cortex_a9: add source files for Cortex A9 support.
authorAaron Carroll <aaronc@cse.unsw.edu.au>
Mon, 24 Jan 2011 07:06:46 +0000 (18:06 +1100)
committerØyvind Harboe <oyvind.harboe@zylin.com>
Mon, 31 Jan 2011 07:57:38 +0000 (08:57 +0100)
add target and build support for A9

Signed-off-by: Aaron Carroll <aaronc@cse.unsw.edu.au>
src/target/Makefile.am
src/target/cortex_a9.c [new file with mode: 0644]
src/target/cortex_a9.h [new file with mode: 0644]
src/target/target.c

index 537c8c31754900779ac2bb2cd16b6bdad6c280f9..b6e8d4bea8589555966c03cab73f86b0ac8d0f04 100644 (file)
@@ -73,7 +73,8 @@ ARMV7_SRC = \
        armv7m.c \
        cortex_m3.c \
        armv7a.c \
-       cortex_a8.c
+       cortex_a8.c \
+       cortex_a9.c
 
 ARM_DEBUG_SRC = \
        arm_dpm.c \
@@ -136,6 +137,7 @@ noinst_HEADERS = \
        breakpoints.h \
        cortex_m3.h \
        cortex_a8.h \
+       cortex_a9.h \
        embeddedice.h \
        etb.h \
        etm.h \
diff --git a/src/target/cortex_a9.c b/src/target/cortex_a9.c
new file mode 100644 (file)
index 0000000..0ede1d5
--- /dev/null
@@ -0,0 +1,2181 @@
+/***************************************************************************
+ *   Copyright (C) 2005 by Dominic Rath                                    *
+ *   Dominic.Rath@gmx.de                                                   *
+ *                                                                         *
+ *   Copyright (C) 2006 by Magnus Lundin                                   *
+ *   lundin@mlu.mine.nu                                                    *
+ *                                                                         *
+ *   Copyright (C) 2008 by Spencer Oliver                                  *
+ *   spen@spen-soft.co.uk                                                  *
+ *                                                                         *
+ *   Copyright (C) 2009 by Dirk Behme                                      *
+ *   dirk.behme@gmail.com - copy from cortex_m3                            *
+ *                                                                         *
+ *   Copyright (C) 2010 Øyvind Harboe                                      *
+ *   oyvind.harboe@zylin.com                                               *
+ *                                                                         *
+ *   This program is free software; you can redistribute it and/or modify  *
+ *   it under the terms of the GNU General Public License as published by  *
+ *   the Free Software Foundation; either version 2 of the License, or     *
+ *   (at your option) any later version.                                   *
+ *                                                                         *
+ *   This program is distributed in the hope that it will be useful,       *
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of        *
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         *
+ *   GNU General Public License for more details.                          *
+ *                                                                         *
+ *   You should have received a copy of the GNU General Public License     *
+ *   along with this program; if not, write to the                         *
+ *   Free Software Foundation, Inc.,                                       *
+ *   59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.             *
+ *                                                                         *
+ *   Cortex-A9(tm) TRM, ARM DDI 0407F                                      *
+ *                                                                         *
+ ***************************************************************************/
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include "breakpoints.h"
+#include "cortex_a9.h"
+#include "register.h"
+#include "target_request.h"
+#include "target_type.h"
+#include "arm_opcodes.h"
+#include <helper/time_support.h>
+
+static int cortex_a9_poll(struct target *target);
+static int cortex_a9_debug_entry(struct target *target);
+static int cortex_a9_restore_context(struct target *target, bool bpwp);
+static int cortex_a9_set_breakpoint(struct target *target,
+               struct breakpoint *breakpoint, uint8_t matchmode);
+static int cortex_a9_unset_breakpoint(struct target *target,
+               struct breakpoint *breakpoint);
+static int cortex_a9_dap_read_coreregister_u32(struct target *target,
+               uint32_t *value, int regnum);
+static int cortex_a9_dap_write_coreregister_u32(struct target *target,
+               uint32_t value, int regnum);
+static int cortex_a9_mmu(struct target *target, int *enabled);
+static int cortex_a9_virt2phys(struct target *target,
+                uint32_t virt, uint32_t *phys);
+static int cortex_a9_disable_mmu_caches(struct target *target, int mmu,
+                int d_u_cache, int i_cache);
+static int cortex_a9_enable_mmu_caches(struct target *target, int mmu,
+                int d_u_cache, int i_cache);
+static int cortex_a9_get_ttb(struct target *target, uint32_t *result);
+
+
+/*
+ * FIXME do topology discovery using the ROM; don't
+ * assume this is an OMAP3.   Also, allow for multiple ARMv7-A
+ * cores, with different AP numbering ... don't use a #define
+ * for these numbers, use per-core armv7a state.
+ */
+#define swjdp_memoryap 0
+#define swjdp_debugap 1
+
+/*
+ * Cortex-A9 Basic debug access, very low level assumes state is saved
+ */
+static int cortex_a9_init_debug_access(struct target *target)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+
+       int retval;
+       uint32_t dummy;
+
+       LOG_DEBUG(" ");
+
+       /* Unlocking the debug registers for modification */
+       /* The debugport might be uninitialised so try twice */
+       retval = mem_ap_write_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_LOCKACCESS, 0xC5ACCE55);
+       if (retval != ERROR_OK)
+       {
+               /* try again */
+               retval = mem_ap_write_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_LOCKACCESS, 0xC5ACCE55);
+               if (retval == ERROR_OK)
+               {
+                       LOG_USER("Locking debug access failed on first, but succeeded on second try.");
+               }
+       }
+       if (retval != ERROR_OK)
+               return retval;
+       /* Clear Sticky Power Down status Bit in PRSR to enable access to
+          the registers in the Core Power Domain */
+       retval = mem_ap_read_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_PRSR, &dummy);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* Enabling of instruction execution in debug mode is done in debug_entry code */
+
+       /* Resync breakpoint registers */
+
+       /* Since this is likely called from init or reset, update target state information*/
+       retval = cortex_a9_poll(target);
+
+       return retval;
+}
+
+/* To reduce needless round-trips, pass in a pointer to the current
+ * DSCR value.  Initialize it to zero if you just need to know the
+ * value on return from this function; or DSCR_INSTR_COMP if you
+ * happen to know that no instruction is pending.
+ */
+static int cortex_a9_exec_opcode(struct target *target,
+               uint32_t opcode, uint32_t *dscr_p)
+{
+       uint32_t dscr;
+       int retval;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+
+       dscr = dscr_p ? *dscr_p : 0;
+
+       LOG_DEBUG("exec opcode 0x%08" PRIx32, opcode);
+
+       /* Wait for InstrCompl bit to be set */
+       long long then = timeval_ms();
+       while ((dscr & DSCR_INSTR_COMP) == 0)
+       {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                               armv7a->debug_base + CPUDBG_DSCR, &dscr);
+               if (retval != ERROR_OK)
+               {
+                       LOG_ERROR("Could not read DSCR register, opcode = 0x%08" PRIx32, opcode);
+                       return retval;
+               }
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for cortex_a9_exec_opcode");
+                       return ERROR_FAIL;
+               }
+       }
+
+       retval = mem_ap_write_u32(swjdp, armv7a->debug_base + CPUDBG_ITR, opcode);
+       if (retval != ERROR_OK)
+               return retval;
+
+       then = timeval_ms();
+       do
+       {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                               armv7a->debug_base + CPUDBG_DSCR, &dscr);
+               if (retval != ERROR_OK)
+               {
+                       LOG_ERROR("Could not read DSCR register");
+                       return retval;
+               }
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for cortex_a9_exec_opcode");
+                       return ERROR_FAIL;
+               }
+       }
+       while ((dscr & DSCR_INSTR_COMP) == 0); /* Wait for InstrCompl bit to be set */
+
+       if (dscr_p)
+               *dscr_p = dscr;
+
+       return retval;
+}
+
+/**************************************************************************
+Read core register with very few exec_opcode, fast but needs work_area.
+This can cause problems with MMU active.
