scsi: hisi_sas: optimise DMA slot memory

Currently we allocate 3 sets of DMA memories from separate pools for
each slot. This is inefficient in terms of memory usage
(buffers are less than 1 page in size, so we lose due to alignment),
and also time spent in doing 3 allocations + de-allocations per slot,
instead of 1.

To optimise, combine the 3 DMA buffers into a single buffer from a
single pool.

Signed-off-by: Xiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

5 files changed, 104 insertions, 122 deletions

diff --git a/drivers/scsi/hisi_sas/hisi_sas.h b/drivers/scsi/hisi_sas/hisi_sas.h
index 22dd48b..a722f2b 100644
--- a/drivers/scsi/hisi_sas/hisi_sas.h
+++ b/drivers/scsi/hisi_sas/hisi_sas.h
@@ -34,10 +34,24 @@
 #define HISI_SAS_MAX_DEVICES HISI_SAS_MAX_ITCT_ENTRIES
 #define HISI_SAS_RESET_BIT	0
 
-#define HISI_SAS_STATUS_BUF_SZ \
-		(sizeof(struct hisi_sas_err_record) + 1024)
-#define HISI_SAS_COMMAND_TABLE_SZ \
-		(((sizeof(union hisi_sas_command_table)+3)/4)*4)
+#define HISI_SAS_STATUS_BUF_SZ (sizeof(struct hisi_sas_status_buffer))
+#define HISI_SAS_COMMAND_TABLE_SZ (sizeof(union hisi_sas_command_table))
+
+#define hisi_sas_status_buf_addr(buf) \
+	(buf + offsetof(struct hisi_sas_slot_buf_table, status_buffer))
+#define hisi_sas_status_buf_addr_mem(slot) hisi_sas_status_buf_addr(slot->buf)
+#define hisi_sas_status_buf_addr_dma(slot) \
+	hisi_sas_status_buf_addr(slot->buf_dma)
+
+#define hisi_sas_cmd_hdr_addr(buf) \
+	(buf + offsetof(struct hisi_sas_slot_buf_table, command_header))
+#define hisi_sas_cmd_hdr_addr_mem(slot) hisi_sas_cmd_hdr_addr(slot->buf)
+#define hisi_sas_cmd_hdr_addr_dma(slot) hisi_sas_cmd_hdr_addr(slot->buf_dma)
+
+#define hisi_sas_sge_addr(buf) \
+	(buf + offsetof(struct hisi_sas_slot_buf_table, sge_page))
+#define hisi_sas_sge_addr_mem(slot) hisi_sas_sge_addr(slot->buf)
+#define hisi_sas_sge_addr_dma(slot) hisi_sas_sge_addr(slot->buf_dma)
 
 #define HISI_SAS_MAX_SSP_RESP_SZ (sizeof(struct ssp_frame_hdr) + 1024)
 #define HISI_SAS_MAX_SMP_RESP_SZ 1028
@@ -139,14 +153,10 @@ struct hisi_sas_slot {
 	int	cmplt_queue_slot;
 	int	idx;
 	int	abort;
+	void	*buf;
+	dma_addr_t buf_dma;
 	void	*cmd_hdr;
 	dma_addr_t cmd_hdr_dma;
-	void	*status_buffer;
-	dma_addr_t status_buffer_dma;
-	void *command_table;
-	dma_addr_t command_table_dma;
-	struct hisi_sas_sge_page *sge_page;
-	dma_addr_t sge_page_dma;
 	struct work_struct abort_slot;
 	struct timer_list internal_abort_timer;
 };
@@ -232,10 +242,8 @@ struct hisi_hba {
 
 	int	queue_count;
 
