[libvirt] [PATCH 0/7] qapi: add commands to remove the need to parse -help output
This series implements the necessary commands to implements danpb's idea to remove -help parsing in libvirt. We would introduce all of these commands in 1.2 and then change the -help output starting in 1.3. Here is Dan's plan from a previous thread: <danpb> Basically I'd sum up my new idea as "just use QMP". * No new command line arguments like -capabilities * libvirt invokes something like $QEMUBINARY -qmp CHARDEV -nodefault -nodefconfig -nographics * libvirt then runs a number of QMP commands to find out what it needs to know. I'd expect the following existing commands would be used - query-version - already supported - query-commands - already supported - query-events - already supported - query-kvm - already supported - qom-{list,list-types,get} - already supported - query-spice/vnc - already supported And add the following new commands - query-devices - new, -device ?, and/or -device NAME,? data in QMP - query-machines - new, -M ? in QMP - query-cpu-types - new, -cpu ? in QMP The above would take care of probably 50% of the current libvirt capabilities probing, including a portion of the -help stuff. Then there is all the rest of the crap we detect from the -help. We could just take the view, that "as of 1.2", we assume everything we previously detected is just available by default, and thus don't need to probe it. For stuff that is QOM based, I expect we'll be able to detect new features in the future using the qom-XXX monitor commands. For stuff that is non-qdev, and non-qom, libvirt can just do a plain version number check, unless we decide there is specific info worth exposing via other new 'query-XXX' monitor commands. Basically I'd sum up my new idea as "just use QMP". * No new command line arguments like -capabilities * libvirt invokes something like $QEMUBINARY -qmp CHARDEV -nodefault -nodefconfig -nographics * libvirt then runs a number of QMP commands to find out what it needs to know. I'd expect the following existing commands would be used - query-version - already supported - query-commands - already supported - query-events - already supported - query-kvm - already supported - qom-{list,list-types,get} - already supported - query-spice/vnc - already supported And add the following new commands - query-devices - new, -device ?, and/or -device NAME,? data in QMP - query-machines - new, -M ? in QMP - query-cpu-types - new, -cpu ? in QMP The above would take care of probably 50% of the current libvirt capabilities probing, including a portion of the -help stuff. Then there is all the rest of the crap we detect from the -help. We could just take the view, that "as of 1.2", we assume everything we previously detected is just available by default, and thus don't need to probe it. For stuff that is QOM based, I expect we'll be able to detect new features in the future using the qom-XXX monitor commands. For stuff that is non-qdev, and non-qom, libvirt can just do a plain version number check, unless we decide there is specific info worth exposing via other new 'query-XXX' monitor commands. </danpb> The one thing to note is that I didn't add a query-devices command because you can already do: qmp query-devices --implements=device --abstract=False To get the equivalent output of -device ?. Instead, I added a command to list specific properties of a device which is the equivalent of -device FOO,?
This can be used in conjunction with qom-list-types to determine the supported set of devices and their parameters. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 7 +++++++ qmp.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 85 insertions(+), 0 deletions(-) diff --git a/qapi-schema.json b/qapi-schema.json index bc55ed2..015a84a 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1727,6 +1727,34 @@ 'returns': [ 'ObjectTypeInfo' ] } ## +# @DevicePropertyInfo: +# +# Information about device properties. +# +# @name: the name of the property +# @type: the typename of the property +# +# Since: 1.2 +## +{ 'type': 'DevicePropertyInfo', + 'data': { 'name': 'str', 'type': 'str' } } + +## +# @device-list-properties: +# +# List properties associated with a device. +# +# @typename: the type name of a device +# +# Returns: a list of DevicePropertyInfo describing a devices properties +# +# Since: 1.2 +## +{ 'command': 'device-list-properties', + 'data': { 'typename': 'str'}, + 'returns': [ 'DevicePropertyInfo' ] } + +## # @migrate # # Migrates the current running guest to another Virtual Machine. diff --git a/qmp-commands.hx b/qmp-commands.hx index e3cf3c5..5c55528 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2215,3 +2215,10 @@ EQMP .args_type = "implements:s?,abstract:b?", .mhandler.cmd_new = qmp_marshal_input_qom_list_types, }, + + { + .name = "device-list-properties", + .args_type = "typename:s", + .mhandler.cmd_new = qmp_marshal_input_device_list_properties, + }, + diff --git a/qmp.c b/qmp.c index fee9fb2..254a32f 100644 --- a/qmp.c +++ b/qmp.c @@ -417,3 +417,53 @@ ObjectTypeInfoList *qmp_qom_list_types(bool has_implements, return ret; } + +DevicePropertyInfoList *qmp_device_list_properties(const char *typename, + Error **errp) +{ + ObjectClass *klass; + Property *prop; + DevicePropertyInfoList *prop_list = NULL; + + klass = object_class_by_name(typename); + if (klass == NULL) { + error_set(errp, QERR_DEVICE_NOT_FOUND, typename); + return NULL; + } + + klass = object_class_dynamic_cast(klass, TYPE_DEVICE); + if (klass == NULL) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, + "name", TYPE_DEVICE); + return NULL; + } + + do { + for (prop = DEVICE_CLASS(klass)->props; prop && prop->name; prop++) { + DevicePropertyInfoList *entry; + DevicePropertyInfo *info; + + /* + * TODO Properties without a parser are just for dirty hacks. + * qdev_prop_ptr is the only such PropertyInfo. It's marked + * for removal. This conditional should be removed along with + * it. + */ + if (!prop->info->set) { + continue; /* no way to set it, don't show */ + } + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(prop->name); + info->type = g_strdup(prop->info->legacy_name ?: prop->info->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = prop_list; + prop_list = entry; + } + klass = object_class_get_parent(klass); + } while (klass != object_class_by_name(TYPE_DEVICE)); + + return prop_list; +} -- 1.7.5.4
On Fri, 27 Jul 2012 08:37:13 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This can be used in conjunction with qom-list-types to determine the supported set of devices and their parameters.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 7 +++++++ qmp.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 85 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index bc55ed2..015a84a 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1727,6 +1727,34 @@ 'returns': [ 'ObjectTypeInfo' ] }
## +# @DevicePropertyInfo: +# +# Information about device properties. +# +# @name: the name of the property +# @type: the typename of the property +# +# Since: 1.2 +## +{ 'type': 'DevicePropertyInfo', + 'data': { 'name': 'str', 'type': 'str' } } + +## +# @device-list-properties: +# +# List properties associated with a device. +# +# @typename: the type name of a device +# +# Returns: a list of DevicePropertyInfo describing a devices properties +# +# Since: 1.2 +## +{ 'command': 'device-list-properties', + 'data': { 'typename': 'str'}, + 'returns': [ 'DevicePropertyInfo' ] }
I find it a bit weird that the argument is called typename but everything else refers to that same argument as a device. Maybe we should rename this to device-type-list-properties, DeviceTypePropertyInfo and typename to name. But that's minor, I won't oppose to it as it's.
+ +## # @migrate # # Migrates the current running guest to another Virtual Machine. diff --git a/qmp-commands.hx b/qmp-commands.hx index e3cf3c5..5c55528 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2215,3 +2215,10 @@ EQMP .args_type = "implements:s?,abstract:b?", .mhandler.cmd_new = qmp_marshal_input_qom_list_types, }, + + { + .name = "device-list-properties", + .args_type = "typename:s", + .mhandler.cmd_new = qmp_marshal_input_device_list_properties, + }, + diff --git a/qmp.c b/qmp.c index fee9fb2..254a32f 100644 --- a/qmp.c +++ b/qmp.c @@ -417,3 +417,53 @@ ObjectTypeInfoList *qmp_qom_list_types(bool has_implements,
return ret; } + +DevicePropertyInfoList *qmp_device_list_properties(const char *typename, + Error **errp) +{ + ObjectClass *klass; + Property *prop; + DevicePropertyInfoList *prop_list = NULL; + + klass = object_class_by_name(typename); + if (klass == NULL) { + error_set(errp, QERR_DEVICE_NOT_FOUND, typename); + return NULL; + } + + klass = object_class_dynamic_cast(klass, TYPE_DEVICE); + if (klass == NULL) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, + "name", TYPE_DEVICE); + return NULL; + } + + do { + for (prop = DEVICE_CLASS(klass)->props; prop && prop->name; prop++) { + DevicePropertyInfoList *entry; + DevicePropertyInfo *info; + + /* + * TODO Properties without a parser are just for dirty hacks. + * qdev_prop_ptr is the only such PropertyInfo. It's marked + * for removal. This conditional should be removed along with + * it. + */ + if (!prop->info->set) { + continue; /* no way to set it, don't show */ + } + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(prop->name); + info->type = g_strdup(prop->info->legacy_name ?: prop->info->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = prop_list; + prop_list = entry; + } + klass = object_class_get_parent(klass); + } while (klass != object_class_by_name(TYPE_DEVICE)); + + return prop_list; +}
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 27 Jul 2012 08:37:13 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This can be used in conjunction with qom-list-types to determine the supported set of devices and their parameters.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 7 +++++++ qmp.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 85 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index bc55ed2..015a84a 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1727,6 +1727,34 @@ 'returns': [ 'ObjectTypeInfo' ] }
## +# @DevicePropertyInfo: +# +# Information about device properties. +# +# @name: the name of the property +# @type: the typename of the property +# +# Since: 1.2 +## +{ 'type': 'DevicePropertyInfo', + 'data': { 'name': 'str', 'type': 'str' } } + +## +# @device-list-properties: +# +# List properties associated with a device. +# +# @typename: the type name of a device +# +# Returns: a list of DevicePropertyInfo describing a devices properties +# +# Since: 1.2 +## +{ 'command': 'device-list-properties', + 'data': { 'typename': 'str'}, + 'returns': [ 'DevicePropertyInfo' ] }
I find it a bit weird that the argument is called typename but everything else refers to that same argument as a device.
