21 files changed, 602 insertions, 302 deletions

diff --git a/arch/x86/include/asm/uprobes.h b/arch/x86/include/asm/uprobes.h
index 3087ea9c5f2..93bee7b9385 100644
--- a/arch/x86/include/asm/uprobes.h
+++ b/arch/x86/include/asm/uprobes.h
@@ -33,15 +33,27 @@ typedef u8 uprobe_opcode_t;
 #define UPROBE_SWBP_INSN		0xcc
 #define UPROBE_SWBP_INSN_SIZE		   1
 
+struct uprobe_xol_ops;
+
 struct arch_uprobe {
-	u16				fixups;
 	union {
 		u8			insn[MAX_UINSN_BYTES];
 		u8			ixol[MAX_UINSN_BYTES];
 	};
+
+	u16				fixups;
+	const struct uprobe_xol_ops	*ops;
+
+	union {
 #ifdef CONFIG_X86_64
-	unsigned long			rip_rela_target_address;
+		unsigned long			rip_rela_target_address;
 #endif
+		struct {
+			s32	offs;
+			u8	ilen;
+			u8	opc1;
+		}				branch;
+	};
 };
 
 struct arch_uprobe_task {
diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
index 2ed845928b5..ace22916ade 100644
--- a/arch/x86/kernel/uprobes.c
+++ b/arch/x86/kernel/uprobes.c
@@ -53,7 +53,7 @@
 #define OPCODE1(insn)		((insn)->opcode.bytes[0])
 #define OPCODE2(insn)		((insn)->opcode.bytes[1])
 #define OPCODE3(insn)		((insn)->opcode.bytes[2])
-#define MODRM_REG(insn)		X86_MODRM_REG(insn->modrm.value)
+#define MODRM_REG(insn)		X86_MODRM_REG((insn)->modrm.value)
 
 #define W(row, b0, b1, b2, b3, b4, b5, b6, b7, b8, b9, ba, bb, bc, bd, be, bf)\
 	(((b0##UL << 0x0)|(b1##UL << 0x1)|(b2##UL << 0x2)|(b3##UL << 0x3) |   \
@@ -229,63 +229,6 @@ static int validate_insn_32bits(struct arch_uprobe *auprobe, struct insn *insn)
 	return -ENOTSUPP;
 }
 
-/*
- * Figure out which fixups arch_uprobe_post_xol() will need to perform, and
- * annotate arch_uprobe->fixups accordingly.  To start with,
- * arch_uprobe->fixups is either zero or it reflects rip-related fixups.
- */
-static void prepare_fixups(struct arch_uprobe *auprobe, struct insn *insn)
-{
-	bool fix_ip = true, fix_call = false;	/* defaults */
-	int reg;
-
-	insn_get_opcode(insn);	/* should be a nop */
-
-	switch (OPCODE1(insn)) {
-	case 0x9d:
-		/* popf */
-		auprobe->fixups |= UPROBE_FIX_SETF;
-		break;
-	case 0xc3:		/* ret/lret */
-	case 0xcb:
-	case 0xc2:
-	case 0xca:
-		/* ip is correct */
-		fix_ip = false;
-		break;
-	case 0xe8:		/* call relative - Fix return addr */
-		fix_call = true;
-		break;
-	case 0x9a:		/* call absolute - Fix return addr, not ip */
-		fix_call = true;
-		fix_ip = false;
-		break;
-	case 0xff:
-		insn_get_modrm(insn);
-		reg = MODRM_REG(insn);
-		if (reg == 2 || reg == 3) {
-			/* call or lcall, indirect */
-			/* Fix return addr; ip is correct. */
-			fix_call = true;
-			fix_ip = false;
-		} else if (reg == 4 || reg == 5) {
-			/* jmp or ljmp, indirect */
-			/* ip is correct. */
-			fix_ip = false;
-		}
-		break;
-	case 0xea:		/* jmp absolute -- ip is correct */
-		fix_ip = false;
-		break;
-	default:
-		break;
-	}
-	if (fix_ip)
-		auprobe->fixups |= UPROBE_FIX_IP;
-	if (fix_call)
-		auprobe->fixups |= UPROBE_FIX_CALL;
-}
-
 #ifdef CONFIG_X86_64
 /*
  * If arch_uprobe->insn doesn't use rip-relative addressing, return
@@ -310,15 +253,11 @@ static void prepare_fixups(struct arch_uprobe *auprobe, struct insn *insn)
  *  - The displacement is always 4 bytes.
  */
 static void
-handle_riprel_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, struct insn *insn)
+handle_riprel_insn(struct arch_uprobe *auprobe, struct insn *insn)
 {
 	u8 *cursor;
 	u8 reg;
 
-	if (mm->context.ia32_compat)
-		return;
-
-	auprobe->rip_rela_target_address = 0x0;
 	if (!insn_rip_relative(insn))
 		return;
 
@@ -372,7 +311,48 @@ handle_riprel_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, struct ins
 		cursor++;
 		memmove(cursor, cursor + insn->displacement.nbytes, insn->immediate.nbytes);
 	}
-	return;
+}
+
+/*
+ * If we're emulating a rip-relative instruction, save the contents
+ * of the scratch register and store the target address in that register.
+ */
+static void
+pre_xol_rip_insn(struct arch_uprobe *auprobe, struct pt_regs *regs,
+				struct arch_uprobe_task *autask)
+{
+	if (auprobe->fixups & UPROBE_FIX_RIP_AX) {
+		autask->saved_scratch_register = regs->ax;
+		regs->ax = current->utask->vaddr;
+		regs->ax += auprobe->rip_rela_target_address;
+	} else if (auprobe->fixups & UPROBE_FIX_RIP_CX) {
+		autask->saved_scratch_register = regs->cx;
+		regs->cx = current->utask->vaddr;
+		regs->cx += auprobe->rip_rela_target_address;
+	}
+}
+
+static void
+handle_riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs, long *correction)
+{
+	if (auprobe->fixups & (UPROBE_FIX_RIP_AX | UPROBE_FIX_RIP_CX)) {
+		struct arch_uprobe_task *autask;
+
+		autask = &current->utask->autask;
+		if (auprobe->fixups & UPROBE_FIX_RIP_AX)
+			regs->ax = autask->saved_scratch_register;
+		else
+			regs->cx = autask->saved_scratch_register;
+
+		/*
+		 * The original instruction includes a displacement, and so
+		 * is 4 bytes longer than what we've just single-stepped.
+		 * Caller may need to apply other fixups to handle stuff
+		 * like "jmpq *...(%rip)" and "callq *...(%rip)".
+		 */
+		if (correction)
+			*correction += 4;
+	}
 }
 
 static int validate_insn_64bits(struct arch_uprobe *auprobe, struct insn *insn)
@@ -401,9 +381,19 @@ static int validate_insn_bits(struct arch_uprobe *auprobe, struct mm_struct *mm,
 	return validate_insn_64bits(auprobe, insn);
 }
 #else /* 32-bit: */
-static void handle_riprel_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, struct insn *insn)
+/*
+ * No RIP-relative addressing on 32-bit
+ */
+static void handle_riprel_insn(struct arch_uprobe *auprobe, struct insn *insn)
+{
+}
+static void pre_xol_rip_insn(struct arch_uprobe *auprobe, struct pt_regs *regs,
+				struct arch_uprobe_task *autask)
+{
+}
+static void handle_riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs,
+					long *correction)
 {
-	/* No RIP-relative addressing on 32-bit */
 }
 
 static int validate_insn_bits(struct arch_uprobe *auprobe, struct mm_struct *mm,  struct insn *insn)
@@ -412,141 +402,311 @@ static int validate_insn_bits(struct arch_uprobe *auprobe, struct mm_struct *mm,
 }
 #endif /* CONFIG_X86_64 */
 
