Skip to contentSkip to conversation

x86: Provide a common MRC cache library and enable it for FSP

PATCH New Bin Meng 2 participants 48 messages Oct 12, 2015 – Nov 6, 2015 x86

lore
Share & tools
lore

Bin Meng

author
#0 lore raw

cover x86: Provide a common MRC cache library and enable it for FSP

This series moves existing MRC cache codes in the ivybridge cpu diretory to a common place and makes some changes so that every x86 board benefits from it. It also updates FSP support codes to pass MRC cache data to fsp_init() to speed up boot time.

Currently tested on Intel Bayley Bay board. Minnow Max board should work as it uses the same FSP that Bayley Bay uses. For Chromebook I cannot test that, but I see Simon added a TODO in the codes below:

	/*
	 * TODO(sjg at chromium.org): Skip this for now as it causes boot
 	 * problems
 	 */
 	if (0) {
		pei_data->mrc_input = mrc_cache->data;
		pei_data->mrc_input_len = mrc_cache->data_size;
	}

I have no idea what breaks the MRC cache on the Chromebook board. Simon, you may retest this on top of this series to see if things go better.

Bin Meng (13):
x86: Add ENABLE_MRC_CACHE Kconfig option
x86: Move mrccache.[c|h] to a common place
x86: Do sanity test on the cache record in mrccache_update()
x86: Add various minor tidy-ups in mrccache codes
x86: Add more common routines to manipulate mrc cache
x86: ivybridge: Use APIs provided in the mrccache lib
x86: Use struct mrc_region to describe a mrc region
x86: fsp: Pass mrc cache to fsp_init() and save it to gd after

    fsp_init()

x86: baytrail: Save mrc cache to spi flash
x86: Enable mrc cache for bayleybay and minnowmax
x86: baytrail: Issue full system reset in reset_cpu()
x86: Remove unused rw-mrc-cache properties in the link and panther dts
files
x86: ivybridge: Correct two typos for MRC

 arch/x86/Kconfig                               |   8 ++
 arch/x86/cpu/baytrail/valleyview.c             |  25 +++++
 arch/x86/cpu/ivybridge/Makefile                |   1 -
 arch/x86/cpu/ivybridge/sdram.c                 | 120 ++---------------------
 arch/x86/dts/bayleybay.dts                     |   6 ++
 arch/x86/dts/chromebook_link.dts               |   2 -
 arch/x86/dts/chromebox_panther.dts             |   3 -
 arch/x86/dts/minnowmax.dts                     |   6 ++
 arch/x86/include/asm/arch-ivybridge/mrccache.h |  51 ----------
 arch/x86/include/asm/mrccache.h                | 107 +++++++++++++++++++++
 arch/x86/lib/Makefile                          |   1 +
 arch/x86/lib/fsp/fsp_common.c                  |  30 +++++-
 arch/x86/lib/fsp/fsp_dram.c                    |   6 ++
 arch/x86/{cpu/ivybridge => lib}/mrccache.c     | 128 +++++++++++++++++++++----
 configs/bayleybay_defconfig                    |   1 +
 configs/chromebook_link_defconfig              |   1 +
 configs/chromebox_panther_defconfig            |   1 +
 configs/minnowmax_defconfig                    |   1 +
 doc/README.x86                                 |   1 +
 19 files changed, 313 insertions(+), 186 deletions(-)
 delete mode 100644 arch/x86/include/asm/arch-ivybridge/mrccache.h
 create mode 100644 arch/x86/include/asm/mrccache.h
 rename arch/x86/{cpu/ivybridge => lib}/mrccache.c (51%)
Signature · 1 line
1.8.2.1
author
#1 lore raw

PATCH 1/13 x86: Add ENABLE_MRC_CACHE Kconfig option

Create a Kconfig option for enabling MRC cache.

---

 arch/x86/Kconfig | 8 ++++++++
 1 file changed, 8 insertions(+)
arch/x86/Kconfig +8 -0
index bf09b21..f92082d 100644
@@ -239,6 +239,14 @@ config FSP_SYS_MALLOC_F_LEN
 	help
 	  Additional size of malloc() pool before relocation.
 
+config ENABLE_MRC_CACHE
+	bool "Enable MRC cache"
+	depends on !EFI && !SYS_COREBOOT
+	help
+	  Enable this feature to cause MRC data to be cached in NV storage
+	  to be used for speeding up boot time on future reboots and/or
+	  power cycles.
+
 config SMP
 	bool "Enable Symmetric Multiprocessing"
 	default n
Signature · 1 line
1.8.2.1
author
#4 lore raw

PATCH 2/13 x86: Move mrccache.[c|h] to a common place

In reply to Bin Meng

mrccache implementation can be common for all boards. Move it from ivybridge cpu directory to the common lib directory.

---

 arch/x86/cpu/ivybridge/Makefile                      | 1 -
 arch/x86/cpu/ivybridge/sdram.c                       | 2 +-
 arch/x86/include/asm/{arch-ivybridge => }/mrccache.h | 0
 arch/x86/lib/Makefile                                | 1 +
 arch/x86/{cpu/ivybridge => lib}/mrccache.c           | 3 +--
 configs/chromebook_link_defconfig                    | 1 +
 configs/chromebox_panther_defconfig                  | 1 +
 7 files changed, 5 insertions(+), 4 deletions(-)
 rename arch/x86/include/asm/{arch-ivybridge => }/mrccache.h (100%)
 rename arch/x86/{cpu/ivybridge => lib}/mrccache.c (98%)
arch/x86/cpu/ivybridge/Makefile +0 -1
index 3576b83..0c7efae 100644
@@ -14,7 +14,6 @@ obj-y += lpc.o
 obj-y += me_status.o
 obj-y += model_206ax.o
 obj-y += microcode_intel.o
-obj-y += mrccache.o
 obj-y += northbridge.o
 obj-y += pch.o
 obj-y += pci.o
arch/x86/cpu/ivybridge/sdram.c +1 -1
index 7f3b13d..10fbe5f 100644
@@ -21,10 +21,10 @@
 #include <asm/processor.h>
 #include <asm/gpio.h>
 #include <asm/global_data.h>
+#include <asm/mrccache.h>
 #include <asm/mtrr.h>
 #include <asm/pci.h>
 #include <asm/arch/me.h>
-#include <asm/arch/mrccache.h>
 #include <asm/arch/pei_data.h>
 #include <asm/arch/pch.h>
 #include <asm/post.h>
arch/x86/include/asm/mrccache.h renamed from arch/x86/include/asm/arch-ivybridge/mrccache.h +0 -0
similarity index 100%
rename from arch/x86/include/asm/arch-ivybridge/mrccache.h
rename to arch/x86/include/asm/mrccache.h
arch/x86/lib/Makefile +1 -0
index fa95944..2f82a21 100644
@@ -17,6 +17,7 @@ obj-y	+= init_helpers.o
 obj-y	+= interrupts.o
 obj-y	+= lpc-uclass.o
 obj-y	+= mpspec.o
+obj-$(CONFIG_ENABLE_MRC_CACHE) += mrccache.o
 obj-y += cmd_mtrr.o
 obj-$(CONFIG_SYS_PCAT_INTERRUPTS) += pcat_interrupts.o
 obj-$(CONFIG_SYS_PCAT_TIMER) += pcat_timer.o
arch/x86/lib/mrccache.c renamed from arch/x86/cpu/ivybridge/mrccache.c +1 -2
similarity index 98%
rename from arch/x86/cpu/ivybridge/mrccache.c
rename to arch/x86/lib/mrccache.c
index 9205494..ec0d2cb 100644
@@ -12,8 +12,7 @@
 #include <net.h>
 #include <spi.h>
 #include <spi_flash.h>
-#include <asm/arch/mrccache.h>
-#include <asm/arch/sandybridge.h>
+#include <asm/mrccache.h>
 
 static struct mrc_data_container *next_mrc_block(
 	struct mrc_data_container *mrc_cache)
configs/chromebook_link_defconfig +1 -0
index 21e85f3..fbecf8b 100644
@@ -4,6 +4,7 @@ CONFIG_VENDOR_GOOGLE=y
 CONFIG_DEFAULT_DEVICE_TREE="chromebook_link"
 CONFIG_TARGET_CHROMEBOOK_LINK=y
 CONFIG_HAVE_MRC=y
+CONFIG_ENABLE_MRC_CACHE=y
 CONFIG_HAVE_VGA_BIOS=y
 # CONFIG_CMD_IMLS is not set
 # CONFIG_CMD_FLASH is not set
configs/chromebox_panther_defconfig +1 -0
index 70a22a9..663aab0 100644
@@ -3,6 +3,7 @@ CONFIG_VENDOR_GOOGLE=y
 CONFIG_DEFAULT_DEVICE_TREE="chromebox_panther"
 CONFIG_TARGET_CHROMEBOX_PANTHER=y
 CONFIG_HAVE_MRC=y
+CONFIG_ENABLE_MRC_CACHE=y
 CONFIG_HAVE_VGA_BIOS=y
 # CONFIG_CMD_IMLS is not set
 # CONFIG_CMD_FLASH is not set
Signature · 1 line
1.8.2.1
#5 lore raw

PATCH 2/13 x86: Move mrccache.[c|h] to a common place

Quoted 6 lines · Bin Meng

mrccache implementation can be common for all boards. Move it from ivybridge cpu directory to the common lib directory.

