summary refs log tree commit diff stats
path: root/target-arm
diff options
context:
space:
mode:
Diffstat (limited to 'target-arm')
-rw-r--r--target-arm/cpu.c37
-rw-r--r--target-arm/cpu.h2
-rw-r--r--target-arm/helper.c397
-rw-r--r--target-arm/internals.h30
-rw-r--r--target-arm/machine.c3
-rw-r--r--target-arm/op_helper.c188
6 files changed, 573 insertions, 84 deletions
diff --git a/target-arm/cpu.c b/target-arm/cpu.c
index 8199f32e32..7ea12bda1c 100644
--- a/target-arm/cpu.c
+++ b/target-arm/cpu.c
@@ -129,26 +129,38 @@ static void arm_cpu_reset(CPUState *s)
     env->uncached_cpsr = ARM_CPU_MODE_SVC;
     env->daif = PSTATE_D | PSTATE_A | PSTATE_I | PSTATE_F;
     /* On ARMv7-M the CPSR_I is the value of the PRIMASK register, and is
-       clear at reset.  Initial SP and PC are loaded from ROM.  */
+     * clear at reset. Initial SP and PC are loaded from ROM.
+     */
     if (IS_M(env)) {
-        uint32_t pc;
+        uint32_t initial_msp; /* Loaded from 0x0 */
+        uint32_t initial_pc; /* Loaded from 0x4 */
         uint8_t *rom;
+
         env->daif &= ~PSTATE_I;
         rom = rom_ptr(0);
         if (rom) {
-            /* We should really use ldl_phys here, in case the guest
-               modified flash and reset itself.  However images
-               loaded via -kernel have not been copied yet, so load the
-               values directly from there.  */
-            env->regs[13] = ldl_p(rom) & 0xFFFFFFFC;
-            pc = ldl_p(rom + 4);
-            env->thumb = pc & 1;
-            env->regs[15] = pc & ~1;
+            /* Address zero is covered by ROM which hasn't yet been
+             * copied into physical memory.
+             */
+            initial_msp = ldl_p(rom);
+            initial_pc = ldl_p(rom + 4);
+        } else {
+            /* Address zero not covered by a ROM blob, or the ROM blob
+             * is in non-modifiable memory and this is a second reset after
+             * it got copied into memory. In the latter case, rom_ptr
+             * will return a NULL pointer and we should use ldl_phys instead.
+             */
+            initial_msp = ldl_phys(s->as, 0);
+            initial_pc = ldl_phys(s->as, 4);
         }
+
+        env->regs[13] = initial_msp & 0xFFFFFFFC;
+        env->regs[15] = initial_pc & ~1;
+        env->thumb = initial_pc & 1;
     }
 
     if (env->cp15.c1_sys & SCTLR_V) {
-            env->regs[15] = 0xFFFF0000;
+        env->regs[15] = 0xFFFF0000;
     }
 
     env->vfp.xregs[ARM_VFP_FPEXC] = 0;
@@ -172,6 +184,8 @@ static void arm_cpu_reset(CPUState *s)
         kvm_arm_reset_vcpu(cpu);
     }
 #endif
+
+    hw_watchpoint_update_all(cpu);
 }
 
 #ifndef CONFIG_USER_ONLY
@@ -1051,6 +1065,7 @@ static void arm_cpu_class_init(ObjectClass *oc, void *data)
 #endif
     cc->gdb_num_core_regs = 26;
     cc->gdb_core_xml_file = "arm-core.xml";
+    cc->debug_excp_handler = arm_debug_excp_handler;
 }
 
 static void cpu_register(const ARMCPUInfo *info)
diff --git a/target-arm/cpu.h b/target-arm/cpu.h
index 51bedc8262..d1e1ccb605 100644
--- a/target-arm/cpu.h
+++ b/target-arm/cpu.h
@@ -323,6 +323,8 @@ typedef struct CPUARMState {
     int eabi;
 #endif
 
+    struct CPUWatchpoint *cpu_watchpoint[16];
+
     CPU_COMMON
 
     /* These fields after the common ones so they are preserved on reset.  */
diff --git a/target-arm/helper.c b/target-arm/helper.c
index 2b95f33872..ece967397f 100644
--- a/target-arm/helper.c
+++ b/target-arm/helper.c
@@ -304,17 +304,6 @@ void init_cpreg_list(ARMCPU *cpu)
     g_list_free(keys);
 }
 
-/* Return true if extended addresses are enabled.
- * This is always the case if our translation regime is 64 bit,
- * but depends on TTBCR.EAE for 32 bit.
- */
-static inline bool extended_addresses_enabled(CPUARMState *env)
-{
-    return arm_el_is_aa64(env, 1)
-        || ((arm_feature(env, ARM_FEATURE_LPAE)
-             && (env->cp15.c2_control & TTBCR_EAE)));
-}
-
 static void dacr_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value)
 {
     ARMCPU *cpu = arm_env_get_cpu(env);
@@ -388,6 +377,47 @@ static void tlbimvaa_write(CPUARMState *env, const ARMCPRegInfo *ri,
     tlb_flush_page(CPU(cpu), value & TARGET_PAGE_MASK);
 }
 
