diff --git a/bl31/aarch64/runtime_exceptions.S b/bl31/aarch64/runtime_exceptions.S
index d8fbb9b..9b7735f 100644
--- a/bl31/aarch64/runtime_exceptions.S
+++ b/bl31/aarch64/runtime_exceptions.S
@@ -14,6 +14,26 @@
 
 	.globl	runtime_exceptions
 
+	.globl	sync_exception_sp_el0
+	.globl	irq_sp_el0
+	.globl	fiq_sp_el0
+	.globl	serror_sp_el0
+
+	.globl	sync_exception_sp_elx
+	.globl	irq_sp_elx
+	.globl	fiq_sp_elx
+	.globl	serror_sp_elx
+
+	.globl	sync_exception_aarch64
+	.globl	irq_aarch64
+	.globl	fiq_aarch64
+	.globl	serror_aarch64
+
+	.globl	sync_exception_aarch32
+	.globl	irq_aarch32
+	.globl	fiq_aarch32
+	.globl	serror_aarch32
+
 	/* ---------------------------------------------------------------------
 	 * This macro handles Synchronous exceptions.
 	 * Only SMC exceptions are supported.
diff --git a/bl31/bl31.mk b/bl31/bl31.mk
index fdcc931..2db4856 100644
--- a/bl31/bl31.mk
+++ b/bl31/bl31.mk
@@ -51,13 +51,19 @@
 endif
 
 ifeq (${ENABLE_AMU},1)
-BL31_SOURCES		+=	lib/extensions/amu/aarch64/amu.c
+BL31_SOURCES		+=	lib/extensions/amu/aarch64/amu.c		\
+				lib/extensions/amu/aarch64/amu_helpers.S
 endif
 
 ifeq (${ENABLE_SVE_FOR_NS},1)
 BL31_SOURCES		+=	lib/extensions/sve/sve.c
 endif
 
+ifeq (${WORKAROUND_CVE_2017_5715},1)
+BL31_SOURCES		+=	lib/cpus/aarch64/workaround_cve_2017_5715_mmu.S		\
+				lib/cpus/aarch64/workaround_cve_2017_5715_bpiall.S
+endif
+
 BL31_LINKERFILE		:=	bl31/bl31.ld.S
 
 # Flag used to indicate if Crash reporting via console should be included
diff --git a/docs/cpu-specific-build-macros.rst b/docs/cpu-specific-build-macros.rst
index f74b459..014817d 100644
--- a/docs/cpu-specific-build-macros.rst
+++ b/docs/cpu-specific-build-macros.rst
@@ -11,6 +11,15 @@
 operations framework to enable errata workarounds and to enable optimizations
 for a specific CPU on a platform.
 
+Security Vulnerability Workarounds
+----------------------------------
+
+ARM Trusted Firmware exports a series of build flags which control which
+security vulnerability workarounds should be applied at runtime.
+
+-  ``WORKAROUND_CVE_2017_5715``: Enables the security workaround for
+   `CVE-2017-5715`_. Defaults to 1.
+
 CPU Errata Workarounds
 ----------------------
 
@@ -142,6 +151,7 @@
 
 *Copyright (c) 2014-2016, ARM Limited and Contributors. All rights reserved.*
 
+.. _CVE-2017-5715: http://www.cve.mitre.org/cgi-bin/cvename.cgi?name=2017-5715
 .. _Cortex-A53 MPCore Software Developers Errata Notice: http://infocenter.arm.com/help/topic/com.arm.doc.epm048406/Cortex_A53_MPCore_Software_Developers_Errata_Notice.pdf
 .. _Cortex-A57 MPCore Software Developers Errata Notice: http://infocenter.arm.com/help/topic/com.arm.doc.epm049219/cortex_a57_mpcore_software_developers_errata_notice.pdf
 .. _Cortex-A72 MPCore Software Developers Errata Notice: http://infocenter.arm.com/help/topic/com.arm.doc.epm012079/index.html
diff --git a/docs/firmware-design.rst b/docs/firmware-design.rst
index 405964d..3cb004a 100644
--- a/docs/firmware-design.rst
+++ b/docs/firmware-design.rst
@@ -1868,9 +1868,11 @@
 
 The FIP layout consists of a table of contents (ToC) followed by payload data.
 The ToC itself has a header followed by one or more table entries. The ToC is
-terminated by an end marker entry. All ToC entries describe some payload data
-that has been appended to the end of the binary package. With the information
-provided in the ToC entry the corresponding payload data can be retrieved.
+terminated by an end marker entry, and since the size of the ToC is 0 bytes,
+the offset equals the total size of the FIP file. All ToC entries describe some
+payload data that has been appended to the end of the binary package. With the
+information provided in the ToC entry the corresponding payload data can be
+retrieved.
 
 ::
 
diff --git a/docs/porting-guide.rst b/docs/porting-guide.rst
index f020ec9..51d2e64 100644
--- a/docs/porting-guide.rst
+++ b/docs/porting-guide.rst
@@ -549,6 +549,22 @@
    doesn't print anything to the console. If ``PLAT_LOG_LEVEL_ASSERT`` isn't
    defined, it defaults to ``LOG_LEVEL``.
 
+If the platform port uses the Activity Monitor Unit, the following constants
+may be defined:
+
+-  **PLAT\_AMU\_GROUP1\_COUNTERS\_MASK**
+   This mask reflects the set of group counters that should be enabled.  The
+   maximum number of group 1 counters supported by AMUv1 is 16 so the mask
+   can be at most 0xffff. If the platform does not define this mask, no group 1
+   counters are enabled. If the platform defines this mask, the following
+   constant needs to also be defined.
+
+-  **PLAT\_AMU\_GROUP1\_NR\_COUNTERS**
+   This value is used to allocate an array to save and restore the counters
+   specified by ``PLAT_AMU_GROUP1_COUNTERS_MASK`` on CPU suspend.
+   This value should be equal to the highest bit position set in the
+   mask, plus 1.  The maximum number of group 1 counters in AMUv1 is 16.
+
 File : plat\_macros.S [mandatory]
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
diff --git a/docs/user-guide.rst b/docs/user-guide.rst
index 13f0964..0647e70 100644
--- a/docs/user-guide.rst
+++ b/docs/user-guide.rst
@@ -55,7 +55,7 @@
 
     sudo apt-get install build-essential gcc make git libssl-dev
 
-ARM TF has been tested with `Linaro Release 17.04`_.
+ARM TF has been tested with `Linaro Release 17.10`_.
 
 Download and install the AArch32 or AArch64 little-endian GCC cross compiler.
 The `Linaro Release Notes`_ documents which version of the compiler to use for a
@@ -1006,7 +1006,7 @@
    modules by checking out a recent version of the `mbed TLS Repository`_. It
    is important to use a version that is compatible with TF and fixes any
    known security vulnerabilities. See `mbed TLS Security Center`_ for more
-   information. The latest version of TF is tested with tag ``mbedtls-2.4.2``.
+   information. The latest version of TF is tested with tag ``mbedtls-2.6.0``.
 
    The ``drivers/auth/mbedtls/mbedtls_*.mk`` files contain the list of mbed TLS
    source files the modules depend upon.
@@ -1475,10 +1475,10 @@
 The latest version of the AArch64 build of ARM Trusted Firmware has been tested
 on the following ARM FVPs (64-bit host machine only).
 
-NOTE: Unless otherwise stated, the model version is Version 11.1 Build 11.1.22.
+NOTE: Unless otherwise stated, the model version is Version 11.2 Build 11.2.33.
 
 -  ``Foundation_Platform``
--  ``FVP_Base_AEMv8A-AEMv8A`` (Version 8.7, Build 0.8.8702)
+-  ``FVP_Base_AEMv8A-AEMv8A`` (Version 9.0, Build 0.8.9005)
 -  ``FVP_Base_Cortex-A35x4``
 -  ``FVP_Base_Cortex-A53x4``
 -  ``FVP_Base_Cortex-A57x4-A53x4``
@@ -1491,7 +1491,7 @@
 The latest version of the AArch32 build of ARM Trusted Firmware has been tested
 on the following ARM FVPs (64-bit host machine only).
 
--  ``FVP_Base_AEMv8A-AEMv8A`` (Version 8.7, Build 0.8.8702)
+-  ``FVP_Base_AEMv8A-AEMv8A`` (Version 9.0, Build 0.8.9005)
 -  ``FVP_Base_Cortex-A32x4``
 
 NOTE: The build numbers quoted above are those reported by launching the FVP
@@ -1868,10 +1868,10 @@
 
 .. _Linaro: `Linaro Release Notes`_
 .. _Linaro Release: `Linaro Release Notes`_
-.. _Linaro Release Notes: https://community.arm.com/tools/dev-platforms/b/documents/posts/linaro-release-notes-deprecated
-.. _Linaro Release 17.04: https://community.arm.com/tools/dev-platforms/b/documents/posts/linaro-release-notes-deprecated#LinaroRelease17.04
-.. _Linaro instructions: https://community.arm.com/dev-platforms/b/documents/posts/instructions-for-using-the-linaro-software-deliverables
-.. _Instructions for using Linaro's deliverables on Juno: https://community.arm.com/dev-platforms/b/documents/posts/using-linaros-deliverables-on-juno
+.. _Linaro Release Notes: https://community.arm.com/dev-platforms/w/docs/226/old-linaro-release-notes
+.. _Linaro Release 17.10: https://community.arm.com/dev-platforms/w/docs/226/old-linaro-release-notes#1710
+.. _Linaro instructions: https://community.arm.com/dev-platforms/w/docs/304/linaro-software-deliverables
+.. _Instructions for using Linaro's deliverables on Juno: https://community.arm.com/dev-platforms/w/docs/303/juno
 .. _ARM Platforms Portal: https://community.arm.com/dev-platforms/
 .. _Development Studio 5 (DS-5): http://www.arm.com/products/tools/software-tools/ds-5/index.php
 .. _Dia: https://wiki.gnome.org/Apps/Dia/Download
diff --git a/include/common/aarch64/el3_common_macros.S b/include/common/aarch64/el3_common_macros.S
index 63a0fa7..defd4a2 100644
--- a/include/common/aarch64/el3_common_macros.S
+++ b/include/common/aarch64/el3_common_macros.S
@@ -13,7 +13,7 @@
 	/*
 	 * Helper macro to initialise EL3 registers we care about.
 	 */
-	.macro el3_arch_init_common _exception_vectors
+	.macro el3_arch_init_common
 	/* ---------------------------------------------------------------------
 	 * SCTLR_EL3 has already been initialised - read current value before
 	 * modifying.
@@ -50,14 +50,6 @@
 #endif /* IMAGE_BL31 */
 
 	/* ---------------------------------------------------------------------
-	 * Set the exception vectors.
-	 * ---------------------------------------------------------------------
-	 */
-	adr	x0, \_exception_vectors
-	msr	vbar_el3, x0
-	isb
-
-	/* ---------------------------------------------------------------------
 	 * Initialise SCR_EL3, setting all fields rather than relying on hw.
 	 * All fields are architecturally UNKNOWN on reset. The following fields
 	 * do not change during the TF lifetime. The remaining fields are set to
@@ -221,6 +213,14 @@
 	.endif /* _warm_boot_mailbox */
 
 	/* ---------------------------------------------------------------------
+	 * Set the exception vectors.
+	 * ---------------------------------------------------------------------
+	 */
+	adr	x0, \_exception_vectors
+	msr	vbar_el3, x0
+	isb
+
+	/* ---------------------------------------------------------------------
 	 * It is a cold boot.
 	 * Perform any processor specific actions upon reset e.g. cache, TLB
 	 * invalidations etc.
@@ -228,7 +228,7 @@
 	 */
 	bl	reset_handler
 
-	el3_arch_init_common \_exception_vectors
+	el3_arch_init_common
 
 	.if \_secondary_cold_boot
 		/* -------------------------------------------------------------
diff --git a/include/lib/aarch32/arch_helpers.h b/include/lib/aarch32/arch_helpers.h
index 0230195..beae5d0 100644
--- a/include/lib/aarch32/arch_helpers.h
+++ b/include/lib/aarch32/arch_helpers.h
@@ -287,6 +287,11 @@
 DEFINE_COPROCR_RW_FUNCS(amcntenclr0, AMCNTENCLR0)
 DEFINE_COPROCR_RW_FUNCS(amcntenclr1, AMCNTENCLR1)
 
+DEFINE_COPROCR_RW_FUNCS_64(amevcntr00, AMEVCNTR00)
+DEFINE_COPROCR_RW_FUNCS_64(amevcntr01, AMEVCNTR01)
+DEFINE_COPROCR_RW_FUNCS_64(amevcntr02, AMEVCNTR02)
+DEFINE_COPROCR_RW_FUNCS_64(amevcntr03, AMEVCNTR03)
+
 /*
  * TLBI operation prototypes
  */
diff --git a/include/lib/aarch64/arch.h b/include/lib/aarch64/arch.h
index 96e2d5f..91aa484 100644
--- a/include/lib/aarch64/arch.h
+++ b/include/lib/aarch64/arch.h
@@ -117,6 +117,9 @@
 #define ID_AA64PFR0_SVE_SHIFT	U(32)
 #define ID_AA64PFR0_SVE_MASK	U(0xf)
 #define ID_AA64PFR0_SVE_LENGTH	U(4)
+#define ID_AA64PFR0_CSV2_SHIFT	U(56)
+#define ID_AA64PFR0_CSV2_MASK	U(0xf)
+#define ID_AA64PFR0_CSV2_LENGTH	U(4)
 
 /* ID_AA64DFR0_EL1.PMS definitions (for ARMv8.2+) */
 #define ID_AA64DFR0_PMS_SHIFT	U(32)
@@ -337,6 +340,11 @@
 #define SPSR_T_ARM		U(0x0)
 #define SPSR_T_THUMB		U(0x1)
 
+#define SPSR_M_SHIFT		U(4)
+#define SPSR_M_MASK		U(0x1)
+#define SPSR_M_AARCH64		U(0x0)
+#define SPSR_M_AARCH32		U(0x1)
+
 #define DISABLE_ALL_EXCEPTIONS \
 		(DAIF_FIQ_BIT | DAIF_IRQ_BIT | DAIF_ABT_BIT | DAIF_DBG_BIT)
 
@@ -656,4 +664,45 @@
 #define AMEVTYPER02_EL0		S3_3_C13_C6_2
 #define AMEVTYPER03_EL0		S3_3_C13_C6_3
 
