github.com/aquanetwork/aquachain@v1.7.8/trie/sync_test.go (about)

     1  // Copyright 2015 The aquachain Authors
     2  // This file is part of the aquachain library.
     3  //
     4  // The aquachain library is free software: you can redistribute it and/or modify
     5  // it under the terms of the GNU Lesser General Public License as published by
     6  // the Free Software Foundation, either version 3 of the License, or
     7  // (at your option) any later version.
     8  //
     9  // The aquachain library is distributed in the hope that it will be useful,
    10  // but WITHOUT ANY WARRANTY; without even the implied warranty of
    11  // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
    12  // GNU Lesser General Public License for more details.
    13  //
    14  // You should have received a copy of the GNU Lesser General Public License
    15  // along with the aquachain library. If not, see <http://www.gnu.org/licenses/>.
    16  
    17  package trie
    18  
    19  import (
    20  	"bytes"
    21  	"testing"
    22  
    23  	"gitlab.com/aquachain/aquachain/aquadb"
    24  	"gitlab.com/aquachain/aquachain/common"
    25  )
    26  
    27  // makeTestTrie create a sample test trie to test node-wise reconstruction.
    28  func makeTestTrie() (*Database, *Trie, map[string][]byte) {
    29  	// Create an empty trie
    30  	diskdb := aquadb.NewMemDatabase()
    31  	triedb := NewDatabase(diskdb)
    32  	trie, _ := New(common.Hash{}, triedb)
    33  
    34  	// Fill it with some arbitrary data
    35  	content := make(map[string][]byte)
    36  	for i := byte(0); i < 255; i++ {
    37  		// Map the same data under multiple keys
    38  		key, val := common.LeftPadBytes([]byte{1, i}, 32), []byte{i}
    39  		content[string(key)] = val
    40  		trie.Update(key, val)
    41  
    42  		key, val = common.LeftPadBytes([]byte{2, i}, 32), []byte{i}
    43  		content[string(key)] = val
    44  		trie.Update(key, val)
    45  
    46  		// Add some other data to inflate the trie
    47  		for j := byte(3); j < 13; j++ {
    48  			key, val = common.LeftPadBytes([]byte{j, i}, 32), []byte{j, i}
    49  			content[string(key)] = val
    50  			trie.Update(key, val)
    51  		}
    52  	}
    53  	trie.Commit(nil)
    54  
    55  	// Return the generated trie
    56  	return triedb, trie, content
    57  }
    58  
    59  // checkTrieContents cross references a reconstructed trie with an expected data
    60  // content map.
    61  func checkTrieContents(t *testing.T, db *Database, root []byte, content map[string][]byte) {
    62  	// Check root availability and trie contents
    63  	trie, err := New(common.BytesToHash(root), db)
    64  	if err != nil {
    65  		t.Fatalf("failed to create trie at %x: %v", root, err)
    66  	}
    67  	if err := checkTrieConsistency(db, common.BytesToHash(root)); err != nil {
    68  		t.Fatalf("inconsistent trie at %x: %v", root, err)
    69  	}
    70  	for key, val := range content {
    71  		if have := trie.Get([]byte(key)); !bytes.Equal(have, val) {
    72  			t.Errorf("entry %x: content mismatch: have %x, want %x", key, have, val)
    73  		}
    74  	}
    75  }
    76  
    77  // checkTrieConsistency checks that all nodes in a trie are indeed present.
    78  func checkTrieConsistency(db *Database, root common.Hash) error {
    79  	// Create and iterate a trie rooted in a subnode
    80  	trie, err := New(root, db)
    81  	if err != nil {
    82  		return nil // Consider a non existent state consistent
    83  	}
    84  	it := trie.NodeIterator(nil)
    85  	for it.Next(true) {
    86  	}
    87  	return it.Error()
    88  }
    89  
    90  // Tests that an empty trie is not scheduled for syncing.