+/* IS variants of TLB operations must affect all cores */
+static void tlbiall_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                             uint64_t value)
+{
+    CPUState *other_cs;
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush(other_cs, 1);
+    }
+}
+
+static void tlbiasid_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                             uint64_t value)
+{
+    CPUState *other_cs;
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush(other_cs, value == 0);
+    }
+}
+
+static void tlbimva_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                             uint64_t value)
+{
+    CPUState *other_cs;
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush_page(other_cs, value & TARGET_PAGE_MASK);
+    }
+}
+
+static void tlbimvaa_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                             uint64_t value)
+{
+    CPUState *other_cs;
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush_page(other_cs, value & TARGET_PAGE_MASK);
+    }
+}
+
 static const ARMCPRegInfo cp_reginfo[] = {
     { .name = "FCSEIDR", .cp = 15, .crn = 13, .crm = 0, .opc1 = 0, .opc2 = 0,
       .access = PL1_RW, .fieldoffset = offsetof(CPUARMState, cp15.c13_fcse),
@@ -414,21 +444,6 @@ static const ARMCPRegInfo not_v8_cp_reginfo[] = {
      */
     { .name = "TLB_LOCKDOWN", .cp = 15, .crn = 10, .crm = CP_ANY,
       .opc1 = CP_ANY, .opc2 = CP_ANY, .access = PL1_RW, .type = ARM_CP_NOP },
-    /* MMU TLB control. Note that the wildcarding means we cover not just
-     * the unified TLB ops but also the dside/iside/inner-shareable variants.
-     */
-    { .name = "TLBIALL", .cp = 15, .crn = 8, .crm = CP_ANY,
-      .opc1 = CP_ANY, .opc2 = 0, .access = PL1_W, .writefn = tlbiall_write,
-      .type = ARM_CP_NO_MIGRATE },
-    { .name = "TLBIMVA", .cp = 15, .crn = 8, .crm = CP_ANY,
-      .opc1 = CP_ANY, .opc2 = 1, .access = PL1_W, .writefn = tlbimva_write,
-      .type = ARM_CP_NO_MIGRATE },
-    { .name = "TLBIASID", .cp = 15, .crn = 8, .crm = CP_ANY,
-      .opc1 = CP_ANY, .opc2 = 2, .access = PL1_W, .writefn = tlbiasid_write,
-      .type = ARM_CP_NO_MIGRATE },
-    { .name = "TLBIMVAA", .cp = 15, .crn = 8, .crm = CP_ANY,
-      .opc1 = CP_ANY, .opc2 = 3, .access = PL1_W, .writefn = tlbimvaa_write,
-      .type = ARM_CP_NO_MIGRATE },
     /* Cache maintenance ops; some of this space may be overridden later. */
     { .name = "CACHEMAINT", .cp = 15, .crn = 7, .crm = CP_ANY,
       .opc1 = 0, .opc2 = CP_ANY, .access = PL1_W,
@@ -472,6 +487,21 @@ static const ARMCPRegInfo not_v7_cp_reginfo[] = {
      */
     { .name = "DBGDIDR", .cp = 14, .crn = 0, .crm = 0, .opc1 = 0, .opc2 = 0,
       .access = PL0_R, .type = ARM_CP_CONST, .resetvalue = 0 },
+    /* MMU TLB control. Note that the wildcarding means we cover not just
+     * the unified TLB ops but also the dside/iside/inner-shareable variants.
+     */
+    { .name = "TLBIALL", .cp = 15, .crn = 8, .crm = CP_ANY,
+      .opc1 = CP_ANY, .opc2 = 0, .access = PL1_W, .writefn = tlbiall_write,
+      .type = ARM_CP_NO_MIGRATE },
+    { .name = "TLBIMVA", .cp = 15, .crn = 8, .crm = CP_ANY,
+      .opc1 = CP_ANY, .opc2 = 1, .access = PL1_W, .writefn = tlbimva_write,
+      .type = ARM_CP_NO_MIGRATE },
+    { .name = "TLBIASID", .cp = 15, .crn = 8, .crm = CP_ANY,
+      .opc1 = CP_ANY, .opc2 = 2, .access = PL1_W, .writefn = tlbiasid_write,
+      .type = ARM_CP_NO_MIGRATE },
+    { .name = "TLBIMVAA", .cp = 15, .crn = 8, .crm = CP_ANY,
+      .opc1 = CP_ANY, .opc2 = 3, .access = PL1_W, .writefn = tlbimvaa_write,
+      .type = ARM_CP_NO_MIGRATE },
     REGINFO_SENTINEL
 };
 
