Merge tag 'x86-fsgsbase-2020-08-04' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fsgsbase from Thomas Gleixner:
 "Support for FSGSBASE. Almost 5 years after the first RFC to support
  it, this has been brought into a shape which is maintainable and
  actually works.

  This final version was done by Sasha Levin who took it up after Intel
  dropped the ball. Sasha discovered that the SGX (sic!) offerings out
  there ship rogue kernel modules enabling FSGSBASE behind the kernels
  back which opens an instantanious unpriviledged root hole.

  The FSGSBASE instructions provide a considerable speedup of the
  context switch path and enable user space to write GSBASE without
  kernel interaction. This enablement requires careful handling of the
  exception entries which go through the paranoid entry path as they
  can no longer rely on the assumption that user GSBASE is positive (as
  enforced via prctl() on non FSGSBASE enabled systemn).

  All other entries (syscalls, interrupts and exceptions) can still just
  utilize SWAPGS unconditionally when the entry comes from user space.
  Converting these entries to use FSGSBASE has no benefit as SWAPGS is
  only marginally slower than WRGSBASE and locating and retrieving the
  kernel GSBASE value is not a free operation either. The real benefit
  of RD/WRGSBASE is the avoidance of the MSR reads and writes.

  The changes come with appropriate selftests and have held up in field
  testing against the (sanitized) Graphene-SGX driver"

* tag 'x86-fsgsbase-2020-08-04' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (21 commits)
  x86/fsgsbase: Fix Xen PV support
  x86/ptrace: Fix 32-bit PTRACE_SETREGS vs fsbase and gsbase
  selftests/x86/fsgsbase: Add a missing memory constraint
  selftests/x86/fsgsbase: Fix a comment in the ptrace_write_gsbase test
  selftests/x86: Add a syscall_arg_fault_64 test for negative GSBASE
  selftests/x86/fsgsbase: Test ptracer-induced GS base write with FSGSBASE
  selftests/x86/fsgsbase: Test GS selector on ptracer-induced GS base write
  Documentation/x86/64: Add documentation for GS/FS addressing mode
  x86/elf: Enumerate kernel FSGSBASE capability in AT_HWCAP2
  x86/cpu: Enable FSGSBASE on 64bit by default and add a chicken bit
  x86/entry/64: Handle FSGSBASE enabled paranoid entry/exit
  x86/entry/64: Introduce the FIND_PERCPU_BASE macro
  x86/entry/64: Switch CR3 before SWAPGS in paranoid entry
  x86/speculation/swapgs: Check FSGSBASE in enabling SWAPGS mitigation
  x86/process/64: Use FSGSBASE instructions on thread copy and ptrace
  x86/process/64: Use FSBSBASE in switch_to() if available
  x86/process/64: Make save_fsgs_for_kvm() ready for FSGSBASE
  x86/fsgsbase/64: Enable FSGSBASE instructions in helper functions
  x86/fsgsbase/64: Add intrinsics for FSGSBASE instructions
  x86/cpu: Add 'unsafe_fsgsbase' to enable CR4.FSGSBASE
  ...

12 files changed, 391 insertions, 98 deletions

diff --git a/arch/x86/entry/calling.h b/arch/x86/entry/calling.h
index 4208c1e3f601..98e4d8886f11 100644
--- a/arch/x86/entry/calling.h
+++ b/arch/x86/entry/calling.h
@@ -6,6 +6,7 @@
 #include <asm/percpu.h>
 #include <asm/asm-offsets.h>
 #include <asm/processor-flags.h>
+#include <asm/inst.h>
 
 /*
 
@@ -341,6 +342,12 @@ For 32-bit we have the following conventions - kernel is built with
 #endif
 .endm
 
