lttng-event-rule - Man Page

Common LTTng event rule specification

Synopsis

Specify an event rule to match Linux kernel tracepoint or system call events:

--type=(kernel:tracepoint | kernel:syscall[:entry|:exit|:entry+exit])]
[--name=NAME] [--filter=EXPR]

Specify an event rule to match Linux kernel kprobe or user space probe events:

--type=(kernel:kprobe | kernel:uprobe) --location=LOC
[--event-name=EVENTNAME]

Specify an event rule to match user space tracepoint events:

--type=user:tracepoint [--name=NAME] [--exclude-name=XNAME]...
[--log-level=(LOGLEVEL | LOGLEVEL.. | ..)] [--filter=EXPR]

Specify an event rule to match Java/Python logging events:

--type=(jul | log4j | python):logging [--name=NAME]
[--log-level=(LOGLEVEL | LOGLEVEL.. | ..)] [--filter=EXPR]

Description

This manual page shows how to specify an LTTng event rule on the command line.

As of LTTng 2.13.7, the command-line options documented here only apply to the event-rule-matches trigger condition specifier (see lttng-add-trigger(1)).

See lttng-concepts(7) to learn more about instrumentation points, events, and event rules.
Note

This manual page only describes the common event rule options. The lttng(1) commands which require an event rule specification may accept or require other options and arguments, depending on the context.

For example, the lttng-add-trigger(1) command also accepts --capture options with the event-rule-matches trigger condition.

Overview of event rule condtions

For LTTng to emit an event EE must satisfy all the conditions of an event rule, that is:

  • The instrumentation point from which LTTng creates E has a specific type.

    See the “Instrumentation point type condition” section below.

  • A pattern matches the name of E while another pattern doesn’t.

    See the “Event name condition” section below.

  • The log level of the instrumentation point from which LTTng creates E is at least as severe as some value, or is exactly some value.

    See the “Instrumentation point log level condition” section below.

  • The fields of the payload of E and the current context fields satisfy a filter expression.

    See the “Event payload and context filter condition” section below.

The dedicated command-line options of most conditions are optional: if you don’t specify the option, the associated condition is always satisfied.

Instrumentation point type condition

An event E satisfies the instrumentation point type condition of an event rule if the instrumentation point from which LTTng creates E is, depending on the argument of the --type option:

kernel:tracepoint

An LTTng kernel tracepoint, that is, a statically defined point in the source code of the kernel image or of a kernel module with LTTng kernel tracer macros.

List the available Linux kernel tracepoints with lttng list --kernel. See lttng-list(1) to learn more.

kernel:syscall:entry, kernel:syscall:exit, kernel:syscall:entry+exit

The entry, exit, or entry and exit of a Linux kernel system call.

List the available Linux kernel system call instrumentation points with lttng list --kernel --syscall. See lttng-list(1) to learn more.

kernel:kprobe

A Linux kprobe, that is, a single probe dynamically placed in the compiled kernel code.

You must specify the kprobe location with the --location option.

The payload of a Linux kprobe event is empty.

kernel:uprobe

A Linux user space probe, that is, a single probe dynamically placed at the entry of a compiled user space application/library function through the kernel.

LTTng 2.13.7 supports the ELF and SystemTap User-level Statically Defined Tracing (USDT; a DTrace-style marker) probing methods. LTTng only supports USDT probes which are NOT reference-counted.

You must specify the user space probe location with the --location option.

The payload of a Linux user space probe event is empty.

user:tracepoint

An LTTng user space tracepoint, that is, a statically defined point in the source code of a C/C++ application/library with LTTng user space tracer macros.

List the available user space tracepoints with lttng list --userspace. See lttng-list(1) to learn more.

jul:logging

A java.util.logging logging statement.

List the available java.util.logging loggers with lttng list --jul See lttng-list(1) to learn more.

log4j:logging

An Apache log4j logging statement.

