Merge "A5DS: Change boot address to point to DDR address" into integration
diff --git a/Makefile b/Makefile
index 073c2ed..42fdb73 100644
--- a/Makefile
+++ b/Makefile
@@ -736,6 +736,7 @@
 $(eval $(call assert_boolean,RESET_TO_BL31))
 $(eval $(call assert_boolean,SAVE_KEYS))
 $(eval $(call assert_boolean,SEPARATE_CODE_AND_RODATA))
+$(eval $(call assert_boolean,SEPARATE_NOBITS_REGION))
 $(eval $(call assert_boolean,SPIN_ON_BL1_EXIT))
 $(eval $(call assert_boolean,SPM_MM))
 $(eval $(call assert_boolean,TRUSTED_BOARD_BOOT))
@@ -800,6 +801,7 @@
 $(eval $(call add_define,RAS_EXTENSION))
 $(eval $(call add_define,RESET_TO_BL31))
 $(eval $(call add_define,SEPARATE_CODE_AND_RODATA))
+$(eval $(call add_define,SEPARATE_NOBITS_REGION))
 $(eval $(call add_define,RECLAIM_INIT_CODE))
 $(eval $(call add_define,SPD_${SPD}))
 $(eval $(call add_define,SPIN_ON_BL1_EXIT))
diff --git a/bl31/bl31.ld.S b/bl31/bl31.ld.S
index a598e59..42227f0 100644
--- a/bl31/bl31.ld.S
+++ b/bl31/bl31.ld.S
@@ -15,6 +15,11 @@
 
 MEMORY {
     RAM (rwx): ORIGIN = BL31_BASE, LENGTH = BL31_LIMIT - BL31_BASE
+#if SEPARATE_NOBITS_REGION
+    NOBITS (rw!a): ORIGIN = BL31_NOBITS_BASE, LENGTH = BL31_NOBITS_LIMIT - BL31_NOBITS_BASE
+#else
+#define NOBITS RAM
+#endif
 }
 
 #ifdef PLAT_EXTRA_LD_SCRIPT
@@ -198,11 +203,28 @@
     ASSERT(. <= BL31_PROGBITS_LIMIT, "BL31 progbits has exceeded its limit.")
 #endif
 
+#if SEPARATE_NOBITS_REGION
+    /*
+     * Define a linker symbol to mark end of the RW memory area for this
+     * image.
+     */
+    __RW_END__ = .;
+    __BL31_END__ = .;
+
+    ASSERT(. <= BL31_LIMIT, "BL31 image has exceeded its limit.")
+
+    . = BL31_NOBITS_BASE;
+    ASSERT(. == ALIGN(PAGE_SIZE),
+           "BL31 NOBITS base address is not aligned on a page boundary.")
+
+    __NOBITS_START__ = .;
+#endif
+
     stacks (NOLOAD) : {
         __STACKS_START__ = .;
         *(tzfw_normal_stacks)
         __STACKS_END__ = .;
-    } >RAM
+    } >NOBITS
 
     /*
      * The .bss section gets initialised to 0 at runtime.
@@ -262,7 +284,7 @@
         __PMF_TIMESTAMP_END__ = .;
 #endif /* ENABLE_PMF */
         __BSS_END__ = .;
-    } >RAM
+    } >NOBITS
 
     /*
      * The xlat_table section is for full, aligned page tables (4K).
@@ -272,7 +294,7 @@
      */
     xlat_table (NOLOAD) : {
         *(xlat_table)
-    } >RAM
+    } >NOBITS
 
 #if USE_COHERENT_MEM
     /*
@@ -298,9 +320,18 @@
          */
         . = ALIGN(PAGE_SIZE);
         __COHERENT_RAM_END__ = .;
-    } >RAM
+    } >NOBITS
 #endif
 
+#if SEPARATE_NOBITS_REGION
+    /*
+     * Define a linker symbol to mark end of the NOBITS memory area for this
+     * image.
+     */
+    __NOBITS_END__ = .;
+
+    ASSERT(. <= BL31_NOBITS_LIMIT, "BL31 NOBITS region has exceeded its limit.")
+#else
     /*
      * Define a linker symbol to mark end of the RW memory area for this
      * image.
@@ -309,4 +340,5 @@
     __BL31_END__ = .;
 
     ASSERT(. <= BL31_LIMIT, "BL31 image has exceeded its limit.")
+#endif
 }
diff --git a/docs/design/cpu-specific-build-macros.rst b/docs/design/cpu-specific-build-macros.rst
index 891703b..7fa027f 100644
--- a/docs/design/cpu-specific-build-macros.rst
+++ b/docs/design/cpu-specific-build-macros.rst
@@ -227,6 +227,12 @@
 -  ``ERRATA_A76_1275112``: This applies errata 1275112 workaround to Cortex-A76
    CPU. This needs to be enabled only for revision <= r3p0 of the CPU.
 
+For Hercules, the following errata build flags are defined :
+
+-  ``ERRATA_HERCULES_1688305``: This applies errata 1688305 workaround to
+   Hercules CPU. This needs to be enabled only for revision r0p0 - r1p0 of
+   the CPU.
+
 For Neoverse N1, the following errata build flags are defined :
 
 -  ``ERRATA_N1_1073348``: This applies errata 1073348 workaround to Neoverse-N1
diff --git a/docs/design/firmware-design.rst b/docs/design/firmware-design.rst
index cae94b5..5fc1335 100644
--- a/docs/design/firmware-design.rst
+++ b/docs/design/firmware-design.rst
@@ -1500,6 +1500,11 @@
 this NOBITS section, making the image unnecessarily bigger. Smaller images
 allow faster loading from the FIP to the main memory.
