SCSI sg on 20190709

This topic branch covers a fundamental change in how our sg lists are
 allocated to make mq more efficient by reducing the size of the
 preallocated sg list.  This necessitates a large number of driver
 changes because the previous guarantee that if a driver specified
 SG_ALL as the size of its scatter list, it would get a non-chained
 list and didn't need to bother with scatterlist iterators is now
 broken and every driver *must* use scatterlist iterators.
 
 This was broken out as a separate topic because we need to convert all
 the drivers before pulling the trigger and unconverted drivers kept
 being found, necessitating a rebase.
 
 Signed-off-by: James E.J. Bottomley <jejb@linux.ibm.com>
 -----BEGIN PGP SIGNATURE-----
 
 iJwEABMIAEQWIQTnYEDbdso9F2cI+arnQslM7pishQUCXSTzzCYcamFtZXMuYm90
 dG9tbGV5QGhhbnNlbnBhcnRuZXJzaGlwLmNvbQAKCRDnQslM7pishZB+AP9I8j/s
 wWfg0Z3WNuf4D5I3rH4x1J3cQTqPJed+RjwgcQEA1gZvtOTg1ZEn/CYMVnaB92x0
 t6MZSchIaFXeqfD+E7U=
 =cv8o
 -----END PGP SIGNATURE-----

Merge tag 'scsi-sg' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi

Pull SCSI scatter-gather list updates from James Bottomley:
 "This topic branch covers a fundamental change in how our sg lists are
  allocated to make mq more efficient by reducing the size of the
  preallocated sg list.

  This necessitates a large number of driver changes because the
  previous guarantee that if a driver specified SG_ALL as the size of
  its scatter list, it would get a non-chained list and didn't need to
  bother with scatterlist iterators is now broken and every driver
  *must* use scatterlist iterators.

  This was broken out as a separate topic because we need to convert all
  the drivers before pulling the trigger and unconverted drivers kept
  being found, necessitating a rebase"

* tag 'scsi-sg' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (21 commits)
  scsi: core: don't preallocate small SGL in case of NO_SG_CHAIN
  scsi: lib/sg_pool.c: clear 'first_chunk' in case of no preallocation
  scsi: core: avoid preallocating big SGL for data
  scsi: core: avoid preallocating big SGL for protection information
  scsi: lib/sg_pool.c: improve APIs for allocating sg pool
  scsi: esp: use sg helper to iterate over scatterlist
  scsi: NCR5380: use sg helper to iterate over scatterlist
  scsi: wd33c93: use sg helper to iterate over scatterlist
  scsi: ppa: use sg helper to iterate over scatterlist
  scsi: pcmcia: nsp_cs: use sg helper to iterate over scatterlist
  scsi: imm: use sg helper to iterate over scatterlist
  scsi: aha152x: use sg helper to iterate over scatterlist
  scsi: s390: zfcp_fc: use sg helper to iterate over scatterlist
  scsi: staging: unisys: visorhba: use sg helper to iterate over scatterlist
  scsi: usb: image: microtek: use sg helper to iterate over scatterlist
  scsi: pmcraid: use sg helper to iterate over scatterlist
  scsi: ipr: use sg helper to iterate over scatterlist
  scsi: mvumi: use sg helper to iterate over scatterlist
  scsi: lpfc: use sg helper to iterate over scatterlist
  scsi: advansys: use sg helper to iterate over scatterlist
  ...

drivers/nvme/host/fc.c

@@ -2112,7 +2112,8 @@ nvme_fc_map_data(struct nvme_fc_ctrl *ctrl, struct request *rq,
 
 	freq->sg_table.sgl = freq->first_sgl;
 	ret = sg_alloc_table_chained(&freq->sg_table,
-			blk_rq_nr_phys_segments(rq), freq->sg_table.sgl);
+			blk_rq_nr_phys_segments(rq), freq->sg_table.sgl,
+			SG_CHUNK_SIZE);
 	if (ret)
 		return -ENOMEM;
 
@@ -2122,7 +2123,7 @@ nvme_fc_map_data(struct nvme_fc_ctrl *ctrl, struct request *rq,
 	freq->sg_cnt = fc_dma_map_sg(ctrl->lport->dev, freq->sg_table.sgl,
 				op->nents, dir);
 	if (unlikely(freq->sg_cnt <= 0)) {
-		sg_free_table_chained(&freq->sg_table, true);
+		sg_free_table_chained(&freq->sg_table, SG_CHUNK_SIZE);
 		freq->sg_cnt = 0;
 		return -EFAULT;
 	}
@@ -2148,7 +2149,7 @@ nvme_fc_unmap_data(struct nvme_fc_ctrl *ctrl, struct request *rq,
 
 	nvme_cleanup_cmd(rq);
 
-	sg_free_table_chained(&freq->sg_table, true);
+	sg_free_table_chained(&freq->sg_table, SG_CHUNK_SIZE);
 
 	freq->sg_cnt = 0;
 }

drivers/nvme/host/rdma.c

@@ -1144,7 +1144,7 @@ static void nvme_rdma_unmap_data(struct nvme_rdma_queue *queue,
 				    WRITE ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
 
 	nvme_cleanup_cmd(rq);
-	sg_free_table_chained(&req->sg_table, true);
+	sg_free_table_chained(&req->sg_table, SG_CHUNK_SIZE);
 }
 
 static int nvme_rdma_set_sg_null(struct nvme_command *c)
@@ -1259,7 +1259,8 @@ static int nvme_rdma_map_data(struct nvme_rdma_queue *queue,
 
 	req->sg_table.sgl = req->first_sgl;
 	ret = sg_alloc_table_chained(&req->sg_table,
-			blk_rq_nr_phys_segments(rq), req->sg_table.sgl);
+			blk_rq_nr_phys_segments(rq), req->sg_table.sgl,
+			SG_CHUNK_SIZE);
 	if (ret)
 		return -ENOMEM;
 
@@ -1299,7 +1300,7 @@ out_unmap_sg:
 			req->nents, rq_data_dir(rq) ==
 			WRITE ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
 out_free_table:
-	sg_free_table_chained(&req->sg_table, true);
+	sg_free_table_chained(&req->sg_table, SG_CHUNK_SIZE);
 	return ret;
 }
 

drivers/nvme/target/loop.c

@@ -77,7 +77,7 @@ static void nvme_loop_complete_rq(struct request *req)
 	struct nvme_loop_iod *iod = blk_mq_rq_to_pdu(req);
 
 	nvme_cleanup_cmd(req);
-	sg_free_table_chained(&iod->sg_table, true);
+	sg_free_table_chained(&iod->sg_table, SG_CHUNK_SIZE);
 	nvme_complete_rq(req);
 }
 