+**************************************************************************/
+static int cortex_a9_read_regs_through_mem(struct target *target, uint32_t address,
+               uint32_t * regfile)
+{
+       int retval = ERROR_OK;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+
+       retval = cortex_a9_dap_read_coreregister_u32(target, regfile, 0);
+       if (retval != ERROR_OK)
+               return retval;
+       retval = cortex_a9_dap_write_coreregister_u32(target, address, 0);
+       if (retval != ERROR_OK)
+               return retval;
+       retval = cortex_a9_exec_opcode(target, ARMV4_5_STMIA(0, 0xFFFE, 0, 0), NULL);
+       if (retval != ERROR_OK)
+               return retval;
+
+       dap_ap_select(swjdp, swjdp_memoryap);
+       retval = mem_ap_read_buf_u32(swjdp, (uint8_t *)(&regfile[1]), 4*15, address);
+       if (retval != ERROR_OK)
+               return retval;
+       dap_ap_select(swjdp, swjdp_debugap);
+
+       return retval;
+}
+
+static int cortex_a9_dap_read_coreregister_u32(struct target *target,
+               uint32_t *value, int regnum)
+{
+       int retval = ERROR_OK;
+       uint8_t reg = regnum&0xFF;
+       uint32_t dscr = 0;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+
+       if (reg > 17)
+               return retval;
+
+       if (reg < 15)
+       {
+               /* Rn to DCCTX, "MCR p14, 0, Rn, c0, c5, 0"  0xEE00nE15 */
+               retval = cortex_a9_exec_opcode(target,
+                               ARMV4_5_MCR(14, 0, reg, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+       else if (reg == 15)
+       {
+               /* "MOV r0, r15"; then move r0 to DCCTX */
+               retval = cortex_a9_exec_opcode(target, 0xE1A0000F, &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = cortex_a9_exec_opcode(target,
+                               ARMV4_5_MCR(14, 0, 0, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+       else
+       {
+               /* "MRS r0, CPSR" or "MRS r0, SPSR"
+                * then move r0 to DCCTX
+                */
+               retval = cortex_a9_exec_opcode(target, ARMV4_5_MRS(0, reg & 1), &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = cortex_a9_exec_opcode(target,
+                               ARMV4_5_MCR(14, 0, 0, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+
+       /* Wait for DTRRXfull then read DTRRTX */
+       long long then = timeval_ms();
+       while ((dscr & DSCR_DTR_TX_FULL) == 0)
+       {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                               armv7a->debug_base + CPUDBG_DSCR, &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for cortex_a9_exec_opcode");
+                       return ERROR_FAIL;
+               }
+       }
+
+       retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DTRTX, value);
+       LOG_DEBUG("read DCC 0x%08" PRIx32, *value);
+
+       return retval;
+}
+
+static int cortex_a9_dap_write_coreregister_u32(struct target *target,
+               uint32_t value, int regnum)
+{
+       int retval = ERROR_OK;
+       uint8_t Rd = regnum&0xFF;
+       uint32_t dscr;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+
+       LOG_DEBUG("register %i, value 0x%08" PRIx32, regnum, value);
+
+       /* Check that DCCRX is not full */
+       retval = mem_ap_read_atomic_u32(swjdp,
+                               armv7a->debug_base + CPUDBG_DSCR, &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+       if (dscr & DSCR_DTR_RX_FULL)
+       {
+               LOG_ERROR("DSCR_DTR_RX_FULL, dscr 0x%08" PRIx32, dscr);
+               /* Clear DCCRX with MCR(p14, 0, Rd, c0, c5, 0), opcode  0xEE000E15 */
+               retval = cortex_a9_exec_opcode(target, ARMV4_5_MRC(14, 0, 0, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+
+       if (Rd > 17)
+               return retval;
+
+       /* Write DTRRX ... sets DSCR.DTRRXfull but exec_opcode() won't care */
+       LOG_DEBUG("write DCC 0x%08" PRIx32, value);
+       retval = mem_ap_write_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DTRRX, value);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if (Rd < 15)
+       {
+               /* DCCRX to Rn, "MCR p14, 0, Rn, c0, c5, 0", 0xEE00nE15 */
+               retval = cortex_a9_exec_opcode(target, ARMV4_5_MRC(14, 0, Rd, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+       else if (Rd == 15)
+       {
+               /* DCCRX to R0, "MCR p14, 0, R0, c0, c5, 0", 0xEE000E15
+                * then "mov r15, r0"
+                */
+               retval = cortex_a9_exec_opcode(target, ARMV4_5_MRC(14, 0, 0, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = cortex_a9_exec_opcode(target, 0xE1A0F000, &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+       else
+       {
+               /* DCCRX to R0, "MCR p14, 0, R0, c0, c5, 0", 0xEE000E15
+                * then "MSR CPSR_cxsf, r0" or "MSR SPSR_cxsf, r0" (all fields)
+                */
+               retval = cortex_a9_exec_opcode(target, ARMV4_5_MRC(14, 0, 0, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = cortex_a9_exec_opcode(target, ARMV4_5_MSR_GP(0, 0xF, Rd & 1),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+
+               /* "Prefetch flush" after modifying execution status in CPSR */
+               if (Rd == 16)
+               {
+                       retval = cortex_a9_exec_opcode(target,
+                                       ARMV4_5_MCR(15, 0, 0, 7, 5, 4),
+                                       &dscr);
+                       if (retval != ERROR_OK)
+                               return retval;
+               }
+       }
+
+       return retval;
+}
+
+/* Write to memory mapped registers directly with no cache or mmu handling */
+static int cortex_a9_dap_write_memap_register_u32(struct target *target, uint32_t address, uint32_t value)
+{
+       int retval;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+
+       retval = mem_ap_write_atomic_u32(swjdp, address, value);
+
+       return retval;
+}
+
+/*
+ * Cortex-A9 implementation of Debug Programmer's Model
+ *
+ * NOTE the invariant:  these routines return with DSCR_INSTR_COMP set,
+ * so there's no need to poll for it before executing an instruction.
+ *
+ * NOTE that in several of these cases the "stall" mode might be useful.
+ * It'd let us queue a few operations together... prepare/finish might
+ * be the places to enable/disable that mode.
+ */
+
+static inline struct cortex_a9_common *dpm_to_a9(struct arm_dpm *dpm)
+{
+       return container_of(dpm, struct cortex_a9_common, armv7a_common.dpm);
+}
+
+static int cortex_a9_write_dcc(struct cortex_a9_common *a9, uint32_t data)
+{
+       LOG_DEBUG("write DCC 0x%08" PRIx32, data);
+       return mem_ap_write_u32(&a9->armv7a_common.dap,
+                       a9->armv7a_common.debug_base + CPUDBG_DTRRX, data);
+}
+
+static int cortex_a9_read_dcc(struct cortex_a9_common *a9, uint32_t *data,
+               uint32_t *dscr_p)
+{
+       struct adiv5_dap *swjdp = &a9->armv7a_common.dap;
+       uint32_t dscr = DSCR_INSTR_COMP;
+       int retval;
+
+       if (dscr_p)
+               dscr = *dscr_p;
+
+       /* Wait for DTRRXfull */
+       long long then = timeval_ms();
+       while ((dscr & DSCR_DTR_TX_FULL) == 0) {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                               a9->armv7a_common.debug_base + CPUDBG_DSCR,
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for read dcc");
+                       return ERROR_FAIL;
+               }
+       }
+
+       retval = mem_ap_read_atomic_u32(swjdp,
+                       a9->armv7a_common.debug_base + CPUDBG_DTRTX, data);
+       if (retval != ERROR_OK)
+               return retval;
+       //LOG_DEBUG("read DCC 0x%08" PRIx32, *data);
+
+       if (dscr_p)
+               *dscr_p = dscr;
+
+       return retval;
+}
+
+static int cortex_a9_dpm_prepare(struct arm_dpm *dpm)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       struct adiv5_dap *swjdp = &a9->armv7a_common.dap;
+       uint32_t dscr;
+       int retval;
+
+       /* set up invariant:  INSTR_COMP is set after ever DPM operation */
+       long long then = timeval_ms();
+       for (;;)
+       {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                               a9->armv7a_common.debug_base + CPUDBG_DSCR,
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               if ((dscr & DSCR_INSTR_COMP) != 0)
+                       break;
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for dpm prepare");
+                       return ERROR_FAIL;
+               }
+       }
+
+       /* this "should never happen" ... */
+       if (dscr & DSCR_DTR_RX_FULL) {
+               LOG_ERROR("DSCR_DTR_RX_FULL, dscr 0x%08" PRIx32, dscr);
+               /* Clear DCCRX */
+               retval = cortex_a9_exec_opcode(
+                               a9->armv7a_common.armv4_5_common.target,
+                               ARMV4_5_MRC(14, 0, 0, 0, 5, 0),
+                               &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+
+       return retval;
+}
+
+static int cortex_a9_dpm_finish(struct arm_dpm *dpm)
+{
+       /* REVISIT what could be done here? */
+       return ERROR_OK;
+}
+
+static int cortex_a9_instr_write_data_dcc(struct arm_dpm *dpm,
+               uint32_t opcode, uint32_t data)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       int retval;
+       uint32_t dscr = DSCR_INSTR_COMP;
+
+       retval = cortex_a9_write_dcc(a9, data);
+       if (retval != ERROR_OK)
+               return retval;
+
+       return cortex_a9_exec_opcode(
+                       a9->armv7a_common.armv4_5_common.target,
+                       opcode,
+                       &dscr);
+}
+
+static int cortex_a9_instr_write_data_r0(struct arm_dpm *dpm,
+               uint32_t opcode, uint32_t data)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       uint32_t dscr = DSCR_INSTR_COMP;
+       int retval;
+
+       retval = cortex_a9_write_dcc(a9, data);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* DCCRX to R0, "MCR p14, 0, R0, c0, c5, 0", 0xEE000E15 */
+       retval = cortex_a9_exec_opcode(
+                       a9->armv7a_common.armv4_5_common.target,
+                       ARMV4_5_MRC(14, 0, 0, 0, 5, 0),
+                       &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* then the opcode, taking data from R0 */
+       retval = cortex_a9_exec_opcode(
+                       a9->armv7a_common.armv4_5_common.target,
+                       opcode,
+                       &dscr);
+
+       return retval;
+}
+
+static int cortex_a9_instr_cpsr_sync(struct arm_dpm *dpm)
+{
+       struct target *target = dpm->arm->target;
+       uint32_t dscr = DSCR_INSTR_COMP;
+
+       /* "Prefetch flush" after modifying execution status in CPSR */
+       return cortex_a9_exec_opcode(target,
+                       ARMV4_5_MCR(15, 0, 0, 7, 5, 4),
+                       &dscr);
+}
+
+static int cortex_a9_instr_read_data_dcc(struct arm_dpm *dpm,
+               uint32_t opcode, uint32_t *data)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       int retval;
+       uint32_t dscr = DSCR_INSTR_COMP;
+
+       /* the opcode, writing data to DCC */
+       retval = cortex_a9_exec_opcode(
