docs/tracing.txt

   1 = Tracing =
   2
   3 == Introduction ==
   4
   5 This document describes the tracing infrastructure in QEMU and how to use it
   6 for debugging, profiling, and observing execution.
   7
   8 == Quickstart ==
   9
  10 1. Build with the 'simple' trace backend:
  11
  12     ./configure --trace-backend=simple
  13     make
  14
  15 2. Create a file with the events you want to trace:
  16
  17    echo bdrv_aio_readv   > /tmp/events
  18    echo bdrv_aio_writev >> /tmp/events
  19
  20 3. Run the virtual machine to produce a trace file:
  21
  22     qemu -trace events=/tmp/events ... # your normal QEMU invocation
  23
  24 4. Pretty-print the binary trace file:
  25
  26     ./simpletrace.py trace-events trace-*
  27
  28 == Trace events ==
  29
  30 There is a set of static trace events declared in the "trace-events" source
  31 file.  Each trace event declaration names the event, its arguments, and the
  32 format string which can be used for pretty-printing:
  33
  34     qemu_vmalloc(size_t size, void *ptr) "size %zu ptr %p"
  35     qemu_vfree(void *ptr) "ptr %p"
  36
  37 The "trace-events" file is processed by the "tracetool" script during build to
  38 generate code for the trace events.  Trace events are invoked directly from
  39 source code like this:
  40
  41     #include "trace.h"  /* needed for trace event prototype */
  42
  43     void *qemu_vmalloc(size_t size)
  44     {
  45         void *ptr;
  46         size_t align = QEMU_VMALLOC_ALIGN;
  47
  48         if (size < align) {
  49             align = getpagesize();
  50         }
  51         ptr = qemu_memalign(align, size);
  52         trace_qemu_vmalloc(size, ptr);
  53         return ptr;
  54     }
  55
  56 === Declaring trace events ===
  57
  58 The "tracetool" script produces the trace.h header file which is included by
  59 every source file that uses trace events.  Since many source files include
  60 trace.h, it uses a minimum of types and other header files included to keep the
  61 namespace clean and compile times and dependencies down.
  62
  63 Trace events should use types as follows:
  64
  65  * Use stdint.h types for fixed-size types.  Most offsets and guest memory
  66    addresses are best represented with uint32_t or uint64_t.  Use fixed-size
  67    types over primitive types whose size may change depending on the host
  68    (32-bit versus 64-bit) so trace events don't truncate values or break
  69    the build.
  70
  71  * Use void * for pointers to structs or for arrays.  The trace.h header
  72    cannot include all user-defined struct declarations and it is therefore
  73    necessary to use void * for pointers to structs.
  74
  75  * For everything else, use primitive scalar types (char, int, long) with the
  76    appropriate signedness.
  77
  78 Format strings should reflect the types defined in the trace event.  Take
  79 special care to use PRId64 and PRIu64 for int64_t and uint64_t types,
  80 respectively.  This ensures portability between 32- and 64-bit platforms.
  81
  82 === Hints for adding new trace events ===
  83
  84 1. Trace state changes in the code.  Interesting points in the code usually
  85    involve a state change like starting, stopping, allocating, freeing.  State
  86    changes are good trace events because they can be used to understand the
  87    execution of the system.
  88
  89 2. Trace guest operations.  Guest I/O accesses like reading device registers
  90    are good trace events because they can be used to understand guest
  91    interactions.
  92
  93 3. Use correlator fields so the context of an individual line of trace output
  94    can be understood.  For example, trace the pointer returned by malloc and
  95    used as an argument to free.  This way mallocs and frees can be matched up.
  96    Trace events with no context are not very useful.
  97
  98 4. Name trace events after their function.  If there are multiple trace events
  99    in one function, append a unique distinguisher at the end of the name.
 100
 101 5. If specific trace events are going to be called a huge number of times, this
 102    might have a noticeable performance impact even when the trace events are
 103    programmatically disabled. In this case you should declare the trace event
 104    with the "disable" property, which will effectively disable it at compile
 105    time (using the "nop" backend).
 106
 107 == Generic interface and monitor commands ==
 108
 109 You can programmatically query and control the dynamic state of trace events
 110 through a backend-agnostic interface:
 111
 112 * trace_print_events
 113
 114 * trace_event_set_state
 115   Enables or disables trace events at runtime inside QEMU.
 116   The function returns "true" if the state of the event has been successfully
 117   changed, or "false" otherwise:
 118
 119     #include "trace/control.h"
 120
 121     trace_event_set_state("virtio_irq", true); /* enable */
 122     [...]
