Skip to contentSkip to conversation

Flatten simple unions where we still can

PATCH In review Markus Armbruster 2 participants 6 messages Feb 21, 2017 – Feb 22, 2017 Networking

lore
Share & tools
lore
#0 lore raw

v2 cover Flatten simple unions where we still can

Simple unions are simpler than flat unions in the schema, but more complicated in C and on the QMP wire: there's extra indirection in C and extra nesting on the wire, both pointless. They're best avoided in new code.

But we can do more: we can flatten the simple unions that aren't used in QMP.

In the longer run, I'd love to get rid of simple unions entirely, and reduce the notational overhead of flat unions in the schema.

v2:
* PATCH 1+2: Since: added [Eric], commit message tweaked
* PATCH 3: new

Markus Armbruster (3):
numa: Flatten simple union NumaOptions
net: Flatten simple union NetLegacyOptions i386/cpu: net: Flatten simple union GuestPanicInformationType

 net/net.c         | 44 ++++++++++++++++++++++----------------------
 numa.c            |  4 ++--
 qapi-schema.json  | 35 +++++++++++++++++++++++++++++++++--
 target/i386/cpu.c | 17 ++++++-----------
 vl.c              | 12 ++++++------
 5 files changed, 69 insertions(+), 43 deletions(-)
Signature · 1 line
2.7.4
#1 lore raw

PATCH v2 1/3 numa: Flatten simple union NumaOptions

Simple unions are simpler than flat unions in the schema, but more complicated in C and on the QMP wire: there's extra indirection in C and extra nesting on the wire, both pointless. They're best avoided in new code.

NumaOptions isn't new, but it's only used internally, not in QMP. Convert it to a flat union.

---
 numa.c           |  4 ++--
 qapi-schema.json | 10 ++++++++++
 2 files changed, 12 insertions(+), 2 deletions(-)
numa.c +2 -2
index 9f56be9..e01cb54 100644
@@ -228,8 +228,8 @@ static int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
     }
 
     switch (object->type) {
-    case NUMA_OPTIONS_KIND_NODE:
-        numa_node_parse(object->u.node.data, opts, &err);
+    case NUMA_OPTIONS_TYPE_NODE:
+        numa_node_parse(&object->u.node, opts, &err);
         if (err) {
             goto end;
         }
qapi-schema.json +10 -0
index e9a6364..a448ea8 100644
@@ -5571,6 +5571,14 @@
             'events' : [ 'InputEvent' ] } }
 
 ##
+# @NumaOptionsType:
+#
+# Since: 2.1
+##
+{ 'enum': 'NumaOptionsType',
+  'data': [ 'node' ] }
+
+##
 # @NumaOptions:
 #
 # A discriminated record of NUMA options. (for OptsVisitor)
@@ -5578,6 +5586,8 @@
 # Since: 2.1
 ##
 { 'union': 'NumaOptions',
+  'base': { 'type': 'NumaOptionsType' },
+  'discriminator': 'type',
   'data': {
     'node': 'NumaNodeOptions' }}
 
Signature · 1 line
2.7.4
#2 lore raw

PATCH v2 2/3 net: Flatten simple union NetLegacyOptions

In reply to Markus Armbruster

Simple unions are simpler than flat unions in the schema, but more complicated in C and on the QMP wire: there's extra indirection in C and extra nesting on the wire, both pointless. They're best avoided in new code.

NetLegacyOptions isn't new, but it's only used internally, not in QMP. Convert it to a flat union.

