Skip to contentSkip to conversation

x86: Protected Processor Inventory Number (PPIN) support

PATCH In review Jan Beulich 2 participants 14 messages Nov 8, 2019 – Dec 16, 2019

lore
Share & tools
lore
#0 lore raw

v2 cover x86: Protected Processor Inventory Number (PPIN) support

1: include the PPIN in MCE records when available 2: explicitly disallow guest access to PPIN 3: provide Dom0 access to PPIN via XENPF_resource_op

I have yet to get around to post the Linux side consumer patch of the interface addition in patch 1.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#1 lore raw

PATCH v2 1/3 x86: include the PPIN in MCE records when available

Quoting the respective Linux commit:

Intel Xeons from Ivy Bridge onwards support a processor identification number set in the factory. To the user this is a handy unique number to identify a particular CPU. Intel can decode this to the fab/production run to track errors. On systems that have it, include it in the machine check record. I'm told that this would be helpful for users that run large data centers with multi-socket servers to keep track of which CPUs are seeing errors.

Newer AMD CPUs support this too, at different MSR numbers.

Take the opportunity and hide __MC_NMSRS from the public interface going forward.

[Linux commit 3f5a7896a5096fd50030a04d4c3f28a7441e30a5]

---
v2: Also update xen-cpuid and libxl_cpuid.c.

tools/libxl/libxl_cpuid.c +1 -0
@@ -261,6 +261,7 @@ int libxl_cpuid_parse_config(libxl_cpuid
         {"rstr-fp-err-ptrs", 0x80000008, NA, CPUID_REG_EBX, 2, 1},
         {"wbnoinvd",     0x80000008, NA, CPUID_REG_EBX,  9,  1},
         {"ibpb",         0x80000008, NA, CPUID_REG_EBX, 12,  1},
+        {"ppin",         0x80000008, NA, CPUID_REG_EBX, 23,  1},
 
         {"nc",           0x80000008, NA, CPUID_REG_ECX,  0,  8},
         {"apicidsize",   0x80000008, NA, CPUID_REG_ECX, 12,  4},
tools/misc/xen-cpuid.c +2 -0
@@ -150,6 +150,8 @@ static const char *const str_e8b[32] =
     /* [ 8] */            [ 9] = "wbnoinvd",
 
     [12] = "ibpb",
+
+    /* [22] */                 [23] = "ppin",
 };
 
 static const char *const str_7d0[32] =
xen/arch/x86/cpu/mcheck/mce.c +9 -1
@@ -35,6 +35,7 @@ bool __read_mostly mce_broadcast;
 bool is_mc_panic;
 DEFINE_PER_CPU_READ_MOSTLY(unsigned int, nr_mce_banks);
 unsigned int __read_mostly firstbank;
+unsigned int __read_mostly ppin_msr;
 uint8_t __read_mostly cmci_apic_vector;
 
 DEFINE_PER_CPU_READ_MOSTLY(struct mca_banks *, poll_bankmask);
@@ -999,10 +1000,17 @@ static void do_mc_get_cpu_info(void *v)
     /*
      * This part needs to run on the CPU itself.
      */
-    xcp->mc_nmsrvals = __MC_NMSRS;
+    xcp->mc_nmsrvals = 1;
     xcp->mc_msrvalues[0].reg = MSR_IA32_MCG_CAP;
     rdmsrl(MSR_IA32_MCG_CAP, xcp->mc_msrvalues[0].value);
 
+    if ( ppin_msr && xcp->mc_nmsrvals < ARRAY_SIZE(xcp->mc_msrvalues) )
+    {
+        xcp->mc_msrvalues[xcp->mc_nmsrvals].reg = ppin_msr;
+        rdmsrl(ppin_msr, xcp->mc_msrvalues[xcp->mc_nmsrvals].value);
+        ++xcp->mc_nmsrvals;
+    }
+
     if ( c->cpuid_level >= 1 )
     {
         cpuid(1, &junk, &ebx, &junk, &junk);
xen/arch/x86/cpu/mcheck/mce.h +1 -0
@@ -49,6 +49,7 @@ enum mcheck_type intel_mcheck_init(struc
 void amd_nonfatal_mcheck_init(struct cpuinfo_x86 *c);
 
 extern unsigned int firstbank;
+extern unsigned int ppin_msr;
 
 struct mcinfo_extended *intel_get_extended_msrs(
     struct mcinfo_global *mig, struct mc_info *mi);
xen/arch/x86/cpu/mcheck/mce_amd.c +20 -0
@@ -315,6 +315,26 @@ amd_mcheck_init(struct cpuinfo_x86 *ci)
     if ( quirkflag == MCEQUIRK_F10_GART )
         mcequirk_amd_apply(quirkflag);
 
