Skip to contentSkip to conversation

amd/msr: implement MSR_VIRT_SPEC_CTRL for HVM guests

PATCH Superseded Roger Pau Monne 3 participants 13 messages Mar 31, 2022 – Apr 25, 2022

v3 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v3 cover amd/msr: implement MSR_VIRT_SPEC_CTRL for HVM guests

Hello,

The following series implements support for MSR_VIRT_SPEC_CTRL (VIRT_SSBD) on different AMD CPU families.

Note that the support is added backwards, starting with the newer CPUs that support MSR_SPEC_CTRL and moving to the older ones either using MSR_VIRT_SPEC_CTRL or the SSBD bit in LS_CFG.

Xen is still free to use it's own SSBD setting, as the selection is context switched on vm{entry,exit}.

On Zen2 and later, SPEC_CTRL.SSBD exists and should be used in preference to VIRT_SPEC_CTRL.SSBD.  However, for migration compatibility, Xen offers VIRT_SSBD to guests (in the max CPUID policy, not default) implemented in terms of SPEC_CTRL.SSBD.

On Fam15h thru Zen1, Xen exposes VIRT_SSBD to guests by default to abstract away the model and/or hypervisor specific differences in MSR_LS_CFG/MSR_VIRT_SPEC_CTRL.

Note that if the hardware itself does offer VIRT_SSBD (ie: very likely when running virtualized on < Zen2 hardware) and not AMD_SSBD Xen will allow untrapped access to MSR_VIRT_SPEC_CTRL for HVM guests.

So the implementation of VIRT_SSBD exposed to HVM guests will use one of the following underlying mechanisms, in the preference order listed below:

* SPEC_CTRL.SSBD. (patch 1)
* VIRT_SPEC_CTRL.SSBD (untrapped). (patch 2).
* Non-architectural way using LS_CFG. (patch 3)

This has survived a XenRT basic set of tests on AMD machines.

Roger Pau Monne (3):
amd/msr: implement VIRT_SPEC_CTRL for HVM guests on top of SPEC_CTRL amd/msr: allow passthrough of VIRT_SPEC_CTRL for HVM guests amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

 xen/arch/x86/cpu/amd.c                      | 116 +++++++++++++++++---
 xen/arch/x86/cpuid.c                        |  28 +++++
 xen/arch/x86/hvm/hvm.c                      |   1 +
 xen/arch/x86/hvm/svm/entry.S                |   6 +
 xen/arch/x86/hvm/svm/svm.c                  |  50 +++++++++
 xen/arch/x86/include/asm/amd.h              |   4 +
 xen/arch/x86/include/asm/cpufeatures.h      |   1 +
 xen/arch/x86/include/asm/msr.h              |  14 +++
 xen/arch/x86/msr.c                          |  26 +++++
 xen/arch/x86/spec_ctrl.c                    |  12 +-
 xen/include/public/arch-x86/cpufeatureset.h |   2 +-
 11 files changed, 241 insertions(+), 19 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH v3 1/3 amd/msr: implement VIRT_SPEC_CTRL for HVM guests on top of SPEC_CTRL

Use the logic to set shadow SPEC_CTRL values in order to implement support for VIRT_SPEC_CTRL (signaled by VIRT_SSBD CPUID flag) for HVM guests. This includes using the spec_ctrl vCPU MSR variable to store the guest set value of VIRT_SPEC_CTRL.SSBD, which will be OR'ed with any SPEC_CTRL values being set by the guest.

On hardware having SPEC_CTRL VIRT_SPEC_CTRL will not be offered by default to guests. VIRT_SPEC_CTRL will only be part of the max CPUID policy so it can be enabled for compatibility purposes.

Some reasoning regarding why '!s' is used to annotate the feature:
* '!': the feature might be exposed to guests even when not present on the host hardware.
* 's': the feature won't be exposed by default.

---
Changes since v2:
- Reword reasoning for using '!s'.
- Trim comment about only setting SSBD bit in spec_ctrl.raw.

Changes since v1:
- Only expose VIRT_SSBD if AMD_SSBD is available on the host.
- Revert change to msr-sc= command line option documentation.
- Only set or clear the SSBD bit of spec_ctrl.

---
 xen/arch/x86/cpuid.c                        |  7 +++++++
 xen/arch/x86/hvm/hvm.c                      |  1 +
 xen/arch/x86/include/asm/msr.h              |  4 ++++
 xen/arch/x86/msr.c                          | 18 ++++++++++++++++++
 xen/arch/x86/spec_ctrl.c                    |  3 ++-
 xen/include/public/arch-x86/cpufeatureset.h |  2 +-
 6 files changed, 33 insertions(+), 2 deletions(-)
xen/arch/x86/cpuid.c +7 -0
index bb554b06a7..4ca77ea870 100644
@@ -543,6 +543,13 @@ static void __init calculate_hvm_max_policy(void)
         __clear_bit(X86_FEATURE_IBRSB, hvm_featureset);
         __clear_bit(X86_FEATURE_IBRS, hvm_featureset);
     }
+    else if ( boot_cpu_has(X86_FEATURE_AMD_SSBD) )
+        /*
+         * If SPEC_CTRL.SSBD is available VIRT_SPEC_CTRL.SSBD can be exposed
+         * and implemented using the former. Expose in the max policy only as
+         * the preference is for guests to use SPEC_CTRL.SSBD if available.
+         */
+        __set_bit(X86_FEATURE_VIRT_SSBD, hvm_featureset);
 
     /*
      * With VT-x, some features are only supported by Xen if dedicated
xen/arch/x86/hvm/hvm.c +1 -0
index 709a4191ef..595858f2a7 100644
@@ -1334,6 +1334,7 @@ static const uint32_t msrs_to_send[] = {
     MSR_INTEL_MISC_FEATURES_ENABLES,
     MSR_IA32_BNDCFGS,
     MSR_IA32_XSS,
+    MSR_VIRT_SPEC_CTRL,
     MSR_AMD64_DR0_ADDRESS_MASK,
     MSR_AMD64_DR1_ADDRESS_MASK,
     MSR_AMD64_DR2_ADDRESS_MASK,
xen/arch/x86/include/asm/msr.h +4 -0
index ce4fe51afe..ab6fbb5051 100644
@@ -291,6 +291,7 @@ struct vcpu_msrs
 {
     /*
      * 0x00000048 - MSR_SPEC_CTRL
+     * 0xc001011f - MSR_VIRT_SPEC_CTRL (if X86_FEATURE_AMD_SSBD)
      *
      * For PV guests, this holds the guest kernel value.  It is accessed on
      * every entry/exit path.
@@ -306,6 +307,9 @@ struct vcpu_msrs
      * We must clear/restore Xen's value before/after VMRUN to avoid unduly
      * influencing the guest.  In order to support "behind the guest's back"
      * protections, we load this value (commonly 0) before VMRUN.
+     *
+     * Once of such "behind the guest's back" usages is setting SPEC_CTRL.SSBD
+     * if the guest sets VIRT_SPEC_CTRL.SSBD.
      */
     struct {
         uint32_t raw;
xen/arch/x86/msr.c +18 -0
index 01a15857b7..72c175fd8b 100644
@@ -381,6 +381,13 @@ int guest_rdmsr(struct vcpu *v, uint32_t msr, uint64_t *val)
                ? K8_HWCR_TSC_FREQ_SEL : 0;
         break;
 
+    case MSR_VIRT_SPEC_CTRL:
+        if ( !cp->extd.virt_ssbd )
+            goto gp_fault;
+
+        *val = msrs->spec_ctrl.raw & SPEC_CTRL_SSBD;
+        break;
+
     case MSR_AMD64_DE_CFG:
         if ( !(cp->x86_vendor & (X86_VENDOR_AMD | X86_VENDOR_HYGON)) )
             goto gp_fault;
@@ -666,6 +673,17 @@ int guest_wrmsr(struct vcpu *v, uint32_t msr, uint64_t val)
             wrmsr_tsc_aux(val);
         break;
 
