target/i386: Add support for perfmon-v2, RAS bits and EPYC-Turin CPU model
v3 cover target/i386: Add support for perfmon-v2, RAS bits and EPYC-Turin CPU model
This series adds the support for following features in qemu.
1. RAS feature bits (SUCCOR, McaOverflowRecov)
2. perfmon-v2
3. Update EPYC-Genoa to support perfmon-v2 and RAS bits
4. Support for bits related to SRSO (sbpb, ibpb-brtype, srso-user-kernel-no)
5. Added support for feature bits CPUID_Fn80000021_EAX/CPUID_Fn80000021_EBX to address CPUID enforcement requirement in Turin platforms.
6. Add support for EPYC-Turin.
---
v3: Added SBPB, IBPB_BRTYPE, SRSO_USER_KERNEL_NO, ERAPS and RAPSIZE bits
to EPYC-Turin.
v2: Fixed couple of typos.
Added Reviewed-by tag from Zhao.
Rebased on top of 6d00c6f98256 ("Merge tag 'for-upstream' of https://repo.or.cz/qemu/kevin into staging")
v2: https://lore.kernel.org/kvm/[email protected]/
v1: https://lore.kernel.org/qemu-devel/[email protected]/
Babu Moger (6):
target/i386: Fix minor typo in NO_NESTED_DATA_BP feature bit target/i386: Add RAS feature bits on EPYC CPU models target/i386: Enable perfmon-v2 and RAS feature bits on EPYC-Genoa target/i386: Expose bits related to SRSO vulnerability target/i386: Expose new feature bits in CPUID 8000_0021_EAX/EBX target/i386: Add support for EPYC-Turin model
Sandipan Das (1):
target/i386: Add PerfMonV2 feature bit
target/i386/cpu.c | 222 +++++++++++++++++++++++++++++++++++++++++++++- target/i386/cpu.h | 27 +++++- 2 files changed, 242 insertions(+), 7 deletions(-)
Signature · 1 line
2.34.1
PATCH v3 1/7 target/i386: Fix minor typo in NO_NESTED_DATA_BP feature bit
Rename CPUID_8000_0021_EAX_No_NESTED_DATA_BP to
CPUID_8000_0021_EAX_NO_NESTED_DATA_BP.
No functional change intended.
---
v3: New patch.
--- target/i386/cpu.c | 2 +- target/i386/cpu.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
target/i386/cpu.c +1 -1
@@ -5226,7 +5226,7 @@ static const X86CPUDefinition builtin_x86_defs[] = {CPUID_8000_0008_EBX_STIBP_ALWAYS_ON | CPUID_8000_0008_EBX_AMD_SSBD | CPUID_8000_0008_EBX_AMD_PSFD, .features[FEAT_8000_0021_EAX ] = - CPUID_8000_0021_EAX_No_NESTED_DATA_BP | +CPUID_8000_0021_EAX_NO_NESTED_DATA_BP |CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING |CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE | CPUID_8000_0021_EAX_AUTO_IBRS,
target/i386/cpu.h +1 -1
@@ -1014,7 +1014,7 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);#define CPUID_8000_0008_EBX_AMD_PSFD (1U << 28)/* Processor ignores nested data breakpoints */ -#define CPUID_8000_0021_EAX_No_NESTED_DATA_BP (1U << 0) +#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0)/* LFENCE is always serializing */ #define CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING (1U << 2)/* Null Selector Clears Base */
Signature · 1 line
2.34.1
PATCH v3 2/7 target/i386: Add RAS feature bits on EPYC CPU models
In reply to Babu MogerAdd the support for following RAS features bits on AMD guests.
SUCCOR: Software uncorrectable error containment and recovery capability.
The processor supports software containment of uncorrectable errors through context synchronizing data poisoning and deferred error interrupts.
McaOverflowRecov: MCA overflow recovery support.
---
v3: No changes
v2: Added reviewed by from Zhao.
--- target/i386/cpu.c | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+)
target/i386/cpu.c +30 -0
@@ -4974,6 +4974,16 @@ static const X86CPUDefinition builtin_x86_defs[] = { }, .cache_info = &epyc_v4_cache_info }, + { + .version = 5, + .props = (PropValue[]) { + { "overflow-recov", "on" }, + { "succor", "on" }, + { "model-id", + "AMD EPYC-v5 Processor" }, + { /* end of list */ } + }, + }, {/* end of list */ } } },
@@ -5112,6 +5122,16 @@ static const X86CPUDefinition builtin_x86_defs[] = { {/* end of list */ } }, }, + { + .version = 5, + .props = (PropValue[]) { + { "overflow-recov", "on" }, + { "succor", "on" }, + { "model-id", + "AMD EPYC-Rome-v5 Processor" }, + { /* end of list */ } + }, + }, {/* end of list */ } } },
@@ -5187,6 +5207,16 @@ static const X86CPUDefinition builtin_x86_defs[] = { }, .cache_info = &epyc_milan_v2_cache_info }, + { + .version = 3, + .props = (PropValue[]) { + { "overflow-recov", "on" }, + { "succor", "on" }, + { "model-id", + "AMD EPYC-Milan-v3 Processor" }, + { /* end of list */ } + }, + }, {/* end of list */ } } },
Signature · 1 line
2.34.1
PATCH v3 3/7 target/i386: Add PerfMonV2 feature bit
In reply to Babu MogerFrom: Sandipan Das <[email protected]>
CPUID leaf 0x80000022, i.e. ExtPerfMonAndDbg, advertises new performance monitoring features for AMD processors. Bit 0 of EAX indicates support for Performance Monitoring Version 2 (PerfMonV2) features. If found to be set during PMU initialization, the EBX bits can be used to determine the number of available counters for different PMUs. It also denotes the availability of global control and status registers.
