Merge pull request #678 from soby-mathew/sm/PSCI_AArch32

Introduce AArch32 support for PSCI library

15 files changed, 1329 insertions, 35 deletions

diff --git a/lib/aarch32/cache_helpers.S b/lib/aarch32/cache_helpers.S
new file mode 100644
index 00000000..d0e5cd06
--- /dev/null
+++ b/lib/aarch32/cache_helpers.S
@@ -0,0 +1,237 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <arch.h>
+#include <asm_macros.S>
+
+	.globl	flush_dcache_range
+	.globl	clean_dcache_range
+	.globl	inv_dcache_range
+	.globl	dcsw_op_louis
+	.globl	dcsw_op_all
+	.globl	dcsw_op_level1
+	.globl	dcsw_op_level2
+	.globl	dcsw_op_level3
+
+/*
+ * This macro can be used for implementing various data cache operations `op`
+ */
+.macro do_dcache_maintenance_by_mva op, coproc, opc1, CRn, CRm, opc2
+	dcache_line_size r2, r3
+	add	r1, r0, r1
+	sub	r3, r2, #1
+	bic	r0, r0, r3
+loop_\op:
+	stcopr	r0, \coproc, \opc1, \CRn, \CRm, \opc2
+	add	r0, r0, r2
+	cmp	r0, r1
+	blo	loop_\op
+	dsb	sy
+	bx	lr
+.endm
+
+	/* ------------------------------------------
+	 * Clean+Invalidate from base address till
+	 * size. 'r0' = addr, 'r1' = size
+	 * ------------------------------------------
+	 */
+func flush_dcache_range
+	do_dcache_maintenance_by_mva cimvac, DCCIMVAC
+endfunc flush_dcache_range
+
+	/* ------------------------------------------
+	 * Clean from base address till size.
+	 * 'r0' = addr, 'r1' = size
+	 * ------------------------------------------
+	 */
+func clean_dcache_range
+	do_dcache_maintenance_by_mva cmvac, DCCMVAC
+endfunc clean_dcache_range
+
+	/* ------------------------------------------
+	 * Invalidate from base address till
+	 * size. 'r0' = addr, 'r1' = size
+	 * ------------------------------------------
+	 */
+func inv_dcache_range
+	do_dcache_maintenance_by_mva imvac, DCIMVAC
+endfunc inv_dcache_range
+
+	/* ----------------------------------------------------------------
+	 * Data cache operations by set/way to the level specified
+	 *
+	 * The main function, do_dcsw_op requires:
+	 * r0: The operation type (DC_OP_ISW, DC_OP_CISW, DC_OP_CSW),
+	 * as defined in arch.h
+	 * r1: The cache level to begin operation from
+	 * r2: clidr_el1
+	 * r3: The last cache level to operate on
+	 * and will carry out the operation on each data cache from level 0
+	 * to the level in r3 in sequence
+	 *
+	 * The dcsw_op macro sets up the r2 and r3 parameters based on
+	 * clidr_el1 cache information before invoking the main function
+	 * ----------------------------------------------------------------
+	 */
+
+	.macro	dcsw_op shift, fw, ls
+	ldcopr	r2, CLIDR
+	ubfx	r3, r2, \shift, \fw
+	lsl	r3, r3, \ls
+	mov	r1, #0
+	b	do_dcsw_op
+	.endm
+
+func do_dcsw_op
+	push	{r4-r12,lr}
+	adr	r11, dcsw_loop_table	// compute cache op based on the operation type
+	add	r6, r11, r0, lsl #3	// cache op is 2x32-bit instructions
+loop1:
+	add	r10, r1, r1, LSR #1	// Work out 3x current cache level
+	mov	r12, r2, LSR r10	// extract cache type bits from clidr
+	and	r12, r12, #7   		// mask the bits for current cache only
+	cmp	r12, #2			// see what cache we have at this level
+	blt	level_done      	// no cache or only instruction cache at this level
+
+	stcopr	r1, CSSELR		// select current cache level in csselr
+	isb				// isb to sych the new cssr&csidr
+	ldcopr	r12, CCSIDR		// read the new ccsidr
+	and	r10, r12, #7   		// extract the length of the cache lines
+	add	r10, r10, #4        	// add 4 (r10 = line length offset)
+	ubfx	r4, r12, #3, #10	// r4 = maximum way number (right aligned)
+	clz	r5, r4            	// r5 = the bit position of the way size increment
+	mov	r9, r4			// r9 working copy of the aligned max way number
+
+loop2:
+	ubfx	r7, r12, #13, #15	// r7 = max set number (right aligned)
+
+loop3:
+	orr	r0, r1, r9, LSL r5	// factor in the way number and cache level into r0
+	orr	r0, r0, r7, LSL r10	// factor in the set number
+
+	blx	r6
+	subs	r7, r7, #1              // decrement the set number
+	bge	loop3
+	subs	r9, r9, #1              // decrement the way number
+	bge	loop2
+level_done:
+	add	r1, r1, #2		// increment the cache number
+	cmp	r3, r1
+	dsb	sy			// ensure completion of previous cache maintenance instruction
+	bgt	loop1
+
+	mov	r6, #0
+	stcopr	r6, CSSELR		//select cache level 0 in csselr
+	dsb	sy
+	isb
+	pop	{r4-r12,pc}
+
+dcsw_loop_table:
+	stcopr	r0, DCISW
+	bx	lr
+	stcopr	r0, DCCISW
+	bx	lr
+	stcopr	r0, DCCSW
+	bx	lr
+
+endfunc do_dcsw_op
+
+	/* ---------------------------------------------------------------
+	 * Data cache operations by set/way till PoU.
+	 *
+	 * The function requires :
+	 * r0: The operation type (DC_OP_ISW, DC_OP_CISW, DC_OP_CSW),
+	 * as defined in arch.h
+	 * ---------------------------------------------------------------
+	 */
+func dcsw_op_louis
+	dcsw_op #LOUIS_SHIFT, #CLIDR_FIELD_WIDTH, #LEVEL_SHIFT
+endfunc	dcsw_op_louis
+
+	/* ---------------------------------------------------------------
+	 * Data cache operations by set/way till PoC.
+	 *
+	 * The function requires :
