trace-cmd: Version 2.5.6

trace-cmd: Version 2.5.5

trace-cmd: Version 2.6.2

trace-cmd: Version 2.6.1

Changes from version v2.5.2: (where 2.6 was forked)

Difference in output:

 o  Show actual event paths instead of glob when starting

 o  Simplify status output at end of trace

 o  Sort events in trace-cmd profile

 o  Show instance names in trace-cmd stat

Options added:

 o  Implented the -g option for trace-cmd profile on custom events

 o  Add -G option to make irqs global events in trace-cmd profile

 o  Add -B and -t options to trace-cmd extract to handle instances

 o  Add -a option for all instances in trace-cmd extract

 o  Add -a option for all instances for trace-cmd stop, reset, and restart

 o  Add --debug to trace-cmd report that shows markup of the trace.dat file.

 o  Add --by-comm to group tasks together by their names in trace-cmd profile.

Plugins:

 o  Added plugin for tlb_flush event

 o  Plugins are now allowed to have options

Misc:

 o  Have trace-cmd hist work with instances

And other various fixes that are not user visible unless they hit the bug
that was fixed.

trace-cmd: Version 2.5.4

Changes since Version 2.5.3:

 o Two bug fixes to trace-cmd profile

    - Fixed bug to read save -H option in trace.dat

    - Fixed spurious stack traces to random events

trace-cmd: Version 2.1.6

Changes since Version 2.1.5:

 o Bug fix

    - tsdate check can get blocked forever in  find_time_stamp

trace-cmd: Version 2.2.4

Changes since Version 2.2.3:

 o Bug fix

    - tsdate check can get blocked forever in  find_time_stamp

trace-cmd: Version 2.3.4

Changes since Version 2.3.3:

 o Two bug fixes

    - CPU 0 was hardcoded for record boundary check

    - tsdate check can get blocked forever in  find_time_stamp

trace-cmd: Version 2.4.2

Changes since Version 2.4.1:

 o Two bug fixes

    - CPU 0 was hardcoded for record boundary check

    - tsdate check can get blocked forever in  find_time_stamp

trace-cmd: Version 2.5.3

Changes since Version 2.5.2:

 o Two bug fixes

    - CPU 0 was hardcoded for record boundary check

    - tsdate check can get blocked forever in  find_time_stamp

trace-cmd v2.5.2 broke the kernelshark build by adding the trace-hooks.o
object to the wrong set of files.

Because this update does not change any of the functionality of trace-cmd
itself, it is staying at 2.5.2. This tag is only to fix the build of
kernelshark.

trace-cmd: Version 2.1.5

Changes since Version 2.1.4:

 o Add support for tracefs file system

 o Add support for IP address formats

 o Handle '%z' in bprint format

 o Add scsi_plugin

 o Add xen plugin

 o Add cfg80211 plugin

 o Handle __print_hex(__get_dynamic_array(fieldname), len)

 o Print event name whan showing warnings (if possible)

 o Report better error messages on bad function args

 o Report unknown VMX exit reasons with code

 o Add a "make dist" target

 o Update VMX exit reasons in KVM plugin

 o Several fixes

 o Several clean ups

trace-cmd: Version 2.2.3

Changes since Version 2.2.2:

 o Add support for tracefs file system

 o Add support for IP address formats

 o Handle '%z' in bprint format

 o Add scsi_plugin

 o Add xen_plugin

 o Add cfg80211 plugin

 o Print event names on warning (if possible)

 o Report better error messages on bad function args

 o Handle __print_hex(__get_dynamic_array(fieldname))

 o Add "make dist" target

 o Have bprintk output the same as the kernel does

 o Several fixes

 o Several clean ups

trace-cmd: Version 2.3.3

Changes since v2.3.2:

 o Added support for tracefs file system

 o Add support for IP address formats

 o Handle '%z' in bprint format

 o Print event name when showing warings (if possible)

 o Have recording to instances not affect top level buffer

 o Have trace-on reset for existing instances

 o Show better error messages on bad function args

 o Several fixes

 o Clean ups

trace-cmd: Version 2.4.1

 o Add support for tracefs file system

 o Add support for IP address formats

 o Handle %z in bprint format

 o Several fixes

 o Some clean ups

trace-cmd: Version 2.5.2

Changes since Version 2.5.1:

 o Added support for tracefs filesystem.

 o profile: Show timestamps of max and mins

 o profile: Add-H for user specified event hooking

 o Have -H in record be saved for report --profile to use

 o Save uname in trace.dat.

 o Cleanups

 o Fixes

trace-cmd: Version 2.5.1

Changes since 2.5:

 o Various bug fixes.