+    if ( cpu_has(ci, X86_FEATURE_AMD_PPIN) &&
+         (ci == &boot_cpu_data || ppin_msr) )
+    {
+        uint64_t val;
+
+        rdmsrl(MSR_AMD_PPIN_CTL, val);
+
+        /* If PPIN is disabled, but not locked, try to enable. */
+        if ( !(val & (PPIN_ENABLE | PPIN_LOCKOUT)) )
+        {
+            wrmsr_safe(MSR_PPIN_CTL, val | PPIN_ENABLE);
+            rdmsrl(MSR_AMD_PPIN_CTL, val);
+        }
+
+        if ( (val & (PPIN_ENABLE | PPIN_LOCKOUT)) != PPIN_ENABLE )
+            ppin_msr = 0;
+        else if ( ci == &boot_cpu_data )
+            ppin_msr = MSR_AMD_PPIN;
+    }
+
     x86_mce_callback_register(amd_f10_handler);
     mce_recoverable_register(mc_amd_recoverable_scan);
     mce_register_addrcheck(mc_amd_addrcheck);
xen/arch/x86/cpu/mcheck/mce_intel.c +39 -0
@@ -853,6 +853,43 @@ static void intel_init_mce(void)
     mce_uhandler_num = ARRAY_SIZE(intel_mce_uhandlers);
 }
 
+static void intel_init_ppin(const struct cpuinfo_x86 *c)
+{
+    /*
+     * Even if testing the presence of the MSR would be enough, we don't
+     * want to risk the situation where other models reuse this MSR for
+     * other purposes.
+     */
+    switch ( c->x86_model )
+    {
+        uint64_t val;
+
+    case 0x3e: /* IvyBridge X */
+    case 0x3f: /* Haswell X */
+    case 0x4f: /* Broadwell X */
+    case 0x55: /* Skylake X */
+    case 0x56: /* Broadwell Xeon D */
+    case 0x57: /* Knights Landing */
+    case 0x85: /* Knights Mill */
+
+        if ( (c != &boot_cpu_data && !ppin_msr) ||
+             rdmsr_safe(MSR_PPIN_CTL, val) )
+            return;
+
+        /* If PPIN is disabled, but not locked, try to enable. */
+        if ( !(val & (PPIN_ENABLE | PPIN_LOCKOUT)) )
+        {
+            wrmsr_safe(MSR_PPIN_CTL, val | PPIN_ENABLE);
+            rdmsr_safe(MSR_PPIN_CTL, val);
+        }
+
+        if ( (val & (PPIN_ENABLE | PPIN_LOCKOUT)) != PPIN_ENABLE )
+            ppin_msr = 0;
+        else if ( c == &boot_cpu_data )
+            ppin_msr = MSR_PPIN;
+    }
+}
+
 static void cpu_mcabank_free(unsigned int cpu)
 {
     struct mca_banks *cmci = per_cpu(no_cmci_banks, cpu);
@@ -941,6 +978,8 @@ enum mcheck_type intel_mcheck_init(struc
 
     intel_init_thermal(c);
 
+    intel_init_ppin(c);
+
     return mcheck_intel;
 }
 
xen/include/asm-x86/msr-index.h +11 -0
@@ -45,6 +45,13 @@
 #define MSR_PRED_CMD			0x00000049
 #define PRED_CMD_IBPB			(_AC(1, ULL) << 0)
 
+/* Intel Protected Processor Inventory Number */
+#define MSR_PPIN_CTL			0x0000004e
+#define MSR_PPIN			0x0000004f
+
+#define PPIN_LOCKOUT			(_AC(1, ULL) << 0)
+#define PPIN_ENABLE			(_AC(1, ULL) << 1)
+
 #define MSR_ARCH_CAPABILITIES		0x0000010a
 #define ARCH_CAPS_RDCL_NO		(_AC(1, ULL) << 0)
 #define ARCH_CAPS_IBRS_ALL		(_AC(1, ULL) << 1)
@@ -278,6 +285,10 @@
 #define MSR_AMD_OSVW_ID_LENGTH          0xc0010140
 #define MSR_AMD_OSVW_STATUS             0xc0010141
 