+    case MSR_VIRT_SPEC_CTRL:
+        if ( !cp->extd.virt_ssbd )
+            goto gp_fault;
+
+        /* Only supports SSBD bit, the rest are ignored. */
+        if ( val & SPEC_CTRL_SSBD )
+            msrs->spec_ctrl.raw |= SPEC_CTRL_SSBD;
+        else
+            msrs->spec_ctrl.raw &= ~SPEC_CTRL_SSBD;
+        break;
+
     case MSR_AMD64_DE_CFG:
         /*
          * OpenBSD 6.7 will panic if writing to DE_CFG triggers a #GP:
xen/arch/x86/spec_ctrl.c +2 -1
index 1408e4c7ab..f338bfe292 100644
@@ -402,12 +402,13 @@ static void __init print_details(enum ind_thunk thunk, uint64_t caps)
      * mitigation support for guests.
      */
 #ifdef CONFIG_HVM
-    printk("  Support for HVM VMs:%s%s%s%s%s\n",
+    printk("  Support for HVM VMs:%s%s%s%s%s%s\n",
            (boot_cpu_has(X86_FEATURE_SC_MSR_HVM) ||
             boot_cpu_has(X86_FEATURE_SC_RSB_HVM) ||
             boot_cpu_has(X86_FEATURE_MD_CLEAR)   ||
             opt_eager_fpu)                           ? ""               : " None",
            boot_cpu_has(X86_FEATURE_SC_MSR_HVM)      ? " MSR_SPEC_CTRL" : "",
+           boot_cpu_has(X86_FEATURE_SC_MSR_HVM)      ? " MSR_VIRT_SPEC_CTRL" : "",
            boot_cpu_has(X86_FEATURE_SC_RSB_HVM)      ? " RSB"           : "",
            opt_eager_fpu                             ? " EAGER_FPU"     : "",
            boot_cpu_has(X86_FEATURE_MD_CLEAR)        ? " MD_CLEAR"      : "");
xen/include/public/arch-x86/cpufeatureset.h +1 -1
index 9cee4b439e..b797c6bea1 100644
@@ -265,7 +265,7 @@ XEN_CPUFEATURE(IBRS_SAME_MODE, 8*32+19) /*S  IBRS provides same-mode protection
 XEN_CPUFEATURE(NO_LMSL,       8*32+20) /*S  EFER.LMSLE no longer supported. */
 XEN_CPUFEATURE(AMD_PPIN,      8*32+23) /*   Protected Processor Inventory Number */
 XEN_CPUFEATURE(AMD_SSBD,      8*32+24) /*S  MSR_SPEC_CTRL.SSBD available */
-XEN_CPUFEATURE(VIRT_SSBD,     8*32+25) /*   MSR_VIRT_SPEC_CTRL.SSBD */
+XEN_CPUFEATURE(VIRT_SSBD,     8*32+25) /*!s MSR_VIRT_SPEC_CTRL.SSBD */
 XEN_CPUFEATURE(SSB_NO,        8*32+26) /*A  Hardware not vulnerable to SSB */
 XEN_CPUFEATURE(PSFD,          8*32+28) /*S  MSR_SPEC_CTRL.PSFD */
 
Signature · 1 line
2.35.1
Reviewed-by on v3 1/3
#2 lore raw

Re: [PATCH v3 1/3] amd/msr: implement VIRT_SPEC_CTRL for HVM guests on top of SPEC_CTRL

Quoted 6 lines · Roger Pau Monne

Use the logic to set shadow SPEC_CTRL values in order to implement support for VIRT_SPEC_CTRL (signaled by VIRT_SSBD CPUID flag) for HVM guests. This includes using the spec_ctrl vCPU MSR variable to store the guest set value of VIRT_SPEC_CTRL.SSBD, which will be OR'ed with any SPEC_CTRL values being set by the guest.

Quoted 7 lines

On hardware having SPEC_CTRL VIRT_SPEC_CTRL will not be offered by default to guests. VIRT_SPEC_CTRL will only be part of the max CPUID policy so it can be enabled for compatibility purposes.

Some reasoning regarding why '!s' is used to annotate the feature:
* '!': the feature might be exposed to guests even when not present on the host hardware.

As mentioned before I don't think ! is to be used for this particular purpose. But this needs input from Andrew. Hence ...

Quoted 4 lines

* 's': the feature won't be exposed by default.

... only tentatively:

Jan

#3 lore raw

PATCH v3 2/3 amd/msr: allow passthrough of VIRT_SPEC_CTRL for HVM guests

In reply to Roger Pau Monne

Allow HVM guests untrapped access to MSR_VIRT_SPEC_CTRL if the hardware has support for it. This requires adding logic in the vm{entry,exit} paths for SVM in order to context switch between the hypervisor value and the guest one. The added handlers for context switch will also be used for the legacy SSBD support.

Introduce a new synthetic feature leaf (X86_FEATURE_VIRT_SC_MSR_HVM) to signal whether VIRT_SPEC_CTRL needs to be handled on guest vm{entry,exit}.

This patch changes the annotation 's' to 'S' because it introduces support to expose VIRT_SSBD to guests by default when the host (virtual) hardware also supports it.

---
Changes since v2:
- Reword part of the commit message regarding annotation change.
- Fix MSR intercept.
- Add handling of VIRT_SPEC_CTRL to guest_{rd,wr}msr when using VIRT_SSBD also.

Changes since v1:
- Introduce virt_spec_ctrl vCPU field.
- Context switch VIRT_SPEC_CTRL on vmentry/vmexit separately from SPEC_CTRL.

---
 xen/arch/x86/cpuid.c                        | 11 ++++++
 xen/arch/x86/hvm/svm/entry.S                |  6 ++++
 xen/arch/x86/hvm/svm/svm.c                  | 40 +++++++++++++++++++++
 xen/arch/x86/include/asm/cpufeatures.h      |  1 +
 xen/arch/x86/include/asm/msr.h              | 10 ++++++
 xen/arch/x86/msr.c                          | 16 ++++++---
 xen/arch/x86/spec_ctrl.c                    |  9 ++++-
 xen/include/public/arch-x86/cpufeatureset.h |  2 +-
 8 files changed, 89 insertions(+), 6 deletions(-)
xen/arch/x86/cpuid.c +11 -0
index 4ca77ea870..91e53284fc 100644
@@ -534,6 +534,10 @@ static void __init calculate_hvm_max_policy(void)
          raw_cpuid_policy.basic.sep )
         __set_bit(X86_FEATURE_SEP, hvm_featureset);
 
+    if ( !boot_cpu_has(X86_FEATURE_VIRT_SC_MSR_HVM) )
+        /* Clear VIRT_SSBD if VIRT_SPEC_CTRL is not exposed to guests. */
+        __clear_bit(X86_FEATURE_VIRT_SSBD, hvm_featureset);
+
     /*
      * If Xen isn't virtualising MSR_SPEC_CTRL for HVM guests (functional
      * availability, or admin choice), hide the feature.
@@ -590,6 +594,13 @@ static void __init calculate_hvm_def_policy(void)
     guest_common_feature_adjustments(hvm_featureset);
     guest_common_default_feature_adjustments(hvm_featureset);
 