+	 * r0: The operation type (DC_OP_ISW, DC_OP_CISW, DC_OP_CSW),
+	 * as defined in arch.h
+	 * ---------------------------------------------------------------
+	 */
+func dcsw_op_all
+	dcsw_op #LOC_SHIFT, #CLIDR_FIELD_WIDTH, #LEVEL_SHIFT
+endfunc	dcsw_op_all
+
+
+	/* ---------------------------------------------------------------
+	 *  Helper macro for data cache operations by set/way for the
+	 *  level specified
+	 * ---------------------------------------------------------------
+	 */
+	.macro	dcsw_op_level level
+	ldcopr	r2, CLIDR
+	mov	r3, \level
+	sub	r1, r3, #2
+	b	do_dcsw_op
+	.endm
+
+	/* ---------------------------------------------------------------
+	 * Data cache operations by set/way for level 1 cache
+	 *
+	 * The main function, do_dcsw_op requires:
+	 * r0: The operation type (DC_OP_ISW, DC_OP_CISW, DC_OP_CSW),
+	 * as defined in arch.h
+	 * ---------------------------------------------------------------
+	 */
+func dcsw_op_level1
+	dcsw_op_level #(1 << LEVEL_SHIFT)
+endfunc dcsw_op_level1
+
+	/* ---------------------------------------------------------------
+	 * Data cache operations by set/way for level 2 cache
+	 *
+	 * The main function, do_dcsw_op requires:
+	 * r0: The operation type (DC_OP_ISW, DC_OP_CISW, DC_OP_CSW),
+	 * as defined in arch.h
+	 * ---------------------------------------------------------------
+	 */
+func dcsw_op_level2
+	dcsw_op_level #(2 << LEVEL_SHIFT)
+endfunc dcsw_op_level2
+
+	/* ---------------------------------------------------------------
+	 * Data cache operations by set/way for level 3 cache
+	 *
+	 * The main function, do_dcsw_op requires:
+	 * r0: The operation type (DC_OP_ISW, DC_OP_CISW, DC_OP_CSW),
+	 * as defined in arch.h
+	 * ---------------------------------------------------------------
+	 */
+func dcsw_op_level3
+	dcsw_op_level #(3 << LEVEL_SHIFT)
+endfunc dcsw_op_level3
diff --git a/lib/aarch32/misc_helpers.S b/lib/aarch32/misc_helpers.S
new file mode 100644
index 00000000..63ac1a7e
--- /dev/null
+++ b/lib/aarch32/misc_helpers.S
@@ -0,0 +1,60 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <arch.h>
+#include <asm_macros.S>
+#include <assert_macros.S>
+
+	.globl	zeromem
+
+/* -----------------------------------------------------------------------
+ * void zeromem(void *mem, unsigned int length);
+ *
+ * Initialise a memory region to 0.
+ * The memory address and length must be 4-byte aligned.
+ * -----------------------------------------------------------------------
+ */
+func zeromem
+#if ASM_ASSERTION
+	tst	r0, #0x3
+	ASM_ASSERT(eq)
+	tst	r1, #0x3
+	ASM_ASSERT(eq)
+#endif
+	add	r2, r0, r1
+	mov	r1, #0
+z_loop:
+	cmp	r2, r0
+	beq	z_end
+	str	r1, [r0], #4
+	b	z_loop
+z_end:
+	bx	lr
+endfunc zeromem
diff --git a/lib/cpus/aarch32/aem_generic.S b/lib/cpus/aarch32/aem_generic.S
new file mode 100644
index 00000000..10ea4e47
--- /dev/null
+++ b/lib/cpus/aarch32/aem_generic.S
@@ -0,0 +1,68 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <aem_generic.h>
+#include <arch.h>
+#include <asm_macros.S>
+#include <assert_macros.S>
+#include <cpu_macros.S>
+
+func aem_generic_core_pwr_dwn
+	/* Assert if cache is enabled */
+#if ASM_ASSERTION
+	ldcopr	r0, SCTLR
+	tst	r0, #SCTLR_C_BIT
+	ASM_ASSERT(eq)
+#endif
+	/* ---------------------------------------------
+	 * Flush L1 cache to PoU.
+	 * ---------------------------------------------
+	 */
+	mov	r0, #DC_OP_CISW
+	b	dcsw_op_louis
+endfunc aem_generic_core_pwr_dwn
+
+
+func aem_generic_cluster_pwr_dwn
+	/* Assert if cache is enabled */
+#if ASM_ASSERTION
+	ldcopr	r0, SCTLR
+	tst	r0, #SCTLR_C_BIT
+	ASM_ASSERT(eq)
+#endif
+	/* ---------------------------------------------
+	 * Flush L1 and L2 caches to PoC.
+	 * ---------------------------------------------
+	 */
+	mov	r0, #DC_OP_CISW
+	b	dcsw_op_all
+endfunc aem_generic_cluster_pwr_dwn
+
+/* cpu_ops for Base AEM FVP */
+declare_cpu_ops aem_generic, BASE_AEM_MIDR, 1
diff --git a/lib/cpus/aarch32/cpu_helpers.S b/lib/cpus/aarch32/cpu_helpers.S
new file mode 100644
index 00000000..927a6f50
--- /dev/null
+++ b/lib/cpus/aarch32/cpu_helpers.S
@@ -0,0 +1,177 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <arch.h>
+#include <asm_macros.S>
+#include <assert_macros.S>
+#include <cpu_data.h>
+#include <cpu_macros.S>
+
+	/*
+	 * The reset handler common to all platforms.  After a matching
+	 * cpu_ops structure entry is found, the correponding reset_handler
+	 * in the cpu_ops is invoked. The reset handler is invoked very early
+	 * in the boot sequence and it is assumed that we can clobber r0 - r10
+	 * without the need to follow AAPCS.
+	 * Clobbers: r0 - r10
+	 */
+	.globl	reset_handler
+func reset_handler
+	mov	r10, lr
+
+	/* The plat_reset_handler can clobber r0 - r9 */
+	bl	plat_reset_handler
+
+	/* Get the matching cpu_ops pointer (clobbers: r0 - r5) */
+	bl	get_cpu_ops_ptr
+
+#if ASM_ASSERTION
+	cmp	r0, #0
+	ASM_ASSERT(ne)
+#endif
+
+	/* Get the cpu_ops reset handler */
+	ldr	r1, [r0, #CPU_RESET_FUNC]
+	cmp	r1, #0
+	mov	lr, r10
+	bxne	r1
+	bx	lr
+endfunc reset_handler
+