+.macro SAVE_AND_SET_GSBASE scratch_reg:req save_reg:req
+	rdgsbase \save_reg
+	GET_PERCPU_BASE \scratch_reg
+	wrgsbase \scratch_reg
+.endm
+
 #else /* CONFIG_X86_64 */
 # undef		UNWIND_HINT_IRET_REGS
 # define	UNWIND_HINT_IRET_REGS
@@ -351,3 +358,36 @@ For 32-bit we have the following conventions - kernel is built with
 	call stackleak_erase
 #endif
 .endm
+
+#ifdef CONFIG_SMP
+
+/*
+ * CPU/node NR is loaded from the limit (size) field of a special segment
+ * descriptor entry in GDT.
+ */
+.macro LOAD_CPU_AND_NODE_SEG_LIMIT reg:req
+	movq	$__CPUNODE_SEG, \reg
+	lsl	\reg, \reg
+.endm
+
+/*
+ * Fetch the per-CPU GSBASE value for this processor and put it in @reg.
+ * We normally use %gs for accessing per-CPU data, but we are setting up
+ * %gs here and obviously can not use %gs itself to access per-CPU data.
+ */
+.macro GET_PERCPU_BASE reg:req
+	ALTERNATIVE \
+		"LOAD_CPU_AND_NODE_SEG_LIMIT \reg", \
+		"RDPID	\reg", \
+		X86_FEATURE_RDPID
+	andq	$VDSO_CPUNODE_MASK, \reg
+	movq	__per_cpu_offset(, \reg, 8), \reg
+.endm
+
+#else
+
+.macro GET_PERCPU_BASE reg:req
+	movq	pcpu_unit_offsets(%rip), \reg
+.endm
+
+#endif /* CONFIG_SMP */
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index d2599c627d38..70dea9337816 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -38,6 +38,7 @@
 #include <asm/frame.h>
 #include <asm/trapnr.h>
 #include <asm/nospec-branch.h>
+#include <asm/fsgsbase.h>
 #include <linux/err.h>
 
 #include "calling.h"
@@ -426,10 +427,7 @@ SYM_CODE_START(\asmsym)
 	testb	$3, CS-ORIG_RAX(%rsp)
 	jnz	.Lfrom_usermode_switch_stack_\@
 
-	/*
-	 * paranoid_entry returns SWAPGS flag for paranoid_exit in EBX.
-	 * EBX == 0 -> SWAPGS, EBX == 1 -> no SWAPGS
-	 */
+	/* paranoid_entry returns GS information for paranoid_exit in EBX. */
 	call	paranoid_entry
 
 	UNWIND_HINT_REGS
@@ -458,10 +456,7 @@ SYM_CODE_START(\asmsym)
 	UNWIND_HINT_IRET_REGS offset=8
 	ASM_CLAC
 
-	/*
-	 * paranoid_entry returns SWAPGS flag for paranoid_exit in EBX.
-	 * EBX == 0 -> SWAPGS, EBX == 1 -> no SWAPGS
-	 */
+	/* paranoid_entry returns GS information for paranoid_exit in EBX. */
 	call	paranoid_entry
 	UNWIND_HINT_REGS
 
@@ -798,24 +793,21 @@ SYM_CODE_END(xen_failsafe_callback)
 #endif /* CONFIG_XEN_PV */
 
 /*
- * Save all registers in pt_regs, and switch gs if needed.
- * Use slow, but surefire "are we in kernel?" check.
- * Return: ebx=0: need swapgs on exit, ebx=1: otherwise
+ * Save all registers in pt_regs. Return GSBASE related information
+ * in EBX depending on the availability of the FSGSBASE instructions:
+ *
+ * FSGSBASE	R/EBX
+ *     N        0 -> SWAPGS on exit
+ *              1 -> no SWAPGS on exit
+ *
+ *     Y        GSBASE value at entry, must be restored in paranoid_exit
  */
 SYM_CODE_START_LOCAL(paranoid_entry)
 	UNWIND_HINT_FUNC
 	cld
 	PUSH_AND_CLEAR_REGS save_ret=1
 	ENCODE_FRAME_POINTER 8
-	movl	$1, %ebx
-	movl	$MSR_GS_BASE, %ecx
-	rdmsr
-	testl	%edx, %edx
-	js	1f				/* negative -> in kernel */
-	SWAPGS
-	xorl	%ebx, %ebx
 
-1:
 	/*
 	 * Always stash CR3 in %r14.  This value will be restored,
 	 * verbatim, at exit.  Needed if paranoid_entry interrupted
@@ -825,16 +817,60 @@ SYM_CODE_START_LOCAL(paranoid_entry)
 	 * This is also why CS (stashed in the "iret frame" by the
 	 * hardware at entry) can not be used: this may be a return
 	 * to kernel code, but with a user CR3 value.
+	 *
+	 * Switching CR3 does not depend on kernel GSBASE so it can
+	 * be done before switching to the kernel GSBASE. This is
+	 * required for FSGSBASE because the kernel GSBASE has to
+	 * be retrieved from a kernel internal table.
 	 */
 	SAVE_AND_SWITCH_TO_KERNEL_CR3 scratch_reg=%rax save_reg=%r14
 
