MIPS: fastdata bulk write fallback
[fw/openocd] / src / target / cortex_a8.c
index 08e546012efa849f0a330a314c53ea87d21c2fae..424263d0cbdb926e3b7dd7134454463314ec20d7 100644 (file)
 #include "register.h"
 #include "target_request.h"
 #include "target_type.h"
+#include "arm_opcodes.h"
 
 static int cortex_a8_poll(struct target *target);
 static int cortex_a8_debug_entry(struct target *target);
-static int cortex_a8_restore_context(struct target *target);
+static int cortex_a8_restore_context(struct target *target, bool bpwp);
 static int cortex_a8_set_breakpoint(struct target *target,
                struct breakpoint *breakpoint, uint8_t matchmode);
 static int cortex_a8_unset_breakpoint(struct target *target,
@@ -91,8 +92,8 @@ static int cortex_a8_init_debug_access(struct target *target)
 
 /* 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 (1 << DSCR_INSTR_COMP) if
- * you happen to know that no instruction is pending.
+ * value on return from this function; or DSCR_INSTR_COMP if you
+ * happen to know that no instruction is pending.
  */
 static int cortex_a8_exec_opcode(struct target *target,
                uint32_t opcode, uint32_t *dscr_p)
@@ -107,7 +108,7 @@ static int cortex_a8_exec_opcode(struct target *target,
        LOG_DEBUG("exec opcode 0x%08" PRIx32, opcode);
 
        /* Wait for InstrCompl bit to be set */
-       while ((dscr & (1 << DSCR_INSTR_COMP)) == 0)
+       while ((dscr & DSCR_INSTR_COMP) == 0)
        {
                retval = mem_ap_read_atomic_u32(swjdp,
                                armv7a->debug_base + CPUDBG_DSCR, &dscr);
@@ -130,7 +131,7 @@ static int cortex_a8_exec_opcode(struct target *target,
                        return retval;
                }
        }
-       while ((dscr & (1 << DSCR_INSTR_COMP)) == 0); /* Wait for InstrCompl bit to be set */
+       while ((dscr & DSCR_INSTR_COMP) == 0); /* Wait for InstrCompl bit to be set */
 
        if (dscr_p)
                *dscr_p = dscr;
@@ -159,97 +160,6 @@ static int cortex_a8_read_regs_through_mem(struct target *target, uint32_t addre
        return retval;
 }
 