Maybe we should rename this to device-type-list-properties, DeviceTypePropertyInfo and typename to name.
I see where you're coming from but I think I'd prefer to leave it as-is. 'DeviceType' isn't really a concept in QEMU. Regards, Anthony Liguori
But that's minor, I won't oppose to it as it's.
+ +## # @migrate # # Migrates the current running guest to another Virtual Machine. diff --git a/qmp-commands.hx b/qmp-commands.hx index e3cf3c5..5c55528 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2215,3 +2215,10 @@ EQMP .args_type = "implements:s?,abstract:b?", .mhandler.cmd_new = qmp_marshal_input_qom_list_types, }, + + { + .name = "device-list-properties", + .args_type = "typename:s", + .mhandler.cmd_new = qmp_marshal_input_device_list_properties, + }, + diff --git a/qmp.c b/qmp.c index fee9fb2..254a32f 100644 --- a/qmp.c +++ b/qmp.c @@ -417,3 +417,53 @@ ObjectTypeInfoList *qmp_qom_list_types(bool has_implements,
return ret; } + +DevicePropertyInfoList *qmp_device_list_properties(const char *typename, + Error **errp) +{ + ObjectClass *klass; + Property *prop; + DevicePropertyInfoList *prop_list = NULL; + + klass = object_class_by_name(typename); + if (klass == NULL) { + error_set(errp, QERR_DEVICE_NOT_FOUND, typename); + return NULL; + } + + klass = object_class_dynamic_cast(klass, TYPE_DEVICE); + if (klass == NULL) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, + "name", TYPE_DEVICE); + return NULL; + } + + do { + for (prop = DEVICE_CLASS(klass)->props; prop && prop->name; prop++) { + DevicePropertyInfoList *entry; + DevicePropertyInfo *info; + + /* + * TODO Properties without a parser are just for dirty hacks. + * qdev_prop_ptr is the only such PropertyInfo. It's marked + * for removal. This conditional should be removed along with + * it. + */ + if (!prop->info->set) { + continue; /* no way to set it, don't show */ + } + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(prop->name); + info->type = g_strdup(prop->info->legacy_name ?: prop->info->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = prop_list; + prop_list = entry; + } + klass = object_class_get_parent(klass); + } while (klass != object_class_by_name(TYPE_DEVICE)); + + return prop_list; +}
We've had a cycle to tweak. It is time to commit to supporting them. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 19 ++++--------------- 1 files changed, 4 insertions(+), 15 deletions(-) diff --git a/qapi-schema.json b/qapi-schema.json index 015a84a..28e9914 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1360,9 +1360,7 @@ # 4) A link type in the form 'link<subtype>' where subtype is a qdev # device type name. Link properties form the device model graph. # -# Since: 1.1 -# -# Notes: This type is experimental. Its syntax may change in future releases. +# Since: 1.2 ## { 'type': 'ObjectPropertyInfo', 'data': { 'name': 'str', 'type': 'str' } } @@ -1379,10 +1377,7 @@ # Returns: a list of @ObjectPropertyInfo that describe the properties of the # object. # -# Since: 1.1 -# -# Notes: This command is experimental. It's syntax may change in future -# releases. +# Since: 1.2 ## { 'command': 'qom-list', 'data': { 'path': 'str' }, @@ -1418,9 +1413,7 @@ # returns as #str pathnames. All integer property types (u8, u16, etc) # are returned as #int. # -# Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. +# Since: 1.2 ## { 'command': 'qom-get', 'data': { 'path': 'str', 'property': 'str' }, @@ -1439,9 +1432,7 @@ # @value: a value who's type is appropriate for the property type. See @qom-get # for a description of type mapping. # -# Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. +# Since: 1.2 ## { 'command': 'qom-set', 'data': { 'path': 'str', 'property': 'str', 'value': 'visitor' }, @@ -1719,8 +1710,6 @@ # Returns: a list of @ObjectTypeInfo or an empty list if no results are found # # Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. ## { 'command': 'qom-list-types', 'data': { '*implements': 'str', '*abstract': 'bool' }, -- 1.7.5.4
On Fri, 27 Jul 2012 08:37:14 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
We've had a cycle to tweak. It is time to commit to supporting them.
qmp_qom_get() and qpm_qom_set() still use the legacy monitor interface, can't we convert it to the qapi?
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 19 ++++--------------- 1 files changed, 4 insertions(+), 15 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 015a84a..28e9914 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1360,9 +1360,7 @@ # 4) A link type in the form 'link<subtype>' where subtype is a qdev # device type name. Link properties form the device model graph. # -# Since: 1.1 -# -# Notes: This type is experimental. Its syntax may change in future releases. +# Since: 1.2 ## { 'type': 'ObjectPropertyInfo', 'data': { 'name': 'str', 'type': 'str' } } @@ -1379,10 +1377,7 @@ # Returns: a list of @ObjectPropertyInfo that describe the properties of the # object. # -# Since: 1.1 -# -# Notes: This command is experimental. It's syntax may change in future -# releases. +# Since: 1.2 ## { 'command': 'qom-list', 'data': { 'path': 'str' }, @@ -1418,9 +1413,7 @@ # returns as #str pathnames. All integer property types (u8, u16, etc) # are returned as #int. # -# Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. +# Since: 1.2 ## { 'command': 'qom-get', 'data': { 'path': 'str', 'property': 'str' }, @@ -1439,9 +1432,7 @@ # @value: a value who's type is appropriate for the property type. See @qom-get # for a description of type mapping. # -# Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. +# Since: 1.2 ## { 'command': 'qom-set', 'data': { 'path': 'str', 'property': 'str', 'value': 'visitor' }, @@ -1719,8 +1710,6 @@ # Returns: a list of @ObjectTypeInfo or an empty list if no results are found # # Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. ## { 'command': 'qom-list-types', 'data': { '*implements': 'str', '*abstract': 'bool' },
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 27 Jul 2012 08:37:14 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
We've had a cycle to tweak. It is time to commit to supporting them.
qmp_qom_get() and qpm_qom_set() still use the legacy monitor interface, can't we convert it to the qapi?
qapi doesn't have a concept of type passthrough yet. If it was added, it could be converted. Regards, Anthony Liguori
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 19 ++++--------------- 1 files changed, 4 insertions(+), 15 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 015a84a..28e9914 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1360,9 +1360,7 @@ # 4) A link type in the form 'link<subtype>' where subtype is a qdev # device type name. Link properties form the device model graph. # -# Since: 1.1 -# -# Notes: This type is experimental. Its syntax may change in future releases. +# Since: 1.2 ## { 'type': 'ObjectPropertyInfo', 'data': { 'name': 'str', 'type': 'str' } } @@ -1379,10 +1377,7 @@ # Returns: a list of @ObjectPropertyInfo that describe the properties of the # object. # -# Since: 1.1 -# -# Notes: This command is experimental. It's syntax may change in future -# releases. +# Since: 1.2 ## { 'command': 'qom-list', 'data': { 'path': 'str' }, @@ -1418,9 +1413,7 @@ # returns as #str pathnames. All integer property types (u8, u16, etc) # are returned as #int. # -# Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. +# Since: 1.2 ## { 'command': 'qom-get', 'data': { 'path': 'str', 'property': 'str' }, @@ -1439,9 +1432,7 @@ # @value: a value who's type is appropriate for the property type. See @qom-get # for a description of type mapping. # -# Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. +# Since: 1.2 ## { 'command': 'qom-set', 'data': { 'path': 'str', 'property': 'str', 'value': 'visitor' }, @@ -1719,8 +1710,6 @@ # Returns: a list of @ObjectTypeInfo or an empty list if no results are found # # Since: 1.1 -# -# Notes: This command is experimental and may change syntax in future releases. ## { 'command': 'qom-list-types', 'data': { '*implements': 'str', '*abstract': 'bool' },
This provides the same output as -M ? but in a structured way. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-) diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name +# +# @default: #optional whether the machine is default +# +# Since: 1.2.0 +## +{ 'type': 'MachineInfo', + 'data': { 'name': 'str', '*alias': 'str', + '*is-default': 'bool' } } + +## +# @query-machines: +# +# Return a list of supported machines +# +# Returns: a list of MachineInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-machines', 'returns': ['MachineInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index 5c55528..a6f82fc 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2222,3 +2222,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_device_list_properties, }, + { + .name = "query-machines", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_machines, + }, + diff --git a/vl.c b/vl.c index 8904db1..cd900e0 100644 --- a/vl.c +++ b/vl.c @@ -1209,6 +1209,37 @@ QEMUMachine *find_default_machine(void) return NULL; } +MachineInfoList *qmp_query_machines(Error **errp) +{ + MachineInfoList *mach_list = NULL; + QEMUMachine *m; + + for (m = first_machine; m; m = m->next) { + MachineInfoList *entry; + MachineInfo *info; + + info = g_malloc0(sizeof(*info)); + if (m->is_default) { + info->has_is_default = true; + info->is_default = true; + } + + if (m->alias) { + info->has_alias = true; + info->alias = g_strdup(m->alias); + } + + info->name = g_strdup(m->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = mach_list; + mach_list = entry; + } + + return mach_list; +} + /***********************************************************/ /* main execution loop */ -- 1.7.5.4
On Fri, 27 Jul 2012 08:37:15 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name +# +# @default: #optional whether the machine is default
Why is default optional?