+	/*
+	 * The prepare core power down function for all platforms.  After
+	 * the cpu_ops pointer is retrieved from cpu_data, the corresponding
+	 * pwr_dwn_core in the cpu_ops is invoked. Follows AAPCS.
+	 */
+	.globl	prepare_core_pwr_dwn
+func prepare_core_pwr_dwn
+	push	{lr}
+	bl	_cpu_data
+	pop	{lr}
+
+	ldr	r1, [r0, #CPU_DATA_CPU_OPS_PTR]
+#if ASM_ASSERTION
+	cmp	r1, #0
+	ASM_ASSERT(ne)
+#endif
+
+	/* Get the cpu_ops core_pwr_dwn handler */
+	ldr	r0, [r1, #CPU_PWR_DWN_CORE]
+	bx	r0
+endfunc prepare_core_pwr_dwn
+
+	/*
+	 * The prepare cluster power down function for all platforms.  After
+	 * the cpu_ops pointer is retrieved from cpu_data, the corresponding
+	 * pwr_dwn_cluster in the cpu_ops is invoked. Follows AAPCS.
+	 */
+	.globl	prepare_cluster_pwr_dwn
+func prepare_cluster_pwr_dwn
+	push	{lr}
+	bl	_cpu_data
+	pop	{lr}
+
+	ldr	r1, [r0, #CPU_DATA_CPU_OPS_PTR]
+#if ASM_ASSERTION
+	cmp	r1, #0
+	ASM_ASSERT(ne)
+#endif
+
+	/* Get the cpu_ops cluster_pwr_dwn handler */
+	ldr	r0, [r1, #CPU_PWR_DWN_CLUSTER]
+	bx	r0
+endfunc prepare_cluster_pwr_dwn
+
+	/*
+	 * Initializes the cpu_ops_ptr if not already initialized
+	 * in cpu_data. This must only be called after the data cache
+	 * is enabled. AAPCS is followed.
+	 */
+	.globl	init_cpu_ops
+func init_cpu_ops
+	push	{r4 - r6, lr}
+	bl	_cpu_data
+	mov	r6, r0
+	ldr	r1, [r0, #CPU_DATA_CPU_OPS_PTR]
+	cmp	r1, #0
+	bne	1f
+	bl	get_cpu_ops_ptr
+#if ASM_ASSERTION
+	cmp	r0, #0
+	ASM_ASSERT(ne)
+#endif
+	str	r0, [r6, #CPU_DATA_CPU_OPS_PTR]!
+1:
+	pop	{r4 - r6, pc}
+endfunc init_cpu_ops
+
+	/*
+	 * The below function returns the cpu_ops structure matching the
+	 * midr of the core. It reads the MIDR and finds the matching
+	 * entry in cpu_ops entries. Only the implementation and part number
+	 * are used to match the entries.
+	 * Return :
+	 *     r0 - The matching cpu_ops pointer on Success
+	 *     r0 - 0 on failure.
+	 * Clobbers: r0 - r5
+	 */
+	.globl	get_cpu_ops_ptr
+func get_cpu_ops_ptr
+	/* Get the cpu_ops start and end locations */
+	ldr	r4, =(__CPU_OPS_START__ + CPU_MIDR)
+	ldr	r5, =(__CPU_OPS_END__ + CPU_MIDR)
+
+	/* Initialize the return parameter */
+	mov	r0, #0
+
+	/* Read the MIDR_EL1 */
+	ldcopr	r2, MIDR
+	ldr	r3, =CPU_IMPL_PN_MASK
+
+	/* Retain only the implementation and part number using mask */
+	and	r2, r2, r3
+1:
+	/* Check if we have reached end of list */
+	cmp	r4, r5
+	bge	error_exit
+
+	/* load the midr from the cpu_ops */
+	ldr	r1, [r4], #CPU_OPS_SIZE
+	and	r1, r1, r3
+
+	/* Check if midr matches to midr of this core */
+	cmp	r1, r2
+	bne	1b
+
+	/* Subtract the increment and offset to get the cpu-ops pointer */
+	sub	r0, r4, #(CPU_OPS_SIZE + CPU_MIDR)
+error_exit:
+	bx	lr
+endfunc get_cpu_ops_ptr
diff --git a/lib/el3_runtime/aarch32/context_mgmt.c b/lib/el3_runtime/aarch32/context_mgmt.c
new file mode 100644
index 00000000..6915ded7
--- /dev/null
+++ b/lib/el3_runtime/aarch32/context_mgmt.c
@@ -0,0 +1,235 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <arch.h>
+#include <arch_helpers.h>
+#include <assert.h>
+#include <bl_common.h>
+#include <context.h>
+#include <context_mgmt.h>
+#include <platform.h>
+#include <platform_def.h>
+#include <smcc_helpers.h>
+#include <string.h>
+
+/*******************************************************************************
+ * Context management library initialisation routine. This library is used by
+ * runtime services to share pointers to 'cpu_context' structures for the secure
+ * and non-secure states. Management of the structures and their associated
+ * memory is not done by the context management library e.g. the PSCI service
+ * manages the cpu context used for entry from and exit to the non-secure state.
+ * The Secure payload manages the context(s) corresponding to the secure state.
+ * It also uses this library to get access to the non-secure
+ * state cpu context pointers.
+ ******************************************************************************/
+void cm_init(void)
+{
+	/*
+	 * The context management library has only global data to initialize, but
+	 * that will be done when the BSS is zeroed out
+	 */
+}
+
+/*******************************************************************************
+ * The following function initializes the cpu_context 'ctx' for
+ * first use, and sets the initial entrypoint state as specified by the
+ * entry_point_info structure.
+ *
+ * The security state to initialize is determined by the SECURE attribute
+ * of the entry_point_info. The function returns a pointer to the initialized
+ * context and sets this as the next context to return to.
+ *
+ * The EE and ST attributes are used to configure the endianness and secure
+ * timer availability for the new execution context.
+ *
+ * To prepare the register state for entry call cm_prepare_el3_exit() and
+ * el3_exit(). For Secure-EL1 cm_prepare_el3_exit() is equivalent to
+ * cm_e1_sysreg_context_restore().
+ ******************************************************************************/