List the available Apache log4j loggers with lttng list --log4j See lttng-list(1) to learn more.

python:logging

A Python logging statement.

List the available Python loggers with lttng list --python See lttng-list(1) to learn more.

Event name condition

An event E satisfies the event name condition of an event rule ER if the two following statements are true:

  • You don’t specify the --name=NAME option or, depending on the instrumentation type condition (see the “Instrumentation point type condition” section above) of ER, NAME matches:
kernel:tracepoint, user:tracepoint

The full name of the tracepoint from which LTTng creates E.

Note that the full name of a user space tracepoint is PROVIDER:NAME, where PROVIDER is the tracepoint provider name and NAME is the tracepoint name.

jul:logging, log4j:logging, python:logging

The name of the Java or Python logger from which LTTng creates E.

kernel:syscall:entry, kernel:syscall:exit, kernel:syscall:entry+exit

The name of the system call, without any sys_ prefix, from which LTTng creates E.

  • You don’t specify any --exclude-name=XNAME option or none of the XNAME arguments matches the full name of the user space tracepoint from which LTTng creates E.

    The --exclude-name option is only available with the --type=user:tracepoint option.

This condition is only meaningful for the LTTng tracepoint, logging statement, and Linux system call instrumentation point types: it’s always satisfied for the other types.

In all cases, NAME and XNAME are globbing patterns: the * character means “match anything”. To match a literal * character, use \*.
Important

Make sure to single-quote NAME and XNAME when they contain the * character and when you run an lttng(1) command from a shell.

As of LTTng 2.13.7, not specifying the --name option is equivalent to specifying --name='*´, but this default may change in the future.

Instrumentation point log level condition

An event E satisfies the instrumentation point log level condition of an event rule if either:

  • The log level of the LTTng user space tracepoint or logging statement from which LTTng creates E is:

With the --log-level=LOGLEVEL.. option

At least as severe as LOGLEVEL.

With the --log-level=LOGLEVEL option

Exactly LOGLEVEL.

As of LTTng 2.13.7, the ..LOGLEVEL and LOGLEVEL..LOGLEVEL formats are NOT supported.

This condition is only meaningful for the LTTng user space tracepoint and logging statement instrumentation point types: it’s always satisfied for other types.

The available values of LOGLEVEL are, depending on the argument of the --type option, from the most to the least severe:

user:tracepoint
  • EMERG (0)
  • ALERT (1)
  • CRIT (2)
  • ERR (3)
  • WARNING (4)
  • NOTICE (5)
  • INFO (6)
  • DEBUG_SYSTEM (7)
  • DEBUG_PROGRAM (8)
  • DEBUG_PROCESS (9)
  • DEBUG_MODULE (10)
  • DEBUG_UNIT (11)
  • DEBUG_FUNCTION (12)
  • DEBUG_LINE (13)
  • DEBUG (14)
jul:logging
  • OFF (INT32_MAX)
  • SEVERE (1000)
  • WARNING (900)
  • INFO (800)
  • CONFIG (700)
  • FINE (500)
  • FINER (400)
  • FINEST (300)
  • ALL (INT32_MIN)
log4j:logging
  • OFF (INT32_MAX)
  • FATAL (50000)
  • ERROR (40000)
  • WARN (30000)
  • INFO (20000)
  • DEBUG (10000)
  • TRACE (5000)
  • ALL (INT32_MIN)
python:logging
  • CRITICAL (50)
  • ERROR (40)
  • WARNING (30)
  • INFO (20)
  • DEBUG (10)
  • NOTSET (0)

Event payload and context filter condition

An event E satisfies the event payload and context filter condition of an event rule if the --filter=EXPR option is missing or if EXPR is true.

This condition is only meaningful for the LTTng tracepoint and Linux system call instrumentation point types: it’s always satisfied for other types.