-	struct dma_pool *sge_page_pool;
+	struct dma_pool *buffer_pool;
 	struct hisi_sas_device	devices[HISI_SAS_MAX_DEVICES];
-	struct dma_pool *command_table_pool;
-	struct dma_pool *status_buffer_pool;
 	struct hisi_sas_cmd_hdr	*cmd_hdr[HISI_SAS_MAX_QUEUES];
 	dma_addr_t cmd_hdr_dma[HISI_SAS_MAX_QUEUES];
 	void *complete_hdr[HISI_SAS_MAX_QUEUES];
@@ -347,7 +355,7 @@ struct hisi_sas_command_table_stp {
 #define HISI_SAS_SGE_PAGE_CNT SG_CHUNK_SIZE
 struct hisi_sas_sge_page {
 	struct hisi_sas_sge sge[HISI_SAS_SGE_PAGE_CNT];
-};
+}  __aligned(16);
 
 struct hisi_sas_command_table_ssp {
 	struct ssp_frame_hdr hdr;
@@ -366,6 +374,17 @@ union hisi_sas_command_table {
 	struct hisi_sas_command_table_ssp ssp;
 	struct hisi_sas_command_table_smp smp;
 	struct hisi_sas_command_table_stp stp;
+}  __aligned(16);
+
+struct hisi_sas_status_buffer {
+	struct hisi_sas_err_record err;
+	u8	iu[1024];
+}  __aligned(16);
+
+struct hisi_sas_slot_buf_table {
+	struct hisi_sas_status_buffer status_buffer;
+	union hisi_sas_command_table command_header;
+	struct hisi_sas_sge_page sge_page;
 };
 
 extern struct scsi_transport_template *hisi_sas_stt;
diff --git a/drivers/scsi/hisi_sas/hisi_sas_main.c b/drivers/scsi/hisi_sas/hisi_sas_main.c
index e2f8d92..4022c3f 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_main.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_main.c
@@ -87,8 +87,10 @@ void hisi_sas_sata_done(struct sas_task *task,
 {
 	struct task_status_struct *ts = &task->task_status;
 	struct ata_task_resp *resp = (struct ata_task_resp *)ts->buf;
-	struct dev_to_host_fis *d2h = slot->status_buffer +
-				      sizeof(struct hisi_sas_err_record);
+	struct hisi_sas_status_buffer *status_buf =
+			hisi_sas_status_buf_addr_mem(slot);
+	u8 *iu = &status_buf->iu[0];
+	struct dev_to_host_fis *d2h =  (struct dev_to_host_fis *)iu;
 
 	resp->frame_len = sizeof(struct dev_to_host_fis);
 	memcpy(&resp->ending_fis[0], d2h, sizeof(struct dev_to_host_fis));
@@ -183,17 +185,9 @@ void hisi_sas_slot_task_free(struct hisi_hba *hisi_hba, struct sas_task *task,
 			atomic64_dec(&sas_dev->running_req);
 	}
 
-	if (slot->command_table)
-		dma_pool_free(hisi_hba->command_table_pool,
-			      slot->command_table, slot->command_table_dma);
+	if (slot->buf)
+		dma_pool_free(hisi_hba->buffer_pool, slot->buf, slot->buf_dma);
 
-	if (slot->status_buffer)
-		dma_pool_free(hisi_hba->status_buffer_pool,
-			      slot->status_buffer, slot->status_buffer_dma);
-
-	if (slot->sge_page)
-		dma_pool_free(hisi_hba->sge_page_pool, slot->sge_page,
-			      slot->sge_page_dma);
 
 	list_del_init(&slot->entry);
 	slot->task = NULL;
@@ -362,24 +356,15 @@ static int hisi_sas_task_prep(struct sas_task *task, struct hisi_sas_dq
 	task->lldd_task = slot;
 	INIT_WORK(&slot->abort_slot, hisi_sas_slot_abort);
 
-	slot->status_buffer = dma_pool_alloc(hisi_hba->status_buffer_pool,
-					     GFP_ATOMIC,
-					     &slot->status_buffer_dma);
-	if (!slot->status_buffer) {
+	slot->buf = dma_pool_alloc(hisi_hba->buffer_pool,
+				   GFP_ATOMIC, &slot->buf_dma);
+	if (!slot->buf) {
 		rc = -ENOMEM;
 		goto err_out_slot_buf;
 	}
-	memset(slot->status_buffer, 0, HISI_SAS_STATUS_BUF_SZ);
-
-	slot->command_table = dma_pool_alloc(hisi_hba->command_table_pool,
-					     GFP_ATOMIC,
-					     &slot->command_table_dma);
-	if (!slot->command_table) {
-		rc = -ENOMEM;
-		goto err_out_status_buf;
-	}
-	memset(slot->command_table, 0, HISI_SAS_COMMAND_TABLE_SZ);
 	memset(slot->cmd_hdr, 0, sizeof(struct hisi_sas_cmd_hdr));
+	memset(hisi_sas_cmd_hdr_addr_mem(slot), 0, HISI_SAS_COMMAND_TABLE_SZ);
+	memset(hisi_sas_status_buf_addr_mem(slot), 0, HISI_SAS_STATUS_BUF_SZ);
 