+static void cm_init_context_common(cpu_context_t *ctx, const entry_point_info_t *ep)
+{
+	unsigned int security_state;
+	uint32_t scr, sctlr;
+	regs_t *reg_ctx;
+
+	assert(ctx);
+
+	security_state = GET_SECURITY_STATE(ep->h.attr);
+
+	/* Clear any residual register values from the context */
+	memset(ctx, 0, sizeof(*ctx));
+
+	/*
+	 * Base the context SCR on the current value, adjust for entry point
+	 * specific requirements
+	 */
+	scr = read_scr();
+	scr &= ~(SCR_NS_BIT | SCR_HCE_BIT);
+
+	if (security_state != SECURE)
+		scr |= SCR_NS_BIT;
+
+	/*
+	 * Set up SCTLR for the Non Secure context.
+	 * EE bit is taken from the entrypoint attributes
+	 * M, C and I bits must be zero (as required by PSCI specification)
+	 *
+	 * The target exception level is based on the spsr mode requested.
+	 * If execution is requested to hyp mode, HVC is enabled
+	 * via SCR.HCE.
+	 *
+	 * Always compute the SCTLR_EL1 value and save in the cpu_context
+	 * - the HYP registers are set up by cm_preapre_ns_entry() as they
+	 * are not part of the stored cpu_context
+	 *
+	 * TODO: In debug builds the spsr should be validated and checked
+	 * against the CPU support, security state, endianness and pc
+	 */
+	if (security_state != SECURE) {
+		sctlr = EP_GET_EE(ep->h.attr) ? SCTLR_EE_BIT : 0;
+		sctlr |= SCTLR_RES1;
+		write_ctx_reg(reg_ctx, CTX_NS_SCTLR, sctlr);
+	}
+
+	if (GET_M32(ep->spsr) == MODE32_hyp)
+		scr |= SCR_HCE_BIT;
+
+	reg_ctx = get_regs_ctx(ctx);
+
+	write_ctx_reg(reg_ctx, CTX_SCR, scr);
+	write_ctx_reg(reg_ctx, CTX_LR, ep->pc);
+	write_ctx_reg(reg_ctx, CTX_SPSR, ep->spsr);
+
+	/*
+	 * Store the r0-r3 value from the entrypoint into the context
+	 * Use memcpy as we are in control of the layout of the structures
+	 */
+	memcpy((void *)reg_ctx, (void *)&ep->args, sizeof(aapcs32_params_t));
+}
+
+/*******************************************************************************
+ * The following function initializes the cpu_context for a CPU specified by
+ * its `cpu_idx` for first use, and sets the initial entrypoint state as
+ * specified by the entry_point_info structure.
+ ******************************************************************************/
+void cm_init_context_by_index(unsigned int cpu_idx,
+			      const entry_point_info_t *ep)
+{
+	cpu_context_t *ctx;
+	ctx = cm_get_context_by_index(cpu_idx, GET_SECURITY_STATE(ep->h.attr));
+	cm_init_context_common(ctx, ep);
+}
+
+/*******************************************************************************
+ * The following function initializes the cpu_context for the current CPU
+ * for first use, and sets the initial entrypoint state as specified by the
+ * entry_point_info structure.
+ ******************************************************************************/
+void cm_init_my_context(const entry_point_info_t *ep)
+{
+	cpu_context_t *ctx;
+	ctx = cm_get_context(GET_SECURITY_STATE(ep->h.attr));
+	cm_init_context_common(ctx, ep);
+}
+
+/*******************************************************************************
+ * Prepare the CPU system registers for first entry into secure or normal world
+ *
+ * If execution is requested to hyp mode, HSCTLR is initialized
+ * If execution is requested to non-secure PL1, and the CPU supports
+ * HYP mode then HYP mode is disabled by configuring all necessary HYP mode
+ * registers.
+ ******************************************************************************/
+void cm_prepare_el3_exit(uint32_t security_state)
+{
+	uint32_t sctlr, scr, hcptr;
+	cpu_context_t *ctx = cm_get_context(security_state);
+
+	assert(ctx);
+
+	if (security_state == NON_SECURE) {
+		scr = read_ctx_reg(get_regs_ctx(ctx), CTX_SCR);
+		if (scr & SCR_HCE_BIT) {
+			/* Use SCTLR value to initialize HSCTLR */
+			sctlr = read_ctx_reg(get_regs_ctx(ctx),
+						 CTX_NS_SCTLR);
+			sctlr |= HSCTLR_RES1;
+			/* Temporarily set the NS bit to access HSCTLR */
+			write_scr(read_scr() | SCR_NS_BIT);
+			/*
+			 * Make sure the write to SCR is complete so that
+			 * we can access HSCTLR
+			 */
+			isb();
+			write_hsctlr(sctlr);
+			isb();
+
+			write_scr(read_scr() & ~SCR_NS_BIT);
+			isb();
+		} else if (read_id_pfr1() &
+			(ID_PFR1_VIRTEXT_MASK << ID_PFR1_VIRTEXT_SHIFT)) {
+			/* Set the NS bit to access HCR, HCPTR, CNTHCTL, VPIDR, VMPIDR */
+			write_scr(read_scr() | SCR_NS_BIT);
+			isb();
+
+			/* PL2 present but unused, need to disable safely */
+			write_hcr(0);
+
+			/* HSCTLR : can be ignored when bypassing */
+
+			/* HCPTR : disable all traps TCPAC, TTA, TCP */
+			hcptr = read_hcptr();
+			hcptr &= ~(TCPAC_BIT | TTA_BIT | TCP11_BIT | TCP10_BIT);
+			write_hcptr(hcptr);
+
+			/* Enable EL1 access to timer */
+			write_cnthctl(PL1PCEN_BIT | PL1PCTEN_BIT);
+
+			/* Reset CNTVOFF_EL2 */
+			write64_cntvoff(0);
+
+			/* Set VPIDR, VMPIDR to match MIDR, MPIDR */
+			write_vpidr(read_midr());
+			write_vmpidr(read_mpidr());
+
+			/*
+			 * Reset VTTBR.
+			 * Needed because cache maintenance operations depend on
+			 * the VMID even when non-secure EL1&0 stage 2 address