Add the required CPUID feature word and feature bit to allow guests to make use of the PerfMonV2 features.
---
v3: No changes
v2: Used OR instead of overwrite for eax and ebx.
Added Zhao's Reviewed-by.
--- target/i386/cpu.c | 26 ++++++++++++++++++++++++++ target/i386/cpu.h | 4 ++++ 2 files changed, 30 insertions(+)
target/i386/cpu.c +26 -0
@@ -1227,6 +1227,22 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = { .tcg_features = 0, .unmigratable_flags = 0, }, + [FEAT_8000_0022_EAX] = { + .type = CPUID_FEATURE_WORD, + .feat_names = { +"perfmon-v2" ,NULL ,NULL ,NULL , +NULL ,NULL ,NULL ,NULL , +NULL ,NULL ,NULL ,NULL , +NULL ,NULL ,NULL ,NULL , +NULL ,NULL ,NULL ,NULL , +NULL ,NULL ,NULL ,NULL , +NULL ,NULL ,NULL ,NULL , + NULL, NULL, NULL, NULL, + }, + .cpuid = { .eax = 0x80000022, .reg = R_EAX, }, + .tcg_features =0 , + .unmigratable_flags = 0, + }, [FEAT_XSAVE] = { .type =CPUID_FEATURE_WORD , .feat_names = {
@@ -7040,6 +7056,16 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count, *edx =0 ; }break ; +case 0x80000022 : + *eax = *ebx = *ecx = *edx =0 ; +/* AMD Extended Performance Monitoring and Debug */ +if (kvm_enabled () && cpu->enable_pmu && + (env->features[FEAT_8000_0022_EAX ] &CPUID_8000_0022_EAX_PERFMON_V2 )) { + *eax |=CPUID_8000_0022_EAX_PERFMON_V2 ; + *ebx |=kvm_arch_get_supported_cpuid (cs->kvm_state, index, count, +R_EBX ) &0xf ; + } +break ;case 0xC0000000 : *eax = env->cpuid_xlevel2; *ebx =0 ;
target/i386/cpu.h +4 -0
@@ -634,6 +634,7 @@ typedef enum FeatureWord {FEAT_8000_0007_EDX ,/* CPUID[8000_0007].EDX */ FEAT_8000_0008_EBX ,/* CPUID[8000_0008].EBX */ FEAT_8000_0021_EAX ,/* CPUID[8000_0021].EAX */ +FEAT_8000_0022_EAX ,/* CPUID[8000_0022].EAX */ FEAT_C000_0001_EDX ,/* CPUID[C000_0001].EDX */ FEAT_KVM ,/* CPUID[4000_0001].EAX (KVM_CPUID_FEATURES) */ FEAT_KVM_HINTS, /* CPUID[4000_0001].EDX */
@@ -1022,6 +1023,9 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);/* Automatic IBRS */ #define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +/* Performance Monitoring Version 2 */ +#define CPUID_8000_0022_EAX_PERFMON_V2 (1U << 0) +#define CPUID_XSAVE_XSAVEOPT (1U << 0)#define CPUID_XSAVE_XSAVEC (1U << 1)#define CPUID_XSAVE_XGETBV1 (1U << 2)
Signature · 1 line
2.34.1
PATCH v3 4/7 target/i386: Enable perfmon-v2 and RAS feature bits on EPYC-Genoa
In reply to Babu MogerFollowing feature bits are added on EPYC-Genoa-v2 model.
perfmon-v2: Allow guests to make use of the PerfMonV2 features.
SUCCOR : Software uncorrectable error containment and recovery capability.
The processor supports software containment of uncorrectable errors through context synchronizing data poisoning and deferred error interrupts.
McaOverflowRecov: MCA overflow recovery support.
The feature details are available in APM listed below [1].
[1] AMD64 Architecture Programmer's Manual Volume 2: System Programming Publication # 24593 Revision 3.41.
---
v3: No changes
v2: Minor typo.
Added Reviewed-by from Zhao.
--- target/i386/cpu.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+)
target/i386/cpu.c +15 -0
@@ -5307,6 +5307,21 @@ static const X86CPUDefinition builtin_x86_defs[] = { .xlevel = 0x80000022, .model_id = "AMD EPYC-Genoa Processor", .cache_info = &epyc_genoa_cache_info, + .versions = (X86CPUVersionDefinition []) { + { .version =1 }, + { + .version =2 , + .props = (PropValue []) { + {"overflow-recov" ,"on" }, + {"succor" ,"on" }, + {"perfmon-v2" ,"on" }, + {"model-id" , +"AMD EPYC-Genoa-v2 Processor" }, + {/* end of list */ } + }, + }, + {/* end of list */ } + } }, };
Signature · 1 line
2.34.1
PATCH v3 5/7 target/i386: Expose bits related to SRSO vulnerability
In reply to Babu MogerAdd following bits related Speculative Return Stack Overflow (SRSO). Guests can make use of these bits if supported.
These bits are reported via CPUID Fn8000_0021_EAX.
===================================================================
Bit Feature Description
===================================================================
27 SBPB Indicates support for the Selective Branch Predictor Barrier. 28 IBPB_BRTYPE MSR_PRED_CMD[IBPB] flushes all branch type predictions. 29 SRSO_NO Not vulnerable to SRSO. 30 SRSO_USER_KERNEL_NO Not vulnerable to SRSO at the user-kernel boundary.
