s390x: CPU Topology
v4 cover s390x: CPU Topology
Hi,
This series is a first part of the implementation of CPU topology for S390 greatly reduced from the first spin.
In particular, we reduced the scope to the S390x specificities, removing all code touching to SMP or NUMA, with the goal to:
- facilitate review and acceptance
- let for later the SMP part currently actively discussed in mainline
- be able despite the reduction of code to handle CPU topology for S390 using the current S390 topology provided by QEMU with cores and sockets only.
To use these patches, you will need the Linux series version 4. You find it there:
https://lkml.org/lkml/2021/9/16/576
Currently this code is for KVM only, I have no idea if it is interesting to provide a TCG patch. If ever it will be done in another series.
A short introduction
====================
CPU Topology is described in the S390 POP with essentially the description of two instructions:
PTF Perform Topology function used to poll for topology change
and used to set the polarization but this part is not part of this item.
STSI Store System Information and the SYSIB 15.1.x providing the Topology
configuration.
S390 Topology is a 6 levels hierarchical topology with up to 5 level
of containers. The last topology level, specifying the CPU cores.
This patch series only uses the two lower levels sockets and cores.
To get the information on the topology, S390 provides the STSI instruction, which stores a structures providing the list of the containers used in the Machine topology: the SYSIB.
A selector within the STSI instruction allow to chose how many topology levels will be provide in the SYSIB.
Using the Topology List Entries (TLE) provided inside the SYSIB we the Linux kernel is able to compute the information about the cache distance between two cores and can use this information to take scheduling decisions.
Note:
-----
Z15 reports 3 levels of containers, drawers, book, sockets as Container-TLEs above the core description inside CPU-TLEs.
The Topology can be seen at several places inside zLinux:
- sysfs: /sys/devices/system/cpu/cpuX/topology
- procfs: /proc/sysinfo and /proc/cpuinfo
- lscpu -e : gives toplogy information
The different Topology levels have names:
- Node - Drawer - Book - sockets or physical package - core
Threads:
Multithreading, is not part of the topology as described by the SYSIB 15.1.x
The interest of the guest to know the CPU topology is obviously to be able to optimise the load balancing and the migration of threads. KVM will have the same interest concerning vCPUs scheduling and cache optimisation.
The design
==========
1) To be ready for hotplug, I chose an Object oriented design of the topology containers:
- A node is a bridge on the SYSBUS and defines a "node bus"
- A drawer is hotplug on the "node bus"
- A book on the "drawer bus"
- A socket on the "book bus"
- And the CPU Topology List Entry (CPU-TLE)sits on the socket bus. These objects will be enhanced with the cache information when NUMA is implemented.
This also allows for easy retrieval when building the different SYSIB for Store Topology System Information (STSI)
2) Perform Topology Function (PTF) instruction is made available to the guest with a new KVM capability and intercepted in QEMU, allowing the guest to pool for topology changes.
Features and TBD list
=====================
- There is no direct match between IDs shown by:
- lscpu (unrelated numbered list),
- SYSIB 15.1.x (topology ID)
- The CPU number, left column of lscpu, is used to reference a CPU by Linux tools
While the CPU address is used by QEMU for hotplug.
- Effect of -smp parsing on the topology with an example:
-smp 9,sockets=4,cores=4,maxcpus=16
We have 4 socket each holding 4 cores so that we have a maximum of 16 CPU, 9 of them are active on boot. (Should be obvious)
# lscpu -e CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2d:L2i ONLINE CONFIGURED POLARIZATION ADDRESS 0 0 0 0 0 0 0:0:0:0 yes yes horizontal 0 1 0 0 0 0 1 1:1:1:1 yes yes horizontal 1 2 0 0 0 0 2 2:2:2:2 yes yes horizontal 2 3 0 0 0 0 3 3:3:3:3 yes yes horizontal 3 4 0 0 0 1 4 4:4:4:4 yes yes horizontal 4 5 0 0 0 1 5 5:5:5:5 yes yes horizontal 5 6 0 0 0 1 6 6:6:6:6 yes yes horizontal 6 7 0 0 0 1 7 7:7:7:7 yes yes horizontal 7 8 0 0 0 2 8 8:8:8:8 yes yes horizontal 8 #
- To plug a new CPU inside the topology one can simply use the CPU address like in:
(qemu) device_add host-s390x-cpu,core-id=12
# lscpu -e CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2d:L2i ONLINE CONFIGURED POLARIZATION ADDRESS 0 0 0 0 0 0 0:0:0:0 yes yes horizontal 0 1 0 0 0 0 1 1:1:1:1 yes yes horizontal 1 2 0 0 0 0 2 2:2:2:2 yes yes horizontal 2 3 0 0 0 0 3 3:3:3:3 yes yes horizontal 3 4 0 0 0 1 4 4:4:4:4 yes yes horizontal 4 5 0 0 0 1 5 5:5:5:5 yes yes horizontal 5 6 0 0 0 1 6 6:6:6:6 yes yes horizontal 6 7 0 0 0 1 7 7:7:7:7 yes yes horizontal 7 8 0 0 0 2 8 8:8:8:8 yes yes horizontal 8 9 - - - - - ::: no yes horizontal 12 # chcpu -e 9 CPU 9 enabled # lscpu -e CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2d:L2i ONLINE CONFIGURED POLARIZATION ADDRESS 0 0 0 0 0 0 0:0:0:0 yes yes horizontal 0 1 0 0 0 0 1 1:1:1:1 yes yes horizontal 1 2 0 0 0 0 2 2:2:2:2 yes yes horizontal 2 3 0 0 0 0 3 3:3:3:3 yes yes horizontal 3 4 0 0 0 1 4 4:4:4:4 yes yes horizontal 4 5 0 0 0 1 5 5:5:5:5 yes yes horizontal 5 6 0 0 0 1 6 6:6:6:6 yes yes horizontal 6 7 0 0 0 1 7 7:7:7:7 yes yes horizontal 7 8 0 0 0 2 8 8:8:8:8 yes yes horizontal 8 9 0 0 0 3 9 9:9:9:9 yes yes horizontal 12 #
It is up to the admin level, Libvirt for example, to pin the righ CPU to the right vCPU, but as we can see without NUMA, chosing separate sockets for CPUs is not easy without hotplug because without information the code will assign the vCPU and fill the sockets one after the other.
