From d5c6191cc94b358de183cc8c88a5722a79445202 Mon Sep 17 00:00:00 2001
From: Phil Elwell <phil@raspberrypi.org>
Date: Thu, 1 Nov 2018 17:31:37 +0000
Subject: [PATCH] vchiq: Add 36-bit address support

Conditional on a new compatible string, change the pagelist encoding
such that the top 24 bits are the pfn, leaving 8 bits for run length
(-1).

Signed-off-by: Phil Elwell <phil@raspberrypi.org>
---
 .../interface/vchiq_arm/vchiq_2835_arm.c      | 90 ++++++++++++++-----
 .../interface/vchiq_arm/vchiq_arm.c           |  6 ++
 .../interface/vchiq_arm/vchiq_arm.h           |  1 +
 3 files changed, 75 insertions(+), 22 deletions(-)

--- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c
+++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c
@@ -47,6 +47,8 @@
 #include <soc/bcm2835/raspberrypi-firmware.h>
 
 #define TOTAL_SLOTS (VCHIQ_SLOT_ZERO_SLOTS + 2 * 32)
+#define VC_SAFE(x) (g_use_36bit_addrs ? ((u32)(x) | 0xc0000000) : (u32)(x))
+#define IS_VC_SAFE(x) (g_use_36bit_addrs ? !((x) & ~0x3fffffffull) : 1)
 
 #include "vchiq_arm.h"
 #include "vchiq_connected.h"
@@ -96,6 +98,7 @@ static void __iomem *g_regs;
  */
 static unsigned int g_cache_line_size;
 static struct dma_pool *g_dma_pool;
+static unsigned int g_use_36bit_addrs = 0;
 static unsigned int g_fragments_size;
 static char *g_fragments_base;
 static char *g_free_fragments;
@@ -139,6 +142,8 @@ int vchiq_platform_init(struct platform_
 	g_cache_line_size = drvdata->cache_line_size;
 	g_fragments_size = 2 * g_cache_line_size;
 
+	g_use_36bit_addrs = (dev->dma_pfn_offset == 0);
+
 	/* Allocate space for the channels in coherent memory */
 	slot_mem_size = PAGE_ALIGN(TOTAL_SLOTS * VCHIQ_SLOT_SIZE);
 	frag_mem_size = PAGE_ALIGN(g_fragments_size * MAX_FRAGMENTS);
@@ -150,14 +155,21 @@ int vchiq_platform_init(struct platform_
 		return -ENOMEM;
 	}
 
+	if (!IS_VC_SAFE(slot_phys)) {
+		dev_err(dev, "allocated DMA memory %pad is not VC-safe\n",
+			&slot_phys);
+		return -ENOMEM;
+	}
+
 	WARN_ON(((unsigned long)slot_mem & (PAGE_SIZE - 1)) != 0);
+	channelbase = VC_SAFE(slot_phys);
 
 	vchiq_slot_zero = vchiq_init_slots(slot_mem, slot_mem_size);
 	if (!vchiq_slot_zero)
 		return -EINVAL;
 
 	vchiq_slot_zero->platform_data[VCHIQ_PLATFORM_FRAGMENTS_OFFSET_IDX] =
-		(int)slot_phys + slot_mem_size;
+		channelbase + slot_mem_size;
 	vchiq_slot_zero->platform_data[VCHIQ_PLATFORM_FRAGMENTS_COUNT_IDX] =
 		MAX_FRAGMENTS;
 
@@ -193,7 +205,6 @@ int vchiq_platform_init(struct platform_
 	}
 
 	/* Send the base address of the slots to VideoCore */
-	channelbase = slot_phys;
 	err = rpi_firmware_property(fw, RPI_FIRMWARE_VCHIQ_INIT,
 				    &channelbase, sizeof(channelbase));
 	if (err || channelbase) {
@@ -282,7 +293,7 @@ vchiq_prepare_bulk_data(VCHIQ_BULK_T *bu
 		return VCHIQ_ERROR;
 
 	bulk->handle = memhandle;
-	bulk->data = (void *)(unsigned long)pagelistinfo->dma_addr;
+	bulk->data = (void *)VC_SAFE(pagelistinfo->dma_addr);
 
 	/*
 	 * Store the pagelistinfo address in remote_data,
@@ -570,25 +581,60 @@ create_pagelist(char __user *buf, size_t
 