 
+For BL31, a platform can specify an alternate location for NOBITS sections
+(other than immediately following PROGBITS sections) by setting
+``SEPARATE_NOBITS_REGION`` to 1 and defining ``BL31_NOBITS_BASE`` and
+``BL31_NOBITS_LIMIT``.
+
 Linker scripts and symbols
 ~~~~~~~~~~~~~~~~~~~~~~~~~~
 
diff --git a/docs/getting_started/build-options.rst b/docs/getting_started/build-options.rst
index 37c28a5..d7bb044 100644
--- a/docs/getting_started/build-options.rst
+++ b/docs/getting_started/build-options.rst
@@ -491,6 +491,13 @@
    pages" section in :ref:`Firmware Design`. This flag is disabled by default and
    affects all BL images.
 
+-  ``SEPARATE_NOBITS_REGION``: Setting this option to ``1`` allows the NOBITS
+   sections of BL31 (.bss, stacks, page tables, and coherent memory) to be
+   allocated in RAM discontiguous from the loaded firmware image. When set, the
+   platform is expected to provide definitons for ``BL31_NOBITS_BASE`` and
+   ``BL31_NOBITS_LIMIT``. When the option is ``0`` (the default), NOBITS
+   sections are placed in RAM immediately following the loaded firmware image.
+
 -  ``SPD``: Choose a Secure Payload Dispatcher component to be built into TF-A.
    This build option is only valid if ``ARCH=aarch64``. The value should be
    the path to the directory containing the SPD source, relative to
diff --git a/docs/getting_started/initial-build.rst b/docs/getting_started/initial-build.rst
index 41cd4d1..893aba2 100644
--- a/docs/getting_started/initial-build.rst
+++ b/docs/getting_started/initial-build.rst
@@ -8,13 +8,13 @@
 
    .. code:: shell
 
-       export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-linux-gnu-
+       export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-none-elf-
 
    For AArch32:
 
    .. code:: shell
 
-       export CROSS_COMPILE=<path-to-aarch32-gcc>/bin/arm-eabi-
+       export CROSS_COMPILE=<path-to-aarch32-gcc>/bin/arm-none-eabi-
 
    It is possible to build TF-A using Clang or Arm Compiler 6. To do so
    ``CC`` needs to point to the clang or armclang binary, which will
@@ -32,7 +32,7 @@
 
    .. code:: shell
 
-       export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-linux-gnu-
+       export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-none-elf-
        make CC=<path-to-armclang>/bin/armclang PLAT=<platform> all
 
    Clang will be selected when the base name of the path assigned to ``CC``
@@ -43,7 +43,7 @@
 
    .. code:: shell
 
-       export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-linux-gnu-
+       export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-none-elf-
        make CC=<path-to-clang>/bin/clang PLAT=<platform> all
 
 -  Change to the root directory of the TF-A source tree and build.
diff --git a/docs/getting_started/prerequisites.rst b/docs/getting_started/prerequisites.rst
index 27ad0ed..3e0c8ff 100644
--- a/docs/getting_started/prerequisites.rst
+++ b/docs/getting_started/prerequisites.rst
@@ -26,7 +26,7 @@
 |TF-A| can be built with any of the following *cross-compiler* toolchains that
 target the Armv7-A or Armv8-A architectures:
 
-- GCC >= 8.3-2019.03 (from the `Arm Developer website`_)
+- GCC >= 9.2-2019.12 (from the `Arm Developer website`_)
 - Clang >= 4.0
 - Arm Compiler >= 6.0
 
diff --git a/docs/plat/allwinner.rst b/docs/plat/allwinner.rst
index a1e0659..5c905d8 100644
--- a/docs/plat/allwinner.rst
+++ b/docs/plat/allwinner.rst
@@ -26,13 +26,13 @@
 
 .. code:: shell
 
-    make CROSS_COMPILE=aarch64-linux-gnu- PLAT=sun50i_a64 DEBUG=1 bl31
+    make CROSS_COMPILE=aarch64-none-elf- PLAT=sun50i_a64 DEBUG=1 bl31
 
 To build for machines with an H6 SoC:
 
 .. code:: shell
 
-    make CROSS_COMPILE=aarch64-linux-gnu- PLAT=sun50i_h6 DEBUG=1 bl31
+    make CROSS_COMPILE=aarch64-none-elf- PLAT=sun50i_h6 DEBUG=1 bl31
 
 .. _U-Boot documentation: http://git.denx.de/?p=u-boot.git;f=board/sunxi/README.sunxi64;hb=HEAD
 
diff --git a/docs/plat/arm/fvp/index.rst b/docs/plat/arm/fvp/index.rst
index b6396b9..37010e1 100644
--- a/docs/plat/arm/fvp/index.rst
+++ b/docs/plat/arm/fvp/index.rst
@@ -223,7 +223,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu-  \
+    CROSS_COMPILE=aarch64-none-elf-  \
     make PLAT=fvp DEBUG=1             \
     RESET_TO_BL31=1                   \
     ARM_LINUX_KERNEL_AS_BL33=1        \
diff --git a/docs/plat/arm/juno/index.rst b/docs/plat/arm/juno/index.rst
index 6429ede..cf328fa 100644
--- a/docs/plat/arm/juno/index.rst
+++ b/docs/plat/arm/juno/index.rst
@@ -136,7 +136,7 @@
 
       .. code:: shell
 
-          export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-linux-gnu-
+          export CROSS_COMPILE=<path-to-aarch64-gcc>/bin/aarch64-none-elf-
 
    -  The following parameters should be used to build BL1 and BL2 in AArch64
       and point to the BL32 file.