+                       a9->armv7a_common.armv4_5_common.target,
+                       opcode,
+                       &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       return cortex_a9_read_dcc(a9, data, &dscr);
+}
+
+
+static int cortex_a9_instr_read_data_r0(struct arm_dpm *dpm,
+               uint32_t opcode, uint32_t *data)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       uint32_t dscr = DSCR_INSTR_COMP;
+       int retval;
+
+       /* the opcode, writing data to R0 */
+       retval = cortex_a9_exec_opcode(
+                       a9->armv7a_common.armv4_5_common.target,
+                       opcode,
+                       &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* write R0 to DCC */
+       retval = cortex_a9_exec_opcode(
+                       a9->armv7a_common.armv4_5_common.target,
+                       ARMV4_5_MCR(14, 0, 0, 0, 5, 0),
+                       &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       return cortex_a9_read_dcc(a9, data, &dscr);
+}
+
+static int cortex_a9_bpwp_enable(struct arm_dpm *dpm, unsigned index_t,
+               uint32_t addr, uint32_t control)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       uint32_t vr = a9->armv7a_common.debug_base;
+       uint32_t cr = a9->armv7a_common.debug_base;
+       int retval;
+
+       switch (index_t) {
+       case 0 ... 15:          /* breakpoints */
+               vr += CPUDBG_BVR_BASE;
+               cr += CPUDBG_BCR_BASE;
+               break;
+       case 16 ... 31:         /* watchpoints */
+               vr += CPUDBG_WVR_BASE;
+               cr += CPUDBG_WCR_BASE;
+               index_t -= 16;
+               break;
+       default:
+               return ERROR_FAIL;
+       }
+       vr += 4 * index_t;
+       cr += 4 * index_t;
+
+       LOG_DEBUG("A9: bpwp enable, vr %08x cr %08x",
+                       (unsigned) vr, (unsigned) cr);
+
+       retval = cortex_a9_dap_write_memap_register_u32(dpm->arm->target,
+                       vr, addr);
+       if (retval != ERROR_OK)
+               return retval;
+       retval = cortex_a9_dap_write_memap_register_u32(dpm->arm->target,
+                       cr, control);
+       return retval;
+}
+
+static int cortex_a9_bpwp_disable(struct arm_dpm *dpm, unsigned index_t)
+{
+       struct cortex_a9_common *a9 = dpm_to_a9(dpm);
+       uint32_t cr;
+
+       switch (index_t) {
+       case 0 ... 15:
+               cr = a9->armv7a_common.debug_base + CPUDBG_BCR_BASE;
+               break;
+       case 16 ... 31:
+               cr = a9->armv7a_common.debug_base + CPUDBG_WCR_BASE;
+               index_t -= 16;
+               break;
+       default:
+               return ERROR_FAIL;
+       }
+       cr += 4 * index_t;
+
+       LOG_DEBUG("A9: bpwp disable, cr %08x", (unsigned) cr);
+
+       /* clear control register */
+       return cortex_a9_dap_write_memap_register_u32(dpm->arm->target, cr, 0);
+}
+
+static int cortex_a9_dpm_setup(struct cortex_a9_common *a9, uint32_t didr)
+{
+       struct arm_dpm *dpm = &a9->armv7a_common.dpm;
+       int retval;
+
+       dpm->arm = &a9->armv7a_common.armv4_5_common;
+       dpm->didr = didr;
+
+       dpm->prepare = cortex_a9_dpm_prepare;
+       dpm->finish = cortex_a9_dpm_finish;
+
+       dpm->instr_write_data_dcc = cortex_a9_instr_write_data_dcc;
+       dpm->instr_write_data_r0 = cortex_a9_instr_write_data_r0;
+       dpm->instr_cpsr_sync = cortex_a9_instr_cpsr_sync;
+
+       dpm->instr_read_data_dcc = cortex_a9_instr_read_data_dcc;
+       dpm->instr_read_data_r0 = cortex_a9_instr_read_data_r0;
+
+       dpm->bpwp_enable = cortex_a9_bpwp_enable;
+       dpm->bpwp_disable = cortex_a9_bpwp_disable;
+
+       retval = arm_dpm_setup(dpm);
+       if (retval == ERROR_OK)
+               retval = arm_dpm_initialize(dpm);
+
+       return retval;
+}
+
+
+/*
+ * Cortex-A9 Run control
+ */
+
+static int cortex_a9_poll(struct target *target)
+{
+       int retval = ERROR_OK;
+       uint32_t dscr;
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       enum target_state prev_target_state = target->state;
+       uint8_t saved_apsel = dap_ap_get_select(swjdp);
+
+       dap_ap_select(swjdp, swjdp_debugap);
+       retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DSCR, &dscr);
+       if (retval != ERROR_OK)
+       {
+               dap_ap_select(swjdp, saved_apsel);
+               return retval;
+       }
+       cortex_a9->cpudbg_dscr = dscr;
+
+       if (DSCR_RUN_MODE(dscr) == (DSCR_CORE_HALTED | DSCR_CORE_RESTARTED))
+       {
+               if (prev_target_state != TARGET_HALTED)
+               {
+                       /* We have a halting debug event */
+                       LOG_DEBUG("Target halted");
+                       target->state = TARGET_HALTED;
+                       if ((prev_target_state == TARGET_RUNNING)
+                                       || (prev_target_state == TARGET_RESET))
+                       {
+                               retval = cortex_a9_debug_entry(target);
+                               if (retval != ERROR_OK)
+                                       return retval;
+
+                               target_call_event_callbacks(target,
+                                               TARGET_EVENT_HALTED);
+                       }
+                       if (prev_target_state == TARGET_DEBUG_RUNNING)
+                       {
+                               LOG_DEBUG(" ");
+
+                               retval = cortex_a9_debug_entry(target);
+                               if (retval != ERROR_OK)
+                                       return retval;
+
+                               target_call_event_callbacks(target,
+                                               TARGET_EVENT_DEBUG_HALTED);
+                       }
+               }
+       }
+       else if (DSCR_RUN_MODE(dscr) == DSCR_CORE_RESTARTED)
+       {
+               target->state = TARGET_RUNNING;
+       }
+       else
+       {
+               LOG_DEBUG("Unknown target state dscr = 0x%08" PRIx32, dscr);
+               target->state = TARGET_UNKNOWN;
+       }
+
+       dap_ap_select(swjdp, saved_apsel);
+
+       return retval;
+}
+
+static int cortex_a9_halt(struct target *target)
+{
+       int retval = ERROR_OK;
+       uint32_t dscr;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       uint8_t saved_apsel = dap_ap_get_select(swjdp);
+       dap_ap_select(swjdp, swjdp_debugap);
+
+       /*
+        * Tell the core to be halted by writing DRCR with 0x1
+        * and then wait for the core to be halted.
+        */
+       retval = mem_ap_write_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DRCR, DRCR_HALT);
+       if (retval != ERROR_OK)
+               goto out;
+
+       /*
+        * enter halting debug mode
+        */
+       retval = mem_ap_read_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_DSCR, &dscr);
+       if (retval != ERROR_OK)
+               goto out;
+
+       retval = mem_ap_write_atomic_u32(swjdp,
+               armv7a->debug_base + CPUDBG_DSCR, dscr | DSCR_HALT_DBG_MODE);
+       if (retval != ERROR_OK)
+               goto out;
+
+       long long then = timeval_ms();
+       for (;;)
+       {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DSCR, &dscr);
+               if (retval != ERROR_OK)
+                       goto out;
+               if ((dscr & DSCR_CORE_HALTED) != 0)
+               {
+                       break;
+               }
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for halt");
+                       return ERROR_FAIL;
+               }
+       }
+
+       target->debug_reason = DBG_REASON_DBGRQ;
+
+out:
+       dap_ap_select(swjdp, saved_apsel);
+       return retval;
+}
+
+static int cortex_a9_resume(struct target *target, int current,
+               uint32_t address, int handle_breakpoints, int debug_execution)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct arm *armv4_5 = &armv7a->armv4_5_common;
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       int retval;
+
+//     struct breakpoint *breakpoint = NULL;
+       uint32_t resume_pc, dscr;
+
+       uint8_t saved_apsel = dap_ap_get_select(swjdp);
+       dap_ap_select(swjdp, swjdp_debugap);
+
+       if (!debug_execution)
+               target_free_all_working_areas(target);
+
+#if 0
+       if (debug_execution)
+       {
+               /* Disable interrupts */
+               /* We disable interrupts in the PRIMASK register instead of
+                * masking with C_MASKINTS,
+                * This is probably the same issue as Cortex-M3 Errata 377493:
+                * C_MASKINTS in parallel with disabled interrupts can cause
+                * local faults to not be taken. */
+               buf_set_u32(armv7m->core_cache->reg_list[ARMV7M_PRIMASK].value, 0, 32, 1);
+               armv7m->core_cache->reg_list[ARMV7M_PRIMASK].dirty = 1;
+               armv7m->core_cache->reg_list[ARMV7M_PRIMASK].valid = 1;
+
+               /* Make sure we are in Thumb mode */
+               buf_set_u32(armv7m->core_cache->reg_list[ARMV7M_xPSR].value, 0, 32,
+                       buf_get_u32(armv7m->core_cache->reg_list[ARMV7M_xPSR].value, 0, 32) | (1 << 24));
+               armv7m->core_cache->reg_list[ARMV7M_xPSR].dirty = 1;
+               armv7m->core_cache->reg_list[ARMV7M_xPSR].valid = 1;
+       }
+#endif
+
+       /* current = 1: continue on current pc, otherwise continue at <address> */
+       resume_pc = buf_get_u32(armv4_5->pc->value, 0, 32);
+       if (!current)
+               resume_pc = address;
+
+       /* Make sure that the Armv7 gdb thumb fixups does not
+        * kill the return address
+        */
+       switch (armv4_5->core_state)
+       {
+       case ARM_STATE_ARM:
+               resume_pc &= 0xFFFFFFFC;
+               break;
+       case ARM_STATE_THUMB:
+       case ARM_STATE_THUMB_EE:
+               /* When the return address is loaded into PC
+                * bit 0 must be 1 to stay in Thumb state
+                */
+               resume_pc |= 0x1;
+               break;
+       case ARM_STATE_JAZELLE:
+               LOG_ERROR("How do I resume into Jazelle state??");
+               return ERROR_FAIL;
+       }
+       LOG_DEBUG("resume pc = 0x%08" PRIx32, resume_pc);
+       buf_set_u32(armv4_5->pc->value, 0, 32, resume_pc);
+       armv4_5->pc->dirty = 1;
+       armv4_5->pc->valid = 1;
+
+       retval = cortex_a9_restore_context(target, handle_breakpoints);
+       if (retval != ERROR_OK)
+               return retval;
+
+#if 0
+       /* the front-end may request us not to handle breakpoints */
+       if (handle_breakpoints)
+       {
+               /* Single step past breakpoint at current address */
+               if ((breakpoint = breakpoint_find(target, resume_pc)))
+               {
+                       LOG_DEBUG("unset breakpoint at 0x%8.8x", breakpoint->address);
+                       cortex_m3_unset_breakpoint(target, breakpoint);
+                       cortex_m3_single_step_core(target);
+                       cortex_m3_set_breakpoint(target, breakpoint);
+               }
+       }
+
+#endif
+
+       /*
+        * Restart core and wait for it to be started.  Clear ITRen and sticky
+        * exception flags: see ARMv7 ARM, C5.9.
