Currently, when a module event is enabled, when that module is removed, it
clears all ring buffers. This is to prevent another module from being loaded
and having one of its trace event IDs from reusing a trace event ID of the
removed module. This could cause undesirable effects as the trace event of
the new module would be using its own processing algorithms to process raw
data of another event. To prevent this, when a module is loaded, if any of
its events have been used (signified by the WAS_ENABLED event call flag,
which is never cleared), all ring buffers are cleared, just in case any one
of them contains event data of the removed event.

The problem is, there's no reason to clear all ring buffers if only one (or
less than all of them) uses one of the events. Instead, only clear the ring
buffers that recorded the events of a module that is being removed.

To do this, instead of keeping the WAS_ENABLED flag with the trace event
call, move it to the per instance (per ring buffer) event file descriptor.
The event file descriptor maps each event to a separate ring buffer
instance. Then when the module is removed, only the ring buffers that
activated one of the module's events get cleared. The rest are not touched.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Inorder to support recording of tgid, the following changes are made:

* Introduce a new API (tracing_record_taskinfo) to additionally record the tgid
  along with the task's comm at the same time. This has has the benefit of not
  setting trace_cmdline_save before all the information for a task is saved.
* Add a new API tracing_record_taskinfo_sched_switch to record task information
  for 2 tasks at a time (previous and next) and use it from sched_switch probe.
* Preserve the old API (tracing_record_cmdline) and create it as a wrapper
  around the new one so that existing callers aren't affected.
* Reuse the existing sched_switch and sched_wakeup probes to record tgid
  information and add a new option 'record-tgid' to enable recording of tgid

When record-tgid option isn't enabled to being with, we take care to make sure
that there's isn't memory or runtime overhead.

Link: http://lkml.kernel.org/r/20170627020155.5139-1-joelaf@google.com

Cc: kernel-team@android.com
Cc: Ingo Molnar <mingo@redhat.com>
Tested-by: NMichael Sartain <mikesart@gmail.com>
Signed-off-by: NJoel Fernandes <joelaf@google.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

The enum_replace stanza works as is for sizeof()
calls as well as enums. Rename it as well.

Link: http://lkml.kernel.org/r/20170531215653.3240-9-jeremy.linton@arm.comSigned-off-by: NJeremy Linton <jeremy.linton@arm.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Rename the core trace enum routines to use eval, to
reflect their use by more than just enum to value mapping.

Link: http://lkml.kernel.org/r/20170531215653.3240-8-jeremy.linton@arm.comSigned-off-by: NJeremy Linton <jeremy.linton@arm.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Each enum is loaded into the trace_enum_map, as we
are now using this for more than enums rename it.

Link: http://lkml.kernel.org/r/20170531215653.3240-3-jeremy.linton@arm.comSigned-off-by: NJeremy Linton <jeremy.linton@arm.com>
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

With the redesign of the registration and execution of the function probes
(triggers), data can now be passed from the setup of the probe to the probe
callers that are specific to the trace_array it is on. Although, all probes
still only affect the toplevel trace array, this change will allow for
instances to have their own probes separated from other instances and the
top array.

That is, something like the stacktrace probe can be set to trace only in an
instance and not the toplevel trace array. This isn't implement yet, but
this change sets the ground work for the change.

When a probe callback is triggered (someone writes the probe format into
set_ftrace_filter), it calls register_ftrace_function_probe() passing in
init_data that will be used to initialize the probe. Then for every matching
function, register_ftrace_function_probe() will call the probe_ops->init()
function with the init data that was passed to it, as well as an address to
a place holder that is associated with the probe and the instance. The first
occurrence will have a NULL in the pointer. The init() function will then
initialize it. If other probes are added, or more functions are part of the
probe, the place holder will be passed to the init() function with the place
holder data that it was initialized to the last time.

Then this place_holder is passed to each of the other probe_ops functions,
where it can be used in the function callback. When the probe_ops free()
function is called, it can be called either with the rip of the function
that is being removed from the probe, or zero, indicating that there are no
more functions attached to the probe, and the place holder is about to be
freed. This gives the probe_ops a way to free the data it assigned to the
place holder if it was allocade during the first init call.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

In order to eventually have each trace_array instance have its own unique
set of function probes (triggers), the trace array needs to hold the ops and
the filters for the probes.

