From 18b415be9d631b3e0c3a3caacc5f02edb9413f6b Mon Sep 17 00:00:00 2001
From: Yann Gautier <yann.gautier@foss.st.com>
Date: Fri, 18 Jun 2021 11:33:26 +0200
Subject: [PATCH] feat(plat/st): improve FIP image loading from MMC

Instead of using a scratch buffer of 512 bytes, we can directly use the
image address and max size. The mmc_block_dev_spec struct info is then
overwritten for each image with this info, except FW_CONFIG and GPT
table which will still use the scratch buffer.
This allows using multiple blocks read on MMC, and so improves the boot
time.
A cache invalidate is required for the remaining data not used from the
first and last blocks read. It is not required for FW_CONFIG_ID,
as it is in scratch buffer in SYSRAM, and also because bl_mem_params
struct is overwritten in this case. This should also not be done if
the image is not found (OP-TEE extra binaries when using SP_min).

Change-Id: If3ecfdfe35bb9db66284036ca49c4bd1be4fd121
Signed-off-by: Yann Gautier <yann.gautier@foss.st.com>
---
 plat/st/common/bl2_io_storage.c   |  8 +++++++-
 plat/st/stm32mp1/bl2_plat_setup.c | 14 ++++++++++++++
 2 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/plat/st/common/bl2_io_storage.c b/plat/st/common/bl2_io_storage.c
index 01c289db8..6c7618aba 100644
--- a/plat/st/common/bl2_io_storage.c
+++ b/plat/st/common/bl2_io_storage.c
@@ -9,6 +9,7 @@
 
 #include <arch_helpers.h>
 #include <common/debug.h>
+#include <common/desc_image_load.h>
 #include <drivers/io/io_block.h>
 #include <drivers/io/io_driver.h>
 #include <drivers/io/io_fip.h>
@@ -45,7 +46,7 @@ static io_block_spec_t gpt_block_spec = {
 
 static uint32_t block_buffer[MMC_BLOCK_SIZE] __aligned(MMC_BLOCK_SIZE);
 
-static const io_block_dev_spec_t mmc_block_dev_spec = {
+static io_block_dev_spec_t mmc_block_dev_spec = {
 	/* It's used as temp buffer in block driver */
 	.buffer = {
 		.offset = (size_t)&block_buffer,
@@ -423,6 +424,11 @@ int bl2_plat_handle_pre_image_load(unsigned int image_id)
 			image_block_spec.length = entry->length;
 
 			gpt_init_done = true;
+		} else {
+			bl_mem_params_node_t *bl_mem_params = get_bl_mem_params_node(image_id);
+
+			mmc_block_dev_spec.buffer.offset = bl_mem_params->image_info.image_base;
+			mmc_block_dev_spec.buffer.length = bl_mem_params->image_info.image_max_size;
 		}
 
 		break;
diff --git a/plat/st/stm32mp1/bl2_plat_setup.c b/plat/st/stm32mp1/bl2_plat_setup.c
index 83e5cd136..53177f628 100644
--- a/plat/st/stm32mp1/bl2_plat_setup.c
+++ b/plat/st/stm32mp1/bl2_plat_setup.c
@@ -15,6 +15,7 @@
 #include <common/desc_image_load.h>
 #include <drivers/delay_timer.h>
 #include <drivers/generic_delay_timer.h>
+#include <drivers/mmc.h>
 #include <drivers/st/bsec.h>
 #include <drivers/st/stm32_console.h>
 #include <drivers/st/stm32_iwdg.h>
@@ -395,6 +396,19 @@ int bl2_plat_handle_post_image_load(unsigned int image_id)
 		break;
 	}
 
+#if STM32MP_SDMMC || STM32MP_EMMC
+	/*
+	 * Invalidate remaining data read from MMC but not flushed by load_image_flush().
+	 * We take the worst case which is 2 MMC blocks.
+	 */
+	if ((image_id != FW_CONFIG_ID) &&
+	    ((bl_mem_params->image_info.h.attr & IMAGE_ATTRIB_SKIP_LOADING) == 0U)) {
+		inv_dcache_range(bl_mem_params->image_info.image_base +
+				 bl_mem_params->image_info.image_size,
+				 2U * MMC_BLOCK_SIZE);
+	}
+#endif /* STM32MP_SDMMC || STM32MP_EMMC */
+
 	return err;
 }