+    /*
+     * AMD_SSBD if preferred over VIRT_SSBD, so don't expose the later by
+     * default if the former is available.
+     */
+    if ( boot_cpu_has(X86_FEATURE_AMD_SSBD) )
+        __clear_bit(X86_FEATURE_VIRT_SSBD, hvm_featureset);
+
     sanitise_featureset(hvm_featureset);
     cpuid_featureset_to_policy(hvm_featureset, p);
     recalculate_xstate(p);
xen/arch/x86/hvm/svm/entry.S +6 -0
index 4ae55a2ef6..e2c104bb1e 100644
@@ -57,6 +57,9 @@ __UNLIKELY_END(nsvm_hap)
 
         clgi
 
+        ALTERNATIVE "", STR(call vmentry_virt_spec_ctrl), \
+                        X86_FEATURE_VIRT_SC_MSR_HVM
+
         /* WARNING! `ret`, `call *`, `jmp *` not safe beyond this point. */
         /* SPEC_CTRL_EXIT_TO_SVM       Req: b=curr %rsp=regs/cpuinfo, Clob: acd */
         .macro svm_vmentry_spec_ctrl
@@ -114,6 +117,9 @@ __UNLIKELY_END(nsvm_hap)
         ALTERNATIVE "", svm_vmexit_spec_ctrl, X86_FEATURE_SC_MSR_HVM
         /* WARNING! `ret`, `call *`, `jmp *` not safe before this point. */
 
+        ALTERNATIVE "", STR(call vmexit_virt_spec_ctrl), \
+                        X86_FEATURE_VIRT_SC_MSR_HVM
+
         stgi
 GLOBAL(svm_stgi_label)
         mov  %rsp,%rdi
xen/arch/x86/hvm/svm/svm.c +40 -0
index 64a45045da..40ff28ecf1 100644
@@ -52,6 +52,7 @@
 #include <asm/hvm/svm/svmdebug.h>
 #include <asm/hvm/svm/nestedsvm.h>
 #include <asm/hvm/nestedhvm.h>
+#include <asm/spec_ctrl.h>
 #include <asm/x86_emulate.h>
 #include <public/sched.h>
 #include <asm/hvm/vpt.h>
@@ -610,6 +611,15 @@ static void cf_check svm_cpuid_policy_changed(struct vcpu *v)
     svm_intercept_msr(v, MSR_SPEC_CTRL,
                       cp->extd.ibrs ? MSR_INTERCEPT_NONE : MSR_INTERCEPT_RW);
 
+    /*
+     * Give access to MSR_VIRT_SPEC_CTRL if the guest has been told about it
+     * and the hardware implements it.
+     */
+    svm_intercept_msr(v, MSR_VIRT_SPEC_CTRL,
+                      cp->extd.virt_ssbd && cpu_has_virt_ssbd &&
+                      !cpu_has_amd_ssbd ?
+                      MSR_INTERCEPT_NONE : MSR_INTERCEPT_RW);
+
     /* Give access to MSR_PRED_CMD if the guest has been told about it. */
     svm_intercept_msr(v, MSR_PRED_CMD,
                       cp->extd.ibpb ? MSR_INTERCEPT_NONE : MSR_INTERCEPT_RW);
@@ -3105,6 +3115,36 @@ void svm_vmexit_handler(struct cpu_user_regs *regs)
     vmcb_set_vintr(vmcb, intr);
 }
 
+/* Called with GIF=0. */
+void vmexit_virt_spec_ctrl(void)
+{
+    unsigned int val = opt_ssbd ? SPEC_CTRL_SSBD : 0;
+
+    if ( cpu_has_virt_ssbd )
+    {
+        unsigned int lo, hi;
+
+        /*
+         * Need to read from the hardware because VIRT_SPEC_CTRL is not context
+         * switched by the hardware, and we allow the guest untrapped access to
+         * the register.
+         */
+        rdmsr(MSR_VIRT_SPEC_CTRL, lo, hi);
+        if ( val != lo )
+            wrmsr(MSR_VIRT_SPEC_CTRL, val, 0);
+        current->arch.msrs->virt_spec_ctrl.raw = lo;
+    }
+}
+
+/* Called with GIF=0. */
+void vmentry_virt_spec_ctrl(void)
+{
+    unsigned int val = current->arch.msrs->virt_spec_ctrl.raw;
+
+    if ( val != (opt_ssbd ? SPEC_CTRL_SSBD : 0) )
+        wrmsr(MSR_VIRT_SPEC_CTRL, val, 0);
+}
+
 /*
  * Local variables:
  * mode: C
xen/arch/x86/include/asm/cpufeatures.h +1 -0
index 7413febd7a..2240547b64 100644
@@ -40,6 +40,7 @@ XEN_CPUFEATURE(SC_VERW_HVM,       X86_SYNTH(24)) /* VERW used by Xen for HVM */
 XEN_CPUFEATURE(SC_VERW_IDLE,      X86_SYNTH(25)) /* VERW used by Xen for idle */
 XEN_CPUFEATURE(XEN_SHSTK,         X86_SYNTH(26)) /* Xen uses CET Shadow Stacks */
 XEN_CPUFEATURE(XEN_IBT,           X86_SYNTH(27)) /* Xen uses CET Indirect Branch Tracking */
+XEN_CPUFEATURE(VIRT_SC_MSR_HVM,   X86_SYNTH(28)) /* MSR_VIRT_SPEC_CTRL exposed to HVM */
 
 /* Bug words follow the synthetic words. */
 #define X86_NR_BUG 1
xen/arch/x86/include/asm/msr.h +10 -0
index ab6fbb5051..460aabe84f 100644
@@ -375,6 +375,16 @@ struct vcpu_msrs
      */
     uint32_t tsc_aux;
 
+    /*
+     * 0xc001011f - MSR_VIRT_SPEC_CTRL (if !X86_FEATURE_AMD_SSBD)
+     *
+     * AMD only. Guest selected value, saved and restored on guest VM
+     * entry/exit.
+     */
+    struct {
+        uint32_t raw;
+    } virt_spec_ctrl;
+
     /*
      * 0xc00110{27,19-1b} MSR_AMD64_DR{0-3}_ADDRESS_MASK
      *
xen/arch/x86/msr.c +12 -4
index 72c175fd8b..a1e268eea9 100644
@@ -385,7 +385,10 @@ int guest_rdmsr(struct vcpu *v, uint32_t msr, uint64_t *val)
         if ( !cp->extd.virt_ssbd )
             goto gp_fault;
 
-        *val = msrs->spec_ctrl.raw & SPEC_CTRL_SSBD;
+        if ( cpu_has_amd_ssbd )
+            *val = msrs->spec_ctrl.raw & SPEC_CTRL_SSBD;
+        else
+            *val = msrs->virt_spec_ctrl.raw;
         break;
 
     case MSR_AMD64_DE_CFG:
@@ -678,10 +681,15 @@ int guest_wrmsr(struct vcpu *v, uint32_t msr, uint64_t val)
             goto gp_fault;
 
         /* Only supports SSBD bit, the rest are ignored. */
-        if ( val & SPEC_CTRL_SSBD )
-            msrs->spec_ctrl.raw |= SPEC_CTRL_SSBD;
+        if ( cpu_has_amd_ssbd )
+        {
+            if ( val & SPEC_CTRL_SSBD )
+                msrs->spec_ctrl.raw |= SPEC_CTRL_SSBD;
+            else
+                msrs->spec_ctrl.raw &= ~SPEC_CTRL_SSBD;
+        }
         else
-            msrs->spec_ctrl.raw &= ~SPEC_CTRL_SSBD;
+            msrs->virt_spec_ctrl.raw = val & SPEC_CTRL_SSBD;
         break;
 
     case MSR_AMD64_DE_CFG:
xen/arch/x86/spec_ctrl.c +8 -1
index f338bfe292..0d5ec877d1 100644
@@ -406,9 +406,12 @@ static void __init print_details(enum ind_thunk thunk, uint64_t caps)
            (boot_cpu_has(X86_FEATURE_SC_MSR_HVM) ||
             boot_cpu_has(X86_FEATURE_SC_RSB_HVM) ||
             boot_cpu_has(X86_FEATURE_MD_CLEAR)   ||
+            boot_cpu_has(X86_FEATURE_VIRT_SC_MSR_HVM) ||
             opt_eager_fpu)                           ? ""               : " None",
            boot_cpu_has(X86_FEATURE_SC_MSR_HVM)      ? " MSR_SPEC_CTRL" : "",
-           boot_cpu_has(X86_FEATURE_SC_MSR_HVM)      ? " MSR_VIRT_SPEC_CTRL" : "",
+           (boot_cpu_has(X86_FEATURE_SC_MSR_HVM) ||
+            boot_cpu_has(X86_FEATURE_VIRT_SC_MSR_HVM)) ? " MSR_VIRT_SPEC_CTRL"
+                                                       : "",
            boot_cpu_has(X86_FEATURE_SC_RSB_HVM)      ? " RSB"           : "",
            opt_eager_fpu                             ? " EAGER_FPU"     : "",
            boot_cpu_has(X86_FEATURE_MD_CLEAR)        ? " MD_CLEAR"      : "");
@@ -1069,6 +1072,10 @@ void __init init_speculation_mitigations(void)
             setup_force_cpu_cap(X86_FEATURE_SC_MSR_HVM);
     }
 