diff --git a/docs/plat/hikey.rst b/docs/plat/hikey.rst
index 372d388..d7a01a8 100644
--- a/docs/plat/hikey.rst
+++ b/docs/plat/hikey.rst
@@ -78,7 +78,7 @@
        EDK2_OUTPUT_DIR=${EDK2_DIR}/Build/HiKey/${BUILD_OPTION}_${AARCH64_TOOLCHAIN}
        # Build fastboot for Trusted Firmware-A. It's used for recovery mode.
        cd ${BUILD_PATH}/atf-fastboot
-       CROSS_COMPILE=aarch64-linux-gnu- make PLAT=hikey DEBUG=1
+       CROSS_COMPILE=aarch64-none-elf- make PLAT=hikey DEBUG=1
        # Convert DEBUG/RELEASE to debug/release
        FASTBOOT_BUILD_OPTION=$(echo ${BUILD_OPTION} | tr '[A-Z]' '[a-z]')
        cd ${EDK2_DIR}
diff --git a/docs/plat/imx8.rst b/docs/plat/imx8.rst
index 49ba374..cec1a39 100644
--- a/docs/plat/imx8.rst
+++ b/docs/plat/imx8.rst
@@ -40,7 +40,7 @@
 
    .. code:: shell
 
-       CROSS_COMPILE=aarch64-linux-gnu- make PLAT=<Target_SoC> bl31
+       CROSS_COMPILE=aarch64-none-elf- make PLAT=<Target_SoC> bl31
 
    Target_SoC should be "imx8qm" for i.MX8QM SoC.
    Target_SoC should be "imx8qx" for i.MX8QX SoC.
diff --git a/docs/plat/imx8m.rst b/docs/plat/imx8m.rst
index 8acd13c..0116b34 100644
--- a/docs/plat/imx8m.rst
+++ b/docs/plat/imx8m.rst
@@ -28,7 +28,7 @@
 
    .. code:: shell
 
-       CROSS_COMPILE=aarch64-linux-gnu- make PLAT=<Target_SoC> bl31
+       CROSS_COMPILE=aarch64-none-elf- make PLAT=<Target_SoC> bl31
 
    Target_SoC should be "imx8mq" for i.MX8MQ SoC.
    Target_SoC should be "imx8mm" for i.MX8MM SoC.
diff --git a/docs/plat/intel-agilex.rst b/docs/plat/intel-agilex.rst
index ff27b6b..f60bf14 100644
--- a/docs/plat/intel-agilex.rst
+++ b/docs/plat/intel-agilex.rst
@@ -41,7 +41,7 @@
 
 .. code:: bash
 
-       make CROSS_COMPILE=aarch64-linux-gnu- bl2 fip PLAT=agilex
+       make CROSS_COMPILE=aarch64-none-elf- bl2 fip PLAT=agilex
        BL33=PEI.ROM
 
 Install Procedure
diff --git a/docs/plat/intel-stratix10.rst b/docs/plat/intel-stratix10.rst
index 7f8d18e..5f33d12 100644
--- a/docs/plat/intel-stratix10.rst
+++ b/docs/plat/intel-stratix10.rst
@@ -41,7 +41,7 @@
 
 .. code:: bash
 
-       make CROSS_COMPILE=aarch64-linux-gnu- bl2 fip PLAT=stratix10
+       make CROSS_COMPILE=aarch64-none-elf- bl2 fip PLAT=stratix10
        BL33=PEI.ROM
 
 Install Procedure
diff --git a/docs/plat/ls1043a.rst b/docs/plat/ls1043a.rst
index 72a51f3..5a8f755 100644
--- a/docs/plat/ls1043a.rst
+++ b/docs/plat/ls1043a.rst
@@ -59,13 +59,13 @@
 
    .. code:: shell
 
-       CROSS_COMPILE=aarch64-linux-gnu- make PLAT=ls1043 bl1
+       CROSS_COMPILE=aarch64-none-elf- make PLAT=ls1043 bl1
 
    Build fip:
 
    .. code:: shell
 
-       CROSS_COMPILE=aarch64-linux-gnu- make PLAT=ls1043 fip \
+       CROSS_COMPILE=aarch64-none-elf- make PLAT=ls1043 fip \
        BL33=u-boot.bin NEED_BL32=yes BL32=tee.bin SPD=opteed
 
 Deploy TF-A Images
diff --git a/docs/plat/marvell/build.rst b/docs/plat/marvell/build.rst
index c10bcff..d0e37c6 100644
--- a/docs/plat/marvell/build.rst
+++ b/docs/plat/marvell/build.rst
@@ -9,7 +9,7 @@
 
     .. code:: shell
 
-        > export CROSS_COMPILE=/path/to/toolchain/aarch64-linux-gnu-
+        > export CROSS_COMPILE=/path/to/toolchain/aarch64-none-elf-
 
 (2) Set path for FIP images:
 
diff --git a/docs/plat/meson-g12a.rst b/docs/plat/meson-g12a.rst
index 7cd1bf7..8b06cc2 100644
--- a/docs/plat/meson-g12a.rst
+++ b/docs/plat/meson-g12a.rst
@@ -17,7 +17,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make DEBUG=1 PLAT=g12a
+    CROSS_COMPILE=aarch64-none-elf- make DEBUG=1 PLAT=g12a
 
 This port has been tested on a SEI510 board. After building it, follow the
 instructions in the `gxlimg repository` or `U-Boot repository`_, replacing the
diff --git a/docs/plat/meson-gxbb.rst b/docs/plat/meson-gxbb.rst
index 2cd8342..f891ecd 100644
--- a/docs/plat/meson-gxbb.rst
+++ b/docs/plat/meson-gxbb.rst
@@ -17,7 +17,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make DEBUG=1 PLAT=gxbb bl31
+    CROSS_COMPILE=aarch64-none-elf- make DEBUG=1 PLAT=gxbb bl31
 
 This port has been tested in a ODROID-C2. After building it, follow the
