Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

5 files changed, 1033 insertions, 0 deletions

diff --git a/samples/trace_events/Makefile b/samples/trace_events/Makefile
new file mode 100644
index 000000000..b3808bb4c
--- /dev/null
+++ b/samples/trace_events/Makefile
@@ -0,0 +1,17 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# builds the trace events example kernel modules;
+# then to use one (as root):  insmod <module_name.ko>
+
+# If you include a trace header outside of include/trace/events
+# then the file that does the #define CREATE_TRACE_POINTS must
+# have that tracer file in its main search path. This is because
+# define_trace.h will include it, and must be able to find it from
+# the include/trace directory.
+#
+# Here trace-events-sample.c does the CREATE_TRACE_POINTS.
+#
+CFLAGS_trace-events-sample.o := -I$(src)
+CFLAGS_trace_custom_sched.o := -I$(src)
+
+obj-$(CONFIG_SAMPLE_TRACE_EVENTS) += trace-events-sample.o
+obj-$(CONFIG_SAMPLE_TRACE_CUSTOM_EVENTS) += trace_custom_sched.o
diff --git a/samples/trace_events/trace-events-sample.c b/samples/trace_events/trace-events-sample.c
new file mode 100644
index 000000000..ca5d98c36
--- /dev/null
+++ b/samples/trace_events/trace-events-sample.c
@@ -0,0 +1,193 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/module.h>
+#include <linux/kthread.h>
+
+/*
+ * Any file that uses trace points, must include the header.
+ * But only one file, must include the header by defining
+ * CREATE_TRACE_POINTS first.  This will make the C code that
+ * creates the handles for the trace points.
+ */
+#define CREATE_TRACE_POINTS
+#include "trace-events-sample.h"
+
+static const char *random_strings[] = {
+	"Mother Goose",
+	"Snoopy",
+	"Gandalf",
+	"Frodo",
+	"One ring to rule them all"
+};
+
+static void do_simple_thread_func(int cnt, const char *fmt, ...)
+{
+	unsigned long bitmask[1] = {0xdeadbeefUL};
+	va_list va;
+	int array[6];
+	int len = cnt % 5;
+	int i;
+
+	set_current_state(TASK_INTERRUPTIBLE);
+	schedule_timeout(HZ);
+
+	for (i = 0; i < len; i++)
+		array[i] = i + 1;
+	array[i] = 0;
+
+	va_start(va, fmt);
+
+	/* Silly tracepoints */
+	trace_foo_bar("hello", cnt, array, random_strings[len],
+		      current->cpus_ptr, fmt, &va);
+
+	va_end(va);
+
+	trace_foo_with_template_simple("HELLO", cnt);
+
+	trace_foo_bar_with_cond("Some times print", cnt);
+
+	trace_foo_with_template_cond("prints other times", cnt);
+
+	trace_foo_with_template_print("I have to be different", cnt);
+
+	trace_foo_rel_loc("Hello __rel_loc", cnt, bitmask, current->cpus_ptr);
+}
+
+static void simple_thread_func(int cnt)
+{
+	do_simple_thread_func(cnt, "iter=%d", cnt);
+}
+
+static int simple_thread(void *arg)
+{
+	int cnt = 0;
+
+	while (!kthread_should_stop())
+		simple_thread_func(cnt++);
+
+	return 0;
+}
+
+static struct task_struct *simple_tsk;
+static struct task_struct *simple_tsk_fn;
+
+static void simple_thread_func_fn(int cnt)
+{
+	set_current_state(TASK_INTERRUPTIBLE);
+	schedule_timeout(HZ);
+
+	/* More silly tracepoints */
+	trace_foo_bar_with_fn("Look at me", cnt);
+	trace_foo_with_template_fn("Look at me too", cnt);
+}
+
+static int simple_thread_fn(void *arg)
+{
+	int cnt = 0;
+
+	while (!kthread_should_stop())
+		simple_thread_func_fn(cnt++);
+
+	return 0;
+}
+
+static DEFINE_MUTEX(thread_mutex);
+static int simple_thread_cnt;
+
+static struct foo_timer_data *foo_timer_data;
+
+static void sample_timer_cb(struct timer_list *t)
+{
+	struct foo_timer_data *data = container_of(t, struct foo_timer_data, timer);
+
+	get_cpu();
+	trace_foo_timer_fn(data);
+	(*this_cpu_ptr(data->counter))++;
+	put_cpu();
+
+	mod_timer(t, jiffies + HZ);
+}
+
+int foo_bar_reg(void)
+{
+	mutex_lock(&thread_mutex);
+	if (simple_thread_cnt++)
+		goto out;
+
+	pr_info("Starting thread for foo_bar_fn\n");
+	/*
+	 * We shouldn't be able to start a trace when the module is
+	 * unloading (there's other locks to prevent that). But
+	 * for consistency sake, we still take the thread_mutex.
