Merge branch 'for-4.3/sg' of git://git.kernel.dk/linux-block

Pull SG updates from Jens Axboe:
"This contains a set of scatter-gather related changes/fixes for 4.3:

- Add support for limited chaining of sg tables even for
architectures that do not set ARCH_HAS_SG_CHAIN. From Christoph.

- Add sg chain support to target_rd. From Christoph.

- Fixup open coded sg->page_link in crypto/omap-sham. From
Christoph.

- Fixup open coded crypto ->page_link manipulation. From Dan.

- Also from Dan, automated fixup of manual sg_unmark_end()
manipulations.

- Also from Dan, automated fixup of open coded sg_phys()
implementations.

- From Robert Jarzmik, addition of an sg table splitting helper that
drivers can use"

* 'for-4.3/sg' of git://git.kernel.dk/linux-block:
lib: scatterlist: add sg splitting function
scatterlist: use sg_phys()
crypto/omap-sham: remove an open coded access to ->page_link
scatterlist: remove open coded sg_unmark_end instances
crypto: replace scatterwalk_sg_chain with sg_chain
target/rd: always chain S/G list
scatterlist: allow limited chaining without ARCH_HAS_SG_CHAIN

arch/arm/mm/dma-mapping.c

···
 		return -ENOMEM;
 
 	for (count = 0, s = sg; count < (size >> PAGE_SHIFT); s = sg_next(s)) {
-		phys_addr_t phys = page_to_phys(sg_page(s));
+		phys_addr_t phys = sg_phys(s) & PAGE_MASK;
 		unsigned int len = PAGE_ALIGN(s->offset + s->length);
 
 		if (!is_coherent &&

arch/microblaze/kernel/dma.c

···
 	/* FIXME this part of code is untested */
 	for_each_sg(sgl, sg, nents, i) {
 		sg->dma_address = sg_phys(sg);
-		__dma_sync(page_to_phys(sg_page(sg)) + sg->offset,
-							sg->length, direction);
+		__dma_sync(sg_phys(sg), sg->length, direction);
 	}
 
 	return nents;

block/blk-merge.c

···
 		if (rq->cmd_flags & REQ_WRITE)
 			memset(q->dma_drain_buffer, 0, q->dma_drain_size);
 
-		sg->page_link &= ~0x02;
+		sg_unmark_end(sg);
 		sg = sg_next(sg);
 		sg_set_page(sg, virt_to_page(q->dma_drain_buffer),
 			    q->dma_drain_size,

crypto/algif_skcipher.c

···
 		sgl->cur = 0;
 
 		if (sg)
-			scatterwalk_sg_chain(sg, MAX_SGL_ENTS + 1, sgl->sg);
+			sg_chain(sg, MAX_SGL_ENTS + 1, sgl->sg);
 
 		list_add_tail(&sgl->list, &ctx->tsgl);
 	}

crypto/gcm.c

···
 	sg_set_buf(pctx->src, pctx->auth_tag, sizeof(pctx->auth_tag));
 	sg = scatterwalk_ffwd(pctx->src + 1, req->src, req->assoclen);
 	if (sg != pctx->src + 1)
-		scatterwalk_sg_chain(pctx->src, 2, sg);
+		sg_chain(pctx->src, 2, sg);
 
 	if (req->src != req->dst) {
 		sg_init_table(pctx->dst, 3);
 		sg_set_buf(pctx->dst, pctx->auth_tag, sizeof(pctx->auth_tag));
 		sg = scatterwalk_ffwd(pctx->dst + 1, req->dst, req->assoclen);
 		if (sg != pctx->dst + 1)
-			scatterwalk_sg_chain(pctx->dst, 2, sg);
+			sg_chain(pctx->dst, 2, sg);
 	}
 }
 

drivers/crypto/bfin_crc.c

···
 			sg_init_table(ctx->bufsl, nsg);
 			sg_set_buf(ctx->bufsl, ctx->buflast, ctx->buflast_len);
 			if (nsg > 1)
-				scatterwalk_sg_chain(ctx->bufsl, nsg,
-						req->src);
+				sg_chain(ctx->bufsl, nsg, req->src);
 			ctx->sg = ctx->bufsl;
 		} else
 			ctx->sg = req->src;

