diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c
index 4ebe25461a..293ee86ea4 100644
--- a/accel/tcg/user-exec.c
+++ b/accel/tcg/user-exec.c
@@ -49,7 +49,8 @@ __thread uintptr_t helper_retaddr;
 /* exit the current TB from a signal handler. The host registers are
    restored in a state compatible with the CPU emulator
  */
-static void cpu_exit_tb_from_sighandler(CPUState *cpu, sigset_t *old_set)
+static void QEMU_NORETURN cpu_exit_tb_from_sighandler(CPUState *cpu,
+                                                      sigset_t *old_set)
 {
     /* XXX: use siglongjmp ? */
     sigprocmask(SIG_SETMASK, old_set, NULL);
diff --git a/bsd-user/main.c b/bsd-user/main.c
index 0a918e8f74..9c700c6234 100644
--- a/bsd-user/main.c
+++ b/bsd-user/main.c
@@ -512,6 +512,7 @@ void cpu_loop(CPUSPARCState *env)
         case 0x141:
             if (bsd_type != target_freebsd)
                 goto badtrap;
+            /* fallthrough */
         case 0x100:
 #endif
             syscall_nr = env->gregs[1];
diff --git a/configure b/configure
index c228f7c21e..881af4b6be 100755
--- a/configure
+++ b/configure
@@ -2023,6 +2023,7 @@ add_to warn_flags -Wempty-body
 add_to warn_flags -Wnested-externs
 add_to warn_flags -Wendif-labels
 add_to warn_flags -Wexpansion-to-defined
+add_to warn_flags -Wimplicit-fallthrough=2
 
 nowarn_flags=
 add_to nowarn_flags -Wno-initializer-overrides
diff --git a/disas/libvixl/vixl/a64/disasm-a64.cc b/disas/libvixl/vixl/a64/disasm-a64.cc
index 7a58a5c087..f34d1d68da 100644
--- a/disas/libvixl/vixl/a64/disasm-a64.cc
+++ b/disas/libvixl/vixl/a64/disasm-a64.cc
@@ -2985,6 +2985,10 @@ int Disassembler::SubstituteImmediateField(const Instruction* instr,
           }
           return 3;
         }
+        default: {
+          VIXL_UNIMPLEMENTED();
+          return 0;
+        }
       }
     }
     case 'C': {  // ICondB - Immediate Conditional Branch.
diff --git a/disas/libvixl/vixl/globals.h b/disas/libvixl/vixl/globals.h
index 61dc9f7f7e..7099aa599f 100644
--- a/disas/libvixl/vixl/globals.h
+++ b/disas/libvixl/vixl/globals.h
@@ -108,10 +108,12 @@ inline void USE(T1, T2, T3, T4) {}
   #define __has_warning(x)  0
 #endif
 
