tracing: Remove special traces
[linux-2.6.git] / kernel / trace / trace.h
index e685ac2..638a588 100644 (file)
@@ -7,10 +7,11 @@
 #include <linux/clocksource.h>
 #include <linux/ring_buffer.h>
 #include <linux/mmiotrace.h>
+#include <linux/tracepoint.h>
 #include <linux/ftrace.h>
-#include <trace/boot.h>
-#include <trace/kmemtrace.h>
-#include <trace/power.h>
+#include <linux/hw_breakpoint.h>
+#include <linux/trace_seq.h>
+#include <linux/ftrace_event.h>
 
 enum trace_type {
        __TRACE_FIRST_TYPE = 0,
@@ -21,191 +22,58 @@ enum trace_type {
        TRACE_STACK,
        TRACE_PRINT,
        TRACE_BPRINT,
-       TRACE_SPECIAL,
        TRACE_MMIO_RW,
        TRACE_MMIO_MAP,
        TRACE_BRANCH,
-       TRACE_BOOT_CALL,
-       TRACE_BOOT_RET,
        TRACE_GRAPH_RET,
        TRACE_GRAPH_ENT,
        TRACE_USER_STACK,
-       TRACE_HW_BRANCHES,
-       TRACE_SYSCALL_ENTER,
-       TRACE_SYSCALL_EXIT,
-       TRACE_KMEM_ALLOC,
-       TRACE_KMEM_FREE,
-       TRACE_POWER,
        TRACE_BLK,
 
        __TRACE_LAST_TYPE,
 };
 
-/*
- * The trace entry - the most basic unit of tracing. This is what
- * is printed in the end as a single line in the trace output, such as:
- *
- *     bash-15816 [01]   235.197585: idle_cpu <- irq_enter
- */
-struct trace_entry {
-       unsigned char           type;
-       unsigned char           flags;
-       unsigned char           preempt_count;
-       int                     pid;
-       int                     tgid;
-};
-
-/*
- * Function trace entry - function address and parent function addres:
- */
-struct ftrace_entry {
-       struct trace_entry      ent;
-       unsigned long           ip;
-       unsigned long           parent_ip;
-};
-
-/* Function call entry */
-struct ftrace_graph_ent_entry {
-       struct trace_entry              ent;
-       struct ftrace_graph_ent         graph_ent;
-};
-
-/* Function return entry */
-struct ftrace_graph_ret_entry {
-       struct trace_entry              ent;
-       struct ftrace_graph_ret         ret;
-};
-extern struct tracer boot_tracer;
-
-/*
- * Context switch trace entry - which task (and prio) we switched from/to:
- */
-struct ctx_switch_entry {
-       struct trace_entry      ent;
-       unsigned int            prev_pid;
-       unsigned char           prev_prio;
-       unsigned char           prev_state;
-       unsigned int            next_pid;
-       unsigned char           next_prio;
-       unsigned char           next_state;
-       unsigned int            next_cpu;
-};
-
-/*
- * Special (free-form) trace entry:
- */
-struct special_entry {
-       struct trace_entry      ent;
-       unsigned long           arg1;
-       unsigned long           arg2;
-       unsigned long           arg3;
-};
-
-/*
- * Stack-trace entry:
- */
-
-#define FTRACE_STACK_ENTRIES   8
-
-struct stack_entry {
-       struct trace_entry      ent;
-       unsigned long           caller[FTRACE_STACK_ENTRIES];
-};
-
-struct userstack_entry {
-       struct trace_entry      ent;
-       unsigned long           caller[FTRACE_STACK_ENTRIES];
-};
-
-/*
- * trace_printk entry:
- */
-struct bprint_entry {
-       struct trace_entry      ent;
-       unsigned long           ip;
-       const char              *fmt;
-       u32                     buf[];
-};
-
-struct print_entry {
-       struct trace_entry      ent;
-       unsigned long           ip;
-       char                    buf[];
-};
 
-#define TRACE_OLD_SIZE         88
+#undef __field
+#define __field(type, item)            type    item;
 
-struct trace_field_cont {
-       unsigned char           type;
-       /* Temporary till we get rid of this completely */
-       char                    buf[TRACE_OLD_SIZE - 1];
-};
+#undef __field_struct
+#define __field_struct(type, item)     __field(type, item)
 