 	switch (task->task_proto) {
 	case SAS_PROTOCOL_SMP:
@@ -402,9 +387,7 @@ static int hisi_sas_task_prep(struct sas_task *task, struct hisi_sas_dq
 
 	if (rc) {
 		dev_err(dev, "task prep: rc = 0x%x\n", rc);
-		if (slot->sge_page)
-			goto err_out_sge;
-		goto err_out_command_table;
+		goto err_out_buf;
 	}
 
 	list_add_tail(&slot->entry, &sas_dev->list);
@@ -419,15 +402,9 @@ static int hisi_sas_task_prep(struct sas_task *task, struct hisi_sas_dq
 
 	return 0;
 
-err_out_sge:
-	dma_pool_free(hisi_hba->sge_page_pool, slot->sge_page,
-		slot->sge_page_dma);
-err_out_command_table:
-	dma_pool_free(hisi_hba->command_table_pool, slot->command_table,
-		slot->command_table_dma);
-err_out_status_buf:
-	dma_pool_free(hisi_hba->status_buffer_pool, slot->status_buffer,
-		slot->status_buffer_dma);
+err_out_buf:
+	dma_pool_free(hisi_hba->buffer_pool, slot->buf,
+		slot->buf_dma);
 err_out_slot_buf:
 	/* Nothing to be done */
 err_out_tag:
@@ -1608,16 +1585,9 @@ int hisi_sas_alloc(struct hisi_hba *hisi_hba, struct Scsi_Host *shost)
 			goto err_out;
 	}
 
-	s = HISI_SAS_STATUS_BUF_SZ;
-	hisi_hba->status_buffer_pool = dma_pool_create("status_buffer",
-						       dev, s, 16, 0);
-	if (!hisi_hba->status_buffer_pool)
-		goto err_out;
-
-	s = HISI_SAS_COMMAND_TABLE_SZ;
-	hisi_hba->command_table_pool = dma_pool_create("command_table",
-						       dev, s, 16, 0);
-	if (!hisi_hba->command_table_pool)
+	s = sizeof(struct hisi_sas_slot_buf_table);
+	hisi_hba->buffer_pool = dma_pool_create("dma_buffer", dev, s, 16, 0);
+	if (!hisi_hba->buffer_pool)
 		goto err_out;
 
 	s = HISI_SAS_MAX_ITCT_ENTRIES * sizeof(struct hisi_sas_itct);
@@ -1652,11 +1622,6 @@ int hisi_sas_alloc(struct hisi_hba *hisi_hba, struct Scsi_Host *shost)
 	if (!hisi_hba->slot_index_tags)
 		goto err_out;
 
-	hisi_hba->sge_page_pool = dma_pool_create("status_sge", dev,
-				sizeof(struct hisi_sas_sge_page), 16, 0);
-	if (!hisi_hba->sge_page_pool)
-		goto err_out;
-
 	s = sizeof(struct hisi_sas_initial_fis) * HISI_SAS_MAX_PHYS;
 	hisi_hba->initial_fis = dma_alloc_coherent(dev, s,
 				&hisi_hba->initial_fis_dma, GFP_KERNEL);
@@ -1703,9 +1668,7 @@ void hisi_sas_free(struct hisi_hba *hisi_hba)
 					  hisi_hba->complete_hdr_dma[i]);
 	}
 
