Common Design Flaws
Deadlock Livelock, starvation
Underspecification Unexpected reception of messages
Overspecification Dead code
Violations of Constraints Buffer overruns Array bounds violations
3
In designing distributed systems:network applications,
data communication protocols,multithreaded code,
client-server applications.
Designing concurrent (software) systems is so hard that these flaws are
mostly overlookedFortunately, most of these
design errors can be detected using model checking techniques
What is model checking?
[Clarke & Emerson 1981]:“Model checking is an automated technique that, given a finite-state model of a system and a logical property, systematically checks whether this property holds for (a given initial state in) that model.”
Model checking tools automatically verify whetherholds, where M is a (finite-state) model of a system and property F is stated in some formal notation.
Problem: state space explosion! SPIN [Holzmann 1991] is one of the most powerful model
checkers. (based on [Vardi & Wolper 1986]
4
M j= ©M
©
Clarke, Emerson, and SifakisReceive
2007 ACM Turing Award
“Modern” Model Checking
Abstraction is the key activity in both approaches
Here, we deal with pure SPIN, i.e. the classic model checking approach
6
Verification vs Debugging
Two (extreme) approaches with respect to the application of model checkers. verification approach: tries to assure the correctness of a
detailed model M of the system under validation. debugging approach: tries to find errors in a model M.
Model checking is most effective in combination with the debugging approach.
Automatic verification is not about proving correctness, but about finding bugs much earlier in the development of a system. (there are exceptions: BIP approach [Henzinger and Sifakis])
7
Spin and Promela
SPIN = Simple Promela Interpreter
Promela = Process Meta Language
The modeling language of SPIN So it is not a language to build an application!
Strong features : Powerful constructs to synchronize concurrent processes Cutting edge model checking technology Simulation to support analysis (of the models)
9
System, process, and action.
A system in SPIN consists of a set of interacting and concurrent processes.
Each process is sequential, but possibly non-deterministic.
Each process is built from atomic actions (transition).
Concurrent execution is modeled by interleaving.
Fairness can be impossed.
Promela model
Promela model consists of: Type declarations Channel declarations Variable declarations Process declarations [init process]
A Promela model corresponds with a (usually very large, but) finite transition system, so No unbounded data No unbounded channels No unbounded processes No unbounded process creation
Processes
A process type (proctype) consist of A name A list of formal parameters Local variable declarations body
12
Processes
A process Is defined by a proctype definition Executes concurrently with all other processes,
independent of speed behavior Communicate with other processes
Using global (shared) variables Using channels
There may be several processes of the same type Each process has its own local state:
Process counter (process identifier) Contents of the local variables
13
Processes
Processes are created using the run statement
Processes can be created at any point in the execution
Processes start executing after the run statement
Processes can also be created by adding active in front of the proctype declaration
14
Statements
The body of a process consists of a sequence of statements. A statement is either: executable: the statement can be executed immediately blocked: the statement cannot be executed
An assignment is always executable
An expression is also a statement; it is executable if it evaluates to non-zero.
15
Statements
The skip statement is always executable. “does nothing”, only changes process’ process counter
• A run statement is only executable if a new process can be created (remember: the number of processes is bounded). A printf statement is always executable
16
Statements
Assert expression; The assert-statement is always executable. If <expr> evaluates to zero, SPIN will exit with an error,
as the <expr> “has been violated”. The assert-statement is often used within Promela
models, to check whether certain properties are valid in a state.
17
(Enabledness) Expression
Example :
This process has 3 atomic actions. The action “y==0”
only enabled in a state where the expression is true it can only be executed when it is enabled; the effect is skip so, as long as it is disabled, the process will block if it is not enabled in the current state, due to the interleaving
semantics it may become enabled in the next state (by a transition caused by another process)
even if it is enabled in the current state, there is no guarantee the action will be selected for execution; but there is a way in SPIN to impose fairness.
active proctype P { x++ ; (y==0) ; x-- }
Example
Use it to synchronize between processes :
// both will terminate, but forcing Q to finish last
byte x=0 , y=0
active proctype P { x++ ; (y>0) ; x-- }
active proctype Q { (x>0) ; y++ ; (x==0) ; y-- }
Multiprogramming is tricky….
