Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull rdma updates from Doug Ledford:
 "This is a fairly plain pull request. Lots of driver updates across the
  stack, a huge number of static analysis cleanups including a close to
  50 patch series from Bart Van Assche, and a number of new features
  inside the stack such as general CQ moderation support.

  Nothing really stands out, but there might be a few conflicts as you
  take things in. In particular, the cleanups touched some of the same
  lines as the new timer_setup changes.

  Everything in this pull request has been through 0day and at least two
  days of linux-next (since Stephen doesn't necessarily flag new
  errors/warnings until day2). A few more items (about 30 patches) from
  Intel and Mellanox showed up on the list on Tuesday. I've excluded
  those from this pull request, and I'm sure some of them qualify as
  fixes suitable to send any time, but I still have to review them
  fully. If they contain mostly fixes and little or no new development,
  then I will probably send them through by the end of the week just to
  get them out of the way.

  There was a break in my acceptance of patches which coincides with the
  computer problems I had, and then when I got things mostly back under
  control I had a backlog of patches to process, which I did mostly last
  Friday and Monday. So there is a larger number of patches processed in
  that timeframe than I was striving for.

  Summary:
   - Add iWARP support to qedr driver
   - Lots of misc fixes across subsystem
   - Multiple update series to hns roce driver
   - Multiple update series to hfi1 driver
   - Updates to vnic driver
   - Add kref to wait struct in cxgb4 driver
   - Updates to i40iw driver
   - Mellanox shared pull request
   - timer_setup changes
   - massive cleanup series from Bart Van Assche
   - Two series of SRP/SRPT changes from Bart Van Assche
   - Core updates from Mellanox
   - i40iw updates
   - IPoIB updates
   - mlx5 updates
   - mlx4 updates
   - hns updates
   - bnxt_re fixes
   - PCI write padding support
   - Sparse/Smatch/warning cleanups/fixes
   - CQ moderation support
   - SRQ support in vmw_pvrdma"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (296 commits)
  RDMA/core: Rename kernel modify_cq to better describe its usage
  IB/mlx5: Add CQ moderation capability to query_device
  IB/mlx4: Add CQ moderation capability to query_device
  IB/uverbs: Add CQ moderation capability to query_device
  IB/mlx5: Exposing modify CQ callback to uverbs layer
  IB/mlx4: Exposing modify CQ callback to uverbs layer
  IB/uverbs: Allow CQ moderation with modify CQ
  iw_cxgb4: atomically flush the qp
  iw_cxgb4: only call the cq comp_handler when the cq is armed
  iw_cxgb4: Fix possible circular dependency locking warning
  RDMA/bnxt_re: report vlan_id and sl in qp1 recv completion
  IB/core: Only maintain real QPs in the security lists
  IB/ocrdma_hw: remove unnecessary code in ocrdma_mbx_dealloc_lkey
  RDMA/core: Make function rdma_copy_addr return void
  RDMA/vmw_pvrdma: Add shared receive queue support
  RDMA/core: avoid uninitialized variable warning in create_udata
  RDMA/bnxt_re: synchronize poll_cq and req_notify_cq verbs
  RDMA/bnxt_re: Flush CQ notification Work Queue before destroying QP
  RDMA/bnxt_re: Set QP state in case of response completion errors
  RDMA/bnxt_re: Add memory barriers when processing CQ/EQ entries
  ...

1 files changed, 260 insertions, 24 deletions

diff --git a/drivers/infiniband/hw/mlx4/mr.c b/drivers/infiniband/hw/mlx4/mr.c
index e6f77f63da75..313bfb9ccb71 100644
--- a/drivers/infiniband/hw/mlx4/mr.c
+++ b/drivers/infiniband/hw/mlx4/mr.c
@@ -87,50 +87,286 @@ err_free:
 	return ERR_PTR(err);
 }
 
