->acpi_processor_id;
>> + if (level == PPTT_ABORT_PACKAGE)
>> + pr_notice_once("Physical package node Processor ID valid
>> not set, will use table offset as substitute\n");
Hi Jeremy,
>
> What happens in the find_acpi_cpu_topology_hetro_id() case, if the last
> IDENTICAL node isn't a socket/etc. Are we expecting to warn of a missing
> processor container there as well?
ok, I could just use this then:
if (flags & ACPI_PPTT_PHYSCIAL_PACKAGE)
pr_notice_once("Physical package node Processor ID ...")
As I recall, this tag for hetro id is not exposed to userspace (yet).
Thanks,
John
Hi,
On 4/2/20 9:51 AM, John Garry wrote:
> On 02/04/2020 14:55, Jeremy Linton wrote:
>> Hi,
>>
>> On 4/2/20 7:22 AM, John Garry wrote:
>>> If the the Processor ID valid is not set for a Physical Processor
>>> Package
>>> node, then the node table offset is used as a substitute. As such, we
>>> may get info like this from sysfs:
>>>
>>> root@(none)$ pwd
>>> /sys/devices/system/cpu/cpu0/topology
>>> root@(none)$ more physical_package_id
>>> 56
>>>
>>> Inform the user of this in the bootlog, as it is much less than
>>> ideal, and
>>> they can remedy this in their FW.
>>>
>>> This topic was originally discussed in:
>>> https://lore.kernel.org/linux-acpi/[email protected]/T/#m0ec18637d8586f832084a8a6af22580e6174669a
>>>
>>>
>>> Signed-off-by: John Garry <[email protected]>
>>>
>>> diff --git a/drivers/acpi/pptt.c b/drivers/acpi/pptt.c
>>> index 4ae93350b70d..b4ed3c818e00 100644
>>> --- a/drivers/acpi/pptt.c
>>> +++ b/drivers/acpi/pptt.c
>>> @@ -515,6 +515,8 @@ static int topology_get_acpi_cpu_tag(struct
>>> acpi_table_header *table,
>>> if (level == 0 ||
>>> cpu_node->flags & ACPI_PPTT_ACPI_PROCESSOR_ID_VALID)
>>> return cpu_node->acpi_processor_id;
>>> + if (level == PPTT_ABORT_PACKAGE)
>>> + pr_notice_once("Physical package node Processor ID valid
>>> not set, will use table offset as substitute\n");
>
> Hi Jeremy,
>
>>
>> What happens in the find_acpi_cpu_topology_hetro_id() case, if the
>> last IDENTICAL node isn't a socket/etc. Are we expecting to warn of a
>> missing processor container there as well?
>
> ok, I could just use this then:
>
> if (flags & ACPI_PPTT_PHYSCIAL_PACKAGE)
> pr_notice_once("Physical package node Processor ID ...")
I think that is a better plan.
>
> As I recall, this tag for hetro id is not exposed to userspace (yet).
Correct.