Re: [PATCH 1/4] acpi: tables: Add CDAT table parsing support

[Jonathan Cameron <Jonathan.Cameron@Huawei.com>]

On Fri, 05 May 2023 10:32:56 -0700
Dave Jiang <dave.jiang@intel.com> wrote:

> The CDAT table is very similar to ACPI tables when it comes to sub-table
> and entry structures. The helper functions can be also used to parse the
> CDAT table. Add support to the helper functions to deal with an external
> CDAT table, and also handle the endieness since CDAT can be processed by a
> BE host. Export a function acpi_table_parse_cdat() for CXL driver to parse
> a CDAT table.
> 
> In order to minimize ACPI code changes, __force is being utilized to deal
> with the case of a big endien (BE) host parsing a CDAT. All CDAT data
> structure variables are being force casted to __leX as appropriate.

Hi Dave,

This falls into the annoyance that CDAT doesn't have a standard table header.
Whilst I understand that was done deliberately it means some odd things happen
in this code.

Just how bad is the duplication if we don't do this at all, but instead roll
a version for CDAT that doesn't force things through pointers of the wrong types?

Otherwise, maybe we need some unions so that the type mashups don't happen.

> 
> Cc: Rafael J. Wysocki <rafael@kernel.org>
> Cc: Len Brown <lenb@kernel.org>
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
> ---
>  drivers/acpi/tables.c |   47 +++++++++++++++++++++++++++++++++++++++++++++--
>  include/acpi/actbl1.h |    3 +++
>  include/linux/acpi.h  |    4 ++++
>  3 files changed, 52 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/acpi/tables.c b/drivers/acpi/tables.c
> index 7b4680da57d7..08486f6df442 100644
> --- a/drivers/acpi/tables.c
> +++ b/drivers/acpi/tables.c
> @@ -42,6 +42,7 @@ enum acpi_subtable_type {
>  	ACPI_SUBTABLE_HMAT,
>  	ACPI_SUBTABLE_PRMT,
>  	ACPI_SUBTABLE_CEDT,
> +	ACPI_SUBTABLE_CDAT,
>  };
>  
>  struct acpi_subtable_entry {
> @@ -239,6 +240,8 @@ acpi_get_entry_type(struct acpi_subtable_entry *entry)
>  		return 0;
>  	case ACPI_SUBTABLE_CEDT:
>  		return entry->hdr->cedt.type;
> +	case ACPI_SUBTABLE_CDAT:
> +		return entry->hdr->cdat.type;
>  	}
>  	return 0;
>  }
> @@ -255,6 +258,8 @@ acpi_get_entry_length(struct acpi_subtable_entry *entry)
>  		return entry->hdr->prmt.length;
>  	case ACPI_SUBTABLE_CEDT:
>  		return entry->hdr->cedt.length;
> +	case ACPI_SUBTABLE_CDAT:
> +		return le16_to_cpu((__force __le16)entry->hdr->cdat.length);
>  	}
>  	return 0;
>  }
> @@ -271,6 +276,8 @@ acpi_get_subtable_header_length(struct acpi_subtable_entry *entry)
>  		return sizeof(entry->hdr->prmt);
>  	case ACPI_SUBTABLE_CEDT:
>  		return sizeof(entry->hdr->cedt);
> +	case ACPI_SUBTABLE_CDAT:
> +		return sizeof(entry->hdr->cdat);
>  	}
>  	return 0;
>  }
> @@ -284,9 +291,22 @@ acpi_get_subtable_type(char *id)
>  		return ACPI_SUBTABLE_PRMT;
>  	if (strncmp(id, ACPI_SIG_CEDT, 4) == 0)
>  		return ACPI_SUBTABLE_CEDT;
> +	if (strncmp(id, ACPI_SIG_CDAT, 4) == 0)
> +		return ACPI_SUBTABLE_CDAT;

I'm not super keen on inventing a SIG when the CDAT 'table'
doesn't actually have one.

>  	return ACPI_SUBTABLE_COMMON;
>  }
>  
> +static unsigned long __init_or_acpilib
> +acpi_table_get_length(enum acpi_subtable_type type,
> +		      struct acpi_table_header *hdr)

I don't like parsing in an acpi_table_header type here when it may not be one.
I think this length decision needs to be pushed up a level to where we can see
if we have a CDAT table or not.


> +{
> +	if (type == ACPI_SUBTABLE_CDAT)
> +		return le32_to_cpu(
> +			(__force __le32)((struct acpi_table_cdat *)hdr)->length);

Perhaps a local variable in here somewhere would make it more readable.
	__le32 length = (__force__le32)((struct acpi_table_cdat *)hdr)->length;

	return le32_to_cpu(length)?