+    /* Support VIRT_SPEC_CTRL.SSBD if AMD_SSBD is not available. */
+    if ( opt_msr_sc_hvm && !cpu_has_amd_ssbd && cpu_has_virt_ssbd )
+        setup_force_cpu_cap(X86_FEATURE_VIRT_SC_MSR_HVM);
+
     /* If we have IBRS available, see whether we should use it. */
     if ( has_spec_ctrl && ibrs )
         default_xen_spec_ctrl |= SPEC_CTRL_IBRS;
xen/include/public/arch-x86/cpufeatureset.h +1 -1
index b797c6bea1..0639b9faf2 100644
@@ -265,7 +265,7 @@ XEN_CPUFEATURE(IBRS_SAME_MODE, 8*32+19) /*S  IBRS provides same-mode protection
 XEN_CPUFEATURE(NO_LMSL,       8*32+20) /*S  EFER.LMSLE no longer supported. */
 XEN_CPUFEATURE(AMD_PPIN,      8*32+23) /*   Protected Processor Inventory Number */
 XEN_CPUFEATURE(AMD_SSBD,      8*32+24) /*S  MSR_SPEC_CTRL.SSBD available */
-XEN_CPUFEATURE(VIRT_SSBD,     8*32+25) /*!s MSR_VIRT_SPEC_CTRL.SSBD */
+XEN_CPUFEATURE(VIRT_SSBD,     8*32+25) /*!S MSR_VIRT_SPEC_CTRL.SSBD */
 XEN_CPUFEATURE(SSB_NO,        8*32+26) /*A  Hardware not vulnerable to SSB */
 XEN_CPUFEATURE(PSFD,          8*32+28) /*S  MSR_SPEC_CTRL.PSFD */
 
Signature · 1 line
2.35.1
Reviewed-by on v3 2/3
#4 lore raw

Re: [PATCH v3 2/3] amd/msr: allow passthrough of VIRT_SPEC_CTRL for HVM guests

Quoted 10 lines · Roger Pau Monne

Allow HVM guests untrapped access to MSR_VIRT_SPEC_CTRL if the hardware has support for it. This requires adding logic in the vm{entry,exit} paths for SVM in order to context switch between the hypervisor value and the guest one. The added handlers for context switch will also be used for the legacy SSBD support.

Introduce a new synthetic feature leaf (X86_FEATURE_VIRT_SC_MSR_HVM) to signal whether VIRT_SPEC_CTRL needs to be handled on guest vm{entry,exit}.

Quoted 6 lines

This patch changes the annotation 's' to 'S' because it introduces support to expose VIRT_SSBD to guests by default when the host (virtual) hardware also supports it.

#5 lore raw

PATCH v3 3/3 amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

In reply to Roger Pau Monne

Expose VIRT_SSBD to guests if the hardware supports setting SSBD in the LS_CFG MSR (a.k.a. non-architectural way). Different AMD CPU families use different bits in LS_CFG, so exposing VIRT_SPEC_CTRL.SSBD allows for an unified way of exposing SSBD support to guests on AMD hardware that's compatible migration wise, regardless of what underlying mechanism is used to set SSBD.

Note that on AMD Family 17h (Zen 1) the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time.

Do the context switching of the SSBD selection in LS_CFG between hypervisor and guest in the same handler that's already used to switch the value of VIRT_SPEC_CTRL in the hardware when supported.

---
Changes since v2:
- Fix codding style issues.
- Use AMD_ZEN1_MAX_SOCKETS to define the max number of possible sockets in Zen1 systems.

Changes since v1:
- Report legacy SSBD support using a global variable.
- Use ro_after_init for ssbd_max_cores.
- Handle boot_cpu_data.x86_num_siblings < 1.
- Add comment regarding _irqsave usage in amd_set_legacy_ssbd.

---
 xen/arch/x86/cpu/amd.c         | 116 ++++++++++++++++++++++++++++-----
 xen/arch/x86/cpuid.c           |  10 +++
 xen/arch/x86/hvm/svm/svm.c     |  12 +++-
 xen/arch/x86/include/asm/amd.h |   4 ++
 xen/arch/x86/spec_ctrl.c       |   4 +-
 5 files changed, 127 insertions(+), 19 deletions(-)
xen/arch/x86/cpu/amd.c +99 -17
index 4999f8be2b..a256a9d882 100644
@@ -48,6 +48,7 @@ boolean_param("allow_unsafe", opt_allow_unsafe);
 
 /* Signal whether the ACPI C1E quirk is required. */
 bool __read_mostly amd_acpi_c1e_quirk;
+bool __ro_after_init amd_legacy_ssbd;
 
 static inline int rdmsr_amd_safe(unsigned int msr, unsigned int *lo,
 				 unsigned int *hi)
@@ -685,23 +686,10 @@ void amd_init_lfence(struct cpuinfo_x86 *c)
  * Refer to the AMD Speculative Store Bypass whitepaper:
  * https://developer.amd.com/wp-content/resources/124441_AMD64_SpeculativeStoreBypassDisable_Whitepaper_final.pdf
  */
-void amd_init_ssbd(const struct cpuinfo_x86 *c)
+static bool set_legacy_ssbd(const struct cpuinfo_x86 *c, bool enable)
 {
 	int bit = -1;
 
-	if (cpu_has_ssb_no)
-		return;
-
-	if (cpu_has_amd_ssbd) {
-		/* Handled by common MSR_SPEC_CTRL logic */
-		return;
-	}
-
-	if (cpu_has_virt_ssbd) {
-		wrmsrl(MSR_VIRT_SPEC_CTRL, opt_ssbd ? SPEC_CTRL_SSBD : 0);
-		return;
-	}
-
 	switch (c->x86) {
 	case 0x15: bit = 54; break;
 	case 0x16: bit = 33; break;
@@ -715,20 +703,114 @@ void amd_init_ssbd(const struct cpuinfo_x86 *c)
 		if (rdmsr_safe(MSR_AMD64_LS_CFG, val) ||
 		    ({
 			    val &= ~mask;
-			    if (opt_ssbd)
+			    if (enable)
 				    val |= mask;
 			    false;
 		    }) ||
 		    wrmsr_safe(MSR_AMD64_LS_CFG, val) ||
 		    ({
 			    rdmsrl(MSR_AMD64_LS_CFG, val);
-			    (val & mask) != (opt_ssbd * mask);
+			    (val & mask) != (enable * mask);
 		    }))
 			bit = -1;
 	}
 