-struct trace_mmiotrace_rw {
-       struct trace_entry      ent;
-       struct mmiotrace_rw     rw;
-};
+#undef __field_desc
+#define __field_desc(type, container, item)
 
-struct trace_mmiotrace_map {
-       struct trace_entry      ent;
-       struct mmiotrace_map    map;
-};
+#undef __array
+#define __array(type, item, size)      type    item[size];
 
-struct trace_boot_call {
-       struct trace_entry      ent;
-       struct boot_trace_call boot_call;
-};
+#undef __array_desc
+#define __array_desc(type, container, item, size)
 
-struct trace_boot_ret {
-       struct trace_entry      ent;
-       struct boot_trace_ret boot_ret;
-};
-
-#define TRACE_FUNC_SIZE 30
-#define TRACE_FILE_SIZE 20
-struct trace_branch {
-       struct trace_entry      ent;
-       unsigned                line;
-       char                    func[TRACE_FUNC_SIZE+1];
-       char                    file[TRACE_FILE_SIZE+1];
-       char                    correct;
-};
+#undef __dynamic_array
+#define __dynamic_array(type, item)    type    item[];
 
-struct hw_branch_entry {
-       struct trace_entry      ent;
-       u64                     from;
-       u64                     to;
-};
+#undef F_STRUCT
+#define F_STRUCT(args...)              args
 
-struct trace_power {
-       struct trace_entry      ent;
-       struct power_trace      state_data;
-};
+#undef FTRACE_ENTRY
+#define FTRACE_ENTRY(name, struct_name, id, tstruct, print)    \
+       struct struct_name {                                    \
+               struct trace_entry      ent;                    \
+               tstruct                                         \
+       }
 
-enum kmemtrace_type_id {
-       KMEMTRACE_TYPE_KMALLOC = 0,     /* kmalloc() or kfree(). */
-       KMEMTRACE_TYPE_CACHE,           /* kmem_cache_*(). */
-       KMEMTRACE_TYPE_PAGES,           /* __get_free_pages() and friends. */
-};
+#undef TP_ARGS
+#define TP_ARGS(args...)       args
 
-struct kmemtrace_alloc_entry {
-       struct trace_entry      ent;
-       enum kmemtrace_type_id type_id;
-       unsigned long call_site;
-       const void *ptr;
-       size_t bytes_req;
-       size_t bytes_alloc;
-       gfp_t gfp_flags;
-       int node;
-};
+#undef FTRACE_ENTRY_DUP
+#define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk)
 
-struct kmemtrace_free_entry {
-       struct trace_entry      ent;
-       enum kmemtrace_type_id type_id;
-       unsigned long call_site;
-       const void *ptr;
-};
+#include "trace_entries.h"
 
+/*
+ * syscalls are special, and need special handling, this is why
+ * they are not included in trace_entries.h
+ */
 struct syscall_trace_enter {
        struct trace_entry      ent;
        int                     nr;
@@ -215,16 +83,26 @@ struct syscall_trace_enter {
 struct syscall_trace_exit {
        struct trace_entry      ent;
        int                     nr;
-       unsigned long           ret;
+       long                    ret;
 };
 
+struct kprobe_trace_entry_head {
+       struct trace_entry      ent;
+       unsigned long           ip;
+};
+
+struct kretprobe_trace_entry_head {
+       struct trace_entry      ent;
+       unsigned long           func;
+       unsigned long           ret_ip;
+};
 
 /*
  * trace_flag_type is an enumeration that holds different
  * states when a trace occurs. These are:
  *  IRQS_OFF           - interrupts were disabled
  *  IRQS_NOSUPPORT     - arch does not support irqs_disabled_flags
- *  NEED_RESCED                - reschedule is requested
+ *  NEED_RESCHED       - reschedule is requested
  *  HARDIRQ            - inside an interrupt handler
  *  SOFTIRQ            - inside a softirq handler
  */
@@ -247,9 +125,6 @@ struct trace_array_cpu {
        atomic_t                disabled;
        void                    *buffer_page;   /* ring buffer spare */
 
-       /* these fields get copied into max-trace: */
-       unsigned long           trace_idx;
-       unsigned long           overrun;
        unsigned long           saved_latency;
        unsigned long           critical_start;
        unsigned long           critical_end;
@@ -257,14 +132,13 @@ struct trace_array_cpu {
        unsigned long           nice;
        unsigned long           policy;
        unsigned long           rt_priority;
+       unsigned long           skipped_entries;
        cycle_t                 preempt_timestamp;
        pid_t                   pid;
        uid_t                   uid;
        char                    comm[TASK_COMM_LEN];
 };
 