---
 net/net.c        | 44 ++++++++++++++++++++++----------------------
 qapi-schema.json | 11 +++++++++++
 2 files changed, 33 insertions(+), 22 deletions(-)
net/net.c +22 -22
index fb7af3a..0ac3b9e 100644
@@ -993,47 +993,47 @@ static int net_client_init1(const void *object, bool is_netdev, Error **errp)
 
         /* Map the old options to the new flat type */
         switch (opts->type) {
-        case NET_LEGACY_OPTIONS_KIND_NONE:
+        case NET_LEGACY_OPTIONS_TYPE_NONE:
             return 0; /* nothing to do */
-        case NET_LEGACY_OPTIONS_KIND_NIC:
+        case NET_LEGACY_OPTIONS_TYPE_NIC:
             legacy.type = NET_CLIENT_DRIVER_NIC;
-            legacy.u.nic = *opts->u.nic.data;
+            legacy.u.nic = opts->u.nic;
             break;
-        case NET_LEGACY_OPTIONS_KIND_USER:
+        case NET_LEGACY_OPTIONS_TYPE_USER:
             legacy.type = NET_CLIENT_DRIVER_USER;
-            legacy.u.user = *opts->u.user.data;
+            legacy.u.user = opts->u.user;
             break;
-        case NET_LEGACY_OPTIONS_KIND_TAP:
+        case NET_LEGACY_OPTIONS_TYPE_TAP:
             legacy.type = NET_CLIENT_DRIVER_TAP;
-            legacy.u.tap = *opts->u.tap.data;
+            legacy.u.tap = opts->u.tap;
             break;
-        case NET_LEGACY_OPTIONS_KIND_L2TPV3:
+        case NET_LEGACY_OPTIONS_TYPE_L2TPV3:
             legacy.type = NET_CLIENT_DRIVER_L2TPV3;
-            legacy.u.l2tpv3 = *opts->u.l2tpv3.data;
+            legacy.u.l2tpv3 = opts->u.l2tpv3;
             break;
-        case NET_LEGACY_OPTIONS_KIND_SOCKET:
+        case NET_LEGACY_OPTIONS_TYPE_SOCKET:
             legacy.type = NET_CLIENT_DRIVER_SOCKET;
-            legacy.u.socket = *opts->u.socket.data;
+            legacy.u.socket = opts->u.socket;
             break;
-        case NET_LEGACY_OPTIONS_KIND_VDE:
+        case NET_LEGACY_OPTIONS_TYPE_VDE:
             legacy.type = NET_CLIENT_DRIVER_VDE;
-            legacy.u.vde = *opts->u.vde.data;
+            legacy.u.vde = opts->u.vde;
             break;
-        case NET_LEGACY_OPTIONS_KIND_DUMP:
+        case NET_LEGACY_OPTIONS_TYPE_DUMP:
             legacy.type = NET_CLIENT_DRIVER_DUMP;
-            legacy.u.dump = *opts->u.dump.data;
+            legacy.u.dump = opts->u.dump;
             break;
-        case NET_LEGACY_OPTIONS_KIND_BRIDGE:
+        case NET_LEGACY_OPTIONS_TYPE_BRIDGE:
             legacy.type = NET_CLIENT_DRIVER_BRIDGE;
-            legacy.u.bridge = *opts->u.bridge.data;
+            legacy.u.bridge = opts->u.bridge;
             break;
-        case NET_LEGACY_OPTIONS_KIND_NETMAP:
+        case NET_LEGACY_OPTIONS_TYPE_NETMAP:
             legacy.type = NET_CLIENT_DRIVER_NETMAP;
-            legacy.u.netmap = *opts->u.netmap.data;
+            legacy.u.netmap = opts->u.netmap;
             break;
-        case NET_LEGACY_OPTIONS_KIND_VHOST_USER:
+        case NET_LEGACY_OPTIONS_TYPE_VHOST_USER:
             legacy.type = NET_CLIENT_DRIVER_VHOST_USER;
-            legacy.u.vhost_user = *opts->u.vhost_user.data;
+            legacy.u.vhost_user = opts->u.vhost_user;
             break;
         default:
             abort();
@@ -1048,7 +1048,7 @@ static int net_client_init1(const void *object, bool is_netdev, Error **errp)
 
         /* Do not add to a vlan if it's a nic with a netdev= parameter. */
         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
-            !opts->u.nic.data->has_netdev) {
+            !opts->u.nic.has_netdev) {
             peer = net_hub_add_port(net->has_vlan ? net->vlan : 0, NULL);
         }
 
qapi-schema.json +11 -0
index a448ea8..5347781 100644
@@ -3959,6 +3959,15 @@
     'opts':  'NetLegacyOptions' } }
 
