[PATCH v4 0/2] ARM: Embedding Position Independent Executables

[PATCH v4 0/2] ARM: Embedding Position Independent Executables

[Alexandre Belloni]

This series introduces Position Independent Executables (PIEs) for the
ARM architecture.
The main goal is to avoid having to write low level code in assembly as
this is currently the case for suspend/resume. Multiple platforms will
benefit from this infrastructure: at91, rockchip, am335x.

It would also be beneficial for the LPDDR support on at91 to avoid
having the handle the DDR controller from 4 different drivers.

Russell, do you see any remaining issues with the current approach?

Changes in v4:
 - rebased on v4.9-rc1
 - Fixed an issue with the alignment reported by Mylene

Changes in v3:
 - rebased on v4.8-rc1

Changes in v2:
 - handle big endian
 - handle gcov and ftrace by disabling them before compilling the PIE
 - Get the alignment from the original ELF to ensure the PIE is
   properly aligned in SRAM.
 - stop using fncpy
 - rebased on v4.7-rc1


Alexandre Belloni (2):
  ARM: PIE infrastructure
  ARM: at91: pm: switch to the PIE infrastructure

 arch/arm/Kconfig                 |   2 +
 arch/arm/Makefile                |   1 +
 arch/arm/mach-at91/Kconfig       |   1 +
 arch/arm/mach-at91/Makefile      |   2 +-
 arch/arm/mach-at91/pm.c          |  31 ++--
 arch/arm/mach-at91/pm/.gitignore |   2 +
 arch/arm/mach-at91/pm/Makefile   |   3 +
 arch/arm/mach-at91/pm/atmel_pm.c |  97 +++++++++++
 arch/arm/mach-at91/pm_suspend.S  | 338 ---------------------------------------
 arch/arm/pie/Kconfig             |   8 +
 arch/arm/pie/Makefile            |   1 +
 arch/arm/pie/Makefile.pie        |  86 ++++++++++
 arch/arm/pie/lib/empty.c         |  15 ++
 arch/arm/pie/pie.c               |  97 +++++++++++
 arch/arm/pie/pie.lds.S           |  40 +++++
 include/linux/pie.h              | 146 +++++++++++++++++
 16 files changed, 510 insertions(+), 360 deletions(-)
 create mode 100644 arch/arm/mach-at91/pm/.gitignore
 create mode 100644 arch/arm/mach-at91/pm/Makefile
 create mode 100644 arch/arm/mach-at91/pm/atmel_pm.c
 delete mode 100644 arch/arm/mach-at91/pm_suspend.S
 create mode 100644 arch/arm/pie/Kconfig
 create mode 100644 arch/arm/pie/Makefile
 create mode 100644 arch/arm/pie/Makefile.pie
 create mode 100644 arch/arm/pie/lib/empty.c
 create mode 100644 arch/arm/pie/pie.c
 create mode 100644 arch/arm/pie/pie.lds.S
 create mode 100644 include/linux/pie.h

-- 
2.10.2

[PATCH v4 1/2] ARM: PIE infrastructure

[Alexandre Belloni]

Add support for embedding position independent executables (PIE) in the
kernel. Those PIEs can then be loaded into memory allocated using genalloc.
For example, this allows running code from SRAM which is usually needed for
suspend/resume or to change the DDR timings. That code is usually written
in assembly and can now be developed in C.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
---
 arch/arm/Kconfig          |   2 +
 arch/arm/Makefile         |   1 +
 arch/arm/pie/Kconfig      |   8 +++
 arch/arm/pie/Makefile     |   1 +
 arch/arm/pie/Makefile.pie |  86 +++++++++++++++++++++++++++
 arch/arm/pie/lib/empty.c  |  15 +++++
 arch/arm/pie/pie.c        |  97 ++++++++++++++++++++++++++++++
 arch/arm/pie/pie.lds.S    |  40 +++++++++++++
 include/linux/pie.h       | 146 ++++++++++++++++++++++++++++++++++++++++++++++
 9 files changed, 396 insertions(+)
 create mode 100644 arch/arm/pie/Kconfig
 create mode 100644 arch/arm/pie/Makefile
 create mode 100644 arch/arm/pie/Makefile.pie
 create mode 100644 arch/arm/pie/lib/empty.c
 create mode 100644 arch/arm/pie/pie.c
 create mode 100644 arch/arm/pie/pie.lds.S
 create mode 100644 include/linux/pie.h

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index b5d529fdffab..aba849d7a7cf 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -2209,3 +2209,5 @@ endif
 source "lib/Kconfig"
 
 source "arch/arm/kvm/Kconfig"
+
+source "arch/arm/pie/Kconfig"
diff --git a/arch/arm/Makefile b/arch/arm/Makefile
index 6be9ee148b78..9acb9581853a 100644
--- a/arch/arm/Makefile
+++ b/arch/arm/Makefile
@@ -282,6 +282,7 @@ core-$(CONFIG_VFP)		+= arch/arm/vfp/
 core-$(CONFIG_XEN)		+= arch/arm/xen/
 core-$(CONFIG_KVM_ARM_HOST) 	+= arch/arm/kvm/
 core-$(CONFIG_VDSO)		+= arch/arm/vdso/
+core-$(CONFIG_PIE)		+= arch/arm/pie/
 