+        *
+        * REVISIT: for single stepping, we probably want to
+        * disable IRQs by default, with optional override...
+        */
+
+       retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DSCR, &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if ((dscr & DSCR_INSTR_COMP) == 0)
+               LOG_ERROR("DSCR InstrCompl must be set before leaving debug!");
+
+       retval = mem_ap_write_atomic_u32(swjdp,
+               armv7a->debug_base + CPUDBG_DSCR, dscr & ~DSCR_ITR_EN);
+       if (retval != ERROR_OK)
+               return retval;
+
+       retval = mem_ap_write_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_DRCR,
+                       DRCR_RESTART | DRCR_CLEAR_EXCEPTIONS);
+       if (retval != ERROR_OK)
+               return retval;
+
+       long long then = timeval_ms();
+       for (;;)
+       {
+               retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DSCR, &dscr);
+               if (retval != ERROR_OK)
+                       return retval;
+               if ((dscr & DSCR_CORE_RESTARTED) != 0)
+                       break;
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("Timeout waiting for resume");
+                       return ERROR_FAIL;
+               }
+       }
+
+       target->debug_reason = DBG_REASON_NOTHALTED;
+       target->state = TARGET_RUNNING;
+
+       /* registers are now invalid */
+       register_cache_invalidate(armv4_5->core_cache);
+
+       if (!debug_execution)
+       {
+               target->state = TARGET_RUNNING;
+               target_call_event_callbacks(target, TARGET_EVENT_RESUMED);
+               LOG_DEBUG("target resumed at 0x%" PRIx32, resume_pc);
+       }
+       else
+       {
+               target->state = TARGET_DEBUG_RUNNING;
+               target_call_event_callbacks(target, TARGET_EVENT_DEBUG_RESUMED);
+               LOG_DEBUG("target debug resumed at 0x%" PRIx32, resume_pc);
+       }
+
+       dap_ap_select(swjdp, saved_apsel);
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_debug_entry(struct target *target)
+{
+       int i;
+       uint32_t regfile[16], cpsr, dscr;
+       int retval = ERROR_OK;
+       struct working_area *regfile_working_area = NULL;
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct arm *armv4_5 = &armv7a->armv4_5_common;
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       struct reg *reg;
+
+       LOG_DEBUG("dscr = 0x%08" PRIx32, cortex_a9->cpudbg_dscr);
+
+       /* REVISIT surely we should not re-read DSCR !! */
+       retval = mem_ap_read_atomic_u32(swjdp,
+                               armv7a->debug_base + CPUDBG_DSCR, &dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* REVISIT see A9 TRM 12.11.4 steps 2..3 -- make sure that any
+        * imprecise data aborts get discarded by issuing a Data
+        * Synchronization Barrier:  ARMV4_5_MCR(15, 0, 0, 7, 10, 4).
+        */
+
+       /* Enable the ITR execution once we are in debug mode */
+       dscr |= DSCR_ITR_EN;
+       retval = mem_ap_write_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DSCR, dscr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* Examine debug reason */
+       arm_dpm_report_dscr(&armv7a->dpm, cortex_a9->cpudbg_dscr);
+
+       /* save address of instruction that triggered the watchpoint? */
+       if (target->debug_reason == DBG_REASON_WATCHPOINT) {
+               uint32_t wfar;
+
+               retval = mem_ap_read_atomic_u32(swjdp,
+                               armv7a->debug_base + CPUDBG_WFAR,
+                               &wfar);
+               if (retval != ERROR_OK)
+                       return retval;
+               arm_dpm_report_wfar(&armv7a->dpm, wfar);
+       }
+
+       /* REVISIT fast_reg_read is never set ... */
+
+       /* Examine target state and mode */
+       if (cortex_a9->fast_reg_read)
+               target_alloc_working_area(target, 64, &regfile_working_area);
+
+       /* First load register acessible through core debug port*/
+       if (!regfile_working_area)
+       {
+               retval = arm_dpm_read_current_registers(&armv7a->dpm);
+       }
+       else
+       {
+               dap_ap_select(swjdp, swjdp_memoryap);
+               retval = cortex_a9_read_regs_through_mem(target,
+                               regfile_working_area->address, regfile);
+               dap_ap_select(swjdp, swjdp_memoryap);
+               target_free_working_area(target, regfile_working_area);
+               if (retval != ERROR_OK)
+               {
+                       return retval;
+               }
+
+               /* read Current PSR */
+               retval = cortex_a9_dap_read_coreregister_u32(target, &cpsr, 16);
+               if (retval != ERROR_OK)
+                       return retval;
+               dap_ap_select(swjdp, swjdp_debugap);
+               LOG_DEBUG("cpsr: %8.8" PRIx32, cpsr);
+
+               arm_set_cpsr(armv4_5, cpsr);
+
+               /* update cache */
+               for (i = 0; i <= ARM_PC; i++)
+               {
+                       reg = arm_reg_current(armv4_5, i);
+
+                       buf_set_u32(reg->value, 0, 32, regfile[i]);
+                       reg->valid = 1;
+                       reg->dirty = 0;
+               }
+
+               /* Fixup PC Resume Address */
+               if (cpsr & (1 << 5))
+               {
+                       // T bit set for Thumb or ThumbEE state
+                       regfile[ARM_PC] -= 4;
+               }
+               else
+               {
+                       // ARM state
+                       regfile[ARM_PC] -= 8;
+               }
+
+               reg = armv4_5->pc;
+               buf_set_u32(reg->value, 0, 32, regfile[ARM_PC]);
+               reg->dirty = reg->valid;
+       }
+
+#if 0
+/* TODO, Move this */
+       uint32_t cp15_control_register, cp15_cacr, cp15_nacr;
+       cortex_a9_read_cp(target, &cp15_control_register, 15, 0, 1, 0, 0);
+       LOG_DEBUG("cp15_control_register = 0x%08x", cp15_control_register);
+
+       cortex_a9_read_cp(target, &cp15_cacr, 15, 0, 1, 0, 2);
+       LOG_DEBUG("cp15 Coprocessor Access Control Register = 0x%08x", cp15_cacr);
+
+       cortex_a9_read_cp(target, &cp15_nacr, 15, 0, 1, 1, 2);
+       LOG_DEBUG("cp15 Nonsecure Access Control Register = 0x%08x", cp15_nacr);
+#endif
+
+       /* Are we in an exception handler */
+//     armv4_5->exception_number = 0;
+       if (armv7a->post_debug_entry)
+       {
+               retval = armv7a->post_debug_entry(target);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+
+       return retval;
+}
+
+static int cortex_a9_post_debug_entry(struct target *target)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       int retval;
+
+       /* MRC p15,0,<Rt>,c1,c0,0 ; Read CP15 System Control Register */
+       retval = armv7a->armv4_5_common.mrc(target, 15,
+                       0, 0,   /* op1, op2 */
+                       1, 0,   /* CRn, CRm */
+                       &cortex_a9->cp15_control_reg);
+       if (retval != ERROR_OK)
+               return retval;
+       LOG_DEBUG("cp15_control_reg: %8.8" PRIx32, cortex_a9->cp15_control_reg);
+
+       if (armv7a->armv4_5_mmu.armv4_5_cache.ctype == -1)
+       {
+               uint32_t cache_type_reg;
+
+               /* MRC p15,0,<Rt>,c0,c0,1 ; Read CP15 Cache Type Register */
+               retval = armv7a->armv4_5_common.mrc(target, 15,
+                               0, 1,   /* op1, op2 */
+                               0, 0,   /* CRn, CRm */
+                               &cache_type_reg);
+               if (retval != ERROR_OK)
+                       return retval;
+               LOG_DEBUG("cp15 cache type: %8.8x", (unsigned) cache_type_reg);
+
+               /* FIXME the armv4_4 cache info DOES NOT APPLY to Cortex-A9 */
+               armv4_5_identify_cache(cache_type_reg,
+                               &armv7a->armv4_5_mmu.armv4_5_cache);
+       }
+
+       armv7a->armv4_5_mmu.mmu_enabled =
+                       (cortex_a9->cp15_control_reg & 0x1U) ? 1 : 0;
+       armv7a->armv4_5_mmu.armv4_5_cache.d_u_cache_enabled =
+                       (cortex_a9->cp15_control_reg & 0x4U) ? 1 : 0;
+       armv7a->armv4_5_mmu.armv4_5_cache.i_cache_enabled =
+                       (cortex_a9->cp15_control_reg & 0x1000U) ? 1 : 0;
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_step(struct target *target, int current, uint32_t address,
+               int handle_breakpoints)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct arm *armv4_5 = &armv7a->armv4_5_common;
+       struct breakpoint *breakpoint = NULL;
+       struct breakpoint stepbreakpoint;
+       struct reg *r;
+       int retval;
+
+       if (target->state != TARGET_HALTED)
+       {
+               LOG_WARNING("target not halted");
+               return ERROR_TARGET_NOT_HALTED;
+       }
+
+       /* current = 1: continue on current pc, otherwise continue at <address> */
+       r = armv4_5->pc;
+       if (!current)
+       {
+               buf_set_u32(r->value, 0, 32, address);
+       }
+       else
+       {
+               address = buf_get_u32(r->value, 0, 32);
+       }
+
+       /* The front-end may request us not to handle breakpoints.