@@ -890,6 +920,44 @@ static const ARMCPRegInfo v7_cp_reginfo[] = {
     { .name = "ISR_EL1", .state = ARM_CP_STATE_BOTH,
       .opc0 = 3, .opc1 = 0, .crn = 12, .crm = 1, .opc2 = 0,
       .type = ARM_CP_NO_MIGRATE, .access = PL1_R, .readfn = isr_read },
+    /* 32 bit ITLB invalidates */
+    { .name = "ITLBIALL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 5, .opc2 = 0,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
+    { .name = "ITLBIMVA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 5, .opc2 = 1,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
+    { .name = "ITLBIASID", .cp = 15, .opc1 = 0, .crn = 8, .crm = 5, .opc2 = 2,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
+    /* 32 bit DTLB invalidates */
+    { .name = "DTLBIALL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 6, .opc2 = 0,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
+    { .name = "DTLBIMVA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 6, .opc2 = 1,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
+    { .name = "DTLBIASID", .cp = 15, .opc1 = 0, .crn = 8, .crm = 6, .opc2 = 2,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
+    /* 32 bit TLB invalidates */
+    { .name = "TLBIALL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 0,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
+    { .name = "TLBIMVA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 1,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
+    { .name = "TLBIASID", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 2,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
+    { .name = "TLBIMVAA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 3,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimvaa_write },
+    REGINFO_SENTINEL
+};
+
+static const ARMCPRegInfo v7mp_cp_reginfo[] = {
+    /* 32 bit TLB invalidates, Inner Shareable */
+    { .name = "TLBIALLIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 0,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_is_write },
+    { .name = "TLBIMVAIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 1,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_is_write },
+    { .name = "TLBIASIDIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 2,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W,
+      .writefn = tlbiasid_is_write },
+    { .name = "TLBIMVAAIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 3,
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W,
+      .writefn = tlbimvaa_is_write },
     REGINFO_SENTINEL
 };
 
@@ -1879,6 +1947,39 @@ static void tlbi_aa64_asid_write(CPUARMState *env, const ARMCPRegInfo *ri,
     tlb_flush(CPU(cpu), asid == 0);
 }
 