drivers/crypto/omap-sham.c

···
 		 * the dmaengine may try to DMA the incorrect amount of data.
 		 */
 		sg_init_table(&ctx->sgl, 1);
-		ctx->sgl.page_link = ctx->sg->page_link;
+		sg_assign_page(&ctx->sgl, sg_page(ctx->sg));
 		ctx->sgl.offset = ctx->sg->offset;
 		sg_dma_len(&ctx->sgl) = len32;
 		sg_dma_address(&ctx->sgl) = sg_dma_address(ctx->sg);

drivers/crypto/qce/sha.c

···
 	if (rctx->buflen) {
 		sg_init_table(rctx->sg, 2);
 		sg_set_buf(rctx->sg, rctx->tmpbuf, rctx->buflen);
-		scatterwalk_sg_chain(rctx->sg, 2, req->src);
+		sg_chain(rctx->sg, 2, req->src);
 		req->src = rctx->sg;
 	}
 

drivers/crypto/sahara.c

···
 		sg_init_table(rctx->in_sg_chain, 2);
 		sg_set_buf(rctx->in_sg_chain, rctx->rembuf, rctx->buf_cnt);
 
-		scatterwalk_sg_chain(rctx->in_sg_chain, 2, req->src);
+		sg_chain(rctx->in_sg_chain, 2, req->src);
 
 		rctx->total = req->nbytes + rctx->buf_cnt;
 		rctx->in_sg = rctx->in_sg_chain;

drivers/crypto/talitos.c

···
 		sg_init_table(req_ctx->bufsl, nsg);
 		sg_set_buf(req_ctx->bufsl, req_ctx->buf, req_ctx->nbuf);
 		if (nsg > 1)
-			scatterwalk_sg_chain(req_ctx->bufsl, 2, areq->src);
+			sg_chain(req_ctx->bufsl, 2, areq->src);
 		req_ctx->psrc = req_ctx->bufsl;
 	} else
 		req_ctx->psrc = areq->src;

drivers/iommu/intel-iommu.c

···
 			sg_res = aligned_nrpages(sg->offset, sg->length);
 			sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
 			sg->dma_length = sg->length;
-			pteval = page_to_phys(sg_page(sg)) | prot;
+			pteval = (sg_phys(sg) & PAGE_MASK) | prot;
 			phys_pfn = pteval >> VTD_PAGE_SHIFT;
 		}
 
···
 
 	for_each_sg(sglist, sg, nelems, i) {
 		BUG_ON(!sg_page(sg));
-		sg->dma_address = page_to_phys(sg_page(sg)) + sg->offset;
+		sg->dma_address = sg_phys(sg);
 		sg->dma_length = sg->length;
 	}
 	return nelems;

drivers/iommu/iommu.c

···
 	min_pagesz = 1 << __ffs(domain->ops->pgsize_bitmap);
 
 	for_each_sg(sg, s, nents, i) {
-		phys_addr_t phys = page_to_phys(sg_page(s)) + s->offset;
+		phys_addr_t phys = sg_phys(s);
 
 		/*
 		 * We are mapping on IOMMU page boundaries, so offset within

drivers/mmc/card/queue.c

···
 			sg_set_buf(__sg, buf + offset, len);
 			offset += len;
 			remain -= len;
-			(__sg++)->page_link &= ~0x02;
+			sg_unmark_end(__sg++);
 			sg_len++;
 		} while (remain);
 	}
···
 	list_for_each_entry(req, &packed->list, queuelist) {
 		sg_len += blk_rq_map_sg(mq->queue, req, __sg);
 		__sg = sg + (sg_len - 1);
-		(__sg++)->page_link &= ~0x02;
+		sg_unmark_end(__sg++);
 	}
 	sg_mark_end(sg + (sg_len - 1));
 	return sg_len;

drivers/staging/android/ion/ion_chunk_heap.c

···
 err:
 	sg = table->sgl;
 	for (i -= 1; i >= 0; i--) {
-		gen_pool_free(chunk_heap->pool, page_to_phys(sg_page(sg)),
+		gen_pool_free(chunk_heap->pool, sg_phys(sg) & PAGE_MASK,
 			      sg->length);
 		sg = sg_next(sg);
 	}
···
 							DMA_BIDIRECTIONAL);
 
 	for_each_sg(table->sgl, sg, table->nents, i) {
-		gen_pool_free(chunk_heap->pool, page_to_phys(sg_page(sg)),
+		gen_pool_free(chunk_heap->pool, sg_phys(sg) & PAGE_MASK,
 			      sg->length);
 	}
 	chunk_heap->allocated -= allocated_size;

drivers/target/target_core_rd.c

···
 		sg_per_table = (total_sg_needed > max_sg_per_table) ?
 			max_sg_per_table : total_sg_needed;
 