+        * But since Cortex-A9 uses breakpoint for single step,
+        * we MUST handle breakpoints.
+        */
+       handle_breakpoints = 1;
+       if (handle_breakpoints) {
+               breakpoint = breakpoint_find(target, address);
+               if (breakpoint)
+                       cortex_a9_unset_breakpoint(target, breakpoint);
+       }
+
+       /* Setup single step breakpoint */
+       stepbreakpoint.address = address;
+       stepbreakpoint.length = (armv4_5->core_state == ARM_STATE_THUMB)
+                       ? 2 : 4;
+       stepbreakpoint.type = BKPT_HARD;
+       stepbreakpoint.set = 0;
+
+       /* Break on IVA mismatch */
+       cortex_a9_set_breakpoint(target, &stepbreakpoint, 0x04);
+
+       target->debug_reason = DBG_REASON_SINGLESTEP;
+
+       retval = cortex_a9_resume(target, 1, address, 0, 0);
+       if (retval != ERROR_OK)
+               return retval;
+
+       long long then = timeval_ms();
+       while (target->state != TARGET_HALTED)
+       {
+               retval = cortex_a9_poll(target);
+               if (retval != ERROR_OK)
+                       return retval;
+               if (timeval_ms() > then + 1000)
+               {
+                       LOG_ERROR("timeout waiting for target halt");
+                       return ERROR_FAIL;
+               }
+       }
+
+       cortex_a9_unset_breakpoint(target, &stepbreakpoint);
+
+       target->debug_reason = DBG_REASON_BREAKPOINT;
+
+       if (breakpoint)
+               cortex_a9_set_breakpoint(target, breakpoint, 0);
+
+       if (target->state != TARGET_HALTED)
+               LOG_DEBUG("target stepped");
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_restore_context(struct target *target, bool bpwp)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+
+       LOG_DEBUG(" ");
+
+       if (armv7a->pre_restore_context)
+               armv7a->pre_restore_context(target);
+
+       return arm_dpm_write_dirty_registers(&armv7a->dpm, bpwp);
+}
+
+
+/*
+ * Cortex-A9 Breakpoint and watchpoint functions
+ */
+
+/* Setup hardware Breakpoint Register Pair */
+static int cortex_a9_set_breakpoint(struct target *target,
+               struct breakpoint *breakpoint, uint8_t matchmode)
+{
+       int retval;
+       int brp_i=0;
+       uint32_t control;
+       uint8_t byte_addr_select = 0x0F;
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       struct cortex_a9_brp * brp_list = cortex_a9->brp_list;
+
+       if (breakpoint->set)
+       {
+               LOG_WARNING("breakpoint already set");
+               return ERROR_OK;
+       }
+
+       if (breakpoint->type == BKPT_HARD)
+       {
+               while (brp_list[brp_i].used && (brp_i < cortex_a9->brp_num))
+                       brp_i++ ;
+               if (brp_i >= cortex_a9->brp_num)
+               {
+                       LOG_ERROR("ERROR Can not find free Breakpoint Register Pair");
+                       return ERROR_TARGET_RESOURCE_NOT_AVAILABLE;
+               }
+               breakpoint->set = brp_i + 1;
+               if (breakpoint->length == 2)
+               {
+                       byte_addr_select = (3 << (breakpoint->address & 0x02));
+               }
+               control = ((matchmode & 0x7) << 20)
+                               | (byte_addr_select << 5)
+                               | (3 << 1) | 1;
+               brp_list[brp_i].used = 1;
+               brp_list[brp_i].value = (breakpoint->address & 0xFFFFFFFC);
+               brp_list[brp_i].control = control;
+               retval = cortex_a9_dap_write_memap_register_u32(target, armv7a->debug_base
+                               + CPUDBG_BVR_BASE + 4 * brp_list[brp_i].BRPn,
+                               brp_list[brp_i].value);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = cortex_a9_dap_write_memap_register_u32(target, armv7a->debug_base
+                               + CPUDBG_BCR_BASE + 4 * brp_list[brp_i].BRPn,
+                               brp_list[brp_i].control);
+               if (retval != ERROR_OK)
+                       return retval;
+               LOG_DEBUG("brp %i control 0x%0" PRIx32 " value 0x%0" PRIx32, brp_i,
+                               brp_list[brp_i].control,
+                               brp_list[brp_i].value);
+       }
+       else if (breakpoint->type == BKPT_SOFT)
+       {
+               uint8_t code[4];
+               if (breakpoint->length == 2)
+               {
+                       buf_set_u32(code, 0, 32, ARMV5_T_BKPT(0x11));
+               }
+               else
+               {
+                       buf_set_u32(code, 0, 32, ARMV5_BKPT(0x11));
+               }
+               retval = target->type->read_memory(target,
+                               breakpoint->address & 0xFFFFFFFE,
+                               breakpoint->length, 1,
+                               breakpoint->orig_instr);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = target->type->write_memory(target,
+                               breakpoint->address & 0xFFFFFFFE,
+                               breakpoint->length, 1, code);
+               if (retval != ERROR_OK)
+                       return retval;
+               breakpoint->set = 0x11; /* Any nice value but 0 */
+       }
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_unset_breakpoint(struct target *target, struct breakpoint *breakpoint)
+{
+       int retval;
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       struct cortex_a9_brp * brp_list = cortex_a9->brp_list;
+
+       if (!breakpoint->set)
+       {
+               LOG_WARNING("breakpoint not set");
+               return ERROR_OK;
+       }
+
+       if (breakpoint->type == BKPT_HARD)
+       {
+               int brp_i = breakpoint->set - 1;
+               if ((brp_i < 0) || (brp_i >= cortex_a9->brp_num))
+               {
+                       LOG_DEBUG("Invalid BRP number in breakpoint");
+                       return ERROR_OK;
+               }
+               LOG_DEBUG("rbp %i control 0x%0" PRIx32 " value 0x%0" PRIx32, brp_i,
+                               brp_list[brp_i].control, brp_list[brp_i].value);
+               brp_list[brp_i].used = 0;
+               brp_list[brp_i].value = 0;
+               brp_list[brp_i].control = 0;
+               retval = cortex_a9_dap_write_memap_register_u32(target, armv7a->debug_base
+                               + CPUDBG_BCR_BASE + 4 * brp_list[brp_i].BRPn,
+                               brp_list[brp_i].control);
+               if (retval != ERROR_OK)
+                       return retval;
+               retval = cortex_a9_dap_write_memap_register_u32(target, armv7a->debug_base
+                               + CPUDBG_BVR_BASE + 4 * brp_list[brp_i].BRPn,
+                               brp_list[brp_i].value);
+               if (retval != ERROR_OK)
+                       return retval;
+       }
+       else
+       {
+               /* restore original instruction (kept in target endianness) */
+               if (breakpoint->length == 4)
+               {
+                       retval = target->type->write_memory(target,
+                                       breakpoint->address & 0xFFFFFFFE,
+                                       4, 1, breakpoint->orig_instr);
+                       if (retval != ERROR_OK)
+                               return retval;
+               }
+               else
+               {
+                       retval = target->type->write_memory(target,
+                                       breakpoint->address & 0xFFFFFFFE,
+                                       2, 1, breakpoint->orig_instr);
+                       if (retval != ERROR_OK)
+                               return retval;
+               }
+       }
+       breakpoint->set = 0;
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_add_breakpoint(struct target *target,
+               struct breakpoint *breakpoint)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+
+       if ((breakpoint->type == BKPT_HARD) && (cortex_a9->brp_num_available < 1))
+       {
+               LOG_INFO("no hardware breakpoint available");
+               return ERROR_TARGET_RESOURCE_NOT_AVAILABLE;
+       }
+
+       if (breakpoint->type == BKPT_HARD)
+               cortex_a9->brp_num_available--;
+
+       return cortex_a9_set_breakpoint(target, breakpoint, 0x00); /* Exact match */
+}
+
+static int cortex_a9_remove_breakpoint(struct target *target, struct breakpoint *breakpoint)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+
+#if 0
+/* It is perfectly possible to remove breakpoints while the target is running */
+       if (target->state != TARGET_HALTED)
+       {
+               LOG_WARNING("target not halted");
+               return ERROR_TARGET_NOT_HALTED;
+       }
+#endif
+
+       if (breakpoint->set)
+       {
+               cortex_a9_unset_breakpoint(target, breakpoint);
+               if (breakpoint->type == BKPT_HARD)
+                       cortex_a9->brp_num_available++ ;
+       }
+
+
+       return ERROR_OK;
+}
+
+
+
+/*
+ * Cortex-A9 Reset functions
+ */
+
+static int cortex_a9_assert_reset(struct target *target)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+
+       LOG_DEBUG(" ");
+
+       /* FIXME when halt is requested, make it work somehow... */
+
+       /* Issue some kind of warm reset. */
+       if (target_has_event_action(target, TARGET_EVENT_RESET_ASSERT)) {
+               target_handle_event(target, TARGET_EVENT_RESET_ASSERT);
+       } else if (jtag_get_reset_config() & RESET_HAS_SRST) {
+               /* REVISIT handle "pulls" cases, if there's
+                * hardware that needs them to work.