 ##
+# @NetLegacyOptionsType:
+#
+# Since: 1.2
+##
+{ 'enum': 'NetLegacyOptionsType',
+  'data': ['none', 'nic', 'user', 'tap', 'l2tpv3', 'socket', 'vde',
+           'dump', 'bridge', 'netmap', 'vhost-user'] }
+
+##
 # @NetLegacyOptions:
 #
 # Like Netdev, but for use only by the legacy command line options
@@ -3966,6 +3975,8 @@
 # Since: 1.2
 ##
 { 'union': 'NetLegacyOptions',
+  'base': { 'type': 'NetLegacyOptionsType' },
+  'discriminator': 'type',
   'data': {
     'none':     'NetdevNoneOptions',
     'nic':      'NetLegacyNicOptions',
Signature · 1 line
2.7.4
#3 lore raw

PATCH v2 3/3 i386/cpu: net: Flatten simple union GuestPanicInformationType

In reply to Markus Armbruster

Simple unions are simpler than flat unions in the schema, but more complicated in C and on the QMP wire: there's extra indirection in C and extra nesting on the wire, both pointless. They're best avoided in new code. Fix up recent commit d187e08 accordingly.

Cc Anton Nefedov <[email protected]>
Cc Denis V. Lunev <[email protected]>
---
 qapi-schema.json  | 14 ++++++++++++--
 target/i386/cpu.c | 17 ++++++-----------
 vl.c              | 12 ++++++------
 3 files changed, 24 insertions(+), 19 deletions(-)
qapi-schema.json +12 -2
index 5347781..0eef37c 100644
@@ -5893,14 +5893,24 @@
   'data': [ 'pause', 'poweroff' ] }
 
 ##
+# @GuestPanicInformationType:
+#
+# Since: 2.9
+##
+{ 'enum': 'GuestPanicInformationType',
+  'data': [ 'hyper-v' ] }
+
+##
 # @GuestPanicInformation:
 #
 # Information about a guest panic
 #
 # Since: 2.9
 ##
-{'union': 'GuestPanicInformation',
- 'data': { 'hyper-v': 'GuestPanicInformationHyperV' } }
+{ 'union': 'GuestPanicInformation',
+  'base': { 'type': 'GuestPanicInformationType' },
+  'discriminator': 'type',
+  'data': { 'hyper-v': 'GuestPanicInformationHyperV' } }
 
 ##
 # @GuestPanicInformationHyperV:
target/i386/cpu.c +6 -11
index fd7add2..c3e6b74 100644
@@ -3502,19 +3502,14 @@ static GuestPanicInformation *x86_cpu_get_crash_info(CPUState *cs)
     GuestPanicInformation *panic_info = NULL;
 
     if (env->features[FEAT_HYPERV_EDX] & HV_X64_GUEST_CRASH_MSR_AVAILABLE) {
-        GuestPanicInformationHyperV *panic_info_hv =
-            g_malloc0(sizeof(GuestPanicInformationHyperV));
         panic_info = g_malloc0(sizeof(GuestPanicInformation));
-
-        panic_info->type = GUEST_PANIC_INFORMATION_KIND_HYPER_V;
-        panic_info->u.hyper_v.data = panic_info_hv;
-
+        panic_info->type = GUEST_PANIC_INFORMATION_TYPE_HYPER_V;
         assert(HV_X64_MSR_CRASH_PARAMS >= 5);
-        panic_info_hv->arg1 = env->msr_hv_crash_params[0];
-        panic_info_hv->arg2 = env->msr_hv_crash_params[1];
-        panic_info_hv->arg3 = env->msr_hv_crash_params[2];
-        panic_info_hv->arg4 = env->msr_hv_crash_params[3];
-        panic_info_hv->arg5 = env->msr_hv_crash_params[4];
+        panic_info->u.hyper_v.arg1 = env->msr_hv_crash_params[0];
+        panic_info->u.hyper_v.arg2 = env->msr_hv_crash_params[1];
+        panic_info->u.hyper_v.arg3 = env->msr_hv_crash_params[2];
+        panic_info->u.hyper_v.arg4 = env->msr_hv_crash_params[3];
+        panic_info->u.hyper_v.arg5 = env->msr_hv_crash_params[4];
     }
 