+/* Activity Monitor Group 1 Event Counter Registers */
+#define AMEVCNTR10_EL0		S3_3_C13_C12_0
+#define AMEVCNTR11_EL0		S3_3_C13_C12_1
+#define AMEVCNTR12_EL0		S3_3_C13_C12_2
+#define AMEVCNTR13_EL0		S3_3_C13_C12_3
+#define AMEVCNTR14_EL0		S3_3_C13_C12_4
+#define AMEVCNTR15_EL0		S3_3_C13_C12_5
+#define AMEVCNTR16_EL0		S3_3_C13_C12_6
+#define AMEVCNTR17_EL0		S3_3_C13_C12_7
+#define AMEVCNTR18_EL0		S3_3_C13_C13_0
+#define AMEVCNTR19_EL0		S3_3_C13_C13_1
+#define AMEVCNTR1A_EL0		S3_3_C13_C13_2
+#define AMEVCNTR1B_EL0		S3_3_C13_C13_3
+#define AMEVCNTR1C_EL0		S3_3_C13_C13_4
+#define AMEVCNTR1D_EL0		S3_3_C13_C13_5
+#define AMEVCNTR1E_EL0		S3_3_C13_C13_6
+#define AMEVCNTR1F_EL0		S3_3_C13_C13_7
+
+/* Activity Monitor Group 1 Event Type Registers */
+#define AMEVTYPER10_EL0		S3_3_C13_C14_0
+#define AMEVTYPER11_EL0		S3_3_C13_C14_1
+#define AMEVTYPER12_EL0		S3_3_C13_C14_2
+#define AMEVTYPER13_EL0		S3_3_C13_C14_3
+#define AMEVTYPER14_EL0		S3_3_C13_C14_4
+#define AMEVTYPER15_EL0		S3_3_C13_C14_5
+#define AMEVTYPER16_EL0		S3_3_C13_C14_6
+#define AMEVTYPER17_EL0		S3_3_C13_C14_7
+#define AMEVTYPER18_EL0		S3_3_C13_C15_0
+#define AMEVTYPER19_EL0		S3_3_C13_C15_1
+#define AMEVTYPER1A_EL0		S3_3_C13_C15_2
+#define AMEVTYPER1B_EL0		S3_3_C13_C15_3
+#define AMEVTYPER1C_EL0		S3_3_C13_C15_4
+#define AMEVTYPER1D_EL0		S3_3_C13_C15_5
+#define AMEVTYPER1E_EL0		S3_3_C13_C15_6
+#define AMEVTYPER1F_EL0		S3_3_C13_C15_7
+
+/* AMCGCR_EL0 definitions */
+#define AMCGCR_EL0_CG1NC_SHIFT	U(8)
+#define AMCGCR_EL0_CG1NC_LENGTH	U(8)
+#define AMCGCR_EL0_CG1NC_MASK	U(0xff)
+
 #endif /* __ARCH_H__ */
diff --git a/include/lib/aarch64/arch_helpers.h b/include/lib/aarch64/arch_helpers.h
index 831dfb0..485ed43 100644
--- a/include/lib/aarch64/arch_helpers.h
+++ b/include/lib/aarch64/arch_helpers.h
@@ -322,6 +322,7 @@
 DEFINE_RENAME_SYSREG_WRITE_FUNC(icc_eoir1_el1, ICC_EOIR1_EL1)
 DEFINE_RENAME_SYSREG_WRITE_FUNC(icc_sgi0r_el1, ICC_SGI0R_EL1)
 
+DEFINE_RENAME_SYSREG_RW_FUNCS(amcgcr_el0, AMCGCR_EL0)
 DEFINE_RENAME_SYSREG_RW_FUNCS(amcntenclr0_el0, AMCNTENCLR0_EL0)
 DEFINE_RENAME_SYSREG_RW_FUNCS(amcntenset0_el0, AMCNTENSET0_EL0)
 DEFINE_RENAME_SYSREG_RW_FUNCS(amcntenclr1_el0, AMCNTENCLR1_EL0)
diff --git a/include/lib/cpus/aarch64/cortex_a75.h b/include/lib/cpus/aarch64/cortex_a75.h
index d68c957..940125d 100644
--- a/include/lib/cpus/aarch64/cortex_a75.h
+++ b/include/lib/cpus/aarch64/cortex_a75.h
@@ -50,7 +50,19 @@
  * CPUAMEVTYPER<n> register and are disabled by default. Platforms may
  * enable this with suitable programming.
  */
+#define CORTEX_A75_AMU_NR_COUNTERS	5
 #define CORTEX_A75_AMU_GROUP0_MASK	0x7
 #define CORTEX_A75_AMU_GROUP1_MASK	(0 << 3)
 
+#ifndef __ASSEMBLY__
+#include <stdint.h>
+
+uint64_t cortex_a75_amu_cnt_read(int idx);
+void cortex_a75_amu_cnt_write(int idx, uint64_t val);
+unsigned int cortex_a75_amu_read_cpuamcntenset_el0(void);
+unsigned int cortex_a75_amu_read_cpuamcntenclr_el0(void);
+void cortex_a75_amu_write_cpuamcntenset_el0(unsigned int mask);
+void cortex_a75_amu_write_cpuamcntenclr_el0(unsigned int mask);
+#endif /* __ASSEMBLY__ */
+
 #endif /* __CORTEX_A75_H__ */
diff --git a/include/lib/el3_runtime/aarch64/context.h b/include/lib/el3_runtime/aarch64/context.h
index 5889904..5e212ec 100644
--- a/include/lib/el3_runtime/aarch64/context.h
+++ b/include/lib/el3_runtime/aarch64/context.h
@@ -46,12 +46,26 @@
 #define CTX_GPREG_SP_EL0	U(0xf8)
 #define CTX_GPREGS_END		U(0x100)
 
+#if WORKAROUND_CVE_2017_5715
+#define CTX_CVE_2017_5715_OFFSET	(CTX_GPREGS_OFFSET + CTX_GPREGS_END)
+#define CTX_CVE_2017_5715_QUAD0		U(0x0)
+#define CTX_CVE_2017_5715_QUAD1		U(0x8)
+#define	CTX_CVE_2017_5715_QUAD2		U(0x10)
+#define CTX_CVE_2017_5715_QUAD3		U(0x18)
+#define CTX_CVE_2017_5715_QUAD4		U(0x20)
+#define CTX_CVE_2017_5715_QUAD5		U(0x28)
+#define CTX_CVE_2017_5715_END		U(0x30)
+#else
+#define CTX_CVE_2017_5715_OFFSET	CTX_GPREGS_OFFSET
+#define CTX_CVE_2017_5715_END		CTX_GPREGS_END
+#endif
+
 /*******************************************************************************
  * Constants that allow assembler code to access members of and the 'el3_state'
  * structure at their correct offsets. Note that some of the registers are only
  * 32-bits wide but are stored as 64-bit values for convenience
  ******************************************************************************/
-#define CTX_EL3STATE_OFFSET	(CTX_GPREGS_OFFSET + CTX_GPREGS_END)
+#define CTX_EL3STATE_OFFSET	(CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_END)
 #define CTX_SCR_EL3		U(0x0)
 #define CTX_RUNTIME_SP		U(0x8)
 #define CTX_SPSR_EL3		U(0x10)
@@ -186,6 +200,9 @@
 
 /* Constants to determine the size of individual context structures */
 #define CTX_GPREG_ALL		(CTX_GPREGS_END >> DWORD_SHIFT)
+#if WORKAROUND_CVE_2017_5715
+#define CTX_CVE_2017_5715_ALL	(CTX_CVE_2017_5715_END >> DWORD_SHIFT)
+#endif
 #define CTX_SYSREG_ALL		(CTX_SYSREGS_END >> DWORD_SHIFT)
 #if CTX_INCLUDE_FPREGS
 #define CTX_FPREG_ALL		(CTX_FPREGS_END >> DWORD_SHIFT)
@@ -201,6 +218,10 @@
  */
 DEFINE_REG_STRUCT(gp_regs, CTX_GPREG_ALL);
 
+#if WORKAROUND_CVE_2017_5715
+DEFINE_REG_STRUCT(cve_2017_5715_regs, CTX_CVE_2017_5715_ALL);
+#endif
+
 /*
  * AArch64 EL1 system register context structure for preserving the
  * architectural state during switches from one security state to
@@ -242,6 +263,9 @@
  */
 typedef struct cpu_context {
 	gp_regs_t gpregs_ctx;
+#if WORKAROUND_CVE_2017_5715
+	cve_2017_5715_regs_t cve_2017_5715_regs_ctx;
+#endif
 	el3_state_t el3state_ctx;
 	el1_sys_regs_t sysregs_ctx;
 #if CTX_INCLUDE_FPREGS
diff --git a/include/lib/el3_runtime/pubsub_events.h b/include/lib/el3_runtime/pubsub_events.h
index 9cfedb4..64b3f63 100644
--- a/include/lib/el3_runtime/pubsub_events.h
+++ b/include/lib/el3_runtime/pubsub_events.h
@@ -17,6 +17,13 @@
  */
 REGISTER_PUBSUB_EVENT(psci_cpu_on_finish);
 
+/*
+ * These events are published before/after a CPU has been powered down/up
+ * via the PSCI CPU SUSPEND API.
+ */
+REGISTER_PUBSUB_EVENT(psci_suspend_pwrdown_start);
+REGISTER_PUBSUB_EVENT(psci_suspend_pwrdown_finish);
+
 #ifdef AARCH64
 /*
  * These events are published by the AArch64 context management framework
diff --git a/include/lib/extensions/amu.h b/include/lib/extensions/amu.h
index bbefe8f..faa0ee1 100644
--- a/include/lib/extensions/amu.h
+++ b/include/lib/extensions/amu.h
@@ -7,9 +7,39 @@
 #ifndef __AMU_H__
 #define __AMU_H__
 
-/* Enable all group 0 counters */
+#include <sys/cdefs.h> /* for CASSERT() */
+#include <cassert.h>
+#include <platform_def.h>
+#include <stdint.h>
+
+/* All group 0 counters */
 #define AMU_GROUP0_COUNTERS_MASK	0xf
 
+#ifdef PLAT_AMU_GROUP1_COUNTERS_MASK
+#define AMU_GROUP1_COUNTERS_MASK	PLAT_AMU_GROUP1_COUNTERS_MASK
+#else
+#define AMU_GROUP1_COUNTERS_MASK	0
+#endif
+
+#ifdef PLAT_AMU_GROUP1_NR_COUNTERS
+#define AMU_GROUP1_NR_COUNTERS		PLAT_AMU_GROUP1_NR_COUNTERS
+#else
+#define AMU_GROUP1_NR_COUNTERS		0
+#endif
+
+CASSERT(AMU_GROUP1_COUNTERS_MASK <= 0xffff, invalid_amu_group1_counters_mask);
+CASSERT(AMU_GROUP1_NR_COUNTERS <= 16, invalid_amu_group1_nr_counters);
+
+int amu_supported(void);
 void amu_enable(int el2_unused);
 
+/* Group 0 configuration helpers */
+uint64_t amu_group0_cnt_read(int idx);
+void amu_group0_cnt_write(int idx, uint64_t val);
+
+/* Group 1 configuration helpers */
+uint64_t amu_group1_cnt_read(int idx);
+void amu_group1_cnt_write(int idx, uint64_t val);
+void amu_group1_set_evtype(int idx, unsigned int val);
+
 #endif /* __AMU_H__ */
diff --git a/include/lib/extensions/amu_private.h b/include/lib/extensions/amu_private.h
new file mode 100644
index 0000000..0c660bb
--- /dev/null
+++ b/include/lib/extensions/amu_private.h
@@ -0,0 +1,19 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#ifndef __AMU_PRIVATE_H__
+#define __AMU_PRIVATE_H__
+
+#include <stdint.h>
+
+uint64_t amu_group0_cnt_read_internal(int idx);
+void amu_group0_cnt_write_internal(int idx, uint64_t);
+
+uint64_t amu_group1_cnt_read_internal(int idx);
+void amu_group1_cnt_write_internal(int idx, uint64_t);
+void amu_group1_set_evtype_internal(int idx, unsigned int val);
+
+#endif /* __AMU_PRIVATE_H__ */
diff --git a/include/plat/arm/board/common/board_arm_def.h b/include/plat/arm/board/common/board_arm_def.h
index e0c3c86..888629e 100644
--- a/include/plat/arm/board/common/board_arm_def.h
+++ b/include/plat/arm/board/common/board_arm_def.h
@@ -90,7 +90,7 @@
  * PLAT_ARM_MAX_BL31_SIZE is calculated using the current BL31 debug size plus a
  * little space for growth.
  */
-#define PLAT_ARM_MAX_BL31_SIZE		0x1E000
+#define PLAT_ARM_MAX_BL31_SIZE		0x20000
 
 #ifdef AARCH32
 /*
diff --git a/lib/cpus/aarch64/cortex_a57.S b/lib/cpus/aarch64/cortex_a57.S
index a720e98..683be47 100644
--- a/lib/cpus/aarch64/cortex_a57.S
+++ b/lib/cpus/aarch64/cortex_a57.S
@@ -383,6 +383,11 @@
 	bl	errata_a57_859972_wa
 #endif
 
+#if IMAGE_BL31 && WORKAROUND_CVE_2017_5715
+	adr	x0, workaround_mmu_runtime_exceptions
+	msr	vbar_el3, x0
+#endif
+
 	/* ---------------------------------------------
 	 * Enable the SMP bit.
 	 * ---------------------------------------------
diff --git a/lib/cpus/aarch64/cortex_a72.S b/lib/cpus/aarch64/cortex_a72.S
index b034125..93821b7 100644
--- a/lib/cpus/aarch64/cortex_a72.S
+++ b/lib/cpus/aarch64/cortex_a72.S
@@ -110,6 +110,12 @@
 	mov	x0, x18
 	bl	errata_a72_859971_wa
 #endif
+
+#if IMAGE_BL31 && WORKAROUND_CVE_2017_5715
+	adr	x0, workaround_mmu_runtime_exceptions
+	msr	vbar_el3, x0
+#endif
+
 	/* ---------------------------------------------
 	 * Enable the SMP bit.
 	 * ---------------------------------------------
diff --git a/lib/cpus/aarch64/cortex_a73.S b/lib/cpus/aarch64/cortex_a73.S
index f642816..c43f07e 100644
--- a/lib/cpus/aarch64/cortex_a73.S
+++ b/lib/cpus/aarch64/cortex_a73.S
@@ -36,6 +36,11 @@
 endfunc cortex_a73_disable_smp
 
 func cortex_a73_reset_func
+#if IMAGE_BL31 && WORKAROUND_CVE_2017_5715
+	adr	x0, workaround_bpiall_vbar0_runtime_exceptions
+	msr	vbar_el3, x0
+#endif
+
 	/* ---------------------------------------------
 	 * Enable the SMP bit.
 	 * Clobbers : x0
diff --git a/lib/cpus/aarch64/cortex_a75.S b/lib/cpus/aarch64/cortex_a75.S
index 4cab9e4..e66ad06 100644
--- a/lib/cpus/aarch64/cortex_a75.S
+++ b/lib/cpus/aarch64/cortex_a75.S
@@ -11,7 +11,120 @@
 #include <plat_macros.S>
 #include <cortex_a75.h>
 