 # If we have a machine-specific directory, then include it in the build.
 core-y				+= arch/arm/kernel/ arch/arm/mm/ arch/arm/common/
diff --git a/arch/arm/pie/Kconfig b/arch/arm/pie/Kconfig
new file mode 100644
index 000000000000..d76122140561
--- /dev/null
+++ b/arch/arm/pie/Kconfig
@@ -0,0 +1,8 @@
+config PIE
+	bool
+	help
+	  This option adds support for embedding position indepentant (PIE)
+	  executables into the kernel. The PIEs can then be copied into
+	  genalloc regions such as SRAM and executed. Some platforms require
+	  this for suspend/resume support.
+
diff --git a/arch/arm/pie/Makefile b/arch/arm/pie/Makefile
new file mode 100644
index 000000000000..d1a6a823e431
--- /dev/null
+++ b/arch/arm/pie/Makefile
@@ -0,0 +1 @@
+obj-y += pie.o
diff --git a/arch/arm/pie/Makefile.pie b/arch/arm/pie/Makefile.pie
new file mode 100644
index 000000000000..4125da72bcaa
--- /dev/null
+++ b/arch/arm/pie/Makefile.pie
@@ -0,0 +1,86 @@
+obj-y		:= pie.bin.elf
+
+# Report unresolved symbol references
+ldflags-y	+= --no-undefined
+# Delete all temporary local symbols
+ldflags-y	+= -X
+
+GCOV_PROFILE := n
+
+KBUILD_CFLAGS += -DDISABLE_BRANCH_PROFILING
+
+ifeq ($(CONFIG_FUNCTION_TRACER),y)
+ORIG_CFLAGS := $(KBUILD_CFLAGS)
+KBUILD_CFLAGS = $(subst -pg, , $(ORIG_CFLAGS))
+endif
+
+# Reset objcopy flags
+OBJCOPYFLAGS	=
+
+$(obj)/empty.o: arch/arm/pie/lib/empty.c FORCE
+	$(call if_changed_rule,cc_o_c)
+
+# Reference gcc builtins for use in PIE with __pie_
+$(obj)/pie_rename.syms: $(obj)/empty.o
+	@$(NM) $^ | awk '{if ($$3) print $$3,"__pie_"$(PIE_NAME)$$3}' > $@
+
+# For embedding address of the symbols copied from the PIE into the kernel
+$(obj)/pie.syms: $(obj)/pie.elf
+	@$(NM) $^ | awk '{if ($$3 && $$2 == toupper($$2)) print $$3,"=","0x"$$1";"}' > $@
+
+# Collect together the libpie objects
+LDFLAGS_libpie_stage1.o += -r
+
+$(obj)/libpie_stage1.o: $(obj)/empty.o
+	$(call if_changed,ld)
+
+# Rename the libpie gcc builtins with a __pie_ prefix
+OBJCOPYFLAGS_libpie_stage2.o += --redefine-syms=$(obj)/pie_rename.syms
+
+$(obj)/libpie_stage2.o: $(obj)/libpie_stage1.o
+	$(call if_changed,objcopy)
+
+CFLAGS_$(PIE_NAME).o += -fPIE
+
+OBJCOPYFLAGS_pie_stage1.o += --redefine-syms=$(obj)/pie_rename.syms
+$(obj)/pie_stage1.o: $(obj)/$(PIE_NAME).o $(obj)/pie_rename.syms
+	$(call if_changed,objcopy)
+
+LDFLAGS_pie_stage2.o += -r
+
+$(obj)/pie_stage2.o: $(obj)/pie_stage1.o $(obj)/libpie_stage2.o
+	$(call if_changed,ld)
+
+SEDFLAGS_lds	= s/PIE_NAME/$(PIE_NAME)/
+$(obj)/pie.lds.S: arch/arm/pie/pie.lds.S
+	@sed "$(SEDFLAGS_lds)" < $< > $@
+
+# Create the position independent executable
+LDFLAGS_pie.elf += -Bstatic -T $(obj)/pie.lds
+
+$(obj)/pie.elf: $(obj)/pie_stage2.o $(obj)/pie.lds
+	$(call if_changed,ld)
+
+# Create binary data for the kernel
+OBJCOPYFLAGS_pie.bin += -O binary
+
+$(obj)/pie.bin: $(obj)/pie.elf
+	$(call if_changed,objcopy)
+
+# Import the data into the kernel
+ifeq ($(CONFIG_CPU_BIG_ENDIAN),y)
+OBJCOPYFLAGS_pie.bin.o += -B $(ARCH) -I binary -O elf32-bigarm
+else
+OBJCOPYFLAGS_pie.bin.o += -B $(ARCH) -I binary -O elf32-littlearm
+endif
+
+$(obj)/pie.al.c: $(obj)/pie.elf
+	@$(OBJDUMP) -h --section .text $^ | awk '{sub(/2\*\*/,"(1 << ",$$7);if ($$2 == ".text") print "unsigned int __pie_$(PIE_NAME)_align = "$$7");"}' > $@
+
+$(obj)/pie.bin.o: $(obj)/pie.bin
+	$(call if_changed,objcopy)
+
+LDFLAGS_pie.bin.elf += --just-symbols=$(obj)/pie.syms -r
+$(obj)/pie.bin.elf: $(obj)/pie.bin.o $(obj)/pie.al.o $(obj)/pie.syms
+	$(call if_changed,ld)
+
diff --git a/arch/arm/pie/lib/empty.c b/arch/arm/pie/lib/empty.c
new file mode 100644
index 000000000000..9a6d54956379
--- /dev/null
+++ b/arch/arm/pie/lib/empty.c
@@ -0,0 +1,15 @@
+void __div0(void)
+{
+};
+
+void __aeabi_unwind_cpp_pr0(void)
+{
+};
+
+void __aeabi_unwind_cpp_pr1(void)
+{
+};
+
+void __aeabi_unwind_cpp_pr2(void)
+{
+};
diff --git a/arch/arm/pie/pie.c b/arch/arm/pie/pie.c
new file mode 100644
index 000000000000..b32aa4cb11a8
--- /dev/null
+++ b/arch/arm/pie/pie.c
@@ -0,0 +1,97 @@
+/*
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/genalloc.h>
+#include <linux/pie.h>
+#include <asm/cacheflush.h>
+
+struct pie_chunk {
+	struct gen_pool *pool;
+	unsigned long addr;
+	unsigned int align_offset;
+	size_t sz;
+};
+
+struct pie_chunk *__pie_load_data(struct gen_pool *pool, void *code_start,
+				  void *code_end, unsigned int align)
+{
+	struct pie_chunk *chunk;
+	unsigned long offset;
+	int ret;
+	size_t code_sz;
+	unsigned long base;
+	phys_addr_t pbase;
+
+	chunk = kzalloc(sizeof(*chunk), GFP_KERNEL);
+	if (!chunk) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	code_sz = code_end - code_start;
+	chunk->pool = pool;
+	chunk->sz = code_sz;
+
+	base = gen_pool_alloc(pool, chunk->sz + align);
+	if (!base) {
+		ret = -ENOMEM;
+		goto err_free;
+	}
+
+	pbase = gen_pool_virt_to_phys(pool, base);
+	chunk->addr = (unsigned long)__arm_ioremap_exec(pbase, code_sz, false);
+	if (!chunk->addr) {
+		ret = -ENOMEM;
+		goto err_remap;
+	}
+
+	chunk->align_offset = chunk->addr % align;
+	chunk->addr += chunk->align_offset;
+
+	memcpy((char *)chunk->addr, code_start, code_sz);
+	flush_icache_range(chunk->addr, chunk->addr + code_sz);
+
+	offset = gen_pool_virt_to_phys(pool, chunk->addr);
+
+	return chunk;
+
+err_remap:
+	gen_pool_free(chunk->pool, chunk->addr, chunk->sz);
+
+err_free:
+	kfree(chunk);
+err:
+	return ERR_PTR(ret);
+}
+EXPORT_SYMBOL_GPL(__pie_load_data);
+
+phys_addr_t pie_to_phys(struct pie_chunk *chunk, unsigned long addr)
+{
