tracing: fix command line to pid reverse map
[pandora-kernel.git] / kernel / trace / trace.c
index bdaf60d..305c562 100644 (file)
  *  Copyright (C) 2004-2006 Ingo Molnar
  *  Copyright (C) 2004 William Lee Irwin III
  */
+#include <linux/ring_buffer.h>
 #include <linux/utsrelease.h>
+#include <linux/stacktrace.h>
+#include <linux/writeback.h>
 #include <linux/kallsyms.h>
 #include <linux/seq_file.h>
 #include <linux/notifier.h>
+#include <linux/irqflags.h>
 #include <linux/debugfs.h>
 #include <linux/pagemap.h>
 #include <linux/hardirq.h>
 #include <linux/linkage.h>
 #include <linux/uaccess.h>
+#include <linux/kprobes.h>
 #include <linux/ftrace.h>
 #include <linux/module.h>
 #include <linux/percpu.h>
+#include <linux/splice.h>
 #include <linux/kdebug.h>
 #include <linux/ctype.h>
 #include <linux/init.h>
 #include <linux/poll.h>
 #include <linux/gfp.h>
 #include <linux/fs.h>
-#include <linux/kprobes.h>
-#include <linux/writeback.h>
-#include <linux/splice.h>
-
-#include <linux/stacktrace.h>
-#include <linux/ring_buffer.h>
-#include <linux/irqflags.h>
 
 #include "trace.h"
 #include "trace_output.h"
 unsigned long __read_mostly    tracing_max_latency;
 unsigned long __read_mostly    tracing_thresh;
 
+/*
+ * On boot up, the ring buffer is set to the minimum size, so that
+ * we do not waste memory on systems that are not using tracing.
+ */
+static int ring_buffer_expanded;
+
 /*
  * We need to change this state when a selftest is running.
  * A selftest will lurk into the ring-buffer to count the
  * entries inserted during the selftest although some concurrent
- * insertions into the ring-buffer such as ftrace_printk could occurred
+ * insertions into the ring-buffer such as trace_printk could occurred
  * at the same time, giving false positive or negative results.
  */
 static bool __read_mostly tracing_selftest_running;
@@ -129,6 +134,8 @@ static int __init set_ftrace(char *str)
 {
        strncpy(bootup_tracer_buf, str, BOOTUP_TRACER_SIZE);
        default_bootup_tracer = bootup_tracer_buf;
+       /* We are using ftrace early, expand it */
+       ring_buffer_expanded = 1;
        return 1;
 }
 __setup("ftrace=", set_ftrace);
@@ -292,7 +299,7 @@ static const char *trace_options[] = {
        "block",
        "stacktrace",
        "sched-tree",
-       "ftrace_printk",
+       "trace_printk",
        "ftrace_preempt",
        "branch",
        "annotate",
@@ -300,6 +307,7 @@ static const char *trace_options[] = {
        "sym-userobj",
        "printk-msg-only",
        "context-info",
+       "latency-format",
        NULL
 };
 
@@ -342,18 +350,14 @@ __update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
        tracing_record_cmdline(tsk);
 }
 
-static void
-trace_seq_reset(struct trace_seq *s)
-{
-       s->len = 0;
-       s->readpos = 0;
-}
-
 ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, size_t cnt)
 {
        int len;
        int ret;
 
+       if (!cnt)
+               return 0;
+
        if (s->len <= s->readpos)
                return -EBUSY;
 
@@ -361,10 +365,12 @@ ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, size_t cnt)
        if (cnt > len)
                cnt = len;
        ret = copy_to_user(ubuf, s->buffer + s->readpos, cnt);
-       if (ret)
+       if (ret == cnt)
                return -EFAULT;
 
-       s->readpos += len;
+       cnt -= ret;
+
+       s->readpos += cnt;
        return cnt;
 }
 
@@ -383,7 +389,7 @@ ssize_t trace_seq_to_buffer(struct trace_seq *s, void *buf, size_t cnt)
        if (!ret)
                return -EFAULT;
 
-       s->readpos += len;
+       s->readpos += cnt;
        return cnt;
 }
 
@@ -395,7 +401,7 @@ trace_print_seq(struct seq_file *m, struct trace_seq *s)
        s->buffer[len] = 0;
        seq_puts(m, s->buffer);
 
-       trace_seq_reset(s);
+       trace_seq_init(s);
 }
 
 /**
@@ -627,19 +633,20 @@ void tracing_reset_online_cpus(struct trace_array *tr)
 }
 
 #define SAVED_CMDLINES 128
+#define NO_CMDLINE_MAP UINT_MAX
 static unsigned map_pid_to_cmdline[PID_MAX_DEFAULT+1];
 static unsigned map_cmdline_to_pid[SAVED_CMDLINES];
 static char saved_cmdlines[SAVED_CMDLINES][TASK_COMM_LEN];
 static int cmdline_idx;
-static DEFINE_SPINLOCK(trace_cmdline_lock);
+static raw_spinlock_t trace_cmdline_lock = __RAW_SPIN_LOCK_UNLOCKED;
 
 /* temporary disable recording */
 static atomic_t trace_record_cmdline_disabled __read_mostly;
 
 static void trace_init_cmdlines(void)
 {
-       memset(&map_pid_to_cmdline, -1, sizeof(map_pid_to_cmdline));
-       memset(&map_cmdline_to_pid, -1, sizeof(map_cmdline_to_pid));
+       memset(&map_pid_to_cmdline, NO_CMDLINE_MAP, sizeof(map_pid_to_cmdline));
+       memset(&map_cmdline_to_pid, NO_CMDLINE_MAP, sizeof(map_cmdline_to_pid));
        cmdline_idx = 0;
 }
 
@@ -731,8 +738,7 @@ void trace_stop_cmdline_recording(void);
 
 static void trace_save_cmdline(struct task_struct *tsk)
 {
-       unsigned map;
-       unsigned idx;
+       unsigned pid, idx;
 
        if (!tsk->pid || unlikely(tsk->pid > PID_MAX_DEFAULT))
                return;
@@ -743,17 +749,24 @@ static void trace_save_cmdline(struct task_struct *tsk)
         * nor do we want to disable interrupts,
         * so if we miss here, then better luck next time.
         */
-       if (!spin_trylock(&trace_cmdline_lock))
+       if (!__raw_spin_trylock(&trace_cmdline_lock))
                return;
 
        idx = map_pid_to_cmdline[tsk->pid];
-       if (idx >= SAVED_CMDLINES) {
+       if (idx == NO_CMDLINE_MAP) {
                idx = (cmdline_idx + 1) % SAVED_CMDLINES;
 
-               map = map_cmdline_to_pid[idx];
-               if (map <= PID_MAX_DEFAULT)
-                       map_pid_to_cmdline[map] = (unsigned)-1;
+               /*
+                * Check whether the cmdline buffer at idx has a pid
+                * mapped. We are going to overwrite that entry so we
+                * need to clear the map_pid_to_cmdline. Otherwise we
+                * would read the new comm for the old pid.
+                */
+               pid = map_cmdline_to_pid[idx];
+               if (pid != NO_CMDLINE_MAP)
+                       map_pid_to_cmdline[pid] = NO_CMDLINE_MAP;
 
