Skip to contentSkip to conversation

Introduce cpu die topology and enable CPUID.1F for i386

PATCH New Like Xu 1 participant 4 messages Jun 20, 2019

lore
Share & tools
lore
author
#2 lore raw

PATCH v4 2/3 machine: Refactor smp_parse() in vl.c as MachineClass::smp_parse()

In reply to Like Xu

To make smp_parse() more flexible and expansive, a smp_parse function pointer is added to MachineClass that machine types could override.

The generic smp_parse() code in vl.c is moved to hw/core/machine.c, and become the default implementation of MachineClass::smp_parse. A PC-specific function called pc_smp_parse() has been added to hw/i386/pc.c, which in this patch changes nothing against the default one .

---
 hw/core/machine.c    | 76 ++++++++++++++++++++++++++++++++++++++++++
 hw/i386/pc.c         | 79 ++++++++++++++++++++++++++++++++++++++++++++
 include/hw/boards.h  |  5 +++
 include/hw/i386/pc.h |  1 +
 vl.c                 | 75 ++---------------------------------------
 5 files changed, 163 insertions(+), 73 deletions(-)
hw/core/machine.c +76 -0
index 8b8d263afe..36a838f1cb 100644
@@ -11,6 +11,9 @@
  */
 
 #include "qemu/osdep.h"
+#include "qemu/option.h"
+#include "qapi/qmp/qerror.h"
+#include "sysemu/replay.h"
 #include "qemu/units.h"
 #include "hw/boards.h"
 #include "qapi/error.h"
@@ -728,6 +731,78 @@ void machine_set_cpu_numa_node(MachineState *machine,
     }
 }
 
+static void smp_parse(MachineState *ms, QemuOpts *opts)
+{
+    if (opts) {
+        unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
+        unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
+        unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
+        unsigned threads = qemu_opt_get_number(opts, "threads", 0);
+
+        /* compute missing values, prefer sockets over cores over threads */
+        if (cpus == 0 || sockets == 0) {
+            cores = cores > 0 ? cores : 1;
+            threads = threads > 0 ? threads : 1;
+            if (cpus == 0) {
+                sockets = sockets > 0 ? sockets : 1;
+                cpus = cores * threads * sockets;
+            } else {
+                ms->smp.max_cpus =
+                        qemu_opt_get_number(opts, "maxcpus", cpus);
+                sockets = ms->smp.max_cpus / (cores * threads);
+            }
+        } else if (cores == 0) {
+            threads = threads > 0 ? threads : 1;
+            cores = cpus / (sockets * threads);
+            cores = cores > 0 ? cores : 1;
+        } else if (threads == 0) {
+            threads = cpus / (cores * sockets);
+            threads = threads > 0 ? threads : 1;
+        } else if (sockets * cores * threads < cpus) {
+            error_report("cpu topology: "
+                         "sockets (%u) * cores (%u) * threads (%u) < "
+                         "smp_cpus (%u)",
+                         sockets, cores, threads, cpus);
+            exit(1);
+        }
+
+        ms->smp.max_cpus =
+                qemu_opt_get_number(opts, "maxcpus", cpus);
+
+        if (ms->smp.max_cpus < cpus) {
+            error_report("maxcpus must be equal to or greater than smp");
+            exit(1);
+        }
+
+        if (sockets * cores * threads > ms->smp.max_cpus) {
+            error_report("cpu topology: "
+                         "sockets (%u) * cores (%u) * threads (%u) > "
+                         "maxcpus (%u)",
+                         sockets, cores, threads,
+                         ms->smp.max_cpus);
+            exit(1);
+        }
+
+        if (sockets * cores * threads != ms->smp.max_cpus) {
+            warn_report("Invalid CPU topology deprecated: "
+                        "sockets (%u) * cores (%u) * threads (%u) "
+                        "!= maxcpus (%u)",
+                        sockets, cores, threads,
+                        ms->smp.max_cpus);
+        }
+
+        ms->smp.cpus = cpus;
+        ms->smp.cores = cores;
+        ms->smp.threads = threads;
+    }
+
+    if (ms->smp.cpus > 1) {
+        Error *blocker = NULL;
+        error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
+        replay_add_blocker(blocker);
+    }
+}
+
 static void machine_class_init(ObjectClass *oc, void *data)
 {
     MachineClass *mc = MACHINE_CLASS(oc);
@@ -735,6 +810,7 @@ static void machine_class_init(ObjectClass *oc, void *data)
     /* Default 128 MB as guest ram size */
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
+    mc->smp_parse = smp_parse;
 