@@ -157,7 +157,7 @@ static blk_status_t nvme_loop_queue_rq(struct blk_mq_hw_ctx *hctx,
 		iod->sg_table.sgl = iod->first_sgl;
 		if (sg_alloc_table_chained(&iod->sg_table,
 				blk_rq_nr_phys_segments(req),
-				iod->sg_table.sgl))
+				iod->sg_table.sgl, SG_CHUNK_SIZE))
 			return BLK_STS_RESOURCE;
 
 		iod->req.sg = iod->sg_table.sgl;

drivers/s390/scsi/zfcp_fc.c

@@ -620,7 +620,7 @@ static void zfcp_fc_sg_free_table(struct scatterlist *sg, int count)
 {
 	int i;
 
-	for (i = 0; i < count; i++, sg++)
+	for (i = 0; i < count; i++, sg = sg_next(sg))
 		if (sg)
 			free_page((unsigned long) sg_virt(sg));
 		else
@@ -641,7 +641,7 @@ static int zfcp_fc_sg_setup_table(struct scatterlist *sg, int count)
 	int i;
 
 	sg_init_table(sg, count);
-	for (i = 0; i < count; i++, sg++) {
+	for (i = 0; i < count; i++, sg = sg_next(sg)) {
 		addr = (void *) get_zeroed_page(GFP_KERNEL);
 		if (!addr) {
 			zfcp_fc_sg_free_table(sg, i);

drivers/scsi/NCR5380.c

@@ -149,12 +149,10 @@ static inline void initialize_SCp(struct scsi_cmnd *cmd)
 
 	if (scsi_bufflen(cmd)) {
 		cmd->SCp.buffer = scsi_sglist(cmd);
-		cmd->SCp.buffers_residual = scsi_sg_count(cmd) - 1;
 		cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);
 		cmd->SCp.this_residual = cmd->SCp.buffer->length;
 	} else {
 		cmd->SCp.buffer = NULL;
-		cmd->SCp.buffers_residual = 0;
 		cmd->SCp.ptr = NULL;
 		cmd->SCp.this_residual = 0;
 	}
@@ -163,6 +161,17 @@ static inline void initialize_SCp(struct scsi_cmnd *cmd)
 	cmd->SCp.Message = 0;
 }
 
+static inline void advance_sg_buffer(struct scsi_cmnd *cmd)
+{
+	struct scatterlist *s = cmd->SCp.buffer;
+
+	if (!cmd->SCp.this_residual && s && !sg_is_last(s)) {
+		cmd->SCp.buffer = sg_next(s);
+		cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);
+		cmd->SCp.this_residual = cmd->SCp.buffer->length;
+	}
+}
+
 /**
  * NCR5380_poll_politely2 - wait for two chip register values
  * @hostdata: host private data
@@ -1670,12 +1679,7 @@ static void NCR5380_information_transfer(struct Scsi_Host *instance)
 			    sun3_dma_setup_done != cmd) {
 				int count;
 
-				if (!cmd->SCp.this_residual && cmd->SCp.buffers_residual) {
+				advance_sg_buffer(cmd);
-					++cmd->SCp.buffer;
-					--cmd->SCp.buffers_residual;
-					cmd->SCp.this_residual = cmd->SCp.buffer->length;
-					cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);
-				}
 
 				count = sun3scsi_dma_xfer_len(hostdata, cmd);
 
@@ -1725,15 +1729,11 @@ static void NCR5380_information_transfer(struct Scsi_Host *instance)
 				 * scatter-gather list, move onto the next one.
 				 */
 
-				if (!cmd->SCp.this_residual && cmd->SCp.buffers_residual) {
+				advance_sg_buffer(cmd);
-					++cmd->SCp.buffer;
+				dsprintk(NDEBUG_INFORMATION, instance,
-					--cmd->SCp.buffers_residual;
+					"this residual %d, sg ents %d\n",
-					cmd->SCp.this_residual = cmd->SCp.buffer->length;
+					cmd->SCp.this_residual,
-					cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);
+					sg_nents(cmd->SCp.buffer));
-					dsprintk(NDEBUG_INFORMATION, instance, "%d bytes and %d buffers left\n",
-					         cmd->SCp.this_residual,
-					         cmd->SCp.buffers_residual);
-				}
 
 				/*
 				 * The preferred transfer method is going to be
@@ -2126,12 +2126,7 @@ static void NCR5380_reselect(struct Scsi_Host *instance)
 	if (sun3_dma_setup_done != tmp) {
 		int count;
 
-		if (!tmp->SCp.this_residual && tmp->SCp.buffers_residual) {
+		advance_sg_buffer(tmp);
-			++tmp->SCp.buffer;
-			--tmp->SCp.buffers_residual;
-			tmp->SCp.this_residual = tmp->SCp.buffer->length;
-			tmp->SCp.ptr = sg_virt(tmp->SCp.buffer);
-		}
 
 		count = sun3scsi_dma_xfer_len(hostdata, tmp);
 

drivers/scsi/advansys.c

@@ -7710,7 +7710,7 @@ adv_get_sglist(struct asc_board *boardp, adv_req_t *reqp,
 				sg_block->sg_ptr = 0L; /* Last ADV_SG_BLOCK in list. */
 				return ADV_SUCCESS;
 			}
-			slp++;
+			slp = sg_next(slp);
 		}
 		sg_block->sg_cnt = NO_OF_SG_PER_BLOCK;
 		prev_sg_block = sg_block;

drivers/scsi/aha152x.c

@@ -937,7 +937,6 @@ static int aha152x_internal_queue(struct scsi_cmnd *SCpnt,
 	   SCp.ptr              : buffer pointer
 	   SCp.this_residual    : buffer length
 	   SCp.buffer           : next buffer
-	   SCp.buffers_residual : left buffers in list
 	   SCp.phase            : current state of the command */
 
 	if ((phase & resetting) || !scsi_sglist(SCpnt)) {
@@ -945,13 +944,11 @@ static int aha152x_internal_queue(struct scsi_cmnd *SCpnt,
 		SCpnt->SCp.this_residual = 0;
 		scsi_set_resid(SCpnt, 0);
 		SCpnt->SCp.buffer           = NULL;
-		SCpnt->SCp.buffers_residual = 0;
 	} else {
 		scsi_set_resid(SCpnt, scsi_bufflen(SCpnt));
 		SCpnt->SCp.buffer           = scsi_sglist(SCpnt);
 		SCpnt->SCp.ptr              = SG_ADDRESS(SCpnt->SCp.buffer);
 		SCpnt->SCp.this_residual    = SCpnt->SCp.buffer->length;
-		SCpnt->SCp.buffers_residual = scsi_sg_count(SCpnt) - 1;
 	}
 
 	DO_LOCK(flags);
@@ -2019,10 +2016,9 @@ static void datai_run(struct Scsi_Host *shpnt)
 				}
 
 				if (CURRENT_SC->SCp.this_residual == 0 &&
-				    CURRENT_SC->SCp.buffers_residual > 0) {
+				    !sg_is_last(CURRENT_SC->SCp.buffer)) {
 					/* advance to next buffer */
-					CURRENT_SC->SCp.buffers_residual--;
+					CURRENT_SC->SCp.buffer = sg_next(CURRENT_SC->SCp.buffer);
-					CURRENT_SC->SCp.buffer++;
 					CURRENT_SC->SCp.ptr           = SG_ADDRESS(CURRENT_SC->SCp.buffer);
 					CURRENT_SC->SCp.this_residual = CURRENT_SC->SCp.buffer->length;
 				}
@@ -2125,10 +2121,10 @@ static void datao_run(struct Scsi_Host *shpnt)
 			CMD_INC_RESID(CURRENT_SC, -2 * data_count);
 		}
 
-		if(CURRENT_SC->SCp.this_residual==0 && CURRENT_SC->SCp.buffers_residual>0) {
+		if (CURRENT_SC->SCp.this_residual == 0 &&
+		    !sg_is_last(CURRENT_SC->SCp.buffer)) {
 			/* advance to next buffer */
-			CURRENT_SC->SCp.buffers_residual--;
+			CURRENT_SC->SCp.buffer = sg_next(CURRENT_SC->SCp.buffer);
-			CURRENT_SC->SCp.buffer++;
 			CURRENT_SC->SCp.ptr           = SG_ADDRESS(CURRENT_SC->SCp.buffer);
 			CURRENT_SC->SCp.this_residual = CURRENT_SC->SCp.buffer->length;
 		}
@@ -2147,22 +2143,26 @@ static void datao_run(struct Scsi_Host *shpnt)
 static void datao_end(struct Scsi_Host *shpnt)
 {
 	if(TESTLO(DMASTAT, DFIFOEMP)) {
-		int data_count = (DATA_LEN - scsi_get_resid(CURRENT_SC)) -
+		u32 datao_cnt = GETSTCNT();
-			GETSTCNT();
+		int datao_out = DATA_LEN - scsi_get_resid(CURRENT_SC);
+		int done;
+		struct scatterlist *sg = scsi_sglist(CURRENT_SC);
 
-		CMD_INC_RESID(CURRENT_SC, data_count);
+		CMD_INC_RESID(CURRENT_SC, datao_out - datao_cnt);
 
-		data_count -= CURRENT_SC->SCp.ptr -
+		done = scsi_bufflen(CURRENT_SC) - scsi_get_resid(CURRENT_SC);
-			SG_ADDRESS(CURRENT_SC->SCp.buffer);
+		/* Locate the first SG entry not yet sent */
-		while(data_count>0) {
+		while (done > 0 && !sg_is_last(sg)) {
-			CURRENT_SC->SCp.buffer--;
+			if (done < sg->length)
-			CURRENT_SC->SCp.buffers_residual++;
+				break;
-			data_count -= CURRENT_SC->SCp.buffer->length;
+			done -= sg->length;
+			sg = sg_next(sg);
 		}
-		CURRENT_SC->SCp.ptr = SG_ADDRESS(CURRENT_SC->SCp.buffer) -
+
-			data_count;
+		CURRENT_SC->SCp.buffer = sg;
-		CURRENT_SC->SCp.this_residual = CURRENT_SC->SCp.buffer->length +
+		CURRENT_SC->SCp.ptr = SG_ADDRESS(CURRENT_SC->SCp.buffer) + done;
-			data_count;
+		CURRENT_SC->SCp.this_residual = CURRENT_SC->SCp.buffer->length -
+			done;
 	}
 
 	SETPORT(SXFRCTL0, CH1|CLRCH1|CLRSTCNT);
@@ -2490,7 +2490,7 @@ static void get_command(struct seq_file *m, struct scsi_cmnd * ptr)
 
 	seq_printf(m, "); resid=%d; residual=%d; buffers=%d; phase |",
 		scsi_get_resid(ptr), ptr->SCp.this_residual,
-		ptr->SCp.buffers_residual);
+		sg_nents(ptr->SCp.buffer) - 1);
 
 	if (ptr->SCp.phase & not_issued)
 		seq_puts(m, "not issued|");

drivers/scsi/esp_scsi.c

@@ -371,6 +371,7 @@ static void esp_map_dma(struct esp *esp, struct scsi_cmnd *cmd)
 	struct esp_cmd_priv *spriv = ESP_CMD_PRIV(cmd);
 	struct scatterlist *sg = scsi_sglist(cmd);
 	int total = 0, i;
+	struct scatterlist *s;
 
 	if (cmd->sc_data_direction == DMA_NONE)
 		return;
@@ -381,16 +382,18 @@ static void esp_map_dma(struct esp *esp, struct scsi_cmnd *cmd)
 		 * a dma address, so perform an identity mapping.
 		 */
 		spriv->num_sg = scsi_sg_count(cmd);
-		for (i = 0; i < spriv->num_sg; i++) {
+
-			sg[i].dma_address = (uintptr_t)sg_virt(&sg[i]);
+		scsi_for_each_sg(cmd, s, spriv->num_sg, i) {
-			total += sg_dma_len(&sg[i]);
+			s->dma_address = (uintptr_t)sg_virt(s);
+			total += sg_dma_len(s);
 		}
 	} else {
 		spriv->num_sg = scsi_dma_map(cmd);
-		for (i = 0; i < spriv->num_sg; i++)
+		scsi_for_each_sg(cmd, s, spriv->num_sg, i)
-			total += sg_dma_len(&sg[i]);
+			total += sg_dma_len(s);
 	}
 	spriv->cur_residue = sg_dma_len(sg);
+	spriv->prv_sg = NULL;
 	spriv->cur_sg = sg;
 	spriv->tot_residue = total;
 }
@@ -444,7 +447,8 @@ static void esp_advance_dma(struct esp *esp, struct esp_cmd_entry *ent,
 		p->tot_residue = 0;
 	}
 	if (!p->cur_residue && p->tot_residue) {
-		p->cur_sg++;
+		p->prv_sg = p->cur_sg;
+		p->cur_sg = sg_next(p->cur_sg);
 		p->cur_residue = sg_dma_len(p->cur_sg);
 	}
 }
@@ -465,6 +469,7 @@ static void esp_save_pointers(struct esp *esp, struct esp_cmd_entry *ent)
 		return;
 	}
 	ent->saved_cur_residue = spriv->cur_residue;
+	ent->saved_prv_sg = spriv->prv_sg;
 	ent->saved_cur_sg = spriv->cur_sg;
 	ent->saved_tot_residue = spriv->tot_residue;
 }
@@ -479,6 +484,7 @@ static void esp_restore_pointers(struct esp *esp, struct esp_cmd_entry *ent)
 		return;
 	}
 	spriv->cur_residue = ent->saved_cur_residue;
+	spriv->prv_sg = ent->saved_prv_sg;
 	spriv->cur_sg = ent->saved_cur_sg;
 	spriv->tot_residue = ent->saved_tot_residue;
 }
@@ -1647,7 +1653,7 @@ static int esp_msgin_process(struct esp *esp)
 		spriv = ESP_CMD_PRIV(ent->cmd);
 
