[PATCH 06/17] lib: sbi: Use scratch space to save per-HART domain pointer

[Andrew Jones]

On Tue, Apr 25, 2023 at 06:02:19PM +0530, Anup Patel wrote:
> Instead of using a global array indexed by hartid, we should use
> scratch space to save per-HART domain pointer.
> 
> Signed-off-by: Anup Patel <apatel at ventanamicro.com>
> ---
>  include/sbi/sbi_domain.h |  6 +-----
>  lib/sbi/sbi_domain.c     | 38 +++++++++++++++++++++++++++++++++++---
>  2 files changed, 36 insertions(+), 8 deletions(-)
> 
> diff --git a/include/sbi/sbi_domain.h b/include/sbi/sbi_domain.h
> index 124ea90..b05bcf4 100644
> --- a/include/sbi/sbi_domain.h
> +++ b/include/sbi/sbi_domain.h
> @@ -129,12 +129,8 @@ struct sbi_domain {
>  /** The root domain instance */
>  extern struct sbi_domain root;
>  
> -/** HART id to domain table */
> -extern struct sbi_domain *hartid_to_domain_table[];
> -
>  /** Get pointer to sbi_domain from HART id */
> -#define sbi_hartid_to_domain(__hartid) \
> -	hartid_to_domain_table[__hartid]
> +struct sbi_domain *sbi_hartid_to_domain(u32 hartid);
>  
>  /** Get pointer to sbi_domain for current HART */
>  #define sbi_domain_thishart_ptr() \
> diff --git a/lib/sbi/sbi_domain.c b/lib/sbi/sbi_domain.c
> index b07d310..531737c 100644
> --- a/lib/sbi/sbi_domain.c
> +++ b/lib/sbi/sbi_domain.c
> @@ -23,7 +23,6 @@
>   * the array to be null-terminated.
>   */
>  struct sbi_domain *domidx_to_domain_table[SBI_DOMAIN_MAX_INDEX + 1] = { 0 };
> -struct sbi_domain *hartid_to_domain_table[SBI_HARTMASK_MAX_BITS] = { 0 };
>  static u32 domain_count = 0;
>  static bool domain_finalized = false;
>  
> @@ -39,6 +38,32 @@ struct sbi_domain root = {
>  	.fw_region_inited = false,
>  };
>  
> +static unsigned long domain_hart_ptr_offset;
> +
> +struct sbi_domain *sbi_hartid_to_domain(u32 hartid)
> +{
> +	struct sbi_scratch *scratch;
> +
> +	scratch = sbi_hartid_to_scratch(hartid);
> +	if (!scratch || !domain_hart_ptr_offset)
> +		return NULL;
> +
> +	return (void *)(*((ulong *)sbi_scratch_offset_ptr(scratch,
> +						   domain_hart_ptr_offset)));
> +}
> +
> +static void update_hartid_to_domain(u32 hartid, struct sbi_domain *dom)
> +{
> +	struct sbi_scratch *scratch;
> +
> +	scratch = sbi_hartid_to_scratch(hartid);
> +	if (!scratch)
> +		return;
> +
> +	*((ulong *)sbi_scratch_offset_ptr(scratch, domain_hart_ptr_offset))
> +					= (ulong)dom;
> +}
> +
>  bool sbi_domain_is_assigned_hart(const struct sbi_domain *dom, u32 hartid)
>  {
>  	if (dom)
> @@ -519,11 +544,11 @@ int sbi_domain_register(struct sbi_domain *dom,
>  		if (!sbi_hartmask_test_hart(i, dom->possible_harts))
>  			continue;
>  
> -		tdom = hartid_to_domain_table[i];
> +		tdom = sbi_hartid_to_domain(i);
>  		if (tdom)
>  			sbi_hartmask_clear_hart(i,
>  					&tdom->assigned_harts);
> -		hartid_to_domain_table[i] = dom;
> +		update_hartid_to_domain(i, dom);
>  		sbi_hartmask_set_hart(i, &dom->assigned_harts);
>  
>  		/*
> @@ -715,9 +740,14 @@ int sbi_domain_init(struct sbi_scratch *scratch, u32 cold_hartid)
>  		return SBI_EINVAL;
>  	}
>  
> +	domain_hart_ptr_offset = sbi_scratch_alloc_offset(sizeof(ulong));
> +	if (!domain_hart_ptr_offset)
> +		return SBI_ENOMEM;
> +
>  	root_memregs = sbi_calloc(sizeof(*root_memregs), ROOT_REGION_MAX + 1);
>  	if (!root_memregs) {
>  		sbi_printf("%s: no memory for root regions\n", __func__);
> +		sbi_scratch_free_offset(domain_hart_ptr_offset);
>  		return SBI_ENOMEM;
>  	}
>  	root.regions = root_memregs;
> @@ -726,6 +756,7 @@ int sbi_domain_init(struct sbi_scratch *scratch, u32 cold_hartid)
>  	if (!root_hmask) {
>  		sbi_printf("%s: no memory for root hartmask\n", __func__);
>  		sbi_free(root_memregs);
> +		sbi_scratch_free_offset(domain_hart_ptr_offset);
>  		return SBI_ENOMEM;
>  	}
>  	root.possible_harts = root_hmask;
> @@ -774,6 +805,7 @@ int sbi_domain_init(struct sbi_scratch *scratch, u32 cold_hartid)
>  	if (rc) {
>  		sbi_free(root_hmask);
>  		sbi_free(root_memregs);
> +		sbi_scratch_free_offset(domain_hart_ptr_offset);
>  		return rc;
>  	}

Rather than duplicating all the free()'s we could use a sequence of error
labels and gotos.

Otherwise,

Reviewed-by: Andrew Jones <ajones at ventanamicro.com>

Thanks,
drew