+	 */
+	simple_tsk_fn = kthread_run(simple_thread_fn, NULL, "event-sample-fn");
+	if (IS_ERR_OR_NULL(simple_tsk_fn)) {
+		pr_err("Failed to create simple_thread_fn\n");
+		simple_tsk_fn = NULL;
+	}
+ out:
+	mutex_unlock(&thread_mutex);
+	return 0;
+}
+
+void foo_bar_unreg(void)
+{
+	mutex_lock(&thread_mutex);
+	if (--simple_thread_cnt)
+		goto out;
+
+	pr_info("Killing thread for foo_bar_fn\n");
+	if (simple_tsk_fn)
+		kthread_stop(simple_tsk_fn);
+	simple_tsk_fn = NULL;
+ out:
+	mutex_unlock(&thread_mutex);
+}
+
+static int __init trace_event_init(void)
+{
+	foo_timer_data = kzalloc_obj(*foo_timer_data, GFP_KERNEL);
+	if (!foo_timer_data)
+		return -ENOMEM;
+
+	foo_timer_data->name = "sample_timer_counter";
+	foo_timer_data->counter = alloc_percpu(int);
+	if (!foo_timer_data->counter) {
+		kfree(foo_timer_data);
+		return -ENOMEM;
+	}
+
+	timer_setup(&foo_timer_data->timer, sample_timer_cb, 0);
+	mod_timer(&foo_timer_data->timer, jiffies + HZ);
+
+	simple_tsk = kthread_run(simple_thread, NULL, "event-sample");
+	if (IS_ERR(simple_tsk)) {
+		timer_shutdown_sync(&foo_timer_data->timer);
+		free_percpu(foo_timer_data->counter);
+		kfree(foo_timer_data);
+		return PTR_ERR(simple_tsk);
+	}
+
+	return 0;
+}
+
+static void __exit trace_event_exit(void)
+{
+	kthread_stop(simple_tsk);
+	mutex_lock(&thread_mutex);
+	if (simple_tsk_fn)
+		kthread_stop(simple_tsk_fn);
+	simple_tsk_fn = NULL;
+	mutex_unlock(&thread_mutex);
+
+	timer_shutdown_sync(&foo_timer_data->timer);
+	free_percpu(foo_timer_data->counter);
+	kfree(foo_timer_data);
+}
+
+module_init(trace_event_init);
+module_exit(trace_event_exit);
+
+MODULE_AUTHOR("Steven Rostedt");
+MODULE_DESCRIPTION("trace-events-sample");
+MODULE_LICENSE("GPL");
diff --git a/samples/trace_events/trace-events-sample.h b/samples/trace_events/trace-events-sample.h
new file mode 100644
index 000000000..816848a45
--- /dev/null
+++ b/samples/trace_events/trace-events-sample.h
@@ -0,0 +1,668 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * If TRACE_SYSTEM is defined, that will be the directory created
+ * in the ftrace directory under /sys/kernel/tracing/events/<system>
+ *
+ * The define_trace.h below will also look for a file name of
+ * TRACE_SYSTEM.h where TRACE_SYSTEM is what is defined here.
+ * In this case, it would look for sample-trace.h
+ *
+ * If the header name will be different than the system name
+ * (as in this case), then you can override the header name that
+ * define_trace.h will look up by defining TRACE_INCLUDE_FILE
+ *
+ * This file is called trace-events-sample.h but we want the system
+ * to be called "sample-trace". Therefore we must define the name of this
+ * file:
+ *
+ * #define TRACE_INCLUDE_FILE trace-events-sample
+ *
+ * As we do an the bottom of this file.
+ *
+ * Notice that TRACE_SYSTEM should be defined outside of #if
+ * protection, just like TRACE_INCLUDE_FILE.
+ */
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM sample-trace
+
+/*
+ * TRACE_SYSTEM is expected to be a C valid variable (alpha-numeric
+ * and underscore), although it may start with numbers. If for some
+ * reason it is not, you need to add the following lines:
+ */
+#undef TRACE_SYSTEM_VAR
+#define TRACE_SYSTEM_VAR sample_trace
+/*
+ * But the above is only needed if TRACE_SYSTEM is not alpha-numeric
+ * and underscored. By default, TRACE_SYSTEM_VAR will be equal to
+ * TRACE_SYSTEM. As TRACE_SYSTEM_VAR must be alpha-numeric, if
+ * TRACE_SYSTEM is not, then TRACE_SYSTEM_VAR must be defined with
+ * only alpha-numeric and underscores.
+ *
+ * The TRACE_SYSTEM_VAR is only used internally and not visible to
+ * user space.
+ */
+
+/*
+ * Notice that this file is not protected like a normal header.
+ * We also must allow for rereading of this file. The
+ *
+ *  || defined(TRACE_HEADER_MULTI_READ)
+ *
+ * serves this purpose.
