Merge branch 'tracing/ftrace' into tracing/core

Documentation/ftrace.txt

@@ -165,6 +165,8 @@ Here is the list of current tracers that may be configured.
   nop - This is not a tracer. To remove all tracers from tracing
 		simply echo "nop" into current_tracer.
 
+  hw-branch-tracer - traces branches on all cpu's in a circular buffer.
+
 
 Examples of using the tracer
 ----------------------------
@@ -1152,6 +1154,78 @@ int main (int argc, char **argv)
         return 0;
 }
 
+
+hw-branch-tracer (x86 only)
+---------------------------
+
+This tracer uses the x86 last branch tracing hardware feature to
+collect a branch trace on all cpus with relatively low overhead.
+
+The tracer uses a fixed-size circular buffer per cpu and only
+traces ring 0 branches. The trace file dumps that buffer in the
+following format:
+
+# tracer: hw-branch-tracer
+#
+# CPU#        TO  <-  FROM
+   0  scheduler_tick+0xb5/0x1bf	  <-  task_tick_idle+0x5/0x6
+   2  run_posix_cpu_timers+0x2b/0x72a	  <-  run_posix_cpu_timers+0x25/0x72a
+   0  scheduler_tick+0x139/0x1bf	  <-  scheduler_tick+0xed/0x1bf
+   0  scheduler_tick+0x17c/0x1bf	  <-  scheduler_tick+0x148/0x1bf
+   2  run_posix_cpu_timers+0x9e/0x72a	  <-  run_posix_cpu_timers+0x5e/0x72a
+   0  scheduler_tick+0x1b6/0x1bf	  <-  scheduler_tick+0x1aa/0x1bf
+
+
+The tracer may be used to dump the trace for the oops'ing cpu on a
+kernel oops into the system log. To enable this, ftrace_dump_on_oops
+must be set. To set ftrace_dump_on_oops, one can either use the sysctl
+function or set it via the proc system interface.
+
+  sysctl kernel.ftrace_dump_on_oops=1
+
+or
+
+  echo 1 > /proc/sys/kernel/ftrace_dump_on_oops
+
+
+Here's an example of such a dump after a null pointer dereference in a
+kernel module:
+
+[57848.105921] BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
+[57848.106019] IP: [<ffffffffa0000006>] open+0x6/0x14 [oops]
+[57848.106019] PGD 2354e9067 PUD 2375e7067 PMD 0
+[57848.106019] Oops: 0002 [#1] SMP
+[57848.106019] last sysfs file: /sys/devices/pci0000:00/0000:00:1e.0/0000:20:05.0/local_cpus
+[57848.106019] Dumping ftrace buffer:
+[57848.106019] ---------------------------------
+[...]
+[57848.106019]    0  chrdev_open+0xe6/0x165	  <-  cdev_put+0x23/0x24
+[57848.106019]    0  chrdev_open+0x117/0x165	  <-  chrdev_open+0xfa/0x165
+[57848.106019]    0  chrdev_open+0x120/0x165	  <-  chrdev_open+0x11c/0x165
+[57848.106019]    0  chrdev_open+0x134/0x165	  <-  chrdev_open+0x12b/0x165
+[57848.106019]    0  open+0x0/0x14 [oops]	  <-  chrdev_open+0x144/0x165
+[57848.106019]    0  page_fault+0x0/0x30	  <-  open+0x6/0x14 [oops]
+[57848.106019]    0  error_entry+0x0/0x5b	  <-  page_fault+0x4/0x30
+[57848.106019]    0  error_kernelspace+0x0/0x31	  <-  error_entry+0x59/0x5b
+[57848.106019]    0  error_sti+0x0/0x1	  <-  error_kernelspace+0x2d/0x31
+[57848.106019]    0  page_fault+0x9/0x30	  <-  error_sti+0x0/0x1
+[57848.106019]    0  do_page_fault+0x0/0x881	  <-  page_fault+0x1a/0x30
+[...]
+[57848.106019]    0  do_page_fault+0x66b/0x881	  <-  is_prefetch+0x1ee/0x1f2
+[57848.106019]    0  do_page_fault+0x6e0/0x881	  <-  do_page_fault+0x67a/0x881
+[57848.106019]    0  oops_begin+0x0/0x96	  <-  do_page_fault+0x6e0/0x881
+[57848.106019]    0  trace_hw_branch_oops+0x0/0x2d	  <-  oops_begin+0x9/0x96
+[...]
+[57848.106019]    0  ds_suspend_bts+0x2a/0xe3	  <-  ds_suspend_bts+0x1a/0xe3
+[57848.106019] ---------------------------------
+[57848.106019] CPU 0
+[57848.106019] Modules linked in: oops
+[57848.106019] Pid: 5542, comm: cat Tainted: G        W  2.6.28 #23
+[57848.106019] RIP: 0010:[<ffffffffa0000006>]  [<ffffffffa0000006>] open+0x6/0x14 [oops]
+[57848.106019] RSP: 0018:ffff880235457d48  EFLAGS: 00010246
+[...]
+
+
 dynamic ftrace
 --------------
 

arch/x86/kernel/ds.c

@@ -15,8 +15,8 @@
  * - buffer allocation (memory accounting)
  *
  *
- * Copyright (C) 2007-2008 Intel Corporation.
- * Markus Metzger <markus.t.metzger@intel.com>, 2007-2008
+ * Copyright (C) 2007-2009 Intel Corporation.
+ * Markus Metzger <markus.t.metzger@intel.com>, 2007-2009
  */
 
