diff --git a/README b/README
index 5c343da..85e5a8d 100644
--- a/README
+++ b/README
@@ -1392,6 +1392,12 @@ The following options need to be configured:
 		CONFIG_DFU_NAND
 		This enables support for exposing NAND devices via DFU.
 
+		CONFIG_SYS_DFU_DATA_BUF_SIZE
+		Dfu transfer uses a buffer before writing data to the
+		raw storage device. Make the size (in bytes) of this buffer
+		configurable. The size of this buffer is also configurable
+		through the "dfu_bufsiz" environment variable.
+
 		CONFIG_SYS_DFU_MAX_FILE_SIZE
 		When updating files rather than the raw storage device,
 		we use a static buffer to copy the file into and then write
diff --git a/drivers/dfu/dfu.c b/drivers/dfu/dfu.c
index 6af6890..e429d74 100644
--- a/drivers/dfu/dfu.c
+++ b/drivers/dfu/dfu.c
@@ -20,6 +20,7 @@
  */
 
 #include <common.h>
+#include <errno.h>
 #include <malloc.h>
 #include <mmc.h>
 #include <fat.h>
@@ -41,8 +42,34 @@ static int dfu_find_alt_num(const char *s)
 	return ++i;
 }
 
-static unsigned char __aligned(CONFIG_SYS_CACHELINE_SIZE)
-				     dfu_buf[DFU_DATA_BUF_SIZE];
+static unsigned char *dfu_buf;
+static unsigned long dfu_buf_size = CONFIG_SYS_DFU_DATA_BUF_SIZE;
+
+static unsigned char *dfu_free_buf(void)
+{
+	free(dfu_buf);
+	dfu_buf = NULL;
+	return dfu_buf;
+}
+
+static unsigned char *dfu_get_buf(void)
+{
+	char *s;
+
+	if (dfu_buf != NULL)
+		return dfu_buf;
+
+	s = getenv("dfu_bufsiz");
+	dfu_buf_size = s ? (unsigned long)simple_strtol(s, NULL, 16) :
+			CONFIG_SYS_DFU_DATA_BUF_SIZE;
+
+	dfu_buf = memalign(CONFIG_SYS_CACHELINE_SIZE, dfu_buf_size);
+	if (dfu_buf == NULL)
+		printf("%s: Could not memalign 0x%lx bytes\n",
+		       __func__, dfu_buf_size);
+
+	return dfu_buf;
+}
 
 static int dfu_write_buffer_drain(struct dfu_entity *dfu)
 {
@@ -87,8 +114,10 @@ int dfu_write(struct dfu_entity *dfu, void *buf, int size, int blk_seq_num)
 		dfu->offset = 0;
 		dfu->bad_skip = 0;
 		dfu->i_blk_seq_num = 0;
-		dfu->i_buf_start = dfu_buf;
-		dfu->i_buf_end = dfu_buf + sizeof(dfu_buf);
+		dfu->i_buf_start = dfu_get_buf();
+		if (dfu->i_buf_start == NULL)
+			return -ENOMEM;
+		dfu->i_buf_end = dfu_get_buf() + dfu_buf_size;
 		dfu->i_buf = dfu->i_buf_start;
 
 		dfu->inited = 1;
@@ -148,11 +177,12 @@ int dfu_write(struct dfu_entity *dfu, void *buf, int size, int blk_seq_num)
 		printf("\nDFU complete CRC32: 0x%08x\n", dfu->crc);
 
 		/* clear everything */
+		dfu_free_buf();
 		dfu->crc = 0;
 		dfu->offset = 0;
 		dfu->i_blk_seq_num = 0;
 		dfu->i_buf_start = dfu_buf;
-		dfu->i_buf_end = dfu_buf + sizeof(dfu_buf);
+		dfu->i_buf_end = dfu_buf;
 		dfu->i_buf = dfu->i_buf_start;
 
 		dfu->inited = 0;
@@ -229,8 +259,10 @@ int dfu_read(struct dfu_entity *dfu, void *buf, int size, int blk_seq_num)
 		dfu->i_blk_seq_num = 0;
 		dfu->crc = 0;
 		dfu->offset = 0;
-		dfu->i_buf_start = dfu_buf;
-		dfu->i_buf_end = dfu_buf + sizeof(dfu_buf);
+		dfu->i_buf_start = dfu_get_buf();
+		if (dfu->i_buf_start == NULL)
+			return -ENOMEM;
+		dfu->i_buf_end = dfu_get_buf() + dfu_buf_size;
 		dfu->i_buf = dfu->i_buf_start;
 		dfu->b_left = 0;
 
@@ -257,11 +289,12 @@ int dfu_read(struct dfu_entity *dfu, void *buf, int size, int blk_seq_num)
 		debug("%s: %s CRC32: 0x%x\n", __func__, dfu->name, dfu->crc);
 		puts("\nUPLOAD ... done\nCtrl+C to exit ...\n");
 
+		dfu_free_buf();
 		dfu->i_blk_seq_num = 0;
 		dfu->crc = 0;
 		dfu->offset = 0;
 		dfu->i_buf_start = dfu_buf;
-		dfu->i_buf_end = dfu_buf + sizeof(dfu_buf);
+		dfu->i_buf_end = dfu_buf;
 		dfu->i_buf = dfu->i_buf_start;
 		dfu->b_left = 0;
 
diff --git a/include/dfu.h b/include/dfu.h
index a107f4b..124653c 100644
--- a/include/dfu.h
+++ b/include/dfu.h
@@ -68,7 +68,9 @@ static inline unsigned int get_mmc_blk_size(int dev)
 
 #define DFU_NAME_SIZE			32
 #define DFU_CMD_BUF_SIZE		128
-#define DFU_DATA_BUF_SIZE		(1024*1024*8)	/* 8 MiB */
+#ifndef CONFIG_SYS_DFU_DATA_BUF_SIZE
+#define CONFIG_SYS_DFU_DATA_BUF_SIZE		(1024*1024*8)	/* 8 MiB */
+#endif
 #ifndef CONFIG_SYS_DFU_MAX_FILE_SIZE
 #define CONFIG_SYS_DFU_MAX_FILE_SIZE	(4 << 20)	/* 4 MiB */
 #endif