diff --git a/arch/powerpc/boot/Makefile b/arch/powerpc/boot/Makefile
index bede555d78cf..861348c72519 100644
--- a/arch/powerpc/boot/Makefile
+++ b/arch/powerpc/boot/Makefile
@@ -63,13 +63,28 @@ $(obj)/treeboot-currituck.o: BOOTCFLAGS += -mcpu=405
 $(obj)/treeboot-akebono.o: BOOTCFLAGS += -mcpu=405
 $(obj)/virtex405-head.o: BOOTAFLAGS += -mcpu=405
 
-# the kernel's version of zlib pulls in a lot of other kernel headers
-# which we don't provide inside the wrapper.
+# The pre-boot decompressors pull in a lot of kernel headers and other source
+# files. This creates a bit of a dependency headache since we need to copy
+# these files into the build dir, fix up any includes and ensure that dependent
+# files are copied in the right order.
+
+# these need to be seperate variables because they are copied out of different
+# directories in the kernel tree. Sure you COULd merge them, but it's a
+# cure-is-worse-than-disease situation.
+zlib-decomp-$(CONFIG_KERNEL_GZIP) := decompress_inflate.c
 zlib-$(CONFIG_KERNEL_GZIP) := inffast.c inflate.c inftrees.c
 zlibheader-$(CONFIG_KERNEL_GZIP) := inffast.h inffixed.h inflate.h inftrees.h infutil.h
 zliblinuxheader-$(CONFIG_KERNEL_GZIP) := zlib.h zconf.h zutil.h
 
-$(addprefix $(obj)/,$(zlib-y) cuboot-c2k.o gunzip_util.o main.o): \
+$(addprefix $(obj)/, decompress.o): \
+	$(addprefix $(obj)/,$(zlib-decomp-y))
+
+$(addprefix $(obj)/, $(zlib-decomp-y)): \
+	$(addprefix $(obj)/,$(zliblinuxheader-y)) \
+	$(addprefix $(obj)/,$(zlibheader-y)) \
+	$(addprefix $(obj)/,$(zlib-y))
+
+$(addprefix $(obj)/,$(zlib-y)): \
 	$(addprefix $(obj)/,$(zliblinuxheader-y)) \
 	$(addprefix $(obj)/,$(zlibheader-y))
 
@@ -79,10 +94,10 @@ libfdtheader := fdt.h libfdt.h libfdt_internal.h
 $(addprefix $(obj)/,$(libfdt) libfdt-wrapper.o simpleboot.o epapr.o opal.o): \
 	$(addprefix $(obj)/,$(libfdtheader))
 
-src-wlib-y := string.S crt0.S crtsavres.S stdio.c main.c \
+src-wlib-y := string.S crt0.S crtsavres.S stdio.c decompress.c main.c \
 		$(libfdt) libfdt-wrapper.c \
 		ns16550.c serial.c simple_alloc.c div64.S util.S \
-		gunzip_util.c elf_util.c $(zlib-y) devtree.c stdlib.c \
+		elf_util.c $(zlib-y) devtree.c stdlib.c \
 		oflib.c ofconsole.c cuboot.c mpsc.c cpm-serial.c \
 		uartlite.c mpc52xx-psc.c opal.c opal-calls.S
 src-wlib-$(CONFIG_40x) += 4xx.c planetcore.c
@@ -143,6 +158,9 @@ $(addprefix $(obj)/,$(zlibheader-y)): $(obj)/%: $(srctree)/lib/zlib_inflate/%
 $(addprefix $(obj)/,$(zliblinuxheader-y)): $(obj)/%: $(srctree)/include/linux/%
 	$(call cmd,copy_kern_src)
 
+$(addprefix $(obj)/,$(zlib-decomp-y)): $(obj)/%: $(srctree)/lib/%
+	$(call cmd,copy_kern_src)
+
 quiet_cmd_copy_libfdt = COPY    $@
       cmd_copy_libfdt = cp $< $@
 
@@ -160,7 +178,7 @@ $(obj)/zImage.coff.lds $(obj)/zImage.ps3.lds : $(obj)/%: $(srctree)/$(src)/%.S
 	$(Q)cp $< $@
 
 clean-files := $(zlib-) $(zlibheader-) $(zliblinuxheader-) \
-		$(libfdt) $(libfdtheader) \
+		$(zlib-decomp-) $(libfdt) $(libfdtheader) \
 		empty.c zImage.coff.lds zImage.ps3.lds zImage.lds
 
 quiet_cmd_bootcc = BOOTCC  $@
@@ -410,8 +428,8 @@ clean-files += $(image-) $(initrd-) cuImage.* dtbImage.* treeImage.* \
 	zImage.maple simpleImage.* otheros.bld *.dtb
 