 instructions in the `U-Boot repository`_, replacing the mentioned **bl31.bin**
diff --git a/docs/plat/meson-gxl.rst b/docs/plat/meson-gxl.rst
index c6d8504..170d7c7 100644
--- a/docs/plat/meson-gxl.rst
+++ b/docs/plat/meson-gxl.rst
@@ -17,7 +17,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make DEBUG=1 PLAT=gxl
+    CROSS_COMPILE=aarch64-none-elf- make DEBUG=1 PLAT=gxl
 
 This port has been tested on a Lepotato. After building it, follow the
 instructions in the `gxlimg repository` or `U-Boot repository`_, replacing the
diff --git a/docs/plat/mt8183.rst b/docs/plat/mt8183.rst
index c639be1..7a0830f 100644
--- a/docs/plat/mt8183.rst
+++ b/docs/plat/mt8183.rst
@@ -17,4 +17,4 @@
 
 .. code:: shell
 
-    make CROSS_COMPILE=aarch64-linux-gnu- PLAT=mt8183 DEBUG=1
+    make CROSS_COMPILE=aarch64-none-elf- PLAT=mt8183 DEBUG=1
diff --git a/docs/plat/poplar.rst b/docs/plat/poplar.rst
index 215f551..040b593 100644
--- a/docs/plat/poplar.rst
+++ b/docs/plat/poplar.rst
@@ -68,7 +68,7 @@
 
 .. code:: bash
 
-       make CROSS_COMPILE=aarch64-linux-gnu-  all fip SPD=none PLAT=poplar
+       make CROSS_COMPILE=aarch64-none-elf- all fip SPD=none PLAT=poplar
        BL33=u-boot.bin
 
 -  Build l-loader (generated the final fastboot.bin)
diff --git a/docs/plat/qemu-sbsa.rst b/docs/plat/qemu-sbsa.rst
index 51fe414..5e8535c 100644
--- a/docs/plat/qemu-sbsa.rst
+++ b/docs/plat/qemu-sbsa.rst
@@ -27,7 +27,7 @@
 
     git clone https://git.trustedfirmware.org/TF-A/trusted-firmware-a.git tfa
     cd tfa
-    export CROSS_COMPILE=aarch64-linux-gnu-
+    export CROSS_COMPILE=aarch64-none-elf-
     make PLAT=qemu_sbsa all fip
 
 Images will be placed at build/qemu_sbsa/release (bl1.bin and fip.bin).
diff --git a/docs/plat/rockchip.rst b/docs/plat/rockchip.rst
index b7c43fb..2c3c38f 100644
--- a/docs/plat/rockchip.rst
+++ b/docs/plat/rockchip.rst
@@ -35,7 +35,7 @@
 
 For AARCH64 architectures the build command looks like
 
-    make CROSS_COMPILE=aarch64-linux-gnu- PLAT=rk3399 bl32
+    make CROSS_COMPILE=aarch64-none-elf- PLAT=rk3399 bl32
 
 while AARCH32 needs a slightly different command
 
diff --git a/docs/plat/rpi3.rst b/docs/plat/rpi3.rst
index 38c3dfa..5c7b6d5 100644
--- a/docs/plat/rpi3.rst
+++ b/docs/plat/rpi3.rst
@@ -315,7 +315,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make PLAT=rpi3             \
+    CROSS_COMPILE=aarch64-none-elf- make PLAT=rpi3             \
     RPI3_BL33_IN_AARCH32=1                                      \
     BL33=../rpi3-arm-tf-bootstrap/aarch32/el2-bootstrap.bin
 
@@ -323,7 +323,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make PLAT=rpi3             \
+    CROSS_COMPILE=aarch64-none-elf- make PLAT=rpi3             \
     BL33=../rpi3-arm-tf-bootstrap/aarch64/el2-bootstrap.bin
 
 However, enabling PSCI support in a 64-bit kernel is really easy. In the
@@ -340,7 +340,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make PLAT=rpi3             \
+    CROSS_COMPILE=aarch64-none-elf- make PLAT=rpi3             \
     PRELOADED_BL33_BASE=0x02000000                              \
     RPI3_PRELOADED_DTB_BASE=0x01000000                          \
     RPI3_DIRECT_LINUX_BOOT=1
@@ -349,7 +349,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make PLAT=rpi3             \
+    CROSS_COMPILE=aarch64-none-elf- make PLAT=rpi3             \
     PRELOADED_BL33_BASE=0x02000000                              \
     RPI3_PRELOADED_DTB_BASE=0x01000000                          \
     RPI3_DIRECT_LINUX_BOOT=1                                    \
diff --git a/docs/plat/rpi4.rst b/docs/plat/rpi4.rst
index 0f529c1..e87aef7 100644
--- a/docs/plat/rpi4.rst
+++ b/docs/plat/rpi4.rst
@@ -22,7 +22,7 @@
 
 .. code:: shell
 
-    CROSS_COMPILE=aarch64-linux-gnu- make PLAT=rpi4 DEBUG=1
+    CROSS_COMPILE=aarch64-none-elf- make PLAT=rpi4 DEBUG=1
 
 Copy the generated build/rpi4/debug/bl31.bin to the SD card, either
 renaming it to ``armstub8.bin`` or adding an entry starting with ``armstub=``,
diff --git a/docs/plat/ti-k3.rst b/docs/plat/ti-k3.rst
index 4843227..2626e67 100644
--- a/docs/plat/ti-k3.rst
+++ b/docs/plat/ti-k3.rst
