Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20150528' into staging

A set of patches add support for vector registers on s390x.
Notable: Floating point registers and vector registers overlap,
so extra care is needed so that we end up with a consistent state
in all cases.

# gpg: Signature made Thu May 28 09:37:27 2015 BST using RSA key ID C6F02FAF
# gpg: Good signature from "Cornelia Huck <huckc@linux.vnet.ibm.com>"
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>"

* remotes/cohuck/tags/s390x-20150528:
  s390x: Enable vector processing capability
  s390x: Migrate vector registers
  s390x: Add vector registers to ELF dump
  linux/elf.h update
  s390x: Add vector registers to HMP output
  s390x: gdb updates for vector registers
  gdb-xml: Include XML for s390 vector registers
  s390x: Store Additional Status SIGP order
  s390x: Vector Register IOCTLs
  s390x: Common access to floating point registers

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

11 files changed, 357 insertions, 43 deletions

diff --git a/configure b/configure
index 9852aef..b707429 100755
--- a/configure
+++ b/configure
@@ -5292,7 +5292,7 @@ case "$target_name" in
     echo "TARGET_ABI32=y" >> $config_target_mak
   ;;
   s390x)
-    gdb_xml_files="s390x-core64.xml s390-acr.xml s390-fpr.xml"
+    gdb_xml_files="s390x-core64.xml s390-acr.xml s390-fpr.xml s390-vx.xml"
   ;;
   tricore)
   ;;
diff --git a/gdb-xml/s390-vx.xml b/gdb-xml/s390-vx.xml
new file mode 100644
index 0000000..8239c11
--- /dev/null
+++ b/gdb-xml/s390-vx.xml
@@ -0,0 +1,59 @@
+<?xml version="1.0"?>
+<!-- Copyright (C) 2010-2014 Free Software Foundation, Inc.
+
+     Copying and distribution of this file, with or without modification,
+     are permitted in any medium without royalty provided the copyright
+     notice and this notice are preserved.  -->
+
+<!DOCTYPE feature SYSTEM "gdb-target.dtd">
+<feature name="org.gnu.gdb.s390.vx">
+  <vector id="v4f" type="ieee_single" count="4"/>
+  <vector id="v2d" type="ieee_double" count="2"/>
+  <vector id="v16i8" type="int8" count="16"/>
+  <vector id="v8i16" type="int16" count="8"/>
+  <vector id="v4i32" type="int32" count="4"/>
+  <vector id="v2i64" type="int64" count="2"/>
+  <union id="vec128">
+    <field name="v4_float" type="v4f"/>
+    <field name="v2_double" type="v2d"/>
+    <field name="v16_int8" type="v16i8"/>
+    <field name="v8_int16" type="v8i16"/>
+    <field name="v4_int32" type="v4i32"/>
+    <field name="v2_int64" type="v2i64"/>
+    <field name="uint128" type="uint128"/>
+  </union>
+
+  <reg name="v0l" bitsize="64" type="uint64"/>
+  <reg name="v1l" bitsize="64" type="uint64"/>
+  <reg name="v2l" bitsize="64" type="uint64"/>
+  <reg name="v3l" bitsize="64" type="uint64"/>
+  <reg name="v4l" bitsize="64" type="uint64"/>
+  <reg name="v5l" bitsize="64" type="uint64"/>
+  <reg name="v6l" bitsize="64" type="uint64"/>
+  <reg name="v7l" bitsize="64" type="uint64"/>
+  <reg name="v8l" bitsize="64" type="uint64"/>
+  <reg name="v9l" bitsize="64" type="uint64"/>
+  <reg name="v10l" bitsize="64" type="uint64"/>
+  <reg name="v11l" bitsize="64" type="uint64"/>
+  <reg name="v12l" bitsize="64" type="uint64"/>
+  <reg name="v13l" bitsize="64" type="uint64"/>
+  <reg name="v14l" bitsize="64" type="uint64"/>
+  <reg name="v15l" bitsize="64" type="uint64"/>
+
+  <reg name="v16" bitsize="128" type="vec128"/>
+  <reg name="v17" bitsize="128" type="vec128"/>
+  <reg name="v18" bitsize="128" type="vec128"/>
+  <reg name="v19" bitsize="128" type="vec128"/>
+  <reg name="v20" bitsize="128" type="vec128"/>
+  <reg name="v21" bitsize="128" type="vec128"/>
+  <reg name="v22" bitsize="128" type="vec128"/>
+  <reg name="v23" bitsize="128" type="vec128"/>
+  <reg name="v24" bitsize="128" type="vec128"/>
+  <reg name="v25" bitsize="128" type="vec128"/>
+  <reg name="v26" bitsize="128" type="vec128"/>
+  <reg name="v27" bitsize="128" type="vec128"/>
+  <reg name="v28" bitsize="128" type="vec128"/>
+  <reg name="v29" bitsize="128" type="vec128"/>
+  <reg name="v30" bitsize="128" type="vec128"/>
+  <reg name="v31" bitsize="128" type="vec128"/>
+</feature>
diff --git a/include/elf.h b/include/elf.h
index 3e75f05..4afd474 100644
--- a/include/elf.h
+++ b/include/elf.h
@@ -1456,6 +1456,8 @@ typedef struct elf64_shdr {
 #define NT_TASKSTRUCT	4
 #define NT_AUXV		6
 #define NT_PRXFPREG     0x46e62b7f      /* copied from gdb5.1/include/elf/common.h */
+#define NT_S390_VXRS_HIGH 0x30a         /* s390 vector registers 16-31 */
+#define NT_S390_VXRS_LOW  0x309         /* s390 vector registers 0-15 (lower half) */
 #define NT_S390_PREFIX  0x305           /* s390 prefix register */
 #define NT_S390_CTRS    0x304           /* s390 control registers */
 #define NT_S390_TODPREG 0x303           /* s390 TOD programmable register */
diff --git a/linux-user/signal.c b/linux-user/signal.c
index 5bb399e..1166f2f 100644
--- a/linux-user/signal.c
+++ b/linux-user/signal.c
@@ -4098,7 +4098,7 @@ static void save_sigregs(CPUS390XState *env, target_sigregs *sregs)
      */
     //save_fp_regs(&current->thread.fp_regs); FIXME
     for (i = 0; i < 16; i++) {
-        __put_user(env->fregs[i].ll, &sregs->fpregs.fprs[i]);
+        __put_user(get_freg(env, i)->ll, &sregs->fpregs.fprs[i]);
     }
 }
 
