Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.6-20160324' into staging

ppc patch queue for 2016-03-24

Accumulated patches for target-ppc, pseries machine type and related
devices.

* Preliminary patches from BenH & Cédric Le Goater's powernv code
    * We don't want the full machine type before 2.7
    * Adding some of the SPRs also fixes migration corner cases for
      spapr (when qemu has no knowledge of the registers, they're
      obviously not migrated)
    * We include some patches that aren't strictly fixes, but make
      applying the others easier, and they're low risk
* Fix to buffer management which significantly improves throughput in
  the spapr-llan virtual network device
* Start with 64-bit mode enabled on spapr.  This is the way it's
  supposed to be but we broke it a while back and didn't notice
  because Linux guests cope anyway.
    * Picked up by kvm-unit-tests
    * Still some bugs here that I'm working on

# gpg: Signature made Thu 24 Mar 2016 04:29:42 GMT using RSA key ID 20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>"
# gpg:                 aka "David Gibson (Red Hat) <dgibson@redhat.com>"
# gpg:                 aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E  87DC 6C38 CACA 20D9 B392

* remotes/dgibson/tags/ppc-for-2.6-20160324:
  ppc: move POWER8 Book4 regs in their own routine
  hw/net/spapr_llan: Enable the RX buffer pools by default for new machines
  hw/net/spapr_llan: Fix receive buffer handling for better performance
  hw/net/spapr_llan: Extract rx buffer code into separate functions
  ppc: A couple more dummy POWER8 Book4 regs
  ppc: Add dummy CIABR SPR
  ppc: Add POWER8 IAMR register
  ppc: Fix writing to AMR/UAMOR
  ppc: Initialize AMOR in PAPR mode
  ppc: Add dummy SPR_IC for POWER8
  ppc: Create cpu_ppc_set_papr() helper
  ppc: Add a bunch of hypervisor SPRs to Book3s
  ppc: Add macros to register hypervisor mode SPRs
  ppc: Update SPR definitions
  spapr/target-ppc/kvm: Only add hcall-instructions if KVM supports it
  ppc64: set MSR_SF bit

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

2 files changed, 297 insertions, 50 deletions

diff --git a/hw/net/spapr_llan.c b/hw/net/spapr_llan.c
index 5237b4d863..efc31cbfba 100644
--- a/hw/net/spapr_llan.c
+++ b/hw/net/spapr_llan.c
@@ -45,6 +45,10 @@
 #define DPRINTF(fmt...)
 #endif
 
+/* Compatibility flags for migration */
+#define SPAPRVLAN_FLAG_RX_BUF_POOLS_BIT  0
+#define SPAPRVLAN_FLAG_RX_BUF_POOLS      (1 << SPAPRVLAN_FLAG_RX_BUF_POOLS_BIT)
+
 /*
  * Virtual LAN device
  */
@@ -86,6 +90,15 @@ typedef uint64_t vlan_bd_t;
 #define VIO_SPAPR_VLAN_DEVICE(obj) \
      OBJECT_CHECK(VIOsPAPRVLANDevice, (obj), TYPE_VIO_SPAPR_VLAN_DEVICE)
 
+#define RX_POOL_MAX_BDS 4096
+#define RX_MAX_POOLS 5
+
+typedef struct {
+    int32_t bufsize;
+    int32_t count;
+    vlan_bd_t bds[RX_POOL_MAX_BDS];
+} RxBufPool;
+
 typedef struct VIOsPAPRVLANDevice {
     VIOsPAPRDevice sdev;
     NICConf nicconf;
@@ -94,6 +107,8 @@ typedef struct VIOsPAPRVLANDevice {
     target_ulong buf_list;
     uint32_t add_buf_ptr, use_buf_ptr, rx_bufs;
     target_ulong rxq_ptr;
+    uint32_t compat_flags;             /* Compatability flags for migration */
+    RxBufPool *rx_pool[RX_MAX_POOLS];  /* Receive buffer descriptor pools */
 } VIOsPAPRVLANDevice;
 
 static int spapr_vlan_can_receive(NetClientState *nc)
@@ -103,6 +118,73 @@ static int spapr_vlan_can_receive(NetClientState *nc)
     return (dev->isopen && dev->rx_bufs > 0);
 }
 
