#ifdef ONLINE_JUDGE
#pragma GCC target("avx2,avx")
#pragma GCC optimize("O3")
#pragma GCC optimize("unroll-loops")
#endif
#include <bits/stdc++.h>
using namespace std;
using ll = long long;
using ull = unsigned long long;
using i128 = __int128_t;
using pii = pair<int, int>;
using pll = pair<long long, long long>;
template<class T> using vec = vector<T>;
template<class T> using vvec = vector<vector<T>>;
#define rep(i, n) for (int i = 0; i < (n); i++)
#define rrep(i, n) for (int i = int(n) - 1; i >= 0; i--)
#define all(x) (x).begin(), (x).end()
constexpr char ln = '\n';
istream& operator>>(istream& is, __int128_t& x) {
    x = 0;
    string s;
    is >> s;
    int n = int(s.size()), it = 0;
    if (s[0] == '-') it++;
    for (; it < n; it++) x = (x * 10 + s[it] - '0');
    if (s[0] == '-') x = -x;
    return is;
}
ostream& operator<<(ostream& os, __int128_t x) {
    if (x == 0) return os << 0;
    if (x < 0) os << '-', x = -x;
    deque<int> deq;
    while (x) deq.emplace_front(x % 10), x /= 10;
    for (int e : deq) os << e;
    return os;
}
template<class T1, class T2>
ostream& operator<<(ostream& os, const pair<T1, T2>& p) {
    return os << "(" << p.first << ", " << p.second << ")";
}
template<class T> 
ostream& operator<<(ostream& os, const vector<T>& v) {
    os << "{";
    for (int i = 0; i < int(v.size()); i++) {
        if (i) os << ", ";
        os << v[i];
    }
    return os << "}";
}
template<class Container> inline int SZ(Container& v) { return int(v.size()); }
template<class T> inline void UNIQUE(vector<T>& v) { v.erase(unique(v.begin(), v.end()), v.end()); }
template<class T1, class T2> inline bool chmax(T1& a, T2 b) { if (a < b) {a = b; return true ;} return false ;}
template<class T1, class T2> inline bool chmin(T1& a, T2 b) { if (a > b) {a = b; return true ;} return false ;}
inline int topbit(ull x) { return x == 0 ? -1 : 63 - __builtin_clzll(x); }
inline int botbit(ull x) { return x == 0 ? 64 : __builtin_ctzll(x); }
inline int popcount(ull x) { return __builtin_popcountll(x); }
inline int kthbit(ull x, int k) { return (x>>k) & 1; }
inline constexpr long long TEN(int x) { return x == 0 ? 1 : TEN(x-1) * 10; }
inline void print() { cout << "\n"; }
template<class T>
inline void print(const vector<T>& v) {
    for (int i = 0; i < int(v.size()); i++) {
        if (i) cout << " ";
        cout << v[i];
    }
    print();
}
template<class T, class... Args>
inline void print(const T& x, const Args& ... args) {
    cout << x << " ";
    print(args...);
}
#ifdef MINATO_LOCAL
inline void debug_out() { cerr << endl; }
template <class T, class... Args>
inline void debug_out(const T& x, const Args& ... args) {
    cerr << " " << x;
    debug_out(args...);
}
#define debug(...) cerr << __LINE__ << " : [" << #__VA_ARGS__ << "] =", debug_out(__VA_ARGS__)
#define dump(x) cerr << __LINE__ << " : " << #x << " = " << (x) << endl
#else
#define debug(...) (void(0))
#define dump(x) (void(0))
#endif
struct fast_ios { fast_ios() { cin.tie(nullptr); ios::sync_with_stdio(false); cout << fixed << setprecision(20); }; } fast_ios_;
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////


template<typename T>
struct Compress {
    vector<T> xs;

    Compress(){};
    Compress(const vector<T> &vs) {add(vs);}
    Compress(const initializer_list<vector<T>> &vs) {for(auto &p : vs) add(p);}

    void add(const vector<T> &vs) {copy(begin(vs), end(vs), back_inserter(xs));}
    void add(const T &x) {xs.emplace_back(x);}

    int build() {
        sort(begin(xs), end(xs));
        xs.erase(unique(begin(xs), end(xs)), end(xs));
        return (int)xs.size();
    }

    vector<int> get(const vector<T> &vs) const {
        vector<int> ret;
        transform(begin(vs), end(vs), back_inserter(ret), [&](const T &x) {
            return lower_bound(begin(xs), end(xs), x) - begin(xs);
        });
        return ret;
    }

    int get(const T &x) const {return lower_bound(begin(xs), end(xs), x) - begin(xs);}

    const T &operator[](int k) const {return xs[k];}
};

template<int m> 
struct ModInt {
  public:
    static constexpr int mod() { return m; }
    static ModInt raw(int v) {
        ModInt x;
        x._v = v;
        return x;
    }

    ModInt() : _v(0) {}
    ModInt(long long v) {
        long long x = (long long)(v % (long long)(umod()));
        if (x < 0) x += umod();
        _v = (unsigned int)(x);
    }

    unsigned int val() const { return _v; }

    ModInt& operator++() {
        _v++;
        if (_v == umod()) _v = 0;
        return *this;
    }
    ModInt& operator--() {
        if (_v == 0) _v = umod();
        _v--;
        return *this;
    }
    ModInt operator++(int) {
        ModInt result = *this;
        ++*this;
        return result;
    }
    ModInt operator--(int) {
        ModInt result = *this;
        --*this;
        return result;
    }