@@ -4239,7 +4239,7 @@ restore_sigregs(CPUS390XState *env, target_sigregs *sc)
         __get_user(env->aregs[i], &sc->regs.acrs[i]);
     }
     for (i = 0; i < 16; i++) {
-        __get_user(env->fregs[i].ll, &sc->fpregs.fprs[i]);
+        __get_user(get_freg(env, i)->ll, &sc->fpregs.fprs[i]);
     }
 
     return err;
diff --git a/target-s390x/arch_dump.c b/target-s390x/arch_dump.c
index a1554f5..dab63eb 100644
--- a/target-s390x/arch_dump.c
+++ b/target-s390x/arch_dump.c
@@ -44,6 +44,18 @@ struct S390xElfFpregsetStruct {
 
 typedef struct S390xElfFpregsetStruct S390xElfFpregset;
 
+struct S390xElfVregsLoStruct {
+    uint64_t vregs[16];
+} QEMU_PACKED;
+
+typedef struct S390xElfVregsLoStruct S390xElfVregsLo;
+
+struct S390xElfVregsHiStruct {
+    uint64_t vregs[16][2];
+} QEMU_PACKED;
+
+typedef struct S390xElfVregsHiStruct S390xElfVregsHi;
+
 typedef struct noteStruct {
     Elf64_Nhdr hdr;
     char name[5];
@@ -51,6 +63,8 @@ typedef struct noteStruct {
     union {
         S390xElfPrstatus prstatus;
         S390xElfFpregset fpregset;
+        S390xElfVregsLo vregslo;
+        S390xElfVregsHi vregshi;
         uint32_t prefix;
         uint64_t timer;
         uint64_t todcmp;
@@ -78,14 +92,38 @@ static void s390x_write_elf64_prstatus(Note *note, S390CPU *cpu)
 static void s390x_write_elf64_fpregset(Note *note, S390CPU *cpu)
 {
     int i;
+    CPUS390XState *cs = &cpu->env;
 