-static int cortex_a8_read_cp(struct target *target, uint32_t *value, uint8_t CP,
-               uint8_t op1, uint8_t CRn, uint8_t CRm, uint8_t op2)
-{
-       int retval;
-       struct armv7a_common *armv7a = target_to_armv7a(target);
-       struct swjdp_common *swjdp = &armv7a->swjdp_info;
-       uint32_t dscr = 0;
-
-       /* MRC(...) to read coprocessor register into r0 */
-       cortex_a8_exec_opcode(target, ARMV4_5_MRC(CP, op1, 0, CRn, CRm, op2),
-                       &dscr);
-
-       /* Move R0 to DTRTX */
-       cortex_a8_exec_opcode(target, ARMV4_5_MCR(14, 0, 0, 0, 5, 0),
-                       &dscr);
-
-       /* Read DCCTX */
-       retval = mem_ap_read_atomic_u32(swjdp,
-                       armv7a->debug_base + CPUDBG_DTRTX, value);
-
-       return retval;
-}
-
-static int cortex_a8_write_cp(struct target *target, uint32_t value,
-       uint8_t CP, uint8_t op1, uint8_t CRn, uint8_t CRm, uint8_t op2)
-{
-       int retval;
-       uint32_t dscr;
-       struct armv7a_common *armv7a = target_to_armv7a(target);
-       struct swjdp_common *swjdp = &armv7a->swjdp_info;
-
-       LOG_DEBUG("CP%i, CRn %i, value 0x%08" PRIx32, CP, CRn, value);
-
-       /* Check that DCCRX is not full */
-       retval = mem_ap_read_atomic_u32(swjdp,
-                               armv7a->debug_base + CPUDBG_DSCR, &dscr);
-       if (dscr & (1 << 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 */
-               cortex_a8_exec_opcode(target, ARMV4_5_MRC(14, 0, 0, 0, 5, 0),
-                               &dscr);
-       }
-
-       /* Write DTRRX ... sets DSCR.DTRRXfull but exec_opcode() won't care */
-       retval = mem_ap_write_u32(swjdp,
-                       armv7a->debug_base + CPUDBG_DTRRX, value);
-
-       /* Move DTRRX to r0 */
-       cortex_a8_exec_opcode(target, ARMV4_5_MRC(14, 0, 0, 0, 5, 0), &dscr);
-
-       /* MCR(...) to write r0 to coprocessor */
-       return cortex_a8_exec_opcode(target,
-                       ARMV4_5_MCR(CP, op1, 0, CRn, CRm, op2),
-                       &dscr);
-}
-
-static int cortex_a8_read_cp15(struct target *target, uint32_t op1, uint32_t op2,
-               uint32_t CRn, uint32_t CRm, uint32_t *value)
-{
-       return cortex_a8_read_cp(target, value, 15, op1, CRn, CRm, op2);
-}
-
-static int cortex_a8_write_cp15(struct target *target, uint32_t op1, uint32_t op2,
-               uint32_t CRn, uint32_t CRm, uint32_t value)
-{
-       return cortex_a8_write_cp(target, value, 15, op1, CRn, CRm, op2);
-}
-
-static int cortex_a8_mrc(struct target *target, int cpnum, uint32_t op1, uint32_t op2, uint32_t CRn, uint32_t CRm, uint32_t *value)
-{
-       if (cpnum!=15)
-       {
-               LOG_ERROR("Only cp15 is supported");
-               return ERROR_FAIL;
-       }
-       return cortex_a8_read_cp15(target, op1, op2, CRn, CRm, value);
-}
-
-static int cortex_a8_mcr(struct target *target, int cpnum, uint32_t op1, uint32_t op2, uint32_t CRn, uint32_t CRm, uint32_t value)
-{
-       if (cpnum!=15)
-       {
-               LOG_ERROR("Only cp15 is supported");
-               return ERROR_FAIL;
-       }
-       return cortex_a8_write_cp15(target, op1, op2, CRn, CRm, value);
-}
-
-
-
 static int cortex_a8_dap_read_coreregister_u32(struct target *target,
                uint32_t *value, int regnum)
 {
@@ -289,7 +199,7 @@ static int cortex_a8_dap_read_coreregister_u32(struct target *target,
        }
 
        /* Wait for DTRRXfull then read DTRRTX */
-       while ((dscr & (1 << DSCR_DTR_TX_FULL)) == 0)
+       while ((dscr & DSCR_DTR_TX_FULL) == 0)
        {
                retval = mem_ap_read_atomic_u32(swjdp,
                                armv7a->debug_base + CPUDBG_DSCR, &dscr);
@@ -316,7 +226,7 @@ static int cortex_a8_dap_write_coreregister_u32(struct target *target,
        /* Check that DCCRX is not full */
        retval = mem_ap_read_atomic_u32(swjdp,
                                armv7a->debug_base + CPUDBG_DSCR, &dscr);
-       if (dscr & (1 << DSCR_DTR_RX_FULL))
+       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 */
@@ -406,14 +316,14 @@ static int cortex_a8_read_dcc(struct cortex_a8_common *a8, uint32_t *data,
                uint32_t *dscr_p)
 {
        struct swjdp_common *swjdp = &a8->armv7a_common.swjdp_info;
-       uint32_t dscr = 1 << DSCR_INSTR_COMP;
+       uint32_t dscr = DSCR_INSTR_COMP;
        int retval;
 
        if (dscr_p)
                dscr = *dscr_p;
 