-/**
- * arch_uprobe_analyze_insn - instruction analysis including validity and fixups.
- * @mm: the probed address space.
- * @arch_uprobe: the probepoint information.
- * @addr: virtual address at which to install the probepoint
- * Return 0 on success or a -ve number on error.
+struct uprobe_xol_ops {
+	bool	(*emulate)(struct arch_uprobe *, struct pt_regs *);
+	int	(*pre_xol)(struct arch_uprobe *, struct pt_regs *);
+	int	(*post_xol)(struct arch_uprobe *, struct pt_regs *);
+};
+
+static inline int sizeof_long(void)
+{
+	return is_ia32_task() ? 4 : 8;
+}
+
+static int default_pre_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
+{
+	pre_xol_rip_insn(auprobe, regs, &current->utask->autask);
+	return 0;
+}
+
+/*
+ * Adjust the return address pushed by a call insn executed out of line.
  */
-int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long addr)
+static int adjust_ret_addr(unsigned long sp, long correction)
 {
-	int ret;
-	struct insn insn;
+	int rasize = sizeof_long();
+	long ra;
 
-	auprobe->fixups = 0;
-	ret = validate_insn_bits(auprobe, mm, &insn);
-	if (ret != 0)
-		return ret;
+	if (copy_from_user(&ra, (void __user *)sp, rasize))
+		return -EFAULT;
 
-	handle_riprel_insn(auprobe, mm, &insn);
-	prepare_fixups(auprobe, &insn);
+	ra += correction;
+	if (copy_to_user((void __user *)sp, &ra, rasize))
+		return -EFAULT;
 
 	return 0;
 }
 
-#ifdef CONFIG_X86_64
-/*
- * If we're emulating a rip-relative instruction, save the contents
- * of the scratch register and store the target address in that register.
- */
-static void
-pre_xol_rip_insn(struct arch_uprobe *auprobe, struct pt_regs *regs,
-				struct arch_uprobe_task *autask)
+static int default_post_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
-	if (auprobe->fixups & UPROBE_FIX_RIP_AX) {
-		autask->saved_scratch_register = regs->ax;
-		regs->ax = current->utask->vaddr;
-		regs->ax += auprobe->rip_rela_target_address;
-	} else if (auprobe->fixups & UPROBE_FIX_RIP_CX) {
-		autask->saved_scratch_register = regs->cx;
-		regs->cx = current->utask->vaddr;
-		regs->cx += auprobe->rip_rela_target_address;
+	struct uprobe_task *utask = current->utask;
+	long correction = (long)(utask->vaddr - utask->xol_vaddr);
+
+	handle_riprel_post_xol(auprobe, regs, &correction);
+	if (auprobe->fixups & UPROBE_FIX_IP)
+		regs->ip += correction;
+
+	if (auprobe->fixups & UPROBE_FIX_CALL) {
+		if (adjust_ret_addr(regs->sp, correction)) {
+			regs->sp += sizeof_long();
+			return -ERESTART;
+		}
 	}
+
+	return 0;
 }
-#else
-static void
-pre_xol_rip_insn(struct arch_uprobe *auprobe, struct pt_regs *regs,
-				struct arch_uprobe_task *autask)
+
+static struct uprobe_xol_ops default_xol_ops = {
+	.pre_xol  = default_pre_xol_op,
+	.post_xol = default_post_xol_op,
+};
+
+static bool branch_is_call(struct arch_uprobe *auprobe)
 {
-	/* No RIP-relative addressing on 32-bit */
+	return auprobe->branch.opc1 == 0xe8;
 }
-#endif
 
-/*
- * arch_uprobe_pre_xol - prepare to execute out of line.
- * @auprobe: the probepoint information.
- * @regs: reflects the saved user state of current task.
- */
-int arch_uprobe_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
-{
-	struct arch_uprobe_task *autask;
+#define CASE_COND					\
+	COND(70, 71, XF(OF))				\
+	COND(72, 73, XF(CF))				\
+	COND(74, 75, XF(ZF))				\
+	COND(78, 79, XF(SF))				\
+	COND(7a, 7b, XF(PF))				\
+	COND(76, 77, XF(CF) || XF(ZF))			\
+	COND(7c, 7d, XF(SF) != XF(OF))			\
+	COND(7e, 7f, XF(ZF) || XF(SF) != XF(OF))
 
-	autask = &current->utask->autask;
-	autask->saved_trap_nr = current->thread.trap_nr;
-	current->thread.trap_nr = UPROBE_TRAP_NR;
-	regs->ip = current->utask->xol_vaddr;
-	pre_xol_rip_insn(auprobe, regs, autask);
+#define COND(op_y, op_n, expr)				\
+	case 0x ## op_y: DO((expr) != 0)		\
+	case 0x ## op_n: DO((expr) == 0)
 
-	autask->saved_tf = !!(regs->flags & X86_EFLAGS_TF);
-	regs->flags |= X86_EFLAGS_TF;
-	if (test_tsk_thread_flag(current, TIF_BLOCKSTEP))
-		set_task_blockstep(current, false);
+#define XF(xf)	(!!(flags & X86_EFLAGS_ ## xf))
 
-	return 0;
+static bool is_cond_jmp_opcode(u8 opcode)
+{
+	switch (opcode) {
+	#define DO(expr)	\
+		return true;
+	CASE_COND
+	#undef	DO
+
+	default:
+		return false;
+	}
 }
 
-/*
- * This function is called by arch_uprobe_post_xol() to adjust the return
- * address pushed by a call instruction executed out of line.
- */
-static int adjust_ret_addr(unsigned long sp, long correction)
+static bool check_jmp_cond(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
-	int rasize, ncopied;
-	long ra = 0;
+	unsigned long flags = regs->flags;
 
-	if (is_ia32_task())
-		rasize = 4;
-	else
-		rasize = 8;
+	switch (auprobe->branch.opc1) {
+	#define DO(expr)	\
+		return expr;
+	CASE_COND
+	#undef	DO
 
-	ncopied = copy_from_user(&ra, (void __user *)sp, rasize);
-	if (unlikely(ncopied))
-		return -EFAULT;
+	default:	/* not a conditional jmp */
+		return true;
+	}
+}
 
-	ra += correction;
-	ncopied = copy_to_user((void __user *)sp, &ra, rasize);
-	if (unlikely(ncopied))
-		return -EFAULT;
+#undef	XF
+#undef	COND
+#undef	CASE_COND
 
-	return 0;
+static bool branch_emulate_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
+{
+	unsigned long new_ip = regs->ip += auprobe->branch.ilen;
+	unsigned long offs = (long)auprobe->branch.offs;
+
+	if (branch_is_call(auprobe)) {
+		unsigned long new_sp = regs->sp - sizeof_long();
+		/*
+		 * If it fails we execute this (mangled, see the comment in
+		 * branch_clear_offset) insn out-of-line. In the likely case
+		 * this should trigger the trap, and the probed application
+		 * should die or restart the same insn after it handles the
+		 * signal, arch_uprobe_post_xol() won't be even called.
+		 *
+		 * But there is corner case, see the comment in ->post_xol().
+		 */
+		if (copy_to_user((void __user *)new_sp, &new_ip, sizeof_long()))
+			return false;
+		regs->sp = new_sp;
+	} else if (!check_jmp_cond(auprobe, regs)) {
+		offs = 0;
+	}
+
+	regs->ip = new_ip + offs;
+	return true;
 }
 