+/* AMD Protected Processor Inventory Number */
+#define MSR_AMD_PPIN_CTL                0xc00102f0
+#define MSR_AMD_PPIN                    0xc00102f1
+
 /* K6 MSRs */
 #define MSR_K6_EFER			0xc0000080
 #define MSR_K6_STAR			0xc0000081
xen/include/public/arch-x86/cpufeatureset.h +1 -0
@@ -247,6 +247,7 @@ XEN_CPUFEATURE(RDPRU,         8*32+ 4) /
 XEN_CPUFEATURE(RSTR_FP_ERR_PTRS, 8*32+ 2) /*A  (F)X{SAVE,RSTOR} always saves/restores FPU Error pointers */
 XEN_CPUFEATURE(WBNOINVD,      8*32+ 9) /*   WBNOINVD instruction */
 XEN_CPUFEATURE(IBPB,          8*32+12) /*A  IBPB support only (no IBRS, used by AMD) */
+XEN_CPUFEATURE(AMD_PPIN,      8*32+23) /*   Protected Processor Inventory Number */
 
 /* Intel-defined CPU features, CPUID level 0x00000007:0.edx, word 9 */
 XEN_CPUFEATURE(AVX512_4VNNIW, 9*32+ 2) /*A  AVX512 Neural Network Instructions */
xen/include/public/arch-x86/xen-mca.h +2 -0
@@ -246,7 +246,9 @@ typedef struct mc_info mc_info_t;
 DEFINE_XEN_GUEST_HANDLE(mc_info_t);
 
 #define __MC_MSR_ARRAYSIZE 8
+#if __XEN_INTERFACE_VERSION__ <= 0x00040d00
 #define __MC_NMSRS 1
+#endif
 #define MC_NCAPS	7	/* 7 CPU feature flag words */
 #define MC_CAPS_STD_EDX	0	/* cpuid level 0x00000001 (%edx) */
 #define MC_CAPS_AMD_EDX	1	/* cpuid level 0x80000001 (%edx) */
xen/include/public/xen-compat.h +1 -1
@@ -27,7 +27,7 @@
 #ifndef __XEN_PUBLIC_XEN_COMPAT_H__
 #define __XEN_PUBLIC_XEN_COMPAT_H__
 
-#define __XEN_LATEST_INTERFACE_VERSION__ 0x00040d00
+#define __XEN_LATEST_INTERFACE_VERSION__ 0x00040e00
 
 #if defined(__XEN__) || defined(__XEN_TOOLS__)
 /* Xen is built with matching headers and implements the latest interface. */

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

Acked-by on v2 1/3
#2 lore raw

Re: [Xen-devel] [PATCH v2 1/3] x86: include the PPIN in MCE records when available

Quoted 10 lines · Jan Beulich

Quoting the respective Linux commit:

Intel Xeons from Ivy Bridge onwards support a processor identification number set in the factory. To the user this is a handy unique number to identify a particular CPU. Intel can decode this to the fab/production run to track errors. On systems that have it, include it in the machine check record. I'm told that this would be helpful for users that run large data centers with multi-socket servers to keep track of which CPUs are seeing errors.

Quoted 7 lines

Newer AMD CPUs support this too, at different MSR numbers.

Take the opportunity and hide __MC_NMSRS from the public interface going forward.

[Linux commit 3f5a7896a5096fd50030a04d4c3f28a7441e30a5]

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#3 lore raw

PATCH v2 2/3 x86: explicitly disallow guest access to PPIN

In reply to Jan Beulich

To fulfill the "protected" in its name, don't let the real hardware values "shine through". Report a control register value expressing this.