 123     trace_event_set_state("virtio_irq", false); /* disable */
 124
 125 Note that some of the backends do not provide an implementation for this
 126 interface, in which case QEMU will just print a warning.
 127
 128 This functionality is also provided through monitor commands:
 129
 130 * info trace-events
 131   View available trace events and their state.  State 1 means enabled, state 0
 132   means disabled.
 133
 134 * trace-event NAME on|off
 135   Enable/disable a given trace event or a group of events having common prefix
 136   through wildcard.
 137
 138 The "-trace events=<file>" command line argument can be used to enable the
 139 events listed in <file> from the very beginning of the program. This file must
 140 contain one event name per line.
 141
 142 A basic wildcard matching is supported in both the monitor command "trace
 143 -event" and the events list file. That means you can enable/disable the events
 144 having a common prefix in a batch. For example, virtio-blk trace events could
 145 be enabled using:
 146   trace-event virtio_blk_* on
 147
 148 == Trace backends ==
 149
 150 The "tracetool" script automates tedious trace event code generation and also
 151 keeps the trace event declarations independent of the trace backend.  The trace
 152 events are not tightly coupled to a specific trace backend, such as LTTng or
 153 SystemTap.  Support for trace backends can be added by extending the "tracetool"
 154 script.
 155
 156 The trace backend is chosen at configure time and only one trace backend can
 157 be built into the binary:
 158
 159     ./configure --trace-backend=simple
 160
 161 For a list of supported trace backends, try ./configure --help or see below.
 162
 163 The following subsections describe the supported trace backends.
 164
 165 === Nop ===
 166
 167 The "nop" backend generates empty trace event functions so that the compiler
 168 can optimize out trace events completely.  This is the default and imposes no
 169 performance penalty.
 170
 171 Note that regardless of the selected trace backend, events with the "disable"
 172 property will be generated with the "nop" backend.
 173
 174 === Stderr ===
 175
 176 The "stderr" backend sends trace events directly to standard error.  This
 177 effectively turns trace events into debug printfs.
 178
 179 This is the simplest backend and can be used together with existing code that
 180 uses DPRINTF().
 181
 182 === Simpletrace ===
 183
 184 The "simple" backend supports common use cases and comes as part of the QEMU
 185 source tree.  It may not be as powerful as platform-specific or third-party
 186 trace backends but it is portable.  This is the recommended trace backend
 187 unless you have specific needs for more advanced backends.
 188
 189 The "simple" backend currently does not capture string arguments, it simply
 190 records the char* pointer value instead of the string that is pointed to.
 191
 192 ==== Monitor commands ====
 193
 194 * info trace
 195   Display the contents of trace buffer.  This command dumps the trace buffer
 196   with simple formatting.  For full pretty-printing, use the simpletrace.py
 197   script on a binary trace file.
 198
 199   The trace buffer is written into until full.  The full trace buffer is
 200   flushed and emptied.  This means the 'info trace' will display few or no
 201   entries if the buffer has just been flushed.
 202
 203 * trace-file on|off|flush|set <path>
 204   Enable/disable/flush the trace file or set the trace file name.
 205
 206 ==== Analyzing trace files ====
 207
 208 The "simple" backend produces binary trace files that can be formatted with the
 209 simpletrace.py script.  The script takes the "trace-events" file and the binary
 210 trace:
 211
 212     ./simpletrace.py trace-events trace-12345
 213
 214 You must ensure that the same "trace-events" file was used to build QEMU,
 215 otherwise trace event declarations may have changed and output will not be
 216 consistent.
 217
 218 === LTTng Userspace Tracer ===
 219
 220 The "ust" backend uses the LTTng Userspace Tracer library.  There are no
 221 monitor commands built into QEMU, instead UST utilities should be used to list,
 222 enable/disable, and dump traces.
 223
 224 === SystemTap ===
 225
 226 The "dtrace" backend uses DTrace sdt probes but has only been tested with
 227 SystemTap.  When SystemTap support is detected a .stp file with wrapper probes
 228 is generated to make use in scripts more convenient.  This step can also be
 229 performed manually after a build in order to change the binary name in the .stp
 230 probes:
 231
 232     scripts/tracetool --dtrace --stap \
 233                       --binary path/to/qemu-binary \
 234                       --target-type system \
 235                       --target-arch x86_64 \
 236                       <trace-events >qemu.stp