 	/*
+	 * Handling GSBASE depends on the availability of FSGSBASE.
+	 *
+	 * Without FSGSBASE the kernel enforces that negative GSBASE
+	 * values indicate kernel GSBASE. With FSGSBASE no assumptions
+	 * can be made about the GSBASE value when entering from user
+	 * space.
+	 */
+	ALTERNATIVE "jmp .Lparanoid_entry_checkgs", "", X86_FEATURE_FSGSBASE
+
+	/*
+	 * Read the current GSBASE and store it in %rbx unconditionally,
+	 * retrieve and set the current CPUs kernel GSBASE. The stored value
+	 * has to be restored in paranoid_exit unconditionally.
+	 *
+	 * The MSR write ensures that no subsequent load is based on a
+	 * mispredicted GSBASE. No extra FENCE required.
+	 */
+	SAVE_AND_SET_GSBASE scratch_reg=%rax save_reg=%rbx
+	ret
+
+.Lparanoid_entry_checkgs:
+	/* EBX = 1 -> kernel GSBASE active, no restore required */
+	movl	$1, %ebx
+	/*
+	 * The kernel-enforced convention is a negative GSBASE indicates
+	 * a kernel value. No SWAPGS needed on entry and exit.
+	 */
+	movl	$MSR_GS_BASE, %ecx
+	rdmsr
+	testl	%edx, %edx
+	jns	.Lparanoid_entry_swapgs
+	ret
+
+.Lparanoid_entry_swapgs:
+	SWAPGS
+
+	/*
 	 * The above SAVE_AND_SWITCH_TO_KERNEL_CR3 macro doesn't do an
 	 * unconditional CR3 write, even in the PTI case.  So do an lfence
 	 * to prevent GS speculation, regardless of whether PTI is enabled.
 	 */
 	FENCE_SWAPGS_KERNEL_ENTRY
 
+	/* EBX = 0 -> SWAPGS required on exit */
+	xorl	%ebx, %ebx
 	ret
 SYM_CODE_END(paranoid_entry)
 
@@ -845,23 +881,45 @@ SYM_CODE_END(paranoid_entry)
  *
  * We may be returning to very strange contexts (e.g. very early
  * in syscall entry), so checking for preemption here would
- * be complicated.  Fortunately, we there's no good reason
- * to try to handle preemption here.
+ * be complicated.  Fortunately, there's no good reason to try
+ * to handle preemption here.
+ *
+ * R/EBX contains the GSBASE related information depending on the
+ * availability of the FSGSBASE instructions:
+ *
+ * FSGSBASE	R/EBX
+ *     N        0 -> SWAPGS on exit
+ *              1 -> no SWAPGS on exit
  *
- * On entry, ebx is "no swapgs" flag (1: don't need swapgs, 0: need it)
+ *     Y        User space GSBASE, must be restored unconditionally
  */
 SYM_CODE_START_LOCAL(paranoid_exit)
 	UNWIND_HINT_REGS
-	testl	%ebx, %ebx			/* swapgs needed? */
-	jnz	.Lparanoid_exit_no_swapgs
-	/* Always restore stashed CR3 value (see paranoid_entry) */
-	RESTORE_CR3	scratch_reg=%rbx save_reg=%r14
+	/*
+	 * The order of operations is important. RESTORE_CR3 requires
+	 * kernel GSBASE.
+	 *
+	 * NB to anyone to try to optimize this code: this code does
+	 * not execute at all for exceptions from user mode. Those
+	 * exceptions go through error_exit instead.
+	 */
+	RESTORE_CR3	scratch_reg=%rax save_reg=%r14
+
+	/* Handle the three GSBASE cases */
+	ALTERNATIVE "jmp .Lparanoid_exit_checkgs", "", X86_FEATURE_FSGSBASE
+
+	/* With FSGSBASE enabled, unconditionally restore GSBASE */
+	wrgsbase	%rbx
+	jmp		restore_regs_and_return_to_kernel
+
+.Lparanoid_exit_checkgs:
+	/* On non-FSGSBASE systems, conditionally do SWAPGS */
+	testl		%ebx, %ebx
+	jnz		restore_regs_and_return_to_kernel
+
+	/* We are returning to a context with user GSBASE */
 	SWAPGS_UNSAFE_STACK
-	jmp	restore_regs_and_return_to_kernel
-.Lparanoid_exit_no_swapgs:
-	/* Always restore stashed CR3 value (see paranoid_entry) */
-	RESTORE_CR3	scratch_reg=%rbx save_reg=%r14
-	jmp restore_regs_and_return_to_kernel
+	jmp		restore_regs_and_return_to_kernel
 SYM_CODE_END(paranoid_exit)
 
