Merge tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
authorLinus Torvalds <torvalds@linux-foundation.org>
Sun, 5 Jun 2022 16:25:12 +0000 (09:25 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sun, 5 Jun 2022 16:25:12 +0000 (09:25 -0700)
Pull more SCSI updates from James Bottomley:
 "Mostly small bug fixes plus other trivial updates.

  The major change of note is moving ufs out of scsi and a minor update
  to lpfc vmid handling"

* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (24 commits)
  scsi: qla2xxx: Remove unused 'ql_dm_tgt_ex_pct' parameter
  scsi: qla2xxx: Remove setting of 'req' and 'rsp' parameters
  scsi: mpi3mr: Fix kernel-doc
  scsi: lpfc: Add support for ATTO Fibre Channel devices
  scsi: core: Return BLK_STS_TRANSPORT for ALUA transitioning
  scsi: sd_zbc: Prevent zone information memory leak
  scsi: sd: Fix potential NULL pointer dereference
  scsi: mpi3mr: Rework mrioc->bsg_device model to fix warnings
  scsi: myrb: Fix up null pointer access on myrb_cleanup()
  scsi: core: Unexport scsi_bus_type
  scsi: sd: Don't call blk_cleanup_disk() in sd_probe()
  scsi: ufs: ufshcd: Delete unnecessary NULL check
  scsi: isci: Fix typo in comment
  scsi: pmcraid: Fix typo in comment
  scsi: smartpqi: Fix typo in comment
  scsi: qedf: Fix typo in comment
  scsi: esas2r: Fix typo in comment
  scsi: storvsc: Fix typo in comment
  scsi: ufs: Split the drivers/scsi/ufs directory
  scsi: qla1280: Remove redundant variable
  ...

1  2 
MAINTAINERS
drivers/Kconfig
drivers/Makefile
drivers/nvme/host/fc.c
drivers/scsi/lpfc/lpfc_scsi.c
drivers/scsi/qla2xxx/qla_target.c
drivers/scsi/sd.c
drivers/scsi/storvsc_drv.c
drivers/ufs/core/ufshpb.c

diff --cc MAINTAINERS
Simple merge
diff --cc drivers/Kconfig
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
index 0000000000000000000000000000000000000000,aee0ec4cee70eac41d8769b806fbddce0714e1ff..de2bb8401bc4e4fa0b06bc4a124bff27e3522032
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,2670 +1,2667 @@@
 -      blk_execute_rq_nowait(req, true, ufshpb_umap_req_compl_fn);
+ // SPDX-License-Identifier: GPL-2.0
+ /*
+  * Universal Flash Storage Host Performance Booster
+  *
+  * Copyright (C) 2017-2021 Samsung Electronics Co., Ltd.
+  *
+  * Authors:
+  *    Yongmyung Lee <ymhungry.lee@samsung.com>
+  *    Jinyoung Choi <j-young.choi@samsung.com>
+  */
+ #include <asm/unaligned.h>
+ #include <linux/delay.h>
+ #include <linux/device.h>
+ #include <linux/module.h>
+ #include <scsi/scsi_cmnd.h>
+ #include "ufshcd-priv.h"
+ #include "ufshpb.h"
+ #include "../../scsi/sd.h"
+ #define ACTIVATION_THRESHOLD 8 /* 8 IOs */
+ #define READ_TO_MS 1000
+ #define READ_TO_EXPIRIES 100
+ #define POLLING_INTERVAL_MS 200
+ #define THROTTLE_MAP_REQ_DEFAULT 1
+ /* memory management */
+ static struct kmem_cache *ufshpb_mctx_cache;
+ static mempool_t *ufshpb_mctx_pool;
+ static mempool_t *ufshpb_page_pool;
+ /* A cache size of 2MB can cache ppn in the 1GB range. */
+ static unsigned int ufshpb_host_map_kbytes = 2048;
+ static int tot_active_srgn_pages;
+ static struct workqueue_struct *ufshpb_wq;
+ static void ufshpb_update_active_info(struct ufshpb_lu *hpb, int rgn_idx,
+                                     int srgn_idx);
+ bool ufshpb_is_allowed(struct ufs_hba *hba)
+ {
+       return !(hba->ufshpb_dev.hpb_disabled);
+ }
+ /* HPB version 1.0 is called as legacy version. */
+ bool ufshpb_is_legacy(struct ufs_hba *hba)
+ {
+       return hba->ufshpb_dev.is_legacy;
+ }
+ static struct ufshpb_lu *ufshpb_get_hpb_data(struct scsi_device *sdev)
+ {
+       return sdev->hostdata;
+ }
+ static int ufshpb_get_state(struct ufshpb_lu *hpb)
+ {
+       return atomic_read(&hpb->hpb_state);
+ }
+ static void ufshpb_set_state(struct ufshpb_lu *hpb, int state)
+ {
+       atomic_set(&hpb->hpb_state, state);
+ }
+ static int ufshpb_is_valid_srgn(struct ufshpb_region *rgn,
+                               struct ufshpb_subregion *srgn)
+ {
+       return rgn->rgn_state != HPB_RGN_INACTIVE &&
+               srgn->srgn_state == HPB_SRGN_VALID;
+ }
+ static bool ufshpb_is_read_cmd(struct scsi_cmnd *cmd)
+ {
+       return req_op(scsi_cmd_to_rq(cmd)) == REQ_OP_READ;
+ }
+ static bool ufshpb_is_write_or_discard(struct scsi_cmnd *cmd)
+ {
+       return op_is_write(req_op(scsi_cmd_to_rq(cmd))) ||
+              op_is_discard(req_op(scsi_cmd_to_rq(cmd)));
+ }
+ static bool ufshpb_is_supported_chunk(struct ufshpb_lu *hpb, int transfer_len)
+ {
+       return transfer_len <= hpb->pre_req_max_tr_len;
+ }
+ static bool ufshpb_is_general_lun(int lun)
+ {
+       return lun < UFS_UPIU_MAX_UNIT_NUM_ID;
+ }
+ static bool ufshpb_is_pinned_region(struct ufshpb_lu *hpb, int rgn_idx)
+ {
+       return hpb->lu_pinned_end != PINNED_NOT_SET &&
+              rgn_idx >= hpb->lu_pinned_start && rgn_idx <= hpb->lu_pinned_end;
+ }
+ static void ufshpb_kick_map_work(struct ufshpb_lu *hpb)
+ {
+       bool ret = false;
+       unsigned long flags;
+       if (ufshpb_get_state(hpb) != HPB_PRESENT)
+               return;
+       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       if (!list_empty(&hpb->lh_inact_rgn) || !list_empty(&hpb->lh_act_srgn))
+               ret = true;
+       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+       if (ret)
+               queue_work(ufshpb_wq, &hpb->map_work);
+ }
+ static bool ufshpb_is_hpb_rsp_valid(struct ufs_hba *hba,
+                                   struct ufshcd_lrb *lrbp,
+                                   struct utp_hpb_rsp *rsp_field)
+ {
+       /* Check HPB_UPDATE_ALERT */
+       if (!(lrbp->ucd_rsp_ptr->header.dword_2 &
+             UPIU_HEADER_DWORD(0, 2, 0, 0)))
+               return false;
+       if (be16_to_cpu(rsp_field->sense_data_len) != DEV_SENSE_SEG_LEN ||
+           rsp_field->desc_type != DEV_DES_TYPE ||
+           rsp_field->additional_len != DEV_ADDITIONAL_LEN ||
+           rsp_field->active_rgn_cnt > MAX_ACTIVE_NUM ||
+           rsp_field->inactive_rgn_cnt > MAX_INACTIVE_NUM ||
+           rsp_field->hpb_op == HPB_RSP_NONE ||
+           (rsp_field->hpb_op == HPB_RSP_REQ_REGION_UPDATE &&
+            !rsp_field->active_rgn_cnt && !rsp_field->inactive_rgn_cnt))
+               return false;
+       if (!ufshpb_is_general_lun(rsp_field->lun)) {
+               dev_warn(hba->dev, "ufshpb: lun(%d) not supported\n",
+                        lrbp->lun);
+               return false;
+       }
+       return true;
+ }
+ static void ufshpb_iterate_rgn(struct ufshpb_lu *hpb, int rgn_idx, int srgn_idx,
+                              int srgn_offset, int cnt, bool set_dirty)
+ {
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn, *prev_srgn = NULL;
+       int set_bit_len;
+       int bitmap_len;
+       unsigned long flags;
+ next_srgn:
+       rgn = hpb->rgn_tbl + rgn_idx;
+       srgn = rgn->srgn_tbl + srgn_idx;
+       if (likely(!srgn->is_last))
+               bitmap_len = hpb->entries_per_srgn;
+       else
+               bitmap_len = hpb->last_srgn_entries;
+       if ((srgn_offset + cnt) > bitmap_len)
+               set_bit_len = bitmap_len - srgn_offset;
+       else
+               set_bit_len = cnt;
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       if (rgn->rgn_state != HPB_RGN_INACTIVE) {
+               if (set_dirty) {
+                       if (srgn->srgn_state == HPB_SRGN_VALID)
+                               bitmap_set(srgn->mctx->ppn_dirty, srgn_offset,
+                                          set_bit_len);
+               } else if (hpb->is_hcm) {
+                        /* rewind the read timer for lru regions */
+                       rgn->read_timeout = ktime_add_ms(ktime_get(),
+                                       rgn->hpb->params.read_timeout_ms);
+                       rgn->read_timeout_expiries =
+                               rgn->hpb->params.read_timeout_expiries;
+               }
+       }
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       if (hpb->is_hcm && prev_srgn != srgn) {
+               bool activate = false;
+               spin_lock(&rgn->rgn_lock);
+               if (set_dirty) {
+                       rgn->reads -= srgn->reads;
+                       srgn->reads = 0;
+                       set_bit(RGN_FLAG_DIRTY, &rgn->rgn_flags);
+               } else {
+                       srgn->reads++;
+                       rgn->reads++;
+                       if (srgn->reads == hpb->params.activation_thld)
+                               activate = true;
+               }
+               spin_unlock(&rgn->rgn_lock);
+               if (activate ||
+                   test_and_clear_bit(RGN_FLAG_UPDATE, &rgn->rgn_flags)) {
+                       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+                       ufshpb_update_active_info(hpb, rgn_idx, srgn_idx);
+                       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+                       dev_dbg(&hpb->sdev_ufs_lu->sdev_dev,
+                               "activate region %d-%d\n", rgn_idx, srgn_idx);
+               }
+               prev_srgn = srgn;
+       }
+       srgn_offset = 0;
+       if (++srgn_idx == hpb->srgns_per_rgn) {
+               srgn_idx = 0;
+               rgn_idx++;
+       }
+       cnt -= set_bit_len;
+       if (cnt > 0)
+               goto next_srgn;
+ }
+ static bool ufshpb_test_ppn_dirty(struct ufshpb_lu *hpb, int rgn_idx,
+                                 int srgn_idx, int srgn_offset, int cnt)
+ {
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       int bitmap_len;
+       int bit_len;
+ next_srgn:
+       rgn = hpb->rgn_tbl + rgn_idx;
+       srgn = rgn->srgn_tbl + srgn_idx;
+       if (likely(!srgn->is_last))
+               bitmap_len = hpb->entries_per_srgn;
+       else
+               bitmap_len = hpb->last_srgn_entries;
+       if (!ufshpb_is_valid_srgn(rgn, srgn))
+               return true;
+       /*
+        * If the region state is active, mctx must be allocated.
+        * In this case, check whether the region is evicted or
+        * mctx allocation fail.