-// Note: This option is only available for Clang. And will only be enabled for
-// C++11(201103L).
+// Fallthrough annotation for Clang and C++11(201103L).
 #if __has_warning("-Wimplicit-fallthrough") && __cplusplus >= 201103L
   #define VIXL_FALLTHROUGH() [[clang::fallthrough]] //NOLINT
+// Fallthrough annotation for GCC >= 7.
+#elif __GNUC__ >= 7
+  #define VIXL_FALLTHROUGH() __attribute__((fallthrough))
 #else
   #define VIXL_FALLTHROUGH() do {} while (0)
 #endif
diff --git a/hw/intc/arm_gicv3_kvm.c b/hw/intc/arm_gicv3_kvm.c
index 187eb054e0..d040a5d1e9 100644
--- a/hw/intc/arm_gicv3_kvm.c
+++ b/hw/intc/arm_gicv3_kvm.c
@@ -478,9 +478,11 @@ static void kvm_arm_gicv3_put(GICv3State *s)
             kvm_gicc_access(s, ICC_AP0R_EL1(3), ncpu, &reg64, true);
             reg64 = c->icc_apr[GICV3_G0][2];
             kvm_gicc_access(s, ICC_AP0R_EL1(2), ncpu, &reg64, true);
+            /* fall through */
         case 6:
             reg64 = c->icc_apr[GICV3_G0][1];
             kvm_gicc_access(s, ICC_AP0R_EL1(1), ncpu, &reg64, true);
+            /* fall through */
         default:
             reg64 = c->icc_apr[GICV3_G0][0];
             kvm_gicc_access(s, ICC_AP0R_EL1(0), ncpu, &reg64, true);
@@ -492,9 +494,11 @@ static void kvm_arm_gicv3_put(GICv3State *s)
             kvm_gicc_access(s, ICC_AP1R_EL1(3), ncpu, &reg64, true);
             reg64 = c->icc_apr[GICV3_G1NS][2];
             kvm_gicc_access(s, ICC_AP1R_EL1(2), ncpu, &reg64, true);
+            /* fall through */
         case 6:
             reg64 = c->icc_apr[GICV3_G1NS][1];
             kvm_gicc_access(s, ICC_AP1R_EL1(1), ncpu, &reg64, true);
+            /* fall through */
         default:
             reg64 = c->icc_apr[GICV3_G1NS][0];
             kvm_gicc_access(s, ICC_AP1R_EL1(0), ncpu, &reg64, true);
@@ -631,9 +635,11 @@ static void kvm_arm_gicv3_get(GICv3State *s)
             c->icc_apr[GICV3_G0][3] = reg64;
             kvm_gicc_access(s, ICC_AP0R_EL1(2), ncpu, &reg64, false);
             c->icc_apr[GICV3_G0][2] = reg64;
+            /* fall through */
         case 6:
             kvm_gicc_access(s, ICC_AP0R_EL1(1), ncpu, &reg64, false);
             c->icc_apr[GICV3_G0][1] = reg64;
+            /* fall through */
         default:
             kvm_gicc_access(s, ICC_AP0R_EL1(0), ncpu, &reg64, false);
             c->icc_apr[GICV3_G0][0] = reg64;
@@ -645,9 +651,11 @@ static void kvm_arm_gicv3_get(GICv3State *s)
             c->icc_apr[GICV3_G1NS][3] = reg64;
             kvm_gicc_access(s, ICC_AP1R_EL1(2), ncpu, &reg64, false);
             c->icc_apr[GICV3_G1NS][2] = reg64;
+            /* fall through */
         case 6:
             kvm_gicc_access(s, ICC_AP1R_EL1(1), ncpu, &reg64, false);
             c->icc_apr[GICV3_G1NS][1] = reg64;
+            /* fall through */
         default:
             kvm_gicc_access(s, ICC_AP1R_EL1(0), ncpu, &reg64, false);
             c->icc_apr[GICV3_G1NS][0] = reg64;
diff --git a/hw/rtc/twl92230.c b/hw/rtc/twl92230.c
index f838913b37..006d75e174 100644
--- a/hw/rtc/twl92230.c
+++ b/hw/rtc/twl92230.c
@@ -271,37 +271,23 @@ static void menelaus_gpio_set(void *opaque, int line, int level)
 static uint8_t menelaus_read(void *opaque, uint8_t addr)
 {
     MenelausState *s = (MenelausState *) opaque;
-    int reg = 0;
 
     switch (addr) {
     case MENELAUS_REV:
         return 0x22;
 
-    case MENELAUS_VCORE_CTRL5: reg ++;
-    case MENELAUS_VCORE_CTRL4: reg ++;
-    case MENELAUS_VCORE_CTRL3: reg ++;
-    case MENELAUS_VCORE_CTRL2: reg ++;
-    case MENELAUS_VCORE_CTRL1:
-        return s->vcore[reg];
+    case MENELAUS_VCORE_CTRL1 ... MENELAUS_VCORE_CTRL5:
+        return s->vcore[addr - MENELAUS_VCORE_CTRL1];
 
-    case MENELAUS_DCDC_CTRL3: reg ++;
-    case MENELAUS_DCDC_CTRL2: reg ++;
-    case MENELAUS_DCDC_CTRL1:
-        return s->dcdc[reg];
+    case MENELAUS_DCDC_CTRL1 ... MENELAUS_DCDC_CTRL3:
+        return s->dcdc[addr - MENELAUS_DCDC_CTRL1];
 
-    case MENELAUS_LDO_CTRL8: reg ++;
-    case MENELAUS_LDO_CTRL7: reg ++;
-    case MENELAUS_LDO_CTRL6: reg ++;
-    case MENELAUS_LDO_CTRL5: reg ++;
-    case MENELAUS_LDO_CTRL4: reg ++;
-    case MENELAUS_LDO_CTRL3: reg ++;
-    case MENELAUS_LDO_CTRL2: reg ++;
-    case MENELAUS_LDO_CTRL1:
-        return s->ldo[reg];
+    case MENELAUS_LDO_CTRL1 ... MENELAUS_LDO_CTRL8:
+        return s->ldo[addr - MENELAUS_LDO_CTRL1];
 
