mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-27 23:46:45 +07:00
bf09fb6cba
Remove support for context switching between the guest's and host's
desired UMWAIT_CONTROL. Propagating the guest's value to hardware isn't
required for correct functionality, e.g. KVM intercepts reads and writes
to the MSR, and the latency effects of the settings controlled by the
MSR are not architecturally visible.
As a general rule, KVM should not allow the guest to control power
management settings unless explicitly enabled by userspace, e.g. see
KVM_CAP_X86_DISABLE_EXITS. E.g. Intel's SDM explicitly states that C0.2
can improve the performance of SMT siblings. A devious guest could
disable C0.2 so as to improve the performance of their workloads at the
detriment to workloads running in the host or on other VMs.
Wholesale removal of UMWAIT_CONTROL context switching also fixes a race
condition where updates from the host may cause KVM to enter the guest
with the incorrect value. Because updates are are propagated to all
CPUs via IPI (SMP function callback), the value in hardware may be
stale with respect to the cached value and KVM could enter the guest
with the wrong value in hardware. As above, the guest can't observe the
bad value, but it's a weird and confusing wart in the implementation.
Removal also fixes the unnecessary usage of VMX's atomic load/store MSR
lists. Using the lists is only necessary for MSRs that are required for
correct functionality immediately upon VM-Enter/VM-Exit, e.g. EFER on
old hardware, or for MSRs that need to-the-uop precision, e.g. perf
related MSRs. For UMWAIT_CONTROL, the effects are only visible in the
kernel via TPAUSE/delay(), and KVM doesn't do any form of delay in
vcpu_vmx_run(). Using the atomic lists is undesirable as they are more
expensive than direct RDMSR/WRMSR.
Furthermore, even if giving the guest control of the MSR is legitimate,
e.g. in pass-through scenarios, it's not clear that the benefits would
outweigh the overhead. E.g. saving and restoring an MSR across a VMX
roundtrip costs ~250 cycles, and if the guest diverged from the host
that cost would be paid on every run of the guest. In other words, if
there is a legitimate use case then it should be enabled by a new
per-VM capability.
Note, KVM still needs to emulate MSR_IA32_UMWAIT_CONTROL so that it can
correctly expose other WAITPKG features to the guest, e.g. TPAUSE,
UMWAIT and UMONITOR.
Fixes: 6e3ba4abce
("KVM: vmx: Emulate MSR IA32_UMWAIT_CONTROL")
Cc: stable@vger.kernel.org
Cc: Jingqi Liu <jingqi.liu@intel.com>
Cc: Tao Xu <tao3.xu@intel.com>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20200623005135.10414-1-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
146 lines
4.4 KiB
C
146 lines
4.4 KiB
C
/* SPDX-License-Identifier: GPL-2.0 */
|
|
#ifndef _ASM_X86_MWAIT_H
|
|
#define _ASM_X86_MWAIT_H
|
|
|
|
#include <linux/sched.h>
|
|
#include <linux/sched/idle.h>
|
|
|
|
#include <asm/cpufeature.h>
|
|
#include <asm/nospec-branch.h>
|
|
|
|
#define MWAIT_SUBSTATE_MASK 0xf
|
|
#define MWAIT_CSTATE_MASK 0xf
|
|
#define MWAIT_SUBSTATE_SIZE 4
|
|
#define MWAIT_HINT2CSTATE(hint) (((hint) >> MWAIT_SUBSTATE_SIZE) & MWAIT_CSTATE_MASK)
|
|
#define MWAIT_HINT2SUBSTATE(hint) ((hint) & MWAIT_CSTATE_MASK)
|
|
|
|
#define CPUID_MWAIT_LEAF 5
|
|
#define CPUID5_ECX_EXTENSIONS_SUPPORTED 0x1
|
|
#define CPUID5_ECX_INTERRUPT_BREAK 0x2
|
|
|
|
#define MWAIT_ECX_INTERRUPT_BREAK 0x1
|
|
#define MWAITX_ECX_TIMER_ENABLE BIT(1)
|
|
#define MWAITX_MAX_WAIT_CYCLES UINT_MAX
|
|
#define MWAITX_DISABLE_CSTATES 0xf0
|
|
#define TPAUSE_C01_STATE 1
|
|
#define TPAUSE_C02_STATE 0
|
|
|
|
static inline void __monitor(const void *eax, unsigned long ecx,
|
|
unsigned long edx)
|
|
{
|
|
/* "monitor %eax, %ecx, %edx;" */
|
|
asm volatile(".byte 0x0f, 0x01, 0xc8;"
|
|
:: "a" (eax), "c" (ecx), "d"(edx));
|
|
}
|
|
|
|
static inline void __monitorx(const void *eax, unsigned long ecx,
|
|
unsigned long edx)
|
|
{
|
|
/* "monitorx %eax, %ecx, %edx;" */
|
|
asm volatile(".byte 0x0f, 0x01, 0xfa;"
|
|
:: "a" (eax), "c" (ecx), "d"(edx));
|
|
}
|
|
|
|
static inline void __mwait(unsigned long eax, unsigned long ecx)
|
|
{
|
|
mds_idle_clear_cpu_buffers();
|
|
|
|
/* "mwait %eax, %ecx;" */
|
|
asm volatile(".byte 0x0f, 0x01, 0xc9;"
|
|
:: "a" (eax), "c" (ecx));
|
|
}
|
|
|
|
/*
|
|
* MWAITX allows for a timer expiration to get the core out a wait state in
|
|
* addition to the default MWAIT exit condition of a store appearing at a
|
|
* monitored virtual address.