 # clean up files cached by wrapper
-clean-kernel := vmlinux.strip vmlinux.bin
-clean-kernel += $(addsuffix .gz,$(clean-kernel))
+clean-kernel-base := vmlinux.strip vmlinux.bin
+clean-kernel := $(addsuffix .gz,$(clean-kernel-base))
 # If not absolute clean-files are relative to $(obj).
 clean-files += $(addprefix $(objtree)/, $(clean-kernel))
 
diff --git a/arch/powerpc/boot/decompress.c b/arch/powerpc/boot/decompress.c
new file mode 100644
index 000000000000..3b414d425f16
--- /dev/null
+++ b/arch/powerpc/boot/decompress.c
@@ -0,0 +1,143 @@
+/*
+ * Wrapper around the kernel's pre-boot decompression library.
+ *
+ * Copyright (C) IBM Corporation 2016.
+ *
+ * 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.
+ */
+
+#include "elf.h"
+#include "page.h"
+#include "string.h"
+#include "stdio.h"
+#include "ops.h"
+#include "reg.h"
+#include "types.h"
+
+/*
+ * The decompressor_*.c files play #ifdef games so they can be used in both
+ * pre-boot and regular kernel code. We need these definitions to make the
+ * includes work.
+ */
+
+#define STATIC static
+#define INIT
+#define __always_inline inline
+
+/*
+ * The build process will copy the required zlib source files and headers
+ * out of lib/ and "fix" the includes so they do not pull in other kernel
+ * headers.
+ */
+
+#ifdef CONFIG_KERNEL_GZIP
+#	include "decompress_inflate.c"
+#endif
+
+/* globals for tracking the state of the decompression */
+static unsigned long decompressed_bytes;
+static unsigned long limit;
+static unsigned long skip;
+static char *output_buffer;
+
+/*
+ * flush() is called by __decompress() when the decompressor's scratch buffer is
+ * full.
+ */
+static long flush(void *v, unsigned long buffer_size)
+{
+	unsigned long end = decompressed_bytes + buffer_size;
+	unsigned long size = buffer_size;
+	unsigned long offset = 0;
+	char *in = v;
+	char *out;
+
+	/*
+	 * if we hit our decompression limit, we need to fake an error to abort
+	 * the in-progress decompression.
+	 */
+	if (decompressed_bytes >= limit)
+		return -1;
+
+	/* skip this entire block */
+	if (end <= skip) {
+		decompressed_bytes += buffer_size;
+		return buffer_size;
+	}
+
+	/* skip some data at the start, but keep the rest of the block */
+	if (decompressed_bytes < skip && end > skip) {
+		offset = skip - decompressed_bytes;
+
+		in += offset;
+		size -= offset;
+		decompressed_bytes += offset;
+	}
+
+	out = &output_buffer[decompressed_bytes - skip];
+	size = min(decompressed_bytes + size, limit) - decompressed_bytes;
+
+	memcpy(out, in, size);
+	decompressed_bytes += size;
+
+	return buffer_size;
+}
+
+static void print_err(char *s)
+{
+	/* suppress the "error" when we terminate the decompressor */
+	if (decompressed_bytes >= limit)
+		return;
+
+	printf("Decompression error: '%s'\n\r", s);
+}
+
+/**
+ * partial_decompress - decompresses part or all of a compressed buffer
+ * @inbuf:       input buffer
+ * @input_size:  length of the input buffer
+ * @outbuf:      input buffer
+ * @output_size: length of the input buffer
+ * @skip         number of output bytes to ignore
+ *
+ * This function takes compressed data from inbuf, decompresses and write it to
+ * outbuf. Once output_size bytes are written to the output buffer, or the
+ * stream is exhausted the function will return the number of bytes that were
+ * decompressed. Otherwise it will return whatever error code the decompressor
+ * reported (NB: This is specific to each decompressor type).
+ *
+ * The skip functionality is mainly there so the program and discover
+ * the size of the compressed image so that it can ask firmware (if present)
+ * for an appropriately sized buffer.
+ */
+long partial_decompress(void *inbuf, unsigned long input_size,
+	void *outbuf, unsigned long output_size, unsigned long _skip)
+{
+	int ret;
+
+	/*
+	 * The skipped bytes needs to be included in the size of data we want
+	 * to decompress.
+	 */
+	output_size += _skip;
+
+	decompressed_bytes = 0;
+	output_buffer = outbuf;
+	limit = output_size;
+	skip = _skip;
+
+	ret = __decompress(inbuf, input_size, NULL, flush, outbuf,
+		output_size, NULL, print_err);
+
+	/*
+	 * If decompression was aborted due to an actual error rather than
+	 * a fake error that we used to abort, then we should report it.
+	 */
+	if (decompressed_bytes < limit)
+		return ret;
+
+	return decompressed_bytes - skip;
+}
diff --git a/arch/powerpc/boot/main.c b/arch/powerpc/boot/main.c
index d80161b633f4..f7a184b6c35b 100644
--- a/arch/powerpc/boot/main.c
+++ b/arch/powerpc/boot/main.c
@@ -15,11 +15,8 @@
 #include "string.h"
 #include "stdio.h"
 #include "ops.h"
-#include "gunzip_util.h"
 #include "reg.h"
 