+               map_cmdline_to_pid[idx] = tsk->pid;
                map_pid_to_cmdline[tsk->pid] = idx;
 
                cmdline_idx = idx;
@@ -761,33 +774,37 @@ static void trace_save_cmdline(struct task_struct *tsk)
 
        memcpy(&saved_cmdlines[idx], tsk->comm, TASK_COMM_LEN);
 
-       spin_unlock(&trace_cmdline_lock);
+       __raw_spin_unlock(&trace_cmdline_lock);
 }
 
-char *trace_find_cmdline(int pid)
+void trace_find_cmdline(int pid, char comm[])
 {
-       char *cmdline = "<...>";
        unsigned map;
 
-       if (!pid)
-               return "<idle>";
+       if (!pid) {
+               strcpy(comm, "<idle>");
+               return;
+       }
 
-       if (pid > PID_MAX_DEFAULT)
-               goto out;
+       if (pid > PID_MAX_DEFAULT) {
+               strcpy(comm, "<...>");
+               return;
+       }
 
+       __raw_spin_lock(&trace_cmdline_lock);
        map = map_pid_to_cmdline[pid];
-       if (map >= SAVED_CMDLINES)
-               goto out;
-
-       cmdline = saved_cmdlines[map];
+       if (map != NO_CMDLINE_MAP)
+               strcpy(comm, saved_cmdlines[map]);
+       else
+               strcpy(comm, "<...>");
 
- out:
-       return cmdline;
+       __raw_spin_unlock(&trace_cmdline_lock);
 }
 
 void tracing_record_cmdline(struct task_struct *tsk)
 {
-       if (atomic_read(&trace_record_cmdline_disabled))
+       if (atomic_read(&trace_record_cmdline_disabled) || !tracer_enabled ||
+           !tracing_is_on())
                return;
 
        trace_save_cmdline(tsk);
@@ -801,7 +818,7 @@ tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags,
 
        entry->preempt_count            = pc & 0xff;
        entry->pid                      = (tsk) ? tsk->pid : 0;
-       entry->tgid                     = (tsk) ? tsk->tgid : 0;
+       entry->tgid                     = (tsk) ? tsk->tgid : 0;
        entry->flags =
 #ifdef CONFIG_TRACE_IRQFLAGS_SUPPORT
                (irqs_disabled_flags(flags) ? TRACE_FLAG_IRQS_OFF : 0) |
@@ -846,6 +863,20 @@ void trace_buffer_unlock_commit(struct trace_array *tr,
        trace_wake_up();
 }
 
+struct ring_buffer_event *
+trace_current_buffer_lock_reserve(unsigned char type, unsigned long len,
+                                 unsigned long flags, int pc)
+{
+       return trace_buffer_lock_reserve(&global_trace,
+                                        type, len, flags, pc);
+}
+
+void trace_current_buffer_unlock_commit(struct ring_buffer_event *event,
+                                       unsigned long flags, int pc)
+{
+       return trace_buffer_unlock_commit(&global_trace, event, flags, pc);
+}
+
 void
 trace_function(struct trace_array *tr,
               unsigned long ip, unsigned long parent_ip, unsigned long flags,
@@ -1157,6 +1188,125 @@ void trace_graph_return(struct ftrace_graph_ret *trace)
 }
 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
 
+
+/**
+ * trace_vbprintk - write binary msg to tracing buffer
+ *
+ */
+int trace_vbprintk(unsigned long ip, int depth, const char *fmt, va_list args)
+{
+       static raw_spinlock_t trace_buf_lock =
+               (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
+       static u32 trace_buf[TRACE_BUF_SIZE];
+
+       struct ring_buffer_event *event;
+       struct trace_array *tr = &global_trace;
+       struct trace_array_cpu *data;
+       struct bprint_entry *entry;
+       unsigned long flags;
+       int resched;
+       int cpu, len = 0, size, pc;
+
+       if (unlikely(tracing_selftest_running || tracing_disabled))
+               return 0;
+
+       /* Don't pollute graph traces with trace_vprintk internals */
+       pause_graph_tracing();
+
+       pc = preempt_count();
+       resched = ftrace_preempt_disable();
+       cpu = raw_smp_processor_id();
+       data = tr->data[cpu];
+
+       if (unlikely(atomic_read(&data->disabled)))
+               goto out;
+
+       /* Lockdep uses trace_printk for lock tracing */
+       local_irq_save(flags);
+       __raw_spin_lock(&trace_buf_lock);
+       len = vbin_printf(trace_buf, TRACE_BUF_SIZE, fmt, args);
+
+       if (len > TRACE_BUF_SIZE || len < 0)
+               goto out_unlock;
+
+       size = sizeof(*entry) + sizeof(u32) * len;
+       event = trace_buffer_lock_reserve(tr, TRACE_BPRINT, size, flags, pc);
+       if (!event)
+               goto out_unlock;
+       entry = ring_buffer_event_data(event);
+       entry->ip                       = ip;
+       entry->depth                    = depth;
+       entry->fmt                      = fmt;
+
+       memcpy(entry->buf, trace_buf, sizeof(u32) * len);
+       ring_buffer_unlock_commit(tr->buffer, event);
+
+out_unlock:
+       __raw_spin_unlock(&trace_buf_lock);
+       local_irq_restore(flags);
+
+out:
+       ftrace_preempt_enable(resched);
+       unpause_graph_tracing();
+
+       return len;
+}
+EXPORT_SYMBOL_GPL(trace_vbprintk);
+
+int trace_vprintk(unsigned long ip, int depth, const char *fmt, va_list args)
+{
+       static raw_spinlock_t trace_buf_lock = __RAW_SPIN_LOCK_UNLOCKED;
+       static char trace_buf[TRACE_BUF_SIZE];
+
+       struct ring_buffer_event *event;
+       struct trace_array *tr = &global_trace;
+       struct trace_array_cpu *data;
+       int cpu, len = 0, size, pc;
+       struct print_entry *entry;
+       unsigned long irq_flags;
+
+       if (tracing_disabled || tracing_selftest_running)
+               return 0;
+
+       pc = preempt_count();
+       preempt_disable_notrace();
+       cpu = raw_smp_processor_id();
+       data = tr->data[cpu];
+
+       if (unlikely(atomic_read(&data->disabled)))
+               goto out;
+
+       pause_graph_tracing();
+       raw_local_irq_save(irq_flags);
+       __raw_spin_lock(&trace_buf_lock);
+       len = vsnprintf(trace_buf, TRACE_BUF_SIZE, fmt, args);
+
+       len = min(len, TRACE_BUF_SIZE-1);
+       trace_buf[len] = 0;
+
+       size = sizeof(*entry) + len + 1;
+       event = trace_buffer_lock_reserve(tr, TRACE_PRINT, size, irq_flags, pc);
+       if (!event)
+               goto out_unlock;
+       entry = ring_buffer_event_data(event);
+       entry->ip                       = ip;
+       entry->depth                    = depth;
+
+       memcpy(&entry->buf, trace_buf, len);
+       entry->buf[len] = 0;
+       ring_buffer_unlock_commit(tr->buffer, event);
+
+ out_unlock:
+       __raw_spin_unlock(&trace_buf_lock);
+       raw_local_irq_restore(irq_flags);
+       unpause_graph_tracing();
+ out:
+       preempt_enable_notrace();
+
+       return len;
+}
+EXPORT_SYMBOL_GPL(trace_vprintk);
+
 enum trace_file_type {
        TRACE_FILE_LAT_FMT      = 1,
        TRACE_FILE_ANNOTATE     = 2,
@@ -1393,11 +1543,11 @@ print_trace_header(struct seq_file *m, struct trace_iterator *iter)
        total = entries +
                ring_buffer_overruns(iter->tr->buffer);
 
