summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2020-03-31 12:09:51 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2020-03-31 12:09:51 -0700
commit42595ce90b9d4a6b9d8c5a1ea78da4eeaf7e086a (patch)
treeadb5dda1bd278e920c2e680c24fb84d7c7c67cc8
parentd9d76778927dc953c553b83ab52287dfbd15ac6a (diff)
parent8fefe9dacdb0a1347d3dac871bb1bba3cbc32945 (diff)
Merge branch 'x86-vmware-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 vmware updates from Ingo Molnar: "The main change in this tree is the addition of 'steal time clock support' for VMware guests" * 'x86-vmware-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/vmware: Use bool type for vmw_sched_clock x86/vmware: Enable steal time accounting x86/vmware: Add steal time clock support for VMware guests x86/vmware: Remove vmware_sched_clock_setup() x86/vmware: Make vmware_select_hypercall() __init
-rw-r--r--Documentation/admin-guide/kernel-parameters.txt2
-rw-r--r--arch/x86/kernel/cpu/vmware.c229
2 files changed, 222 insertions, 9 deletions
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index cbd8bc7cf21d..24799f1d29fe 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3201,7 +3201,7 @@
[X86,PV_OPS] Disable paravirtualized VMware scheduler
clock and use the default one.
- no-steal-acc [X86,KVM,ARM64] Disable paravirtualized steal time
+ no-steal-acc [X86,PV_OPS,ARM64] Disable paravirtualized steal time
accounting. steal time is computed, but won't
influence scheduler behaviour
diff --git a/arch/x86/kernel/cpu/vmware.c b/arch/x86/kernel/cpu/vmware.c
index 46d732696c1c..9b6fafa69be9 100644
--- a/arch/x86/kernel/cpu/vmware.c
+++ b/arch/x86/kernel/cpu/vmware.c
@@ -25,6 +25,8 @@
#include <linux/init.h>
#include <linux/export.h>
#include <linux/clocksource.h>
+#include <linux/cpu.h>
+#include <linux/reboot.h>
#include <asm/div64.h>
#include <asm/x86_init.h>
#include <asm/hypervisor.h>
@@ -47,6 +49,11 @@
#define VMWARE_CMD_GETVCPU_INFO 68
#define VMWARE_CMD_LEGACY_X2APIC 3
#define VMWARE_CMD_VCPU_RESERVED 31
+#define VMWARE_CMD_STEALCLOCK 91
+
+#define STEALCLOCK_NOT_AVAILABLE (-1)
+#define STEALCLOCK_DISABLED 0
+#define STEALCLOCK_ENABLED 1
#define VMWARE_PORT(cmd, eax, ebx, ecx, edx) \
__asm__("inl (%%dx), %%eax" : \
@@ -86,6 +93,18 @@
} \
} while (0)
+struct vmware_steal_time {
+ union {
+ uint64_t clock; /* stolen time counter in units of vtsc */
+ struct {
+ /* only for little-endian */
+ uint32_t clock_low;
+ uint32_t clock_high;
+ };
+ };
+ uint64_t reserved[7];
+};
+
static unsigned long vmware_tsc_khz __ro_after_init;
static u8 vmware_hypercall_mode __ro_after_init;
@@ -103,15 +122,25 @@ static unsigned long vmware_get_tsc_khz(void)
#ifdef CONFIG_PARAVIRT
static struct cyc2ns_data vmware_cyc2ns __ro_after_init;
-static int vmw_sched_clock __initdata = 1;
+static bool vmw_sched_clock __initdata = true;
+static DEFINE_PER_CPU_DECRYPTED(struct vmware_steal_time, vmw_steal_time) __aligned(64);
+static bool has_steal_clock;
+static bool steal_acc __initdata = true; /* steal time accounting */
static __init int setup_vmw_sched_clock(char *s)
{
- vmw_sched_clock = 0;
+ vmw_sched_clock = false;
return 0;
}
early_param("no-vmw-sched-clock", setup_vmw_sched_clock);
+static __init int parse_no_stealacc(char *arg)
+{
+ steal_acc = false;
+ return 0;
+}
+early_param("no-steal-acc", parse_no_stealacc);
+
static unsigned long long notrace vmware_sched_clock(void)
{
unsigned long long ns;
@@ -122,7 +151,7 @@ static unsigned long long notrace vmware_sched_clock(void)
return ns;
}
-static void __init vmware_sched_clock_setup(void)
+static void __init vmware_cyc2ns_setup(void)
{
struct cyc2ns_data *d = &vmware_cyc2ns;
unsigned long long tsc_now = rdtsc();
@@ -132,17 +161,201 @@ static void __init vmware_sched_clock_setup(void)
d->cyc2ns_offset = mul_u64_u32_shr(tsc_now, d->cyc2ns_mul,
d->cyc2ns_shift);
- pv_ops.time.sched_clock = vmware_sched_clock;
- pr_info("using sched offset of %llu ns\n", d->cyc2ns_offset);
+ pr_info("using clock offset of %llu ns\n", d->cyc2ns_offset);
+}
+
+static int vmware_cmd_stealclock(uint32_t arg1, uint32_t arg2)
+{
+ uint32_t result, info;
+
+ asm volatile (VMWARE_HYPERCALL :
+ "=a"(result),
+ "=c"(info) :
+ "a"(VMWARE_HYPERVISOR_MAGIC),
+ "b"(0),
+ "c"(VMWARE_CMD_STEALCLOCK),
+ "d"(0),
+ "S"(arg1),
+ "D"(arg2) :
+ "memory");
+ return result;
+}
+
+static bool stealclock_enable(phys_addr_t pa)
+{
+ return vmware_cmd_stealclock(upper_32_bits(pa),
+ lower_32_bits(pa)) == STEALCLOCK_ENABLED;
+}
+
+static int __stealclock_disable(void)
+{
+ return vmware_cmd_stealclock(0, 1);
+}
+
+static void stealclock_disable(void)
+{
+ __stealclock_disable();
+}
+
+static bool vmware_is_stealclock_available(void)
+{
+ return __stealclock_disable() != STEALCLOCK_NOT_AVAILABLE;
+}
+
+/**
+ * vmware_steal_clock() - read the per-cpu steal clock
+ * @cpu: the cpu number whose steal clock we want to read
+ *
+ * The function reads the steal clock if we are on a 64-bit system, otherwise
+ * reads it in parts, checking that the high part didn't change in the
+ * meantime.