 		if (spriv->cur_residue == sg_dma_len(spriv->cur_sg)) {
-			spriv->cur_sg--;
+			spriv->cur_sg = spriv->prv_sg;
 			spriv->cur_residue = 1;
 		} else
 			spriv->cur_residue++;

drivers/scsi/esp_scsi.h

@@ -251,6 +251,7 @@
 struct esp_cmd_priv {
 	int			num_sg;
 	int			cur_residue;
+	struct scatterlist	*prv_sg;
 	struct scatterlist	*cur_sg;
 	int			tot_residue;
 };
@@ -273,6 +274,7 @@ struct esp_cmd_entry {
 	struct scsi_cmnd	*cmd;
 
 	unsigned int		saved_cur_residue;
+	struct scatterlist	*saved_prv_sg;
 	struct scatterlist	*saved_cur_sg;
 	unsigned int		saved_tot_residue;
 

drivers/scsi/imm.c

@@ -687,7 +687,7 @@ static int imm_completion(struct scsi_cmnd *cmd)
 		if (cmd->SCp.buffer && !cmd->SCp.this_residual) {
 			/* if scatter/gather, advance to the next segment */
 			if (cmd->SCp.buffers_residual--) {
-				cmd->SCp.buffer++;
+				cmd->SCp.buffer = sg_next(cmd->SCp.buffer);
 				cmd->SCp.this_residual =
 				    cmd->SCp.buffer->length;
 				cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);

drivers/scsi/ipr.c

@@ -3901,22 +3901,23 @@ static int ipr_copy_ucode_buffer(struct ipr_sglist *sglist,
 				 u8 *buffer, u32 len)
 {
 	int bsize_elem, i, result = 0;
-	struct scatterlist *scatterlist;
+	struct scatterlist *sg;
 	void *kaddr;
 
 	/* Determine the actual number of bytes per element */
 	bsize_elem = PAGE_SIZE * (1 << sglist->order);
 
-	scatterlist = sglist->scatterlist;
+	sg = sglist->scatterlist;
 
-	for (i = 0; i < (len / bsize_elem); i++, buffer += bsize_elem) {
+	for (i = 0; i < (len / bsize_elem); i++, sg = sg_next(sg),
-		struct page *page = sg_page(&scatterlist[i]);
+			buffer += bsize_elem) {
+		struct page *page = sg_page(sg);
 
 		kaddr = kmap(page);
 		memcpy(kaddr, buffer, bsize_elem);
 		kunmap(page);
 
-		scatterlist[i].length = bsize_elem;
+		sg->length = bsize_elem;
 
 		if (result != 0) {
 			ipr_trace;
@@ -3925,13 +3926,13 @@ static int ipr_copy_ucode_buffer(struct ipr_sglist *sglist,
 	}
 
 	if (len % bsize_elem) {
-		struct page *page = sg_page(&scatterlist[i]);
+		struct page *page = sg_page(sg);
 
 		kaddr = kmap(page);
 		memcpy(kaddr, buffer, len % bsize_elem);
 		kunmap(page);
 
-		scatterlist[i].length = len % bsize_elem;
+		sg->length = len % bsize_elem;
 	}
 
 	sglist->buffer_len = len;
@@ -3952,6 +3953,7 @@ static void ipr_build_ucode_ioadl64(struct ipr_cmnd *ipr_cmd,
 	struct ipr_ioarcb *ioarcb = &ipr_cmd->ioarcb;
 	struct ipr_ioadl64_desc *ioadl64 = ipr_cmd->i.ioadl64;
 	struct scatterlist *scatterlist = sglist->scatterlist;
+	struct scatterlist *sg;
 	int i;
 
 	ipr_cmd->dma_use_sg = sglist->num_dma_sg;
@@ -3960,10 +3962,10 @@ static void ipr_build_ucode_ioadl64(struct ipr_cmnd *ipr_cmd,
 
 	ioarcb->ioadl_len =
 		cpu_to_be32(sizeof(struct ipr_ioadl64_desc) * ipr_cmd->dma_use_sg);
-	for (i = 0; i < ipr_cmd->dma_use_sg; i++) {
+	for_each_sg(scatterlist, sg, ipr_cmd->dma_use_sg, i) {
 		ioadl64[i].flags = cpu_to_be32(IPR_IOADL_FLAGS_WRITE);
-		ioadl64[i].data_len = cpu_to_be32(sg_dma_len(&scatterlist[i]));
+		ioadl64[i].data_len = cpu_to_be32(sg_dma_len(sg));
-		ioadl64[i].address = cpu_to_be64(sg_dma_address(&scatterlist[i]));
+		ioadl64[i].address = cpu_to_be64(sg_dma_address(sg));
 	}
 
 	ioadl64[i-1].flags |= cpu_to_be32(IPR_IOADL_FLAGS_LAST);
@@ -3983,6 +3985,7 @@ static void ipr_build_ucode_ioadl(struct ipr_cmnd *ipr_cmd,
 	struct ipr_ioarcb *ioarcb = &ipr_cmd->ioarcb;
 	struct ipr_ioadl_desc *ioadl = ipr_cmd->i.ioadl;
 	struct scatterlist *scatterlist = sglist->scatterlist;
+	struct scatterlist *sg;
 	int i;
 
 	ipr_cmd->dma_use_sg = sglist->num_dma_sg;
@@ -3992,11 +3995,11 @@ static void ipr_build_ucode_ioadl(struct ipr_cmnd *ipr_cmd,
 	ioarcb->ioadl_len =
 		cpu_to_be32(sizeof(struct ipr_ioadl_desc) * ipr_cmd->dma_use_sg);
 