-       seq_printf(m, "%s latency trace v1.1.5 on %s\n",
+       seq_printf(m, "%s latency trace v1.1.5 on %s\n",
                   name, UTS_RELEASE);
-       seq_puts(m, "-----------------------------------"
+       seq_puts(m, "-----------------------------------"
                 "---------------------------------\n");
-       seq_printf(m, " latency: %lu us, #%lu/%lu, CPU#%d |"
+       seq_printf(m, "# latency: %lu us, #%lu/%lu, CPU#%d |"
                   " (M:%s VP:%d, KP:%d, SP:%d HP:%d",
                   nsecs_to_usecs(data->saved_latency),
                   entries,
@@ -1419,24 +1569,24 @@ print_trace_header(struct seq_file *m, struct trace_iterator *iter)
 #else
        seq_puts(m, ")\n");
 #endif
-       seq_puts(m, "    -----------------\n");
-       seq_printf(m, "    | task: %.16s-%d "
+       seq_puts(m, "#    -----------------\n");
+       seq_printf(m, "#    | task: %.16s-%d "
                   "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n",
                   data->comm, data->pid, data->uid, data->nice,
                   data->policy, data->rt_priority);
-       seq_puts(m, "    -----------------\n");
+       seq_puts(m, "#    -----------------\n");
 
        if (data->critical_start) {
-               seq_puts(m, " => started at: ");
+               seq_puts(m, " => started at: ");
                seq_print_ip_sym(&iter->seq, data->critical_start, sym_flags);
                trace_print_seq(m, &iter->seq);
-               seq_puts(m, "\n => ended at:   ");
+               seq_puts(m, "\n => ended at:   ");
                seq_print_ip_sym(&iter->seq, data->critical_end, sym_flags);
                trace_print_seq(m, &iter->seq);
-               seq_puts(m, "\n");
+               seq_puts(m, "#\n");
        }
 
-       seq_puts(m, "\n");
+       seq_puts(m, "#\n");
 }
 
 static void test_cpu_buff_start(struct trace_iterator *iter)
@@ -1456,33 +1606,6 @@ static void test_cpu_buff_start(struct trace_iterator *iter)
        trace_seq_printf(s, "##### CPU %u buffer started ####\n", iter->cpu);
 }
 
-static enum print_line_t print_lat_fmt(struct trace_iterator *iter)
-{
-       struct trace_seq *s = &iter->seq;
-       unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
-       struct trace_event *event;
-       struct trace_entry *entry = iter->ent;
-
-       test_cpu_buff_start(iter);
-
-       event = ftrace_find_event(entry->type);
-
-       if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
-               if (!trace_print_lat_context(iter))
-                       goto partial;
-       }
-
-       if (event)
-               return event->latency_trace(iter, sym_flags);
-
-       if (!trace_seq_printf(s, "Unknown type %d\n", entry->type))
-               goto partial;
-
-       return TRACE_TYPE_HANDLED;
-partial:
-       return TRACE_TYPE_PARTIAL_LINE;
-}
-
 static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
 {
        struct trace_seq *s = &iter->seq;
@@ -1497,8 +1620,13 @@ static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
        event = ftrace_find_event(entry->type);
 
        if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
-               if (!trace_print_context(iter))
-                       goto partial;
+               if (iter->iter_flags & TRACE_FILE_LAT_FMT) {
+                       if (!trace_print_lat_context(iter))
+                               goto partial;
+               } else {
+                       if (!trace_print_context(iter))
+                               goto partial;
+               }
        }
 
        if (event)
@@ -1565,6 +1693,22 @@ static enum print_line_t print_hex_fmt(struct trace_iterator *iter)
        return TRACE_TYPE_HANDLED;
 }
 
+static enum print_line_t print_bprintk_msg_only(struct trace_iterator *iter)
+{
+       struct trace_seq *s = &iter->seq;
+       struct trace_entry *entry = iter->ent;
+       struct bprint_entry *field;
+       int ret;
+
+       trace_assign_type(field, entry);
+
+       ret = trace_seq_bprintf(s, field->fmt, field->buf);
+       if (!ret)
+               return TRACE_TYPE_PARTIAL_LINE;
+
+       return TRACE_TYPE_HANDLED;
+}
+
 static enum print_line_t print_printk_msg_only(struct trace_iterator *iter)
 {
        struct trace_seq *s = &iter->seq;
@@ -1603,6 +1747,19 @@ static int trace_empty(struct trace_iterator *iter)
 {
        int cpu;
 
+       /* If we are looking at one CPU buffer, only check that one */
+       if (iter->cpu_file != TRACE_PIPE_ALL_CPU) {
+               cpu = iter->cpu_file;
+               if (iter->buffer_iter[cpu]) {
+                       if (!ring_buffer_iter_empty(iter->buffer_iter[cpu]))
+                               return 0;
+               } else {
+                       if (!ring_buffer_empty_cpu(iter->tr->buffer, cpu))
+                               return 0;
+               }
+               return 1;
+       }
+
        for_each_tracing_cpu(cpu) {
                if (iter->buffer_iter[cpu]) {
                        if (!ring_buffer_iter_empty(iter->buffer_iter[cpu]))
@@ -1626,6 +1783,11 @@ static enum print_line_t print_trace_line(struct trace_iterator *iter)
                        return ret;
        }
 
+       if (iter->ent->type == TRACE_BPRINT &&
+                       trace_flags & TRACE_ITER_PRINTK &&
+                       trace_flags & TRACE_ITER_PRINTK_MSGONLY)
+               return print_bprintk_msg_only(iter);
+
        if (iter->ent->type == TRACE_PRINT &&
                        trace_flags & TRACE_ITER_PRINTK &&
                        trace_flags & TRACE_ITER_PRINTK_MSGONLY)
@@ -1640,9 +1802,6 @@ static enum print_line_t print_trace_line(struct trace_iterator *iter)
        if (trace_flags & TRACE_ITER_RAW)
                return print_raw_fmt(iter);
 
-       if (iter->iter_flags & TRACE_FILE_LAT_FMT)
-               return print_lat_fmt(iter);
-
        return print_trace_fmt(iter);
 }
 
@@ -1684,23 +1843,20 @@ static struct seq_operations tracer_seq_ops = {
 };
 
 static struct trace_iterator *
-__tracing_open(struct inode *inode, struct file *file, int *ret)
+__tracing_open(struct inode *inode, struct file *file)
 {
        long cpu_file = (long) inode->i_private;
+       void *fail_ret = ERR_PTR(-ENOMEM);
        struct trace_iterator *iter;
        struct seq_file *m;
-       int cpu;
+       int cpu, ret;
 
-       if (tracing_disabled) {
-               *ret = -ENODEV;
-               return NULL;
-       }
+       if (tracing_disabled)
+               return ERR_PTR(-ENODEV);
 
        iter = kzalloc(sizeof(*iter), GFP_KERNEL);
-       if (!iter) {
-               *ret = -ENOMEM;
-               goto out;
-       }
+       if (!iter)
+               return ERR_PTR(-ENOMEM);
 