     return panic_info;
vl.c +6 -6
index b5d0a19..e307ae0 100644
@@ -1697,14 +1697,14 @@ void qemu_system_guest_panicked(GuestPanicInformation *info)
     }
 
     if (info) {
-        if (info->type == GUEST_PANIC_INFORMATION_KIND_HYPER_V) {
+        if (info->type == GUEST_PANIC_INFORMATION_TYPE_HYPER_V) {
             qemu_log_mask(LOG_GUEST_ERROR, "HV crash parameters: (%#"PRIx64
                           " %#"PRIx64" %#"PRIx64" %#"PRIx64" %#"PRIx64")\n",
-                          info->u.hyper_v.data->arg1,
-                          info->u.hyper_v.data->arg2,
-                          info->u.hyper_v.data->arg3,
-                          info->u.hyper_v.data->arg4,
-                          info->u.hyper_v.data->arg5);
+                          info->u.hyper_v.arg1,
+                          info->u.hyper_v.arg2,
+                          info->u.hyper_v.arg3,
+                          info->u.hyper_v.arg4,
+                          info->u.hyper_v.arg5);
         }
         qapi_free_GuestPanicInformation(info);
     }
Signature · 1 line
2.7.4
Reviewed-by on v2 3/3
#4 lore raw

Re: [Qemu-devel] [PATCH v2 3/3] i386/cpu: net: Flatten simple union GuestPanicInformationType

Quoted 5 lines · Markus Armbruster

Simple unions are simpler than flat unions in the schema, but more complicated in C and on the QMP wire: there's extra indirection in C and extra nesting on the wire, both pointless. They're best avoided in new code. Fix up recent commit d187e08 accordingly.

Quoted 8 lines
Cc Anton Nefedov <[email protected]>
Cc Denis V. Lunev <[email protected]>
---
 qapi-schema.json  | 14 ++++++++++++--
 target/i386/cpu.c | 17 ++++++-----------
 vl.c              | 12 ++++++------
 3 files changed, 24 insertions(+), 19 deletions(-)

Very similar to Anton's patch which is on Paolo's queue:
https://lists.gnu.org/archive/html/qemu-devel/2017-02/msg04448.html

I don't care which version goes in, but will leave it to Markus and Paolo to decide which queue it should go through.

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
#5 lore raw

Re: [Qemu-devel] [PATCH v2 3/3] i386/cpu: net: Flatten simple union GuestPanicInformationType

Quoted 15 lines · Eric Blake
Quoted 5 lines · Markus Armbruster

Simple unions are simpler than flat unions in the schema, but more complicated in C and on the QMP wire: there's extra indirection in C and extra nesting on the wire, both pointless. They're best avoided in new code. Fix up recent commit d187e08 accordingly.

Quoted 8 lines
Cc Anton Nefedov <[email protected]>
Cc Denis V. Lunev <[email protected]>
---
 qapi-schema.json  | 14 ++++++++++++--
 target/i386/cpu.c | 17 ++++++-----------
 vl.c              | 12 ++++++------
 3 files changed, 24 insertions(+), 19 deletions(-)
Quoted 2 lines

Very similar to Anton's patch which is on Paolo's queue:
https://lists.gnu.org/archive/html/qemu-devel/2017-02/msg04448.html

Wasn't cc'ed, so I missed it. No biggie.

Quoted 2 lines

I don't care which version goes in, but will leave it to Markus and Paolo to decide which queue it should go through.

My commit message is more verbose, and I fix space style in the QAPI schema. On the other hand, Anton has an extra QAPI schema comment line.

Picking Anton's is probably simpler, because it's 2/3, and mine looks like it conflicts with his 1/3.

Quoted 1 line

Thanks!