[S390] ftrace: add function graph tracer support

Function graph tracer support for s390.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/Kconfig

@@ -85,6 +85,7 @@ config S390
 	select HAVE_FUNCTION_TRACE_MCOUNT_TEST
 	select HAVE_FTRACE_MCOUNT_RECORD
 	select HAVE_DYNAMIC_FTRACE
+	select HAVE_FUNCTION_GRAPH_TRACER
 	select HAVE_DEFAULT_NO_SPIN_MUTEXES
 	select HAVE_OPROFILE
 	select HAVE_KPROBES

arch/s390/include/asm/ftrace.h

@@ -11,11 +11,13 @@ struct dyn_arch_ftrace { };
 #define MCOUNT_ADDR ((long)_mcount)
 
 #ifdef CONFIG_64BIT
-#define MCOUNT_INSN_SIZE 24
-#define MCOUNT_OFFSET	 14
+#define MCOUNT_OFFSET_RET 18
+#define MCOUNT_INSN_SIZE  24
+#define MCOUNT_OFFSET	  14
 #else
-#define MCOUNT_INSN_SIZE 30
-#define MCOUNT_OFFSET	  8
+#define MCOUNT_OFFSET_RET 26
+#define MCOUNT_INSN_SIZE  30
+#define MCOUNT_OFFSET	   8
 #endif
 
 static inline unsigned long ftrace_call_adjust(unsigned long addr)

arch/s390/kernel/Makefile

@@ -3,11 +3,8 @@
 #
 
 ifdef CONFIG_FUNCTION_TRACER
-# Do not trace early boot code
+# Don't trace early setup code and tracing code
 CFLAGS_REMOVE_early.o = -pg
-endif
-
-ifdef CONFIG_DYNAMIC_FTRACE
 CFLAGS_REMOVE_ftrace.o = -pg
 endif
 
@@ -46,6 +43,7 @@ obj-$(CONFIG_STACKTRACE)	+= stacktrace.o
 obj-$(CONFIG_KPROBES)		+= kprobes.o
 obj-$(CONFIG_FUNCTION_TRACER)	+= mcount.o
 obj-$(CONFIG_DYNAMIC_FTRACE)	+= ftrace.o
+obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o
 
 # Kexec part
 S390_KEXEC_OBJS := machine_kexec.o crash.o

arch/s390/kernel/ftrace.c

@@ -7,13 +7,17 @@
  *
  */
 
+#include <linux/hardirq.h>
 #include <linux/uaccess.h>
 #include <linux/ftrace.h>
 #include <linux/kernel.h>
 #include <linux/types.h>
 #include <asm/lowcore.h>
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+
 void ftrace_disable_code(void);
+void ftrace_disable_return(void);
 void ftrace_call_code(void);
 void ftrace_nop_code(void);
 