-#ifdef CONFIG_ARCH_HAS_SG_CHAIN
-
 		/*
 		 * Reserve extra element for chain entry
 		 */
 		if (sg_per_table < total_sg_needed)
 			chain_entry = 1;
-
-#endif /* CONFIG_ARCH_HAS_SG_CHAIN */
 
 		sg = kcalloc(sg_per_table + chain_entry, sizeof(*sg),
 				GFP_KERNEL);
···
 
 		sg_init_table(sg, sg_per_table + chain_entry);
 
-#ifdef CONFIG_ARCH_HAS_SG_CHAIN
-
 		if (i > 0) {
 			sg_chain(sg_table[i - 1].sg_table,
 				 max_sg_per_table + 1, sg);
 		}
-
-#endif /* CONFIG_ARCH_HAS_SG_CHAIN */
 
 		sg_table[i].sg_table = sg;
 		sg_table[i].rd_sg_count = sg_per_table;
···
 
 	prot_sg = &prot_table->sg_table[prot_page -
 					prot_table->page_start_offset];
-
-#ifndef CONFIG_ARCH_HAS_SG_CHAIN
-
-	prot_npages = DIV_ROUND_UP(prot_offset + sectors * se_dev->prot_length,
-				   PAGE_SIZE);
-
-	/*
-	 * Allocate temporaly contiguous scatterlist entries if prot pages
-	 * straddles multiple scatterlist tables.
-	 */
-	if (prot_table->page_end_offset < prot_page + prot_npages - 1) {
-		int i;
-
-		prot_sg = kcalloc(prot_npages, sizeof(*prot_sg), GFP_KERNEL);
-		if (!prot_sg)
-			return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
-
-		need_to_release = true;
-		sg_init_table(prot_sg, prot_npages);
-
-		for (i = 0; i < prot_npages; i++) {
-			if (prot_page + i > prot_table->page_end_offset) {
-				prot_table = rd_get_prot_table(dev,
-								prot_page + i);
-				if (!prot_table) {
-					kfree(prot_sg);
-					return rc;
-				}
-				sg_unmark_end(&prot_sg[i - 1]);
-			}
-			prot_sg[i] = prot_table->sg_table[prot_page + i -
-						prot_table->page_start_offset];
-		}
-	}
-
-#endif /* !CONFIG_ARCH_HAS_SG_CHAIN */
 
 	if (is_read)
 		rc = sbc_dif_verify(cmd, cmd->t_task_lba, sectors, 0,

include/crypto/scatterwalk.h

···
 #include <linux/scatterlist.h>
 #include <linux/sched.h>
 
-static inline void scatterwalk_sg_chain(struct scatterlist *sg1, int num,
-					struct scatterlist *sg2)
-{
-	sg_set_page(&sg1[num - 1], (void *)sg2, 0, 0);
-	sg1[num - 1].page_link &= ~0x02;
-	sg1[num - 1].page_link |= 0x01;
-}
-
 static inline void scatterwalk_crypto_chain(struct scatterlist *head,
 					    struct scatterlist *sg,
 					    int chain, int num)
···
 	}
 
 	if (sg)
-		scatterwalk_sg_chain(head, num, sg);
+		sg_chain(head, num, sg);
 	else
 		sg_mark_end(head);
 }

include/linux/scatterlist.h

···
 static inline void sg_chain(struct scatterlist *prv, unsigned int prv_nents,
 			    struct scatterlist *sgl)
 {
-#ifndef CONFIG_ARCH_HAS_SG_CHAIN
-	BUG();
-#endif
-
 	/*
 	 * offset and length are unused for chain entry.  Clear them.
 	 */
···
 struct scatterlist *sg_last(struct scatterlist *s, unsigned int);
 void sg_init_table(struct scatterlist *, unsigned int);
 void sg_init_one(struct scatterlist *, const void *, unsigned int);
+int sg_split(struct scatterlist *in, const int in_mapped_nents,
+	     const off_t skip, const int nb_splits,
+	     const size_t *split_sizes,
+	     struct scatterlist **out, int *out_mapped_nents,
+	     gfp_t gfp_mask);
 
 typedef struct scatterlist *(sg_alloc_fn)(unsigned int, gfp_t);
 typedef void (sg_free_fn)(struct scatterlist *, unsigned int);

lib/Kconfig

···
 
 source "lib/fonts/Kconfig"
 