     note->hdr.n_type = cpu_to_be32(NT_FPREGSET);
     note->contents.fpregset.fpc = cpu_to_be32(cpu->env.fpc);
     for (i = 0; i <= 15; i++) {
-        note->contents.fpregset.fprs[i] = cpu_to_be64(cpu->env.fregs[i].ll);
+        note->contents.fpregset.fprs[i] = cpu_to_be64(get_freg(cs, i)->ll);
+    }
+}
+
+static void s390x_write_elf64_vregslo(Note *note, S390CPU *cpu)
+{
+    int i;
+
+    note->hdr.n_type = cpu_to_be32(NT_S390_VXRS_LOW);
+    for (i = 0; i <= 15; i++) {
+        note->contents.vregslo.vregs[i] = cpu_to_be64(cpu->env.vregs[i][1].ll);
     }
 }
 
+static void s390x_write_elf64_vregshi(Note *note, S390CPU *cpu)
+{
+    int i;
+    S390xElfVregsHi *temp_vregshi;
+
+    temp_vregshi = &note->contents.vregshi;
+
+    note->hdr.n_type = cpu_to_be32(NT_S390_VXRS_HIGH);
+    for (i = 0; i <= 15; i++) {
+        temp_vregshi->vregs[i][0] = cpu_to_be64(cpu->env.vregs[i + 16][0].ll);
+        temp_vregshi->vregs[i][1] = cpu_to_be64(cpu->env.vregs[i + 16][1].ll);
+    }
+}
 
 static void s390x_write_elf64_timer(Note *note, S390CPU *cpu)
 {
@@ -134,6 +172,8 @@ static const struct NoteFuncDescStruct {
     {sizeof(((Note *)0)->contents.timer),    s390x_write_elf64_timer},
     {sizeof(((Note *)0)->contents.todcmp),   s390x_write_elf64_todcmp},
     {sizeof(((Note *)0)->contents.todpreg),  s390x_write_elf64_todpreg},
+    {sizeof(((Note *)0)->contents.vregslo),  s390x_write_elf64_vregslo},
+    {sizeof(((Note *)0)->contents.vregshi),  s390x_write_elf64_vregshi},
     { 0, NULL}
 };
 
diff --git a/target-s390x/cpu.h b/target-s390x/cpu.h
index c557211..a71abae 100644
--- a/target-s390x/cpu.h
+++ b/target-s390x/cpu.h
@@ -81,7 +81,11 @@ typedef struct MchkQueue {
 
 typedef struct CPUS390XState {
     uint64_t regs[16];     /* GP registers */
-    CPU_DoubleU fregs[16]; /* FP registers */
+    /*
+     * The floating point registers are part of the vector registers.
+     * vregs[0][0] -> vregs[15][0] are 16 floating point registers
+     */
+    CPU_DoubleU vregs[32][2];  /* vector registers */
     uint32_t aregs[16];    /* access registers */
 
     uint32_t fpc;          /* floating-point control register */
@@ -162,6 +166,11 @@ typedef struct CPUS390XState {
 
 } CPUS390XState;
 
+static inline CPU_DoubleU *get_freg(CPUS390XState *cs, int nr)
+{
+    return &cs->vregs[nr][0];
+}
+
 #include "cpu-qom.h"
 #include <sysemu/kvm.h>
 
@@ -936,6 +945,7 @@ struct sysib_322 {
 #define SIGP_SET_PREFIX        0x0d
 #define SIGP_STORE_STATUS_ADDR 0x0e
 #define SIGP_SET_ARCH          0x12
+#define SIGP_STORE_ADTL_STATUS 0x17
 