+	.globl	cortex_a75_amu_cnt_read
+	.globl	cortex_a75_amu_cnt_write
+	.globl	cortex_a75_amu_read_cpuamcntenset_el0
+	.globl	cortex_a75_amu_read_cpuamcntenclr_el0
+	.globl	cortex_a75_amu_write_cpuamcntenset_el0
+	.globl	cortex_a75_amu_write_cpuamcntenclr_el0
+
+/*
+ * uint64_t cortex_a75_amu_cnt_read(int idx);
+ *
+ * Given `idx`, read the corresponding AMU counter
+ * and return it in `x0`.
+ */
+func cortex_a75_amu_cnt_read
+	adr	x1, 1f
+	lsl	x0, x0, #3
+	add	x1, x1, x0
+	br	x1
+
+1:
+	mrs	x0, CPUAMEVCNTR0_EL0
+	ret
+	mrs	x0, CPUAMEVCNTR1_EL0
+	ret
+	mrs	x0, CPUAMEVCNTR2_EL0
+	ret
+	mrs	x0, CPUAMEVCNTR3_EL0
+	ret
+	mrs	x0, CPUAMEVCNTR4_EL0
+	ret
+endfunc cortex_a75_amu_cnt_read
+
+/*
+ * void cortex_a75_amu_cnt_write(int idx, uint64_t val);
+ *
+ * Given `idx`, write `val` to the corresponding AMU counter.
+ */
+func cortex_a75_amu_cnt_write
+	adr	x2, 1f
+	lsl	x0, x0, #3
+	add	x2, x2, x0
+	br	x2
+
+1:
+	msr	CPUAMEVCNTR0_EL0, x0
+	ret
+	msr	CPUAMEVCNTR1_EL0, x0
+	ret
+	msr	CPUAMEVCNTR2_EL0, x0
+	ret
+	msr	CPUAMEVCNTR3_EL0, x0
+	ret
+	msr	CPUAMEVCNTR4_EL0, x0
+	ret
+endfunc cortex_a75_amu_cnt_write
+
+/*
+ * unsigned int cortex_a75_amu_read_cpuamcntenset_el0(void);
+ *
+ * Read the `CPUAMCNTENSET_EL0` CPU register and return
+ * it in `x0`.
+ */
+func cortex_a75_amu_read_cpuamcntenset_el0
+	mrs	x0, CPUAMCNTENSET_EL0
+	ret
+endfunc cortex_a75_amu_read_cpuamcntenset_el0
+
+/*
+ * unsigned int cortex_a75_amu_read_cpuamcntenclr_el0(void);
+ *
+ * Read the `CPUAMCNTENCLR_EL0` CPU register and return
+ * it in `x0`.
+ */
+func cortex_a75_amu_read_cpuamcntenclr_el0
+	mrs	x0, CPUAMCNTENCLR_EL0
+	ret
+endfunc cortex_a75_amu_read_cpuamcntenclr_el0
+
+/*
+ * void cortex_a75_amu_write_cpuamcntenset_el0(unsigned int mask);
+ *
+ * Write `mask` to the `CPUAMCNTENSET_EL0` CPU register.
+ */
+func cortex_a75_amu_write_cpuamcntenset_el0
+	msr	CPUAMCNTENSET_EL0, x0
+	ret
+endfunc cortex_a75_amu_write_cpuamcntenset_el0
+
+/*
+ * void cortex_a75_amu_write_cpuamcntenclr_el0(unsigned int mask);
+ *
+ * Write `mask` to the `CPUAMCNTENCLR_EL0` CPU register.
+ */
+func cortex_a75_amu_write_cpuamcntenclr_el0
+	mrs	x0, CPUAMCNTENCLR_EL0
+	ret
+endfunc cortex_a75_amu_write_cpuamcntenclr_el0
+
 func cortex_a75_reset_func
+#if IMAGE_BL31 && WORKAROUND_CVE_2017_5715
+	mrs	x0, id_aa64pfr0_el1
+	ubfx	x0, x0, #ID_AA64PFR0_CSV2_SHIFT, #ID_AA64PFR0_CSV2_LENGTH
+	/*
+	 * If the field equals to 1 then branch targets trained in one
+	 * context cannot affect speculative execution in a different context.
+	 */
+	cmp	x0, #1
+	beq	1f
+
+	adr	x0, workaround_bpiall_vbar0_runtime_exceptions
+	msr	vbar_el3, x0
+1:
+#endif
+
 #if ENABLE_AMU
 	/* Make sure accesses from EL0/EL1 and EL2 are not trapped to EL3 */
 	mrs	x0, actlr_el3
diff --git a/lib/cpus/aarch64/cortex_a75_pubsub.c b/lib/cpus/aarch64/cortex_a75_pubsub.c
new file mode 100644
index 0000000..c1089a6
--- /dev/null
+++ b/lib/cpus/aarch64/cortex_a75_pubsub.c
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include <cortex_a75.h>
+#include <pubsub_events.h>
+#include <platform.h>
+
+struct amu_ctx {
+	uint64_t cnts[CORTEX_A75_AMU_NR_COUNTERS];
+	uint16_t mask;
+};
+
+static struct amu_ctx amu_ctxs[PLATFORM_CORE_COUNT];
+
+static void *cortex_a75_context_save(const void *arg)
+{
+	struct amu_ctx *ctx = &amu_ctxs[plat_my_core_pos()];
+	unsigned int midr;
+	unsigned int midr_mask;
+	int i;
+
+	midr = read_midr();
+	midr_mask = (MIDR_IMPL_MASK << MIDR_IMPL_SHIFT) |
+		(MIDR_PN_MASK << MIDR_PN_SHIFT);
+	if ((midr & midr_mask) != (CORTEX_A75_MIDR & midr_mask))
+		return 0;
+
+	/* Save counter configuration */
+	ctx->mask = cortex_a75_amu_read_cpuamcntenset_el0();
+
+	/* Ensure counters are disabled */
+	cortex_a75_amu_write_cpuamcntenclr_el0(ctx->mask);
+	isb();
+
+	/* Save counters */
+	for (i = 0; i < CORTEX_A75_AMU_NR_COUNTERS; i++)
+		ctx->cnts[i] = cortex_a75_amu_cnt_read(i);
+
+	return 0;
+}
+
+static void *cortex_a75_context_restore(const void *arg)
+{
+	struct amu_ctx *ctx = &amu_ctxs[plat_my_core_pos()];
+	unsigned int midr;
+	unsigned int midr_mask;
+	int i;
+
+	midr = read_midr();
+	midr_mask = (MIDR_IMPL_MASK << MIDR_IMPL_SHIFT) |
+		(MIDR_PN_MASK << MIDR_PN_SHIFT);
+	if ((midr & midr_mask) != (CORTEX_A75_MIDR & midr_mask))
+		return 0;
+
+	ctx = &amu_ctxs[plat_my_core_pos()];
+
+	/* Counters were disabled in `cortex_a75_context_save()` */
+	assert(cortex_a75_amu_read_cpuamcntenset_el0() == 0);
+
+	/* Restore counters */
+	for (i = 0; i < CORTEX_A75_AMU_NR_COUNTERS; i++)
+		cortex_a75_amu_cnt_write(i, ctx->cnts[i]);
+	isb();
+
+	/* Restore counter configuration */
+	cortex_a75_amu_write_cpuamcntenset_el0(ctx->mask);
+
+	return 0;
+}
+
+SUBSCRIBE_TO_EVENT(psci_suspend_pwrdown_start, cortex_a75_context_save);
+SUBSCRIBE_TO_EVENT(psci_suspend_pwrdown_finish, cortex_a75_context_restore);
diff --git a/lib/cpus/aarch64/workaround_cve_2017_5715_bpiall.S b/lib/cpus/aarch64/workaround_cve_2017_5715_bpiall.S
new file mode 100644
index 0000000..cd29266
--- /dev/null
+++ b/lib/cpus/aarch64/workaround_cve_2017_5715_bpiall.S
@@ -0,0 +1,372 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include <arch.h>
+#include <asm_macros.S>
+#include <context.h>
+
+	.globl	workaround_bpiall_vbar0_runtime_exceptions
+
+#define EMIT_BPIALL		0xee070fd5
+#define EMIT_MOV_R0_IMM(v)	0xe3a0000##v
+#define EMIT_SMC		0xe1600070
+
+	.macro	enter_workaround _stub_name
+	/* Save GP regs */
+	stp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	stp	x2, x3, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X2]
+	stp	x4, x5, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X4]
+	stp	x6, x7, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X6]
+	stp	x8, x9, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X8]
+	stp	x10, x11, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X10]
+	stp	x12, x13, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X12]
+	stp	x14, x15, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X14]
+	stp	x16, x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X16]
+	stp	x18, x19, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
+	stp	x20, x21, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X20]
+	stp	x22, x23, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X22]
+	stp	x24, x25, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X24]
+	stp	x26, x27, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X26]
+	stp	x28, x29, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X28]
+
+	adr	x4, \_stub_name
+
+	/*
+	 * Load SPSR_EL3 and VBAR_EL3.  SPSR_EL3 is set up to have
+	 * all interrupts masked in preparation to running the workaround
+	 * stub in S-EL1.  VBAR_EL3 points to the vector table that
+	 * will handle the SMC back from the workaround stub.
+	 */
+	ldp	x0, x1, [x4, #0]
+
+	/*
+	 * Load SCTLR_EL1 and ELR_EL3.  SCTLR_EL1 is configured to disable
+	 * the MMU in S-EL1.  ELR_EL3 points to the appropriate stub in S-EL1.
+	 */
+	ldp	x2, x3, [x4, #16]
+
+	mrs	x4, scr_el3
+	mrs	x5, spsr_el3
+	mrs	x6, elr_el3
+	mrs	x7, sctlr_el1
+	mrs	x8, esr_el3
+
+	/* Preserve system registers in the workaround context */
+	stp	x4, x5, [sp, #CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_QUAD0]
+	stp	x6, x7, [sp, #CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_QUAD2]
+	stp	x8, x30, [sp, #CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_QUAD4]
+
+	/*
+	 * Setting SCR_EL3 to all zeroes means that the NS, RW
+	 * and SMD bits are configured as expected.
+	 */
+	msr	scr_el3, xzr
+
+	/*
+	 * Reload system registers with the crafted values
+	 * in preparation for entry in S-EL1.
+	 */
+	msr	spsr_el3, x0
+	msr	vbar_el3, x1
+	msr	sctlr_el1, x2
+	msr	elr_el3, x3
+
+	eret
+	.endm
+
+	/* ---------------------------------------------------------------------
+	 * This vector table is used at runtime to enter the workaround at
+	 * AArch32 S-EL1 for Sync/IRQ/FIQ/SError exceptions.  If the workaround
+	 * is not enabled, the existing runtime exception vector table is used.
+	 * ---------------------------------------------------------------------
+	 */
+vector_base workaround_bpiall_vbar0_runtime_exceptions
+
+	/* ---------------------------------------------------------------------
+	 * Current EL with SP_EL0 : 0x0 - 0x200
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar0_sync_exception_sp_el0
+	b	sync_exception_sp_el0
+	/*
+	 * Since each vector table entry is 128 bytes, we can store the
+	 * stub context in the unused space to minimize memory footprint.
+	 */
+aarch32_stub_smc:
+	.word	EMIT_BPIALL
+	.word	EMIT_MOV_R0_IMM(1)
+	.word	EMIT_SMC
+aarch32_stub_ctx_smc:
+	/* Mask all interrupts and set AArch32 Supervisor mode */
+	.quad	(SPSR_AIF_MASK << SPSR_AIF_SHIFT | \
+	         SPSR_M_AARCH32 << SPSR_M_SHIFT | \
+	         MODE32_svc << MODE32_SHIFT)
+
+	/*
+	 * VBAR_EL3 points to vbar1 which is the vector table
+	 * used while the workaround is executing.
+	 */
+	.quad	workaround_bpiall_vbar1_runtime_exceptions
+
+	/* Setup SCTLR_EL1 with MMU off and I$ on */
+	.quad	SCTLR_AARCH32_EL1_RES1 | SCTLR_I_BIT
+
+	/* ELR_EL3 is setup to point to the sync exception stub in AArch32 */
+	.quad	aarch32_stub_smc
+	check_vector_size workaround_bpiall_vbar0_sync_exception_sp_el0
+
+vector_entry workaround_bpiall_vbar0_irq_sp_el0
+	b	irq_sp_el0
+aarch32_stub_irq:
+	.word	EMIT_BPIALL
+	.word	EMIT_MOV_R0_IMM(2)
+	.word	EMIT_SMC
+aarch32_stub_ctx_irq:
+	.quad	(SPSR_AIF_MASK << SPSR_AIF_SHIFT | \
+	         SPSR_M_AARCH32 << SPSR_M_SHIFT | \
+	         MODE32_svc << MODE32_SHIFT)
+	.quad	workaround_bpiall_vbar1_runtime_exceptions
+	.quad	SCTLR_AARCH32_EL1_RES1 | SCTLR_I_BIT
+	.quad	aarch32_stub_irq
+	check_vector_size workaround_bpiall_vbar0_irq_sp_el0
+
+vector_entry workaround_bpiall_vbar0_fiq_sp_el0
+	b	fiq_sp_el0
+aarch32_stub_fiq:
+	.word	EMIT_BPIALL
+	.word	EMIT_MOV_R0_IMM(4)
+	.word	EMIT_SMC
+aarch32_stub_ctx_fiq:
+	.quad	(SPSR_AIF_MASK << SPSR_AIF_SHIFT | \
+	         SPSR_M_AARCH32 << SPSR_M_SHIFT | \
+	         MODE32_svc << MODE32_SHIFT)
+	.quad	workaround_bpiall_vbar1_runtime_exceptions
+	.quad	SCTLR_AARCH32_EL1_RES1 | SCTLR_I_BIT
+	.quad	aarch32_stub_fiq
+	check_vector_size workaround_bpiall_vbar0_fiq_sp_el0
+
+vector_entry workaround_bpiall_vbar0_serror_sp_el0
+	b	serror_sp_el0
+aarch32_stub_serror:
+	.word	EMIT_BPIALL
+	.word	EMIT_MOV_R0_IMM(8)
+	.word	EMIT_SMC
+aarch32_stub_ctx_serror:
+	.quad	(SPSR_AIF_MASK << SPSR_AIF_SHIFT | \
+	         SPSR_M_AARCH32 << SPSR_M_SHIFT | \
+	         MODE32_svc << MODE32_SHIFT)
+	.quad	workaround_bpiall_vbar1_runtime_exceptions
+	.quad	SCTLR_AARCH32_EL1_RES1 | SCTLR_I_BIT
+	.quad	aarch32_stub_serror
+	check_vector_size workaround_bpiall_vbar0_serror_sp_el0
+
+	/* ---------------------------------------------------------------------
+	 * Current EL with SP_ELx: 0x200 - 0x400
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar0_sync_exception_sp_elx
+	b	sync_exception_sp_elx
+	check_vector_size workaround_bpiall_vbar0_sync_exception_sp_elx
+
+vector_entry workaround_bpiall_vbar0_irq_sp_elx
+	b	irq_sp_elx
+	check_vector_size workaround_bpiall_vbar0_irq_sp_elx
+
+vector_entry workaround_bpiall_vbar0_fiq_sp_elx
+	b	fiq_sp_elx
+	check_vector_size workaround_bpiall_vbar0_fiq_sp_elx
+
+vector_entry workaround_bpiall_vbar0_serror_sp_elx
+	b	serror_sp_elx
+	check_vector_size workaround_bpiall_vbar0_serror_sp_elx
+
+	/* ---------------------------------------------------------------------
+	 * Lower EL using AArch64 : 0x400 - 0x600
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar0_sync_exception_aarch64
+	enter_workaround aarch32_stub_ctx_smc
+	check_vector_size workaround_bpiall_vbar0_sync_exception_aarch64
+
+vector_entry workaround_bpiall_vbar0_irq_aarch64
+	enter_workaround aarch32_stub_ctx_irq
+	check_vector_size workaround_bpiall_vbar0_irq_aarch64
+
+vector_entry workaround_bpiall_vbar0_fiq_aarch64
+	enter_workaround aarch32_stub_ctx_fiq
+	check_vector_size workaround_bpiall_vbar0_fiq_aarch64
+
+vector_entry workaround_bpiall_vbar0_serror_aarch64
+	enter_workaround aarch32_stub_ctx_serror
+	check_vector_size workaround_bpiall_vbar0_serror_aarch64
+
+	/* ---------------------------------------------------------------------
+	 * Lower EL using AArch32 : 0x600 - 0x800
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar0_sync_exception_aarch32
+	enter_workaround aarch32_stub_ctx_smc
+	check_vector_size workaround_bpiall_vbar0_sync_exception_aarch32
+
+vector_entry workaround_bpiall_vbar0_irq_aarch32
+	enter_workaround aarch32_stub_ctx_irq
+	check_vector_size workaround_bpiall_vbar0_irq_aarch32
+
+vector_entry workaround_bpiall_vbar0_fiq_aarch32
+	enter_workaround aarch32_stub_ctx_fiq
+	check_vector_size workaround_bpiall_vbar0_fiq_aarch32
+
+vector_entry workaround_bpiall_vbar0_serror_aarch32
+	enter_workaround aarch32_stub_ctx_serror
+	check_vector_size workaround_bpiall_vbar0_serror_aarch32
+
+	/* ---------------------------------------------------------------------
+	 * This vector table is used while the workaround is executing.  It
+	 * installs a simple SMC handler to allow the Sync/IRQ/FIQ/SError
+	 * workaround stubs to enter EL3 from S-EL1.  It restores the previous
+	 * EL3 state before proceeding with the normal runtime exception vector.
+	 * ---------------------------------------------------------------------
+	 */
+vector_base workaround_bpiall_vbar1_runtime_exceptions
+
+	/* ---------------------------------------------------------------------
+	 * Current EL with SP_EL0 : 0x0 - 0x200 (UNUSED)
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar1_sync_exception_sp_el0
+	b	report_unhandled_exception
+	check_vector_size workaround_bpiall_vbar1_sync_exception_sp_el0
+
+vector_entry workaround_bpiall_vbar1_irq_sp_el0
+	b	report_unhandled_interrupt
+	check_vector_size workaround_bpiall_vbar1_irq_sp_el0
+
+vector_entry workaround_bpiall_vbar1_fiq_sp_el0
+	b	report_unhandled_interrupt
+	check_vector_size workaround_bpiall_vbar1_fiq_sp_el0
+
+vector_entry workaround_bpiall_vbar1_serror_sp_el0
+	b	report_unhandled_exception
+	check_vector_size workaround_bpiall_vbar1_serror_sp_el0
+
+	/* ---------------------------------------------------------------------
+	 * Current EL with SP_ELx: 0x200 - 0x400 (UNUSED)
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar1_sync_exception_sp_elx
+	b	report_unhandled_exception
+	check_vector_size workaround_bpiall_vbar1_sync_exception_sp_elx
+
+vector_entry workaround_bpiall_vbar1_irq_sp_elx
+	b	report_unhandled_interrupt
+	check_vector_size workaround_bpiall_vbar1_irq_sp_elx
+
+vector_entry workaround_bpiall_vbar1_fiq_sp_elx
+	b	report_unhandled_interrupt
+	check_vector_size workaround_bpiall_vbar1_fiq_sp_elx
+
+vector_entry workaround_bpiall_vbar1_serror_sp_elx
+	b	report_unhandled_exception
+	check_vector_size workaround_bpiall_vbar1_serror_sp_elx
+
+	/* ---------------------------------------------------------------------
+	 * Lower EL using AArch64 : 0x400 - 0x600 (UNUSED)
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar1_sync_exception_aarch64
+	b	report_unhandled_exception
+	check_vector_size workaround_bpiall_vbar1_sync_exception_aarch64
+
+vector_entry workaround_bpiall_vbar1_irq_aarch64
+	b	report_unhandled_interrupt
+	check_vector_size workaround_bpiall_vbar1_irq_aarch64
+
+vector_entry workaround_bpiall_vbar1_fiq_aarch64
+	b	report_unhandled_interrupt
+	check_vector_size workaround_bpiall_vbar1_fiq_aarch64
+
+vector_entry workaround_bpiall_vbar1_serror_aarch64
+	b	report_unhandled_exception
+	check_vector_size workaround_bpiall_vbar1_serror_aarch64
+
+	/* ---------------------------------------------------------------------
+	 * Lower EL using AArch32 : 0x600 - 0x800
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_bpiall_vbar1_sync_exception_aarch32
+	/* Restore register state from the workaround context */
+	ldp	x2, x3, [sp, #CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_QUAD0]
+	ldp	x4, x5, [sp, #CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_QUAD2]
+	ldp	x6, x30, [sp, #CTX_CVE_2017_5715_OFFSET + CTX_CVE_2017_5715_QUAD4]
+
+	/* Apply the restored system register state */
+	msr	scr_el3, x2
+	msr	spsr_el3, x3
+	msr	elr_el3, x4
+	msr	sctlr_el1, x5
+	msr	esr_el3, x6
+
+	/*
+	 * Workaround is complete, so swap VBAR_EL3 to point
+	 * to workaround entry table in preparation for subsequent
+	 * Sync/IRQ/FIQ/SError exceptions.
+	 */
+	adr	x2, workaround_bpiall_vbar0_runtime_exceptions
+	msr	vbar_el3, x2
+
+	/*
+	 * Restore all GP regs except x0 and x1.  The value in x0
+	 * indicates the type of the original exception.
+	 */
+	ldp	x2, x3, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X2]
+	ldp	x4, x5, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X4]
+	ldp	x6, x7, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X6]
+	ldp	x8, x9, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X8]
+	ldp	x10, x11, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X10]
+	ldp	x12, x13, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X12]
+	ldp	x14, x15, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X14]
+	ldp	x16, x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X16]
+	ldp	x18, x19, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
+	ldp	x20, x21, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X20]
+	ldp	x22, x23, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X22]
+	ldp	x24, x25, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X24]
+	ldp	x26, x27, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X26]
+	ldp	x28, x29, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X28]
+
+	/*
+	 * Each of these handlers will first restore x0 and x1 from
+	 * the context and the branch to the common implementation for
+	 * each of the exception types.
+	 */
+	tbnz	x0, #1, workaround_bpiall_vbar1_irq
+	tbnz	x0, #2, workaround_bpiall_vbar1_fiq
+	tbnz	x0, #3, workaround_bpiall_vbar1_serror
+
+	/* Fallthrough case for Sync exception */
+	ldp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	b	sync_exception_aarch64
+	check_vector_size workaround_bpiall_vbar1_sync_exception_aarch32
+
+vector_entry workaround_bpiall_vbar1_irq_aarch32
+	b	report_unhandled_interrupt
+workaround_bpiall_vbar1_irq:
+	ldp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	b	irq_aarch64
+	check_vector_size workaround_bpiall_vbar1_irq_aarch32
+
+vector_entry workaround_bpiall_vbar1_fiq_aarch32
+	b	report_unhandled_interrupt
+workaround_bpiall_vbar1_fiq:
+	ldp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	b	fiq_aarch64
+	check_vector_size workaround_bpiall_vbar1_fiq_aarch32
+
+vector_entry workaround_bpiall_vbar1_serror_aarch32
+	b	report_unhandled_exception
+workaround_bpiall_vbar1_serror:
+	ldp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	b	serror_aarch64
+	check_vector_size workaround_bpiall_vbar1_serror_aarch32
diff --git a/lib/cpus/aarch64/workaround_cve_2017_5715_mmu.S b/lib/cpus/aarch64/workaround_cve_2017_5715_mmu.S
new file mode 100644
index 0000000..f478148
--- /dev/null
+++ b/lib/cpus/aarch64/workaround_cve_2017_5715_mmu.S
@@ -0,0 +1,114 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include <arch.h>
+#include <asm_macros.S>
+#include <context.h>
+
+	.globl	workaround_mmu_runtime_exceptions
+
+vector_base workaround_mmu_runtime_exceptions
+
+	.macro	apply_workaround
+	stp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	mrs	x0, sctlr_el3
+	/* Disable MMU */
+	bic	x1, x0, #SCTLR_M_BIT
+	msr	sctlr_el3, x1
+	isb
+	/* Restore MMU config */
+	msr	sctlr_el3, x0
+	isb
+	ldp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
+	.endm
+
+	/* ---------------------------------------------------------------------
+	 * Current EL with SP_EL0 : 0x0 - 0x200
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_mmu_sync_exception_sp_el0
+	b	sync_exception_sp_el0
+	check_vector_size workaround_mmu_sync_exception_sp_el0
+
+vector_entry workaround_mmu_irq_sp_el0
+	b	irq_sp_el0
+	check_vector_size workaround_mmu_irq_sp_el0
+
+vector_entry workaround_mmu_fiq_sp_el0
+	b	fiq_sp_el0
+	check_vector_size workaround_mmu_fiq_sp_el0
+
+vector_entry workaround_mmu_serror_sp_el0
+	b	serror_sp_el0
+	check_vector_size workaround_mmu_serror_sp_el0
+
+	/* ---------------------------------------------------------------------
+	 * Current EL with SP_ELx: 0x200 - 0x400
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_mmu_sync_exception_sp_elx
+	b	sync_exception_sp_elx
+	check_vector_size workaround_mmu_sync_exception_sp_elx
+
+vector_entry workaround_mmu_irq_sp_elx
+	b	irq_sp_elx
+	check_vector_size workaround_mmu_irq_sp_elx
+
+vector_entry workaround_mmu_fiq_sp_elx
+	b	fiq_sp_elx
+	check_vector_size workaround_mmu_fiq_sp_elx
+
+vector_entry workaround_mmu_serror_sp_elx
+	b	serror_sp_elx
+	check_vector_size workaround_mmu_serror_sp_elx
+
+	/* ---------------------------------------------------------------------
+	 * Lower EL using AArch64 : 0x400 - 0x600
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_mmu_sync_exception_aarch64
+	apply_workaround
+	b	sync_exception_aarch64
+	check_vector_size workaround_mmu_sync_exception_aarch64
+
+vector_entry workaround_mmu_irq_aarch64
+	apply_workaround
+	b	irq_aarch64
+	check_vector_size workaround_mmu_irq_aarch64
+
+vector_entry workaround_mmu_fiq_aarch64
+	apply_workaround
+	b	fiq_aarch64
+	check_vector_size workaround_mmu_fiq_aarch64
+
+vector_entry workaround_mmu_serror_aarch64
+	apply_workaround
+	b	serror_aarch64
+	check_vector_size workaround_mmu_serror_aarch64
+
+	/* ---------------------------------------------------------------------
+	 * Lower EL using AArch32 : 0x600 - 0x800
+	 * ---------------------------------------------------------------------
+	 */
+vector_entry workaround_mmu_sync_exception_aarch32
+	apply_workaround
+	b	sync_exception_aarch32
+	check_vector_size workaround_mmu_sync_exception_aarch32
+
+vector_entry workaround_mmu_irq_aarch32
+	apply_workaround
+	b	irq_aarch32
+	check_vector_size workaround_mmu_irq_aarch32
+
+vector_entry workaround_mmu_fiq_aarch32
+	apply_workaround
+	b	fiq_aarch32
+	check_vector_size workaround_mmu_fiq_aarch32
+
+vector_entry workaround_mmu_serror_aarch32
+	apply_workaround
+	b	serror_aarch32
+	check_vector_size workaround_mmu_serror_aarch32
diff --git a/lib/cpus/cpu-ops.mk b/lib/cpus/cpu-ops.mk
index 31adfb4..3ba8c1f 100644
--- a/lib/cpus/cpu-ops.mk
+++ b/lib/cpus/cpu-ops.mk
@@ -16,6 +16,8 @@
 # It is enabled by default.
 A57_DISABLE_NON_TEMPORAL_HINT	?=1
 
