//Let's join Kaede Takagaki Fan Club !!
#pragma GCC optimize("Ofast")
#pragma GCC optimize("unroll-loops")
#include <cstdio>
#include <cstring>
#include <cstdlib>
#include <cmath>
#include <ctime>
#include <cassert>
#include <string>
#include <algorithm>
#include <vector>
#include <queue>
#include <stack>
#include <functional>
#include <iostream>
#include <map>
#include <set>
#include <unordered_map>
#include <unordered_set>
#include <cassert>
#include <iomanip>
#include <chrono>
#include <random>
#include <bitset>
#include <complex>
#include <ext/pb_ds/assoc_container.hpp>
#include <ext/pb_ds/tree_policy.hpp>
using namespace std;
#define int long long
//#define L __int128
typedef long long ll;
typedef pair<int,int> P;
typedef pair<int,P> P1;
typedef pair<P,P> P2;
#define pu push
#define pb push_back
#define eb emplace_back
#define mp make_pair
#define eps 1e-7
#define INF 1000000000
#define a first
#define b second
#define fi first
#define sc second
//#define rng(i,a,b) for(int i=(int)(a);i<(int)(b);i++)
#define rep(i,x) for(int i=0;i<x;i++)
#define repn(i,x) for(int i=1;i<=x;i++)
#define SORT(x) sort(x.begin(),x.end())
#define ERASE(x) x.erase(unique(x.begin(),x.end()),x.end())
#define POSL(x,v) (lower_bound(x.begin(),x.end(),v)-x.begin())
#define POSU(x,v) (upper_bound(x.begin(),x.end(),v)-x.begin())
#define all(x) x.begin(),x.end()
#define si(x) int(x.size())
#define pcnt(x) __builtin_popcountll(x)

#ifdef LOCAL
#define dmp(x) cerr<<__LINE__<<" "<<#x<<" "<<x<<endl
#else
#define dmp(x) void(0)
#endif
 
template<class t,class u> bool chmax(t&a,u b){if(a<b){a=b;return true;}else return false;}
template<class t,class u> bool chmin(t&a,u b){if(b<a){a=b;return true;}else return false;}
 
template<class t> using vc=vector<t>;
 
template<class t,class u>
ostream& operator<<(ostream& os,const pair<t,u>& p){
	return os<<"{"<<p.fi<<","<<p.sc<<"}";
}
 
template<class t> ostream& operator<<(ostream& os,const vc<t>& v){
	os<<"{";
	for(auto e:v)os<<e<<",";
	return os<<"}";
}
 
 //https://codeforces.com/blog/entry/62393
struct custom_hash {
	static uint64_t splitmix64(uint64_t x) {
		// http://xorshift.di.unimi.it/splitmix64.c
		x += 0x9e3779b97f4a7c15;
		x = (x ^ (x >> 30)) * 0xbf58476d1ce4e5b9;
		x = (x ^ (x >> 27)) * 0x94d049bb133111eb;
		return x ^ (x >> 31);
	}
 
	size_t operator()(uint64_t x) const {
		static const uint64_t FIXED_RANDOM = chrono::steady_clock::now().time_since_epoch().count();
		return splitmix64(x + FIXED_RANDOM);
	}
	//don't make x negative!
	size_t operator()(pair<int,int> x)const{
		return operator()(uint64_t(x.first)<<32|x.second);
	}
};
//unordered_set -> dtype, null_type
//unordered_map -> dtype(key), dtype(value)
using namespace __gnu_pbds;
template<class t,class u>
using hash_table=gp_hash_table<t,u,custom_hash>;

template<class T>
void g(T &a){
	cin >> a;
}
template<class T>
void o(const T &a,bool space=false){
	cout << a << (space?' ':'\n');
}
//ios::sync_with_stdio(false);
const ll mod = 1000000007;
mt19937_64 mt(chrono::steady_clock::now().time_since_epoch().count());
template<class T>
void add(T&a,T b){
	a+=b;
	if(a >= mod) a-=mod;
}
ll modpow(ll x,ll n){
	ll res=1;
	while(n>0){
		if(n&1) res=res*x%mod;
		x=x*x%mod;
		n>>=1;
	}
	return res;
}
#define _sz 1005
ll F[_sz],R[_sz];
void make(){
	F[0] = 1;
	for(int i=1;i<_sz;i++) F[i] = F[i-1]*i%mod;
	R[_sz-1] = modpow(F[_sz-1], mod-2);
	for(int i=_sz-2;i>=0;i--) R[i] = R[i+1] * (i+1) % mod;
}
ll C(int a,int b){
	if(b < 0 || a < b) return 0;
	return F[a]*R[b]%mod*R[a-b]%mod;
}
//o(ans?"Yes":"No");
#define sz 1005
int par[sz],ran[sz];
void init(){ for(int i=0;i<sz;i++) par[i] = i, ran[i] = 0; }
int find(int x){ if(x == par[x]) return x; else return par[x] = find(par[x]); }
void unite(int x,int y){
	x = find(x); y = find(y); if(x==y) return;
	if(ran[x] < ran[y]) par[x] = y;
	else{
		par[y] = x;
		if(ran[x] == ran[y]) ran[x]++;
	}
}
bool same(int x,int y){ return find(x)==find(y); }
int n, m, a[1005], num[1005];

int dp[1005][1005];
int calc(int nn, int cd){
	rep(i,nn+3)rep(j,nn+3) dp[i][j] = 0;
	dp[nn+1][0] = 1;
	for(int i=nn;i>=1;i--){
		for(int j=0;j<=nn;j++){
			if(dp[i+1][j] == 0) continue;
			
			add(dp[i][j], dp[i+1][j]);
			if(__gcd(cd*i, m) != i) continue;
			int fc = 1, tan = modpow(cd, i-1);
			for(int x=1;x*i<=nn-j;x++){
				fc = fc * F[i-1] % mod * C(nn-j-(x-1)*i, i) % mod * tan % mod;
				add(dp[i][j+x*i], dp[i+1][j]*fc%mod*R[x]%mod);
			}
		}
	}
	return dp[1][nn];
}
void solve(){
	cin>>n>>m;
	make();
	init();
	repn(i,n){
		cin >> a[i];
		unite(i, a[i]);
	}
	vc<int>cnt(n+1, 0);
	repn(i,n) cnt[find(i)] ++;
	repn(i,n) if(cnt[i]){
		/*if(__gcd(m, cnt[i]) > 1){
			o(0);
			return;
		}*/
		num[cnt[i]] ++;
		//cout << cnt[i] << endl;
	}
	int ans = 1;
	repn(i,n){
		if(!num[i]) continue;
		ans = ans * calc(num[i], i) % mod;
		//cout << num[i] << " " << i << endl;
	}
	o(ans);
}
signed main(){
	cin.tie(0);
	ios::sync_with_stdio(0);
	cout<<fixed<<setprecision(20);
	int t; t = 1; //cin >> t;
	while(t--) solve();
}