-    case MENELAUS_SLEEP_CTRL2: reg ++;
     case MENELAUS_SLEEP_CTRL1:
-        return s->sleep[reg];
+    case MENELAUS_SLEEP_CTRL2:
+        return s->sleep[addr - MENELAUS_SLEEP_CTRL1];
 
     case MENELAUS_DEVICE_OFF:
         return 0;
@@ -395,10 +381,8 @@ static uint8_t menelaus_read(void *opaque, uint8_t addr)
     case MENELAUS_S2_PULL_DIR:
         return s->pull[3];
 
-    case MENELAUS_MCT_CTRL3: reg ++;
-    case MENELAUS_MCT_CTRL2: reg ++;
-    case MENELAUS_MCT_CTRL1:
-        return s->mmc_ctrl[reg];
+    case MENELAUS_MCT_CTRL1 ... MENELAUS_MCT_CTRL3:
+        return s->mmc_ctrl[addr - MENELAUS_MCT_CTRL1];
     case MENELAUS_MCT_PIN_ST:
         /* TODO: return the real Card Detect */
         return 0;
@@ -418,7 +402,6 @@ static void menelaus_write(void *opaque, uint8_t addr, uint8_t value)
 {
     MenelausState *s = (MenelausState *) opaque;
     int line;
-    int reg = 0;
     struct tm tm;
 
     switch (addr) {
@@ -496,9 +479,9 @@ static void menelaus_write(void *opaque, uint8_t addr, uint8_t value)
         s->ldo[7] = value & 3;
         break;
 
-    case MENELAUS_SLEEP_CTRL2: reg ++;
     case MENELAUS_SLEEP_CTRL1:
-        s->sleep[reg] = value;
+    case MENELAUS_SLEEP_CTRL2:
+        s->sleep[addr - MENELAUS_SLEEP_CTRL1] = value;
         break;
 
     case MENELAUS_DEVICE_OFF:
@@ -714,6 +697,7 @@ static void menelaus_write(void *opaque, uint8_t addr, uint8_t value)
 #ifdef VERBOSE
         printf("%s: unknown register %02x\n", __func__, addr);
 #endif
+        break;
     }
 }
 
diff --git a/hw/timer/renesas_tmr.c b/hw/timer/renesas_tmr.c
index 446f2eacdd..e03a8155b2 100644
--- a/hw/timer/renesas_tmr.c
+++ b/hw/timer/renesas_tmr.c
@@ -221,6 +221,7 @@ static uint64_t tmr_read(void *opaque, hwaddr addr, unsigned size)
         } else if (ch == 0) {
             return concat_reg(tmr->tcora);
         }
+        /* fall through */
     case A_TCORB:
         if (size == 1) {
             return tmr->tcorb[ch];
diff --git a/include/qemu/compiler.h b/include/qemu/compiler.h
index 1b9e58e82b..df9ec08f8a 100644
--- a/include/qemu/compiler.h
+++ b/include/qemu/compiler.h
@@ -222,4 +222,15 @@ extern void QEMU_NORETURN QEMU_ERROR("code path is reachable")
 #define qemu_build_not_reached()  g_assert_not_reached()
 #endif
 