---

Quoted 10 lines
 arch/x86/cpu/ivybridge/Makefile                      | 1 -
 arch/x86/cpu/ivybridge/sdram.c                       | 2 +-
 arch/x86/include/asm/{arch-ivybridge => }/mrccache.h | 0
 arch/x86/lib/Makefile                                | 1 +
 arch/x86/{cpu/ivybridge => lib}/mrccache.c           | 3 +--
 configs/chromebook_link_defconfig                    | 1 +
 configs/chromebox_panther_defconfig                  | 1 +
 7 files changed, 5 insertions(+), 4 deletions(-)
 rename arch/x86/include/asm/{arch-ivybridge => }/mrccache.h (100%)
 rename arch/x86/{cpu/ivybridge => lib}/mrccache.c (98%)
#6 lore raw

PATCH 2/13 x86: Move mrccache.[c|h] to a common place

Quoted 18 lines · Simon Glass
Quoted 6 lines · Bin Meng

mrccache implementation can be common for all boards. Move it from ivybridge cpu directory to the common lib directory.

---

Quoted 10 lines
 arch/x86/cpu/ivybridge/Makefile                      | 1 -
 arch/x86/cpu/ivybridge/sdram.c                       | 2 +-
 arch/x86/include/asm/{arch-ivybridge => }/mrccache.h | 0
 arch/x86/lib/Makefile                                | 1 +
 arch/x86/{cpu/ivybridge => lib}/mrccache.c           | 3 +--
 configs/chromebook_link_defconfig                    | 1 +
 configs/chromebox_panther_defconfig                  | 1 +
 7 files changed, 5 insertions(+), 4 deletions(-)
 rename arch/x86/include/asm/{arch-ivybridge => }/mrccache.h (100%)
 rename arch/x86/{cpu/ivybridge => lib}/mrccache.c (98%)
Quoted 1 line

Applied to u-boot-x86, thanks!

author
#7 lore raw

PATCH 3/13 x86: Do sanity test on the cache record in mrccache_update()

In reply to Bin Meng

For the cache record to write in mrccache_update(), we should perform a sanity test to see if it is a valid one.

---

 arch/x86/include/asm/mrccache.h | 2 +-
 arch/x86/lib/mrccache.c         | 3 +++
 2 files changed, 4 insertions(+), 1 deletion(-)
arch/x86/include/asm/mrccache.h +1 -1
index 1d50ebb..ff41b8a 100644
@@ -43,7 +43,7 @@ struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
  * @entry:	Position and size of MRC cache in SPI flash
  * @cur:	Record to write
  * @return 0 if updated, -EEXIST if the record is the same as the latest
- * record, other error if SPI write failed
+ * record, -EINVAL if the record is not valid, other error if SPI write failed
  */
 int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
 		    struct mrc_data_container *cur);
arch/x86/lib/mrccache.c +3 -0
index ec0d2cb..6dd3b5e 100644
@@ -112,6 +112,9 @@ int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
 	ulong base_addr;
 	int ret;
 
+	if (!is_mrc_cache(cur))
+		return -EINVAL;
+
 	/* Find the last used block */
 	base_addr = (1ULL << 32) - CONFIG_ROM_SIZE + entry->offset;
 	debug("Updating MRC cache data\n");
Signature · 1 line
1.8.2.1
#8 lore raw

PATCH 3/13 x86: Do sanity test on the cache record in mrccache_update()

Quoted 9 lines · Bin Meng

For the cache record to write in mrccache_update(), we should perform a sanity test to see if it is a valid one.

---

 arch/x86/include/asm/mrccache.h | 2 +-
 arch/x86/lib/mrccache.c         | 3 +++
 2 files changed, 4 insertions(+), 1 deletion(-)
#9 lore raw

PATCH 3/13 x86: Do sanity test on the cache record in mrccache_update()

Quoted 4 lines · Simon Glass
Quoted 3 lines · Bin Meng

For the cache record to write in mrccache_update(), we should perform a sanity test to see if it is a valid one.

Quoted 8 lines
Quoted 6 lines

---

 arch/x86/include/asm/mrccache.h | 2 +-
 arch/x86/lib/mrccache.c         | 3 +++
 2 files changed, 4 insertions(+), 1 deletion(-)

Applied to u-boot-x86, thanks!

author
#10 lore raw

PATCH 4/13 x86: Add various minor tidy-ups in mrccache codes

In reply to Bin Meng

Fix some nits, improve some comments and reorder some codes a little bit.

---

 arch/x86/include/asm/mrccache.h | 20 ++++++++++----------
 arch/x86/lib/mrccache.c         | 14 ++++++--------
 2 files changed, 16 insertions(+), 18 deletions(-)
arch/x86/include/asm/mrccache.h +10 -10
index ff41b8a..2fd9082 100644
@@ -1,17 +1,17 @@
 /*
- * Copyright (c) 2014 Google, Inc
+ * Copyright (C) 2014 Google, Inc
  *
  * SPDX-License-Identifier:	GPL-2.0+
  */
 
-#ifndef _ASM_ARCH_MRCCACHE_H
-#define _ASM_ARCH_MRCCACHE_H
+#ifndef _ASM_MRCCACHE_H
+#define _ASM_MRCCACHE_H
 
 #define MRC_DATA_ALIGN		0x1000
-#define MRC_DATA_SIGNATURE	(('M' << 0) | ('R' << 8) | ('C' << 16) | \
-					('D'<<24))
+#define MRC_DATA_SIGNATURE	(('M' << 0) | ('R' << 8) | \
+				 ('C' << 16) | ('D'<<24))
 
-__packed struct mrc_data_container {
+struct __packed mrc_data_container {
 	u32	signature;	/* "MRCD" */
 	u32	data_size;	/* Size of the 'data' field */
 	u32	checksum;	/* IP style checksum */
@@ -28,7 +28,7 @@ struct udevice;
  * This searches the MRC cache region looking for the latest record to use
  * for setting up SDRAM
  *
- * @entry:	Information about the position and size of the MRC cache
+ * @entry:	Position and size of MRC cache in SPI flash
  * @return pointer to latest record, or NULL if none
  */
 struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
@@ -36,8 +36,8 @@ struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
 /**
  * mrccache_update() - update the MRC cache with a new record
  *
- * This writes a new record to the end of the MRC cache. If the new record is
- * the same as the latest record then the write is skipped
+ * This writes a new record to the end of the MRC cache region. If the new
+ * record is the same as the latest record then the write is skipped
  *
  * @sf:		SPI flash to write to
  * @entry:	Position and size of MRC cache in SPI flash
@@ -48,4 +48,4 @@ struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
 int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
 		    struct mrc_data_container *cur);
 
-#endif
+#endif /* _ASM_MRCCACHE_H */
arch/x86/lib/mrccache.c +6 -8
index 6dd3b5e..1c4b5a8 100644
@@ -1,5 +1,5 @@
 /*
- * From Coreboot src/southbridge/intel/bd82x6x/mrccache.c
+ * From coreboot src/southbridge/intel/bd82x6x/mrccache.c
  *
  * Copyright (C) 2014 Google Inc.
  *
@@ -15,17 +15,19 @@
 #include <asm/mrccache.h>
 
 static struct mrc_data_container *next_mrc_block(
-	struct mrc_data_container *mrc_cache)
+	struct mrc_data_container *cache)
 {
 	/* MRC data blocks are aligned within the region */
-	u32 mrc_size = sizeof(*mrc_cache) + mrc_cache->data_size;
+	u32 mrc_size = sizeof(*cache) + cache->data_size;
+	u8 *region_ptr = (u8 *)cache;
+
 	if (mrc_size & (MRC_DATA_ALIGN - 1UL)) {
 		mrc_size &= ~(MRC_DATA_ALIGN - 1UL);
 		mrc_size += MRC_DATA_ALIGN;
 	}
 
-	u8 *region_ptr = (u8 *)mrc_cache;
 	region_ptr += mrc_size;
+
 	return (struct mrc_data_container *)region_ptr;
 }
 
@@ -34,10 +36,6 @@ static int is_mrc_cache(struct mrc_data_container *cache)
 	return cache && (cache->signature == MRC_DATA_SIGNATURE);
 }
 
-/*
- * Find the largest index block in the MRC cache. Return NULL if none is
- * found.
- */
 struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry)
 {
 	struct mrc_data_container *cache, *next;
Signature · 1 line
1.8.2.1
#11 lore raw

PATCH 4/13 x86: Add various minor tidy-ups in mrccache codes

Quoted 9 lines · Bin Meng

Fix some nits, improve some comments and reorder some codes a little bit.