        /*
         * We make a copy of the current tracer to avoid concurrent
@@ -1708,10 +1864,9 @@ __tracing_open(struct inode *inode, struct file *file, int *ret)
         */
        mutex_lock(&trace_types_lock);
        iter->trace = kzalloc(sizeof(*iter->trace), GFP_KERNEL);
-       if (!iter->trace) {
-               *ret = -ENOMEM;
+       if (!iter->trace)
                goto fail;
-       }
+
        if (current_trace)
                *iter->trace = *current_trace;
 
@@ -1736,23 +1891,19 @@ __tracing_open(struct inode *inode, struct file *file, int *ret)
 
                        iter->buffer_iter[cpu] =
                                ring_buffer_read_start(iter->tr->buffer, cpu);
-
-                       if (!iter->buffer_iter[cpu])
-                               goto fail_buffer;
                }
        } else {
                cpu = iter->cpu_file;
                iter->buffer_iter[cpu] =
                                ring_buffer_read_start(iter->tr->buffer, cpu);
-
-               if (!iter->buffer_iter[cpu])
-                       goto fail;
        }
 
        /* TODO stop tracer */
-       *ret = seq_open(file, &tracer_seq_ops);
-       if (*ret)
+       ret = seq_open(file, &tracer_seq_ops);
+       if (ret < 0) {
+               fail_ret = ERR_PTR(ret);
                goto fail_buffer;
+       }
 
        m = file->private_data;
        m->private = iter;
@@ -1762,7 +1913,6 @@ __tracing_open(struct inode *inode, struct file *file, int *ret)
 
        mutex_unlock(&trace_types_lock);
 
- out:
        return iter;
 
  fail_buffer:
@@ -1775,7 +1925,7 @@ __tracing_open(struct inode *inode, struct file *file, int *ret)
        kfree(iter->trace);
        kfree(iter);
 
-       return ERR_PTR(-ENOMEM);
+       return fail_ret;
 }
 
 int tracing_open_generic(struct inode *inode, struct file *filp)
@@ -1814,28 +1964,19 @@ static int tracing_release(struct inode *inode, struct file *file)
 }
 
 static int tracing_open(struct inode *inode, struct file *file)
-{
-       int ret;
-
-       __tracing_open(inode, file, &ret);
-
-       return ret;
-}
-
-static int tracing_lt_open(struct inode *inode, struct file *file)
 {
        struct trace_iterator *iter;
-       int ret;
-
-       iter = __tracing_open(inode, file, &ret);
+       int ret = 0;
 
-       if (!ret)
+       iter = __tracing_open(inode, file);
+       if (IS_ERR(iter))
+               ret = PTR_ERR(iter);
+       else if (trace_flags & TRACE_ITER_LATENCY_FMT)
                iter->iter_flags |= TRACE_FILE_LAT_FMT;
 
        return ret;
 }
 
-
 static void *
 t_next(struct seq_file *m, void *v, loff_t *pos)
 {
@@ -1907,21 +2048,14 @@ static int show_traces_open(struct inode *inode, struct file *file)
        return ret;
 }
 
-static struct file_operations tracing_fops = {
+static const struct file_operations tracing_fops = {
        .open           = tracing_open,
        .read           = seq_read,
        .llseek         = seq_lseek,
        .release        = tracing_release,
 };
 
-static struct file_operations tracing_lt_fops = {
-       .open           = tracing_lt_open,
-       .read           = seq_read,
-       .llseek         = seq_lseek,
-       .release        = tracing_release,
-};
-
-static struct file_operations show_traces_fops = {
+static const struct file_operations show_traces_fops = {
        .open           = show_traces_open,
        .read           = seq_read,
        .release        = seq_release,
@@ -2014,7 +2148,7 @@ err_unlock:
        return err;
 }
 
-static struct file_operations tracing_cpumask_fops = {
+static const struct file_operations tracing_cpumask_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_cpumask_read,
        .write          = tracing_cpumask_write,
@@ -2024,57 +2158,62 @@ static ssize_t
 tracing_trace_options_read(struct file *filp, char __user *ubuf,
                       size_t cnt, loff_t *ppos)
 {
-       int i;
+       struct tracer_opt *trace_opts;
+       u32 tracer_flags;
+       int len = 0;
        char *buf;
        int r = 0;
-       int len = 0;
-       u32 tracer_flags = current_trace->flags->val;
-       struct tracer_opt *trace_opts = current_trace->flags->opts;
+       int i;
 
 
        /* calculate max size */
        for (i = 0; trace_options[i]; i++) {
                len += strlen(trace_options[i]);
-               len += 3; /* "no" and space */
+               len += 3; /* "no" and newline */
        }
 
+       mutex_lock(&trace_types_lock);
+       tracer_flags = current_trace->flags->val;
+       trace_opts = current_trace->flags->opts;
+
        /*
         * Increase the size with names of options specific
         * of the current tracer.
         */
        for (i = 0; trace_opts[i].name; i++) {
                len += strlen(trace_opts[i].name);
-               len += 3; /* "no" and space */
+               len += 3; /* "no" and newline */
        }
 
        /* +2 for \n and \0 */
        buf = kmalloc(len + 2, GFP_KERNEL);
-       if (!buf)
+       if (!buf) {
+               mutex_unlock(&trace_types_lock);
                return -ENOMEM;
+       }
 
        for (i = 0; trace_options[i]; i++) {
                if (trace_flags & (1 << i))
-                       r += sprintf(buf + r, "%s ", trace_options[i]);
+                       r += sprintf(buf + r, "%s\n", trace_options[i]);
                else
-                       r += sprintf(buf + r, "no%s ", trace_options[i]);
+                       r += sprintf(buf + r, "no%s\n", trace_options[i]);
        }
 
        for (i = 0; trace_opts[i].name; i++) {
                if (tracer_flags & trace_opts[i].bit)
-                       r += sprintf(buf + r, "%s ",
+                       r += sprintf(buf + r, "%s\n",
                                trace_opts[i].name);
                else
-                       r += sprintf(buf + r, "no%s ",
+                       r += sprintf(buf + r, "no%s\n",
                                trace_opts[i].name);
        }
+       mutex_unlock(&trace_types_lock);
 
-       r += sprintf(buf + r, "\n");
        WARN_ON(r >= len + 2);
 
        r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
 
        kfree(buf);
-
        return r;
 }
 
@@ -2149,7 +2288,9 @@ tracing_trace_options_write(struct file *filp, const char __user *ubuf,
 
        /* If no option could be set, test the specific tracer options */
        if (!trace_options[i]) {
+               mutex_lock(&trace_types_lock);
                ret = set_tracer_option(current_trace, cmp, neg);
+               mutex_unlock(&trace_types_lock);
                if (ret)
                        return ret;
        }
@@ -2159,7 +2300,7 @@ tracing_trace_options_write(struct file *filp, const char __user *ubuf,
        return cnt;
 }
 
-static struct file_operations tracing_iter_fops = {
+static const struct file_operations tracing_iter_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_trace_options_read,
        .write          = tracing_trace_options_write,
@@ -2192,7 +2333,7 @@ tracing_readme_read(struct file *filp, char __user *ubuf,
                                        readme_msg, strlen(readme_msg));
 }
 
-static struct file_operations tracing_readme_fops = {
+static const struct file_operations tracing_readme_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_readme_read,
 };
@@ -2275,29 +2416,120 @@ int tracer_init(struct tracer *t, struct trace_array *tr)
        return t->init(tr);
 }
 