+config SG_SPLIT
+	def_bool n
+	help
+	 Provides a heler to split scatterlists into chunks, each chunk being a
+	 scatterlist. This should be selected by a driver or an API which
+	 whishes to split a scatterlist amongst multiple DMA channel.
+
 #
 # sg chaining option
 #

lib/Makefile

···
 
 obj-$(CONFIG_GENERIC_NET_UTILS) += net_utils.o
 
+obj-$(CONFIG_SG_SPLIT) += sg_split.o
 obj-$(CONFIG_STMP_DEVICE) += stmp_device.o
 
 libfdt_files = fdt.o fdt_ro.o fdt_wip.o fdt_rw.o fdt_sw.o fdt_strerror.o \

lib/scatterlist.c

···
  **/
 struct scatterlist *sg_last(struct scatterlist *sgl, unsigned int nents)
 {
-#ifndef CONFIG_ARCH_HAS_SG_CHAIN
-	struct scatterlist *ret = &sgl[nents - 1];
-#else
 	struct scatterlist *sg, *ret = NULL;
 	unsigned int i;
 
 	for_each_sg(sgl, sg, nents, i)
 		ret = sg;
 
-#endif
 #ifdef CONFIG_DEBUG_SG
 	BUG_ON(sgl[0].sg_magic != SG_MAGIC);
 	BUG_ON(!sg_is_last(ret));

