[PATCH v3] cxl/acpi: Verify CHBS length for CXL2.0

[PATCH v3] cxl/acpi: Verify CHBS length for CXL2.0

[Li Zhijian]

Per CXL Spec r3.1 Table 9-21, both CXL1.1 and CXL2.0 have defined their
own length, verify it to avoid an invalid CHBS.

Additionally, this patch also removes the redundant macro CXL_RCRB_SIZE,
favoring the use of `ACPI_CEDT_CHBS_LENGTH_CXL11`, which offers clearer
semantic meaning.

Reviewed-by: Gregory Price <gourry@gourry.net>
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
---
V3:
 - say more words in removing CXL_RCRB_SIZE # Alison
 - Collected Reviewed-by
V2: don't factor out, just validate # Dan
---
 drivers/cxl/acpi.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
index cb14829bb9be..2e63e50b2c40 100644
--- a/drivers/cxl/acpi.c
+++ b/drivers/cxl/acpi.c
@@ -11,8 +11,6 @@
 #include "cxlpci.h"
 #include "cxl.h"
 
-#define CXL_RCRB_SIZE	SZ_8K
-
 struct cxl_cxims_data {
 	int nr_maps;
 	u64 xormaps[] __counted_by(nr_maps);
@@ -478,8 +476,10 @@ static int cxl_get_chbs_iter(union acpi_subtable_headers *header, void *arg,
 
 	chbs = (struct acpi_cedt_chbs *) header;
 
-	if (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
-	    chbs->length != CXL_RCRB_SIZE)
+	if ((chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
+	    chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL11) ||
+	   (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL20 &&
+	    chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL20))
 		return 0;
 
 	if (!chbs->base)
-- 
2.47.0

Re: [PATCH v3] cxl/acpi: Verify CHBS length for CXL2.0

[Dan Williams]

Li Zhijian wrote:
> Per CXL Spec r3.1 Table 9-21, both CXL1.1 and CXL2.0 have defined their
> own length, verify it to avoid an invalid CHBS.
> 
> Additionally, this patch also removes the redundant macro CXL_RCRB_SIZE,
> favoring the use of `ACPI_CEDT_CHBS_LENGTH_CXL11`, which offers clearer
> semantic meaning.

Right, it was a mistake to allow the CXL_RCRB_SIZE definition to leak
into the kernel in v6.2 when ACPI_CEDT_CHBS_LENGTH_CXL11 had existed
since v5.14.

> 
> Reviewed-by: Gregory Price <gourry@gourry.net>
> Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
> ---
> V3:
>  - say more words in removing CXL_RCRB_SIZE # Alison
>  - Collected Reviewed-by
> V2: don't factor out, just validate # Dan
> ---
>  drivers/cxl/acpi.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
> index cb14829bb9be..2e63e50b2c40 100644
> --- a/drivers/cxl/acpi.c
> +++ b/drivers/cxl/acpi.c
> @@ -11,8 +11,6 @@
>  #include "cxlpci.h"
>  #include "cxl.h"
>  
> -#define CXL_RCRB_SIZE	SZ_8K
> -
>  struct cxl_cxims_data {
>  	int nr_maps;
>  	u64 xormaps[] __counted_by(nr_maps);
> @@ -478,8 +476,10 @@ static int cxl_get_chbs_iter(union acpi_subtable_headers *header, void *arg,
>  
>  	chbs = (struct acpi_cedt_chbs *) header;
>  
> -	if (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
> -	    chbs->length != CXL_RCRB_SIZE)
> +	if ((chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
> +	    chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL11) ||
> +	   (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL20 &&
> +	    chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL20))
>  		return 0;

I have a small readability preference for:

        if (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
            chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL11)
                return 0;
        if (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL20 &&
            chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL20)
                return 0;

...but not enough to ask you to respin this again.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>

Re: [PATCH v3] cxl/acpi: Verify CHBS length for CXL2.0

[Dave Jiang]

On 4/9/25 6:05 PM, Li Zhijian wrote:
> Per CXL Spec r3.1 Table 9-21, both CXL1.1 and CXL2.0 have defined their
> own length, verify it to avoid an invalid CHBS.
> 
> Additionally, this patch also removes the redundant macro CXL_RCRB_SIZE,
> favoring the use of `ACPI_CEDT_CHBS_LENGTH_CXL11`, which offers clearer
> semantic meaning.
> 
> Reviewed-by: Gregory Price <gourry@gourry.net>
> Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>

Applied to cxl/next with Dan's suggestion.

> ---
> V3:
>  - say more words in removing CXL_RCRB_SIZE # Alison
>  - Collected Reviewed-by
> V2: don't factor out, just validate # Dan
> ---
>  drivers/cxl/acpi.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
> index cb14829bb9be..2e63e50b2c40 100644
> --- a/drivers/cxl/acpi.c
> +++ b/drivers/cxl/acpi.c
> @@ -11,8 +11,6 @@
>  #include "cxlpci.h"
>  #include "cxl.h"
>  
> -#define CXL_RCRB_SIZE	SZ_8K
> -
>  struct cxl_cxims_data {
>  	int nr_maps;
>  	u64 xormaps[] __counted_by(nr_maps);
> @@ -478,8 +476,10 @@ static int cxl_get_chbs_iter(union acpi_subtable_headers *header, void *arg,
>  
>  	chbs = (struct acpi_cedt_chbs *) header;
>  
> -	if (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
> -	    chbs->length != CXL_RCRB_SIZE)
> +	if ((chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL11 &&
> +	    chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL11) ||
> +	   (chbs->cxl_version == ACPI_CEDT_CHBS_VERSION_CXL20 &&
> +	    chbs->length != ACPI_CEDT_CHBS_LENGTH_CXL20))
>  		return 0;
>  
>  	if (!chbs->base)