+WORKAROUND_CVE_2017_5715	?=1
+
 # Process SKIP_A57_L1_FLUSH_PWR_DWN flag
 $(eval $(call assert_boolean,SKIP_A57_L1_FLUSH_PWR_DWN))
 $(eval $(call add_define,SKIP_A57_L1_FLUSH_PWR_DWN))
@@ -28,6 +30,9 @@
 $(eval $(call assert_boolean,A57_DISABLE_NON_TEMPORAL_HINT))
 $(eval $(call add_define,A57_DISABLE_NON_TEMPORAL_HINT))
 
+# Process WORKAROUND_CVE_2017_5715 flag
+$(eval $(call assert_boolean,WORKAROUND_CVE_2017_5715))
+$(eval $(call add_define,WORKAROUND_CVE_2017_5715))
 
 # CPU Errata Build flags.
 # These should be enabled by the platform if the erratum workaround needs to be
diff --git a/lib/extensions/amu/aarch32/amu.c b/lib/extensions/amu/aarch32/amu.c
index d450bd6..effc5bd 100644
--- a/lib/extensions/amu/aarch32/amu.c
+++ b/lib/extensions/amu/aarch32/amu.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2017-2018, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -7,26 +7,100 @@
 #include <amu.h>
 #include <arch.h>
 #include <arch_helpers.h>
+#include <platform.h>
+#include <pubsub_events.h>
+
+#define AMU_GROUP0_NR_COUNTERS	4
+
+struct amu_ctx {
+	uint64_t group0_cnts[AMU_GROUP0_NR_COUNTERS];
+};
+
+static struct amu_ctx amu_ctxs[PLATFORM_CORE_COUNT];
 
 void amu_enable(int el2_unused)
 {
 	uint64_t features;
 
 	features = read_id_pfr0() >> ID_PFR0_AMU_SHIFT;
-	if ((features & ID_PFR0_AMU_MASK) == 1) {
-		if (el2_unused) {
-			uint64_t v;
+	if ((features & ID_PFR0_AMU_MASK) != 1)
+		return;
 
-			/*
-			 * Non-secure access from EL0 or EL1 to the Activity Monitor
-			 * registers do not trap to EL2.
-			 */
-			v = read_hcptr();
-			v &= ~TAM_BIT;
-			write_hcptr(v);
-		}
+	if (el2_unused) {
+		uint64_t v;
 
-		/* Enable group 0 counters */
-		write_amcntenset0(AMU_GROUP0_COUNTERS_MASK);
+		/*
+		 * Non-secure access from EL0 or EL1 to the Activity Monitor
+		 * registers do not trap to EL2.
+		 */
+		v = read_hcptr();
+		v &= ~TAM_BIT;
+		write_hcptr(v);
 	}
+
+	/* Enable group 0 counters */
+	write_amcntenset0(AMU_GROUP0_COUNTERS_MASK);
 }
+
+static void *amu_context_save(const void *arg)
+{
+	struct amu_ctx *ctx;
+	uint64_t features;
+
+	features = read_id_pfr0() >> ID_PFR0_AMU_SHIFT;
+	if ((features & ID_PFR0_AMU_MASK) != 1)
+		return (void *)-1;
+
+	ctx = &amu_ctxs[plat_my_core_pos()];
+
+	/* Assert that group 0 counter configuration is what we expect */
+	assert(read_amcntenset0() == AMU_GROUP0_COUNTERS_MASK);
+
+	/*
+	 * Disable group 0 counters to avoid other observers like SCP sampling
+	 * counter values from the future via the memory mapped view.
+	 */
+	write_amcntenclr0(AMU_GROUP0_COUNTERS_MASK);
+	isb();
+
+	ctx->group0_cnts[0] = read64_amevcntr00();
+	ctx->group0_cnts[1] = read64_amevcntr01();
+	ctx->group0_cnts[2] = read64_amevcntr02();
+	ctx->group0_cnts[3] = read64_amevcntr03();
+
+	return 0;
+}
+
+static void *amu_context_restore(const void *arg)
+{
+	struct amu_ctx *ctx;
+	uint64_t features;
+
+	features = read_id_pfr0() >> ID_PFR0_AMU_SHIFT;
+	if ((features & ID_PFR0_AMU_MASK) != 1)
+		return (void *)-1;
+
+	ctx = &amu_ctxs[plat_my_core_pos()];
+
+	/* Counters were disabled in `amu_context_save()` */
+	assert(read_amcntenset0() == 0);
+
+	/* Restore group 0 counters */
+	if (AMU_GROUP0_COUNTERS_MASK & (1U << 0))
+		write64_amevcntr00(ctx->group0_cnts[0]);
+	if (AMU_GROUP0_COUNTERS_MASK & (1U << 1))
+		write64_amevcntr01(ctx->group0_cnts[1]);
+	if (AMU_GROUP0_COUNTERS_MASK & (1U << 2))
+		write64_amevcntr02(ctx->group0_cnts[2]);
+	if (AMU_GROUP0_COUNTERS_MASK & (1U << 3))
+		write64_amevcntr03(ctx->group0_cnts[3]);
+	isb();
+
+	/* Enable group 0 counters */
+	write_amcntenset0(AMU_GROUP0_COUNTERS_MASK);
+
+	return 0;
+}
+
+SUBSCRIBE_TO_EVENT(psci_suspend_pwrdown_start, amu_context_save);
+SUBSCRIBE_TO_EVENT(psci_suspend_pwrdown_finish, amu_context_restore);
diff --git a/lib/extensions/amu/aarch64/amu.c b/lib/extensions/amu/aarch64/amu.c
index 007b349..d7645a9 100644
--- a/lib/extensions/amu/aarch64/amu.c
+++ b/lib/extensions/amu/aarch64/amu.c
@@ -1,40 +1,185 @@
 /*
- * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2017-2018, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
 
 #include <amu.h>
+#include <amu_private.h>
 #include <arch.h>
 #include <arch_helpers.h>
+#include <assert.h>
+#include <platform.h>
+#include <pubsub_events.h>
 