+			 * translation are disabled.
+			 */
+			write64_vttbr(0);
+			isb();
+
+			write_scr(read_scr() & ~SCR_NS_BIT);
+			isb();
+		}
+	}
+}
diff --git a/lib/el3_runtime/aarch32/cpu_data.S b/lib/el3_runtime/aarch32/cpu_data.S
new file mode 100644
index 00000000..b97911fb
--- /dev/null
+++ b/lib/el3_runtime/aarch32/cpu_data.S
@@ -0,0 +1,63 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <asm_macros.S>
+#include <cpu_data.h>
+
+	.globl	_cpu_data
+	.globl	_cpu_data_by_index
+
+/* -----------------------------------------------------------------
+ * cpu_data_t *_cpu_data(void)
+ *
+ * Return the cpu_data structure for the current CPU.
+ * -----------------------------------------------------------------
+ */
+func _cpu_data
+	push	{lr}
+	bl	plat_my_core_pos
+	pop	{lr}
+	b	_cpu_data_by_index
+endfunc _cpu_data
+
+/* -----------------------------------------------------------------
+ * cpu_data_t *_cpu_data_by_index(uint32_t cpu_index)
+ *
+ * Return the cpu_data structure for the CPU with given linear index
+ *
+ * This can be called without a valid stack.
+ * clobbers: r0, r1
+ * -----------------------------------------------------------------
+ */
+func _cpu_data_by_index
+	ldr	r1, =percpu_data
+	add	r0, r1, r0, LSL #CPU_DATA_LOG2SIZE
+	bx	lr
+endfunc _cpu_data_by_index
diff --git a/lib/locks/exclusive/aarch32/spinlock.S b/lib/locks/exclusive/aarch32/spinlock.S
new file mode 100644
index 00000000..f3a2bc36
--- /dev/null
+++ b/lib/locks/exclusive/aarch32/spinlock.S
@@ -0,0 +1,55 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <asm_macros.S>
+
+	.globl	spin_lock
+	.globl	spin_unlock
+
+
+func spin_lock
+	mov	r2, #1
+1:
+	ldrex	r1, [r0]
+	cmp	r1, #0
+	wfene
+	strexeq	r1, r2, [r0]
+	cmpeq	r1, #0
+	bne	1b
+	dmb
+	bx	lr
+endfunc spin_lock
+
+
+func spin_unlock
+	mov	r1, #0
+	stl	r1, [r0]
+	bx	lr
+endfunc spin_unlock
diff --git a/lib/locks/exclusive/aarch64/spinlock.S b/lib/locks/exclusive/aarch64/spinlock.S
new file mode 100644
index 00000000..1ca59123
--- /dev/null
+++ b/lib/locks/exclusive/aarch64/spinlock.S
@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2013-2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <asm_macros.S>
+
+	.globl	spin_lock
+	.globl	spin_unlock
+
+
+func spin_lock
+	mov	w2, #1
+	sevl
+l1:	wfe
+l2:	ldaxr	w1, [x0]
+	cbnz	w1, l1
+	stxr	w1, w2, [x0]
+	cbnz	w1, l2
+	ret
+endfunc spin_lock
+
+
+func spin_unlock
+	stlr	wzr, [x0]
+	ret
+endfunc spin_unlock
diff --git a/lib/locks/exclusive/spinlock.S b/lib/locks/exclusive/spinlock.S
index 772f14e9..9c945f9b 100644
--- a/lib/locks/exclusive/spinlock.S
+++ b/lib/locks/exclusive/spinlock.S
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013-2014, ARM Limited and Contributors. All rights reserved.
+ * Copyright (c) 2013-2016, ARM Limited and Contributors. All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
@@ -28,25 +28,6 @@
  * POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <asm_macros.S>
-
-	.globl	spin_lock
-	.globl	spin_unlock
-
-
-func spin_lock
-	mov	w2, #1
-	sevl
-l1:	wfe
-l2:	ldaxr	w1, [x0]
-	cbnz	w1, l1
-	stxr	w1, w2, [x0]
-	cbnz	w1, l2
-	ret
-endfunc spin_lock
-
-
-func spin_unlock
-	stlr	wzr, [x0]
-	ret
-endfunc spin_unlock
+#if !ERROR_DEPRECATED
+#include "./aarch64/spinlock.S"
+#endif
diff --git a/lib/psci/aarch32/psci_helpers.S b/lib/psci/aarch32/psci_helpers.S
new file mode 100644
index 00000000..36d5d7d9
--- /dev/null
+++ b/lib/psci/aarch32/psci_helpers.S
@@ -0,0 +1,180 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <asm_macros.S>
+#include <platform_def.h>
+#include <psci.h>
+
+	.globl	psci_do_pwrdown_cache_maintenance
+	.globl	psci_do_pwrup_cache_maintenance
+	.globl	psci_power_down_wfi
+
+/* -----------------------------------------------------------------------
+ * void psci_do_pwrdown_cache_maintenance(unsigned int power level);
+ *
+ * This function performs cache maintenance for the specified power
+ * level. The levels of cache affected are determined by the power
+ * level which is passed as the argument i.e. level 0 results
+ * in a flush of the L1 cache. Both the L1 and L2 caches are flushed
+ * for a higher power level.
+ *
+ * Additionally, this function also ensures that stack memory is correctly
+ * flushed out to avoid coherency issues due to a change in its memory
+ * attributes after the data cache is disabled.
+ * -----------------------------------------------------------------------
+ */
+func psci_do_pwrdown_cache_maintenance
+	push	{r4, lr}
+
+	/* ----------------------------------------------
+	 * Turn OFF cache and do stack maintenance
+	 * prior to cpu operations . This sequence is
+	 * different from AArch64 because in AArch32 the
+	 * assembler routines for cpu operations utilize
+	 * the stack whereas in AArch64 it doesn't.
+	 * ----------------------------------------------
+	 */
+	mov	r4, r0
+	bl	do_stack_maintenance
+
+	/* ---------------------------------------------
+	 * Determine how many levels of cache will be