 /* SIGP condition codes */
 #define SIGP_CC_ORDER_CODE_ACCEPTED 0
@@ -1180,4 +1190,18 @@ static inline int s390_assign_subch_ioeventfd(EventNotifier *notifier,
     }
 }
 
+#ifdef CONFIG_KVM
+static inline bool vregs_needed(void *opaque)
+{
+    if (kvm_enabled()) {
+        return kvm_check_extension(kvm_state, KVM_CAP_S390_VECTOR_REGISTERS);
+    }
+    return 0;
+}
+#else
+static inline bool vregs_needed(void *opaque)
+{
+    return 0;
+}
+#endif
 #endif
diff --git a/target-s390x/gdbstub.c b/target-s390x/gdbstub.c
index 8945f02..ddc14a6 100644
--- a/target-s390x/gdbstub.c
+++ b/target-s390x/gdbstub.c
@@ -111,7 +111,7 @@ static int cpu_read_fp_reg(CPUS390XState *env, uint8_t *mem_buf, int n)
     case S390_FPC_REGNUM:
         return gdb_get_reg32(mem_buf, env->fpc);
     case S390_F0_REGNUM ... S390_F15_REGNUM:
-        return gdb_get_reg64(mem_buf, env->fregs[n - S390_F0_REGNUM].ll);
+        return gdb_get_reg64(mem_buf, get_freg(env, n - S390_F0_REGNUM)->ll);
     default:
         return 0;
     }
@@ -124,13 +124,55 @@ static int cpu_write_fp_reg(CPUS390XState *env, uint8_t *mem_buf, int n)
         env->fpc = ldl_p(mem_buf);
         return 4;
     case S390_F0_REGNUM ... S390_F15_REGNUM:
-        env->fregs[n - S390_F0_REGNUM].ll = ldtul_p(mem_buf);
+        get_freg(env, n - S390_F0_REGNUM)->ll = ldtul_p(mem_buf);
         return 8;
     default:
         return 0;
     }
 }
 
+/* the values represent the positions in s390-vx.xml */
+#define S390_V0L_REGNUM 0
+#define S390_V15L_REGNUM 15
+#define S390_V16_REGNUM 16
+#define S390_V31_REGNUM 31
+/* total number of registers in s390-vx.xml */
+#define S390_NUM_VREGS 32
+
+static int cpu_read_vreg(CPUS390XState *env, uint8_t *mem_buf, int n)
+{
+    int ret;
+
+    switch (n) {
+    case S390_V0L_REGNUM ... S390_V15L_REGNUM:
+        ret = gdb_get_reg64(mem_buf, env->vregs[n][1].ll);
+        break;
+    case S390_V16_REGNUM ... S390_V31_REGNUM:
+        ret = gdb_get_reg64(mem_buf, env->vregs[n][0].ll);
+        ret += gdb_get_reg64(mem_buf + 8, env->vregs[n][1].ll);
+        break;
+    default:
+        ret = 0;
+    }
+
+    return ret;
+}
+
+static int cpu_write_vreg(CPUS390XState *env, uint8_t *mem_buf, int n)
+{
+    switch (n) {
+    case S390_V0L_REGNUM ... S390_V15L_REGNUM:
+        env->vregs[n][1].ll = ldtul_p(mem_buf + 8);
+        return 8;
+    case S390_V16_REGNUM ... S390_V31_REGNUM:
+        env->vregs[n][0].ll = ldtul_p(mem_buf);
+        env->vregs[n][1].ll = ldtul_p(mem_buf + 8);
+        return 16;
+    default:
+        return 0;
+    }
+}
+
 void s390_cpu_gdb_init(CPUState *cs)
 {
     gdb_register_coprocessor(cs, cpu_read_ac_reg,
@@ -140,4 +182,8 @@ void s390_cpu_gdb_init(CPUState *cs)
     gdb_register_coprocessor(cs, cpu_read_fp_reg,
                              cpu_write_fp_reg,
                              S390_NUM_FP_REGS, "s390-fpr.xml", 0);
+
+    gdb_register_coprocessor(cs, cpu_read_vreg,
+                             cpu_write_vreg,
+                             S390_NUM_VREGS, "s390-vx.xml", 0);
 }
diff --git a/target-s390x/helper.c b/target-s390x/helper.c
index 041c9c7..6b47766 100644
--- a/target-s390x/helper.c
+++ b/target-s390x/helper.c
@@ -445,7 +445,7 @@ static void do_mchk_interrupt(CPUS390XState *env)
     lowcore = cpu_map_lowcore(env);
 