===================================================================
---
v3: New patch
--- target/i386/cpu.c | 2 +- target/i386/cpu.h | 14 +++++++++++--- 2 files changed, 12 insertions(+), 4 deletions(-)
target/i386/cpu.c +1 -1
@@ -1221,7 +1221,7 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,"sbpb" , - "ibpb-brtype", NULL, NULL, NULL, + "ibpb-brtype", "srso-no", "srso-user-kernel-no", NULL, }, .cpuid = { .eax = 0x80000021, .reg = R_EAX, }, .tcg_features = 0,
target/i386/cpu.h +11 -3
@@ -1015,13 +1015,21 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);#define CPUID_8000_0008_EBX_AMD_PSFD (1U << 28)/* Processor ignores nested data breakpoints */ -#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0) +#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0)/* LFENCE is always serializing */ #define CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING (1U << 2)/* Null Selector Clears Base */ -#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6) +#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6)/* Automatic IBRS */ -#define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +#define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +/* Selective Branch Predictor Barrier */ +#define CPUID_8000_0021_EAX_SBPB (1U << 27) +/* IBPB includes branch type prediction flushing */ +#define CPUID_8000_0021_EAX_IBPB_BRTYPE (1U << 28) +/* Not vulnerable to Speculative Return Stack Overflow */ +#define CPUID_8000_0021_EAX_SRSO_NO (1U << 29) +/* Not vulnerable to SRSO at the user-kernel boundary */ +#define CPUID_8000_0021_EAX_SRSO_USER_KERNEL_NO (1U << 30)/* Performance Monitoring Version 2 */ #define CPUID_8000_0022_EAX_PERFMON_V2 (1U << 0)
Signature · 1 line
2.34.1
PATCH v3 6/7 target/i386: Expose new feature bits in CPUID 8000_0021_EAX/EBX
In reply to Babu MogerNewer AMD CPUs support ERAPS (Enhanced Return Address Prediction Security) feature that enables the auto-clear of RSB entries on a TLB flush, context switches and VMEXITs. The number of default RSP entries is reflected in RapSize.
Add the feature bit and feature word to support these features.
CPUID_Fn80000021_EAX
Bits Feature Description
24 ERAPS:
Indicates support for enhanced return address predictor security.
CPUID_Fn80000021_EBX
Bits Feature Description 31-24 Reserved 23:16 RapSize:
Return Address Predictor size. RapSize x 8 is the minimum number of CALL instructions software needs to execute to flush the RAP.
15-00 MicrocodePatchSize. Read-only.
Reports the size of the Microcode patch in 16-byte multiples. If 0, the size of the patch is at most 5568 (15C0h) bytes.
---
v3: New patch
--- target/i386/cpu.c | 11 +++++++++-- target/i386/cpu.h | 9 +++++++++ 2 files changed, 18 insertions(+), 2 deletions(-)
target/i386/cpu.c +9 -2
@@ -1220,13 +1220,19 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL , -NULL ,NULL ,NULL ,"sbpb" , +"eraps" ,NULL ,NULL ,"sbpb" , "ibpb-brtype", "srso-no", "srso-user-kernel-no", NULL, }, .cpuid = { .eax = 0x80000021, .reg = R_EAX, }, .tcg_features =0 , .unmigratable_flags = 0, }, + [FEAT_8000_0021_EBX] = { + .type =CPUID_FEATURE_WORD , + .cpuid = { .eax = 0x80000021, .reg = R_EBX, }, + .tcg_features =0 , + .unmigratable_flags = 0, + }, [FEAT_8000_0022_EAX] = { .type =CPUID_FEATURE_WORD , .feat_names = {
@@ -7114,8 +7120,9 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count, }break ;case 0x80000021 : + *eax = *ebx = *ecx = *edx =0 ; *eax = env->features[FEAT_8000_0021_EAX ]; - *ebx = *ecx = *edx =0 ; + *ebx = env->features[FEAT_8000_0021_EBX ];break ;default :/* reserved values: zero */
target/i386/cpu.h +9 -0
@@ -634,6 +634,7 @@ typedef enum FeatureWord {FEAT_8000_0007_EDX ,/* CPUID[8000_0007].EDX */ FEAT_8000_0008_EBX ,/* CPUID[8000_0008].EBX */ FEAT_8000_0021_EAX ,/* CPUID[8000_0021].EAX */ +FEAT_8000_0021_EBX ,/* CPUID[8000_0021].EBX */ FEAT_8000_0022_EAX ,/* CPUID[8000_0022].EAX */ FEAT_C000_0001_EDX ,/* CPUID[C000_0001].EDX */ FEAT_KVM, /* CPUID[4000_0001].EAX (KVM_CPUID_FEATURES) */
@@ -1022,6 +1023,8 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6)/* Automatic IBRS */ #define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +/* Enhanced Return Address Predictor Scurity */ +#define CPUID_8000_0021_EAX_ERAPS (1U << 24)/* Selective Branch Predictor Barrier */ #define CPUID_8000_0021_EAX_SBPB (1U << 27)/* IBPB includes branch type prediction flushing */
@@ -1031,6 +1034,12 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);/* Not vulnerable to SRSO at the user-kernel boundary */ #define CPUID_8000_0021_EAX_SRSO_USER_KERNEL_NO (1U << 30) +/* +* Return Address Predictor size. RapSize x 8 is the minimum number of +* CALL instructions software needs to execute to flush the RAP. +*/ +#define CPUID_8000_0021_EBX_RAPSIZE (8U << 16) +/* Performance Monitoring Version 2 */ #define CPUID_8000_0022_EAX_PERFMON_V2 (1U << 0)
Signature · 1 line
2.34.1
PATCH v3 7/7 target/i386: Add support for EPYC-Turin model
In reply to Babu MogerAdd the support for AMD EPYC zen 5 processors (EPYC-Turin).