+ */
+#if !defined(_TRACE_EVENT_SAMPLE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_EVENT_SAMPLE_H
+
+/*
+ * All trace headers should include tracepoint.h, until we finally
+ * make it into a standard header.
+ */
+#include <linux/tracepoint.h>
+
+/*
+ * The TRACE_EVENT macro is broken up into 5 parts.
+ *
+ * name: name of the trace point. This is also how to enable the tracepoint.
+ *   A function called trace_foo_bar() will be created.
+ *
+ * proto: the prototype of the function trace_foo_bar()
+ *   Here it is trace_foo_bar(char *foo, int bar).
+ *
+ * args:  must match the arguments in the prototype.
+ *    Here it is simply "foo, bar".
+ *
+ * struct:  This defines the way the data will be stored in the ring buffer.
+ *          The items declared here become part of a special structure
+ *          called "__entry", which can be used in the fast_assign part of the
+ *          TRACE_EVENT macro.
+ *
+ *      Here are the currently defined types you can use:
+ *
+ *   __field : Is broken up into type and name. Where type can be any
+ *         primitive type (integer, long or pointer).
+ *
+ *        __field(int, foo)
+ *
+ *        __entry->foo = 5;
+ *
+ *   __field_struct : This can be any static complex data type (struct, union
+ *         but not an array). Be careful using complex types, as each
+ *         event is limited in size, and copying large amounts of data
+ *         into the ring buffer can slow things down.
+ *
+ *         __field_struct(struct bar, foo)
+ *
+ *         __entry->bar.x = y;
+
+ *   __array: There are three fields (type, name, size). The type is the
+ *         type of elements in the array, the name is the name of the array.
+ *         size is the number of items in the array (not the total size).
+ *
+ *         __array( char, foo, 10) is the same as saying: char foo[10];
+ *
+ *         Assigning arrays can be done like any array:
+ *
+ *         __entry->foo[0] = 'a';
+ *
+ *         memcpy(__entry->foo, bar, 10);
+ *
+ *   __dynamic_array: This is similar to array, but can vary its size from
+ *         instance to instance of the tracepoint being called.
+ *         Like __array, this too has three elements (type, name, size);
+ *         type is the type of the element, name is the name of the array.
+ *         The size is different than __array. It is not a static number,
+ *         but the algorithm to figure out the length of the array for the
+ *         specific instance of tracepoint. Again, size is the number of
+ *         items in the array, not the total length in bytes.
+ *
+ *         __dynamic_array( int, foo, bar) is similar to: int foo[bar];
+ *
+ *         Note, unlike arrays, you must use the __get_dynamic_array() macro
+ *         to access the array.
+ *
+ *         memcpy(__get_dynamic_array(foo), bar, 10);
+ *
+ *         Notice, that "__entry" is not needed here.
+ *
+ *   __string: This is a special kind of __dynamic_array. It expects to
+ *         have a null terminated character array passed to it (it allows
+ *         for NULL too, which would be converted into "(null)"). __string
+ *         takes two parameter (name, src), where name is the name of
+ *         the string saved, and src is the string to copy into the
+ *         ring buffer.
+ *
+ *         __string(foo, bar)  is similar to:  strcpy(foo, bar)
+ *
+ *         To assign a string, use the helper macro __assign_str().
+ *
+ *         __assign_str(foo);
+ *
+ *	   The __string() macro saves off the string that is passed into
+ *         the second parameter, and the __assign_str() will store than
+ *         saved string into the "foo" field.
+ *
+ *   __vstring: This is similar to __string() but instead of taking a
+ *         dynamic length, it takes a variable list va_list 'va' variable.
+ *         Some event callers already have a message from parameters saved
+ *         in a va_list. Passing in the format and the va_list variable
+ *         will save just enough on the ring buffer for that string.
+ *         Note, the va variable used is a pointer to a va_list, not
+ *         to the va_list directly.
+ *
+ *           (va_list *va)
+ *
+ *         __vstring(foo, fmt, va)  is similar to:  vsnprintf(foo, fmt, va)
+ *
+ *         To assign the string, use the helper macro __assign_vstr().
+ *
+ *         __assign_vstr(foo, fmt, va);
+ *
+ *         In most cases, the __assign_vstr() macro will take the same
+ *         parameters as the __vstring() macro had to declare the string.
+ *         Use __get_str() to retrieve the __vstring() just like it would for
+ *         __string().
+ *
+ *   __string_len: This is a helper to a __dynamic_array, but it understands
+ *	   that the array has characters in it, it will allocate 'len' + 1 bytes
+ *         in the ring buffer and add a '\0' to the string. This is
+ *         useful if the string being saved has no terminating '\0' byte.
+ *         It requires that the length of the string is known as it acts
+ *         like a memcpy().
+ *
+ *         Declared with:
+ *
+ *         __string_len(foo, bar, len)
+ *
+ *         To assign this string, use the helper macro __assign_str().