        /* Wait for DTRRXfull */
-       while ((dscr & (1 << DSCR_DTR_TX_FULL)) == 0) {
+       while ((dscr & DSCR_DTR_TX_FULL) == 0) {
                retval = mem_ap_read_atomic_u32(swjdp,
                                a8->armv7a_common.debug_base + CPUDBG_DSCR,
                                &dscr);
@@ -421,7 +331,7 @@ static int cortex_a8_read_dcc(struct cortex_a8_common *a8, uint32_t *data,
 
        retval = mem_ap_read_atomic_u32(swjdp,
                        a8->armv7a_common.debug_base + CPUDBG_DTRTX, data);
-       LOG_DEBUG("read DCC 0x%08" PRIx32, *data);
+       //LOG_DEBUG("read DCC 0x%08" PRIx32, *data);
 
        if (dscr_p)
                *dscr_p = dscr;
@@ -441,10 +351,10 @@ static int cortex_a8_dpm_prepare(struct arm_dpm *dpm)
                retval = mem_ap_read_atomic_u32(swjdp,
                                a8->armv7a_common.debug_base + CPUDBG_DSCR,
                                &dscr);
-       } while ((dscr & (1 << DSCR_INSTR_COMP)) == 0);
+       } while ((dscr & DSCR_INSTR_COMP) == 0);
 
        /* this "should never happen" ... */
-       if (dscr & (1 << DSCR_DTR_RX_FULL)) {
+       if (dscr & DSCR_DTR_RX_FULL) {
                LOG_ERROR("DSCR_DTR_RX_FULL, dscr 0x%08" PRIx32, dscr);
                /* Clear DCCRX */
                retval = cortex_a8_exec_opcode(
@@ -467,7 +377,7 @@ static int cortex_a8_instr_write_data_dcc(struct arm_dpm *dpm,
 {
        struct cortex_a8_common *a8 = dpm_to_a8(dpm);
        int retval;
-       uint32_t dscr = 1 << DSCR_INSTR_COMP;
+       uint32_t dscr = DSCR_INSTR_COMP;
 
        retval = cortex_a8_write_dcc(a8, data);
 
@@ -481,7 +391,7 @@ static int cortex_a8_instr_write_data_r0(struct arm_dpm *dpm,
                uint32_t opcode, uint32_t data)
 {
        struct cortex_a8_common *a8 = dpm_to_a8(dpm);
-       uint32_t dscr = 1 << DSCR_INSTR_COMP;
+       uint32_t dscr = DSCR_INSTR_COMP;
        int retval;
 
        retval = cortex_a8_write_dcc(a8, data);
@@ -504,7 +414,7 @@ static int cortex_a8_instr_write_data_r0(struct arm_dpm *dpm,
 static int cortex_a8_instr_cpsr_sync(struct arm_dpm *dpm)
 {
        struct target *target = dpm->arm->target;
-       uint32_t dscr = 1 << DSCR_INSTR_COMP;
+       uint32_t dscr = DSCR_INSTR_COMP;
 
        /* "Prefetch flush" after modifying execution status in CPSR */
        return cortex_a8_exec_opcode(target,
@@ -517,7 +427,7 @@ static int cortex_a8_instr_read_data_dcc(struct arm_dpm *dpm,
 {
        struct cortex_a8_common *a8 = dpm_to_a8(dpm);
        int retval;
-       uint32_t dscr = 1 << DSCR_INSTR_COMP;
+       uint32_t dscr = DSCR_INSTR_COMP;
 
        /* the opcode, writing data to DCC */
        retval = cortex_a8_exec_opcode(
@@ -533,7 +443,7 @@ static int cortex_a8_instr_read_data_r0(struct arm_dpm *dpm,
                uint32_t opcode, uint32_t *data)
 {
        struct cortex_a8_common *a8 = dpm_to_a8(dpm);
-       uint32_t dscr = 1 << DSCR_INSTR_COMP;
+       uint32_t dscr = DSCR_INSTR_COMP;
        int retval;
 
        /* the opcode, writing data to R0 */
@@ -551,9 +461,70 @@ static int cortex_a8_instr_read_data_r0(struct arm_dpm *dpm,
        return cortex_a8_read_dcc(a8, data, &dscr);
 }
 