+	 * subject to cache maintenance. Power level
+	 * 0 implies that only the cpu is being powered
+	 * down. Only the L1 data cache needs to be
+	 * flushed to the PoU in this case. For a higher
+	 * power level we are assuming that a flush
+	 * of L1 data and L2 unified cache is enough.
+	 * This information should be provided by the
+	 * platform.
+	 * ---------------------------------------------
+	 */
+	cmp	r4, #PSCI_CPU_PWR_LVL
+	pop	{r4,lr}
+
+	beq	prepare_core_pwr_dwn
+	b	prepare_cluster_pwr_dwn
+endfunc psci_do_pwrdown_cache_maintenance
+
+
+/* -----------------------------------------------------------------------
+ * void psci_do_pwrup_cache_maintenance(void);
+ *
+ * This function performs cache maintenance after this cpu is powered up.
+ * Currently, this involves managing the used stack memory before turning
+ * on the data cache.
+ * -----------------------------------------------------------------------
+ */
+func psci_do_pwrup_cache_maintenance
+	push	{lr}
+
+	/* ---------------------------------------------
+	 * Ensure any inflight stack writes have made it
+	 * to main memory.
+	 * ---------------------------------------------
+	 */
+	dmb	st
+
+	/* ---------------------------------------------
+	 * Calculate and store the size of the used
+	 * stack memory in r1. Calculate and store the
+	 * stack base address in r0.
+	 * ---------------------------------------------
+	 */
+	bl	plat_get_my_stack
+	mov	r1, sp
+	sub	r1, r0, r1
+	mov	r0, sp
+	bl	inv_dcache_range
+
+	/* ---------------------------------------------
+	 * Enable the data cache.
+	 * ---------------------------------------------
+	 */
+	ldcopr	r0, SCTLR
+	orr	r0, r0, #SCTLR_C_BIT
+	stcopr	r0, SCTLR
+	isb
+
+	pop	{pc}
+endfunc psci_do_pwrup_cache_maintenance
+
+	/* ---------------------------------------------
+	 * void do_stack_maintenance(void)
+	 * Do stack maintenance by flushing the used
+	 * stack to the main memory and invalidating the
+	 * remainder.
+	 * ---------------------------------------------
+	 */
+func do_stack_maintenance
+	push	{r4, lr}
+	bl	plat_get_my_stack
+
+	/* Turn off the D-cache */
+	ldcopr	r1, SCTLR
+	bic	r1, #SCTLR_C_BIT
+	stcopr	r1, SCTLR
+	isb
+
+	/* ---------------------------------------------
+	 * Calculate and store the size of the used
+	 * stack memory in r1.
+	 * ---------------------------------------------
+	 */
+	mov	r4, r0
+	mov	r1, sp
+	sub	r1, r0, r1
+	mov	r0, sp
+	bl	flush_dcache_range
+
+	/* ---------------------------------------------
+	 * Calculate and store the size of the unused
+	 * stack memory in r1. Calculate and store the
+	 * stack base address in r0.
+	 * ---------------------------------------------
+	 */
+	sub	r0, r4, #PLATFORM_STACK_SIZE
+	sub	r1, sp, r0
+	bl	inv_dcache_range
+
+	pop	{r4, pc}
+endfunc do_stack_maintenance
+
+/* -----------------------------------------------------------------------
+ * This function is called to indicate to the power controller that it
+ * is safe to power down this cpu. It should not exit the wfi and will
+ * be released from reset upon power up.
+ * -----------------------------------------------------------------------
+ */
+func psci_power_down_wfi
+	dsb	sy		// ensure write buffer empty
+	wfi
+	bl	plat_panic_handler
+endfunc psci_power_down_wfi
diff --git a/lib/psci/psci_common.c b/lib/psci/psci_common.c
index e87e8c05..68cdd6eb 100644
--- a/lib/psci/psci_common.c
+++ b/lib/psci/psci_common.c
@@ -592,6 +592,53 @@ int psci_validate_mpidr(u_register_t mpidr)
  * This function determines the full entrypoint information for the requested
  * PSCI entrypoint on power on/resume and returns it.
  ******************************************************************************/
+#ifdef AARCH32
+static int psci_get_ns_ep_info(entry_point_info_t *ep,
+			       uintptr_t entrypoint,
+			       u_register_t context_id)
+{
+	u_register_t ep_attr;
+	unsigned int aif, ee, mode;
+	u_register_t scr = read_scr();
+	u_register_t ns_sctlr, sctlr;
+
+	/* Switch to non secure state */
+	write_scr(scr | SCR_NS_BIT);
+	isb();
+	ns_sctlr = read_sctlr();
+
+	sctlr = scr & SCR_HCE_BIT ? read_hsctlr() : ns_sctlr;
+
+	/* Return to original state */
+	write_scr(scr);
+	isb();
+	ee = 0;
+
+	ep_attr = NON_SECURE | EP_ST_DISABLE;
+	if (sctlr & SCTLR_EE_BIT) {
+		ep_attr |= EP_EE_BIG;
+		ee = 1;
+	}
+	SET_PARAM_HEAD(ep, PARAM_EP, VERSION_1, ep_attr);
+
+	ep->pc = entrypoint;
+	memset(&ep->args, 0, sizeof(ep->args));
+	ep->args.arg0 = context_id;
+
+	mode = scr & SCR_HCE_BIT ? MODE32_hyp : MODE32_svc;
+
+	/*
+	 * TODO: Choose async. exception bits if HYP mode is not
+	 * implemented according to the values of SCR.{AW, FW} bits
+	 */
+	aif = SPSR_ABT_BIT | SPSR_IRQ_BIT | SPSR_FIQ_BIT;
+
+	ep->spsr = SPSR_MODE32(mode, entrypoint & 0x1, ee, aif);
+
+	return PSCI_E_SUCCESS;
+}
+
+#else
 static int psci_get_ns_ep_info(entry_point_info_t *ep,
 			       uintptr_t entrypoint,
 			       u_register_t context_id)
@@ -646,6 +693,7 @@ static int psci_get_ns_ep_info(entry_point_info_t *ep,
 