+ *         The length is saved via the __string_len() and is retrieved in
+ *         __assign_str().
+ *
+ *         __assign_str(foo);
+ *
+ *         Then len + 1 is allocated to the ring buffer, and a nul terminating
+ *         byte is added. This is similar to:
+ *
+ *         memcpy(__get_str(foo), bar, len);
+ *         __get_str(foo)[len] = 0;
+ *
+ *        The advantage of using this over __dynamic_array, is that it
+ *        takes care of allocating the extra byte on the ring buffer
+ *        for the '\0' terminating byte, and __get_str(foo) can be used
+ *        in the TP_printk().
+ *
+ *   __bitmask: This is another kind of __dynamic_array, but it expects
+ *         an array of longs, and the number of bits to parse. It takes
+ *         two parameters (name, nr_bits), where name is the name of the
+ *         bitmask to save, and the nr_bits is the number of bits to record.
+ *
+ *         __bitmask(target_cpu, nr_cpumask_bits)
+ *
+ *         To assign a bitmask, use the __assign_bitmask() helper macro.
+ *
+ *         __assign_bitmask(target_cpus, cpumask_bits(bar), nr_cpumask_bits);
+ *
+ *   __cpumask: This is pretty much the same as __bitmask but is specific for
+ *         CPU masks. The type displayed to the user via the format files will
+ *         be "cpumaks_t" such that user space may deal with them differently
+ *         if they choose to do so, and the bits is always set to nr_cpumask_bits.
+ *
+ *         __cpumask(target_cpu)
+ *
+ *         To assign a cpumask, use the __assign_cpumask() helper macro.
+ *
+ *         __assign_cpumask(target_cpus, cpumask_bits(bar));
+ *
+ * fast_assign: This is a C like function that is used to store the items
+ *    into the ring buffer. A special variable called "__entry" will be the
+ *    structure that points into the ring buffer and has the same fields as
+ *    described by the struct part of TRACE_EVENT above.
+ *
+ * printk: This is a way to print out the data in pretty print. This is
+ *    useful if the system crashes and you are logging via a serial line,
+ *    the data can be printed to the console using this "printk" method.
+ *    This is also used to print out the data from the trace files.
+ *    Again, the __entry macro is used to access the data from the ring buffer.
+ *
+ *    Note, __dynamic_array, __string, __bitmask and __cpumask require special
+ *       helpers to access the data.
+ *
+ *      For __dynamic_array(int, foo, bar) use __get_dynamic_array(foo)
+ *            Use __get_dynamic_array_len(foo) to get the length of the array
+ *            saved. Note, __get_dynamic_array_len() returns the total allocated
+ *            length of the dynamic array; __print_array() expects the second
+ *            parameter to be the number of elements. To get that, the array length
+ *            needs to be divided by the element size.
+ *
+ *      For __string(foo, bar) use __get_str(foo)
+ *
+ *      For __bitmask(target_cpus, nr_cpumask_bits) use __get_bitmask(target_cpus)
+ *
+ *      For __cpumask(target_cpus) use __get_cpumask(target_cpus)
+ *
+ *
+ * Note, that for both the assign and the printk, __entry is the handler
+ * to the data structure in the ring buffer, and is defined by the
+ * TP_STRUCT__entry.
+ */
+
+/*
+ * It is OK to have helper functions and data structures in the file, but they
+ * need to be protected from being defined more than once. Remember, this file
+ * gets included more than once.
+ */
+#ifndef __TRACE_EVENT_SAMPLE_HELPER_FUNCTIONS
+#define __TRACE_EVENT_SAMPLE_HELPER_FUNCTIONS
+#include <linux/timer.h>
+
+static inline int __length_of(const int *list)
+{
+	int i;
+
+	if (!list)
+		return 0;
+
+	for (i = 0; list[i]; i++)
+		;
+	return i;
+}
+
+enum {
+	TRACE_SAMPLE_FOO = 2,
+	TRACE_SAMPLE_BAR = 4,
+	TRACE_SAMPLE_ZOO = 8,
+};
+
+struct foo_timer_data {
+	const char		*name;
+	struct timer_list	timer;
+	int __percpu		*counter;
+};
+
+#endif
+
+/*
+ * If enums are used in the TP_printk(), their names will be shown in
+ * format files and not their values. This can cause problems with user
+ * space programs that parse the format files to know how to translate
+ * the raw binary trace output into human readable text.
+ *
+ * To help out user space programs, any enum that is used in the TP_printk()
+ * should be defined by TRACE_DEFINE_ENUM() macro. All that is needed to
+ * be done is to add this macro with the enum within it in the trace
+ * header file, and it will be converted in the output.