-#ifdef CONFIG_X86_64
-static bool is_riprel_insn(struct arch_uprobe *auprobe)
+static int branch_post_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
-	return ((auprobe->fixups & (UPROBE_FIX_RIP_AX | UPROBE_FIX_RIP_CX)) != 0);
+	BUG_ON(!branch_is_call(auprobe));
+	/*
+	 * We can only get here if branch_emulate_op() failed to push the ret
+	 * address _and_ another thread expanded our stack before the (mangled)
+	 * "call" insn was executed out-of-line. Just restore ->sp and restart.
+	 * We could also restore ->ip and try to call branch_emulate_op() again.
+	 */
+	regs->sp += sizeof_long();
+	return -ERESTART;
 }
 
-static void
-handle_riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs, long *correction)
+static void branch_clear_offset(struct arch_uprobe *auprobe, struct insn *insn)
 {
-	if (is_riprel_insn(auprobe)) {
-		struct arch_uprobe_task *autask;
+	/*
+	 * Turn this insn into "call 1f; 1:", this is what we will execute
+	 * out-of-line if ->emulate() fails. We only need this to generate
+	 * a trap, so that the probed task receives the correct signal with
+	 * the properly filled siginfo.
+	 *
+	 * But see the comment in ->post_xol(), in the unlikely case it can
+	 * succeed. So we need to ensure that the new ->ip can not fall into
+	 * the non-canonical area and trigger #GP.
+	 *
+	 * We could turn it into (say) "pushf", but then we would need to
+	 * divorce ->insn[] and ->ixol[]. We need to preserve the 1st byte
+	 * of ->insn[] for set_orig_insn().
+	 */
+	memset(auprobe->insn + insn_offset_immediate(insn),
+		0, insn->immediate.nbytes);
+}
 
-		autask = &current->utask->autask;
-		if (auprobe->fixups & UPROBE_FIX_RIP_AX)
-			regs->ax = autask->saved_scratch_register;
-		else
-			regs->cx = autask->saved_scratch_register;
+static struct uprobe_xol_ops branch_xol_ops = {
+	.emulate  = branch_emulate_op,
+	.post_xol = branch_post_xol_op,
+};
+
+/* Returns -ENOSYS if branch_xol_ops doesn't handle this insn */
+static int branch_setup_xol_ops(struct arch_uprobe *auprobe, struct insn *insn)
+{
+	u8 opc1 = OPCODE1(insn);
+
+	/* has the side-effect of processing the entire instruction */
+	insn_get_length(insn);
+	if (WARN_ON_ONCE(!insn_complete(insn)))
+		return -ENOEXEC;
+
+	switch (opc1) {
+	case 0xeb:	/* jmp 8 */
+	case 0xe9:	/* jmp 32 */
+	case 0x90:	/* prefix* + nop; same as jmp with .offs = 0 */
+		break;
+
+	case 0xe8:	/* call relative */
+		branch_clear_offset(auprobe, insn);
+		break;
 
+	case 0x0f:
+		if (insn->opcode.nbytes != 2)
+			return -ENOSYS;
 		/*
-		 * The original instruction includes a displacement, and so
-		 * is 4 bytes longer than what we've just single-stepped.
-		 * Fall through to handle stuff like "jmpq *...(%rip)" and
-		 * "callq *...(%rip)".
+		 * If it is a "near" conditional jmp, OPCODE2() - 0x10 matches
+		 * OPCODE1() of the "short" jmp which checks the same condition.
 		 */
-		if (correction)
-			*correction += 4;
+		opc1 = OPCODE2(insn) - 0x10;
+	default:
+		if (!is_cond_jmp_opcode(opc1))
+			return -ENOSYS;
 	}
+
+	auprobe->branch.opc1 = opc1;
+	auprobe->branch.ilen = insn->length;
+	auprobe->branch.offs = insn->immediate.value;
+
+	auprobe->ops = &branch_xol_ops;
+	return 0;
 }
-#else
-static void
-handle_riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs, long *correction)
+
+/**
+ * arch_uprobe_analyze_insn - instruction analysis including validity and fixups.
+ * @mm: the probed address space.
+ * @arch_uprobe: the probepoint information.
+ * @addr: virtual address at which to install the probepoint
+ * Return 0 on success or a -ve number on error.
+ */
+int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long addr)
+{
+	struct insn insn;
+	bool fix_ip = true, fix_call = false;
+	int ret;
+
+	ret = validate_insn_bits(auprobe, mm, &insn);
+	if (ret)
+		return ret;
+
+	ret = branch_setup_xol_ops(auprobe, &insn);
+	if (ret != -ENOSYS)
+		return ret;
+
+	/*
+	 * Figure out which fixups arch_uprobe_post_xol() will need to perform,
+	 * and annotate arch_uprobe->fixups accordingly. To start with, ->fixups
+	 * is either zero or it reflects rip-related fixups.
+	 */
+	switch (OPCODE1(&insn)) {
+	case 0x9d:		/* popf */
+		auprobe->fixups |= UPROBE_FIX_SETF;
+		break;
+	case 0xc3:		/* ret or lret -- ip is correct */
+	case 0xcb:
+	case 0xc2:
+	case 0xca:
+		fix_ip = false;
+		break;
+	case 0x9a:		/* call absolute - Fix return addr, not ip */
+		fix_call = true;
+		fix_ip = false;
+		break;
+	case 0xea:		/* jmp absolute -- ip is correct */
+		fix_ip = false;
+		break;
+	case 0xff:
+		insn_get_modrm(&insn);
+		switch (MODRM_REG(&insn)) {
+		case 2: case 3:			/* call or lcall, indirect */
+			fix_call = true;
+		case 4: case 5:			/* jmp or ljmp, indirect */
+			fix_ip = false;
+		}
+		/* fall through */
+	default:
+		handle_riprel_insn(auprobe, &insn);
+	}
+
+	if (fix_ip)
+		auprobe->fixups |= UPROBE_FIX_IP;
+	if (fix_call)
+		auprobe->fixups |= UPROBE_FIX_CALL;
+
+	auprobe->ops = &default_xol_ops;
+	return 0;
+}
+
+/*
+ * arch_uprobe_pre_xol - prepare to execute out of line.
+ * @auprobe: the probepoint information.
+ * @regs: reflects the saved user state of current task.
+ */
+int arch_uprobe_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
-	/* No RIP-relative addressing on 32-bit */
+	struct uprobe_task *utask = current->utask;
+
+	regs->ip = utask->xol_vaddr;
+	utask->autask.saved_trap_nr = current->thread.trap_nr;
+	current->thread.trap_nr = UPROBE_TRAP_NR;
+
+	utask->autask.saved_tf = !!(regs->flags & X86_EFLAGS_TF);
+	regs->flags |= X86_EFLAGS_TF;
+	if (test_tsk_thread_flag(current, TIF_BLOCKSTEP))
+		set_task_blockstep(current, false);
+
+	if (auprobe->ops->pre_xol)
+		return auprobe->ops->pre_xol(auprobe, regs);
+	return 0;
 }
-#endif
 
 /*
  * If xol insn itself traps and generates a signal(Say,
@@ -592,22 +752,25 @@ bool arch_uprobe_xol_was_trapped(struct task_struct *t)
  */
 int arch_uprobe_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
-	struct uprobe_task *utask;
-	long correction;
-	int result = 0;
+	struct uprobe_task *utask = current->utask;
 
 	WARN_ON_ONCE(current->thread.trap_nr != UPROBE_TRAP_NR);
 