EXPR can contain references to the payload fields of E and to the current context fields.
Important

Make sure to single-quote EXPR when you run an lttng(1) command from a shell, as filter expressions typically include characters having a special meaning for most shells.

The expected syntax of EXPR is similar to the syntax of a C language conditional expression (an expression which an if statement can evaluate), but there are a few differences:

  • A NAME expression identifies an event payload field named NAME (a C identifier).

    Use the C language dot and square bracket notations to access nested structure and array/sequence fields. You can only use a constant, positive integer number within square brackets. If the index is out of bounds, EXPR is false.

    The value of an enumeration field is an integer.

    When a field expression doesn’t exist, EXPR is false.

    Examples: my_field, target_cpu, seq[7], msg.user[1].data[2][17].

  • A $ctx.TYPE expression identifies the statically-known context field having the type TYPE (a C identifier).

    List the available statically-known context field names with the lttng-add-context(1) command.

    When a field expression doesn’t exist, EXPR is false.

    Examples: $ctx.prio, $ctx.preemptible, $ctx.perf:cpu:stalled-cycles-frontend.

  • A $app.PROVIDER:TYPE expression identifies the application-specific context field having the type TYPE (a C identifier) from the provider PROVIDER (a C identifier).

    When a field expression doesn’t exist, EXPR is false.

    Example: $app.server:cur_user.

  • Compare strings, either string fields or string literals (double-quoted), with the == and != operators.

    When comparing to a string literal, the * character means “match anything”. To match a literal * character, use \*.

    Examples: my_field == "user34", my_field == my_other_field, my_field == "192.168.*".

  • The precedence table of the operators which are supported in EXPR is as follows. In this table, the highest precedence is 1:

    PrecedenceOperatorDescriptionAssociativity
    1-Unary minusRight-to-left
    1+Unary plusRight-to-left
    1!Logical NOTRight-to-left
    1~Bitwise NOTRight-to-left
    2<<Bitwise left shiftLeft-to-right
    2>>Bitwise right shiftLeft-to-right
    3&Bitwise ANDLeft-to-right
    4^Bitwise XORLeft-to-right
    5|Bitwise ORLeft-to-right
    6<Less thanLeft-to-right
    6<=Less than or equal toLeft-to-right
    6>Greater thanLeft-to-right
    6>=Greater than or equal toLeft-to-right
    7==Equal toLeft-to-right
    7!=Not equal toLeft-to-right
    8&&Logical ANDLeft-to-right
    9||Logical ORLeft-to-right

    Parentheses are supported to bypass the default order.
    Important

    Unlike the C language, the bitwise AND and OR operators (& and |) in EXPR take precedence over relational operators (<, <=, >, >=, ==, and !=). This means the expression 2 & 2 == 2 is true while the equivalent C expression is false.

    The arithmetic operators are NOT supported.

    LTTng first casts all integer constants and fields to signed 64-bit integers. The representation of negative integers is two’s complement. This means that, for example, the signed 8-bit integer field 0xff (-1) becomes 0xffffffffffffffff (still -1) once casted.

    Before a bitwise operator is applied, LTTng casts all its operands to unsigned 64-bit integers, and then casts the result back to a signed 64-bit integer. For the bitwise NOT operator, it’s the equivalent of this C expression:

    (int64_t) ~((uint64_t) val)

    For the binary bitwise operators, it’s the equivalent of those C expressions:

    (int64_t) ((uint64_t) lhs >> (uint64_t) rhs)
    (int64_t) ((uint64_t) lhs << (uint64_t) rhs)
    (int64_t) ((uint64_t) lhs & (uint64_t) rhs)
    (int64_t) ((uint64_t) lhs ^ (uint64_t) rhs)
    (int64_t) ((uint64_t) lhs | (uint64_t) rhs)

    If the right-hand side of a bitwise shift operator (<< and >>) is not in the [0, 63] range, then EXPR is false.