+# +# Since: 1.2.0 +## +{ 'type': 'MachineInfo', + 'data': { 'name': 'str', '*alias': 'str', + '*is-default': 'bool' } } + +## +# @query-machines: +# +# Return a list of supported machines +# +# Returns: a list of MachineInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-machines', 'returns': ['MachineInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index 5c55528..a6f82fc 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2222,3 +2222,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_device_list_properties, },
+ { + .name = "query-machines", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_machines, + }, + diff --git a/vl.c b/vl.c index 8904db1..cd900e0 100644 --- a/vl.c +++ b/vl.c @@ -1209,6 +1209,37 @@ QEMUMachine *find_default_machine(void) return NULL; }
+MachineInfoList *qmp_query_machines(Error **errp) +{ + MachineInfoList *mach_list = NULL; + QEMUMachine *m; + + for (m = first_machine; m; m = m->next) { + MachineInfoList *entry; + MachineInfo *info; + + info = g_malloc0(sizeof(*info)); + if (m->is_default) { + info->has_is_default = true; + info->is_default = true; + } + + if (m->alias) { + info->has_alias = true; + info->alias = g_strdup(m->alias); + } + + info->name = g_strdup(m->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = mach_list; + mach_list = entry; + } + + return mach_list; +} + /***********************************************************/ /* main execution loop */
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 27 Jul 2012 08:37:15 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name +# +# @default: #optional whether the machine is default
Why is default optional?
Brievity. Regards, Anthony Liguori
+# +# Since: 1.2.0 +## +{ 'type': 'MachineInfo', + 'data': { 'name': 'str', '*alias': 'str', + '*is-default': 'bool' } } + +## +# @query-machines: +# +# Return a list of supported machines +# +# Returns: a list of MachineInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-machines', 'returns': ['MachineInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index 5c55528..a6f82fc 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2222,3 +2222,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_device_list_properties, },
+ { + .name = "query-machines", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_machines, + }, + diff --git a/vl.c b/vl.c index 8904db1..cd900e0 100644 --- a/vl.c +++ b/vl.c @@ -1209,6 +1209,37 @@ QEMUMachine *find_default_machine(void) return NULL; }
+MachineInfoList *qmp_query_machines(Error **errp) +{ + MachineInfoList *mach_list = NULL; + QEMUMachine *m; + + for (m = first_machine; m; m = m->next) { + MachineInfoList *entry; + MachineInfo *info; + + info = g_malloc0(sizeof(*info)); + if (m->is_default) { + info->has_is_default = true; + info->is_default = true; + } + + if (m->alias) { + info->has_alias = true; + info->alias = g_strdup(m->alias); + } + + info->name = g_strdup(m->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = mach_list; + mach_list = entry; + } + + return mach_list; +} + /***********************************************************/ /* main execution loop */
On Fri, 10 Aug 2012 09:41:20 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 27 Jul 2012 08:37:15 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name +# +# @default: #optional whether the machine is default
Why is default optional?
Brievity.
Can you elaborate, please?
Regards,
Anthony Liguori
+# +# Since: 1.2.0 +## +{ 'type': 'MachineInfo', + 'data': { 'name': 'str', '*alias': 'str', + '*is-default': 'bool' } } + +## +# @query-machines: +# +# Return a list of supported machines +# +# Returns: a list of MachineInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-machines', 'returns': ['MachineInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index 5c55528..a6f82fc 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2222,3 +2222,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_device_list_properties, },
+ { + .name = "query-machines", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_machines, + }, + diff --git a/vl.c b/vl.c index 8904db1..cd900e0 100644 --- a/vl.c +++ b/vl.c @@ -1209,6 +1209,37 @@ QEMUMachine *find_default_machine(void) return NULL; }
+MachineInfoList *qmp_query_machines(Error **errp) +{ + MachineInfoList *mach_list = NULL; + QEMUMachine *m; + + for (m = first_machine; m; m = m->next) { + MachineInfoList *entry; + MachineInfo *info; + + info = g_malloc0(sizeof(*info)); + if (m->is_default) { + info->has_is_default = true; + info->is_default = true; + } + + if (m->alias) { + info->has_alias = true; + info->alias = g_strdup(m->alias); + } + + info->name = g_strdup(m->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = mach_list; + mach_list = entry; + } + + return mach_list; +} + /***********************************************************/ /* main execution loop */
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 10 Aug 2012 09:41:20 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 27 Jul 2012 08:37:15 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name +# +# @default: #optional whether the machine is default
Why is default optional?
Brievity.
Can you elaborate, please?
There is only one machine that is default. Having default=false for all of the rest just adds a lot of unnecessary information in the response. Regards, Anthony Liguori
Regards,
Anthony Liguori
+# +# Since: 1.2.0 +## +{ 'type': 'MachineInfo', + 'data': { 'name': 'str', '*alias': 'str', + '*is-default': 'bool' } } + +## +# @query-machines: +# +# Return a list of supported machines +# +# Returns: a list of MachineInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-machines', 'returns': ['MachineInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index 5c55528..a6f82fc 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2222,3 +2222,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_device_list_properties, },
+ { + .name = "query-machines", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_machines, + }, + diff --git a/vl.c b/vl.c index 8904db1..cd900e0 100644 --- a/vl.c +++ b/vl.c @@ -1209,6 +1209,37 @@ QEMUMachine *find_default_machine(void) return NULL; }
+MachineInfoList *qmp_query_machines(Error **errp) +{ + MachineInfoList *mach_list = NULL; + QEMUMachine *m; + + for (m = first_machine; m; m = m->next) { + MachineInfoList *entry; + MachineInfo *info; + + info = g_malloc0(sizeof(*info)); + if (m->is_default) { + info->has_is_default = true; + info->is_default = true; + } + + if (m->alias) { + info->has_alias = true; + info->alias = g_strdup(m->alias); + } + + info->name = g_strdup(m->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = mach_list; + mach_list = entry; + } + + return mach_list; +} + /***********************************************************/ /* main execution loop */
On Fri, 10 Aug 2012 11:06:14 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 10 Aug 2012 09:41:20 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
Luiz Capitulino <lcapitulino@redhat.com> writes:
On Fri, 27 Jul 2012 08:37:15 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name +# +# @default: #optional whether the machine is default
Why is default optional?
Brievity.
Can you elaborate, please?
There is only one machine that is default. Having default=false for all of the rest just adds a lot of unnecessary information in the response.
I think it's more consistent to have the key (also, there are better ways to save bytes on the wire if this is an issue), but I don't mind much though.
On 07/27/2012 07:37 AM, Anthony Liguori wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name
Can a machine have more than one alias? In that case, it might be better to have: @name: the name of the machine @alias-of: #optional if present, name is an alias, and this lists the canonical form that it resolves to -- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
Eric Blake <eblake@redhat.com> writes:
On 07/27/2012 07:37 AM, Anthony Liguori wrote:
This provides the same output as -M ? but in a structured way.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 28 ++++++++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ vl.c | 31 +++++++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 28e9914..5b47026 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2200,3 +2200,31 @@ # Since: 0.14.0 ## { 'command': 'closefd', 'data': {'fdname': 'str'} } + +## +# @MachineInfo: +# +# Information describing a machine. +# +# @name: the name of the machine +# +# @alias: #optional an alias for the machine name
Can a machine have more than one alias?
No, it can't. Really, alias is only used to map 'pc' -> 'pc-1.1'. It's never more complicated than that. Regards, Anthony Liguori
In that case, it might be better to have:
@name: the name of the machine @alias-of: #optional if present, name is an alias, and this lists the canonical form that it resolves to
-- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
On 07/27/2012 12:12 PM, Anthony Liguori wrote:
Eric Blake <eblake@redhat.com> writes:
+# @name: the name of the machine +# +# @alias: #optional an alias for the machine name
Can a machine have more than one alias?
No, it can't. Really, alias is only used to map 'pc' -> 'pc-1.1'. It's never more complicated than that.
