x86: use {rd,wr}{fs,gs}base when available

... as being intended to be faster than MSR reads/writes.

In the case of emulate_privileged_op() also use these in favor of the
cached (but possibly stale) addresses from arch.pv_vcpu. This allows
entirely removing the code that was the subject of XSA-67.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Keir Fraser <keir@xen.org>

7 files changed, 79 insertions, 29 deletions

diff --git a/xen/arch/x86/Rules.mk b/xen/arch/x86/Rules.mk
index c93d2affd3..576985e710 100644
--- a/xen/arch/x86/Rules.mk
+++ b/xen/arch/x86/Rules.mk
@@ -31,6 +31,7 @@ $(call cc-options-add,CFLAGS,CC,$(EMBEDDED_EXTRA_CFLAGS))
 $(call cc-option-add,CFLAGS,CC,-Wnested-externs)
 $(call as-insn-check,CFLAGS,CC,"vmcall",-DHAVE_GAS_VMX)
 $(call as-insn-check,CFLAGS,CC,"invept (%rax)$$(comma)%rax",-DHAVE_GAS_EPT)
+$(call as-insn-check,CFLAGS,CC,"rdfsbase %rax",-DHAVE_GAS_FSGSBASE)
 
 ifeq ($(supervisor_mode_kernel),y)
 CFLAGS += -DCONFIG_X86_SUPERVISOR_MODE_KERNEL=1
diff --git a/xen/arch/x86/acpi/suspend.c b/xen/arch/x86/acpi/suspend.c
index c690b45345..9c3cf38fe5 100644
--- a/xen/arch/x86/acpi/suspend.c
+++ b/xen/arch/x86/acpi/suspend.c
@@ -27,8 +27,8 @@ void save_rest_processor_state(void)
     asm volatile (
         "movw %%ds,(%0); movw %%es,2(%0); movw %%fs,4(%0); movw %%gs,6(%0)"
         : : "r" (saved_segs) : "memory" );
-    rdmsrl(MSR_FS_BASE, saved_fs_base);
-    rdmsrl(MSR_GS_BASE, saved_gs_base);
+    saved_fs_base = rdfsbase();
+    saved_gs_base = rdgsbase();
     rdmsrl(MSR_SHADOW_GS_BASE, saved_kernel_gs_base);
     rdmsrl(MSR_CSTAR, saved_cstar);
     rdmsrl(MSR_LSTAR, saved_lstar);
@@ -56,8 +56,8 @@ void restore_rest_processor_state(void)
           X86_EFLAGS_DF|X86_EFLAGS_IF|X86_EFLAGS_TF,
           0U);
 
-    wrmsrl(MSR_FS_BASE, saved_fs_base);
-    wrmsrl(MSR_GS_BASE, saved_gs_base);
+    wrfsbase(saved_fs_base);
+    wrgsbase(saved_gs_base);
     wrmsrl(MSR_SHADOW_GS_BASE, saved_kernel_gs_base);
 
     if ( boot_cpu_data.x86_vendor == X86_VENDOR_INTEL ||
diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index 52b7a37fec..b67fcb8f78 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -1092,7 +1092,7 @@ static void load_segments(struct vcpu *n)
     {
         /* This can only be non-zero if selector is NULL. */
         if ( n->arch.pv_vcpu.fs_base )
-            wrmsrl(MSR_FS_BASE, n->arch.pv_vcpu.fs_base);
+            wrfsbase(n->arch.pv_vcpu.fs_base);
 
         /* Most kernels have non-zero GS base, so don't bother testing. */
         /* (This is also a serialising instruction, avoiding AMD erratum #88.) */
@@ -1100,7 +1100,7 @@ static void load_segments(struct vcpu *n)
 
         /* This can only be non-zero if selector is NULL. */
         if ( n->arch.pv_vcpu.gs_base_user )
-            wrmsrl(MSR_GS_BASE, n->arch.pv_vcpu.gs_base_user);
+            wrgsbase(n->arch.pv_vcpu.gs_base_user);
 
         /* If in kernel mode then switch the GS bases around. */
         if ( (n->arch.flags & TF_kernel_mode) )
diff --git a/xen/arch/x86/setup.c b/xen/arch/x86/setup.c
index 637f1f97f8..30a4fd6b88 100644
--- a/xen/arch/x86/setup.c
+++ b/xen/arch/x86/setup.c
@@ -1264,6 +1264,9 @@ void __init __start_xen(unsigned long mbi_p)
     if ( cpu_has_smep )
         set_in_cr4(X86_CR4_SMEP);
 