-	for (i = 0; i < ipr_cmd->dma_use_sg; i++) {
+	for_each_sg(scatterlist, sg, ipr_cmd->dma_use_sg, i) {
 		ioadl[i].flags_and_data_len =
-			cpu_to_be32(IPR_IOADL_FLAGS_WRITE | sg_dma_len(&scatterlist[i]));
+			cpu_to_be32(IPR_IOADL_FLAGS_WRITE | sg_dma_len(sg));
 		ioadl[i].address =
-			cpu_to_be32(sg_dma_address(&scatterlist[i]));
+			cpu_to_be32(sg_dma_address(sg));
 	}
 
 	ioadl[i-1].flags_and_data_len |=

drivers/scsi/lpfc/lpfc_nvmet.c

@@ -2904,8 +2904,7 @@ lpfc_nvmet_prep_fcp_wqe(struct lpfc_hba *phba,
 	nvmewqe->drvrTimeout = (phba->fc_ratov * 3) + LPFC_DRVR_TIMEOUT;
 	nvmewqe->context1 = ndlp;
 
-	for (i = 0; i < rsp->sg_cnt; i++) {
+	for_each_sg(rsp->sg, sgel, rsp->sg_cnt, i) {
-		sgel = &rsp->sg[i];
 		physaddr = sg_dma_address(sgel);
 		cnt = sg_dma_len(sgel);
 		sgl->addr_hi = putPaddrHigh(physaddr);

drivers/scsi/mvumi.c

@@ -195,23 +195,22 @@ static int mvumi_make_sgl(struct mvumi_hba *mhba, struct scsi_cmnd *scmd,
 	unsigned int sgnum = scsi_sg_count(scmd);
 	dma_addr_t busaddr;
 
-	sg = scsi_sglist(scmd);
+	*sg_count = dma_map_sg(&mhba->pdev->dev, scsi_sglist(scmd), sgnum,
-	*sg_count = dma_map_sg(&mhba->pdev->dev, sg, sgnum,
 			       scmd->sc_data_direction);
 	if (*sg_count > mhba->max_sge) {
 		dev_err(&mhba->pdev->dev,
 			"sg count[0x%x] is bigger than max sg[0x%x].\n",
 			*sg_count, mhba->max_sge);
-		dma_unmap_sg(&mhba->pdev->dev, sg, sgnum,
+		dma_unmap_sg(&mhba->pdev->dev, scsi_sglist(scmd), sgnum,
 			     scmd->sc_data_direction);
 		return -1;
 	}
-	for (i = 0; i < *sg_count; i++) {
+	scsi_for_each_sg(scmd, sg, *sg_count, i) {
-		busaddr = sg_dma_address(&sg[i]);
+		busaddr = sg_dma_address(sg);
 		m_sg->baseaddr_l = cpu_to_le32(lower_32_bits(busaddr));
 		m_sg->baseaddr_h = cpu_to_le32(upper_32_bits(busaddr));
 		m_sg->flags = 0;
-		sgd_setsz(mhba, m_sg, cpu_to_le32(sg_dma_len(&sg[i])));
+		sgd_setsz(mhba, m_sg, cpu_to_le32(sg_dma_len(sg)));
 		if ((i + 1) == *sg_count)
 			m_sg->flags |= 1U << mhba->eot_flag;
 

drivers/scsi/pcmcia/nsp_cs.c

@@ -789,7 +789,7 @@ static void nsp_pio_read(struct scsi_cmnd *SCpnt)
 		    SCpnt->SCp.buffers_residual != 0 ) {
 			//nsp_dbg(NSP_DEBUG_DATA_IO, "scatterlist next timeout=%d", time_out);
 			SCpnt->SCp.buffers_residual--;
-			SCpnt->SCp.buffer++;
+			SCpnt->SCp.buffer = sg_next(SCpnt->SCp.buffer);
 			SCpnt->SCp.ptr		 = BUFFER_ADDR;
 			SCpnt->SCp.this_residual = SCpnt->SCp.buffer->length;
 			time_out = 1000;
@@ -887,7 +887,7 @@ static void nsp_pio_write(struct scsi_cmnd *SCpnt)
 		    SCpnt->SCp.buffers_residual != 0 ) {
 			//nsp_dbg(NSP_DEBUG_DATA_IO, "scatterlist next");
 			SCpnt->SCp.buffers_residual--;
-			SCpnt->SCp.buffer++;
+			SCpnt->SCp.buffer = sg_next(SCpnt->SCp.buffer);
 			SCpnt->SCp.ptr		 = BUFFER_ADDR;
 			SCpnt->SCp.this_residual = SCpnt->SCp.buffer->length;
 			time_out = 1000;

drivers/scsi/pmcraid.c

@@ -3255,7 +3255,7 @@ static int pmcraid_copy_sglist(
 	int direction
 )
 {
-	struct scatterlist *scatterlist;
+	struct scatterlist *sg;
 	void *kaddr;
 	int bsize_elem;
 	int i;
@@ -3264,10 +3264,10 @@ static int pmcraid_copy_sglist(
 	/* Determine the actual number of bytes per element */
 	bsize_elem = PAGE_SIZE * (1 << sglist->order);
 
-	scatterlist = sglist->scatterlist;
+	sg = sglist->scatterlist;
 
-	for (i = 0; i < (len / bsize_elem); i++, buffer += bsize_elem) {
+	for (i = 0; i < (len / bsize_elem); i++, sg = sg_next(sg), buffer += bsize_elem) {
-		struct page *page = sg_page(&scatterlist[i]);
+		struct page *page = sg_page(sg);
 
 		kaddr = kmap(page);
 		if (direction == DMA_TO_DEVICE)
@@ -3282,11 +3282,11 @@ static int pmcraid_copy_sglist(
 			return -EFAULT;
 		}
 
-		scatterlist[i].length = bsize_elem;
+		sg->length = bsize_elem;
 	}
 
 	if (len % bsize_elem) {
-		struct page *page = sg_page(&scatterlist[i]);
+		struct page *page = sg_page(sg);
 
 		kaddr = kmap(page);
 
@@ -3297,7 +3297,7 @@ static int pmcraid_copy_sglist(
 
 		kunmap(page);
 
-		scatterlist[i].length = len % bsize_elem;
+		sg->length = len % bsize_elem;
 	}
 
 	if (rc) {

drivers/scsi/ppa.c

@@ -590,7 +590,7 @@ static int ppa_completion(struct scsi_cmnd *cmd)
 		if (cmd->SCp.buffer && !cmd->SCp.this_residual) {
 			/* if scatter/gather, advance to the next segment */
 			if (cmd->SCp.buffers_residual--) {
-				cmd->SCp.buffer++;
+				cmd->SCp.buffer = sg_next(cmd->SCp.buffer);
 				cmd->SCp.this_residual =
 				    cmd->SCp.buffer->length;
 				cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);

drivers/scsi/scsi_lib.c

@@ -40,6 +40,18 @@
 #include "scsi_priv.h"
 #include "scsi_logging.h"
 