+static int cortex_a8_bpwp_enable(struct arm_dpm *dpm, unsigned index,
+               uint32_t addr, uint32_t control)
+{
+       struct cortex_a8_common *a8 = dpm_to_a8(dpm);
+       uint32_t vr = a8->armv7a_common.debug_base;
+       uint32_t cr = a8->armv7a_common.debug_base;
+       int retval;
+
+       switch (index) {
+       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 -= 16;
+               break;
+       default:
+               return ERROR_FAIL;
+       }
+       vr += 4 * index;
+       cr += 4 * index;
+
+       LOG_DEBUG("A8: bpwp enable, vr %08x cr %08x",
+                       (unsigned) vr, (unsigned) cr);
+
+       retval = cortex_a8_dap_write_memap_register_u32(dpm->arm->target,
+                       vr, addr);
+       if (retval != ERROR_OK)
+               return retval;
+       retval = cortex_a8_dap_write_memap_register_u32(dpm->arm->target,
+                       cr, control);
+       return retval;
+}
+
+static int cortex_a8_bpwp_disable(struct arm_dpm *dpm, unsigned index)
+{
+       struct cortex_a8_common *a8 = dpm_to_a8(dpm);
+       uint32_t cr;
+
+       switch (index) {
+       case 0 ... 15:
+               cr = a8->armv7a_common.debug_base + CPUDBG_BCR_BASE;
+               break;
+       case 16 ... 31:
+               cr = a8->armv7a_common.debug_base + CPUDBG_WCR_BASE;
+               index -= 16;
+               break;
+       default:
+               return ERROR_FAIL;
+       }
+       cr += 4 * index;
+
+       LOG_DEBUG("A8: bpwp disable, cr %08x", (unsigned) cr);
+
+       /* clear control register */
+       return cortex_a8_dap_write_memap_register_u32(dpm->arm->target, cr, 0);
+}
+
 static int cortex_a8_dpm_setup(struct cortex_a8_common *a8, uint32_t didr)
 {
        struct arm_dpm *dpm = &a8->armv7a_common.dpm;
+       int retval;
 
        dpm->arm = &a8->armv7a_common.armv4_5_common;
        dpm->didr = didr;
@@ -568,7 +539,14 @@ static int cortex_a8_dpm_setup(struct cortex_a8_common *a8, uint32_t didr)
        dpm->instr_read_data_dcc = cortex_a8_instr_read_data_dcc;
        dpm->instr_read_data_r0 = cortex_a8_instr_read_data_r0;
 
-       return arm_dpm_setup(dpm);
+       dpm->bpwp_enable = cortex_a8_bpwp_enable;
+       dpm->bpwp_disable = cortex_a8_bpwp_disable;
+
+       retval = arm_dpm_setup(dpm);
+       if (retval == ERROR_OK)
+               retval = arm_dpm_initialize(dpm);
+
+       return retval;
 }
 
 
@@ -662,7 +640,7 @@ static int cortex_a8_halt(struct target *target)
         */
        mem_ap_read_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_DSCR, &dscr);
        retval = mem_ap_write_atomic_u32(swjdp,
-               armv7a->debug_base + CPUDBG_DSCR, dscr | (1 << DSCR_HALT_DBG_MODE));
+               armv7a->debug_base + CPUDBG_DSCR, dscr | DSCR_HALT_DBG_MODE);
 
        if (retval != ERROR_OK)
                goto out;
@@ -670,7 +648,7 @@ static int cortex_a8_halt(struct target *target)
        do {
                mem_ap_read_atomic_u32(swjdp,
                        armv7a->debug_base + CPUDBG_DSCR, &dscr);
-       } while ((dscr & (1 << DSCR_CORE_HALTED)) == 0);
+       } while ((dscr & DSCR_CORE_HALTED) == 0);
 
        target->debug_reason = DBG_REASON_DBGRQ;
 
@@ -693,11 +671,7 @@ static int cortex_a8_resume(struct target *target, int current,
        dap_ap_select(swjdp, swjdp_debugap);
 
        if (!debug_execution)
-       {
                target_free_all_working_areas(target);
-//             cortex_m3_enable_breakpoints(target);
-//             cortex_m3_enable_watchpoints(target);
-       }
 