+/**
+ * Get buffer descriptor from one of our receive buffer pools
+ */
+static vlan_bd_t spapr_vlan_get_rx_bd_from_pool(VIOsPAPRVLANDevice *dev,
+                                                size_t size)
+{
+    vlan_bd_t bd;
+    int pool;
+
+    for (pool = 0; pool < RX_MAX_POOLS; pool++) {
+        if (dev->rx_pool[pool]->count > 0 &&
+            dev->rx_pool[pool]->bufsize >= size + 8) {
+            break;
+        }
+    }
+    if (pool == RX_MAX_POOLS) {
+        /* Failed to find a suitable buffer */
+        return 0;
+    }
+
+    DPRINTF("Found buffer: pool=%d count=%d rxbufs=%d\n", pool,
+            dev->rx_pool[pool]->count, dev->rx_bufs);
+
+    /* Remove the buffer from the pool */
+    dev->rx_pool[pool]->count--;
+    bd = dev->rx_pool[pool]->bds[dev->rx_pool[pool]->count];
+    dev->rx_pool[pool]->bds[dev->rx_pool[pool]->count] = 0;
+
+    return bd;
+}
+
+/**
+ * Get buffer descriptor from the receive buffer list page that has been
+ * supplied by the guest with the H_REGISTER_LOGICAL_LAN call
+ */
+static vlan_bd_t spapr_vlan_get_rx_bd_from_page(VIOsPAPRVLANDevice *dev,
+                                                size_t size)
+{
+    int buf_ptr = dev->use_buf_ptr;
+    vlan_bd_t bd;
+
+    do {
+        buf_ptr += 8;
+        if (buf_ptr >= VLAN_RX_BDS_LEN + VLAN_RX_BDS_OFF) {
+            buf_ptr = VLAN_RX_BDS_OFF;
+        }
+
+        bd = vio_ldq(&dev->sdev, dev->buf_list + buf_ptr);
+        DPRINTF("use_buf_ptr=%d bd=0x%016llx\n",
+                buf_ptr, (unsigned long long)bd);
+    } while ((!(bd & VLAN_BD_VALID) || VLAN_BD_LEN(bd) < size + 8)
+             && buf_ptr != dev->use_buf_ptr);
+
+    if (!(bd & VLAN_BD_VALID) || VLAN_BD_LEN(bd) < size + 8) {
+        /* Failed to find a suitable buffer */
+        return 0;
+    }
+
+    /* Remove the buffer from the pool */
+    dev->use_buf_ptr = buf_ptr;
+    vio_stq(&dev->sdev, dev->buf_list + dev->use_buf_ptr, 0);
+
+    DPRINTF("Found buffer: ptr=%d rxbufs=%d\n", dev->use_buf_ptr, dev->rx_bufs);
+
+    return bd;
+}
+
 static ssize_t spapr_vlan_receive(NetClientState *nc, const uint8_t *buf,
                                   size_t size)
 {
@@ -110,7 +192,6 @@ static ssize_t spapr_vlan_receive(NetClientState *nc, const uint8_t *buf,
     VIOsPAPRDevice *sdev = VIO_SPAPR_DEVICE(dev);
     vlan_bd_t rxq_bd = vio_ldq(sdev, dev->buf_list + VLAN_RXQ_BD_OFF);
     vlan_bd_t bd;
-    int buf_ptr = dev->use_buf_ptr;
     uint64_t handle;
     uint8_t control;
 
@@ -125,29 +206,16 @@ static ssize_t spapr_vlan_receive(NetClientState *nc, const uint8_t *buf,
         return -1;
     }
 