E.g. one or more processes can become stuck (deadlocked) :
(hence the need for verification!)
byte x=0 , y=0 active proctype P { x++ ; (y>0) ; (y==0) }
active proctype Q { y++ ; (x>0) ; (x==0) ; y-- }
Non-determinism
Non-determinism can be used to abstractly model alternate behavior:
active proctype client1() { do :: r ! REQ1 // spamming requests
:: g1 ? GRANTED ; ... ; rstatus = 0
:: g1 ? GRANTED ; rstatus= err // sometimes error
:: break // sometimes customer is impatient od ...
Scope
There are only 2 levels of scope: global var (visible in the entire sys) local var (visible only to the process that contains the
declaration)
Channels
for exchanging messages between processes finite sized and asynchronously, unless you set it to size 0
synchronous channel Syntax :
c ! 0 sending over channel c; blocking if c is full c ? x receives from c, transfer it to x; blocking if c is empty
There are some more exotic channel operations : checking empty/full, testing head-value, copying instead of receiving, sorted send, random receive ... check out the Manual
chan c = [0] of {bit};chan d = [2] of {mtype, bit, byte};chan e[2] = [1] of {mtype, record};
Assertion
Thanks to built-in non-determinism in the interleaving semantics, we can also use assertion to specify a global invariant !
// implying that at any time during the run x is either 0 or 1
byte x=0 , y=0
active proctype P { x++ ; (y>0) ; x-- }
active proctype Q { (x>0) ; y++ ; (x==0) ; y--}
active proctype Monitor { assert ((x==0 || x==1)) }
End state
When a process P fails to reach its terminal (end) state: Then it was locked error. Maybe this P is not supposed to reach end-state
suppress end-state checking with SPIN option.
The terminal state is by default just the textual end of of P’s code.
You can specify additional terminal states by using end-label: Of the form “end*”
25
State
Each (global) state of a system is a “product” of the states of its processes.
E.g. Suppose we have: One global var byte x Process P with byte y Process Q with byte z
Each system state should describe: all these variables Program counter of each process Other SPIN predefined vars
Represent each global state as a tuple … this tuple can be quite big.
26
Watch out for state explosion!
Size of each state: > 96 bits Number of possible states (232) 3 = 296
Far too huge Focus on the critical aspect of your model; abstract from
data when possible.
27
int x,y,z ;
P { do :: x++ od }Q { do :: y++ od }R { do :: x/=y z++ od }
(X)SPIN Architecture
LTL Translater
Spin
Simulator
VerifierGenerator
spinrandomguided
interactive
Xspin
ϕ
•deadlocks•safety properties•liveness properties
Promelamodel M
editing windowsimulation optionsverification optionsMSC simulation window
C program
checker
pan.*
pan.execounterexample
false
The stack
To save space SPIN does not keep a stack of states (large!), but rather a stack of action-IDs (small!)
Though this requires computing action-undo when backtracking
29
Verifier’s output
assertion violated !((crit[0]&&crit[1])) (at depth 5) // computation depth...Warning: Search not completed
Full statespace search for: ...
never-claim - (not selected)assertion violations +invalid endstates +
State-vector 20 byte, depth reached 7, errors: 1 // max. stack depth 24 states, stored // states stored in hash table 17 states, matched // states found re-revisited 41 transitions (= stored+matched) 0 atomic stepshash conflicts: 0 (resolved)(max size 2^19 states)
2.542 memory usage (Mbyte)
Specifying LTL properties
In Xspin via the LTL manager; available operators ; somewhat silly interface
SPIN then generates the Buchi automaton for this LTL formula; called “Never Claim” in SPIN.
[](ok1 && !ok2)
#define ok1 crit[1]#define ok2 crit[2]
Example of a Never Claim
Here is the never-claim of []<>p (the Buchi of []<>p = <>[]p)
This is automatically generated by SPIN
never { do :: p ; break :: skip od ; accept : do :: !p ; skip od }
Error if accept is reachable in the lock-step execution, and from there a cyclic run can be found.