This is the first step to accomplish this. Instead of having the private
data of the probe ops point to the trace_array, create a separate list that
the trace_array holds. There's only one private_data for a probe, we need
one per trace_array. The probe ftrace_ops will be dynamically created for
each instance, instead of being static.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Pass the trace_array associated to a ftrace_probe_ops into the probe_ops
func(), init() and free() functions. The trace_array is the descriptor that
describes a tracing instance. This will help create the infrastructure that
will allow having function probes unique to tracing instances.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Add a link list to the trace_array to hold func probes that are registered.
Currently, all function probes are the same for all instances as it was
before, that is, only the top level trace_array holds the function probes.
But this lays the ground work to have function probes be attached to
individual instances, and having the event trigger only affect events in the
given instance. But that work is still to be done.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

Currently unregister_ftrace_function_probe_func() is a void function. It
does not give any feedback if an error occurred or no item was found to
remove and nothing was done.

Change it to return status and success if it removed something. Also update
the callers to return that feedback to the user.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

No users of the function probes uses the data field anymore. Remove it, and
change the init function to take a void *data parameter instead of a
void **data, because the init will just get the data that the registering
function was received, and there's no state after it is called.

The other functions for ftrace_probe_ops still take the data parameter, but
it will currently only be passed NULL. It will stay as a parameter for
future data to be passed to these functions.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

In order to move the ops to the function probes directly, they need a way to
map function ips to their own data without depending on the infrastructure
of the function probes, as the data field will be going away.

New helper functions are added that are based on the ftrace_hash code.
ftrace_func_mapper functions are there to let the probes map ips to their
data. These can be allocated by the probe ops, and referenced in the
function callbacks.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

In preparation to cleaning up the probe function registration code, the
"data" parameter will eventually be removed from the probe->func() call.
Instead it will receive its own "ops" function, in which it can set up its
own data that it needs to map.
Signed-off-by: NSteven Rostedt (VMware) <rostedt@goodmis.org>

If one of the events within a system fails to enable when "1" is written
to the system "enable" file, it should return an error. Note, some events
may still be enabled, but the user should know that something did go wrong.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Currently, when tracepoint_printk is set (enabled by the "tp_printk" kernel
command line), it causes trace events to print via printk(). This is a very
dangerous operation, but is useful for debugging.

The issue is, it's seldom used, but it is always checked even if it's not
enabled by the kernel command line. Instead of having this feature called by
a branch against a variable, turn that variable into a static key, and this
will remove the test and jump.

To simplify things, the functions output_printk() and
trace_event_buffer_commit() were moved from trace_events.c to trace.c.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The creation of the set_event_pid file was assigned to a variable "entry"
but that variable was never used. Ideally, it should be used to check if the
file was created and warn if it was not.

The files header_page, header_event should also be checked and a warning if
they fail to be created.

The "enable" file was moved up, as it is a more crucial file to have and a
hard failure (return -ENOMEM) should be returned if it is not created.
Reported-by: NDavid Binderman <dcb314@hotmail.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Because tracepoint callbacks are done with preemption enabled, the trace
events are always called with preempt disable due to the
rcu_read_lock_sched_notrace() in __DO_TRACE(). This causes the preempt count
shown in the recorded trace event to be inaccurate. It is always one more
that what the preempt_count was when the tracepoint was called.

If CONFIG_PREEMPT is enabled, subtract 1 from the preempt_count before
recording it in the trace buffer.

Link: http://lkml.kernel.org/r/20160525132537.GA10808@linutronix.deReported-by: NSebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The addition of PIDs into a pid_list via the write operation of
set_event_pid is a bit complex. The same operation will be needed for
function tracing pids. Move the code into its own generic function in
trace.c, so that we can avoid duplication of this code.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

To allow other aspects of ftrace to use the pid_list logic, we need to reuse
the seq_file functions. Making the generic part into functions that can be
called by other files will help in this regard.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

As the filtered_pid functions are going to be used by function tracer as
well as trace_events, move the code into the generic trace.c file.

The functions moved are:

 trace_find_filtered_pid()
 trace_ignore_this_task()
 trace_filter_add_remove_task()

Kernel Doc text was also added.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Make the functions used for pid filtering global for tracing, such that the
function tracer can use the pid code as well.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Filtering of events requires the data to be written to the ring buffer
before it can be decided to filter or not. This is because the parameters of
the filter are based on the result that is written to the ring buffer and
not on the parameters that are passed into the trace functions.