+static void tlbi_aa64_va_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                                  uint64_t value)
+{
+    CPUState *other_cs;
+    uint64_t pageaddr = sextract64(value << 12, 0, 56);
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush_page(other_cs, pageaddr);
+    }
+}
+
+static void tlbi_aa64_vaa_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                                  uint64_t value)
+{
+    CPUState *other_cs;
+    uint64_t pageaddr = sextract64(value << 12, 0, 56);
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush_page(other_cs, pageaddr);
+    }
+}
+
+static void tlbi_aa64_asid_is_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                                  uint64_t value)
+{
+    CPUState *other_cs;
+    int asid = extract64(value, 48, 16);
+
+    CPU_FOREACH(other_cs) {
+        tlb_flush(other_cs, asid == 0);
+    }
+}
+
 static CPAccessResult aa64_zva_access(CPUARMState *env, const ARMCPRegInfo *ri)
 {
     /* We don't implement EL2, so the only control on DC ZVA is the
@@ -1996,27 +2097,27 @@ static const ARMCPRegInfo v8_cp_reginfo[] = {
     { .name = "TLBI_VMALLE1IS", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 0,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
-      .writefn = tlbiall_write },
+      .writefn = tlbiall_is_write },
     { .name = "TLBI_VAE1IS", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 1,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
-      .writefn = tlbi_aa64_va_write },
+      .writefn = tlbi_aa64_va_is_write },
     { .name = "TLBI_ASIDE1IS", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 2,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
-      .writefn = tlbi_aa64_asid_write },
+      .writefn = tlbi_aa64_asid_is_write },
     { .name = "TLBI_VAAE1IS", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 3,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
-      .writefn = tlbi_aa64_vaa_write },
+      .writefn = tlbi_aa64_vaa_is_write },
     { .name = "TLBI_VALE1IS", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 5,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
-      .writefn = tlbi_aa64_va_write },
+      .writefn = tlbi_aa64_va_is_write },
     { .name = "TLBI_VAALE1IS", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 7,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
-      .writefn = tlbi_aa64_vaa_write },
+      .writefn = tlbi_aa64_vaa_is_write },
     { .name = "TLBI_VMALLE1", .state = ARM_CP_STATE_AA64,
       .opc0 = 1, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 0,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE,
@@ -2056,42 +2157,12 @@ static const ARMCPRegInfo v8_cp_reginfo[] = {
       .opc0 = 1, .opc1 = 0, .crn = 7, .crm = 8, .opc2 = 3,
       .access = PL1_W, .type = ARM_CP_NO_MIGRATE, .writefn = ats_write },
 #endif
-    /* 32 bit TLB invalidates, Inner Shareable */
-    { .name = "TLBIALLIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 0,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
-    { .name = "TLBIMVAIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 1,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
-    { .name = "TLBIASIDIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 2,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
-    { .name = "TLBIMVAAIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 3,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimvaa_write },
+    /* TLB invalidate last level of translation table walk */
     { .name = "TLBIMVALIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 5,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_is_write },
     { .name = "TLBIMVAALIS", .cp = 15, .opc1 = 0, .crn = 8, .crm = 3, .opc2 = 7,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimvaa_write },
-    /* 32 bit ITLB invalidates */
-    { .name = "ITLBIALL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 5, .opc2 = 0,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
-    { .name = "ITLBIMVA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 5, .opc2 = 1,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
-    { .name = "ITLBIASID", .cp = 15, .opc1 = 0, .crn = 8, .crm = 5, .opc2 = 2,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
-    /* 32 bit DTLB invalidates */
-    { .name = "DTLBIALL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 6, .opc2 = 0,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
-    { .name = "DTLBIMVA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 6, .opc2 = 1,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
-    { .name = "DTLBIASID", .cp = 15, .opc1 = 0, .crn = 8, .crm = 6, .opc2 = 2,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
-    /* 32 bit TLB invalidates */
-    { .name = "TLBIALL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 0,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiall_write },
-    { .name = "TLBIMVA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 1,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
-    { .name = "TLBIASID", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 2,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbiasid_write },
-    { .name = "TLBIMVAA", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 3,
-      .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimvaa_write },
+      .type = ARM_CP_NO_MIGRATE, .access = PL1_W,
+      .writefn = tlbimvaa_is_write },
     { .name = "TLBIMVAL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 5,
       .type = ARM_CP_NO_MIGRATE, .access = PL1_W, .writefn = tlbimva_write },
     { .name = "TLBIMVAAL", .cp = 15, .opc1 = 0, .crn = 8, .crm = 7, .opc2 = 7,
@@ -2255,18 +2326,35 @@ static const ARMCPRegInfo debug_cp_reginfo[] = {
       .access = PL1_R, .type = ARM_CP_CONST, .resetvalue = 0 },
     { .name = "DBGDSAR", .cp = 14, .crn = 2, .crm = 0, .opc1 = 0, .opc2 = 0,
       .access = PL0_R, .type = ARM_CP_CONST, .resetvalue = 0 },
-    /* Dummy implementation of monitor debug system control register:
-     * we don't support debug. (The 32-bit alias is DBGDSCRext.)
-     */
+    /* Monitor debug system control register; the 32-bit alias is DBGDSCRext. */
     { .name = "MDSCR_EL1", .state = ARM_CP_STATE_BOTH,
       .cp = 14, .opc0 = 2, .opc1 = 0, .crn = 0, .crm = 2, .opc2 = 2,
       .access = PL1_RW,
       .fieldoffset = offsetof(CPUARMState, cp15.mdscr_el1),
       .resetvalue = 0 },
+    /* MDCCSR_EL0, aka DBGDSCRint. This is a read-only mirror of MDSCR_EL1.
+     * We don't implement the configurable EL0 access.
+     */
+    { .name = "MDCCSR_EL0", .state = ARM_CP_STATE_BOTH,
+      .cp = 14, .opc0 = 2, .opc1 = 0, .crn = 0, .crm = 1, .opc2 = 0,
+      .type = ARM_CP_NO_MIGRATE,
+      .access = PL1_R,
+      .fieldoffset = offsetof(CPUARMState, cp15.mdscr_el1),
+      .resetfn = arm_cp_reset_ignore },
     /* We define a dummy WI OSLAR_EL1, because Linux writes to it. */
     { .name = "OSLAR_EL1", .state = ARM_CP_STATE_BOTH,
       .cp = 14, .opc0 = 2, .opc1 = 0, .crn = 1, .crm = 0, .opc2 = 4,
       .access = PL1_W, .type = ARM_CP_NOP },
+    /* Dummy OSDLR_EL1: 32-bit Linux will read this */
+    { .name = "OSDLR_EL1", .state = ARM_CP_STATE_BOTH,
+      .cp = 14, .opc0 = 2, .opc1 = 0, .crn = 1, .crm = 3, .opc2 = 4,
+      .access = PL1_RW, .type = ARM_CP_NOP },
+    /* Dummy DBGVCR: Linux wants to clear this on startup, but we don't
+     * implement vector catch debug events yet.
+     */
+    { .name = "DBGVCR",
+      .cp = 14, .opc1 = 0, .crn = 0, .crm = 7, .opc2 = 0,
+      .access = PL1_RW, .type = ARM_CP_NOP },
     REGINFO_SENTINEL
 };
 