-	dma_pool_destroy(hisi_hba->status_buffer_pool);
-	dma_pool_destroy(hisi_hba->command_table_pool);
-	dma_pool_destroy(hisi_hba->sge_page_pool);
+	dma_pool_destroy(hisi_hba->buffer_pool);
 
 	s = HISI_SAS_MAX_ITCT_ENTRIES * sizeof(struct hisi_sas_itct);
 	if (hisi_hba->itct)
diff --git a/drivers/scsi/hisi_sas/hisi_sas_v1_hw.c b/drivers/scsi/hisi_sas/hisi_sas_v1_hw.c
index afa87d4..08eca20 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_v1_hw.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_v1_hw.c
@@ -939,6 +939,7 @@ static int prep_prd_sge_v1_hw(struct hisi_hba *hisi_hba,
 			      struct scatterlist *scatter,
 			      int n_elem)
 {
+	struct hisi_sas_sge_page *sge_page = hisi_sas_sge_addr_mem(slot);
 	struct device *dev = hisi_hba->dev;
 	struct scatterlist *sg;
 	int i;
@@ -949,13 +950,8 @@ static int prep_prd_sge_v1_hw(struct hisi_hba *hisi_hba,
 		return -EINVAL;
 	}
 
-	slot->sge_page = dma_pool_alloc(hisi_hba->sge_page_pool, GFP_ATOMIC,
-					&slot->sge_page_dma);
-	if (!slot->sge_page)
-		return -ENOMEM;
-
 	for_each_sg(scatter, sg, n_elem, i) {
-		struct hisi_sas_sge *entry = &slot->sge_page->sge[i];
+		struct hisi_sas_sge *entry = &sge_page->sge[i];
 
 		entry->addr = cpu_to_le64(sg_dma_address(sg));
 		entry->page_ctrl_0 = entry->page_ctrl_1 = 0;
@@ -963,7 +959,7 @@ static int prep_prd_sge_v1_hw(struct hisi_hba *hisi_hba,
 		entry->data_off = 0;
 	}
 
-	hdr->prd_table_addr = cpu_to_le64(slot->sge_page_dma);
+	hdr->prd_table_addr = cpu_to_le64(hisi_sas_sge_addr_dma(slot));
 
 	hdr->sg_len = cpu_to_le32(n_elem << CMD_HDR_DATA_SGL_LEN_OFF);
 
@@ -1026,7 +1022,7 @@ static int prep_smp_v1_hw(struct hisi_hba *hisi_hba,
 	hdr->transfer_tags = cpu_to_le32(slot->idx << CMD_HDR_IPTT_OFF);
 
 	hdr->cmd_table_addr = cpu_to_le64(req_dma_addr);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
 	return 0;
 
@@ -1107,10 +1103,11 @@ static int prep_ssp_v1_hw(struct hisi_hba *hisi_hba,
 	}
 
 	hdr->data_transfer_len = cpu_to_le32(task->total_xfer_len);
-	hdr->cmd_table_addr = cpu_to_le64(slot->command_table_dma);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->cmd_table_addr = cpu_to_le64(hisi_sas_cmd_hdr_addr_dma(slot));
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
-	buf_cmd = slot->command_table + sizeof(struct ssp_frame_hdr);
+	buf_cmd = hisi_sas_cmd_hdr_addr_mem(slot) +
+		sizeof(struct ssp_frame_hdr);
 	if (task->ssp_task.enable_first_burst) {
 		fburst = (1 << 7);
 		dw2 |= 1 << CMD_HDR_FIRST_BURST_OFF;
@@ -1147,7 +1144,8 @@ static void slot_err_v1_hw(struct hisi_hba *hisi_hba,
 			   struct hisi_sas_slot *slot)
 {
 	struct task_status_struct *ts = &task->task_status;
-	struct hisi_sas_err_record_v1 *err_record = slot->status_buffer;
+	struct hisi_sas_err_record_v1 *err_record =
+			hisi_sas_status_buf_addr_mem(slot);
 	struct device *dev = hisi_hba->dev;
 