-static int tracing_set_tracer(const char *buf)
+static int tracing_resize_ring_buffer(unsigned long size)
 {
-       struct trace_array *tr = &global_trace;
-       struct tracer *t;
-       int ret = 0;
+       int ret;
 
-       mutex_lock(&trace_types_lock);
-       for (t = trace_types; t; t = t->next) {
-               if (strcmp(t->name, buf) == 0)
-                       break;
-       }
-       if (!t) {
-               ret = -EINVAL;
-               goto out;
+       /*
+        * If kernel or user changes the size of the ring buffer
+        * we use the size that was given, and we can forget about
+        * expanding it later.
+        */
+       ring_buffer_expanded = 1;
+
+       ret = ring_buffer_resize(global_trace.buffer, size);
+       if (ret < 0)
+               return ret;
+
+       ret = ring_buffer_resize(max_tr.buffer, size);
+       if (ret < 0) {
+               int r;
+
+               r = ring_buffer_resize(global_trace.buffer,
+                                      global_trace.entries);
+               if (r < 0) {
+                       /*
+                        * AARGH! We are left with different
+                        * size max buffer!!!!
+                        * The max buffer is our "snapshot" buffer.
+                        * When a tracer needs a snapshot (one of the
+                        * latency tracers), it swaps the max buffer
+                        * with the saved snap shot. We succeeded to
+                        * update the size of the main buffer, but failed to
+                        * update the size of the max buffer. But when we tried
+                        * to reset the main buffer to the original size, we
+                        * failed there too. This is very unlikely to
+                        * happen, but if it does, warn and kill all
+                        * tracing.
+                        */
+                       WARN_ON(1);
+                       tracing_disabled = 1;
+               }
+               return ret;
        }
-       if (t == current_trace)
-               goto out;
 
-       trace_branch_disable();
-       if (current_trace && current_trace->reset)
-               current_trace->reset(tr);
+       global_trace.entries = size;
+
+       return ret;
+}
+
+/**
+ * tracing_update_buffers - used by tracing facility to expand ring buffers
+ *
+ * To save on memory when the tracing is never used on a system with it
+ * configured in. The ring buffers are set to a minimum size. But once
+ * a user starts to use the tracing facility, then they need to grow
+ * to their default size.
+ *
+ * This function is to be called when a tracer is about to be used.
+ */
+int tracing_update_buffers(void)
+{
+       int ret = 0;
+
+       mutex_lock(&trace_types_lock);
+       if (!ring_buffer_expanded)
+               ret = tracing_resize_ring_buffer(trace_buf_size);
+       mutex_unlock(&trace_types_lock);
+
+       return ret;
+}
+
+struct trace_option_dentry;
+
+static struct trace_option_dentry *
+create_trace_option_files(struct tracer *tracer);
+
+static void
+destroy_trace_option_files(struct trace_option_dentry *topts);
+
+static int tracing_set_tracer(const char *buf)
+{
+       static struct trace_option_dentry *topts;
+       struct trace_array *tr = &global_trace;
+       struct tracer *t;
+       int ret = 0;
+
+       mutex_lock(&trace_types_lock);
+
+       if (!ring_buffer_expanded) {
+               ret = tracing_resize_ring_buffer(trace_buf_size);
+               if (ret < 0)
+                       goto out;
+               ret = 0;
+       }
+
+       for (t = trace_types; t; t = t->next) {
+               if (strcmp(t->name, buf) == 0)
+                       break;
+       }
+       if (!t) {
+               ret = -EINVAL;
+               goto out;
+       }
+       if (t == current_trace)
+               goto out;
+
+       trace_branch_disable();
+       if (current_trace && current_trace->reset)
+               current_trace->reset(tr);
+
+       destroy_trace_option_files(topts);
+
+       current_trace = t;
+
+       topts = create_trace_option_files(current_trace);
 
-       current_trace = t;
        if (t->init) {
                ret = tracer_init(t, tr);
                if (ret)
@@ -2583,7 +2815,7 @@ tracing_read_pipe(struct file *filp, char __user *ubuf,
        if (sret != -EBUSY)
                return sret;
 
-       trace_seq_reset(&iter->seq);
+       trace_seq_init(&iter->seq);
 
        /* copy the tracer to avoid using a global lock all around */
        mutex_lock(&trace_types_lock);
@@ -2645,7 +2877,7 @@ waitagain:
        /* Now copy what we have to the user */
        sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
        if (iter->seq.readpos >= iter->seq.len)
-               trace_seq_reset(&iter->seq);
+               trace_seq_init(&iter->seq);
 
        /*
         * If there was nothing to send to user, inspite of consuming trace
@@ -2782,7 +3014,7 @@ static ssize_t tracing_splice_read_pipe(struct file *filp,
                partial[i].offset = 0;
                partial[i].len = iter->seq.len;
 
-               trace_seq_reset(&iter->seq);
+               trace_seq_init(&iter->seq);
        }
 
        mutex_unlock(&iter->mutex);
@@ -2802,10 +3034,18 @@ tracing_entries_read(struct file *filp, char __user *ubuf,
                     size_t cnt, loff_t *ppos)
 {
        struct trace_array *tr = filp->private_data;
-       char buf[64];
+       char buf[96];
        int r;
 
-       r = sprintf(buf, "%lu\n", tr->entries >> 10);
+       mutex_lock(&trace_types_lock);
+       if (!ring_buffer_expanded)
+               r = sprintf(buf, "%lu (expanded: %lu)\n",
+                           tr->entries >> 10,
+                           trace_buf_size >> 10);
+       else
+               r = sprintf(buf, "%lu\n", tr->entries >> 10);
+       mutex_unlock(&trace_types_lock);
+
        return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
 }
 
@@ -2849,28 +3089,11 @@ tracing_entries_write(struct file *filp, const char __user *ubuf,
        val <<= 10;
 
        if (val != global_trace.entries) {
-               ret = ring_buffer_resize(global_trace.buffer, val);
+               ret = tracing_resize_ring_buffer(val);
                if (ret < 0) {
                        cnt = ret;
                        goto out;
                }
-
-               ret = ring_buffer_resize(max_tr.buffer, val);
-               if (ret < 0) {
-                       int r;
-                       cnt = ret;
-                       r = ring_buffer_resize(global_trace.buffer,
-                                              global_trace.entries);
-                       if (r < 0) {
-                               /* AARGH! We are left with different
-                                * size max buffer!!!! */
-                               WARN_ON(1);
-                               tracing_disabled = 1;
-                       }
-                       goto out;
-               }
-
-               global_trace.entries = val;
        }
 
        filp->f_pos += cnt;
@@ -2938,25 +3161,25 @@ tracing_mark_write(struct file *filp, const char __user *ubuf,
        return cnt;
 }
 
-static struct file_operations tracing_max_lat_fops = {
+static const struct file_operations tracing_max_lat_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_max_lat_read,
        .write          = tracing_max_lat_write,
 };
 
-static struct file_operations tracing_ctrl_fops = {
+static const struct file_operations tracing_ctrl_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_ctrl_read,
        .write          = tracing_ctrl_write,
 };
 
-static struct file_operations set_tracer_fops = {
+static const struct file_operations set_tracer_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_set_trace_read,
        .write          = tracing_set_trace_write,
 };
 