-void amu_enable(int el2_unused)
+#define AMU_GROUP0_NR_COUNTERS	4
+
+struct amu_ctx {
+	uint64_t group0_cnts[AMU_GROUP0_NR_COUNTERS];
+	uint64_t group1_cnts[AMU_GROUP1_NR_COUNTERS];
+};
+
+static struct amu_ctx amu_ctxs[PLATFORM_CORE_COUNT];
+
+int amu_supported(void)
 {
 	uint64_t features;
 
 	features = read_id_aa64pfr0_el1() >> ID_AA64PFR0_AMU_SHIFT;
-	if ((features & ID_AA64PFR0_AMU_MASK) == 1) {
-		uint64_t v;
+	return (features & ID_AA64PFR0_AMU_MASK) == 1;
+}
+
+/*
+ * Enable counters.  This function is meant to be invoked
+ * by the context management library before exiting from EL3.
+ */
+void amu_enable(int el2_unused)
+{
+	uint64_t v;
 
-		if (el2_unused) {
-			/*
-			 * CPTR_EL2.TAM: Set to zero so any accesses to
-			 * the Activity Monitor registers do not trap to EL2.
-			 */
-			v = read_cptr_el2();
-			v &= ~CPTR_EL2_TAM_BIT;
-			write_cptr_el2(v);
-		}
+	if (!amu_supported())
+		return;
 
+	if (el2_unused) {
 		/*
-		 * CPTR_EL3.TAM: Set to zero so that any accesses to
-		 * the Activity Monitor registers do not trap to EL3.
+		 * CPTR_EL2.TAM: Set to zero so any accesses to
+		 * the Activity Monitor registers do not trap to EL2.
 		 */
-		v = read_cptr_el3();
-		v &= ~TAM_BIT;
-		write_cptr_el3(v);
-
-		/* Enable group 0 counters */
-		write_amcntenset0_el0(AMU_GROUP0_COUNTERS_MASK);
+		v = read_cptr_el2();
+		v &= ~CPTR_EL2_TAM_BIT;
+		write_cptr_el2(v);
 	}
+
+	/*
+	 * CPTR_EL3.TAM: Set to zero so that any accesses to
+	 * the Activity Monitor registers do not trap to EL3.
+	 */
+	v = read_cptr_el3();
+	v &= ~TAM_BIT;
+	write_cptr_el3(v);
+
+	/* Enable group 0 counters */
+	write_amcntenset0_el0(AMU_GROUP0_COUNTERS_MASK);
+	/* Enable group 1 counters */
+	write_amcntenset1_el0(AMU_GROUP1_COUNTERS_MASK);
+}
+
+/* Read the group 0 counter identified by the given `idx`. */
+uint64_t amu_group0_cnt_read(int idx)
+{
+	assert(amu_supported());
+	assert(idx >= 0 && idx < AMU_GROUP0_NR_COUNTERS);
+
+	return amu_group0_cnt_read_internal(idx);
+}
+
+/* Write the group 0 counter identified by the given `idx` with `val`. */
+void amu_group0_cnt_write(int idx, uint64_t val)
+{
+	assert(amu_supported());
+	assert(idx >= 0 && idx < AMU_GROUP0_NR_COUNTERS);
+
+	amu_group0_cnt_write_internal(idx, val);
+	isb();
+}
+
+/* Read the group 1 counter identified by the given `idx`. */
+uint64_t amu_group1_cnt_read(int idx)
+{
+	assert(amu_supported());
+	assert(idx >= 0 && idx < AMU_GROUP1_NR_COUNTERS);
+
+	return amu_group1_cnt_read_internal(idx);
+}
+
+/* Write the group 1 counter identified by the given `idx` with `val`. */
+void amu_group1_cnt_write(int idx, uint64_t val)
+{
+	assert(amu_supported());
+	assert(idx >= 0 && idx < AMU_GROUP1_NR_COUNTERS);
+
+	amu_group1_cnt_write_internal(idx, val);
+	isb();
 }
+
+/*
+ * Program the event type register for the given `idx` with
+ * the event number `val`.
+ */
+void amu_group1_set_evtype(int idx, unsigned int val)
+{
+	assert(amu_supported());
+	assert (idx >= 0 && idx < AMU_GROUP1_NR_COUNTERS);
+
+	amu_group1_set_evtype_internal(idx, val);
+	isb();
+}
+
+static void *amu_context_save(const void *arg)
+{
+	struct amu_ctx *ctx = &amu_ctxs[plat_my_core_pos()];
+	int i;
+
+	if (!amu_supported())
+		return (void *)-1;
+
+	/* Assert that group 0/1 counter configuration is what we expect */
+	assert(read_amcntenset0_el0() == AMU_GROUP0_COUNTERS_MASK &&
+	       read_amcntenset1_el0() == AMU_GROUP1_COUNTERS_MASK);
+
+	assert((sizeof(int) * 8) - __builtin_clz(AMU_GROUP1_COUNTERS_MASK)
+		<= AMU_GROUP1_NR_COUNTERS);
+
+	/*
+	 * Disable group 0/1 counters to avoid other observers like SCP sampling
+	 * counter values from the future via the memory mapped view.
+	 */
+	write_amcntenclr0_el0(AMU_GROUP0_COUNTERS_MASK);
+	write_amcntenclr1_el0(AMU_GROUP1_COUNTERS_MASK);
+	isb();
+
+	/* Save group 0 counters */
+	for (i = 0; i < AMU_GROUP0_NR_COUNTERS; i++)
+		ctx->group0_cnts[i] = amu_group0_cnt_read(i);
+
+	/* Save group 1 counters */
+	for (i = 0; i < AMU_GROUP1_NR_COUNTERS; i++)
+		ctx->group1_cnts[i] = amu_group1_cnt_read(i);
+
+	return 0;
+}
+
+static void *amu_context_restore(const void *arg)
+{
+	struct amu_ctx *ctx = &amu_ctxs[plat_my_core_pos()];
+	int i;
+
+	if (!amu_supported())
+		return (void *)-1;
+
+	/* Counters were disabled in `amu_context_save()` */
+	assert(read_amcntenset0_el0() == 0 && read_amcntenset1_el0() == 0);
+
+	assert((sizeof(int) * 8) - __builtin_clz(AMU_GROUP1_COUNTERS_MASK)
+		<= AMU_GROUP1_NR_COUNTERS);
+
+	/* Restore group 0 counters */
+	for (i = 0; i < AMU_GROUP0_NR_COUNTERS; i++)
+		if (AMU_GROUP0_COUNTERS_MASK & (1U << i))
+			amu_group0_cnt_write(i, ctx->group0_cnts[i]);
+
+	/* Restore group 1 counters */
+	for (i = 0; i < AMU_GROUP1_NR_COUNTERS; i++)
+		if (AMU_GROUP1_COUNTERS_MASK & (1U << i))
+			amu_group1_cnt_write(i, ctx->group1_cnts[i]);
+	isb();
+
+	/* Restore group 0/1 counter configuration */
+	write_amcntenset0_el0(AMU_GROUP0_COUNTERS_MASK);
+	write_amcntenset1_el0(AMU_GROUP1_COUNTERS_MASK);
+
+	return 0;
+}
+
+SUBSCRIBE_TO_EVENT(psci_suspend_pwrdown_start, amu_context_save);
+SUBSCRIBE_TO_EVENT(psci_suspend_pwrdown_finish, amu_context_restore);
diff --git a/lib/extensions/amu/aarch64/amu_helpers.S b/lib/extensions/amu/aarch64/amu_helpers.S
new file mode 100644
index 0000000..e0b1f56
--- /dev/null
+++ b/lib/extensions/amu/aarch64/amu_helpers.S
@@ -0,0 +1,281 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include <arch.h>
+#include <assert_macros.S>
+#include <asm_macros.S>
+
+	.globl	amu_group0_cnt_read_internal
+	.globl	amu_group0_cnt_write_internal
+	.globl	amu_group1_cnt_read_internal
+	.globl	amu_group1_cnt_write_internal
+	.globl	amu_group1_set_evtype_internal
+
+/*
+ * uint64_t amu_group0_cnt_read_internal(int idx);
+ *
+ * Given `idx`, read the corresponding AMU counter
+ * and return it in `x0`.
+ */
+func amu_group0_cnt_read_internal
+#if ENABLE_ASSERTIONS
+	/*
+	 * It can be dangerous to call this function with an
+	 * out of bounds index.  Ensure `idx` is valid.
+	 */
+	mov	x1, x0
+	lsr	x1, x1, #2
+	cmp	x1, #0
+	ASM_ASSERT(eq)
+#endif
+
+	/*
+	 * Given `idx` calculate address of mrs/ret instruction pair
+	 * in the table below.
+	 */
+	adr	x1, 1f
+	lsl	x0, x0, #3		/* each mrs/ret sequence is 8 bytes */
+	add	x1, x1, x0
+	br	x1
+
+1:
+	mrs	x0, AMEVCNTR00_EL0	/* index 0 */
+	ret
+	mrs	x0, AMEVCNTR01_EL0	/* index 1 */
+	ret
+	mrs	x0, AMEVCNTR02_EL0	/* index 2 */
+	ret
+	mrs	x0, AMEVCNTR03_EL0	/* index 3 */
+	ret
+endfunc amu_group0_cnt_read_internal
+
+/*
+ * void amu_group0_cnt_write_internal(int idx, uint64_t val);
+ *
+ * Given `idx`, write `val` to the corresponding AMU counter.
+ */
+func amu_group0_cnt_write_internal
+#if ENABLE_ASSERTIONS
+	/*
+	 * It can be dangerous to call this function with an
+	 * out of bounds index.  Ensure `idx` is valid.
+	 */
+	mov	x2, x0
+	lsr	x2, x2, #2
+	cmp	x2, #0
+	ASM_ASSERT(eq)
+#endif
+
+	/*
+	 * Given `idx` calculate address of mrs/ret instruction pair
+	 * in the table below.
+	 */
+	adr	x2, 1f
+	lsl	x0, x0, #3		/* each msr/ret sequence is 8 bytes */
+	add	x2, x2, x0
+	br	x2
+
+1:
+	msr	AMEVCNTR00_EL0, x1	/* index 0 */
+	ret
+	msr	AMEVCNTR01_EL0, x1	/* index 1 */
+	ret
+	msr	AMEVCNTR02_EL0, x1	/* index 2 */
+	ret
+	msr	AMEVCNTR03_EL0, x1	/* index 3 */
+	ret
+endfunc amu_group0_cnt_write_internal
+
+/*
+ * uint64_t amu_group1_cnt_read_internal(int idx);
+ *
+ * Given `idx`, read the corresponding AMU counter
+ * and return it in `x0`.
+ */
+func amu_group1_cnt_read_internal
+#if ENABLE_ASSERTIONS
+	/*
+	 * It can be dangerous to call this function with an
+	 * out of bounds index.  Ensure `idx` is valid.
+	 */
+	mov	x1, x0
+	lsr	x1, x1, #4
+	cmp	x1, #0
+	ASM_ASSERT(eq)
+#endif
+
+	/*
+	 * Given `idx` calculate address of mrs/ret instruction pair
+	 * in the table below.
+	 */
+	adr	x1, 1f
+	lsl	x0, x0, #3		/* each mrs/ret sequence is 8 bytes */
+	add	x1, x1, x0
+	br	x1
+
+1:
+	mrs	x0, AMEVCNTR10_EL0	/* index 0 */
+	ret
+	mrs	x0, AMEVCNTR11_EL0	/* index 1 */
+	ret
+	mrs	x0, AMEVCNTR12_EL0	/* index 2 */
+	ret
+	mrs	x0, AMEVCNTR13_EL0	/* index 3 */
+	ret
+	mrs	x0, AMEVCNTR14_EL0	/* index 4 */
+	ret
+	mrs	x0, AMEVCNTR15_EL0	/* index 5 */
+	ret
+	mrs	x0, AMEVCNTR16_EL0	/* index 6 */
+	ret
+	mrs	x0, AMEVCNTR17_EL0	/* index 7 */
+	ret
+	mrs	x0, AMEVCNTR18_EL0	/* index 8 */
+	ret
+	mrs	x0, AMEVCNTR19_EL0	/* index 9 */
+	ret
+	mrs	x0, AMEVCNTR1A_EL0	/* index 10 */
+	ret
+	mrs	x0, AMEVCNTR1B_EL0	/* index 11 */
+	ret
+	mrs	x0, AMEVCNTR1C_EL0	/* index 12 */
+	ret
+	mrs	x0, AMEVCNTR1D_EL0	/* index 13 */
+	ret
+	mrs	x0, AMEVCNTR1E_EL0	/* index 14 */
+	ret
+	mrs	x0, AMEVCNTR1F_EL0	/* index 15 */
+	ret
+endfunc amu_group1_cnt_read_internal
+
+/*
+ * void amu_group1_cnt_write_internal(int idx, uint64_t val);
+ *
+ * Given `idx`, write `val` to the corresponding AMU counter.
+ */
+func amu_group1_cnt_write_internal
+#if ENABLE_ASSERTIONS
+	/*
+	 * It can be dangerous to call this function with an
+	 * out of bounds index.  Ensure `idx` is valid.
+	 */
+	mov	x2, x0
+	lsr	x2, x2, #4
+	cmp	x2, #0
+	ASM_ASSERT(eq)
+#endif
+
+	/*
+	 * Given `idx` calculate address of mrs/ret instruction pair
+	 * in the table below.
+	 */
+	adr	x2, 1f
+	lsl	x0, x0, #3		/* each msr/ret sequence is 8 bytes */
+	add	x2, x2, x0
+	br	x2
+
+1:
+	msr	AMEVCNTR10_EL0, x1	/* index 0 */
+	ret
+	msr	AMEVCNTR11_EL0, x1	/* index 1 */
+	ret
+	msr	AMEVCNTR12_EL0, x1	/* index 2 */
+	ret
+	msr	AMEVCNTR13_EL0, x1	/* index 3 */
+	ret
+	msr	AMEVCNTR14_EL0, x1	/* index 4 */
+	ret
+	msr	AMEVCNTR15_EL0, x1	/* index 5 */
+	ret
+	msr	AMEVCNTR16_EL0, x1	/* index 6 */
+	ret
+	msr	AMEVCNTR17_EL0, x1	/* index 7 */
+	ret
+	msr	AMEVCNTR18_EL0, x1	/* index 8 */
+	ret
+	msr	AMEVCNTR19_EL0, x1	/* index 9 */
+	ret
+	msr	AMEVCNTR1A_EL0, x1	/* index 10 */
+	ret
+	msr	AMEVCNTR1B_EL0, x1	/* index 11 */
+	ret
+	msr	AMEVCNTR1C_EL0, x1	/* index 12 */
+	ret
+	msr	AMEVCNTR1D_EL0, x1	/* index 13 */
+	ret
+	msr	AMEVCNTR1E_EL0, x1	/* index 14 */
+	ret
+	msr	AMEVCNTR1F_EL0, x1	/* index 15 */
+	ret
+endfunc amu_group1_cnt_write_internal
+
+/*
+ * void amu_group1_set_evtype_internal(int idx, unsigned int val);
+ *
+ * Program the AMU event type register indexed by `idx`
+ * with the value `val`.
+ */
+func amu_group1_set_evtype_internal
+#if ENABLE_ASSERTIONS
+	/*
+	 * It can be dangerous to call this function with an
+	 * out of bounds index.  Ensure `idx` is valid.
+	 */
+	mov	x2, x0
+	lsr	x2, x2, #4
+	cmp	x2, #0
+	ASM_ASSERT(eq)
+
+	/* val should be between [0, 65535] */
+	mov	x2, x1
+	lsr	x2, x2, #16
+	cmp	x2, #0
+	ASM_ASSERT(eq)
+#endif
+
+	/*
+	 * Given `idx` calculate address of msr/ret instruction pair
+	 * in the table below.
+	 */
+	adr	x2, 1f
+	lsl	x0, x0, #3		/* each msr/ret sequence is 8 bytes */
+	add	x2, x2, x0
+	br	x2
+
+1:
+	msr	AMEVTYPER10_EL0, x1	/* index 0 */
+	ret
+	msr	AMEVTYPER11_EL0, x1	/* index 1 */
+	ret
+	msr	AMEVTYPER12_EL0, x1	/* index 2 */
+	ret
+	msr	AMEVTYPER13_EL0, x1	/* index 3 */
+	ret
+	msr	AMEVTYPER14_EL0, x1	/* index 4 */
+	ret
+	msr	AMEVTYPER15_EL0, x1	/* index 5 */
+	ret
+	msr	AMEVTYPER16_EL0, x1	/* index 6 */
+	ret
+	msr	AMEVTYPER17_EL0, x1	/* index 7 */
+	ret
+	msr	AMEVTYPER18_EL0, x1	/* index 8 */
+	ret
+	msr	AMEVTYPER19_EL0, x1	/* index 9 */
+	ret
+	msr	AMEVTYPER1A_EL0, x1	/* index 10 */
+	ret
+	msr	AMEVTYPER1B_EL0, x1	/* index 11 */
+	ret
+	msr	AMEVTYPER1C_EL0, x1	/* index 12 */
+	ret
+	msr	AMEVTYPER1D_EL0, x1	/* index 13 */
+	ret
+	msr	AMEVTYPER1E_EL0, x1	/* index 14 */
+	ret
+	msr	AMEVTYPER1F_EL0, x1	/* index 15 */
+	ret
+endfunc amu_group1_set_evtype_internal
diff --git a/lib/psci/psci_suspend.c b/lib/psci/psci_suspend.c
index d949067..a77972d 100644
--- a/lib/psci/psci_suspend.c
+++ b/lib/psci/psci_suspend.c
@@ -14,6 +14,7 @@
 #include <debug.h>
 #include <platform.h>
 #include <pmf.h>
+#include <pubsub_events.h>
 #include <runtime_instr.h>
 #include <stddef.h>
 #include "psci_private.h"
@@ -68,6 +69,8 @@
 {
 	unsigned int max_off_lvl = psci_find_max_off_lvl(state_info);
 