-	if (bit < 0)
+	return bit >= 0;
+}
+
+void amd_init_ssbd(const struct cpuinfo_x86 *c)
+{
+	if (cpu_has_ssb_no)
+		return;
+
+	if (cpu_has_amd_ssbd) {
+		/* Handled by common MSR_SPEC_CTRL logic */
+		return;
+	}
+
+	if (cpu_has_virt_ssbd) {
+		wrmsrl(MSR_VIRT_SPEC_CTRL, opt_ssbd ? SPEC_CTRL_SSBD : 0);
+		return;
+	}
+
+	if (!set_legacy_ssbd(c, opt_ssbd)) {
 		printk_once(XENLOG_ERR "No SSBD controls available\n");
+		if (amd_legacy_ssbd)
+			panic("CPU feature mismatch: no legacy SSBD\n");
+	} else if (c == &boot_cpu_data)
+		amd_legacy_ssbd = true;
+}
+
+static struct ssbd_core {
+    spinlock_t lock;
+    unsigned int count;
+} *ssbd_core;
+static unsigned int __ro_after_init ssbd_max_cores;
+#define AMD_ZEN1_MAX_SOCKETS 2
+
+bool __init amd_setup_legacy_ssbd(void)
+{
+	unsigned int i;
+
+	if (boot_cpu_data.x86 != 0x17 || boot_cpu_data.x86_num_siblings <= 1)
+		return true;
+
+	/*
+	 * One could be forgiven for thinking that c->x86_max_cores is the
+	 * correct value to use here.
+	 *
+	 * However, that value is derived from the current configuration, and
+	 * c->cpu_core_id is sparse on all but the top end CPUs.  Derive
+	 * max_cpus from ApicIdCoreIdSize which will cover any sparseness.
+	 */
+	if (boot_cpu_data.extended_cpuid_level >= 0x80000008) {
+		ssbd_max_cores = 1u << MASK_EXTR(cpuid_ecx(0x80000008), 0xf000);
+		ssbd_max_cores /= boot_cpu_data.x86_num_siblings;
+	}
+	if (!ssbd_max_cores)
+		return false;
+
+	/* Max is two sockets for Fam17h hardware. */
+	ssbd_core = xzalloc_array(struct ssbd_core,
+	                          ssbd_max_cores * AMD_ZEN1_MAX_SOCKETS);
+	if (!ssbd_core)
+		return false;
+
+	for (i = 0; i < ssbd_max_cores * AMD_ZEN1_MAX_SOCKETS; i++) {
+		spin_lock_init(&ssbd_core[i].lock);
+		/* Record initial state, also applies to any hotplug CPU. */
+		if (opt_ssbd)
+			ssbd_core[i].count = boot_cpu_data.x86_num_siblings;
+	}
+
+	return true;
+}
+
+void amd_set_legacy_ssbd(bool enable)
+{
+	const struct cpuinfo_x86 *c = &current_cpu_data;
+	struct ssbd_core *core;
+	unsigned long flags;
+
+	if (c->x86 != 0x17 || c->x86_num_siblings <= 1) {
+		BUG_ON(!set_legacy_ssbd(c, enable));
+		return;
+	}
+
+	BUG_ON(c->phys_proc_id >= AMD_ZEN1_MAX_SOCKETS);
+	BUG_ON(c->cpu_core_id >= ssbd_max_cores);
+	core = &ssbd_core[c->phys_proc_id * ssbd_max_cores + c->cpu_core_id];
+	/*
+	 * Use irqsave variant to make check_lock() happy. When called from
+	 * vm{exit,entry}_virt_spec_ctrl GIF=0, but the state of IF could be 1,
+	 * thus fooling the spinlock check.
+	 */
+	spin_lock_irqsave(&core->lock, flags);
+	core->count += enable ? 1 : -1;
+	ASSERT(core->count <= c->x86_num_siblings);
+	if (enable ? core->count == 1 : !core->count)
+		BUG_ON(!set_legacy_ssbd(c, enable));
+	spin_unlock_irqrestore(&core->lock, flags);
 }
 
 void __init detect_zen2_null_seg_behaviour(void)
xen/arch/x86/cpuid.c +10 -0
index 91e53284fc..e278fee689 100644
@@ -537,6 +537,16 @@ static void __init calculate_hvm_max_policy(void)
     if ( !boot_cpu_has(X86_FEATURE_VIRT_SC_MSR_HVM) )
         /* Clear VIRT_SSBD if VIRT_SPEC_CTRL is not exposed to guests. */
         __clear_bit(X86_FEATURE_VIRT_SSBD, hvm_featureset);
+    else
+        /*
+         * Expose VIRT_SSBD if VIRT_SPEC_CTRL is supported, as that implies the
+         * underlying hardware is capable of setting SSBD using
+         * non-architectural way or VIRT_SSBD is available.
+         *
+         * Note that if the hardware supports VIRT_SSBD natively this setting
+         * will just override an already set bit.
+         */
+        __set_bit(X86_FEATURE_VIRT_SSBD, hvm_featureset);
 
     /*
      * If Xen isn't virtualising MSR_SPEC_CTRL for HVM guests (functional
xen/arch/x86/hvm/svm/svm.c +11 -1
index 40ff28ecf1..9b8f8d21bd 100644
@@ -3133,7 +3133,12 @@ void vmexit_virt_spec_ctrl(void)
         if ( val != lo )
             wrmsr(MSR_VIRT_SPEC_CTRL, val, 0);
         current->arch.msrs->virt_spec_ctrl.raw = lo;
+
+        return;
     }
+
+    if ( val != current->arch.msrs->virt_spec_ctrl.raw )
+        amd_set_legacy_ssbd(val & SPEC_CTRL_SSBD);
 }
 
 /* Called with GIF=0. */
@@ -3142,7 +3147,12 @@ void vmentry_virt_spec_ctrl(void)
     unsigned int val = current->arch.msrs->virt_spec_ctrl.raw;
 
     if ( val != (opt_ssbd ? SPEC_CTRL_SSBD : 0) )
-        wrmsr(MSR_VIRT_SPEC_CTRL, val, 0);
+    {
+        if ( cpu_has_virt_ssbd )
+            wrmsr(MSR_VIRT_SPEC_CTRL, val, 0);
+        else
+            amd_set_legacy_ssbd(val & SPEC_CTRL_SSBD);
+    }
 }
 
 /*
xen/arch/x86/include/asm/amd.h +4 -0
index a82382e6bf..6a42f68542 100644
@@ -151,4 +151,8 @@ void check_enable_amd_mmconf_dmi(void);
 extern bool amd_acpi_c1e_quirk;
 void amd_check_disable_c1e(unsigned int port, u8 value);
 
+extern bool amd_legacy_ssbd;
+bool amd_setup_legacy_ssbd(void);
+void amd_set_legacy_ssbd(bool enable);
+
 #endif /* __AMD_H__ */
xen/arch/x86/spec_ctrl.c +3 -1
index 0d5ec877d1..495e6f9405 100644
@@ -22,6 +22,7 @@
 #include <xen/param.h>
 #include <xen/warning.h>
 
+#include <asm/amd.h>
 #include <asm/hvm/svm/svm.h>
 #include <asm/microcode.h>
 #include <asm/msr.h>
@@ -1073,7 +1074,8 @@ void __init init_speculation_mitigations(void)
     }
 
     /* Support VIRT_SPEC_CTRL.SSBD if AMD_SSBD is not available. */
-    if ( opt_msr_sc_hvm && !cpu_has_amd_ssbd && cpu_has_virt_ssbd )
+    if ( opt_msr_sc_hvm && !cpu_has_amd_ssbd &&
+         (cpu_has_virt_ssbd || (amd_legacy_ssbd && amd_setup_legacy_ssbd())) )
         setup_force_cpu_cap(X86_FEATURE_VIRT_SC_MSR_HVM);
 
     /* If we have IBRS available, see whether we should use it. */
Signature · 1 line
2.35.1
#6 lore raw

Re: [PATCH v3 3/3] amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

Quoted 7 lines · Roger Pau Monne

Expose VIRT_SSBD to guests if the hardware supports setting SSBD in the LS_CFG MSR (a.k.a. non-architectural way). Different AMD CPU families use different bits in LS_CFG, so exposing VIRT_SPEC_CTRL.SSBD allows for an unified way of exposing SSBD support to guests on AMD hardware that's compatible migration wise, regardless of what underlying mechanism is used to set SSBD.

