Merge branch 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:
 "Yet another pile of melted spectrum related changes:

   - sanitize the array_index_nospec protection mechanism: Remove the
     overengineered array_index_nospec_mask_check() magic and allow
     const-qualified types as index to avoid temporary storage in a
     non-const local variable.

   - make the microcode loader more robust by properly propagating error
     codes. Provide information about new feature bits after micro code
     was updated so administrators can act upon.

   - optimizations of the entry ASM code which reduce code footprint and
     make the code simpler and faster.

   - fix the {pmd,pud}_{set,clear}_flags() implementations to work
     properly on paravirt kernels by removing the address translation
     operations.

   - revert the harmful vmexit_fill_RSB() optimization

   - use IBRS around firmware calls

   - teach objtool about retpolines and add annotations for indirect
     jumps and calls.

   - explicitly disable jumplabel patching in __init code and handle
     patching failures properly instead of silently ignoring them.

   - remove indirect paravirt calls for writing the speculation control
     MSR as these calls are obviously proving the same attack vector
     which is tried to be mitigated.

   - a few small fixes which address build issues with recent compiler
     and assembler versions"

* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (38 commits)
  KVM/VMX: Optimize vmx_vcpu_run() and svm_vcpu_run() by marking the RDMSR path as unlikely()
  KVM/x86: Remove indirect MSR op calls from SPEC_CTRL
  objtool, retpolines: Integrate objtool with retpoline support more closely
  x86/entry/64: Simplify ENCODE_FRAME_POINTER
  extable: Make init_kernel_text() global
  jump_label: Warn on failed jump_label patching attempt
  jump_label: Explicitly disable jump labels in __init code
  x86/entry/64: Open-code switch_to_thread_stack()
  x86/entry/64: Move ASM_CLAC to interrupt_entry()
  x86/entry/64: Remove 'interrupt' macro
  x86/entry/64: Move the switch_to_thread_stack() call to interrupt_entry()
  x86/entry/64: Move ENTER_IRQ_STACK from interrupt macro to interrupt_entry
  x86/entry/64: Move PUSH_AND_CLEAR_REGS from interrupt macro to helper function
  x86/speculation: Move firmware_restrict_branch_speculation_*() from C to CPP
  objtool: Add module specific retpoline rules
  objtool: Add retpoline validation
  objtool: Use existing global variables for options
  x86/mm/sme, objtool: Annotate indirect call in sme_encrypt_execute()
  x86/boot, objtool: Annotate indirect jump in secondary_startup_64()
  x86/paravirt, objtool: Annotate indirect calls
  ...

1 files changed, 118 insertions, 20 deletions

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 81a1be326571..d0dabeae0505 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -8,6 +8,50 @@
 #include <asm/cpufeatures.h>
 #include <asm/msr-index.h>
 
+/*
+ * Fill the CPU return stack buffer.
+ *
+ * Each entry in the RSB, if used for a speculative 'ret', contains an
+ * infinite 'pause; lfence; jmp' loop to capture speculative execution.
+ *
+ * This is required in various cases for retpoline and IBRS-based
+ * mitigations for the Spectre variant 2 vulnerability. Sometimes to
+ * eliminate potentially bogus entries from the RSB, and sometimes
+ * purely to ensure that it doesn't get empty, which on some CPUs would
+ * allow predictions from other (unwanted!) sources to be used.
+ *
+ * We define a CPP macro such that it can be used from both .S files and
+ * inline assembly. It's possible to do a .macro and then include that
+ * from C via asm(".include <asm/nospec-branch.h>") but let's not go there.
+ */
+
+#define RSB_CLEAR_LOOPS		32	/* To forcibly overwrite all entries */
+#define RSB_FILL_LOOPS		16	/* To avoid underflow */
+
+/*
+ * Google experimented with loop-unrolling and this turned out to be
+ * the optimal version — two calls, each with their own speculation
+ * trap should their return address end up getting used, in a loop.
+ */
+#define __FILL_RETURN_BUFFER(reg, nr, sp)	\
+	mov	$(nr/2), reg;			\
+771:						\
+	call	772f;				\
+773:	/* speculation trap */			\
+	pause;					\
+	lfence;					\
+	jmp	773b;				\
+772:						\
+	call	774f;				\
+775:	/* speculation trap */			\
+	pause;					\
+	lfence;					\
+	jmp	775b;				\
+774:						\
+	dec	reg;				\
+	jnz	771b;				\
+	add	$(BITS_PER_LONG/8) * nr, sp;
+
 #ifdef __ASSEMBLY__
 