---

 arch/x86/include/asm/mrccache.h | 20 ++++++++++----------
 arch/x86/lib/mrccache.c         | 14 ++++++--------
 2 files changed, 16 insertions(+), 18 deletions(-)
#12 lore raw

PATCH 4/13 x86: Add various minor tidy-ups in mrccache codes

Quoted 4 lines · Simon Glass
Quoted 3 lines · Bin Meng

Fix some nits, improve some comments and reorder some codes a little bit.

Quoted 8 lines
Quoted 6 lines

---

 arch/x86/include/asm/mrccache.h | 20 ++++++++++----------
 arch/x86/lib/mrccache.c         | 14 ++++++--------
 2 files changed, 16 insertions(+), 18 deletions(-)

Applied to u-boot-x86, thanks!

author
#13 lore raw

PATCH 5/13 x86: Add more common routines to manipulate mrc cache

In reply to Bin Meng

This adds mrccache_reserve(), mrccache_get_region() and mrccache_save() APIs to the mrccache codes. They are ported from the ivybridge implementation, but with some changes. For example, in the mrccache_reserve(), ivybridge version only reserves the pure MRC data, which causes additional malloc() when saving the cache as the save API needs some meta data. Now we change it to save the whole MRC date plus the meta data to elinimate the need for the malloc() later.

---

 arch/x86/include/asm/mrccache.h | 51 +++++++++++++++++++++++
 arch/x86/lib/mrccache.c         | 89 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 140 insertions(+)
arch/x86/include/asm/mrccache.h +51 -0
index 2fd9082..bcf7117 100644
@@ -1,5 +1,6 @@
 /*
  * Copyright (C) 2014 Google, Inc
+ * Copyright (C) 2015 Bin Meng <[email protected]>
  *
  * SPDX-License-Identifier:	GPL-2.0+
  */
@@ -11,6 +12,8 @@
 #define MRC_DATA_SIGNATURE	(('M' << 0) | ('R' << 8) | \
 				 ('C' << 16) | ('D'<<24))
 
+#define MRC_DATA_HEADER_SIZE	32
+
 struct __packed mrc_data_container {
 	u32	signature;	/* "MRCD" */
 	u32	data_size;	/* Size of the 'data' field */
@@ -48,4 +51,52 @@ struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
 int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
 		    struct mrc_data_container *cur);
 
+/**
+ * mrccache_reserve() - reserve MRC data on the stack
+ *
+ * This copies MRC data pointed by gd->arch.mrc_output to a new place on the
+ * stack with length gd->arch.mrc_output_len, and updates gd->arch.mrc_output
+ * to point to the new place once the migration is done.
+ *
+ * This routine should be called by reserve_arch() before U-Boot is relocated
+ * when MRC cache is enabled.
+ *
+ * @return 0 always
+ */
+int mrccache_reserve(void);
+
+/**
+ * mrccache_get_region() - get MRC region on the SPI flash
+ *
+ * This gets MRC region whose offset and size are described in the device tree
+ * as a subnode to the SPI flash. If a non-NULL device pointer is supplied,
+ * this also probes the SPI flash device and returns its device pointer for
+ * the caller to use later.
+ *
+ * Be careful when calling this routine with a non-NULL device pointer:
+ * - driver model initialization must be complete
+ * - calling in the pre-relocation phase may bring some side effects during
+ *   the SPI flash device probe (eg: for SPI controllers on a PCI bus, it
+ *   triggers PCI bus enumeration during which insufficient memory issue
+ *   might be exposed and it causes subsequent SPI flash probe fails).
+ *
+ * @devp:	Returns pointer to the SPI flash device
+ * @entry:	Position and size of MRC cache in SPI flash
+ * @return 0 if success, -ENOENT if SPI flash node does not exist in the
+ * device tree, -EPERM if MRC region subnode does not exist in the device
+ * tree, -EINVAL if MRC region properties format is incorrect, other error
+ * if SPI flash probe failed.
+ */
+int mrccache_get_region(struct udevice **devp, struct fmap_entry *entry);
+
+/**
+ * mrccache_save() - save MRC data to the SPI flash
+ *
+ * This saves MRC data stored previously by gd->arch.mrc_output to a proper
+ * place within the MRC region on the SPI flash.
+ *
+ * @return 0 if saved to SPI flash successfully, other error if failed
+ */
+int mrccache_save(void);
+
 #endif /* _ASM_MRCCACHE_H */
arch/x86/lib/mrccache.c +89 -0

109 lines not on this page · raw message

Signature · 1 line
1.8.2.1
#14 lore raw

PATCH 5/13 x86: Add more common routines to manipulate mrc cache

Quoted 9 lines · Bin Meng

This adds mrccache_reserve(), mrccache_get_region() and mrccache_save() APIs to the mrccache codes. They are ported from the ivybridge implementation, but with some changes. For example, in the mrccache_reserve(), ivybridge version only reserves the pure MRC data, which causes additional malloc() when saving the cache as the save API needs some meta data. Now we change it to save the whole MRC date plus the meta data to elinimate the need for the malloc() later.

Quoted 6 lines

---

 arch/x86/include/asm/mrccache.h | 51 +++++++++++++++++++++++
 arch/x86/lib/mrccache.c         | 89 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 140 insertions(+)
#15 lore raw

PATCH 5/13 x86: Add more common routines to manipulate mrc cache

Quoted 10 lines · Simon Glass
Quoted 9 lines · Bin Meng

This adds mrccache_reserve(), mrccache_get_region() and mrccache_save() APIs to the mrccache codes. They are ported from the ivybridge implementation, but with some changes. For example, in the mrccache_reserve(), ivybridge version only reserves the pure MRC data, which causes additional malloc() when saving the cache as the save API needs some meta data. Now we change it to save the whole MRC date plus the meta data to elinimate the need for the malloc() later.

Quoted 8 lines
Quoted 6 lines

---

 arch/x86/include/asm/mrccache.h | 51 +++++++++++++++++++++++
 arch/x86/lib/mrccache.c         | 89 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 140 insertions(+)

Applied to u-boot-x86, thanks!

author
#16 lore raw

PATCH 6/13 x86: ivybridge: Use APIs provided in the mrccache lib

In reply to Bin Meng

Remove the call to custom mrc cache APIs, and use the ones provided in the mrccache lib.

---

 arch/x86/cpu/ivybridge/sdram.c | 112 ++---------------------------------------
 1 file changed, 4 insertions(+), 108 deletions(-)
arch/x86/cpu/ivybridge/sdram.c +4 -108
index 10fbe5f..20c2e68 100644
@@ -89,42 +89,6 @@ void dram_init_banksize(void)
 	}
 }
 
-static int get_mrc_entry(struct udevice **devp, struct fmap_entry *entry)
-{
-	const void *blob = gd->fdt_blob;
-	int node, spi_node, mrc_node;
-	int upto;
-	int ret;
-
-	/* Find the flash chip within the SPI controller node */
-	upto = 0;
-	spi_node = fdtdec_next_alias(blob, "spi", COMPAT_INTEL_ICH_SPI, &upto);
-	if (spi_node < 0)
-		return -ENOENT;
-	node = fdt_first_subnode(blob, spi_node);
-	if (node < 0)
-		return -ECHILD;
-
-	/* Find the place where we put the MRC cache */
-	mrc_node = fdt_subnode_offset(blob, node, "rw-mrc-cache");
-	if (mrc_node < 0)
-		return -EPERM;
-
-	if (fdtdec_read_fmap_entry(blob, mrc_node, "rm-mrc-cache", entry))
-		return -EINVAL;
-
-	if (devp) {
-		debug("getting sf\n");
-		ret = uclass_get_device_by_of_offset(UCLASS_SPI_FLASH, node,
-						     devp);
-		debug("ret = %d\n", ret);
-		if (ret)
-			return ret;
-	}
-
-	return 0;
-}
-
 static int read_seed_from_cmos(struct pei_data *pei_data)
 {
 	u16 c1, c2, checksum, seed_checksum;
@@ -180,7 +144,7 @@ static int prepare_mrc_cache(struct pei_data *pei_data)
 	ret = read_seed_from_cmos(pei_data);
 	if (ret)
 		return ret;
-	ret = get_mrc_entry(NULL, &entry);
+	ret = mrccache_get_region(NULL, &entry);
 	if (ret)
 		return ret;
 	mrc_cache = mrccache_find_current(&entry);
@@ -202,32 +166,6 @@ static int prepare_mrc_cache(struct pei_data *pei_data)
 	return 0;
 }
 