 	switch (task->task_proto) {
@@ -1364,8 +1362,11 @@ static int slot_complete_v1_hw(struct hisi_hba *hisi_hba,
 	switch (task->task_proto) {
 	case SAS_PROTOCOL_SSP:
 	{
-		struct ssp_response_iu *iu = slot->status_buffer +
-			sizeof(struct hisi_sas_err_record);
+		struct hisi_sas_status_buffer *status_buffer =
+				hisi_sas_status_buf_addr_mem(slot);
+		struct ssp_response_iu *iu = (struct ssp_response_iu *)
+				&status_buffer->iu[0];
+
 		sas_ssp_task_response(dev, task, iu);
 		break;
 	}
@@ -1382,7 +1383,7 @@ static int slot_complete_v1_hw(struct hisi_hba *hisi_hba,
 		dma_unmap_sg(dev, &task->smp_task.smp_req, 1,
 			     DMA_TO_DEVICE);
 		memcpy(to + sg_resp->offset,
-		       slot->status_buffer +
+		       hisi_sas_status_buf_addr_mem(slot) +
 		       sizeof(struct hisi_sas_err_record),
 		       sg_dma_len(sg_resp));
 		kunmap_atomic(to);
diff --git a/drivers/scsi/hisi_sas/hisi_sas_v2_hw.c b/drivers/scsi/hisi_sas/hisi_sas_v2_hw.c
index 341a0bf..551d103 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_v2_hw.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_v2_hw.c
@@ -1492,6 +1492,7 @@ static int prep_prd_sge_v2_hw(struct hisi_hba *hisi_hba,
 			      struct scatterlist *scatter,
 			      int n_elem)
 {
+	struct hisi_sas_sge_page *sge_page = hisi_sas_sge_addr_mem(slot);
 	struct device *dev = hisi_hba->dev;
 	struct scatterlist *sg;
 	int i;
@@ -1502,13 +1503,8 @@ static int prep_prd_sge_v2_hw(struct hisi_hba *hisi_hba,
 		return -EINVAL;
 	}
 
-	slot->sge_page = dma_pool_alloc(hisi_hba->sge_page_pool, GFP_ATOMIC,
-					&slot->sge_page_dma);
-	if (!slot->sge_page)
-		return -ENOMEM;
-
 	for_each_sg(scatter, sg, n_elem, i) {
-		struct hisi_sas_sge *entry = &slot->sge_page->sge[i];
+		struct hisi_sas_sge *entry = &sge_page->sge[i];
 
 		entry->addr = cpu_to_le64(sg_dma_address(sg));
 		entry->page_ctrl_0 = entry->page_ctrl_1 = 0;
@@ -1516,7 +1512,7 @@ static int prep_prd_sge_v2_hw(struct hisi_hba *hisi_hba,
 		entry->data_off = 0;
 	}
 
-	hdr->prd_table_addr = cpu_to_le64(slot->sge_page_dma);
+	hdr->prd_table_addr = cpu_to_le64(hisi_sas_sge_addr_dma(slot));
 
 	hdr->sg_len = cpu_to_le32(n_elem << CMD_HDR_DATA_SGL_LEN_OFF);
 
@@ -1580,7 +1576,7 @@ static int prep_smp_v2_hw(struct hisi_hba *hisi_hba,
 	hdr->transfer_tags = cpu_to_le32(slot->idx << CMD_HDR_IPTT_OFF);
 
 	hdr->cmd_table_addr = cpu_to_le64(req_dma_addr);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
 	return 0;
 
@@ -1654,10 +1650,11 @@ static int prep_ssp_v2_hw(struct hisi_hba *hisi_hba,
 	}
 
 	hdr->data_transfer_len = cpu_to_le32(task->total_xfer_len);
-	hdr->cmd_table_addr = cpu_to_le64(slot->command_table_dma);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->cmd_table_addr = cpu_to_le64(hisi_sas_cmd_hdr_addr_dma(slot));
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
-	buf_cmd = slot->command_table + sizeof(struct ssp_frame_hdr);
+	buf_cmd = hisi_sas_cmd_hdr_addr_mem(slot) +
+		sizeof(struct ssp_frame_hdr);
 