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned
hw/i386/pc.c +79 -0

97 lines not on this page · raw message

include/hw/boards.h +5 -0
index 04f52c1335..6dd0761562 100644
@@ -159,6 +159,10 @@ typedef struct {
  * @kvm_type:
  *    Return the type of KVM corresponding to the kvm-type string option or
  *    computed based on other criteria such as the host kernel capabilities.
+ * @smp_parse:
+ *    The function pointer to hook different machine specific functions for
+ *    parsing "smp-opts" from QemuOpts to MachineState::CpuTopology and more
+ *    machine specific topology fields, such as smp_dies for PCMachine.
  */
 struct MachineClass {
     /*< private >*/
@@ -175,6 +179,7 @@ struct MachineClass {
     void (*reset)(MachineState *state);
     void (*hot_add_cpu)(MachineState *state, const int64_t id, Error **errp);
     int (*kvm_type)(MachineState *machine, const char *arg);
+    void (*smp_parse)(MachineState *ms, QemuOpts *opts);
 
     BlockInterfaceType block_default_type;
     int units_per_default_bus;
include/hw/i386/pc.h +1 -0
index 2ae86e202d..e96e449cef 100644
@@ -189,6 +189,7 @@ void pc_acpi_smi_interrupt(void *opaque, int irq, int level);
 
 void pc_cpus_init(PCMachineState *pcms);
 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp);
+void pc_smp_parse(MachineState *ms, QemuOpts *opts);
 
 void pc_guest_info_init(PCMachineState *pcms);
 
vl.c +2 -73

87 lines not on this page · raw message

Signature · 1 line
2.21.0

Like Xu

author
#0 lore raw

v4 cover Introduce cpu die topology and enable CPUID.1F for i386

This patch series introduces a new cpu topolgy 'die' for PCMachine, which extends virtual cpu topology to the socket/die/core/thread model, allowing the setting of dies number per one socket via -smp qemu command.

For i386, it upgrades APIC-IDs generation and reversion functions with a new exposed leaf called CPUID.1F, which is a preferred superset to leaf 0BH. The CPUID.1F spec is on the latest Inetl SDM, 3-190 Vol 2A.

Guest system could discover multi-die/package topology through CPUID.1F. and its benefit is primarily for _reporting_ of the guest CPU topology. The guest kernel with multi-die/package support have no impact on its cache topology, NUMA topology, Linux scheduler, or system performance.

==changelog==

v4:

- base commit: 22fa84da on github.com/ehabkost/qemu.git:machine-next
- refine comments for pc_smp_parse()
- remove the use of cpu->enable_cpuid_0x1f
- apply new logic for cpuid_min_level adjustment and drop the legacy one
- refine the way of MachineState casting in pc_smp_parse()
- [QUEUED] move test_topo_bits to the previous patch for bisectability

v3: https://patchwork.kernel.org/cover/10989013/

- add a MachineClass::smp_parse function pointer
- place the PC-specific function inside hw/i386/pc.c
- introduce die_id in a separate patch with default value 0
- set env->nr_dies in pc_new_cpu() and pc_cpu_pre_plug()
- fix a circular dependency between target/i386/cpu.c and hw/i386/pc.c
- fix cpu->die_id check in pc_cpu_pre_plug()
- Based on "[PATCH v3 00/10] Refactor cpu topo into machine properties"
- Rebase to commit 219dca61ebf41625831d4f96a720852baf44b762

v2: https://patchwork.kernel.org/cover/10953191/

- Enable cpu die-level topolgy only for PCMachine and X86CPU
- Minimize cpuid.0.eax to the setting value actually used by guest
- Update cmd line -smps docs for die-level configurations
- Refactoring topo-bit tests for x86_apicid_from_cpu_idx() with nr_dies
- Based on "[PATCH v3 00/10] Refactor cpu topo into machine properties"
- Rebase to commit 2259637b95bef3116cc262459271de08e038cc66