+	PUBLISH_EVENT(psci_suspend_pwrdown_start);
+
 	/* Save PSCI target power level for the suspend finisher handler */
 	psci_set_suspend_pwrlvl(end_pwrlvl);
 
@@ -308,6 +311,8 @@
 	/* Invalidate the suspend level for the cpu */
 	psci_set_suspend_pwrlvl(PSCI_INVALID_PWR_LVL);
 
+	PUBLISH_EVENT(psci_suspend_pwrdown_finish);
+
 	/*
 	 * Generic management: Now we just need to retrieve the
 	 * information that we had stashed away during the suspend
diff --git a/make_helpers/build_macros.mk b/make_helpers/build_macros.mk
index e1bfbbe..a4fbc5a 100644
--- a/make_helpers/build_macros.mk
+++ b/make_helpers/build_macros.mk
@@ -101,7 +101,7 @@
 # FIP_ADD_PAYLOAD appends the command line arguments required by fiptool
 # to package a new payload. Optionally, it adds the dependency on this payload
 #   $(1) = payload filename (i.e. bl31.bin)
-#   $(2) = command line option for the specified payload (i.e. --bl31)
+#   $(2) = command line option for the specified payload (i.e. --soc-fw)
 #   $(3) = fip target dependency (optional) (i.e. bl31)
 define FIP_ADD_PAYLOAD
     $(eval FIP_ARGS += $(2) $(1))
@@ -121,14 +121,15 @@
 # using a build option. It also adds a dependency on the image file, aborting
 # the build if the file does not exist.
 #   $(1) = build option to specify the image filename (SCP_BL2, BL33, etc)
-#   $(2) = command line option for fiptool (scp_bl2, bl33, etc)
+#   $(2) = command line option for fiptool (--scp-fw, --nt-fw, etc)
 # Example:
-#   $(eval $(call FIP_ADD_IMG,BL33,--bl33))
+#   $(eval $(call FIP_ADD_IMG,BL33,--nt-fw))
 define FIP_ADD_IMG
     CRT_DEPS += check_$(1)
     FIP_DEPS += check_$(1)
     $(call FIP_ADD_PAYLOAD,$(value $(1)),$(2))
 
+.PHONY: check_$(1)
 check_$(1):
 	$$(if $(value $(1)),,$$(error "Platform '${PLAT}' requires $(1). Please set $(1) to point to the right file"))
 endef
@@ -154,14 +155,15 @@
 
 # FWU_FIP_ADD_IMG allows the platform to pack a binary image in the FWU FIP
 #   $(1) build option to specify the image filename (BL2U, NS_BL2U, etc)
-#   $(2) command line option for fiptool (bl2u, ns_bl2u, etc)
+#   $(2) command line option for fiptool (--ap-fwu-cfg, --fwu, etc)
 # Example:
-#   $(eval $(call FWU_FIP_ADD_IMG,BL2U,--bl2u))
+#   $(eval $(call FWU_FIP_ADD_IMG,BL2U,--ap-fwu-cfg))
 define FWU_FIP_ADD_IMG
     FWU_CRT_DEPS += check_$(1)
     FWU_FIP_DEPS += check_$(1)
     $(call FWU_FIP_ADD_PAYLOAD,$(value $(1)),$(2))
 
+.PHONY: check_$(1)
 check_$(1):
 	$$(if $(value $(1)),,$$(error "Platform '${PLAT}' requires $(1). Please set $(1) to point to the right file"))
 endef
diff --git a/plat/arm/board/fvp/platform.mk b/plat/arm/board/fvp/platform.mk
index 7edbd3d..9d3c5f6 100644
--- a/plat/arm/board/fvp/platform.mk
+++ b/plat/arm/board/fvp/platform.mk
@@ -150,6 +150,10 @@
 # Enable Activity Monitor Unit extensions by default
 ENABLE_AMU			:=	1
 
+ifeq (${ENABLE_AMU},1)
+BL31_SOURCES		+= lib/cpus/aarch64/cortex_a75_pubsub.c
+endif
+
 ifneq (${ENABLE_STACK_PROTECTOR},0)
 PLAT_BL_COMMON_SOURCES	+=	plat/arm/board/fvp/fvp_stack_protector.c
 endif
diff --git a/plat/hisilicon/hikey/platform.mk b/plat/hisilicon/hikey/platform.mk
index 18b5e15..524fa6a 100644
--- a/plat/hisilicon/hikey/platform.mk
+++ b/plat/hisilicon/hikey/platform.mk
@@ -120,3 +120,5 @@
 ERRATA_A53_836870		:=	1
 ERRATA_A53_843419		:=	1
 ERRATA_A53_855873		:=	1
+
+FIP_ALIGN			:=	512
diff --git a/plat/hisilicon/hikey960/hikey960_bl1_setup.c b/plat/hisilicon/hikey960/hikey960_bl1_setup.c
index 6dfada7..ae33bd2 100644
--- a/plat/hisilicon/hikey960/hikey960_bl1_setup.c
+++ b/plat/hisilicon/hikey960/hikey960_bl1_setup.c
@@ -519,6 +519,11 @@
 	set_audio_power_up();
 	set_pcie_power_up();
 	set_isp_srt_power_up();
+
+	/* set ISP_CORE_CTRL_S to unsecure mode */
+	mmio_write_32(0xe8583800, 0x7);
+	/* set ISP_SUB_CTRL_S to unsecure mode */
+	mmio_write_32(0xe8583804, 0xf);
 }
 
 static void hikey960_ufs_reset(void)
diff --git a/plat/hisilicon/hikey960/platform.mk b/plat/hisilicon/hikey960/platform.mk
index 695f092..cb97deb 100644
--- a/plat/hisilicon/hikey960/platform.mk
+++ b/plat/hisilicon/hikey960/platform.mk
@@ -101,3 +101,5 @@
 ERRATA_A53_836870		:=	1
 ERRATA_A53_843419		:=	1
 ERRATA_A53_855873		:=	1
+
+FIP_ALIGN			:=	512
diff --git a/plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.c b/plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.c
new file mode 100644
index 0000000..bfc19d3
--- /dev/null
+++ b/plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.c
@@ -0,0 +1,129 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+/*
+ * Top-level SMC handler for ZynqMP IPI Mailbox doorbell functions.
+ */
+
+#include <bakery_lock.h>
+#include <debug.h>
+#include <errno.h>
+#include <mmio.h>
+#include <runtime_svc.h>
+#include <string.h>
+#include "ipi_mailbox_svc.h"
+#include "../zynqmp_ipi.h"
+#include "../zynqmp_private.h"
+#include "../../../services/spd/trusty/smcall.h"
+
+/*********************************************************************
+ * Macros definitions
+ ********************************************************************/
+
+/* IPI SMC calls macros: */
+#define IPI_SMC_OPEN_IRQ_MASK		0x00000001U /* IRQ enable bit in IPI
+						     * open SMC call
+						     */
+#define IPI_SMC_NOTIFY_BLOCK_MASK	0x00000001U /* Flag to indicate if
+						     * IPI notification needs
+						     * to be blocking.
+						     */
+#define IPI_SMC_ENQUIRY_DIRQ_MASK	0x00000001U /* Flag to indicate if
+						     * notification interrupt
+						     * to be disabled.
+						     */
+#define IPI_SMC_ACK_EIRQ_MASK		0x00000001U /* Flag to indicate if
+						     * notification interrupt
+						     * to be enable.
+						     */
+
+#define UNSIGNED32_MASK			0xFFFFFFFFU /* 32bit mask */
+
+/**
+ * ipi_smc_handler() - SMC handler for IPI SMC calls
+ *
+ * @smc_fid - Function identifier
+ * @x1 - x4 - Arguments
+ * @cookie  - Unused
+ * @handler - Pointer to caller's context structure
+ *
+ * @return  - Unused
+ *
+ * Determines that smc_fid is valid and supported PM SMC Function ID from the
+ * list of pm_api_ids, otherwise completes the request with
+ * the unknown SMC Function ID
+ *
+ * The SMC calls for PM service are forwarded from SIP Service SMC handler
+ * function with rt_svc_handle signature
+ */
+uint64_t ipi_smc_handler(uint32_t smc_fid, uint64_t x1, uint64_t x2,
+			 uint64_t x3, uint64_t x4, void *cookie,
+			 void *handle, uint64_t flags)
+{
+	int ret;
+	uint32_t ipi_local_id;
+	uint32_t ipi_remote_id;
+	unsigned int is_secure;
+
+	ipi_local_id = x1 & UNSIGNED32_MASK;
+	ipi_remote_id = x2 & UNSIGNED32_MASK;
+
+	if (SMC_ENTITY(smc_fid) >= SMC_ENTITY_TRUSTED_APP)
+		is_secure = 1;
+	else
+		is_secure = 0;
+
+	/* Validate IPI mailbox access */
+	ret = ipi_mb_validate(ipi_local_id, ipi_remote_id, is_secure);
+	if (ret)
+		SMC_RET1(handle, ret);
+
+	switch (SMC_FUNCTION(smc_fid)) {
+	case IPI_MAILBOX_OPEN:
+		ipi_mb_open(ipi_local_id, ipi_remote_id);
+		SMC_RET1(handle, 0);
+	case IPI_MAILBOX_RELEASE:
+		ipi_mb_release(ipi_local_id, ipi_remote_id);
+		SMC_RET1(handle, 0);
+	case IPI_MAILBOX_STATUS_ENQUIRY:
+	{
+		int disable_irq;
+
+		disable_irq = (x3 & IPI_SMC_ENQUIRY_DIRQ_MASK) ? 1 : 0;
+		ret = ipi_mb_enquire_status(ipi_local_id, ipi_remote_id);
+		if ((ret & IPI_MB_STATUS_RECV_PENDING) && disable_irq)
+			ipi_mb_disable_irq(ipi_local_id, ipi_remote_id);
+		SMC_RET1(handle, ret);
+	}
+	case IPI_MAILBOX_NOTIFY:
+	{
+		uint32_t is_blocking;
+
+		is_blocking = (x3 & IPI_SMC_NOTIFY_BLOCK_MASK) ? 1 : 0;
+		ipi_mb_notify(ipi_local_id, ipi_remote_id, is_blocking);
+		SMC_RET1(handle, 0);
+	}
+	case IPI_MAILBOX_ACK:
+	{
+		int enable_irq;
+
+		enable_irq = (x3 & IPI_SMC_ACK_EIRQ_MASK) ? 1 : 0;
+		ipi_mb_ack(ipi_local_id, ipi_remote_id);
+		if (enable_irq)
+			ipi_mb_enable_irq(ipi_local_id, ipi_remote_id);
+		SMC_RET1(handle, 0);
+	}
+	case IPI_MAILBOX_ENABLE_IRQ:
+		ipi_mb_enable_irq(ipi_local_id, ipi_remote_id);
+		SMC_RET1(handle, 0);
+	case IPI_MAILBOX_DISABLE_IRQ:
+		ipi_mb_disable_irq(ipi_local_id, ipi_remote_id);
+		SMC_RET1(handle, 0);
+	default:
+		WARN("Unimplemented IPI service call: 0x%x\n", smc_fid);
+		SMC_RET1(handle, SMC_UNK);
+	}
+}
diff --git a/plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.h b/plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.h
new file mode 100644
index 0000000..387ffd2
--- /dev/null
+++ b/plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+/* ZynqMP IPI mailbox doorbell service enums and defines */
+
+#ifndef _IPI_MAILBOX_SVC_H_
+#define _IPI_MAILBOX_SVC_H_
+
+#include <stdint.h>
+
+/*********************************************************************
+ * Enum definitions
+ ********************************************************************/
+
+/* IPI SMC function numbers enum definition */
+enum ipi_api_id {
+	/* IPI mailbox operations functions: */
+	IPI_MAILBOX_OPEN = 0x1000,
+	IPI_MAILBOX_RELEASE,
+	IPI_MAILBOX_STATUS_ENQUIRY,
+	IPI_MAILBOX_NOTIFY,
+	IPI_MAILBOX_ACK,
+	IPI_MAILBOX_ENABLE_IRQ,
+	IPI_MAILBOX_DISABLE_IRQ
+};
+
+/*********************************************************************
+ * IPI mailbox service APIs declarations
+ ********************************************************************/
+
+/* IPI SMC handler */
+uint64_t ipi_smc_handler(uint32_t smc_fid, uint64_t x1, uint64_t x2,
+			 uint64_t x3, uint64_t x4, void *cookie, void *handle,
+			 uint64_t flags);
+
+#endif /* _IPI_MAILBOX_SVC_H_ */
diff --git a/plat/xilinx/zynqmp/platform.mk b/plat/xilinx/zynqmp/platform.mk
index cb3b442..bdd194b 100644
--- a/plat/xilinx/zynqmp/platform.mk
+++ b/plat/xilinx/zynqmp/platform.mk
@@ -42,7 +42,8 @@
 PLAT_INCLUDES		:=	-Iinclude/plat/arm/common/			\
 				-Iinclude/plat/arm/common/aarch64/		\
 				-Iplat/xilinx/zynqmp/include/			\
-				-Iplat/xilinx/zynqmp/pm_service/
+				-Iplat/xilinx/zynqmp/pm_service/		\
+				-Iplat/xilinx/zynqmp/ipi_mailbox_service/
 
 PLAT_BL_COMMON_SOURCES	:=	lib/xlat_tables/xlat_tables_common.c		\
 				lib/xlat_tables/aarch64/xlat_tables.c		\
@@ -71,7 +72,9 @@
 				plat/xilinx/zynqmp/plat_startup.c		\
 				plat/xilinx/zynqmp/plat_topology.c		\
 				plat/xilinx/zynqmp/sip_svc_setup.c		\
+				plat/xilinx/zynqmp/zynqmp_ipi.c		\
 				plat/xilinx/zynqmp/pm_service/pm_svc_main.c	\
 				plat/xilinx/zynqmp/pm_service/pm_api_sys.c	\
 				plat/xilinx/zynqmp/pm_service/pm_ipi.c		\
-				plat/xilinx/zynqmp/pm_service/pm_client.c
+				plat/xilinx/zynqmp/pm_service/pm_client.c	\
+				plat/xilinx/zynqmp/ipi_mailbox_service/ipi_mailbox_svc.c
diff --git a/plat/xilinx/zynqmp/pm_service/pm_api_sys.c b/plat/xilinx/zynqmp/pm_service/pm_api_sys.c
index 90c670d..9e21067 100644
--- a/plat/xilinx/zynqmp/pm_service/pm_api_sys.c
+++ b/plat/xilinx/zynqmp/pm_service/pm_api_sys.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013-2015, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2013-2017, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -542,7 +542,6 @@
  */
 void pm_get_callbackdata(uint32_t *data, size_t count)
 {
-
 	pm_ipi_buff_read_callb(data, count);
-	pm_ipi_irq_clear();
+	pm_ipi_irq_clear(primary_proc);
 }
diff --git a/plat/xilinx/zynqmp/pm_service/pm_common.h b/plat/xilinx/zynqmp/pm_service/pm_common.h
index 03351c2..5dcbb0d 100644
--- a/plat/xilinx/zynqmp/pm_service/pm_common.h
+++ b/plat/xilinx/zynqmp/pm_service/pm_common.h
@@ -21,13 +21,13 @@
 