Note that this is also the default behavior on the LPAR.
Conclusion
==========
This patch, together with the associated KVM patch allows to provide CPU topology information to the guest.
Currently, only dedicated vCPU and CPU are supported and a NUMA topology can only be handled using CPU hotplug inside the guest.
Next extensions are to provide:
- adding books and drawers levels
- NUMA using the -numa QEMU parameter.
- Topology information change for shared CPU
Regards,
Pierre
Pierre Morel (5):
linux-headers update
s390x: topology: CPU topology objects and structures
s390x: topology: implementating Store Topology System Information
s390x: CPU topology: CPU topology migration
s390x: kvm: topology: interception of PTF instruction
hw/s390x/cpu-topology.c | 361 ++++++++++++++++++++++++++++ hw/s390x/meson.build | 1 + hw/s390x/s390-virtio-ccw.c | 54 +++++ include/hw/s390x/cpu-topology.h | 74 ++++++ include/hw/s390x/s390-virtio-ccw.h | 6 + linux-headers/linux/kvm.h | 1 + target/s390x/cpu.h | 50 ++++ target/s390x/cpu_features_def.h.inc | 1 + target/s390x/cpu_models.c | 2 + target/s390x/cpu_topology.c | 113 +++++++++ target/s390x/gen-features.c | 3 + target/s390x/kvm/kvm.c | 26 ++ target/s390x/machine.c | 48 ++++ target/s390x/meson.build | 1 + 14 files changed, 741 insertions(+) create mode 100644 hw/s390x/cpu-topology.c create mode 100644 include/hw/s390x/cpu-topology.h create mode 100644 target/s390x/cpu_topology.c
Signature · 20 lines
2.27.0 Changelog: - since v3 - Added migration (Thomas) - Separated STSI instruction from KVM to prepare TCG (Thomas) - Take care of endianess to prepare TCG (Thomas) - Added comments on STSI CPU container and PFT instruction (Thomas) - Moved enabling the instructions as the last patch (Thomas)
PATCH v4 1/5 linux-headers update
--- linux-headers/linux/kvm.h | 1 + 1 file changed, 1 insertion(+)
linux-headers/linux/kvm.h +1 -0
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {#define KVM_CAP_BINARY_STATS_FD 203#define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204#define KVM_CAP_ARM_MTE 205 +#define KVM_CAP_S390_CPU_TOPOLOGY 206#ifdef KVM_CAP_IRQ_ROUTING
Signature · 1 line
2.27.0
PATCH v4 2/5 s390x: topology: CPU topology objects and structures
In reply to Pierre MorelWe use new objects to have a dynamic administration of the CPU topology. The highest level object in this implementation is the s390 book and in this first implementation of CPU topology for S390 we have a single book.
The book is built as a SYSBUS bridge during the CPU initialization.
Every object under this single book will be build dynamically immediately after a CPU has be realized if it is needed. The CPU will fill the sockets once after the other, according to the number of core per socket defined during the smp parsing.
Each CPU inside a socket will be represented by a bit in a 64bit unsigned long. Set on plug and clear on unplug of a CPU.
--- hw/s390x/cpu-topology.c | 361 ++++++++++++++++++++++++++++++++ hw/s390x/meson.build | 1 + hw/s390x/s390-virtio-ccw.c | 4 + include/hw/s390x/cpu-topology.h | 74 +++++++ target/s390x/cpu.h | 47 +++++ 5 files changed, 487 insertions(+) create mode 100644 hw/s390x/cpu-topology.c create mode 100644 include/hw/s390x/cpu-topology.h
hw/s390x/cpu-topology.c new file +361 -0
@@ -0,0 +1,361 @@ +/* +* CPU Topology +* +* Copyright 2021 IBM Corp. +* Author(s): Pierre Morel <[email protected]> + +* This work is licensed under the terms of the GNU GPL, version 2 or (at +* your option) any later version. See the COPYING file in the top-level +* directory. +*/ + +#include "qemu/osdep.h" +#include "qapi/error.h" +#include "qemu/error-report.h" +#include "hw/sysbus.h" +#include "hw/s390x/cpu-topology.h" +#include "hw/qdev-properties.h" +#include "hw/boards.h" +#include "qemu/typedefs.h" +#include "target/s390x/cpu.h" +#include "hw/s390x/s390-virtio-ccw.h" + +static S390TopologyCores *s390_create_cores (S390TopologySocket *socket, +int origin) +{ +DeviceState *dev; +S390TopologyCores *cores; +const MachineState *ms =MACHINE (qdev_get_machine ()); + +if (socket->bus->num_children >= ms->smp.cores) { +return NULL ; + } + + dev =qdev_new (TYPE_S390_TOPOLOGY_CORES ); +qdev_realize_and_unref (dev, socket->bus, &error_fatal); + + cores =S390_TOPOLOGY_CORES (dev); + cores->origin = origin; + socket->cnt +=1 ; + +return cores; +} + +static S390TopologySocket *s390_create_socket (S390TopologyBook *book,int id) +{ +DeviceState *dev; +S390TopologySocket *socket; +const MachineState *ms =MACHINE (qdev_get_machine ()); + +if (book->bus->num_children >= ms->smp.sockets) { +return NULL ; + } + + dev =qdev_new (TYPE_S390_TOPOLOGY_SOCKET ); +qdev_realize_and_unref (dev, book->bus, &error_fatal); + + socket =S390_TOPOLOGY_SOCKET (dev); + socket->socket_id = id; + book->cnt++; + +return socket; +} + +/* +* s390_get_cores: +* @socket: the socket to search into +* @origin: the origin specified for the S390TopologyCores +* +* returns a pointer to a S390TopologyCores structure within a socket having +* the specified origin. +* First search if the socket is already containing the S390TopologyCores +* structure and if not create one with this origin. +*/ +static S390TopologyCores *s390_get_cores (S390TopologySocket *socket,int origin) +{ +S390TopologyCores *cores; +BusChild *kid; + + QTAILQ_FOREACH(kid, &socket->bus->children, sibling) { + cores =S390_TOPOLOGY_CORES (kid->child); +if (cores->origin == origin) { +return cores; + } + } +return s390_create_cores (socket, origin); +} + +/* +* s390_get_socket: +* @book: The book to search into +* @socket_id: the identifier of the socket to search for +* +* returns a pointer to a S390TopologySocket structure within a book having +* the specified socket_id. +* First search if the book is already containing the S390TopologySocket +* structure and if not create one with this socket_id. +*/ +static S390TopologySocket *s390_get_socket (S390TopologyBook *book, +int socket_id) +{ +S390TopologySocket *socket; +BusChild *kid; + + QTAILQ_FOREACH(kid, &book->bus->children, sibling) { + socket =S390_TOPOLOGY_SOCKET (kid->child); +if (socket->socket_id == socket_id) { +return socket; + } + } +return s390_create_socket (book, socket_id); +} + +/* +* s390_topology_new_cpu: +* @core_id: the core ID is machine wide +* +* We have a single book returned by s390_get_topology(), +* then we build the hierarchy on demand. +* Note that we do not destroy the hierarchy on error creating +* an entry in the topology, we just keep it empty. +* We do not need to worry about not finding a topology level +* entry this would have been caught during smp parsing. +*/ +void s390_topology_new_cpu (int core_id) +{ +const MachineState *ms =MACHINE (qdev_get_machine ()); +S390TopologyBook *book; +S390TopologySocket *socket; +S390TopologyCores *cores; +int cores_per_socket, sock_idx; +int origin, bit; + + book =s390_get_topology (); + + cores_per_socket = ms->smp.max_cpus / ms->smp.sockets; + + sock_idx = (core_id / cores_per_socket); + socket =s390_get_socket (book, sock_idx); + +/* +* At the core level, each CPU is represented by a bit in a 64bit +* unsigned long. Set on plug and clear on unplug of a CPU. +* The firmware assume that all CPU in the core description have the same +* type, polarization and are all dedicated or shared. +* In the case a socket contains CPU with different type, polarization +* or dedication then they will be defined in different CPU containers. +* Currently we assume all CPU are identical and the only reason to have +* several S390TopologyCores inside a socket is to have more than 64 CPUs +* in that case the origin field, representing the offset of the first CPU +* in the CPU container allows to represent up to the maximal number of +* CPU inside several CPU containers inside the socket container. +*/ + origin =64 * (core_id /64 ); + + cores =s390_get_cores (socket, origin); + + bit =63 - (core_id - origin); +set_bit (bit, &cores->mask); + cores->origin = origin; +} + +/* +* Setting the first topology: 1 book, 1 socket +* This is enough for 64 cores if the topology is flat (single socket) +*/ +void s390_topology_setup (MachineState *ms) +{ +DeviceState *dev; + +/* Create BOOK bridge device */ + dev =qdev_new (TYPE_S390_TOPOLOGY_BOOK ); +object_property_add_child (qdev_get_machine (), +TYPE_S390_TOPOLOGY_BOOK ,OBJECT (dev)); +sysbus_realize_and_unref (SYS_BUS_DEVICE (dev), &error_fatal); +} + +S390TopologyBook *s390_get_topology (void ) +{ +static S390TopologyBook *book; + +if (!book) { + book =S390_TOPOLOGY_BOOK ( +object_resolve_path (TYPE_S390_TOPOLOGY_BOOK ,NULL )); +assert (book !