diff --git a/src/include/b64_decode.h b/src/include/b64_decode.h
new file mode 100644
index 0000000..b527196
--- /dev/null
+++ b/src/include/b64_decode.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2015 Google, Inc.
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * 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.
+ */
+
+#ifndef __INCLUDE_B64_DECODE_H__
+#define __INCLUDE_B64_DECODE_H__
+
+#include <stddef.h>
+#include <stdint.h>
+
+/*
+ * A function to convert a buffer of base64 format data into its source.
+ *
+ * The user provides output buffer of the size guaranteed to fit the result.
+ *
+ * Returns the size of the decoded data or zero if invalid charactes were
+ * encountered in the input buffer.
+ */
+size_t b64_decode(const uint8_t *input_data,
+		  size_t input_length,
+		  uint8_t *output_data);
+
+/* A macro to derive decoded size of a base64 encoded blob. */
+#define B64_DECODED_SIZE(encoded_size) (((encoded_size) * 3)/4)
+
+#endif
diff --git a/src/lib/Makefile.inc b/src/lib/Makefile.inc
index d4f0452..19ff2b2 100644
--- a/src/lib/Makefile.inc
+++ b/src/lib/Makefile.inc
@@ -97,6 +97,7 @@
 ramstage-$(CONFIG_TIMER_QUEUE) += timer_queue.c
 ramstage-$(CONFIG_GENERIC_GPIO_LIB) += gpio.c
 ramstage-$(CONFIG_GENERIC_UDELAY) += timer.c
+ramstage-y += b64_decode.c
 
 romstage-y += cbmem_common.c dynamic_cbmem.c
 ramstage-y += cbmem_common.c dynamic_cbmem.c
diff --git a/src/lib/b64_decode.c b/src/lib/b64_decode.c
new file mode 100644
index 0000000..85f43c0
--- /dev/null
+++ b/src/lib/b64_decode.c
@@ -0,0 +1,146 @@
+/*
+ * Copyright (C) 2015 Google, Inc.
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * 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 <b64_decode.h>
+#include <console/console.h>
+
+/*
+ * Translation Table to decode base64 ASCII stream into binary. Borrowed from
+ *
+ * http://base64.sourceforge.net/b64.c.
+ *
+ */
+static const char cd64[] = "|$$$}rstuvwxyz{$$$$$$$>?@ABCDEFGHIJKLMN"
+	"OPQRSTUVW$$$$$$XYZ[\\]^_`abcdefghijklmnopq";
+
+struct buffer_descriptor {
+	const uint8_t *input_buffer;
+	size_t data_size;
+	size_t input_index;
+};
+
+#define isalnum(c) ((((c) >= 'a') && ((c) <= 'z')) || \
+		    (((c) >= 'A') && ((c) <= 'Z')) || \
+		    (((c) >= '0') && ((c) <= '9')))
+
+/*
+ * On each invocation this function returns the next valid base64 character
+ * from the encoded message, ignoring padding and line breaks.
+ *
+ * Once all input is consumed, 0 is returned on all following invocations. In
+ * case any other than expected characters is found in the encoded message, -1
+ * is returned for error.
+ */
+static int get_next_char(struct buffer_descriptor *bd)
+{
+	uint8_t c;
+
+	/*
+	 * The canonical base64 encoded messages include the following
+	 * characters:
+	 * - '0..9A..Za..z+/' to represent 64 values
+	 * - '=' for padding
+	 * - '<CR><LF>' to split the message into lines.
+	 */
+	while (bd->input_index < bd->data_size) {
+		c = bd->input_buffer[bd->input_index++];
+
+		switch (c) {
+		case '=':
+		case 0xa:
+		case 0xd:
+			continue;
+
+		default:
+			break;
+		}
+
+		if (!isalnum(c) && (c != '+') && (c != '/'))
+			return -1;
+
+		return c;
+	}
+
+	return 0;
+}
+
+/*
+** decode
+**
+** decode a base64 encoded stream discarding padding and line breaks.
+*/
+size_t b64_decode(const uint8_t *input_data,
+		  size_t input_length,
+		  uint8_t *output_data)
+{
+	struct buffer_descriptor bd;
+	unsigned interim = 0;
+	size_t output_size = 0;
+	/* count of processed input bits, modulo log2(64) */
+	unsigned bit_count = 0;
+
+	/*
+	 * Keep the context on the stack to make things easier if this needs
+	 * to run with CAR.
+	 */
+	bd.input_buffer = input_data;
+	bd.data_size = input_length;
+	bd.input_index = 0;
+
+	while (1) { /* Until input is exausted. */
+		int v = get_next_char(&bd);
+
+		if (v < 0) {
+			printk(BIOS_ERR,
+			       "Incompatible character at offset %zd.\n",
+			       bd.input_index);
+			return 0;
+		}
+
+		if (!v)
+			break;
+
+		/*
+		 * v is guaranteed to be in the proper range for cd64, the
+		 * result is a 6 bit number.
+		 */
+		v = cd64[v - 43] - 62;
+
+		if (bit_count >= 2) {
+			/*
+			 * Once 6 more bits are added to the output, there is
+			 * going to be at least a full byte.
+			 *
+			 * 'remaining_bits' is the exact number of bits which
+			 * need to be added to the output to have another full
+			 * byte ready.
+			 */
+			int remaining_bits = 8 - bit_count;
+
+			interim <<= remaining_bits;
+			interim |= v >> (6 - remaining_bits);
+
+			/* Pass the new full byte to the output. */
+			output_data[output_size++] = interim & 0xff;
+
+			interim = v;
+			bit_count -= 2;
+		} else {
+			interim <<= 6;
+			interim |= v;
+			bit_count += 6;
+		}
+	}
+
+	return output_size;
+}