+        */
+       if (unlikely(!srgn->mctx)) {
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                       "no mctx in region %d subregion %d.\n",
+                       srgn->rgn_idx, srgn->srgn_idx);
+               return true;
+       }
+       if ((srgn_offset + cnt) > bitmap_len)
+               bit_len = bitmap_len - srgn_offset;
+       else
+               bit_len = cnt;
+       if (find_next_bit(srgn->mctx->ppn_dirty, bit_len + srgn_offset,
+                         srgn_offset) < bit_len + srgn_offset)
+               return true;
+       srgn_offset = 0;
+       if (++srgn_idx == hpb->srgns_per_rgn) {
+               srgn_idx = 0;
+               rgn_idx++;
+       }
+       cnt -= bit_len;
+       if (cnt > 0)
+               goto next_srgn;
+       return false;
+ }
+ static inline bool is_rgn_dirty(struct ufshpb_region *rgn)
+ {
+       return test_bit(RGN_FLAG_DIRTY, &rgn->rgn_flags);
+ }
+ static int ufshpb_fill_ppn_from_page(struct ufshpb_lu *hpb,
+                                    struct ufshpb_map_ctx *mctx, int pos,
+                                    int len, __be64 *ppn_buf)
+ {
+       struct page *page;
+       int index, offset;
+       int copied;
+       index = pos / (PAGE_SIZE / HPB_ENTRY_SIZE);
+       offset = pos % (PAGE_SIZE / HPB_ENTRY_SIZE);
+       if ((offset + len) <= (PAGE_SIZE / HPB_ENTRY_SIZE))
+               copied = len;
+       else
+               copied = (PAGE_SIZE / HPB_ENTRY_SIZE) - offset;
+       page = mctx->m_page[index];
+       if (unlikely(!page)) {
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                       "error. cannot find page in mctx\n");
+               return -ENOMEM;
+       }
+       memcpy(ppn_buf, page_address(page) + (offset * HPB_ENTRY_SIZE),
+              copied * HPB_ENTRY_SIZE);
+       return copied;
+ }
+ static void
+ ufshpb_get_pos_from_lpn(struct ufshpb_lu *hpb, unsigned long lpn, int *rgn_idx,
+                       int *srgn_idx, int *offset)
+ {
+       int rgn_offset;
+       *rgn_idx = lpn >> hpb->entries_per_rgn_shift;
+       rgn_offset = lpn & hpb->entries_per_rgn_mask;
+       *srgn_idx = rgn_offset >> hpb->entries_per_srgn_shift;
+       *offset = rgn_offset & hpb->entries_per_srgn_mask;
+ }
+ static void
+ ufshpb_set_hpb_read_to_upiu(struct ufs_hba *hba, struct ufshcd_lrb *lrbp,
+                           __be64 ppn, u8 transfer_len)
+ {
+       unsigned char *cdb = lrbp->cmd->cmnd;
+       __be64 ppn_tmp = ppn;
+       cdb[0] = UFSHPB_READ;
+       if (hba->dev_quirks & UFS_DEVICE_QUIRK_SWAP_L2P_ENTRY_FOR_HPB_READ)
+               ppn_tmp = (__force __be64)swab64((__force u64)ppn);
+       /* ppn value is stored as big-endian in the host memory */
+       memcpy(&cdb[6], &ppn_tmp, sizeof(__be64));
+       cdb[14] = transfer_len;
+       cdb[15] = 0;
+       lrbp->cmd->cmd_len = UFS_CDB_SIZE;
+ }
+ /*
+  * This function will set up HPB read command using host-side L2P map data.
+  */
+ int ufshpb_prep(struct ufs_hba *hba, struct ufshcd_lrb *lrbp)
+ {
+       struct ufshpb_lu *hpb;
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       struct scsi_cmnd *cmd = lrbp->cmd;
+       u32 lpn;
+       __be64 ppn;
+       unsigned long flags;
+       int transfer_len, rgn_idx, srgn_idx, srgn_offset;
+       int err = 0;
+       hpb = ufshpb_get_hpb_data(cmd->device);
+       if (!hpb)
+               return -ENODEV;
+       if (ufshpb_get_state(hpb) == HPB_INIT)
+               return -ENODEV;
+       if (ufshpb_get_state(hpb) != HPB_PRESENT) {
+               dev_notice(&hpb->sdev_ufs_lu->sdev_dev,
+                          "%s: ufshpb state is not PRESENT", __func__);
+               return -ENODEV;
+       }
+       if (blk_rq_is_passthrough(scsi_cmd_to_rq(cmd)) ||
+           (!ufshpb_is_write_or_discard(cmd) &&
+            !ufshpb_is_read_cmd(cmd)))
+               return 0;
+       transfer_len = sectors_to_logical(cmd->device,
+                                         blk_rq_sectors(scsi_cmd_to_rq(cmd)));
+       if (unlikely(!transfer_len))
+               return 0;
+       lpn = sectors_to_logical(cmd->device, blk_rq_pos(scsi_cmd_to_rq(cmd)));
+       ufshpb_get_pos_from_lpn(hpb, lpn, &rgn_idx, &srgn_idx, &srgn_offset);
+       rgn = hpb->rgn_tbl + rgn_idx;
+       srgn = rgn->srgn_tbl + srgn_idx;
+       /* If command type is WRITE or DISCARD, set bitmap as drity */
+       if (ufshpb_is_write_or_discard(cmd)) {
+               ufshpb_iterate_rgn(hpb, rgn_idx, srgn_idx, srgn_offset,
+                                  transfer_len, true);
+               return 0;
+       }
+       if (!ufshpb_is_supported_chunk(hpb, transfer_len))
+               return 0;
+       if (hpb->is_hcm) {
+               /*
+                * in host control mode, reads are the main source for
+                * activation trials.
+                */
+               ufshpb_iterate_rgn(hpb, rgn_idx, srgn_idx, srgn_offset,
+                                  transfer_len, false);
+               /* keep those counters normalized */
+               if (rgn->reads > hpb->entries_per_srgn)
+                       schedule_work(&hpb->ufshpb_normalization_work);
+       }
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       if (ufshpb_test_ppn_dirty(hpb, rgn_idx, srgn_idx, srgn_offset,
+                                  transfer_len)) {
+               hpb->stats.miss_cnt++;
+               spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+               return 0;
+       }
+       err = ufshpb_fill_ppn_from_page(hpb, srgn->mctx, srgn_offset, 1, &ppn);
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       if (unlikely(err < 0)) {
+               /*
+                * In this case, the region state is active,
+                * but the ppn table is not allocated.
+                * Make sure that ppn table must be allocated on
+                * active state.
+                */
+               dev_err(hba->dev, "get ppn failed. err %d\n", err);
+               return err;
+       }
+       ufshpb_set_hpb_read_to_upiu(hba, lrbp, ppn, transfer_len);
+       hpb->stats.hit_cnt++;
+       return 0;
+ }
+ static struct ufshpb_req *ufshpb_get_req(struct ufshpb_lu *hpb,
+                                        int rgn_idx, enum req_opf dir,
+                                        bool atomic)
+ {
+       struct ufshpb_req *rq;
+       struct request *req;
+       int retries = HPB_MAP_REQ_RETRIES;
+       rq = kmem_cache_alloc(hpb->map_req_cache, GFP_KERNEL);
+       if (!rq)
+               return NULL;
+ retry:
+       req = blk_mq_alloc_request(hpb->sdev_ufs_lu->request_queue, dir,
+                             BLK_MQ_REQ_NOWAIT);
+       if (!atomic && (PTR_ERR(req) == -EWOULDBLOCK) && (--retries > 0)) {
+               usleep_range(3000, 3100);
+               goto retry;
+       }
+       if (IS_ERR(req))
+               goto free_rq;
+       rq->hpb = hpb;
+       rq->req = req;
+       rq->rb.rgn_idx = rgn_idx;
+       return rq;
+ free_rq:
+       kmem_cache_free(hpb->map_req_cache, rq);
+       return NULL;
+ }
+ static void ufshpb_put_req(struct ufshpb_lu *hpb, struct ufshpb_req *rq)
+ {
+       blk_mq_free_request(rq->req);
+       kmem_cache_free(hpb->map_req_cache, rq);
+ }
+ static struct ufshpb_req *ufshpb_get_map_req(struct ufshpb_lu *hpb,
+                                            struct ufshpb_subregion *srgn)
+ {
+       struct ufshpb_req *map_req;
+       struct bio *bio;
+       unsigned long flags;
+       if (hpb->is_hcm &&
+           hpb->num_inflight_map_req >= hpb->params.inflight_map_req) {
+               dev_info(&hpb->sdev_ufs_lu->sdev_dev,
+                        "map_req throttle. inflight %d throttle %d",
+                        hpb->num_inflight_map_req,
+                        hpb->params.inflight_map_req);
+               return NULL;
+       }
+       map_req = ufshpb_get_req(hpb, srgn->rgn_idx, REQ_OP_DRV_IN, false);
+       if (!map_req)
+               return NULL;
+       bio = bio_alloc(NULL, hpb->pages_per_srgn, 0, GFP_KERNEL);
+       if (!bio) {
+               ufshpb_put_req(hpb, map_req);
+               return NULL;
+       }
+       map_req->bio = bio;
+       map_req->rb.srgn_idx = srgn->srgn_idx;
+       map_req->rb.mctx = srgn->mctx;
+       spin_lock_irqsave(&hpb->param_lock, flags);
+       hpb->num_inflight_map_req++;
+       spin_unlock_irqrestore(&hpb->param_lock, flags);
+       return map_req;
+ }
+ static void ufshpb_put_map_req(struct ufshpb_lu *hpb,
+                              struct ufshpb_req *map_req)
+ {
+       unsigned long flags;
+       bio_put(map_req->bio);
+       ufshpb_put_req(hpb, map_req);
+       spin_lock_irqsave(&hpb->param_lock, flags);
+       hpb->num_inflight_map_req--;
+       spin_unlock_irqrestore(&hpb->param_lock, flags);
+ }
+ static int ufshpb_clear_dirty_bitmap(struct ufshpb_lu *hpb,
+                                    struct ufshpb_subregion *srgn)
+ {
+       struct ufshpb_region *rgn;
+       u32 num_entries = hpb->entries_per_srgn;
+       if (!srgn->mctx) {
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                       "no mctx in region %d subregion %d.\n",
+                       srgn->rgn_idx, srgn->srgn_idx);
+               return -1;
+       }
+       if (unlikely(srgn->is_last))
+               num_entries = hpb->last_srgn_entries;
+       bitmap_zero(srgn->mctx->ppn_dirty, num_entries);
+       rgn = hpb->rgn_tbl + srgn->rgn_idx;
+       clear_bit(RGN_FLAG_DIRTY, &rgn->rgn_flags);
+       return 0;
+ }
+ static void ufshpb_update_active_info(struct ufshpb_lu *hpb, int rgn_idx,
+                                     int srgn_idx)
+ {
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       rgn = hpb->rgn_tbl + rgn_idx;
+       srgn = rgn->srgn_tbl + srgn_idx;
+       list_del_init(&rgn->list_inact_rgn);
+       if (list_empty(&srgn->list_act_srgn))
+               list_add_tail(&srgn->list_act_srgn, &hpb->lh_act_srgn);
+       hpb->stats.rcmd_active_cnt++;
+ }
+ static void ufshpb_update_inactive_info(struct ufshpb_lu *hpb, int rgn_idx)
+ {
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       int srgn_idx;
+       rgn = hpb->rgn_tbl + rgn_idx;
+       for_each_sub_region(rgn, srgn_idx, srgn)
+               list_del_init(&srgn->list_act_srgn);
+       if (list_empty(&rgn->list_inact_rgn))
+               list_add_tail(&rgn->list_inact_rgn, &hpb->lh_inact_rgn);
+       hpb->stats.rcmd_inactive_cnt++;
+ }
+ static void ufshpb_activate_subregion(struct ufshpb_lu *hpb,
+                                     struct ufshpb_subregion *srgn)
+ {
+       struct ufshpb_region *rgn;
+       /*
+        * If there is no mctx in subregion
+        * after I/O progress for HPB_READ_BUFFER, the region to which the
+        * subregion belongs was evicted.