---
v2: Use "cp" consistently. Re-base.

xen/arch/x86/msr.c +18 -0
@@ -135,6 +135,8 @@ int guest_rdmsr(struct vcpu *v, uint32_t
     case MSR_TSX_FORCE_ABORT:
     case MSR_AMD64_LWP_CFG:
     case MSR_AMD64_LWP_CBADDR:
+    case MSR_PPIN:
+    case MSR_AMD_PPIN:
         /* Not offered to guests. */
         goto gp_fault;
 
@@ -237,6 +239,18 @@ int guest_rdmsr(struct vcpu *v, uint32_t
                                    ARRAY_SIZE(msrs->dr_mask))];
         break;
 
+    case MSR_PPIN_CTL:
+        if ( cp->x86_vendor != X86_VENDOR_INTEL )
+            goto gp_fault;
+        *val = PPIN_LOCKOUT;
+        break;
+
+    case MSR_AMD_PPIN_CTL:
+        if ( !cp->extd.amd_ppin )
+            goto gp_fault;
+        *val = PPIN_LOCKOUT;
+        break;
+
         /*
          * TODO: Implement when we have better topology representation.
     case MSR_INTEL_CORE_THREAD_COUNT:
@@ -273,10 +287,14 @@ int guest_wrmsr(struct vcpu *v, uint32_t
     case MSR_INTEL_CORE_THREAD_COUNT:
     case MSR_INTEL_PLATFORM_INFO:
     case MSR_ARCH_CAPABILITIES:
+    case MSR_PPIN:
+    case MSR_AMD_PPIN:
         /* Read-only */
     case MSR_TSX_FORCE_ABORT:
     case MSR_AMD64_LWP_CFG:
     case MSR_AMD64_LWP_CBADDR:
+    case MSR_PPIN_CTL:
+    case MSR_AMD_PPIN_CTL:
         /* Not offered to guests. */
         goto gp_fault;
 

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#4 lore raw

Re: [Xen-devel] [PATCH v2 2/3] x86: explicitly disallow guest access to PPIN

Quoted 2 lines · Jan Beulich

To fulfill the "protected" in its name, don't let the real hardware values "shine through". Report a control register value expressing this.

Why not call it as it is?  They leak through due to bugs in MSR handling.

Quoted 21 lines

---
v2: Use "cp" consistently. Re-base.

xen/arch/x86/msr.c +4 -0
@@ -135,6 +135,8 @@ int guest_rdmsr(struct vcpu *v, uint32_t
     case MSR_TSX_FORCE_ABORT:
     case MSR_AMD64_LWP_CFG:
     case MSR_AMD64_LWP_CBADDR:
+    case MSR_PPIN:
+    case MSR_AMD_PPIN:
         /* Not offered to guests. */
         goto gp_fault;
 
@@ -237,6 +239,18 @@ int guest_rdmsr(struct vcpu *v, uint32_t
                                    ARRAY_SIZE(msrs->dr_mask))];
         break;
 
+    case MSR_PPIN_CTL:
+        if ( cp->x86_vendor != X86_VENDOR_INTEL )

xen/arch/x86/msr.c int guest_rdmsr(struct vcpu *v, uint32_t · line 253

Quoted 10 lines
xen/arch/x86/msr.c +10 -0
+            goto gp_fault;
+        *val = PPIN_LOCKOUT;
+        break;
+
+    case MSR_AMD_PPIN_CTL:
+        if ( !cp->extd.amd_ppin )
+            goto gp_fault;
+        *val = PPIN_LOCKOUT;
+        break;
+

The "not offered to guests" blocks should always be symmetric.  What you've done here is half-virtualise something we have no intention to ever virtualise for guests.

Both of these should be blanket #GP faults.  AMD should never be in the position of seeing amd_ppin clear but PPIN_CTL returning LOCKOUT, and while Intel does have model specific behaviour, whatever else might be behind that MSR obviously shouldn't be leaking though either.

xen/arch/x86/msr.c int guest_wrmsr(struct vcpu *v, uint32_t · line 291

Quoted 9 lines
xen/arch/x86/msr.c +2 -0
         /*
          * TODO: Implement when we have better topology representation.
     case MSR_INTEL_CORE_THREAD_COUNT:
@@ -273,10 +287,14 @@ int guest_wrmsr(struct vcpu *v, uint32_t
     case MSR_INTEL_CORE_THREAD_COUNT:
     case MSR_INTEL_PLATFORM_INFO:
     case MSR_ARCH_CAPABILITIES:
+    case MSR_PPIN:
+    case MSR_AMD_PPIN:

... these should be in the lower block, as "not offered to guests" is logically different from "we virtualise them, but they are read only".