@@ -2279,20 +2367,149 @@ static const ARMCPRegInfo debug_lpae_cp_reginfo[] = {
     REGINFO_SENTINEL
 };
 
+void hw_watchpoint_update(ARMCPU *cpu, int n)
+{
+    CPUARMState *env = &cpu->env;
+    vaddr len = 0;
+    vaddr wvr = env->cp15.dbgwvr[n];
+    uint64_t wcr = env->cp15.dbgwcr[n];
+    int mask;
+    int flags = BP_CPU | BP_STOP_BEFORE_ACCESS;
+
+    if (env->cpu_watchpoint[n]) {
+        cpu_watchpoint_remove_by_ref(CPU(cpu), env->cpu_watchpoint[n]);
+        env->cpu_watchpoint[n] = NULL;
+    }
+
+    if (!extract64(wcr, 0, 1)) {
+        /* E bit clear : watchpoint disabled */
+        return;
+    }
+
+    switch (extract64(wcr, 3, 2)) {
+    case 0:
+        /* LSC 00 is reserved and must behave as if the wp is disabled */
+        return;
+    case 1:
+        flags |= BP_MEM_READ;
+        break;
+    case 2:
+        flags |= BP_MEM_WRITE;
+        break;
+    case 3:
+        flags |= BP_MEM_ACCESS;
+        break;
+    }
+
+    /* Attempts to use both MASK and BAS fields simultaneously are
+     * CONSTRAINED UNPREDICTABLE; we opt to ignore BAS in this case,
+     * thus generating a watchpoint for every byte in the masked region.
+     */
+    mask = extract64(wcr, 24, 4);
+    if (mask == 1 || mask == 2) {
+        /* Reserved values of MASK; we must act as if the mask value was
+         * some non-reserved value, or as if the watchpoint were disabled.
+         * We choose the latter.
+         */
+        return;
+    } else if (mask) {
+        /* Watchpoint covers an aligned area up to 2GB in size */
+        len = 1ULL << mask;
+        /* If masked bits in WVR are not zero it's CONSTRAINED UNPREDICTABLE
+         * whether the watchpoint fires when the unmasked bits match; we opt
+         * to generate the exceptions.
+         */
+        wvr &= ~(len - 1);
+    } else {
+        /* Watchpoint covers bytes defined by the byte address select bits */
+        int bas = extract64(wcr, 5, 8);
+        int basstart;
+
+        if (bas == 0) {
+            /* This must act as if the watchpoint is disabled */
+            return;
+        }
+
+        if (extract64(wvr, 2, 1)) {
+            /* Deprecated case of an only 4-aligned address. BAS[7:4] are
+             * ignored, and BAS[3:0] define which bytes to watch.
+             */
+            bas &= 0xf;
+        }
+        /* The BAS bits are supposed to be programmed to indicate a contiguous
+         * range of bytes. Otherwise it is CONSTRAINED UNPREDICTABLE whether
+         * we fire for each byte in the word/doubleword addressed by the WVR.
+         * We choose to ignore any non-zero bits after the first range of 1s.
+         */
+        basstart = ctz32(bas);
+        len = cto32(bas >> basstart);
+        wvr += basstart;
+    }
+
+    cpu_watchpoint_insert(CPU(cpu), wvr, len, flags,
+                          &env->cpu_watchpoint[n]);
+}
+
+void hw_watchpoint_update_all(ARMCPU *cpu)
+{
+    int i;
+    CPUARMState *env = &cpu->env;
+
+    /* Completely clear out existing QEMU watchpoints and our array, to
+     * avoid possible stale entries following migration load.
+     */
+    cpu_watchpoint_remove_all(CPU(cpu), BP_CPU);
+    memset(env->cpu_watchpoint, 0, sizeof(env->cpu_watchpoint));
+
+    for (i = 0; i < ARRAY_SIZE(cpu->env.cpu_watchpoint); i++) {
+        hw_watchpoint_update(cpu, i);
+    }
+}
+
+static void dbgwvr_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                         uint64_t value)
+{
+    ARMCPU *cpu = arm_env_get_cpu(env);
+    int i = ri->crm;
+
+    /* Bits [63:49] are hardwired to the value of bit [48]; that is, the
+     * register reads and behaves as if values written are sign extended.
+     * Bits [1:0] are RES0.
+     */
+    value = sextract64(value, 0, 49) & ~3ULL;
+
+    raw_write(env, ri, value);
+    hw_watchpoint_update(cpu, i);
+}
+
+static void dbgwcr_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                         uint64_t value)
+{
+    ARMCPU *cpu = arm_env_get_cpu(env);
+    int i = ri->crm;
+
+    raw_write(env, ri, value);
+    hw_watchpoint_update(cpu, i);
+}
+
 static void define_debug_regs(ARMCPU *cpu)
 {
     /* Define v7 and v8 architectural debug registers.
      * These are just dummy implementations for now.
      */
     int i;
-    int wrps, brps;
+    int wrps, brps, ctx_cmps;
     ARMCPRegInfo dbgdidr = {
         .name = "DBGDIDR", .cp = 14, .crn = 0, .crm = 0, .opc1 = 0, .opc2 = 0,
         .access = PL0_R, .type = ARM_CP_CONST, .resetvalue = cpu->dbgdidr,
     };
 
