#include <bits/stdc++.h>
#pragma GCC optimize("Ofast")
#pragma GCC optimize("unroll-loops")
#pragma GCC target("avx2")
using namespace std;
const int N = 1 << 18;
const int LGN = 18;
const int R = 25001;
const int B = 500;
int p[N], r[N];
int tl[N], tr[N], preord[N], cnt[R];
int prec[N / B][R], id[R], ids = 0;
vector<int> G[N];
vector<int> emp[N];
vector<int> pos[R];
int n, rn, q;
int ct = 0;
void dfs(int v){
tl[v] = ++ct; preord[ct] = v;
for(auto u : G[v]){
dfs(u);
}
tr[v] = ct;
}
bool vis[N];
void dfs_prec(int Co, int v, int acc = 0){
if(r[v] == Co) vis[v] = true;
for(auto u : G[v]){
dfs_prec(Co, u, acc + (r[v] == Co));
}
prec[id[Co]][r[v]] += acc;
}
int main() {
ios_base::sync_with_stdio(false);
cin >> n >> rn >> q;
cin >> r[1];
for(int i = 2; i <= n; i++){
cin >> p[i] >> r[i];
G[p[i]].push_back(i);
cnt[r[i]]++;
}
for(int i = 1; i <= n; i++){
emp[r[i]].push_back(i);
}
dfs(1);
for(int i = 1; i <= rn; i++){
if(cnt[i] >= B){
id[i] = ids;
ids++;
}
}
for(int i = 1; i <= n; i++){
if(cnt[r[i]] >= B && !vis[i]){
dfs_prec(r[i], i);
}
}
for(int i = 1; i <= n; i++){
pos[r[preord[i]]].push_back(i);
}
for(int i = 0; i < q; i++){
int r1, r2; cin >> r1 >> r2;
if(cnt[r1] >= B){
cout << prec[id[r1]][r2] << endl;
}else{
int ans = 0;
for(auto i : emp[r1]){
ans += upper_bound(pos[r2].begin(), pos[r2].end(), tr[i]) - lower_bound(pos[r2].begin(), pos[r2].end(), tl[i]);
}
cout << ans << endl;
}
}
return 0;
}