Add the following new feature bits on top of the feature bits from the previous generation EPYC models.
movdiri : Move Doubleword as Direct Store Instruction
movdir64b : Move 64 Bytes as Direct Store Instruction
avx512-vp2intersect : AVX512 Vector Pair Intersection to a Pair
of Mask Register
avx-vnni : AVX VNNI Instruction
sbpb : Selective Branch Predictor Barrier
ibpb-brtype : IBPB includes branch type prediction flushing
srso-user-kernel-no : Not vulnerable to SRSO at the user-kernel boundary
eraps : Enhanced Return Address Predictor Security
rapsize : Return Address Predictor size. RapSize x 8 is the
minimum number of CALL instructions software needs to execute to flush the RAP
---
v3: Removed Zhao's Reviewed-by as the patch has changed,
Added sbpb, ibpb-brtype, srso-user-kernel-no, eraps, rapsize
v2: Fixed minor typo.
Added Zhao's Reviewed-by.
--- target/i386/cpu.c | 136 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 136 insertions(+)
target/i386/cpu.c +136 -0
@@ -2411,6 +2411,60 @@ static const CPUCaches epyc_genoa_cache_info = { }, }; +static const CPUCaches epyc_turin_cache_info = { + .l1d_cache = &(CPUCacheInfo ) { + .type =DATA_CACHE , + .level =1 , + .size =48 * KiB, + .line_size =64 , + .associativity =12 , + .partitions =1 , + .sets =64 , + .lines_per_tag =1 , + .self_init =1 , + .no_invd_sharing = true, + .share_level =CPU_TOPO_LEVEL_CORE , + }, + .l1i_cache = &(CPUCacheInfo ) { + .type =INSTRUCTION_CACHE , + .level =1 , + .size =32 * KiB, + .line_size =64 , + .associativity =8 , + .partitions =1 , + .sets =64 , + .lines_per_tag =1 , + .self_init =1 , + .no_invd_sharing = true, + .share_level =CPU_TOPO_LEVEL_CORE , + }, + .l2_cache = &(CPUCacheInfo ) { + .type =UNIFIED_CACHE , + .level =2 , + .size =1 * MiB, + .line_size =64 , + .associativity =16 , + .partitions =1 , + .sets =1024 , + .lines_per_tag =1 , + .share_level =CPU_TOPO_LEVEL_CORE , + }, + .l3_cache = &(CPUCacheInfo ) { + .type =UNIFIED_CACHE , + .level =3 , + .size =32 * MiB, + .line_size =64 , + .associativity =16 , + .partitions =1 , + .sets =32768 , + .lines_per_tag =1 , + .self_init = true, + .inclusive = true, + .complex_indexing = false, + .share_level =CPU_TOPO_LEVEL_DIE , + }, +}; +/* The following VMX features are not supported by KVM and are left out in the * CPU definitions: *
@@ -5329,6 +5383,88 @@ static const X86CPUDefinition builtin_x86_defs[] = { {/* end of list */ } } }, + { + .name = "EPYC-Turin", + .level = 0xd, + .vendor = CPUID_VENDOR_AMD, + .family = 26, + .model = 0, + .stepping = 0, + .features[FEAT_1_ECX ] = +CPUID_EXT_RDRAND |CPUID_EXT_F16C |CPUID_EXT_AVX | +CPUID_EXT_XSAVE |CPUID_EXT_AES |CPUID_EXT_POPCNT | +CPUID_EXT_MOVBE |CPUID_EXT_SSE42 |CPUID_EXT_SSE41 | +CPUID_EXT_PCID |CPUID_EXT_CX16 |CPUID_EXT_FMA | +CPUID_EXT_SSSE3 |CPUID_EXT_MONITOR |CPUID_EXT_PCLMULQDQ | + CPUID_EXT_SSE3, + .features[FEAT_1_EDX ] = +CPUID_SSE2 |CPUID_SSE |CPUID_FXSR |CPUID_MMX |CPUID_CLFLUSH | +CPUID_PSE36 |CPUID_PAT |CPUID_CMOV |CPUID_MCA |CPUID_PGE | +CPUID_MTRR |CPUID_SEP |CPUID_APIC |CPUID_CX8 |CPUID_MCE | +CPUID_PAE |CPUID_MSR |CPUID_TSC |CPUID_PSE |CPUID_DE | + CPUID_VME | CPUID_FP87, + .features[FEAT_6_EAX ] = + CPUID_6_EAX_ARAT, + .features[FEAT_7_0_EBX ] = +CPUID_7_0_EBX_FSGSBASE |CPUID_7_0_EBX_BMI1 |CPUID_7_0_EBX_AVX2 | +CPUID_7_0_EBX_SMEP |CPUID_7_0_EBX_BMI2 |CPUID_7_0_EBX_ERMS | +CPUID_7_0_EBX_INVPCID |CPUID_7_0_EBX_AVX512F | +CPUID_7_0_EBX_AVX512DQ |CPUID_7_0_EBX_RDSEED |CPUID_7_0_EBX_ADX | +CPUID_7_0_EBX_SMAP |CPUID_7_0_EBX_AVX512IFMA | +CPUID_7_0_EBX_CLFLUSHOPT |CPUID_7_0_EBX_CLWB | +CPUID_7_0_EBX_AVX512CD |CPUID_7_0_EBX_SHA_NI | + CPUID_7_0_EBX_AVX512BW | CPUID_7_0_EBX_AVX512VL, + .features[FEAT_7_0_ECX ] = +CPUID_7_0_ECX_AVX512_VBMI |CPUID_7_0_ECX_UMIP |CPUID_7_0_ECX_PKU | +CPUID_7_0_ECX_AVX512_VBMI2 |CPUID_7_0_ECX_GFNI | +CPUID_7_0_ECX_VAES |CPUID_7_0_ECX_VPCLMULQDQ | +CPUID_7_0_ECX_AVX512VNNI |CPUID_7_0_ECX_AVX512BITALG | +CPUID_7_0_ECX_AVX512_VPOPCNTDQ |CPUID_7_0_ECX_LA57 | +CPUID_7_0_ECX_RDPID |CPUID_7_0_ECX_MOVDIRI | + CPUID_7_0_ECX_MOVDIR64B, + .features[FEAT_7_0_EDX ] = + CPUID_7_0_EDX_FSRM | CPUID_7_0_EDX_AVX512_VP2INTERSECT, + .features[FEAT_7_1_EAX ] = + CPUID_7_1_EAX_AVX_VNNI | CPUID_7_1_EAX_AVX512_BF16, + .features[FEAT_8000_0001_ECX ] = +CPUID_EXT3_OSVW |CPUID_EXT3_3DNOWPREFETCH | +CPUID_EXT3_MISALIGNSSE |CPUID_EXT3_SSE4A |CPUID_EXT3_ABM | +CPUID_EXT3_CR8LEG |CPUID_EXT3_SVM |CPUID_EXT3_LAHF_LM | + CPUID_EXT3_TOPOEXT | CPUID_EXT3_PERFCORE, + .features[FEAT_8000_0001_EDX ] = +CPUID_EXT2_LM |CPUID_EXT2_RDTSCP |CPUID_EXT2_PDPE1GB | +CPUID_EXT2_FFXSR |CPUID_EXT2_MMXEXT |CPUID_EXT2_NX | + CPUID_EXT2_SYSCALL, + .features[FEAT_8000_0007_EBX ] = + CPUID_8000_0007_EBX_OVERFLOW_RECOV | CPUID_8000_0007_EBX_SUCCOR, + .features[FEAT_8000_0008_EBX ] = +CPUID_8000_0008_EBX_CLZERO |CPUID_8000_0008_EBX_XSAVEERPTR | +CPUID_8000_0008_EBX_WBNOINVD |CPUID_8000_0008_EBX_IBPB | +CPUID_8000_0008_EBX_IBRS |CPUID_8000_0008_EBX_STIBP | +CPUID_8000_0008_EBX_STIBP_ALWAYS_ON | + CPUID_8000_0008_EBX_AMD_SSBD | CPUID_8000_0008_EBX_AMD_PSFD, + .features[FEAT_8000_0021_EAX ] = +CPUID_8000_0021_EAX_NO_NESTED_DATA_BP | +CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING | +CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE | +CPUID_8000_0021_EAX_AUTO_IBRS | +CPUID_8000_0021_EAX_ERAPS |CPUID_8000_0021_EAX_SBPB | +CPUID_8000_0021_EAX_IBPB_BRTYPE | + CPUID_8000_0021_EAX_SRSO_USER_KERNEL_NO, + .features[FEAT_8000_0021_EBX ] = + CPUID_8000_0021_EBX_RAPSIZE, + .features[FEAT_8000_0022_EAX ] = + CPUID_8000_0022_EAX_PERFMON_V2, + .features[FEAT_XSAVE ] = +CPUID_XSAVE_XSAVEOPT |CPUID_XSAVE_XSAVEC | + CPUID_XSAVE_XGETBV1 | CPUID_XSAVE_XSAVES, + .features[FEAT_SVM ] = +CPUID_SVM_NPT |CPUID_SVM_NRIPSAVE |CPUID_SVM_VNMI | + CPUID_SVM_SVME_ADDR_CHK, + .xlevel = 0x80000022, + .model_id = "AMD EPYC-Turin Processor", + .cache_info = &epyc_turin_cache_info, + }, };/*
Signature · 1 line
2.34.1
Re: [PATCH v3 1/7] target/i386: Fix minor typo in NO_NESTED_DATA_BP feature bit
In reply to Babu MogerQuoted 11 lines · Babu Moger
Date: Thu, 24 Oct 2024 17:18:19 -0500
From: Babu Moger <[email protected]>
Subject: [PATCH v3 1/7] target/i386: Fix minor typo inNO_NESTED_DATA_BP
feature bit
X-Mailer: git-send-email 2.34.1Rename
CPUID_8000_0021_EAX_No_NESTED_DATA_BPtoCPUID_8000_0021_EAX_NO_NESTED_DATA_BP.No functional change intended.
Quoted 7 lines
Signed-off-by Babu Moger <[email protected]>---
v3: New patch.--- target/i386/cpu.c | 2 +- target/i386/cpu.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Re: [PATCH v3 2/7] target/i386: Add RAS feature bits on EPYC CPU models
In reply to Babu Moger(+John)
Hi Babu,
This patch is fine for me.
However, users recently reported an issue with SUCCOR support on AMD
hosts: https://gitlab.com/qemu-project/qemu/-/issues/2571.
Could you please double check and clarify that issue on AMD host?