 /**
  * pm_ipi - struct for capturing IPI-channel specific info
- * @mask	mask for enabling/disabling and triggering the IPI
- * @base	base address for IPI
+ * @apu_ipi_id	APU IPI agent ID
+ * @pmu_ipi_id	PMU Agent ID
  * @buffer_base	base address for payload buffer
  */
 struct pm_ipi {
-	const unsigned int mask;
-	const uintptr_t base;
+	const uint32_t apu_ipi_id;
+	const uint32_t pmu_ipi_id;
 	const uintptr_t buffer_base;
 };
 
diff --git a/plat/xilinx/zynqmp/pm_service/pm_ipi.c b/plat/xilinx/zynqmp/pm_service/pm_ipi.c
index fdffde7..58faf0e 100644
--- a/plat/xilinx/zynqmp/pm_service/pm_ipi.c
+++ b/plat/xilinx/zynqmp/pm_service/pm_ipi.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013-2016, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2013-2017, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -8,28 +8,17 @@
 #include <bakery_lock.h>
 #include <mmio.h>
 #include <platform.h>
+#include "../zynqmp_ipi.h"
 #include "../zynqmp_private.h"
 #include "pm_ipi.h"
 
 /* IPI message buffers */
 #define IPI_BUFFER_BASEADDR	0xFF990000U
 
-#define IPI_BUFFER_RPU_0_BASE	(IPI_BUFFER_BASEADDR + 0x0U)
-#define IPI_BUFFER_RPU_1_BASE	(IPI_BUFFER_BASEADDR + 0x200U)
 #define IPI_BUFFER_APU_BASE	(IPI_BUFFER_BASEADDR + 0x400U)
-#define IPI_BUFFER_PL_0_BASE	(IPI_BUFFER_BASEADDR + 0x600U)
-#define IPI_BUFFER_PL_1_BASE	(IPI_BUFFER_BASEADDR + 0x800U)
-#define IPI_BUFFER_PL_2_BASE	(IPI_BUFFER_BASEADDR + 0xA00U)
-#define IPI_BUFFER_PL_3_BASE	(IPI_BUFFER_BASEADDR + 0xC00U)
 #define IPI_BUFFER_PMU_BASE	(IPI_BUFFER_BASEADDR + 0xE00U)
 
-#define IPI_BUFFER_TARGET_RPU_0_OFFSET	0x0U
-#define IPI_BUFFER_TARGET_RPU_1_OFFSET	0x40U
 #define IPI_BUFFER_TARGET_APU_OFFSET	0x80U
-#define IPI_BUFFER_TARGET_PL_0_OFFSET	0xC0U
-#define IPI_BUFFER_TARGET_PL_1_OFFSET	0x100U
-#define IPI_BUFFER_TARGET_PL_2_OFFSET	0x140U
-#define IPI_BUFFER_TARGET_PL_3_OFFSET	0x180U
 #define IPI_BUFFER_TARGET_PMU_OFFSET	0x1C0U
 
 #define IPI_BUFFER_MAX_WORDS	8
@@ -37,76 +26,33 @@
 #define IPI_BUFFER_REQ_OFFSET	0x0U
 #define IPI_BUFFER_RESP_OFFSET	0x20U
 
-/* IPI Base Address */
-#define IPI_BASEADDR		0XFF300000
-
-/* APU's IPI registers */
-#define IPI_APU_ISR		(IPI_BASEADDR + 0X00000010)
-#define IPI_APU_IER		(IPI_BASEADDR + 0X00000018)
-#define IPI_APU_IDR		(IPI_BASEADDR + 0X0000001C)
-#define IPI_APU_IXR_PMU_0_MASK		(1 << 16)
-
-#define IPI_TRIG_OFFSET		0
-#define IPI_OBS_OFFSET		4
-
-/* Power Management IPI interrupt number */
-#define PM_INT_NUM		0
-#define IPI_PMU_PM_INT_BASE	(IPI_PMU_0_TRIG + (PM_INT_NUM * 0x1000))
-#define IPI_PMU_PM_INT_MASK	(IPI_APU_IXR_PMU_0_MASK << PM_INT_NUM)
-#if (PM_INT_NUM < 0 || PM_INT_NUM > 3)
-	#error PM_INT_NUM value out of range
-#endif
-
-#define IPI_APU_MASK		1U
-
 DEFINE_BAKERY_LOCK(pm_secure_lock);
 
 const struct pm_ipi apu_ipi = {
-	.mask = IPI_APU_MASK,
-	.base = IPI_BASEADDR,
+	.apu_ipi_id = IPI_ID_APU,
+	.pmu_ipi_id = IPI_ID_PMU0,
 	.buffer_base = IPI_BUFFER_APU_BASE,
 };
 
 /**
  * pm_ipi_init() - Initialize IPI peripheral for communication with PMU
  *
+ * @proc	Pointer to the processor who is initiating request
  * @return	On success, the initialization function must return 0.
  *		Any other return value will cause the framework to ignore
  *		the service
  *
  * Called from pm_setup initialization function
  */
-int pm_ipi_init(void)
+int pm_ipi_init(const struct pm_proc *proc)
 {
 	bakery_lock_init(&pm_secure_lock);
-
-	/* IPI Interrupts Clear & Disable */
-	mmio_write_32(IPI_APU_ISR, 0xffffffff);
-	mmio_write_32(IPI_APU_IDR, 0xffffffff);
+	ipi_mb_open(proc->ipi->apu_ipi_id, proc->ipi->pmu_ipi_id);
 
 	return 0;
 }
 
 /**
- * pm_ipi_wait() - wait for pmu to handle request
- * @proc	proc which is waiting for PMU to handle request
- */
-static enum pm_ret_status pm_ipi_wait(const struct pm_proc *proc)
-{
-	int status;
-
-	/* Wait until previous interrupt is handled by PMU */
-	do {
-		status = mmio_read_32(proc->ipi->base + IPI_OBS_OFFSET) &
-					IPI_PMU_PM_INT_MASK;
-		/* TODO: 1) Use timer to add delay between read attempts */
-		/* TODO: 2) Return PM_RET_ERR_TIMEOUT if this times out */
-	} while (status);
-
-	return PM_RET_SUCCESS;
-}
-
-/**
  * pm_ipi_send_common() - Sends IPI request to the PMU
  * @proc	Pointer to the processor who is initiating request
  * @payload	API id and call arguments to be written in IPI buffer
@@ -124,16 +70,13 @@
 					IPI_BUFFER_TARGET_PMU_OFFSET +
 					IPI_BUFFER_REQ_OFFSET;
 
-	/* Wait until previous interrupt is handled by PMU */
-	pm_ipi_wait(proc);
-
 	/* Write payload into IPI buffer */
 	for (size_t i = 0; i < PAYLOAD_ARG_CNT; i++) {
 		mmio_write_32(buffer_base + offset, payload[i]);
 		offset += PAYLOAD_ARG_SIZE;
 	}
 	/* Generate IPI to PMU */
-	mmio_write_32(proc->ipi->base + IPI_TRIG_OFFSET, IPI_PMU_PM_INT_MASK);
+	ipi_mb_notify(proc->ipi->apu_ipi_id, proc->ipi->pmu_ipi_id, 1);
 
 	return PM_RET_SUCCESS;
 }
@@ -178,8 +121,6 @@
 				IPI_BUFFER_TARGET_PMU_OFFSET +
 				IPI_BUFFER_RESP_OFFSET;
 
-	pm_ipi_wait(proc);
-
 	/*
 	 * Read response from IPI buffer
 	 * buf-0: success or error+reason
@@ -250,17 +191,12 @@
 	return ret;
 }
 
-void pm_ipi_irq_enable(void)
-{
-	mmio_write_32(IPI_APU_IER, IPI_APU_IXR_PMU_0_MASK);
-}
-
-void pm_ipi_irq_disable(void)
+void pm_ipi_irq_enable(const struct pm_proc *proc)
 {
-	mmio_write_32(IPI_APU_IDR, IPI_APU_IXR_PMU_0_MASK);
+	ipi_mb_enable_irq(proc->ipi->apu_ipi_id, proc->ipi->pmu_ipi_id);
 }
 
-void pm_ipi_irq_clear(void)
+void pm_ipi_irq_clear(const struct pm_proc *proc)
 {
-	mmio_write_32(IPI_APU_ISR, IPI_APU_IXR_PMU_0_MASK);
+	ipi_mb_ack(proc->ipi->apu_ipi_id, proc->ipi->pmu_ipi_id);
 }
diff --git a/plat/xilinx/zynqmp/pm_service/pm_ipi.h b/plat/xilinx/zynqmp/pm_service/pm_ipi.h
index a76298b..e6b36f5 100644
--- a/plat/xilinx/zynqmp/pm_service/pm_ipi.h
+++ b/plat/xilinx/zynqmp/pm_service/pm_ipi.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013-2015, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2013-2017, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -9,7 +9,7 @@
 
 #include "pm_common.h"
 
-int pm_ipi_init(void);
+int pm_ipi_init(const struct pm_proc *proc);
 
 enum pm_ret_status pm_ipi_send(const struct pm_proc *proc,
 			       uint32_t payload[PAYLOAD_ARG_CNT]);
@@ -17,8 +17,7 @@
 				    uint32_t payload[PAYLOAD_ARG_CNT],
 				    unsigned int *value, size_t count);
 void pm_ipi_buff_read_callb(unsigned int *value, size_t count);
-void pm_ipi_irq_enable(void);
-void pm_ipi_irq_disable(void);
-void pm_ipi_irq_clear(void);
+void pm_ipi_irq_enable(const struct pm_proc *proc);
+void pm_ipi_irq_clear(const struct pm_proc *proc);
 
 #endif /* _PM_IPI_H_ */
diff --git a/plat/xilinx/zynqmp/pm_service/pm_svc_main.c b/plat/xilinx/zynqmp/pm_service/pm_svc_main.c
index f4e679b..fb64bc5 100644
--- a/plat/xilinx/zynqmp/pm_service/pm_svc_main.c
+++ b/plat/xilinx/zynqmp/pm_service/pm_svc_main.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013-2015, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2013-2017, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -49,22 +49,25 @@
  */
 int pm_setup(void)
 {
-	int status;
+	int status, ret;
 
 	if (!zynqmp_is_pmu_up())
 		return -ENODEV;
 
-	status = pm_ipi_init();
+	status = pm_ipi_init(primary_proc);
 
-	if (status == 0)
+	if (status >= 0) {
 		INFO("BL31: PM Service Init Complete: API v%d.%d\n",
 		     PM_VERSION_MAJOR, PM_VERSION_MINOR);
-	else
+		ret = 0;
+	} else {
 		INFO("BL31: PM Service Init Failed, Error Code %d!\n", status);
+		ret = status;
+	}
 
 	pm_down = status;
 
-	return status;
+	return ret;
 }
 
 /**
@@ -163,7 +166,7 @@
 		 * Even if we were wrong, it would not enable the IRQ in
 		 * the GIC.
 		 */
-		pm_ipi_irq_enable();
+		pm_ipi_irq_enable(primary_proc);
 		SMC_RET1(handle, (uint64_t)ret |
 			 ((uint64_t)pm_ctx.api_version << 32));
 
diff --git a/plat/xilinx/zynqmp/sip_svc_setup.c b/plat/xilinx/zynqmp/sip_svc_setup.c
index ae6ecaf..8b44eaa 100644
--- a/plat/xilinx/zynqmp/sip_svc_setup.c
+++ b/plat/xilinx/zynqmp/sip_svc_setup.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013-2015, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2013-2017, ARM Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -8,7 +8,9 @@
 
 #include <runtime_svc.h>
 #include <uuid.h>
+#include "ipi_mailbox_svc.h"
 #include "pm_svc_main.h"
+#include "zynqmp_ipi.h"
 
 /* SMC function IDs for SiP Service queries */
 #define ZYNQMP_SIP_SVC_CALL_COUNT	0x8200ff00
@@ -19,10 +21,12 @@
 #define SIP_SVC_VERSION_MAJOR	0
 #define SIP_SVC_VERSION_MINOR	1
 
-/* These macros are used to identify PM calls from the SMC function ID */
+/* These macros are used to identify PM, IPI calls from the SMC function ID */
 #define PM_FID_MASK	0xf000u
 #define PM_FID_VALUE	0u
+#define IPI_FID_VALUE	0x1000u
 #define is_pm_fid(_fid) (((_fid) & PM_FID_MASK) == PM_FID_VALUE)
+#define is_ipi_fid(_fid) (((_fid) & PM_FID_MASK) == IPI_FID_VALUE)
 
 /* SiP Service UUID */
 DEFINE_SVC_UUID(zynqmp_sip_uuid,
@@ -63,6 +67,12 @@
 				      flags);
 	}
 
