RE: [PATCH v22 4/4] scsi: ufs: Add HPB 2.0 support

[Avri Altman]

> @@ -7447,8 +7452,14 @@ static int ufs_get_device_desc(struct ufs_hba *hba)
> 
>         if (dev_info->wspecversion >= UFS_DEV_HPB_SUPPORT_VERSION &&
>             (b_ufs_feature_sup & UFS_DEV_HPB_SUPPORT)) {
> -               dev_info->hpb_enabled = true;
> -               ufshpb_get_dev_info(hba, desc_buf);
> +               bool hpb_en = false;
> +
> +               err = ufshcd_query_flag_retry(hba,
> UPIU_QUERY_OPCODE_READ_FLAG,
> +                                             QUERY_FLAG_IDN_HPB_EN, 0, &hpb_en);
> +               if (!err && hpb_en) {
> +                       dev_info->hpb_enabled = true;
> +                       ufshpb_get_dev_info(hba, desc_buf);
QUERY_FLAG_IDN_HPB_EN only apply to HPB2.0

> +               }
>         }
> 
> +
> +/*
> + * WRITE_BUFFER CMD support 36K (len=9) ~ 512K (len=128) default.
> + * it is possible to change range of transfer_len through sysfs.
> + */
Actually the transfer length is limited by its (and read id) single byte.
Fixing MAX_HPB_READ_ID = 128  is IMO a reasonable choice,
But not limited by spec.  Maybe make note of that ?

> +static inline bool ufshpb_is_required_wb(struct ufshpb_lu *hpb, int len)
> +{
> +       return (len >= hpb->pre_req_min_tr_len &&
> +               len <= hpb->pre_req_max_tr_len);
>  }
Maybe also check HPB2.0 as well?

> -void ufshpb_prep(struct ufs_hba *hba, struct ufshcd_lrb *lrbp)
> +int ufshpb_prep(struct ufs_hba *hba, struct ufshcd_lrb *lrbp)
>  {
>         struct ufshpb_lu *hpb;
>         struct ufshpb_region *rgn;
> @@ -282,26 +546,27 @@ void ufshpb_prep(struct ufs_hba *hba, struct
> ufshcd_lrb *lrbp)
>         u64 ppn;
>         unsigned long flags;
>         int transfer_len, rgn_idx, srgn_idx, srgn_offset;
> +       int read_id = MAX_HPB_READ_ID;
Should be 0 if wb is not used?

> +
> +       hpb->pre_req = kcalloc(qd, sizeof(struct ufshpb_req), GFP_KERNEL);
> +       hpb->throttle_pre_req = qd;
What is the point in throttling if you are allowing 32 simultaneous commands?
There can't be more than qd/2 anyway?
On the contrary, it makes much more sense to control the inflight map requests, instead?

> +       hpb->num_inflight_pre_req = 0;
> +

> -#define HPB_SUPPORT_VERSION                    0x100
> +#define HPB_SUPPORT_VERSION                    0x200
In ufshpb_get_dev_info you are bailing out if version != HPB_SUPPORT_VERSION
Meaning you are no longer backward supporting HPB1.0?

Maybe it would be more constructive to allow a day or 2 for more people to comment this new patch?
After all, it is a lot of code.

Thanks,
Avri