+    if ( cpu_has_fsgsbase )
+        set_in_cr4(X86_CR4_FSGSBASE);
+
     local_irq_enable();
 
     pt_pci_init();
diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
index 35be017dc8..8dcb70a98f 100644
--- a/xen/arch/x86/traps.c
+++ b/xen/arch/x86/traps.c
@@ -1985,28 +1985,18 @@ static int emulate_privileged_op(struct cpu_user_regs *regs)
         }
         else
         {
-            if ( lm_ovr == lm_seg_none || data_sel < 4 )
+            switch ( lm_ovr )
             {
-                switch ( lm_ovr )
-                {
-                case lm_seg_none:
-                    data_base = 0UL;
-                    break;
-                case lm_seg_fs:
-                    data_base = v->arch.pv_vcpu.fs_base;
-                    break;
-                case lm_seg_gs:
-                    if ( guest_kernel_mode(v, regs) )
-                        data_base = v->arch.pv_vcpu.gs_base_kernel;
-                    else
-                        data_base = v->arch.pv_vcpu.gs_base_user;
-                    break;
-                }
+            default:
+                data_base = 0UL;
+                break;
+            case lm_seg_fs:
+                data_base = rdfsbase();
+                break;
+            case lm_seg_gs:
+                data_base = rdgsbase();
+                break;
             }
-            else if ( !read_descriptor(data_sel, v, regs,
-                                       &data_base, &data_limit, &ar, 0) ||
-                      !(ar & _SEGMENT_S) || !(ar & _SEGMENT_P) )
-                goto fail;
             data_limit = ~0UL;
             ar = _SEGMENT_WR|_SEGMENT_S|_SEGMENT_DPL|_SEGMENT_P;
         }
diff --git a/xen/include/asm-x86/domain.h b/xen/include/asm-x86/domain.h
index 909f4490e5..e42651ef7f 100644
--- a/xen/include/asm-x86/domain.h
+++ b/xen/include/asm-x86/domain.h
@@ -457,12 +457,12 @@ unsigned long pv_guest_cr4_fixup(const struct vcpu *, unsigned long guest_cr4);
     (((v)->arch.pv_vcpu.ctrlreg[4]                          \
       | (mmu_cr4_features                                   \
          & (X86_CR4_PGE | X86_CR4_PSE | X86_CR4_SMEP |      \
-            X86_CR4_OSXSAVE))                               \
+            X86_CR4_OSXSAVE | X86_CR4_FSGSBASE))            \
       | ((v)->domain->arch.vtsc ? X86_CR4_TSD : 0))         \
      & ~X86_CR4_DE)
 #define real_cr4_to_pv_guest_cr4(c)                         \
-    ((c) & ~(X86_CR4_PGE | X86_CR4_PSE | X86_CR4_TSD        \
-             | X86_CR4_OSXSAVE | X86_CR4_SMEP))
+    ((c) & ~(X86_CR4_PGE | X86_CR4_PSE | X86_CR4_TSD |      \
+             X86_CR4_OSXSAVE | X86_CR4_SMEP | X86_CR4_FSGSBASE))
 
 void domain_cpuid(struct domain *d,
                   unsigned int  input,
diff --git a/xen/include/asm-x86/msr.h b/xen/include/asm-x86/msr.h
index 06c46a8bd2..5277c2b6fb 100644
--- a/xen/include/asm-x86/msr.h
+++ b/xen/include/asm-x86/msr.h
@@ -9,6 +9,7 @@
 #include <xen/percpu.h>
 #include <xen/errno.h>
 #include <asm/asm_defns.h>
+#include <asm/cpufeature.h>
 
 #define rdmsr(msr,val1,val2) \
      __asm__ __volatile__("rdmsr" \
@@ -97,6 +98,61 @@ static inline int wrmsr_safe(unsigned int msr, uint64_t val)
 			  : "=a" (low), "=d" (high) \
 			  : "c" (counter))
 
+static inline unsigned long rdfsbase(void)
+{
+    unsigned long base;
+
+    if ( cpu_has_fsgsbase )
+#ifdef HAVE_GAS_FSGSBASE
+        asm volatile ( "rdfsbase %0" : "=r" (base) );
+#else
+        asm volatile ( ".byte 0xf3, 0x48, 0x0f, 0xae, 0xc0" : "=a" (base) );
+#endif
+    else
+        rdmsrl(MSR_FS_BASE, base);
+
+    return base;
+}
+
+static inline unsigned long rdgsbase(void)
+{
+    unsigned long base;
+
+    if ( cpu_has_fsgsbase )
+#ifdef HAVE_GAS_FSGSBASE
+        asm volatile ( "rdgsbase %0" : "=r" (base) );
+#else
+        asm volatile ( ".byte 0xf3, 0x48, 0x0f, 0xae, 0xc8" : "=a" (base) );
+#endif
+    else
+        rdmsrl(MSR_GS_BASE, base);
+
+    return base;
+}
+
+static inline void wrfsbase(unsigned long base)
+{
+    if ( cpu_has_fsgsbase )
+#ifdef HAVE_GAS_FSGSBASE
+        asm volatile ( "wrfsbase %0" :: "r" (base) );
+#else
+        asm volatile ( ".byte 0xf3, 0x48, 0x0f, 0xae, 0xd0" :: "a" (base) );
+#endif
+    else
+        wrmsrl(MSR_FS_BASE, base);
+}
+
+static inline void wrgsbase(unsigned long base)
+{
+    if ( cpu_has_fsgsbase )
+#ifdef HAVE_GAS_FSGSBASE
+        asm volatile ( "wrgsbase %0" :: "r" (base) );
+#else
+        asm volatile ( ".byte 0xf3, 0x48, 0x0f, 0xae, 0xd8" :: "a" (base) );
+#endif
+    else
+        wrmsrl(MSR_GS_BASE, base);
+}
 
 DECLARE_PER_CPU(u64, efer);
 u64 read_efer(void);