+        * Make sure the region must not evict in I/O progress
+        */
+       if (!srgn->mctx) {
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                       "no mctx in region %d subregion %d.\n",
+                       srgn->rgn_idx, srgn->srgn_idx);
+               srgn->srgn_state = HPB_SRGN_INVALID;
+               return;
+       }
+       rgn = hpb->rgn_tbl + srgn->rgn_idx;
+       if (unlikely(rgn->rgn_state == HPB_RGN_INACTIVE)) {
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                       "region %d subregion %d evicted\n",
+                       srgn->rgn_idx, srgn->srgn_idx);
+               srgn->srgn_state = HPB_SRGN_INVALID;
+               return;
+       }
+       srgn->srgn_state = HPB_SRGN_VALID;
+ }
+ static void ufshpb_umap_req_compl_fn(struct request *req, blk_status_t error)
+ {
+       struct ufshpb_req *umap_req = (struct ufshpb_req *)req->end_io_data;
+       ufshpb_put_req(umap_req->hpb, umap_req);
+ }
+ static void ufshpb_map_req_compl_fn(struct request *req, blk_status_t error)
+ {
+       struct ufshpb_req *map_req = (struct ufshpb_req *) req->end_io_data;
+       struct ufshpb_lu *hpb = map_req->hpb;
+       struct ufshpb_subregion *srgn;
+       unsigned long flags;
+       srgn = hpb->rgn_tbl[map_req->rb.rgn_idx].srgn_tbl +
+               map_req->rb.srgn_idx;
+       ufshpb_clear_dirty_bitmap(hpb, srgn);
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       ufshpb_activate_subregion(hpb, srgn);
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       ufshpb_put_map_req(map_req->hpb, map_req);
+ }
+ static void ufshpb_set_unmap_cmd(unsigned char *cdb, struct ufshpb_region *rgn)
+ {
+       cdb[0] = UFSHPB_WRITE_BUFFER;
+       cdb[1] = rgn ? UFSHPB_WRITE_BUFFER_INACT_SINGLE_ID :
+                         UFSHPB_WRITE_BUFFER_INACT_ALL_ID;
+       if (rgn)
+               put_unaligned_be16(rgn->rgn_idx, &cdb[2]);
+       cdb[9] = 0x00;
+ }
+ static void ufshpb_set_read_buf_cmd(unsigned char *cdb, int rgn_idx,
+                                   int srgn_idx, int srgn_mem_size)
+ {
+       cdb[0] = UFSHPB_READ_BUFFER;
+       cdb[1] = UFSHPB_READ_BUFFER_ID;
+       put_unaligned_be16(rgn_idx, &cdb[2]);
+       put_unaligned_be16(srgn_idx, &cdb[4]);
+       put_unaligned_be24(srgn_mem_size, &cdb[6]);
+       cdb[9] = 0x00;
+ }
+ static void ufshpb_execute_umap_req(struct ufshpb_lu *hpb,
+                                  struct ufshpb_req *umap_req,
+                                  struct ufshpb_region *rgn)
+ {
+       struct request *req = umap_req->req;
+       struct scsi_cmnd *scmd = blk_mq_rq_to_pdu(req);
+       req->timeout = 0;
+       req->end_io_data = umap_req;
++      req->end_io = ufshpb_umap_req_compl_fn;
+       ufshpb_set_unmap_cmd(scmd->cmnd, rgn);
+       scmd->cmd_len = HPB_WRITE_BUFFER_CMD_LENGTH;
 -      blk_execute_rq_nowait(req, true, ufshpb_map_req_compl_fn);
++      blk_execute_rq_nowait(req, true);
+       hpb->stats.umap_req_cnt++;
+ }
+ static int ufshpb_execute_map_req(struct ufshpb_lu *hpb,
+                                 struct ufshpb_req *map_req, bool last)
+ {
+       struct request_queue *q;
+       struct request *req;
+       struct scsi_cmnd *scmd;
+       int mem_size = hpb->srgn_mem_size;
+       int ret = 0;
+       int i;
+       q = hpb->sdev_ufs_lu->request_queue;
+       for (i = 0; i < hpb->pages_per_srgn; i++) {
+               ret = bio_add_pc_page(q, map_req->bio, map_req->rb.mctx->m_page[i],
+                                     PAGE_SIZE, 0);
+               if (ret != PAGE_SIZE) {
+                       dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                                  "bio_add_pc_page fail %d - %d\n",
+                                  map_req->rb.rgn_idx, map_req->rb.srgn_idx);
+                       return ret;
+               }
+       }
+       req = map_req->req;
+       blk_rq_append_bio(req, map_req->bio);
+       req->end_io_data = map_req;
++      req->end_io = ufshpb_map_req_compl_fn;
+       if (unlikely(last))
+               mem_size = hpb->last_srgn_entries * HPB_ENTRY_SIZE;
+       scmd = blk_mq_rq_to_pdu(req);
+       ufshpb_set_read_buf_cmd(scmd->cmnd, map_req->rb.rgn_idx,
+                               map_req->rb.srgn_idx, mem_size);
+       scmd->cmd_len = HPB_READ_BUFFER_CMD_LENGTH;
 -      data_seg_len = be32_to_cpu(lrbp->ucd_rsp_ptr->header.dword_2)
 -              & MASK_RSP_UPIU_DATA_SEG_LEN;
 -
 -      /* To flush remained rsp_list, we queue the map_work task */
 -      if (!data_seg_len) {
 -              if (!ufshpb_is_general_lun(hpb->lun))
 -                      return;
 -
 -              ufshpb_kick_map_work(hpb);
 -              return;
 -      }
 -
++      blk_execute_rq_nowait(req, true);
+       hpb->stats.map_req_cnt++;
+       return 0;
+ }
+ static struct ufshpb_map_ctx *ufshpb_get_map_ctx(struct ufshpb_lu *hpb,
+                                                bool last)
+ {
+       struct ufshpb_map_ctx *mctx;
+       u32 num_entries = hpb->entries_per_srgn;
+       int i, j;
+       mctx = mempool_alloc(ufshpb_mctx_pool, GFP_KERNEL);
+       if (!mctx)
+               return NULL;
+       mctx->m_page = kmem_cache_alloc(hpb->m_page_cache, GFP_KERNEL);
+       if (!mctx->m_page)
+               goto release_mctx;
+       if (unlikely(last))
+               num_entries = hpb->last_srgn_entries;
+       mctx->ppn_dirty = bitmap_zalloc(num_entries, GFP_KERNEL);
+       if (!mctx->ppn_dirty)
+               goto release_m_page;
+       for (i = 0; i < hpb->pages_per_srgn; i++) {
+               mctx->m_page[i] = mempool_alloc(ufshpb_page_pool, GFP_KERNEL);
+               if (!mctx->m_page[i]) {
+                       for (j = 0; j < i; j++)
+                               mempool_free(mctx->m_page[j], ufshpb_page_pool);
+                       goto release_ppn_dirty;
+               }
+               clear_page(page_address(mctx->m_page[i]));
+       }
+       return mctx;
+ release_ppn_dirty:
+       bitmap_free(mctx->ppn_dirty);
+ release_m_page:
+       kmem_cache_free(hpb->m_page_cache, mctx->m_page);
+ release_mctx:
+       mempool_free(mctx, ufshpb_mctx_pool);
+       return NULL;
+ }
+ static void ufshpb_put_map_ctx(struct ufshpb_lu *hpb,
+                              struct ufshpb_map_ctx *mctx)
+ {
+       int i;
+       for (i = 0; i < hpb->pages_per_srgn; i++)
+               mempool_free(mctx->m_page[i], ufshpb_page_pool);
+       bitmap_free(mctx->ppn_dirty);
+       kmem_cache_free(hpb->m_page_cache, mctx->m_page);
+       mempool_free(mctx, ufshpb_mctx_pool);
+ }
+ static int ufshpb_check_srgns_issue_state(struct ufshpb_lu *hpb,
+                                         struct ufshpb_region *rgn)
+ {
+       struct ufshpb_subregion *srgn;
+       int srgn_idx;
+       for_each_sub_region(rgn, srgn_idx, srgn)
+               if (srgn->srgn_state == HPB_SRGN_ISSUED)
+                       return -EPERM;
+       return 0;
+ }
+ static void ufshpb_read_to_handler(struct work_struct *work)
+ {
+       struct ufshpb_lu *hpb = container_of(work, struct ufshpb_lu,
+                                            ufshpb_read_to_work.work);
+       struct victim_select_info *lru_info = &hpb->lru_info;
+       struct ufshpb_region *rgn, *next_rgn;
+       unsigned long flags;
+       unsigned int poll;
+       LIST_HEAD(expired_list);
+       if (test_and_set_bit(TIMEOUT_WORK_RUNNING, &hpb->work_data_bits))
+               return;
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       list_for_each_entry_safe(rgn, next_rgn, &lru_info->lh_lru_rgn,
+                                list_lru_rgn) {
+               bool timedout = ktime_after(ktime_get(), rgn->read_timeout);
+               if (timedout) {
+                       rgn->read_timeout_expiries--;
+                       if (is_rgn_dirty(rgn) ||
+                           rgn->read_timeout_expiries == 0)
+                               list_add(&rgn->list_expired_rgn, &expired_list);
+                       else
+                               rgn->read_timeout = ktime_add_ms(ktime_get(),
+                                               hpb->params.read_timeout_ms);
+               }
+       }
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       list_for_each_entry_safe(rgn, next_rgn, &expired_list,
+                                list_expired_rgn) {
+               list_del_init(&rgn->list_expired_rgn);
+               spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+               ufshpb_update_inactive_info(hpb, rgn->rgn_idx);
+               spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+       }
+       ufshpb_kick_map_work(hpb);
+       clear_bit(TIMEOUT_WORK_RUNNING, &hpb->work_data_bits);
+       poll = hpb->params.timeout_polling_interval_ms;
+       schedule_delayed_work(&hpb->ufshpb_read_to_work,
+                             msecs_to_jiffies(poll));
+ }
+ static void ufshpb_add_lru_info(struct victim_select_info *lru_info,
+                               struct ufshpb_region *rgn)
+ {
+       rgn->rgn_state = HPB_RGN_ACTIVE;
+       list_add_tail(&rgn->list_lru_rgn, &lru_info->lh_lru_rgn);
+       atomic_inc(&lru_info->active_cnt);
+       if (rgn->hpb->is_hcm) {
+               rgn->read_timeout =
+                       ktime_add_ms(ktime_get(),
+                                    rgn->hpb->params.read_timeout_ms);
+               rgn->read_timeout_expiries =
+                       rgn->hpb->params.read_timeout_expiries;
+       }
+ }
+ static void ufshpb_hit_lru_info(struct victim_select_info *lru_info,
+                               struct ufshpb_region *rgn)
+ {
+       list_move_tail(&rgn->list_lru_rgn, &lru_info->lh_lru_rgn);
+ }
+ static struct ufshpb_region *ufshpb_victim_lru_info(struct ufshpb_lu *hpb)
+ {
+       struct victim_select_info *lru_info = &hpb->lru_info;
+       struct ufshpb_region *rgn, *victim_rgn = NULL;
+       list_for_each_entry(rgn, &lru_info->lh_lru_rgn, list_lru_rgn) {
+               if (ufshpb_check_srgns_issue_state(hpb, rgn))
+                       continue;
+               /*
+                * in host control mode, verify that the exiting region
+                * has fewer reads
+                */
+               if (hpb->is_hcm &&
+                   rgn->reads > hpb->params.eviction_thld_exit)
+                       continue;
+               victim_rgn = rgn;
+               break;
+       }
+       if (!victim_rgn)
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                       "%s: no region allocated\n",
+                       __func__);
+       return victim_rgn;
+ }
+ static void ufshpb_cleanup_lru_info(struct victim_select_info *lru_info,
+                                   struct ufshpb_region *rgn)
+ {
+       list_del_init(&rgn->list_lru_rgn);
+       rgn->rgn_state = HPB_RGN_INACTIVE;
+       atomic_dec(&lru_info->active_cnt);
+ }
+ static void ufshpb_purge_active_subregion(struct ufshpb_lu *hpb,
+                                         struct ufshpb_subregion *srgn)
+ {
+       if (srgn->srgn_state != HPB_SRGN_UNUSED) {
+               ufshpb_put_map_ctx(hpb, srgn->mctx);
+               srgn->srgn_state = HPB_SRGN_UNUSED;
+               srgn->mctx = NULL;
+       }
+ }
+ static int ufshpb_issue_umap_req(struct ufshpb_lu *hpb,
+                                struct ufshpb_region *rgn,
+                                bool atomic)
+ {
+       struct ufshpb_req *umap_req;
+       int rgn_idx = rgn ? rgn->rgn_idx : 0;
+       umap_req = ufshpb_get_req(hpb, rgn_idx, REQ_OP_DRV_OUT, atomic);
+       if (!umap_req)
+               return -ENOMEM;
+       ufshpb_execute_umap_req(hpb, umap_req, rgn);
+       return 0;
+ }
+ static int ufshpb_issue_umap_single_req(struct ufshpb_lu *hpb,
+                                       struct ufshpb_region *rgn)
+ {
+       return ufshpb_issue_umap_req(hpb, rgn, true);
+ }
+ static void __ufshpb_evict_region(struct ufshpb_lu *hpb,
+                                struct ufshpb_region *rgn)
+ {
+       struct victim_select_info *lru_info;
+       struct ufshpb_subregion *srgn;
+       int srgn_idx;
+       lru_info = &hpb->lru_info;
+       dev_dbg(&hpb->sdev_ufs_lu->sdev_dev, "evict region %d\n", rgn->rgn_idx);
+       ufshpb_cleanup_lru_info(lru_info, rgn);
+       for_each_sub_region(rgn, srgn_idx, srgn)
+               ufshpb_purge_active_subregion(hpb, srgn);
+ }
+ static int ufshpb_evict_region(struct ufshpb_lu *hpb, struct ufshpb_region *rgn)
+ {
+       unsigned long flags;
+       int ret = 0;
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       if (rgn->rgn_state == HPB_RGN_PINNED) {
+               dev_warn(&hpb->sdev_ufs_lu->sdev_dev,
+                        "pinned region cannot drop-out. region %d\n",
+                        rgn->rgn_idx);
+               goto out;
+       }
+       if (!list_empty(&rgn->list_lru_rgn)) {
+               if (ufshpb_check_srgns_issue_state(hpb, rgn)) {
+                       ret = -EBUSY;
+                       goto out;
+               }
+               if (hpb->is_hcm) {
+                       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+                       ret = ufshpb_issue_umap_single_req(hpb, rgn);
+                       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+                       if (ret)
+                               goto out;
+               }
+               __ufshpb_evict_region(hpb, rgn);
+       }
+ out:
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       return ret;
+ }
+ static int ufshpb_issue_map_req(struct ufshpb_lu *hpb,
+                               struct ufshpb_region *rgn,
+                               struct ufshpb_subregion *srgn)
+ {
+       struct ufshpb_req *map_req;
+       unsigned long flags;
+       int ret;
+       int err = -EAGAIN;
+       bool alloc_required = false;
+       enum HPB_SRGN_STATE state = HPB_SRGN_INVALID;
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       if (ufshpb_get_state(hpb) != HPB_PRESENT) {
+               dev_notice(&hpb->sdev_ufs_lu->sdev_dev,
+                          "%s: ufshpb state is not PRESENT\n", __func__);
+               goto unlock_out;
+       }
+       if ((rgn->rgn_state == HPB_RGN_INACTIVE) &&
+           (srgn->srgn_state == HPB_SRGN_INVALID)) {
+               err = 0;
+               goto unlock_out;
+       }
+       if (srgn->srgn_state == HPB_SRGN_UNUSED)
+               alloc_required = true;
+       /*
+        * If the subregion is already ISSUED state,
+        * a specific event (e.g., GC or wear-leveling, etc.) occurs in
+        * the device and HPB response for map loading is received.