-static int build_mrc_data(struct mrc_data_container **datap)
-{
-	struct mrc_data_container *data;
-	int orig_len;
-	int output_len;
-
-	orig_len = gd->arch.mrc_output_len;
-	output_len = ALIGN(orig_len, 16);
-	data = malloc(output_len + sizeof(*data));
-	if (!data)
-		return -ENOMEM;
-	data->signature = MRC_DATA_SIGNATURE;
-	data->data_size = output_len;
-	data->reserved = 0;
-	memcpy(data->data, gd->arch.mrc_output, orig_len);
-
-	/* Zero the unused space in aligned buffer. */
-	if (output_len > orig_len)
-		memset(data->data + orig_len, 0, output_len - orig_len);
-
-	data->checksum = compute_ip_checksum(data->data, output_len);
-	*datap = data;
-
-	return 0;
-}
-
 static int write_seeds_to_cmos(struct pei_data *pei_data)
 {
 	u16 c1, c2, checksum;
@@ -262,42 +200,12 @@ static int write_seeds_to_cmos(struct pei_data *pei_data)
 	return 0;
 }
 
-static int sdram_save_mrc_data(void)
-{
-	struct mrc_data_container *data;
-	struct fmap_entry entry;
-	struct udevice *sf;
-	int ret;
-
-	if (!gd->arch.mrc_output_len)
-		return 0;
-	debug("Saving %d bytes of MRC output data to SPI flash\n",
-	      gd->arch.mrc_output_len);
-
-	ret = get_mrc_entry(&sf, &entry);
-	if (ret)
-		goto err_entry;
-	ret = build_mrc_data(&data);
-	if (ret)
-		goto err_data;
-	ret = mrccache_update(sf, &entry, data);
-	if (!ret)
-		debug("Saved MRC data with checksum %04x\n", data->checksum);
-
-	free(data);
-err_data:
-err_entry:
-	if (ret)
-		debug("%s: Failed: %d\n", __func__, ret);
-	return ret;
-}
-
 /* Use this hook to save our SDRAM parameters */
 int misc_init_r(void)
 {
 	int ret;
 
-	ret = sdram_save_mrc_data();
+	ret = mrccache_save();
 	if (ret)
 		printf("Unable to save MRC data: %d\n", ret);
 
@@ -476,7 +384,7 @@ int sdram_initialise(struct pei_data *pei_data)
 	if (pei_data->boot_mode != PEI_BOOT_RESUME) {
 		/*
 		 * This will be copied to SDRAM in reserve_arch(), then written
-		 * to SPI flash in sdram_save_mrc_data()
+		 * to SPI flash in mrccache_save()
 		 */
 		gd->arch.mrc_output = (char *)pei_data->mrc_output;
 		gd->arch.mrc_output_len = pei_data->mrc_output_len;
@@ -490,19 +398,7 @@ int sdram_initialise(struct pei_data *pei_data)
 
 int reserve_arch(void)
 {
-	u16 checksum;
-
-	checksum = compute_ip_checksum(gd->arch.mrc_output,
-				       gd->arch.mrc_output_len);
-	debug("Saving %d bytes for MRC output data, checksum %04x\n",
-	      gd->arch.mrc_output_len, checksum);
-	gd->start_addr_sp -= gd->arch.mrc_output_len;
-	memcpy((void *)gd->start_addr_sp, gd->arch.mrc_output,
-	       gd->arch.mrc_output_len);
-	gd->arch.mrc_output = (char *)gd->start_addr_sp;
-	gd->start_addr_sp &= ~0xf;
-
-	return 0;
+	return mrccache_reserve();
 }
 
 static int copy_spd(struct pei_data *peid)
Signature · 1 line
1.8.2.1
#17 lore raw

PATCH 6/13 x86: ivybridge: Use APIs provided in the mrccache lib

Quoted 8 lines · Bin Meng

Remove the call to custom mrc cache APIs, and use the ones provided in the mrccache lib.

---

 arch/x86/cpu/ivybridge/sdram.c | 112 ++---------------------------------------
 1 file changed, 4 insertions(+), 108 deletions(-)
#18 lore raw

PATCH 6/13 x86: ivybridge: Use APIs provided in the mrccache lib

Quoted 11 lines · Simon Glass
Quoted 8 lines · Bin Meng

Remove the call to custom mrc cache APIs, and use the ones provided in the mrccache lib.

---

 arch/x86/cpu/ivybridge/sdram.c | 112 ++---------------------------------------
 1 file changed, 4 insertions(+), 108 deletions(-)

Applied to u-boot-x86, thanks!

author
#19 lore raw

PATCH 7/13 x86: Use struct mrc_region to describe a mrc region

In reply to Bin Meng

Currently struct fmap_entry is used to describe a mrc region. However this structure contains some other fields that are not related to mrc cache and causes confusion. Besides, it does not include a base address field to store SPI flash's base address. Instead in the mrccache.c it tries to use CONFIG_ROM_SIZE to calculate the SPI flash base address, which unfortunately is not 100% correct as CONFIG_ROM_SIZE may not match the whole SPI flash size.

Define a new struct mrc_region and use it instead.

---

 arch/x86/cpu/ivybridge/sdram.c  |  2 +-
 arch/x86/include/asm/mrccache.h | 13 +++++++++----
 arch/x86/lib/mrccache.c         | 27 +++++++++++++++++----------
 3 files changed, 27 insertions(+), 15 deletions(-)
arch/x86/cpu/ivybridge/sdram.c +1 -1
index 20c2e68..9121426 100644
@@ -138,7 +138,7 @@ static int read_seed_from_cmos(struct pei_data *pei_data)
 static int prepare_mrc_cache(struct pei_data *pei_data)
 {
 	struct mrc_data_container *mrc_cache;
-	struct fmap_entry entry;
+	struct mrc_region entry;
 	int ret;
 
 	ret = read_seed_from_cmos(pei_data);
arch/x86/include/asm/mrccache.h +9 -4
index bcf7117..e35b5ed 100644
@@ -22,7 +22,12 @@ struct __packed mrc_data_container {
 	u8	data[0];	/* Variable size, platform/run time dependent */
 };
 
-struct fmap_entry;
+struct mrc_region {
+	u32	base;
+	u32	offset;
+	u32	length;
+};
+
 struct udevice;
 
 /**
@@ -34,7 +39,7 @@ struct udevice;
  * @entry:	Position and size of MRC cache in SPI flash
  * @return pointer to latest record, or NULL if none
  */
-struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
+struct mrc_data_container *mrccache_find_current(struct mrc_region *entry);
 
 /**
  * mrccache_update() - update the MRC cache with a new record
@@ -48,7 +53,7 @@ struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry);
  * @return 0 if updated, -EEXIST if the record is the same as the latest
  * record, -EINVAL if the record is not valid, other error if SPI write failed
  */
-int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
+int mrccache_update(struct udevice *sf, struct mrc_region *entry,
 		    struct mrc_data_container *cur);
 
 /**
@@ -87,7 +92,7 @@ int mrccache_reserve(void);
  * tree, -EINVAL if MRC region properties format is incorrect, other error
  * if SPI flash probe failed.
  */
-int mrccache_get_region(struct udevice **devp, struct fmap_entry *entry);
+int mrccache_get_region(struct udevice **devp, struct mrc_region *entry);
 
 /**
  * mrccache_save() - save MRC data to the SPI flash
arch/x86/lib/mrccache.c +17 -10
index 0efae6d..53a1259 100644
@@ -40,13 +40,13 @@ static int is_mrc_cache(struct mrc_data_container *cache)
 	return cache && (cache->signature == MRC_DATA_SIGNATURE);
 }
 
-struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry)
+struct mrc_data_container *mrccache_find_current(struct mrc_region *entry)
 {
 	struct mrc_data_container *cache, *next;
 	ulong base_addr, end_addr;
 	uint id;
 
-	base_addr = (1ULL << 32) - CONFIG_ROM_SIZE + entry->offset;
+	base_addr = entry->base + entry->offset;
 	end_addr = base_addr + entry->length;
 	cache = NULL;
 
@@ -85,12 +85,12 @@ struct mrc_data_container *mrccache_find_current(struct fmap_entry *entry)
  *
  * @return next cache entry if found, NULL if we got to the end
  */
-static struct mrc_data_container *find_next_mrc_cache(struct fmap_entry *entry,
+static struct mrc_data_container *find_next_mrc_cache(struct mrc_region *entry,
 		struct mrc_data_container *cache)
 {
 	ulong base_addr, end_addr;
 
-	base_addr = (1ULL << 32) - CONFIG_ROM_SIZE + entry->offset;
+	base_addr = entry->base + entry->offset;
 	end_addr = base_addr + entry->length;
 
 	cache = next_mrc_block(cache);
@@ -106,7 +106,7 @@ static struct mrc_data_container *find_next_mrc_cache(struct fmap_entry *entry,
 	return cache;
 }
 
-int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
+int mrccache_update(struct udevice *sf, struct mrc_region *entry,
 		    struct mrc_data_container *cur)
 {
 	struct mrc_data_container *cache;
@@ -118,7 +118,7 @@ int mrccache_update(struct udevice *sf, struct fmap_entry *entry,
 		return -EINVAL;
 
 	/* Find the last used block */
-	base_addr = (1ULL << 32) - CONFIG_ROM_SIZE + entry->offset;
+	base_addr = entry->base + entry->offset;
 	debug("Updating MRC cache data\n");
 	cache = mrccache_find_current(entry);
 	if (cache && (cache->data_size == cur->data_size) &&
@@ -189,10 +189,11 @@ int mrccache_reserve(void)
 	return 0;
 }
 
-int mrccache_get_region(struct udevice **devp, struct fmap_entry *entry)
+int mrccache_get_region(struct udevice **devp, struct mrc_region *entry)
 {
 	const void *blob = gd->fdt_blob;
 	int node, mrc_node;
+	u32 reg[2];
 	int ret;
 
 	/* Find the flash chip within the SPI controller node */
@@ -200,13 +201,19 @@ int mrccache_get_region(struct udevice **devp, struct fmap_entry *entry)
 	if (node < 0)
 		return -ENOENT;
 
+	if (fdtdec_get_int_array(blob, node, "memory-map", reg, 2))
+		return -FDT_ERR_NOTFOUND;
+	entry->base = reg[0];
+
 	/* Find the place where we put the MRC cache */
 	mrc_node = fdt_subnode_offset(blob, node, "rw-mrc-cache");
 	if (mrc_node < 0)
 		return -EPERM;
 
-	if (fdtdec_read_fmap_entry(blob, mrc_node, "rm-mrc-cache", entry))
-		return -EINVAL;
+	if (fdtdec_get_int_array(blob, mrc_node, "reg", reg, 2))
+		return -FDT_ERR_NOTFOUND;
+	entry->offset = reg[0];
+	entry->length = reg[1];
 
 	if (devp) {
 		ret = uclass_get_device_by_of_offset(UCLASS_SPI_FLASH, node,
@@ -222,7 +229,7 @@ int mrccache_get_region(struct udevice **devp, struct fmap_entry *entry)
 int mrccache_save(void)
 {
 	struct mrc_data_container *data;
-	struct fmap_entry entry;
+	struct mrc_region entry;
 	struct udevice *sf;
 	int ret;
 
Signature · 1 line
1.8.2.1
#20 lore raw

PATCH 7/13 x86: Use struct mrc_region to describe a mrc region

Hi Bin,

Quoted 8 lines · Bin Meng

Currently struct fmap_entry is used to describe a mrc region. However this structure contains some other fields that are not related to mrc cache and causes confusion. Besides, it does not include a base address field to store SPI flash's base address. Instead in the mrccache.c it tries to use CONFIG_ROM_SIZE to calculate the SPI flash base address, which unfortunately is not 100% correct as CONFIG_ROM_SIZE may not match the whole SPI flash size.

I'm still not comfortable with CONFIG_ROM_SIZE not being the size of the SPI flash. It just seems wrong. Maybe we should have a u-boot.rom.min with a minimal images?

Quoted 10 lines

Define a new struct mrc_region and use it instead.

---

 arch/x86/cpu/ivybridge/sdram.c  |  2 +-
 arch/x86/include/asm/mrccache.h | 13 +++++++++----
 arch/x86/lib/mrccache.c         | 27 +++++++++++++++++----------
 3 files changed, 27 insertions(+), 15 deletions(-)
#21 lore raw

PATCH 7/13 x86: Use struct mrc_region to describe a mrc region

Quoted 22 lines · Simon Glass

Hi Bin,

Quoted 8 lines · Bin Meng

Currently struct fmap_entry is used to describe a mrc region. However this structure contains some other fields that are not related to mrc cache and causes confusion. Besides, it does not include a base address field to store SPI flash's base address. Instead in the mrccache.c it tries to use CONFIG_ROM_SIZE to calculate the SPI flash base address, which unfortunately is not 100% correct as CONFIG_ROM_SIZE may not match the whole SPI flash size.

I'm still not comfortable with CONFIG_ROM_SIZE not being the size of the SPI flash. It just seems wrong. Maybe we should have a u-boot.rom.min with a minimal images?

Quoted 6 lines

Define a new struct mrc_region and use it instead.

---

Quoted 6 lines
Quoted 4 lines
 arch/x86/cpu/ivybridge/sdram.c  |  2 +-
 arch/x86/include/asm/mrccache.h | 13 +++++++++----
 arch/x86/lib/mrccache.c         | 27 +++++++++++++++++----------
 3 files changed, 27 insertions(+), 15 deletions(-)

Applied to u-boot-x86, thanks!

author
#22 lore raw

PATCH 7/13 x86: Use struct mrc_region to describe a mrc region

In reply to Simon Glass

Hi Simon,

Quoted 12 lines · Simon Glass

Hi Bin,

Quoted 8 lines · Bin Meng

Currently struct fmap_entry is used to describe a mrc region. However this structure contains some other fields that are not related to mrc cache and causes confusion. Besides, it does not include a base address field to store SPI flash's base address. Instead in the mrccache.c it tries to use CONFIG_ROM_SIZE to calculate the SPI flash base address, which unfortunately is not 100% correct as CONFIG_ROM_SIZE may not match the whole SPI flash size.

Quoted 4 lines

I'm still not comfortable with CONFIG_ROM_SIZE not being the size of the SPI flash. It just seems wrong. Maybe we should have a u-boot.rom.min with a minimal images?

As we discussed before, we can just document this in README.x86?

Quoted 6 lines
Quoted 6 lines

Define a new struct mrc_region and use it instead.

---

Quoted 6 lines
Quoted 4 lines
 arch/x86/cpu/ivybridge/sdram.c  |  2 +-
 arch/x86/include/asm/mrccache.h | 13 +++++++++----
 arch/x86/lib/mrccache.c         | 27 +++++++++++++++++----------
 3 files changed, 27 insertions(+), 15 deletions(-)

Regards,
Bin

#23 lore raw

PATCH 7/13 x86: Use struct mrc_region to describe a mrc region

Hi Bin,

Quoted 15 lines · Bin Meng

Hi Simon,

Quoted 12 lines · Simon Glass

Hi Bin,

Quoted 8 lines · Bin Meng

Currently struct fmap_entry is used to describe a mrc region. However this structure contains some other fields that are not related to mrc cache and causes confusion. Besides, it does not include a base address field to store SPI flash's base address. Instead in the mrccache.c it tries to use CONFIG_ROM_SIZE to calculate the SPI flash base address, which unfortunately is not 100% correct as CONFIG_ROM_SIZE may not match the whole SPI flash size.

Quoted 7 lines
Quoted 4 lines

I'm still not comfortable with CONFIG_ROM_SIZE not being the size of the SPI flash. It just seems wrong. Maybe we should have a u-boot.rom.min with a minimal images?

As we discussed before, we can just document this in README.x86?

Yes although it irks me when I see a commit message suggesting that it is causing problems!

Quoted 11 lines
Quoted 11 lines
Quoted 10 lines

Define a new struct mrc_region and use it instead.

---

 arch/x86/cpu/ivybridge/sdram.c  |  2 +-
 arch/x86/include/asm/mrccache.h | 13 +++++++++----
 arch/x86/lib/mrccache.c         | 27 +++++++++++++++++----------
 3 files changed, 27 insertions(+), 15 deletions(-)
Quoted 4 lines
Quoted 1 line

Regards,
Bin

Regards,
Simon

author
#24 lore raw

PATCH 8/13 x86: fsp: Pass mrc cache to fsp_init() and save it to gd after fsp_init()

In reply to Bin Meng

fsp_init() call has a parameter nvs_buf which is used by FSP as the MRC cache but currently is blindly set to NULL. Retreive the MRC cache from SPI flash and pass it to fsp_init() call. After the call, save FSP produced MRC cache to SPI flash too.