    ModInt& operator+=(const ModInt& rhs) {
        _v += rhs._v;
        if (_v >= umod()) _v -= umod();
        return *this;
    }
    ModInt& operator-=(const ModInt& rhs) {
        _v -= rhs._v;
        if (_v >= umod()) _v += umod();
        return *this;
    }
    ModInt& operator*=(const ModInt& rhs) {
        unsigned long long z = _v;
        z *= rhs._v;
        _v = (unsigned int)(z % umod());
        return *this;
    }
    ModInt& operator^=(long long n) {
        ModInt x = *this;
        *this = 1;
        if (n < 0) x = x.inv(), n = -n;
        while (n) {
            if (n & 1) *this *= x;
            x *= x;
            n >>= 1;
        }
        return *this;
    }
    ModInt& operator/=(const ModInt& rhs) { return *this = *this * rhs.inv(); }

    ModInt operator+() const { return *this; }
    ModInt operator-() const { return ModInt() - *this; }

    ModInt pow(long long n) const {
        ModInt r = *this;
        r ^= n;
        return r;
    }
    ModInt inv() const {
        int a = _v, b = umod(), y = 1, z = 0, t;
        for (; ; ) {
            t = a / b; a -= t * b;
            if (a == 0) {
                assert(b == 1 || b == -1);
                return ModInt(b * z);
            }
            y -= t * z;
            t = b / a; b -= t * a;
            if (b == 0) {
                assert(a == 1 || a == -1);
                return ModInt(a * y);
            }
            z -= t * y;
        }
    }

    friend ModInt operator+(const ModInt& lhs, const ModInt& rhs) { return ModInt(lhs) += rhs; }
    friend ModInt operator-(const ModInt& lhs, const ModInt& rhs) { return ModInt(lhs) -= rhs; }
    friend ModInt operator*(const ModInt& lhs, const ModInt& rhs) { return ModInt(lhs) *= rhs; }
    friend ModInt operator/(const ModInt& lhs, const ModInt& rhs) { return ModInt(lhs) /= rhs; }
    friend ModInt operator^(const ModInt& lhs, long long rhs) { return ModInt(lhs) ^= rhs; }
    friend bool operator==(const ModInt& lhs, const ModInt& rhs) { return lhs._v == rhs._v; }
    friend bool operator!=(const ModInt& lhs, const ModInt& rhs) { return lhs._v != rhs._v; }
    friend ModInt operator+(long long lhs, const ModInt& rhs) { return (ModInt(lhs) += rhs); }
    friend ModInt operator-(long long lhs, const ModInt& rhs) { return (ModInt(lhs) -= rhs); }
    friend ModInt operator*(long long lhs, const ModInt& rhs) { return (ModInt(lhs) *= rhs); }
    friend ostream &operator<<(ostream& os, const ModInt& rhs) { return os << rhs._v; }

  private:
    unsigned int _v;
    static constexpr unsigned int umod() { return m; }
};
 
//constexpr int MOD = 1000000007;
constexpr int MOD = 998244353;

using mint = ModInt<MOD>;


struct ModCombination {
    vector<mint> Fac,Facinv;

    ModCombination(int x) {
        Fac.resize(x+1);
        Facinv.resize(x+1);
        Fac[0] = 1;
        for (int i = 0; i < x; ++i) Fac[i+1] = Fac[i]*(i+1);
        Facinv[x] = Fac[0]/Fac[x];
        for (int i = x; i > 0; --i) Facinv[i-1] = Facinv[i]*i;
    }

    mint get(int n, int k) {
        if (k < 0 || k > n) return 0;
        return Fac[n]*Facinv[k]*Facinv[n-k];
    }
};

template<typename T>
struct BinaryIndexedTree {
    int N;
    vector<T> data;

    BinaryIndexedTree(){}
    BinaryIndexedTree(int N) : N(N), data(N+1,0) {}

    // [0,k) (0-indexed) a[0] + … + a[k-1]
    T sum(int k) {
        T ret = 0;
        for(; k > 0; k -= k & -k) ret += data[k];
        return ret;
    }

    // [l,r) (0-indexed) a[l] + …　+ a[r-1]
    T sum(int l, int r) {
        if (l >= r) return 0;
        T vl = sum(l);
        T vr = sum(r);
        return vr - vl;
    }
    // (0-indexed) a[k] += x;
    void add(int k, T x) {
        for(++k; k <= N; k += k & -k) data[k] += x;
    }

    // (0-indexed)
    int lowerbound(T x) {
        int k = 1;
        int ret = 0;
        while ((k<<1) <= N) k <<= 1;
        while (k) {
            if (ret + k <= N && data[ret+k] < x) {
                x -= data[ret+k];
                ret += k;
            }
            k >>= 1;
        }
        return ret;
    }

    // (0-indexed)
    int upperbound(T x) {return lowerbound(x+1);}
};

int main() {
    int N; cin >> N;
    ModCombination MC(N);
    vec<int> A_(N);
    rep(i,N) cin >> A_[i];
    Compress<int> Com(A_);
    int M = Com.build();
    auto A = Com.get(A_);

    mint T = 0;
    BinaryIndexedTree<mint> BIT(M);
    rep(i,N) {
        T += BIT.sum(A[i]+1,M);
        BIT.add(A[i],1);
    }

    mint S = 0;
    rep(i,N) S += BIT.sum(A[i]+1,M);

    mint ans = 0;
    mint ALL = 1;
    for (int i = 1; i <= N; i++) ALL *= MC.get(N,i);
    mint sm = 0;
    for (int i = 1; i <= N; i++) {
        ans += MC.get(N-2,i-2)*T*ALL/MC.get(N,i);
        ans += sm*MC.get(N-1,i-1)*S*ALL/MC.get(N,i);
        sm += mint(i)/N;
    }

    cout << ans << ln;    
}