On Mon, Dec 18, 2023 at 08:02:34PM +0530, Avnish Chouhan wrote:
> This patch adds support for Radix, Xive and Radix_gtse in Options
> vector5 which is required for KVM LPARs. KVM LPARs ONLY support
> Radix and not the Hash. Not enabling Radix on any PowerVM KVM LPARs
> will result in boot failure.
>
> Signed-off-by: Avnish Chouhan <avn...@linux.vnet.ibm.com>
> ---
> grub-core/kern/ieee1275/init.c | 68
> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 67 insertions(+), 1 deletion(-)
>
> diff --git a/grub-core/kern/ieee1275/init.c b/grub-core/kern/ieee1275/init.c
> index fb7d1a3..94bbf86 100644
> --- a/grub-core/kern/ieee1275/init.c
> +++ b/grub-core/kern/ieee1275/init.c
> @@ -113,6 +113,17 @@ grub_addr_t grub_ieee1275_original_stack;
> #define DRC_INFO 0x40
> #define BYTE22 (DY_MEM_V2 | DRC_INFO)
>
> +/* For ibm,arch-vec-5-platform-support */
> +
> +#define XIVE_INDEX 0x17
> +#define MMU_INDEX 0x18
> +#define RADIX_GTSE_INDEX 0x1a
> +#define RADIX_ENABLED 0x40
> +#define XIVE_ENABLED 0x40
> +#define HASH_ENABLED 0x0
This looks like a mistake. Is not it? If it is correct I would do s/0x0/0x00/.
> +#define MAX_SUPPORTED 0xC0
> +#define RADIX_GTSE_ENABLED 0x40
> +
> void
> grub_exit (void)
> {
> @@ -737,6 +748,10 @@ struct option_vector5
> grub_uint32_t platform_facilities;
> grub_uint8_t sub_processors;
> grub_uint8_t byte22;
> + grub_uint8_t xive;
> + grub_uint8_t mmu;
> + grub_uint8_t hpt_ext;
> + grub_uint8_t radix_gtse;
> } GRUB_PACKED;
>
> struct pvr_entry
> @@ -775,6 +790,13 @@ grub_ieee1275_ibm_cas (void)
> {
> int rc;
> grub_ieee1275_ihandle_t root;
> + grub_uint8_t ibm_arch_platform_support[8];
> + grub_ssize_t actual;
> + grub_uint8_t xive_support = 0;
> + grub_uint8_t mmu_support = 0;
> + grub_uint8_t radix_gtse_support = 0;
> + int i=0;
> + int prop_len=8;
Wrong coding style, some missing spaces before and after "=".
> struct cas_args
> {
> struct grub_ieee1275_common_hdr common;
> @@ -783,6 +805,50 @@ grub_ieee1275_ibm_cas (void)
> grub_ieee1275_cell_t cas_addr;
> grub_ieee1275_cell_t result;
> } args;
> +
> + grub_ieee1275_get_integer_property (grub_ieee1275_chosen,
> + "ibm,arch-vec-5-platform-support",
> + ibm_arch_platform_support,
> + sizeof (ibm_arch_platform_support),
> + &actual);
> +
> + for (i=0; i<prop_len; i++)
> + {
> + switch (ibm_arch_platform_support[i])
> + {
> + case XIVE_INDEX:
> + if (ibm_arch_platform_support[i+1] & MAX_SUPPORTED)
Similar thing here...
> + xive_support = XIVE_ENABLED;
> + else
> + xive_support = 0;
> + break;
> +
> + case MMU_INDEX:
> + if (ibm_arch_platform_support[i+1] & MAX_SUPPORTED)
Ditto.
> + mmu_support = RADIX_ENABLED;
> + else
> + mmu_support = HASH_ENABLED;
> + break;
> +
> + case RADIX_GTSE_INDEX:
> + if (mmu_support == RADIX_ENABLED)
> + {
> + radix_gtse_support = ibm_arch_platform_support[i+1] &
> RADIX_GTSE_ENABLED;
Ditto. And you can drop curly braces here and below.
> + }
> + else
> + {
> + radix_gtse_support = 0;
> + }
> + break;
> +
> + default:
> + /* Ignoring the other indexes of ibm,arch-vec-5-platform-support
> */
> + break;
> + }
> + /* Skipping the property value */
> + i++;
> + }
> +
> struct cas_vector vector =
> {
> .pvr_list = { { 0x00000000, 0xffffffff } }, /* any processor */
> @@ -799,7 +865,7 @@ grub_ieee1275_ibm_cas (void)
> .vec4 = 0x0001, /* set required minimum capacity % to the lowest value */
> .vec5_size = 1 + sizeof (struct option_vector5) - 2,
> .vec5 = {
> - 0, BYTE2, 0, CMO, ASSOCIATIVITY, BIN_OPTS, 0, 0, MAX_CPU, 0, 0,
> PLATFORM_FACILITIES, SUB_PROCESSORS, BYTE22
> + 0, BYTE2, 0, CMO, ASSOCIATIVITY, BIN_OPTS, 0, 0, MAX_CPU, 0, 0,
> PLATFORM_FACILITIES, SUB_PROCESSORS, BYTE22, xive_support, mmu_support, 0,
> radix_gtse_support
> }
> };
Daniel
_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel
