lynxpoint: Add cbfs_load_payload() implementation

SPI accesses can be slow depending on the setup and the access pattern.
The current SPI hardware setup to cache and prefetch. The alternative
cbfs_load_payload() function takes advantage of the caching in the CPU
because the ROM is cached as write protected as well as the SPI's
hardware's caching/prefetching implementation. The CPU will fetch
consecutive aligned cachelines which will hit the ROM as
cacheline-aligned addresses. Once the payload is mirrored into RAM the
segment loading can take place by reading RAM instead of ROM.

With the alternative cbfs_load_payload() the boot time on a baskingridge
board saves ~100ms. This savings is observed using cbmem.py after
performing warm reboots and looking at TS_SELFBOOT_JUMP (99) entries.
This is booting with a depthcharge payload whose payload file fits
within the SMM_DEFAULT_SIZE (0x10000 bytes).

Datapoints with TS_LOAD_PAYLOAD (90) & TS_SELFBOOT_JUMP (99) cbmem entries:

Baseline                          Alt
--------                          --------
90:3,859,310  (473)               90:3,863,647  (454)
99:3,989,578  (130,268)           99:3,888,709  (25,062)

90:3,899,450  (477)               90:3,860,926  (463)
99:4,029,459  (130,008)           99:3,890,583  (29,657)

90:3,834,600  (466)               90:3,890,564  (465)
99:3,964,535  (129,934)           99:3,920,213  (29,649)

Booted baskingridge many times and observed 100ms reduction in
TS_SELFBOOT_JUMP times (time to load payload).

Change-Id: I27b2dec59ecd469a4906b4179b39928e9201db81
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: http://review.coreboot.org/2783
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>

1 files changed, 85 insertions, 0 deletions

diff --git a/src/southbridge/intel/lynxpoint/spi_loading.c b/src/southbridge/intel/lynxpoint/spi_loading.c
new file mode 100644
index 0000000000..eb3082bb5a
--- /dev/null
+++ b/src/southbridge/intel/lynxpoint/spi_loading.c
@@ -0,0 +1,85 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2013 ChromeOS Authors
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+
+#include <stdlib.h>
+#include <string.h>
+#include <arch/byteorder.h>
+#include <cbfs.h>
+#include <console/console.h>
+#include <cpu/x86/smm.h>
+
+#define CACHELINE_SIZE 64
+#define INTRA_CACHELINE_MASK (CACHELINE_SIZE - 1)
+#define CACHELINE_MASK (~INTRA_CACHELINE_MASK)
+
+/* Mirror the payload file to the default SMM location if it is small enough.
+ * The default SMM region can be used since no one is using the memory at this
+ * location at this stage in the boot. */
+static inline void *spi_mirror(void *file_start, int file_len)
+{
+	int alignment_diff;
+	char *src;
+	char *dest = (void *)SMM_DEFAULT_BASE;
+
+	alignment_diff = (INTRA_CACHELINE_MASK & (long)file_start);
+
+	/* Adjust file length so that the start and end points are aligned to a
+	 * cacheline. Coupled with the ROM caching in the CPU the SPI hardware
+	 * will read and cache full length cachelines. It will also prefetch
+	 * data as well. Once things are mirrored in memory all accesses should
+	 * hit the CPUs cache. */
+	file_len += alignment_diff;
+	file_len = ALIGN(file_len, CACHELINE_SIZE);
+
+	printk(BIOS_DEBUG, "Payload aligned size: 0x%x\n", file_len);
+
+	/* Just pass back the pointer to ROM space if the file is larger
+	 * than the RAM mirror region. */
+	if (file_len > SMM_DEFAULT_SIZE)
+		return file_start;
+
+	src = (void *)(CACHELINE_MASK & (long)file_start);
+	/* Note that if mempcy is not using 32-bit moves the performance will
+	 * degrade because the SPI hardware prefetchers look for
+	 * cacheline-aligned 32-bit accesses to kick in. */
+	memcpy(dest, src, file_len);
+
+	/* Provide pointer into mirrored space. */
+	return &dest[alignment_diff];
+}
+
+void *cbfs_load_payload(struct cbfs_media *media, const char *name)
+{
+	int file_len;
+	void *file_start;
+	struct cbfs_file *file = cbfs_get_file(media, name);
+
+	if (file == NULL)
+		return NULL;
+
+	if (ntohl(file->type) != CBFS_TYPE_PAYLOAD)
+		return NULL;
+
+	file_len = ntohl(file->len);
+
+	file_start = CBFS_SUBHEADER(file);
+
+	return spi_mirror(file_start, file_len);
+}