Quoted 6 lines

Note that on AMD Family 17h (Zen 1) the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time.

So where exactly is the boundary? If I'm not mistaken Zen2 is also Fam17 (and only Zen3 is Fam19), yet here and elsewhere you look to take Zen1 == Fam17.

Just one further nit on the code:

xen/arch/x86/cpu/amd.c void amd_init_ssbd(const struct cpuinfo_x86 *c) · line 749

Quoted 6 lines
xen/arch/x86/cpu/amd.c +6 -0
+static struct ssbd_core {
+    spinlock_t lock;
+    unsigned int count;
+} *ssbd_core;
+static unsigned int __ro_after_init ssbd_max_cores;
+#define AMD_ZEN1_MAX_SOCKETS 2

This #define looks to render ...

Quoted 15 lines
xen/arch/x86/cpu/amd.c +15 -0
+bool __init amd_setup_legacy_ssbd(void)
+{
+	unsigned int i;
+
+	if (boot_cpu_data.x86 != 0x17 || boot_cpu_data.x86_num_siblings <= 1)
+		return true;
+
+	/*
+	 * One could be forgiven for thinking that c->x86_max_cores is the
+	 * correct value to use here.
+	 *
+	 * However, that value is derived from the current configuration, and
+	 * c->cpu_core_id is sparse on all but the top end CPUs.  Derive
+	 * max_cpus from ApicIdCoreIdSize which will cover any sparseness.
+	 */

xen/arch/x86/cpu/amd.c void amd_init_ssbd(const struct cpuinfo_x86 *c) · line 775

Quoted 10 lines
xen/arch/x86/cpu/amd.c +10 -0
+	if (boot_cpu_data.extended_cpuid_level >= 0x80000008) {
+		ssbd_max_cores = 1u << MASK_EXTR(cpuid_ecx(0x80000008), 0xf000);
+		ssbd_max_cores /= boot_cpu_data.x86_num_siblings;
+	}
+	if (!ssbd_max_cores)
+		return false;
+
+	/* Max is two sockets for Fam17h hardware. */
+	ssbd_core = xzalloc_array(struct ssbd_core,
+	                          ssbd_max_cores * AMD_ZEN1_MAX_SOCKETS);

... largely redundant the comment here; if anywhere I think the comment would want to go next to the #define.

Jan

#7 lore raw

Re: [PATCH v3 3/3] amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

Quoted 15 lines · Jan Beulich
Quoted 7 lines · Roger Pau Monne

Expose VIRT_SSBD to guests if the hardware supports setting SSBD in the LS_CFG MSR (a.k.a. non-architectural way). Different AMD CPU families use different bits in LS_CFG, so exposing VIRT_SPEC_CTRL.SSBD allows for an unified way of exposing SSBD support to guests on AMD hardware that's compatible migration wise, regardless of what underlying mechanism is used to set SSBD.

Quoted 6 lines

Note that on AMD Family 17h (Zen 1) the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time.

Quoted 3 lines

So where exactly is the boundary? If I'm not mistaken Zen2 is also Fam17 (and only Zen3 is Fam19), yet here and elsewhere you look to take Zen1 == Fam17.

Right, but Zen2 already has AMD_SSBD (ie: SPEC_CTRL), so it's not using this logic.

The AMD whitepaper is more clear about this: any Fam17h processor that is using the non-architectural MSRs to set SSBD and has more than 1 logical processor for each logical core must synchronize the setting of SSBD.

I think just dropping the mention of Zen 1 in the commit message should remove your concerns?

Quoted 31 lines

Just one further nit on the code:

xen/arch/x86/cpu/amd.c void amd_init_ssbd(const struct cpuinfo_x86 *c) · line 749

Quoted 6 lines
xen/arch/x86/cpu/amd.c +6 -0
+static struct ssbd_core {
+    spinlock_t lock;
+    unsigned int count;
+} *ssbd_core;
+static unsigned int __ro_after_init ssbd_max_cores;
+#define AMD_ZEN1_MAX_SOCKETS 2

This #define looks to render ...

Quoted 20 lines
xen/arch/x86/cpu/amd.c +20 -0
+bool __init amd_setup_legacy_ssbd(void)
+{
+	unsigned int i;
+
+	if (boot_cpu_data.x86 != 0x17 || boot_cpu_data.x86_num_siblings <= 1)
+		return true;
+
+	/*
+	 * One could be forgiven for thinking that c->x86_max_cores is the
+	 * correct value to use here.
+	 *
+	 * However, that value is derived from the current configuration, and
+	 * c->cpu_core_id is sparse on all but the top end CPUs.  Derive
+	 * max_cpus from ApicIdCoreIdSize which will cover any sparseness.
+	 */
+	if (boot_cpu_data.extended_cpuid_level >= 0x80000008) {
+		ssbd_max_cores = 1u << MASK_EXTR(cpuid_ecx(0x80000008), 0xf000);
+		ssbd_max_cores /= boot_cpu_data.x86_num_siblings;
+	}
+	if (!ssbd_max_cores)
Quoted 8 lines

xen/arch/x86/cpu/amd.c void amd_init_ssbd(const struct cpuinfo_x86 *c) · line 775

Quoted 5 lines
xen/arch/x86/cpu/amd.c +5 -0
+		return false;
+
+	/* Max is two sockets for Fam17h hardware. */
+	ssbd_core = xzalloc_array(struct ssbd_core,
+	                          ssbd_max_cores * AMD_ZEN1_MAX_SOCKETS);

... largely redundant the comment here; if anywhere I think the comment would want to go next to the #define.

I guess I should also rename the define to FAM17H instead of ZEN1, as all Fam17h is limited to two sockets, then the comment can be removed as I think the define is self-explanatory.

Thanks, Roger.

#8 lore raw

Re: [PATCH v3 3/3] amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

Quoted 23 lines · Roger Pau Monné
Quoted 15 lines · Jan Beulich
Quoted 7 lines · Roger Pau Monne

Expose VIRT_SSBD to guests if the hardware supports setting SSBD in the LS_CFG MSR (a.k.a. non-architectural way). Different AMD CPU families use different bits in LS_CFG, so exposing VIRT_SPEC_CTRL.SSBD allows for an unified way of exposing SSBD support to guests on AMD hardware that's compatible migration wise, regardless of what underlying mechanism is used to set SSBD.

Quoted 6 lines

Note that on AMD Family 17h (Zen 1) the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time.

Quoted 3 lines

So where exactly is the boundary? If I'm not mistaken Zen2 is also Fam17 (and only Zen3 is Fam19), yet here and elsewhere you look to take Zen1 == Fam17.

Right, but Zen2 already has AMD_SSBD (ie: SPEC_CTRL), so it's not using this logic.

Quoted 7 lines

The AMD whitepaper is more clear about this: any Fam17h processor that is using the non-architectural MSRs to set SSBD and has more than 1 logical processor for each logical core must synchronize the setting of SSBD.

I think just dropping the mention of Zen 1 in the commit message should remove your concerns?

Or keep it, but qualify it by saying that Zen2 isn't expected to take this path because of having SSBD. But iirc SSBD was introduced to Zen2 only by a ucode update, so such a description should not be wrong wrt such not-up-to-date systems.

Quoted 40 lines
Quoted 31 lines

Just one further nit on the code:

xen/arch/x86/cpu/amd.c void amd_init_ssbd(const struct cpuinfo_x86 *c) · line 749

Quoted 6 lines
xen/arch/x86/cpu/amd.c +6 -0
+static struct ssbd_core {
+    spinlock_t lock;
+    unsigned int count;
+} *ssbd_core;
+static unsigned int __ro_after_init ssbd_max_cores;
+#define AMD_ZEN1_MAX_SOCKETS 2

This #define looks to render ...