+                */
+               jtag_add_reset(0, 1);
+       } else {
+               LOG_ERROR("%s: how to reset?", target_name(target));
+               return ERROR_FAIL;
+       }
+
+       /* registers are now invalid */
+       register_cache_invalidate(armv7a->armv4_5_common.core_cache);
+
+       target->state = TARGET_RESET;
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_deassert_reset(struct target *target)
+{
+       int retval;
+
+       LOG_DEBUG(" ");
+
+       /* be certain SRST is off */
+       jtag_add_reset(0, 0);
+
+       retval = cortex_a9_poll(target);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if (target->reset_halt) {
+               if (target->state != TARGET_HALTED) {
+                       LOG_WARNING("%s: ran after reset and before halt ...",
+                                       target_name(target));
+                       if ((retval = target_halt(target)) != ERROR_OK)
+                               return retval;
+               }
+       }
+
+       return ERROR_OK;
+}
+
+/*
+ * Cortex-A9 Memory access
+ *
+ * This is same Cortex M3 but we must also use the correct
+ * ap number for every access.
+ */
+
+static int cortex_a9_read_phys_memory(struct target *target,
+               uint32_t address, uint32_t size,
+               uint32_t count, uint8_t *buffer)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       int retval = ERROR_INVALID_ARGUMENTS;
+       uint8_t saved_apsel = dap_ap_get_select(swjdp);
+
+       /* cortex_a9 handles unaligned memory access */
+
+       dap_ap_select(swjdp, swjdp_memoryap);
+
+       LOG_DEBUG("Reading memory at real address 0x%x; size %d; count %d", address, size, count);
+       if (count && buffer) {
+               switch (size) {
+                       case 4:
+                               retval = mem_ap_read_buf_u32(swjdp, buffer, 4 * count, address);
+                               break;
+                       case 2:
+                               retval = mem_ap_read_buf_u16(swjdp, buffer, 2 * count, address);
+                               break;
+                       case 1:
+                               retval = mem_ap_read_buf_u8(swjdp, buffer, count, address);
+                               break;
+               }
+       }
+
+       dap_ap_select(swjdp, saved_apsel);
+
+       return retval;
+}
+
+static int cortex_a9_read_memory(struct target *target, uint32_t address,
+               uint32_t size, uint32_t count, uint8_t *buffer)
+{
+       int enabled = 0;
+       uint32_t virt, phys;
+       int retval;
+
+       /* cortex_a9 handles unaligned memory access */
+
+       LOG_DEBUG("Reading memory at address 0x%x; size %d; count %d", address, size, count);
+       retval = cortex_a9_mmu(target, &enabled);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if (enabled)
+       {
+               virt = address;
+               retval = cortex_a9_virt2phys(target, virt, &phys);
+               if (retval != ERROR_OK)
+                       return retval;
+
+               LOG_DEBUG("Reading at virtual address. Translating v:0x%x to r:0x%x", virt, phys);
+               address = phys;
+       }
+
+       return cortex_a9_read_phys_memory(target, address, size, count, buffer);
+}
+
+static int cortex_a9_write_phys_memory(struct target *target,
+               uint32_t address, uint32_t size,
+               uint32_t count, uint8_t *buffer)
+{
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       int retval = ERROR_INVALID_ARGUMENTS;
+
+       LOG_DEBUG("Writing memory to real address 0x%x; size %d; count %d", address, size, count);
+
+       if (count && buffer) {
+               uint8_t saved_apsel = dap_ap_get_select(swjdp);
+               dap_ap_select(swjdp, swjdp_memoryap);
+
+               switch (size) {
+                       case 4:
+                               retval = mem_ap_write_buf_u32(swjdp, buffer, 4 * count, address);
+                               break;
+                       case 2:
+                               retval = mem_ap_write_buf_u16(swjdp, buffer, 2 * count, address);
+                               break;
+                       case 1:
+                               retval = mem_ap_write_buf_u8(swjdp, buffer, count, address);
+                               break;
+               }
+
+               dap_ap_select(swjdp, saved_apsel);
+       }
+
+
+       /* REVISIT this op is generic ARMv7-A/R stuff */
+       if (retval == ERROR_OK && target->state == TARGET_HALTED)
+       {
+               struct arm_dpm *dpm = armv7a->armv4_5_common.dpm;
+
+               retval = dpm->prepare(dpm);
+               if (retval != ERROR_OK)
+                       return retval;
+
+               /* The Cache handling will NOT work with MMU active, the
+                * wrong addresses will be invalidated!
+                *
+                * For both ICache and DCache, walk all cache lines in the
+                * address range. Cortex-A9 has fixed 64 byte line length.
+                *
+                * REVISIT per ARMv7, these may trigger watchpoints ...
+                */
+
+               /* invalidate I-Cache */
+               if (armv7a->armv4_5_mmu.armv4_5_cache.i_cache_enabled)
+               {
+                       /* ICIMVAU - Invalidate Cache single entry
+                        * with MVA to PoU
+                        *      MCR p15, 0, r0, c7, c5, 1
+                        */
+                       for (uint32_t cacheline = address;
+                                       cacheline < address + size * count;
+                                       cacheline += 64) {
+                               retval = dpm->instr_write_data_r0(dpm,
+                                               ARMV4_5_MCR(15, 0, 0, 7, 5, 1),
+                                               cacheline);
+                               if (retval != ERROR_OK)
+                                       return retval;
+                       }
+               }
+
+               /* invalidate D-Cache */
+               if (armv7a->armv4_5_mmu.armv4_5_cache.d_u_cache_enabled)
+               {
+                       /* DCIMVAC - Invalidate data Cache line
+                        * with MVA to PoC
+                        *      MCR p15, 0, r0, c7, c6, 1
+                        */
+                       for (uint32_t cacheline = address;
+                                       cacheline < address + size * count;
+                                       cacheline += 64) {
+                               retval = dpm->instr_write_data_r0(dpm,
+                                               ARMV4_5_MCR(15, 0, 0, 7, 6, 1),
+                                               cacheline);
+                               if (retval != ERROR_OK)
+                                       return retval;
+                       }
+               }
+
+               /* (void) */ dpm->finish(dpm);
+       }
+
+       return retval;
+}
+
+static int cortex_a9_write_memory(struct target *target, uint32_t address,
+                uint32_t size, uint32_t count, uint8_t *buffer)
+{
+       int enabled = 0;
+       uint32_t virt, phys;
+       int retval;
+
+       LOG_DEBUG("Writing memory to address 0x%x; size %d; count %d", address, size, count);
+       retval = cortex_a9_mmu(target, &enabled);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if (enabled)
+       {
+               virt = address;
+               retval = cortex_a9_virt2phys(target, virt, &phys);
+               if (retval != ERROR_OK)
+                       return retval;
+               LOG_DEBUG("Writing to virtual address. Translating v:0x%x to r:0x%x", virt, phys);
+               address = phys;
+       }
+
+       return cortex_a9_write_phys_memory(target, address, size,
+                       count, buffer);
+}
+
+static int cortex_a9_bulk_write_memory(struct target *target, uint32_t address,
+               uint32_t count, uint8_t *buffer)
+{
+       return cortex_a9_write_memory(target, address, 4, count, buffer);
+}
+
+static int cortex_a9_dcc_read(struct adiv5_dap *swjdp, uint8_t *value, uint8_t *ctrl)
+{
+#if 0
+       u16 dcrdr;
+
+       mem_ap_read_buf_u16(swjdp, (uint8_t*)&dcrdr, 1, DCB_DCRDR);
+       *ctrl = (uint8_t)dcrdr;
+       *value = (uint8_t)(dcrdr >> 8);
+
+       LOG_DEBUG("data 0x%x ctrl 0x%x", *value, *ctrl);
+
+       /* write ack back to software dcc register
+        * signify we have read data */
+       if (dcrdr & (1 << 0))
+       {
+               dcrdr = 0;
+               mem_ap_write_buf_u16(swjdp, (uint8_t*)&dcrdr, 1, DCB_DCRDR);
+       }
+#endif
+       return ERROR_OK;
+}
+
+
+static int cortex_a9_handle_target_request(void *priv)
+{
+       struct target *target = priv;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       int retval;
+
+       if (!target_was_examined(target))
+               return ERROR_OK;
+       if (!target->dbg_msg_enabled)
+               return ERROR_OK;
+
+       if (target->state == TARGET_RUNNING)
+       {
+               uint8_t data = 0;
+               uint8_t ctrl = 0;
+
+               retval = cortex_a9_dcc_read(swjdp, &data, &ctrl);
+               if (retval != ERROR_OK)
+                       return retval;
+
+               /* check if we have data */
+               if (ctrl & (1 << 0))
+               {
+                       uint32_t request;
+
+                       /* we assume target is quick enough */
+                       request = data;
+                       retval = cortex_a9_dcc_read(swjdp, &data, &ctrl);
+                       if (retval != ERROR_OK)
+                               return retval;
+                       request |= (data << 8);
+                       retval = cortex_a9_dcc_read(swjdp, &data, &ctrl);
+                       if (retval != ERROR_OK)
+                               return retval;
+                       request |= (data << 16);
+                       retval = cortex_a9_dcc_read(swjdp, &data, &ctrl);
+                       if (retval != ERROR_OK)
+                               return retval;
+                       request |= (data << 24);
+                       target_request(target, request);
+               }
+       }
+
+       return ERROR_OK;
+}
+
+/*
+ * Cortex-A9 target information and configuration
+ */
+
+static int cortex_a9_examine_first(struct target *target)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       struct adiv5_dap *swjdp = &armv7a->dap;
+       int i;
+       int retval = ERROR_OK;
+       uint32_t didr, ctypr, ttypr, cpuid;
+
+       /* We do one extra read to ensure DAP is configured,
+        * we call ahbap_debugport_init(swjdp) instead
+        */
+       retval = ahbap_debugport_init(swjdp);
+       if (retval != ERROR_OK)
+               return retval;
+
+       dap_ap_select(swjdp, swjdp_debugap);
+
+       /*
+        * FIXME: assuming omap4430
+        *
+        * APB DBGBASE reads 0x80040000, but this points to an empty ROM table.