@@ -27,7 +27,7 @@
 
 .. code:: shell
 
-    make CROSS_COMPILE=aarch64-linux-gnu- PLAT=k3 SPD=opteed all
+    make CROSS_COMPILE=aarch64-none-elf- PLAT=k3 SPD=opteed all
 
 OP-TEE:
 
diff --git a/drivers/arm/scu/scu.c b/drivers/arm/scu/scu.c
new file mode 100644
index 0000000..aceac92
--- /dev/null
+++ b/drivers/arm/scu/scu.c
@@ -0,0 +1,51 @@
+/*
+ * Copyright (c) 2019, Arm Limited. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include <assert.h>
+#include <drivers/arm/scu.h>
+#include <lib/mmio.h>
+#include <plat/common/platform.h>
+#include <stdint.h>
+
+/*******************************************************************************
+ * Turn ON snoop control unit. This is needed to synchronize the data between
+ * CPU's.
+ ******************************************************************************/
+void enable_snoop_ctrl_unit(uintptr_t base)
+{
+	uint32_t scu_ctrl;
+
+	INFO("[SCU]: enabling snoop control unit ... \n");
+
+	assert(base != 0U);
+	scu_ctrl = mmio_read_32(base + SCU_CTRL_REG);
+
+	/* already enabled? */
+	if ((scu_ctrl & SCU_ENABLE_BIT) != 0) {
+		return;
+	}
+
+	scu_ctrl |= SCU_ENABLE_BIT;
+	mmio_write_32(base + SCU_CTRL_REG, scu_ctrl);
+}
+
+/*******************************************************************************
+ * Snoop Control Unit configuration register. This is read-only register and
+ * contains information such as
+ * - number of CPUs present
+ * - is a particular CPU operating in SMP mode or AMP mode
+ * - data cache size of a particular CPU
+ * - does SCU has ACP port
+ * - is L2CPRESENT
+ * NOTE: user of this API should interpert the bits in this register according
+ * to the TRM
+ ******************************************************************************/
+uint32_t read_snoop_ctrl_unit_cfg(uintptr_t base)
+{
+	assert(base != 0U);
+
+	return mmio_read_32(base + SCU_CFG_REG);
+}
diff --git a/include/arch/aarch64/el3_common_macros.S b/include/arch/aarch64/el3_common_macros.S
index b14b7b6..156b18a 100644
--- a/include/arch/aarch64/el3_common_macros.S
+++ b/include/arch/aarch64/el3_common_macros.S
@@ -382,6 +382,14 @@
 		add	x1, x1, :lo12:__RW_END__
 		sub	x1, x1, x0
 		bl	inv_dcache_range
+#if defined(IMAGE_BL31) && SEPARATE_NOBITS_REGION
+		adrp	x0, __NOBITS_START__
+		add	x0, x0, :lo12:__NOBITS_START__
+		adrp	x1, __NOBITS_END__
+		add	x1, x1, :lo12:__NOBITS_END__
+		sub	x1, x1, x0
+		bl	inv_dcache_range
+#endif
 #endif
 		adrp	x0, __BSS_START__
 		add	x0, x0, :lo12:__BSS_START__
diff --git a/include/drivers/arm/scu.h b/include/drivers/arm/scu.h
new file mode 100644
index 0000000..992539f
--- /dev/null
+++ b/include/drivers/arm/scu.h
@@ -0,0 +1,20 @@
+/*
+ * Copyright (c) 2019, Arm Limited. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#ifndef SCU_H
+#define SCU_H
+
+#include <stdint.h>
+
+#define SCU_CTRL_REG	0x00
+#define SCU_CFG_REG	0x04
+
+#define SCU_ENABLE_BIT	(1 << 0)
+
+void enable_snoop_ctrl_unit(uintptr_t base);
+uint32_t read_snoop_ctrl_unit_cfg(uintptr_t base);
+
+#endif /* SCU_H */
diff --git a/include/lib/cpus/aarch64/cortex_a72.h b/include/lib/cpus/aarch64/cortex_a72.h
index 4a444c6..28b440e 100644
--- a/include/lib/cpus/aarch64/cortex_a72.h
+++ b/include/lib/cpus/aarch64/cortex_a72.h
@@ -43,7 +43,14 @@
  ******************************************************************************/
 #define CORTEX_A72_L2ACTLR_EL1					S3_1_C15_C0_0
 
+#define CORTEX_A72_L2ACTLR_FORCE_TAG_BANK_CLK_ACTIVE		(ULL(1) << 28)
+#define CORTEX_A72_L2ACTLR_FORCE_L2_LOGIC_CLK_ACTIVE		(ULL(1) << 27)
+#define CORTEX_A72_L2ACTLR_FORCE_L2_GIC_TIMER_RCG_CLK_ACTIVE	(ULL(1) << 26)
 #define CORTEX_A72_L2ACTLR_ENABLE_UNIQUE_CLEAN			(ULL(1) << 14)
+#define CORTEX_A72_L2ACTLR_DISABLE_DSB_WITH_NO_DVM_SYNC		(ULL(1) << 11)
+#define CORTEX_A72_L2ACTLR_DISABLE_DVM_CMO_BROADCAST		(ULL(1) << 8)
+#define CORTEX_A72_L2ACTLR_ENABLE_HAZARD_DETECT_TIMEOUT		(ULL(1) << 7)
+#define CORTEX_A72_L2ACTLR_DISABLE_ACE_SH_OR_CHI		(ULL(1) << 6)
 
 /*******************************************************************************
  * L2 Control register specific definitions.