-struct trace_iterator;
-
 /*
  * The trace array - an array of per-CPU trace arrays. This is the
  * highest level data structure that individual tracers deal with.
@@ -314,40 +188,18 @@ extern void __ftrace_bad_type(void);
                IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
                IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT);   \
                IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT); \
-               IF_ASSIGN(var, ent, struct special_entry, 0);           \
                IF_ASSIGN(var, ent, struct trace_mmiotrace_rw,          \
                          TRACE_MMIO_RW);                               \
                IF_ASSIGN(var, ent, struct trace_mmiotrace_map,         \
                          TRACE_MMIO_MAP);                              \
-               IF_ASSIGN(var, ent, struct trace_boot_call, TRACE_BOOT_CALL);\
-               IF_ASSIGN(var, ent, struct trace_boot_ret, TRACE_BOOT_RET);\
                IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
                IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry,      \
                          TRACE_GRAPH_ENT);             \
                IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,      \
                          TRACE_GRAPH_RET);             \
-               IF_ASSIGN(var, ent, struct hw_branch_entry, TRACE_HW_BRANCHES);\
-               IF_ASSIGN(var, ent, struct trace_power, TRACE_POWER); \
-               IF_ASSIGN(var, ent, struct kmemtrace_alloc_entry,       \
-                         TRACE_KMEM_ALLOC);    \
-               IF_ASSIGN(var, ent, struct kmemtrace_free_entry,        \
-                         TRACE_KMEM_FREE);     \
-               IF_ASSIGN(var, ent, struct syscall_trace_enter,         \
-                         TRACE_SYSCALL_ENTER);                         \
-               IF_ASSIGN(var, ent, struct syscall_trace_exit,          \
-                         TRACE_SYSCALL_EXIT);                          \
                __ftrace_bad_type();                                    \
        } while (0)
 
-/* Return values for print_line callback */
-enum print_line_t {
-       TRACE_TYPE_PARTIAL_LINE = 0,    /* Retry after flushing the seq */
-       TRACE_TYPE_HANDLED      = 1,
-       TRACE_TYPE_UNHANDLED    = 2,    /* Relay to other output functions */
-       TRACE_TYPE_NO_CONSUME   = 3     /* Handled but ask to not consume */
-};
-
-
 /*
  * An option specific to a tracer. This is a boolean value.
  * The bit is the bit index that sets its value on the
@@ -382,6 +234,7 @@ struct tracer_flags {
  * @pipe_open: called when the trace_pipe file is opened
  * @wait_pipe: override how the user waits for traces on trace_pipe
  * @close: called when the trace file is released
+ * @pipe_close: called when the trace_pipe file is released
  * @read: override the default read callback on trace_pipe
  * @splice_read: override the default splice_read callback on trace_pipe
  * @selftest: selftest to run on boot (see trace_selftest.c)
@@ -400,6 +253,7 @@ struct tracer {
        void                    (*pipe_open)(struct trace_iterator *iter);
        void                    (*wait_pipe)(struct trace_iterator *iter);
        void                    (*close)(struct trace_iterator *iter);
+       void                    (*pipe_close)(struct trace_iterator *iter);
        ssize_t                 (*read)(struct trace_iterator *iter,
                                        struct file *filp, char __user *ubuf,
                                        size_t cnt, loff_t *ppos);
@@ -420,88 +274,45 @@ struct tracer {
        struct tracer           *next;
        int                     print_max;
        struct tracer_flags     *flags;
-       struct tracer_stat      *stats;
-};
-
-struct trace_seq {
-       unsigned char           buffer[PAGE_SIZE];
-       unsigned int            len;
-       unsigned int            readpos;
 };
 
-static inline void
-trace_seq_init(struct trace_seq *s)
-{
-       s->len = 0;
-       s->readpos = 0;
-}
-
 