     for (i = 0; i < 16; i++) {
-        lowcore->floating_pt_save_area[i] = cpu_to_be64(env->fregs[i].ll);
+        lowcore->floating_pt_save_area[i] = cpu_to_be64(get_freg(env, i)->ll);
         lowcore->gpregs_save_area[i] = cpu_to_be64(env->regs[i]);
         lowcore->access_regs_save_area[i] = cpu_to_be32(env->aregs[i]);
         lowcore->cregs_save_area[i] = cpu_to_be64(env->cregs[i]);
diff --git a/target-s390x/kvm.c b/target-s390x/kvm.c
index ea18015..6de7759 100644
--- a/target-s390x/kvm.c
+++ b/target-s390x/kvm.c
@@ -269,6 +269,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
     }
 
     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_SIGP, 0);
+    kvm_vm_enable_cap(s, KVM_CAP_S390_VECTOR_REGISTERS, 0);
     kvm_vm_enable_cap(s, KVM_CAP_S390_USER_STSI, 0);
 
     return 0;
@@ -337,15 +338,24 @@ int kvm_arch_put_registers(CPUState *cs, int level)
         }
     }
 
-    /* Floating point */
-    for (i = 0; i < 16; i++) {
-        fpu.fprs[i] = env->fregs[i].ll;
-    }
-    fpu.fpc = env->fpc;
+    if (can_sync_regs(cs, KVM_SYNC_VRS)) {
+        for (i = 0; i < 32; i++) {
+            cs->kvm_run->s.regs.vrs[i][0] = env->vregs[i][0].ll;
+            cs->kvm_run->s.regs.vrs[i][1] = env->vregs[i][1].ll;
+        }
+        cs->kvm_run->s.regs.fpc = env->fpc;
+        cs->kvm_run->kvm_dirty_regs |= KVM_SYNC_VRS;
+    } else {
+        /* Floating point */
+        for (i = 0; i < 16; i++) {
+            fpu.fprs[i] = get_freg(env, i)->ll;
+        }
+        fpu.fpc = env->fpc;
 
-    r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
-    if (r < 0) {
-        return r;
+        r = kvm_vcpu_ioctl(cs, KVM_SET_FPU, &fpu);
+        if (r < 0) {
+            return r;
+        }
     }
 
     /* Do we need to save more than that? */
@@ -468,15 +478,23 @@ int kvm_arch_get_registers(CPUState *cs)
         }
     }
 
-    /* Floating point */
-    r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
-    if (r < 0) {
-        return r;
-    }
-    for (i = 0; i < 16; i++) {
-        env->fregs[i].ll = fpu.fprs[i];
+    /* Floating point and vector registers */
+    if (can_sync_regs(cs, KVM_SYNC_VRS)) {
+        for (i = 0; i < 32; i++) {
+            env->vregs[i][0].ll = cs->kvm_run->s.regs.vrs[i][0];
+            env->vregs[i][1].ll = cs->kvm_run->s.regs.vrs[i][1];
+        }
+        env->fpc = cs->kvm_run->s.regs.fpc;
+    } else {
+        r = kvm_vcpu_ioctl(cs, KVM_GET_FPU, &fpu);
+        if (r < 0) {
+            return r;
+        }
+        for (i = 0; i < 16; i++) {
+            get_freg(env, i)->ll = fpu.fprs[i];
+        }
+        env->fpc = fpu.fpc;
     }
-    env->fpc = fpu.fpc;
 
     /* The prefix */
     if (can_sync_regs(cs, KVM_SYNC_PREFIX)) {
@@ -1356,6 +1374,28 @@ static void sigp_stop(void *arg)
     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
 }
 