+ */
+
+TRACE_DEFINE_ENUM(TRACE_SAMPLE_FOO);
+TRACE_DEFINE_ENUM(TRACE_SAMPLE_BAR);
+TRACE_DEFINE_ENUM(TRACE_SAMPLE_ZOO);
+
+TRACE_EVENT(foo_bar,
+
+	TP_PROTO(const char *foo, int bar, const int *lst,
+		 const char *string, const struct cpumask *mask,
+		 const char *fmt, va_list *va),
+
+	TP_ARGS(foo, bar, lst, string, mask, fmt, va),
+
+	TP_STRUCT__entry(
+		__array(	char,	foo,    10		)
+		__field(	int,	bar			)
+		__dynamic_array(int,	list,   __length_of(lst))
+		__string(	str,	string			)
+		__bitmask(	cpus,	num_possible_cpus()	)
+		__cpumask(	cpum				)
+		__vstring(	vstr,	fmt,	va		)
+		__string_len(	lstr,	foo,	bar / 2 < strlen(foo) ? bar / 2 : strlen(foo) )
+	),
+
+	TP_fast_assign(
+		strscpy(__entry->foo, foo, 10);
+		__entry->bar	= bar;
+		memcpy(__get_dynamic_array(list), lst,
+		       __length_of(lst) * sizeof(int));
+		__assign_str(str);
+		__assign_str(lstr);
+		__assign_vstr(vstr, fmt, va);
+		__assign_bitmask(cpus, cpumask_bits(mask), num_possible_cpus());
+		__assign_cpumask(cpum, cpumask_bits(mask));
+	),
+
+	TP_printk("foo %s %d %s %s %s %s %s %s (%s) (%s) %s [%d] %*pbl",
+		  __entry->foo, __entry->bar,
+
+/*
+ * Notice here the use of some helper functions. This includes:
+ *
+ *  __print_symbolic( variable, { value, "string" }, ... ),
+ *
+ *    The variable is tested against each value of the { } pair. If
+ *    the variable matches one of the values, then it will print the
+ *    string in that pair. If non are matched, it returns a string
+ *    version of the number (if __entry->bar == 7 then "7" is returned).
+ */
+		  __print_symbolic(__entry->bar,
+				   { 0, "zero" },
+				   { TRACE_SAMPLE_FOO, "TWO" },
+				   { TRACE_SAMPLE_BAR, "FOUR" },
+				   { TRACE_SAMPLE_ZOO, "EIGHT" },
+				   { 10, "TEN" }
+			  ),
+
+/*
+ *  __print_flags( variable, "delim", { value, "flag" }, ... ),
+ *
+ *    This is similar to __print_symbolic, except that it tests the bits
+ *    of the value. If ((FLAG & variable) == FLAG) then the string is
+ *    printed. If more than one flag matches, then each one that does is
+ *    also printed with delim in between them.
+ *    If not all bits are accounted for, then the not found bits will be
+ *    added in hex format: 0x506 will show BIT2|BIT4|0x500
+ */
+		  __print_flags(__entry->bar, "|",
+				{ 1, "BIT1" },
+				{ 2, "BIT2" },
+				{ 4, "BIT3" },
+				{ 8, "BIT4" }
+			  ),
+/*
+ *  __print_array( array, len, element_size )
+ *
+ *    This prints out the array that is defined by __array in a nice format.
+ */
+		  __print_array(__get_dynamic_array(list),
+				__get_dynamic_array_len(list) / sizeof(int),
+				sizeof(int)),
+
+/*     A shortcut is to use __print_dynamic_array for dynamic arrays */
+
+		  __print_dynamic_array(list, sizeof(int)),
+
+		  __get_str(str), __get_str(lstr),
+		  __get_bitmask(cpus), __get_cpumask(cpum),
+		  __get_str(vstr),
+		  __get_dynamic_array_len(cpus),
+		  __get_dynamic_array_len(cpus),
+		  __get_dynamic_array(cpus))
+);
+
+/*
+ * There may be a case where a tracepoint should only be called if
+ * some condition is set. Otherwise the tracepoint should not be called.
+ * But to do something like:
+ *
+ *  if (cond)
+ *     trace_foo();
+ *
+ * Would cause a little overhead when tracing is not enabled, and that
+ * overhead, even if small, is not something we want. As tracepoints
+ * use static branch (aka jump_labels), where no branch is taken to
+ * skip the tracepoint when not enabled, and a jmp is placed to jump
+ * to the tracepoint code when it is enabled, having a if statement
+ * nullifies that optimization. It would be nice to place that
+ * condition within the static branch. This is where TRACE_EVENT_CONDITION
+ * comes in.
+ *
+ * TRACE_EVENT_CONDITION() is just like TRACE_EVENT, except it adds another
+ * parameter just after args. Where TRACE_EVENT has:
+ *
+ * TRACE_EVENT(name, proto, args, struct, assign, printk)
+ *
+ * the CONDITION version has:
+ *
+ * TRACE_EVENT_CONDITION(name, proto, args, cond, struct, assign, printk)
+ *
+ * Everything is the same as TRACE_EVENT except for the new cond. Think
+ * of the cond variable as:
+ *
+ *   if (cond)
+ *      trace_foo_bar_with_cond();
+ *
+ * Except that the logic for the if branch is placed after the static branch.
