#include <bits/stdc++.h>
#pragma GCC optimize("Ofast")
#pragma GCC optimize("unroll-loops")
using namespace std;
using std::cout;
using std::cin;
using std::endl;
using ll=long long;
using ld=long double;
ll ILL=2167167167167167167;
const int INF=2100000000;
const int mod=998244353;
#define rep(i,a,b) for (ll i=a;i<b;i++)
#define all(p) p.begin(),p.end()
template<class T> using _pq = priority_queue<T, vector<T>, greater<T>>;
template<class T> ll LB(vector<T> &v,T a){return lower_bound(v.begin(),v.end(),a)-v.begin();}
template<class T> ll UB(vector<T> &v,T a){return upper_bound(v.begin(),v.end(),a)-v.begin();}
template<class T> bool chmin(T &a,const T &b){if(a>b){a=b;return 1;}else return 0;}
template<class T> bool chmax(T &a,const T &b){if(a<b){a=b;return 1;}else return 0;}
template<class T> void So(vector<T> &v) {sort(v.begin(),v.end());}
template<class T> void Sore(vector<T> &v) {sort(v.begin(),v.end(),[](T x,T y){return x>y;});}
void yneos(bool a){if(a) cout<<"Yes\n"; else cout<<"No\n";}
template<class T> void vec_out(vector<T> &p){for(int i=0;i<(int)(p.size());i++){if(i) cout<<" ";cout<<p[i];}cout<<"\n";}
template<class T> T vec_min(vector<T> &a){assert(!a.empty());T ans=a[0];for(auto &x:a) chmin(ans,x);return ans;}
template<class T> T vec_max(vector<T> &a){assert(!a.empty());T ans=a[0];for(auto &x:a) chmax(ans,x);return ans;}
template<class T> T vec_sum(vector<T> &a){assert(!a.empty());T ans=a[0]-a[0];for(auto &x:a) ans+=x;return ans;}
int pop_count(long long a){int res=0;while(a){res+=(a&1),a>>=1;}return res;}

namespace po167{
struct combination{
	int upper;
	long long MOD;
	std::vector<long long> fact;
	std::vector<long long> rev;
	std::vector<long long> fact_rev;
	combination(int max,long long mod):upper(max),MOD(mod),fact(max+1),rev(max+1),fact_rev(max+1){
		for(long long i=0;i<=max;i++){
			if(i<2){
				fact[i]=1;
				fact_rev[i]=1;
				rev[i]=1;
				continue;
			}
			fact[i]=(fact[i-1]*i)%mod;
			rev[i]=mod-((mod/i)*rev[mod%i])%mod;
			fact_rev[i]=(fact_rev[i-1]*rev[i])%mod;
		}
	}
	long long Comb(int x,int y){
		assert(upper>=x);
		if (x<y||y<0||x<0) return 0;
		return (((fact_rev[y]*fact_rev[x-y])%MOD)*fact[x])%MOD;
	}
	long long P(int x,int y){
		assert(upper>=x);
		if (x<y||y<0||x<0) return 0;
		return (fact_rev[x-y]*fact[x])%MOD;
	}
};
}
using po167::combination;

namespace po167{
template <class T,T (*op)(T,T),T(*e)()>
struct segment_tree{
	int _n,size;
	std::vector<T> seg;
	int ceil_pow2(int a){
		int b=1;
		while(a>b){
			b<<=1;
		}
		return b;
	}
	void update(int k){seg[k]=op(seg[k*2],seg[k*2+1]);};
	segment_tree(int n) :_n(n){
		size=ceil_pow2(n);
		seg=std::vector<T>(size*2,e());
	}
	segment_tree(std::vector<T> &p) :_n((int) p.size()){
		size=ceil_pow2(_n);
		seg=std::vector<T>(size*2,e());
		for(int i=0;i<_n;i++) seg[i+size]=p[i];
		for(int i=size-1;i>0;i--) update(i);
	}
	void set(int ind,T val){
		assert(0<=ind&&ind<_n);
		ind+=size;
		seg[ind]=val;
		while(ind!=1){
			ind>>=1;
			update(ind);
		}
	}
    void addl(int ind,T val){
        set(ind,op(get(ind),val));
    }
    void addr(int ind,T val){
        set(ind,op(val,get(ind)));
    }
	T get(int ind){
		assert(0<=ind&&ind<_n);
		return seg[ind+size];
	}
	T query(int l,int r){
		assert(0<=l&&l<=r&&r<=_n);
		T l_val=e();
		T r_val=e();
		l+=size,r+=size;
		while(l<r){
			if(l&1) l_val=op(l_val,seg[l]),l+=1;
			if(r&1) r-=1,r_val=op(seg[r],r_val);
			r>>=1;
			l>>=1;
		}
		return op(l_val,r_val);
	}
	template <bool (*f)(T)> int max_right(int l) {
        return max_right(l, [](T x) { return f(x); });
    }
	template <class F> int max_right(int l, F f) {
		assert(0<=l&&l<=_n);
		assert(f(e()));
		if(f(query(l,_n))) return _n;
		T val=e();
		l+=size;
		while(true){
			while(l%2==0) l>>=1;
			if(!f(op(val,seg[l]))){
				while(l<size){
					l*=2;
					if(f(op(val,seg[l]))){
						val=op(val,seg[l]);
						l++;
					}
				}
				return l-size;
			}
			val=op(val,seg[l]);
			l++;
		}
	}
	template <bool (*f)(T)> int min_left(int r) {
        return min_left(r, [](T x) { return f(x); });
    }
    template <class F> int min_left(int r, F f) {
        assert(0 <= r && r <= _n);
        assert(f(e()));
        if (r == 0) return 0;
        r += size;
        T sm = e();
        do {
            r--;
            while (r > 1 && (r % 2)) r >>= 1;
            if (!f(op(seg[r], sm))) {
                while (r < size) {
                    r = (2 * r + 1);
                    if (f(op(seg[r], sm))) {
                        sm = op(seg[r], sm);
                        r--;
                    }
                }
                return r + 1 - size;
            }
            sm = op(seg[r], sm);
        } while ((r & -r) != r);
        return 0;
    }

};
}
using po167::segment_tree;


using F= ll;
F op(F a,F b){return a+b;}
F e(){return 0;}
int target;
bool f(F x){
	return (x%target==0);
}




void solve();
// oddloop
int main() {
	ios::sync_with_stdio(false);
	cin.tie(nullptr);
	
	int t=1;
	//cin>>t;
	rep(i,0,t) solve();
}

void solve(){
	ll N,M;
    cin>>N>>M;
    combination table(N,mod);
    vector<ll> p(N+1,1),q(N,-1);
    rep(i,0,M){
        ll a,b;
        cin>>a>>b;
        a--,b--;
        q[b]=a;
        p[a]=0;
    }
    rep(i,0,N) p[i+1]+=p[i];
    ll ans=((N-M)*(N-M-1))/2;
    ans%=mod;
    ans=(ans*table.fact_rev[2])%mod;
    ll X=0,Y=N-M;
    segment_tree<F,op,e> seg(N);
    rep(i,0,N){
        if(q[i]==-1){
            X++,Y--;
            continue;
        }
        ll B=p[q[i]],A=N-M-B;
        ll tmp=(A*X)%mod;
        tmp=(tmp+(B*Y))%mod;
        tmp=(tmp*table.rev[N-M])%mod;
        ans=(ans+tmp)%mod;
        ans=(ans+seg.query(q[i],N))%mod;
        seg.set(q[i],1);
    }
    ans=(ans*table.fact[N-M])%mod;
    cout<<(ans+mod)%mod<<"\n";
}