The ftrace ring buffer is optimized for writing into the ring buffer and
committing. The discard procedure used when filtering decides the event
should be discarded is much more heavy weight. Thus, using a temporary
filter when filtering events can speed things up drastically.

Without a temp buffer we have:

 # trace-cmd start -p nop
 # perf stat -r 10 hackbench 50
       0.790706626 seconds time elapsed ( +-  0.71% )

 # trace-cmd start -e all
 # perf stat -r 10 hackbench 50
       1.566904059 seconds time elapsed ( +-  0.27% )

 # trace-cmd start -e all -f 'common_preempt_count==20'
 # perf stat -r 10 hackbench 50
       1.690598511 seconds time elapsed ( +-  0.19% )

 # trace-cmd start -e all -f 'common_preempt_count!=20'
 # perf stat -r 10 hackbench 50
       1.707486364 seconds time elapsed ( +-  0.30% )

The first run above is without any tracing, just to get a based figure.
hackbench takes ~0.79 seconds to run on the system.

The second run enables tracing all events where nothing is filtered. This
increases the time by 100% and hackbench takes 1.57 seconds to run.

The third run filters all events where the preempt count will equal "20"
(this should never happen) thus all events are discarded. This takes 1.69
seconds to run. This is 10% slower than just committing the events!

The last run enables all events and filters where the filter will commit all
events, and this takes 1.70 seconds to run. The filtering overhead is
approximately 10%. Thus, the discard and commit of an event from the ring
buffer may be about the same time.

With this patch, the numbers change:

 # trace-cmd start -p nop
 # perf stat -r 10 hackbench 50
       0.778233033 seconds time elapsed ( +-  0.38% )

 # trace-cmd start -e all
 # perf stat -r 10 hackbench 50
       1.582102692 seconds time elapsed ( +-  0.28% )

 # trace-cmd start -e all -f 'common_preempt_count==20'
 # perf stat -r 10 hackbench 50
       1.309230710 seconds time elapsed ( +-  0.22% )

 # trace-cmd start -e all -f 'common_preempt_count!=20'
 # perf stat -r 10 hackbench 50
       1.786001924 seconds time elapsed ( +-  0.20% )

The first run is again the base with no tracing.

The second run is all tracing with no filtering. It is a little slower, but
that may be well within the noise.

The third run shows that discarding all events only took 1.3 seconds. This
is a speed up of 23%! The discard is much faster than even the commit.

The one downside is shown in the last run. Events that are not discarded by
the filter will take longer to add, this is due to the extra copy of the
event.

Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Currently register functions for events will be called
through the 'reg' field of event class directly without
any check when seting up triggers.

Triggers for events that don't support register through
debug fs (events under events/ftrace are for trace-cmd to
read event format, and most of them don't have a register
function except events/ftrace/functionx) can't be enabled
at all, and an oops will be hit when setting up trigger
for those events, so just not creating them is an easy way
to avoid the oops.

Link: http://lkml.kernel.org/r/1462275274-3911-1-git-send-email-chuhu@redhat.com

Cc: stable@vger.kernel.org # 3.14+
Fixes: 85f2b082 ("tracing: Add basic event trigger framework")
Signed-off-by: NChunyu Hu <chuhu@redhat.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The only user of trace_current_buffer_lock_reserve() is in the boot up self
tests. Restructure the code a little to have that code use what everything
else uses: trace_event_buffer_lock_reserve().
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

'hist' triggers allow users to continually aggregate trace events,
which can then be viewed afterwards by simply reading a 'hist' file
containing the aggregation in a human-readable format.

The basic idea is very simple and boils down to a mechanism whereby
trace events, rather than being exhaustively dumped in raw form and
viewed directly, are automatically 'compressed' into meaningful tables
completely defined by the user.

This is done strictly via single-line command-line commands and
without the aid of any kind of programming language or interpreter.

A surprising number of typical use cases can be accomplished by users
via this simple mechanism.  In fact, a large number of the tasks that
users typically do using the more complicated script-based tracing
tools, at least during the initial stages of an investigation, can be
accomplished by simply specifying a set of keys and values to be used
in the creation of a hash table.