@@ -51,8 +58,12 @@
 #define CORTEX_A72_L2CTLR_EL1				S3_1_C11_C0_2
 
 #define CORTEX_A72_L2CTLR_DATA_RAM_LATENCY_SHIFT	U(0)
+#define CORTEX_A72_L2CTLR_DATA_RAM_SETUP_SHIFT		U(5)
 #define CORTEX_A72_L2CTLR_TAG_RAM_LATENCY_SHIFT		U(6)
+#define CORTEX_A72_L2CTLR_TAG_RAM_SETUP_SHIFT		U(9)
 
+#define CORTEX_A72_L2_DATA_RAM_LATENCY_MASK		U(0x7)
+#define CORTEX_A72_L2_TAG_RAM_LATENCY_MASK		U(0x7)
 #define CORTEX_A72_L2_DATA_RAM_LATENCY_3_CYCLES		U(0x2)
 #define CORTEX_A72_L2_TAG_RAM_LATENCY_2_CYCLES		U(0x1)
 #define CORTEX_A72_L2_TAG_RAM_LATENCY_3_CYCLES		U(0x2)
diff --git a/include/lib/cpus/aarch64/cortex_hercules.h b/include/lib/cpus/aarch64/cortex_hercules.h
index b943e7a..d5ca85e 100644
--- a/include/lib/cpus/aarch64/cortex_hercules.h
+++ b/include/lib/cpus/aarch64/cortex_hercules.h
@@ -27,6 +27,9 @@
  ******************************************************************************/
 #define CORTEX_HERCULES_ACTLR_TAM_BIT				(ULL(1) << 30)
 
+#define CORTEX_HERCULES_ACTLR2_EL1				S3_0_C15_C1_1
+#define CORTEX_HERCULES_ACTLR2_EL1_BIT_1			(ULL(1) << 1)
+
 /*******************************************************************************
  * CPU Activity Monitor Unit register specific definitions.
  ******************************************************************************/
diff --git a/lib/cpus/aarch64/cortex_hercules.S b/lib/cpus/aarch64/cortex_hercules.S
index 4e04814..a239196 100644
--- a/lib/cpus/aarch64/cortex_hercules.S
+++ b/lib/cpus/aarch64/cortex_hercules.S
@@ -16,12 +16,49 @@
 #error "cortex_hercules must be compiled with HW_ASSISTED_COHERENCY enabled"
 #endif
 
+
+/* --------------------------------------------------
+ * Errata Workaround for Hercules Erratum 1688305.
+ * This applies to revision r0p0 and r1p0 of Hercules.
+ * Inputs:
+ * x0: variant[4:7] and revision[0:3] of current cpu.
+ * Shall clobber: x0-x17
+ * --------------------------------------------------
+ */
+func errata_hercules_1688305_wa
+	/* Compare x0 against revision r1p0 */
+	mov	x17, x30
+	bl	check_errata_1688305
+	cbz	x0, 1f
+	mrs     x1, CORTEX_HERCULES_ACTLR2_EL1
+	orr	x1, x1, CORTEX_HERCULES_ACTLR2_EL1_BIT_1
+	msr     CORTEX_HERCULES_ACTLR2_EL1, x1
+	isb
+1:
+	ret	x17
+endfunc errata_hercules_1688305_wa
+
+func check_errata_1688305
+	/* Applies to r0p0 and r1p0 */
+	mov	x1, #0x10
+	b	cpu_rev_var_ls
+endfunc check_errata_1688305
+
 	/* -------------------------------------------------
 	 * The CPU Ops reset function for Cortex-Hercules
 	 * -------------------------------------------------
 	 */
-#if ENABLE_AMU
 func cortex_hercules_reset_func
+	mov	x19, x30
+	bl	cpu_get_rev_var
+	mov	x18, x0
+
+#if ERRATA_HERCULES_1688305
+	mov     x0, x18
+	bl	errata_hercules_1688305_wa
+#endif
+
+#if ENABLE_AMU
 	/* Make sure accesses from EL0/EL1 and EL2 are not trapped to EL3 */
 	mrs	x0, actlr_el3
 	bic	x0, x0, #CORTEX_HERCULES_ACTLR_TAM_BIT
@@ -39,11 +76,11 @@
 	/* Enable group1 counters */
 	mov	x0, #CORTEX_HERCULES_AMU_GROUP1_MASK
 	msr	CPUAMCNTENSET1_EL0, x0
-	isb
+#endif
 
-	ret
+	isb
+	ret	x19
 endfunc cortex_hercules_reset_func
-#endif
 
 	/* ---------------------------------------------
 	 * HW will do the cache maintenance while powering down
@@ -66,6 +103,18 @@
 	 */
 #if REPORT_ERRATA
 func cortex_hercules_errata_report
+	stp	x8, x30, [sp, #-16]!