 /*
@@ -1266,10 +1324,27 @@ end_repeat_nmi:
 	/* Always restore stashed CR3 value (see paranoid_entry) */
 	RESTORE_CR3 scratch_reg=%r15 save_reg=%r14
 
-	testl	%ebx, %ebx			/* swapgs needed? */
+	/*
+	 * The above invocation of paranoid_entry stored the GSBASE
+	 * related information in R/EBX depending on the availability
+	 * of FSGSBASE.
+	 *
+	 * If FSGSBASE is enabled, restore the saved GSBASE value
+	 * unconditionally, otherwise take the conditional SWAPGS path.
+	 */
+	ALTERNATIVE "jmp nmi_no_fsgsbase", "", X86_FEATURE_FSGSBASE
+
+	wrgsbase	%rbx
+	jmp	nmi_restore
+
+nmi_no_fsgsbase:
+	/* EBX == 0 -> invoke SWAPGS */
+	testl	%ebx, %ebx
 	jnz	nmi_restore
+
 nmi_swapgs:
 	SWAPGS_UNSAFE_STACK
+
 nmi_restore:
 	POP_REGS
 
diff --git a/arch/x86/include/asm/fsgsbase.h b/arch/x86/include/asm/fsgsbase.h
index bca4c743de77..d552646411a9 100644
--- a/arch/x86/include/asm/fsgsbase.h
+++ b/arch/x86/include/asm/fsgsbase.h
@@ -19,36 +19,65 @@ extern unsigned long x86_gsbase_read_task(struct task_struct *task);
 extern void x86_fsbase_write_task(struct task_struct *task, unsigned long fsbase);
 extern void x86_gsbase_write_task(struct task_struct *task, unsigned long gsbase);
 
-/* Helper functions for reading/writing FS/GS base */
+/* Must be protected by X86_FEATURE_FSGSBASE check. */
 
-static inline unsigned long x86_fsbase_read_cpu(void)
+static __always_inline unsigned long rdfsbase(void)
 {
 	unsigned long fsbase;
 
-	rdmsrl(MSR_FS_BASE, fsbase);
+	asm volatile("rdfsbase %0" : "=r" (fsbase) :: "memory");
 
 	return fsbase;
 }
 
-static inline unsigned long x86_gsbase_read_cpu_inactive(void)
+static __always_inline unsigned long rdgsbase(void)
 {
 	unsigned long gsbase;
 
-	rdmsrl(MSR_KERNEL_GS_BASE, gsbase);
+	asm volatile("rdgsbase %0" : "=r" (gsbase) :: "memory");
 
 	return gsbase;
 }
 
-static inline void x86_fsbase_write_cpu(unsigned long fsbase)
+static __always_inline void wrfsbase(unsigned long fsbase)
 {
-	wrmsrl(MSR_FS_BASE, fsbase);
+	asm volatile("wrfsbase %0" :: "r" (fsbase) : "memory");
 }
 
-static inline void x86_gsbase_write_cpu_inactive(unsigned long gsbase)
+static __always_inline void wrgsbase(unsigned long gsbase)
 {
-	wrmsrl(MSR_KERNEL_GS_BASE, gsbase);
+	asm volatile("wrgsbase %0" :: "r" (gsbase) : "memory");
 }
 
+#include <asm/cpufeature.h>
+
+/* Helper functions for reading/writing FS/GS base */
+
+static inline unsigned long x86_fsbase_read_cpu(void)
+{
+	unsigned long fsbase;
+
+	if (static_cpu_has(X86_FEATURE_FSGSBASE))
+		fsbase = rdfsbase();
+	else
+		rdmsrl(MSR_FS_BASE, fsbase);
+
+	return fsbase;
+}
+
+static inline void x86_fsbase_write_cpu(unsigned long fsbase)
+{
+	if (static_cpu_has(X86_FEATURE_FSGSBASE))
+		wrfsbase(fsbase);
+	else
+		wrmsrl(MSR_FS_BASE, fsbase);
+}
+
+extern unsigned long x86_gsbase_read_cpu_inactive(void);
+extern void x86_gsbase_write_cpu_inactive(unsigned long gsbase);
+extern unsigned long x86_fsgsbase_read_task(struct task_struct *task,
+					    unsigned short selector);
+
 #endif /* CONFIG_X86_64 */
 