 	return PSCI_E_SUCCESS;
 }
+#endif
 
 /*******************************************************************************
  * This function validates the entrypoint with the platform layer if the
diff --git a/lib/psci/psci_lib.mk b/lib/psci/psci_lib.mk
index 662e14a2..8daa8318 100644
--- a/lib/psci/psci_lib.mk
+++ b/lib/psci/psci_lib.mk
@@ -29,11 +29,10 @@
 #
 
 PSCI_LIB_SOURCES	:=	lib/el3_runtime/cpu_data_array.c	\
-				lib/el3_runtime/aarch64/context.S	\
-				lib/el3_runtime/aarch64/cpu_data.S	\
-				lib/el3_runtime/aarch64/context_mgmt.c	\
-				lib/cpus/aarch64/cpu_helpers.S		\
-				lib/locks/exclusive/spinlock.S		\
+				lib/el3_runtime/${ARCH}/cpu_data.S	\
+				lib/el3_runtime/${ARCH}/context_mgmt.c	\
+				lib/cpus/${ARCH}/cpu_helpers.S		\
+				lib/locks/exclusive/${ARCH}/spinlock.S	\
 				lib/psci/psci_off.c			\
 				lib/psci/psci_on.c			\
 				lib/psci/psci_suspend.c			\
@@ -41,7 +40,11 @@ PSCI_LIB_SOURCES	:=	lib/el3_runtime/cpu_data_array.c	\
 				lib/psci/psci_main.c			\
 				lib/psci/psci_setup.c			\
 				lib/psci/psci_system_off.c		\
-				lib/psci/aarch64/psci_helpers.S
+				lib/psci/${ARCH}/psci_helpers.S
+
+ifeq (${ARCH}, aarch64)
+PSCI_LIB_SOURCES	+=	lib/el3_runtime/aarch64/context.S
+endif
 