+ *
+ * Return:
+ * The steal clock reading in ns.
+ */
+static uint64_t vmware_steal_clock(int cpu)
+{
+ struct vmware_steal_time *steal = &per_cpu(vmw_steal_time, cpu);
+ uint64_t clock;
+
+ if (IS_ENABLED(CONFIG_64BIT))
+ clock = READ_ONCE(steal->clock);
+ else {
+ uint32_t initial_high, low, high;
+
+ do {
+ initial_high = READ_ONCE(steal->clock_high);
+ /* Do not reorder initial_high and high readings */
+ virt_rmb();
+ low = READ_ONCE(steal->clock_low);
+ /* Keep low reading in between */
+ virt_rmb();
+ high = READ_ONCE(steal->clock_high);
+ } while (initial_high != high);
+
+ clock = ((uint64_t)high << 32) | low;
+ }
+
+ return mul_u64_u32_shr(clock, vmware_cyc2ns.cyc2ns_mul,
+ vmware_cyc2ns.cyc2ns_shift);
+}
+
+static void vmware_register_steal_time(void)
+{
+ int cpu = smp_processor_id();
+ struct vmware_steal_time *st = &per_cpu(vmw_steal_time, cpu);
+
+ if (!has_steal_clock)
+ return;
+
+ if (!stealclock_enable(slow_virt_to_phys(st))) {
+ has_steal_clock = false;
+ return;
+ }
+
+ pr_info("vmware-stealtime: cpu %d, pa %llx\n",
+ cpu, (unsigned long long) slow_virt_to_phys(st));
}
+static void vmware_disable_steal_time(void)
+{
+ if (!has_steal_clock)
+ return;
+
+ stealclock_disable();
+}
+
+static void vmware_guest_cpu_init(void)
+{
+ if (has_steal_clock)
+ vmware_register_steal_time();
+}
+
+static void vmware_pv_guest_cpu_reboot(void *unused)
+{
+ vmware_disable_steal_time();
+}
+
+static int vmware_pv_reboot_notify(struct notifier_block *nb,
+ unsigned long code, void *unused)
+{
+ if (code == SYS_RESTART)
+ on_each_cpu(vmware_pv_guest_cpu_reboot, NULL, 1);
+ return NOTIFY_DONE;
+}
+
+static struct notifier_block vmware_pv_reboot_nb = {
+ .notifier_call = vmware_pv_reboot_notify,
+};
+
+#ifdef CONFIG_SMP
+static void __init vmware_smp_prepare_boot_cpu(void)
+{
+ vmware_guest_cpu_init();
+ native_smp_prepare_boot_cpu();
+}
+
+static int vmware_cpu_online(unsigned int cpu)
+{
+ local_irq_disable();
+ vmware_guest_cpu_init();
+ local_irq_enable();
+ return 0;
+}
+
+static int vmware_cpu_down_prepare(unsigned int cpu)
+{
+ local_irq_disable();
+ vmware_disable_steal_time();
+ local_irq_enable();
+ return 0;
+}
+#endif
+
+static __init int activate_jump_labels(void)
+{
+ if (has_steal_clock) {
+ static_key_slow_inc(&paravirt_steal_enabled);
+ if (steal_acc)
+ static_key_slow_inc(&paravirt_steal_rq_enabled);
+ }
+
+ return 0;
+}
+arch_initcall(activate_jump_labels);
+
static void __init vmware_paravirt_ops_setup(void)
{
pv_info.name = "VMware hypervisor";
pv_ops.cpu.io_delay = paravirt_nop;
- if (vmware_tsc_khz && vmw_sched_clock)
- vmware_sched_clock_setup();
+ if (vmware_tsc_khz == 0)
+ return;
+
+ vmware_cyc2ns_setup();
+
+ if (vmw_sched_clock)
+ pv_ops.time.sched_clock = vmware_sched_clock;
+
+ if (vmware_is_stealclock_available()) {
+ has_steal_clock = true;
+ pv_ops.time.steal_clock = vmware_steal_clock;
+
+ /* We use reboot notifier only to disable steal clock */
+ register_reboot_notifier(&vmware_pv_reboot_nb);
+
+#ifdef CONFIG_SMP
+ smp_ops.smp_prepare_boot_cpu =
+ vmware_smp_prepare_boot_cpu;
+ if (cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
+ "x86/vmware:online",
+ vmware_cpu_online,
+ vmware_cpu_down_prepare) < 0)
+ pr_err("vmware_guest: Failed to install cpu hotplug callbacks\n");
+#else
+ vmware_guest_cpu_init();
+#endif
+ }
}
#else
#define vmware_paravirt_ops_setup() do {} while (0)
@@ -213,7 +426,7 @@ static void __init vmware_platform_setup(void)
vmware_set_capabilities();
}
-static u8 vmware_select_hypercall(void)
+static u8 __init vmware_select_hypercall(void)
{
int eax, ebx, ecx, edx;