Quoted 20 lines
xen/arch/x86/cpu/amd.c +20 -0
+bool __init amd_setup_legacy_ssbd(void)
+{
+	unsigned int i;
+
+	if (boot_cpu_data.x86 != 0x17 || boot_cpu_data.x86_num_siblings <= 1)
+		return true;
+
+	/*
+	 * One could be forgiven for thinking that c->x86_max_cores is the
+	 * correct value to use here.
+	 *
+	 * However, that value is derived from the current configuration, and
+	 * c->cpu_core_id is sparse on all but the top end CPUs.  Derive
+	 * max_cpus from ApicIdCoreIdSize which will cover any sparseness.
+	 */
+	if (boot_cpu_data.extended_cpuid_level >= 0x80000008) {
+		ssbd_max_cores = 1u << MASK_EXTR(cpuid_ecx(0x80000008), 0xf000);
+		ssbd_max_cores /= boot_cpu_data.x86_num_siblings;
+	}
+	if (!ssbd_max_cores)
Quoted 8 lines

xen/arch/x86/cpu/amd.c void amd_init_ssbd(const struct cpuinfo_x86 *c) · line 775

Quoted 5 lines
xen/arch/x86/cpu/amd.c +5 -0
+		return false;
+
+	/* Max is two sockets for Fam17h hardware. */
+	ssbd_core = xzalloc_array(struct ssbd_core,
+	                          ssbd_max_cores * AMD_ZEN1_MAX_SOCKETS);

... largely redundant the comment here; if anywhere I think the comment would want to go next to the #define.

Quoted 3 lines

I guess I should also rename the define to FAM17H instead of ZEN1, as all Fam17h is limited to two sockets, then the comment can be removed as I think the define is self-explanatory.

Yes, this rename would help both eliminate my confusion as well as the need for the comment.

Jan

#9 lore raw

Re: [PATCH v3 3/3] amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

Quoted 29 lines · Jan Beulich
Quoted 28 lines · Roger Pau Monné
Quoted 15 lines · Jan Beulich
Quoted 7 lines · Roger Pau Monne

Expose VIRT_SSBD to guests if the hardware supports setting SSBD in the LS_CFG MSR (a.k.a. non-architectural way). Different AMD CPU families use different bits in LS_CFG, so exposing VIRT_SPEC_CTRL.SSBD allows for an unified way of exposing SSBD support to guests on AMD hardware that's compatible migration wise, regardless of what underlying mechanism is used to set SSBD.

Quoted 6 lines

Note that on AMD Family 17h (Zen 1) the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time.

Quoted 3 lines

So where exactly is the boundary? If I'm not mistaken Zen2 is also Fam17 (and only Zen3 is Fam19), yet here and elsewhere you look to take Zen1 == Fam17.

Right, but Zen2 already has AMD_SSBD (ie: SPEC_CTRL), so it's not using this logic.

The AMD whitepaper is more clear about this: any Fam17h processor that is using the non-architectural MSRs to set SSBD and has more than 1 logical processor for each logical core must synchronize the setting of SSBD.

Quoted 7 lines
Quoted 2 lines

I think just dropping the mention of Zen 1 in the commit message should remove your concerns?

Or keep it, but qualify it by saying that Zen2 isn't expected to take this path because of having SSBD. But iirc SSBD was introduced to Zen2 only by a ucode update, so such a description should not be wrong wrt such not-up-to-date systems.

FTAOD I've worded this as:

"Note that on AMD Family 17h and Hygon Family 18h processors the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time."

Which I think is correct in all cases. Iff Zen2 was to resort to using the non-architectural way of setting SSBD (if that's even possible) it should synchronize it between threads according to my read of the AMD whitepaper.

I've also added handling for Hygon Fam18h, seeing as those also make use of the non-architectural way of setting SSBD.

Thanks, Roger.

#10 lore raw

Re: [PATCH v3 3/3] amd/msr: implement VIRT_SPEC_CTRL for HVM guests using legacy SSBD

Quoted 38 lines · Roger Pau Monné
Quoted 29 lines · Jan Beulich
Quoted 28 lines · Roger Pau Monné
Quoted 15 lines · Jan Beulich
Quoted 7 lines · Roger Pau Monne

Expose VIRT_SSBD to guests if the hardware supports setting SSBD in the LS_CFG MSR (a.k.a. non-architectural way). Different AMD CPU families use different bits in LS_CFG, so exposing VIRT_SPEC_CTRL.SSBD allows for an unified way of exposing SSBD support to guests on AMD hardware that's compatible migration wise, regardless of what underlying mechanism is used to set SSBD.

Quoted 6 lines

Note that on AMD Family 17h (Zen 1) the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time.

Quoted 3 lines

So where exactly is the boundary? If I'm not mistaken Zen2 is also Fam17 (and only Zen3 is Fam19), yet here and elsewhere you look to take Zen1 == Fam17.

Right, but Zen2 already has AMD_SSBD (ie: SPEC_CTRL), so it's not using this logic.

The AMD whitepaper is more clear about this: any Fam17h processor that is using the non-architectural MSRs to set SSBD and has more than 1 logical processor for each logical core must synchronize the setting of SSBD.

Quoted 7 lines
Quoted 2 lines

I think just dropping the mention of Zen 1 in the commit message should remove your concerns?

Or keep it, but qualify it by saying that Zen2 isn't expected to take this path because of having SSBD. But iirc SSBD was introduced to Zen2 only by a ucode update, so such a description should not be wrong wrt such not-up-to-date systems.

Quoted 8 lines

FTAOD I've worded this as:

"Note that on AMD Family 17h and Hygon Family 18h processors the value of SSBD in LS_CFG is shared between threads on the same core, so there's extra logic in order to synchronize the value and have SSBD set as long as one of the threads in the core requires it to be set. Such logic also requires extra storage for each thread state, which is allocated at initialization time."

Thanks.

Quoted 7 lines

Which I think is correct in all cases. Iff Zen2 was to resort to using the non-architectural way of setting SSBD (if that's even possible) it should synchronize it between threads according to my read of the AMD whitepaper.

I've also added handling for Hygon Fam18h, seeing as those also make use of the non-architectural way of setting SSBD.

Right, better be on the safe side.

Jan

#11 lore raw
In reply to Roger Pau Monne
Quoted 29 lines · Roger Pau Monne

Hello,

The following series implements support for MSR_VIRT_SPEC_CTRL (VIRT_SSBD) on different AMD CPU families.

Note that the support is added backwards, starting with the newer CPUs that support MSR_SPEC_CTRL and moving to the older ones either using MSR_VIRT_SPEC_CTRL or the SSBD bit in LS_CFG.

Xen is still free to use it's own SSBD setting, as the selection is context switched on vm{entry,exit}.

On Zen2 and later, SPEC_CTRL.SSBD exists and should be used in preference to VIRT_SPEC_CTRL.SSBD.  However, for migration compatibility, Xen offers VIRT_SSBD to guests (in the max CPUID policy, not default) implemented in terms of SPEC_CTRL.SSBD.

On Fam15h thru Zen1, Xen exposes VIRT_SSBD to guests by default to abstract away the model and/or hypervisor specific differences in MSR_LS_CFG/MSR_VIRT_SPEC_CTRL.

Note that if the hardware itself does offer VIRT_SSBD (ie: very likely when running virtualized on < Zen2 hardware) and not AMD_SSBD Xen will allow untrapped access to MSR_VIRT_SPEC_CTRL for HVM guests.

So the implementation of VIRT_SSBD exposed to HVM guests will use one of the following underlying mechanisms, in the preference order listed below:

Quoted 5 lines

* SPEC_CTRL.SSBD. (patch 1)
* VIRT_SPEC_CTRL.SSBD (untrapped). (patch 2).
* Non-architectural way using LS_CFG. (patch 3)

This has survived a XenRT basic set of tests on AMD machines.

