faults.xml

<faults> <!--a faults.xml root-->
	<definitions> <!--faults are defined here and can be referenced with a <fault_ref> node-->
		<fault name="meta"> <!--a <fault> has <triggers> and <actions>-->
			<triggers>
				<trigger type="META_COUNTER"> <!--a <trigger> must have a type of {META_COUNTER, VARIABLEVALUE, TIME, TIMERELATIVE, ASAP, NOP}-->
					<count>1</count> <!--the META_COUNTER here will trigger once its subtrigger fired <count> times-->
					<trigger type="VARIABLEVALUE"> <!--a <trigger> can have a sub <trigger>, awesome!-->
						<injector>core%i%</injector> <!--the trigger type VARIABLEVALUE needs an <injector> that supplies the triggerable object, bare_etiss_processor constructs "core0" ETISS-FI will replace %i% to the core id "0" by default-->
						<field>instructionPointer</field> <!--the <field> of "core%i%" should be listened for, here, the instruction pointer-->
						<value>b0</value> <!--the <value> that <field>'s value will be compared to as a hexadecimal-->
					</trigger>
				</trigger>
			</triggers>
			<actions>
				<action type="COMMAND">
					<injector>core%i%::system</injector>
					<command>rmw 0x80000 OR 0xFFFFFFFF</command>
				</action>
				<action type="COMMAND"> <!--an <action> must have a type of {BITFLIP, MASK, COMMAND, NOP, INJECTION, EJECTION, EVENT}, here, COMMAND-->
					<injector>core%i%</injector> <!--an <action> can have a different <injector> than its <trigger>s, which means, you could trigger on one Injector and inject into another-->
					<command>some string</command> <!--the COMMAND type is a custom string encoded action the <injector> has to implement, by default no custom actions are supported. Set VirtualStructs::applyCustomAction member to a std::function of your choice and handle the passed string-->
				</action>
				<action type="BITFLIP"> <!--why not have more than one <action>: Here, an additional "BITFLIP"-->
					<injector>core%i%</injector> <!--again we need an <injector>-->
					<field>X1</field> <!--BITFLIP and MASK takes a <field>, here R1 aka X1 in RISC-V, with the CPUCore all ISA-GPRs and the instructionPointer are <fields>-->
					<bit>1</bit> <!--the target <bit> number-->
				</action>
				<action type="INJECTION"> <!--now, we inject a new fault as our action.-->
						<fault_ref name="injected_fault"></fault_ref> <!--INJECTION and EJECTION <actions> take a <fault>'s name as the fault reference-->
				</action>
				<action type="INJECTION"> <!--now, we inject another fault that will be responsible for removing the "injected_fault" as our action.-->
						<fault_ref name="injected_fault_remove"></fault_ref> <!--"injected_fault_remove" will remove "injected_fault"-->
				</action>
			</actions>
		</fault>
		<fault name="injected_fault_remove"> <!-- here, we define the previously referenced (<fault_ref>) fault "injected_fault_remove"-->
			<triggers>
				<trigger type="META_COUNTER">
					<count>9</count>
					<trigger type="ASAP"> <!-- the "as soon as possible" <trigger> type ASAP does exactly this, it fires immediately the next time its <injector> checks all its <trigger>s-->
						<injector>core%i%</injector>
					</trigger>
				</trigger>
			</triggers>
			<actions>
					<action type="EJECTION"> <!-- the remover <action> is EJECTION, which is the antonym to the INJECTION type. Once applied all the referenced <fault>'s <trigger>s are deactivated-->
						<fault_ref name="injected_fault"></fault_ref> <!-- the fault "injected_fault" is virtually removed (because its <trigger>s are deactivated)-->
					</action>
			</actions>
		</fault>
		<fault name="injected_fault"> <!-- "injected_fault_remove" already referenced "injected_fault", although the latter was not yet defined in the XML, this is allowed.-->
			<triggers>
				<trigger type="ASAP"> <!--ASAP can also be used to model permanent faults, because ASAP holds true until the fault is EJECTED. In our case, it is a semi-permanent fault as "injected_fault_remove" ejects "injected_fault" after 10 callbacks to core%i%'s injector'-->
					<injector>core%i%</injector>
				</trigger>
			</triggers>
			<actions>
				<action type="MASK"> <!--MASK action is similar to BITFLIP as it takes a <field> as target, however, the manipulation is equal to "<field> = <field> <op> <value>" in C primitive.-->
					<injector>core%i%</injector>
					<field>X2</field>
					<op>OR</op> <!-- the MASK operation can be either of {AND, OR, XOR, NAND, NOR}, e.g. use AND to reset a <field>'s bits, OR to set, XOR to flip, .. etc.-->
					<value>0x1</value> <!--specify the mask <value> as hexadecimal-->
				</action>
			</actions>
		</fault>
	</definitions>
	<initial> <!--these are the initial <fault>s that are added to the simulation on start. "meta" is the only one here, but it adds additional faults during the simulation.-->
		<fault_ref name="meta"></fault_ref> <!-- like the INJECTION <action>, <initial> only takes <fault_ref>-->
	</initial>
</faults>