+ * That is, the if statement that processes the condition will not be
+ * executed unless that traecpoint is enabled. Otherwise it still remains
+ * a nop.
+ */
+TRACE_EVENT_CONDITION(foo_bar_with_cond,
+
+	TP_PROTO(const char *foo, int bar),
+
+	TP_ARGS(foo, bar),
+
+	TP_CONDITION(!(bar % 10)),
+
+	TP_STRUCT__entry(
+		__string(	foo,    foo		)
+		__field(	int,	bar			)
+	),
+
+	TP_fast_assign(
+		__assign_str(foo);
+		__entry->bar	= bar;
+	),
+
+	TP_printk("foo %s %d", __get_str(foo), __entry->bar)
+);
+
+int foo_bar_reg(void);
+void foo_bar_unreg(void);
+
+/*
+ * Now in the case that some function needs to be called when the
+ * tracepoint is enabled and/or when it is disabled, the
+ * TRACE_EVENT_FN() serves this purpose. This is just like TRACE_EVENT()
+ * but adds two more parameters at the end:
+ *
+ * TRACE_EVENT_FN( name, proto, args, struct, assign, printk, reg, unreg)
+ *
+ * reg and unreg are functions with the prototype of:
+ *
+ *    void reg(void)
+ *
+ * The reg function gets called before the tracepoint is enabled, and
+ * the unreg function gets called after the tracepoint is disabled.
+ *
+ * Note, reg and unreg are allowed to be NULL. If you only need to
+ * call a function before enabling, or after disabling, just set one
+ * function and pass in NULL for the other parameter.
+ */
+TRACE_EVENT_FN(foo_bar_with_fn,
+
+	TP_PROTO(const char *foo, int bar),
+
+	TP_ARGS(foo, bar),
+
+	TP_STRUCT__entry(
+		__string(	foo,    foo		)
+		__field(	int,	bar		)
+	),
+
+	TP_fast_assign(
+		__assign_str(foo);
+		__entry->bar	= bar;
+	),
+
+	TP_printk("foo %s %d", __get_str(foo), __entry->bar),
+
+	foo_bar_reg, foo_bar_unreg
+);
+
+/*
+ * Each TRACE_EVENT macro creates several helper functions to produce
+ * the code to add the tracepoint, create the files in the trace
+ * directory, hook it to perf, assign the values and to print out
+ * the raw data from the ring buffer. To prevent too much bloat,
+ * if there are more than one tracepoint that uses the same format
+ * for the proto, args, struct, assign and printk, and only the name
+ * is different, it is highly recommended to use the DECLARE_EVENT_CLASS
+ *
+ * DECLARE_EVENT_CLASS() macro creates most of the functions for the
+ * tracepoint. Then DEFINE_EVENT() is use to hook a tracepoint to those
+ * functions. This DEFINE_EVENT() is an instance of the class and can
+ * be enabled and disabled separately from other events (either TRACE_EVENT
+ * or other DEFINE_EVENT()s).
+ *
+ * Note, TRACE_EVENT() itself is simply defined as:
+ *
+ * #define TRACE_EVENT(name, proto, args, tstruct, assign, printk)  \
+ *  DECLARE_EVENT_CLASS(name, proto, args, tstruct, assign, printk); \
+ *  DEFINE_EVENT(name, name, proto, args)
+ *
+ * The DEFINE_EVENT() also can be declared with conditions and reg functions:
+ *
+ * DEFINE_EVENT_CONDITION(template, name, proto, args, cond);
+ * DEFINE_EVENT_FN(template, name, proto, args, reg, unreg);
+ */
+DECLARE_EVENT_CLASS(foo_template,
+
+	TP_PROTO(const char *foo, int bar),
+
+	TP_ARGS(foo, bar),
+
+	TP_STRUCT__entry(
+		__string(	foo,    foo		)
+		__field(	int,	bar		)
+	),
+
+	TP_fast_assign(
+		__assign_str(foo);
+		__entry->bar	= bar;
+	),
+
+	TP_printk("foo %s %d", __get_str(foo), __entry->bar)
+);
+
+/*
+ * Here's a better way for the previous samples (except, the first
+ * example had more fields and could not be used here).
+ */
+DEFINE_EVENT(foo_template, foo_with_template_simple,
+	TP_PROTO(const char *foo, int bar),
+	TP_ARGS(foo, bar));
+
+DEFINE_EVENT_CONDITION(foo_template, foo_with_template_cond,
+	TP_PROTO(const char *foo, int bar),
+	TP_ARGS(foo, bar),
+	TP_CONDITION(!(bar % 8)));
+
+
+DEFINE_EVENT_FN(foo_template, foo_with_template_fn,
+	TP_PROTO(const char *foo, int bar),
+	TP_ARGS(foo, bar),
+	foo_bar_reg, foo_bar_unreg);
+
+/*
+ * Anytime two events share basically the same values and have
+ * the same output, use the DECLARE_EVENT_CLASS() and DEFINE_EVENT()
+ * when ever possible.