 #define TRACE_PIPE_ALL_CPU     -1
 
-/*
- * Trace iterator - used by printout routines who present trace
- * results to users and which routines might sleep, etc:
- */
-struct trace_iterator {
-       struct trace_array      *tr;
-       struct tracer           *trace;
-       void                    *private;
-       int                     cpu_file;
-       struct mutex            mutex;
-       struct ring_buffer_iter *buffer_iter[NR_CPUS];
-
-       /* The below is zeroed out in pipe_read */
-       struct trace_seq        seq;
-       struct trace_entry      *ent;
-       int                     cpu;
-       u64                     ts;
-
-       unsigned long           iter_flags;
-       loff_t                  pos;
-       long                    idx;
-
-       cpumask_var_t           started;
-};
-
 int tracer_init(struct tracer *t, struct trace_array *tr);
 int tracing_is_enabled(void);
 void trace_wake_up(void);
 void tracing_reset(struct trace_array *tr, int cpu);
 void tracing_reset_online_cpus(struct trace_array *tr);
+void tracing_reset_current(int cpu);
+void tracing_reset_current_online_cpus(void);
 int tracing_open_generic(struct inode *inode, struct file *filp);
+struct dentry *trace_create_file(const char *name,
+                                mode_t mode,
+                                struct dentry *parent,
+                                void *data,
+                                const struct file_operations *fops);
+
 struct dentry *tracing_init_dentry(void);
-void init_tracer_sysprof_debugfs(struct dentry *d_tracer);
 
 struct ring_buffer_event;
 
-struct ring_buffer_event *trace_buffer_lock_reserve(struct trace_array *tr,
-                                                   unsigned char type,
-                                                   unsigned long len,
-                                                   unsigned long flags,
-                                                   int pc);
-void trace_buffer_unlock_commit(struct trace_array *tr,
+struct ring_buffer_event *
+trace_buffer_lock_reserve(struct ring_buffer *buffer,
+                         int type,
+                         unsigned long len,
+                         unsigned long flags,
+                         int pc);
+void trace_buffer_unlock_commit(struct ring_buffer *buffer,
                                struct ring_buffer_event *event,
                                unsigned long flags, int pc);
 
-struct ring_buffer_event *
-trace_current_buffer_lock_reserve(unsigned char type, unsigned long len,
-                                 unsigned long flags, int pc);
-void trace_current_buffer_unlock_commit(struct ring_buffer_event *event,
-                                       unsigned long flags, int pc);
-void trace_nowake_buffer_unlock_commit(struct ring_buffer_event *event,
-                                       unsigned long flags, int pc);
-
 struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
                                                struct trace_array_cpu *data);
 
 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
                                          int *ent_cpu, u64 *ent_ts);
 
-void tracing_generic_entry_update(struct trace_entry *entry,
-                                 unsigned long flags,
-                                 int pc);
-
 void default_wait_pipe(struct trace_iterator *iter);
 void poll_wait_pipe(struct trace_iterator *iter);
 
@@ -514,24 +325,22 @@ void tracing_sched_switch_trace(struct trace_array *tr,
                                struct task_struct *prev,
                                struct task_struct *next,
                                unsigned long flags, int pc);
-void tracing_record_cmdline(struct task_struct *tsk);
 
 void tracing_sched_wakeup_trace(struct trace_array *tr,
                                struct task_struct *wakee,
                                struct task_struct *cur,
                                unsigned long flags, int pc);
-void trace_special(struct trace_array *tr,
-                  struct trace_array_cpu *data,
-                  unsigned long arg1,
-                  unsigned long arg2,
-                  unsigned long arg3, int pc);
 void trace_function(struct trace_array *tr,
                    unsigned long ip,
                    unsigned long parent_ip,
                    unsigned long flags, int pc);
+void trace_default_header(struct seq_file *m);
+void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
+int trace_empty(struct trace_iterator *iter);
 
 void trace_graph_return(struct ftrace_graph_ret *trace);
 int trace_graph_entry(struct ftrace_graph_ent *trace);
+void set_graph_array(struct trace_array *tr);
 
 void tracing_start_cmdline_record(void);
 void tracing_stop_cmdline_record(void);
@@ -540,35 +349,47 @@ void tracing_stop_sched_switch_record(void);
 void tracing_start_sched_switch_record(void);
 int register_tracer(struct tracer *type);
 void unregister_tracer(struct tracer *type);
+int is_tracing_stopped(void);
 
 extern unsigned long nsecs_to_usecs(unsigned long nsecs);
 
-extern unsigned long tracing_max_latency;
 extern unsigned long tracing_thresh;
 