=NULL ); + } + +return book; +} + +/* --- CORES Definitions --- */ + +static Property s390_topology_cores_properties[] = { +DEFINE_PROP_BOOL ("dedicated" , S390TopologyCores, dedicated, false), +DEFINE_PROP_UINT8 ("polarity" , S390TopologyCores, polarity, +S390_TOPOLOGY_POLARITY_H ), +DEFINE_PROP_UINT8 ("cputype" , S390TopologyCores, cputype, +S390_TOPOLOGY_CPU_TYPE ), +DEFINE_PROP_UINT16 ("origin" , S390TopologyCores, origin,0 ), +DEFINE_PROP_UINT64 ("mask" , S390TopologyCores, mask,0 ), +DEFINE_PROP_UINT8 ("id" , S390TopologyCores, id,0 ), +DEFINE_PROP_END_OF_LIST (), +}; + +static void cpu_cores_class_init (ObjectClass *oc,void *data) +{ +DeviceClass *dc =DEVICE_CLASS (oc); +HotplugHandlerClass *hc =HOTPLUG_HANDLER_CLASS (oc); + +device_class_set_props (dc, s390_topology_cores_properties); + hc->unplug = qdev_simple_device_unplug_cb; + dc->bus_type =TYPE_S390_TOPOLOGY_SOCKET_BUS ; + dc->desc ="topology cpu entry" ; +} + +static const TypeInfo cpu_cores_info = { + .name =TYPE_S390_TOPOLOGY_CORES , + .parent =TYPE_DEVICE , + .instance_size =sizeof (S390TopologyCores), + .class_init = cpu_cores_class_init, + .interfaces = (InterfaceInfo []) { + {TYPE_HOTPLUG_HANDLER }, + { } + } +}; + +/* --- SOCKETS Definitions --- */ +static Property s390_topology_socket_properties[] = { +DEFINE_PROP_UINT8 ("socket_id" , S390TopologySocket, socket_id,0 ), +DEFINE_PROP_END_OF_LIST (), +}; + +static char *socket_bus_get_dev_path (DeviceState *dev) +{ +S390TopologySocket *socket =S390_TOPOLOGY_SOCKET (dev); +DeviceState *book = dev->parent_bus->parent; +char *id =qdev_get_dev_path (book); +char *ret; + +if (id) { + ret =g_strdup_printf ("%s:%02d" , id, socket->socket_id); +g_free (id); + }else { + ret =g_strdup_printf ("_:%02d" , socket->socket_id); + } + +return ret; +} + +static void socket_bus_class_init (ObjectClass *oc,void *data) +{ +BusClass *k =BUS_CLASS (oc); + + k->get_dev_path = socket_bus_get_dev_path; + k->max_dev =S390_MAX_SOCKETS ; +} + +static const TypeInfo socket_bus_info = { + .name =TYPE_S390_TOPOLOGY_SOCKET_BUS , + .parent =TYPE_BUS , + .instance_size =0 , + .class_init = socket_bus_class_init, +}; + +static void s390_socket_device_realize (DeviceState *dev,Error **errp) +{ +S390TopologySocket *socket =S390_TOPOLOGY_SOCKET (dev); +BusState *bus; + + bus =qbus_new (TYPE_S390_TOPOLOGY_SOCKET_BUS , dev, +TYPE_S390_TOPOLOGY_SOCKET_BUS ); +qbus_set_hotplug_handler (bus,OBJECT (dev)); + socket->bus = bus; +} + +static void socket_class_init (ObjectClass *oc,void *data) +{ +DeviceClass *dc =DEVICE_CLASS (oc); +HotplugHandlerClass *hc =HOTPLUG_HANDLER_CLASS (oc); + + hc->unplug = qdev_simple_device_unplug_cb; +set_bit (DEVICE_CATEGORY_BRIDGE , dc->categories); + dc->bus_type =TYPE_S390_TOPOLOGY_BOOK_BUS ; + dc->realize = s390_socket_device_realize; +device_class_set_props (dc, s390_topology_socket_properties); + dc->desc ="topology socket" ; +} + +static const TypeInfo socket_info = { + .name =TYPE_S390_TOPOLOGY_SOCKET , + .parent =TYPE_DEVICE , + .instance_size =sizeof (S390TopologySocket), + .class_init = socket_class_init, + .interfaces = (InterfaceInfo []) { + {TYPE_HOTPLUG_HANDLER }, + { } + } +}; + +static char *book_bus_get_dev_path (DeviceState *dev) +{ +return g_strdup_printf ("00" ); +} + +static void book_bus_class_init (ObjectClass *oc,void *data) +{ +BusClass *k =BUS_CLASS (oc); + + k->get_dev_path = book_bus_get_dev_path; + k->max_dev =S390_MAX_BOOKS ; +} + +static const TypeInfo book_bus_info = { + .name =TYPE_S390_TOPOLOGY_BOOK_BUS , + .parent =TYPE_BUS , + .instance_size =0 , + .class_init = book_bus_class_init, +}; + +static void s390_book_device_realize (DeviceState *dev,Error **errp) +{ +S390TopologyBook *book =S390_TOPOLOGY_BOOK (dev); +BusState *bus; + + bus =qbus_new (TYPE_S390_TOPOLOGY_BOOK_BUS , dev, +TYPE_S390_TOPOLOGY_BOOK_BUS ); +qbus_set_hotplug_handler (bus,OBJECT (dev)); + book->bus = bus; +} + +static void book_class_init (ObjectClass *oc,void *data) +{ +DeviceClass *dc =DEVICE_CLASS (oc); +HotplugHandlerClass *hc =HOTPLUG_HANDLER_CLASS (oc); + + hc->unplug = qdev_simple_device_unplug_cb; +set_bit (DEVICE_CATEGORY_BRIDGE , dc->categories); + dc->realize = s390_book_device_realize; + dc->desc ="topology book" ; +} + +static const TypeInfo book_info = { + .name =TYPE_S390_TOPOLOGY_BOOK , + .parent =TYPE_SYS_BUS_DEVICE , + .instance_size =sizeof (S390TopologyBook), + .class_init = book_class_init, + .interfaces = (InterfaceInfo []) { + {TYPE_HOTPLUG_HANDLER }, + { } + } +}; + +static void topology_register (void ) +{ +type_register_static (&cpu_cores_info); +type_register_static (&socket_bus_info); +type_register_static (&socket_info); +type_register_static (&book_bus_info); +type_register_static (&book_info); +} + +type_init (topology_register);
hw/s390x/meson.build +1 -0
@@ -2,6 +2,7 @@ s390x_ss = ss.source_set() s390x_ss.add(files( 'ap-bridge.c', 'ap-device.c', + 'cpu-topology.c', 'ccw-device.c', 'css-bridge.c', 'css.c',
hw/s390x/s390-virtio-ccw.c +4 -0
@@ -42,6 +42,7 @@#include "sysemu/sysemu.h" #include "hw/s390x/pv.h" #include "migration/blocker.h" +#include "hw/s390x/cpu-topology.h" static Error *pv_mig_blocker;
@@ -88,6 +89,7 @@ static void s390_init_cpus(MachineState *machine)/* initialize possible_cpus */ mc->possible_cpu_arch_ids (machine); +s390_topology_setup (machine);for (i =0 ; i < machine->smp.cpus; i++) {s390x_new_cpu (machine->cpu_type, i, &error_fatal); }
@@ -305,6 +307,8 @@ static void s390_cpu_plug(HotplugHandler *hotplug_dev,g_assert (!ms->possible_cpus->cpus[cpu->env.core_id].cpu); ms->possible_cpus->cpus[cpu->env.core_id].cpu =OBJECT (dev); +s390_topology_new_cpu (cpu->env.core_id); +if (dev->hotplugged) {raise_irq_cpu_hotplug (); }
include/hw/s390x/cpu-topology.h new file +74 -0
74 lines not on this page · raw message
target/s390x/cpu.h +47 -0
53 lines not on this page · raw message
Signature · 1 line
2.27.0
PATCH v4 3/5 s390x: topology: implementating Store Topology System Information
In reply to Pierre MorelThe handling of STSI is enhanced with the interception of the function code 15 for storing CPU topology.
Using the objects built during the pluging of CPU, we build the SYSIB 15_1_x structures.
With this patch the maximum MNEST level is 2, this is also the only level allowed and only SYSIB 15_1_2 will be built.
--- target/s390x/cpu.h | 1 + target/s390x/cpu_topology.c | 113 ++++++++++++++++++++++++++++++++++++ target/s390x/kvm/kvm.c | 5 ++ target/s390x/meson.build | 1 + 4 files changed, 120 insertions(+) create mode 100644 target/s390x/cpu_topology.c
target/s390x/cpu.h +1 -0
@@ -897,4 +897,5 @@ typedef S390CPU ArchCPU;#include "exec/cpu-all.h" +void insert_stsi_15_1_x (S390CPU *cpu,int sel2,__u64 addr,uint8_t ar);#endif
target/s390x/cpu_topology.c new file +113 -0
113 lines not on this page · raw message
target/s390x/kvm/kvm.c +5 -0
@@ -52,6 +52,7 @@#include "hw/s390x/s390-virtio-ccw.h" #include "hw/s390x/s390-virtio-hcall.h" #include "hw/s390x/pv.h" +#include "hw/s390x/cpu-topology.h" #ifndef DEBUG_KVM #define DEBUG_KVM 0
@@ -1906,6 +1907,10 @@ static int handle_stsi(S390CPU *cpu)/* Only sysib 3.2.2 needs post-handling for now. */ insert_stsi_3_2_2 (cpu, run->s390_stsi.addr, run->s390_stsi.ar);return 0 ; +case 15 : +insert_stsi_15_1_x (cpu, run->s390_stsi.sel2, run->s390_stsi.addr, + run->s390_stsi.ar); +return 0 ;default :return 0 ; }
target/s390x/meson.build +1 -0
@@ -29,6 +29,7 @@ s390x_softmmu_ss.add(files( 'sigp.c', 'cpu-sysemu.c', 'cpu_models_sysemu.c', + 'cpu_topology.c', )) s390x_user_ss = ss.source_set()
Signature · 1 line
2.27.0
PATCH v4 4/5 s390x: CPU topology: CPU topology migration
In reply to Pierre MorelBoth source and target must have the same configuration regarding the activation of Perform Topology Function and Store Topology System Information.