|
|
*
|
|
* Registers:
|
|
*
|
|
* MWAITX ECX[1]: enable timer if set
|
|
* MWAITX EBX[31:0]: max wait time expressed in SW P0 clocks. The software P0
|
|
* frequency is the same as the TSC frequency.
|
|
*
|
|
* Below is a comparison between MWAIT and MWAITX on AMD processors:
|
|
*
|
|
* MWAIT MWAITX
|
|
* opcode 0f 01 c9 | 0f 01 fb
|
|
* ECX[0] value of RFLAGS.IF seen by instruction
|
|
* ECX[1] unused/#GP if set | enable timer if set
|
|
* ECX[31:2] unused/#GP if set
|
|
* EAX unused (reserve for hint)
|
|
* EBX[31:0] unused | max wait time (P0 clocks)
|
|
*
|
|
* MONITOR MONITORX
|
|
* opcode 0f 01 c8 | 0f 01 fa
|
|
* EAX (logical) address to monitor
|
|
* ECX #GP if not zero
|
|
*/
|
|
static inline void __mwaitx(unsigned long eax, unsigned long ebx,
|
|
unsigned long ecx)
|
|
{
|
|
/* No MDS buffer clear as this is AMD/HYGON only */
|
|
|
|
/* "mwaitx %eax, %ebx, %ecx;" */
|
|
asm volatile(".byte 0x0f, 0x01, 0xfb;"
|
|
:: "a" (eax), "b" (ebx), "c" (ecx));
|
|
}
|
|
|
|
static inline void __sti_mwait(unsigned long eax, unsigned long ecx)
|
|
{
|
|
trace_hardirqs_on();
|
|
|
|
mds_idle_clear_cpu_buffers();
|
|
/* "mwait %eax, %ecx;" */
|
|
asm volatile("sti; .byte 0x0f, 0x01, 0xc9;"
|
|
:: "a" (eax), "c" (ecx));
|
|
}
|
|
|
|
/*
|
|
* This uses new MONITOR/MWAIT instructions on P4 processors with PNI,
|
|
* which can obviate IPI to trigger checking of need_resched.
|
|
* We execute MONITOR against need_resched and enter optimized wait state
|
|
* through MWAIT. Whenever someone changes need_resched, we would be woken
|
|
* up from MWAIT (without an IPI).
|
|
*
|
|
* New with Core Duo processors, MWAIT can take some hints based on CPU
|
|
* capability.
|
|
*/
|
|
static inline void mwait_idle_with_hints(unsigned long eax, unsigned long ecx)
|
|
{
|
|
if (static_cpu_has_bug(X86_BUG_MONITOR) || !current_set_polling_and_test()) {
|
|
if (static_cpu_has_bug(X86_BUG_CLFLUSH_MONITOR)) {
|
|
mb();
|
|
clflush((void *)¤t_thread_info()->flags);
|
|
mb();
|
|
}
|
|
|
|
__monitor((void *)¤t_thread_info()->flags, 0, 0);
|
|
if (!need_resched())
|
|
__mwait(eax, ecx);
|
|
}
|
|
current_clr_polling();
|
|
}
|
|
|
|
/*
|
|
* Caller can specify whether to enter C0.1 (low latency, less
|
|
* power saving) or C0.2 state (saves more power, but longer wakeup
|
|
* latency). This may be overridden by the IA32_UMWAIT_CONTROL MSR
|
|
* which can force requests for C0.2 to be downgraded to C0.1.
|
|
*/
|
|
static inline void __tpause(u32 ecx, u32 edx, u32 eax)
|
|
{
|
|
/* "tpause %ecx, %edx, %eax;" */
|
|
#ifdef CONFIG_AS_TPAUSE
|
|
asm volatile("tpause %%ecx\n"
|
|
:
|
|
: "c"(ecx), "d"(edx), "a"(eax));
|
|
#else
|
|
asm volatile(".byte 0x66, 0x0f, 0xae, 0xf1\t\n"
|
|
:
|
|
: "c"(ecx), "d"(edx), "a"(eax));
|
|
#endif
|
|
}
|
|
|
|
#endif /* _ASM_X86_MWAIT_H */
|