-static struct file_operations tracing_pipe_fops = {
+static const struct file_operations tracing_pipe_fops = {
        .open           = tracing_open_pipe,
        .poll           = tracing_poll_pipe,
        .read           = tracing_read_pipe,
@@ -2964,17 +3187,262 @@ static struct file_operations tracing_pipe_fops = {
        .release        = tracing_release_pipe,
 };
 
-static struct file_operations tracing_entries_fops = {
+static const struct file_operations tracing_entries_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_entries_read,
        .write          = tracing_entries_write,
 };
 
-static struct file_operations tracing_mark_fops = {
+static const struct file_operations tracing_mark_fops = {
        .open           = tracing_open_generic,
        .write          = tracing_mark_write,
 };
 
+struct ftrace_buffer_info {
+       struct trace_array      *tr;
+       void                    *spare;
+       int                     cpu;
+       unsigned int            read;
+};
+
+static int tracing_buffers_open(struct inode *inode, struct file *filp)
+{
+       int cpu = (int)(long)inode->i_private;
+       struct ftrace_buffer_info *info;
+
+       if (tracing_disabled)
+               return -ENODEV;
+
+       info = kzalloc(sizeof(*info), GFP_KERNEL);
+       if (!info)
+               return -ENOMEM;
+
+       info->tr        = &global_trace;
+       info->cpu       = cpu;
+       info->spare     = ring_buffer_alloc_read_page(info->tr->buffer);
+       /* Force reading ring buffer for first read */
+       info->read      = (unsigned int)-1;
+       if (!info->spare)
+               goto out;
+
+       filp->private_data = info;
+
+       return 0;
+
+ out:
+       kfree(info);
+       return -ENOMEM;
+}
+
+static ssize_t
+tracing_buffers_read(struct file *filp, char __user *ubuf,
+                    size_t count, loff_t *ppos)
+{
+       struct ftrace_buffer_info *info = filp->private_data;
+       unsigned int pos;
+       ssize_t ret;
+       size_t size;
+
+       if (!count)
+               return 0;
+
+       /* Do we have previous read data to read? */
+       if (info->read < PAGE_SIZE)
+               goto read;
+
+       info->read = 0;
+
+       ret = ring_buffer_read_page(info->tr->buffer,
+                                   &info->spare,
+                                   count,
+                                   info->cpu, 0);
+       if (ret < 0)
+               return 0;
+
+       pos = ring_buffer_page_len(info->spare);
+
+       if (pos < PAGE_SIZE)
+               memset(info->spare + pos, 0, PAGE_SIZE - pos);
+
+read:
+       size = PAGE_SIZE - info->read;
+       if (size > count)
+               size = count;
+
+       ret = copy_to_user(ubuf, info->spare + info->read, size);
+       if (ret == size)
+               return -EFAULT;
+       size -= ret;
+
+       *ppos += size;
+       info->read += size;
+
+       return size;
+}
+
+static int tracing_buffers_release(struct inode *inode, struct file *file)
+{
+       struct ftrace_buffer_info *info = file->private_data;
+
+       ring_buffer_free_read_page(info->tr->buffer, info->spare);
+       kfree(info);
+
+       return 0;
+}
+
+struct buffer_ref {
+       struct ring_buffer      *buffer;
+       void                    *page;
+       int                     ref;
+};
+
+static void buffer_pipe_buf_release(struct pipe_inode_info *pipe,
+                                   struct pipe_buffer *buf)
+{
+       struct buffer_ref *ref = (struct buffer_ref *)buf->private;
+
+       if (--ref->ref)
+               return;
+
+       ring_buffer_free_read_page(ref->buffer, ref->page);
+       kfree(ref);
+       buf->private = 0;
+}
+
+static int buffer_pipe_buf_steal(struct pipe_inode_info *pipe,
+                                struct pipe_buffer *buf)
+{
+       return 1;
+}
+
+static void buffer_pipe_buf_get(struct pipe_inode_info *pipe,
+                               struct pipe_buffer *buf)
+{
+       struct buffer_ref *ref = (struct buffer_ref *)buf->private;
+
+       ref->ref++;
+}
+
+/* Pipe buffer operations for a buffer. */
+static struct pipe_buf_operations buffer_pipe_buf_ops = {
+       .can_merge              = 0,
+       .map                    = generic_pipe_buf_map,
+       .unmap                  = generic_pipe_buf_unmap,
+       .confirm                = generic_pipe_buf_confirm,
+       .release                = buffer_pipe_buf_release,
+       .steal                  = buffer_pipe_buf_steal,
+       .get                    = buffer_pipe_buf_get,
+};
+
+/*
+ * Callback from splice_to_pipe(), if we need to release some pages
+ * at the end of the spd in case we error'ed out in filling the pipe.
+ */
+static void buffer_spd_release(struct splice_pipe_desc *spd, unsigned int i)
+{
+       struct buffer_ref *ref =
+               (struct buffer_ref *)spd->partial[i].private;
+
+       if (--ref->ref)
+               return;
+
+       ring_buffer_free_read_page(ref->buffer, ref->page);
+       kfree(ref);
+       spd->partial[i].private = 0;
+}
+
+static ssize_t
+tracing_buffers_splice_read(struct file *file, loff_t *ppos,
+                           struct pipe_inode_info *pipe, size_t len,
+                           unsigned int flags)
+{
+       struct ftrace_buffer_info *info = file->private_data;
+       struct partial_page partial[PIPE_BUFFERS];
+       struct page *pages[PIPE_BUFFERS];
+       struct splice_pipe_desc spd = {
+               .pages          = pages,
+               .partial        = partial,
+               .flags          = flags,
+               .ops            = &buffer_pipe_buf_ops,
+               .spd_release    = buffer_spd_release,
+       };
+       struct buffer_ref *ref;
+       int size, i;
+       size_t ret;
+
+       /*
+        * We can't seek on a buffer input
+        */
+       if (unlikely(*ppos))
+               return -ESPIPE;
+
+
+       for (i = 0; i < PIPE_BUFFERS && len; i++, len -= size) {
+               struct page *page;
+               int r;
+
+               ref = kzalloc(sizeof(*ref), GFP_KERNEL);
+               if (!ref)
+                       break;
+
+               ref->buffer = info->tr->buffer;
+               ref->page = ring_buffer_alloc_read_page(ref->buffer);
+               if (!ref->page) {
+                       kfree(ref);
+                       break;
+               }
+
+               r = ring_buffer_read_page(ref->buffer, &ref->page,
+                                         len, info->cpu, 0);
+               if (r < 0) {
+                       ring_buffer_free_read_page(ref->buffer,
+                                                  ref->page);
+                       kfree(ref);
+                       break;
+               }
+
+               /*
+                * zero out any left over data, this is going to
+                * user land.
+                */
+               size = ring_buffer_page_len(ref->page);
+               if (size < PAGE_SIZE)
+                       memset(ref->page + size, 0, PAGE_SIZE - size);
+
+               page = virt_to_page(ref->page);
+
+               spd.pages[i] = page;
+               spd.partial[i].len = PAGE_SIZE;
+               spd.partial[i].offset = 0;
+               spd.partial[i].private = (unsigned long)ref;
+               spd.nr_pages++;
+       }
+
+       spd.nr_pages = i;
+
+       /* did we read anything? */
+       if (!spd.nr_pages) {
+               if (flags & SPLICE_F_NONBLOCK)
+                       ret = -EAGAIN;
+               else
+                       ret = 0;
+               /* TODO: block */
+               return ret;
+       }
+
+       ret = splice_to_pipe(pipe, &spd);
+
+       return ret;
+}
+
+static const struct file_operations tracing_buffers_fops = {
+       .open           = tracing_buffers_open,
+       .read           = tracing_buffers_read,
+       .release        = tracing_buffers_release,
+       .splice_read    = tracing_buffers_splice_read,
+       .llseek         = no_llseek,
+};
+
 #ifdef CONFIG_DYNAMIC_FTRACE
 
 int __weak ftrace_arch_read_dyn_info(char *buf, int size)
@@ -3006,7 +3474,7 @@ tracing_read_dyn_info(struct file *filp, char __user *ubuf,
        return r;
 }
 