Fair enough. Just making sure that in this case, we would see: { "name":"pc-1.1", "alias":"pc" } and not the other way around, and in a future release, it may change to: { "name":"pc-1.1" }, { "name":"pc-1.2", "alias":"pc" } -- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
This lets us provide a default implementation of a symbol which targets can override. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- compiler.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/compiler.h b/compiler.h index 736e770..f76921e 100644 --- a/compiler.h +++ b/compiler.h @@ -45,6 +45,7 @@ # define GCC_ATTR __attribute__((__unused__, format(gnu_printf, 1, 2))) # define GCC_FMT_ATTR(n, m) __attribute__((format(gnu_printf, n, m))) # endif +#define GCC_WEAK __attribute__((weak)) #else #define GCC_ATTR /**/ #define GCC_FMT_ATTR(n, m) -- 1.7.5.4
On 27 July 2012 14:37, Anthony Liguori <aliguori@us.ibm.com> wrote:
--- a/compiler.h +++ b/compiler.h @@ -45,6 +45,7 @@ # define GCC_ATTR __attribute__((__unused__, format(gnu_printf, 1, 2))) # define GCC_FMT_ATTR(n, m) __attribute__((format(gnu_printf, n, m))) # endif +#define GCC_WEAK __attribute__((weak)) #else #define GCC_ATTR /**/ #define GCC_FMT_ATTR(n, m)
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ? (Also, no version of the macro in the not-GCC branch of the #if.) -- PMM
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 14:37, Anthony Liguori <aliguori@us.ibm.com> wrote:
--- a/compiler.h +++ b/compiler.h @@ -45,6 +45,7 @@ # define GCC_ATTR __attribute__((__unused__, format(gnu_printf, 1, 2))) # define GCC_FMT_ATTR(n, m) __attribute__((format(gnu_printf, n, m))) # endif +#define GCC_WEAK __attribute__((weak)) #else #define GCC_ATTR /**/ #define GCC_FMT_ATTR(n, m)
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
Weak symbols are supposed to be supported by mingw32. I have no idea about MacOS X. I have no way to develop or test for MacOS X using free software so I honestly don't care about it. Regards, Anthony Liguori
(Also, no version of the macro in the not-GCC branch of the #if.)
-- PMM
On 27 July 2012 15:27, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
Weak symbols are supposed to be supported by mingw32.
I have no idea about MacOS X.
I have no way to develop or test for MacOS X using free software so I honestly don't care about it.
My approach to this is to avoid non-standard things -- if I write a patch which is pretty much standard C then it's the platform's problem if it mishandles it. If I write a patch that uses a compiler-specific or OS-specific thing then I have to also provide the relevant alternatives...so I try to avoid doing that :-) -- PMM
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 15:27, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
Weak symbols are supposed to be supported by mingw32.
I have no idea about MacOS X.
I have no way to develop or test for MacOS X using free software so I honestly don't care about it.
My approach to this is to avoid non-standard things
http://en.wikipedia.org/wiki/C99#Implementations So unless you plan on compiling QEMU with xlc, pgi, or icc, I don't think relying on "standard things" really helps. QEMU doesn't support C99, it supports GCC. There's no point in debating about whether we should rely on extensions or not. We already do--extensively. Regards, Anthony Liguori
-- if I write a patch which is pretty much standard C then it's the platform's problem if it mishandles it. If I write a patch that uses a compiler-specific or OS-specific thing then I have to also provide the relevant alternatives...so I try to avoid doing that :-)
-- PMM
On Fri, Jul 27, 2012 at 3:31 PM, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 15:27, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
Weak symbols are supposed to be supported by mingw32.
I have no idea about MacOS X.
I have no way to develop or test for MacOS X using free software so I honestly don't care about it.
My approach to this is to avoid non-standard things
http://en.wikipedia.org/wiki/C99#Implementations
So unless you plan on compiling QEMU with xlc, pgi, or icc, I don't think relying on "standard things" really helps.
LLVM/Clang should definitely be in the plan.
QEMU doesn't support C99, it supports GCC. There's no point in debating about whether we should rely on extensions or not. We already do--extensively.
Not so extensively. There are a few extensions for which there is no simple alternative (like QEMU_PACKED) but other compilers likely need similar extensions. Then there are other extensions (like :? without middle expression) which can be easily avoided. We should avoid to use the non-standard extensions whenever possible.
Regards,
Anthony Liguori
-- if I write a patch which is pretty much standard C then it's the platform's problem if it mishandles it. If I write a patch that uses a compiler-specific or OS-specific thing then I have to also provide the relevant alternatives...so I try to avoid doing that :-)
-- PMM
Blue Swirl <blauwirbel@gmail.com> writes:
On Fri, Jul 27, 2012 at 3:31 PM, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 15:27, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
Weak symbols are supposed to be supported by mingw32.
I have no idea about MacOS X.
I have no way to develop or test for MacOS X using free software so I honestly don't care about it.
My approach to this is to avoid non-standard things
http://en.wikipedia.org/wiki/C99#Implementations
So unless you plan on compiling QEMU with xlc, pgi, or icc, I don't think relying on "standard things" really helps.
LLVM/Clang should definitely be in the plan.
weak symbols are supported by clang.
QEMU doesn't support C99, it supports GCC. There's no point in debating about whether we should rely on extensions or not. We already do--extensively.
Not so extensively. There are a few extensions for which there is no simple alternative (like QEMU_PACKED) but other compilers likely need similar extensions. Then there are other extensions (like :? without middle expression) which can be easily avoided. We should avoid to use the non-standard extensions whenever possible.
I disagree. I don't see a point to it. QEMU has never been routinely built on anything other than GCC. Why go to a lot of trouble to support a user base that doesn't exist? If someone comes along and actively maintains support for another compiler, we can revisit. But otherwise, there's no practical reason to avoid extensions. Regards, Anthony Liguori
Regards,
Anthony Liguori
-- if I write a patch which is pretty much standard C then it's the platform's problem if it mishandles it. If I write a patch that uses a compiler-specific or OS-specific thing then I have to also provide the relevant alternatives...so I try to avoid doing that :-)
-- PMM
On Fri, Jul 27, 2012 at 8:51 PM, Anthony Liguori <aliguori@us.ibm.com> wrote:
Blue Swirl <blauwirbel@gmail.com> writes:
On Fri, Jul 27, 2012 at 3:31 PM, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 15:27, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
Weak symbols are supposed to be supported by mingw32.
I have no idea about MacOS X.
I have no way to develop or test for MacOS X using free software so I honestly don't care about it.
My approach to this is to avoid non-standard things
http://en.wikipedia.org/wiki/C99#Implementations
So unless you plan on compiling QEMU with xlc, pgi, or icc, I don't think relying on "standard things" really helps.
LLVM/Clang should definitely be in the plan.
weak symbols are supported by clang.
QEMU doesn't support C99, it supports GCC. There's no point in debating about whether we should rely on extensions or not. We already do--extensively.
Not so extensively. There are a few extensions for which there is no simple alternative (like QEMU_PACKED) but other compilers likely need similar extensions. Then there are other extensions (like :? without middle expression) which can be easily avoided. We should avoid to use the non-standard extensions whenever possible.
I disagree. I don't see a point to it. QEMU has never been routinely built on anything other than GCC. Why go to a lot of trouble to support a user base that doesn't exist?
If someone comes along and actively maintains support for another compiler, we can revisit. But otherwise, there's no practical reason to avoid extensions.
Because it's more compliant to standards. There's also very little benefit from using the nonessential extensions.
Regards,
Anthony Liguori
Regards,
Anthony Liguori
-- if I write a patch which is pretty much standard C then it's the platform's problem if it mishandles it. If I write a patch that uses a compiler-specific or OS-specific thing then I have to also provide the relevant alternatives...so I try to avoid doing that :-)
-- PMM
On 27 July 2012 16:31, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
My approach to this is to avoid non-standard things
http://en.wikipedia.org/wiki/C99#Implementations
So unless you plan on compiling QEMU with xlc, pgi, or icc, I don't think relying on "standard things" really helps.
QEMU doesn't support C99, it supports GCC.
OK, you could perhaps rephrase that as 'mainstream' rather than 'standards-compliant'. I don't think we need to be strict C99; I do think we have more than one working host OS and that patches that use functionality that's documented not to work on all gcc targets ought to come attached to a statement that they've been tested. (MacOSX isn't actually in MAINTAINERS as a host so is a bit of a red herring. Windows is listed.) So if you really like weak symbols, go ahead. I'm just saying you're imposing a bigger testing burden on yourself than if you handled this some other way. (I do think it would be nice to care about building with CLANG, because there are some static analysis tools that we would then be able to run. That doesn't mean dropping all GCC extensions, though, because CLANG does support a lot of them.) -- PMM
On Sat, Jul 28, 2012 at 6:50 AM, Peter Maydell <peter.maydell@linaro.org> wrote:
On 27 July 2012 16:31, Anthony Liguori <aliguori@us.ibm.com> wrote:
Peter Maydell <peter.maydell@linaro.org> writes:
My approach to this is to avoid non-standard things
http://en.wikipedia.org/wiki/C99#Implementations
So unless you plan on compiling QEMU with xlc, pgi, or icc, I don't think relying on "standard things" really helps.
