Sophie

Sophie

distrib > Scientific%20Linux > 5x > x86_64 > by-pkgid > 27922b4260f65d317aabda37e42bbbff > files > 1569

kernel-2.6.18-238.el5.src.rpm

From: Mike Christie <mchristi@redhat.com>
Date: Tue, 1 Dec 2009 19:43:54 -0500
Subject: [infiniband] fix iser sg aligment handling
Message-id: <1259696634-12224-1-git-send-email-mchristi@redhat.com>
Patchwork-id: 21595
O-Subject: [PATCH RHEL 5.5] Fix iser sg aligment handling
Bugzilla: 540686
RH-Acked-by: Doug Ledford <dledford@redhat.com>

From: Mike Christie <mchristi@redhat.com>

This is for BZ: 540686

>From the upstream commit:

Rewrite SG handling for RDMA logic

After dma-mapping an SG list provided by the SCSI midlayer, iser has
to make sure the mapped SG is "aligned for RDMA" in the sense that its
possible to produce one mapping in the HCA IOMMU which represents the
whole SG. Next, the mapped SG is formatted for registration with the
HCA.

This patch re-writes the logic that does the above, to make it clearer
and simpler. It also fixes a bug in the being aligned for RDMA checks,
where a "start" check wasn't done but rather only "end" check.

The git commit is here:
http://git.kernel.org/?p=linux/kernel/git/roland/infiniband.git;a=patch;h=c1ccaf2478f84c2665cf57f981db143aa582d646;hp=91d3f9bacdb4950d2f79fe2ba296aa249f60d06c

This is also a potential data corruptor. From the iser maintainer:

Under heavy load, without the patch, the HCA can be programmed to write
(corrupt) into pages/location which doesn't belong to the SG associated
with
the actual I/O and cause a kernel oops. It took us some time to get into
reproducing the problem and fixing it, basically if one runs fio with 64
threads each submitting 64 requests of 2K each using AIO you hit that
always,
e.g (replace sdXXX with the underline actual device)

fio --rw=read --bs=2k --numjobs=64 --iodepth=64 --sync=0 --direct=1
--randrepeat=0 \
--ioengine=libaio --filename=/dev/sdXXX --name=test --loops=10000
--size=103809024 \
--runtime=600 --group_reporting

I have only compile tested the patch. The iser guys from Voltaire build
and tested the patch and sent it upstream.

diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c
index b9453d0..274c883 100644
--- a/drivers/infiniband/ulp/iser/iser_memory.c
+++ b/drivers/infiniband/ulp/iser/iser_memory.c
@@ -209,6 +209,8 @@ void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *iser_task,
 	mem_copy->copy_buf = NULL;
 }
 
+#define IS_4K_ALIGNED(addr)	((((unsigned long)addr) & ~MASK_4K) == 0)
+
 /**
  * iser_sg_to_page_vec - Translates scatterlist entries to physical addresses
  * and returns the length of resulting physical address array (may be less than
@@ -221,62 +223,52 @@ void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *iser_task,
  * where --few fragments of the same page-- are present in the SG as
  * consecutive elements. Also, it handles one entry SG.
  */
+
 static int iser_sg_to_page_vec(struct iser_data_buf *data,
 			       struct iser_page_vec *page_vec,
 			       struct ib_device *ibdev)
 {
-	struct scatterlist *sgl = (struct scatterlist *)data->buf;
-	struct scatterlist *sg;
-	u64 first_addr, last_addr, page;
-	int end_aligned;
-	unsigned int cur_page = 0;
+	struct scatterlist *sg, *sgl = (struct scatterlist *)data->buf;
+	u64 start_addr, end_addr, page, chunk_start = 0;
 	unsigned long total_sz = 0;
-	int i;
+	unsigned int dma_len;
+	int i, new_chunk, cur_page, last_ent = data->dma_nents - 1;
 
 	/* compute the offset of first element */
 	page_vec->offset = (u64) sgl[0].offset & ~MASK_4K;
 
+	new_chunk = 1;
+	cur_page  = 0;
 	for_each_sg(sgl, sg, data->dma_nents, i) {
-		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
-
+		start_addr = ib_sg_dma_address(ibdev, sg);
+		if (new_chunk)
+			chunk_start = start_addr;
+		dma_len = ib_sg_dma_len(ibdev, sg);
+		end_addr = start_addr + dma_len;
 		total_sz += dma_len;
 