+enum {
+	MLX4_MAX_MTT_SHIFT = 31
+};
+
+static int mlx4_ib_umem_write_mtt_block(struct mlx4_ib_dev *dev,
+					struct mlx4_mtt *mtt,
+					u64 mtt_size, u64 mtt_shift, u64 len,
+					u64 cur_start_addr, u64 *pages,
+					int *start_index, int *npages)
+{
+	u64 cur_end_addr = cur_start_addr + len;
+	u64 cur_end_addr_aligned = 0;
+	u64 mtt_entries;
+	int err = 0;
+	int k;
+
+	len += (cur_start_addr & (mtt_size - 1ULL));
+	cur_end_addr_aligned = round_up(cur_end_addr, mtt_size);
+	len += (cur_end_addr_aligned - cur_end_addr);
+	if (len & (mtt_size - 1ULL)) {
+		pr_warn("write_block: len %llx is not aligned to mtt_size %llx\n",
+			len, mtt_size);
+		return -EINVAL;
+	}
+
+	mtt_entries = (len >> mtt_shift);
+
+	/*
+	 * Align the MTT start address to the mtt_size.
+	 * Required to handle cases when the MR starts in the middle of an MTT
+	 * record. Was not required in old code since the physical addresses
+	 * provided by the dma subsystem were page aligned, which was also the
+	 * MTT size.
+	 */
+	cur_start_addr = round_down(cur_start_addr, mtt_size);
+	/* A new block is started ... */
+	for (k = 0; k < mtt_entries; ++k) {
+		pages[*npages] = cur_start_addr + (mtt_size * k);
+		(*npages)++;
+		/*
+		 * Be friendly to mlx4_write_mtt() and pass it chunks of
+		 * appropriate size.
+		 */
+		if (*npages == PAGE_SIZE / sizeof(u64)) {
+			err = mlx4_write_mtt(dev->dev, mtt, *start_index,
+					     *npages, pages);
+			if (err)
+				return err;
+
+			(*start_index) += *npages;
+			*npages = 0;
+		}
+	}
+
+	return 0;
+}
+
+static inline u64 alignment_of(u64 ptr)
+{
+	return ilog2(ptr & (~(ptr - 1)));
+}
+
+static int mlx4_ib_umem_calc_block_mtt(u64 next_block_start,
+				       u64 current_block_end,
+				       u64 block_shift)
+{
+	/* Check whether the alignment of the new block is aligned as well as
+	 * the previous block.
+	 * Block address must start with zeros till size of entity_size.
+	 */
+	if ((next_block_start & ((1ULL << block_shift) - 1ULL)) != 0)
+		/*
+		 * It is not as well aligned as the previous block-reduce the
+		 * mtt size accordingly. Here we take the last right bit which
+		 * is 1.
+		 */
+		block_shift = alignment_of(next_block_start);
+
+	/*
+	 * Check whether the alignment of the end of previous block - is it
+	 * aligned as well as the start of the block
+	 */
+	if (((current_block_end) & ((1ULL << block_shift) - 1ULL)) != 0)
+		/*
+		 * It is not as well aligned as the start of the block -
+		 * reduce the mtt size accordingly.
+		 */
+		block_shift = alignment_of(current_block_end);
+
+	return block_shift;
+}
+
 int mlx4_ib_umem_write_mtt(struct mlx4_ib_dev *dev, struct mlx4_mtt *mtt,
 			   struct ib_umem *umem)
 {
 	u64 *pages;
-	int i, k, entry;
-	int n;
-	int len;
+	u64 len = 0;
 	int err = 0;
+	u64 mtt_size;
+	u64 cur_start_addr = 0;
+	u64 mtt_shift;
+	int start_index = 0;
+	int npages = 0;
 	struct scatterlist *sg;
+	int i;
 
 	pages = (u64 *) __get_free_page(GFP_KERNEL);
 	if (!pages)
 		return -ENOMEM;
 
-	i = n = 0;
+	mtt_shift = mtt->page_shift;
+	mtt_size = 1ULL << mtt_shift;
 
-	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
-		len = sg_dma_len(sg) >> mtt->page_shift;
-		for (k = 0; k < len; ++k) {
-			pages[i++] = sg_dma_address(sg) +
-				(k << umem->page_shift);
-			/*
-			 * Be friendly to mlx4_write_mtt() and
-			 * pass it chunks of appropriate size.
-			 */
-			if (i == PAGE_SIZE / sizeof (u64)) {
-				err = mlx4_write_mtt(dev->dev, mtt, n,
-						     i, pages);
-				if (err)
-					goto out;
-				n += i;
-				i = 0;
-			}
+	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, i) {
+		if (cur_start_addr + len == sg_dma_address(sg)) {
+			/* still the same block */
+			len += sg_dma_len(sg);
+			continue;
 		}
+		/*
+		 * A new block is started ...
+		 * If len is malaligned, write an extra mtt entry to cover the
+		 * misaligned area (round up the division)
+		 */
+		err = mlx4_ib_umem_write_mtt_block(dev, mtt, mtt_size,
+						   mtt_shift, len,
+						   cur_start_addr,
+						   pages, &start_index,
+						   &npages);
+		if (err)
+			goto out;
+
+		cur_start_addr = sg_dma_address(sg);
+		len = sg_dma_len(sg);
 	}
 
