lib: Add FIT payload support

* Add support for parsing and booting FIT payloads.
* Build fit loader code from depthcharge.
* Fix coding style.
* Add Kconfig option to add compiletime support for FIT.
* Add support for initrd.
* Add default compat strings
* Apply optional devicetree fixups using dt_apply_fixups

Starting at this point the CBFS payload/ can be either SELF or FIT.

Tested on Cavium SoC: Parses and loads a Linux kernel 4.16.3.
Tested on Cavium SoC: Parses and loads a Linux kernel 4.15.0.
Tested on Cavium SoC: Parses and loads a Linux kernel 4.1.52.

Change-Id: I0f27b92a5e074966f893399eb401eb97d784850d
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/25019
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Philipp Deppenwiese <zaolin.daisuki@gmail.com>

2 files changed, 263 insertions, 0 deletions

diff --git a/src/arch/arm64/Makefile.inc b/src/arch/arm64/Makefile.inc
index f57ef720b9..3c3cd20370 100644
--- a/src/arch/arm64/Makefile.inc
+++ b/src/arch/arm64/Makefile.inc
@@ -136,6 +136,7 @@ ramstage-y += memcpy.S
 ramstage-y += memmove.S
 ramstage-$(CONFIG_ARM64_USE_ARM_TRUSTED_FIRMWARE) += arm_tf.c
 ramstage-y += transition.c transition_asm.S
+ramstage-$(CONFIG_PAYLOAD_FIT_SUPPORT) += fit_payload.c
 