+/*
+ * Size of integrity metadata is usually small, 1 inline sg should
+ * cover normal cases.
+ */
+#ifdef CONFIG_ARCH_NO_SG_CHAIN
+#define  SCSI_INLINE_PROT_SG_CNT  0
+#define  SCSI_INLINE_SG_CNT  0
+#else
+#define  SCSI_INLINE_PROT_SG_CNT  1
+#define  SCSI_INLINE_SG_CNT  2
+#endif
+
 static struct kmem_cache *scsi_sdb_cache;
 static struct kmem_cache *scsi_sense_cache;
 static struct kmem_cache *scsi_sense_isadma_cache;
@@ -542,9 +554,11 @@ static void scsi_uninit_cmd(struct scsi_cmnd *cmd)
 static void scsi_mq_free_sgtables(struct scsi_cmnd *cmd)
 {
 	if (cmd->sdb.table.nents)
-		sg_free_table_chained(&cmd->sdb.table, true);
+		sg_free_table_chained(&cmd->sdb.table,
+				SCSI_INLINE_SG_CNT);
 	if (scsi_prot_sg_count(cmd))
-		sg_free_table_chained(&cmd->prot_sdb->table, true);
+		sg_free_table_chained(&cmd->prot_sdb->table,
+				SCSI_INLINE_PROT_SG_CNT);
 }
 
 static void scsi_mq_uninit_cmd(struct scsi_cmnd *cmd)
@@ -977,7 +991,8 @@ static blk_status_t scsi_init_sgtable(struct request *req,
 	 * If sg table allocation fails, requeue request later.
 	 */
 	if (unlikely(sg_alloc_table_chained(&sdb->table,
-			blk_rq_nr_phys_segments(req), sdb->table.sgl)))
+			blk_rq_nr_phys_segments(req), sdb->table.sgl,
+			SCSI_INLINE_SG_CNT)))
 		return BLK_STS_RESOURCE;
 
 	/* 
@@ -1031,7 +1046,8 @@ blk_status_t scsi_init_io(struct scsi_cmnd *cmd)
 		ivecs = blk_rq_count_integrity_sg(rq->q, rq->bio);
 
 		if (sg_alloc_table_chained(&prot_sdb->table, ivecs,
-				prot_sdb->table.sgl)) {
+				prot_sdb->table.sgl,
+				SCSI_INLINE_PROT_SG_CNT)) {
 			ret = BLK_STS_RESOURCE;
 			goto out_free_sgtables;
 		}
@@ -1542,9 +1558,9 @@ static int scsi_dispatch_cmd(struct scsi_cmnd *cmd)
 }
 
 /* Size in bytes of the sg-list stored in the scsi-mq command-private data. */
-static unsigned int scsi_mq_sgl_size(struct Scsi_Host *shost)
+static unsigned int scsi_mq_inline_sgl_size(struct Scsi_Host *shost)
 {
-	return min_t(unsigned int, shost->sg_tablesize, SG_CHUNK_SIZE) *
+	return min_t(unsigned int, shost->sg_tablesize, SCSI_INLINE_SG_CNT) *
 		sizeof(struct scatterlist);
 }
 
@@ -1726,7 +1742,7 @@ static int scsi_mq_init_request(struct blk_mq_tag_set *set, struct request *rq,
 	if (scsi_host_get_prot(shost)) {
 		sg = (void *)cmd + sizeof(struct scsi_cmnd) +
 			shost->hostt->cmd_size;
-		cmd->prot_sdb = (void *)sg + scsi_mq_sgl_size(shost);
+		cmd->prot_sdb = (void *)sg + scsi_mq_inline_sgl_size(shost);
 	}
 
 	return 0;
@@ -1820,10 +1836,11 @@ int scsi_mq_setup_tags(struct Scsi_Host *shost)
 {
 	unsigned int cmd_size, sgl_size;
 
-	sgl_size = scsi_mq_sgl_size(shost);
+	sgl_size = scsi_mq_inline_sgl_size(shost);
 	cmd_size = sizeof(struct scsi_cmnd) + shost->hostt->cmd_size + sgl_size;
 	if (scsi_host_get_prot(shost))
-		cmd_size += sizeof(struct scsi_data_buffer) + sgl_size;
+		cmd_size += sizeof(struct scsi_data_buffer) +
+			sizeof(struct scatterlist) * SCSI_INLINE_PROT_SG_CNT;
 
 	memset(&shost->tag_set, 0, sizeof(shost->tag_set));
 	shost->tag_set.ops = &scsi_mq_ops;

drivers/scsi/vmw_pvscsi.c

@@ -335,7 +335,7 @@ static void pvscsi_create_sg(struct pvscsi_ctx *ctx,
 	BUG_ON(count > PVSCSI_MAX_NUM_SG_ENTRIES_PER_SEGMENT);
 
 	sge = &ctx->sgl->sge[0];
-	for (i = 0; i < count; i++, sg++) {
+	for (i = 0; i < count; i++, sg = sg_next(sg)) {
 		sge[i].addr   = sg_dma_address(sg);
 		sge[i].length = sg_dma_len(sg);
 		sge[i].flags  = 0;

drivers/scsi/wd33c93.c

@@ -735,7 +735,7 @@ transfer_bytes(const wd33c93_regs regs, struct scsi_cmnd *cmd,
  * source or destination for THIS transfer.
  */
 	if (!cmd->SCp.this_residual && cmd->SCp.buffers_residual) {
-		++cmd->SCp.buffer;
+		cmd->SCp.buffer = sg_next(cmd->SCp.buffer);
 		--cmd->SCp.buffers_residual;
 		cmd->SCp.this_residual = cmd->SCp.buffer->length;
 		cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);

drivers/staging/unisys/visorhba/visorhba_main.c

@@ -871,12 +871,11 @@ static void do_scsi_nolinuxstat(struct uiscmdrsp *cmdrsp,
 			return;
 		}
 
-		sg = scsi_sglist(scsicmd);
+		scsi_for_each_sg(scsicmd, sg, scsi_sg_count(scsicmd), i) {
-		for (i = 0; i < scsi_sg_count(scsicmd); i++) {
+			this_page_orig = kmap_atomic(sg_page(sg));
-			this_page_orig = kmap_atomic(sg_page(sg + i));
 			this_page = (void *)((unsigned long)this_page_orig |
-					     sg[i].offset);
+					     sg->offset);
-			memcpy(this_page, buf + bufind, sg[i].length);
+			memcpy(this_page, buf + bufind, sg->length);
 			kunmap_atomic(this_page_orig);
 		}
 		kfree(buf);

drivers/usb/image/microtek.c

@@ -488,7 +488,6 @@ static void mts_command_done( struct urb *transfer )
 
 static void mts_do_sg (struct urb* transfer)
 {
-	struct scatterlist * sg;
 	int status = transfer->status;
 	MTS_INT_INIT();
 