+	return gen_pool_virt_to_phys(chunk->pool, addr);
+}
+EXPORT_SYMBOL_GPL(pie_to_phys);
+
+void __iomem *__kern_to_pie(struct pie_chunk *chunk, void *ptr)
+{
+	uintptr_t offset = (uintptr_t)ptr;
+
+	if (offset >= chunk->sz)
+		return NULL;
+	else
+		return (void *)(chunk->addr + offset);
+}
+EXPORT_SYMBOL_GPL(__kern_to_pie);
+
+void pie_free(struct pie_chunk *chunk)
+{
+	gen_pool_free(chunk->pool, chunk->addr, chunk->sz);
+	kfree(chunk);
+}
+EXPORT_SYMBOL_GPL(pie_free);
diff --git a/arch/arm/pie/pie.lds.S b/arch/arm/pie/pie.lds.S
new file mode 100644
index 000000000000..e640c93f5bc8
--- /dev/null
+++ b/arch/arm/pie/pie.lds.S
@@ -0,0 +1,40 @@
+OUTPUT_FORMAT("elf32-littlearm", "elf32-bigarm", "elf32-littlearm")
+OUTPUT_ARCH(arm)
+
+#include <linux/export.h>
+
+SECTIONS
+{
+	/* Don't need unwind tables */
+	/DISCARD/ : {
+		*(.ARM.exidx*)
+		*(.ARM.extab*)
+		*(.comment)
+	}
+
+	. = 0x0;
+
+	____pie_PIE_NAME_start : {
+		VMLINUX_SYMBOL(__pie_PIE_NAME_start) = .;
+	}
+
+	.text : {
+		. = ALIGN(4);
+		KEEP(*(.text))
+	}
+
+	.rodata : { *(SORT_BY_ALIGNMENT(.rodata*)) }
+	. = ALIGN(4);
+
+	.data : {
+		*(SORT_BY_ALIGNMENT(.data*))
+		. = ALIGN(4);
+
+		*(SORT_BY_ALIGNMENT(.bss*))
+		. = ALIGN(4);
+	}
+
+	____pie_PIE_NAME_end : {
+		VMLINUX_SYMBOL(__pie_PIE_NAME_end) = .;
+	}
+}
diff --git a/include/linux/pie.h b/include/linux/pie.h
new file mode 100644
index 000000000000..cec58384dda2
--- /dev/null
+++ b/include/linux/pie.h
@@ -0,0 +1,146 @@
+/*
+ * Copyright 2013 Texas Instruments, Inc.
+ *      Russ Dill <russ.dill@ti.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ */
+
+#ifndef _LINUX_PIE_H
+#define _LINUX_PIE_H
+
+#include <linux/kernel.h>
+#include <linux/err.h>
+
+#include <asm/fncpy.h>
+#include <linux/bug.h>
+
+struct gen_pool;
+struct pie_chunk;
+
+#ifdef CONFIG_PIE
+
+/**
+ * __pie_load_data - load and fixup PIE code from kernel data
+ * @pool:	pool to allocate memory from and copy code into
+ * @start:	virtual start address in kernel of chunk specific code
+ * @end:	virtual end address in kernel of chunk specific code
+ * @phys:	%true to fixup to physical address of destination, %false to
+ *		fixup to virtual address of destination
+ *
+ * Returns 0 on success, -EERROR otherwise
+ */
+struct pie_chunk *__pie_load_data(struct gen_pool *pool, void *start, void *end,
+				  unsigned int align);
+
+/**
+ * pie_to_phys - translate a virtual PIE address into a physical one
+ * @chunk:	identifier returned by pie_load_sections
+ * @addr:	virtual address within pie chunk
+ *
+ * Returns physical address on success, -1 otherwise
+ */
+phys_addr_t pie_to_phys(struct pie_chunk *chunk, unsigned long addr);
+
+void __iomem *__kern_to_pie(struct pie_chunk *chunk, void *ptr);
+
+/**
+ * pie_free - free the pool space used by an pie chunk
+ * @chunk:	identifier returned by pie_load_sections
+ */
+void pie_free(struct pie_chunk *chunk);
+
+#define __pie_load_sections(pool, name, folder) ({			\
+	extern char _binary_##folder##_pie_bin_start[];			\
+	extern char __pie_##name##_start[];				\
+	extern char __pie_##name##_end[];				\
+	extern unsigned int __pie_##name##_align;			\
+	char *start = _binary_##folder##_pie_bin_start +		\
+		(unsigned long)__pie_##name##_start;			\
+	char *end = _binary_##folder##_pie_bin_start +			\
+		(unsigned long)__pie_##name##_end;			\
+	__pie_load_data(pool, start, end, __pie_##name##_align);	\
+})
+
+/*
+ * Required for any symbol within an PIE section that is referenced by the
+ * kernel
+ */
+#define EXPORT_PIE_SYMBOL(sym)		extern typeof(sym) sym __weak
+
+#else
+
+static inline struct pie_chunk *__pie_load_data(struct gen_pool *pool,
+						void *start, void *end,
+						unsigned int algin)
+{
+	return ERR_PTR(-EINVAL);
+}
+
+static inline phys_addr_t pie_to_phys(struct pie_chunk *chunk,
+				      unsigned long addr)
+{
+	return -1;
+}
+
+static inline void __iomem *__kern_to_pie(struct pie_chunk *chunk, void *ptr)
+{
+	return NULL;
+}
+
+static inline void pie_free(struct pie_chunk *chunk)
+{
+}
+
+#define __pie_load_sections(pool, name, folder, phys) ({ ERR_PTR(-EINVAL); })
+
+#endif
+
+/**
+ * pie_load_sections - load and fixup sections associated with the given name
+ * @pool:	pool to allocate memory from and copy code into
+ *		fixup to virtual address of destination
+ * @name:	the name given to __pie() and __pie_data() when marking
+ *		data and code
+ *
+ * Returns 0 on success, -EERROR otherwise
+ */
+#define pie_load_sections(pool, name, folder) ({			\
+	__pie_load_sections(pool, name, folder);			\
+})
+
+/**
+ * kern_to_pie - convert a kernel symbol to the virtual address of where
+ * that symbol is loaded into the given PIE chunk.
+ *
+ * @chunk:	identifier returned by pie_load_sections
+ * @p:		symbol to convert
+ *
+ * Return type is the same as type passed
+ */
+#define kern_to_pie(chunk, p) ({					\
+	void *__ptr = (void *)(p);					\
+	typeof(p) __result = (typeof(p))__kern_to_pie(chunk, __ptr);	\
+	__result;							\
+})
+
+/**
+ * kern_to_fn - convert a kernel function symbol to the virtual address of where
+ * that symbol is loaded into the given PIE chunk
+ *
+ * @chunk:	identifier returned by pie_load_sections
+ * @funcp:	function to convert
+ *
+ * Return type is the same as type passed
+ */
+#define fn_to_pie(chunk, funcp) ({					\
+	uintptr_t __kern_addr, __pie_addr;				\
+									\
+	__kern_addr = (uintptr_t)funcp;					\
+	__pie_addr = kern_to_pie(chunk, __kern_addr);			\
+									\
+	(typeof(&funcp))(__pie_addr);					\
+})
+
+#endif
-- 
2.10.2