+/**
+ * In most cases, normal "fallthrough" comments are good enough for
+ * switch-case statements, but sometimes the compiler has problems
+ * with those. In that case you can use QEMU_FALLTHROUGH instead.
+ */
+#if __has_attribute(fallthrough)
+# define QEMU_FALLTHROUGH __attribute__((fallthrough))
+#else
+# define QEMU_FALLTHROUGH do {} while (0) /* fallthrough */
+#endif
+
 #endif /* COMPILER_H */
diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c
index 750f75c257..11db2f3c8d 100644
--- a/target/i386/tcg/translate.c
+++ b/target/i386/tcg/translate.c
@@ -1778,9 +1778,12 @@ static void gen_shiftd_rm_T1(DisasContext *s, MemOp ot, int op1,
         } else {
             tcg_gen_deposit_tl(s->T1, s->T0, s->T1, 16, 16);
         }
-        /* FALLTHRU */
-#ifdef TARGET_X86_64
+        /*
+         * If TARGET_X86_64 defined then fall through into MO_32 case,
+         * otherwise fall through default case.
+         */
     case MO_32:
+#ifdef TARGET_X86_64
         /* Concatenate the two 32-bit values and use a 64-bit shift.  */
         tcg_gen_subi_tl(s->tmp0, count, 1);
         if (is_right) {
diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index 30c73f8d2e..4bfa3179f8 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -2324,8 +2324,8 @@ static void gen_st_asi(DisasContext *dc, TCGv src, TCGv addr,
         }
         /* in OpenSPARC T1+ CPUs TWINX ASIs in store instructions
          * are ST_BLKINIT_ ASIs */
-        /* fall through */
 #endif
+        /* fall through */
     case GET_ASI_DIRECT:
         gen_address_mask(dc, addr);
         tcg_gen_qemu_st_tl(src, addr, da.mem_idx, da.memop);
diff --git a/target/sparc/win_helper.c b/target/sparc/win_helper.c
index 5b57892a10..3a7c0ff943 100644
--- a/target/sparc/win_helper.c
+++ b/target/sparc/win_helper.c
@@ -302,7 +302,7 @@ static inline uint64_t *get_gregset(CPUSPARCState *env, uint32_t pstate)
     switch (pstate) {
     default:
         trace_win_helper_gregset_error(pstate);
-        /* pass through to normal set of global registers */
+        /* fall through to normal set of global registers */
     case 0:
         return env->bgregs;
     case PS_AG:
diff --git a/target/unicore32/translate.c b/target/unicore32/translate.c
index d4b06df672..962f9877a0 100644
--- a/target/unicore32/translate.c
+++ b/target/unicore32/translate.c
@@ -1801,6 +1801,7 @@ static void disas_uc32_insn(CPUUniCore32State *env, DisasContext *s)
             do_misc(env, s, insn);
             break;
         }
+        /* fallthrough */
     case 0x1:
         if (((UCOP_OPCODES >> 2) == 2) && !UCOP_SET_S) {
             do_misc(env, s, insn);
@@ -1817,6 +1818,7 @@ static void disas_uc32_insn(CPUUniCore32State *env, DisasContext *s)
         if (UCOP_SET(8) || UCOP_SET(5)) {
             ILLEGAL;
         }
+        /* fallthrough */
     case 0x3:
         do_ldst_ir(env, s, insn);
         break;
diff --git a/tcg/optimize.c b/tcg/optimize.c
index 220f4601d5..7ca71de956 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -855,6 +855,7 @@ void tcg_optimize(TCGContext *s)
             if ((arg_info(op->args[1])->mask & 0x80) != 0) {
                 break;
             }
+            QEMU_FALLTHROUGH;
         CASE_OP_32_64(ext8u):
             mask = 0xff;
             goto and_const;
@@ -862,6 +863,7 @@ void tcg_optimize(TCGContext *s)
             if ((arg_info(op->args[1])->mask & 0x8000) != 0) {
                 break;
             }
+            QEMU_FALLTHROUGH;
         CASE_OP_32_64(ext16u):
             mask = 0xffff;
             goto and_const;
@@ -869,6 +871,7 @@ void tcg_optimize(TCGContext *s)
             if ((arg_info(op->args[1])->mask & 0x80000000) != 0) {
                 break;
             }
+            QEMU_FALLTHROUGH;
         case INDEX_op_ext32u_i64:
             mask = 0xffffffffU;
             goto and_const;
@@ -886,6 +889,7 @@ void tcg_optimize(TCGContext *s)
             if ((arg_info(op->args[1])->mask & 0x80000000) != 0) {
                 break;
             }
+            QEMU_FALLTHROUGH;
         case INDEX_op_extu_i32_i64:
             /* We do not compute affected as it is a size changing op.  */
             mask = (uint32_t)arg_info(op->args[1])->mask;
diff --git a/tests/fp/meson.build b/tests/fp/meson.build
index 3d4fb00f9d..8d739c4d59 100644
--- a/tests/fp/meson.build
+++ b/tests/fp/meson.build
@@ -27,6 +27,7 @@ tfdir = 'berkeley-testfloat-3/source'
 sfinc = include_directories(sfdir / 'include', sfspedir)
 
 tfcflags = [
+  '-Wno-implicit-fallthrough',
   '-Wno-strict-prototypes',
   '-Wno-unknown-pragmas',
   '-Wno-uninitialized',
@@ -209,6 +210,7 @@ libtestfloat = static_library(
 )
 
 sfcflags = [
+  '-Wno-implicit-fallthrough',
   '-Wno-missing-prototypes',
   '-Wno-redundant-decls',
   '-Wno-return-type',