+
+	bl	cpu_get_rev_var
+	mov	x8, x0
+
+	/*
+	 * Report all errata. The revision-variant information is passed to
+	 * checking functions of each errata.
+	 */
+	report_errata ERRATA_HERCULES_1688305, cortex_hercules, 1688305
+
+	ldp	x8, x30, [sp], #16
 	ret
 endfunc cortex_hercules_errata_report
 #endif
@@ -89,12 +138,6 @@
 	ret
 endfunc cortex_hercules_cpu_reg_dump
 
-#if ENABLE_AMU
-#define HERCULES_RESET_FUNC cortex_hercules_reset_func
-#else
-#define HERCULES_RESET_FUNC CPU_NO_RESET_FUNC
-#endif
-
 declare_cpu_ops cortex_hercules, CORTEX_HERCULES_MIDR, \
-	HERCULES_RESET_FUNC, \
+	cortex_hercules_reset_func, \
 	cortex_hercules_core_pwr_dwn
diff --git a/lib/cpus/cpu-ops.mk b/lib/cpus/cpu-ops.mk
index 078888e..c9b9b38 100644
--- a/lib/cpus/cpu-ops.mk
+++ b/lib/cpus/cpu-ops.mk
@@ -234,6 +234,10 @@
 # only to revision <= r3p0 of the Cortex A76 cpu.
 ERRATA_A76_1286807	?=0
 
+# Flag to apply erratum 1688305 workaround during reset. This erratum applies
+# to revisions r0p0 - r1p0 of the Hercules cpu.
+ERRATA_HERCULES_1688305	?=0
+
 # Flag to apply T32 CLREX workaround during reset. This erratum applies
 # only to r0p0 and r1p0 of the Neoverse N1 cpu.
 ERRATA_N1_1043202	?=1
@@ -467,6 +471,10 @@
 $(eval $(call assert_boolean,ERRATA_A76_1286807))
 $(eval $(call add_define,ERRATA_A76_1286807))
 
+# Process ERRATA_HERCULES_1688305 flag
+$(eval $(call assert_boolean,ERRATA_HERCULES_1688305))
+$(eval $(call add_define,ERRATA_HERCULES_1688305))
+
 # Process ERRATA_N1_1043202 flag
 $(eval $(call assert_boolean,ERRATA_N1_1043202))
 $(eval $(call add_define,ERRATA_N1_1043202))
diff --git a/make_helpers/defaults.mk b/make_helpers/defaults.mk
index a1f9db9..53832c5 100644
--- a/make_helpers/defaults.mk
+++ b/make_helpers/defaults.mk
@@ -171,6 +171,10 @@
 # platform Makefile is free to override this value.
 SEPARATE_CODE_AND_RODATA	:= 0
 
+# Put NOBITS sections (.bss, stacks, page tables, and coherent memory) in a
+# separate memory region, which may be discontiguous from the rest of BL31.
+SEPARATE_NOBITS_REGION		:= 0
+
 # If the BL31 image initialisation code is recalimed after use for the secondary
 # cores stack
 RECLAIM_INIT_CODE		:= 0
diff --git a/plat/allwinner/common/allwinner-common.mk b/plat/allwinner/common/allwinner-common.mk
index 5e8885d..d281011 100644
--- a/plat/allwinner/common/allwinner-common.mk
+++ b/plat/allwinner/common/allwinner-common.mk
@@ -8,8 +8,7 @@
 
 AW_PLAT			:=	plat/allwinner
 
-PLAT_INCLUDES		:=	-Iinclude/plat/arm/common		\
-				-Iinclude/plat/arm/common/aarch64	\
+PLAT_INCLUDES		:=	-Iinclude/plat/arm/common/aarch64	\
 				-I${AW_PLAT}/common/include		\
 				-I${AW_PLAT}/${PLAT}/include
 
diff --git a/plat/arm/board/a5ds/a5ds_pm.c b/plat/arm/board/a5ds/a5ds_pm.c
index cc734b0..7774002 100644
--- a/plat/arm/board/a5ds/a5ds_pm.c
+++ b/plat/arm/board/a5ds/a5ds_pm.c
@@ -4,11 +4,10 @@
  * SPDX-License-Identifier: BSD-3-Clause
  */
 #include <assert.h>
-
+#include <drivers/arm/gicv2.h>
 #include <lib/psci/psci.h>
 #include <plat/arm/common/plat_arm.h>
 #include <plat/common/platform.h>
-#include <drivers/arm/gicv2.h>
 
 /*******************************************************************************
  * Platform handler called when a power domain is about to be turned on. The
diff --git a/plat/arm/board/a5ds/include/platform_def.h b/plat/arm/board/a5ds/include/platform_def.h
index 322645f..4c87c22 100644
--- a/plat/arm/board/a5ds/include/platform_def.h