---

 arch/x86/lib/fsp/fsp_common.c | 30 +++++++++++++++++++++++++++++-
 arch/x86/lib/fsp/fsp_dram.c   |  6 ++++++
 2 files changed, 35 insertions(+), 1 deletion(-)
arch/x86/lib/fsp/fsp_common.c +29 -1
index 658f32d..c78df94 100644
@@ -7,6 +7,7 @@
 #include <common.h>
 #include <errno.h>
 #include <asm/io.h>
+#include <asm/mrccache.h>
 #include <asm/post.h>
 #include <asm/processor.h>
 #include <asm/fsp/fsp_support.h>
@@ -54,15 +55,42 @@ void board_final_cleanup(void)
 	return;
 }
 
+static __maybe_unused void *fsp_prepare_mrc_cache(void)
+{
+	struct mrc_data_container *cache;
+	struct mrc_region entry;
+	int ret;
+
+	ret = mrccache_get_region(NULL, &entry);
+	if (ret)
+		return NULL;
+
+	cache = mrccache_find_current(&entry);
+	if (!cache)
+		return NULL;
+
+	debug("%s: mrc cache at %p, size %x checksum %04x\n", __func__,
+	      cache->data, cache->data_size, cache->checksum);
+
+	return cache->data;
+}
+
 int x86_fsp_init(void)
 {
+	void *nvs;
+
 	if (!gd->arch.hob_list) {
+#ifdef CONFIG_ENABLE_MRC_CACHE
+		nvs = fsp_prepare_mrc_cache();
+#else
+		nvs = NULL;
+#endif
 		/*
 		 * The first time we enter here, call fsp_init().
 		 * Note the execution does not return to this function,
 		 * instead it jumps to fsp_continue().
 		 */
-		fsp_init(CONFIG_FSP_TEMP_RAM_ADDR, BOOT_FULL_CONFIG, NULL);
+		fsp_init(CONFIG_FSP_TEMP_RAM_ADDR, BOOT_FULL_CONFIG, nvs);
 	} else {
 		/*
 		 * The second time we enter here, adjust the size of malloc()
arch/x86/lib/fsp/fsp_dram.c +6 -0
index e51ca96..fcfe693 100644
@@ -7,6 +7,7 @@
 #include <common.h>
 #include <asm/fsp/fsp_support.h>
 #include <asm/e820.h>
+#include <asm/mrccache.h>
 #include <asm/post.h>
 
 DECLARE_GLOBAL_DATA_PTR;
@@ -32,6 +33,11 @@ int dram_init(void)
 	gd->ram_size = ram_size;
 	post_code(POST_DRAM);
 
+#ifdef CONFIG_ENABLE_MRC_CACHE
+	gd->arch.mrc_output = fsp_get_nvs_data(gd->arch.hob_list,
+					       &gd->arch.mrc_output_len);
+#endif
+
 	return 0;
 }
 
Signature · 1 line
1.8.2.1
#25 lore raw

PATCH 8/13 x86: fsp: Pass mrc cache to fsp_init() and save it to gd after fsp_init()

Quoted 11 lines · Bin Meng

fsp_init() call has a parameter nvs_buf which is used by FSP as the MRC cache but currently is blindly set to NULL. Retreive the MRC cache from SPI flash and pass it to fsp_init() call. After the call, save FSP produced MRC cache to SPI flash too.

---

 arch/x86/lib/fsp/fsp_common.c | 30 +++++++++++++++++++++++++++++-
 arch/x86/lib/fsp/fsp_dram.c   |  6 ++++++
 2 files changed, 35 insertions(+), 1 deletion(-)
#26 lore raw

PATCH 8/13 x86: fsp: Pass mrc cache to fsp_init() and save it to gd after fsp_init()

Quoted 6 lines · Simon Glass
Quoted 5 lines · Bin Meng

fsp_init() call has a parameter nvs_buf which is used by FSP as the MRC cache but currently is blindly set to NULL. Retreive the MRC cache from SPI flash and pass it to fsp_init() call. After the call, save FSP produced MRC cache to SPI flash too.

Quoted 8 lines
Quoted 6 lines

---

 arch/x86/lib/fsp/fsp_common.c | 30 +++++++++++++++++++++++++++++-
 arch/x86/lib/fsp/fsp_dram.c   |  6 ++++++
 2 files changed, 35 insertions(+), 1 deletion(-)

Applied to u-boot-x86, thanks!

author
#27 lore raw

PATCH 9/13 x86: baytrail: Save mrc cache to spi flash

In reply to Bin Meng

Save MRC cache to SPI flash in arch_misc_init().

---

 arch/x86/cpu/baytrail/valleyview.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
arch/x86/cpu/baytrail/valleyview.c +19 -0
index 4baaae6..215e0d0 100644
@@ -8,6 +8,7 @@
 #include <mmc.h>
 #include <pci_ids.h>
 #include <asm/irq.h>
+#include <asm/mrccache.h>
 #include <asm/post.h>
 
 static struct pci_device_id mmc_supported[] = {
@@ -43,6 +44,24 @@ int arch_misc_init(void)
 	if (!ll_boot_init())
 		return 0;
 
+#ifdef CONFIG_ENABLE_MRC_CACHE
+	/*
+	 * We intend not to check any return value here, as even MRC cache
+	 * is not saved successfully, it is not a severe error that will
+	 * prevent system from continuing to boot.
+	 */
+	mrccache_save();
+#endif
+
 	return pirq_init();
 }
+
+int reserve_arch(void)
+{
+#ifdef CONFIG_ENABLE_MRC_CACHE
+	return mrccache_reserve();
+#else
+	return 0;
+#endif
+}
 #endif
Signature · 1 line
1.8.2.1
#29 lore raw

PATCH 9/13 x86: baytrail: Save mrc cache to spi flash

Quoted 10 lines · Simon Glass
Quoted 7 lines · Bin Meng

Save MRC cache to SPI flash in arch_misc_init().

---

 arch/x86/cpu/baytrail/valleyview.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

Applied to u-boot-x86, thanks!

author
#30 lore raw

PATCH 10/13 x86: Enable mrc cache for bayleybay and minnowmax

In reply to Bin Meng

Now that we have added MRC cache for Intel FSP and BayTrail codes, enable it for all BayTrail boards (Bayley Bay and Minnow Max).

Note it turns out that FSP for Intel Atom E6xx does not produce the HOB for NV storage, so we don't have such functionality on Intel Crown Bay board.

---

 arch/x86/dts/bayleybay.dts  | 6 ++++++
 arch/x86/dts/minnowmax.dts  | 6 ++++++
 configs/bayleybay_defconfig | 1 +
 configs/minnowmax_defconfig | 1 +
 doc/README.x86              | 1 +
 5 files changed, 15 insertions(+)
arch/x86/dts/bayleybay.dts +6 -0
index d646987..52d0999 100644
@@ -68,9 +68,15 @@
 		#size-cells = <0>;
 		compatible = "intel,ich-spi";
 		spi-flash at 0 {
+			#address-cells = <1>;
+			#size-cells = <1>;
 			reg = <0>;
 			compatible = "winbond,w25q64dw", "spi-flash";
 			memory-map = <0xff800000 0x00800000>;
+			rw-mrc-cache {
+				label = "rw-mrc-cache";
+				reg = <0x006e0000 0x00010000>;
+			};
 		};
 	};
 
arch/x86/dts/minnowmax.dts +6 -0
index e917f0f..b03f987 100644
@@ -273,9 +273,15 @@
 		#size-cells = <0>;
 		compatible = "intel,ich-spi";
 		spi-flash at 0 {
+			#address-cells = <1>;
+			#size-cells = <1>;
 			reg = <0>;
 			compatible = "stmicro,n25q064a", "spi-flash";
 			memory-map = <0xff800000 0x00800000>;
+			rw-mrc-cache {
+				label = "rw-mrc-cache";
+				reg = <0x006f0000 0x00010000>;
+			};
 		};
 	};
 
configs/bayleybay_defconfig +1 -0
index 56316ee..fc40da8 100644
@@ -3,6 +3,7 @@ CONFIG_VENDOR_INTEL=y
 CONFIG_DEFAULT_DEVICE_TREE="bayleybay"
 CONFIG_TARGET_BAYLEYBAY=y
 CONFIG_HAVE_INTEL_ME=y
+CONFIG_ENABLE_MRC_CACHE=y
 CONFIG_SMP=y
 CONFIG_HAVE_VGA_BIOS=y
 CONFIG_VGA_BIOS_ADDR=0xfffa0000
configs/minnowmax_defconfig +1 -0
index 0d5bd4e..8f99f0e 100644
@@ -3,6 +3,7 @@ CONFIG_VENDOR_INTEL=y
 CONFIG_DEFAULT_DEVICE_TREE="minnowmax"
 CONFIG_TARGET_MINNOWMAX=y
 CONFIG_HAVE_INTEL_ME=y
+CONFIG_ENABLE_MRC_CACHE=y
 CONFIG_SMP=y
 CONFIG_HAVE_VGA_BIOS=y
 CONFIG_GENERATE_PIRQ_TABLE=y
doc/README.x86 +1 -0
index a9d0e0f..1271e5e 100644
@@ -190,6 +190,7 @@ Offset   Description         Controlling config
 000000   descriptor.bin      Hard-coded to 0 in ifdtool
 001000   me.bin              Set by the descriptor
 500000   <spare>
+6f0000   MRC cache           CONFIG_ENABLE_MRC_CACHE
 700000   u-boot-dtb.bin      CONFIG_SYS_TEXT_BASE
 790000   vga.bin             CONFIG_VGA_BIOS_ADDR
 7c0000   fsp.bin             CONFIG_FSP_ADDR
Signature · 1 line
1.8.2.1
#31 lore raw

PATCH 10/13 x86: Enable mrc cache for bayleybay and minnowmax

Quoted 8 lines · Bin Meng

Now that we have added MRC cache for Intel FSP and BayTrail codes, enable it for all BayTrail boards (Bayley Bay and Minnow Max).

