Re: [PATCH v2 2/2] acpi, gicv3-its, numa: Adding numa node mapping for gic-its units

[Ganapatrao Kulkarni]

Hi Lorenzo,

On Tue, Jun 20, 2017 at 8:47 PM, Lorenzo Pieralisi
<lorenzo.pieralisi@xxxxxxx> wrote:
> Hi Ganapatrao,
>
> On Tue, Jun 20, 2017 at 12:37:17PM +0530, Ganapatrao Kulkarni wrote:
>> Add code to parse SRAT ITS Affinity sub table as defined in ACPI 6.2
>> Later in per device probe, ITS devices are mapped to
>> numa node using ITS id to proximity domain mapping.
>>
>> Signed-off-by: Ganapatrao Kulkarni <ganapatrao.kulkarni@xxxxxxxxxx>
>> ---
>>  drivers/irqchip/irq-gic-v3-its.c | 76 +++++++++++++++++++++++++++++++++++++++-
>>  1 file changed, 75 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
>> index 45ea1933..5865a75 100644
>> --- a/drivers/irqchip/irq-gic-v3-its.c
>> +++ b/drivers/irqchip/irq-gic-v3-its.c
>> @@ -1833,6 +1833,78 @@ static int __init its_of_probe(struct device_node *node)
>>
>>  #define ACPI_GICV3_ITS_MEM_SIZE (SZ_128K)
>>
>> +#ifdef CONFIG_ACPI_NUMA
>> +struct its_srat_map {
>> +     u32                     numa_node;  /* numa node id */
>> +     u32                     its_id;  /* GIC ITS ID */
>> +     struct list_head        entry;
>> +};
>> +
>> +static LIST_HEAD(its_srat_maps);
>> +
>> +static int acpi_get_its_numa_node(u32 its_id)
>> +{
>> +     struct its_srat_map *srat_map;
>> +
>> +     list_for_each_entry(srat_map, &its_srat_maps, entry) {
>> +             if (its_id == srat_map->its_id)
>> +                     return srat_map->numa_node;
>> +     }
>> +     return NUMA_NO_NODE;
>> +}
>
> I do not want to come across as pedantic and I understand that the API
> to parse static ACPI tables sucks (ie in theory you could avoid stashing
> nodes altogether and just use a global variable to parse an ITS ID at
> a time but it sucks to go through the SRAT multiple times) but would not
> it be easier to just declare an __initdata static array and counter and
> be done with this ? We do not need this stuff after boot so I do not see
> a point in keeping it in memory and in allocating the nodes dynamically.

yes, it was my initial thought (as done v1), and i have replaced to
dynamic array in this series. I might have misunderstood your comments.
if you have no objection for the static array of size MAX_NUMNODES
as done v1 [1], i can send v3 with that change?

[1] https://patchwork.kernel.org/patch/9768869/

>
> Apologies if I was not clear in my comments to the previous set, I did
> not like stashing ITS PXM values in arch/arm64 code but I understand you
> need an array (__initdata) to avoid reading the SRAT multiple times.

yes, having static array to hold parsed table is required to avoid
multiple iterations
of parsing functions.
This array can go in __initdata.

>
> Thanks,
> Lorenzo
>
>> +
>> +static int __init
>> +gic_acpi_parse_srat_its(struct acpi_subtable_header *header,
>> +                      const unsigned long end)
>> +{
>> +     int pxm, node;
>> +     struct its_srat_map *srat_map;
>> +     struct acpi_srat_its_affinity *its_affinity;
>> +
>> +     its_affinity = (struct acpi_srat_its_affinity *)header;
>> +     if (!its_affinity)
>> +             return -EINVAL;
>> +
>> +     if (its_affinity->header.length <
>> +                     sizeof(struct acpi_srat_its_affinity)) {
>> +             pr_err("SRAT:ITS: Invalid SRAT header length: %d\n",
>> +                     its_affinity->header.length);
>> +             return -EINVAL;
>> +     }
>> +
>> +     pxm = its_affinity->proximity_domain;
>> +     node = acpi_map_pxm_to_node(pxm);
>> +
>> +     if (node == NUMA_NO_NODE || node >= MAX_NUMNODES) {
>> +             pr_err("SRAT:ITS Invalid numa node %d\n", node);
>> +             return -EINVAL;
>> +     }
>> +
>> +     srat_map = kzalloc(sizeof(*srat_map), GFP_KERNEL);
>> +     if (!srat_map)
>> +             return -ENOMEM;
>> +
>> +     srat_map->numa_node = node;
>> +     srat_map->its_id = its_affinity->its_id;
>> +     list_add(&srat_map->entry, &its_srat_maps);
>> +     pr_info("ACPI: NUMA: SRAT: ITS: PXM %d -> ITS_ID %d -> NODE %d\n",
>> +             srat_map->its_id, pxm, node);
>> +
>> +     return 0;
>> +}
>> +
>> +static int __init acpi_table_parse_srat_its(void)
>> +{
>> +     return acpi_table_parse_entries(ACPI_SIG_SRAT,
>> +                                     sizeof(struct acpi_table_srat),
>> +                                     ACPI_SRAT_TYPE_GIC_ITS_AFFINITY,
>> +                                     gic_acpi_parse_srat_its, 0);
>> +}
>> +#else
>> +#define acpi_table_parse_srat_its(void) { }
>> +#define acpi_get_its_numa_node(its_id) NUMA_NO_NODE
>> +#endif
>> +
>>  static int __init gic_acpi_parse_madt_its(struct acpi_subtable_header *header,
>>                                         const unsigned long end)
>>  {
>> @@ -1861,7 +1933,8 @@ static int __init gic_acpi_parse_madt_its(struct acpi_subtable_header *header,
>>               goto dom_err;
>>       }
>>
>> -     err = its_probe_one(&res, dom_handle, NUMA_NO_NODE);
>> +     err = its_probe_one(&res, dom_handle,
>> +                     acpi_get_its_numa_node(its_entry->translation_id));
>>       if (!err)
>>               return 0;
>>
>> @@ -1873,6 +1946,7 @@ static int __init gic_acpi_parse_madt_its(struct acpi_subtable_header *header,
>>
>>  static void __init its_acpi_probe(void)
>>  {
>> +     acpi_table_parse_srat_its();
>>       acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_TRANSLATOR,
>>                             gic_acpi_parse_madt_its, 0);
>>  }
>> --
>> 1.8.1.4
>>

thanks
Ganapat