-    do {
-        buf_ptr += 8;
-        if (buf_ptr >= (VLAN_RX_BDS_LEN + VLAN_RX_BDS_OFF)) {
-            buf_ptr = VLAN_RX_BDS_OFF;
-        }
-
-        bd = vio_ldq(sdev, dev->buf_list + buf_ptr);
-        DPRINTF("use_buf_ptr=%d bd=0x%016llx\n",
-                buf_ptr, (unsigned long long)bd);
-    } while ((!(bd & VLAN_BD_VALID) || (VLAN_BD_LEN(bd) < (size + 8)))
-             && (buf_ptr != dev->use_buf_ptr));
-
-    if (!(bd & VLAN_BD_VALID) || (VLAN_BD_LEN(bd) < (size + 8))) {
-        /* Failed to find a suitable buffer */
+    if (dev->compat_flags & SPAPRVLAN_FLAG_RX_BUF_POOLS) {
+        bd = spapr_vlan_get_rx_bd_from_pool(dev, size);
+    } else {
+        bd = spapr_vlan_get_rx_bd_from_page(dev, size);
+    }
+    if (!bd) {
         return -1;
     }
 
-    /* Remove the buffer from the pool */
     dev->rx_bufs--;
-    dev->use_buf_ptr = buf_ptr;
-    vio_stq(sdev, dev->buf_list + dev->use_buf_ptr, 0);
-
-    DPRINTF("Found buffer: ptr=%d num=%d\n", dev->use_buf_ptr, dev->rx_bufs);
 
     /* Transfer the packet data */
     if (spapr_vio_dma_write(sdev, VLAN_BD_ADDR(bd) + 8, buf, size) < 0) {
@@ -195,13 +263,31 @@ static NetClientInfo net_spapr_vlan_info = {
     .receive = spapr_vlan_receive,
 };
 
+static void spapr_vlan_reset_rx_pool(RxBufPool *rxp)
+{
+    /*
+     * Use INT_MAX as bufsize so that unused buffers are moved to the end
+     * of the list during the qsort in spapr_vlan_add_rxbuf_to_pool() later.
+     */
+    rxp->bufsize = INT_MAX;
+    rxp->count = 0;
+    memset(rxp->bds, 0, sizeof(rxp->bds));
+}
+
 static void spapr_vlan_reset(VIOsPAPRDevice *sdev)
 {
     VIOsPAPRVLANDevice *dev = VIO_SPAPR_VLAN_DEVICE(sdev);
+    int i;
 
     dev->buf_list = 0;
     dev->rx_bufs = 0;
     dev->isopen = 0;
+
+    if (dev->compat_flags & SPAPRVLAN_FLAG_RX_BUF_POOLS) {
+        for (i = 0; i < RX_MAX_POOLS; i++) {
+            spapr_vlan_reset_rx_pool(dev->rx_pool[i]);
+        }
+    }
 }
 
 static void spapr_vlan_realize(VIOsPAPRDevice *sdev, Error **errp)
@@ -218,10 +304,31 @@ static void spapr_vlan_realize(VIOsPAPRDevice *sdev, Error **errp)
 static void spapr_vlan_instance_init(Object *obj)
 {
     VIOsPAPRVLANDevice *dev = VIO_SPAPR_VLAN_DEVICE(obj);
+    int i;
 
     device_add_bootindex_property(obj, &dev->nicconf.bootindex,
                                   "bootindex", "",
                                   DEVICE(dev), NULL);
+
+    if (dev->compat_flags & SPAPRVLAN_FLAG_RX_BUF_POOLS) {
+        for (i = 0; i < RX_MAX_POOLS; i++) {
+            dev->rx_pool[i] = g_new(RxBufPool, 1);
+            spapr_vlan_reset_rx_pool(dev->rx_pool[i]);
+        }
+    }
+}
+
+static void spapr_vlan_instance_finalize(Object *obj)
+{
+    VIOsPAPRVLANDevice *dev = VIO_SPAPR_VLAN_DEVICE(obj);
+    int i;
+
+    if (dev->compat_flags & SPAPRVLAN_FLAG_RX_BUF_POOLS) {
+        for (i = 0; i < RX_MAX_POOLS; i++) {
+            g_free(dev->rx_pool[i]);
+            dev->rx_pool[i] = NULL;
+        }
+    }
 }
 