Thanks,
Zhao
Quoted 23 lines · Babu Moger
Date: Thu, 24 Oct 2024 17:18:20 -0500
From: Babu Moger <[email protected]>
Subject: [PATCH v3 2/7] target/i386: Add RAS feature bits on EPYC CPU models
X-Mailer: git-send-email 2.34.1Add the support for following RAS features bits on AMD guests.
SUCCOR: Software uncorrectable error containment and recovery capability.
The processor supports software containment of uncorrectable errors through context synchronizing data poisoning and deferred error interrupts.McaOverflowRecov: MCA overflow recovery support.
Reviewed-by Zhao Liu <[email protected]>Signed-off-by Babu Moger <[email protected]>---
v3: No changesv2: Added reviewed by from Zhao.
--- target/i386/cpu.c | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+)
Quoted 9 lines · Babu Moger
This series adds the support for following features in qemu.
1. RAS feature bits (SUCCOR, McaOverflowRecov)
2. perfmon-v2
3. Update EPYC-Genoa to support perfmon-v2 and RAS bits
4. Support for bits related to SRSO (sbpb, ibpb-brtype, srso-user-kernel-no)
5. Added support for feature bits CPUID_Fn80000021_EAX/CPUID_Fn80000021_EBX to address CPUID enforcement requirement in Turin platforms.
6. Add support for EPYC-Turin.
Queued, thanks. I looked at https://gitlab.com/qemu-project/qemu/-/issues/2571 and I think it's caused by the ignore_msrs=1 parameter on the KVM kernel module.
However, can you look into adding new CPUID_SVM_* bits?
Thanks,
Paolo
Quoted 29 lines
---
v3: Added SBPB,IBPB_BRTYPE,SRSO_USER_KERNEL_NO, ERAPS and RAPSIZE bits
to EPYC-Turin.v2: Fixed couple of typos.
Added Reviewed-by tag from Zhao.
Rebased on top of6d00c6f98256("Merge tag 'for-upstream' of https://repo.or.cz/qemu/kevin into staging")v2: https://lore.kernel.org/kvm/[email protected]/
v1: https://lore.kernel.org/qemu-devel/[email protected]/
Babu Moger (6):
target/i386: Fix minor typo inNO_NESTED_DATA_BPfeature bit target/i386: Add RAS feature bits on EPYC CPU models target/i386: Enable perfmon-v2 and RAS feature bits on EPYC-Genoa target/i386: Expose bits related to SRSO vulnerability target/i386: Expose new feature bits in CPUID 8000_0021_EAX/EBX target/i386: Add support for EPYC-Turin modelSandipan Das (1):
target/i386: Add PerfMonV2 feature bittarget/i386/cpu.c | 222 +++++++++++++++++++++++++++++++++++++++++++++- target/i386/cpu.h | 27 +++++- 2 files changed, 242 insertions(+), 7 deletions(-)
Re: [PATCH v3 5/7] target/i386: Expose bits related to SRSO vulnerability
In reply to Babu MogerHi Babu
Quoted 13 lines · Babu Moger
Date: Thu, 24 Oct 2024 17:18:23 -0500
From: Babu Moger <[email protected]>
Subject: [PATCH v3 5/7] target/i386: Expose bits related to SRSO
vulnerability
X-Mailer: git-send-email 2.34.1Add following bits related Speculative Return Stack Overflow (SRSO). Guests can make use of these bits if supported.
These bits are reported via CPUID
Fn8000_0021_EAX.
===================================================================
Bit Feature Description
===================================================================
Quoted 8 lines
27 SBPB Indicates support for the Selective Branch Predictor Barrier. 28 IBPB_BRTYPE MSR_PRED_CMD[IBPB] flushes all branch type predictions. 29 SRSO_NO Not vulnerable to SRSO. 30 SRSO_USER_KERNEL_NO Not vulnerable to SRSO at the user-kernel boundary.===================================================================
I suggest updating the description of SRSO-related mitigations in the "Important CPU features for AMD x86 hosts" section of docs/system/ cpu-models-x86.rst.inc.
If you could also synchronize the CPU model (you added in this series) in the "Preferred CPU models for AMD x86 hosts" section, that would be even better. :-)
Quoted 37 lines
Signed-off-by Babu Moger <[email protected]>---
v3: New patch--- target/i386/cpu.c | 2 +- target/i386/cpu.h | 14 +++++++++++--- 2 files changed, 12 insertions(+), 4 deletions(-)target/i386/cpu.c +1
-1@@ -1221,7 +1221,7 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,"sbpb" , - "ibpb-brtype", NULL, NULL, NULL, + "ibpb-brtype", "srso-no", "srso-user-kernel-no", NULL, }, .cpuid = { .eax = 0x80000021, .reg = R_EAX, }, .tcg_features = 0,target/i386/cpu.h +2
-2@@ -1015,13 +1015,21 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);#define CPUID_8000_0008_EBX_AMD_PSFD (1U << 28)/* Processor ignores nested data breakpoints */ -#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0) +#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0)/* LFENCE is always serializing */ #define CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING (1U << 2)/* Null Selector Clears Base */ -#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6) +#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6)/* Automatic IBRS */
target/i386/cpu.h uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w); · line 1032
Quoted 10 lines
target/i386/cpu.h +9
-1-#define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +#define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +/* Selective Branch Predictor Barrier */ +#define CPUID_8000_0021_EAX_SBPB (1U << 27) +/* IBPB includes branch type prediction flushing */ +#define CPUID_8000_0021_EAX_IBPB_BRTYPE (1U << 28) +/* Not vulnerable to Speculative Return Stack Overflow */ +#define CPUID_8000_0021_EAX_SRSO_NO (1U << 29) +/* Not vulnerable to SRSO at the user-kernel boundary */ +#define CPUID_8000_0021_EAX_SRSO_USER_KERNEL_NO (1U << 30)
These feature bits defination could be added in patch 7 because only patch 7 uses these macros.