+        * In this case, after finishing the HPB_READ_BUFFER,
+        * the next HPB_READ_BUFFER is performed again to obtain the latest
+        * map data.
+        */
+       if (srgn->srgn_state == HPB_SRGN_ISSUED)
+               goto unlock_out;
+       srgn->srgn_state = HPB_SRGN_ISSUED;
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       if (alloc_required) {
+               srgn->mctx = ufshpb_get_map_ctx(hpb, srgn->is_last);
+               if (!srgn->mctx) {
+                       dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                           "get map_ctx failed. region %d - %d\n",
+                           rgn->rgn_idx, srgn->srgn_idx);
+                       state = HPB_SRGN_UNUSED;
+                       goto change_srgn_state;
+               }
+       }
+       map_req = ufshpb_get_map_req(hpb, srgn);
+       if (!map_req)
+               goto change_srgn_state;
+       ret = ufshpb_execute_map_req(hpb, map_req, srgn->is_last);
+       if (ret) {
+               dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                          "%s: issue map_req failed: %d, region %d - %d\n",
+                          __func__, ret, srgn->rgn_idx, srgn->srgn_idx);
+               goto free_map_req;
+       }
+       return 0;
+ free_map_req:
+       ufshpb_put_map_req(hpb, map_req);
+ change_srgn_state:
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       srgn->srgn_state = state;
+ unlock_out:
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       return err;
+ }
+ static int ufshpb_add_region(struct ufshpb_lu *hpb, struct ufshpb_region *rgn)
+ {
+       struct ufshpb_region *victim_rgn = NULL;
+       struct victim_select_info *lru_info = &hpb->lru_info;
+       unsigned long flags;
+       int ret = 0;
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       /*
+        * If region belongs to lru_list, just move the region
+        * to the front of lru list because the state of the region
+        * is already active-state.
+        */
+       if (!list_empty(&rgn->list_lru_rgn)) {
+               ufshpb_hit_lru_info(lru_info, rgn);
+               goto out;
+       }
+       if (rgn->rgn_state == HPB_RGN_INACTIVE) {
+               if (atomic_read(&lru_info->active_cnt) ==
+                   lru_info->max_lru_active_cnt) {
+                       /*
+                        * If the maximum number of active regions
+                        * is exceeded, evict the least recently used region.
+                        * This case may occur when the device responds
+                        * to the eviction information late.
+                        * It is okay to evict the least recently used region,
+                        * because the device could detect this region
+                        * by not issuing HPB_READ
+                        *
+                        * in host control mode, verify that the entering
+                        * region has enough reads
+                        */
+                       if (hpb->is_hcm &&
+                           rgn->reads < hpb->params.eviction_thld_enter) {
+                               ret = -EACCES;
+                               goto out;
+                       }
+                       victim_rgn = ufshpb_victim_lru_info(hpb);
+                       if (!victim_rgn) {
+                               dev_warn(&hpb->sdev_ufs_lu->sdev_dev,
+                                   "cannot get victim region %s\n",
+                                   hpb->is_hcm ? "" : "error");
+                               ret = -ENOMEM;
+                               goto out;
+                       }
+                       dev_dbg(&hpb->sdev_ufs_lu->sdev_dev,
+                               "LRU full (%d), choose victim %d\n",
+                               atomic_read(&lru_info->active_cnt),
+                               victim_rgn->rgn_idx);
+                       if (hpb->is_hcm) {
+                               spin_unlock_irqrestore(&hpb->rgn_state_lock,
+                                                      flags);
+                               ret = ufshpb_issue_umap_single_req(hpb,
+                                                               victim_rgn);
+                               spin_lock_irqsave(&hpb->rgn_state_lock,
+                                                 flags);
+                               if (ret)
+                                       goto out;
+                       }
+                       __ufshpb_evict_region(hpb, victim_rgn);
+               }
+               /*
+                * When a region is added to lru_info list_head,
+                * it is guaranteed that the subregion has been
+                * assigned all mctx. If failed, try to receive mctx again
+                * without being added to lru_info list_head
+                */
+               ufshpb_add_lru_info(lru_info, rgn);
+       }
+ out:
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+       return ret;
+ }
+ /**
+  *ufshpb_submit_region_inactive() - submit a region to be inactivated later
+  *@hpb: per-LU HPB instance
+  *@region_index: the index associated with the region that will be inactivated later
+  */
+ static void ufshpb_submit_region_inactive(struct ufshpb_lu *hpb, int region_index)
+ {
+       int subregion_index;
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       /*
+        * Remove this region from active region list and add it to inactive list
+        */
+       spin_lock(&hpb->rsp_list_lock);
+       ufshpb_update_inactive_info(hpb, region_index);
+       spin_unlock(&hpb->rsp_list_lock);
+       rgn = hpb->rgn_tbl + region_index;
+       /*
+        * Set subregion state to be HPB_SRGN_INVALID, there will no HPB read on this subregion
+        */
+       spin_lock(&hpb->rgn_state_lock);
+       if (rgn->rgn_state != HPB_RGN_INACTIVE) {
+               for (subregion_index = 0; subregion_index < rgn->srgn_cnt; subregion_index++) {
+                       srgn = rgn->srgn_tbl + subregion_index;
+                       if (srgn->srgn_state == HPB_SRGN_VALID)
+                               srgn->srgn_state = HPB_SRGN_INVALID;
+               }
+       }
+       spin_unlock(&hpb->rgn_state_lock);
+ }
+ static void ufshpb_rsp_req_region_update(struct ufshpb_lu *hpb,
+                                        struct utp_hpb_rsp *rsp_field)
+ {
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       int i, rgn_i, srgn_i;
+       BUILD_BUG_ON(sizeof(struct ufshpb_active_field) != HPB_ACT_FIELD_SIZE);
+       /*
+        * If the active region and the inactive region are the same,
+        * we will inactivate this region.
+        * The device could check this (region inactivated) and
+        * will response the proper active region information
+        */
+       for (i = 0; i < rsp_field->active_rgn_cnt; i++) {
+               rgn_i =
+                       be16_to_cpu(rsp_field->hpb_active_field[i].active_rgn);
+               srgn_i =
+                       be16_to_cpu(rsp_field->hpb_active_field[i].active_srgn);
+               rgn = hpb->rgn_tbl + rgn_i;
+               if (hpb->is_hcm &&
+                   (rgn->rgn_state != HPB_RGN_ACTIVE || is_rgn_dirty(rgn))) {
+                       /*
+                        * in host control mode, subregion activation
+                        * recommendations are only allowed to active regions.
+                        * Also, ignore recommendations for dirty regions - the
+                        * host will make decisions concerning those by himself
+                        */
+                       continue;
+               }
+               dev_dbg(&hpb->sdev_ufs_lu->sdev_dev,
+                       "activate(%d) region %d - %d\n", i, rgn_i, srgn_i);
+               spin_lock(&hpb->rsp_list_lock);
+               ufshpb_update_active_info(hpb, rgn_i, srgn_i);
+               spin_unlock(&hpb->rsp_list_lock);
+               srgn = rgn->srgn_tbl + srgn_i;
+               /* blocking HPB_READ */
+               spin_lock(&hpb->rgn_state_lock);
+               if (srgn->srgn_state == HPB_SRGN_VALID)
+                       srgn->srgn_state = HPB_SRGN_INVALID;
+               spin_unlock(&hpb->rgn_state_lock);
+       }
+       if (hpb->is_hcm) {
+               /*
+                * in host control mode the device is not allowed to inactivate
+                * regions
+                */
+               goto out;
+       }
+       for (i = 0; i < rsp_field->inactive_rgn_cnt; i++) {
+               rgn_i = be16_to_cpu(rsp_field->hpb_inactive_field[i]);
+               dev_dbg(&hpb->sdev_ufs_lu->sdev_dev, "inactivate(%d) region %d\n", i, rgn_i);
+               ufshpb_submit_region_inactive(hpb, rgn_i);
+       }
+ out:
+       dev_dbg(&hpb->sdev_ufs_lu->sdev_dev, "Noti: #ACT %u #INACT %u\n",
+               rsp_field->active_rgn_cnt, rsp_field->inactive_rgn_cnt);
+       if (ufshpb_get_state(hpb) == HPB_PRESENT)
+               queue_work(ufshpb_wq, &hpb->map_work);
+ }
+ /*
+  * Set the flags of all active regions to RGN_FLAG_UPDATE to let host side reload L2P entries later
+  */
+ static void ufshpb_set_regions_update(struct ufshpb_lu *hpb)
+ {
+       struct victim_select_info *lru_info = &hpb->lru_info;
+       struct ufshpb_region *rgn;
+       unsigned long flags;
+       spin_lock_irqsave(&hpb->rgn_state_lock, flags);
+       list_for_each_entry(rgn, &lru_info->lh_lru_rgn, list_lru_rgn)
+               set_bit(RGN_FLAG_UPDATE, &rgn->rgn_flags);
+       spin_unlock_irqrestore(&hpb->rgn_state_lock, flags);
+ }
+ static void ufshpb_dev_reset_handler(struct ufs_hba *hba)
+ {
+       struct scsi_device *sdev;
+       struct ufshpb_lu *hpb;
+       __shost_for_each_device(sdev, hba->host) {
+               hpb = ufshpb_get_hpb_data(sdev);
+               if (!hpb)
+                       continue;
+               if (hpb->is_hcm) {
+                       /*
+                        * For the HPB host control mode, in case device powered up and lost HPB
+                        * information, we will set the region flag to be RGN_FLAG_UPDATE, it will
+                        * let host reload its L2P entries(reactivate region in the UFS device).
