#include <string.h>

#include <algorithm>
#include <array>
#include <bitset>
#include <cassert>
#include <chrono>
#include <ciso646>
#include <climits>
#include <cmath>
#include <complex>
#include <cstdio>
#include <functional>
#include <iomanip>
#include <iostream>
#include <map>
#include <numeric>
#include <optional>
#include <queue>
#include <random>
#include <set>
#include <stack>
#include <string>
#include <unordered_map>
#include <unordered_set>
#include <utility>
#include <vector>

#define REP_OVERLOAD(arg1, arg2, arg3, arg4, NAME, ...) NAME
#define REP3(i, l, r, s)                                               \
    for (int i = int(l), rep3_r = int(r), rep3_s = int(s); i < rep3_r; \
         i += rep3_s)
#define REP2(i, l, r) REP3(i, l, r, 1)
#define REP1(i, n) REP2(i, 0, n)
#define rep(...) REP_OVERLOAD(__VA_ARGS__, REP3, REP2, REP1, )(__VA_ARGS__)
#define repin(i, l, r) for (int i = int(l), repin_r = int(r); i <= repin_r; ++i)

#define RREP_OVERLOAD(arg1, arg2, arg3, arg4, NAME, ...) NAME
#define RREP3(i, l, r, s)                                                      \
    for (int i = int(r) - 1, rrep3_l = int(l), rrep3_s = int(s); i >= rrep3_l; \
         i -= rrep3_s)
#define RREP2(i, l, r) RREP3(i, l, r, 1)
#define RREP1(i, n) RREP2(i, 0, n)
#define rrep(...) RREP_OVERLOAD(__VA_ARGS__, RREP3, RREP2, RREP1, )(__VA_ARGS__)
#define rrepin(i, l, r) \
    for (int i = int(r), rrepin_l = int(l); i >= rrepin_l; --i)

#define fi first
#define se second

#include <atcoder/fenwicktree>
#include <atcoder/modint>

#include <atcoder/modint>
#include <vector>

namespace rklib {

template <class T>
struct Factorial {
   public:
    Factorial() : Factorial(0) {}
    Factorial(int n) {
        fc.resize(n + 1);
        inv_fc.resize(n + 1);
        fc[0] = 1;
        for (int i = 0; i < n; ++i) fc[i + 1] = fc[i] * (i + 1);

        inv_fc[n] = 1 / fc[n];
        for (int i = n - 1; i >= 0; --i) inv_fc[i] = inv_fc[i + 1] * (i + 1);
    }

    T fact(int n) {
        if (n >= (int)fc.size()) extend(n);
        return fc[n];
    }

    T inv_fact(int n) {
        if (n >= (int)fc.size()) extend(n);
        return inv_fc[n];
    }

    T inv(int n) {
        assert(n > 0);
        if (n >= (int)fc.size()) extend(n);
        return inv_fc[n] * fc[n - 1];
    }

    T comb(int n, int r) {
        if (n < r || r < 0) return 0;
        if (n >= (int)fc.size()) extend(n);
        return fc[n] * inv_fc[r] * inv_fc[n - r];
    }

    T perm(int n, int r) {
        if (n < r || r < 0) return 0;
        if (n >= (int)fc.size()) extend(n);
        return fc[n] * inv_fc[n - r];
    }

   private:
    std::vector<T> fc;
    std::vector<T> inv_fc;

    void extend(int n) {
        int l = fc.size();
        int r = l;
        while (r <= n) r *= 2;

        fc.resize(r);
        inv_fc.resize(r);

        for (int i = l; i < r; ++i) fc[i] = fc[i - 1] * i;

        inv_fc[r - 1] = 1 / fc[r - 1];
        for (int i = r - 2; i >= l; --i) inv_fc[i] = inv_fc[i + 1] * (i + 1);
    }
};

}  // namespace rklib


#include <algorithm>
#include <cassert>
#include <vector>

namespace rklib {

template <class T>
bool chmax(T &a, const T &b) {
    if (a < b) {
        a = b;
        return true;
    }
    return false;
}

template <class T>
bool chmin(T &a, const T &b) {
    if (a > b) {
        a = b;
        return true;
    }
    return false;
}

template <class T>
bool chmin_non_negative(T &a, const T &b) {
    if (a < 0 || a > b) {
        a = b;
        return true;
    }
    return false;
}

template <class T>
T div_floor(T num, T den) {
    if (den < 0) num = -num, den = -den;
    return num >= 0 ? num / den : (num + 1) / den - 1;
}

template <class T>
T div_ceil(T num, T den) {
    if (den < 0) num = -num, den = -den;
    return num <= 0 ? num / den : (num - 1) / den + 1;
}

}  // namespace rklib


using namespace std;
using namespace rklib;

using lint = long long;
using pii = pair<int, int>;
using pll = pair<lint, lint>;

using mint = atcoder::modint998244353;

int main() {
    int n, m;
    scanf("%d%d", &n, &m);
    int a[n];
    fill(a, a + n, -1);
    bool used[n];
    memset(used, false, sizeof(used));
    rep(i, m) {
        int p, k;
        scanf("%d%d", &p, &k);
        --p;
        --k;
        a[k] = p;
        used[p] = true;
    }

    mint ans = 0;
    Factorial<mint> fc(n);
    if (n - m >= 2) {
        ans += fc.comb(n - m, 2) * fc.comb(n - m, 2) * fc.fact(n - m - 2);
    }

    {
        atcoder::fenwick_tree<int> ft(n);
        rep(i, n) {
            if (a[i] == -1) continue;
            ans += ft.sum(a[i] + 1, n) * fc.fact(n - m);
            ft.add(a[i], 1);
        }
    }

    vector<int> yet;
    rep(i, n) if (!used[i]) yet.push_back(i);
    int vacant = 0;
    rep(i, n) {
        if (a[i] == -1) {
            ++vacant;
            continue;
        }
        mint tmp = 0;
        tmp += mint(vacant) *
               (yet.end() - upper_bound(yet.begin(), yet.end(), a[i]));
        tmp += mint(n - m - vacant) *
               (lower_bound(yet.begin(), yet.end(), a[i]) - yet.begin());
        ans += tmp * fc.fact(n - m - 1);
    }
    printf("%u\n", ans.val());
}