+#ifdef CONFIG_TRACER_MAX_TRACE
+extern unsigned long tracing_max_latency;
+
 void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
 void update_max_tr_single(struct trace_array *tr,
                          struct task_struct *tsk, int cpu);
+#endif /* CONFIG_TRACER_MAX_TRACE */
 
-void __trace_stack(struct trace_array *tr,
-                  unsigned long flags,
-                  int skip, int pc);
+#ifdef CONFIG_STACKTRACE
+void ftrace_trace_stack(struct ring_buffer *buffer, unsigned long flags,
+                       int skip, int pc);
 
-extern cycle_t ftrace_now(int cpu);
+void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
+                           int pc);
 
-#ifdef CONFIG_CONTEXT_SWITCH_TRACER
-typedef void
-(*tracer_switch_func_t)(void *private,
-                       void *__rq,
-                       struct task_struct *prev,
-                       struct task_struct *next);
-
-struct tracer_switch_ops {
-       tracer_switch_func_t            func;
-       void                            *private;
-       struct tracer_switch_ops        *next;
-};
-#endif /* CONFIG_CONTEXT_SWITCH_TRACER */
+void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
+                  int pc);
+#else
+static inline void ftrace_trace_stack(struct ring_buffer *buffer,
+                                     unsigned long flags, int skip, int pc)
+{
+}
+
+static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
+                                         unsigned long flags, int pc)
+{
+}
+
+static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
+                                int skip, int pc)
+{
+}
+#endif /* CONFIG_STACKTRACE */
+
+extern cycle_t ftrace_now(int cpu);
 
 extern void trace_find_cmdline(int pid, char comm[]);
 
@@ -578,6 +399,10 @@ extern unsigned long ftrace_update_tot_cnt;
 extern int DYN_FTRACE_TEST_NAME(void);
 #endif
 
+extern int ring_buffer_expanded;
+extern bool tracing_selftest_disabled;
+DECLARE_PER_CPU(int, ftrace_cpu_disabled);
+
 #ifdef CONFIG_FTRACE_STARTUP_TEST
 extern int trace_selftest_startup_function(struct tracer *trace,
                                           struct trace_array *tr);
@@ -595,8 +420,6 @@ extern int trace_selftest_startup_nop(struct tracer *trace,
                                         struct trace_array *tr);
 extern int trace_selftest_startup_sched_switch(struct tracer *trace,
                                               struct trace_array *tr);
-extern int trace_selftest_startup_sysprof(struct tracer *trace,
-                                              struct trace_array *tr);
 extern int trace_selftest_startup_branch(struct tracer *trace,
                                         struct trace_array *tr);
 #endif /* CONFIG_FTRACE_STARTUP_TEST */
@@ -607,16 +430,46 @@ extern int
 trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
 extern int
 trace_vprintk(unsigned long ip, const char *fmt, va_list args);
+extern int
+trace_array_vprintk(struct trace_array *tr,
+                   unsigned long ip, const char *fmt, va_list args);
+int trace_array_printk(struct trace_array *tr,
+                      unsigned long ip, const char *fmt, ...);
 
 extern unsigned long trace_flags;
 
+extern int trace_clock_id;
+
 /* Standard output formatting function used for function return traces */
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
-extern enum print_line_t print_graph_function(struct trace_iterator *iter);
+
+/* Flag options */
+#define TRACE_GRAPH_PRINT_OVERRUN       0x1
+#define TRACE_GRAPH_PRINT_CPU           0x2
+#define TRACE_GRAPH_PRINT_OVERHEAD      0x4
+#define TRACE_GRAPH_PRINT_PROC          0x8
+#define TRACE_GRAPH_PRINT_DURATION      0x10
+#define TRACE_GRAPH_PRINT_ABS_TIME      0x20
+
+extern enum print_line_t
+print_graph_function_flags(struct trace_iterator *iter, u32 flags);
+extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
+extern enum print_line_t
+trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
+extern void graph_trace_open(struct trace_iterator *iter);
+extern void graph_trace_close(struct trace_iterator *iter);
+extern int __trace_graph_entry(struct trace_array *tr,
+                              struct ftrace_graph_ent *trace,
+                              unsigned long flags, int pc);
+extern void __trace_graph_return(struct trace_array *tr,
+                                struct ftrace_graph_ret *trace,
+                                unsigned long flags, int pc);
+
 