QEMU doesn't support C99, it supports GCC.
OK, you could perhaps rephrase that as 'mainstream' rather than 'standards-compliant'. I don't think we need to be strict C99; I do think we have more than one working host OS and that patches that use functionality that's documented not to work on all gcc targets ought to come attached to a statement that they've been tested. (MacOSX isn't actually in MAINTAINERS as a host so is a bit of a red herring. Windows is listed.)
I'd also like to avoid a world where everything only targets GCC on x86_64 on Linux with KVM. "Embrace and extend" may also be seen to apply to GCC extensions.
So if you really like weak symbols, go ahead. I'm just saying you're imposing a bigger testing burden on yourself than if you handled this some other way.
(I do think it would be nice to care about building with CLANG, because there are some static analysis tools that we would then be able to run. That doesn't mean dropping all GCC extensions, though, because CLANG does support a lot of them.)
-- PMM
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 14:37, Anthony Liguori <aliguori@us.ibm.com> wrote:
--- a/compiler.h +++ b/compiler.h @@ -45,6 +45,7 @@ # define GCC_ATTR __attribute__((__unused__, format(gnu_printf, 1, 2))) # define GCC_FMT_ATTR(n, m) __attribute__((format(gnu_printf, n, m))) # endif +#define GCC_WEAK __attribute__((weak)) #else #define GCC_ATTR /**/ #define GCC_FMT_ATTR(n, m)
The GCC manual says "Weak symbols are supported for ELF targets, and also for a.out targets when using the GNU assembler and linker". Have you tested this on Windows and MacOSX ?
(Also, no version of the macro in the not-GCC branch of the #if.)
(We don't support any compiler other than GCC). Not really sure why it is even in a branch. Regards, Anthony Liguori
-- PMM
This command attempts to map to the behavior of -cpu ?. Unfortunately, the output of this command differs wildly across targets. To accomodate this, we use a weak symbol to implement a default version of the command that fails with a QERR_NOT_SUPPORTED error code. Targets can then override and implement this command if it makes sense for them. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 23 +++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ qmp.c | 6 ++++++ 3 files changed, 35 insertions(+), 0 deletions(-) diff --git a/qapi-schema.json b/qapi-schema.json index 5b47026..768fb44 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2228,3 +2228,26 @@ # Since: 1.2.0 ## { 'command': 'query-machines', 'returns': ['MachineInfo'] } + +## +# @CpuDefInfo: +# +# Virtual CPU definition. +# +# @name: the name of the CPU definition +# +# Since: 1.2.0 +## +{ 'type': 'CpuDefInfo', + 'data': { 'name': 'str' } } + +## +# @query-cpudefs: +# +# Return a list of supported virtual CPU definitions +# +# Returns: a list of CpuDefInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-cpudefs', 'returns': ['CpuDefInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index a6f82fc..73dfeab 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2228,3 +2228,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_query_machines, }, + { + .name = "query-cpudefs", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_cpudefs, + }, + diff --git a/qmp.c b/qmp.c index 254a32f..51b4f75 100644 --- a/qmp.c +++ b/qmp.c @@ -467,3 +467,9 @@ DevicePropertyInfoList *qmp_device_list_properties(const char *typename, return prop_list; } + +CpuDefInfoList GCC_WEAK *qmp_query_cpudefs(Error **errp) +{ + error_set(errp, QERR_NOT_SUPPORTED); + return NULL; +} -- 1.7.5.4
On 27 July 2012 14:37, Anthony Liguori <aliguori@us.ibm.com> wrote:
This command attempts to map to the behavior of -cpu ?. Unfortunately, the output of this command differs wildly across targets.
I've never really understood why so much of the cpu selection logic is deferred to target-*...
To accomodate this, we use a weak symbol to implement a default version of the command that fails with a QERR_NOT_SUPPORTED error code. Targets can then override and implement this command if it makes sense for them.
This is a bit of a weak reason (boom boom!) for requiring a platform specific thing like weak symbols, though, and it's not how we handle similar existing cases (eg see the configure/makefile logic for memory_mapping.c vs memory_mapping-stub.c). If having separate configure/make stuff for each of these things sounds a bit heavyweight, we could just have a target-stubs.c which #includes cpu.h and has a lot of #ifndef TARGET_QUERY_CPUDEFS [stub version] #endif #ifndef TARGET_GET_MEMORY_MAPPING [stub version] #endif etc. -- PMM
Peter Maydell <peter.maydell@linaro.org> writes:
On 27 July 2012 14:37, Anthony Liguori <aliguori@us.ibm.com> wrote:
This command attempts to map to the behavior of -cpu ?. Unfortunately, the output of this command differs wildly across targets.
I've never really understood why so much of the cpu selection logic is deferred to target-*...
It will be fixed as part of the QOM conversion.
To accomodate this, we use a weak symbol to implement a default version of the command that fails with a QERR_NOT_SUPPORTED error code. Targets can then override and implement this command if it makes sense for them.
This is a bit of a weak reason (boom boom!) for requiring a platform specific thing like weak symbols, though, and it's not how we handle similar existing cases (eg see the configure/makefile logic for memory_mapping.c vs memory_mapping-stub.c).
I don't think we have a consistent approach today FWIW. I think using weak symbols is sufficiently compelling that it will become consistent.
If having separate configure/make stuff for each of these things sounds a bit heavyweight, we could just have a target-stubs.c which #includes cpu.h and has a lot of #ifndef TARGET_QUERY_CPUDEFS [stub version] #endif #ifndef TARGET_GET_MEMORY_MAPPING [stub version] #endif
This is pretty hideous. FWIW, weak symbols are supported on OS X as of 10.2. Regards, Anthony Liguori
etc.
-- PMM
On Fri, 27 Jul 2012 08:37:17 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This command attempts to map to the behavior of -cpu ?. Unfortunately, the output of this command differs wildly across targets.
To accomodate this, we use a weak symbol to implement a default version of the command that fails with a QERR_NOT_SUPPORTED error code. Targets can then override and implement this command if it makes sense for them.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- qapi-schema.json | 23 +++++++++++++++++++++++ qmp-commands.hx | 6 ++++++ qmp.c | 6 ++++++ 3 files changed, 35 insertions(+), 0 deletions(-)
diff --git a/qapi-schema.json b/qapi-schema.json index 5b47026..768fb44 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -2228,3 +2228,26 @@ # Since: 1.2.0 ## { 'command': 'query-machines', 'returns': ['MachineInfo'] } + +## +# @CpuDefInfo: +# +# Virtual CPU definition. +# +# @name: the name of the CPU definition +# +# Since: 1.2.0 +## +{ 'type': 'CpuDefInfo', + 'data': { 'name': 'str' } } + +## +# @query-cpudefs:
I'd call this query-cpu-defs or even query-cpu-difinitions. The latter makes it self-documenting.
+# +# Return a list of supported virtual CPU definitions +# +# Returns: a list of CpuDefInfo +# +# Since: 1.2.0 +## +{ 'command': 'query-cpudefs', 'returns': ['CpuDefInfo'] } diff --git a/qmp-commands.hx b/qmp-commands.hx index a6f82fc..73dfeab 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -2228,3 +2228,9 @@ EQMP .mhandler.cmd_new = qmp_marshal_input_query_machines, },
+ { + .name = "query-cpudefs", + .args_type = "", + .mhandler.cmd_new = qmp_marshal_input_query_cpudefs, + }, + diff --git a/qmp.c b/qmp.c index 254a32f..51b4f75 100644 --- a/qmp.c +++ b/qmp.c @@ -467,3 +467,9 @@ DevicePropertyInfoList *qmp_device_list_properties(const char *typename,
return prop_list; } + +CpuDefInfoList GCC_WEAK *qmp_query_cpudefs(Error **errp) +{ + error_set(errp, QERR_NOT_SUPPORTED); + return NULL; +}
On 07/27/2012 07:37 AM, Anthony Liguori wrote:
This command attempts to map to the behavior of -cpu ?. Unfortunately, the output of this command differs wildly across targets.
To accomodate this, we use a weak symbol to implement a default version of the
s/accomodate/accommodate/
command that fails with a QERR_NOT_SUPPORTED error code. Targets can then override and implement this command if it makes sense for them.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
+## +# @CpuDefInfo: +# +# Virtual CPU definition. +# +# @name: the name of the CPU definition +# +# Since: 1.2.0
We are inconsistent in this file whether to use '1.2.0' or just '1.2'; probably worth a followup patch to clean whichever version is deemed inappropriate.