-	utask = current->utask;
-	current->thread.trap_nr = utask->autask.saved_trap_nr;
-	correction = (long)(utask->vaddr - utask->xol_vaddr);
-	handle_riprel_post_xol(auprobe, regs, &correction);
-	if (auprobe->fixups & UPROBE_FIX_IP)
-		regs->ip += correction;
-
-	if (auprobe->fixups & UPROBE_FIX_CALL)
-		result = adjust_ret_addr(regs->sp, correction);
+	if (auprobe->ops->post_xol) {
+		int err = auprobe->ops->post_xol(auprobe, regs);
+		if (err) {
+			arch_uprobe_abort_xol(auprobe, regs);
+			/*
+			 * Restart the probed insn. ->post_xol() must ensure
+			 * this is really possible if it returns -ERESTART.
+			 */
+			if (err == -ERESTART)
+				return 0;
+			return err;
+		}
+	}
 
+	current->thread.trap_nr = utask->autask.saved_trap_nr;
 	/*
 	 * arch_uprobe_pre_xol() doesn't save the state of TIF_BLOCKSTEP
 	 * so we can get an extra SIGTRAP if we do not clear TF. We need
@@ -618,7 +781,7 @@ int arch_uprobe_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 	else if (!(auprobe->fixups & UPROBE_FIX_SETF))
 		regs->flags &= ~X86_EFLAGS_TF;
 
-	return result;
+	return 0;
 }
 
 /* callback routine for handling exceptions. */
@@ -652,8 +815,9 @@ int arch_uprobe_exception_notify(struct notifier_block *self, unsigned long val,
 
 /*
  * This function gets called when XOL instruction either gets trapped or
- * the thread has a fatal signal, so reset the instruction pointer to its
- * probed address.
+ * the thread has a fatal signal, or if arch_uprobe_post_xol() failed.
+ * Reset the instruction pointer to its probed address for the potential
+ * restart or for post mortem analysis.
  */
 void arch_uprobe_abort_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
@@ -668,25 +832,10 @@ void arch_uprobe_abort_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
 		regs->flags &= ~X86_EFLAGS_TF;
 }
 
-/*
- * Skip these instructions as per the currently known x86 ISA.
- * rep=0x66*; nop=0x90
- */
 static bool __skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs)
 {
-	int i;
-
-	for (i = 0; i < MAX_UINSN_BYTES; i++) {
-		if (auprobe->insn[i] == 0x66)
-			continue;
-
-		if (auprobe->insn[i] == 0x90) {
-			regs->ip += i + 1;
-			return true;
-		}
-
-		break;
-	}
+	if (auprobe->ops->emulate)
+		return auprobe->ops->emulate(auprobe, regs);
 	return false;
 }
 
@@ -701,23 +850,21 @@ bool arch_uprobe_skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs)
 unsigned long
 arch_uretprobe_hijack_return_addr(unsigned long trampoline_vaddr, struct pt_regs *regs)
 {
-	int rasize, ncopied;
+	int rasize = sizeof_long(), nleft;
 	unsigned long orig_ret_vaddr = 0; /* clear high bits for 32-bit apps */
 
-	rasize = is_ia32_task() ? 4 : 8;
-	ncopied = copy_from_user(&orig_ret_vaddr, (void __user *)regs->sp, rasize);
-	if (unlikely(ncopied))
+	if (copy_from_user(&orig_ret_vaddr, (void __user *)regs->sp, rasize))
 		return -1;
 
 	/* check whether address has been already hijacked */
 	if (orig_ret_vaddr == trampoline_vaddr)
 		return orig_ret_vaddr;
 
-	ncopied = copy_to_user((void __user *)regs->sp, &trampoline_vaddr, rasize);
-	if (likely(!ncopied))
+	nleft = copy_to_user((void __user *)regs->sp, &trampoline_vaddr, rasize);
+	if (likely(!nleft))
 		return orig_ret_vaddr;
 
-	if (ncopied != rasize) {
+	if (nleft != rasize) {
 		pr_err("uprobe: return address clobbered: pid=%d, %%sp=%#lx, "
 			"%%ip=%#lx\n", current->pid, regs->sp, regs->ip);
 
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 04709b66369..d1edc5e6fd0 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -60,8 +60,6 @@ static struct percpu_rw_semaphore dup_mmap_sem;
 
 /* Have a copy of original instruction */
 #define UPROBE_COPY_INSN	0
-/* Can skip singlestep */
-#define UPROBE_SKIP_SSTEP	1
 
 struct uprobe {
 	struct rb_node		rb_node;	/* node in the rb tree */
@@ -491,12 +489,9 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset)
 	uprobe->offset = offset;
 	init_rwsem(&uprobe->register_rwsem);
 	init_rwsem(&uprobe->consumer_rwsem);
-	/* For now assume that the instruction need not be single-stepped */
-	__set_bit(UPROBE_SKIP_SSTEP, &uprobe->flags);
 
 	/* add to uprobes_tree, sorted on inode:offset */
 	cur_uprobe = insert_uprobe(uprobe);
-
 	/* a uprobe exists for this inode:offset combination */
 	if (cur_uprobe) {
 		kfree(uprobe);
@@ -1628,20 +1623,6 @@ bool uprobe_deny_signal(void)
 	return true;
 }
 
-/*
- * Avoid singlestepping the original instruction if the original instruction
- * is a NOP or can be emulated.
- */
-static bool can_skip_sstep(struct uprobe *uprobe, struct pt_regs *regs)
-{
-	if (test_bit(UPROBE_SKIP_SSTEP, &uprobe->flags)) {
-		if (arch_uprobe_skip_sstep(&uprobe->arch, regs))
-			return true;
-		clear_bit(UPROBE_SKIP_SSTEP, &uprobe->flags);
-	}
-	return false;
-}
-
 static void mmf_recalc_uprobes(struct mm_struct *mm)
 {
 	struct vm_area_struct *vma;
@@ -1868,13 +1849,13 @@ static void handle_swbp(struct pt_regs *regs)
 
 	handler_chain(uprobe, regs);
 
-	if (can_skip_sstep(uprobe, regs))
+	if (arch_uprobe_skip_sstep(&uprobe->arch, regs))
 		goto out;
 
 	if (!pre_ssout(uprobe, regs, bp_vaddr))
 		return;
 
-	/* can_skip_sstep() succeeded, or restart if can't singlestep */
+	/* arch_uprobe_skip_sstep() succeeded, or restart if can't singlestep */
 out:
 	put_uprobe(uprobe);
 }
@@ -1886,10 +1867,11 @@ out:
 static void handle_singlestep(struct uprobe_task *utask, struct pt_regs *regs)
 {
 	struct uprobe *uprobe;
+	int err = 0;
 
 	uprobe = utask->active_uprobe;
 	if (utask->state == UTASK_SSTEP_ACK)
-		arch_uprobe_post_xol(&uprobe->arch, regs);
+		err = arch_uprobe_post_xol(&uprobe->arch, regs);
 	else if (utask->state == UTASK_SSTEP_TRAPPED)
 		arch_uprobe_abort_xol(&uprobe->arch, regs);
 	else
@@ -1903,6 +1885,11 @@ static void handle_singlestep(struct uprobe_task *utask, struct pt_regs *regs)
 	spin_lock_irq(&current->sighand->siglock);
 	recalc_sigpending(); /* see uprobe_deny_signal() */
 	spin_unlock_irq(&current->sighand->siglock);
+
+	if (unlikely(err)) {
+		uprobe_warn(current, "execute the probed insn, sending SIGILL.");
+		force_sig_info(SIGILL, SEND_SIG_FORCED, current);
+	}
 }
 
 /*
diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt
index fdfceee0ffd..fbfa1192923 100644
--- a/tools/perf/Documentation/perf-diff.txt
+++ b/tools/perf/Documentation/perf-diff.txt
@@ -33,17 +33,20 @@ OPTIONS
 -d::
 --dsos=::
 	Only consider symbols in these dsos. CSV that understands
-	file://filename entries.
+	file://filename entries.  This option will affect the percentage
+	of the Baseline/Delta column.  See --percentage for more info.
 