> +
> +	return hdr->length;
> +}
> +
>  static __init_or_acpilib bool has_handler(struct acpi_subtable_proc *proc)
>  {
>  	return proc->handler || proc->handler_arg;
> @@ -332,16 +352,19 @@ static int __init_or_acpilib acpi_parse_entries_array(
>  	int proc_num, unsigned int max_entries)
>  {
>  	struct acpi_subtable_entry entry;
> +	enum acpi_subtable_type type;
>  	unsigned long table_end, subtable_len, entry_len;
>  	int count = 0;
>  	int errs = 0;
>  	int i;
>  
> -	table_end = (unsigned long)table_header + table_header->length;
> +	type = acpi_get_subtable_type(id);
> +	table_end = (unsigned long)table_header +
> +		    acpi_table_get_length(type, table_header);
As above, I don't like carrying CDAT which doesn't have an acpi_table_header
section around as that type of pointer.

>  
>  	/* Parse all entries looking for a match. */
>  
> -	entry.type = acpi_get_subtable_type(id);
> +	entry.type = type;
>  	entry.hdr = (union acpi_subtable_headers *)
>  	    ((unsigned long)table_header + table_size);
>  	subtable_len = acpi_get_subtable_header_length(&entry);
> @@ -464,6 +487,26 @@ int __init acpi_table_parse_madt(enum acpi_madt_type id,
>  					    handler, max_entries);
>  }
>  
> +int acpi_table_parse_cdat(enum acpi_cdat_type type,
> +			  acpi_tbl_entry_handler_arg handler_arg, void *arg,
> +			  struct acpi_table_cdat *table_header)
> +{
> +	struct acpi_subtable_proc proc = {
> +		.id		= type,
> +		.handler_arg	= handler_arg,
> +		.arg		= arg,
> +	};
> +
> +	if (!table_header)
> +		return -EINVAL;
> +
> +	return acpi_parse_entries_array(ACPI_SIG_CDAT,
> +			sizeof(struct acpi_table_cdat),
> +			(struct acpi_table_header *)table_header,
> +			&proc, 1, 0);
> +}
> +EXPORT_SYMBOL_NS_GPL(acpi_table_parse_cdat, CXL);
> +
>  /**
>   * acpi_table_parse - find table with @id, run @handler on it
>   * @id: table id to find
> diff --git a/include/acpi/actbl1.h b/include/acpi/actbl1.h
> index 81b9e794424d..3119be093cfe 100644
> --- a/include/acpi/actbl1.h
> +++ b/include/acpi/actbl1.h
> @@ -66,6 +66,9 @@
>  #define ACPI_SIG_IEIT           "IEIT"
>  #endif
>  
> +/* External to ACPI */
> +#define ACPI_SIG_CDAT		"CDAT" /* Coherent Device Attribute Table */

Worse that that, fictional signature :)
It's the nameof the 'table', but it's not a signature as it's never
used as they are in ACPI and doesn't appear anywhere in the table.

> +
>  /*
>   * All tables must be byte-packed to match the ACPI specification, since
>   * the tables are provided by the system BIOS.
> diff --git a/include/linux/acpi.h b/include/linux/acpi.h
> index efff750f326d..4c3dfe7587e9 100644
> --- a/include/linux/acpi.h
> +++ b/include/linux/acpi.h
> @@ -135,6 +135,7 @@ union acpi_subtable_headers {
>  	struct acpi_hmat_structure hmat;
>  	struct acpi_prmt_module_header prmt;
>  	struct acpi_cedt_header cedt;
> +	struct acpi_cdat_header cdat;
>  };
>  
>  typedef int (*acpi_tbl_table_handler)(struct acpi_table_header *table);
> @@ -266,6 +267,9 @@ acpi_table_parse_cedt(enum acpi_cedt_type id,
>  
>  int acpi_parse_mcfg (struct acpi_table_header *header);
>  void acpi_table_print_madt_entry (struct acpi_subtable_header *madt);
> +int acpi_table_parse_cdat(enum acpi_cdat_type type,
> +			  acpi_tbl_entry_handler_arg handler, void *arg,
> +			  struct acpi_table_cdat *table_header);
How did we end up with an 'acpi_' table that isn't in ACPI?
(I'm not looking as I fear I might be responsible :)
Should perhaps consider renaming all the CDAT entries so it doesn't looks like they
are.

>  
>  /* the following numa functions are architecture-dependent */
>  void acpi_numa_slit_init (struct acpi_table_slit *slit);
> 
> 
>