[PATCH v4 2/2] ARM: at91: pm: switch to the PIE infrastructure

[Alexandre Belloni]

Using the PIE infrastructure allows to write the whole suspend/resume
functions in C instead of assembly.

The only remaining assembly instruction is wfi for armv5
It makes the code shorter and clearer.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
---
 arch/arm/mach-at91/Kconfig       |   1 +
 arch/arm/mach-at91/Makefile      |   2 +-
 arch/arm/mach-at91/pm.c          |  31 ++--
 arch/arm/mach-at91/pm/.gitignore |   2 +
 arch/arm/mach-at91/pm/Makefile   |   3 +
 arch/arm/mach-at91/pm/atmel_pm.c |  97 +++++++++++
 arch/arm/mach-at91/pm_suspend.S  | 338 ---------------------------------------
 7 files changed, 114 insertions(+), 360 deletions(-)
 create mode 100644 arch/arm/mach-at91/pm/.gitignore
 create mode 100644 arch/arm/mach-at91/pm/Makefile
 create mode 100644 arch/arm/mach-at91/pm/atmel_pm.c
 delete mode 100644 arch/arm/mach-at91/pm_suspend.S

diff --git a/arch/arm/mach-at91/Kconfig b/arch/arm/mach-at91/Kconfig
index 841e924143f9..cecb20efbf3c 100644
--- a/arch/arm/mach-at91/Kconfig
+++ b/arch/arm/mach-at91/Kconfig
@@ -5,6 +5,7 @@ menuconfig ARCH_AT91
 	select GPIOLIB
 	select PINCTRL
 	select SOC_BUS
+	select PIE
 
 if ARCH_AT91
 config SOC_SAMA5D2
diff --git a/arch/arm/mach-at91/Makefile b/arch/arm/mach-at91/Makefile
index c5bbf8bb8c0f..062336de4f66 100644
--- a/arch/arm/mach-at91/Makefile
+++ b/arch/arm/mach-at91/Makefile
@@ -10,7 +10,7 @@ obj-$(CONFIG_SOC_SAMA5)		+= sama5.o
 