 #endif /* __ASSEMBLY__ */
diff --git a/arch/x86/include/asm/inst.h b/arch/x86/include/asm/inst.h
index 438ccd4f3cc4..bd7f02480ca1 100644
--- a/arch/x86/include/asm/inst.h
+++ b/arch/x86/include/asm/inst.h
@@ -143,6 +143,21 @@
 	.macro MODRM mod opd1 opd2
 	.byte \mod | (\opd1 & 7) | ((\opd2 & 7) << 3)
 	.endm
+
+.macro RDPID opd
+	REG_TYPE rdpid_opd_type \opd
+	.if rdpid_opd_type == REG_TYPE_R64
+	R64_NUM rdpid_opd \opd
+	.else
+	R32_NUM rdpid_opd \opd
+	.endif
+	.byte 0xf3
+	.if rdpid_opd > 7
+	PFX_REX rdpid_opd 0
+	.endif
+	.byte 0x0f, 0xc7
+	MODRM 0xc0 rdpid_opd 0x7
+.endm
 #endif
 
 #endif
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 68ba42fdd184..97143d87994c 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -457,10 +457,8 @@ static inline unsigned long cpu_kernelmode_gs_base(int cpu)
 DECLARE_PER_CPU(unsigned int, irq_count);
 extern asmlinkage void ignore_sysret(void);
 
-#if IS_ENABLED(CONFIG_KVM)
 /* Save actual FS/GS selectors and bases to current->thread */
-void save_fsgs_for_kvm(void);
-#endif
+void current_save_fsgs(void);
 #else	/* X86_64 */
 #ifdef CONFIG_STACKPROTECTOR
 /*
@@ -575,7 +573,7 @@ native_load_sp0(unsigned long sp0)
 	this_cpu_write(cpu_tss_rw.x86_tss.sp0, sp0);
 }
 
-static inline void native_swapgs(void)
+static __always_inline void native_swapgs(void)
 {
 #ifdef CONFIG_X86_64
 	asm volatile("swapgs" ::: "memory");
diff --git a/arch/x86/include/uapi/asm/hwcap2.h b/arch/x86/include/uapi/asm/hwcap2.h
index 8b2effe6efb8..5fdfcb47000f 100644
--- a/arch/x86/include/uapi/asm/hwcap2.h
+++ b/arch/x86/include/uapi/asm/hwcap2.h
@@ -5,4 +5,7 @@
 /* MONITOR/MWAIT enabled in Ring 3 */
 #define HWCAP2_RING3MWAIT		(1 << 0)
 
+/* Kernel allows FSGSBASE instructions available in Ring 3 */
+#define HWCAP2_FSGSBASE			BIT(1)
+
 #endif
diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
index 7beaefa9d198..f0b743a2fe9c 100644
--- a/arch/x86/kernel/cpu/bugs.c
+++ b/arch/x86/kernel/cpu/bugs.c
@@ -543,14 +543,12 @@ static void __init spectre_v1_select_mitigation(void)
 		 * If FSGSBASE is enabled, the user can put a kernel address in
 		 * GS, in which case SMAP provides no protection.
 		 *
-		 * [ NOTE: Don't check for X86_FEATURE_FSGSBASE until the
-		 *	   FSGSBASE enablement patches have been merged. ]
-		 *
 		 * If FSGSBASE is disabled, the user can only put a user space
 		 * address in GS.  That makes an attack harder, but still
 		 * possible if there's no SMAP protection.
 		 */
-		if (!smap_works_speculatively()) {
+		if (boot_cpu_has(X86_FEATURE_FSGSBASE) ||
+		    !smap_works_speculatively()) {
 			/*
 			 * Mitigation can be provided from SWAPGS itself or
 			 * PTI as the CR3 write in the Meltdown mitigation
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 95c090a45b4b..965474d78cef 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -441,6 +441,22 @@ static void __init setup_cr_pinning(void)
 	static_key_enable(&cr_pinning.key);
 }
 