@@ -500,13 +499,12 @@ static void mts_do_sg (struct urb* transfer)
 		mts_transfer_cleanup(transfer);
         }
 
-	sg = scsi_sglist(context->srb);
+	context->curr_sg = sg_next(context->curr_sg);
-	context->fragment++;
 	mts_int_submit_urb(transfer,
 			   context->data_pipe,
-			   sg_virt(&sg[context->fragment]),
+			   sg_virt(context->curr_sg),
-			   sg[context->fragment].length,
+			   context->curr_sg->length,
-			   context->fragment + 1 == scsi_sg_count(context->srb) ?
+			   sg_is_last(context->curr_sg) ?
 			   mts_data_done : mts_do_sg);
 }
 
@@ -526,22 +524,20 @@ static void
 mts_build_transfer_context(struct scsi_cmnd *srb, struct mts_desc* desc)
 {
 	int pipe;
-	struct scatterlist * sg;
+
-	
 	MTS_DEBUG_GOT_HERE();
 
 	desc->context.instance = desc;
 	desc->context.srb = srb;
-	desc->context.fragment = 0;
 
 	if (!scsi_bufflen(srb)) {
 		desc->context.data = NULL;
 		desc->context.data_length = 0;
 		return;
 	} else {
-		sg = scsi_sglist(srb);
+		desc->context.curr_sg = scsi_sglist(srb);
-		desc->context.data = sg_virt(&sg[0]);
+		desc->context.data = sg_virt(desc->context.curr_sg);
-		desc->context.data_length = sg[0].length;
+		desc->context.data_length = desc->context.curr_sg->length;
 	}
 
 

drivers/usb/image/microtek.h

@@ -21,7 +21,7 @@ struct mts_transfer_context
 	void *data;
 	unsigned data_length;
 	int data_pipe;
-	int fragment;
+	struct scatterlist *curr_sg;
 
 	u8 *scsi_status; /* status returned from ep_response after command completion */
 };

include/linux/scatterlist.h

@@ -266,10 +266,11 @@ int sg_split(struct scatterlist *in, const int in_mapped_nents,
 typedef struct scatterlist *(sg_alloc_fn)(unsigned int, gfp_t);
 typedef void (sg_free_fn)(struct scatterlist *, unsigned int);
 
-void __sg_free_table(struct sg_table *, unsigned int, bool, sg_free_fn *);
+void __sg_free_table(struct sg_table *, unsigned int, unsigned int,
+		     sg_free_fn *);
 void sg_free_table(struct sg_table *);
 int __sg_alloc_table(struct sg_table *, unsigned int, unsigned int,
-		     struct scatterlist *, gfp_t, sg_alloc_fn *);
+		     struct scatterlist *, unsigned int, gfp_t, sg_alloc_fn *);
 int sg_alloc_table(struct sg_table *, unsigned int, gfp_t);
 int __sg_alloc_table_from_pages(struct sg_table *sgt, struct page **pages,
 				unsigned int n_pages, unsigned int offset,
@@ -331,9 +332,11 @@ size_t sg_zero_buffer(struct scatterlist *sgl, unsigned int nents,
 #endif
 
 #ifdef CONFIG_SG_POOL
-void sg_free_table_chained(struct sg_table *table, bool first_chunk);
+void sg_free_table_chained(struct sg_table *table,
+			   unsigned nents_first_chunk);
 int sg_alloc_table_chained(struct sg_table *table, int nents,
-			   struct scatterlist *first_chunk);
+			   struct scatterlist *first_chunk,
+			   unsigned nents_first_chunk);
 #endif
 
 /*

lib/scatterlist.c

@@ -179,7 +179,8 @@ static void sg_kfree(struct scatterlist *sg, unsigned int nents)
  * __sg_free_table - Free a previously mapped sg table
  * @table:	The sg table header to use
  * @max_ents:	The maximum number of entries per single scatterlist
- * @skip_first_chunk: don't free the (preallocated) first scatterlist chunk
+ * @nents_first_chunk: Number of entries int the (preallocated) first
+ * 	scatterlist chunk, 0 means no such preallocated first chunk
  * @free_fn:	Free function
  *
  *  Description:
@@ -189,9 +190,10 @@ static void sg_kfree(struct scatterlist *sg, unsigned int nents)
  *
  **/
 void __sg_free_table(struct sg_table *table, unsigned int max_ents,
-		     bool skip_first_chunk, sg_free_fn *free_fn)
+		     unsigned int nents_first_chunk, sg_free_fn *free_fn)
 {
 	struct scatterlist *sgl, *next;
+	unsigned curr_max_ents = nents_first_chunk ?: max_ents;
 
 	if (unlikely(!table->sgl))
 		return;
@@ -207,9 +209,9 @@ void __sg_free_table(struct sg_table *table, unsigned int max_ents,
 		 * sg_size is then one less than alloc size, since the last
 		 * element is the chain pointer.
 		 */
-		if (alloc_size > max_ents) {
+		if (alloc_size > curr_max_ents) {
-			next = sg_chain_ptr(&sgl[max_ents - 1]);
+			next = sg_chain_ptr(&sgl[curr_max_ents - 1]);
-			alloc_size = max_ents;
+			alloc_size = curr_max_ents;
 			sg_size = alloc_size - 1;
 		} else {
 			sg_size = alloc_size;
@@ -217,11 +219,12 @@ void __sg_free_table(struct sg_table *table, unsigned int max_ents,
 		}
 
 		table->orig_nents -= sg_size;
-		if (skip_first_chunk)
+		if (nents_first_chunk)
-			skip_first_chunk = false;
+			nents_first_chunk = 0;
 		else
 			free_fn(sgl, alloc_size);
 		sgl = next;
+		curr_max_ents = max_ents;
 	}
 
 	table->sgl = NULL;
@@ -244,6 +247,8 @@ EXPORT_SYMBOL(sg_free_table);
  * @table:	The sg table header to use
  * @nents:	Number of entries in sg list
  * @max_ents:	The maximum number of entries the allocator returns per call
+ * @nents_first_chunk: Number of entries int the (preallocated) first
+ * 	scatterlist chunk, 0 means no such preallocated chunk provided by user
  * @gfp_mask:	GFP allocation mask
  * @alloc_fn:	Allocator to use
  *
@@ -260,10 +265,13 @@ EXPORT_SYMBOL(sg_free_table);
  **/
 int __sg_alloc_table(struct sg_table *table, unsigned int nents,
 		     unsigned int max_ents, struct scatterlist *first_chunk,
-		     gfp_t gfp_mask, sg_alloc_fn *alloc_fn)
+		     unsigned int nents_first_chunk, gfp_t gfp_mask,
+		     sg_alloc_fn *alloc_fn)
 {
 	struct scatterlist *sg, *prv;
 	unsigned int left;
+	unsigned curr_max_ents = nents_first_chunk ?: max_ents;
+	unsigned prv_max_ents;
 
 	memset(table, 0, sizeof(*table));
 