 #ifdef CONFIG_DYNAMIC_FTRACE
 /* TODO: make this variable */
 #define FTRACE_GRAPH_MAX_FUNCS         32
+extern int ftrace_graph_filter_enabled;
 extern int ftrace_graph_count;
 extern unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS];
 
@@ -624,7 +477,7 @@ static inline int ftrace_graph_addr(unsigned long addr)
 {
        int i;
 
-       if (!ftrace_graph_count || test_tsk_trace_graph(current))
+       if (!ftrace_graph_filter_enabled)
                return 1;
 
        for (i = 0; i < ftrace_graph_count; i++) {
@@ -635,33 +488,70 @@ static inline int ftrace_graph_addr(unsigned long addr)
        return 0;
 }
 #else
-static inline int ftrace_trace_addr(unsigned long addr)
-{
-       return 1;
-}
 static inline int ftrace_graph_addr(unsigned long addr)
 {
        return 1;
 }
 #endif /* CONFIG_DYNAMIC_FTRACE */
-
 #else /* CONFIG_FUNCTION_GRAPH_TRACER */
 static inline enum print_line_t
-print_graph_function(struct trace_iterator *iter)
+print_graph_function_flags(struct trace_iterator *iter, u32 flags)
 {
        return TRACE_TYPE_UNHANDLED;
 }
 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
 
-extern struct pid *ftrace_pid_trace;
+extern struct list_head ftrace_pids;
 
+#ifdef CONFIG_FUNCTION_TRACER
 static inline int ftrace_trace_task(struct task_struct *task)
 {
-       if (!ftrace_pid_trace)
+       if (list_empty(&ftrace_pids))
                return 1;
 
        return test_tsk_trace_trace(task);
 }
+#else
+static inline int ftrace_trace_task(struct task_struct *task)
+{
+       return 1;
+}
+#endif
+
+/*
+ * struct trace_parser - servers for reading the user input separated by spaces
+ * @cont: set if the input is not complete - no final space char was found
+ * @buffer: holds the parsed user input
+ * @idx: user input length
+ * @size: buffer size
+ */
+struct trace_parser {
+       bool            cont;
+       char            *buffer;
+       unsigned        idx;
+       unsigned        size;
+};
+
+static inline bool trace_parser_loaded(struct trace_parser *parser)
+{
+       return (parser->idx != 0);
+}
+
+static inline bool trace_parser_cont(struct trace_parser *parser)
+{
+       return parser->cont;
+}
+
+static inline void trace_parser_clear(struct trace_parser *parser)
+{
+       parser->cont = false;
+       parser->idx = 0;
+}
+
+extern int trace_parser_get_init(struct trace_parser *parser, int size);
+extern void trace_parser_put(struct trace_parser *parser);
+extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
+       size_t cnt, loff_t *ppos);
 
 /*
  * trace_iterator_flags is an enumeration that defines bit
@@ -680,18 +570,17 @@ enum trace_iterator_flags {
        TRACE_ITER_BIN                  = 0x40,
        TRACE_ITER_BLOCK                = 0x80,
        TRACE_ITER_STACKTRACE           = 0x100,
-       TRACE_ITER_SCHED_TREE           = 0x200,
-       TRACE_ITER_PRINTK               = 0x400,
-       TRACE_ITER_PREEMPTONLY          = 0x800,
-       TRACE_ITER_BRANCH               = 0x1000,
-       TRACE_ITER_ANNOTATE             = 0x2000,
-       TRACE_ITER_USERSTACKTRACE       = 0x4000,
-       TRACE_ITER_SYM_USEROBJ          = 0x8000,
-       TRACE_ITER_PRINTK_MSGONLY       = 0x10000,
-       TRACE_ITER_CONTEXT_INFO         = 0x20000, /* Print pid/cpu/time */
-       TRACE_ITER_LATENCY_FMT          = 0x40000,
-       TRACE_ITER_GLOBAL_CLK           = 0x80000,
-       TRACE_ITER_SLEEP_TIME           = 0x100000,
+       TRACE_ITER_PRINTK               = 0x200,
+       TRACE_ITER_PREEMPTONLY          = 0x400,
+       TRACE_ITER_BRANCH               = 0x800,
+       TRACE_ITER_ANNOTATE             = 0x1000,
+       TRACE_ITER_USERSTACKTRACE       = 0x2000,
+       TRACE_ITER_SYM_USEROBJ          = 0x4000,
+       TRACE_ITER_PRINTK_MSGONLY       = 0x8000,
+       TRACE_ITER_CONTEXT_INFO         = 0x10000, /* Print pid/cpu/time */
+       TRACE_ITER_LATENCY_FMT          = 0x20000,
+       TRACE_ITER_SLEEP_TIME           = 0x40000,
+       TRACE_ITER_GRAPH_TIME           = 0x80000,
 };
 