 
@@ -890,7 +890,7 @@ int ds_set_pebs_reset(struct pebs_tracer *tracer, u64 value)
 }
 
 static const struct ds_configuration ds_cfg_netburst = {
-	.name = "netburst",
+	.name = "Netburst",
 	.ctl[dsf_bts]		= (1 << 2) | (1 << 3),
 	.ctl[dsf_bts_kernel]	= (1 << 5),
 	.ctl[dsf_bts_user]	= (1 << 6),
@@ -904,7 +904,7 @@ static const struct ds_configuration ds_cfg_netburst = {
 #endif
 };
 static const struct ds_configuration ds_cfg_pentium_m = {
-	.name = "pentium m",
+	.name = "Pentium M",
 	.ctl[dsf_bts]		= (1 << 6) | (1 << 7),
 
 	.sizeof_field		= sizeof(long),
@@ -915,8 +915,8 @@ static const struct ds_configuration ds_cfg_pentium_m = {
 	.sizeof_rec[ds_pebs]	= sizeof(long) * 18,
 #endif
 };
-static const struct ds_configuration ds_cfg_core2 = {
-	.name = "core 2",
+static const struct ds_configuration ds_cfg_core2_atom = {
+	.name = "Core 2/Atom",
 	.ctl[dsf_bts]		= (1 << 6) | (1 << 7),
 	.ctl[dsf_bts_kernel]	= (1 << 9),
 	.ctl[dsf_bts_user]	= (1 << 10),
@@ -949,19 +949,22 @@ void __cpuinit ds_init_intel(struct cpuinfo_x86 *c)
 	switch (c->x86) {
 	case 0x6:
 		switch (c->x86_model) {
-		case 0 ... 0xC:
-			/* sorry, don't know about them */
-			break;
-		case 0xD:
-		case 0xE: /* Pentium M */
+		case 0x9:
+		case 0xd: /* Pentium M */
 			ds_configure(&ds_cfg_pentium_m);
 			break;
-		default: /* Core2, Atom, ... */
-			ds_configure(&ds_cfg_core2);
+		case 0xf:
+		case 0x17: /* Core2 */
+		case 0x1c: /* Atom */
+			ds_configure(&ds_cfg_core2_atom);
+			break;
+		case 0x1a: /* i7 */
+		default:
+			/* sorry, don't know about them */
 			break;
 		}
 		break;
-	case 0xF:
+	case 0xf:
 		switch (c->x86_model) {
 		case 0x0:
 		case 0x1:

arch/x86/kernel/dumpstack.c

@@ -14,6 +14,7 @@
 #include <linux/bug.h>
 #include <linux/nmi.h>
 #include <linux/sysfs.h>
+#include <linux/ftrace.h>
 
 #include <asm/stacktrace.h>
 
@@ -195,6 +196,11 @@ unsigned __kprobes long oops_begin(void)
 	int cpu;
 	unsigned long flags;
 
+	/* notify the hw-branch tracer so it may disable tracing and
+	   add the last trace to the trace buffer -
+	   the earlier this happens, the more useful the trace. */
+	trace_hw_branch_oops();
+
 	oops_enter();
 
 	/* racy, but better than risking deadlock. */

include/linux/ftrace.h

@@ -496,4 +496,17 @@ static inline int test_tsk_trace_graph(struct task_struct *tsk)
 
 #endif /* CONFIG_TRACING */
 
+
+#ifdef CONFIG_HW_BRANCH_TRACER
+
+void trace_hw_branch(u64 from, u64 to);
+void trace_hw_branch_oops(void);
+
+#else /* CONFIG_HW_BRANCH_TRACER */
+
+static inline void trace_hw_branch(u64 from, u64 to) {}
+static inline void trace_hw_branch_oops(void) {}
+
+#endif /* CONFIG_HW_BRANCH_TRACER */
+
 #endif /* _LINUX_FTRACE_H */

kernel/trace/trace.h

@@ -438,7 +438,6 @@ void trace_function(struct trace_array *tr,
 
 void trace_graph_return(struct ftrace_graph_ret *trace);
 int trace_graph_entry(struct ftrace_graph_ent *trace);
-void trace_hw_branch(struct trace_array *tr, u64 from, u64 to);
 
 void tracing_start_cmdline_record(void);
 void tracing_stop_cmdline_record(void);