| 1 |
==================== |
| 2 |
Runtime Verification |
| 3 |
==================== |
| 4 |
|
| 5 |
Runtime Verification (RV) is a lightweight (yet rigorous) method that |
| 6 |
complements classical exhaustive verification techniques (such as *model |
| 7 |
checking* and *theorem proving*) with a more practical approach for complex |
| 8 |
systems. |
| 9 |
|
| 10 |
Instead of relying on a fine-grained model of a system (e.g., a |
| 11 |
re-implementation at instruction level), RV works by analyzing the trace of the |
| 12 |
system's actual execution, comparing it against a formal specification of |
| 13 |
the system behavior. |
| 14 |
|
| 15 |
The main advantage is that RV can give precise information on the runtime |
| 16 |
behavior of the monitored system, without the pitfalls of developing models |
| 17 |
that require a re-implementation of the entire system in a modeling language. |
| 18 |
Moreover, given an efficient monitoring method, it is possible to execute an |
| 19 |
*online* verification of a system, enabling the *reaction* for unexpected |
| 20 |
events, avoiding, for example, the propagation of a failure on safety-critical |
| 21 |
systems. |
| 22 |
|
| 23 |
Runtime Monitors and Reactors |
| 24 |
============================= |
| 25 |
|
| 26 |
A monitor is the central part of the runtime verification of a system. The |
| 27 |
monitor stands in between the formal specification of the desired (or |
| 28 |
undesired) behavior, and the trace of the actual system. |
| 29 |
|
| 30 |
In Linux terms, the runtime verification monitors are encapsulated inside the |
| 31 |
*RV monitor* abstraction. A *RV monitor* includes a reference model of the |
| 32 |
system, a set of instances of the monitor (per-cpu monitor, per-task monitor, |
| 33 |
and so on), and the helper functions that glue the monitor to the system via |
| 34 |
trace, as depicted below:: |
| 35 |
|
| 36 |
Linux +---- RV Monitor ----------------------------------+ Formal |
| 37 |
Realm | | Realm |
| 38 |
+-------------------+ +----------------+ +-----------------+ |
| 39 |
| Linux kernel | | Monitor | | Reference | |
| 40 |
| Tracing | -> | Instance(s) | <- | Model | |
| 41 |
| (instrumentation) | | (verification) | | (specification) | |
| 42 |
+-------------------+ +----------------+ +-----------------+ |
| 43 |
| | | |
| 44 |
| V | |
| 45 |
| +----------+ | |
| 46 |
| | Reaction | | |
| 47 |
| +--+--+--+-+ | |
| 48 |
| | | | | |
| 49 |
| | | +-> trace output ? | |
| 50 |
+------------------------|--|----------------------+ |
| 51 |
| +----> panic ? |
| 52 |
+-------> <user-specified> |
| 53 |
|
| 54 |
In addition to the verification and monitoring of the system, a monitor can |
| 55 |
react to an unexpected event. The forms of reaction can vary from logging the |
| 56 |
event occurrence to the enforcement of the correct behavior to the extreme |
| 57 |
action of taking a system down to avoid the propagation of a failure. |
| 58 |
|
| 59 |
In Linux terms, a *reactor* is an reaction method available for *RV monitors*. |
| 60 |
By default, all monitors should provide a trace output of their actions, |
| 61 |
which is already a reaction. In addition, other reactions will be available |
| 62 |
so the user can enable them as needed. |
| 63 |
|
| 64 |
For further information about the principles of runtime verification and |
| 65 |
RV applied to Linux: |
| 66 |
|
| 67 |
Bartocci, Ezio, et al. *Introduction to runtime verification.* In: Lectures on |
| 68 |
Runtime Verification. Springer, Cham, 2018. p. 1-33. |
| 69 |
|
| 70 |
Falcone, Ylies, et al. *A taxonomy for classifying runtime verification tools.* |
| 71 |
In: International Conference on Runtime Verification. Springer, Cham, 2018. p. |
| 72 |
241-262. |
| 73 |
|
| 74 |
De Oliveira, Daniel Bristot. *Automata-based formal analysis and |
| 75 |
verification of the real-time Linux kernel.* Ph.D. Thesis, 2020. |
| 76 |
|
| 77 |
Online RV monitors |
| 78 |
================== |
| 79 |
|
| 80 |
Monitors can be classified as *offline* and *online* monitors. *Offline* |
| 81 |
monitor process the traces generated by a system after the events, generally by |
| 82 |
reading the trace execution from a permanent storage system. *Online* monitors |
| 83 |
process the trace during the execution of the system. Online monitors are said |
| 84 |
to be *synchronous* if the processing of an event is attached to the system |
| 85 |
execution, blocking the system during the event monitoring. On the other hand, |
| 86 |
an *asynchronous* monitor has its execution detached from the system. Each type |
| 87 |
of monitor has a set of advantages. For example, *offline* monitors can be |
| 88 |
executed on different machines but require operations to save the log to a |
| 89 |
file. In contrast, *synchronous online* method can react at the exact moment |
| 90 |
a violation occurs. |
| 91 |
|
| 92 |
Another important aspect regarding monitors is the overhead associated with the |
| 93 |
event analysis. If the system generates events at a frequency higher than the |
