Documentation/trace/uprobetracer.txt

   1             Uprobe-tracer: Uprobe-based Event Tracing
   2             =========================================
   3
   4            Documentation written by Srikar Dronamraju
   5
   6
   7 Overview
   8 --------
   9 Uprobe based trace events are similar to kprobe based trace events.
  10 To enable this feature, build your kernel with CONFIG_UPROBE_EVENT=y.
  11
  12 Similar to the kprobe-event tracer, this doesn't need to be activated via
  13 current_tracer. Instead of that, add probe points via
  14 /sys/kernel/debug/tracing/uprobe_events, and enable it via
  15 /sys/kernel/debug/tracing/events/uprobes/<EVENT>/enabled.
  16
  17 However unlike kprobe-event tracer, the uprobe event interface expects the
  18 user to calculate the offset of the probepoint in the object.
  19
  20 Synopsis of uprobe_tracer
  21 -------------------------
  22   p[:[GRP/]EVENT] PATH:OFFSET [FETCHARGS] : Set a uprobe
  23   r[:[GRP/]EVENT] PATH:OFFSET [FETCHARGS] : Set a return uprobe (uretprobe)
  24   -:[GRP/]EVENT                           : Clear uprobe or uretprobe event
  25
  26   GRP           : Group name. If omitted, "uprobes" is the default value.
  27   EVENT         : Event name. If omitted, the event name is generated based
  28                   on PATH+OFFSET.
  29   PATH          : Path to an executable or a library.
  30   OFFSET        : Offset where the probe is inserted.
  31
  32   FETCHARGS     : Arguments. Each probe can have up to 128 args.
  33    %REG         : Fetch register REG
  34    @ADDR        : Fetch memory at ADDR (ADDR should be in userspace)
  35    @+OFFSET     : Fetch memory at OFFSET (OFFSET from same file as PATH)
  36    $stackN      : Fetch Nth entry of stack (N >= 0)
  37    $stack       : Fetch stack address.
  38    $retval      : Fetch return value.(*)
  39    $comm        : Fetch current task comm.
  40    +|-offs(FETCHARG) : Fetch memory at FETCHARG +|- offs address.(**)
  41    NAME=FETCHARG     : Set NAME as the argument name of FETCHARG.
  42    FETCHARG:TYPE     : Set TYPE as the type of FETCHARG. Currently, basic types
  43                        (u8/u16/u32/u64/s8/s16/s32/s64), "string" and bitfield
  44                        are supported.
  45
  46   (*) only for return probe.
  47   (**) this is useful for fetching a field of data structures.
  48
  49 Types
  50 -----
  51 Several types are supported for fetch-args. Uprobe tracer will access memory
  52 by given type. Prefix 's' and 'u' means those types are signed and unsigned
  53 respectively. Traced arguments are shown in decimal (signed) or hex (unsigned).
  54 String type is a special type, which fetches a "null-terminated" string from
  55 user space.
  56 Bitfield is another special type, which takes 3 parameters, bit-width, bit-
  57 offset, and container-size (usually 32). The syntax is;
  58
  59  b<bit-width>@<bit-offset>/<container-size>
  60
  61 For $comm, the default type is "string"; any other type is invalid.
  62
  63
  64 Event Profiling
  65 ---------------
  66 You can check the total number of probe hits and probe miss-hits via
  67 /sys/kernel/debug/tracing/uprobe_profile.
  68 The first column is event name, the second is the number of probe hits,
  69 the third is the number of probe miss-hits.
  70
  71 Usage examples
  72 --------------
  73  * Add a probe as a new uprobe event, write a new definition to uprobe_events
  74 as below: (sets a uprobe at an offset of 0x4245c0 in the executable /bin/bash)
  75
  76     echo 'p: /bin/bash:0x4245c0' > /sys/kernel/debug/tracing/uprobe_events
  77
  78  * Add a probe as a new uretprobe event:
  79
  80     echo 'r: /bin/bash:0x4245c0' > /sys/kernel/debug/tracing/uprobe_events
  81
  82  * Unset registered event:
  83
  84     echo '-:bash_0x4245c0' >> /sys/kernel/debug/tracing/uprobe_events
  85
  86  * Print out the events that are registered:
  87
  88     cat /sys/kernel/debug/tracing/uprobe_events
  89
  90  * Clear all events:
  91
  92     echo > /sys/kernel/debug/tracing/uprobe_events
  93
  94 Following example shows how to dump the instruction pointer and %ax register
  95 at the probed text address. Probe zfree function in /bin/zsh:
  96
  97     # cd /sys/kernel/debug/tracing/
  98     # cat /proc/`pgrep zsh`/maps | grep /bin/zsh | grep r-xp
  99     00400000-0048a000 r-xp 00000000 08:03 130904 /bin/zsh
 100     # objdump -T /bin/zsh | grep -w zfree
 101     0000000000446420 g    DF .text  0000000000000012  Base        zfree
 102
 103   0x46420 is the offset of zfree in object /bin/zsh that is loaded at
 104   0x00400000. Hence the command to uprobe would be:
 105
 106     # echo 'p:zfree_entry /bin/zsh:0x46420 %ip %ax' > uprobe_events
 107
 108   And the same for the uretprobe would be:
 109
 110     # echo 'r:zfree_exit /bin/zsh:0x46420 %ip %ax' >> uprobe_events
 111
 112 Please note: User has to explicitly calculate the offset of the probe-point
 113 in the object. We can see the events that are registered by looking at the
 114 uprobe_events file.
 115
 116     # cat uprobe_events
 117     p:uprobes/zfree_entry /bin/zsh:0x00046420 arg1=%ip arg2=%ax
 118     r:uprobes/zfree_exit /bin/zsh:0x00046420 arg1=%ip arg2=%ax
 119
 120 Format of events can be seen by viewing the file events/uprobes/zfree_entry/format
 121
 122     # cat events/uprobes/zfree_entry/format
 123     name: zfree_entry
 124     ID: 922
 125     format:
 126          field:unsigned short common_type;         offset:0;  size:2; signed:0;
 127          field:unsigned char common_flags;         offset:2;  size:1; signed:0;
 128          field:unsigned char common_preempt_count; offset:3;  size:1; signed:0;
 129          field:int common_pid;                     offset:4;  size:4; signed:1;
 130          field:int common_padding;                 offset:8;  size:4; signed:1;
 131
 132          field:unsigned long __probe_ip;           offset:12; size:4; signed:0;
 133          field:u32 arg1;                           offset:16; size:4; signed:0;
 134          field:u32 arg2;                           offset:20; size:4; signed:0;
 135
 136     print fmt: "(%lx) arg1=%lx arg2=%lx", REC->__probe_ip, REC->arg1, REC->arg2
 137
 138 Right after definition, each event is disabled by default. For tracing these
 139 events, you need to enable it by:
 140
 141     # echo 1 > events/uprobes/enable
 142
 143 Lets disable the event after sleeping for some time.
 144
 145     # sleep 20
 146     # echo 0 > events/uprobes/enable
 147
 148 And you can see the traced information via /sys/kernel/debug/tracing/trace.
 149
 150     # cat trace
 151     # tracer: nop
 152     #
 153     #           TASK-PID    CPU#    TIMESTAMP  FUNCTION
 154     #              | |       |          |         |
 155                  zsh-24842 [006] 258544.995456: zfree_entry: (0x446420) arg1=446420 arg2=79
 156                  zsh-24842 [007] 258545.000270: zfree_exit:  (0x446540 <- 0x446420) arg1=446540 arg2=0
 157                  zsh-24842 [002] 258545.043929: zfree_entry: (0x446420) arg1=446420 arg2=79
 158                  zsh-24842 [004] 258547.046129: zfree_exit:  (0x446540 <- 0x446420) arg1=446540 arg2=0
 159
 160 Output shows us uprobe was triggered for a pid 24842 with ip being 0x446420
 161 and contents of ax register being 79. And uretprobe was triggered with ip at
 162 0x446540 with counterpart function entry at 0x446420.