+#define ADTL_SAVE_AREA_SIZE 1024
+static int kvm_s390_store_adtl_status(S390CPU *cpu, hwaddr addr)
+{
+    void *mem;
+    hwaddr len = ADTL_SAVE_AREA_SIZE;
+
+    mem = cpu_physical_memory_map(addr, &len, 1);
+    if (!mem) {
+        return -EFAULT;
+    }
+    if (len != ADTL_SAVE_AREA_SIZE) {
+        cpu_physical_memory_unmap(mem, len, 1, 0);
+        return -EFAULT;
+    }
+
+    memcpy(mem, &cpu->env.vregs, 512);
+
+    cpu_physical_memory_unmap(mem, len, 1, len);
+
+    return 0;
+}
+
 #define KVM_S390_STORE_STATUS_DEF_ADDR offsetof(LowCore, floating_pt_save_area)
 #define SAVE_AREA_SIZE 512
 static int kvm_s390_store_status(S390CPU *cpu, hwaddr addr, bool store_arch)
@@ -1363,6 +1403,7 @@ static int kvm_s390_store_status(S390CPU *cpu, hwaddr addr, bool store_arch)
     static const uint8_t ar_id = 1;
     uint64_t ckc = cpu->env.ckc >> 8;
     void *mem;
+    int i;
     hwaddr len = SAVE_AREA_SIZE;
 
     mem = cpu_physical_memory_map(addr, &len, 1);
@@ -1377,7 +1418,9 @@ static int kvm_s390_store_status(S390CPU *cpu, hwaddr addr, bool store_arch)
     if (store_arch) {
         cpu_physical_memory_write(offsetof(LowCore, ar_access_id), &ar_id, 1);
     }
-    memcpy(mem, &cpu->env.fregs, 128);
+    for (i = 0; i < 16; ++i) {
+        *((uint64 *)mem + i) = get_freg(&cpu->env, i)->ll;
+    }
     memcpy(mem + 128, &cpu->env.regs, 128);
     memcpy(mem + 256, &cpu->env.psw, 16);
     memcpy(mem + 280, &cpu->env.psa, 4);
@@ -1441,6 +1484,36 @@ static void sigp_store_status_at_address(void *arg)
     si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
 }
 
+static void sigp_store_adtl_status(void *arg)
+{
+    SigpInfo *si = arg;
+
+    if (!kvm_check_extension(kvm_state, KVM_CAP_S390_VECTOR_REGISTERS)) {
+        set_sigp_status(si, SIGP_STAT_INVALID_ORDER);
+        return;
+    }
+
+    /* cpu has to be stopped */
+    if (s390_cpu_get_state(si->cpu) != CPU_STATE_STOPPED) {
+        set_sigp_status(si, SIGP_STAT_INCORRECT_STATE);
+        return;
+    }
+
+    /* parameter must be aligned to 1024-byte boundary */
+    if (si->param & 0x3ff) {
+        set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
+        return;
+    }
+
+    cpu_synchronize_state(CPU(si->cpu));
+
+    if (kvm_s390_store_adtl_status(si->cpu, si->param)) {
+        set_sigp_status(si, SIGP_STAT_INVALID_PARAMETER);
+        return;
+    }
+    si->cc = SIGP_CC_ORDER_CODE_ACCEPTED;
+}
+
 static void sigp_restart(void *arg)
 {
     SigpInfo *si = arg;
@@ -1558,6 +1631,9 @@ static int handle_sigp_single_dst(S390CPU *dst_cpu, uint8_t order,
     case SIGP_STORE_STATUS_ADDR:
         run_on_cpu(CPU(dst_cpu), sigp_store_status_at_address, &si);
         break;
+    case SIGP_STORE_ADTL_STATUS:
+        run_on_cpu(CPU(dst_cpu), sigp_store_adtl_status, &si);
+        break;
     case SIGP_SET_PREFIX:
         run_on_cpu(CPU(dst_cpu), sigp_set_prefix, &si);
         break;
diff --git a/target-s390x/machine.c b/target-s390x/machine.c
index 7853e3c..e52d760 100644
--- a/target-s390x/machine.c
+++ b/target-s390x/machine.c
@@ -47,22 +47,22 @@ const VMStateDescription vmstate_fpu = {
     .version_id = 1,
     .minimum_version_id = 1,
     .fields = (VMStateField[]) {
-        VMSTATE_UINT64(env.fregs[0].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[1].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[2].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[3].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[4].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[5].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[6].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[7].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[8].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[9].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[10].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[11].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[12].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[13].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[14].ll, S390CPU),
-        VMSTATE_UINT64(env.fregs[15].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[0][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[1][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[2][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[3][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[4][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[5][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[6][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[7][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[8][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[9][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[10][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[11][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[12][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[13][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[14][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[15][0].ll, S390CPU),
         VMSTATE_UINT32(env.fpc, S390CPU),
         VMSTATE_END_OF_LIST()
     }
@@ -73,6 +73,64 @@ static inline bool fpu_needed(void *opaque)
     return true;
 }
 