--- target/s390x/cpu.h | 2 ++ target/s390x/cpu_features_def.h.inc | 1 + target/s390x/cpu_models.c | 2 ++ target/s390x/gen-features.c | 3 ++ target/s390x/kvm/kvm.c | 6 ++++ target/s390x/machine.c | 48 +++++++++++++++++++++++++++++ 6 files changed, 62 insertions(+)
target/s390x/cpu.h +2 -0
@@ -150,6 +150,8 @@ struct CPUS390XState {/* currently processed sigp order */ uint8_t sigp_order; +/* Using Perform CPU Topology Function*/ +bool using_ptf; };static inline uint64_t *get_freg (CPUS390XState *cs,int nr)
target/s390x/cpu_features_def.h.inc +1 -0
@@ -146,6 +146,7 @@ DEF_FEAT(SIE_CEI, "cei", SCLP_CPU, 43, "SIE: Conditional-external-interception f DEF_FEAT(DAT_ENH_2, "dateh2", MISC, 0, "DAT-enhancement facility 2") DEF_FEAT(CMM, "cmm", MISC, 0, "Collaborative-memory-management facility") DEF_FEAT(AP, "ap", MISC, 0, "AP instructions installed") +DEF_FEAT(CPU_TOPOLOGY, "cpu_topology", MISC, 0, "CPU Topology available") /* Features exposed via the PLO instruction. */ DEF_FEAT(PLO_CL, "plo-cl", PLO, 0, "PLO Compare and load (32 bit in general registers)")
target/s390x/cpu_models.c +2 -0
@@ -238,6 +238,7 @@ bool s390_has_feat(S390Feat feat)if (s390_is_pv ()) {switch (feat) { +case S390_FEAT_CPU_TOPOLOGY :case S390_FEAT_DIAG_318 :case S390_FEAT_HPMA2 :case S390_FEAT_SIE_F2 :
@@ -467,6 +468,7 @@ static void check_consistency(const S390CPUModel *model) { S390_FEAT_DIAG_318, S390_FEAT_EXTENDED_LENGTH_SCCB }, { S390_FEAT_NNPA, S390_FEAT_VECTOR }, { S390_FEAT_RDP, S390_FEAT_LOCAL_TLB_CLEARING }, + { S390_FEAT_CPU_TOPOLOGY, S390_FEAT_CONFIGURATION_TOPOLOGY }, };int i;
target/s390x/gen-features.c +3 -0
@@ -488,6 +488,7 @@ static uint16_t full_GEN9_GA3[] = {static uint16_t full_GEN10_GA1[] = {S390_FEAT_EDAT ,S390_FEAT_CONFIGURATION_TOPOLOGY , +S390_FEAT_CPU_TOPOLOGY ,S390_FEAT_GROUP_MSA_EXT_2 ,S390_FEAT_ESOP ,S390_FEAT_SIE_PFMFI ,
@@ -604,6 +605,8 @@ static uint16_t default_GEN9_GA1[] = {static uint16_t default_GEN10_GA1[] = {S390_FEAT_EDAT ,S390_FEAT_GROUP_MSA_EXT_2 , +S390_FEAT_CONFIGURATION_TOPOLOGY , +S390_FEAT_CPU_TOPOLOGY , };#define default_GEN10_GA2 EmptyFeat
target/s390x/kvm/kvm.c +6 -0
@@ -612,6 +612,7 @@ int kvm_arch_put_registers(CPUState *cs, int level) }else {/* prefix is only supported via sync regs */ } +return 0 ; }
@@ -2427,6 +2428,10 @@ void kvm_s390_get_host_cpu_model(S390CPUModel *model, Error **errp)clear_bit (S390_FEAT_CMM_NT , model->features); } +if (kvm_check_extension (kvm_state,KVM_CAP_S390_CPU_TOPOLOGY )) { +set_bit (S390_FEAT_CPU_TOPOLOGY , model->features); + } +/* bpb needs kernel support for migration, VSIE and reset */ if (!kvm_check_extension (kvm_state,KVM_CAP_S390_BPB )) {clear_bit (S390_FEAT_BPB , model->features);
@@ -2535,6 +2540,7 @@ void kvm_s390_apply_cpu_model(const S390CPUModel *model, Error **errp)error_setg (errp,"KVM: Error configuring CPU subfunctions: %d" , rc);return ; } +/* enable CMM via CMMA */ if (test_bit (S390_FEAT_CMM , model->features)) {kvm_s390_enable_cmma ();
target/s390x/machine.c +48 -0
60 lines not on this page · raw message
Signature · 1 line
2.27.0
PATCH v4 5/5 s390x: kvm: topology: interception of PTF instruction
In reply to Pierre MorelWhen the host supports the CPU topology facility, the PTF instruction with function code 2 is interpreted by the SIE, provided that the userland hypervizor activates the interpretation by using the KVM_CAP_S390_CPU_TOPOLOGY KVM extension.
The PTF instructions with function code 0 and 1 are intercepted and must be emulated by the userland hypervizor.
--- hw/s390x/s390-virtio-ccw.c | 50 ++++++++++++++++++++++++++++++ include/hw/s390x/s390-virtio-ccw.h | 6 ++++ target/s390x/kvm/kvm.c | 15 +++++++++ 3 files changed, 71 insertions(+)
hw/s390x/s390-virtio-ccw.c +50 -0
@@ -408,6 +408,56 @@ static void s390_pv_prepare_reset(S390CcwMachineState *ms)s390_pv_prep_reset (); } +/* +* s390_handle_ptf: +* +* @register 1: contains the function code +* +* Function codes 0 and 1 handle the CPU polarization. +* We assume an horizontal topology, the only one supported currently +* by Linux, consequently we answer to function code 0, requesting +* horizontal polarization that it is already the current polarization +* and reject vertical polarization request without further explanation. +* +* Function code 2 is handling topology changes and is interpreted +* by the SIE. +*/ +int s390_handle_ptf (S390CPU *cpu,uint8_t r1,uintptr_t ra) +{ +CPUS390XState *env = &cpu->env; +uint64_t reg = env->regs[r1]; +uint8_t fc = reg &S390_TOPO_FC_MASK ; + +if (!s390_has_feat (S390_FEAT_CONFIGURATION_TOPOLOGY )) { +s390_program_interrupt (env,PGM_OPERATION , ra); +return 0 ; + } + +if (env->psw.mask &PSW_MASK_PSTATE ) { +s390_program_interrupt (env,PGM_PRIVILEGED , ra); +return 0 ; + } + +if (reg & ~S390_TOPO_FC_MASK ) { +s390_program_interrupt (env,PGM_SPECIFICATION , ra); +return 0 ; + } + +switch (fc) { +case 0 :/* Horizontal polarization is already set */ + env->regs[r1] |=S390_PTF_REASON_DONE ; +return 2 ; +case 1 :/* Vertical polarization is not supported */ + env->regs[r1] |=S390_PTF_REASON_NONE ; +return 2 ; +default : +/* Note that fc == 2 is interpreted by the SIE */ +s390_program_interrupt (env,PGM_SPECIFICATION , ra); + } + +return 0 ; +} +static void s390_machine_reset (MachineState *machine) {S390CcwMachineState *ms =S390_CCW_MACHINE (machine);