EXPR examples:

msg_id == 23 && size >= 2048
$ctx.procname == "lttng*" && (!flag || poel < 34)
$app.my_provider:my_context == 17.34e9 || some_enum >= 14
$ctx.cpu_id == 2 && filename != "*.log"
eax_reg & 0xff7 == 0x240 && x[4] >> 12 <= 0x1234

Migration from a recording event rule specification

Since LTTng 2.13, what this manual page documents is the standard, common way to specify an LTTng event rule.

With the lttng-enable-event(1) command, you also specify an event rule, but with deprecated options and arguments.

The following table shows how to translate from the lttng-enable-event(1) options and arguments to the common event rule specification options:

Recording event rule option(s)/argument(s)Common event rule option(s)
--kernel and --tracepoint--type=kernel:tracepoint
--kernel and --syscall--type=kernel:syscall:entry+exit
--probe=LOC and RECORDNAME (non-option)--type=kernel:kprobe, --location=LOC, and --event-name=RECORDNAME
--userspace-probe=LOC and RECORDNAME (non-option)--type=kernel:uprobe, --location=LOC, and --event-name=RECORDNAME
--function=LOC and RECORDNAME (non-option)Not available as of LTTng 2.13.7
--userspace and --tracepoint--type=user:tracepoint
--jul and --tracepoint--type=jul:logging
--log4j and --tracepoint--type=log4j:logging
--python and --tracepoint--type=python:logging
NAME (non-option)--name=NAME
--all--name='*´ or no --name option
--exclude=XNAME[,XNAME]...--exclude-name=XNAME for each XNAME
--loglevel=LOGLEVEL--log-level=LOGLEVEL..
--loglevel-only=LOGLEVEL--log-level=LOGLEVEL
--filter=EXPR--filter=EXPR

Options

Instrumentation point type condition

See the “Instrumentation point type condition” section above.

-E NAME, --event-name=NAME

With the --type=kernel:kprobe or --type=kernel:uprobe option, set the name of the emitted events to NAME instead of the LOC argument of the --location=LOC option.

Defaulting to LOC is specific to LTTng 2.13.7 and may change in the future.

-L LOC, --location=LOC

With the --type=kernel:kprobe option

Set the location of the Linux kprobe to insert to LOC.

LOC is one of:

  • An address (0x hexadecimal prefix supported).
  • A symbol name.
  • A symbol name and an offset (SYMBOL+OFFSET format).

With the --type=kernel:uprobe option

Set the location of the user space probe to insert to LOC.

LOC is one of:

[elf:]PATH:SYMBOL

An available symbol within a user space application or library.

PATH

Application or library path.

One of:

  • An absolute path.
  • A relative path.
  • The name of an application as found in the directories listed in the PATH environment variable.
SYMBOL

Symbol name of the function of which to instrument the entry.

SYMBOL can be any defined code symbol in the output of the nm(1) command, including with its --dynamic option, which lists dynamic symbols.

As of LTTng 2.13.7, not specifying elf: is equivalent to specifying it, but this default may change in the future.

Examples:

  • /usr/lib/libc.so.6:malloc
  • ./myapp:createUser
  • elf:httpd:ap_run_open_htaccess
sdt:PATH:PROVIDER:NAME

A SystemTap User-level Statically Defined Tracing (USDT) probe within a user space application or library.

PATH

Application or library path.

This can be:

  • An absolute path.
  • A relative path.
  • The name of an application as found in the directories listed in the PATH environment variable.
PROVIDER, NAME

USDT provider and probe names.

For example, with the following USDT probe:

DTRACE_PROBE2("server", "accept_request",
              request_id, ip_addr);

The provider/probe name pair is server:accept_request.

Example: sdt:./build/server:server:accept_request

-t TYPE, --type=TYPE

Only match events which LTTng creates from an instrumentation point having the type TYPE.

TYPE is one of:

kernel:tracepoint

LTTng kernel tracepoint.