+    /* Note that all these register fields hold "number of Xs minus 1". */
     brps = extract32(cpu->dbgdidr, 24, 4);
     wrps = extract32(cpu->dbgdidr, 28, 4);
+    ctx_cmps = extract32(cpu->dbgdidr, 20, 4);
+
+    assert(ctx_cmps <= brps);
 
     /* The DBGDIDR and ID_AA64DFR0_EL1 define various properties
      * of the debug registers such as number of breakpoints;
@@ -2301,6 +2518,7 @@ static void define_debug_regs(ARMCPU *cpu)
     if (arm_feature(&cpu->env, ARM_FEATURE_AARCH64)) {
         assert(extract32(cpu->id_aa64dfr0, 12, 4) == brps);
         assert(extract32(cpu->id_aa64dfr0, 20, 4) == wrps);
+        assert(extract32(cpu->id_aa64dfr0, 28, 4) == ctx_cmps);
     }
 
     define_one_arm_cp_reg(cpu, &dbgdidr);
@@ -2330,12 +2548,16 @@ static void define_debug_regs(ARMCPU *cpu)
             { .name = "DBGWVR", .state = ARM_CP_STATE_BOTH,
               .cp = 14, .opc0 = 2, .opc1 = 0, .crn = 0, .crm = i, .opc2 = 6,
               .access = PL1_RW,
-              .fieldoffset = offsetof(CPUARMState, cp15.dbgwvr[i]) },
+              .fieldoffset = offsetof(CPUARMState, cp15.dbgwvr[i]),
+              .writefn = dbgwvr_write, .raw_writefn = raw_write
+            },
             { .name = "DBGWCR", .state = ARM_CP_STATE_BOTH,
               .cp = 14, .opc0 = 2, .opc1 = 0, .crn = 0, .crm = i, .opc2 = 7,
               .access = PL1_RW,
-              .fieldoffset = offsetof(CPUARMState, cp15.dbgwcr[i]) },
-               REGINFO_SENTINEL
+              .fieldoffset = offsetof(CPUARMState, cp15.dbgwcr[i]),
+              .writefn = dbgwcr_write, .raw_writefn = raw_write
+            },
+            REGINFO_SENTINEL
         };
         define_arm_cp_regs(cpu, dbgregs);
     }
@@ -2434,6 +2656,9 @@ void register_cp_regs_for_features(ARMCPU *cpu)
     if (arm_feature(env, ARM_FEATURE_V6K)) {
         define_arm_cp_regs(cpu, v6k_cp_reginfo);
     }
+    if (arm_feature(env, ARM_FEATURE_V7MP)) {
+        define_arm_cp_regs(cpu, v7mp_cp_reginfo);
+    }
     if (arm_feature(env, ARM_FEATURE_V7)) {
         /* v7 performance monitor control register: same implementor
          * field as main ID register, and we implement only the cycle
@@ -3506,11 +3731,37 @@ void arm_cpu_do_interrupt(CPUState *cs)
     uint32_t mask;
     int new_mode;
     uint32_t offset;
+    uint32_t moe;
 
     assert(!IS_M(env));
 
     arm_log_exception(cs->exception_index);
 
+    /* If this is a debug exception we must update the DBGDSCR.MOE bits */
+    switch (env->exception.syndrome >> ARM_EL_EC_SHIFT) {
+    case EC_BREAKPOINT:
+    case EC_BREAKPOINT_SAME_EL:
+        moe = 1;
+        break;
+    case EC_WATCHPOINT:
+    case EC_WATCHPOINT_SAME_EL:
+        moe = 10;
+        break;
+    case EC_AA32_BKPT:
+        moe = 3;
+        break;
+    case EC_VECTORCATCH:
+        moe = 5;
+        break;
+    default:
+        moe = 0;
+        break;
+    }
+
+    if (moe) {
+        env->cp15.mdscr_el1 = deposit64(env->cp15.mdscr_el1, 2, 4, moe);
+    }
+
     /* TODO: Vectored interrupt controller.  */
     switch (cs->exception_index) {
     case EXCP_UDEF:
diff --git a/target-arm/internals.h b/target-arm/internals.h
index 53c2e3cf3e..64751a0798 100644
--- a/target-arm/internals.h
+++ b/target-arm/internals.h
@@ -142,6 +142,17 @@ static inline void update_spsel(CPUARMState *env, uint32_t imm)
     aarch64_restore_sp(env, cur_el);
 }
 