 -C::
 --comms=::
 	Only consider symbols in these comms. CSV that understands
-	file://filename entries.
+	file://filename entries.  This option will affect the percentage
+	of the Baseline/Delta column.  See --percentage for more info.
 
 -S::
 --symbols=::
 	Only consider these symbols. CSV that understands
-	file://filename entries.
+	file://filename entries.  This option will affect the percentage
+	of the Baseline/Delta column.  See --percentage for more info.
 
 -s::
 --sort=::
@@ -89,6 +92,14 @@ OPTIONS
 --order::
        Specify compute sorting column number.
 
+--percentage::
+	Determine how to display the overhead percentage of filtered entries.
+	Filters can be applied by --comms, --dsos and/or --symbols options.
+
+	"relative" means it's relative to filtered entries only so that the
+	sum of shown entries will be always 100%.  "absolute" means it retains
+	the original value before and after the filter is applied.
+
 COMPARISON
 ----------
 The comparison is governed by the baseline file. The baseline perf.data
@@ -157,6 +168,10 @@ with:
   - period_percent being the % of the hist entry period value within
     single data file
 
+  - with filtering by -C, -d and/or -S, period_percent might be changed
+    relative to how entries are filtered.  Use --percentage=absolute to
+    prevent such fluctuation.
+
 ratio
 ~~~~~
 If specified the 'Ratio' column is displayed with value 'r' computed as:
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 8eab8a4bdeb..09af6629856 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -25,10 +25,6 @@ OPTIONS
 --verbose::
         Be more verbose. (show symbol address, etc)
 
--d::
---dsos=::
-	Only consider symbols in these dsos. CSV that understands
-	file://filename entries.
 -n::
 --show-nr-samples::
 	Show the number of samples for each symbol
@@ -42,11 +38,18 @@ OPTIONS
 -c::
 --comms=::
 	Only consider symbols in these comms. CSV that understands
-	file://filename entries.
+	file://filename entries.  This option will affect the percentage of
+	the overhead column.  See --percentage for more info.
+-d::
+--dsos=::
+	Only consider symbols in these dsos. CSV that understands
+	file://filename entries.  This option will affect the percentage of
+	the overhead column.  See --percentage for more info.
 -S::
 --symbols=::
 	Only consider these symbols. CSV that understands
-	file://filename entries.
+	file://filename entries.  This option will affect the percentage of
+	the overhead column.  See --percentage for more info.
 
 --symbol-filter=::
 	Only show symbols that match (partially) with this filter.
@@ -237,6 +240,15 @@ OPTIONS
 	Do not show entries which have an overhead under that percent.
 	(Default: 0).
 
+--percentage::
+	Determine how to display the overhead percentage of filtered entries.
+	Filters can be applied by --comms, --dsos and/or --symbols options and
+	Zoom operations on the TUI (thread, dso, etc).
+
+	"relative" means it's relative to filtered entries only so that the
+	sum of shown entries will be always 100%.  "absolute" means it retains
+	the original value before and after the filter is applied.
+
 --header::
 	Show header information in the perf.data file.  This includes
 	various information like hostname, OS and perf version, cpu/mem
diff --git a/tools/perf/Documentation/perf-top.txt b/tools/perf/Documentation/perf-top.txt
index 976b00c6cdb..64ed79c4363 100644
--- a/tools/perf/Documentation/perf-top.txt
+++ b/tools/perf/Documentation/perf-top.txt
@@ -123,13 +123,16 @@ Default is to monitor all CPUS.
 	Show a column with the sum of periods.
 
 --dsos::
-	Only consider symbols in these dsos.
+	Only consider symbols in these dsos.  This option will affect the
+	percentage of the overhead column.  See --percentage for more info.
 
 --comms::
-	Only consider symbols in these comms.
+	Only consider symbols in these comms.  This option will affect the
+	percentage of the overhead column.  See --percentage for more info.
 
 --symbols::
-	Only consider these symbols.
+	Only consider these symbols.  This option will affect the
+	percentage of the overhead column.  See --percentage for more info.
 
 -M::
 --disassembler-style=:: Set disassembler style for objdump.
@@ -165,6 +168,15 @@ Default is to monitor all CPUS.
 	Do not show entries which have an overhead under that percent.
 	(Default: 0).
 
+--percentage::
+	Determine how to display the overhead percentage of filtered entries.
+	Filters can be applied by --comms, --dsos and/or --symbols options and
+	Zoom operations on the TUI (thread, dso, etc).
+
+	"relative" means it's relative to filtered entries only so that the
+	sum of shown entries will be always 100%. "absolute" means it retains
+	the original value before and after the filter is applied.
+
 INTERACTIVE PROMPTING KEYS
 --------------------------
 
diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c
index 204fffe2253..6ef80f22c1e 100644
--- a/tools/perf/builtin-diff.c
+++ b/tools/perf/builtin-diff.c
@@ -220,7 +220,8 @@ static int setup_compute(const struct option *opt, const char *str,
 
 static double period_percent(struct hist_entry *he, u64 period)
 {
-	u64 total = he->hists->stats.total_period;
+	u64 total = hists__total_period(he->hists);
+
 	return (period * 100.0) / total;
 }
 
@@ -259,11 +260,18 @@ static s64 compute_wdiff(struct hist_entry *he, struct hist_entry *pair)
 static int formula_delta(struct hist_entry *he, struct hist_entry *pair,
 			 char *buf, size_t size)
 {
+	u64 he_total = he->hists->stats.total_period;
+	u64 pair_total = pair->hists->stats.total_period;
+
+	if (symbol_conf.filter_relative) {
+		he_total = he->hists->stats.total_non_filtered_period;
+		pair_total = pair->hists->stats.total_non_filtered_period;
+	}
 	return scnprintf(buf, size,
 			 "(%" PRIu64 " * 100 / %" PRIu64 ") - "
 			 "(%" PRIu64 " * 100 / %" PRIu64 ")",
-			  pair->stat.period, pair->hists->stats.total_period,
-			  he->stat.period, he->hists->stats.total_period);
+			 pair->stat.period, pair_total,
+			 he->stat.period, he_total);
 }
 
 static int formula_ratio(struct hist_entry *he, struct hist_entry *pair,
@@ -327,15 +335,16 @@ static int diff__process_sample_event(struct perf_tool *tool __maybe_unused,
 		return -1;
 	}
 
-	if (al.filtered)
-		return 0;
-
 	if (hists__add_entry(&evsel->hists, &al, sample->period,
 			     sample->weight, sample->transaction)) {
 		pr_warning("problem incrementing symbol period, skipping event\n");
 		return -1;
 	}
 
+	if (al.filtered == 0) {
+		evsel->hists.stats.total_non_filtered_period += sample->period;
+		evsel->hists.nr_non_filtered_entries++;
+	}
 	evsel->hists.stats.total_period += sample->period;
 	return 0;
 }
@@ -565,7 +574,9 @@ static void hists__compute_resort(struct hists *hists)
 	next = rb_first(root);
 
 	hists->nr_entries = 0;
+	hists->nr_non_filtered_entries = 0;