 #if 0
        if (debug_execution)
@@ -732,17 +706,17 @@ static int cortex_a8_resume(struct target *target, int current,
         */
        switch (armv4_5->core_state)
        {
-       case ARMV4_5_STATE_ARM:
+       case ARM_STATE_ARM:
                resume_pc &= 0xFFFFFFFC;
                break;
-       case ARMV4_5_STATE_THUMB:
+       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 ARMV4_5_STATE_JAZELLE:
+       case ARM_STATE_JAZELLE:
                LOG_ERROR("How do I resume into Jazelle state??");
                return ERROR_FAIL;
        }
@@ -752,7 +726,7 @@ static int cortex_a8_resume(struct target *target, int current,
        armv4_5->core_cache->reg_list[15].dirty = 1;
        armv4_5->core_cache->reg_list[15].valid = 1;
 
-       cortex_a8_restore_context(target);
+       cortex_a8_restore_context(target, handle_breakpoints);
 
 #if 0
        /* the front-end may request us not to handle breakpoints */
@@ -769,13 +743,18 @@ static int cortex_a8_resume(struct target *target, int current,
        }
 
 #endif
-       /* Restart core and wait for it to be started */
+       /* Restart core and wait for it to be started
+        * NOTE: this clears DSCR_ITR_EN and other bits.
+        *
+        * REVISIT: for single stepping, we probably want to
+        * disable IRQs by default, with optional override...
+        */
        mem_ap_write_atomic_u32(swjdp, armv7a->debug_base + CPUDBG_DRCR, 0x2);
 
        do {
                mem_ap_read_atomic_u32(swjdp,
                        armv7a->debug_base + CPUDBG_DSCR, &dscr);
-       } while ((dscr & (1 << DSCR_CORE_RESTARTED)) == 0);
+       } while ((dscr & DSCR_CORE_RESTARTED) == 0);
 
        target->debug_reason = DBG_REASON_NOTHALTED;
        target->state = TARGET_RUNNING;
@@ -804,7 +783,7 @@ static int cortex_a8_resume(struct target *target, int current,
 static int cortex_a8_debug_entry(struct target *target)
 {
        int i;
-       uint32_t regfile[16], pc, cpsr, dscr;
+       uint32_t regfile[16], cpsr, dscr;
        int retval = ERROR_OK;
        struct working_area *regfile_working_area = NULL;
        struct cortex_a8_common *cortex_a8 = target_to_cortex_a8(target);
@@ -815,7 +794,7 @@ static int cortex_a8_debug_entry(struct target *target)
 
        LOG_DEBUG("dscr = 0x%08" PRIx32, cortex_a8->cpudbg_dscr);
 
-       /* Enable the ITR execution once we are in debug mode */
+       /* REVISIT surely we should not re-read DSCR !! */
        mem_ap_read_atomic_u32(swjdp,
                                armv7a->debug_base + CPUDBG_DSCR, &dscr);
 
@@ -824,31 +803,22 @@ static int cortex_a8_debug_entry(struct target *target)
         * Synchronization Barrier:  ARMV4_5_MCR(15, 0, 0, 7, 10, 4).
         */
 
-       dscr |= (1 << DSCR_EXT_INT_EN);
+       /* 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);
 
        /* Examine debug reason */
-       switch ((cortex_a8->cpudbg_dscr >> 2)&0xF)
-       {
-               case 0:         /* DRCR[0] write */
-               case 4:         /* EDBGRQ */
-                       target->debug_reason = DBG_REASON_DBGRQ;
-                       break;
-               case 1:         /* HW breakpoint */
-               case 3:         /* SW BKPT */
-               case 5:         /* vector catch */
-                       target->debug_reason = DBG_REASON_BREAKPOINT;
-                       break;
-               case 10:        /* precise watchpoint */
-                       target->debug_reason = DBG_REASON_WATCHPOINT;
-                       /* REVISIT could collect WFAR later, to see just
-                        * which instruction triggered the watchpoint.
-                        */
-                       break;
-               default:
-                       target->debug_reason = DBG_REASON_UNDEFINED;
-                       break;
+       arm_dpm_report_dscr(&armv7a->dpm, cortex_a8->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);
+               arm_dpm_report_wfar(&armv7a->dpm, wfar);
        }
 