+++ b/plat/arm/board/a5ds/include/platform_def.h
@@ -332,6 +332,9 @@
 #define A5DS_HOLD_STATE_WAIT	0
 #define A5DS_HOLD_STATE_GO	1
 
+/* Snoop Control Unit base address */
+#define A5DS_SCU_BASE			0x1C000000
+
 /*
  * GIC related constants to cater for GICv2
  */
diff --git a/plat/arm/board/a5ds/sp_min/a5ds_sp_min_setup.c b/plat/arm/board/a5ds/sp_min/a5ds_sp_min_setup.c
index 8b45af8..a951dc7 100644
--- a/plat/arm/board/a5ds/sp_min/a5ds_sp_min_setup.c
+++ b/plat/arm/board/a5ds/sp_min/a5ds_sp_min_setup.c
@@ -4,12 +4,17 @@
  * SPDX-License-Identifier: BSD-3-Clause
  */
 
+#include <drivers/arm/scu.h>
 #include <plat/arm/common/plat_arm.h>
 
+
 void plat_arm_sp_min_early_platform_setup(u_register_t arg0, u_register_t arg1,
 			u_register_t arg2, u_register_t arg3)
 {
 	arm_sp_min_early_platform_setup((void *)arg0, arg1, arg2, (void *)arg3);
+
+	/* enable snoop control unit */
+	enable_snoop_ctrl_unit(A5DS_SCU_BASE);
 }
 
 /*
diff --git a/plat/arm/board/a5ds/sp_min/sp_min-a5ds.mk b/plat/arm/board/a5ds/sp_min/sp_min-a5ds.mk
index da1d785..4b0c97d 100644
--- a/plat/arm/board/a5ds/sp_min/sp_min-a5ds.mk
+++ b/plat/arm/board/a5ds/sp_min/sp_min-a5ds.mk
@@ -5,7 +5,8 @@
 #
 
 # SP_MIN source files specific to A5DS platform
-BL32_SOURCES	+=	drivers/cfi/v2m/v2m_flash.c			\
+BL32_SOURCES	+=	drivers/arm/scu/scu.c                           \
+			drivers/cfi/v2m/v2m_flash.c			\
 			lib/utils/mem_region.c				\
 			lib/aarch32/arm32_aeabi_divmod.c		\
 			lib/aarch32/arm32_aeabi_divmod_a32.S		\
diff --git a/plat/mediatek/mt8183/drivers/spm/spm.c b/plat/mediatek/mt8183/drivers/spm/spm.c
index 547af57..d6d2344 100644
--- a/plat/mediatek/mt8183/drivers/spm/spm.c
+++ b/plat/mediatek/mt8183/drivers/spm/spm.c
@@ -12,6 +12,12 @@
 
 DEFINE_BAKERY_LOCK(spm_lock);
 
+/* SPM_DVS_LEVEL */
+#define SPM_VMODEM_LEVEL_MASK	(0xff << 16)
+#define SPM_VMODEM_LEVEL	(1U << 18)
+#define SPM_VCORE_LEVEL_MASK	(0xff)
+#define SPM_VCORE_LEVEL		(1U << 1)
+
 /* CLK_SCP_CFG_0 */
 #define SPM_CK_OFF_CONTROL	(0x3FF)
 
@@ -339,6 +345,11 @@
 	spm_lock_init();
 	mt_spm_pmic_wrap_set_phase(PMIC_WRAP_PHASE_ALLINONE);
 
+	/* Set Vmodem / Vcore DVS init level */
+	mmio_clrsetbits_32(SPM_DVS_LEVEL,
+			   SPM_VMODEM_LEVEL_MASK | SPM_VCORE_LEVEL_MASK,
+			   SPM_VMODEM_LEVEL | SPM_VCORE_LEVEL);
+
 	/* switch ck_off/axi_26m control to SPM */
 	mmio_setbits_32(CLK_SCP_CFG_0, SPM_CK_OFF_CONTROL);
 	mmio_setbits_32(CLK_SCP_CFG_1, SPM_AXI_26M_SEL);
diff --git a/plat/rockchip/rk3328/platform.mk b/plat/rockchip/rk3328/platform.mk
index 0c5cfae..0219422 100644
--- a/plat/rockchip/rk3328/platform.mk
+++ b/plat/rockchip/rk3328/platform.mk
@@ -59,6 +59,9 @@
 include lib/coreboot/coreboot.mk
 include lib/libfdt/libfdt.mk
 
+# Enable workarounds for selected Cortex-A53 errata
+ERRATA_A53_855873	:=	1
+
 $(eval $(call add_define,PLAT_EXTRA_LD_SCRIPT))
 $(eval $(call add_define,PLAT_SKIP_OPTEE_S_EL1_INT_REGISTER))
 
diff --git a/plat/socionext/uniphier/uniphier_image_desc.c b/plat/socionext/uniphier/uniphier_image_desc.c
index 9e171e0..817029a 100644
--- a/plat/socionext/uniphier/uniphier_image_desc.c
+++ b/plat/socionext/uniphier/uniphier_image_desc.c
@@ -80,7 +80,7 @@
 				      VERSION_2, entry_point_info_t,
 				      NON_SECURE | EXECUTABLE),
 		.ep_info.pc = UNIPHIER_BL33_BASE,
-		.ep_info.spsr = SPSR_64(MODE_EL1, MODE_SP_ELX,
+		.ep_info.spsr = SPSR_64(MODE_EL2, MODE_SP_ELX,
 					DISABLE_ALL_EXCEPTIONS),
 
 		.next_handoff_image_id = INVALID_IMAGE_ID,