~Andrew

xen/arch/x86/msr.c int guest_wrmsr(struct vcpu *v, uint32_t · line 300

Quoted 10 lines
xen/arch/x86/msr.c +2 -0
         /* Read-only */
     case MSR_TSX_FORCE_ABORT:
     case MSR_AMD64_LWP_CFG:
     case MSR_AMD64_LWP_CBADDR:
+    case MSR_PPIN_CTL:
+    case MSR_AMD_PPIN_CTL:
         /* Not offered to guests. */
         goto gp_fault;
 

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#5 lore raw

Re: [Xen-devel] [PATCH v2 2/3] x86: explicitly disallow guest access to PPIN

Quoted 25 lines · Andrew Cooper
Quoted 2 lines · Jan Beulich

To fulfill the "protected" in its name, don't let the real hardware values "shine through". Report a control register value expressing this.

Why not call it as it is?  They leak through due to bugs in MSR handling.

Quoted 19 lines

---
v2: Use "cp" consistently. Re-base.

xen/arch/x86/msr.c +2 -0
@@ -135,6 +135,8 @@ int guest_rdmsr(struct vcpu *v, uint32_t
     case MSR_TSX_FORCE_ABORT:
     case MSR_AMD64_LWP_CFG:
     case MSR_AMD64_LWP_CBADDR:
+    case MSR_PPIN:
+    case MSR_AMD_PPIN:
         /* Not offered to guests. */
         goto gp_fault;
 
@@ -237,6 +239,18 @@ int guest_rdmsr(struct vcpu *v, uint32_t
                                    ARRAY_SIZE(msrs->dr_mask))];
         break;
 
Quoted 14 lines

xen/arch/x86/msr.c int guest_rdmsr(struct vcpu *v, uint32_t · line 253

Quoted 12 lines
xen/arch/x86/msr.c +12 -0
+    case MSR_PPIN_CTL:
+        if ( cp->x86_vendor != X86_VENDOR_INTEL )
+            goto gp_fault;
+        *val = PPIN_LOCKOUT;
+        break;
+
+    case MSR_AMD_PPIN_CTL:
+        if ( !cp->extd.amd_ppin )
+            goto gp_fault;
+        *val = PPIN_LOCKOUT;
+        break;
+

The "not offered to guests" blocks should always be symmetric.

Well, the wrmsr side of things simply was the wrong way round (see below) - with it flipped they then are symmetric.

Quoted 8 lines

  What
you've done here is half-virtualise something we have no intention to ever virtualise for guests.

Both of these should be blanket #GP faults.  AMD should never be in the position of seeing amd_ppin clear but PPIN_CTL returning LOCKOUT, and while Intel does have model specific behaviour, whatever else might be behind that MSR obviously shouldn't be leaking though either.

In the interest of getting this ack-ed I might switch to the blanket-#GP as you suggest, but I'm having trouble seeing why giving back sane (and safe) values is wrong. This isn't meant to indicate we might virtualize more of this. But why incur an unnecessary #GP(0) in the guest when we can indicate the same in a more "friendly" manner?

Quoted 9 lines

xen/arch/x86/msr.c int guest_wrmsr(struct vcpu *v, uint32_t · line 291

Quoted 6 lines
xen/arch/x86/msr.c +2 -0
@@ -273,10 +287,14 @@ int guest_wrmsr(struct vcpu *v, uint32_t
     case MSR_INTEL_CORE_THREAD_COUNT:
     case MSR_INTEL_PLATFORM_INFO:
     case MSR_ARCH_CAPABILITIES:
+    case MSR_PPIN:
+    case MSR_AMD_PPIN:

... these should be in the lower block, as "not offered to guests" is logically different from "we virtualise them, but they are read only".

Hmm, yes, I got these and ...

Quoted 8 lines

xen/arch/x86/msr.c int guest_wrmsr(struct vcpu *v, uint32_t · line 299

Quoted 8 lines
xen/arch/x86/msr.c +2 -0
         /* Read-only */
     case MSR_TSX_FORCE_ABORT:
     case MSR_AMD64_LWP_CFG:
     case MSR_AMD64_LWP_CBADDR:
+    case MSR_PPIN_CTL:
+    case MSR_AMD_PPIN_CTL:
         /* Not offered to guests. */
         goto gp_fault;

... these the wrong way round.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#6 lore raw

Re: [Xen-devel] [PATCH v2 2/3] x86: explicitly disallow guest access to PPIN

Quoted 8 lines · Jan Beulich
Quoted 8 lines

  What
you've done here is half-virtualise something we have no intention to ever virtualise for guests.