 # Power Management
 obj-$(CONFIG_PM)		+= pm.o
-obj-$(CONFIG_PM)		+= pm_suspend.o
+obj-$(CONFIG_PM)		+= pm/
 
 ifeq ($(CONFIG_CPU_V7),y)
 AFLAGS_pm_suspend.o := -march=armv7-a
diff --git a/arch/arm/mach-at91/pm.c b/arch/arm/mach-at91/pm.c
index b4332b727e9c..f0a3eee73b86 100644
--- a/arch/arm/mach-at91/pm.c
+++ b/arch/arm/mach-at91/pm.c
@@ -21,6 +21,7 @@
 #include <linux/of.h>
 #include <linux/of_platform.h>
 #include <linux/of_address.h>
+#include <linux/pie.h>
 #include <linux/platform_device.h>
 #include <linux/platform_data/atmel.h>
 #include <linux/io.h>
@@ -57,6 +58,8 @@ static struct {
 
 void __iomem *at91_ramc_base[2];
 
+static struct pie_chunk *atmel_pm_pie;
+
 static int at91_pm_valid_state(suspend_state_t state)
 {
 	switch (state) {
@@ -134,10 +137,6 @@ EXPORT_SYMBOL(at91_suspend_entering_slow_clock);
 static void (*at91_suspend_sram_fn)(void __iomem *pmc, void __iomem *ramc0,
 			  void __iomem *ramc1, int memctrl);
 
-extern void at91_pm_suspend_in_sram(void __iomem *pmc, void __iomem *ramc0,
-			    void __iomem *ramc1, int memctrl);
-extern u32 at91_pm_suspend_in_sram_sz;
-
 static void at91_pm_suspend(suspend_state_t state)
 {
 	unsigned int pm_data = at91_pm_data.memctrl;
@@ -371,11 +370,12 @@ static void at91sam9_idle(void)
 	cpu_do_idle();
 }
 
+extern void atmel_pm_suspend(void __iomem *pmc, void __iomem *ramc0,
+			     void __iomem *ramc1, int memctrl);
+
 static void __init at91_pm_sram_init(void)
 {
 	struct gen_pool *sram_pool;
-	phys_addr_t sram_pbase;
-	unsigned long sram_base;
 	struct device_node *node;
 	struct platform_device *pdev = NULL;
 
@@ -398,23 +398,12 @@ static void __init at91_pm_sram_init(void)
 		return;
 	}
 
-	sram_base = gen_pool_alloc(sram_pool, at91_pm_suspend_in_sram_sz);
-	if (!sram_base) {
-		pr_warn("%s: unable to alloc sram!\n", __func__);
+	atmel_pm_pie = pie_load_sections(sram_pool, atmel_pm,
+					 arch_arm_mach_at91_pm);
+	if (IS_ERR(atmel_pm_pie))
 		return;
-	}
-
-	sram_pbase = gen_pool_virt_to_phys(sram_pool, sram_base);
-	at91_suspend_sram_fn = __arm_ioremap_exec(sram_pbase,
-					at91_pm_suspend_in_sram_sz, false);
-	if (!at91_suspend_sram_fn) {
-		pr_warn("SRAM: Could not map\n");
-		return;
-	}
 
-	/* Copy the pm suspend handler to SRAM */
-	at91_suspend_sram_fn = fncpy(at91_suspend_sram_fn,
-			&at91_pm_suspend_in_sram, at91_pm_suspend_in_sram_sz);
+	at91_suspend_sram_fn = fn_to_pie(atmel_pm_pie, atmel_pm_suspend);
 }
 