@@ -28,6 +32,7 @@ asm(
 	"	.word	0x0024\n"
 	"	lg	%r1,"__stringify(__LC_FTRACE_FUNC)"\n"
 	"	basr	%r14,%r1\n"
+	"ftrace_disable_return:\n"
 	"	lg	%r14,8(15)\n"
 	"	lgr	%r0,%r0\n"
 	"0:\n");
@@ -50,6 +55,7 @@ asm(
 	"	j	0f\n"
 	"	l	%r1,"__stringify(__LC_FTRACE_FUNC)"\n"
 	"	basr	%r14,%r1\n"
+	"ftrace_disable_return:\n"
 	"	l	%r14,4(%r15)\n"
 	"	j	0f\n"
 	"	bcr	0,%r7\n"
@@ -130,3 +136,69 @@ int __init ftrace_dyn_arch_init(void *data)
 	*(unsigned long *)data = 0;
 	return 0;
 }
+
+#endif /* CONFIG_DYNAMIC_FTRACE */
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+#ifdef CONFIG_DYNAMIC_FTRACE
+/*
+ * Patch the kernel code at ftrace_graph_caller location:
+ * The instruction there is branch relative on condition. The condition mask
+ * is either all ones (always branch aka disable ftrace_graph_caller) or all
+ * zeroes (nop aka enable ftrace_graph_caller).
+ * Instruction format for brc is a7m4xxxx where m is the condition mask.
+ */
+int ftrace_enable_ftrace_graph_caller(void)
+{
+	unsigned short opcode = 0xa704;
+
+	return probe_kernel_write(ftrace_graph_caller, &opcode, sizeof(opcode));
+}
+
+int ftrace_disable_ftrace_graph_caller(void)
+{
+	unsigned short opcode = 0xa7f4;
+
+	return probe_kernel_write(ftrace_graph_caller, &opcode, sizeof(opcode));
+}
+
+static inline unsigned long ftrace_mcount_call_adjust(unsigned long addr)
+{
+	return addr - (ftrace_disable_return - ftrace_disable_code);
+}
+
+#else /* CONFIG_DYNAMIC_FTRACE */
+
+static inline unsigned long ftrace_mcount_call_adjust(unsigned long addr)
+{
+	return addr - MCOUNT_OFFSET_RET;
+}
+
+#endif /* CONFIG_DYNAMIC_FTRACE */
+
+/*
+ * Hook the return address and push it in the stack of return addresses
+ * in current thread info.
+ */
+unsigned long prepare_ftrace_return(unsigned long ip, unsigned long parent)
+{
+	struct ftrace_graph_ent trace;
+
+	/* Nmi's are currently unsupported. */
+	if (unlikely(in_nmi()))
+		goto out;
+	if (unlikely(atomic_read(&current->tracing_graph_pause)))
+		goto out;
+	if (ftrace_push_return_trace(parent, ip, &trace.depth) == -EBUSY)
+		goto out;
+	trace.func = ftrace_mcount_call_adjust(ip) & PSW_ADDR_INSN;
+	/* Only trace if the calling function expects to. */
+	if (!ftrace_graph_entry(&trace)) {
+		current->curr_ret_stack--;
+		goto out;
+	}
+	parent = (unsigned long)return_to_handler;
+out:
+	return parent;
+}
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */

arch/s390/kernel/mcount.S

@@ -34,6 +34,18 @@ ftrace_caller:
 	larl	%r14,ftrace_dyn_func
 	lg	%r14,0(%r14)
 	basr	%r14,%r14
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	.globl	ftrace_graph_caller
+ftrace_graph_caller:
+	# This unconditional branch gets runtime patched. Change only if
+	# you know what you are doing. See ftrace_enable_graph_caller().
+	j	0f
+	lg	%r2,272(%r15)
+	lg	%r3,168(%r15)
+	brasl	%r14,prepare_ftrace_return
+	stg	%r2,168(%r15)
+0:
+#endif
 	aghi	%r15,160
 	lmg	%r2,%r5,32(%r15)
 	lg	%r14,112(%r15)
@@ -62,6 +74,12 @@ _mcount:
 	larl	%r14,ftrace_trace_function
 	lg	%r14,0(%r14)
 	basr	%r14,%r14
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	lg	%r2,272(%r15)
+	lg	%r3,168(%r15)
+	brasl	%r14,prepare_ftrace_return
+	stg	%r2,168(%r15)
+#endif
 	aghi	%r15,160
 	lmg	%r2,%r5,32(%r15)
 	lg	%r14,112(%r15)
@@ -69,6 +87,22 @@ _mcount:
 
 #endif /* CONFIG_DYNAMIC_FTRACE */
 
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+
+	.globl	return_to_handler
+return_to_handler:
+	stmg	%r2,%r5,32(%r15)
+	lgr	%r1,%r15
+	aghi	%r15,-160
+	stg	%r1,__SF_BACKCHAIN(%r15)
+	brasl	%r14,ftrace_return_to_handler
+	aghi	%r15,160
+	lgr	%r14,%r2
+	lmg	%r2,%r5,32(%r15)
+	br	%r14
+
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+
 #else /* CONFIG_64BIT */
 