-static struct file_operations tracing_dyn_info_fops = {
+static const struct file_operations tracing_dyn_info_fops = {
        .open           = tracing_open_generic,
        .read           = tracing_read_dyn_info,
 };
@@ -3086,6 +3554,11 @@ static void tracing_init_debugfs_percpu(long cpu)
                                (void *) cpu, &tracing_fops);
        if (!entry)
                pr_warning("Could not create debugfs 'trace' entry\n");
+
+       entry = debugfs_create_file("trace_pipe_raw", 0444, d_cpu,
+                                   (void *) cpu, &tracing_buffers_fops);
+       if (!entry)
+               pr_warning("Could not create debugfs 'trace_pipe_raw' entry\n");
 }
 
 #ifdef CONFIG_FTRACE_SELFTEST
@@ -3093,6 +3566,279 @@ static void tracing_init_debugfs_percpu(long cpu)
 #include "trace_selftest.c"
 #endif
 
+struct trace_option_dentry {
+       struct tracer_opt               *opt;
+       struct tracer_flags             *flags;
+       struct dentry                   *entry;
+};
+
+static ssize_t
+trace_options_read(struct file *filp, char __user *ubuf, size_t cnt,
+                       loff_t *ppos)
+{
+       struct trace_option_dentry *topt = filp->private_data;
+       char *buf;
+
+       if (topt->flags->val & topt->opt->bit)
+               buf = "1\n";
+       else
+               buf = "0\n";
+
+       return simple_read_from_buffer(ubuf, cnt, ppos, buf, 2);
+}
+
+static ssize_t
+trace_options_write(struct file *filp, const char __user *ubuf, size_t cnt,
+                        loff_t *ppos)
+{
+       struct trace_option_dentry *topt = filp->private_data;
+       unsigned long val;
+       char buf[64];
+       int ret;
+
+       if (cnt >= sizeof(buf))
+               return -EINVAL;
+
+       if (copy_from_user(&buf, ubuf, cnt))
+               return -EFAULT;
+
+       buf[cnt] = 0;
+
+       ret = strict_strtoul(buf, 10, &val);
+       if (ret < 0)
+               return ret;
+
+       ret = 0;
+       switch (val) {
+       case 0:
+               /* do nothing if already cleared */
+               if (!(topt->flags->val & topt->opt->bit))
+                       break;
+
+               mutex_lock(&trace_types_lock);
+               if (current_trace->set_flag)
+                       ret = current_trace->set_flag(topt->flags->val,
+                                                     topt->opt->bit, 0);
+               mutex_unlock(&trace_types_lock);
+               if (ret)
+                       return ret;
+               topt->flags->val &= ~topt->opt->bit;
+               break;
+       case 1:
+               /* do nothing if already set */
+               if (topt->flags->val & topt->opt->bit)
+                       break;
+
+               mutex_lock(&trace_types_lock);
+               if (current_trace->set_flag)
+                       ret = current_trace->set_flag(topt->flags->val,
+                                                     topt->opt->bit, 1);
+               mutex_unlock(&trace_types_lock);
+               if (ret)
+                       return ret;
+               topt->flags->val |= topt->opt->bit;
+               break;
+
+       default:
+               return -EINVAL;
+       }
+
+       *ppos += cnt;
+
+       return cnt;
+}
+
+
+static const struct file_operations trace_options_fops = {
+       .open = tracing_open_generic,
+       .read = trace_options_read,
+       .write = trace_options_write,
+};
+
+static ssize_t
+trace_options_core_read(struct file *filp, char __user *ubuf, size_t cnt,
+                       loff_t *ppos)
+{
+       long index = (long)filp->private_data;
+       char *buf;
+
+       if (trace_flags & (1 << index))
+               buf = "1\n";
+       else
+               buf = "0\n";
+
+       return simple_read_from_buffer(ubuf, cnt, ppos, buf, 2);
+}
+
+static ssize_t
+trace_options_core_write(struct file *filp, const char __user *ubuf, size_t cnt,
+                        loff_t *ppos)
+{
+       long index = (long)filp->private_data;
+       char buf[64];
+       unsigned long val;
+       int ret;
+
+       if (cnt >= sizeof(buf))
+               return -EINVAL;
+
+       if (copy_from_user(&buf, ubuf, cnt))
+               return -EFAULT;
+
+       buf[cnt] = 0;
+
+       ret = strict_strtoul(buf, 10, &val);
+       if (ret < 0)
+               return ret;
+
+       switch (val) {
+       case 0:
+               trace_flags &= ~(1 << index);
+               break;
+       case 1:
+               trace_flags |= 1 << index;
+               break;
+
+       default:
+               return -EINVAL;
+       }
+
+       *ppos += cnt;
+
+       return cnt;
+}
+
+static const struct file_operations trace_options_core_fops = {
+       .open = tracing_open_generic,
+       .read = trace_options_core_read,
+       .write = trace_options_core_write,
+};
+
+static struct dentry *trace_options_init_dentry(void)
+{
+       struct dentry *d_tracer;
+       static struct dentry *t_options;
+
+       if (t_options)
+               return t_options;
+
+       d_tracer = tracing_init_dentry();
+       if (!d_tracer)
+               return NULL;
+
+       t_options = debugfs_create_dir("options", d_tracer);
+       if (!t_options) {
+               pr_warning("Could not create debugfs directory 'options'\n");
+               return NULL;
+       }
+
+       return t_options;
+}
+
+static void
+create_trace_option_file(struct trace_option_dentry *topt,
+                        struct tracer_flags *flags,
+                        struct tracer_opt *opt)
+{
+       struct dentry *t_options;
+       struct dentry *entry;
+
+       t_options = trace_options_init_dentry();
+       if (!t_options)
+               return;
+
+       topt->flags = flags;
+       topt->opt = opt;
+
+       entry = debugfs_create_file(opt->name, 0644, t_options, topt,
+                                   &trace_options_fops);
+
+       topt->entry = entry;
+
+}
+
+static struct trace_option_dentry *
+create_trace_option_files(struct tracer *tracer)
+{
+       struct trace_option_dentry *topts;
+       struct tracer_flags *flags;
+       struct tracer_opt *opts;
+       int cnt;
+
+       if (!tracer)
+               return NULL;
+
+       flags = tracer->flags;
+
+       if (!flags || !flags->opts)
+               return NULL;
+
+       opts = flags->opts;
+
+       for (cnt = 0; opts[cnt].name; cnt++)
+               ;
+
+       topts = kcalloc(cnt + 1, sizeof(*topts), GFP_KERNEL);
+       if (!topts)
+               return NULL;
+
+       for (cnt = 0; opts[cnt].name; cnt++)
+               create_trace_option_file(&topts[cnt], flags,
+                                        &opts[cnt]);
+
+       return topts;
+}
+
+static void
+destroy_trace_option_files(struct trace_option_dentry *topts)
+{
+       int cnt;
+
+       if (!topts)
+               return;
+
+       for (cnt = 0; topts[cnt].opt; cnt++) {
+               if (topts[cnt].entry)
+                       debugfs_remove(topts[cnt].entry);
+       }
+
+       kfree(topts);
+}
+
+static struct dentry *
+create_trace_option_core_file(const char *option, long index)
+{
+       struct dentry *t_options;
+       struct dentry *entry;
+
+       t_options = trace_options_init_dentry();
+       if (!t_options)
+               return NULL;
+
+       entry = debugfs_create_file(option, 0644, t_options, (void *)index,
+                                   &trace_options_core_fops);
+
+       return entry;
+}
+
+static __init void create_trace_options_dir(void)
+{
+       struct dentry *t_options;
+       struct dentry *entry;
+       int i;
+
+       t_options = trace_options_init_dentry();
+       if (!t_options)
+               return;
+
+       for (i = 0; trace_options[i]; i++) {
+               entry = create_trace_option_core_file(trace_options[i], i);
+               if (!entry)
+                       pr_warning("Could not create debugfs %s entry\n",
+                                  trace_options[i]);
+       }
+}
+
 static __init int tracer_init_debugfs(void)
 {
        struct dentry *d_tracer;
@@ -3111,6 +3857,8 @@ static __init int tracer_init_debugfs(void)
        if (!entry)
                pr_warning("Could not create debugfs 'trace_options' entry\n");
 