BTW, which platform supports CPUID_8000_0021_EAX_SRSO_NO? I found that even the Turin model added in patch 7 does not support this feature.
Thanks,
Zhao
Quoted 6 lines
target/i386/cpu.h +0
-0/* Performance Monitoring Version 2 */ #define CPUID_8000_0022_EAX_PERFMON_V2 (1U << 0)Signature · 1 line
2.34.1
Hi Paolo,
Quoted 12 lines · Paolo Bonzini
Quoted 10 lines · Babu Moger
This series adds the support for following features in qemu.
1. RAS feature bits (SUCCOR, McaOverflowRecov)
2. perfmon-v2
3. Update EPYC-Genoa to support perfmon-v2 and RAS bits
4. Support for bits related to SRSO (sbpb, ibpb-brtype, srso-user-kernel-no)
5. Added support for feature bits CPUID_Fn80000021_EAX/CPUID_Fn80000021_EBX to address CPUID enforcement requirement in Turin platforms.
6. Add support for EPYC-Turin.
Quoted 1 line
Queued, thanks. I looked at
Thanks.
Quoted 2 lines
https://gitlab.com/qemu-project/qemu/-/issues/2571 and I think it's caused by the ignore_msrs=1 parameter on the KVM kernel module.
Thanks again.
Quoted 2 lines
However, can you look into adding new CPUID_SVM_* bits?
I normally pickup bits when it is added in kernel/kvm. Are you thinking of any specific bits here?
Thanks
Babu
Paolo Bonzini
maintainerQuoted 23 lines · Babu
Hi Paolo,
Quoted 12 lines · Paolo Bonzini
Quoted 10 lines · Babu Moger
This series adds the support for following features in qemu.
1. RAS feature bits (SUCCOR, McaOverflowRecov)
2. perfmon-v2
3. Update EPYC-Genoa to support perfmon-v2 and RAS bits
4. Support for bits related to SRSO (sbpb, ibpb-brtype, srso-user-kernel-no)
5. Added support for feature bits CPUID_Fn80000021_EAX/CPUID_Fn80000021_EBX to address CPUID enforcement requirement in Turin platforms.
6. Add support for EPYC-Turin.Quoted 1 line
Queued, thanks. I looked at
Thanks.
Quoted 2 lines
https://gitlab.com/qemu-project/qemu/-/issues/2571 and I think it's caused by the ignore_msrs=1 parameter on the KVM kernel module.
Quoted 7 lines
Thanks again.
Quoted 2 lines
However, can you look into adding new CPUID_SVM_* bits?
I normally pickup bits when it is added in kernel/kvm. Are you thinking of any specific bits here?
Yes, KVM already supports vGIF, virtual VMLOAD/VMSAVE, virtual TSC rate MSR, vNMI, virtual LBR, virtual pause filter and virtual pause filter threshold.
Paolo
Re: [PATCH v3 5/7] target/i386: Expose bits related to SRSO vulnerability
In reply to Zhao LiuHi Zhao,
Thanks for the review.
Quoted 25 lines · Zhao Liu
Hi Babu
Quoted 13 lines · Babu Moger
Date: Thu, 24 Oct 2024 17:18:23 -0500
From: Babu Moger <[email protected]>
Subject: [PATCH v3 5/7] target/i386: Expose bits related to SRSO
vulnerability
X-Mailer: git-send-email 2.34.1Add following bits related Speculative Return Stack Overflow (SRSO). Guests can make use of these bits if supported.
These bits are reported via CPUID
Fn8000_0021_EAX.
===================================================================
Bit Feature Description
===================================================================Quoted 8 lines
27 SBPB Indicates support for the Selective Branch Predictor Barrier. 28 IBPB_BRTYPE MSR_PRED_CMD[IBPB] flushes all branch type predictions. 29 SRSO_NO Not vulnerable to SRSO. 30 SRSO_USER_KERNEL_NO Not vulnerable to SRSO at the user-kernel boundary.===================================================================
Quoted 7 lines
I suggest updating the description of SRSO-related mitigations in the "Important CPU features for AMD x86 hosts" section of docs/system/ cpu-models-x86.rst.inc.
If you could also synchronize the CPU model (you added in this series) in the "Preferred CPU models for AMD x86 hosts" section, that would be even better. :-)
Sure. Will look into both of these.
Quoted 43 lines
Quoted 43 lines
Signed-off-by Babu Moger <[email protected]>---
v3: New patch--- target/i386/cpu.c | 2 +- target/i386/cpu.h | 14 +++++++++++--- 2 files changed, 12 insertions(+), 4 deletions(-)target/i386/cpu.c +1
-1@@ -1221,7 +1221,7 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,NULL ,"sbpb" , - "ibpb-brtype", NULL, NULL, NULL, + "ibpb-brtype", "srso-no", "srso-user-kernel-no", NULL, }, .cpuid = { .eax = 0x80000021, .reg = R_EAX, }, .tcg_features = 0,target/i386/cpu.h +6
-3@@ -1015,13 +1015,21 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);#define CPUID_8000_0008_EBX_AMD_PSFD (1U << 28)/* Processor ignores nested data breakpoints */ -#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0) +#define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP (1U << 0)/* LFENCE is always serializing */ #define CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING (1U << 2)/* Null Selector Clears Base */ -#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6) +#define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE (1U << 6)/* Automatic IBRS */ -#define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +#define CPUID_8000_0021_EAX_AUTO_IBRS (1U << 8) +/* Selective Branch Predictor Barrier */ +#define CPUID_8000_0021_EAX_SBPB (1U << 27) +/* IBPB includes branch type prediction flushing */
Quoted 8 lines
target/i386/cpu.h uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w); · line 1032
Quoted 5 lines
target/i386/cpu.h +5
-0+#define CPUID_8000_0021_EAX_IBPB_BRTYPE (1U << 28) +/* Not vulnerable to Speculative Return Stack Overflow */ +#define CPUID_8000_0021_EAX_SRSO_NO (1U << 29) +/* Not vulnerable to SRSO at the user-kernel boundary */ +#define CPUID_8000_0021_EAX_SRSO_USER_KERNEL_NO (1U << 30)These feature bits defination could be added in patch 7 because only patch 7 uses these macros.