+	/* Let IPI SMC handler deal with IPI-related requests */
+	if (is_ipi_fid(smc_fid)) {
+		return ipi_smc_handler(smc_fid, x1, x2, x3, x4, cookie, handle,
+				      flags);
+	}
+
 	switch (smc_fid) {
 	case ZYNQMP_SIP_SVC_CALL_COUNT:
 		/* PM functions + default functions */
diff --git a/plat/xilinx/zynqmp/zynqmp_ipi.c b/plat/xilinx/zynqmp/zynqmp_ipi.c
new file mode 100644
index 0000000..755a3b7
--- /dev/null
+++ b/plat/xilinx/zynqmp/zynqmp_ipi.c
@@ -0,0 +1,283 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+/*
+ * Zynq UltraScale+ MPSoC IPI agent registers access management
+ */
+
+#include <bakery_lock.h>
+#include <debug.h>
+#include <errno.h>
+#include <mmio.h>
+#include <runtime_svc.h>
+#include <string.h>
+#include "zynqmp_ipi.h"
+#include "../zynqmp_private.h"
+
+/*********************************************************************
+ * Macros definitions
+ ********************************************************************/
+
+/* IPI registers base address */
+#define IPI_REGS_BASE   0xFF300000U
+
+/* IPI registers offsets macros */
+#define IPI_TRIG_OFFSET 0x00U
+#define IPI_OBR_OFFSET  0x04U
+#define IPI_ISR_OFFSET  0x10U
+#define IPI_IMR_OFFSET  0x14U
+#define IPI_IER_OFFSET  0x18U
+#define IPI_IDR_OFFSET  0x1CU
+
+/* IPI register start offset */
+#define IPI_REG_BASE(I) (zynqmp_ipi_table[(I)].ipi_reg_base)
+
+/* IPI register bit mask */
+#define IPI_BIT_MASK(I) (zynqmp_ipi_table[(I)].ipi_bit_mask)
+
+/* IPI secure check */
+#define IPI_SECURE_MASK  0x1U
+#define IPI_IS_SECURE(I) ((zynqmp_ipi_table[(I)].secure_only & \
+			   IPI_SECURE_MASK) ? 1 : 0)
+
+/*********************************************************************
+ * Struct definitions
+ ********************************************************************/
+
+/* structure to maintain IPI configuration information */
+struct zynqmp_ipi_config {
+	unsigned int ipi_bit_mask;
+	unsigned int ipi_reg_base;
+	unsigned char secure_only;
+};
+
+/* Zynqmp ipi configuration table */
+const static struct zynqmp_ipi_config zynqmp_ipi_table[] = {
+	/* APU IPI */
+	{
+		.ipi_bit_mask = 0x1,
+		.ipi_reg_base = 0xFF300000,
+		.secure_only = 0,
+	},
+	/* RPU0 IPI */
+	{
+		.ipi_bit_mask = 0x100,
+		.ipi_reg_base = 0xFF310000,
+		.secure_only = 0,
+	},
+	/* RPU1 IPI */
+	{
+		.ipi_bit_mask = 0x200,
+		.ipi_reg_base = 0xFF320000,
+		.secure_only = 0,
+	},
+	/* PMU0 IPI */
+	{
+		.ipi_bit_mask = 0x10000,
+		.ipi_reg_base = 0xFF330000,
+		.secure_only = IPI_SECURE_MASK,
+	},
+	/* PMU1 IPI */
+	{
+		.ipi_bit_mask = 0x20000,
+		.ipi_reg_base = 0xFF331000,
+		.secure_only = IPI_SECURE_MASK,
+	},
+	/* PMU2 IPI */
+	{
+		.ipi_bit_mask = 0x40000,
+		.ipi_reg_base = 0xFF332000,
+		.secure_only = IPI_SECURE_MASK,
+	},
+	/* PMU3 IPI */
+	{
+		.ipi_bit_mask = 0x80000,
+		.ipi_reg_base = 0xFF333000,
+		.secure_only = IPI_SECURE_MASK,
+	},
+	/* PL0 IPI */
+	{
+		.ipi_bit_mask = 0x1000000,
+		.ipi_reg_base = 0xFF340000,
+		.secure_only = 0,
+	},
+	/* PL1 IPI */
+	{
+		.ipi_bit_mask = 0x2000000,
+		.ipi_reg_base = 0xFF350000,
+		.secure_only = 0,
+	},
+	/* PL2 IPI */
+	{
+		.ipi_bit_mask = 0x4000000,
+		.ipi_reg_base = 0xFF360000,
+		.secure_only = 0,
+	},
+	/* PL3 IPI */
+	{
+		.ipi_bit_mask = 0x8000000,
+		.ipi_reg_base = 0xFF370000,
+		.secure_only = 0,
+	},
+};
+
+/* is_ipi_mb_within_range() - verify if IPI mailbox is within range
+ *
+ * @local  - local IPI ID
+ * @remote - remote IPI ID
+ *
+ * return - 1 if within range, 0 if not
+ */
+static inline int is_ipi_mb_within_range(uint32_t local, uint32_t remote)
+{
+	int ret = 1;
+	uint32_t ipi_total = ARRAY_SIZE(zynqmp_ipi_table);
+
+	if (remote >= ipi_total || local >= ipi_total)
+		ret = 0;
+
+	return ret;
+}
+
+/**
+ * ipi_mb_validate() - validate IPI mailbox access
+ *
+ * @local  - local IPI ID
+ * @remote - remote IPI ID
+ * @is_secure - indicate if the requester is from secure software
+ *
+ * return - 0 success, negative value for errors
+ */
+int ipi_mb_validate(uint32_t local, uint32_t remote, unsigned int is_secure)
+{
+	int ret = 0;
+
+	if (!is_ipi_mb_within_range(local, remote))
+		ret = -EINVAL;
+	else if (IPI_IS_SECURE(local) && !is_secure)
+		ret = -EPERM;
+	else if (IPI_IS_SECURE(remote) && !is_secure)
+		ret = -EPERM;
+
+	return ret;
+}
+
+/**
+ * ipi_mb_open() - Open IPI mailbox.
+ *
+ * @local  - local IPI ID
+ * @remote - remote IPI ID
+ *
+ */
+void ipi_mb_open(uint32_t local, uint32_t remote)
+{
+	mmio_write_32(IPI_REG_BASE(local) + IPI_IDR_OFFSET,
+		      IPI_BIT_MASK(remote));
+	mmio_write_32(IPI_REG_BASE(local) + IPI_ISR_OFFSET,
+		      IPI_BIT_MASK(remote));
+}
+
+/**
+ * ipi_mb_release() - Open IPI mailbox.
+ *
+ * @local  - local IPI ID
+ * @remote - remote IPI ID
+ *
+ */
+void ipi_mb_release(uint32_t local, uint32_t remote)
+{
+	mmio_write_32(IPI_REG_BASE(local) + IPI_IDR_OFFSET,
+		      IPI_BIT_MASK(remote));
+}
+
+/**
+ * ipi_mb_enquire_status() - Enquire IPI mailbox status
+ *
+ * @local  - local IPI ID
+ * @remote - remote IPI ID
+ *
+ * return - 0 idle, positive value for pending sending or receiving,
+ *          negative value for errors
+ */
+int ipi_mb_enquire_status(uint32_t local, uint32_t remote)
+{
+	int ret = 0;
+	uint32_t status;
+
+	status = mmio_read_32(IPI_REG_BASE(local) + IPI_OBR_OFFSET);
+	if (status & IPI_BIT_MASK(remote))
+		ret |= IPI_MB_STATUS_SEND_PENDING;
+	status = mmio_read_32(IPI_REG_BASE(local) + IPI_ISR_OFFSET);
+	if (status & IPI_BIT_MASK(remote))
+		ret |= IPI_MB_STATUS_RECV_PENDING;
+
+	return ret;
+}
+
+/* ipi_mb_notify() - Trigger IPI mailbox notification
+ *
+ * @local - local IPI ID
+ * @remote - remote IPI ID
+ * @is_blocking - if to trigger the notification in blocking mode or not.
+ *
+ * It sets the remote bit in the IPI agent trigger register.
+ *
+ */
+void ipi_mb_notify(uint32_t local, uint32_t remote, uint32_t is_blocking)
+{
+	uint32_t status;
+
+	mmio_write_32(IPI_REG_BASE(local) + IPI_TRIG_OFFSET,
+		      IPI_BIT_MASK(remote));
+	if (is_blocking) {
+		do {
+			status = mmio_read_32(IPI_REG_BASE(local) +
+					      IPI_OBR_OFFSET);
+		} while (status & IPI_BIT_MASK(remote));
+	}
+}
+
+/* ipi_mb_ack() - Ack IPI mailbox notification from the other end
+ *
+ * @local - local IPI ID
+ * @remote - remote IPI ID
+ *
+ * It will clear the remote bit in the isr register.
+ *
+ */
+void ipi_mb_ack(uint32_t local, uint32_t remote)
+{
+	mmio_write_32(IPI_REG_BASE(local) + IPI_ISR_OFFSET,
+		      IPI_BIT_MASK(remote));
+}
+
+/* ipi_mb_disable_irq() - Disable IPI mailbox notification interrupt
+ *
+ * @local - local IPI ID
+ * @remote - remote IPI ID
+ *
+ * It will mask the remote bit in the idr register.
+ *
+ */
+void ipi_mb_disable_irq(uint32_t local, uint32_t remote)
+{
+	mmio_write_32(IPI_REG_BASE(local) + IPI_IDR_OFFSET,
+		      IPI_BIT_MASK(remote));
+}
+
+/* ipi_mb_enable_irq() - Enable IPI mailbox notification interrupt
+ *
+ * @local - local IPI ID
+ * @remote - remote IPI ID
+ *
+ * It will mask the remote bit in the idr register.
+ *
+ */
+void ipi_mb_enable_irq(uint32_t local, uint32_t remote)
+{
+	mmio_write_32(IPI_REG_BASE(local) + IPI_IER_OFFSET,
+		      IPI_BIT_MASK(remote));
+}
diff --git a/plat/xilinx/zynqmp/zynqmp_ipi.h b/plat/xilinx/zynqmp/zynqmp_ipi.h
new file mode 100644
index 0000000..0544ddb
--- /dev/null
+++ b/plat/xilinx/zynqmp/zynqmp_ipi.h
@@ -0,0 +1,70 @@
+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+/* ZynqMP IPI management enums and defines */
+
+#ifndef _ZYNQMP_IPI_H_
+#define _ZYNQMP_IPI_H_
+
+#include <stdint.h>
+
+/*********************************************************************
+ * IPI agent IDs macros
+ ********************************************************************/
+#define IPI_ID_APU	0U
+#define IPI_ID_RPU0	1U
+#define IPI_ID_RPU1	2U
+#define IPI_ID_PMU0	3U
+#define IPI_ID_PMU1	4U
+#define IPI_ID_PMU2	5U
+#define IPI_ID_PMU3	6U
+#define IPI_ID_PL0	7U
+#define IPI_ID_PL1	8U
+#define IPI_ID_PL2	9U
+#define IPI_ID_PL3	10U
+
+/*********************************************************************
+ * IPI mailbox status macros
+ ********************************************************************/
+#define IPI_MB_STATUS_IDLE		0
+#define IPI_MB_STATUS_SEND_PENDING	1
+#define IPI_MB_STATUS_RECV_PENDING	2
+
+/*********************************************************************
+ * IPI mailbox call is secure or not macros
+ ********************************************************************/
+#define IPI_MB_CALL_NOTSECURE	0
+#define IPI_MB_CALL_SECURE	1
+
+/*********************************************************************
+ * IPI APIs declarations
+ ********************************************************************/
+
+/* Validate IPI mailbox access */
+int ipi_mb_validate(uint32_t local, uint32_t remote, unsigned int is_secure);
+
+/* Open the IPI mailbox */
+void ipi_mb_open(uint32_t local, uint32_t remote);
+
+/* Release the IPI mailbox */
+void ipi_mb_release(uint32_t local, uint32_t remote);
+
+/* Enquire IPI mailbox status */
+int ipi_mb_enquire_status(uint32_t local, uint32_t remote);
+
+/* Trigger notification on the IPI mailbox */
+void ipi_mb_notify(uint32_t local, uint32_t remote, uint32_t is_blocking);
+
+/* Ack IPI mailbox notification */
+void ipi_mb_ack(uint32_t local, uint32_t remote);
+
+/* Disable IPI mailbox notification interrupt */
+void ipi_mb_disable_irq(uint32_t local, uint32_t remote);
+
+/* Enable IPI mailbox notification interrupt */
+void ipi_mb_enable_irq(uint32_t local, uint32_t remote);
+
+#endif /* _ZYNQMP_IPI_H_ */
diff --git a/services/std_svc/spm/spm_main.c b/services/std_svc/spm/spm_main.c
index ae71c1d..979b9a8 100644
--- a/services/std_svc/spm/spm_main.c
+++ b/services/std_svc/spm/spm_main.c
@@ -29,7 +29,6 @@
  * Secure Partition context information.
  ******************************************************************************/
 static secure_partition_context_t sp_ctx;
-unsigned int sp_init_in_progress;
 
 /*******************************************************************************
  * Replace the S-EL1 re-entry information with S-EL0 re-entry
@@ -126,12 +125,19 @@
 	secure_partition_setup();
 
 	/*
+	 * Make all CPUs use the same secure context.
+	 */
+	for (unsigned int i = 0; i < PLATFORM_CORE_COUNT; i++) {
+		cm_set_context_by_index(i, &sp_ctx.cpu_ctx, SECURE);
+	}
+
+	/*
 	 * Arrange for an entry into the secure partition.
 	 */
-	sp_init_in_progress = 1;
+	sp_ctx.sp_init_in_progress = 1;
 	rc = spm_synchronous_sp_entry(&sp_ctx);
 	assert(rc == 0);
-	sp_init_in_progress = 0;
+	sp_ctx.sp_init_in_progress = 0;
 	VERBOSE("SP_MEMORY_ATTRIBUTES_SET_AARCH64 availability has been revoked\n");
 
 	return rc;
@@ -358,7 +364,7 @@
 			cm_el1_sysregs_context_save(SECURE);
 			spm_setup_next_eret_into_sel0(handle);
 
-			if (sp_init_in_progress) {
+			if (sp_ctx.sp_init_in_progress) {
 				/*
 				 * SPM reports completion. The SPM must have
 				 * initiated the original request through a
@@ -370,6 +376,9 @@
 				assert(0);
 			}
 
+			/* Release the Secure Partition context */
+			spin_unlock(&sp_ctx.lock);
+
 			/*
 			 * This is the result from the Secure partition of an
 			 * earlier request. Copy the result into the non-secure
@@ -391,7 +400,7 @@
 		case SP_MEMORY_ATTRIBUTES_GET_AARCH64:
 			INFO("Received SP_MEMORY_ATTRIBUTES_GET_AARCH64 SMC\n");
 
-			if (!sp_init_in_progress) {
+			if (!sp_ctx.sp_init_in_progress) {
 				WARN("SP_MEMORY_ATTRIBUTES_GET_AARCH64 is available at boot time only\n");
 				SMC_RET1(handle, SPM_NOT_SUPPORTED);
 			}
@@ -400,7 +409,7 @@
 		case SP_MEMORY_ATTRIBUTES_SET_AARCH64:
 			INFO("Received SP_MEMORY_ATTRIBUTES_SET_AARCH64 SMC\n");
 
-			if (!sp_init_in_progress) {
+			if (!sp_ctx.sp_init_in_progress) {
 				WARN("SP_MEMORY_ATTRIBUTES_SET_AARCH64 is available at boot time only\n");
 				SMC_RET1(handle, SPM_NOT_SUPPORTED);
 			}
@@ -443,6 +452,9 @@
 			/* Save the Normal world context */
 			cm_el1_sysregs_context_save(NON_SECURE);
 
+			/* Lock the Secure Partition context. */
+			spin_lock(&sp_ctx.lock);
+
 			/*
 			 * Restore the secure world context and prepare for
 			 * entry in S-EL0
diff --git a/services/std_svc/spm/spm_private.h b/services/std_svc/spm/spm_private.h
index 16993e8..1d16b45 100644
--- a/services/std_svc/spm/spm_private.h
+++ b/services/std_svc/spm/spm_private.h
@@ -32,6 +32,7 @@
 
 #ifndef __ASSEMBLY__
 
+#include <spinlock.h>
 #include <stdint.h>
 #include <xlat_tables_v2.h>
 
@@ -43,6 +44,8 @@
 typedef struct secure_partition_context {
 	uint64_t c_rt_ctx;
 	cpu_context_t cpu_ctx;
+	unsigned int sp_init_in_progress;
+	spinlock_t lock;
 } secure_partition_context_t;
 
 uint64_t spm_secure_partition_enter(uint64_t *c_rt_ctx);
diff --git a/tools/fiptool/fiptool.c b/tools/fiptool/fiptool.c
index 1dcb7e8..33c451e 100644
--- a/tools/fiptool/fiptool.c
+++ b/tools/fiptool/fiptool.c
@@ -492,7 +492,7 @@
 	fip_toc_header_t *toc_header;
 	fip_toc_entry_t *toc_entry;
 	char *buf;
-	uint64_t entry_offset, buf_size, payload_size = 0;
+	uint64_t entry_offset, buf_size, payload_size = 0, pad_size;
 	size_t nr_images = 0;
 
 	for (desc = image_desc_head; desc != NULL; desc = desc->next)
@@ -526,9 +526,13 @@
 		entry_offset += image->toc_e.size;
 	}
 
-	/* Append a null uuid entry to mark the end of ToC entries. */
+	/*
+	 * Append a null uuid entry to mark the end of ToC entries.
+	 * NOTE the offset address for the last toc_entry must match the fip
+	 * size.
+	 */
 	memset(toc_entry, 0, sizeof(*toc_entry));
-	toc_entry->offset_address = entry_offset;
+	toc_entry->offset_address = (entry_offset + align - 1) & ~(align - 1);
 
 	/* Generate the FIP file. */
 	fp = fopen(filename, "wb");
@@ -555,6 +559,13 @@
 		xfwrite(image->buffer, image->toc_e.size, fp, filename);
 	}
 
+	if (fseek(fp, entry_offset, SEEK_SET))
+		log_errx("Failed to set file position");
+
+	pad_size = toc_entry->offset_address - entry_offset;
+	while (pad_size--)
+		fputc(0x0, fp);
+
 	fclose(fp);
 	return 0;
 }