+static __init int x86_nofsgsbase_setup(char *arg)
+{
+	/* Require an exact match without trailing characters. */
+	if (strlen(arg))
+		return 0;
+
+	/* Do not emit a message if the feature is not present. */
+	if (!boot_cpu_has(X86_FEATURE_FSGSBASE))
+		return 1;
+
+	setup_clear_cpu_cap(X86_FEATURE_FSGSBASE);
+	pr_info("FSGSBASE disabled via kernel command line\n");
+	return 1;
+}
+__setup("nofsgsbase", x86_nofsgsbase_setup);
+
 /*
  * Protection Keys are not available in 32-bit mode.
  */
@@ -1495,6 +1511,12 @@ static void identify_cpu(struct cpuinfo_x86 *c)
 	setup_smap(c);
 	setup_umip(c);
 
+	/* Enable FSGSBASE instructions if available. */
+	if (cpu_has(c, X86_FEATURE_FSGSBASE)) {
+		cr4_set_bits(X86_CR4_FSGSBASE);
+		elf_hwcap2 |= HWCAP2_FSGSBASE;
+	}
+
 	/*
 	 * The vendor-specific functions might have changed features.
 	 * Now we do "generic changes."
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 4298634d3f85..994d8393f2f7 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -140,10 +140,12 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
 	memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
 
 #ifdef CONFIG_X86_64
-	savesegment(gs, p->thread.gsindex);
-	p->thread.gsbase = p->thread.gsindex ? 0 : current->thread.gsbase;
-	savesegment(fs, p->thread.fsindex);
-	p->thread.fsbase = p->thread.fsindex ? 0 : current->thread.fsbase;
+	current_save_fsgs();
+	p->thread.fsindex = current->thread.fsindex;
+	p->thread.fsbase = current->thread.fsbase;
+	p->thread.gsindex = current->thread.gsindex;
+	p->thread.gsbase = current->thread.gsbase;
+
 	savesegment(es, p->thread.es);
 	savesegment(ds, p->thread.ds);
 #else
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 04d201ad3a1e..d6f946707270 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -151,6 +151,56 @@ enum which_selector {
 };
 
 /*
+ * Out of line to be protected from kprobes and tracing. If this would be
+ * traced or probed than any access to a per CPU variable happens with
+ * the wrong GS.
+ *
+ * It is not used on Xen paravirt. When paravirt support is needed, it
+ * needs to be renamed with native_ prefix.
+ */
+static noinstr unsigned long __rdgsbase_inactive(void)
+{
+	unsigned long gsbase;
+
+	lockdep_assert_irqs_disabled();
+
+	if (!static_cpu_has(X86_FEATURE_XENPV)) {
+		native_swapgs();
+		gsbase = rdgsbase();
+		native_swapgs();
+	} else {
+		instrumentation_begin();
+		rdmsrl(MSR_KERNEL_GS_BASE, gsbase);
+		instrumentation_end();
+	}
+
+	return gsbase;
+}
+
+/*
+ * Out of line to be protected from kprobes and tracing. If this would be
+ * traced or probed than any access to a per CPU variable happens with
+ * the wrong GS.
+ *
+ * It is not used on Xen paravirt. When paravirt support is needed, it
+ * needs to be renamed with native_ prefix.
+ */
+static noinstr void __wrgsbase_inactive(unsigned long gsbase)
+{
+	lockdep_assert_irqs_disabled();
+
+	if (!static_cpu_has(X86_FEATURE_XENPV)) {
+		native_swapgs();
+		wrgsbase(gsbase);
+		native_swapgs();
+	} else {
+		instrumentation_begin();
+		wrmsrl(MSR_KERNEL_GS_BASE, gsbase);
+		instrumentation_end();
+	}
+}
+
+/*
  * Saves the FS or GS base for an outgoing thread if FSGSBASE extensions are
  * not available.  The goal is to be reasonably fast on non-FSGSBASE systems.
  * It's forcibly inlined because it'll generate better code and this function
@@ -199,22 +249,35 @@ static __always_inline void save_fsgs(struct task_struct *task)
 {
 	savesegment(fs, task->thread.fsindex);
 	savesegment(gs, task->thread.gsindex);
-	save_base_legacy(task, task->thread.fsindex, FS);
-	save_base_legacy(task, task->thread.gsindex, GS);
+	if (static_cpu_has(X86_FEATURE_FSGSBASE)) {
+		/*
+		 * If FSGSBASE is enabled, we can't make any useful guesses
+		 * about the base, and user code expects us to save the current
+		 * value.  Fortunately, reading the base directly is efficient.
+		 */
+		task->thread.fsbase = rdfsbase();
+		task->thread.gsbase = __rdgsbase_inactive();
+	} else {
+		save_base_legacy(task, task->thread.fsindex, FS);
+		save_base_legacy(task, task->thread.gsindex, GS);
+	}
 }
 
