x86/entry/unwind: Create stack frames for saved interrupt registers

With frame pointers, when a task is interrupted, its stack is no longer
completely reliable because the function could have been interrupted
before it had a chance to save the previous frame pointer on the stack.
So the caller of the interrupted function could get skipped by a stack
trace.

This is problematic for live patching, which needs to know whether a
stack trace of a sleeping task can be relied upon. There's currently no
way to detect if a sleeping task was interrupted by a page fault
exception or preemption before it went to sleep.

Another issue is that when dumping the stack of an interrupted task, the
unwinder has no way of knowing where the saved pt_regs registers are, so
it can't print them.

This solves those issues by encoding the pt_regs pointer in the frame
pointer on entry from an interrupt or an exception.

This patch also updates the unwinder to be able to decode it, because
otherwise the unwinder would be broken by this change.

Note that this causes a change in the behavior of the unwinder: each
instance of a pt_regs on the stack is now considered a "frame". So
callers of unwind_get_return_address() will now get an occasional
'regs->ip' address that would have previously been skipped over.

Suggested-by: Andy Lutomirski <[e-mail]>
Signed-off-by: Josh Poimboeuf <[e-mail]>
Cc: Andy Lutomirski <[e-mail]>
Cc: Borislav Petkov <[e-mail]>
Cc: Brian Gerst <[e-mail]>
Cc: Denys Vlasenko <[e-mail]>
Cc: H. Peter Anvin <[e-mail]>
Cc: Linus Torvalds <[e-mail]>
Cc: Peter Zijlstra <[e-mail]>
Cc: Thomas Gleixner <[e-mail]>
Link: http://lkml.kernel.org/r/[e-mail]
Signed-off-by: Ingo Molnar <[e-mail]>

arch/x86/entry/calling.h

···
 	.byte 0xf1
 	.endm
 
+/*
+ * This is a sneaky trick to help the unwinder find pt_regs on the stack.  The
+ * frame pointer is replaced with an encoded pointer to pt_regs.  The encoding
+ * is just setting the LSB, which makes it an invalid stack address and is also
+ * a signal to the unwinder that it's a pt_regs pointer in disguise.
+ *
+ * NOTE: This macro must be used *after* SAVE_EXTRA_REGS because it corrupts
+ * the original rbp.
+ */
+.macro ENCODE_FRAME_POINTER ptregs_offset=0
+#ifdef CONFIG_FRAME_POINTER
+	.if \ptregs_offset
+		leaq \ptregs_offset(%rsp), %rbp
+	.else
+		mov %rsp, %rbp
+	.endif
+	orq	$0x1, %rbp
+#endif
+.endm
+
 #endif /* CONFIG_X86_64 */
 
 /*

arch/x86/entry/entry_32.S

···
 	SET_KERNEL_GS %edx
 .endm
 
+/*
+ * This is a sneaky trick to help the unwinder find pt_regs on the stack.  The
+ * frame pointer is replaced with an encoded pointer to pt_regs.  The encoding
+ * is just setting the LSB, which makes it an invalid stack address and is also
+ * a signal to the unwinder that it's a pt_regs pointer in disguise.
+ *
+ * NOTE: This macro must be used *after* SAVE_ALL because it corrupts the
+ * original rbp.
+ */
+.macro ENCODE_FRAME_POINTER
+#ifdef CONFIG_FRAME_POINTER
+	mov %esp, %ebp
+	orl $0x1, %ebp
+#endif
+.endm
+
 .macro RESTORE_INT_REGS
 	popl	%ebx
 	popl	%ecx
···
 	ASM_CLAC
 	addl	$-0x80, (%esp)			/* Adjust vector into the [-256, -1] range */
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	TRACE_IRQS_OFF
 	movl	%esp, %eax
 	call	do_IRQ
···
 	ASM_CLAC;			\
 	pushl	$~(nr);			\
 	SAVE_ALL;			\
+	ENCODE_FRAME_POINTER;		\
 	TRACE_IRQS_OFF			\
 	movl	%esp, %eax;		\
 	call	fn;			\
···
 ENTRY(xen_hypervisor_callback)
 	pushl	$-1				/* orig_ax = -1 => not a system call */
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	TRACE_IRQS_OFF
 
 	/*
···
 	jmp	iret_exc
 5:	pushl	$-1				/* orig_ax = -1 => not a system call */
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	jmp	ret_from_exception
 