 static const struct of_device_id atmel_pmc_ids[] __initconst = {
diff --git a/arch/arm/mach-at91/pm/.gitignore b/arch/arm/mach-at91/pm/.gitignore
new file mode 100644
index 000000000000..1cb8879eb0bf
--- /dev/null
+++ b/arch/arm/mach-at91/pm/.gitignore
@@ -0,0 +1,2 @@
+*.lds*
+*.syms
diff --git a/arch/arm/mach-at91/pm/Makefile b/arch/arm/mach-at91/pm/Makefile
new file mode 100644
index 000000000000..c12d54862c10
--- /dev/null
+++ b/arch/arm/mach-at91/pm/Makefile
@@ -0,0 +1,3 @@
+PIE_NAME := atmel_pm
+
+include arch/arm/pie/Makefile.pie
diff --git a/arch/arm/mach-at91/pm/atmel_pm.c b/arch/arm/mach-at91/pm/atmel_pm.c
new file mode 100644
index 000000000000..7f391addd2da
--- /dev/null
+++ b/arch/arm/mach-at91/pm/atmel_pm.c
@@ -0,0 +1,97 @@
+#include <linux/io.h>
+#include <linux/clk/at91_pmc.h>
+#include <linux/mfd/syscon/atmel-mc.h>
+#include <linux/pie.h>
+#include "../pm.h"
+
+#define	SRAMC_SELF_FRESH_ACTIVE		0x01
+#define	SRAMC_SELF_FRESH_EXIT		0x00
+
+static void at91_sramc_self_refresh(unsigned int is_active,
+				    unsigned int memtype,
+				    void __iomem *sdramc_base,
+				    void __iomem *sdramc_base1)
+{
+	static unsigned int lpr, mdr, lpr1, mdr1;
+
+	switch (memtype) {
+	case AT91_MEMCTRL_MC:
+	/*
+	 * at91rm9200 Memory controller
+	 */
+		if (is_active)
+			__raw_writel(1, sdramc_base + AT91_MC_SDRAMC_SRR);
+		break;
+
+	case AT91_MEMCTRL_DDRSDR:
+		if (is_active) {
+			mdr = __raw_readl(sdramc_base + AT91_DDRSDRC_MDR);
+			lpr = __raw_readl(sdramc_base + AT91_DDRSDRC_LPR);
+
+			if ((mdr & AT91_DDRSDRC_MD) == AT91_DDRSDRC_MD_LOW_POWER_DDR)
+				__raw_writel((mdr & ~AT91_DDRSDRC_MD) |
+					     AT91_DDRSDRC_MD_DDR2, sdramc_base +
+					     AT91_DDRSDRC_MDR);
+			__raw_writel((lpr & ~AT91_DDRSDRC_LPCB) |
+				     AT91_DDRSDRC_LPCB_SELF_REFRESH, sdramc_base
+				     + AT91_DDRSDRC_LPR);
+
+			if (sdramc_base1) {
+				mdr1 = __raw_readl(sdramc_base1 + AT91_DDRSDRC_MDR);
+				lpr1 = __raw_readl(sdramc_base1 + AT91_DDRSDRC_LPR);
+				if ((mdr1 & AT91_DDRSDRC_MD) == AT91_DDRSDRC_MD_LOW_POWER_DDR)
+					__raw_writel((mdr1 & ~AT91_DDRSDRC_MD) |
+						     AT91_DDRSDRC_MD_DDR2,
+						     sdramc_base1 +
+						     AT91_DDRSDRC_MDR);
+				__raw_writel((lpr1 & ~AT91_DDRSDRC_LPCB) |
+					     AT91_DDRSDRC_LPCB_SELF_REFRESH,
+					     sdramc_base1 + AT91_DDRSDRC_LPR);
+			}
+		} else {
+			__raw_writel(mdr, sdramc_base + AT91_DDRSDRC_MDR);
+			__raw_writel(lpr, sdramc_base + AT91_DDRSDRC_LPR);
+			if (sdramc_base1) {
+				__raw_writel(mdr, sdramc_base1 + AT91_DDRSDRC_MDR);
+				__raw_writel(lpr, sdramc_base1 + AT91_DDRSDRC_LPR);
+			}
+		}
+		break;
+
+	case AT91_MEMCTRL_SDRAMC:
+		if (is_active) {
+			lpr = __raw_readl(sdramc_base + AT91_SDRAMC_LPR);
+
+			__raw_writel((lpr & ~AT91_SDRAMC_LPCB) |
+				     AT91_SDRAMC_LPCB_SELF_REFRESH, sdramc_base
+				     + AT91_SDRAMC_LPR);
+		} else {
+			__raw_writel(lpr, sdramc_base + AT91_SDRAMC_LPR);
+		}
+		break;
+	}
+}
+
+void atmel_pm_suspend(void __iomem *pmc, void __iomem *ramc0,
+		      void __iomem *ramc1, int memctrl)
+{
+	int memtype, pm_mode;
+
+	memtype = memctrl & AT91_PM_MEMTYPE_MASK;
+	pm_mode = (memctrl >> AT91_PM_MODE_OFFSET) & AT91_PM_MODE_MASK;
+
+	dsb();
+
+	at91_sramc_self_refresh(1, memtype, ramc0, ramc1);
+
+#if defined(CONFIG_CPU_V7)
+	dsb();
+	wfi();
+#else
+	asm volatile ("mcr	p15, 0, %0, c7, c0, 4" \
+		      : : "r" (0) : "memory");
+#endif
+
+	at91_sramc_self_refresh(0, memtype, ramc0, ramc1);
+}
+EXPORT_PIE_SYMBOL(atmel_pm_suspend);
diff --git a/arch/arm/mach-at91/pm_suspend.S b/arch/arm/mach-at91/pm_suspend.S
deleted file mode 100644