v1: https://patchwork.kernel.org/cover/10876667/

Like Xu (3):
target/i386: Add CPUID.1F generation support for multi-dies PCMachine
machine: Refactor smp_parse() in vl.c as MachineClass::smp_parse() vl.c: Add -smp, dies=* command line support and update doc

 hw/core/machine.c    | 76 ++++++++++++++++++++++++++++++++++++++++
 hw/i386/pc.c         | 83 ++++++++++++++++++++++++++++++++++++++++++++
 include/hw/boards.h  |  5 +++
 include/hw/i386/pc.h |  1 +
 qemu-options.hx      | 17 ++++-----
 target/i386/cpu.c    | 41 ++++++++++++++++++++++
 target/i386/cpu.h    |  1 +
 target/i386/kvm.c    | 12 +++++++
 vl.c                 | 78 +++--------------------------------------
 9 files changed, 233 insertions(+), 81 deletions(-)
Signature · 1 line
2.21.0
author
#3 lore raw

PATCH v4 3/3 vl.c: Add -smp, dies=* command line support and update doc

For PC target, users could configure the number of dies per one package via command line with this patch, such as "-smp dies=2,cores=4".

The parsing rules of new cpu-topology model obey the same restrictions/logic as the legacy socket/core/thread model especially on missing values computing.

---
 hw/i386/pc.c    | 30 +++++++++++++++++-------------
 qemu-options.hx | 17 +++++++++--------
 vl.c            |  3 +++
 3 files changed, 29 insertions(+), 21 deletions(-)
hw/i386/pc.c +17 -13
index 092bd10d4d..2ed1b3f8de 100644
@@ -1548,9 +1548,12 @@ static void pc_new_cpu(PCMachineState *pcms, int64_t apic_id, Error **errp)
  */
 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
 {
+    PCMachineState *pcms = PC_MACHINE(ms);
+
     if (opts) {
         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
+        unsigned dies = qemu_opt_get_number(opts, "dies", 1);
         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
 
@@ -1560,24 +1563,24 @@ void pc_smp_parse(MachineState *ms, QemuOpts *opts)
             threads = threads > 0 ? threads : 1;
             if (cpus == 0) {
                 sockets = sockets > 0 ? sockets : 1;
-                cpus = cores * threads * sockets;
+                cpus = cores * threads * dies * sockets;
             } else {
                 ms->smp.max_cpus =
                         qemu_opt_get_number(opts, "maxcpus", cpus);
-                sockets = ms->smp.max_cpus / (cores * threads);
+                sockets = ms->smp.max_cpus / (cores * threads * dies);
             }
         } else if (cores == 0) {
             threads = threads > 0 ? threads : 1;
-            cores = cpus / (sockets * threads);
+            cores = cpus / (sockets * dies * threads);
             cores = cores > 0 ? cores : 1;
         } else if (threads == 0) {
-            threads = cpus / (cores * sockets);
+            threads = cpus / (cores * dies * sockets);
             threads = threads > 0 ? threads : 1;
-        } else if (sockets * cores * threads < cpus) {
+        } else if (sockets * dies * cores * threads < cpus) {
             error_report("cpu topology: "
-                         "sockets (%u) * cores (%u) * threads (%u) < "
+                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
                          "smp_cpus (%u)",
-                         sockets, cores, threads, cpus);
+                         sockets, dies, cores, threads, cpus);
             exit(1);
         }
 
@@ -1589,26 +1592,27 @@ void pc_smp_parse(MachineState *ms, QemuOpts *opts)
             exit(1);
         }
 
-        if (sockets * cores * threads > ms->smp.max_cpus) {
+        if (sockets * dies * cores * threads > ms->smp.max_cpus) {
             error_report("cpu topology: "
-                         "sockets (%u) * cores (%u) * threads (%u) > "
+                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
                          "maxcpus (%u)",
-                         sockets, cores, threads,
+                         sockets, dies, cores, threads,
                          ms->smp.max_cpus);
             exit(1);
         }
 