+const VMStateDescription vmstate_vregs = {
+    .name = "cpu/vregs",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .fields = (VMStateField[]) {
+        /* vregs[0][0] -> vregs[15][0] and fregs are overlays */
+        VMSTATE_UINT64(env.vregs[16][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[17][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[18][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[19][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[20][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[21][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[22][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[23][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[24][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[25][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[26][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[27][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[28][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[29][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[30][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[31][0].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[0][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[1][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[2][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[3][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[4][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[5][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[6][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[7][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[8][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[9][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[10][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[11][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[12][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[13][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[14][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[15][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[16][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[17][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[18][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[19][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[20][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[21][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[22][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[23][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[24][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[25][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[26][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[27][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[28][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[29][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[30][1].ll, S390CPU),
+        VMSTATE_UINT64(env.vregs[31][1].ll, S390CPU),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 const VMStateDescription vmstate_s390_cpu = {
     .name = "cpu",
     .post_load = cpu_post_load,
@@ -106,6 +164,9 @@ const VMStateDescription vmstate_s390_cpu = {
             .vmsd = &vmstate_fpu,
             .needed = fpu_needed,
         } , {
+            .vmsd = &vmstate_vregs,
+            .needed = vregs_needed,
+        } , {
             /* empty */
         }
     },
diff --git a/target-s390x/translate.c b/target-s390x/translate.c
index 80e3a54..fbffd30 100644
--- a/target-s390x/translate.c
+++ b/target-s390x/translate.c
@@ -113,7 +113,7 @@ void s390_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
     }
 
     for (i = 0; i < 16; i++) {
-        cpu_fprintf(f, "F%02d=%016" PRIx64, i, env->fregs[i].ll);
+        cpu_fprintf(f, "F%02d=%016" PRIx64, i, get_freg(env, i)->ll);
         if ((i % 4) == 3) {
             cpu_fprintf(f, "\n");
         } else {
@@ -121,6 +121,12 @@ void s390_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
         }
     }
 
+    for (i = 0; i < 32; i++) {
+        cpu_fprintf(f, "V%02d=%016" PRIx64 "%016" PRIx64, i,
+                    env->vregs[i][0].ll, env->vregs[i][1].ll);
+        cpu_fprintf(f, (i % 2) ? " " : "\n");
+    }
+
 #ifndef CONFIG_USER_ONLY
     for (i = 0; i < 16; i++) {
         cpu_fprintf(f, "C%02d=%016" PRIx64, i, env->cregs[i]);
@@ -187,7 +193,7 @@ void s390x_translate_init(void)
     for (i = 0; i < 16; i++) {
         snprintf(cpu_reg_names[i + 16], sizeof(cpu_reg_names[0]), "f%d", i);
         fregs[i] = tcg_global_mem_new(TCG_AREG0,
-                                      offsetof(CPUS390XState, fregs[i].d),
+                                      offsetof(CPUS390XState, vregs[i][0].d),
                                       cpu_reg_names[i + 16]);
     }
 }