From d5fc5012bbae5e51a9be9f63372a1ef6fef10371 Mon Sep 17 00:00:00 2001
From: Marcin Wojtas <mw@FreeBSD.org>
Date: Wed, 18 Nov 2020 14:30:59 +0000
Subject: [PATCH] ena-com: Fix ena-com to allocate cdesc aligned to 4k

The latest generation hardware requires IO CQ (completion queue)
descriptors memory to be aligned to a 4K. It needs that feature for
the best performance.

Allocating unaligned descriptors will have a big performance impact as
the packet processing in a HW won't be optimized properly.

It's a critical fix, especially for the arm64 EC2 instances.
---
 ena_com.c  | 26 ++++++++++++++------------
 ena_com.h  |  2 ++
 ena_plat.h | 22 +++++++++++++++++-----
 3 files changed, 33 insertions(+), 17 deletions(-)

diff --git a/ena_com.c b/ena_com.c
index dde0c3357f63..266e39859102 100644
--- a/ena_com.c
+++ b/ena_com.c
@@ -449,19 +449,21 @@ static int ena_com_init_io_cq(struct ena_com_dev *ena_dev,
 	size = io_cq->cdesc_entry_size_in_bytes * io_cq->q_depth;
 	io_cq->bus = ena_dev->bus;
 
-	ENA_MEM_ALLOC_COHERENT_NODE(ena_dev->dmadev,
-			size,
-			io_cq->cdesc_addr.virt_addr,
-			io_cq->cdesc_addr.phys_addr,
-			io_cq->cdesc_addr.mem_handle,
-			ctx->numa_node,
-			prev_node);
+	ENA_MEM_ALLOC_COHERENT_NODE_ALIGNED(ena_dev->dmadev,
+					    size,
+					    io_cq->cdesc_addr.virt_addr,
+					    io_cq->cdesc_addr.phys_addr,
+					    io_cq->cdesc_addr.mem_handle,
+					    ctx->numa_node,
+					    prev_node,
+					    ENA_CDESC_RING_SIZE_ALIGNMENT);
 	if (!io_cq->cdesc_addr.virt_addr) {
-		ENA_MEM_ALLOC_COHERENT(ena_dev->dmadev,
-				       size,
-				       io_cq->cdesc_addr.virt_addr,
-				       io_cq->cdesc_addr.phys_addr,
-				       io_cq->cdesc_addr.mem_handle);
+		ENA_MEM_ALLOC_COHERENT_ALIGNED(ena_dev->dmadev,
+					       size,
+					       io_cq->cdesc_addr.virt_addr,
+					       io_cq->cdesc_addr.phys_addr,
+					       io_cq->cdesc_addr.mem_handle,
+					       ENA_CDESC_RING_SIZE_ALIGNMENT);
 	}
 
 	if (!io_cq->cdesc_addr.virt_addr) {
diff --git a/ena_com.h b/ena_com.h
index c1b9540edd0b..b94728310fc9 100644
--- a/ena_com.h
+++ b/ena_com.h
@@ -51,6 +51,8 @@
 #define ADMIN_CQ_SIZE(depth)	((depth) * sizeof(struct ena_admin_acq_entry))
 #define ADMIN_AENQ_SIZE(depth)	((depth) * sizeof(struct ena_admin_aenq_entry))
 
+#define ENA_CDESC_RING_SIZE_ALIGNMENT	(1 << 12) /* 4K */
+
 /*****************************************************************************/
 /*****************************************************************************/
 /* ENA adaptive interrupt moderation settings */
diff --git a/ena_plat.h b/ena_plat.h
index e3536cdf3573..f096a5ff88d1 100644
--- a/ena_plat.h
+++ b/ena_plat.h
@@ -106,6 +106,8 @@ extern struct ena_bus_space ebs;
 #define ENA_ADMQ	(1 << 8) /* Detailed info about admin queue. 	      */
 #define ENA_NETMAP	(1 << 9) /* Detailed info about netmap. 	      */
 
+#define DEFAULT_ALLOC_ALIGNMENT	8
+
 extern int ena_log_level;
 
 #define ena_trace_raw(level, fmt, args...)			\
@@ -285,7 +287,7 @@ typedef uint64_t ena_time_t;
 void	ena_dmamap_callback(void *arg, bus_dma_segment_t *segs, int nseg,
     int error);
 int	ena_dma_alloc(device_t dmadev, bus_size_t size, ena_mem_handle_t *dma,
-    int mapflags);
+    int mapflags, bus_size_t alignment);
 
 static inline uint32_t
 ena_reg_read32(struct ena_bus *bus, bus_size_t offset)
@@ -313,20 +315,30 @@ ena_reg_read32(struct ena_bus *bus, bus_size_t offset)
 		(void)(size);						\
 		free(ptr, M_DEVBUF);					\
 	} while (0)
-#define ENA_MEM_ALLOC_COHERENT_NODE(dmadev, size, virt, phys, handle, node, \
-    dev_node)								\
+#define ENA_MEM_ALLOC_COHERENT_NODE_ALIGNED(dmadev, size, virt, phys,	\
+    handle, node, dev_node, alignment) 					\
 	do {								\
 		((virt) = NULL);					\
 		(void)(dev_node);					\
 	} while (0)
 
-#define ENA_MEM_ALLOC_COHERENT(dmadev, size, virt, phys, dma)		\
+#define ENA_MEM_ALLOC_COHERENT_NODE(dmadev, size, virt, phys, handle,	\
+    node, dev_node)							\
+	ENA_MEM_ALLOC_COHERENT_NODE_ALIGNED(dmadev, size, virt,		\
+	    phys, handle, node, dev_node, DEFAULT_ALLOC_ALIGNMENT)
+
+#define ENA_MEM_ALLOC_COHERENT_ALIGNED(dmadev, size, virt, phys, dma,	\
+    alignment)								\
 	do {								\
-		ena_dma_alloc((dmadev), (size), &(dma), 0);		\
+		ena_dma_alloc((dmadev), (size), &(dma), 0, alignment);	\
 		(virt) = (void *)(dma).vaddr;				\
 		(phys) = (dma).paddr;					\
 	} while (0)
 
+#define ENA_MEM_ALLOC_COHERENT(dmadev, size, virt, phys, dma)		\
+	ENA_MEM_ALLOC_COHERENT_ALIGNED(dmadev, size, virt,		\
+	    phys, dma, DEFAULT_ALLOC_ALIGNMENT)
+
 #define ENA_MEM_FREE_COHERENT(dmadev, size, virt, phys, dma)		\
 	do {								\
 		(void)size;						\