Both of these should be blanket #GP faults.  AMD should never be in the position of seeing amd_ppin clear but PPIN_CTL returning LOCKOUT, and while Intel does have model specific behaviour, whatever else might be behind that MSR obviously shouldn't be leaking though either.

Quoted 6 lines

In the interest of getting this ack-ed I might switch to the blanket-#GP as you suggest, but I'm having trouble seeing why giving back sane (and safe) values is wrong. This isn't meant to indicate we might virtualize more of this. But why incur an unnecessary #GP(0) in the guest when we can indicate the same in a more "friendly" manner?

Why add dead code to Xen?

It is unnecessary complexity in some already-complicated functions which are going to become far more complicated by the time we get Xen's MSR behaviour into a somewhat-reasonable state.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#7 lore raw

Re: [Xen-devel] [PATCH v2 2/3] x86: explicitly disallow guest access to PPIN

Quoted 9 lines · Andrew Cooper
Quoted 8 lines · Jan Beulich
Quoted 8 lines

  What
you've done here is half-virtualise something we have no intention to ever virtualise for guests.

Both of these should be blanket #GP faults.  AMD should never be in the position of seeing amd_ppin clear but PPIN_CTL returning LOCKOUT, and while Intel does have model specific behaviour, whatever else might be behind that MSR obviously shouldn't be leaking though either.

Quoted 8 lines
Quoted 6 lines

In the interest of getting this ack-ed I might switch to the blanket-#GP as you suggest, but I'm having trouble seeing why giving back sane (and safe) values is wrong. This isn't meant to indicate we might virtualize more of this. But why incur an unnecessary #GP(0) in the guest when we can indicate the same in a more "friendly" manner?

Why add dead code to Xen?

Well, as you said yourself - at least the Intel part of this isn't really dead. Of course we _expect_ guest kernels to cope with #GP here, but there are many expectations of ours which get violated ... (To give a concrete example in this very area of code: A customer of ours noticed the x2APIC MSRs having become inaccessible at some point. Clearly we didn't expect PV guests to try to access them.)

Quoted 3 lines

It is unnecessary complexity in some already-complicated functions which are going to become far more complicated by the time we get Xen's MSR behaviour into a somewhat-reasonable state.

If this was really about "complexity", I'd fully agree. But we talk about two instance of almost the same 5-line piece of code.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#8 lore raw

Re: [Xen-devel] [PATCH v2 2/3] x86: explicitly disallow guest access to PPIN

Quoted 16 lines · Jan Beulich
Quoted 15 lines · Andrew Cooper
Quoted 14 lines · Jan Beulich
Quoted 8 lines

  What
you've done here is half-virtualise something we have no intention to ever virtualise for guests.

Both of these should be blanket #GP faults.  AMD should never be in the position of seeing amd_ppin clear but PPIN_CTL returning LOCKOUT, and while Intel does have model specific behaviour, whatever else might be behind that MSR obviously shouldn't be leaking though either.

In the interest of getting this ack-ed I might switch to the blanket-#GP as you suggest, but I'm having trouble seeing why giving back sane (and safe) values is wrong. This isn't meant to indicate we might virtualize more of this. But why incur an unnecessary #GP(0) in the guest when we can indicate the same in a more "friendly" manner?

Quoted 8 lines
Quoted 1 line

Why add dead code to Xen?

Well, as you said yourself - at least the Intel part of this isn't really dead. Of course we _expect_ guest kernels to cope with #GP here, but there are many expectations of ours which get violated ... (To give a concrete example in this very area of code: A customer of ours noticed the x2APIC MSRs having become inaccessible at some point. Clearly we didn't expect PV guests to try to access them.)

This is a concrete example proves my point.

We would never have been in this mess to begin with if Xen didn't have broken MSR handling.  It was an information leak (at best) to offer PV guests even read-only access to the LAPIC.

The rest is incorrect pv-ops in Linux, caused by the fact that things which shouldn't leak through, leaked through.  This breakage is still present - a PV guest has no business accessing MSR_APIC_BASE, let alone depending on its contents.

