From: Peter Zijlstra <peterz@infradead.org>
Date: Tue, 14 Jun 2022 23:15:53 +0200
Subject: x86/entry: Add kernel IBRS implementation
Git-commit: 2dbb887e875b1de3ca8f40ddf26bcfe55798c609
Patch-mainline: v5.19-rc7
References: bsc#1199657 CVE-2022-29900 CVE-2022-29901

Implement Kernel IBRS - currently the only known option to mitigate RSB
underflow speculation issues on Skylake hardware.

Note: since IBRS_ENTER requires fuller context established than
UNTRAIN_RET, it must be placed after it. However, since UNTRAIN_RET
itself implies a RET, it must come after IBRS_ENTER. This means
IBRS_ENTER needs to also move UNTRAIN_RET.

Note 2: KERNEL_IBRS is sub-optimal for XenPV.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
---
 arch/x86/entry/calling.h           |   58 +++++++++++++++++++++++++++++++++++++
 arch/x86/entry/entry_64.S          |   45 ++++++++++++++++++++++++++--
 arch/x86/entry/entry_64_compat.S   |   13 +++++++-
 arch/x86/include/asm/cpufeatures.h |    2 -
 4 files changed, 111 insertions(+), 7 deletions(-)

--- a/arch/x86/entry/calling.h
+++ b/arch/x86/entry/calling.h
@@ -6,6 +6,8 @@
 #include <asm/percpu.h>
 #include <asm/asm-offsets.h>
 #include <asm/processor-flags.h>
+#include <asm/msr.h>
+#include <asm/nospec-branch.h>
 
 /*
 
@@ -309,6 +311,62 @@ For 32-bit we have the following convent
 #endif
 
 /*
+ * IBRS kernel mitigation for Spectre_v2.
+ *
+ * Assumes full context is established (PUSH_REGS, CR3 and GS) and it clobbers
+ * the regs it uses (AX, CX, DX). Must be called before the first RET
+ * instruction (NOTE! UNTRAIN_RET includes a RET instruction)
+ *
+ * The optional argument is used to save/restore the current value,
+ * which is used on the paranoid paths.
+ *
+ * Assumes x86_spec_ctrl_{base,current} to have SPEC_CTRL_IBRS set.
+ */
+.macro IBRS_ENTER save_reg
+	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_KERNEL_IBRS
+	movl	$MSR_IA32_SPEC_CTRL, %ecx
+
+.ifnb \save_reg
+	rdmsr
+	shl	$32, %rdx
+	or	%rdx, %rax
+	mov	%rax, \save_reg
+	test	$SPEC_CTRL_IBRS, %eax
+	jz	.Ldo_wrmsr_\@
+	lfence
+	jmp	.Lend_\@
+.Ldo_wrmsr_\@:
+.endif
+
+	movq	PER_CPU_VAR(x86_spec_ctrl_current), %rdx
+	movl	%edx, %eax
+	shr	$32, %rdx
+	wrmsr
+.Lend_\@:
+.endm
+
+/*
+ * Similar to IBRS_ENTER, requires KERNEL GS,CR3 and clobbers (AX, CX, DX)
+ * regs. Must be called after the last RET.
+ */
+.macro IBRS_EXIT save_reg
+	ALTERNATIVE "jmp .Lend_\@", "", X86_FEATURE_KERNEL_IBRS
+	movl	$MSR_IA32_SPEC_CTRL, %ecx
+
+.ifnb \save_reg
+	mov	\save_reg, %rdx
+.else
+	movq	PER_CPU_VAR(x86_spec_ctrl_current), %rdx
+	andl	$(~SPEC_CTRL_IBRS), %edx
+.endif
+
+	movl	%edx, %eax
+	shr	$32, %rdx
+	wrmsr
+.Lend_\@:
+.endm
+
+/*
  * Mitigate Spectre v1 for conditional swapgs code paths.
  *
  * FENCE_SWAPGS_USER_ENTRY is used in the user entry swapgs code path, to
--- a/arch/x86/entry/entry_64_compat.S
+++ b/arch/x86/entry/entry_64_compat.S
@@ -4,7 +4,6 @@
  *
  * Copyright 2000-2002 Andi Kleen, SuSE Labs.
  */
-#include "calling.h"
 #include <asm/asm-offsets.h>
 #include <asm/current.h>
 #include <asm/errno.h>
@@ -18,6 +17,8 @@
 #include <linux/linkage.h>
 #include <linux/err.h>
 