 .section .fixup, "ax"
···
 	pushl	%edx
 	pushl	%ecx
 	pushl	%ebx
+	ENCODE_FRAME_POINTER
 	cld
 	movl	$(__KERNEL_PERCPU), %ecx
 	movl	%ecx, %fs
···
 	ASM_CLAC
 	pushl	$-1				# mark this as an int
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	xorl	%edx, %edx			# error code 0
 	movl	%esp, %eax			# pt_regs pointer
 
···
 
 .Ldebug_from_sysenter_stack:
 	/* We're on the SYSENTER stack.  Switch off. */
-	movl	%esp, %ebp
+	movl	%esp, %ebx
 	movl	PER_CPU_VAR(cpu_current_top_of_stack), %esp
 	TRACE_IRQS_OFF
 	call	do_debug
-	movl	%ebp, %esp
+	movl	%ebx, %esp
 	jmp	ret_from_exception
 END(debug)
 
···
 
 	pushl	%eax				# pt_regs->orig_ax
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	xorl	%edx, %edx			# zero error code
 	movl	%esp, %eax			# pt_regs pointer
 
···
 	 * We're on the SYSENTER stack.  Switch off.  No one (not even debug)
 	 * is using the thread stack right now, so it's safe for us to use it.
 	 */
-	movl	%esp, %ebp
+	movl	%esp, %ebx
 	movl	PER_CPU_VAR(cpu_current_top_of_stack), %esp
 	call	do_nmi
-	movl	%ebp, %esp
+	movl	%ebx, %esp
 	jmp	.Lrestore_all_notrace
 
 #ifdef CONFIG_X86_ESPFIX32
···
 	.endr
 	pushl	%eax
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	FIXUP_ESPFIX_STACK			# %eax == %esp
 	xorl	%edx, %edx			# zero error code
 	call	do_nmi
···
 	ASM_CLAC
 	pushl	$-1				# mark this as an int
 	SAVE_ALL
+	ENCODE_FRAME_POINTER
 	TRACE_IRQS_OFF
 	xorl	%edx, %edx			# zero error code
 	movl	%esp, %eax			# pt_regs pointer

arch/x86/entry/entry_64.S

···
 	ALLOC_PT_GPREGS_ON_STACK
 	SAVE_C_REGS
 	SAVE_EXTRA_REGS
+	ENCODE_FRAME_POINTER
 
 	testb	$3, CS(%rsp)
 	jz	1f
···
 	ALLOC_PT_GPREGS_ON_STACK
 	SAVE_C_REGS
 	SAVE_EXTRA_REGS
+	ENCODE_FRAME_POINTER
 	jmp	error_exit
 END(xen_failsafe_callback)
 
···
 	cld
 	SAVE_C_REGS 8
 	SAVE_EXTRA_REGS 8
+	ENCODE_FRAME_POINTER 8
 	movl	$1, %ebx
 	movl	$MSR_GS_BASE, %ecx
 	rdmsr
···
 	cld
 	SAVE_C_REGS 8
 	SAVE_EXTRA_REGS 8
+	ENCODE_FRAME_POINTER 8
 	xorl	%ebx, %ebx
 	testb	$3, CS+8(%rsp)
 	jz	.Lerror_kernelspace
···
 	pushq	%r13		/* pt_regs->r13 */
 	pushq	%r14		/* pt_regs->r14 */
 	pushq	%r15		/* pt_regs->r15 */
+	ENCODE_FRAME_POINTER
 
 	/*
 	 * At this point we no longer need to worry about stack damage
···
 
 	/*
 	 * Return back to user mode.  We must *not* do the normal exit
-	 * work, because we don't want to enable interrupts.  Fortunately,
-	 * do_nmi doesn't modify pt_regs.
+	 * work, because we don't want to enable interrupts.
 	 */
 	SWAPGS
-	jmp	restore_c_regs_and_iret
+	jmp	restore_regs_and_iret
 
 .Lnmi_from_kernel:
 	/*

arch/x86/include/asm/unwind.h

···
 	int graph_idx;
 #ifdef CONFIG_FRAME_POINTER
 	unsigned long *bp;
+	struct pt_regs *regs;
 #else
 	unsigned long *sp;
 #endif
···
 	if (unwind_done(state))
 		return NULL;
 