| 94 |
monitor's ability to process them in the same system, only the *offline* |
| 95 |
methods are viable. On the other hand, if the tracing of the events incurs |
| 96 |
on higher overhead than the simple handling of an event by a monitor, then a |
| 97 |
*synchronous online* monitors will incur on lower overhead. |
| 98 |
|
| 99 |
Indeed, the research presented in: |
| 100 |
|
| 101 |
De Oliveira, Daniel Bristot; Cucinotta, Tommaso; De Oliveira, Romulo Silva. |
| 102 |
*Efficient formal verification for the Linux kernel.* In: International |
| 103 |
Conference on Software Engineering and Formal Methods. Springer, Cham, 2019. |
| 104 |
p. 315-332. |
| 105 |
|
| 106 |
Shows that for Deterministic Automata models, the synchronous processing of |
| 107 |
events in-kernel causes lower overhead than saving the same events to the trace |
| 108 |
buffer, not even considering collecting the trace for user-space analysis. |
| 109 |
This motivated the development of an in-kernel interface for online monitors. |
| 110 |
|
| 111 |
For further information about modeling of Linux kernel behavior using automata, |
| 112 |
see: |
| 113 |
|
| 114 |
De Oliveira, Daniel B.; De Oliveira, Romulo S.; Cucinotta, Tommaso. *A thread |
| 115 |
synchronization model for the PREEMPT_RT Linux kernel.* Journal of Systems |
| 116 |
Architecture, 2020, 107: 101729. |
| 117 |
|
| 118 |
The user interface |
| 119 |
================== |
| 120 |
|
| 121 |
The user interface resembles the tracing interface (on purpose). It is |
| 122 |
currently at "/sys/kernel/tracing/rv/". |
| 123 |
|
| 124 |
The following files/folders are currently available: |
| 125 |
|
| 126 |
**available_monitors** |
| 127 |
|
| 128 |
- Reading list the available monitors, one per line |
| 129 |
|
| 130 |
For example:: |
| 131 |
|
| 132 |
# cat available_monitors |
| 133 |
wip |
| 134 |
wwnr |
| 135 |
|
| 136 |
**available_reactors** |
| 137 |
|
| 138 |
- Reading shows the available reactors, one per line. |
| 139 |
|
| 140 |
For example:: |
| 141 |
|
| 142 |
# cat available_reactors |
| 143 |
nop |
| 144 |
panic |
| 145 |
printk |
| 146 |
|
| 147 |
**enabled_monitors**: |
| 148 |
|
| 149 |
- Reading lists the enabled monitors, one per line |
| 150 |
- Writing to it enables a given monitor |
| 151 |
- Writing a monitor name with a '!' prefix disables it |
| 152 |
- Truncating the file disables all enabled monitors |
| 153 |
|
| 154 |
For example:: |
| 155 |
|
| 156 |
# cat enabled_monitors |
| 157 |
# echo wip > enabled_monitors |
| 158 |
# echo wwnr >> enabled_monitors |
| 159 |
# cat enabled_monitors |
| 160 |
wip |
| 161 |
wwnr |
| 162 |
# echo '!wip' >> enabled_monitors |
| 163 |
# cat enabled_monitors |
| 164 |
wwnr |
| 165 |
# echo > enabled_monitors |
| 166 |
# cat enabled_monitors |
| 167 |
# |
| 168 |
|
| 169 |
Note that it is possible to enable more than one monitor concurrently. |
| 170 |
|
| 171 |
**monitoring_on** |
| 172 |
|
| 173 |
This is an on/off general switcher for monitoring. It resembles the |
| 174 |
"tracing_on" switcher in the trace interface. |
| 175 |
|
| 176 |
- Writing "0" stops the monitoring |
| 177 |
- Writing "1" continues the monitoring |
| 178 |
- Reading returns the current status of the monitoring |
| 179 |
|
| 180 |
Note that it does not disable enabled monitors but stop the per-entity |
| 181 |
monitors monitoring the events received from the system. |
| 182 |
|
| 183 |
**reacting_on** |
| 184 |
|
| 185 |
- Writing "0" prevents reactions for happening |
| 186 |
- Writing "1" enable reactions |
| 187 |
- Reading returns the current status of the reaction |
| 188 |
|
| 189 |
**monitors/** |
| 190 |
|
| 191 |
Each monitor will have its own directory inside "monitors/". There the |
| 192 |
monitor-specific files will be presented. The "monitors/" directory resembles |
| 193 |
the "events" directory on tracefs. |
| 194 |
|
| 195 |
For example:: |
| 196 |
|
| 197 |
# cd monitors/wip/ |
| 198 |
# ls |
| 199 |
desc enable |
| 200 |
# cat desc |
| 201 |
wakeup in preemptive per-cpu testing monitor. |
| 202 |
# cat enable |
| 203 |
0 |
| 204 |
|
| 205 |
**monitors/MONITOR/desc** |
| 206 |
|
| 207 |
- Reading shows a description of the monitor *MONITOR* |
| 208 |
|
| 209 |
**monitors/MONITOR/enable** |
| 210 |
|
| 211 |
- Writing "0" disables the *MONITOR* |
| 212 |
- Writing "1" enables the *MONITOR* |
| 213 |
- Reading return the current status of the *MONITOR* |
| 214 |
|
| 215 |
**monitors/MONITOR/reactors** |
| 216 |
|
| 217 |
- List available reactors, with the select reaction for the given *MONITOR* |
| 218 |
inside "[]". The default one is the nop (no operation) reactor. |
| 219 |
- Writing the name of a reactor enables it to the given MONITOR. |
| 220 |
|
| 221 |
For example:: |
| 222 |
|
| 223 |
# cat monitors/wip/reactors |
| 224 |
[nop] |
| 225 |
panic |
| 226 |
printk |
| 227 |
# echo panic > monitors/wip/reactors |
| 228 |
# cat monitors/wip/reactors |
| 229 |
nop |
| 230 |
[panic] |
| 231 |
printk |