 rmodules_arm64-y += memset.S
 rmodules_arm64-y += memcpy.S
diff --git a/src/arch/arm64/fit_payload.c b/src/arch/arm64/fit_payload.c
new file mode 100644
index 0000000000..c4bbceee4c
--- /dev/null
+++ b/src/arch/arm64/fit_payload.c
@@ -0,0 +1,262 @@
+/*
+ * Copyright 2013 Google Inc.
+ * Copyright 2018 Facebook, Inc.
+ *
+ * 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; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * 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.
+ */
+
+#include <console/console.h>
+#include <bootmem.h>
+#include <stdlib.h>
+#include <program_loading.h>
+#include <string.h>
+#include <commonlib/compression.h>
+#include <commonlib/cbfs_serialized.h>
+#include <lib.h>
+#include <fit.h>
+#include <endian.h>
+
+#define MAX_KERNEL_SIZE (64*MiB)
+
+struct arm64_kernel_header {
+	u32 code0;
+	u32 code1;
+	u64 text_offset;
+	u64 image_size;
+	u64 flags;
+	u64 res2;
+	u64 res3;
+	u64 res4;
+	u32 magic;
+#define KERNEL_HEADER_MAGIC  0x644d5241
+	u32 res5;
+};
+
+static struct {
+	union {
+		struct arm64_kernel_header header;
+		u8 raw[sizeof(struct arm64_kernel_header) + 0x100];
+	};
+#define SCRATCH_CANARY_VALUE 0xdeadbeef
+	u32 canary;
+} scratch;
+
+/* Returns true if decompressing was successful and it looks like a kernel. */
+static bool decompress_kernel_header(const struct fit_image_node *node)
+{
+	/* Partially decompress to get text_offset. Can't check for errors. */
+	scratch.canary = SCRATCH_CANARY_VALUE;
+	switch (node->compression) {
+	case CBFS_COMPRESS_NONE:
+		memcpy(scratch.raw, node->data, sizeof(scratch.raw));
+		break;
+	case CBFS_COMPRESS_LZMA:
+		ulzman(node->data, node->size,
+		       scratch.raw, sizeof(scratch.raw));
+		break;
+	case CBFS_COMPRESS_LZ4:
+		ulz4fn(node->data, node->size,
+		       scratch.raw, sizeof(scratch.raw));
+		break;
+	default:
+		printk(BIOS_ERR, "ERROR: Unsupported compression algorithm!\n");
+		return false;
+	}
+
+	/* Should never happen, but if it does we'll want to know. */
+	if (scratch.canary != SCRATCH_CANARY_VALUE)
+		die("ERROR: Partial decompression ran over scratchbuf!\n");
+
+	if (scratch.header.magic != KERNEL_HEADER_MAGIC) {
+		printk(BIOS_ERR,
+		       "ERROR: Invalid kernel magic: %#.8x\n != %#.8x\n",
+		       scratch.header.magic, KERNEL_HEADER_MAGIC);
+		return false;
+	}
+
+	/**
+	 * Prior to v3.17, the endianness of text_offset was not specified.  In
+	 * these cases image_size is zero and text_offset is 0x80000 in the
+	 * endianness of the kernel.  Where image_size is non-zero image_size is
+	 * little-endian and must be respected.  Where image_size is zero,
+	 * text_offset can be assumed to be 0x80000.
+	 */
+	if (!scratch.header.image_size)
+		scratch.header.text_offset = cpu_to_le64(0x80000);
+
+	return true;
+}
+
+static size_t get_kernel_size(const struct fit_image_node *node)
+{
+	if (scratch.header.image_size)
+		return le64_to_cpu(scratch.header.image_size);
+
+	/**
+	 * When image_size is zero, a bootloader should attempt to keep as much
+	 * memory as possible free for use by the kernel immediately after the
+	 * end of the kernel image. The amount of space required will vary
+	 * depending on selected features, and is effectively unbound.
+	 */
+
+	printk(BIOS_WARNING, "FIT: image_size not set in kernel header.\n"
+	       "Leaving additional %u MiB of free space after kernel.\n",
+	       MAX_KERNEL_SIZE >> 20);
+
+	return node->size + MAX_KERNEL_SIZE;
+}
+
+static bool fit_place_kernel(const struct range_entry *r, void *arg)
+{
+	struct region *region = arg;
+	resource_t start;
+
+	if (range_entry_tag(r) != BM_MEM_RAM)
+		return true;
+
+	/**
+	 * The Image must be placed text_offset bytes from a 2MB aligned base
+	 * address anywhere in usable system RAM and called there. The region
+	 * between the 2 MB aligned base address and the start of the image has
+	 * no special significance to the kernel, and may be used for other
+	 * purposes.
+	 *
+	 * If the reserved memory (BL31 for example) is smaller than text_offset
+	 * we can use the 2 MiB base address, otherwise use the next 2 MiB page.
+	 * It's not mandatory, but wastes less memory below the kernel.
+	 */
+	start = ALIGN_DOWN(range_entry_base(r), 2 * MiB) +
+		le64_to_cpu(scratch.header.text_offset);
+
+	if (start < range_entry_base(r))
+		start += 2 * MiB;
+	/**
+	 * At least image_size bytes from the start of the image must be free
+	 * for use by the kernel.
+	 */
+	if (start + region->size < range_entry_end(r)) {
+		region->offset = (size_t)start;
+		return false;
+	}
+
+	return true;
+}
+
+/**
+ * Place the region in free memory range.
+ *
+ * The caller has to set region->offset to the minimum allowed address.
+ * The region->offset is usually 0 on kernel >v4.6 and kernel_base + kernel_size
+ * on kernel <v4.6.
+ */
+static bool fit_place_mem(const struct range_entry *r, void *arg)
+{
+	struct region *region = arg;
+	resource_t start;
+
+	if (range_entry_tag(r) != BM_MEM_RAM)
+		return true;
+
+	/* Linux 4.15 doesn't like 4KiB alignment. Align to 1 MiB for now. */
+	start = ALIGN_UP(MAX(region->offset, range_entry_base(r)), 1 * MiB);
+
+	if (start + region->size < range_entry_end(r)) {
+		region->offset = (size_t)start;
+		return false;
+	}
+
+	return true;
+}
+
+bool fit_payload_arch(struct prog *payload, struct fit_config_node *config,
+		      struct region *kernel,
+		      struct region *fdt,
+		      struct region *initrd)
+{
+	bool place_anywhere;
+	void *arg = NULL;
+
+	if (!config->fdt || !fdt) {
+		printk(BIOS_CRIT, "CRIT: Providing a valid FDT is mandatory to "
+		       "boot an ARM64 kernel!\n");
+		return false;
+	}
+
+	if (!decompress_kernel_header(config->kernel_node)) {
+		printk(BIOS_CRIT, "CRIT: Payload doesn't look like an ARM64"
+		       " kernel Image.\n");
+		return false;
+	}
+
+	/* Update kernel size from image header, if possible */
+	kernel->size = get_kernel_size(config->kernel_node);
+	printk(BIOS_DEBUG, "FIT: Using kernel size of 0x%zx bytes\n",
+	       kernel->size);
+
+	/**
+	 * The code assumes that bootmem_walk provides a sorted list of memory
+	 * regions, starting from the lowest address.
+	 * The order of the calls here doesn't matter, as the placement is
+	 * enforced in the called functions.
+	 * For details check code on top.
+	 */
+
+	if (!bootmem_walk(fit_place_kernel, kernel))
+		return false;
+
+	/* Mark as reserved for future allocations. */
+	bootmem_add_range(kernel->offset, kernel->size, BM_MEM_PAYLOAD);
+
+	/**
+	 * NOTE: versions prior to v4.6 cannot make use of memory below the
+	 * physical offset of the Image so it is recommended that the Image be
+	 * placed as close as possible to the start of system RAM.
+	 *
+	 * For kernel <v4.6 the INITRD and FDT can't be placed below the kernel.
+	 * In that case set region offset to an address on top of kernel.
+	 */
+	place_anywhere = !!(le64_to_cpu(scratch.header.flags) & (1 << 3));
+	printk(BIOS_DEBUG, "FIT: Placing FDT and INITRD %s\n",
+	       place_anywhere ? "anywhere" : "on top of kernel");
+
+	/* Place INITRD */
+	if (config->ramdisk) {
+		if (place_anywhere)
+			initrd->offset = 0;
+		else
+			initrd->offset = kernel->offset + kernel->size;
+
+		if (!bootmem_walk(fit_place_mem, initrd))
+			return false;
+		/* Mark as reserved for future allocations. */
+		bootmem_add_range(initrd->offset, initrd->size, BM_MEM_PAYLOAD);
+	}
+
+	/* Place FDT */
+	if (place_anywhere)
+		fdt->offset = 0;
+	else
+		fdt->offset = kernel->offset + kernel->size;
+
+	if (!bootmem_walk(fit_place_mem, fdt))
+		return false;
+	/* Mark as reserved for future allocations. */
+	bootmem_add_range(fdt->offset, fdt->size, BM_MEM_PAYLOAD);
+
+	/* Kernel expects FDT as argument */
+	arg = (void *)fdt->offset;
+
+	prog_set_entry(payload, (void *)kernel->offset, arg);
+
+	bootmem_dump_ranges();
+
+	return true;
+}