 	/* Combine adjacent blocks for performance */
 	k = 0;
-	for_each_sg(scatterlist, sg, dma_buffers, i) {
-		u32 len = sg_dma_len(sg);
-		u32 addr = sg_dma_address(sg);
-
-		/* Note: addrs is the address + page_count - 1
-		 * The firmware expects blocks after the first to be page-
-		 * aligned and a multiple of the page size
-		 */
-		WARN_ON(len == 0);
-		WARN_ON(i && (i != (dma_buffers - 1)) && (len & ~PAGE_MASK));
-		WARN_ON(i && (addr & ~PAGE_MASK));
-		if (k > 0 &&
-		    ((addrs[k - 1] & PAGE_MASK) +
-		     (((addrs[k - 1] & ~PAGE_MASK) + 1) << PAGE_SHIFT))
-		    == (addr & PAGE_MASK))
-			addrs[k - 1] += ((len + PAGE_SIZE - 1) >> PAGE_SHIFT);
-		else
-			addrs[k++] = (addr & PAGE_MASK) |
-				(((len + PAGE_SIZE - 1) >> PAGE_SHIFT) - 1);
+	if (g_use_36bit_addrs) {
+		for_each_sg(scatterlist, sg, dma_buffers, i) {
+			u32 len = sg_dma_len(sg);
+			u64 addr = sg_dma_address(sg);
+			u32 page_id = (u32)((addr >> 4) & ~0xff);
+			u32 sg_pages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
+
+			/* Note: addrs is the address + page_count - 1
+			 * The firmware expects blocks after the first to be page-
+			 * aligned and a multiple of the page size
+			 */
+			WARN_ON(len == 0);
+			WARN_ON(i &&
+				(i != (dma_buffers - 1)) && (len & ~PAGE_MASK));
+			WARN_ON(i && (addr & ~PAGE_MASK));
+			WARN_ON(upper_32_bits(addr) > 0xf);
+			if (k > 0 &&
+			    ((addrs[k - 1] & ~0xff) +
+			     (((addrs[k - 1] & 0xff) + 1) << 8)
+			     == page_id)) {
+				u32 inc_pages = min(sg_pages,
+						    0xff - (addrs[k - 1] & 0xff));
+				addrs[k - 1] += inc_pages;
+				page_id += inc_pages << 8;
+				sg_pages -= inc_pages;
+			}
+			while (sg_pages) {
+				u32 inc_pages = min(sg_pages, 0x100u);
+				addrs[k++] = page_id | (inc_pages - 1);
+				page_id += inc_pages << 8;
+				sg_pages -= inc_pages;
+			}
+		}
+	} else {
+		for_each_sg(scatterlist, sg, dma_buffers, i) {
+			u32 len = sg_dma_len(sg);
+			u32 addr = VC_SAFE(sg_dma_address(sg));
+			u32 new_pages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
+
+			/* Note: addrs is the address + page_count - 1
+			 * The firmware expects blocks after the first to be page-
+			 * aligned and a multiple of the page size
+			 */
+			WARN_ON(len == 0);
+			WARN_ON(i && (i != (dma_buffers - 1)) && (len & ~PAGE_MASK));
+			WARN_ON(i && (addr & ~PAGE_MASK));
+			if (k > 0 &&
+			    ((addrs[k - 1] & PAGE_MASK) +
+			     (((addrs[k - 1] & ~PAGE_MASK) + 1) << PAGE_SHIFT))
+			    == (addr & PAGE_MASK))
+				addrs[k - 1] += new_pages;
+			else
+				addrs[k++] = (addr & PAGE_MASK) | (new_pages - 1);
+		}
 	}
 
 	/* Partial cache lines (fragments) require special measures */
--- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_arm.c
+++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_arm.c
@@ -181,6 +181,11 @@ static struct vchiq_drvdata bcm2836_drvd
 	.cache_line_size = 64,
 };
 
+static struct vchiq_drvdata bcm2838_drvdata = {
+	.cache_line_size = 64,
+	.use_36bit_addrs = true,
+};
+
 static const char *const ioctl_names[] = {
 	"CONNECT",
 	"SHUTDOWN",
@@ -3618,6 +3623,7 @@ vchiq_register_child(struct platform_dev
 static const struct of_device_id vchiq_of_match[] = {
 	{ .compatible = "brcm,bcm2835-vchiq", .data = &bcm2835_drvdata },
 	{ .compatible = "brcm,bcm2836-vchiq", .data = &bcm2836_drvdata },
+	{ .compatible = "brcm,bcm2838-vchiq", .data = &bcm2838_drvdata },
 	{},
 };
 MODULE_DEVICE_TABLE(of, vchiq_of_match);
--- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_arm.h
+++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_arm.h
@@ -125,6 +125,7 @@ typedef struct vchiq_arm_state_struct {
 
 struct vchiq_drvdata {
 	const unsigned int cache_line_size;
+	const bool use_36bit_addrs;
 	struct rpi_firmware *fw;
 };