+#include "calling.h"
+
 	.section .entry.text, "ax"
 
 /*
@@ -107,8 +108,8 @@ ENTRY(entry_SYSENTER_compat)
 	xorl	%r15d, %r15d		/* nospec   r15 */
 	cld
 
+	IBRS_ENTER
 	UNTRAIN_RET
-
 	/*
 	 * SYSENTER doesn't filter flags, so we need to clear NT and AC
 	 * ourselves.  To save a few cycles, we can check whether
@@ -256,6 +257,9 @@ GLOBAL(entry_SYSCALL_compat_after_hwfram
 	 */
 	TRACE_IRQS_OFF
 
+	IBRS_ENTER
+	UNTRAIN_RET
+
 	movq	%rsp, %rdi
 	call	do_fast_syscall_32
 	/* XEN PV guests always use IRET path */
@@ -270,6 +274,8 @@ sysret32_from_system_call:
 	 */
 	STACKLEAK_ERASE
 	TRACE_IRQS_ON			/* User mode traces as IRQs on. */
+	IBRS_EXIT
+
 	movq	RBX(%rsp), %rbx		/* pt_regs->rbx */
 	movq	RBP(%rsp), %rbp		/* pt_regs->rbp */
 	movq	EFLAGS(%rsp), %r11	/* pt_regs->flags (in r11) */
@@ -412,6 +418,9 @@ ENTRY(entry_INT80_compat)
 	 */
 	TRACE_IRQS_OFF
 
+	IBRS_ENTER
+	UNTRAIN_RET
+
 	movq	%rsp, %rdi
 	call	do_int80_syscall_32
 .Lsyscall_32_done:
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -155,7 +155,6 @@ ENTRY(entry_SYSCALL_64)
 	movq	%rsp, PER_CPU_VAR(cpu_tss_rw + TSS_sp2)
 	SWITCH_TO_KERNEL_CR3 scratch_reg=%rsp
 	movq	PER_CPU_VAR(cpu_current_top_of_stack), %rsp
-	UNTRAIN_RET
 
 	/* Construct struct pt_regs on stack */
 	pushq	$__USER_DS				/* pt_regs->ss */
@@ -173,6 +172,11 @@ GLOBAL(entry_SYSCALL_64_after_hwframe)
 	/* IRQs are off. */
 	movq	%rax, %rdi
 	movq	%rsp, %rsi
+
+	/* clobbers %rax, make sure it is after saving the syscall nr */
+	IBRS_ENTER
+	UNTRAIN_RET
+
 	call	do_syscall_64		/* returns with IRQs disabled */
 
 	TRACE_IRQS_IRETQ		/* we're about to change IF */
@@ -249,6 +253,7 @@ GLOBAL(entry_SYSCALL_64_after_hwframe)
 	 * perf profiles. Nothing jumps here.
 	 */
 syscall_return_via_sysret:
+	IBRS_EXIT
 	POP_REGS pop_rdi=0
 
 	/*
@@ -328,6 +333,9 @@ ENDPROC(__switch_to_asm)
  * rax: prev task we switched from
  * rbx: kernel thread func (NULL for user thread)
  * r12: kernel thread arg
+ *
+ * R14 - old CR3
+ * R15 - old SPEC_CTRL
  */
 ENTRY(ret_from_fork)
 	UNWIND_HINT_EMPTY
@@ -621,6 +629,7 @@ GLOBAL(retint_user)
 	TRACE_IRQS_IRETQ
 
 GLOBAL(swapgs_restore_regs_and_return_to_usermode)
+	IBRS_EXIT
 #ifdef CONFIG_DEBUG_ENTRY
 	/* Assert that pt_regs indicates user mode. */
 	testb	$3, CS(%rsp)
@@ -1219,6 +1228,9 @@ idtentry machine_check		do_mce			has_err
  * 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
+ *
+ * R14 - old CR3
+ * R15 - old SPEC_CTRL
  */
 ENTRY(paranoid_entry)
 	UNWIND_HINT_FUNC
@@ -1245,7 +1257,6 @@ ENTRY(paranoid_entry)
 	 * to kernel code, but with a user CR3 value.
 	 */
 	SAVE_AND_SWITCH_TO_KERNEL_CR3 scratch_reg=%rax save_reg=%r14
-	UNTRAIN_RET
 