+                        */
+                       ufshpb_set_regions_update(hpb);
+               } else {
+                       /*
+                        * For the HPB device control mode, if host side receives 02h:HPB Operation
+                        * in UPIU response, which means device recommends the host side should
+                        * inactivate all active regions. Here we add all active regions to inactive
+                        * list, they will be inactivated later in ufshpb_map_work_handler().
+                        */
+                       struct victim_select_info *lru_info = &hpb->lru_info;
+                       struct ufshpb_region *rgn;
+                       list_for_each_entry(rgn, &lru_info->lh_lru_rgn, list_lru_rgn)
+                               ufshpb_submit_region_inactive(hpb, rgn->rgn_idx);
+                       if (ufshpb_get_state(hpb) == HPB_PRESENT)
+                               queue_work(ufshpb_wq, &hpb->map_work);
+               }
+       }
+ }
+ /*
+  * This function will parse recommended active subregion information in sense
+  * data field of response UPIU with SAM_STAT_GOOD state.
+  */
+ void ufshpb_rsp_upiu(struct ufs_hba *hba, struct ufshcd_lrb *lrbp)
+ {
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(lrbp->cmd->device);
+       struct utp_hpb_rsp *rsp_field = &lrbp->ucd_rsp_ptr->hr;
+       int data_seg_len;
++      data_seg_len = be32_to_cpu(lrbp->ucd_rsp_ptr->header.dword_2)
++              & MASK_RSP_UPIU_DATA_SEG_LEN;
++
++      /* If data segment length is zero, rsp_field is not valid */
++      if (!data_seg_len)
++              return;
++
+       if (unlikely(lrbp->lun != rsp_field->lun)) {
+               struct scsi_device *sdev;
+               bool found = false;
+               __shost_for_each_device(sdev, hba->host) {
+                       hpb = ufshpb_get_hpb_data(sdev);
+                       if (!hpb)
+                               continue;
+                       if (rsp_field->lun == hpb->lun) {
+                               found = true;
+                               break;
+                       }
+               }
+               if (!found)
+                       return;
+       }
+       if (!hpb)
+               return;
+       if (ufshpb_get_state(hpb) == HPB_INIT)
+               return;
+       if ((ufshpb_get_state(hpb) != HPB_PRESENT) &&
+           (ufshpb_get_state(hpb) != HPB_SUSPEND)) {
+               dev_notice(&hpb->sdev_ufs_lu->sdev_dev,
+                          "%s: ufshpb state is not PRESENT/SUSPEND\n",
+                          __func__);
+               return;
+       }
+       BUILD_BUG_ON(sizeof(struct utp_hpb_rsp) != UTP_HPB_RSP_SIZE);
+       if (!ufshpb_is_hpb_rsp_valid(hba, lrbp, rsp_field))
+               return;
+       hpb->stats.rcmd_noti_cnt++;
+       switch (rsp_field->hpb_op) {
+       case HPB_RSP_REQ_REGION_UPDATE:
+               if (data_seg_len != DEV_DATA_SEG_LEN)
+                       dev_warn(&hpb->sdev_ufs_lu->sdev_dev,
+                                "%s: data seg length is not same.\n",
+                                __func__);
+               ufshpb_rsp_req_region_update(hpb, rsp_field);
+               break;
+       case HPB_RSP_DEV_RESET:
+               dev_warn(&hpb->sdev_ufs_lu->sdev_dev,
+                        "UFS device lost HPB information during PM.\n");
+               ufshpb_dev_reset_handler(hba);
+               break;
+       default:
+               dev_notice(&hpb->sdev_ufs_lu->sdev_dev,
+                          "hpb_op is not available: %d\n",
+                          rsp_field->hpb_op);
+               break;
+       }
+ }
+ static void ufshpb_add_active_list(struct ufshpb_lu *hpb,
+                                  struct ufshpb_region *rgn,
+                                  struct ufshpb_subregion *srgn)
+ {
+       if (!list_empty(&rgn->list_inact_rgn))
+               return;
+       if (!list_empty(&srgn->list_act_srgn)) {
+               list_move(&srgn->list_act_srgn, &hpb->lh_act_srgn);
+               return;
+       }
+       list_add(&srgn->list_act_srgn, &hpb->lh_act_srgn);
+ }
+ static void ufshpb_add_pending_evict_list(struct ufshpb_lu *hpb,
+                                         struct ufshpb_region *rgn,
+                                         struct list_head *pending_list)
+ {
+       struct ufshpb_subregion *srgn;
+       int srgn_idx;
+       if (!list_empty(&rgn->list_inact_rgn))
+               return;
+       for_each_sub_region(rgn, srgn_idx, srgn)
+               if (!list_empty(&srgn->list_act_srgn))
+                       return;
+       list_add_tail(&rgn->list_inact_rgn, pending_list);
+ }
+ static void ufshpb_run_active_subregion_list(struct ufshpb_lu *hpb)
+ {
+       struct ufshpb_region *rgn;
+       struct ufshpb_subregion *srgn;
+       unsigned long flags;
+       int ret = 0;
+       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       while ((srgn = list_first_entry_or_null(&hpb->lh_act_srgn,
+                                               struct ufshpb_subregion,
+                                               list_act_srgn))) {
+               if (ufshpb_get_state(hpb) == HPB_SUSPEND)
+                       break;
+               list_del_init(&srgn->list_act_srgn);
+               spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+               rgn = hpb->rgn_tbl + srgn->rgn_idx;
+               ret = ufshpb_add_region(hpb, rgn);
+               if (ret)
+                       goto active_failed;
+               ret = ufshpb_issue_map_req(hpb, rgn, srgn);
+               if (ret) {
+                       dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                           "issue map_req failed. ret %d, region %d - %d\n",
+                           ret, rgn->rgn_idx, srgn->srgn_idx);
+                       goto active_failed;
+               }
+               spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       }
+       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+       return;
+ active_failed:
+       dev_err(&hpb->sdev_ufs_lu->sdev_dev, "failed to activate region %d - %d, will retry\n",
+                  rgn->rgn_idx, srgn->srgn_idx);
+       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       ufshpb_add_active_list(hpb, rgn, srgn);
+       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+ }
+ static void ufshpb_run_inactive_region_list(struct ufshpb_lu *hpb)
+ {
+       struct ufshpb_region *rgn;
+       unsigned long flags;
+       int ret;
+       LIST_HEAD(pending_list);
+       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       while ((rgn = list_first_entry_or_null(&hpb->lh_inact_rgn,
+                                              struct ufshpb_region,
+                                              list_inact_rgn))) {
+               if (ufshpb_get_state(hpb) == HPB_SUSPEND)
+                       break;
+               list_del_init(&rgn->list_inact_rgn);
+               spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+               ret = ufshpb_evict_region(hpb, rgn);
+               if (ret) {
+                       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+                       ufshpb_add_pending_evict_list(hpb, rgn, &pending_list);
+                       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+               }
+               spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       }
+       list_splice(&pending_list, &hpb->lh_inact_rgn);
+       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+ }
+ static void ufshpb_normalization_work_handler(struct work_struct *work)
+ {
+       struct ufshpb_lu *hpb = container_of(work, struct ufshpb_lu,
+                                            ufshpb_normalization_work);
+       int rgn_idx;
+       u8 factor = hpb->params.normalization_factor;
+       for (rgn_idx = 0; rgn_idx < hpb->rgns_per_lu; rgn_idx++) {
+               struct ufshpb_region *rgn = hpb->rgn_tbl + rgn_idx;
+               int srgn_idx;
+               spin_lock(&rgn->rgn_lock);
+               rgn->reads = 0;
+               for (srgn_idx = 0; srgn_idx < hpb->srgns_per_rgn; srgn_idx++) {
+                       struct ufshpb_subregion *srgn = rgn->srgn_tbl + srgn_idx;
+                       srgn->reads >>= factor;
+                       rgn->reads += srgn->reads;
+               }
+               spin_unlock(&rgn->rgn_lock);
+               if (rgn->rgn_state != HPB_RGN_ACTIVE || rgn->reads)
+                       continue;
+               /* if region is active but has no reads - inactivate it */
+               spin_lock(&hpb->rsp_list_lock);
+               ufshpb_update_inactive_info(hpb, rgn->rgn_idx);
+               spin_unlock(&hpb->rsp_list_lock);
+       }
+ }
+ static void ufshpb_map_work_handler(struct work_struct *work)
+ {
+       struct ufshpb_lu *hpb = container_of(work, struct ufshpb_lu, map_work);
+       if (ufshpb_get_state(hpb) != HPB_PRESENT) {
+               dev_notice(&hpb->sdev_ufs_lu->sdev_dev,
+                          "%s: ufshpb state is not PRESENT\n", __func__);
+               return;
+       }
+       ufshpb_run_inactive_region_list(hpb);
+       ufshpb_run_active_subregion_list(hpb);
+ }
+ /*
+  * this function doesn't need to hold lock due to be called in init.
+  * (rgn_state_lock, rsp_list_lock, etc..)
+  */
+ static int ufshpb_init_pinned_active_region(struct ufs_hba *hba,
+                                           struct ufshpb_lu *hpb,
+                                           struct ufshpb_region *rgn)
+ {
+       struct ufshpb_subregion *srgn;
+       int srgn_idx, i;
+       int err = 0;
+       for_each_sub_region(rgn, srgn_idx, srgn) {
+               srgn->mctx = ufshpb_get_map_ctx(hpb, srgn->is_last);
+               srgn->srgn_state = HPB_SRGN_INVALID;
+               if (!srgn->mctx) {
+                       err = -ENOMEM;
+                       dev_err(hba->dev,
+                               "alloc mctx for pinned region failed\n");
+                       goto release;
+               }
+               list_add_tail(&srgn->list_act_srgn, &hpb->lh_act_srgn);
+       }
+       rgn->rgn_state = HPB_RGN_PINNED;
+       return 0;
+ release:
+       for (i = 0; i < srgn_idx; i++) {
+               srgn = rgn->srgn_tbl + i;
+               ufshpb_put_map_ctx(hpb, srgn->mctx);
+       }
+       return err;
+ }
+ static void ufshpb_init_subregion_tbl(struct ufshpb_lu *hpb,
+                                     struct ufshpb_region *rgn, bool last)
+ {
+       int srgn_idx;
+       struct ufshpb_subregion *srgn;
+       for_each_sub_region(rgn, srgn_idx, srgn) {
+               INIT_LIST_HEAD(&srgn->list_act_srgn);
+               srgn->rgn_idx = rgn->rgn_idx;
+               srgn->srgn_idx = srgn_idx;
+               srgn->srgn_state = HPB_SRGN_UNUSED;
+       }
+       if (unlikely(last && hpb->last_srgn_entries))
+               srgn->is_last = true;
+ }
+ static int ufshpb_alloc_subregion_tbl(struct ufshpb_lu *hpb,
+                                     struct ufshpb_region *rgn, int srgn_cnt)
+ {
+       rgn->srgn_tbl = kvcalloc(srgn_cnt, sizeof(struct ufshpb_subregion),
+                                GFP_KERNEL);
+       if (!rgn->srgn_tbl)
+               return -ENOMEM;
+       rgn->srgn_cnt = srgn_cnt;
+       return 0;
+ }
+ static void ufshpb_lu_parameter_init(struct ufs_hba *hba,
+                                    struct ufshpb_lu *hpb,
+                                    struct ufshpb_dev_info *hpb_dev_info,
+                                    struct ufshpb_lu_info *hpb_lu_info)
+ {
+       u32 entries_per_rgn;
+       u64 rgn_mem_size, tmp;
+       if (ufshpb_is_legacy(hba))
+               hpb->pre_req_max_tr_len = HPB_LEGACY_CHUNK_HIGH;
+       else
+               hpb->pre_req_max_tr_len = hpb_dev_info->max_hpb_single_cmd;
+       hpb->lu_pinned_start = hpb_lu_info->pinned_start;
+       hpb->lu_pinned_end = hpb_lu_info->num_pinned ?