lib/sg_split.c

···
+/*
+ * Copyright (C) 2015 Robert Jarzmik <[e-mail]>
+ *
+ * Scatterlist splitting helpers.
+ *
+ * This source code is licensed under the GNU General Public License,
+ * Version 2. See the file COPYING for more details.
+ */
+
+#include <linux/scatterlist.h>
+#include <linux/slab.h>
+
+struct sg_splitter {
+	struct scatterlist *in_sg0;
+	int nents;
+	off_t skip_sg0;
+	unsigned int length_last_sg;
+
+	struct scatterlist *out_sg;
+};
+
+static int sg_calculate_split(struct scatterlist *in, int nents, int nb_splits,
+			      off_t skip, const size_t *sizes,
+			      struct sg_splitter *splitters, bool mapped)
+{
+	int i;
+	unsigned int sglen;
+	size_t size = sizes[0], len;
+	struct sg_splitter *curr = splitters;
+	struct scatterlist *sg;
+
+	for (i = 0; i < nb_splits; i++) {
+		splitters[i].in_sg0 = NULL;
+		splitters[i].nents = 0;
+	}
+
+	for_each_sg(in, sg, nents, i) {
+		sglen = mapped ? sg_dma_len(sg) : sg->length;
+		if (skip > sglen) {
+			skip -= sglen;
+			continue;
+		}
+
+		len = min_t(size_t, size, sglen - skip);
+		if (!curr->in_sg0) {
+			curr->in_sg0 = sg;
+			curr->skip_sg0 = skip;
+		}
+		size -= len;
+		curr->nents++;
+		curr->length_last_sg = len;
+
+		while (!size && (skip + len < sglen) && (--nb_splits > 0)) {
+			curr++;
+			size = *(++sizes);
+			skip += len;
+			len = min_t(size_t, size, sglen - skip);
+
+			curr->in_sg0 = sg;
+			curr->skip_sg0 = skip;
+			curr->nents = 1;
+			curr->length_last_sg = len;
+			size -= len;
+		}
+		skip = 0;
+
+		if (!size && --nb_splits > 0) {
+			curr++;
+			size = *(++sizes);
+		}
+
+		if (!nb_splits)
+			break;
+	}
+
+	return (size || !splitters[0].in_sg0) ? -EINVAL : 0;
+}
+
+static void sg_split_phys(struct sg_splitter *splitters, const int nb_splits)
+{
+	int i, j;
+	struct scatterlist *in_sg, *out_sg;
+	struct sg_splitter *split;
+
+	for (i = 0, split = splitters; i < nb_splits; i++, split++) {
+		in_sg = split->in_sg0;
+		out_sg = split->out_sg;
+		for (j = 0; j < split->nents; j++, out_sg++) {
+			*out_sg = *in_sg;
+			if (!j) {
+				out_sg->offset += split->skip_sg0;
+				out_sg->length -= split->skip_sg0;
+			} else {
+				out_sg->offset = 0;
+			}
+			sg_dma_address(out_sg) = 0;
+			sg_dma_len(out_sg) = 0;
+			in_sg = sg_next(in_sg);
+		}
+		out_sg[-1].length = split->length_last_sg;
+		sg_mark_end(out_sg - 1);
+	}
+}
+
+static void sg_split_mapped(struct sg_splitter *splitters, const int nb_splits)
+{
+	int i, j;
+	struct scatterlist *in_sg, *out_sg;
+	struct sg_splitter *split;
+
+	for (i = 0, split = splitters; i < nb_splits; i++, split++) {
+		in_sg = split->in_sg0;
+		out_sg = split->out_sg;
+		for (j = 0; j < split->nents; j++, out_sg++) {
+			sg_dma_address(out_sg) = sg_dma_address(in_sg);
+			sg_dma_len(out_sg) = sg_dma_len(in_sg);
+			if (!j) {
+				sg_dma_address(out_sg) += split->skip_sg0;
+				sg_dma_len(out_sg) -= split->skip_sg0;
+			}
+			in_sg = sg_next(in_sg);
+		}
+		sg_dma_len(--out_sg) = split->length_last_sg;
+	}
+}
+
+/**
+ * sg_split - split a scatterlist into several scatterlists
+ * @in: the input sg list
+ * @in_mapped_nents: the result of a dma_map_sg(in, ...), or 0 if not mapped.
+ * @skip: the number of bytes to skip in the input sg list
+ * @nb_splits: the number of desired sg outputs
+ * @split_sizes: the respective size of each output sg list in bytes
+ * @out: an array where to store the allocated output sg lists
+ * @out_mapped_nents: the resulting sg lists mapped number of sg entries. Might
+ *                    be NULL if sglist not already mapped (in_mapped_nents = 0)
+ * @gfp_mask: the allocation flag
+ *
+ * This function splits the input sg list into nb_splits sg lists, which are
+ * allocated and stored into out.
+ * The @in is split into :
+ *  - @out[0], which covers bytes [@skip .. @skip + @split_sizes[0] - 1] of @in
+ *  - @out[1], which covers bytes [@skip + split_sizes[0] ..
+ *                                 @skip + @split_sizes[0] + @split_sizes[1] -1]
+ * etc ...
+ * It will be the caller's duty to kfree() out array members.
+ *
+ * Returns 0 upon success, or error code
+ */
+int sg_split(struct scatterlist *in, const int in_mapped_nents,
+	     const off_t skip, const int nb_splits,
+	     const size_t *split_sizes,
+	     struct scatterlist **out, int *out_mapped_nents,
+	     gfp_t gfp_mask)
+{
+	int i, ret;
+	struct sg_splitter *splitters;
+
+	splitters = kcalloc(nb_splits, sizeof(*splitters), gfp_mask);
+	if (!splitters)
+		return -ENOMEM;
+
+	ret = sg_calculate_split(in, sg_nents(in), nb_splits, skip, split_sizes,
+			   splitters, false);
+	if (ret < 0)
+		goto err;
+
+	ret = -ENOMEM;
+	for (i = 0; i < nb_splits; i++) {
+		splitters[i].out_sg = kmalloc_array(splitters[i].nents,
+						    sizeof(struct scatterlist),
+						    gfp_mask);
+		if (!splitters[i].out_sg)
+			goto err;
+	}
+
+	/*
+	 * The order of these 3 calls is important and should be kept.
+	 */
+	sg_split_phys(splitters, nb_splits);
+	ret = sg_calculate_split(in, in_mapped_nents, nb_splits, skip,
+				 split_sizes, splitters, true);
+	if (ret < 0)
+		goto err;
+	sg_split_mapped(splitters, nb_splits);
+
+	for (i = 0; i < nb_splits; i++) {
+		out[i] = splitters[i].out_sg;
+		if (out_mapped_nents)
+			out_mapped_nents[i] = splitters[i].nents;
+	}
+
+	kfree(splitters);
+	return 0;
+
+err:
+	for (i = 0; i < nb_splits; i++)
+		kfree(splitters[i].out_sg);
+	kfree(splitters);
+	return ret;
+}
+EXPORT_SYMBOL(sg_split);