 void spapr_vlan_create(VIOsPAPRBus *bus, NICInfo *nd)
@@ -372,6 +479,113 @@ static target_ulong h_free_logical_lan(PowerPCCPU *cpu,
     return H_SUCCESS;
 }
 
+/**
+ * Used for qsort, this function compares two RxBufPools by size.
+ */
+static int rx_pool_size_compare(const void *p1, const void *p2)
+{
+    const RxBufPool *pool1 = *(RxBufPool **)p1;
+    const RxBufPool *pool2 = *(RxBufPool **)p2;
+
+    if (pool1->bufsize < pool2->bufsize) {
+        return -1;
+    }
+    return pool1->bufsize > pool2->bufsize;
+}
+
+/**
+ * Search for a matching buffer pool with exact matching size,
+ * or return -1 if no matching pool has been found.
+ */
+static int spapr_vlan_get_rx_pool_id(VIOsPAPRVLANDevice *dev, int size)
+{
+    int pool;
+
+    for (pool = 0; pool < RX_MAX_POOLS; pool++) {
+        if (dev->rx_pool[pool]->bufsize == size) {
+            return pool;
+        }
+    }
+
+    return -1;
+}
+
+/**
+ * Enqueuing receive buffer by adding it to one of our receive buffer pools
+ */
+static target_long spapr_vlan_add_rxbuf_to_pool(VIOsPAPRVLANDevice *dev,
+                                                target_ulong buf)
+{
+    int size = VLAN_BD_LEN(buf);
+    int pool;
+
+    pool = spapr_vlan_get_rx_pool_id(dev, size);
+    if (pool < 0) {
+        /*
+         * No matching pool found? Try to use a new one. If the guest used all
+         * pools before, but changed the size of one pool inbetween, we might
+         * need to recycle that pool here (if it's empty already). Thus scan
+         * all buffer pools now, starting with the last (likely empty) one.
+         */
+        for (pool = RX_MAX_POOLS - 1; pool >= 0 ; pool--) {
+            if (dev->rx_pool[pool]->count == 0) {
+                dev->rx_pool[pool]->bufsize = size;
+                /*
+                 * Sort pools by size so that spapr_vlan_receive()
+                 * can later find the smallest buffer pool easily.
+                 */
+                qsort(dev->rx_pool, RX_MAX_POOLS, sizeof(dev->rx_pool[0]),
+                      rx_pool_size_compare);
+                pool = spapr_vlan_get_rx_pool_id(dev, size);
+                DPRINTF("created RX pool %d for size %lld\n", pool,
+                        VLAN_BD_LEN(buf));
+                break;
+            }
+        }
+    }
+    /* Still no usable pool? Give up */
+    if (pool < 0 || dev->rx_pool[pool]->count >= RX_POOL_MAX_BDS) {
+        return H_RESOURCE;
+    }
+
+    DPRINTF("h_add_llan_buf():  Add buf using pool %i (size %lli, count=%i)\n",
+            pool, VLAN_BD_LEN(buf), dev->rx_pool[pool]->count);
+
+    dev->rx_pool[pool]->bds[dev->rx_pool[pool]->count++] = buf;
+
+    return 0;
+}
+
+/**
+ * This is the old way of enqueuing receive buffers: Add it to the rx queue
+ * page that has been supplied by the guest (which is quite limited in size).
+ */
+static target_long spapr_vlan_add_rxbuf_to_page(VIOsPAPRVLANDevice *dev,
+                                                target_ulong buf)
+{
+    vlan_bd_t bd;
+
+    if (dev->rx_bufs >= VLAN_MAX_BUFS) {
+        return H_RESOURCE;
+    }
+
+    do {
+        dev->add_buf_ptr += 8;
+        if (dev->add_buf_ptr >= VLAN_RX_BDS_LEN + VLAN_RX_BDS_OFF) {
+            dev->add_buf_ptr = VLAN_RX_BDS_OFF;
+        }
+
+        bd = vio_ldq(&dev->sdev, dev->buf_list + dev->add_buf_ptr);
+    } while (bd & VLAN_BD_VALID);
+
+    vio_stq(&dev->sdev, dev->buf_list + dev->add_buf_ptr, buf);
+
+    DPRINTF("h_add_llan_buf():  Added buf  ptr=%d  rx_bufs=%d bd=0x%016llx\n",
+            dev->add_buf_ptr, dev->rx_bufs, (unsigned long long)buf);
+
+    return 0;
+}
+
 static target_ulong h_add_logical_lan_buffer(PowerPCCPU *cpu,
                                              sPAPRMachineState *spapr,
                                              target_ulong opcode,
@@ -381,7 +595,7 @@ static target_ulong h_add_logical_lan_buffer(PowerPCCPU *cpu,
     target_ulong buf = args[1];
     VIOsPAPRDevice *sdev = spapr_vio_find_by_reg(spapr->vio_bus, reg);
     VIOsPAPRVLANDevice *dev = VIO_SPAPR_VLAN_DEVICE(sdev);
-    vlan_bd_t bd;
+    target_long ret;
 