        /* REVISIT fast_reg_read is never set ... */
@@ -872,7 +842,6 @@ static int cortex_a8_debug_entry(struct target *target)
 
                /* read Current PSR */
                cortex_a8_dap_read_coreregister_u32(target, &cpsr, 16);
-               pc = regfile[15];
                dap_ap_select(swjdp, swjdp_debugap);
                LOG_DEBUG("cpsr: %8.8" PRIx32, cpsr);
 
@@ -923,29 +892,33 @@ static int cortex_a8_debug_entry(struct target *target)
        if (armv7a->post_debug_entry)
                armv7a->post_debug_entry(target);
 
-
-
        return retval;
-
 }
 
 static void cortex_a8_post_debug_entry(struct target *target)
 {
        struct cortex_a8_common *cortex_a8 = target_to_cortex_a8(target);
        struct armv7a_common *armv7a = &cortex_a8->armv7a_common;
+       int retval;
 
-//     cortex_a8_read_cp(target, &cp15_control_register, 15, 0, 1, 0, 0);
-       /* examine cp15 control reg */
-       armv7a->read_cp15(target, 0, 0, 1, 0, &cortex_a8->cp15_control_reg);
-       jtag_execute_queue();
+       /* 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_a8->cp15_control_reg);
        LOG_DEBUG("cp15_control_reg: %8.8" PRIx32, cortex_a8->cp15_control_reg);
 
        if (armv7a->armv4_5_mmu.armv4_5_cache.ctype == -1)
        {
                uint32_t cache_type_reg;
-               /* identify caches */
-               armv7a->read_cp15(target, 0, 1, 0, 0, &cache_type_reg);
-               jtag_execute_queue();
+
+               /* 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);
+               LOG_DEBUG("cp15 cache type: %8.8x", (unsigned) cache_type_reg);
+
                /* FIXME the armv4_4 cache info DOES NOT APPLY to Cortex-A8 */
                armv4_5_identify_cache(cache_type_reg,
                                &armv7a->armv4_5_mmu.armv4_5_cache);
@@ -1002,7 +975,7 @@ static int cortex_a8_step(struct target *target, int current, uint32_t address,
 
        /* Setup single step breakpoint */
        stepbreakpoint.address = address;
-       stepbreakpoint.length = (armv4_5->core_state == ARMV4_5_STATE_THUMB)
+       stepbreakpoint.length = (armv4_5->core_state == ARM_STATE_THUMB)
                        ? 2 : 4;
        stepbreakpoint.type = BKPT_HARD;
        stepbreakpoint.set = 0;
@@ -1025,7 +998,8 @@ static int cortex_a8_step(struct target *target, int current, uint32_t address,
        }
 
        cortex_a8_unset_breakpoint(target, &stepbreakpoint);
-       if (timeout > 0) target->debug_reason = DBG_REASON_BREAKPOINT;
+       if (timeout > 0)
+               target->debug_reason = DBG_REASON_BREAKPOINT;
 
        if (breakpoint)
                cortex_a8_set_breakpoint(target, breakpoint, 0);
@@ -1036,7 +1010,7 @@ static int cortex_a8_step(struct target *target, int current, uint32_t address,
        return ERROR_OK;
 }
 
-static int cortex_a8_restore_context(struct target *target)
+static int cortex_a8_restore_context(struct target *target, bool bpwp)
 {
        struct armv7a_common *armv7a = target_to_armv7a(target);
 
@@ -1045,7 +1019,7 @@ static int cortex_a8_restore_context(struct target *target)
        if (armv7a->pre_restore_context)
                armv7a->pre_restore_context(target);
 
-       arm_dpm_write_dirty_registers(&armv7a->dpm);
+       arm_dpm_write_dirty_registers(&armv7a->dpm, bpwp);
 
        if (armv7a->post_restore_context)
                armv7a->post_restore_context(target);
@@ -1246,6 +1220,21 @@ static int cortex_a8_assert_reset(struct target *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);
 