+               (hpb_lu_info->pinned_start + hpb_lu_info->num_pinned - 1)
+               : PINNED_NOT_SET;
+       hpb->lru_info.max_lru_active_cnt =
+               hpb_lu_info->max_active_rgns - hpb_lu_info->num_pinned;
+       rgn_mem_size = (1ULL << hpb_dev_info->rgn_size) * HPB_RGN_SIZE_UNIT
+                       * HPB_ENTRY_SIZE;
+       do_div(rgn_mem_size, HPB_ENTRY_BLOCK_SIZE);
+       hpb->srgn_mem_size = (1ULL << hpb_dev_info->srgn_size)
+               * HPB_RGN_SIZE_UNIT / HPB_ENTRY_BLOCK_SIZE * HPB_ENTRY_SIZE;
+       tmp = rgn_mem_size;
+       do_div(tmp, HPB_ENTRY_SIZE);
+       entries_per_rgn = (u32)tmp;
+       hpb->entries_per_rgn_shift = ilog2(entries_per_rgn);
+       hpb->entries_per_rgn_mask = entries_per_rgn - 1;
+       hpb->entries_per_srgn = hpb->srgn_mem_size / HPB_ENTRY_SIZE;
+       hpb->entries_per_srgn_shift = ilog2(hpb->entries_per_srgn);
+       hpb->entries_per_srgn_mask = hpb->entries_per_srgn - 1;
+       tmp = rgn_mem_size;
+       do_div(tmp, hpb->srgn_mem_size);
+       hpb->srgns_per_rgn = (int)tmp;
+       hpb->rgns_per_lu = DIV_ROUND_UP(hpb_lu_info->num_blocks,
+                               entries_per_rgn);
+       hpb->srgns_per_lu = DIV_ROUND_UP(hpb_lu_info->num_blocks,
+                               (hpb->srgn_mem_size / HPB_ENTRY_SIZE));
+       hpb->last_srgn_entries = hpb_lu_info->num_blocks
+                                % (hpb->srgn_mem_size / HPB_ENTRY_SIZE);
+       hpb->pages_per_srgn = DIV_ROUND_UP(hpb->srgn_mem_size, PAGE_SIZE);
+       if (hpb_dev_info->control_mode == HPB_HOST_CONTROL)
+               hpb->is_hcm = true;
+ }
+ static int ufshpb_alloc_region_tbl(struct ufs_hba *hba, struct ufshpb_lu *hpb)
+ {
+       struct ufshpb_region *rgn_table, *rgn;
+       int rgn_idx, i;
+       int ret = 0;
+       rgn_table = kvcalloc(hpb->rgns_per_lu, sizeof(struct ufshpb_region),
+                           GFP_KERNEL);
+       if (!rgn_table)
+               return -ENOMEM;
+       for (rgn_idx = 0; rgn_idx < hpb->rgns_per_lu; rgn_idx++) {
+               int srgn_cnt = hpb->srgns_per_rgn;
+               bool last_srgn = false;
+               rgn = rgn_table + rgn_idx;
+               rgn->rgn_idx = rgn_idx;
+               spin_lock_init(&rgn->rgn_lock);
+               INIT_LIST_HEAD(&rgn->list_inact_rgn);
+               INIT_LIST_HEAD(&rgn->list_lru_rgn);
+               INIT_LIST_HEAD(&rgn->list_expired_rgn);
+               if (rgn_idx == hpb->rgns_per_lu - 1) {
+                       srgn_cnt = ((hpb->srgns_per_lu - 1) %
+                                   hpb->srgns_per_rgn) + 1;
+                       last_srgn = true;
+               }
+               ret = ufshpb_alloc_subregion_tbl(hpb, rgn, srgn_cnt);
+               if (ret)
+                       goto release_srgn_table;
+               ufshpb_init_subregion_tbl(hpb, rgn, last_srgn);
+               if (ufshpb_is_pinned_region(hpb, rgn_idx)) {
+                       ret = ufshpb_init_pinned_active_region(hba, hpb, rgn);
+                       if (ret)
+                               goto release_srgn_table;
+               } else {
+                       rgn->rgn_state = HPB_RGN_INACTIVE;
+               }
+               rgn->rgn_flags = 0;
+               rgn->hpb = hpb;
+       }
+       hpb->rgn_tbl = rgn_table;
+       return 0;
+ release_srgn_table:
+       for (i = 0; i <= rgn_idx; i++)
+               kvfree(rgn_table[i].srgn_tbl);
+       kvfree(rgn_table);
+       return ret;
+ }
+ static void ufshpb_destroy_subregion_tbl(struct ufshpb_lu *hpb,
+                                        struct ufshpb_region *rgn)
+ {
+       int srgn_idx;
+       struct ufshpb_subregion *srgn;
+       for_each_sub_region(rgn, srgn_idx, srgn)
+               if (srgn->srgn_state != HPB_SRGN_UNUSED) {
+                       srgn->srgn_state = HPB_SRGN_UNUSED;
+                       ufshpb_put_map_ctx(hpb, srgn->mctx);
+               }
+ }
+ static void ufshpb_destroy_region_tbl(struct ufshpb_lu *hpb)
+ {
+       int rgn_idx;
+       for (rgn_idx = 0; rgn_idx < hpb->rgns_per_lu; rgn_idx++) {
+               struct ufshpb_region *rgn;
+               rgn = hpb->rgn_tbl + rgn_idx;
+               if (rgn->rgn_state != HPB_RGN_INACTIVE) {
+                       rgn->rgn_state = HPB_RGN_INACTIVE;
+                       ufshpb_destroy_subregion_tbl(hpb, rgn);
+               }
+               kvfree(rgn->srgn_tbl);
+       }
+       kvfree(hpb->rgn_tbl);
+ }
+ /* SYSFS functions */
+ #define ufshpb_sysfs_attr_show_func(__name)                           \
+ static ssize_t __name##_show(struct device *dev,                      \
+       struct device_attribute *attr, char *buf)                       \
+ {                                                                     \
+       struct scsi_device *sdev = to_scsi_device(dev);                 \
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);              \
+                                                                       \
+       if (!hpb)                                                       \
+               return -ENODEV;                                         \
+                                                                       \
+       return sysfs_emit(buf, "%llu\n", hpb->stats.__name);            \
+ }                                                                     \
+ \
+ static DEVICE_ATTR_RO(__name)
+ ufshpb_sysfs_attr_show_func(hit_cnt);
+ ufshpb_sysfs_attr_show_func(miss_cnt);
+ ufshpb_sysfs_attr_show_func(rcmd_noti_cnt);
+ ufshpb_sysfs_attr_show_func(rcmd_active_cnt);
+ ufshpb_sysfs_attr_show_func(rcmd_inactive_cnt);
+ ufshpb_sysfs_attr_show_func(map_req_cnt);
+ ufshpb_sysfs_attr_show_func(umap_req_cnt);
+ static struct attribute *hpb_dev_stat_attrs[] = {
+       &dev_attr_hit_cnt.attr,
+       &dev_attr_miss_cnt.attr,
+       &dev_attr_rcmd_noti_cnt.attr,
+       &dev_attr_rcmd_active_cnt.attr,
+       &dev_attr_rcmd_inactive_cnt.attr,
+       &dev_attr_map_req_cnt.attr,
+       &dev_attr_umap_req_cnt.attr,
+       NULL,
+ };
+ struct attribute_group ufs_sysfs_hpb_stat_group = {
+       .name = "hpb_stats",
+       .attrs = hpb_dev_stat_attrs,
+ };
+ /* SYSFS functions */
+ #define ufshpb_sysfs_param_show_func(__name)                          \
+ static ssize_t __name##_show(struct device *dev,                      \
+       struct device_attribute *attr, char *buf)                       \
+ {                                                                     \
+       struct scsi_device *sdev = to_scsi_device(dev);                 \
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);              \
+                                                                       \
+       if (!hpb)                                                       \
+               return -ENODEV;                                         \
+                                                                       \
+       return sysfs_emit(buf, "%d\n", hpb->params.__name);             \
+ }
+ ufshpb_sysfs_param_show_func(requeue_timeout_ms);
+ static ssize_t
+ requeue_timeout_ms_store(struct device *dev, struct device_attribute *attr,
+                        const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val < 0)
+               return -EINVAL;
+       hpb->params.requeue_timeout_ms = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(requeue_timeout_ms);
+ ufshpb_sysfs_param_show_func(activation_thld);
+ static ssize_t
+ activation_thld_store(struct device *dev, struct device_attribute *attr,
+                     const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val <= 0)
+               return -EINVAL;
+       hpb->params.activation_thld = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(activation_thld);
+ ufshpb_sysfs_param_show_func(normalization_factor);
+ static ssize_t
+ normalization_factor_store(struct device *dev, struct device_attribute *attr,
+                          const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val <= 0 || val > ilog2(hpb->entries_per_srgn))
+               return -EINVAL;
+       hpb->params.normalization_factor = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(normalization_factor);
+ ufshpb_sysfs_param_show_func(eviction_thld_enter);
+ static ssize_t
+ eviction_thld_enter_store(struct device *dev, struct device_attribute *attr,
+                         const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val <= hpb->params.eviction_thld_exit)
+               return -EINVAL;
+       hpb->params.eviction_thld_enter = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(eviction_thld_enter);
+ ufshpb_sysfs_param_show_func(eviction_thld_exit);
+ static ssize_t
+ eviction_thld_exit_store(struct device *dev, struct device_attribute *attr,
+                        const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val <= hpb->params.activation_thld)
+               return -EINVAL;
+       hpb->params.eviction_thld_exit = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(eviction_thld_exit);
+ ufshpb_sysfs_param_show_func(read_timeout_ms);
+ static ssize_t
+ read_timeout_ms_store(struct device *dev, struct device_attribute *attr,
+                     const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       /* read_timeout >> timeout_polling_interval */
+       if (val < hpb->params.timeout_polling_interval_ms * 2)
+               return -EINVAL;
+       hpb->params.read_timeout_ms = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(read_timeout_ms);
+ ufshpb_sysfs_param_show_func(read_timeout_expiries);
+ static ssize_t
+ read_timeout_expiries_store(struct device *dev, struct device_attribute *attr,
+                           const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val <= 0)
+               return -EINVAL;
+       hpb->params.read_timeout_expiries = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(read_timeout_expiries);
+ ufshpb_sysfs_param_show_func(timeout_polling_interval_ms);
+ static ssize_t
+ timeout_polling_interval_ms_store(struct device *dev,
+                                 struct device_attribute *attr,
+                                 const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       /* timeout_polling_interval << read_timeout */
+       if (val <= 0 || val > hpb->params.read_timeout_ms / 2)
+               return -EINVAL;
+       hpb->params.timeout_polling_interval_ms = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(timeout_polling_interval_ms);
+ ufshpb_sysfs_param_show_func(inflight_map_req);
+ static ssize_t inflight_map_req_store(struct device *dev,