-#if IS_ENABLED(CONFIG_KVM)
 /*
  * While a process is running,current->thread.fsbase and current->thread.gsbase
- * may not match the corresponding CPU registers (see save_base_legacy()). KVM
- * wants an efficient way to save and restore FSBASE and GSBASE.
- * When FSGSBASE extensions are enabled, this will have to use RD{FS,GS}BASE.
+ * may not match the corresponding CPU registers (see save_base_legacy()).
  */
-void save_fsgs_for_kvm(void)
+void current_save_fsgs(void)
 {
+	unsigned long flags;
+
+	/* Interrupts need to be off for FSGSBASE */
+	local_irq_save(flags);
 	save_fsgs(current);
+	local_irq_restore(flags);
 }
-EXPORT_SYMBOL_GPL(save_fsgs_for_kvm);
+#if IS_ENABLED(CONFIG_KVM)
+EXPORT_SYMBOL_GPL(current_save_fsgs);
 #endif
 
 static __always_inline void loadseg(enum which_selector which,
@@ -279,14 +342,26 @@ static __always_inline void load_seg_legacy(unsigned short prev_index,
 static __always_inline void x86_fsgsbase_load(struct thread_struct *prev,
 					      struct thread_struct *next)
 {
-	load_seg_legacy(prev->fsindex, prev->fsbase,
-			next->fsindex, next->fsbase, FS);
-	load_seg_legacy(prev->gsindex, prev->gsbase,
-			next->gsindex, next->gsbase, GS);
+	if (static_cpu_has(X86_FEATURE_FSGSBASE)) {
+		/* Update the FS and GS selectors if they could have changed. */
+		if (unlikely(prev->fsindex || next->fsindex))
+			loadseg(FS, next->fsindex);
+		if (unlikely(prev->gsindex || next->gsindex))
+			loadseg(GS, next->gsindex);
+
+		/* Update the bases. */
+		wrfsbase(next->fsbase);
+		__wrgsbase_inactive(next->gsbase);
+	} else {
+		load_seg_legacy(prev->fsindex, prev->fsbase,
+				next->fsindex, next->fsbase, FS);
+		load_seg_legacy(prev->gsindex, prev->gsbase,
+				next->gsindex, next->gsbase, GS);
+	}
 }
 
-static unsigned long x86_fsgsbase_read_task(struct task_struct *task,
-					    unsigned short selector)
+unsigned long x86_fsgsbase_read_task(struct task_struct *task,
+				     unsigned short selector)
 {
 	unsigned short idx = selector >> 3;
 	unsigned long base;
@@ -328,13 +403,44 @@ static unsigned long x86_fsgsbase_read_task(struct task_struct *task,
 	return base;
 }
 
+unsigned long x86_gsbase_read_cpu_inactive(void)
+{
+	unsigned long gsbase;
+
+	if (static_cpu_has(X86_FEATURE_FSGSBASE)) {
+		unsigned long flags;
+
+		local_irq_save(flags);
+		gsbase = __rdgsbase_inactive();
+		local_irq_restore(flags);
+	} else {
+		rdmsrl(MSR_KERNEL_GS_BASE, gsbase);
+	}
+
+	return gsbase;
+}
+
+void x86_gsbase_write_cpu_inactive(unsigned long gsbase)
+{
+	if (static_cpu_has(X86_FEATURE_FSGSBASE)) {
+		unsigned long flags;
+
+		local_irq_save(flags);
+		__wrgsbase_inactive(gsbase);
+		local_irq_restore(flags);
+	} else {
+		wrmsrl(MSR_KERNEL_GS_BASE, gsbase);
+	}
+}
+
 unsigned long x86_fsbase_read_task(struct task_struct *task)
 {
 	unsigned long fsbase;
 
 	if (task == current)
 		fsbase = x86_fsbase_read_cpu();
-	else if (task->thread.fsindex == 0)
+	else if (static_cpu_has(X86_FEATURE_FSGSBASE) ||
+		 (task->thread.fsindex == 0))
 		fsbase = task->thread.fsbase;
 	else
 		fsbase = x86_fsgsbase_read_task(task, task->thread.fsindex);
@@ -348,7 +454,8 @@ unsigned long x86_gsbase_read_task(struct task_struct *task)
 