include/hw/s390x/s390-virtio-ccw.h +6 -0
@@ -30,6 +30,12 @@ struct S390CcwMachineState {uint8_t loadparm[8 ]; }; +#define S390_PTF_REASON_NONE (0x00 << 8) +#define S390_PTF_REASON_DONE (0x01 << 8) +#define S390_PTF_REASON_BUSY (0x02 << 8) +#define S390_TOPO_FC_MASK 0xffUL +int s390_handle_ptf (S390CPU *cpu,uint8_t r1,uintptr_t ra); +struct S390CcwMachineClass {/*< private >*/ MachineClass parent_class;
target/s390x/kvm/kvm.c +15 -0
@@ -98,6 +98,7 @@#define PRIV_B9_EQBS 0x9c#define PRIV_B9_CLP 0xa0 +#define PRIV_B9_PTF 0xa2#define PRIV_B9_PCISTG 0xd0#define PRIV_B9_PCILG 0xd2#define PRIV_B9_RPCIT 0xd3
@@ -363,6 +364,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s)kvm_vm_enable_cap (s,KVM_CAP_S390_USER_SIGP ,0 );kvm_vm_enable_cap (s,KVM_CAP_S390_VECTOR_REGISTERS ,0 );kvm_vm_enable_cap (s,KVM_CAP_S390_USER_STSI ,0 ); +kvm_vm_enable_cap (s,KVM_CAP_S390_CPU_TOPOLOGY ,0 );if (ri_allowed ()) {if (kvm_vm_enable_cap (s,KVM_CAP_S390_RI ,0 ) ==0 ) { cap_ri =1 ;
@@ -1454,6 +1456,16 @@ static int kvm_mpcifc_service_call(S390CPU *cpu, struct kvm_run *run) } } +static int kvm_handle_ptf (S390CPU *cpu,struct kvm_run *run) +{ +uint8_t r1 = (run->s390_sieic.ipb >>20 ) &0x0f ; +int ret; + + ret =s390_handle_ptf (cpu, r1,RA_IGNORED ); +setcc (cpu, ret); +return 0 ; +} +static int handle_b9 (S390CPU *cpu,struct kvm_run *run,uint8_t ipa1) {int r =0 ;
@@ -1471,6 +1483,9 @@ static int handle_b9(S390CPU *cpu, struct kvm_run *run, uint8_t ipa1)case PRIV_B9_RPCIT : r =kvm_rpcit_service_call (cpu, run);break ; +case PRIV_B9_PTF : + r =kvm_handle_ptf (cpu, run); +break ;case PRIV_B9_EQBS :/* just inject exception */ r =-1 ;
Signature · 1 line
2.27.0
Hi,
This series is updated by a v5 series with documentation and numa extensions. Some changes have been made in some of the patches contained in this series too.
Regards,
Pierre
Quoted 203 lines · Pierre Morel
Hi,
This series is a first part of the implementation of CPU topology for S390 greatly reduced from the first spin.
In particular, we reduced the scope to the S390x specificities, removing all code touching to SMP or NUMA, with the goal to:
- facilitate review and acceptance
- let for later the SMP part currently actively discussed in mainline
- be able despite the reduction of code to handle CPU topology for S390 using the current S390 topology provided by QEMU with cores and sockets only.To use these patches, you will need the Linux series version 4. You find it there:
https://lkml.org/lkml/2021/9/16/576Currently this code is for KVM only, I have no idea if it is interesting to provide a TCG patch. If ever it will be done in another series.
A short introduction
====================CPU Topology is described in the S390 POP with essentially the description of two instructions:
PTF Perform Topology function used to poll for topology change
and used to set the polarization but this part is not part of this item.STSI Store System Information and the SYSIB 15.1.x providing the Topology
configuration.S390 Topology is a 6 levels hierarchical topology with up to 5 level
of containers. The last topology level, specifying the CPU cores.This patch series only uses the two lower levels sockets and cores.
To get the information on the topology, S390 provides the STSI instruction, which stores a structures providing the list of the containers used in the Machine topology: the SYSIB.
A selector within the STSI instruction allow to chose how many topology levels will be provide in the SYSIB.Using the Topology List Entries (TLE) provided inside the SYSIB we the Linux kernel is able to compute the information about the cache distance between two cores and can use this information to take scheduling decisions.
Note:
-----
Z15 reports 3 levels of containers, drawers, book, sockets as Container-TLEs above the core description inside CPU-TLEs.The Topology can be seen at several places inside zLinux:
- sysfs:/sys/devices/system/cpu/cpuX/topology
- procfs:/proc/sysinfoand/proc/cpuinfo
- lscpu -e : gives toplogy informationThe different Topology levels have names:
- Node - Drawer - Book - sockets or physical package - coreThreads:
Multithreading, is not part of the topology as described by the SYSIB 15.1.xThe interest of the guest to know the CPU topology is obviously to be able to optimise the load balancing and the migration of threads. KVM will have the same interest concerning vCPUs scheduling and cache optimisation.