As of LTTng 2.13.7, kernel is an alias, but this may change in the future.

user:tracepoint

LTTng user space tracepoint.

As of LTTng 2.13.7, user is an alias, but this may change in the future.

kernel:syscall:entry

Linux system call entry.

As of LTTng 2.13.7, syscall:entry is an alias, but this may change in the future.

kernel:syscall:exit

Linux system call exit.

As of LTTng 2.13.7, syscall:exit is an alias, but this may change in the future.

kernel:syscall:entry+exit

Linux system call entry and exit (two distinct instrumentation points).

As of LTTng 2.13.7, the following are aliases, but this may change in the future:

  • syscall:entry+exit
  • kernel:syscall
  • syscall
kernel:kprobe

Linux kprobe.

As of LTTng 2.13.7, kprobe is an alias, but this may change in the future.

You must specify the location of the kprobe to insert with the --location option.

You may specify the name of the emitted events with the --event-name option.

kernel:uprobe

Linux user space probe.

You must specify the location of the user space probe to insert with the --location option.

You may specify the name of the emitted events with the --event-name option.

jul:logging

java.util.logging logging statement.

As of LTTng 2.13.7, jul is an alias, but this may change in the future.

log4j:logging

Apache log4j logging statement.

As of LTTng 2.13.7, log4j is an alias, but this may change in the future.

python:logging

Python logging statement.

As of LTTng 2.13.7, python is an alias, but this may change in the future.

Event name condition

See the “Event name condition” section above.

-n NAME, --name=NAME

Only match events of which NAME matches, depending on the argument of the --type option:

kernel:tracepoint, user:tracepoint

The full name of the LTTng tracepoint.

jul:logging, log4j:logging, python:logging

The Java or Python logger name.

kernel:syscall:entry, kernel:syscall:exit, kernel:syscall:entry+exit

The name of the system call, without any sys_ prefix.

This option is NOT available with other instrumentation point types.

As of LTTng 2.13.7, not specifying this option is equivalent to specifying --name='*´ (when it applies), but this default may change in the future.

-x XNAME, --exclude-name=XNAME

Only match events of which XNAME does NOT match the full name of the LTTng user space tracepoint.

Only available with the --type=user:tracepoint option.

NAME and XNAME are globbing patterns: the * character means “match anything”. To match a literal * character, use \*.

Instrumentation point log level condition

See the “Instrumentation point log level condition” section above.

-l LOGLEVELSPEC, --log-level=LOGLEVELSPEC

Only match events of which the log level of the LTTng tracepoint or logging statement is, depending on the format of LOGLEVELSPEC:

LOGLEVEL..

At least as severe as LOGLEVEL.

LOGLEVEL

Exactly LOGLEVEL.

..

Anything.

This option is NOT available with the following options:

  • --type=kernel:tracepoint
  • --type=kernel:syscall:entry
  • --type=kernel:syscall:exit
  • --type=kernel:syscall:entry+exit
  • --type=kernel:kprobe
  • --type=kernel:uprobe

    As of LTTng 2.13.7, not specifying this option is equivalent to specifying --log-level=.. (when it applies), but this default may change in the future.

Event payload and context filter condition

See the “Event payload and context filter condition” section above.

-f EXPR, --filter=EXPR

Only match events of which EXPR, which can contain references to event payload and current context fields, is true.

This option is only available with the following options:

Resources

Thanks

Special thanks to Michel Dagenais and the DORSAL laboratory <http://www.dorsal.polymtl.ca/> at École Polytechnique de Montréal for the LTTng journey.

Also thanks to the Ericsson teams working on tracing which helped us greatly with detailed bug reports and unusual test cases.

See Also

lttng(1), lttng-add-trigger(1), lttng-list(1), lttng-concepts(7)

Referenced By

lttng-add-trigger(1), lttng-concepts(7).

18 May 2021 LTTng 2.13.7 LTTng Manual