Quoted 7 lines
Quoted 4 lines

It is unnecessary complexity in some already-complicated functions which are going to become far more complicated by the time we get Xen's MSR behaviour into a somewhat-reasonable state.

If this was really about "complexity", I'd fully agree. But we talk about two instance of almost the same 5-line piece of code.

We are talking about complexity.  I said "of the function", not of the individual case blocks.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#9 lore raw

PATCH v2 3/3 x86: provide Dom0 access to PPIN via XENPF_resource_op

In reply to Jan Beulich

It was requested that we provide a way independent of the MCE reporting interface that Dom0 software could use to get hold of the values for particular CPUs.

---
v2: New.

TBD: I couldn't identify any suitable utility under tools/misc/ that I
would have felt like making the frontend of this extension.

xen/arch/x86/platform_hypercall.c +4 -0
@@ -30,6 +30,7 @@
 #include <asm/mtrr.h>
 #include <asm/io_apic.h>
 #include <asm/setup.h>
+#include "cpu/mcheck/mce.h"
 #include "cpu/mtrr/mtrr.h"
 #include <xsm/xsm.h>
 
@@ -94,6 +95,9 @@ void check_resource_access(struct resour
         switch ( entry->u.cmd )
         {
         case XEN_RESOURCE_OP_MSR_READ:
+            if ( ppin_msr && entry->idx == ppin_msr )
+                break;
+            /* fall through */
         case XEN_RESOURCE_OP_MSR_WRITE:
             if ( entry->idx >> 32 )
                 ret = -EINVAL;

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

Acked-by on v2 3/3
#10 lore raw

Re: [Xen-devel] [PATCH v2 3/3] x86: provide Dom0 access to PPIN via XENPF_resource_op

Quoted 5 lines · Jan Beulich

It was requested that we provide a way independent of the MCE reporting interface that Dom0 software could use to get hold of the values for particular CPUs.

This hypercall is a total gross bodge, but I suppose it is the least bad place to add more bodges.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#11 lore raw
In reply to Jan Beulich
Quoted 6 lines · Jan Beulich

1: include the PPIN in MCE records when available 2: explicitly disallow guest access to PPIN 3: provide Dom0 access to PPIN via XENPF_resource_op

I have yet to get around to post the Linux side consumer patch of the interface addition in patch 1.

What about printing the PPIN information when Xen writes MCE details to the console?

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#12 lore raw
Quoted 10 lines · Andrew Cooper
Quoted 6 lines · Jan Beulich

1: include the PPIN in MCE records when available 2: explicitly disallow guest access to PPIN 3: provide Dom0 access to PPIN via XENPF_resource_op

I have yet to get around to post the Linux side consumer patch of the interface addition in patch 1.

What about printing the PPIN information when Xen writes MCE details to the console?

Do you mean in x86_mcinfo_dump(), or mc_panic(), or yet somewhere else? In any event if we did so I'd want to arrange for each PPIN to get logged at most once, to reduce unnecessary redundancy. Of course there's a comment ahead of x86_mcinfo_dump() mentioning that the output is to be parseable by mcelog, so I'm not sure if we can reasonably change what we produce. Otoh I'm also not sure the comment hasn't become stale with its (presumed) porting from Linux anyway.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#13 lore raw
Quoted 6 lines · Jan Beulich
Quoted 5 lines · Andrew Cooper
Quoted 4 lines · Jan Beulich

1: include the PPIN in MCE records when available 2: explicitly disallow guest access to PPIN 3: provide Dom0 access to PPIN via XENPF_resource_op

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

I have yet to get around to post the Linux side consumer patch of the interface addition in patch 1.

What about printing the PPIN information when Xen writes MCE details to the console?

Do you mean in x86_mcinfo_dump(), or mc_panic(), or yet somewhere else?

Something which ends up on the console.

Quoted 2 lines

In any event if we did so I'd want to arrange for each PPIN to get logged at most once, to reduce unnecessary redundancy.

Getting it printed, even redundantly is more helpful than not having it printed.

Quoted 5 lines

Of course there's a comment ahead of x86_mcinfo_dump() mentioning that the output is to be parseable by mcelog, so I'm not sure if we can reasonably change what we produce. Otoh I'm also not sure the comment hasn't become stale with its (presumed) porting from Linux anyway.

That comment is long obsolete.  What Xen prints needs substantial rearranging for mcelog to parse it these days.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel