TODO

   1 Please contact Mathieu Desnoyers <mathieu.desnoyers@efficios.com> for
   2 questions about this TODO list. The "Cleanup/Testing" section would be
   3 good to go through before integration into mainline. The "Features"
   4 section is a wish list of features to complete before releasing the
   5 "LTTng 2.0" final version, but are not required to have LTTng working.
   6 These features are mostly performance enhancements and instrumentation
   7 enhancements.
   8
   9 TODO:
  10
  11 A) Cleanup/Testing
  12
  13         1) Remove debugfs "lttng" file (keep only procfs "lttng" file).
  14            The rationale for this is that this file is needed for
  15            user-level tracing support (LTTng-UST 2.0) intended to be
  16            used on production system, and therefore should be present as
  17            part of a "usually mounted" filesystem rather than a debug
  18            filesystem.
  19
  20         2) Test lib ring buffer snapshot feature.
  21            When working on the lttngtop project, Julien Desfossez
  22            reported that he needed to push the consumer position
  23            forward explicitely with lib_ring_buffer_put_next_subbuf.
  24            This means that although the usual case of pairs of
  25            lib_ring_buffer_get_next_subbuf/lib_ring_buffer_put_next_subbuf
  26            work fine, there is probably a problem that needs to be
  27            investigated in
  28            lib_ring_buffer_get_subbuf/lib_ring_buffer_put_subbuf, which
  29            depend on the producer to push the reader position.
  30            Contact: Julien Desfossez <julien.desfossez@polymtl.ca>
  31
  32
  33 B) Features
  34
  35         1) Integration of the LTTng 0.x trace clocks into
  36            LTTng 2.0.
  37              Currently using mainline kernel monotonic clock. NMIs can
  38              therefore not be traced, and this causes a significant
  39              performance degradation compared to the LTTng 0.x trace
  40              clocks. Imply the creation of drivers/staging/lttng/arch to
  41              contain the arch-specific clock support files.
  42              * Dependency: addition of clock descriptions to CTF.
  43            See: http://git.lttng.org/?p=linux-2.6-lttng.git;a=summary
  44                 for the LTTng 0.x git tree.
  45
  46         2) Port OMAP3 LTTng trace clocks to x86 to support systems
  47            without constant TSC.
  48             * Dependency: (B.1)
  49            See: http://git.lttng.org/?p=linux-2.6-lttng.git;a=summary
  50                 for the LTTng 0.x git tree.
  51
  52         3) Implement mmap operation on an anonymous file created by a
  53            LTTNG_KERNEL_CLOCK ioctl to export data to export
  54            synchronized kernel and user-level LTTng trace clocks:
  55            with:
  56               - shared per-cpu data,
  57               - read seqlock.
  58            The content exported by this shared memory area will be
  59            arch-specific.
  60            * Dependency: (B.1) && (B.2)
  61            See: http://git.lttng.org/?p=linux-2.6-lttng.git;a=summary
  62                 for the LTTng 0.x git tree, which has vDSO support for
  63                 LTTng trace clock on the x86 architecture.
  64
  65         3) Integrate the "statedump" module from LTTng 0.x into LTTng
  66            2.0.
  67              * Dependency: addition of "dynamic enumerations" type to CTF.
  68            See: http://git.lttng.org/?p=lttng-modules.git;a=shortlog;h=refs/heads/v0.19-stable
  69                 ltt-statedump.c
  70
  71         4) Generate system call TRACE_EVENT headers for all
  72            architectures (currently done: x86 32/64).
  73
  74         5) Define "unknown" system calls into instrumentation/syscalls
  75            override files / or do SYSCALL_DEFINE improvements to
  76            mainline kernel to allow automatic generation of these
  77            missing system call descriptions.
  78
  79         6) Create missing tracepoint event headers files into
  80            instrumentation/events from headers located in
  81            include/trace/events/. Choice: either do as currently done,
  82            and copy those headers locally into the lttng driver and
  83            perform the modifications locally, or push TRACE_EVENT API
  84            modification into mainline headers, which would require
  85            collaboration from Ftrace/Perf maintainers.
  86
  87         7) Poll: implement a poll and/or epoll exclusive wakeup scheme,
  88            which contradicts POSIX, but protect multiple consumer
  89            threads from thundering herd effect.
  90
  91         8) Re-integrate sample modules from libringbuffer into
  92            lttng driver. Those modules can be used as example of how to
  93            use libringbuffer in other contexts than LTTng, and are
  94            useful to perform benchmarks of the ringbuffer library.
  95            See: http://www.efficios.com/ringbuffer
  96
  97         9) NOHZ support for lib ring buffer. NOHZ infrastructure in the
  98            Linux kernel does not support notifiers chains, which does
  99            not let LTTng play nicely with low power consumption setups
 100            for flight recorder (overwrite mode) live traces. One way to
 101            allow integration between NOHZ and LTTng would be to add
 102            support for such notifiers into NOHZ kernel infrastructure.
 103
 104         10) Turn drivers/staging/lttng/ltt-probes.c probe_list into a
 105             hash table. Turns O(n^2) trace systems registration (cost
 106             for n systems) into O(n). (O(1) per system)
 107
 108         11) drivers/staging/lttng/probes/lttng-ftrace.c:
 109             LTTng currently uses kretprobes for per-function tracing,
 110             not the function tracer. So lttng-ftrace.c should be used
 111             for "all" function tracing.
 112
 113         12) drivers/staging/lttng/probes/lttng-types.c:
 114             This is a currently unused placeholder to export entire C
 115             type declarations into the trace metadata, e.g. for support
 116             of describing the layout of structures/enumeration mapping
 117             along with syscall entry events.  The design of this support
 118             will likely change though, and become integrated with the
 119             TRACE_EVENT support within lttng, by adding new macros, and
 120             support for generation of metadata from these macros, to
 121             allow description of those compound types/enumerations.
 122
 123 Please send patches
 124 To: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>