-        if (sockets * cores * threads != ms->smp.max_cpus) {
+        if (sockets * dies * cores * threads != ms->smp.max_cpus) {
             warn_report("Invalid CPU topology deprecated: "
-                        "sockets (%u) * cores (%u) * threads (%u) "
+                        "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
                         "!= maxcpus (%u)",
-                        sockets, cores, threads,
+                        sockets, dies, cores, threads,
                         ms->smp.max_cpus);
         }
 
         ms->smp.cpus = cpus;
         ms->smp.cores = cores;
         ms->smp.threads = threads;
+        pcms->smp_dies = dies;
     }
 
     if (ms->smp.cpus > 1) {
qemu-options.hx +9 -8
index 0d8beb4afd..a5b314a448 100644
@@ -138,25 +138,26 @@ no incompatible TCG features have been enabled (e.g. icount/replay).
 ETEXI
 
 DEF("smp", HAS_ARG, QEMU_OPTION_smp,
-    "-smp [cpus=]n[,maxcpus=cpus][,cores=cores][,threads=threads][,sockets=sockets]\n"
+    "-smp [cpus=]n[,maxcpus=cpus][,cores=cores][,threads=threads][,dies=dies][,sockets=sockets]\n"
     "                set the number of CPUs to 'n' [default=1]\n"
     "                maxcpus= maximum number of total cpus, including\n"
     "                offline CPUs for hotplug, etc\n"
-    "                cores= number of CPU cores on one socket\n"
+    "                cores= number of CPU cores on one socket (for PC, it's on one die)\n"
     "                threads= number of threads on one CPU core\n"
+    "                dies= number of CPU dies on one socket (for PC only)\n"
     "                sockets= number of discrete sockets in the system\n",
         QEMU_ARCH_ALL)
 STEXI
-@item -smp [cpus=]@var{n}[,cores=@var{cores}][,threads=@var{threads}][,sockets=@var{sockets}][,maxcpus=@var{maxcpus}]
+@item -smp [cpus=]@var{n}[,cores=@var{cores}][,threads=@var{threads}][,dies=dies][,sockets=@var{sockets}][,maxcpus=@var{maxcpus}]
 @findex -smp
 Simulate an SMP system with @var{n} CPUs. On the PC target, up to 255
 CPUs are supported. On Sparc32 target, Linux limits the number of usable CPUs
 to 4.
-For the PC target, the number of @var{cores} per socket, the number
-of @var{threads} per cores and the total number of @var{sockets} can be
-specified. Missing values will be computed. If any on the three values is
-given, the total number of CPUs @var{n} can be omitted. @var{maxcpus}
-specifies the maximum number of hotpluggable CPUs.
+For the PC target, the number of @var{cores} per die, the number of @var{threads}
+per cores, the number of @var{dies} per packages and the total number of
+@var{sockets} can be specified. Missing values will be computed.
+If any on the three values is given, the total number of CPUs @var{n} can be omitted.
+@var{maxcpus} specifies the maximum number of hotpluggable CPUs.
 ETEXI
 
 DEF("numa", HAS_ARG, QEMU_OPTION_numa,
vl.c +3 -0
index 9218a718f3..e3908caa47 100644
@@ -1233,6 +1233,9 @@ static QemuOptsList qemu_smp_opts = {
         }, {
             .name = "sockets",
             .type = QEMU_OPT_NUMBER,
+        }, {
+            .name = "dies",
+            .type = QEMU_OPT_NUMBER,
         }, {
             .name = "cores",
             .type = QEMU_OPT_NUMBER,
Signature · 1 line
2.21.0
author
#1 lore raw

PATCH v4 1/3 target/i386: Add CPUID.1F generation support for multi-dies PCMachine

In reply to Like Xu

The CPUID.1F as Intel V2 Extended Topology Enumeration Leaf would be exposed if guests want to emulate multiple software-visible die within each package. Per Intel's SDM, the 0x1f is a superset of 0xb, thus they can be generated by almost same code as 0xb except die_offset setting.

If the number of dies per package is greater than 1, the cpuid_min_level would be adjusted to 0x1f regardless of whether the host supports CPUID.1F. Likewise, the CPUID.1F wouldn't be exposed if env->nr_dies < 2.