 	memcpy(buf_cmd, &task->ssp_task.LUN, 8);
 	if (!is_tmf) {
@@ -1884,7 +1881,8 @@ static void slot_err_v2_hw(struct hisi_hba *hisi_hba,
 			   int err_phase)
 {
 	struct task_status_struct *ts = &task->task_status;
-	struct hisi_sas_err_record_v2 *err_record = slot->status_buffer;
+	struct hisi_sas_err_record_v2 *err_record =
+			hisi_sas_status_buf_addr_mem(slot);
 	u32 trans_tx_fail_type = cpu_to_le32(err_record->trans_tx_fail_type);
 	u32 trans_rx_fail_type = cpu_to_le32(err_record->trans_rx_fail_type);
 	u16 dma_tx_err_type = cpu_to_le16(err_record->dma_tx_err_type);
@@ -2273,8 +2271,10 @@ slot_complete_v2_hw(struct hisi_hba *hisi_hba, struct hisi_sas_slot *slot)
 	switch (task->task_proto) {
 	case SAS_PROTOCOL_SSP:
 	{
-		struct ssp_response_iu *iu = slot->status_buffer +
-			sizeof(struct hisi_sas_err_record);
+		struct hisi_sas_status_buffer *status_buffer =
+				hisi_sas_status_buf_addr_mem(slot);
+		struct ssp_response_iu *iu = (struct ssp_response_iu *)
+				&status_buffer->iu[0];
 
 		sas_ssp_task_response(dev, task, iu);
 		break;
@@ -2292,7 +2292,7 @@ slot_complete_v2_hw(struct hisi_hba *hisi_hba, struct hisi_sas_slot *slot)
 		dma_unmap_sg(dev, &task->smp_task.smp_req, 1,
 			     DMA_TO_DEVICE);
 		memcpy(to + sg_resp->offset,
-		       slot->status_buffer +
+		       hisi_sas_status_buf_addr_mem(slot) +
 		       sizeof(struct hisi_sas_err_record),
 		       sg_dma_len(sg_resp));
 		kunmap_atomic(to);
@@ -2398,12 +2398,11 @@ static int prep_ata_v2_hw(struct hisi_hba *hisi_hba,
 			return rc;
 	}
 
-
 	hdr->data_transfer_len = cpu_to_le32(task->total_xfer_len);
-	hdr->cmd_table_addr = cpu_to_le64(slot->command_table_dma);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->cmd_table_addr = cpu_to_le64(hisi_sas_cmd_hdr_addr_dma(slot));
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
-	buf_cmd = slot->command_table;
+	buf_cmd = hisi_sas_cmd_hdr_addr_mem(slot);
 
 	if (likely(!task->ata_task.device_control_reg_update))
 		task->ata_task.fis.flags |= 0x80; /* C=1: update ATA cmd reg */
diff --git a/drivers/scsi/hisi_sas/hisi_sas_v3_hw.c b/drivers/scsi/hisi_sas/hisi_sas_v3_hw.c
index c998b81..83d2dca 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_v3_hw.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_v3_hw.c
@@ -743,6 +743,7 @@ static int prep_prd_sge_v3_hw(struct hisi_hba *hisi_hba,
 			      struct scatterlist *scatter,
 			      int n_elem)
 {
+	struct hisi_sas_sge_page *sge_page = hisi_sas_sge_addr_mem(slot);
 	struct device *dev = hisi_hba->dev;
 	struct scatterlist *sg;
 	int i;
@@ -753,13 +754,8 @@ static int prep_prd_sge_v3_hw(struct hisi_hba *hisi_hba,
 		return -EINVAL;
 	}
 
-	slot->sge_page = dma_pool_alloc(hisi_hba->sge_page_pool, GFP_ATOMIC,
-					&slot->sge_page_dma);
-	if (!slot->sge_page)
-		return -ENOMEM;
-
 	for_each_sg(scatter, sg, n_elem, i) {
-		struct hisi_sas_sge *entry = &slot->sge_page->sge[i];
+		struct hisi_sas_sge *entry = &sge_page->sge[i];
 
 		entry->addr = cpu_to_le64(sg_dma_address(sg));
 		entry->page_ctrl_0 = entry->page_ctrl_1 = 0;
@@ -767,7 +763,8 @@ static int prep_prd_sge_v3_hw(struct hisi_hba *hisi_hba,
 		entry->data_off = 0;
 	}
 