Note it turns out that FSP for Intel Atom E6xx does not produce the HOB for NV storage, so we don't have such functionality on Intel Crown Bay board.

Quoted 8 lines

---

 arch/x86/dts/bayleybay.dts  | 6 ++++++
 arch/x86/dts/minnowmax.dts  | 6 ++++++
 configs/bayleybay_defconfig | 1 +
 configs/minnowmax_defconfig | 1 +
 doc/README.x86              | 1 +
 5 files changed, 15 insertions(+)
#32 lore raw

PATCH 10/13 x86: Enable mrc cache for bayleybay and minnowmax

Quoted 11 lines · Simon Glass
Quoted 10 lines · Bin Meng

Now that we have added MRC cache for Intel FSP and BayTrail codes, enable it for all BayTrail boards (Bayley Bay and Minnow Max).

Note it turns out that FSP for Intel Atom E6xx does not produce the HOB for NV storage, so we don't have such functionality on Intel Crown Bay board.

---

Quoted 8 lines
Quoted 6 lines
 arch/x86/dts/bayleybay.dts  | 6 ++++++
 arch/x86/dts/minnowmax.dts  | 6 ++++++
 configs/bayleybay_defconfig | 1 +
 configs/minnowmax_defconfig | 1 +
 doc/README.x86              | 1 +
 5 files changed, 15 insertions(+)

Applied to u-boot-x86, thanks!

author
#33 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

In reply to Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...

Changing reset_cpu() to do a full system reset fixes this issue.

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)
arch/x86/cpu/baytrail/valleyview.c +6 -0
index 215e0d0..a009c14 100644
@@ -65,3 +65,9 @@ int reserve_arch(void)
 #endif
 }
 #endif
+
+void reset_cpu(ulong addr)
+{
+	/* cold reset */
+	x86_full_reset();
+}
Signature · 1 line
1.8.2.1
#34 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

Hi Bin,

Quoted 12 lines · Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...
Quoted 7 lines

Changing reset_cpu() to do a full system reset fixes this issue.

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)
Quoted 16 lines
arch/x86/cpu/baytrail/valleyview.c +6 -1
index 215e0d0..a009c14 100644
@@ -65,3 +65,9 @@ int reserve_arch(void)
 #endif
 }
 #endif
+
+void reset_cpu(ulong addr)
+{
+       /* cold reset */
+       x86_full_reset();
+}
--
Quoted 2 lines

1.8.2.1

Actually on Minnowmax this doesn't fix the hang.

Also I don't see any speed-up from using the cache on Minnowmax.

Regards,
Simon

#35 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

Quoted 17 lines · Simon Glass

Hi Bin,

Quoted 14 lines · Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...

Changing reset_cpu() to do a full system reset fixes this issue.

Quoted 7 lines
Quoted 5 lines

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)

Applied to u-boot-x86, thanks!

author
#36 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

In reply to Simon Glass

Hi Simon,

Quoted 37 lines · Simon Glass

Hi Bin,

Quoted 12 lines · Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...
Quoted 7 lines

Changing reset_cpu() to do a full system reset fixes this issue.

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)
Quoted 12 lines
arch/x86/cpu/baytrail/valleyview.c +3 -0
index 215e0d0..a009c14 100644
@@ -65,3 +65,9 @@ int reserve_arch(void)
 #endif
 }
 #endif
+
+void reset_cpu(ulong addr)
+{
Quoted 8 lines
Quoted 6 lines
arch/x86/cpu/baytrail/valleyview.c +3 -1
+       /* cold reset */
+       x86_full_reset();
+}
--

1.8.2.1

Actually on Minnowmax this doesn't fix the hang.

That's strange. As full reset indicates a complete power cycle. Does MinnowMax boot with MRC cache enabled?

Quoted 2 lines

Also I don't see any speed-up from using the cache on Minnowmax.

On Galileo, I did not see any speed-up either. I think it is because on MinnowMax and Galileo they both uses memory-down configuration. The speed-up seems only helpful for DIMMs (on Intel Bayley Bay)

Quoted 3 lines

Regards,
Simon

Regards,
Bin

#37 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

Hi Bin,

Quoted 49 lines · Bin Meng

Hi Simon,

Quoted 37 lines · Simon Glass

Hi Bin,

Quoted 12 lines · Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...
Quoted 7 lines

Changing reset_cpu() to do a full system reset fixes this issue.

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)
Quoted 12 lines
arch/x86/cpu/baytrail/valleyview.c +3 -0
index 215e0d0..a009c14 100644
@@ -65,3 +65,9 @@ int reserve_arch(void)
 #endif
 }
 #endif
+
+void reset_cpu(ulong addr)
+{
Quoted 8 lines
Quoted 6 lines
arch/x86/cpu/baytrail/valleyview.c +3 -1
+       /* cold reset */
+       x86_full_reset();
+}
--

1.8.2.1

Actually on Minnowmax this doesn't fix the hang.

Quoted 2 lines

That's strange. As full reset indicates a complete power cycle. Does MinnowMax boot with MRC cache enabled?

It's hard to tell because the UART does not work that early. I wish we could fix that. I thought I did have it running at one point, but I may have imagined it.

Re the hang, I'm not sure - it seems to have stopped happening, so let's not worry about it for now.

Quoted 6 lines
Quoted 2 lines

Also I don't see any speed-up from using the cache on Minnowmax.

On Galileo, I did not see any speed-up either. I think it is because on MinnowMax and Galileo they both uses memory-down configuration. The speed-up seems only helpful for DIMMs (on Intel Bayley Bay)

So what is the point of enabling the MRC cache?

Regards,
Simon

author
#38 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

Hi Simon,

Quoted 52 lines · Simon Glass

Hi Bin,

Quoted 49 lines · Bin Meng

Hi Simon,

Quoted 45 lines · Simon Glass

Hi Bin,

Quoted 12 lines · Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...
Quoted 7 lines

Changing reset_cpu() to do a full system reset fixes this issue.

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)
Quoted 16 lines
arch/x86/cpu/baytrail/valleyview.c +6 -1
index 215e0d0..a009c14 100644
@@ -65,3 +65,9 @@ int reserve_arch(void)
 #endif
 }
 #endif
+
+void reset_cpu(ulong addr)
+{
+       /* cold reset */
+       x86_full_reset();
+}
--
Quoted 2 lines

1.8.2.1

Actually on Minnowmax this doesn't fix the hang.

Quoted 6 lines
Quoted 2 lines

That's strange. As full reset indicates a complete power cycle. Does MinnowMax boot with MRC cache enabled?

It's hard to tell because the UART does not work that early. I wish we could fix that. I thought I did have it running at one point, but I may have imagined it.

We need figure out where the hang happens. Is it in FSP, or in U-Boot? You can use a debug version of FSP (the gold4 release includes a debug version FSP) to see where the FSP hangs if it is the FSP hang case.

Quoted 6 lines

Re the hang, I'm not sure - it seems to have stopped happening, so let's not worry about it for now.

Quoted 1 line
Quoted 1 line
Quoted 8 lines
Quoted 5 lines
Quoted 1 line

Also I don't see any speed-up from using the cache on Minnowmax.

On Galileo, I did not see any speed-up either. I think it is because on MinnowMax and Galileo they both uses memory-down configuration. The speed-up seems only helpful for DIMMs (on Intel Bayley Bay)

So what is the point of enabling the MRC cache?

On Bayley Bay it indeed improves the boot time. So, maybe we can just leave this option unselected by the defconfig files for Chromebook and Galileo?

One thing I noticed that on Chromebook the memory SPD is passed by device tree. Is there any DIMM on the Chromebook and the SPD in device tree is just a short-cut for the MRC? If we let MRC reads the SPD, does MRC cache help?

Regards,
Bin

#39 lore raw

PATCH 11/13 x86: baytrail: Issue full system reset in reset_cpu()

Hi Bin,

Quoted 58 lines · Bin Meng

Hi Simon,

Quoted 55 lines · Simon Glass

Hi Bin,

Quoted 51 lines · Bin Meng

Hi Simon,

Quoted 37 lines · Simon Glass

Hi Bin,

Quoted 12 lines · Bin Meng

With MRC cache enabled, when typing 'reset' in the U-Boot shell, BayTrail FSP initialization hangs at "Configuring Memory Start":

  Setting BootMode to 0
  Install PPI: 1F4C6F90-B06B-48D8-A201-BAE5F1CD7D56
  Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
  About to call MrcInit();
  BayleyBay Platform Type
  CurrentMrcData.BootMode = 4
  Taking Fastboot path!
  Configuring Memory Start...
Quoted 7 lines

Changing reset_cpu() to do a full system reset fixes this issue.