     DPRINTF("H_ADD_LOGICAL_LAN_BUFFER(0x" TARGET_FMT_lx
             ", 0x" TARGET_FMT_lx ")\n", reg, buf);
@@ -397,29 +611,23 @@ static target_ulong h_add_logical_lan_buffer(PowerPCCPU *cpu,
         return H_PARAMETER;
     }
 
-    if (!dev->isopen || dev->rx_bufs >= VLAN_MAX_BUFS) {
+    if (!dev->isopen) {
         return H_RESOURCE;
     }
 
-    do {
-        dev->add_buf_ptr += 8;
-        if (dev->add_buf_ptr >= (VLAN_RX_BDS_LEN + VLAN_RX_BDS_OFF)) {
-            dev->add_buf_ptr = VLAN_RX_BDS_OFF;
-        }
-
-        bd = vio_ldq(sdev, dev->buf_list + dev->add_buf_ptr);
-    } while (bd & VLAN_BD_VALID);
-
-    vio_stq(sdev, dev->buf_list + dev->add_buf_ptr, buf);
+    if (dev->compat_flags & SPAPRVLAN_FLAG_RX_BUF_POOLS) {
+        ret = spapr_vlan_add_rxbuf_to_pool(dev, buf);
+    } else {
+        ret = spapr_vlan_add_rxbuf_to_page(dev, buf);
+    }
+    if (ret) {
+        return ret;
+    }
 
     dev->rx_bufs++;
 
     qemu_flush_queued_packets(qemu_get_queue(dev->nic));
 
-    DPRINTF("h_add_logical_lan_buffer():  Added buf  ptr=%d  rx_bufs=%d"
-            " bd=0x%016llx\n", dev->add_buf_ptr, dev->rx_bufs,
-            (unsigned long long)buf);
-
     return H_SUCCESS;
 }
 