 ifeq (${USE_COHERENT_MEM}, 1)
 PSCI_LIB_SOURCES		+=	lib/locks/bakery/bakery_lock_coherent.c
diff --git a/lib/psci/psci_setup.c b/lib/psci/psci_setup.c
index d35e0001..20d06352 100644
--- a/lib/psci/psci_setup.c
+++ b/lib/psci/psci_setup.c
@@ -278,3 +278,15 @@ void psci_arch_setup(void)
 	/* Initialize the cpu_ops pointer. */
 	init_cpu_ops();
 }
+
+/******************************************************************************
+ * PSCI Library interface to initialize the cpu context for the next non
+ * secure image during cold boot. The relevant registers in the cpu context
+ * need to be retrieved and programmed on return from this interface.
+ *****************************************************************************/
+void psci_prepare_next_non_secure_ctx(entry_point_info_t *next_image_info)
+{
+	assert(GET_SECURITY_STATE(next_image_info->h.attr) == NON_SECURE);
+	cm_init_my_context(next_image_info);
+	cm_prepare_el3_exit(NON_SECURE);
+}
diff --git a/lib/xlat_tables/aarch32/xlat_tables.c b/lib/xlat_tables/aarch32/xlat_tables.c
new file mode 100644
index 00000000..a97cf312
--- /dev/null
+++ b/lib/xlat_tables/aarch32/xlat_tables.c
@@ -0,0 +1,123 @@
+/*
+ * Copyright (c) 2016, ARM Limited and Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * Neither the name of ARM nor the names of its contributors may be used
+ * to endorse or promote products derived from this software without specific
+ * prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <arch.h>
+#include <arch_helpers.h>
+#include <assert.h>
+#include <cassert.h>
+#include <platform_def.h>
+#include <utils.h>
+#include <xlat_tables.h>
+#include "../xlat_tables_private.h"
+
+/*
+ * The virtual address space size must be a power of two. As we start the initial
+ * lookup at level 1, it must also be between 2 GB and 4 GB. See section
+ * G4.6.5 in the ARMv8-A Architecture Reference Manual (DDI 0487A.j) for more
+ * information.
+ */
+CASSERT(ADDR_SPACE_SIZE >= (1ull << 31) && ADDR_SPACE_SIZE <= (1ull << 32) &&
+	IS_POWER_OF_TWO(ADDR_SPACE_SIZE), assert_valid_addr_space_size);
+
+#define NUM_L1_ENTRIES (ADDR_SPACE_SIZE >> L1_XLAT_ADDRESS_SHIFT)
+
+static uint64_t l1_xlation_table[NUM_L1_ENTRIES]
+		__aligned(NUM_L1_ENTRIES * sizeof(uint64_t));
+
+void init_xlat_tables(void)
+{
+	unsigned long long max_pa;
+	uintptr_t max_va;
+	print_mmap();
+	init_xlation_table(0, l1_xlation_table, 1, &max_va, &max_pa);
+	assert(max_va < ADDR_SPACE_SIZE);
+}
+
+/*******************************************************************************
+ * Function for enabling the MMU in Secure PL1, assuming that the
+ * page-tables have already been created.
+ ******************************************************************************/
+void enable_mmu_secure(unsigned int flags)
+{
+	unsigned int mair0, ttbcr, sctlr;
+	uint64_t ttbr0;
+
+	assert(IS_IN_SECURE());
+	assert((read_sctlr() & SCTLR_M_BIT) == 0);
+
+	/* Set attributes in the right indices of the MAIR */
+	mair0 = MAIR0_ATTR_SET(ATTR_DEVICE, ATTR_DEVICE_INDEX);
+	mair0 |= MAIR0_ATTR_SET(ATTR_IWBWA_OWBWA_NTR,
+			ATTR_IWBWA_OWBWA_NTR_INDEX);
+	mair0 |= MAIR0_ATTR_SET(ATTR_NON_CACHEABLE,
+			ATTR_NON_CACHEABLE_INDEX);
+	write_mair0(mair0);
+
+	/* Invalidate TLBs at the current exception level */
+	tlbiall();
+
+	/*
+	 * Set TTBCR bits as well. Set TTBR0 table properties as Inner
+	 * & outer WBWA & shareable. Disable TTBR1.
+	 */
+	ttbcr = TTBCR_EAE_BIT |
+		TTBCR_SH0_INNER_SHAREABLE | TTBCR_RGN0_OUTER_WBA |
+		TTBCR_RGN0_INNER_WBA |
+		(32 - __builtin_ctzl((uintptr_t)ADDR_SPACE_SIZE));
+	ttbcr |= TTBCR_EPD1_BIT;
+	write_ttbcr(ttbcr);
+
+	/* Set TTBR0 bits as well */
+	ttbr0 = (uintptr_t) l1_xlation_table;
+	write64_ttbr0(ttbr0);
+	write64_ttbr1(0);
+
+	/*
+	 * Ensure all translation table writes have drained
+	 * into memory, the TLB invalidation is complete,
+	 * and translation register writes are committed
+	 * before enabling the MMU
+	 */
+	dsb();
+	isb();
+
+	sctlr = read_sctlr();
+	sctlr |= SCTLR_WXN_BIT | SCTLR_M_BIT;
+
+	if (flags & DISABLE_DCACHE)
+		sctlr &= ~SCTLR_C_BIT;
+	else
+		sctlr |= SCTLR_C_BIT;
+
+	write_sctlr(sctlr);
+
+	/* Ensure the MMU enable takes effect immediately */
+	isb();
+}
diff --git a/lib/xlat_tables/xlat_tables_common.c b/lib/xlat_tables/xlat_tables_common.c
index 33784c2d..bc7fed78 100644
--- a/lib/xlat_tables/xlat_tables_common.c
+++ b/lib/xlat_tables/xlat_tables_common.c
@@ -289,17 +289,17 @@ static int mmap_region_attr(mmap_region_t *mm, uintptr_t base_va,
 		if (!mm->size)
 			return attr; /* Reached end of list */
 
-		if (mm->base_va >= base_va + size)
+		if (mm->base_va > base_va + size - 1)
 			return attr; /* Next region is after area so end */
 
-		if (mm->base_va + mm->size <= base_va)
+		if (mm->base_va + mm->size - 1 < base_va)
 			continue; /* Next region has already been overtaken */
 
 		if (mm->attr == attr)
 			continue; /* Region doesn't override attribs so skip */
 
 		if (mm->base_va > base_va ||
-			mm->base_va + mm->size < base_va + size)
+			mm->base_va + mm->size - 1 < base_va + size - 1)
 			return -1; /* Region doesn't fully cover our area */
 
 		attr = mm->attr;
@@ -328,7 +328,7 @@ static mmap_region_t *init_xlation_table_inner(mmap_region_t *mm,
 		if (!mm->size) {
 			/* Done mapping regions; finish zeroing the table */
 			desc = INVALID_DESC;
-		} else if (mm->base_va + mm->size <= base_va) {
+		} else if (mm->base_va + mm->size - 1 < base_va) {
 			/* This area is after the region so get next region */
 			++mm;
 			continue;
@@ -337,7 +337,7 @@ static mmap_region_t *init_xlation_table_inner(mmap_region_t *mm,
 		debug_print("%s VA:%p size:0x%x ", get_level_spacer(level),
 				(void *)base_va, level_size);
 
-		if (mm->base_va >= base_va + level_size) {
+		if (mm->base_va > base_va + level_size - 1) {
 			/* Next region is after this area. Nothing to map yet */
 			desc = INVALID_DESC;
 		} else {
@@ -369,7 +369,7 @@ static mmap_region_t *init_xlation_table_inner(mmap_region_t *mm,
 
 		*table++ = desc;
 		base_va += level_size;
-	} while ((base_va & level_index_mask) && (base_va < ADDR_SPACE_SIZE));
+	} while ((base_va & level_index_mask) && (base_va - 1 < ADDR_SPACE_SIZE - 1));
 
 	return mm;
 }