-static struct gunzip_state gzstate;
-
 struct addr_range {
 	void *addr;
 	unsigned long size;
@@ -30,15 +27,14 @@ struct addr_range {
 static struct addr_range prep_kernel(void)
 {
 	char elfheader[256];
-	void *vmlinuz_addr = _vmlinux_start;
+	unsigned char *vmlinuz_addr = (unsigned char *)_vmlinux_start;
 	unsigned long vmlinuz_size = _vmlinux_end - _vmlinux_start;
 	void *addr = 0;
 	struct elf_info ei;
-	int len;
+	long len;
 
-	/* gunzip the ELF header of the kernel */
-	gunzip_start(&gzstate, vmlinuz_addr, vmlinuz_size);
-	gunzip_exactly(&gzstate, elfheader, sizeof(elfheader));
+	partial_decompress(vmlinuz_addr, vmlinuz_size,
+		elfheader, sizeof(elfheader), 0);
 
 	if (!parse_elf64(elfheader, &ei) && !parse_elf32(elfheader, &ei))
 		fatal("Error: not a valid PPC32 or PPC64 ELF file!\n\r");
@@ -51,7 +47,7 @@ static struct addr_range prep_kernel(void)
 	 * the kernel bss must be claimed (it will be zero'd by the
 	 * kernel itself)
 	 */
-	printf("Allocating 0x%lx bytes for kernel ...\n\r", ei.memsize);
+	printf("Allocating 0x%lx bytes for kernel...\n\r", ei.memsize);
 
 	if (platform_ops.vmlinux_alloc) {
 		addr = platform_ops.vmlinux_alloc(ei.memsize);
@@ -71,16 +67,21 @@ static struct addr_range prep_kernel(void)
 					"device tree\n\r");
 	}
 
-	/* Finally, gunzip the kernel */
-	printf("gunzipping (0x%p <- 0x%p:0x%p)...", addr,
+	/* Finally, decompress the kernel */
+	printf("Decompressing (0x%p <- 0x%p:0x%p)...\n\r", addr,
 	       vmlinuz_addr, vmlinuz_addr+vmlinuz_size);
-	/* discard up to the actual load data */
-	gunzip_discard(&gzstate, ei.elfoffset - sizeof(elfheader));
-	len = gunzip_finish(&gzstate, addr, ei.loadsize);
+
+	len = partial_decompress(vmlinuz_addr, vmlinuz_size,
+		addr, ei.loadsize, ei.elfoffset);
+
+	if (len < 0)
+		fatal("Decompression failed with error code %ld\n\r", len);
+
 	if (len != ei.loadsize)
-		fatal("ran out of data!  only got 0x%x of 0x%lx bytes.\n\r",
-				len, ei.loadsize);
-	printf("done 0x%x bytes\n\r", len);
+		 fatal("Decompression error: got 0x%lx bytes, expected 0x%lx.\n\r",
+			 len, ei.loadsize);
+
+	printf("Done! Decompressed 0x%lx bytes\n\r", len);
 
 	flush_cache(addr, ei.loadsize);
 
diff --git a/arch/powerpc/boot/ops.h b/arch/powerpc/boot/ops.h
index e19b64ef977a..309d1b127e96 100644
--- a/arch/powerpc/boot/ops.h
+++ b/arch/powerpc/boot/ops.h
@@ -260,4 +260,7 @@ int __ilog2_u32(u32 n)
 	return 31 - bit;
 }
 
+long partial_decompress(void *inbuf, unsigned long input_size, void *outbuf,
+	unsigned long output_size, unsigned long skip);
+
 #endif /* _PPC_BOOT_OPS_H_ */