@@ -509,9 +717,44 @@ static target_ulong h_multicast_ctrl(PowerPCCPU *cpu, sPAPRMachineState *spapr,
 static Property spapr_vlan_properties[] = {
     DEFINE_SPAPR_PROPERTIES(VIOsPAPRVLANDevice, sdev),
     DEFINE_NIC_PROPERTIES(VIOsPAPRVLANDevice, nicconf),
+    DEFINE_PROP_BIT("use-rx-buffer-pools", VIOsPAPRVLANDevice,
+                    compat_flags, SPAPRVLAN_FLAG_RX_BUF_POOLS_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static bool spapr_vlan_rx_buffer_pools_needed(void *opaque)
+{
+    VIOsPAPRVLANDevice *dev = opaque;
+
+    return (dev->compat_flags & SPAPRVLAN_FLAG_RX_BUF_POOLS) != 0;
+}
+
+static const VMStateDescription vmstate_rx_buffer_pool = {
+    .name = "spapr_llan/rx_buffer_pool",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .needed = spapr_vlan_rx_buffer_pools_needed,
+    .fields = (VMStateField[]) {
+        VMSTATE_INT32(bufsize, RxBufPool),
+        VMSTATE_INT32(count, RxBufPool),
+        VMSTATE_UINT64_ARRAY(bds, RxBufPool, RX_POOL_MAX_BDS),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
+static const VMStateDescription vmstate_rx_pools = {
+    .name = "spapr_llan/rx_pools",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .needed = spapr_vlan_rx_buffer_pools_needed,
+    .fields = (VMStateField[]) {
+        VMSTATE_ARRAY_OF_POINTER_TO_STRUCT(rx_pool, VIOsPAPRVLANDevice,
+                                           RX_MAX_POOLS, 1,
+                                           vmstate_rx_buffer_pool, RxBufPool),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 static const VMStateDescription vmstate_spapr_llan = {
     .name = "spapr_llan",
     .version_id = 1,
@@ -528,6 +771,10 @@ static const VMStateDescription vmstate_spapr_llan = {
 
         VMSTATE_END_OF_LIST()
     },
+    .subsections = (const VMStateDescription * []) {
+        &vmstate_rx_pools,
+        NULL
+    }
 };
 
 static void spapr_vlan_class_init(ObjectClass *klass, void *data)
@@ -554,6 +801,7 @@ static const TypeInfo spapr_vlan_info = {
     .instance_size = sizeof(VIOsPAPRVLANDevice),
     .class_init    = spapr_vlan_class_init,
     .instance_init = spapr_vlan_instance_init,
+    .instance_finalize = spapr_vlan_instance_finalize,
 };
 
 static void spapr_vlan_register_types(void)
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index d0bb42305b..65abccb2f5 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -497,10 +497,11 @@ static void *spapr_create_fdt_skel(hwaddr initrd_base,
              * Older KVM versions with older guest kernels were broken with the
              * magic page, don't allow the guest to map it.
              */
-            kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
-                                 sizeof(hypercall));
-            _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
-                              sizeof(hypercall))));
+            if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
+                                      sizeof(hypercall))) {
+                _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
+                                   sizeof(hypercall))));
+            }
         }
         _FDT((fdt_end_node(fdt)));
     }
@@ -1612,15 +1613,8 @@ static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu,
     /* Set time-base frequency to 512 MHz */
     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
 
-    /* PAPR always has exception vectors in RAM not ROM. To ensure this,
-     * MSR[IP] should never be set.
-     */
-    env->msr_mask &= ~(1 << 6);
-
-    /* Tell KVM that we're in PAPR mode */
-    if (kvm_enabled()) {
-        kvmppc_set_papr(cpu);
-    }
+    /* Enable PAPR mode in TCG or KVM */
+    cpu_ppc_set_papr(cpu);
 
     if (cpu->max_compat) {
         Error *local_err = NULL;
@@ -2362,7 +2356,12 @@ DEFINE_SPAPR_MACHINE(2_6, "2.6", true);
  * pseries-2.5
  */
 #define SPAPR_COMPAT_2_5 \
-        HW_COMPAT_2_5
+    HW_COMPAT_2_5 \
+    { \
+        .driver   = "spapr-vlan", \
+        .property = "use-rx-buffer-pools", \
+        .value    = "off", \
+    },
 
 static void spapr_machine_2_5_instance_options(MachineState *machine)
 {