 	if (task == current)
 		gsbase = x86_gsbase_read_cpu_inactive();
-	else if (task->thread.gsindex == 0)
+	else if (static_cpu_has(X86_FEATURE_FSGSBASE) ||
+		 (task->thread.gsindex == 0))
 		gsbase = task->thread.gsbase;
 	else
 		gsbase = x86_fsgsbase_read_task(task, task->thread.gsindex);
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 44130588987f..3f006489087f 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -281,17 +281,9 @@ static int set_segment_reg(struct task_struct *task,
 		return -EIO;
 
 	/*
-	 * This function has some ABI oddities.
-	 *
-	 * A 32-bit ptracer probably expects that writing FS or GS will change
-	 * FSBASE or GSBASE respectively.  In the absence of FSGSBASE support,
-	 * this code indeed has that effect.  When FSGSBASE is added, this
-	 * will require a special case.
-	 *
-	 * For existing 64-bit ptracers, writing FS or GS *also* currently
-	 * changes the base if the selector is nonzero the next time the task
-	 * is run.  This behavior may not be needed, and trying to preserve it
-	 * when FSGSBASE is added would be complicated at best.
+	 * Writes to FS and GS will change the stored selector.  Whether
+	 * this changes the segment base as well depends on whether
+	 * FSGSBASE is enabled.
 	 */
 
 	switch (offset) {
@@ -379,25 +371,12 @@ static int putreg(struct task_struct *child,
 	case offsetof(struct user_regs_struct,fs_base):
 		if (value >= TASK_SIZE_MAX)
 			return -EIO;
-		/*
-		 * When changing the FS base, use do_arch_prctl_64()
-		 * to set the index to zero and to set the base
-		 * as requested.
-		 *
-		 * NB: This behavior is nonsensical and likely needs to
-		 * change when FSGSBASE support is added.
-		 */
-		if (child->thread.fsbase != value)
-			return do_arch_prctl_64(child, ARCH_SET_FS, value);
+		x86_fsbase_write_task(child, value);
 		return 0;
 	case offsetof(struct user_regs_struct,gs_base):
-		/*
-		 * Exactly the same here as the %fs handling above.
-		 */
 		if (value >= TASK_SIZE_MAX)
 			return -EIO;
-		if (child->thread.gsbase != value)
-			return do_arch_prctl_64(child, ARCH_SET_GS, value);
+		x86_gsbase_write_task(child, value);
 		return 0;
 #endif
 	}
@@ -880,14 +859,39 @@ long arch_ptrace(struct task_struct *child, long request,
 static int putreg32(struct task_struct *child, unsigned regno, u32 value)
 {
 	struct pt_regs *regs = task_pt_regs(child);
+	int ret;
 
 	switch (regno) {
 
 	SEG32(cs);
 	SEG32(ds);
 	SEG32(es);
-	SEG32(fs);
-	SEG32(gs);
+
+	/*
+	 * A 32-bit ptracer on a 64-bit kernel expects that writing
+	 * FS or GS will also update the base.  This is needed for
+	 * operations like PTRACE_SETREGS to fully restore a saved
+	 * CPU state.
+	 */
+
+	case offsetof(struct user32, regs.fs):
+		ret = set_segment_reg(child,
+				      offsetof(struct user_regs_struct, fs),
+				      value);
+		if (ret == 0)
+			child->thread.fsbase =
+				x86_fsgsbase_read_task(child, value);
+		return ret;
+
+	case offsetof(struct user32, regs.gs):
+		ret = set_segment_reg(child,
+				      offsetof(struct user_regs_struct, gs),
+				      value);
+		if (ret == 0)
+			child->thread.gsbase =
+				x86_fsgsbase_read_task(child, value);
+		return ret;
+
 	SEG32(ss);
 
 	R32(ebx, bx);
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 9909375ee1fd..559634b59d2a 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -1170,7 +1170,7 @@ void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
 
 	gs_base = cpu_kernelmode_gs_base(cpu);
 	if (likely(is_64bit_mm(current->mm))) {
-		save_fsgs_for_kvm();
+		current_save_fsgs();
 		fs_sel = current->thread.fsindex;
 		gs_sel = current->thread.gsindex;
 		fs_base = current->thread.fsbase;