The Linux kernel trace event subsystem happens to provide an extensive
list of keys and values ready-made for such a purpose in the form of
the event format files associated with each trace event.  By simply
consulting the format file for field names of interest and by plugging
them into the hist trigger command, users can create an endless number
of useful aggregations to help with investigating various properties
of the system.  See Documentation/trace/events.txt for examples.

hist triggers are implemented on top of the existing event trigger
infrastructure, and as such are consistent with the existing triggers
from a user's perspective as well.

The basic syntax follows the existing trigger syntax.  Users start an
aggregation by writing a 'hist' trigger to the event of interest's
trigger file:

  # echo hist:keys=xxx [ if filter] > event/trigger

Once a hist trigger has been set up, by default it continually
aggregates every matching event into a hash table using the event key
and a value field named 'hitcount'.

To view the aggregation at any point in time, simply read the 'hist'
file in the same directory as the 'trigger' file:

  # cat event/hist

The detailed syntax provides additional options for user control, and
is described exhaustively in Documentation/trace/events.txt and in the
virtual tracing/README file in the tracing subsystem.

Link: http://lkml.kernel.org/r/72d263b5e1853fe9c314953b65833c3aa75479f2.1457029949.git.tom.zanussi@linux.intel.comSigned-off-by: NTom Zanussi <tom.zanussi@linux.intel.com>
Tested-by: NMasami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Reviewed-by: NNamhyung Kim <namhyung@kernel.org>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Add the infrastructure needed to have the PIDs in set_event_pid to
automatically add PIDs of the children of the tasks that have their PIDs in
set_event_pid. This will also remove PIDs from set_event_pid when a task
exits

This is implemented by adding hooks into the fork and exit tracepoints. On
fork, the PIDs are added to the list, and on exit, they are removed.

Add a new option called event_fork that when set, PIDs in set_event_pid will
automatically get their children PIDs added when they fork, as well as any
task that exits will have its PID removed from set_event_pid.

This works for instances as well.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

In order to add the ability to let tasks that are filtered by the events
have their children also be traced on fork (and then not traced on exit),
convert the array into a pid bitmask. Most of the time the number of pids is
only 32768 pids or a 4k bitmask, which is the same size as the default list
currently is, and that list could grow if more pids are listed.

This also greatly simplifies the code.
Suggested-by: N"H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The name "check_ignore_pid" is confusing in trying to figure out if the pid
should be ignored or not. Rename it to "ignore_this_task" which is pretty
straight forward, as a task (not a pid) is passed in, and should if true
should be ignored.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

during bpf program loading remember the last byte of ctx access
and at the time of attaching the program to tracepoint check that
the program doesn't access bytes beyond defined in tracepoint fields

This also disallows access to __dynamic_array fields, but can be
relaxed in the future.
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 9f616680 "tracing: Allow triggers to filter for CPU ids and
process names" added a 'comm' filter that will filter events based on the
current tasks struct 'comm'. But this now hides the ability to filter events
that have a 'comm' field too. For example, sched_migrate_task trace event.
That has a 'comm' field of the task to be migrated.

 echo 'comm == "bash"' > events/sched_migrate_task/filter

will now filter all sched_migrate_task events for tasks named "bash" that
migrates other tasks (in interrupt context), instead of seeing when "bash"
itself gets migrated.

This fix requires a couple of changes.

1) Change the look up order for filter predicates to look at the events
   fields before looking at the generic filters.

2) Instead of basing the filter function off of the "comm" name, have the
   generic "comm" filter have its own filter_type (FILTER_COMM). Test
   against the type instead of the name to assign the filter function.

3) Add a new "COMM" filter that works just like "comm" but will filter based
   on the current task, even if the trace event contains a "comm" field.

Do the same for "cpu" field, adding a FILTER_CPU and a filter "CPU".

Cc: stable@vger.kernel.org # v4.3+
Fixes: 9f616680 "tracing: Allow triggers to filter for CPU ids and process names"
Reported-by: NMatt Fleming <matt@codeblueprint.co.uk>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The ftrace:function event is only displayed for parsing the function tracer
data. It is not used to enable function tracing, and does not include an
"enable" file in its event directory.

Originally, this event was kept separate from other events because it did
not have a ->reg parameter. But perf added a "reg" parameter for its use
which caused issues, because it made the event available to functions where
it was not compatible for.