+## +{ 'type': 'CpuDefInfo', + 'data': { 'name': 'str' } } + +## +# @query-cpudefs:
I though QMP favored English words, as in 'query-cpu-definitions' -- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-i386/cpu.c | 22 ++++++++++++++++++++++ 1 files changed, 22 insertions(+), 0 deletions(-) diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 6b9659f..b398439 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -28,6 +28,7 @@ #include "qemu-config.h" #include "qapi/qapi-visit-core.h" +#include "qmp-commands.h" #include "hyperv.h" @@ -1123,6 +1124,27 @@ void x86_cpu_list(FILE *f, fprintf_function cpu_fprintf, const char *optarg) } } +CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + x86_def_t *def; + + for (def = x86_defs; def; def = def->next) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(def->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +} + int cpu_x86_register(X86CPU *cpu, const char *cpu_model) { CPUX86State *env = &cpu->env; -- 1.7.5.4
On Fri, Jul 27, 2012 at 08:37:18AM -0500, Anthony Liguori wrote:
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-i386/cpu.c | 22 ++++++++++++++++++++++ 1 files changed, 22 insertions(+), 0 deletions(-)
diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 6b9659f..b398439 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -28,6 +28,7 @@ #include "qemu-config.h"
#include "qapi/qapi-visit-core.h" +#include "qmp-commands.h"
#include "hyperv.h"
@@ -1123,6 +1124,27 @@ void x86_cpu_list(FILE *f, fprintf_function cpu_fprintf, const char *optarg) } }
+CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + x86_def_t *def; + + for (def = x86_defs; def; def = def->next) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(def->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +}
How would the interface look like once we: - let libvirt know which features are available on each CPU model (libvirt needs that information[1]); and - add machine-type-specific cpudef compatibility changes? Would the command report different results depending on -machine? Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself? [1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about. -- Eduardo
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Jul 27, 2012 at 08:37:18AM -0500, Anthony Liguori wrote:
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-i386/cpu.c | 22 ++++++++++++++++++++++ 1 files changed, 22 insertions(+), 0 deletions(-)
diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 6b9659f..b398439 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -28,6 +28,7 @@ #include "qemu-config.h"
#include "qapi/qapi-visit-core.h" +#include "qmp-commands.h"
#include "hyperv.h"
@@ -1123,6 +1124,27 @@ void x86_cpu_list(FILE *f, fprintf_function cpu_fprintf, const char *optarg) } }
+CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + x86_def_t *def; + + for (def = x86_defs; def; def = def->next) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(def->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +}
How would the interface look like once we: - let libvirt know which features are available on each CPU model (libvirt needs that information[1]); and
I'm not sure I understand why libvirt needs this information. Can you elaborate?
- add machine-type-specific cpudef compatibility changes?
I think we've discussed this in IRC. I don't think we need to worry about this.
Would the command report different results depending on -machine?
No.
Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself?
I'm not sure what you mean by compatibility data. Regards, Anthony Liguori
[1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about.
-- Eduardo
On Fri, Aug 10, 2012 at 09:43:21AM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Jul 27, 2012 at 08:37:18AM -0500, Anthony Liguori wrote:
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-i386/cpu.c | 22 ++++++++++++++++++++++ 1 files changed, 22 insertions(+), 0 deletions(-)
diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 6b9659f..b398439 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -28,6 +28,7 @@ #include "qemu-config.h"
#include "qapi/qapi-visit-core.h" +#include "qmp-commands.h"
#include "hyperv.h"
@@ -1123,6 +1124,27 @@ void x86_cpu_list(FILE *f, fprintf_function cpu_fprintf, const char *optarg) } }
+CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + x86_def_t *def; + + for (def = x86_defs; def; def = def->next) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(def->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +}
How would the interface look like once we: - let libvirt know which features are available on each CPU model (libvirt needs that information[1]); and
I'm not sure I understand why libvirt needs this information. Can you elaborate?
I see two reasons: - The libvirt API has functions to tell the user which features are going to be enabled for each CPU model, so it needs to know which features are enabled or not, for each machine-type + cpu-model combination, so this information can be reported proeprly. - Also, if libvirt can enable/disable specific CPU features in the command-line, it just makes sens to know which ones are already enabled in each built-in CPU model. - Probing for migration: libvirt needs to know if a given CPU model on a host can be migrated to another host. To know that, two pieces of information are needed: A) Which CPU features are visible to the guest for a specific configuration; B) Which of those features are really supported by the host hardware+kernel+QEMU, on the destination host, so it can know if migration is really possible. I am assuming that libvirt will query for A and B, and then combine it itself. But QEMU could also simply calculate (A&B) itself, and just have a boolean function that tells if a given model can be run on a specific host or not. But then it wouldn't solve the first item above (knowing which features are already enabled on a CPU model). - The problem is: even if QEMU does the check itself, the result of that "can this host run this VM?" probing function depends on the machine-type, too (see explanation below).
- add machine-type-specific cpudef compatibility changes?
I think we've discussed this in IRC. I don't think we need to worry about this.
I remember discussing a lot about the mechanism we will use to add the compatibility changes, but I don t know how the query API will look like, after we implement this mechanism.
Would the command report different results depending on -machine?
No.
The problem is: 1) We need to introduce fixes on a CPU model that changes the set of guest-visible features (add or remove a feature)[1]; 2) The fix has to keep compatibility, so older machine-types will keep exposing the old set of gues-visible features; - That means different machine-types will have different CPU features being exposed. 3) libvirt needs to control/know which guest-visible CPU features are available to the guest (see above); 4) Because of (2), the querying system used by libvirt need to depend on the CPU model and machine-type. [1] Example: The SandyBridge model today has the "tsc-deadline" bit set, but QEMU-1.1 did not expose the tsc-deadline feature properly because of incorrect expectations about the GET_SUPPORTED_CPUID ioctl. This was fixed on qemu-1.2. That means "qemu-1.1 -machine pc-1.1 -cpu SandyBridge" does _not_ expose tsc-deadline to the guest, and we need to make "qemu-1.2 -machine pc-1.1 -cpu SandyBridge" _not_ expose it, too (otherwise migration from qemu-1.1 to qemu-1.2 will be broken).
Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself?
I'm not sure what you mean by compatibility data.
I mean any guest-visible compatibility bit that we will need to introduce on older machine-types, when making changes on CPU models (see the SandyBridge + tsc-deadline example above). I see two options: - Libvirt queries for a [f(machine_type, cpu_model) -> cpu_features] function, that will take into account the machine-type-specific compatibility bits. - Libvirt queries for a [f(cpu_model) -> cpu_features] function and a [f(machine_type) -> compatibility_changes] function, and combine both. - I don't like this approach, I am just including it as a possible alternative.
Regards,
Anthony Liguori
[1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about.
-- Eduardo
-- Eduardo
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Aug 10, 2012 at 09:43:21AM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Jul 27, 2012 at 08:37:18AM -0500, Anthony Liguori wrote:
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-i386/cpu.c | 22 ++++++++++++++++++++++ 1 files changed, 22 insertions(+), 0 deletions(-)
diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 6b9659f..b398439 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -28,6 +28,7 @@ #include "qemu-config.h"
#include "qapi/qapi-visit-core.h" +#include "qmp-commands.h"
#include "hyperv.h"
@@ -1123,6 +1124,27 @@ void x86_cpu_list(FILE *f, fprintf_function cpu_fprintf, const char *optarg) } }
+CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + x86_def_t *def; + + for (def = x86_defs; def; def = def->next) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(def->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +}
How would the interface look like once we: - let libvirt know which features are available on each CPU model (libvirt needs that information[1]); and
I'm not sure I understand why libvirt needs this information. Can you elaborate?
I see two reasons:
- The libvirt API has functions to tell the user which features are going to be enabled for each CPU model, so it needs to know which features are enabled or not, for each machine-type + cpu-model combination, so this information can be reported proeprly.
Ok, step number one is that CPU 'features' need to be defined more formally. By formally, I mean via qapi-schema.json. Then we can extend this command to return the set of features supported by each CPU type. The first step will need to sort out how this maps across architectures.
- Also, if libvirt can enable/disable specific CPU features in the command-line, it just makes sens to know which ones are already enabled in each built-in CPU model.
- Probing for migration: libvirt needs to know if a given CPU model on a host can be migrated to another host. To know that, two pieces of information are needed: A) Which CPU features are visible to the guest for a specific configuration; B) Which of those features are really supported by the host hardware+kernel+QEMU, on the destination host, so it can know if migration is really possible.
Note that what QEMU thinks it exposes is not necessarily what gets exposed. KVM may mask additional features. How is this handled today?
- add machine-type-specific cpudef compatibility changes?
I think we've discussed this in IRC. I don't think we need to worry about this.
I remember discussing a lot about the mechanism we will use to add the compatibility changes, but I don t know how the query API will look like, after we implement this mechanism.
0) User-defined CPU definitions go away - We already made a big step in this direction 1) CPU becomes a DeviceState 2) Features are expressed as properties 3) Same global mechanism used for everything else is used for CPUs Regards, Anthony Liguori
Would the command report different results depending on -machine?
No.
The problem is:
1) We need to introduce fixes on a CPU model that changes the set of guest-visible features (add or remove a feature)[1]; 2) The fix has to keep compatibility, so older machine-types will keep exposing the old set of gues-visible features; - That means different machine-types will have different CPU features being exposed. 3) libvirt needs to control/know which guest-visible CPU features are available to the guest (see above); 4) Because of (2), the querying system used by libvirt need to depend on the CPU model and machine-type.
[1] Example: The SandyBridge model today has the "tsc-deadline" bit set, but QEMU-1.1 did not expose the tsc-deadline feature properly because of incorrect expectations about the GET_SUPPORTED_CPUID ioctl. This was fixed on qemu-1.2.