@@ -279,8 +287,8 @@ int __sg_alloc_table(struct sg_table *table, unsigned int nents,
 	do {
 		unsigned int sg_size, alloc_size = left;
 
-		if (alloc_size > max_ents) {
+		if (alloc_size > curr_max_ents) {
-			alloc_size = max_ents;
+			alloc_size = curr_max_ents;
 			sg_size = alloc_size - 1;
 		} else
 			sg_size = alloc_size;
@@ -314,7 +322,7 @@ int __sg_alloc_table(struct sg_table *table, unsigned int nents,
 		 * If this is not the first mapping, chain previous part.
 		 */
 		if (prv)
-			sg_chain(prv, max_ents, sg);
+			sg_chain(prv, prv_max_ents, sg);
 		else
 			table->sgl = sg;
 
@@ -325,6 +333,8 @@ int __sg_alloc_table(struct sg_table *table, unsigned int nents,
 			sg_mark_end(&sg[sg_size - 1]);
 
 		prv = sg;
+		prv_max_ents = curr_max_ents;
+		curr_max_ents = max_ents;
 	} while (left);
 
 	return 0;
@@ -347,9 +357,9 @@ int sg_alloc_table(struct sg_table *table, unsigned int nents, gfp_t gfp_mask)
 	int ret;
 
 	ret = __sg_alloc_table(table, nents, SG_MAX_SINGLE_ALLOC,
-			       NULL, gfp_mask, sg_kmalloc);
+			       NULL, 0, gfp_mask, sg_kmalloc);
 	if (unlikely(ret))
-		__sg_free_table(table, SG_MAX_SINGLE_ALLOC, false, sg_kfree);
+		__sg_free_table(table, SG_MAX_SINGLE_ALLOC, 0, sg_kfree);
 
 	return ret;
 }

lib/sg_pool.c

@@ -70,18 +70,27 @@ static struct scatterlist *sg_pool_alloc(unsigned int nents, gfp_t gfp_mask)
 /**
  * sg_free_table_chained - Free a previously mapped sg table
  * @table:	The sg table header to use
- * @first_chunk: was first_chunk not NULL in sg_alloc_table_chained?
+ * @nents_first_chunk: size of the first_chunk SGL passed to
+ *		sg_alloc_table_chained
  *
  *  Description:
  *    Free an sg table previously allocated and setup with
  *    sg_alloc_table_chained().
  *
+ *    @nents_first_chunk has to be same with that same parameter passed
+ *    to sg_alloc_table_chained().
+ *
  **/
-void sg_free_table_chained(struct sg_table *table, bool first_chunk)
+void sg_free_table_chained(struct sg_table *table,
+		unsigned nents_first_chunk)
 {
-	if (first_chunk && table->orig_nents <= SG_CHUNK_SIZE)
+	if (table->orig_nents <= nents_first_chunk)
 		return;
-	__sg_free_table(table, SG_CHUNK_SIZE, first_chunk, sg_pool_free);
+
+	if (nents_first_chunk == 1)
+		nents_first_chunk = 0;
+
+	__sg_free_table(table, SG_CHUNK_SIZE, nents_first_chunk, sg_pool_free);
 }
 EXPORT_SYMBOL_GPL(sg_free_table_chained);
 
@@ -90,31 +99,41 @@ EXPORT_SYMBOL_GPL(sg_free_table_chained);
  * @table:	The sg table header to use
  * @nents:	Number of entries in sg list
  * @first_chunk: first SGL
+ * @nents_first_chunk: number of the SGL of @first_chunk
  *
  *  Description:
  *    Allocate and chain SGLs in an sg table. If @nents@ is larger than
- *    SG_CHUNK_SIZE a chained sg table will be setup.
+ *    @nents_first_chunk a chained sg table will be setup. @first_chunk is
+ *    ignored if nents_first_chunk <= 1 because user expects the SGL points
+ *    non-chain SGL.
  *
  **/
 int sg_alloc_table_chained(struct sg_table *table, int nents,
-		struct scatterlist *first_chunk)
+		struct scatterlist *first_chunk, unsigned nents_first_chunk)
 {
 	int ret;
 
 	BUG_ON(!nents);
 
-	if (first_chunk) {
+	if (first_chunk && nents_first_chunk) {
-		if (nents <= SG_CHUNK_SIZE) {
+		if (nents <= nents_first_chunk) {
 			table->nents = table->orig_nents = nents;
 			sg_init_table(table->sgl, nents);
 			return 0;
 		}
 	}
 
+	/* User supposes that the 1st SGL includes real entry */
+	if (nents_first_chunk <= 1) {
+		first_chunk = NULL;
+		nents_first_chunk = 0;
+	}
+
 	ret = __sg_alloc_table(table, nents, SG_CHUNK_SIZE,
-			       first_chunk, GFP_ATOMIC, sg_pool_alloc);
+			       first_chunk, nents_first_chunk,
+			       GFP_ATOMIC, sg_pool_alloc);
 	if (unlikely(ret))
-		sg_free_table_chained(table, (bool)first_chunk);
+		sg_free_table_chained(table, nents_first_chunk);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(sg_alloc_table_chained);

net/sunrpc/xprtrdma/svc_rdma_rw.c

@@ -73,7 +73,8 @@ svc_rdma_get_rw_ctxt(struct svcxprt_rdma *rdma, unsigned int sges)
 
 	ctxt->rw_sg_table.sgl = ctxt->rw_first_sgl;
 	if (sg_alloc_table_chained(&ctxt->rw_sg_table, sges,
-				   ctxt->rw_sg_table.sgl)) {
+				   ctxt->rw_sg_table.sgl,
+				   SG_CHUNK_SIZE)) {
 		kfree(ctxt);
 		ctxt = NULL;
 	}
@@ -84,7 +85,7 @@ out:
 static void svc_rdma_put_rw_ctxt(struct svcxprt_rdma *rdma,
 				 struct svc_rdma_rw_ctxt *ctxt)
 {
-	sg_free_table_chained(&ctxt->rw_sg_table, true);
+	sg_free_table_chained(&ctxt->rw_sg_table, SG_CHUNK_SIZE);
 
 	spin_lock(&rdma->sc_rw_ctxt_lock);
 	list_add(&ctxt->rw_list, &rdma->sc_rw_ctxts);