+                                     struct device_attribute *attr,
+                                     const char *buf, size_t count)
+ {
+       struct scsi_device *sdev = to_scsi_device(dev);
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       int val;
+       if (!hpb)
+               return -ENODEV;
+       if (!hpb->is_hcm)
+               return -EOPNOTSUPP;
+       if (kstrtouint(buf, 0, &val))
+               return -EINVAL;
+       if (val <= 0 || val > hpb->sdev_ufs_lu->queue_depth - 1)
+               return -EINVAL;
+       hpb->params.inflight_map_req = val;
+       return count;
+ }
+ static DEVICE_ATTR_RW(inflight_map_req);
+ static void ufshpb_hcm_param_init(struct ufshpb_lu *hpb)
+ {
+       hpb->params.activation_thld = ACTIVATION_THRESHOLD;
+       hpb->params.normalization_factor = 1;
+       hpb->params.eviction_thld_enter = (ACTIVATION_THRESHOLD << 5);
+       hpb->params.eviction_thld_exit = (ACTIVATION_THRESHOLD << 4);
+       hpb->params.read_timeout_ms = READ_TO_MS;
+       hpb->params.read_timeout_expiries = READ_TO_EXPIRIES;
+       hpb->params.timeout_polling_interval_ms = POLLING_INTERVAL_MS;
+       hpb->params.inflight_map_req = THROTTLE_MAP_REQ_DEFAULT;
+ }
+ static struct attribute *hpb_dev_param_attrs[] = {
+       &dev_attr_requeue_timeout_ms.attr,
+       &dev_attr_activation_thld.attr,
+       &dev_attr_normalization_factor.attr,
+       &dev_attr_eviction_thld_enter.attr,
+       &dev_attr_eviction_thld_exit.attr,
+       &dev_attr_read_timeout_ms.attr,
+       &dev_attr_read_timeout_expiries.attr,
+       &dev_attr_timeout_polling_interval_ms.attr,
+       &dev_attr_inflight_map_req.attr,
+       NULL,
+ };
+ struct attribute_group ufs_sysfs_hpb_param_group = {
+       .name = "hpb_params",
+       .attrs = hpb_dev_param_attrs,
+ };
+ static int ufshpb_pre_req_mempool_init(struct ufshpb_lu *hpb)
+ {
+       struct ufshpb_req *pre_req = NULL, *t;
+       int qd = hpb->sdev_ufs_lu->queue_depth / 2;
+       int i;
+       INIT_LIST_HEAD(&hpb->lh_pre_req_free);
+       hpb->pre_req = kcalloc(qd, sizeof(struct ufshpb_req), GFP_KERNEL);
+       hpb->throttle_pre_req = qd;
+       hpb->num_inflight_pre_req = 0;
+       if (!hpb->pre_req)
+               goto release_mem;
+       for (i = 0; i < qd; i++) {
+               pre_req = hpb->pre_req + i;
+               INIT_LIST_HEAD(&pre_req->list_req);
+               pre_req->req = NULL;
+               pre_req->bio = bio_alloc(NULL, 1, 0, GFP_KERNEL);
+               if (!pre_req->bio)
+                       goto release_mem;
+               pre_req->wb.m_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
+               if (!pre_req->wb.m_page) {
+                       bio_put(pre_req->bio);
+                       goto release_mem;
+               }
+               list_add_tail(&pre_req->list_req, &hpb->lh_pre_req_free);
+       }
+       return 0;
+ release_mem:
+       list_for_each_entry_safe(pre_req, t, &hpb->lh_pre_req_free, list_req) {
+               list_del_init(&pre_req->list_req);
+               bio_put(pre_req->bio);
+               __free_page(pre_req->wb.m_page);
+       }
+       kfree(hpb->pre_req);
+       return -ENOMEM;
+ }
+ static void ufshpb_pre_req_mempool_destroy(struct ufshpb_lu *hpb)
+ {
+       struct ufshpb_req *pre_req = NULL;
+       int i;
+       for (i = 0; i < hpb->throttle_pre_req; i++) {
+               pre_req = hpb->pre_req + i;
+               bio_put(hpb->pre_req[i].bio);
+               if (!pre_req->wb.m_page)
+                       __free_page(hpb->pre_req[i].wb.m_page);
+               list_del_init(&pre_req->list_req);
+       }
+       kfree(hpb->pre_req);
+ }
+ static void ufshpb_stat_init(struct ufshpb_lu *hpb)
+ {
+       hpb->stats.hit_cnt = 0;
+       hpb->stats.miss_cnt = 0;
+       hpb->stats.rcmd_noti_cnt = 0;
+       hpb->stats.rcmd_active_cnt = 0;
+       hpb->stats.rcmd_inactive_cnt = 0;
+       hpb->stats.map_req_cnt = 0;
+       hpb->stats.umap_req_cnt = 0;
+ }
+ static void ufshpb_param_init(struct ufshpb_lu *hpb)
+ {
+       hpb->params.requeue_timeout_ms = HPB_REQUEUE_TIME_MS;
+       if (hpb->is_hcm)
+               ufshpb_hcm_param_init(hpb);
+ }
+ static int ufshpb_lu_hpb_init(struct ufs_hba *hba, struct ufshpb_lu *hpb)
+ {
+       int ret;
+       spin_lock_init(&hpb->rgn_state_lock);
+       spin_lock_init(&hpb->rsp_list_lock);
+       spin_lock_init(&hpb->param_lock);
+       INIT_LIST_HEAD(&hpb->lru_info.lh_lru_rgn);
+       INIT_LIST_HEAD(&hpb->lh_act_srgn);
+       INIT_LIST_HEAD(&hpb->lh_inact_rgn);
+       INIT_LIST_HEAD(&hpb->list_hpb_lu);
+       INIT_WORK(&hpb->map_work, ufshpb_map_work_handler);
+       if (hpb->is_hcm) {
+               INIT_WORK(&hpb->ufshpb_normalization_work,
+                         ufshpb_normalization_work_handler);
+               INIT_DELAYED_WORK(&hpb->ufshpb_read_to_work,
+                                 ufshpb_read_to_handler);
+       }
+       hpb->map_req_cache = kmem_cache_create("ufshpb_req_cache",
+                         sizeof(struct ufshpb_req), 0, 0, NULL);
+       if (!hpb->map_req_cache) {
+               dev_err(hba->dev, "ufshpb(%d) ufshpb_req_cache create fail",
+                       hpb->lun);
+               return -ENOMEM;
+       }
+       hpb->m_page_cache = kmem_cache_create("ufshpb_m_page_cache",
+                         sizeof(struct page *) * hpb->pages_per_srgn,
+                         0, 0, NULL);
+       if (!hpb->m_page_cache) {
+               dev_err(hba->dev, "ufshpb(%d) ufshpb_m_page_cache create fail",
+                       hpb->lun);
+               ret = -ENOMEM;
+               goto release_req_cache;
+       }
+       ret = ufshpb_pre_req_mempool_init(hpb);
+       if (ret) {
+               dev_err(hba->dev, "ufshpb(%d) pre_req_mempool init fail",
+                       hpb->lun);
+               goto release_m_page_cache;
+       }
+       ret = ufshpb_alloc_region_tbl(hba, hpb);
+       if (ret)
+               goto release_pre_req_mempool;
+       ufshpb_stat_init(hpb);
+       ufshpb_param_init(hpb);
+       if (hpb->is_hcm) {
+               unsigned int poll;
+               poll = hpb->params.timeout_polling_interval_ms;
+               schedule_delayed_work(&hpb->ufshpb_read_to_work,
+                                     msecs_to_jiffies(poll));
+       }
+       return 0;
+ release_pre_req_mempool:
+       ufshpb_pre_req_mempool_destroy(hpb);
+ release_m_page_cache:
+       kmem_cache_destroy(hpb->m_page_cache);
+ release_req_cache:
+       kmem_cache_destroy(hpb->map_req_cache);
+       return ret;
+ }
+ static struct ufshpb_lu *
+ ufshpb_alloc_hpb_lu(struct ufs_hba *hba, struct scsi_device *sdev,
+                   struct ufshpb_dev_info *hpb_dev_info,
+                   struct ufshpb_lu_info *hpb_lu_info)
+ {
+       struct ufshpb_lu *hpb;
+       int ret;
+       hpb = kzalloc(sizeof(struct ufshpb_lu), GFP_KERNEL);
+       if (!hpb)
+               return NULL;
+       hpb->lun = sdev->lun;
+       hpb->sdev_ufs_lu = sdev;
+       ufshpb_lu_parameter_init(hba, hpb, hpb_dev_info, hpb_lu_info);
+       ret = ufshpb_lu_hpb_init(hba, hpb);
+       if (ret) {
+               dev_err(hba->dev, "hpb lu init failed. ret %d", ret);
+               goto release_hpb;
+       }
+       sdev->hostdata = hpb;
+       return hpb;
+ release_hpb:
+       kfree(hpb);
+       return NULL;
+ }
+ static void ufshpb_discard_rsp_lists(struct ufshpb_lu *hpb)
+ {
+       struct ufshpb_region *rgn, *next_rgn;
+       struct ufshpb_subregion *srgn, *next_srgn;
+       unsigned long flags;
+       /*
+        * If the device reset occurred, the remaining HPB region information
+        * may be stale. Therefore, by discarding the lists of HPB response
+        * that remained after reset, we prevent unnecessary work.
+        */
+       spin_lock_irqsave(&hpb->rsp_list_lock, flags);
+       list_for_each_entry_safe(rgn, next_rgn, &hpb->lh_inact_rgn,
+                                list_inact_rgn)
+               list_del_init(&rgn->list_inact_rgn);
+       list_for_each_entry_safe(srgn, next_srgn, &hpb->lh_act_srgn,
+                                list_act_srgn)
+               list_del_init(&srgn->list_act_srgn);
+       spin_unlock_irqrestore(&hpb->rsp_list_lock, flags);
+ }
+ static void ufshpb_cancel_jobs(struct ufshpb_lu *hpb)
+ {
+       if (hpb->is_hcm) {
+               cancel_delayed_work_sync(&hpb->ufshpb_read_to_work);
+               cancel_work_sync(&hpb->ufshpb_normalization_work);
+       }
+       cancel_work_sync(&hpb->map_work);
+ }
+ static bool ufshpb_check_hpb_reset_query(struct ufs_hba *hba)
+ {
+       int err = 0;
+       bool flag_res = true;
+       int try;
+       /* wait for the device to complete HPB reset query */
+       for (try = 0; try < HPB_RESET_REQ_RETRIES; try++) {
+               dev_dbg(hba->dev,
+                       "%s start flag reset polling %d times\n",
+                       __func__, try);
+               /* Poll fHpbReset flag to be cleared */
+               err = ufshcd_query_flag(hba, UPIU_QUERY_OPCODE_READ_FLAG,
+                               QUERY_FLAG_IDN_HPB_RESET, 0, &flag_res);
+               if (err) {
+                       dev_err(hba->dev,
+                               "%s reading fHpbReset flag failed with error %d\n",
+                               __func__, err);
+                       return flag_res;
+               }
+               if (!flag_res)
+                       goto out;
+               usleep_range(1000, 1100);
+       }
+       if (flag_res) {
+               dev_err(hba->dev,
+                       "%s fHpbReset was not cleared by the device\n",
+                       __func__);
+       }
+ out:
+       return flag_res;
+ }
+ /**
+  * ufshpb_toggle_state - switch HPB state of all LUs
+  * @hba: per-adapter instance
+  * @src: expected current HPB state
+  * @dest: target HPB state to switch to
+  */
+ void ufshpb_toggle_state(struct ufs_hba *hba, enum UFSHPB_STATE src, enum UFSHPB_STATE dest)
+ {
+       struct ufshpb_lu *hpb;
+       struct scsi_device *sdev;
+       shost_for_each_device(sdev, hba->host) {
+               hpb = ufshpb_get_hpb_data(sdev);
+               if (!hpb || ufshpb_get_state(hpb) != src)
+                       continue;
+               ufshpb_set_state(hpb, dest);
+               if (dest == HPB_RESET) {
+                       ufshpb_cancel_jobs(hpb);
+                       ufshpb_discard_rsp_lists(hpb);
+               }
+       }
+ }
+ void ufshpb_suspend(struct ufs_hba *hba)
+ {
+       struct ufshpb_lu *hpb;
+       struct scsi_device *sdev;
