sanitizes bloom filters to avoid division by zero (#13714)

Pull requests received over the wire can cause a validator to panic
because of division by zero in bloom filters:
https://github.com/solana-labs/solana/blob/af08ba93e/runtime/src/bloom.rs#L86-L88
This commit is contained in:
behzad nouri 2020-11-19 23:35:22 +00:00 committed by GitHub
parent 856693ac1f
commit a8c29505f0
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 18 additions and 10 deletions

View File

@ -79,7 +79,7 @@ impl CrdsFilter {
let seed: u64 = seed.checked_shl(64 - mask_bits).unwrap_or(0x0); let seed: u64 = seed.checked_shl(64 - mask_bits).unwrap_or(0x0);
seed | (!0u64).checked_shr(mask_bits).unwrap_or(!0x0) as u64 seed | (!0u64).checked_shr(mask_bits).unwrap_or(!0x0) as u64
} }
pub fn max_items(max_bits: f64, false_rate: f64, num_keys: f64) -> f64 { fn max_items(max_bits: f64, false_rate: f64, num_keys: f64) -> f64 {
let m = max_bits; let m = max_bits;
let p = false_rate; let p = false_rate;
let k = num_keys; let k = num_keys;
@ -93,28 +93,26 @@ impl CrdsFilter {
let buf = item.as_ref()[..8].try_into().unwrap(); let buf = item.as_ref()[..8].try_into().unwrap();
u64::from_le_bytes(buf) u64::from_le_bytes(buf)
} }
pub fn test_mask_u64(&self, item: u64, ones: u64) -> bool { fn test_mask(&self, item: &Hash) -> bool {
let bits = item | ones;
bits == self.mask
}
pub fn test_mask(&self, item: &Hash) -> bool {
// only consider the highest mask_bits bits from the hash and set the rest to 1. // only consider the highest mask_bits bits from the hash and set the rest to 1.
let ones = (!0u64).checked_shr(self.mask_bits).unwrap_or(!0u64); let ones = (!0u64).checked_shr(self.mask_bits).unwrap_or(!0u64);
let bits = Self::hash_as_u64(item) | ones; let bits = Self::hash_as_u64(item) | ones;
bits == self.mask bits == self.mask
} }
pub fn add(&mut self, item: &Hash) { #[cfg(test)]
fn add(&mut self, item: &Hash) {
if self.test_mask(item) { if self.test_mask(item) {
self.filter.add(item); self.filter.add(item);
} }
} }
pub fn contains(&self, item: &Hash) -> bool { #[cfg(test)]
fn contains(&self, item: &Hash) -> bool {
if !self.test_mask(item) { if !self.test_mask(item) {
return true; return true;
} }
self.filter.contains(item) self.filter.contains(item)
} }
pub fn filter_contains(&self, item: &Hash) -> bool { fn filter_contains(&self, item: &Hash) -> bool {
self.filter.contains(item) self.filter.contains(item)
} }
} }

View File

@ -3,6 +3,7 @@ use bv::BitVec;
use fnv::FnvHasher; use fnv::FnvHasher;
use rand::{self, Rng}; use rand::{self, Rng};
use serde::{Deserialize, Serialize}; use serde::{Deserialize, Serialize};
use solana_sdk::sanitize::{Sanitize, SanitizeError};
use std::fmt; use std::fmt;
use std::sync::atomic::{AtomicU64, Ordering}; use std::sync::atomic::{AtomicU64, Ordering};
use std::{cmp, hash::Hasher, marker::PhantomData}; use std::{cmp, hash::Hasher, marker::PhantomData};
@ -45,7 +46,16 @@ impl<T: BloomHashIndex> fmt::Debug for Bloom<T> {
} }
} }
impl<T: BloomHashIndex> solana_sdk::sanitize::Sanitize for Bloom<T> {} impl<T: BloomHashIndex> Sanitize for Bloom<T> {
fn sanitize(&self) -> Result<(), SanitizeError> {
// Avoid division by zero in self.pos(...).
if self.bits.is_empty() {
Err(SanitizeError::InvalidValue)
} else {
Ok(())
}
}
}
impl<T: BloomHashIndex> Bloom<T> { impl<T: BloomHashIndex> Bloom<T> {
pub fn new(num_bits: usize, keys: Vec<u64>) -> Self { pub fn new(num_bits: usize, keys: Vec<u64>) -> Self {