 /*
@@ -24,6 +68,18 @@
 .endm
 
 /*
+ * This should be used immediately before an indirect jump/call. It tells
+ * objtool the subsequent indirect jump/call is vouched safe for retpoline
+ * builds.
+ */
+.macro ANNOTATE_RETPOLINE_SAFE
+	.Lannotate_\@:
+	.pushsection .discard.retpoline_safe
+	_ASM_PTR .Lannotate_\@
+	.popsection
+.endm
+
+/*
  * These are the bare retpoline primitives for indirect jmp and call.
  * Do not use these directly; they only exist to make the ALTERNATIVE
  * invocation below less ugly.
@@ -59,9 +115,9 @@
 .macro JMP_NOSPEC reg:req
 #ifdef CONFIG_RETPOLINE
 	ANNOTATE_NOSPEC_ALTERNATIVE
-	ALTERNATIVE_2 __stringify(jmp *\reg),				\
+	ALTERNATIVE_2 __stringify(ANNOTATE_RETPOLINE_SAFE; jmp *\reg),	\
 		__stringify(RETPOLINE_JMP \reg), X86_FEATURE_RETPOLINE,	\
-		__stringify(lfence; jmp *\reg), X86_FEATURE_RETPOLINE_AMD
+		__stringify(lfence; ANNOTATE_RETPOLINE_SAFE; jmp *\reg), X86_FEATURE_RETPOLINE_AMD
 #else
 	jmp	*\reg
 #endif
@@ -70,18 +126,25 @@
 .macro CALL_NOSPEC reg:req
 #ifdef CONFIG_RETPOLINE
 	ANNOTATE_NOSPEC_ALTERNATIVE
-	ALTERNATIVE_2 __stringify(call *\reg),				\
+	ALTERNATIVE_2 __stringify(ANNOTATE_RETPOLINE_SAFE; call *\reg),	\
 		__stringify(RETPOLINE_CALL \reg), X86_FEATURE_RETPOLINE,\
-		__stringify(lfence; call *\reg), X86_FEATURE_RETPOLINE_AMD
+		__stringify(lfence; ANNOTATE_RETPOLINE_SAFE; call *\reg), X86_FEATURE_RETPOLINE_AMD
 #else
 	call	*\reg
 #endif
 .endm
 
-/* This clobbers the BX register */
-.macro FILL_RETURN_BUFFER nr:req ftr:req
+ /*
+  * A simpler FILL_RETURN_BUFFER macro. Don't make people use the CPP
+  * monstrosity above, manually.
+  */
+.macro FILL_RETURN_BUFFER reg:req nr:req ftr:req
 #ifdef CONFIG_RETPOLINE
-	ALTERNATIVE "", "call __clear_rsb", \ftr
+	ANNOTATE_NOSPEC_ALTERNATIVE
+	ALTERNATIVE "jmp .Lskip_rsb_\@",				\
+		__stringify(__FILL_RETURN_BUFFER(\reg,\nr,%_ASM_SP))	\
+		\ftr
+.Lskip_rsb_\@:
 #endif
 .endm
 
@@ -93,6 +156,12 @@
 	".long 999b - .\n\t"					\
 	".popsection\n\t"
 
+#define ANNOTATE_RETPOLINE_SAFE					\
+	"999:\n\t"						\
+	".pushsection .discard.retpoline_safe\n\t"		\
+	_ASM_PTR " 999b\n\t"					\
+	".popsection\n\t"
+
 #if defined(CONFIG_X86_64) && defined(RETPOLINE)
 
 /*
@@ -102,6 +171,7 @@
 # define CALL_NOSPEC						\
 	ANNOTATE_NOSPEC_ALTERNATIVE				\
 	ALTERNATIVE(						\
+	ANNOTATE_RETPOLINE_SAFE					\
 	"call *%[thunk_target]\n",				\
 	"call __x86_indirect_thunk_%V[thunk_target]\n",		\
 	X86_FEATURE_RETPOLINE)
@@ -156,26 +226,54 @@ extern char __indirect_thunk_end[];
 static inline void vmexit_fill_RSB(void)
 {
 #ifdef CONFIG_RETPOLINE
-	alternative_input("",
-			  "call __fill_rsb",
-			  X86_FEATURE_RETPOLINE,
-			  ASM_NO_INPUT_CLOBBER(_ASM_BX, "memory"));
+	unsigned long loops;
+
+	asm volatile (ANNOTATE_NOSPEC_ALTERNATIVE
+		      ALTERNATIVE("jmp 910f",
+				  __stringify(__FILL_RETURN_BUFFER(%0, RSB_CLEAR_LOOPS, %1)),
+				  X86_FEATURE_RETPOLINE)
+		      "910:"
+		      : "=r" (loops), ASM_CALL_CONSTRAINT
+		      : : "memory" );
 #endif
 }
 
+#define alternative_msr_write(_msr, _val, _feature)		\
+	asm volatile(ALTERNATIVE("",				\
+				 "movl %[msr], %%ecx\n\t"	\
+				 "movl %[val], %%eax\n\t"	\
+				 "movl $0, %%edx\n\t"		\
+				 "wrmsr",			\
+				 _feature)			\
+		     : : [msr] "i" (_msr), [val] "i" (_val)	\
+		     : "eax", "ecx", "edx", "memory")
+
 static inline void indirect_branch_prediction_barrier(void)
 {
-	asm volatile(ALTERNATIVE("",
-				 "movl %[msr], %%ecx\n\t"
-				 "movl %[val], %%eax\n\t"
-				 "movl $0, %%edx\n\t"
-				 "wrmsr",
-				 X86_FEATURE_USE_IBPB)
-		     : : [msr] "i" (MSR_IA32_PRED_CMD),
-			 [val] "i" (PRED_CMD_IBPB)
-		     : "eax", "ecx", "edx", "memory");
+	alternative_msr_write(MSR_IA32_PRED_CMD, PRED_CMD_IBPB,
+			      X86_FEATURE_USE_IBPB);
 }
 
+/*
+ * With retpoline, we must use IBRS to restrict branch prediction
+ * before calling into firmware.
+ *
+ * (Implemented as CPP macros due to header hell.)
+ */
+#define firmware_restrict_branch_speculation_start()			\
+do {									\
+	preempt_disable();						\
+	alternative_msr_write(MSR_IA32_SPEC_CTRL, SPEC_CTRL_IBRS,	\
+			      X86_FEATURE_USE_IBRS_FW);			\
+} while (0)
+
+#define firmware_restrict_branch_speculation_end()			\
+do {									\
+	alternative_msr_write(MSR_IA32_SPEC_CTRL, 0,			\
+			      X86_FEATURE_USE_IBRS_FW);			\
+	preempt_enable();						\
+} while (0)
+
 #endif /* __ASSEMBLY__ */
 
 /*