@@ -1256,14 +1245,22 @@ static int cortex_a8_assert_reset(struct target *target)
 
 static int cortex_a8_deassert_reset(struct target *target)
 {
+       int retval;
 
        LOG_DEBUG(" ");
 
-       if (target->reset_halt)
-       {
-               int retval;
-               if ((retval = target_halt(target)) != ERROR_OK)
-                       return retval;
+       /* be certain SRST is off */
+       jtag_add_reset(0, 0);
+
+       retval = cortex_a8_poll(target);
+
+       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;
@@ -1327,25 +1324,57 @@ static int cortex_a8_write_memory(struct target *target, uint32_t address,
                }
        }
 
+       /* REVISIT this op is generic ARMv7-A/R stuff */
        if (retval == ERROR_OK && target->state == TARGET_HALTED)
        {
-               /* The Cache handling will NOT work with MMU active, the wrong addresses will be invalidated */
+               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-A8 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)
                {
-                       /* Invalidate ICache single entry with MVA, repeat this for all cache
-                          lines in the address range, Cortex-A8 has fixed 64 byte line length */
-                       /* Invalidate Cache single entry with MVA to PoU */
-                       for (uint32_t cacheline=address; cacheline<address+size*count; cacheline+=64)
-                               armv7a->write_cp15(target, 0, 1, 7, 5, cacheline); /* I-Cache to PoU */
+                       /* 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);
+                       }
                }
+
                /* invalidate D-Cache */
                if (armv7a->armv4_5_mmu.armv4_5_cache.d_u_cache_enabled)
                {
-                       /* Invalidate Cache single entry with MVA to PoC */
-                       for (uint32_t cacheline=address; cacheline<address+size*count; cacheline+=64)
-                               armv7a->write_cp15(target, 0, 1, 7, 6, cacheline); /* U/D cache to PoC */
+                       /* 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);
+                       }
                }
+
+               /* (void) */ dpm->finish(dpm);
        }
 
        return retval;
@@ -1475,6 +1504,7 @@ static int cortex_a8_examine_first(struct target *target)
        LOG_DEBUG("ttypr = 0x%08" PRIx32, ttypr);
        LOG_DEBUG("didr = 0x%08" PRIx32, didr);
 
+       armv7a->armv4_5_common.core_type = ARM_MODE_MON;
        cortex_a8_dpm_setup(cortex_a8, didr);
 
        /* Setup Breakpoint Register Pairs */
@@ -1495,20 +1525,7 @@ static int cortex_a8_examine_first(struct target *target)
                cortex_a8->brp_list[i].BRPn = i;
        }
 
-       /* Setup Watchpoint Register Pairs */
-       cortex_a8->wrp_num = ((didr >> 28) & 0x0F) + 1;
-       cortex_a8->wrp_num_available = cortex_a8->wrp_num;
-       cortex_a8->wrp_list = calloc(cortex_a8->wrp_num, sizeof(struct cortex_a8_wrp));
-       for (i = 0; i < cortex_a8->wrp_num; i++)
-       {
-               cortex_a8->wrp_list[i].used = 0;
-               cortex_a8->wrp_list[i].type = 0;
-               cortex_a8->wrp_list[i].value = 0;
-               cortex_a8->wrp_list[i].control = 0;
-               cortex_a8->wrp_list[i].WRPn = i;
-       }
-       LOG_DEBUG("Configured %i hw breakpoint pairs and %i hw watchpoint pairs",
-                       cortex_a8->brp_num , cortex_a8->wrp_num);
+       LOG_DEBUG("Configured %i hw breakpoints", cortex_a8->brp_num);
 
        target_set_examined(target);
        return ERROR_OK;
@@ -1566,7 +1583,6 @@ static int cortex_a8_init_arch_info(struct target *target,
 
        cortex_a8->fast_reg_read = 0;
 
-
        /* register arch-specific functions */
        armv7a->examine_debug_reason = NULL;
 