 	hists->stats.total_period = 0;
+	hists->stats.total_non_filtered_period = 0;
 	hists__reset_col_len(hists);
 
 	while (next != NULL) {
@@ -732,13 +743,16 @@ static const struct option options[] = {
 	OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory",
 		    "Look for files with symbols relative to this directory"),
 	OPT_UINTEGER('o', "order", &sort_compute, "Specify compute sorting."),
+	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
+		     "How to display percentage of filtered entries", parse_filter_percentage),
 	OPT_END()
 };
 
 static double baseline_percent(struct hist_entry *he)
 {
-	struct hists *hists = he->hists;
-	return 100.0 * he->stat.period / hists->stats.total_period;
+	u64 total = hists__total_period(he->hists);
+
+	return 100.0 * he->stat.period / total;
 }
 
 static int hpp__color_baseline(struct perf_hpp_fmt *fmt,
@@ -1120,6 +1134,8 @@ static int data_init(int argc, const char **argv)
 
 int cmd_diff(int argc, const char **argv, const char *prefix __maybe_unused)
 {
+	perf_config(perf_default_config, NULL);
+
 	sort_order = diff__default_sort_order;
 	argc = parse_options(argc, argv, options, diff_usage, 0);
 
diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c
index 929462aa494..bd91de07d2a 100644
--- a/tools/perf/builtin-kmem.c
+++ b/tools/perf/builtin-kmem.c
@@ -756,11 +756,13 @@ int cmd_kmem(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_BOOLEAN(0, "raw-ip", &raw_ip, "show raw ip instead of symbol"),
 	OPT_END()
 	};
-	const char * const kmem_usage[] = {
-		"perf kmem [<options>] {record|stat}",
+	const char *const kmem_subcommands[] = { "record", "stat", NULL };
+	const char *kmem_usage[] = {
+		NULL,
 		NULL
 	};
-	argc = parse_options(argc, argv, kmem_options, kmem_usage, 0);
+	argc = parse_options_subcommand(argc, argv, kmem_options,
+					kmem_subcommands, kmem_usage, 0);
 
 	if (!argc)
 		usage_with_options(kmem_usage, kmem_options);
diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c
index c852c7a85d3..6148afc995c 100644
--- a/tools/perf/builtin-lock.c
+++ b/tools/perf/builtin-lock.c
@@ -961,8 +961,10 @@ int cmd_lock(int argc, const char **argv, const char *prefix __maybe_unused)
 		"perf lock info [<options>]",
 		NULL
 	};
-	const char * const lock_usage[] = {
-		"perf lock [<options>] {record|report|script|info}",
+	const char *const lock_subcommands[] = { "record", "report", "script",
+						 "info", NULL };
+	const char *lock_usage[] = {
+		NULL,
 		NULL
 	};
 	const char * const report_usage[] = {
@@ -976,8 +978,8 @@ int cmd_lock(int argc, const char **argv, const char *prefix __maybe_unused)
 	for (i = 0; i < LOCKHASH_SIZE; i++)
 		INIT_LIST_HEAD(lockhash_table + i);
 
-	argc = parse_options(argc, argv, lock_options, lock_usage,
-			     PARSE_OPT_STOP_AT_NON_OPTION);
+	argc = parse_options_subcommand(argc, argv, lock_options, lock_subcommands,
+					lock_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 	if (!argc)
 		usage_with_options(lock_usage, lock_options);
 
diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c
index 2e3ade69a58..4a1a6c94a5e 100644
--- a/tools/perf/builtin-mem.c
+++ b/tools/perf/builtin-mem.c
@@ -21,11 +21,6 @@ struct perf_mem {
 	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
 };
 
-static const char * const mem_usage[] = {
-	"perf mem [<options>] {record <command> |report}",
-	NULL
-};
-
 static int __cmd_record(int argc, const char **argv)
 {
 	int rec_argc, i = 0, j;
@@ -220,9 +215,15 @@ int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
 		   " between columns '.' is reserved."),
 	OPT_END()
 	};
+	const char *const mem_subcommands[] = { "record", "report", NULL };
+	const char *mem_usage[] = {
+		NULL,
+		NULL
+	};
+
 
-	argc = parse_options(argc, argv, mem_options, mem_usage,
-			     PARSE_OPT_STOP_AT_NON_OPTION);
+	argc = parse_options_subcommand(argc, argv, mem_options, mem_subcommands,
+					mem_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 
 	if (!argc || !(strncmp(argv[0], "rec", 3) || mem_operation))
 		usage_with_options(mem_usage, mem_options);
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index c8f21137dfd..af8cb7a2c9b 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -123,6 +123,8 @@ static int report__add_mem_hist_entry(struct report *rep, struct addr_location *
 
 	evsel->hists.stats.total_period += cost;
 	hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
+	if (!he->filtered)
+		evsel->hists.stats.nr_non_filtered_samples++;
 	err = hist_entry__append_callchain(he, sample);
 out:
 	return err;
@@ -176,6 +178,8 @@ static int report__add_branch_hist_entry(struct report *rep, struct addr_locatio
 
 			evsel->hists.stats.total_period += 1;
 			hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
+			if (!he->filtered)
+				evsel->hists.stats.nr_non_filtered_samples++;
 		} else
 			goto out;
 	}
@@ -209,6 +213,8 @@ static int report__add_hist_entry(struct report *rep, struct perf_evsel *evsel,
 		err = hist_entry__inc_addr_samples(he, evsel->idx, al->addr);
 
 	evsel->hists.stats.total_period += sample->period;
+	if (!he->filtered)
+		evsel->hists.stats.nr_non_filtered_samples++;
 	hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
 out:
 	return err;
@@ -337,6 +343,11 @@ static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report
 	char buf[512];
 	size_t size = sizeof(buf);
 
+	if (symbol_conf.filter_relative) {
+		nr_samples = hists->stats.nr_non_filtered_samples;
+		nr_events = hists->stats.total_non_filtered_period;
+	}
+
 	if (perf_evsel__is_group_event(evsel)) {
 		struct perf_evsel *pos;
 
@@ -344,8 +355,13 @@ static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report
 		evname = buf;
 
 		for_each_group_member(pos, evsel) {
-			nr_samples += pos->hists.stats.nr_events[PERF_RECORD_SAMPLE];
-			nr_events += pos->hists.stats.total_period;
+			if (symbol_conf.filter_relative) {
+				nr_samples += pos->hists.stats.nr_non_filtered_samples;
+				nr_events += pos->hists.stats.total_non_filtered_period;
+			} else {
+				nr_samples += pos->hists.stats.nr_events[PERF_RECORD_SAMPLE];
+				nr_events += pos->hists.stats.total_period;
+			}
 		}
 	}
 
@@ -823,6 +839,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
 	OPT_CALLBACK(0, "percent-limit", &report, "percent",
 		     "Don't show entries under that percent", parse_percent_limit),
+	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
+		     "how to display percentage of filtered entries", parse_filter_percentage),
 	OPT_END()
 	};
 	struct perf_data_file file = {
diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c
index 9ac0a495c95..d3fb0ed7240 100644
--- a/tools/perf/builtin-sched.c
+++ b/tools/perf/builtin-sched.c
@@ -1713,8 +1713,10 @@ int cmd_sched(int argc, const char **argv, const char *prefix __maybe_unused)
 		"perf sched replay [<options>]",
 		NULL
 	};