 /*
@@ -703,54 +592,6 @@ enum trace_iterator_flags {
 
 extern struct tracer nop_trace;
 
-/**
- * ftrace_preempt_disable - disable preemption scheduler safe
- *
- * When tracing can happen inside the scheduler, there exists
- * cases that the tracing might happen before the need_resched
- * flag is checked. If this happens and the tracer calls
- * preempt_enable (after a disable), a schedule might take place
- * causing an infinite recursion.
- *
- * To prevent this, we read the need_resched flag before
- * disabling preemption. When we want to enable preemption we
- * check the flag, if it is set, then we call preempt_enable_no_resched.
- * Otherwise, we call preempt_enable.
- *
- * The rational for doing the above is that if need_resched is set
- * and we have yet to reschedule, we are either in an atomic location
- * (where we do not need to check for scheduling) or we are inside
- * the scheduler and do not want to resched.
- */
-static inline int ftrace_preempt_disable(void)
-{
-       int resched;
-
-       resched = need_resched();
-       preempt_disable_notrace();
-
-       return resched;
-}
-
-/**
- * ftrace_preempt_enable - enable preemption scheduler safe
- * @resched: the return value from ftrace_preempt_disable
- *
- * This is a scheduler safe way to enable preemption and not miss
- * any preemption checks. The disabled saved the state of preemption.
- * If resched is set, then we are either inside an atomic or
- * are inside the scheduler (we would have already scheduled
- * otherwise). In this case, we do not want to call normal
- * preempt_enable, but preempt_enable_no_resched instead.
- */
-static inline void ftrace_preempt_enable(int resched)
-{
-       if (resched)
-               preempt_enable_no_resched_notrace();
-       else
-               preempt_enable_notrace();
-}
-
 #ifdef CONFIG_BRANCH_TRACER
 extern int enable_branch_tracing(struct trace_array *tr);
 extern void disable_branch_tracing(void);
@@ -788,105 +629,103 @@ struct ftrace_event_field {
        struct list_head        link;
        char                    *name;
        char                    *type;
+       int                     filter_type;
        int                     offset;
        int                     size;
+       int                     is_signed;
 };
 
-struct ftrace_event_call {
-       char                    *name;
-       char                    *system;
-       struct dentry           *dir;
-       int                     enabled;
-       int                     (*regfunc)(void);
-       void                    (*unregfunc)(void);
-       int                     id;
-       int                     (*raw_init)(void);
-       int                     (*show_format)(struct trace_seq *s);
-       int                     (*define_fields)(void);
-       struct list_head        fields;
+struct event_filter {
+       int                     n_preds;
        struct filter_pred      **preds;
-
-#ifdef CONFIG_EVENT_PROFILE
-       atomic_t        profile_count;
-       int             (*profile_enable)(struct ftrace_event_call *);
-       void            (*profile_disable)(struct ftrace_event_call *);
-#endif
+       char                    *filter_string;
 };
 
 struct event_subsystem {
        struct list_head        list;
        const char              *name;
        struct dentry           *entry;
-       struct filter_pred      **preds;
+       struct event_filter     *filter;
+       int                     nr_events;
 };
 
-#define events_for_each(event)                                         \
-       for (event = __start_ftrace_events;                             \
-            (unsigned long)event < (unsigned long)__stop_ftrace_events; \
-            event++)
+struct filter_pred;
+struct regex;
 
-#define MAX_FILTER_PRED 8
+typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event,
+                                int val1, int val2);
 
-struct filter_pred;
+typedef int (*regex_match_func)(char *str, struct regex *r, int len);
+
+enum regex_type {
+       MATCH_FULL = 0,
+       MATCH_FRONT_ONLY,
+       MATCH_MIDDLE_ONLY,
+       MATCH_END_ONLY,
+};
 
-typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
+struct regex {
+       char                    pattern[MAX_FILTER_STR_VAL];
+       int                     len;
+       int                     field_len;
+       regex_match_func        match;
+};
 
 struct filter_pred {
-       filter_pred_fn_t fn;
-       u64 val;
-       char *str_val;
-       int str_len;
-       char *field_name;
-       int offset;
-       int not;
-       int or;
-       int compound;
-       int clear;
+       filter_pred_fn_t        fn;
+       u64                     val;
+       struct regex            regex;
+       char                    *field_name;
+       int                     offset;
+       int                     not;
+       int                     op;
+       int                     pop_n;
 };
 
-int trace_define_field(struct ftrace_event_call *call, char *type,
-                      char *name, int offset, int size);
-extern void filter_free_pred(struct filter_pred *pred);
-extern void filter_print_preds(struct filter_pred **preds,
+extern struct list_head ftrace_common_fields;
+
+extern enum regex_type
+filter_parse_regex(char *buff, int len, char **search, int *not);
+extern void print_event_filter(struct ftrace_event_call *call,
                               struct trace_seq *s);
-extern int filter_parse(char **pbuf, struct filter_pred *pred);
-extern int filter_add_pred(struct ftrace_event_call *call,
-                          struct filter_pred *pred);
-extern void filter_free_preds(struct ftrace_event_call *call);
-extern int filter_match_preds(struct ftrace_event_call *call, void *rec);
-extern void filter_free_subsystem_preds(struct event_subsystem *system);
-extern int filter_add_subsystem_pred(struct event_subsystem *system,
-                                    struct filter_pred *pred);
-
-void event_trace_printk(unsigned long ip, const char *fmt, ...);
-extern struct ftrace_event_call __start_ftrace_events[];
-extern struct ftrace_event_call __stop_ftrace_events[];
-
-#define for_each_event(event)                                          \
-       for (event = __start_ftrace_events;                             \
-            (unsigned long)event < (unsigned long)__stop_ftrace_events; \
-            event++)
+extern int apply_event_filter(struct ftrace_event_call *call,
+                             char *filter_string);
+extern int apply_subsystem_event_filter(struct event_subsystem *system,
+                                       char *filter_string);
+extern void print_subsystem_event_filter(struct event_subsystem *system,
+                                        struct trace_seq *s);
+extern int filter_assign_type(const char *type);
+
+struct list_head *
+trace_get_fields(struct ftrace_event_call *event_call);
+
+static inline int
+filter_check_discard(struct ftrace_event_call *call, void *rec,
+                    struct ring_buffer *buffer,
+                    struct ring_buffer_event *event)
+{
+       if (unlikely(call->flags & TRACE_EVENT_FL_FILTERED) &&
+           !filter_match_preds(call->filter, rec)) {
+               ring_buffer_discard_commit(buffer, event);
+               return 1;
+       }
+
+       return 0;
+}
+
+extern struct mutex event_mutex;
+extern struct list_head ftrace_events;
 
 extern const char *__start___trace_bprintk_fmt[];
 extern const char *__stop___trace_bprintk_fmt[];
 
-/*
- * The double __builtin_constant_p is because gcc will give us an error
- * if we try to allocate the static variable to fmt if it is not a
- * constant. Even with the outer if statement optimizing out.
- */
-#define event_trace_printk(ip, fmt, args...)                           \
-do {                                                                   \
-       __trace_printk_check_format(fmt, ##args);                       \
-       tracing_record_cmdline(current);                                \
-       if (__builtin_constant_p(fmt)) {                                \
-               static const char *trace_printk_fmt                     \
-                 __attribute__((section("__trace_printk_fmt"))) =      \
-                       __builtin_constant_p(fmt) ? fmt : NULL;         \
-                                                                       \
-               __trace_bprintk(ip, trace_printk_fmt, ##args);          \
-       } else                                                          \
-               __trace_printk(ip, fmt, ##args);                        \
-} while (0)
+#undef FTRACE_ENTRY
+#define FTRACE_ENTRY(call, struct_name, id, tstruct, print)            \
+       extern struct ftrace_event_call                                 \
+       __attribute__((__aligned__(4))) event_##call;
+#undef FTRACE_ENTRY_DUP
+#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print)                \
+       FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print))
+#include "trace_entries.h"
 
 #endif /* _LINUX_KERNEL_TRACE_H */