+        * 0x80000000 is cpu0 coresight region
+        */
+       if (target->coreid > 3) {
+               LOG_ERROR("cortex_a9 supports up to 4 cores");
+               return ERROR_INVALID_ARGUMENTS;
+       }
+       armv7a->debug_base = 0x80000000 |
+                       ((target->coreid & 0x3) << CORTEX_A9_PADDRDBG_CPU_SHIFT);
+
+       retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_CPUID, &cpuid);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if ((retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_CPUID, &cpuid)) != ERROR_OK)
+       {
+               LOG_DEBUG("Examine %s failed", "CPUID");
+               return retval;
+       }
+
+       if ((retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_CTYPR, &ctypr)) != ERROR_OK)
+       {
+               LOG_DEBUG("Examine %s failed", "CTYPR");
+               return retval;
+       }
+
+       if ((retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_TTYPR, &ttypr)) != ERROR_OK)
+       {
+               LOG_DEBUG("Examine %s failed", "TTYPR");
+               return retval;
+       }
+
+       if ((retval = mem_ap_read_atomic_u32(swjdp,
+                       armv7a->debug_base + CPUDBG_DIDR, &didr)) != ERROR_OK)
+       {
+               LOG_DEBUG("Examine %s failed", "DIDR");
+               return retval;
+       }
+
+       LOG_DEBUG("cpuid = 0x%08" PRIx32, cpuid);
+       LOG_DEBUG("ctypr = 0x%08" PRIx32, ctypr);
+       LOG_DEBUG("ttypr = 0x%08" PRIx32, ttypr);
+       LOG_DEBUG("didr = 0x%08" PRIx32, didr);
+
+       armv7a->armv4_5_common.core_type = ARM_MODE_MON;
+       retval = cortex_a9_dpm_setup(cortex_a9, didr);
+       if (retval != ERROR_OK)
+               return retval;
+
+       /* Setup Breakpoint Register Pairs */
+       cortex_a9->brp_num = ((didr >> 24) & 0x0F) + 1;
+       cortex_a9->brp_num_context = ((didr >> 20) & 0x0F) + 1;
+       cortex_a9->brp_num_available = cortex_a9->brp_num;
+       cortex_a9->brp_list = calloc(cortex_a9->brp_num, sizeof(struct cortex_a9_brp));
+//     cortex_a9->brb_enabled = ????;
+       for (i = 0; i < cortex_a9->brp_num; i++)
+       {
+               cortex_a9->brp_list[i].used = 0;
+               if (i < (cortex_a9->brp_num-cortex_a9->brp_num_context))
+                       cortex_a9->brp_list[i].type = BRP_NORMAL;
+               else
+                       cortex_a9->brp_list[i].type = BRP_CONTEXT;
+               cortex_a9->brp_list[i].value = 0;
+               cortex_a9->brp_list[i].control = 0;
+               cortex_a9->brp_list[i].BRPn = i;
+       }
+
+       LOG_DEBUG("Configured %i hw breakpoints", cortex_a9->brp_num);
+
+       target_set_examined(target);
+       return ERROR_OK;
+}
+
+static int cortex_a9_examine(struct target *target)
+{
+       int retval = ERROR_OK;
+
+       /* don't re-probe hardware after each reset */
+       if (!target_was_examined(target))
+               retval = cortex_a9_examine_first(target);
+
+       /* Configure core debug access */
+       if (retval == ERROR_OK)
+               retval = cortex_a9_init_debug_access(target);
+
+       return retval;
+}
+
+/*
+ *     Cortex-A9 target creation and initialization
+ */
+
+static int cortex_a9_init_target(struct command_context *cmd_ctx,
+               struct target *target)
+{
+       /* examine_first() does a bunch of this */
+       return ERROR_OK;
+}
+
+static int cortex_a9_init_arch_info(struct target *target,
+               struct cortex_a9_common *cortex_a9, struct jtag_tap *tap)
+{
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       struct arm *armv4_5 = &armv7a->armv4_5_common;
+       struct adiv5_dap *dap = &armv7a->dap;
+
+       armv7a->armv4_5_common.dap = dap;
+
+       /* Setup struct cortex_a9_common */
+       cortex_a9->common_magic = CORTEX_A9_COMMON_MAGIC;
+       armv4_5->arch_info = armv7a;
+
+       /* prepare JTAG information for the new target */
+       cortex_a9->jtag_info.tap = tap;
+       cortex_a9->jtag_info.scann_size = 4;
+
+       /* Leave (only) generic DAP stuff for debugport_init() */
+       dap->jtag_info = &cortex_a9->jtag_info;
+       dap->memaccess_tck = 80;
+
+       /* Number of bits for tar autoincrement, impl. dep. at least 10 */
+       dap->tar_autoincr_block = (1 << 10);
+
+       cortex_a9->fast_reg_read = 0;
+
+       /* Set default value */
+       cortex_a9->current_address_mode = ARM_MODE_ANY;
+
+       /* register arch-specific functions */
+       armv7a->examine_debug_reason = NULL;
+
+       armv7a->post_debug_entry = cortex_a9_post_debug_entry;
+
+       armv7a->pre_restore_context = NULL;
+       armv7a->armv4_5_mmu.armv4_5_cache.ctype = -1;
+       armv7a->armv4_5_mmu.get_ttb = cortex_a9_get_ttb;
+       armv7a->armv4_5_mmu.read_memory = cortex_a9_read_phys_memory;
+       armv7a->armv4_5_mmu.write_memory = cortex_a9_write_phys_memory;
+       armv7a->armv4_5_mmu.disable_mmu_caches = cortex_a9_disable_mmu_caches;
+       armv7a->armv4_5_mmu.enable_mmu_caches = cortex_a9_enable_mmu_caches;
+       armv7a->armv4_5_mmu.has_tiny_pages = 1;
+       armv7a->armv4_5_mmu.mmu_enabled = 0;
+
+
+//     arm7_9->handle_target_request = cortex_a9_handle_target_request;
+
+       /* REVISIT v7a setup should be in a v7a-specific routine */
+       arm_init_arch_info(target, armv4_5);
+       armv7a->common_magic = ARMV7_COMMON_MAGIC;
+
+       target_register_timer_callback(cortex_a9_handle_target_request, 1, 1, target);
+
+       return ERROR_OK;
+}
+
+static int cortex_a9_target_create(struct target *target, Jim_Interp *interp)
+{
+       struct cortex_a9_common *cortex_a9 = calloc(1, sizeof(struct cortex_a9_common));
+
+       return cortex_a9_init_arch_info(target, cortex_a9, target->tap);
+}
+
+static int cortex_a9_get_ttb(struct target *target, uint32_t *result)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+    struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+    uint32_t ttb = 0, retval = ERROR_OK;
+
+    /* current_address_mode is set inside cortex_a9_virt2phys()
+       where we can determine if address belongs to user or kernel */
+    if(cortex_a9->current_address_mode == ARM_MODE_SVC)
+    {
+        /* MRC p15,0,<Rt>,c1,c0,0 ; Read CP15 System Control Register */
+        retval = armv7a->armv4_5_common.mrc(target, 15,
+                    0, 1,   /* op1, op2 */
+                    2, 0,   /* CRn, CRm */
+                    &ttb);
+               if (retval != ERROR_OK)
+                       return retval;
+    }
+    else if(cortex_a9->current_address_mode == ARM_MODE_USR)
+    {
+        /* MRC p15,0,<Rt>,c1,c0,0 ; Read CP15 System Control Register */
+        retval = armv7a->armv4_5_common.mrc(target, 15,
+                    0, 0,   /* op1, op2 */
+                    2, 0,   /* CRn, CRm */
+                    &ttb);
+               if (retval != ERROR_OK)
+                       return retval;
+    }
+    /* we don't know whose address is: user or kernel
+       we assume that if we are in kernel mode then
+       address belongs to kernel else if in user mode
+       - to user */
+    else if(armv7a->armv4_5_common.core_mode == ARM_MODE_SVC)
+    {
+        /* MRC p15,0,<Rt>,c1,c0,0 ; Read CP15 System Control Register */
+        retval = armv7a->armv4_5_common.mrc(target, 15,
+                    0, 1,   /* op1, op2 */
+                    2, 0,   /* CRn, CRm */
+                    &ttb);
+               if (retval != ERROR_OK)
+                       return retval;
+    }
+    else if(armv7a->armv4_5_common.core_mode == ARM_MODE_USR)
+    {
+        /* MRC p15,0,<Rt>,c1,c0,0 ; Read CP15 System Control Register */
+        retval = armv7a->armv4_5_common.mrc(target, 15,
+                    0, 0,   /* op1, op2 */
+                    2, 0,   /* CRn, CRm */
+                    &ttb);
+               if (retval != ERROR_OK)
+                       return retval;
+    }
+    /* finally we don't know whose ttb to use: user or kernel */
+    else
+        LOG_ERROR("Don't know how to get ttb for current mode!!!");
+
+    ttb &= 0xffffc000;
+
+    *result = ttb;
+
+    return ERROR_OK;
+}
+
+static int cortex_a9_disable_mmu_caches(struct target *target, int mmu,
+                int d_u_cache, int i_cache)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       uint32_t cp15_control;
+       int retval;
+
+       /* read cp15 control register */
+       retval = armv7a->armv4_5_common.mrc(target, 15,
+                       0, 0,   /* op1, op2 */
+                       1, 0,   /* CRn, CRm */
+                       &cp15_control);
+       if (retval != ERROR_OK)
+               return retval;
+
+
+       if (mmu)
+               cp15_control &= ~0x1U;
+
+       if (d_u_cache)
+               cp15_control &= ~0x4U;
+
+       if (i_cache)
+               cp15_control &= ~0x1000U;
+
+       retval = armv7a->armv4_5_common.mcr(target, 15,
+                       0, 0,   /* op1, op2 */
+                       1, 0,   /* CRn, CRm */
+                       cp15_control);
+       return retval;
+}
+
+static int cortex_a9_enable_mmu_caches(struct target *target, int mmu,
+               int d_u_cache, int i_cache)
+{
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       uint32_t cp15_control;
+       int retval;
+
+       /* read cp15 control register */
+       retval = armv7a->armv4_5_common.mrc(target, 15,
+                       0, 0,   /* op1, op2 */
+                       1, 0,   /* CRn, CRm */