-		first_addr = ib_sg_dma_address(ibdev, sg);
-		last_addr  = first_addr + dma_len;
-
-		end_aligned   = !(last_addr  & ~MASK_4K);
-
-		/* continue to collect page fragments till aligned or SG ends */
-		while (!end_aligned && (i + 1 < data->dma_nents)) {
-			sg = sg_next(sg);
-			i++;
-			dma_len = ib_sg_dma_len(ibdev, sg);
-			total_sz += dma_len;
-			last_addr = ib_sg_dma_address(ibdev, sg) + dma_len;
-			end_aligned = !(last_addr  & ~MASK_4K);
+		/* collect page fragments until aligned or end of SG list */
+		if (!IS_4K_ALIGNED(end_addr) && i < last_ent) {
+			new_chunk = 0;
+			continue;
 		}
-
-		/* handle the 1st page in the 1st DMA element */
-		if (cur_page == 0) {
-			page = first_addr & MASK_4K;
-			page_vec->pages[cur_page] = page;
-			cur_page++;
+		new_chunk = 1;
+
+		/* address of the first page in the contiguous chunk;
+		   masking relevant for the very first SG entry,
+		   which might be unaligned */
+		page = chunk_start & MASK_4K;
+		do {
+			page_vec->pages[cur_page++] = page;
 			page += SIZE_4K;
-		} else
-			page = first_addr;
-
-		for (; page < last_addr; page += SIZE_4K) {
-			page_vec->pages[cur_page] = page;
-			cur_page++;
-		}
-
+		} while (page < end_addr);
 	}
+
 	page_vec->data_size = total_sz;
 	iser_dbg("page_vec->data_size:%d cur_page %d\n", page_vec->data_size,cur_page);
 	return cur_page;
 }
 
-#define IS_4K_ALIGNED(addr)	((((unsigned long)addr) & ~MASK_4K) == 0)
 
 /**
  * iser_data_buf_aligned_len - Tries to determine the maximal correctly aligned
@@ -284,42 +276,40 @@ static int iser_sg_to_page_vec(struct iser_data_buf *data,
  * the number of entries which are aligned correctly. Supports the case where
  * consecutive SG elements are actually fragments of the same physcial page.
  */
-static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data,
-					      struct ib_device *ibdev)
+static int iser_data_buf_aligned_len(struct iser_data_buf *data,
+				      struct ib_device *ibdev)
 {
-	struct scatterlist *sgl, *sg;
-	u64 end_addr, next_addr;
-	int i, cnt;
-	unsigned int ret_len = 0;
+	struct scatterlist *sgl, *sg, *next_sg = NULL;
+	u64 start_addr, end_addr;
+	int i, ret_len, start_check = 0;
+
+	if (data->dma_nents == 1)
+		return 1;
 
 	sgl = (struct scatterlist *)data->buf;
+	start_addr  = ib_sg_dma_address(ibdev, sgl);
 
-	cnt = 0;
 	for_each_sg(sgl, sg, data->dma_nents, i) {
-		/* iser_dbg("Checking sg iobuf [%d]: phys=0x%08lX "
-		   "offset: %ld sz: %ld\n", i,
-		   (unsigned long)sg_phys(sg),
-		   (unsigned long)sg->offset,
-		   (unsigned long)sg->length); */
-		end_addr = ib_sg_dma_address(ibdev, sg) +
-			   ib_sg_dma_len(ibdev, sg);
-		/* iser_dbg("Checking sg iobuf end address "
-		       "0x%08lX\n", end_addr); */
-		if (i + 1 < data->dma_nents) {
-			next_addr = ib_sg_dma_address(ibdev, sg_next(sg));
-			/* are i, i+1 fragments of the same page? */
-			if (end_addr == next_addr) {
-				cnt++;
-				continue;
-			} else if (!IS_4K_ALIGNED(end_addr)) {
-				ret_len = cnt + 1;
-				break;
-			}
-		}
-		cnt++;
+		if (start_check && !IS_4K_ALIGNED(start_addr))
+			break;
+
+		next_sg = sg_next(sg);
+		if (!next_sg)
+			break;
+
+		end_addr    = start_addr + ib_sg_dma_len(ibdev, sg);
+		start_addr  = ib_sg_dma_address(ibdev, next_sg);
+
+		if (end_addr == start_addr) {
+			start_check = 0;
+			continue;
+		} else
+			start_check = 1;
+
+		if (!IS_4K_ALIGNED(end_addr))
+			break;
 	}
-	if (i == data->dma_nents)
-		ret_len = cnt;	/* loop ended */
+	ret_len = (next_sg) ? i : i+1;
 	iser_dbg("Found %d aligned entries out of %d in sg:0x%p\n",
 		 ret_len, data->dma_nents, data);
 	return ret_len;