+       create_trace_options_dir();
+
        entry = debugfs_create_file("tracing_cpumask", 0644, d_tracer,
                                    NULL, &tracing_cpumask_fops);
        if (!entry)
@@ -3184,82 +3932,6 @@ static __init int tracer_init_debugfs(void)
        return 0;
 }
 
-int trace_vprintk(unsigned long ip, int depth, const char *fmt, va_list args)
-{
-       static DEFINE_SPINLOCK(trace_buf_lock);
-       static char trace_buf[TRACE_BUF_SIZE];
-
-       struct ring_buffer_event *event;
-       struct trace_array *tr = &global_trace;
-       struct trace_array_cpu *data;
-       int cpu, len = 0, size, pc;
-       struct print_entry *entry;
-       unsigned long irq_flags;
-
-       if (tracing_disabled || tracing_selftest_running)
-               return 0;
-
-       pc = preempt_count();
-       preempt_disable_notrace();
-       cpu = raw_smp_processor_id();
-       data = tr->data[cpu];
-
-       if (unlikely(atomic_read(&data->disabled)))
-               goto out;
-
-       pause_graph_tracing();
-       spin_lock_irqsave(&trace_buf_lock, irq_flags);
-       len = vsnprintf(trace_buf, TRACE_BUF_SIZE, fmt, args);
-
-       len = min(len, TRACE_BUF_SIZE-1);
-       trace_buf[len] = 0;
-
-       size = sizeof(*entry) + len + 1;
-       event = trace_buffer_lock_reserve(tr, TRACE_PRINT, size, irq_flags, pc);
-       if (!event)
-               goto out_unlock;
-       entry = ring_buffer_event_data(event);
-       entry->ip                       = ip;
-       entry->depth                    = depth;
-
-       memcpy(&entry->buf, trace_buf, len);
-       entry->buf[len] = 0;
-       ring_buffer_unlock_commit(tr->buffer, event);
-
- out_unlock:
-       spin_unlock_irqrestore(&trace_buf_lock, irq_flags);
-       unpause_graph_tracing();
- out:
-       preempt_enable_notrace();
-
-       return len;
-}
-EXPORT_SYMBOL_GPL(trace_vprintk);
-
-int __ftrace_printk(unsigned long ip, const char *fmt, ...)
-{
-       int ret;
-       va_list ap;
-
-       if (!(trace_flags & TRACE_ITER_PRINTK))
-               return 0;
-
-       va_start(ap, fmt);
-       ret = trace_vprintk(ip, task_curr_ret_stack(current), fmt, ap);
-       va_end(ap);
-       return ret;
-}
-EXPORT_SYMBOL_GPL(__ftrace_printk);
-
-int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap)
-{
-       if (!(trace_flags & TRACE_ITER_PRINTK))
-               return 0;
-
-       return trace_vprintk(ip, task_curr_ret_stack(current), fmt, ap);
-}
-EXPORT_SYMBOL_GPL(__ftrace_vprintk);
-
 static int trace_panic_handler(struct notifier_block *this,
                               unsigned long event, void *unused)
 {
@@ -3319,7 +3991,7 @@ trace_printk_seq(struct trace_seq *s)
 
        printk(KERN_TRACE "%s", s->buffer);
 
-       trace_seq_reset(s);
+       trace_seq_init(s);
 }
 
 void ftrace_dump(void)
@@ -3351,8 +4023,10 @@ void ftrace_dump(void)
 
        printk(KERN_TRACE "Dumping ftrace buffer:\n");
 
+       /* Simulate the iterator */
        iter.tr = &global_trace;
        iter.trace = current_trace;
+       iter.cpu_file = TRACE_PIPE_ALL_CPU;
 
        /*
         * We need to stop all tracing on all CPUS to read the
@@ -3395,6 +4069,7 @@ void ftrace_dump(void)
 __init static int tracer_alloc_buffers(void)
 {
        struct trace_array_cpu *data;
+       int ring_buf_size;
        int i;
        int ret = -ENOMEM;
 
@@ -3407,12 +4082,18 @@ __init static int tracer_alloc_buffers(void)
        if (!alloc_cpumask_var(&tracing_reader_cpumask, GFP_KERNEL))
                goto out_free_tracing_cpumask;
 
+       /* To save memory, keep the ring buffer size to its minimum */
+       if (ring_buffer_expanded)
+               ring_buf_size = trace_buf_size;
+       else
+               ring_buf_size = 1;
+
        cpumask_copy(tracing_buffer_mask, cpu_possible_mask);
        cpumask_copy(tracing_cpumask, cpu_all_mask);
        cpumask_clear(tracing_reader_cpumask);
 
        /* TODO: make the number of buffers hot pluggable with CPUS */
-       global_trace.buffer = ring_buffer_alloc(trace_buf_size,
+       global_trace.buffer = ring_buffer_alloc(ring_buf_size,
                                                   TRACE_BUFFER_FLAGS);
        if (!global_trace.buffer) {
                printk(KERN_ERR "tracer: failed to allocate ring buffer!\n");
@@ -3423,7 +4104,7 @@ __init static int tracer_alloc_buffers(void)
 
 
 #ifdef CONFIG_TRACER_MAX_TRACE
-       max_tr.buffer = ring_buffer_alloc(trace_buf_size,
+       max_tr.buffer = ring_buffer_alloc(ring_buf_size,
                                             TRACE_BUFFER_FLAGS);
        if (!max_tr.buffer) {
                printk(KERN_ERR "tracer: failed to allocate max ring buffer!\n");
@@ -3455,7 +4136,8 @@ __init static int tracer_alloc_buffers(void)
                                       &trace_panic_notifier);
 
        register_die_notifier(&trace_die_notifier);
-       ret = 0;
+
+       return 0;
 
 out_free_cpumask:
        free_cpumask_var(tracing_reader_cpumask);