index a25defda3d22..000000000000
--- a/arch/arm/mach-at91/pm_suspend.S
+++ /dev/null
@@ -1,338 +0,0 @@
-/*
- * arch/arm/mach-at91/pm_slow_clock.S
- *
- *  Copyright (C) 2006 Savin Zlobec
- *
- * AT91SAM9 support:
- *  Copyright (C) 2007 Anti Sullin <anti.sullin@artecdesign.ee
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
- *
- */
-#include <linux/linkage.h>
-#include <linux/clk/at91_pmc.h>
-#include "pm.h"
-
-#define	SRAMC_SELF_FRESH_ACTIVE		0x01
-#define	SRAMC_SELF_FRESH_EXIT		0x00
-
-pmc	.req	r0
-tmp1	.req	r4
-tmp2	.req	r5
-
-/*
- * Wait until master clock is ready (after switching master clock source)
- */
-	.macro wait_mckrdy
-1:	ldr	tmp1, [pmc, #AT91_PMC_SR]
-	tst	tmp1, #AT91_PMC_MCKRDY
-	beq	1b
-	.endm
-
-/*
- * Wait until master oscillator has stabilized.
- */
-	.macro wait_moscrdy
-1:	ldr	tmp1, [pmc, #AT91_PMC_SR]
-	tst	tmp1, #AT91_PMC_MOSCS
-	beq	1b
-	.endm
-
-/*
- * Wait until PLLA has locked.
- */
-	.macro wait_pllalock
-1:	ldr	tmp1, [pmc, #AT91_PMC_SR]
-	tst	tmp1, #AT91_PMC_LOCKA
-	beq	1b
-	.endm
-
-/*
- * Put the processor to enter the idle state
- */
-	.macro at91_cpu_idle
-
-#if defined(CONFIG_CPU_V7)
-	mov	tmp1, #AT91_PMC_PCK
-	str	tmp1, [pmc, #AT91_PMC_SCDR]
-
-	dsb
-
-	wfi		@ Wait For Interrupt
-#else
-	mcr	p15, 0, tmp1, c7, c0, 4
-#endif
-
-	.endm
-
-	.text
-
-	.arm
-
-/*
- * void at91_pm_suspend_in_sram(void __iomem *pmc, void __iomem *sdramc,
- *			void __iomem *ramc1, int memctrl)
- * @input param:
- * 	@r0: base address of AT91_PMC
- *  	@r1: base address of SDRAM Controller (SDRAM, DDRSDR, or AT91_SYS)
- *	@r2: base address of second SDRAM Controller or 0 if not present
- *	@r3: pm information
- */
-/* at91_pm_suspend_in_sram must be 8-byte aligned per the requirements of fncpy() */
-	.align 3
-ENTRY(at91_pm_suspend_in_sram)
-	/* Save registers on stack */
-	stmfd	sp!, {r4 - r12, lr}
-
-	/* Drain write buffer */
-	mov	tmp1, #0
-	mcr	p15, 0, tmp1, c7, c10, 4
-
-	str	r0, .pmc_base
-	str	r1, .sramc_base
-	str	r2, .sramc1_base
-
-	and	r0, r3, #AT91_PM_MEMTYPE_MASK
-	str	r0, .memtype
-
-	lsr	r0, r3, #AT91_PM_MODE_OFFSET
-	and	r0, r0, #AT91_PM_MODE_MASK
-	str	r0, .pm_mode
-
-	/* Active the self-refresh mode */
-	mov	r0, #SRAMC_SELF_FRESH_ACTIVE
-	bl	at91_sramc_self_refresh
-
-	ldr	r0, .pm_mode
-	tst	r0, #AT91_PM_SLOW_CLOCK
-	beq	skip_disable_main_clock
-
-	ldr	pmc, .pmc_base
-
-	/* Save Master clock setting */
-	ldr	tmp1, [pmc, #AT91_PMC_MCKR]
-	str	tmp1, .saved_mckr
-
-	/*
-	 * Set the Master clock source to slow clock
-	 */
-	bic	tmp1, tmp1, #AT91_PMC_CSS
-	str	tmp1, [pmc, #AT91_PMC_MCKR]
-
-	wait_mckrdy
-
-	/* Save PLLA setting and disable it */
-	ldr	tmp1, [pmc, #AT91_CKGR_PLLAR]
-	str	tmp1, .saved_pllar
-
-	mov	tmp1, #AT91_PMC_PLLCOUNT
-	orr	tmp1, tmp1, #(1 << 29)		/* bit 29 always set */
-	str	tmp1, [pmc, #AT91_CKGR_PLLAR]
-
-	/* Turn off the main oscillator */
-	ldr	tmp1, [pmc, #AT91_CKGR_MOR]
-	bic	tmp1, tmp1, #AT91_PMC_MOSCEN
-	orr	tmp1, tmp1, #AT91_PMC_KEY
-	str	tmp1, [pmc, #AT91_CKGR_MOR]
-
-skip_disable_main_clock:
-	ldr	pmc, .pmc_base
-
-	/* Wait for interrupt */
-	at91_cpu_idle
-
-	ldr	r0, .pm_mode
-	tst	r0, #AT91_PM_SLOW_CLOCK
-	beq	skip_enable_main_clock
-
-	ldr	pmc, .pmc_base
-
-	/* Turn on the main oscillator */
-	ldr	tmp1, [pmc, #AT91_CKGR_MOR]
-	orr	tmp1, tmp1, #AT91_PMC_MOSCEN
-	orr	tmp1, tmp1, #AT91_PMC_KEY
-	str	tmp1, [pmc, #AT91_CKGR_MOR]
-
-	wait_moscrdy
-
-	/* Restore PLLA setting */
-	ldr	tmp1, .saved_pllar
-	str	tmp1, [pmc, #AT91_CKGR_PLLAR]
-
-	tst	tmp1, #(AT91_PMC_MUL &  0xff0000)
-	bne	3f
-	tst	tmp1, #(AT91_PMC_MUL & ~0xff0000)
-	beq	4f
-3:
-	wait_pllalock
-4:
-
-	/*
-	 * Restore master clock setting