+/* Return true if extended addresses are enabled.
+ * This is always the case if our translation regime is 64 bit,
+ * but depends on TTBCR.EAE for 32 bit.
+ */
+static inline bool extended_addresses_enabled(CPUARMState *env)
+{
+    return arm_el_is_aa64(env, 1)
+        || ((arm_feature(env, ARM_FEATURE_LPAE)
+             && (env->cp15.c2_control & TTBCR_EAE)));
+}
+
 /* Valid Syndrome Register EC field values */
 enum arm_exception_class {
     EC_UNCATEGORIZED          = 0x00,
@@ -296,4 +307,23 @@ static inline uint32_t syn_swstep(int same_el, int isv, int ex)
         | (isv << 24) | (ex << 6) | 0x22;
 }
 
+static inline uint32_t syn_watchpoint(int same_el, int cm, int wnr)
+{
+    return (EC_WATCHPOINT << ARM_EL_EC_SHIFT) | (same_el << ARM_EL_EC_SHIFT)
+        | (cm << 8) | (wnr << 6) | 0x22;
+}
+
+/* Update a QEMU watchpoint based on the information the guest has set in the
+ * DBGWCR<n>_EL1 and DBGWVR<n>_EL1 registers.
+ */
+void hw_watchpoint_update(ARMCPU *cpu, int n);
+/* Update the QEMU watchpoints for every guest watchpoint. This does a
+ * complete delete-and-reinstate of the QEMU watchpoint list and so is
+ * suitable for use after migration or on reset.
+ */
+void hw_watchpoint_update_all(ARMCPU *cpu);
+
+/* Callback function for when a watchpoint or breakpoint triggers. */
+void arm_debug_excp_handler(CPUState *cs);
+
 #endif
diff --git a/target-arm/machine.c b/target-arm/machine.c
index 3bcc7cc833..8dfe87cb6b 100644
--- a/target-arm/machine.c
+++ b/target-arm/machine.c
@@ -2,6 +2,7 @@
 #include "hw/boards.h"
 #include "sysemu/kvm.h"
 #include "kvm_arm.h"
+#include "internals.h"
 
 static bool vfp_needed(void *opaque)
 {
@@ -213,6 +214,8 @@ static int cpu_post_load(void *opaque, int version_id)
         }
     }
 
+    hw_watchpoint_update_all(cpu);
+
     return 0;
 }
 
diff --git a/target-arm/op_helper.c b/target-arm/op_helper.c
index fe40358c96..b956216c4b 100644
--- a/target-arm/op_helper.c
+++ b/target-arm/op_helper.c
@@ -456,6 +456,194 @@ illegal_return:
     }
 }
 