@@ -1582,14 +1598,12 @@ static int cortex_a8_init_arch_info(struct target *target,
 //     armv7a->armv4_5_mmu.enable_mmu_caches = armv7a_enable_mmu_caches;
        armv7a->armv4_5_mmu.has_tiny_pages = 1;
        armv7a->armv4_5_mmu.mmu_enabled = 0;
-       armv7a->read_cp15 = cortex_a8_read_cp15;
-       armv7a->write_cp15 = cortex_a8_write_cp15;
 
 
 //     arm7_9->handle_target_request = cortex_a8_handle_target_request;
 
        /* REVISIT v7a setup should be in a v7a-specific routine */
-       armv4_5_init_arch_info(target, armv4_5);
+       arm_init_arch_info(target, armv4_5);
        armv7a->common_magic = ARMV7_COMMON_MAGIC;
 
        target_register_timer_callback(cortex_a8_handle_target_request, 1, 1, target);
@@ -1625,29 +1639,36 @@ COMMAND_HANDLER(cortex_a8_handle_dbginit_command)
        return ERROR_OK;
 }
 
-
-static int cortex_a8_register_commands(struct command_context *cmd_ctx)
-{
-       struct command *cortex_a8_cmd;
-       int retval = ERROR_OK;
-
-       armv4_5_register_commands(cmd_ctx);
-       armv7a_register_commands(cmd_ctx);
-
-       cortex_a8_cmd = register_command(cmd_ctx, NULL, "cortex_a8",
-                       NULL, COMMAND_ANY,
-                       "cortex_a8 specific commands");
-
-       register_command(cmd_ctx, cortex_a8_cmd, "cache_info",
-                       cortex_a8_handle_cache_info_command, COMMAND_EXEC,
-                       "display information about target caches");
-
-       register_command(cmd_ctx, cortex_a8_cmd, "dbginit",
-                       cortex_a8_handle_dbginit_command, COMMAND_EXEC,
-                       "Initialize core debug");
-
-       return retval;
-}
+static const struct command_registration cortex_a8_exec_command_handlers[] = {
+       {
+               .name = "cache_info",
+               .handler = &cortex_a8_handle_cache_info_command,
+               .mode = COMMAND_EXEC,
+               .help = "display information about target caches",
+       },
+       {
+               .name = "dbginit",
+               .handler = &cortex_a8_handle_dbginit_command,
+               .mode = COMMAND_EXEC,
+               .help = "Initialize core debug",
+       },
+       COMMAND_REGISTRATION_DONE
+};
+static const struct command_registration cortex_a8_command_handlers[] = {
+       {
+               .chain = arm_command_handlers,
+       },
+       {
+               .chain = armv7a_command_handlers,
+       },
+       {
+               .name = "cortex_a8",
+               .mode = COMMAND_ANY,
+               .help = "Cortex-A8 command group",
+               .chain = cortex_a8_exec_command_handlers,
+       },
+       COMMAND_REGISTRATION_DONE
+};
 
 struct target_type cortexa8_target = {
        .name = "cortex_a8",
@@ -1665,7 +1686,8 @@ struct target_type cortexa8_target = {
        .deassert_reset = cortex_a8_deassert_reset,
        .soft_reset_halt = NULL,
 
-       .get_gdb_reg_list = armv4_5_get_gdb_reg_list,
+       /* REVISIT allow exporting VFP3 registers ... */
+       .get_gdb_reg_list = arm_get_gdb_reg_list,
 
        .read_memory = cortex_a8_read_memory,
        .write_memory = cortex_a8_write_memory,
@@ -1681,10 +1703,8 @@ struct target_type cortexa8_target = {
        .add_watchpoint = NULL,
        .remove_watchpoint = NULL,
 
-       .register_commands = cortex_a8_register_commands,
+       .commands = cortex_a8_command_handlers,
        .target_create = cortex_a8_target_create,
        .init_target = cortex_a8_init_target,
        .examine = cortex_a8_examine,
-       .mrc = cortex_a8_mrc,
-       .mcr = cortex_a8_mcr,
 };