-	if (i)
-		err = mlx4_write_mtt(dev->dev, mtt, n, i, pages);
+	/* Handle the last block */
+	if (len > 0) {
+		/*
+		 * If len is malaligned, write an extra mtt entry to cover
+		 * the misaligned area (round up the division)
+		 */
+		err = mlx4_ib_umem_write_mtt_block(dev, mtt, mtt_size,
+						   mtt_shift, len,
+						   cur_start_addr, pages,
+						   &start_index, &npages);
+		if (err)
+			goto out;
+	}
+
+	if (npages)
+		err = mlx4_write_mtt(dev->dev, mtt, start_index, npages, pages);
 
 out:
 	free_page((unsigned long) pages);
 	return err;
 }
 
+/*
+ * Calculate optimal mtt size based on contiguous pages.
+ * Function will return also the number of pages that are not aligned to the
+ * calculated mtt_size to be added to total number of pages. For that we should
+ * check the first chunk length & last chunk length and if not aligned to
+ * mtt_size we should increment the non_aligned_pages number. All chunks in the
+ * middle already handled as part of mtt shift calculation for both their start
+ * & end addresses.
+ */
+int mlx4_ib_umem_calc_optimal_mtt_size(struct ib_umem *umem, u64 start_va,
+				       int *num_of_mtts)
+{
+	u64 block_shift = MLX4_MAX_MTT_SHIFT;
+	u64 min_shift = umem->page_shift;
+	u64 last_block_aligned_end = 0;
+	u64 current_block_start = 0;
+	u64 first_block_start = 0;
+	u64 current_block_len = 0;
+	u64 last_block_end = 0;
+	struct scatterlist *sg;
+	u64 current_block_end;
+	u64 misalignment_bits;
+	u64 next_block_start;
+	u64 total_len = 0;
+	int i;
+
+	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, i) {
+		/*
+		 * Initialization - save the first chunk start as the
+		 * current_block_start - block means contiguous pages.
+		 */
+		if (current_block_len == 0 && current_block_start == 0) {
+			current_block_start = sg_dma_address(sg);
+			first_block_start = current_block_start;
+			/*
+			 * Find the bits that are different between the physical
+			 * address and the virtual address for the start of the
+			 * MR.
+			 * umem_get aligned the start_va to a page boundary.
+			 * Therefore, we need to align the start va to the same
+			 * boundary.
+			 * misalignment_bits is needed to handle the  case of a
+			 * single memory region. In this case, the rest of the
+			 * logic will not reduce the block size.  If we use a
+			 * block size which is bigger than the alignment of the
+			 * misalignment bits, we might use the virtual page
+			 * number instead of the physical page number, resulting
+			 * in access to the wrong data.
+			 */
+			misalignment_bits =
+			(start_va & (~(((u64)(BIT(umem->page_shift))) - 1ULL)))
+			^ current_block_start;
+			block_shift = min(alignment_of(misalignment_bits),
+					  block_shift);
+		}
+
+		/*
+		 * Go over the scatter entries and check if they continue the
+		 * previous scatter entry.
+		 */
+		next_block_start = sg_dma_address(sg);
+		current_block_end = current_block_start	+ current_block_len;
+		/* If we have a split (non-contig.) between two blocks */
+		if (current_block_end != next_block_start) {
+			block_shift = mlx4_ib_umem_calc_block_mtt
+					(next_block_start,
+					 current_block_end,
+					 block_shift);
+
+			/*
+			 * If we reached the minimum shift for 4k page we stop
+			 * the loop.
+			 */
+			if (block_shift <= min_shift)
+				goto end;
+
+			/*
+			 * If not saved yet we are in first block - we save the
+			 * length of first block to calculate the
+			 * non_aligned_pages number at the end.
+			 */
+			total_len += current_block_len;
+
+			/* Start a new block */
+			current_block_start = next_block_start;
+			current_block_len = sg_dma_len(sg);
+			continue;
+		}
+		/* The scatter entry is another part of the current block,
+		 * increase the block size.
+		 * An entry in the scatter can be larger than 4k (page) as of
+		 * dma mapping which merge some blocks together.
+		 */
+		current_block_len += sg_dma_len(sg);
+	}
+
+	/* Account for the last block in the total len */
+	total_len += current_block_len;
+	/* Add to the first block the misalignment that it suffers from. */
+	total_len += (first_block_start & ((1ULL << block_shift) - 1ULL));
+	last_block_end = current_block_start + current_block_len;
+	last_block_aligned_end = round_up(last_block_end, 1 << block_shift);
+	total_len += (last_block_aligned_end - last_block_end);
+
+	if (total_len & ((1ULL << block_shift) - 1ULL))
+		pr_warn("misaligned total length detected (%llu, %llu)!",
+			total_len, block_shift);
+
+	*num_of_mtts = total_len >> block_shift;
+end:
+	if (block_shift < min_shift) {
+		/*
+		 * If shift is less than the min we set a warning and return the
+		 * min shift.
+		 */
+		pr_warn("umem_calc_optimal_mtt_size - unexpected shift %lld\n", block_shift);
+
+		block_shift = min_shift;
+	}
+	return block_shift;
+}
+
 struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata)
@@ -155,7 +391,7 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	}
 
 	n = ib_umem_page_count(mr->umem);
-	shift = mr->umem->page_shift;
+	shift = mlx4_ib_umem_calc_optimal_mtt_size(mr->umem, start, &n);
 
 	err = mlx4_mr_alloc(dev->dev, to_mpd(pd)->pdn, virt_addr, length,
 			    convert_access(access_flags), n, shift, &mr->mmr);