 o Add -C option to list command to show what trace clocks are available

 o Add -C option to record command to set trace clock

 o Add --strerr to profile command to set output to stderr

 o Add -o option to profile command to redirect output to a specific file

 o Add support to print ip addresses in TP_printk()

Changes from version 2.4:

Added three new commands:

 o trace-cmd stream
    This starts ftrace tracing and reads the binary files and processes it
    to produce text output without the need to save to disk (or network).

 o trace-cmd profile
    This uses the stream infrastructure from above, but also keeps track
    of the data and analyzes it to produce various useful statistics.
    By default it enables function graph tracing with a depth of 1, to only
    show where user tasks entered the kernel. It also enables sched switch
    and sched wakeup events to see how long it took to wake up a task.
    It also enables stack tracing to display where the latencies are happening.
    There's lots of options to change the default settings to show more
    about how the kernel is working.

 o trace-cmd stat
    Shows what may be enabled in ftrace. There's times when events are enabled
    or kprobes are added, or even when filters are used, and it may be difficult
    to see what state ftrace is in. Using trace-cmd stat will show you the state
    that ftrace is in. It will display if a tracer is enabled and much more.

New options:

In trace-cmd record

  Added --profile, that enables the default profiling events (and function graph
  tracing with depth of 1). These are the same events that are enabled with the
  trace-cmd profile command.

In trace-cmd report

  Added --profile option that will read the data of the trace.dat file and report
  it the same as the trace-cmd profile command would.

Some improvements:

 o trace-cmd record will reset everything it touches after it is done recording.
    Before it use to leave some things still active, like options and such.
    Now, if you enable options for recording, it will reset them when finished,
    as well as triggers and filters. trace-cmd start, or trace-cmd record with
    the -k option will leave the changes as is when done.

 o trace-cmd reset now resets triggers and filters.

Changes from version 2.3:

In trace-cmd list

  o Regex expresions can be passed to the -e option to limit the events
    that are displayed to what matches the regex.

  o New -F option to list the event's format file. (following -e)

  o New -l option to show the event's filters (following -e)

  o New -R option to show the event's triggers (following -e)

  o New -B option to show what buffer instances exist on the machine

In trace-cmd record (and start)

  o New -R option to specify a trigger to the previous event

In trace-cmd report

  o New --event <regex> option to list the event formats in the trace.dat
    file that match the regex.

In trace-cmd reset

  o New -B option to reset a specified buffer. Does not reset buffers
    not listed (multiple -B options may be specified).

  o New -t option to reset the top level buffer when a -B option is used.
    If no -B option is used, the top level is reset regardless if the -t
    option is specified or not.

  o New -d option to delete the previously specified buffer that came after
    the -B option.

In trace-cmd show

  o New --tracing-on option to show the value of tracing_on (in the
    debugfs/tracing directory)

  o New --curent-tracer option to show the value of current_tracer
    (in the debugfs/tracing directory)

  o New --buffer_size option to show the value of buffer_size_kb (from
    the debugfs/tracing directory).

  o New --buffer_total_size opton to show the value of the
    buffer_total_size_kb (from the debugfs/tracing directory).

  o New --ftrace_filter option, to show the value of the
    set_ftrace_filter file (from the debugfs/tracing directory).

  o New --ftrace_notrace option, to show the value of the
    set_ftrace_notrace file (from the debugfs/tracing directory).

  o New --ftrace_pid option to show the value of the set_ftrace_pid
    (from the debugfs/tracing directory).

  o New --graph_function option to show the value of the set_graph_function
    (from the debugfs/tracing directory)

  o New --graph_notrace option to show the value of set_graph_notrace
    (from the debugfs/tracing directory)

  o New --cpumask option to show the value of tracing_cpumask (from the
    debugfs/tracing directory).

In trace-cmd split

  o New -C <cpu> option to split out the cpu buffers from a trace.dat file.

In trace-cmd stop

  o New -B option to stop a specific buffer. No other buffers are affected.
    Multiple buffers may be specified with multiple -B options.

  o New -t option to stop the top level buffer when the -B option is used,
    otherwise the top level is not affected.  This is the default when -B
    is not used.

New trace-cmd restart option

  As trace-cmd stop just disables the ring buffer (writes '0' into tracing_on)
  but trace-cmd start will reset the ring buffer and several options removing
  whatever was there previously. The 'restart' command was added that is basically
  the inverse of the 'stop' command (writes '1' into tracing_on). This way if
  you execute trace-cmd stop and want to restart the trace without reseting anything
  you can use trace-cmd restart. It includes the -B and -t options that are similar
  to the new trace-cmd stop options.