 #ifdef CONFIG_DYNAMIC_FTRACE
@@ -96,6 +130,21 @@ ftrace_caller:
 	l	%r14,0b-0b(%r1)
 	l	%r14,0(%r14)
 	basr	%r14,%r14
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	.globl	ftrace_graph_caller
+ftrace_graph_caller:
+	# This unconditional branch gets runtime patched. Change only if
+	# you know what you are doing. See ftrace_enable_graph_caller().
+	j	1f
+	bras	%r1,0f
+	.long	prepare_ftrace_return
+0:	l	%r2,152(%r15)
+	l	%r4,0(%r1)
+	l	%r3,100(%r15)
+	basr	%r14,%r4
+	st	%r2,100(%r15)
+1:
+#endif
 	ahi	%r15,96
 	l	%r14,56(%r15)
 3:	lm	%r2,%r5,16(%r15)
@@ -128,10 +177,40 @@ _mcount:
 	l	%r14,0b-0b(%r1)
 	l	%r14,0(%r14)
 	basr	%r14,%r14
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	bras	%r1,0f
+	.long	prepare_ftrace_return
+0:	l	%r2,152(%r15)
+	l	%r4,0(%r1)
+	l	%r3,100(%r15)
+	basr	%r14,%r4
+	st	%r2,100(%r15)
+#endif
 	ahi	%r15,96
 	l	%r14,56(%r15)
 3:	lm	%r2,%r5,16(%r15)
 	br	%r14
 
 #endif /* CONFIG_DYNAMIC_FTRACE */
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+
+	.globl	return_to_handler
+return_to_handler:
+	stm	%r2,%r5,16(%r15)
+	st	%r14,56(%r15)
+	lr	%r0,%r15
+	ahi	%r15,-96
+	st	%r0,__SF_BACKCHAIN(%r15)
+	bras	%r1,0f
+	.long	ftrace_return_to_handler
+0:	l	%r2,0b-0b(%r1)
+	basr	%r14,%r2
+	lr	%r14,%r2
+	ahi	%r15,96
+	lm	%r2,%r5,16(%r15)
+	br	%r14
+
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+
 #endif /* CONFIG_64BIT */

arch/s390/kernel/s390_ext.c

@@ -10,6 +10,7 @@
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/slab.h>
+#include <linux/ftrace.h>
 #include <linux/errno.h>
 #include <linux/kernel_stat.h>
 #include <linux/interrupt.h>
@@ -112,7 +113,7 @@ int unregister_early_external_interrupt(__u16 code, ext_int_handler_t handler,
 	return 0;
 }
 
-void do_extint(struct pt_regs *regs, unsigned short code)
+void __irq_entry do_extint(struct pt_regs *regs, unsigned short code)
 {
         ext_int_info_t *p;
         int index;

arch/s390/kernel/time.c

@@ -70,7 +70,7 @@ static DEFINE_PER_CPU(struct clock_event_device, comparators);
 /*
  * Scheduler clock - returns current time in nanosec units.
  */
-unsigned long long sched_clock(void)
+unsigned long long notrace sched_clock(void)
 {
 	return ((get_clock_xt() - sched_clock_base_cc) * 125) >> 9;
 }

arch/s390/kernel/vmlinux.lds.S

@@ -34,6 +34,7 @@ SECTIONS
 		SCHED_TEXT
 		LOCK_TEXT
 		KPROBES_TEXT
+		IRQENTRY_TEXT
 		*(.fixup)
 		*(.gnu.warning)
 	} :text = 0x0700

drivers/s390/cio/cio.c

@@ -12,6 +12,7 @@
 #define KMSG_COMPONENT "cio"
 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
 
+#include <linux/ftrace.h>
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/slab.h>
@@ -626,8 +627,7 @@ out:
  *	    handlers).
  *
  */
-void
-do_IRQ (struct pt_regs *regs)
+void __irq_entry do_IRQ(struct pt_regs *regs)
 {
 	struct tpi_info *tpi_info;
 	struct subchannel *sch;