-	return state->bp + 1;
+	return state->regs ? &state->regs->ip : state->bp + 1;
+}
+
+static inline struct pt_regs *unwind_get_entry_regs(struct unwind_state *state)
+{
+	if (unwind_done(state))
+		return NULL;
+
+	return state->regs;
 }
 
 #else /* !CONFIG_FRAME_POINTER */
 
 static inline
 unsigned long *unwind_get_return_address_ptr(struct unwind_state *state)
+{
+	return NULL;
+}
+
+static inline struct pt_regs *unwind_get_entry_regs(struct unwind_state *state)
 {
 	return NULL;
 }

arch/x86/kernel/unwind_frame.c

···
 	if (unwind_done(state))
 		return 0;
 
+	if (state->regs && user_mode(state->regs))
+		return 0;
+
 	addr = ftrace_graph_ret_addr(state->task, &state->graph_idx, *addr_p,
 				     addr_p);
 
 	return __kernel_text_address(addr) ? addr : 0;
 }
 EXPORT_SYMBOL_GPL(unwind_get_return_address);
+
+/*
+ * This determines if the frame pointer actually contains an encoded pointer to
+ * pt_regs on the stack.  See ENCODE_FRAME_POINTER.
+ */
+static struct pt_regs *decode_frame_pointer(unsigned long *bp)
+{
+	unsigned long regs = (unsigned long)bp;
+
+	if (!(regs & 0x1))
+		return NULL;
+
+	return (struct pt_regs *)(regs & ~0x1);
+}
 
 static bool update_stack_state(struct unwind_state *state, void *addr,
 			       size_t len)
···
 
 bool unwind_next_frame(struct unwind_state *state)
 {
-	unsigned long *next_bp;
+	struct pt_regs *regs;
+	unsigned long *next_bp, *next_frame;
+	size_t next_len;
 
 	if (unwind_done(state))
 		return false;
 
-	next_bp = (unsigned long *)*state->bp;
+	/* have we reached the end? */
+	if (state->regs && user_mode(state->regs))
+		goto the_end;
+
+	/* get the next frame pointer */
+	if (state->regs)
+		next_bp = (unsigned long *)state->regs->bp;
+	else
+		next_bp = (unsigned long *)*state->bp;
+
+	/* is the next frame pointer an encoded pointer to pt_regs? */
+	regs = decode_frame_pointer(next_bp);
+	if (regs) {
+		next_frame = (unsigned long *)regs;
+		next_len = sizeof(*regs);
+	} else {
+		next_frame = next_bp;
+		next_len = FRAME_HEADER_SIZE;
+	}
 
 	/* make sure the next frame's data is accessible */
-	if (!update_stack_state(state, next_bp, FRAME_HEADER_SIZE))
+	if (!update_stack_state(state, next_frame, next_len))
 		return false;
-
 	/* move to the next frame */
-	state->bp = next_bp;
+	if (regs) {
+		state->regs = regs;
+		state->bp = NULL;
+	} else {
+		state->bp = next_bp;
+		state->regs = NULL;
+	}
+
 	return true;
+
+the_end:
+	state->stack_info.type = STACK_TYPE_UNKNOWN;
+	return false;
 }
 EXPORT_SYMBOL_GPL(unwind_next_frame);
 
 void __unwind_start(struct unwind_state *state, struct task_struct *task,
 		    struct pt_regs *regs, unsigned long *first_frame)
 {
+	unsigned long *bp, *frame;
+	size_t len;
+
 	memset(state, 0, sizeof(*state));
 	state->task = task;
 
···
 	}
 
 	/* set up the starting stack frame */
-	state->bp = get_frame_pointer(task, regs);
+	bp = get_frame_pointer(task, regs);
+	regs = decode_frame_pointer(bp);
+	if (regs) {
+		state->regs = regs;
+		frame = (unsigned long *)regs;
+		len = sizeof(*regs);
+	} else {
+		state->bp = bp;
+		frame = bp;
+		len = FRAME_HEADER_SIZE;
+	}
 
 	/* initialize stack info and make sure the frame data is accessible */
-	get_stack_info(state->bp, state->task, &state->stack_info,
+	get_stack_info(frame, state->task, &state->stack_info,
 		       &state->stack_mask);
-	update_stack_state(state, state->bp, FRAME_HEADER_SIZE);
+	update_stack_state(state, frame, len);
 
 	/*
 	 * The caller can provide the address of the first frame directly