That means "qemu-1.1 -machine pc-1.1 -cpu SandyBridge" does _not_ expose tsc-deadline to the guest, and we need to make "qemu-1.2 -machine pc-1.1 -cpu SandyBridge" _not_ expose it, too (otherwise migration from qemu-1.1 to qemu-1.2 will be broken).
Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself?
I'm not sure what you mean by compatibility data.
I mean any guest-visible compatibility bit that we will need to introduce on older machine-types, when making changes on CPU models (see the SandyBridge + tsc-deadline example above).
I see two options: - Libvirt queries for a [f(machine_type, cpu_model) -> cpu_features] function, that will take into account the machine-type-specific compatibility bits. - Libvirt queries for a [f(cpu_model) -> cpu_features] function and a [f(machine_type) -> compatibility_changes] function, and combine both. - I don't like this approach, I am just including it as a possible alternative.
Regards,
Anthony Liguori
[1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about.
-- Eduardo
-- Eduardo
On Fri, Aug 10, 2012 at 11:37:30AM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Aug 10, 2012 at 09:43:21AM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Jul 27, 2012 at 08:37:18AM -0500, Anthony Liguori wrote:
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-i386/cpu.c | 22 ++++++++++++++++++++++ 1 files changed, 22 insertions(+), 0 deletions(-)
diff --git a/target-i386/cpu.c b/target-i386/cpu.c index 6b9659f..b398439 100644 --- a/target-i386/cpu.c +++ b/target-i386/cpu.c @@ -28,6 +28,7 @@ #include "qemu-config.h"
#include "qapi/qapi-visit-core.h" +#include "qmp-commands.h"
#include "hyperv.h"
@@ -1123,6 +1124,27 @@ void x86_cpu_list(FILE *f, fprintf_function cpu_fprintf, const char *optarg) } }
+CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + x86_def_t *def; + + for (def = x86_defs; def; def = def->next) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(def->name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +}
How would the interface look like once we: - let libvirt know which features are available on each CPU model (libvirt needs that information[1]); and
I'm not sure I understand why libvirt needs this information. Can you elaborate?
I see two reasons:
- The libvirt API has functions to tell the user which features are going to be enabled for each CPU model, so it needs to know which features are enabled or not, for each machine-type + cpu-model combination, so this information can be reported proeprly.
Ok, step number one is that CPU 'features' need to be defined more formally. By formally, I mean via qapi-schema.json.
Then we can extend this command to return the set of features supported by each CPU type.
The first step will need to sort out how this maps across architectures.
- Also, if libvirt can enable/disable specific CPU features in the command-line, it just makes sens to know which ones are already enabled in each built-in CPU model.
- Probing for migration: libvirt needs to know if a given CPU model on a host can be migrated to another host. To know that, two pieces of information are needed: A) Which CPU features are visible to the guest for a specific configuration; B) Which of those features are really supported by the host hardware+kernel+QEMU, on the destination host, so it can know if migration is really possible.
Note that what QEMU thinks it exposes is not necessarily what gets exposed. KVM may mask additional features. How is this handled today?
No, what QEMU thinks it exposes actually is what gets exposed (and if it is not, it's a bug we have to fix). This is handled using the KVM GET_SUPPORTED_CPUID ioctl().
- add machine-type-specific cpudef compatibility changes?
I think we've discussed this in IRC. I don't think we need to worry about this.
I remember discussing a lot about the mechanism we will use to add the compatibility changes, but I don t know how the query API will look like, after we implement this mechanism.
0) User-defined CPU definitions go away - We already made a big step in this direction
1) CPU becomes a DeviceState
1.1) CPU models become classes
2) Features are expressed as properties
3) Same global mechanism used for everything else is used for CPUs
This is basically the compatibility mechanism we agreed upon, yes, but what about the probing mechanism to allow libvirt to know what will be the result of "-machine M -cpu C"[1] before actually starting a VM? [1] By "result" I mean: - Whether that combination can be run properly on that host; - Which CPU features will be visible to the guest in case it runs. Both items depend on CPU model _and_ machine-type, that's why we need some probing mechanism that depends on the machine-type or use the machine-type as input.
Regards,
Anthony Liguori
Would the command report different results depending on -machine?
No.
The problem is:
1) We need to introduce fixes on a CPU model that changes the set of guest-visible features (add or remove a feature)[1]; 2) The fix has to keep compatibility, so older machine-types will keep exposing the old set of gues-visible features; - That means different machine-types will have different CPU features being exposed. 3) libvirt needs to control/know which guest-visible CPU features are available to the guest (see above); 4) Because of (2), the querying system used by libvirt need to depend on the CPU model and machine-type.
[1] Example: The SandyBridge model today has the "tsc-deadline" bit set, but QEMU-1.1 did not expose the tsc-deadline feature properly because of incorrect expectations about the GET_SUPPORTED_CPUID ioctl. This was fixed on qemu-1.2.
That means "qemu-1.1 -machine pc-1.1 -cpu SandyBridge" does _not_ expose tsc-deadline to the guest, and we need to make "qemu-1.2 -machine pc-1.1 -cpu SandyBridge" _not_ expose it, too (otherwise migration from qemu-1.1 to qemu-1.2 will be broken).
Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself?
I'm not sure what you mean by compatibility data.
I mean any guest-visible compatibility bit that we will need to introduce on older machine-types, when making changes on CPU models (see the SandyBridge + tsc-deadline example above).
I see two options: - Libvirt queries for a [f(machine_type, cpu_model) -> cpu_features] function, that will take into account the machine-type-specific compatibility bits. - Libvirt queries for a [f(cpu_model) -> cpu_features] function and a [f(machine_type) -> compatibility_changes] function, and combine both. - I don't like this approach, I am just including it as a possible alternative.
Regards,
Anthony Liguori
[1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about.
-- Eduardo
-- Eduardo
-- Eduardo
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Aug 10, 2012 at 11:37:30AM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
- add machine-type-specific cpudef compatibility changes?
I think we've discussed this in IRC. I don't think we need to worry about this.
I remember discussing a lot about the mechanism we will use to add the compatibility changes, but I don t know how the query API will look like, after we implement this mechanism.
0) User-defined CPU definitions go away - We already made a big step in this direction
1) CPU becomes a DeviceState
1.1) CPU models become classes
2) Features are expressed as properties
3) Same global mechanism used for everything else is used for CPUs
This is basically the compatibility mechanism we agreed upon, yes, but what about the probing mechanism to allow libvirt to know what will be the result of "-machine M -cpu C"[1] before actually starting a VM?
I think that the requirement of "before actually starting a VM" is unreasonable. Presumably migration compatibility checking would happen after launching a guest so libvirt could surely delay querying the CPUID info until after the guest has started. There's a lot of logic involved in deciding what gets exposed to the guest. We don't really fully know until we've created the VCPU. It's a whole lot easier and saner to just create the VCPU. Regards, Anthony Liguori
[1] By "result" I mean: - Whether that combination can be run properly on that host; - Which CPU features will be visible to the guest in case it runs. Both items depend on CPU model _and_ machine-type, that's why we need some probing mechanism that depends on the machine-type or use the machine-type as input.
Regards,
Anthony Liguori
Would the command report different results depending on -machine?
No.
The problem is:
1) We need to introduce fixes on a CPU model that changes the set of guest-visible features (add or remove a feature)[1]; 2) The fix has to keep compatibility, so older machine-types will keep exposing the old set of gues-visible features; - That means different machine-types will have different CPU features being exposed. 3) libvirt needs to control/know which guest-visible CPU features are available to the guest (see above); 4) Because of (2), the querying system used by libvirt need to depend on the CPU model and machine-type.
[1] Example: The SandyBridge model today has the "tsc-deadline" bit set, but QEMU-1.1 did not expose the tsc-deadline feature properly because of incorrect expectations about the GET_SUPPORTED_CPUID ioctl. This was fixed on qemu-1.2.
That means "qemu-1.1 -machine pc-1.1 -cpu SandyBridge" does _not_ expose tsc-deadline to the guest, and we need to make "qemu-1.2 -machine pc-1.1 -cpu SandyBridge" _not_ expose it, too (otherwise migration from qemu-1.1 to qemu-1.2 will be broken).
Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself?
I'm not sure what you mean by compatibility data.
I mean any guest-visible compatibility bit that we will need to introduce on older machine-types, when making changes on CPU models (see the SandyBridge + tsc-deadline example above).
I see two options: - Libvirt queries for a [f(machine_type, cpu_model) -> cpu_features] function, that will take into account the machine-type-specific compatibility bits. - Libvirt queries for a [f(cpu_model) -> cpu_features] function and a [f(machine_type) -> compatibility_changes] function, and combine both. - I don't like this approach, I am just including it as a possible alternative.
Regards,
Anthony Liguori
[1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about.
-- Eduardo
-- Eduardo
-- Eduardo
On Fri, Aug 10, 2012 at 12:09:44PM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
On Fri, Aug 10, 2012 at 11:37:30AM -0500, Anthony Liguori wrote:
Eduardo Habkost <ehabkost@redhat.com> writes:
- add machine-type-specific cpudef compatibility changes?