+ */
+
+/*
+ * If the event is similar to the DECLARE_EVENT_CLASS, but you need
+ * to have a different output, then use DEFINE_EVENT_PRINT() which
+ * lets you override the TP_printk() of the class.
+ */
+
+DEFINE_EVENT_PRINT(foo_template, foo_with_template_print,
+	TP_PROTO(const char *foo, int bar),
+	TP_ARGS(foo, bar),
+	TP_printk("bar %s %d", __get_str(foo), __entry->bar));
+
+/*
+ * There are yet another __rel_loc dynamic data attribute. If you
+ * use __rel_dynamic_array() and __rel_string() etc. macros, you
+ * can use this attribute. There is no difference from the viewpoint
+ * of functionality with/without 'rel' but the encoding is a bit
+ * different. This is expected to be used with user-space event,
+ * there is no reason that the kernel event use this, but only for
+ * testing.
+ */
+
+TRACE_EVENT(foo_rel_loc,
+
+	TP_PROTO(const char *foo, int bar, unsigned long *mask, const cpumask_t *cpus),
+
+	TP_ARGS(foo, bar, mask, cpus),
+
+	TP_STRUCT__entry(
+		__rel_string(	foo,	foo	)
+		__field(	int,	bar	)
+		__rel_bitmask(	bitmask,
+			BITS_PER_BYTE * sizeof(unsigned long)	)
+		__rel_cpumask(	cpumask )
+	),
+
+	TP_fast_assign(
+		__assign_rel_str(foo);
+		__entry->bar = bar;
+		__assign_rel_bitmask(bitmask, mask,
+			BITS_PER_BYTE * sizeof(unsigned long));
+		__assign_rel_cpumask(cpumask, cpus);
+	),
+
+	TP_printk("foo_rel_loc %s, %d, %s, %s", __get_rel_str(foo), __entry->bar,
+		  __get_rel_bitmask(bitmask),
+		  __get_rel_cpumask(cpumask))
+);
+
+TRACE_EVENT(foo_timer_fn,
+
+	TP_PROTO(struct foo_timer_data *data),
+
+	TP_ARGS(data),
+
+	TP_STRUCT__entry(
+		__string(	name,			data->name	)
+		__field(	int,			count		)
+	),
+
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->count	= *this_cpu_ptr(data->counter);
+	),
+
+	TP_printk("name=%s count=%d", __get_str(name), __entry->count)
+);
+#endif
+
+/***** NOTICE! The #if protection ends here. *****/
+
+
+/*
+ * There are several ways I could have done this. If I left out the
+ * TRACE_INCLUDE_PATH, then it would default to the kernel source
+ * include/trace/events directory.
+ *
+ * I could specify a path from the define_trace.h file back to this
+ * file.
+ *
+ * #define TRACE_INCLUDE_PATH ../../samples/trace_events
+ *
+ * But the safest and easiest way to simply make it use the directory
+ * that the file is in is to add in the Makefile:
+ *
+ * CFLAGS_trace-events-sample.o := -I$(src)
+ *
+ * This will make sure the current path is part of the include
+ * structure for our file so that define_trace.h can find it.
+ *
+ * I could have made only the top level directory the include:
+ *
+ * CFLAGS_trace-events-sample.o := -I$(PWD)
+ *
+ * And then let the path to this directory be the TRACE_INCLUDE_PATH:
+ *
+ * #define TRACE_INCLUDE_PATH samples/trace_events
+ *
+ * But then if something defines "samples" or "trace_events" as a macro
+ * then we could risk that being converted too, and give us an unexpected
+ * result.
+ */
+#undef TRACE_INCLUDE_PATH
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_PATH .
+/*
+ * TRACE_INCLUDE_FILE is not needed if the filename and TRACE_SYSTEM are equal
+ */
+#define TRACE_INCLUDE_FILE trace-events-sample
+#include <trace/define_trace.h>
diff --git a/samples/trace_events/trace_custom_sched.c b/samples/trace_events/trace_custom_sched.c
new file mode 100644
index 000000000..dd409b704
--- /dev/null
+++ b/samples/trace_events/trace_custom_sched.c
@@ -0,0 +1,59 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * event tracer
+ *
+ * Copyright (C) 2022 Google Inc, Steven Rostedt <rostedt@goodmis.org>
+ */
+
+#define pr_fmt(fmt) fmt
+
+#include <linux/trace_events.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+
+/*
+ * Must include the event header that the custom event will attach to,
+ * from the C file, and not in the custom header file.