---

 arch/x86/cpu/baytrail/valleyview.c | 6 ++++++
 1 file changed, 6 insertions(+)
Quoted 12 lines
arch/x86/cpu/baytrail/valleyview.c +3 -0
index 215e0d0..a009c14 100644
@@ -65,3 +65,9 @@ int reserve_arch(void)
 #endif
 }
 #endif
+
+void reset_cpu(ulong addr)
+{
Quoted 8 lines
Quoted 6 lines
arch/x86/cpu/baytrail/valleyview.c +3 -1
+       /* cold reset */
+       x86_full_reset();
+}
--

1.8.2.1

Actually on Minnowmax this doesn't fix the hang.

That's strange. As full reset indicates a complete power cycle. Does MinnowMax boot with MRC cache enabled?

Quoted 7 lines
Quoted 3 lines

It's hard to tell because the UART does not work that early. I wish we could fix that. I thought I did have it running at one point, but I may have imagined it.

We need figure out where the hang happens. Is it in FSP, or in U-Boot? You can use a debug version of FSP (the gold4 release includes a debug version FSP) to see where the FSP hangs if it is the FSP hang case.

Ah yes, that's a good idea.

Quoted 13 lines
Quoted 13 lines

Re the hang, I'm not sure - it seems to have stopped happening, so let's not worry about it for now.

Quoted 6 lines
Quoted 2 lines

Also I don't see any speed-up from using the cache on Minnowmax.

On Galileo, I did not see any speed-up either. I think it is because on MinnowMax and Galileo they both uses memory-down configuration. The speed-up seems only helpful for DIMMs (on Intel Bayley Bay)

Quoted 6 lines
Quoted 2 lines

So what is the point of enabling the MRC cache?

On Bayley Bay it indeed improves the boot time. So, maybe we can just leave this option unselected by the defconfig files for Chromebook and Galileo?

OK. It does speed up link.

Quoted 5 lines

One thing I noticed that on Chromebook the memory SPD is passed by device tree. Is there any DIMM on the Chromebook and the SPD in device tree is just a short-cut for the MRC? If we let MRC reads the SPD, does MRC cache help?

Yes there are DIMMs but we know exactly what they are. The MRC cache does seem to help on this platform.

Regards,
Simon

author
#40 lore raw

PATCH 12/13 x86: Remove unused rw-mrc-cache properties in the link and panther dts files

In reply to Bin Meng

"type" and "wipe-value" are never used by the mrccache codes. Remove them to avoid confusion. This also removes the alignment comment in the panther dts file.

---

 arch/x86/dts/chromebook_link.dts   | 2 --
 arch/x86/dts/chromebox_panther.dts | 3 ---
 2 files changed, 5 deletions(-)
arch/x86/dts/chromebook_link.dts +0 -2
index 4291141..f27263a 100644
@@ -208,8 +208,6 @@
 					rw-mrc-cache {
 						label = "rw-mrc-cache";
 						reg = <0x003e0000 0x00010000>;
-						type = "wiped";
-						wipe-value = [ff];
 					};
 				};
 			};
arch/x86/dts/chromebox_panther.dts +0 -3
index c60ab71..61e8f2f 100644
@@ -64,10 +64,7 @@
 			memory-map = <0xff800000 0x00800000>;
 			rw-mrc-cache {
 				label = "rw-mrc-cache";
-				/* Alignment: 4k (for updating) */
 				reg = <0x003e0000 0x00010000>;
-				type = "wiped";
-				wipe-value = [ff];
 			};
 		};
 	};
Signature · 1 line
1.8.2.1
#41 lore raw

PATCH 12/13 x86: Remove unused rw-mrc-cache properties in the link and panther dts files

Hi Bin,

Quoted 10 lines · Bin Meng

"type" and "wipe-value" are never used by the mrccache codes. Remove them to avoid confusion. This also removes the alignment comment in the panther dts file.

---

 arch/x86/dts/chromebook_link.dts   | 2 --
 arch/x86/dts/chromebox_panther.dts | 3 ---
 2 files changed, 5 deletions(-)

These are used by the cros_bundle_firmware tool in Chrome OS, which builds the image. But it does not matter for upstream.

Regards,
Simon

#42 lore raw

PATCH 12/13 x86: Remove unused rw-mrc-cache properties in the link and panther dts files

Quoted 14 lines · Simon Glass

Hi Bin,

Quoted 10 lines · Bin Meng

"type" and "wipe-value" are never used by the mrccache codes. Remove them to avoid confusion. This also removes the alignment comment in the panther dts file.

---

 arch/x86/dts/chromebook_link.dts   | 2 --
 arch/x86/dts/chromebox_panther.dts | 3 ---
 2 files changed, 5 deletions(-)
Quoted 7 lines

These are used by the cros_bundle_firmware tool in Chrome OS, which builds the image. But it does not matter for upstream.

Regards,
Simon

Applied to u-boot-x86, thanks!

author
#43 lore raw

PATCH 13/13 x86: ivybridge: Correct two typos for MRC

In reply to Bin Meng

It should be MRC, not MCR.

---

 arch/x86/cpu/ivybridge/sdram.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/cpu/ivybridge/sdram.c +2 -2
index 9121426..fc66a3c 100644
@@ -363,7 +363,7 @@ int sdram_initialise(struct pei_data *pei_data)
 	debug("System Agent Version %d.%d.%d Build %d\n",
 	      version >> 24 , (version >> 16) & 0xff,
 	      (version >> 8) & 0xff, version & 0xff);
-	debug("MCR output data length %#x at %p\n", pei_data->mrc_output_len,
+	debug("MRC output data length %#x at %p\n", pei_data->mrc_output_len,
 	      pei_data->mrc_output);
 
 	/*
@@ -723,7 +723,7 @@ int dram_init(void)
 	int ret;
 
 	debug("Boot mode %d\n", gd->arch.pei_boot_mode);
-	debug("mcr_input %p\n", pei_data.mrc_input);
+	debug("mrc_input %p\n", pei_data.mrc_input);
 	pei_data.boot_mode = gd->arch.pei_boot_mode;
 	ret = copy_spd(&pei_data);
 	if (!ret)
Signature · 1 line
1.8.2.1
author
#46 lore raw

cover x86: Provide a common MRC cache library and enable it for FSP

In reply to Bin Meng

Hi Simon,

Quoted 18 lines · Bin Meng

This series moves existing MRC cache codes in the ivybridge cpu diretory to a common place and makes some changes so that every x86 board benefits from it. It also updates FSP support codes to pass MRC cache data to fsp_init() to speed up boot time.

Currently tested on Intel Bayley Bay board. Minnow Max board should work as it uses the same FSP that Bayley Bay uses. For Chromebook I cannot test that, but I see Simon added a TODO in the codes below:

        /*
         * TODO(sjg at chromium.org): Skip this for now as it causes boot
         * problems
         */
        if (0) {
                pei_data->mrc_input = mrc_cache->data;
                pei_data->mrc_input_len = mrc_cache->data_size;
        }
Quoted 4 lines

I have no idea what breaks the MRC cache on the Chromebook board. Simon, you may retest this on top of this series to see if things go better.

Can you please try this patch on Chromebook?
http://patchwork.ozlabs.org/patch/528970/

I suspect this patch fixes the boot failures you saw before.

[snip]

Regards,
Bin

#47 lore raw

cover x86: Provide a common MRC cache library and enable it for FSP

Hi Bin,

Quoted 26 lines · Bin Meng

Hi Simon,

Quoted 18 lines · Bin Meng

This series moves existing MRC cache codes in the ivybridge cpu diretory to a common place and makes some changes so that every x86 board benefits from it. It also updates FSP support codes to pass MRC cache data to fsp_init() to speed up boot time.

Currently tested on Intel Bayley Bay board. Minnow Max board should work as it uses the same FSP that Bayley Bay uses. For Chromebook I cannot test that, but I see Simon added a TODO in the codes below:

        /*
         * TODO(sjg at chromium.org): Skip this for now as it causes boot
         * problems
         */
        if (0) {
                pei_data->mrc_input = mrc_cache->data;
                pei_data->mrc_input_len = mrc_cache->data_size;
        }
Quoted 4 lines

I have no idea what breaks the MRC cache on the Chromebook board. Simon, you may retest this on top of this series to see if things go better.

Quoted 6 lines

Can you please try this patch on Chromebook?
http://patchwork.ozlabs.org/patch/528970/

I suspect this patch fixes the boot failures you saw before.

[snip]

Thanks for the prod - actually the problem turns out to be in car_uninit() so I will send a series to fix that, and some RTC problems.

Regards,
Simon