+/* Return true if the linked breakpoint entry lbn passes its checks */
+static bool linked_bp_matches(ARMCPU *cpu, int lbn)
+{
+    CPUARMState *env = &cpu->env;
+    uint64_t bcr = env->cp15.dbgbcr[lbn];
+    int brps = extract32(cpu->dbgdidr, 24, 4);
+    int ctx_cmps = extract32(cpu->dbgdidr, 20, 4);
+    int bt;
+    uint32_t contextidr;
+
+    /* Links to unimplemented or non-context aware breakpoints are
+     * CONSTRAINED UNPREDICTABLE: either behave as if disabled, or
+     * as if linked to an UNKNOWN context-aware breakpoint (in which
+     * case DBGWCR<n>_EL1.LBN must indicate that breakpoint).
+     * We choose the former.
+     */
+    if (lbn > brps || lbn < (brps - ctx_cmps)) {
+        return false;
+    }
+
+    bcr = env->cp15.dbgbcr[lbn];
+
+    if (extract64(bcr, 0, 1) == 0) {
+        /* Linked breakpoint disabled : generate no events */
+        return false;
+    }
+
+    bt = extract64(bcr, 20, 4);
+
+    /* We match the whole register even if this is AArch32 using the
+     * short descriptor format (in which case it holds both PROCID and ASID),
+     * since we don't implement the optional v7 context ID masking.
+     */
+    contextidr = extract64(env->cp15.contextidr_el1, 0, 32);
+
+    switch (bt) {
+    case 3: /* linked context ID match */
+        if (arm_current_pl(env) > 1) {
+            /* Context matches never fire in EL2 or (AArch64) EL3 */
+            return false;
+        }
+        return (contextidr == extract64(env->cp15.dbgbvr[lbn], 0, 32));
+    case 5: /* linked address mismatch (reserved in AArch64) */
+    case 9: /* linked VMID match (reserved if no EL2) */
+    case 11: /* linked context ID and VMID match (reserved if no EL2) */
+    default:
+        /* Links to Unlinked context breakpoints must generate no
+         * events; we choose to do the same for reserved values too.
+         */
+        return false;
+    }
+
+    return false;
+}
+
+static bool wp_matches(ARMCPU *cpu, int n)
+{
+    CPUARMState *env = &cpu->env;
+    uint64_t wcr = env->cp15.dbgwcr[n];
+    int pac, hmc, ssc, wt, lbn;
+    /* TODO: check against CPU security state when we implement TrustZone */
+    bool is_secure = false;
+
+    if (!env->cpu_watchpoint[n]
+        || !(env->cpu_watchpoint[n]->flags & BP_WATCHPOINT_HIT)) {
+        return false;
+    }
+
+    /* The WATCHPOINT_HIT flag guarantees us that the watchpoint is
+     * enabled and that the address and access type match; check the
+     * remaining fields, including linked breakpoints.
+     * Note that some combinations of {PAC, HMC SSC} are reserved and
+     * must act either like some valid combination or as if the watchpoint
+     * were disabled. We choose the former, and use this together with
+     * the fact that EL3 must always be Secure and EL2 must always be
+     * Non-Secure to simplify the code slightly compared to the full
+     * table in the ARM ARM.
+     */
+    pac = extract64(wcr, 1, 2);
+    hmc = extract64(wcr, 13, 1);
+    ssc = extract64(wcr, 14, 2);
+
+    switch (ssc) {
+    case 0:
+        break;
+    case 1:
+    case 3:
+        if (is_secure) {
+            return false;
+        }
+        break;
+    case 2:
+        if (!is_secure) {
+            return false;
+        }
+        break;
+    }
+
+    /* TODO: this is not strictly correct because the LDRT/STRT/LDT/STT
+     * "unprivileged access" instructions should match watchpoints as if
+     * they were accesses done at EL0, even if the CPU is at EL1 or higher.
+     * Implementing this would require reworking the core watchpoint code
+     * to plumb the mmu_idx through to this point. Luckily Linux does not
+     * rely on this behaviour currently.
+     */
+    switch (arm_current_pl(env)) {
+    case 3:
+    case 2:
+        if (!hmc) {
+            return false;
+        }
+        break;
+    case 1:
+        if (extract32(pac, 0, 1) == 0) {
+            return false;
+        }
+        break;
+    case 0:
+        if (extract32(pac, 1, 1) == 0) {
+            return false;
+        }
+        break;
+    default:
+        g_assert_not_reached();
+    }
+
+    wt = extract64(wcr, 20, 1);
+    lbn = extract64(wcr, 16, 4);
+
+    if (wt && !linked_bp_matches(cpu, lbn)) {
+        return false;
+    }
+
+    return true;
+}
+
+static bool check_watchpoints(ARMCPU *cpu)
+{
+    CPUARMState *env = &cpu->env;
+    int n;
+
+    /* If watchpoints are disabled globally or we can't take debug
+     * exceptions here then watchpoint firings are ignored.
+     */
+    if (extract32(env->cp15.mdscr_el1, 15, 1) == 0
+        || !arm_generate_debug_exceptions(env)) {
+        return false;
+    }
+
+    for (n = 0; n < ARRAY_SIZE(env->cpu_watchpoint); n++) {
+        if (wp_matches(cpu, n)) {
+            return true;
+        }
+    }
+    return false;
+}
+
+void arm_debug_excp_handler(CPUState *cs)
+{
+    /* Called by core code when a watchpoint or breakpoint fires;
+     * need to check which one and raise the appropriate exception.
+     */
+    ARMCPU *cpu = ARM_CPU(cs);
+    CPUARMState *env = &cpu->env;
+    CPUWatchpoint *wp_hit = cs->watchpoint_hit;
+
+    if (wp_hit) {
+        if (wp_hit->flags & BP_CPU) {
+            cs->watchpoint_hit = NULL;
+            if (check_watchpoints(cpu)) {
+                bool wnr = (wp_hit->flags & BP_WATCHPOINT_HIT_WRITE) != 0;
+                bool same_el = arm_debug_target_el(env) == arm_current_pl(env);
+
+                env->exception.syndrome = syn_watchpoint(same_el, 0, wnr);
+                if (extended_addresses_enabled(env)) {
+                    env->exception.fsr = (1 << 9) | 0x22;
+                } else {
+                    env->exception.fsr = 0x2;
+                }
+                env->exception.vaddress = wp_hit->hitaddr;
+                raise_exception(env, EXCP_DATA_ABORT);
+            } else {
+                cpu_resume_from_signal(cs, NULL);
+            }
+        }
+    }
+}
+
 /* ??? Flag setting arithmetic is awkward because we need to do comparisons.
    The only way to do that in TCG is a conditional branch, which clobbers
    all our temporaries.  For now implement these as helper functions.  */