Sure. Looks like Paolo already pulled this in. I will take care of this if there is another revision.
Quoted 2 lines
BTW, which platform supports
CPUID_8000_0021_EAX_SRSO_NO? I found that even the Turin model added in patch 7 does not support this feature.
SRSO_NO is not supported in Turin. I added it for completion as it is in the same CPUID fn.
Quoted 11 lines
Thanks,
ZhaoQuoted 6 lines
target/i386/cpu.h +0
-0/* Performance Monitoring Version 2 */ #define CPUID_8000_0022_EAX_PERFMON_V2 (1U << 0)Signature · 1 line
2.34.1
Signature · 2 lines
Thanks Babu Moger
Quoted 27 lines · Paolo Bonzini
Quoted 26 lines · Babu
Hi Paolo,
Quoted 12 lines · Paolo Bonzini
Quoted 10 lines · Babu Moger
This series adds the support for following features in qemu.
1. RAS feature bits (SUCCOR, McaOverflowRecov)
2. perfmon-v2
3. Update EPYC-Genoa to support perfmon-v2 and RAS bits
4. Support for bits related to SRSO (sbpb, ibpb-brtype, srso-user-kernel-no)
5. Added support for feature bits CPUID_Fn80000021_EAX/CPUID_Fn80000021_EBX to address CPUID enforcement requirement in Turin platforms.
6. Add support for EPYC-Turin.Quoted 1 line
Queued, thanks. I looked at
Thanks.
Quoted 2 lines
https://gitlab.com/qemu-project/qemu/-/issues/2571 and I think it's caused by the ignore_msrs=1 parameter on the KVM kernel module.
Thanks again.
Quoted 1 line
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
However, can you look into adding new CPUID_SVM_* bits?
I normally pickup bits when it is added in kernel/kvm. Are you thinking of any specific bits here?
Yes, KVM already supports vGIF, virtual VMLOAD/VMSAVE, virtual TSC rate MSR, vNMI, virtual LBR, virtual pause filter and virtual pause filter threshold.
Oh ok. Sure. I will look into this.
Signature · 2 lines
Thanks Babu Moger
Paolo Bonzini
maintainerQuoted 9 lines · Babu Moger
This series adds the support for following features in qemu.
1. RAS feature bits (SUCCOR, McaOverflowRecov)
2. perfmon-v2
3. Update EPYC-Genoa to support perfmon-v2 and RAS bits
4. Support for bits related to SRSO (sbpb, ibpb-brtype, srso-user-kernel-no)
5. Added support for feature bits CPUID_Fn80000021_EAX/CPUID_Fn80000021_EBX to address CPUID enforcement requirement in Turin platforms.
6. Add support for EPYC-Turin.
Queued the following
- target/i386: Fix minor typo in NO_NESTED_DATA_BP feature bit
- target/i386: Add PerfMonV2 feature bit
- target/i386: Expose bits related to SRSO vulnerability
- target/i386: Expose new feature bits in CPUID 8000_0021_EAX/EBX
I am leaving out the CPU model updates so that they are updated just once instead of twice (especially Turin should have the right CPUID bits for nested SVM).
Paolo
Re: [PATCH v3 2/7] target/i386: Add RAS feature bits on EPYC CPU models
In reply to Zhao LiuQuoted 10 lines · Zhao Liu
(+John)
Hi Babu,
This patch is fine for me.
However, users recently reported an issue with SUCCOR support on AMD
hosts: https://gitlab.com/qemu-project/qemu/-/issues/2571.Could you please double check and clarify that issue on AMD host?
Thanks for the heads up. I can reproduce this on an AMD host with kvm.ignore_msrs=1. It seems like kvm will need to block this feature on the guest when ignore_msrs is set, but it's not clear to me how that should be done yet.
Paolo,
Do you have any ideas about how we should handle this situation?
Thanks,
John
Quoted 29 lines
Thanks,
ZhaoQuoted 23 lines · Babu Moger
Date: Thu, 24 Oct 2024 17:18:20 -0500
From: Babu Moger <[email protected]>
Subject: [PATCH v3 2/7] target/i386: Add RAS feature bits on EPYC CPU models
X-Mailer: git-send-email 2.34.1Add the support for following RAS features bits on AMD guests.
SUCCOR: Software uncorrectable error containment and recovery capability.
The processor supports software containment of uncorrectable errors through context synchronizing data poisoning and deferred error interrupts.McaOverflowRecov: MCA overflow recovery support.
Reviewed-by Zhao Liu <[email protected]>Signed-off-by Babu Moger <[email protected]>---
v3: No changesv2: Added reviewed by from Zhao.
--- target/i386/cpu.c | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+)