Commit 9b63776f "tracing: Do not enable function event with enable"
added a TRACE_EVENT_FL_IGNORE_ENABLE flag that prevented the function event
from being enabled by normal trace events. But this commit missed keeping
the function event from being displayed by the "available_events" directory,
which is used to show what events can be enabled by set_event.

One documented way to enable all events is to:

 cat available_events > set_event

But because the function event is displayed in the available_events, this
now causes an INVALID error:

 cat: write error: Invalid argument
Reported-by: NChunyu Hu <chuhu@redhat.com>
Fixes: 9b63776f "tracing: Do not enable function event with enable"
Cc: stable@vger.kernel.org # 3.4+
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

The set_event_pid filter relies on attaching to the sched_switch and
sched_wakeup tracepoints to see if it should filter the tracing on schedule
tracepoints. By adding the callbacks to sched_wakeup, pids in the
set_event_pid file will trace the wakeups of those tasks with those pids.

But sched_wakeup_new and sched_waking were missed. These two should also be
traced. Luckily, these tracepoints share the same class as sched_wakeup
which means they can use the same pre and post callbacks as sched_wakeup
does.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Both early_enable_events() and apply_trace_boot_options() parse a boot
string that may get parsed later on. They both use strsep() which converts a
comma into a nul character. To still allow the boot string to be parsed
again the same way, the nul character gets converted back to a comma after
the token is processed.

The problem is that these two functions check for an empty parameter (two
commas in a row ",,"), and continue the loop if the parameter is empty, but
fails to place the comma back. In this case, the second parsing will end at
this blank field, and not process fields afterward.

In most cases, users should not have an empty field, but if its going to be
checked, the code might as well be correct.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

For the case where pids are already in set_event_pid, and one is added or
removed then each CPU should be checked to make sure that the new or old pid
is on or not on a CPU.

 For example:

 # echo 123 >> set_event_pid

or

 # echo '!123' >> set_event_pid

Link: http://lkml.kernel.org/r/20151030061643.GA19480@cacSuggested-by: NJiaxing Wang <hello.wjx@gmail.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

p_start() and p_stop() are seq_file functions that match. Teach sparse to
know that rcu_read_lock_sched() that is taken by p_start() is released by
p_stop.
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

My tests found that if a task is running but not filtered when set_event_pid
is modified, then it can still be traced.

Call on_each_cpu() to check if the current running task should be filtered
and update the per cpu flags of tr->data appropriately.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Add the necessary hooks to use the pids loaded in set_event_pid to filter
all the events enabled in the tracing instance that match the pids listed.

Two probes are added to both sched_switch and sched_wakeup tracepoints to be
called before other probes are called and after the other probes are called.
The first is used to set the necessary flags to let the probes know to test
if they should be traced or not.

The sched_switch pre probe will set the "ignore_pid" flag if neither the
previous or next task has a matching pid.

The sched_switch probe will set the "ignore_pid" flag if the next task
does not match the matching pid.

The pre probe allows for probes tracing sched_switch to be traced if
necessary.

The sched_wakeup pre probe will set the "ignore_pid" flag if neither the
current task nor the wakee task has a matching pid.

The sched_wakeup post probe will set the "ignore_pid" flag if the current
task does not have a matching pid.

Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Create a tracing directory called set_event_pid, which currently has no
function, but will be used to filter all events for the tracing instance or
the pids that are added to the file.

The reason no functionality is added with this commit is that this commit
focuses on the creation and removal of the pids in a safe manner. And tests
can be made against this change to make sure things are correct before
hooking features to the list of pids.

Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The kernel now has kstrdup_const/kfree_const for reusing .rodata
(typically string literals) when possible; there's no reason to
duplicate that logic in the tracing system. Moreover, as the comment
above core_kernel_data states, it may not always return true for
.rodata - that is for example the case on x86_64, where we thus end up
kstrdup'ing all the passed-in strings.

Arguably, testing for .rodata explicitly (as kstrdup_const does) is
also more correct: I don't think one is supposed to be able to change
the name after creating the event_subsystem by passing the address of
a static char (but non-const) array.

Link: http://lkml.kernel.org/r/1441833841-12955-1-git-send-email-linux@rasmusvillemoes.dkSigned-off-by: NRasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>