+       shost_for_each_device(sdev, hba->host) {
+               hpb = ufshpb_get_hpb_data(sdev);
+               if (!hpb || ufshpb_get_state(hpb) != HPB_PRESENT)
+                       continue;
+               ufshpb_set_state(hpb, HPB_SUSPEND);
+               ufshpb_cancel_jobs(hpb);
+       }
+ }
+ void ufshpb_resume(struct ufs_hba *hba)
+ {
+       struct ufshpb_lu *hpb;
+       struct scsi_device *sdev;
+       shost_for_each_device(sdev, hba->host) {
+               hpb = ufshpb_get_hpb_data(sdev);
+               if (!hpb || ufshpb_get_state(hpb) != HPB_SUSPEND)
+                       continue;
+               ufshpb_set_state(hpb, HPB_PRESENT);
+               ufshpb_kick_map_work(hpb);
+               if (hpb->is_hcm) {
+                       unsigned int poll = hpb->params.timeout_polling_interval_ms;
+                       schedule_delayed_work(&hpb->ufshpb_read_to_work, msecs_to_jiffies(poll));
+               }
+       }
+ }
+ static int ufshpb_get_lu_info(struct ufs_hba *hba, int lun,
+                             struct ufshpb_lu_info *hpb_lu_info)
+ {
+       u16 max_active_rgns;
+       u8 lu_enable;
+       int size;
+       int ret;
+       char desc_buf[QUERY_DESC_MAX_SIZE];
+       ufshcd_map_desc_id_to_length(hba, QUERY_DESC_IDN_UNIT, &size);
+       ufshcd_rpm_get_sync(hba);
+       ret = ufshcd_query_descriptor_retry(hba, UPIU_QUERY_OPCODE_READ_DESC,
+                                           QUERY_DESC_IDN_UNIT, lun, 0,
+                                           desc_buf, &size);
+       ufshcd_rpm_put_sync(hba);
+       if (ret) {
+               dev_err(hba->dev,
+                       "%s: idn: %d lun: %d  query request failed",
+                       __func__, QUERY_DESC_IDN_UNIT, lun);
+               return ret;
+       }
+       lu_enable = desc_buf[UNIT_DESC_PARAM_LU_ENABLE];
+       if (lu_enable != LU_ENABLED_HPB_FUNC)
+               return -ENODEV;
+       max_active_rgns = get_unaligned_be16(
+                       desc_buf + UNIT_DESC_PARAM_HPB_LU_MAX_ACTIVE_RGNS);
+       if (!max_active_rgns) {
+               dev_err(hba->dev,
+                       "lun %d wrong number of max active regions\n", lun);
+               return -ENODEV;
+       }
+       hpb_lu_info->num_blocks = get_unaligned_be64(
+                       desc_buf + UNIT_DESC_PARAM_LOGICAL_BLK_COUNT);
+       hpb_lu_info->pinned_start = get_unaligned_be16(
+                       desc_buf + UNIT_DESC_PARAM_HPB_PIN_RGN_START_OFF);
+       hpb_lu_info->num_pinned = get_unaligned_be16(
+                       desc_buf + UNIT_DESC_PARAM_HPB_NUM_PIN_RGNS);
+       hpb_lu_info->max_active_rgns = max_active_rgns;
+       return 0;
+ }
+ void ufshpb_destroy_lu(struct ufs_hba *hba, struct scsi_device *sdev)
+ {
+       struct ufshpb_lu *hpb = ufshpb_get_hpb_data(sdev);
+       if (!hpb)
+               return;
+       ufshpb_set_state(hpb, HPB_FAILED);
+       sdev = hpb->sdev_ufs_lu;
+       sdev->hostdata = NULL;
+       ufshpb_cancel_jobs(hpb);
+       ufshpb_pre_req_mempool_destroy(hpb);
+       ufshpb_destroy_region_tbl(hpb);
+       kmem_cache_destroy(hpb->map_req_cache);
+       kmem_cache_destroy(hpb->m_page_cache);
+       list_del_init(&hpb->list_hpb_lu);
+       kfree(hpb);
+ }
+ static void ufshpb_hpb_lu_prepared(struct ufs_hba *hba)
+ {
+       int pool_size;
+       struct ufshpb_lu *hpb;
+       struct scsi_device *sdev;
+       bool init_success;
+       if (tot_active_srgn_pages == 0) {
+               ufshpb_remove(hba);
+               return;
+       }
+       init_success = !ufshpb_check_hpb_reset_query(hba);
+       pool_size = PAGE_ALIGN(ufshpb_host_map_kbytes * 1024) / PAGE_SIZE;
+       if (pool_size > tot_active_srgn_pages) {
+               mempool_resize(ufshpb_mctx_pool, tot_active_srgn_pages);
+               mempool_resize(ufshpb_page_pool, tot_active_srgn_pages);
+       }
+       shost_for_each_device(sdev, hba->host) {
+               hpb = ufshpb_get_hpb_data(sdev);
+               if (!hpb)
+                       continue;
+               if (init_success) {
+                       ufshpb_set_state(hpb, HPB_PRESENT);
+                       if ((hpb->lu_pinned_end - hpb->lu_pinned_start) > 0)
+                               queue_work(ufshpb_wq, &hpb->map_work);
+               } else {
+                       dev_err(hba->dev, "destroy HPB lu %d\n", hpb->lun);
+                       ufshpb_destroy_lu(hba, sdev);
+               }
+       }
+       if (!init_success)
+               ufshpb_remove(hba);
+ }
+ void ufshpb_init_hpb_lu(struct ufs_hba *hba, struct scsi_device *sdev)
+ {
+       struct ufshpb_lu *hpb;
+       int ret;
+       struct ufshpb_lu_info hpb_lu_info = { 0 };
+       int lun = sdev->lun;
+       if (lun >= hba->dev_info.max_lu_supported)
+               goto out;
+       ret = ufshpb_get_lu_info(hba, lun, &hpb_lu_info);
+       if (ret)
+               goto out;
+       hpb = ufshpb_alloc_hpb_lu(hba, sdev, &hba->ufshpb_dev,
+                                 &hpb_lu_info);
+       if (!hpb)
+               goto out;
+       tot_active_srgn_pages += hpb_lu_info.max_active_rgns *
+                       hpb->srgns_per_rgn * hpb->pages_per_srgn;
+ out:
+       /* All LUs are initialized */
+       if (atomic_dec_and_test(&hba->ufshpb_dev.slave_conf_cnt))
+               ufshpb_hpb_lu_prepared(hba);
+ }
+ static int ufshpb_init_mem_wq(struct ufs_hba *hba)
+ {
+       int ret;
+       unsigned int pool_size;
+       ufshpb_mctx_cache = kmem_cache_create("ufshpb_mctx_cache",
+                                       sizeof(struct ufshpb_map_ctx),
+                                       0, 0, NULL);
+       if (!ufshpb_mctx_cache) {
+               dev_err(hba->dev, "ufshpb: cannot init mctx cache\n");
+               return -ENOMEM;
+       }
+       pool_size = PAGE_ALIGN(ufshpb_host_map_kbytes * 1024) / PAGE_SIZE;
+       dev_info(hba->dev, "%s:%d ufshpb_host_map_kbytes %u pool_size %u\n",
+              __func__, __LINE__, ufshpb_host_map_kbytes, pool_size);
+       ufshpb_mctx_pool = mempool_create_slab_pool(pool_size,
+                                                   ufshpb_mctx_cache);
+       if (!ufshpb_mctx_pool) {
+               dev_err(hba->dev, "ufshpb: cannot init mctx pool\n");
+               ret = -ENOMEM;
+               goto release_mctx_cache;
+       }
+       ufshpb_page_pool = mempool_create_page_pool(pool_size, 0);
+       if (!ufshpb_page_pool) {
+               dev_err(hba->dev, "ufshpb: cannot init page pool\n");
+               ret = -ENOMEM;
+               goto release_mctx_pool;
+       }
+       ufshpb_wq = alloc_workqueue("ufshpb-wq",
+                                       WQ_UNBOUND | WQ_MEM_RECLAIM, 0);
+       if (!ufshpb_wq) {
+               dev_err(hba->dev, "ufshpb: alloc workqueue failed\n");
+               ret = -ENOMEM;
+               goto release_page_pool;
+       }
+       return 0;
+ release_page_pool:
+       mempool_destroy(ufshpb_page_pool);
+ release_mctx_pool:
+       mempool_destroy(ufshpb_mctx_pool);
+ release_mctx_cache:
+       kmem_cache_destroy(ufshpb_mctx_cache);
+       return ret;
+ }
+ void ufshpb_get_geo_info(struct ufs_hba *hba, u8 *geo_buf)
+ {
+       struct ufshpb_dev_info *hpb_info = &hba->ufshpb_dev;
+       int max_active_rgns = 0;
+       int hpb_num_lu;
+       hpb_num_lu = geo_buf[GEOMETRY_DESC_PARAM_HPB_NUMBER_LU];
+       if (hpb_num_lu == 0) {
+               dev_err(hba->dev, "No HPB LU supported\n");
+               hpb_info->hpb_disabled = true;
+               return;
+       }
+       hpb_info->rgn_size = geo_buf[GEOMETRY_DESC_PARAM_HPB_REGION_SIZE];
+       hpb_info->srgn_size = geo_buf[GEOMETRY_DESC_PARAM_HPB_SUBREGION_SIZE];
+       max_active_rgns = get_unaligned_be16(geo_buf +
+                         GEOMETRY_DESC_PARAM_HPB_MAX_ACTIVE_REGS);
+       if (hpb_info->rgn_size == 0 || hpb_info->srgn_size == 0 ||
+           max_active_rgns == 0) {
+               dev_err(hba->dev, "No HPB supported device\n");
+               hpb_info->hpb_disabled = true;
+               return;
+       }
+ }
+ void ufshpb_get_dev_info(struct ufs_hba *hba, u8 *desc_buf)
+ {
+       struct ufshpb_dev_info *hpb_dev_info = &hba->ufshpb_dev;
+       int version, ret;
+       int max_single_cmd;
+       hpb_dev_info->control_mode = desc_buf[DEVICE_DESC_PARAM_HPB_CONTROL];
+       version = get_unaligned_be16(desc_buf + DEVICE_DESC_PARAM_HPB_VER);
+       if ((version != HPB_SUPPORT_VERSION) &&
+           (version != HPB_SUPPORT_LEGACY_VERSION)) {
+               dev_err(hba->dev, "%s: HPB %x version is not supported.\n",
+                       __func__, version);
+               hpb_dev_info->hpb_disabled = true;
+               return;
+       }
+       if (version == HPB_SUPPORT_LEGACY_VERSION)
+               hpb_dev_info->is_legacy = true;
+       /*
+        * Get the number of user logical unit to check whether all
+        * scsi_device finish initialization
+        */
+       hpb_dev_info->num_lu = desc_buf[DEVICE_DESC_PARAM_NUM_LU];
+       if (hpb_dev_info->is_legacy)
+               return;
+       ret = ufshcd_query_attr_retry(hba, UPIU_QUERY_OPCODE_READ_ATTR,
+               QUERY_ATTR_IDN_MAX_HPB_SINGLE_CMD, 0, 0, &max_single_cmd);
+       if (ret)
+               hpb_dev_info->max_hpb_single_cmd = HPB_LEGACY_CHUNK_HIGH;
+       else
+               hpb_dev_info->max_hpb_single_cmd = min(max_single_cmd + 1, HPB_MULTI_CHUNK_HIGH);
+ }
+ void ufshpb_init(struct ufs_hba *hba)
+ {
+       struct ufshpb_dev_info *hpb_dev_info = &hba->ufshpb_dev;
+       int try;
+       int ret;
+       if (!ufshpb_is_allowed(hba) || !hba->dev_info.hpb_enabled)
+               return;
+       if (ufshpb_init_mem_wq(hba)) {
+               hpb_dev_info->hpb_disabled = true;
+               return;
+       }
+       atomic_set(&hpb_dev_info->slave_conf_cnt, hpb_dev_info->num_lu);
+       tot_active_srgn_pages = 0;
+       /* issue HPB reset query */
+       for (try = 0; try < HPB_RESET_REQ_RETRIES; try++) {
+               ret = ufshcd_query_flag(hba, UPIU_QUERY_OPCODE_SET_FLAG,
+                                       QUERY_FLAG_IDN_HPB_RESET, 0, NULL);
+               if (!ret)
+                       break;
+       }
+ }
+ void ufshpb_remove(struct ufs_hba *hba)
+ {
+       mempool_destroy(ufshpb_page_pool);
+       mempool_destroy(ufshpb_mctx_pool);
+       kmem_cache_destroy(ufshpb_mctx_cache);
+       destroy_workqueue(ufshpb_wq);
+ }
+ module_param(ufshpb_host_map_kbytes, uint, 0644);
+ MODULE_PARM_DESC(ufshpb_host_map_kbytes,
+       "ufshpb host mapping memory kilo-bytes for ufshpb memory-pool");