-	 */
-	ldr	tmp1, .saved_mckr
-	str	tmp1, [pmc, #AT91_PMC_MCKR]
-
-	wait_mckrdy
-
-skip_enable_main_clock:
-	/* Exit the self-refresh mode */
-	mov	r0, #SRAMC_SELF_FRESH_EXIT
-	bl	at91_sramc_self_refresh
-
-	/* Restore registers, and return */
-	ldmfd	sp!, {r4 - r12, pc}
-ENDPROC(at91_pm_suspend_in_sram)
-
-/*
- * void at91_sramc_self_refresh(unsigned int is_active)
- *
- * @input param:
- *	@r0: 1 - active self-refresh mode
- *	     0 - exit self-refresh mode
- * register usage:
- * 	@r1: memory type
- *	@r2: base address of the sram controller
- */
-
-ENTRY(at91_sramc_self_refresh)
-	ldr	r1, .memtype
-	ldr	r2, .sramc_base
-
-	cmp	r1, #AT91_MEMCTRL_MC
-	bne	ddrc_sf
-
-	/*
-	 * at91rm9200 Memory controller
-	 */
-
-	 /*
-	  * For exiting the self-refresh mode, do nothing,
-	  * automatically exit the self-refresh mode.
-	  */
-	tst	r0, #SRAMC_SELF_FRESH_ACTIVE
-	beq	exit_sramc_sf
-
-	/* Active SDRAM self-refresh mode */
-	mov	r3, #1
-	str	r3, [r2, #AT91_MC_SDRAMC_SRR]
-	b	exit_sramc_sf
-
-ddrc_sf:
-	cmp	r1, #AT91_MEMCTRL_DDRSDR
-	bne	sdramc_sf
-
-	/*
-	 * DDR Memory controller
-	 */
-	tst	r0, #SRAMC_SELF_FRESH_ACTIVE
-	beq	ddrc_exit_sf
-
-	/* LPDDR1 --> force DDR2 mode during self-refresh */
-	ldr	r3, [r2, #AT91_DDRSDRC_MDR]
-	str	r3, .saved_sam9_mdr
-	bic	r3, r3, #~AT91_DDRSDRC_MD
-	cmp	r3, #AT91_DDRSDRC_MD_LOW_POWER_DDR
-	ldreq	r3, [r2, #AT91_DDRSDRC_MDR]
-	biceq	r3, r3, #AT91_DDRSDRC_MD
-	orreq	r3, r3, #AT91_DDRSDRC_MD_DDR2
-	streq	r3, [r2, #AT91_DDRSDRC_MDR]
-
-	/* Active DDRC self-refresh mode */
-	ldr	r3, [r2, #AT91_DDRSDRC_LPR]
-	str	r3, .saved_sam9_lpr
-	bic	r3, r3, #AT91_DDRSDRC_LPCB
-	orr	r3, r3, #AT91_DDRSDRC_LPCB_SELF_REFRESH
-	str	r3, [r2, #AT91_DDRSDRC_LPR]
-
-	/* If using the 2nd ddr controller */
-	ldr	r2, .sramc1_base
-	cmp	r2, #0
-	beq	no_2nd_ddrc
-
-	ldr	r3, [r2, #AT91_DDRSDRC_MDR]
-	str	r3, .saved_sam9_mdr1
-	bic	r3, r3, #~AT91_DDRSDRC_MD
-	cmp	r3, #AT91_DDRSDRC_MD_LOW_POWER_DDR
-	ldreq	r3, [r2, #AT91_DDRSDRC_MDR]
-	biceq	r3, r3, #AT91_DDRSDRC_MD
-	orreq	r3, r3, #AT91_DDRSDRC_MD_DDR2
-	streq	r3, [r2, #AT91_DDRSDRC_MDR]
-
-	/* Active DDRC self-refresh mode */
-	ldr	r3, [r2, #AT91_DDRSDRC_LPR]
-	str	r3, .saved_sam9_lpr1
-	bic	r3, r3, #AT91_DDRSDRC_LPCB
-	orr	r3, r3, #AT91_DDRSDRC_LPCB_SELF_REFRESH
-	str	r3, [r2, #AT91_DDRSDRC_LPR]
-
-no_2nd_ddrc:
-	b	exit_sramc_sf
-
-ddrc_exit_sf:
-	/* Restore MDR in case of LPDDR1 */
-	ldr	r3, .saved_sam9_mdr
-	str	r3, [r2, #AT91_DDRSDRC_MDR]
-	/* Restore LPR on AT91 with DDRAM */
-	ldr	r3, .saved_sam9_lpr
-	str	r3, [r2, #AT91_DDRSDRC_LPR]
-
-	/* If using the 2nd ddr controller */
-	ldr	r2, .sramc1_base
-	cmp	r2, #0
-	ldrne	r3, .saved_sam9_mdr1
-	strne	r3, [r2, #AT91_DDRSDRC_MDR]
-	ldrne	r3, .saved_sam9_lpr1
-	strne	r3, [r2, #AT91_DDRSDRC_LPR]
-
-	b	exit_sramc_sf
-
-	/*
-	 * SDRAMC Memory controller
-	 */
-sdramc_sf:
-	tst	r0, #SRAMC_SELF_FRESH_ACTIVE
-	beq	sdramc_exit_sf
-
-	/* Active SDRAMC self-refresh mode */
-	ldr	r3, [r2, #AT91_SDRAMC_LPR]
-	str	r3, .saved_sam9_lpr
-	bic	r3, r3, #AT91_SDRAMC_LPCB
-	orr	r3, r3, #AT91_SDRAMC_LPCB_SELF_REFRESH
-	str	r3, [r2, #AT91_SDRAMC_LPR]
-
-sdramc_exit_sf:
-	ldr	r3, .saved_sam9_lpr
-	str	r3, [r2, #AT91_SDRAMC_LPR]
-
-exit_sramc_sf:
-	mov	pc, lr
-ENDPROC(at91_sramc_self_refresh)
-
-.pmc_base:
-	.word 0
-.sramc_base:
-	.word 0
-.sramc1_base:
-	.word 0
-.memtype:
-	.word 0
-.pm_mode:
-	.word 0
-.saved_mckr:
-	.word 0
-.saved_pllar:
-	.word 0
-.saved_sam9_lpr:
-	.word 0
-.saved_sam9_lpr1:
-	.word 0
-.saved_sam9_mdr:
-	.word 0
-.saved_sam9_mdr1:
-	.word 0
-
-ENTRY(at91_pm_suspend_in_sram_sz)
-	.word .-at91_pm_suspend_in_sram
-- 
2.10.2