Sorry for the mixed feedback, but some is applicable across multiple patches.

First, it is important to know why MSR_VIRT_SPEC_CTRL exists, because that informs what is, and is not, sensible to do with it.

It exists to be a FMS-invariant abstraction of the DE_CFG interface, emulated by the hypervisor.  At the time, we experimented with emulating MSR_SPEC_CTRL directly, but the results were unusable slow (legacy IBRS causing a vmexit on every syscall/interrupt entry&exit) so MSR_VIRT_SPEC_CTRL is also an explicit statement that it is an expensive operation and shouldn't be used frequently.

In practice, this means "only for very very important processes, and not to be used more frequently than process context switch".  Also, there is no hardware which implements MSR_VIRT_SPEC_CTRL, nor will there be.

Patch 2 has added an extra two vmexits around each vmexit, in an effort to let L2 vmexit to L0 rather than L1 for what is likely to be 0 times in an L1 timeslice.  It's not a credible optimisation, for something which isn't a production usecase.  Yes - nested virt does exist, and is useful for dev, but noone runs a fully fat server virt hypervisor at L1 in production if they care in the slightest about performance.  Either way, patch 2 is premature optimisation with a massive complexity cost.

Furthermore, writes to LS_CFG are also incredibly expensive, even if you're not changing any bits.  The AMD recommended algorithm specifically avoids rewriting it with the same value as before.

Another thing is that Xen shouldn't touch LS_CFG like this if there is any hint of a hypervisor on the system.  If there is a hypervisor and it doesn't offer VIRT_SPEC_CTRL, trying to play with LS_CFG isn't going to make the situation any better.

As to the CPUID bit handling, on consideration of the whole series, it wants to be "!" only.  ! is there to indicate "something complicated is going on with this bit", and life is too short to try and get the derivation logic right with both implicit and explicit conditions. Leave it without an s/S (so no auto propagation from the host policy), and set it in the max policy for LS_CFG || VIRT_SPEC_CTRL || SPEC_CTRL, and set it in the default policy for LS_CFG || VIRT_SPEC_CTRL, which will be far clearer to follow.

For struct ssbd_core, the name isn't great.  It's more ssbd_ls_cfg/state.  Also, each array element wants 64 byte alignment, because that's the only way to avoid atomic cacheline pingpong from the spinlocks.  Also, the accessors need to be raw, because GIF=0 context is weird and working around checklock with irqsave variants is not a clever move.  It is not safe to printk()/bug/etc from GIF=0 context, so logic needs to be kept to an absolute bare minimum.

~Andrew

#12 lore raw
Quoted 49 lines · Andrew Cooper
Quoted 29 lines · Roger Pau Monne

Hello,

The following series implements support for MSR_VIRT_SPEC_CTRL (VIRT_SSBD) on different AMD CPU families.

Note that the support is added backwards, starting with the newer CPUs that support MSR_SPEC_CTRL and moving to the older ones either using MSR_VIRT_SPEC_CTRL or the SSBD bit in LS_CFG.

Xen is still free to use it's own SSBD setting, as the selection is context switched on vm{entry,exit}.

On Zen2 and later, SPEC_CTRL.SSBD exists and should be used in preference to VIRT_SPEC_CTRL.SSBD.  However, for migration compatibility, Xen offers VIRT_SSBD to guests (in the max CPUID policy, not default) implemented in terms of SPEC_CTRL.SSBD.

On Fam15h thru Zen1, Xen exposes VIRT_SSBD to guests by default to abstract away the model and/or hypervisor specific differences in MSR_LS_CFG/MSR_VIRT_SPEC_CTRL.

Note that if the hardware itself does offer VIRT_SSBD (ie: very likely when running virtualized on < Zen2 hardware) and not AMD_SSBD Xen will allow untrapped access to MSR_VIRT_SPEC_CTRL for HVM guests.

So the implementation of VIRT_SSBD exposed to HVM guests will use one of the following underlying mechanisms, in the preference order listed below:

Quoted 5 lines

* SPEC_CTRL.SSBD. (patch 1)
* VIRT_SPEC_CTRL.SSBD (untrapped). (patch 2).
* Non-architectural way using LS_CFG. (patch 3)

This has survived a XenRT basic set of tests on AMD machines.

Sorry for the mixed feedback, but some is applicable across multiple patches.

First, it is important to know why MSR_VIRT_SPEC_CTRL exists, because that informs what is, and is not, sensible to do with it.

It exists to be a FMS-invariant abstraction of the DE_CFG interface, emulated by the hypervisor.  At the time, we experimented with emulating MSR_SPEC_CTRL directly, but the results were unusable slow (legacy IBRS causing a vmexit on every syscall/interrupt entry&exit) so MSR_VIRT_SPEC_CTRL is also an explicit statement that it is an expensive operation and shouldn't be used frequently.

Quoted 5 lines

In practice, this means "only for very very important processes, and not to be used more frequently than process context switch".  Also, there is no hardware which implements MSR_VIRT_SPEC_CTRL, nor will there be.

Patch 2 has added an extra two vmexits around each vmexit, in an effort

No, it adds just one extra unconditional vmexit, the wrmsr is conditional to the value in the guest and Xen differing, and it's not unconditionally executed.

Quoted 6 lines

to let L2 vmexit to L0 rather than L1 for what is likely to be 0 times in an L1 timeslice.  It's not a credible optimisation, for something which isn't a production usecase.  Yes - nested virt does exist, and is useful for dev, but noone runs a fully fat server virt hypervisor at L1 in production if they care in the slightest about performance.  Either way, patch 2 is premature optimisation with a massive complexity cost.

OK, so your recommendation is to trap writes to VIRT_SPEC_CTRL so the unconditional rdmsr on vmexit is avoided at the cost of taking a vmexit on all writes to VIRT_SPEC_CTRL, that's indeed fine.

Quoted 4 lines

Furthermore, writes to LS_CFG are also incredibly expensive, even if you're not changing any bits.  The AMD recommended algorithm specifically avoids rewriting it with the same value as before.

Writes to LS_CFG will only happen if the guest and Xen selection of SSBD differ, as the guest value is cached.

Quoted 4 lines

Another thing is that Xen shouldn't touch LS_CFG like this if there is any hint of a hypervisor on the system.  If there is a hypervisor and it doesn't offer VIRT_SPEC_CTRL, trying to play with LS_CFG isn't going to make the situation any better.

I would assume that no hypervisor will offer LS_CFG support for setting SSBD if the guest shouldn't use it, and hence set_legacy_ssbd will already return false.

Quoted 8 lines

As to the CPUID bit handling, on consideration of the whole series, it wants to be "!" only.  ! is there to indicate "something complicated is going on with this bit", and life is too short to try and get the derivation logic right with both implicit and explicit conditions. Leave it without an s/S (so no auto propagation from the host policy), and set it in the max policy for LS_CFG || VIRT_SPEC_CTRL || SPEC_CTRL, and set it in the default policy for LS_CFG || VIRT_SPEC_CTRL, which will be far clearer to follow.

OK, as both Jan and you agree that using just '!' is fine I can work with that.

Quoted 7 lines

For struct ssbd_core, the name isn't great.  It's more ssbd_ls_cfg/state.  Also, each array element wants 64 byte alignment, because that's the only way to avoid atomic cacheline pingpong from the spinlocks.  Also, the accessors need to be raw, because GIF=0 context is weird and working around checklock with irqsave variants is not a clever move.  It is not safe to printk()/bug/etc from GIF=0 context, so logic needs to be kept to an absolute bare minimum.

Sorry I'm a bit dense today, but I don't seem to be able to find any raw accessors for our spinlock implementation. _spin_lock_cb will unconditionally call check_lock and there doesn't seem to be a way to bypass the checks if lock debugging is enabled. Am I missing something?

Thanks, Roger.