+ */
+#include <trace/events/sched.h>
+
+/* Declare CREATE_CUSTOM_TRACE_EVENTS before including custom header */
+#define CREATE_CUSTOM_TRACE_EVENTS
+
+#include "trace_custom_sched.h"
+
+/*
+ * As the trace events are not exported to modules, the use of
+ * for_each_kernel_tracepoint() is needed to find the trace event
+ * to attach to. The fct() function below, is a callback that
+ * will be called for every event.
+ *
+ * Helper functions are created by the TRACE_CUSTOM_EVENT() macro
+ * update the event. Those are of the form:
+ *
+ *    trace_custom_event_<event>_update()
+ *
+ * Where <event> is the event to attach.
+ */
+static void fct(struct tracepoint *tp, void *priv)
+{
+	trace_custom_event_sched_switch_update(tp);
+	trace_custom_event_sched_waking_update(tp);
+}
+
+static int __init trace_sched_init(void)
+{
+	for_each_kernel_tracepoint(fct, NULL);
+	return 0;
+}
+
+static void __exit trace_sched_exit(void)
+{
+}
+
+module_init(trace_sched_init);
+module_exit(trace_sched_exit);
+
+MODULE_AUTHOR("Steven Rostedt");
+MODULE_DESCRIPTION("Custom scheduling events");
+MODULE_LICENSE("GPL");
diff --git a/samples/trace_events/trace_custom_sched.h b/samples/trace_events/trace_custom_sched.h
new file mode 100644
index 000000000..951388334
--- /dev/null
+++ b/samples/trace_events/trace_custom_sched.h
@@ -0,0 +1,96 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+/*
+ * Like the headers that use TRACE_EVENT(), the TRACE_CUSTOM_EVENT()
+ * needs a header that allows for multiple inclusions.
+ *
+ * Test for a unique name (here we have _TRACE_CUSTOM_SCHED_H),
+ * also allowing to continue if TRACE_CUSTOM_MULTI_READ is defined.
+ */
+#if !defined(_TRACE_CUSTOM_SCHED_H) || defined(TRACE_CUSTOM_MULTI_READ)
+#define _TRACE_CUSTOM_SCHED_H
+
+/* Include linux/trace_events.h for initial defines of TRACE_CUSTOM_EVENT() */
+#include <linux/trace_events.h>
+
+/*
+ * TRACE_CUSTOM_EVENT() is just like TRACE_EVENT(). The first parameter
+ * is the event name of an existing event where the TRACE_EVENT has been included
+ * in the C file before including this file.
+ */
+TRACE_CUSTOM_EVENT(sched_switch,
+
+	/*
+	 * The TP_PROTO() and TP_ARGS must match the trace event
+	 * that the custom event is using.
+	 */
+	TP_PROTO(bool preempt,
+		 struct task_struct *prev,
+		 struct task_struct *next,
+		 unsigned int prev_state),
+
+	TP_ARGS(preempt, prev, next, prev_state),
+
+	/*
+	 * The next fields are where the customization happens.
+	 * The TP_STRUCT__entry() defines what will be recorded
+	 * in the ring buffer when the custom event triggers.
+	 *
+	 * The rest is just like the TRACE_EVENT() macro except that
+	 * it uses the custom entry.
+	 */
+	TP_STRUCT__entry(
+		__field(	unsigned short,		prev_prio	)
+		__field(	unsigned short,		next_prio	)
+		__field(	pid_t,	next_pid			)
+	),
+
+	TP_fast_assign(
+		__entry->prev_prio	= prev->prio;
+		__entry->next_pid	= next->pid;
+		__entry->next_prio	= next->prio;
+	),
+
+	TP_printk("prev_prio=%d next_pid=%d next_prio=%d",
+		  __entry->prev_prio, __entry->next_pid, __entry->next_prio)
+)
+
+
+TRACE_CUSTOM_EVENT(sched_waking,
+
+	TP_PROTO(struct task_struct *p),
+
+	TP_ARGS(p),
+
+	TP_STRUCT__entry(
+		__field(	pid_t,			pid	)
+		__field(	unsigned short,		prio	)
+	),
+
+	TP_fast_assign(
+		__entry->pid	= p->pid;
+		__entry->prio	= p->prio;
+	),
+
+	TP_printk("pid=%d prio=%d", __entry->pid, __entry->prio)
+)
+#endif
+/*
+ * Just like the headers that create TRACE_EVENTs, the below must
+ * be outside the protection of the above #if block.
+ */
+
+/*
+ * It is required that the Makefile includes:
+ *    CFLAGS_<c_file>.o := -I$(src)
+ */
+#undef TRACE_INCLUDE_PATH
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_PATH .
+
+/*
+ * It is requred that the TRACE_INCLUDE_FILE be the same
+ * as this file without the ".h".
+ */
+#define TRACE_INCLUDE_FILE trace_custom_sched
+#include <trace/define_custom_trace.h>