-	hdr->prd_table_addr = cpu_to_le64(slot->sge_page_dma);
+	hdr->prd_table_addr = cpu_to_le64(hisi_sas_sge_addr_dma(slot));
+
 	hdr->sg_len = cpu_to_le32(n_elem << CMD_HDR_DATA_SGL_LEN_OFF);
 
 	return 0;
@@ -833,12 +830,13 @@ static int prep_ssp_v3_hw(struct hisi_hba *hisi_hba,
 	}
 
 	hdr->data_transfer_len = cpu_to_le32(task->total_xfer_len);
-	hdr->cmd_table_addr = cpu_to_le64(slot->command_table_dma);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->cmd_table_addr = cpu_to_le64(hisi_sas_cmd_hdr_addr_dma(slot));
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
-	buf_cmd = slot->command_table + sizeof(struct ssp_frame_hdr);
-	memcpy(buf_cmd, ssp_task->LUN, 8);
+	buf_cmd = hisi_sas_cmd_hdr_addr_mem(slot) +
+		sizeof(struct ssp_frame_hdr);
 
+	memcpy(buf_cmd, &task->ssp_task.LUN, 8);
 	if (!is_tmf) {
 		buf_cmd[9] = ssp_task->task_attr | (ssp_task->task_prio << 3);
 		memcpy(buf_cmd + 12, scsi_cmnd->cmnd, scsi_cmnd->cmd_len);
@@ -917,7 +915,7 @@ static int prep_smp_v3_hw(struct hisi_hba *hisi_hba,
 	hdr->transfer_tags = cpu_to_le32(slot->idx << CMD_HDR_IPTT_OFF);
 
 	hdr->cmd_table_addr = cpu_to_le64(req_dma_addr);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
 	return 0;
 
@@ -1012,10 +1010,10 @@ static int prep_ata_v3_hw(struct hisi_hba *hisi_hba,
 	}
 
 	hdr->data_transfer_len = cpu_to_le32(task->total_xfer_len);
-	hdr->cmd_table_addr = cpu_to_le64(slot->command_table_dma);
-	hdr->sts_buffer_addr = cpu_to_le64(slot->status_buffer_dma);
+	hdr->cmd_table_addr = cpu_to_le64(hisi_sas_cmd_hdr_addr_dma(slot));
+	hdr->sts_buffer_addr = cpu_to_le64(hisi_sas_status_buf_addr_dma(slot));
 
-	buf_cmd = slot->command_table;
+	buf_cmd = hisi_sas_cmd_hdr_addr_mem(slot);
 
 	if (likely(!task->ata_task.device_control_reg_update))
 		task->ata_task.fis.flags |= 0x80; /* C=1: update ATA cmd reg */
@@ -1283,7 +1281,8 @@ slot_err_v3_hw(struct hisi_hba *hisi_hba, struct sas_task *task,
 			hisi_hba->complete_hdr[slot->cmplt_queue];
 	struct hisi_sas_complete_v3_hdr *complete_hdr =
 			&complete_queue[slot->cmplt_queue_slot];
-	struct hisi_sas_err_record_v3 *record =	slot->status_buffer;
+	struct hisi_sas_err_record_v3 *record =
+			hisi_sas_status_buf_addr_mem(slot);
 	u32 dma_rx_err_type = record->dma_rx_err_type;
 	u32 trans_tx_fail_type = record->trans_tx_fail_type;
 
@@ -1402,7 +1401,8 @@ slot_complete_v3_hw(struct hisi_hba *hisi_hba, struct hisi_sas_slot *slot)
 
 	switch (task->task_proto) {
 	case SAS_PROTOCOL_SSP: {
-		struct ssp_response_iu *iu = slot->status_buffer +
+		struct ssp_response_iu *iu =
+			hisi_sas_status_buf_addr_mem(slot) +
 			sizeof(struct hisi_sas_err_record);
 
 		sas_ssp_task_response(dev, task, iu);
@@ -1420,7 +1420,7 @@ slot_complete_v3_hw(struct hisi_hba *hisi_hba, struct hisi_sas_slot *slot)
 		dma_unmap_sg(dev, &task->smp_task.smp_req, 1,
 			     DMA_TO_DEVICE);
 		memcpy(to + sg_resp->offset,
-		       slot->status_buffer +
+			hisi_sas_status_buf_addr_mem(slot) +
 		       sizeof(struct hisi_sas_err_record),
 		       sg_dma_len(sg_resp));
 		kunmap_atomic(to);