-	const char * const sched_usage[] = {
-		"perf sched [<options>] {record|latency|map|replay|script}",
+	const char *const sched_subcommands[] = { "record", "latency", "map",
+						  "replay", "script", NULL };
+	const char *sched_usage[] = {
+		NULL,
 		NULL
 	};
 	struct trace_sched_handler lat_ops  = {
@@ -1736,8 +1738,8 @@ int cmd_sched(int argc, const char **argv, const char *prefix __maybe_unused)
 	for (i = 0; i < ARRAY_SIZE(sched.curr_pid); i++)
 		sched.curr_pid[i] = -1;
 
-	argc = parse_options(argc, argv, sched_options, sched_usage,
-			     PARSE_OPT_STOP_AT_NON_OPTION);
+	argc = parse_options_subcommand(argc, argv, sched_options, sched_subcommands,
+					sched_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 	if (!argc)
 		usage_with_options(sched_usage, sched_options);
 
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 65aaa5bbf7e..37d30460bad 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -253,6 +253,9 @@ static struct hist_entry *perf_evsel__add_hist_entry(struct perf_evsel *evsel,
 		return NULL;
 
 	hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
+	if (!he->filtered)
+		evsel->hists.stats.nr_non_filtered_samples++;
+
 	return he;
 }
 
@@ -694,8 +697,7 @@ static void perf_event__process_sample(struct perf_tool *tool,
 	if (event->header.misc & PERF_RECORD_MISC_EXACT_IP)
 		top->exact_samples++;
 
-	if (perf_event__preprocess_sample(event, machine, &al, sample) < 0 ||
-	    al.filtered)
+	if (perf_event__preprocess_sample(event, machine, &al, sample) < 0)
 		return;
 
 	if (!top->kptr_restrict_warned &&
@@ -1116,6 +1118,8 @@ int cmd_top(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_STRING('u', "uid", &target->uid_str, "user", "user to profile"),
 	OPT_CALLBACK(0, "percent-limit", &top, "percent",
 		     "Don't show entries under that percent", parse_percent_limit),
+	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
+		     "How to display percentage of filtered entries", parse_filter_percentage),
 	OPT_END()
 	};
 	const char * const top_usage[] = {
diff --git a/tools/perf/perf-completion.sh b/tools/perf/perf-completion.sh
index ae3a57694b6..33569847fdc 100644
--- a/tools/perf/perf-completion.sh
+++ b/tools/perf/perf-completion.sh
@@ -121,8 +121,8 @@ __perf_main ()
 	elif [[ $prev == "-e" && "${words[1]}" == @(record|stat|top) ]]; then
 		evts=$($cmd list --raw-dump)
 		__perfcomp_colon "$evts" "$cur"
-	# List subcommands for 'perf kvm'
-	elif [[ $prev == "kvm" ]]; then
+	# List subcommands for perf commands
+	elif [[ $prev == @(kvm|kmem|mem|lock|sched) ]]; then
 		subcmds=$($cmd $prev --list-cmds)
 		__perfcomp_colon "$subcmds" "$cur"
 	# List long option names
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 7ec871af3f6..4d416984c59 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -769,12 +769,15 @@ static unsigned int hist_browser__refresh(struct ui_browser *browser)
 
 	for (nd = browser->top; nd; nd = rb_next(nd)) {
 		struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
-		float percent = h->stat.period * 100.0 /
-					hb->hists->stats.total_period;
+		u64 total = hists__total_period(h->hists);
+		float percent = 0.0;
 
 		if (h->filtered)
 			continue;
 
+		if (total)
+			percent = h->stat.period * 100.0 / total;
+
 		if (percent < hb->min_pcnt)
 			continue;
 
@@ -792,8 +795,11 @@ static struct rb_node *hists__filter_entries(struct rb_node *nd,
 {
 	while (nd != NULL) {
 		struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
-		float percent = h->stat.period * 100.0 /
-					hists->stats.total_period;
+		u64 total = hists__total_period(hists);
+		float percent = 0.0;
+
+		if (total)
+			percent = h->stat.period * 100.0 / total;
 
 		if (percent < min_pcnt)
 			return NULL;
@@ -813,8 +819,11 @@ static struct rb_node *hists__filter_prev_entries(struct rb_node *nd,
 {
 	while (nd != NULL) {
 		struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
-		float percent = h->stat.period * 100.0 /
-					hists->stats.total_period;
+		u64 total = hists__total_period(hists);
+		float percent = 0.0;
+
+		if (total)
+			percent = h->stat.period * 100.0 / total;
 
 		if (!h->filtered && percent >= min_pcnt)
 			return nd;
@@ -1189,6 +1198,11 @@ static int hists__browser_title(struct hists *hists, char *bf, size_t size,
 	char buf[512];
 	size_t buflen = sizeof(buf);
 
+	if (symbol_conf.filter_relative) {
+		nr_samples = hists->stats.nr_non_filtered_samples;
+		nr_events = hists->stats.total_non_filtered_period;
+	}
+
 	if (perf_evsel__is_group_event(evsel)) {
 		struct perf_evsel *pos;
 
@@ -1196,8 +1210,13 @@ static int hists__browser_title(struct hists *hists, char *bf, size_t size,
 		ev_name = buf;
 
 		for_each_group_member(pos, evsel) {
-			nr_samples += pos->hists.stats.nr_events[PERF_RECORD_SAMPLE];
-			nr_events += pos->hists.stats.total_period;
+			if (symbol_conf.filter_relative) {
+				nr_samples += pos->hists.stats.nr_non_filtered_samples;
+				nr_events += pos->hists.stats.total_non_filtered_period;
+			} else {
+				nr_samples += pos->hists.stats.nr_events[PERF_RECORD_SAMPLE];
+				nr_events += pos->hists.stats.total_period;
+			}
 		}
 	}
 
@@ -1370,6 +1389,7 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 	"C             Collapse all callchains\n"			\
 	"d             Zoom into current DSO\n"				\
 	"E             Expand all callchains\n"				\
+	"F             Toggle percentage of filtered entries\n"		\
 
 	/* help messages are sorted by lexical order of the hotkey */
 	const char report_help[] = HIST_BROWSER_HELP_COMMON
@@ -1475,6 +1495,9 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 			if (env->arch)
 				tui__header_window(env);
 			continue;
+		case 'F':
+			symbol_conf.filter_relative ^= 1;
+			continue;
 		case K_F1:
 		case 'h':
 		case '?':
diff --git a/tools/perf/ui/gtk/hists.c b/tools/perf/ui/gtk/hists.c
index e395ef9b0ae..91f10f3f6dd 100644
--- a/tools/perf/ui/gtk/hists.c
+++ b/tools/perf/ui/gtk/hists.c
@@ -228,12 +228,15 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 	for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
 		struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
 		GtkTreeIter iter;
-		float percent = h->stat.period * 100.0 /
-					hists->stats.total_period;
+		u64 total = hists__total_period(h->hists);
+		float percent = 0.0;
 
 		if (h->filtered)
 			continue;
 
+		if (total)
+			percent = h->stat.period * 100.0 / total;
+
 		if (percent < min_pcnt)
 			continue;
 
@@ -261,12 +264,8 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 		}
 
 		if (symbol_conf.use_callchain && sort__has_sym) {
-			u64 total;
-
 			if (callchain_param.mode == CHAIN_GRAPH_REL)
 				total = h->stat.period;
-			else
-				total = hists->stats.total_period;
 
 			perf_gtk__add_callchain(&h->sorted_chain, store, &iter,
 						sym_col, total);
diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c