---
 target/i386/cpu.c | 41 +++++++++++++++++++++++++++++++++++++++++
 target/i386/cpu.h |  1 +
 target/i386/kvm.c | 12 ++++++++++++
 3 files changed, 54 insertions(+)
target/i386/cpu.c +41 -0
index 88908a6373..efcbe6a2b2 100644
@@ -4439,6 +4439,42 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,
             *ecx |= CPUID_TOPOLOGY_LEVEL_INVALID;
         }
 
+        assert(!(*eax & ~0x1f));
+        *ebx &= 0xffff; /* The count doesn't need to be reliable. */
+        break;
+    case 0x1F:
+        /* V2 Extended Topology Enumeration Leaf */
+        if (env->nr_dies < 2) {
+            *eax = *ebx = *ecx = *edx = 0;
+            break;
+        }
+
+        *ecx = count & 0xff;
+        *edx = cpu->apic_id;
+        switch (count) {
+        case 0:
+            *eax = apicid_core_offset(env->nr_dies, cs->nr_cores,
+                                                    cs->nr_threads);
+            *ebx = cs->nr_threads;
+            *ecx |= CPUID_TOPOLOGY_LEVEL_SMT;
+            break;
+        case 1:
+            *eax = apicid_die_offset(env->nr_dies, cs->nr_cores,
+                                                   cs->nr_threads);
+            *ebx = cs->nr_cores * cs->nr_threads;
+            *ecx |= CPUID_TOPOLOGY_LEVEL_CORE;
+            break;
+        case 2:
+            *eax = apicid_pkg_offset(env->nr_dies, cs->nr_cores,
+                                                   cs->nr_threads);
+            *ebx = env->nr_dies * cs->nr_cores * cs->nr_threads;
+            *ecx |= CPUID_TOPOLOGY_LEVEL_DIE;
+            break;
+        default:
+            *eax = 0;
+            *ebx = 0;
+            *ecx |= CPUID_TOPOLOGY_LEVEL_INVALID;
+        }
         assert(!(*eax & ~0x1f));
         *ebx &= 0xffff; /* The count doesn't need to be reliable. */
         break;
@@ -5116,6 +5152,11 @@ static void x86_cpu_expand_features(X86CPU *cpu, Error **errp)
             x86_cpu_adjust_level(cpu, &cpu->env.cpuid_min_level, 0x14);
         }
 
+        /* CPU topology with multi-dies support requires CPUID[0x1F] */
+        if (env->nr_dies > 1) {
+            x86_cpu_adjust_level(cpu, &env->cpuid_min_level, 0x1F);
+        }
+
         /* SVM requires CPUID[0x8000000A] */
         if (env->features[FEAT_8000_0001_ECX] & CPUID_EXT3_SVM) {
             x86_cpu_adjust_level(cpu, &env->cpuid_min_xlevel, 0x8000000A);
target/i386/cpu.h +1 -0
index 46dd81f6b7..eec6e4b7b7 100644
@@ -726,6 +726,7 @@ typedef uint32_t FeatureWordArray[FEATURE_WORDS];
 #define CPUID_TOPOLOGY_LEVEL_INVALID  (0U << 8)
 #define CPUID_TOPOLOGY_LEVEL_SMT      (1U << 8)
 #define CPUID_TOPOLOGY_LEVEL_CORE     (2U << 8)
+#define CPUID_TOPOLOGY_LEVEL_DIE      (5U << 8)
 
 /* MSR Feature Bits */
 #define MSR_ARCH_CAP_RDCL_NO    (1U << 0)
target/i386/kvm.c +12 -0
index 6899061b4e..5deb4248ac 100644
@@ -1080,6 +1080,10 @@ int kvm_arch_init_vcpu(CPUState *cs)
             }
             break;
         }
+        case 0x1f:
+            if (env->nr_dies < 2) {
+                break;
+            }
         case 4:
         case 0xb:
         case 0xd:
@@ -1087,6 +1091,11 @@ int kvm_arch_init_vcpu(CPUState *cs)
                 if (i == 0xd && j == 64) {
                     break;
                 }
+
+                if (i == 0x1f && j == 64) {
+                    break;
+                }
+
                 c->function = i;
                 c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
                 c->index = j;
@@ -1098,6 +1107,9 @@ int kvm_arch_init_vcpu(CPUState *cs)
                 if (i == 0xb && !(c->ecx & 0xff00)) {
                     break;
                 }
+                if (i == 0x1f && !(c->ecx & 0xff00)) {
+                    break;
+                }
                 if (i == 0xd && c->eax == 0) {
                     continue;
                 }
Signature · 1 line
2.21.0