 	/*
 	 * The above SAVE_AND_SWITCH_TO_KERNEL_CR3 macro doesn't do an
@@ -1254,7 +1265,14 @@ ENTRY(paranoid_entry)
 	 */
 	FENCE_SWAPGS_KERNEL_ENTRY
 
-	ret
+	/*
+	 * Once we have CR3 and %GS setup save and set SPEC_CTRL. Just like
+	 * CR3 above, keep the old value in a callee saved register.
+	 */
+	IBRS_ENTER save_reg=%r15
+	UNTRAIN_RET
+
+        RET
 END(paranoid_entry)
 
 /*
@@ -1268,9 +1286,19 @@ END(paranoid_entry)
  * to try to handle preemption here.
  *
  * On entry, ebx is "no swapgs" flag (1: don't need swapgs, 0: need it)
+ *
+ * R14 - old CR3
+ * R15 - old SPEC_CTRL
  */
 ENTRY(paranoid_exit)
 	UNWIND_HINT_REGS
+
+	/*
+	 * Must restore IBRS state before both CR3 and %GS since we need access
+	 * to the per-CPU x86_spec_ctrl_shadow variable.
+	 */
+	IBRS_EXIT save_reg=%r15
+
 	DISABLE_INTERRUPTS(CLBR_ANY)
 	TRACE_IRQS_OFF_DEBUG
 	testl	%ebx, %ebx			/* swapgs needed? */
@@ -1307,6 +1335,7 @@ ENTRY(error_entry)
 	FENCE_SWAPGS_USER_ENTRY
 	/* We have user CR3.  Change to kernel CR3. */
 	SWITCH_TO_KERNEL_CR3 scratch_reg=%rax
+	IBRS_ENTER
 	UNTRAIN_RET
 
 .Lerror_entry_from_usermode_after_swapgs:
@@ -1348,6 +1377,9 @@ ENTRY(error_entry)
 	SWAPGS
 	FENCE_SWAPGS_USER_ENTRY
 	SWITCH_TO_KERNEL_CR3 scratch_reg=%rax
+	IBRS_ENTER
+	UNTRAIN_RET
+
 	jmp .Lerror_entry_done
 
 .Lbstep_iret:
@@ -1458,7 +1490,6 @@ ENTRY(nmi)
 	movq	%rsp, %rdx
 	movq	PER_CPU_VAR(cpu_current_top_of_stack), %rsp
 	UNWIND_HINT_IRET_REGS base=%rdx offset=8
-	UNTRAIN_RET
 	pushq	5*8(%rdx)	/* pt_regs->ss */
 	pushq	4*8(%rdx)	/* pt_regs->rsp */
 	pushq	3*8(%rdx)	/* pt_regs->flags */
@@ -1469,6 +1500,9 @@ ENTRY(nmi)
 	PUSH_AND_CLEAR_REGS rdx=(%rdx)
 	ENCODE_FRAME_POINTER
 
+	IBRS_ENTER
+	UNTRAIN_RET
+
 	/*
 	 * At this point we no longer need to worry about stack damage
 	 * due to nesting -- we're on the normal thread stack and we're
@@ -1692,6 +1726,9 @@ end_repeat_nmi:
 	movq	$-1, %rsi
 	call	do_nmi
 
+	/* Always restore stashed SPEC_CTRL value (see paranoid_entry) */
+	IBRS_EXIT save_reg=%r15
+
 	/* Always restore stashed CR3 value (see paranoid_entry) */
 	RESTORE_CR3 scratch_reg=%r15 save_reg=%r14
 
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -202,7 +202,7 @@
 #define X86_FEATURE_PROC_FEEDBACK	( 7*32+ 9) /* AMD ProcFeedbackInterface */
 #define X86_FEATURE_SME			( 7*32+10) /* AMD Secure Memory Encryption */
 #define X86_FEATURE_PTI			( 7*32+11) /* Kernel Page Table Isolation enabled */
-/* FREE!				( 7*32+12) */
+#define X86_FEATURE_KERNEL_IBRS		( 7*32+12) /* "" Set/clear IBRS on kernel entry/exit */
 /* FREE!				( 7*32+13) */
 #define X86_FEATURE_INTEL_PPIN		( 7*32+14) /* Intel Processor Inventory Number */
 #define X86_FEATURE_CDP_L2		( 7*32+15) /* Code and Data Prioritization L2 */