index 0f403b83e9d..0912805c08f 100644
--- a/tools/perf/ui/hist.c
+++ b/tools/perf/ui/hist.c
@@ -32,10 +32,10 @@ int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 
 	if (fmt_percent) {
 		double percent = 0.0;
+		u64 total = hists__total_period(hists);
 
-		if (hists->stats.total_period)
-			percent = 100.0 * get_field(he) /
-				  hists->stats.total_period;
+		if (total)
+			percent = 100.0 * get_field(he) / total;
 
 		ret += hpp__call_print_fn(hpp, print_fn, fmt, percent);
 	} else
@@ -50,7 +50,7 @@ int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 
 		list_for_each_entry(pair, &he->pairs.head, pairs.node) {
 			u64 period = get_field(pair);
-			u64 total = pair->hists->stats.total_period;
+			u64 total = hists__total_period(pair->hists);
 
 			if (!total)
 				continue;
diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c
index 3e0fdd369cc..24519e14ac5 100644
--- a/tools/perf/util/config.c
+++ b/tools/perf/util/config.c
@@ -11,6 +11,7 @@
 #include "util.h"
 #include "cache.h"
 #include "exec_cmd.h"
+#include "util/hist.h"  /* perf_hist_config */
 
 #define MAXNAME (256)
 
@@ -355,6 +356,9 @@ int perf_default_config(const char *var, const char *value,
 	if (!prefixcmp(var, "core."))
 		return perf_default_core_config(var, value);
 
+	if (!prefixcmp(var, "hist."))
+		return perf_hist_config(var, value);
+
 	/* Add other config variables here. */
 	return 0;
 }
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index f38590d7561..5a892477aa5 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -321,9 +321,11 @@ void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h)
 {
 	if (!h->filtered) {
 		hists__calc_col_len(hists, h);
-		++hists->nr_entries;
-		hists->stats.total_period += h->stat.period;
+		hists->nr_non_filtered_entries++;
+		hists->stats.total_non_filtered_period += h->stat.period;
 	}
+	hists->nr_entries++;
+	hists->stats.total_period += h->stat.period;
 }
 
 static u8 symbol__parent_filter(const struct symbol *parent)
@@ -674,8 +676,9 @@ void hists__output_resort(struct hists *hists)
 	next = rb_first(root);
 	hists->entries = RB_ROOT;
 
-	hists->nr_entries = 0;
+	hists->nr_non_filtered_entries = 0;
 	hists->stats.total_period = 0;
+	hists->stats.total_non_filtered_period = 0;
 	hists__reset_col_len(hists);
 
 	while (next) {
@@ -694,12 +697,12 @@ static void hists__remove_entry_filter(struct hists *hists, struct hist_entry *h
 	if (h->filtered)
 		return;
 
-	++hists->nr_entries;
+	++hists->nr_non_filtered_entries;
 	if (h->ms.unfolded)
-		hists->nr_entries += h->nr_rows;
+		hists->nr_non_filtered_entries += h->nr_rows;
 	h->row_offset = 0;
-	hists->stats.total_period += h->stat.period;
-	hists->stats.nr_events[PERF_RECORD_SAMPLE] += h->stat.nr_events;
+	hists->stats.total_non_filtered_period += h->stat.period;
+	hists->stats.nr_non_filtered_samples += h->stat.nr_events;
 
 	hists__calc_col_len(hists, h);
 }
@@ -721,8 +724,9 @@ void hists__filter_by_dso(struct hists *hists)
 {
 	struct rb_node *nd;
 
-	hists->nr_entries = hists->stats.total_period = 0;
-	hists->stats.nr_events[PERF_RECORD_SAMPLE] = 0;
+	hists->nr_non_filtered_entries = 0;
+	hists->stats.total_non_filtered_period = 0;
+	hists->stats.nr_non_filtered_samples = 0;
 	hists__reset_col_len(hists);
 
 	for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
@@ -754,8 +758,9 @@ void hists__filter_by_thread(struct hists *hists)
 {
 	struct rb_node *nd;
 
-	hists->nr_entries = hists->stats.total_period = 0;
-	hists->stats.nr_events[PERF_RECORD_SAMPLE] = 0;
+	hists->nr_non_filtered_entries = 0;
+	hists->stats.total_non_filtered_period = 0;
+	hists->stats.nr_non_filtered_samples = 0;
 	hists__reset_col_len(hists);
 
 	for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
@@ -785,8 +790,9 @@ void hists__filter_by_symbol(struct hists *hists)
 {
 	struct rb_node *nd;
 
-	hists->nr_entries = hists->stats.total_period = 0;
-	hists->stats.nr_events[PERF_RECORD_SAMPLE] = 0;
+	hists->nr_non_filtered_entries = 0;
+	hists->stats.total_non_filtered_period = 0;
+	hists->stats.nr_non_filtered_samples = 0;
 	hists__reset_col_len(hists);
 
 	for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
@@ -931,3 +937,30 @@ int hists__link(struct hists *leader, struct hists *other)
 
 	return 0;
 }
+
+u64 hists__total_period(struct hists *hists)
+{
+	return symbol_conf.filter_relative ? hists->stats.total_non_filtered_period :
+		hists->stats.total_period;
+}
+
+int parse_filter_percentage(const struct option *opt __maybe_unused,
+			    const char *arg, int unset __maybe_unused)
+{
+	if (!strcmp(arg, "relative"))
+		symbol_conf.filter_relative = true;
+	else if (!strcmp(arg, "absolute"))
+		symbol_conf.filter_relative = false;
+	else
+		return -1;
+
+	return 0;
+}
+
+int perf_hist_config(const char *var, const char *value)
+{
+	if (!strcmp(var, "hist.percentage"))
+		return parse_filter_percentage(NULL, value, 0);
+
+	return 0;
+}
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 1f1f513dfe7..5a0343eb22e 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -37,9 +37,11 @@ enum hist_filter {
  */
 struct events_stats {
 	u64 total_period;
+	u64 total_non_filtered_period;
 	u64 total_lost;
 	u64 total_invalid_chains;
 	u32 nr_events[PERF_RECORD_HEADER_MAX];
+	u32 nr_non_filtered_samples;
 	u32 nr_lost_warned;
 	u32 nr_unknown_events;
 	u32 nr_invalid_chains;
@@ -83,6 +85,7 @@ struct hists {
 	struct rb_root		entries;
 	struct rb_root		entries_collapsed;
 	u64			nr_entries;
+	u64			nr_non_filtered_entries;
 	const struct thread	*thread_filter;
 	const struct dso	*dso_filter;
 	const char		*uid_filter_str;
@@ -112,6 +115,7 @@ void hists__collapse_resort(struct hists *hists, struct ui_progress *prog);
 void hists__decay_entries(struct hists *hists, bool zap_user, bool zap_kernel);
 void hists__output_recalc_col_len(struct hists *hists, int max_rows);
 
+u64 hists__total_period(struct hists *hists);
 void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h);
 void hists__inc_nr_events(struct hists *hists, u32 type);
 void events_stats__inc(struct events_stats *stats, u32 type);
@@ -250,4 +254,10 @@ static inline int script_browse(const char *script_opt __maybe_unused)
 #endif
 
 unsigned int hists__sort_list_width(struct hists *hists);
+
+struct option;
+int parse_filter_percentage(const struct option *opt __maybe_unused,
+			    const char *arg, int unset __maybe_unused);
+int perf_hist_config(const char *var, const char *value);
+
 #endif	/* __PERF_HIST_H */
diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h
index 501e4e722e8..ae94e006a52 100644
--- a/tools/perf/util/symbol.h
+++ b/tools/perf/util/symbol.h
@@ -115,7 +115,8 @@ struct symbol_conf {
 			annotate_asm_raw,
 			annotate_src,
 			event_group,
-			demangle;
+			demangle,
+			filter_relative;
 	const char	*vmlinux_name,
 			*kallsyms_name,
 			*source_prefix,