The design
==========1) To be ready for hotplug, I chose an Object oriented design of the topology containers:
- A node is a bridge on the SYSBUS and defines a "node bus"
- A drawer is hotplug on the "node bus"
- A book on the "drawer bus"
- A socket on the "book bus"
- And the CPU Topology List Entry (CPU-TLE)sits on the socket bus. These objects will be enhanced with the cache information when NUMA is implemented.This also allows for easy retrieval when building the different SYSIB for Store Topology System Information (STSI)
2) Perform Topology Function (PTF) instruction is made available to the guest with a new KVM capability and intercepted in QEMU, allowing the guest to pool for topology changes.
Features and TBD list
=====================- There is no direct match between IDs shown by:
- lscpu (unrelated numbered list),
- SYSIB 15.1.x (topology ID)- The CPU number, left column of lscpu, is used to reference a CPU by Linux tools
While the CPU address is used by QEMU for hotplug.- Effect of -smp parsing on the topology with an example:
-smp 9,sockets=4,cores=4,maxcpus=16We have 4 socket each holding 4 cores so that we have a maximum of 16 CPU, 9 of them are active on boot. (Should be obvious)
# lscpu -e CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2d:L2i ONLINE CONFIGURED POLARIZATION ADDRESS 0 0 0 0 0 0 0:0:0:0 yes yes horizontal 0 1 0 0 0 0 1 1:1:1:1 yes yes horizontal 1 2 0 0 0 0 2 2:2:2:2 yes yes horizontal 2 3 0 0 0 0 3 3:3:3:3 yes yes horizontal 3 4 0 0 0 1 4 4:4:4:4 yes yes horizontal 4 5 0 0 0 1 5 5:5:5:5 yes yes horizontal 5 6 0 0 0 1 6 6:6:6:6 yes yes horizontal 6 7 0 0 0 1 7 7:7:7:7 yes yes horizontal 7 8 0 0 0 2 8 8:8:8:8 yes yes horizontal 8 #- To plug a new CPU inside the topology one can simply use the CPU address like in:
(qemu)
device_addhost-s390x-cpu,core-id=12# lscpu -e CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2d:L2i ONLINE CONFIGURED POLARIZATION ADDRESS 0 0 0 0 0 0 0:0:0:0 yes yes horizontal 0 1 0 0 0 0 1 1:1:1:1 yes yes horizontal 1 2 0 0 0 0 2 2:2:2:2 yes yes horizontal 2 3 0 0 0 0 3 3:3:3:3 yes yes horizontal 3 4 0 0 0 1 4 4:4:4:4 yes yes horizontal 4 5 0 0 0 1 5 5:5:5:5 yes yes horizontal 5 6 0 0 0 1 6 6:6:6:6 yes yes horizontal 6 7 0 0 0 1 7 7:7:7:7 yes yes horizontal 7 8 0 0 0 2 8 8:8:8:8 yes yes horizontal 8 9 - - - - - ::: no yes horizontal 12 # chcpu -e 9 CPU 9 enabled # lscpu -e CPU NODE DRAWER BOOK SOCKET CORE L1d:L1i:L2d:L2i ONLINE CONFIGURED POLARIZATION ADDRESS 0 0 0 0 0 0 0:0:0:0 yes yes horizontal 0 1 0 0 0 0 1 1:1:1:1 yes yes horizontal 1 2 0 0 0 0 2 2:2:2:2 yes yes horizontal 2 3 0 0 0 0 3 3:3:3:3 yes yes horizontal 3 4 0 0 0 1 4 4:4:4:4 yes yes horizontal 4 5 0 0 0 1 5 5:5:5:5 yes yes horizontal 5 6 0 0 0 1 6 6:6:6:6 yes yes horizontal 6 7 0 0 0 1 7 7:7:7:7 yes yes horizontal 7 8 0 0 0 2 8 8:8:8:8 yes yes horizontal 8 9 0 0 0 3 9 9:9:9:9 yes yes horizontal 12 #It is up to the admin level, Libvirt for example, to pin the righ CPU to the right vCPU, but as we can see without NUMA, chosing separate sockets for CPUs is not easy without hotplug because without information the code will assign the vCPU and fill the sockets one after the other.
Note that this is also the default behavior on the LPAR.Conclusion
==========This patch, together with the associated KVM patch allows to provide CPU topology information to the guest.
Currently, only dedicated vCPU and CPU are supported and a NUMA topology can only be handled using CPU hotplug inside the guest.Next extensions are to provide:
- adding books and drawers levels
- NUMA using the -numa QEMU parameter.
- Topology information change for shared CPURegards,
PierrePierre Morel (5):
linux-headers update
s390x: topology: CPU topology objects and structures
s390x: topology: implementating Store Topology System Information
s390x: CPU topology: CPU topology migration
s390x: kvm: topology: interception of PTF instructionhw/s390x/cpu-topology.c | 361 ++++++++++++++++++++++++++++ hw/s390x/meson.build | 1 + hw/s390x/s390-virtio-ccw.c | 54 +++++ include/hw/s390x/cpu-topology.h | 74 ++++++ include/hw/s390x/s390-virtio-ccw.h | 6 + linux-headers/linux/kvm.h | 1 + target/s390x/cpu.h | 50 ++++ target/s390x/cpu_features_def.h.inc | 1 + target/s390x/cpu_models.c | 2 + target/s390x/cpu_topology.c | 113 +++++++++ target/s390x/gen-features.c | 3 + target/s390x/kvm/kvm.c | 26 ++ target/s390x/machine.c | 48 ++++ target/s390x/meson.build | 1 + 14 files changed, 741 insertions(+)create mode 100644
hw/s390x/cpu-topology.ccreate mode 100644include/hw/s390x/cpu-topology.hcreate mode 100644target/s390x/cpu_topology.c
Signature · 2 lines
Pierre Morel IBM Lab Boeblingen