Re: [PATCH v3 1/3] acpi: HMAT support in acpi_parse_entries_array()

[Dan Williams]

On Fri, Dec 15, 2017 at 5:53 PM, Rafael J. Wysocki <rjw@xxxxxxxxxxxxx> wrote:
> On Friday, December 15, 2017 2:10:17 AM CET Dan Williams wrote:
>> On Wed, Dec 13, 2017 at 6:10 PM, Ross Zwisler
>> <ross.zwisler@xxxxxxxxxxxxxxx> wrote:
>> > The current implementation of acpi_parse_entries_array() assumes that each
>> > subtable has a standard ACPI subtable entry of type struct
>> > acpi_subtable_header.  This standard subtable header has a one byte length
>> > followed by a one byte type.
>> >
>> > The HMAT subtables have to allow for a longer length so they have subtable
>> > headers of type struct acpi_hmat_structure which has a 2 byte type and a 4
>> > byte length.
>>
>> Hmm, NFIT has a 2 byte type and a 2 byte length, so its one more
>> permutation. I happened to reinvent sub-table parsing in the NFIT
>> driver, but it might be nice in the future to refactor that to use the
>> common parsing.
>>
>> >
>> > Enhance the subtable parsing in acpi_parse_entries_array() so that it can
>> > handle these new HMAT subtables.
>> >
>> > Signed-off-by: Ross Zwisler <ross.zwisler@xxxxxxxxxxxxxxx>
>> > ---
>> >  drivers/acpi/tables.c | 52 ++++++++++++++++++++++++++++++++++++++++-----------
>> >  1 file changed, 41 insertions(+), 11 deletions(-)
>> >
>> > diff --git a/drivers/acpi/tables.c b/drivers/acpi/tables.c
>> > index 80ce2a7d224b..f777b94c234a 100644
>> > --- a/drivers/acpi/tables.c
>> > +++ b/drivers/acpi/tables.c
>> > @@ -218,6 +218,33 @@ void acpi_table_print_madt_entry(struct acpi_subtable_header *header)
>> >         }
>> >  }
>> >
>> > +static unsigned long __init
>> > +acpi_get_entry_type(char *id, void *entry)
>> > +{
>> > +       if (strncmp(id, ACPI_SIG_HMAT, 4) == 0)
>> > +               return ((struct acpi_hmat_structure *)entry)->type;
>> > +       else
>> > +               return ((struct acpi_subtable_header *)entry)->type;
>> > +}
>>
>> It seems inefficient to make all checks keep asking "is HMAT?".
>
> Well, ideally, the signature should be checked once.  I guess that can be
> arranged for here.
>
>> Especially if we want to extend this to other table types should we
>> instead setup and pass a pair of function pointers to parse the
>> sub-table format?
>
> Function pointers may be too much even. :-)

True, how about an enum of acpi sub-table header types?