I think we've discussed this in IRC. I don't think we need to worry about this.
I remember discussing a lot about the mechanism we will use to add the compatibility changes, but I don t know how the query API will look like, after we implement this mechanism.
0) User-defined CPU definitions go away - We already made a big step in this direction
1) CPU becomes a DeviceState
1.1) CPU models become classes
2) Features are expressed as properties
3) Same global mechanism used for everything else is used for CPUs
This is basically the compatibility mechanism we agreed upon, yes, but what about the probing mechanism to allow libvirt to know what will be the result of "-machine M -cpu C"[1] before actually starting a VM?
I think that the requirement of "before actually starting a VM" is unreasonable.
How is it unreasonable to expect an API where you can know what will be the results of an operation before actually running it? Maybe it doesn't fit in the beautiful and elegant model you are trying to push, but that doesn't make it unreasonable.
Presumably migration compatibility checking would happen after launching a guest so libvirt could surely delay querying the CPUID info until after the guest has started.
This is what I call unreasonable. A management layer needs to know if a host can run a VM before trying to migrate it, so the software or the user can take better decisions about migration before asking the VMs to be actually migrated. Note that I don't argue for every single CPUID bit to be available and queriable, but the (un)availability of some features need to be predictable.
There's a lot of logic involved in deciding what gets exposed to the guest. We don't really fully know until we've created the VCPU. It's a whole lot easier and saner to just create the VCPU.
If the logic is too complex and unpredictable, we have to make it clearer and more predictable. It's important to do so even if libvirt didn't need a probing interface, otherwise we would never be sure if the code is migration-safe.
Regards,
Anthony Liguori
[1] By "result" I mean: - Whether that combination can be run properly on that host; - Which CPU features will be visible to the guest in case it runs. Both items depend on CPU model _and_ machine-type, that's why we need some probing mechanism that depends on the machine-type or use the machine-type as input.
Regards,
Anthony Liguori
Would the command report different results depending on -machine?
No.
The problem is:
1) We need to introduce fixes on a CPU model that changes the set of guest-visible features (add or remove a feature)[1]; 2) The fix has to keep compatibility, so older machine-types will keep exposing the old set of gues-visible features; - That means different machine-types will have different CPU features being exposed. 3) libvirt needs to control/know which guest-visible CPU features are available to the guest (see above); 4) Because of (2), the querying system used by libvirt need to depend on the CPU model and machine-type.
[1] Example: The SandyBridge model today has the "tsc-deadline" bit set, but QEMU-1.1 did not expose the tsc-deadline feature properly because of incorrect expectations about the GET_SUPPORTED_CPUID ioctl. This was fixed on qemu-1.2.
That means "qemu-1.1 -machine pc-1.1 -cpu SandyBridge" does _not_ expose tsc-deadline to the guest, and we need to make "qemu-1.2 -machine pc-1.1 -cpu SandyBridge" _not_ expose it, too (otherwise migration from qemu-1.1 to qemu-1.2 will be broken).
Would the command return the latest cpudef without any machine-type hacks, and libvirt would have to query for the cpudef compatibility data for each machine-type and combine both pieces of information itself?
I'm not sure what you mean by compatibility data.
I mean any guest-visible compatibility bit that we will need to introduce on older machine-types, when making changes on CPU models (see the SandyBridge + tsc-deadline example above).
I see two options: - Libvirt queries for a [f(machine_type, cpu_model) -> cpu_features] function, that will take into account the machine-type-specific compatibility bits. - Libvirt queries for a [f(cpu_model) -> cpu_features] function and a [f(machine_type) -> compatibility_changes] function, and combine both. - I don't like this approach, I am just including it as a possible alternative.
Regards,
Anthony Liguori
[1] Note that it doesn't have to be low-level leaf-by-leaf register-by-register CPUID bits (I prefer a more high-level interface, myself), but it has to at least say "feature FOO is enabled/disabled" for a set of features libvirt cares about.
-- Eduardo
-- Eduardo
-- Eduardo
-- Eduardo
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> --- target-ppc/translate_init.c | 25 +++++++++++++++++++++++++ 1 files changed, 25 insertions(+), 0 deletions(-) diff --git a/target-ppc/translate_init.c b/target-ppc/translate_init.c index 5742229..7e025cb 100644 --- a/target-ppc/translate_init.c +++ b/target-ppc/translate_init.c @@ -10345,6 +10345,31 @@ void ppc_cpu_list (FILE *f, fprintf_function cpu_fprintf) } } +CpuDefInfoList *qmp_query_cpudefs(Error **errp) +{ + CpuDefInfoList *cpu_list = NULL; + int i; + + for (i = 0; i < ARRAY_SIZE(ppc_defs); i++) { + CpuDefInfoList *entry; + CpuDefInfo *info; + + if (!ppc_cpu_usable(&ppc_defs[i])) { + continue; + } + + info = g_malloc0(sizeof(*info)); + info->name = g_strdup(pcc_defs[i].name); + + entry = g_malloc0(sizeof(*entry)); + entry->value = info; + entry->next = cpu_list; + cpu_list = entry; + } + + return cpu_list; +} + /* CPUClass::reset() */ static void ppc_cpu_reset(CPUState *s) { -- 1.7.5.4
On Fri, 27 Jul 2012 08:37:12 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This series implements the necessary commands to implements danpb's idea to remove -help parsing in libvirt. We would introduce all of these commands in 1.2 and then change the -help output starting in 1.3.
I've reviewed this and apart from small details, it looks good to me. Would be nice to get an ack from libvirt folks before applying.
Here is Dan's plan from a previous thread:
<danpb>
Basically I'd sum up my new idea as "just use QMP".
* No new command line arguments like -capabilities
* libvirt invokes something like
$QEMUBINARY -qmp CHARDEV -nodefault -nodefconfig -nographics
* libvirt then runs a number of QMP commands to find out what it needs to know. I'd expect the following existing commands would be used
- query-version - already supported - query-commands - already supported - query-events - already supported - query-kvm - already supported - qom-{list,list-types,get} - already supported - query-spice/vnc - already supported
And add the following new commands
- query-devices - new, -device ?, and/or -device NAME,? data in QMP - query-machines - new, -M ? in QMP - query-cpu-types - new, -cpu ? in QMP
The above would take care of probably 50% of the current libvirt capabilities probing, including a portion of the -help stuff. Then there is all the rest of the crap we detect from the -help. We could just take the view, that "as of 1.2", we assume everything we previously detected is just available by default, and thus don't need to probe it. For stuff that is QOM based, I expect we'll be able to detect new features in the future using the qom-XXX monitor commands. For stuff that is non-qdev, and non-qom, libvirt can just do a plain version number check, unless we decide there is specific info worth exposing via other new 'query-XXX' monitor commands. Basically I'd sum up my new idea as "just use QMP".
* No new command line arguments like -capabilities
* libvirt invokes something like
$QEMUBINARY -qmp CHARDEV -nodefault -nodefconfig -nographics
* libvirt then runs a number of QMP commands to find out what it needs to know. I'd expect the following existing commands would be used
- query-version - already supported - query-commands - already supported - query-events - already supported - query-kvm - already supported - qom-{list,list-types,get} - already supported - query-spice/vnc - already supported
And add the following new commands
- query-devices - new, -device ?, and/or -device NAME,? data in QMP - query-machines - new, -M ? in QMP - query-cpu-types - new, -cpu ? in QMP
The above would take care of probably 50% of the current libvirt capabilities probing, including a portion of the -help stuff. Then there is all the rest of the crap we detect from the -help. We could just take the view, that "as of 1.2", we assume everything we previously detected is just available by default, and thus don't need to probe it. For stuff that is QOM based, I expect we'll be able to detect new features in the future using the qom-XXX monitor commands. For stuff that is non-qdev, and non-qom, libvirt can just do a plain version number check, unless we decide there is specific info worth exposing via other new 'query-XXX' monitor commands.
</danpb>
The one thing to note is that I didn't add a query-devices command because you can already do:
qmp query-devices --implements=device --abstract=False
To get the equivalent output of -device ?. Instead, I added a command to list specific properties of a device which is the equivalent of -device FOO,?
On Fri, Jul 27, 2012 at 01:21:01PM -0300, Luiz Capitulino wrote:
On Fri, 27 Jul 2012 08:37:12 -0500 Anthony Liguori <aliguori@us.ibm.com> wrote:
This series implements the necessary commands to implements danpb's idea to remove -help parsing in libvirt. We would introduce all of these commands in 1.2 and then change the -help output starting in 1.3.
I've reviewed this and apart from small details, it looks good to me.
Would be nice to get an ack from libvirt folks before applying.
I think what Anthony has posted looks workable for libvirt. The proof will be in the implementation of course. So we'll aim to get libvirt proof of concept working with this approach asap, and give feedback on what, if anything, was missing so we can get it all finalized before 1.2 GA. Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
participants (7)
-
Anthony Liguori -
Blue Swirl -
Daniel P. Berrange -
Eduardo Habkost -
Eric Blake -
Luiz Capitulino -
Peter Maydell