+                       &cp15_control);
+       if (retval != ERROR_OK)
+               return retval;
+
+       if (mmu)
+               cp15_control |= 0x1U;
+
+       if (d_u_cache)
+               cp15_control |= 0x4U;
+
+       if (i_cache)
+               cp15_control |= 0x1000U;
+
+       retval = armv7a->armv4_5_common.mcr(target, 15,
+                       0, 0,   /* op1, op2 */
+                       1, 0,   /* CRn, CRm */
+                       cp15_control);
+       return retval;
+}
+
+
+static int cortex_a9_mmu(struct target *target, int *enabled)
+{
+       if (target->state != TARGET_HALTED) {
+               LOG_ERROR("%s: target not halted", __func__);
+               return ERROR_TARGET_INVALID;
+       }
+
+       *enabled = target_to_cortex_a9(target)->armv7a_common.armv4_5_mmu.mmu_enabled;
+       return ERROR_OK;
+}
+
+static int cortex_a9_virt2phys(struct target *target,
+               uint32_t virt, uint32_t *phys)
+{
+       uint32_t cb;
+       struct cortex_a9_common *cortex_a9 = target_to_cortex_a9(target);
+       // struct armv7a_common *armv7a = &cortex_a9->armv7a_common;
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+
+    /* We assume that virtual address is separated
+       between user and kernel in Linux style:
+       0x00000000-0xbfffffff - User space
+       0xc0000000-0xffffffff - Kernel space */
+    if( virt < 0xc0000000 ) /* Linux user space */
+        cortex_a9->current_address_mode = ARM_MODE_USR;
+    else /* Linux kernel */
+        cortex_a9->current_address_mode = ARM_MODE_SVC;
+       uint32_t ret;
+       int retval = armv4_5_mmu_translate_va(target,
+                       &armv7a->armv4_5_mmu, virt, &cb, &ret);
+       if (retval != ERROR_OK)
+               return retval;
+    /* Reset the flag. We don't want someone else to use it by error */
+    cortex_a9->current_address_mode = ARM_MODE_ANY;
+
+       *phys = ret;
+       return ERROR_OK;
+}
+
+COMMAND_HANDLER(cortex_a9_handle_cache_info_command)
+{
+       struct target *target = get_current_target(CMD_CTX);
+       struct armv7a_common *armv7a = target_to_armv7a(target);
+
+       return armv4_5_handle_cache_info_command(CMD_CTX,
+                       &armv7a->armv4_5_mmu.armv4_5_cache);
+}
+
+
+COMMAND_HANDLER(cortex_a9_handle_dbginit_command)
+{
+       struct target *target = get_current_target(CMD_CTX);
+       if (!target_was_examined(target))
+       {
+               LOG_ERROR("target not examined yet");
+               return ERROR_FAIL;
+       }
+
+       return cortex_a9_init_debug_access(target);
+}
+
+static const struct command_registration cortex_a9_exec_command_handlers[] = {
+       {
+               .name = "cache_info",
+               .handler = cortex_a9_handle_cache_info_command,
+               .mode = COMMAND_EXEC,
+               .help = "display information about target caches",
+       },
+       {
+               .name = "dbginit",
+               .handler = cortex_a9_handle_dbginit_command,
+               .mode = COMMAND_EXEC,
+               .help = "Initialize core debug",
+       },
+       COMMAND_REGISTRATION_DONE
+};
+static const struct command_registration cortex_a9_command_handlers[] = {
+       {
+               .chain = arm_command_handlers,
+       },
+       {
+               .chain = armv7a_command_handlers,
+       },
+       {
+               .name = "cortex_a9",
+               .mode = COMMAND_ANY,
+               .help = "Cortex-A9 command group",
+               .chain = cortex_a9_exec_command_handlers,
+       },
+       COMMAND_REGISTRATION_DONE
+};
+
+struct target_type cortexa9_target = {
+       .name = "cortex_a9",
+
+       .poll = cortex_a9_poll,
+       .arch_state = armv7a_arch_state,
+
+       .target_request_data = NULL,
+
+       .halt = cortex_a9_halt,
+       .resume = cortex_a9_resume,
+       .step = cortex_a9_step,
+
+       .assert_reset = cortex_a9_assert_reset,
+       .deassert_reset = cortex_a9_deassert_reset,
+       .soft_reset_halt = NULL,
+
+       /* REVISIT allow exporting VFP3 registers ... */
+       .get_gdb_reg_list = arm_get_gdb_reg_list,
+
+       .read_memory = cortex_a9_read_memory,
+       .write_memory = cortex_a9_write_memory,
+       .bulk_write_memory = cortex_a9_bulk_write_memory,
+
+       .checksum_memory = arm_checksum_memory,
+       .blank_check_memory = arm_blank_check_memory,
+
+       .run_algorithm = armv4_5_run_algorithm,
+
+       .add_breakpoint = cortex_a9_add_breakpoint,
+       .remove_breakpoint = cortex_a9_remove_breakpoint,
+       .add_watchpoint = NULL,
+       .remove_watchpoint = NULL,
+
+       .commands = cortex_a9_command_handlers,
+       .target_create = cortex_a9_target_create,
+       .init_target = cortex_a9_init_target,
+       .examine = cortex_a9_examine,
+
+       .read_phys_memory = cortex_a9_read_phys_memory,
+       .write_phys_memory = cortex_a9_write_phys_memory,
+       .mmu = cortex_a9_mmu,
+       .virt2phys = cortex_a9_virt2phys,
+};
diff --git a/src/target/cortex_a9.h b/src/target/cortex_a9.h
new file mode 100644 (file)
index 0000000..82f04f8
--- /dev/null
@@ -0,0 +1,89 @@
+/***************************************************************************
+ *   Copyright (C) 2005 by Dominic Rath                                    *
+ *   Dominic.Rath@gmx.de                                                   *
+ *                                                                         *
+ *   Copyright (C) 2006 by Magnus Lundin                                   *
+ *   lundin@mlu.mine.nu                                                    *
+ *                                                                         *
+ *   Copyright (C) 2008 by Spencer Oliver                                  *
+ *   spen@spen-soft.co.uk                                                  *
+ *                                                                         *
+ *   Copyright (C) 2009 by Dirk Behme                                      *
+ *   dirk.behme@gmail.com - copy from cortex_m3                            *
+ *                                                                         *
+ *   This program is free software; you can redistribute it and/or modify  *
+ *   it under the terms of the GNU General Public License as published by  *
+ *   the Free Software Foundation; either version 2 of the License, or     *
+ *   (at your option) any later version.                                   *
+ *                                                                         *
+ *   This program is distributed in the hope that it will be useful,       *
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of        *
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         *
+ *   GNU General Public License for more details.                          *
+ *                                                                         *
+ *   You should have received a copy of the GNU General Public License     *
+ *   along with this program; if not, write to the                         *
+ *   Free Software Foundation, Inc.,                                       *
+ *   59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.             *
+ ***************************************************************************/
+#ifndef CORTEX_A9_H
+#define CORTEX_A9_H
+
+#include "armv7a.h"
+
+#define CORTEX_A9_COMMON_MAGIC 0x411fc082
+
+#define CPUDBG_CPUID   0xD00
+#define CPUDBG_CTYPR   0xD04
+#define CPUDBG_TTYPR   0xD0C
+#define CPUDBG_LOCKACCESS 0xFB0
+#define CPUDBG_LOCKSTATUS 0xFB4
+
+#define BRP_NORMAL 0
+#define BRP_CONTEXT 1
+
+#define CORTEX_A9_PADDRDBG_CPU_SHIFT 13
+
+struct cortex_a9_brp
+{
+       int used;
+       int type;
+       uint32_t value;
+       uint32_t control;
+       uint8_t BRPn;
+};
+
+struct cortex_a9_common
+{
+       int common_magic;
+       struct arm_jtag jtag_info;
+
+       /* Context information */
+       uint32_t cpudbg_dscr;
+
+       /* Saved cp15 registers */
+       uint32_t cp15_control_reg;
+
+       /* Breakpoint register pairs */
+       int brp_num_context;
+       int brp_num;
+       int brp_num_available;
+       struct cortex_a9_brp *brp_list;
+
+       /* Use cortex_a9_read_regs_through_mem for fast register reads */
+       int fast_reg_read;
+
+       /* Flag that helps to resolve what ttb to use: user or kernel */
+       int current_address_mode;
+
+       struct armv7a_common armv7a_common;
+};
+
+static inline struct cortex_a9_common *
+target_to_cortex_a9(struct target *target)
+{
+       return container_of(target->arch_info, struct cortex_a9_common,
+                       armv7a_common.armv4_5_common);
+}
+
+#endif /* CORTEX_A9_H */
index 31ade6acff60d9908b22734f3ed9092e325a1a17..66aa6bf7bf3d7987b3e31ece61835cd35940f895 100644 (file)
@@ -66,6 +66,7 @@ extern struct target_type dragonite_target;
 extern struct target_type xscale_target;
 extern struct target_type cortexm3_target;
 extern struct target_type cortexa8_target;
+extern struct target_type cortexa9_target;
 extern struct target_type arm11_target;
 extern struct target_type mips_m4k_target;
 extern struct target_type avr_target;
@@ -88,6 +89,7 @@ static struct target_type *target_types[] =
        &xscale_target,
        &cortexm3_target,
        &cortexa8_target,
+       &cortexa9_target,
        &arm11_target,
        &mips_m4k_target,
        &avr_target,