    91  func TestEmptyTrieSync(t *testing.T) {
    92  	diskdbA := aquadb.NewMemDatabase()
    93  	triedbA := NewDatabase(diskdbA)
    94  
    95  	diskdbB := aquadb.NewMemDatabase()
    96  	triedbB := NewDatabase(diskdbB)
    97  
    98  	emptyA, _ := New(common.Hash{}, triedbA)
    99  	emptyB, _ := New(emptyRoot, triedbB)
   100  
   101  	for i, trie := range []*Trie{emptyA, emptyB} {
   102  		diskdb := aquadb.NewMemDatabase()
   103  		if req := NewTrieSync(trie.Hash(), diskdb, nil).Missing(1); len(req) != 0 {
   104  			t.Errorf("test %d: content requested for empty trie: %v", i, req)
   105  		}
   106  	}
   107  }
   108  
   109  // Tests that given a root hash, a trie can sync iteratively on a single thread,
   110  // requesting retrieval tasks and returning all of them in one go.
   111  func TestIterativeTrieSyncIndividual(t *testing.T) { testIterativeTrieSync(t, 1) }
   112  func TestIterativeTrieSyncBatched(t *testing.T)    { testIterativeTrieSync(t, 100) }
   113  
   114  func testIterativeTrieSync(t *testing.T, batch int) {
   115  	// Create a random trie to copy
   116  	srcDb, srcTrie, srcData := makeTestTrie()
   117  
   118  	// Create a destination trie and sync with the scheduler
   119  	diskdb := aquadb.NewMemDatabase()
   120  	triedb := NewDatabase(diskdb)
   121  	sched := NewTrieSync(srcTrie.Hash(), diskdb, nil)
   122  
   123  	queue := append([]common.Hash{}, sched.Missing(batch)...)
   124  	for len(queue) > 0 {
   125  		results := make([]SyncResult, len(queue))
   126  		for i, hash := range queue {
   127  			data, err := srcDb.Node(hash)
   128  			if err != nil {
   129  				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
   130  			}
   131  			results[i] = SyncResult{hash, data}
   132  		}
   133  		if _, index, err := sched.Process(results); err != nil {
   134  			t.Fatalf("failed to process result #%d: %v", index, err)
   135  		}
   136  		if index, err := sched.Commit(diskdb); err != nil {
   137  			t.Fatalf("failed to commit data #%d: %v", index, err)
   138  		}
   139  		queue = append(queue[:0], sched.Missing(batch)...)
   140  	}
   141  	// Cross check that the two tries are in sync
   142  	checkTrieContents(t, triedb, srcTrie.Root(), srcData)
   143  }
   144  
   145  // Tests that the trie scheduler can correctly reconstruct the state even if only
   146  // partial results are returned, and the others sent only later.
   147  func TestIterativeDelayedTrieSync(t *testing.T) {
   148  	// Create a random trie to copy
   149  	srcDb, srcTrie, srcData := makeTestTrie()
   150  
   151  	// Create a destination trie and sync with the scheduler
   152  	diskdb := aquadb.NewMemDatabase()
   153  	triedb := NewDatabase(diskdb)
   154  	sched := NewTrieSync(srcTrie.Hash(), diskdb, nil)
   155  
   156  	queue := append([]common.Hash{}, sched.Missing(10000)...)
   157  	for len(queue) > 0 {
   158  		// Sync only half of the scheduled nodes
   159  		results := make([]SyncResult, len(queue)/2+1)
   160  		for i, hash := range queue[:len(results)] {
   161  			data, err := srcDb.Node(hash)
   162  			if err != nil {
   163  				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
   164  			}
   165  			results[i] = SyncResult{hash, data}
   166  		}
   167  		if _, index, err := sched.Process(results); err != nil {
   168  			t.Fatalf("failed to process result #%d: %v", index, err)
   169  		}
   170  		if index, err := sched.Commit(diskdb); err != nil {
   171  			t.Fatalf("failed to commit data #%d: %v", index, err)
   172  		}
   173  		queue = append(queue[len(results):], sched.Missing(10000)...)
   174  	}
   175  	// Cross check that the two tries are in sync
   176  	checkTrieContents(t, triedb, srcTrie.Root(), srcData)
   177  }
   178  
   179  // Tests that given a root hash, a trie can sync iteratively on a single thread,
   180  // requesting retrieval tasks and returning all of them in one go, however in a
   181  // random order.
   182  func TestIterativeRandomTrieSyncIndividual(t *testing.T) { testIterativeRandomTrieSync(t, 1) }
   183  func TestIterativeRandomTrieSyncBatched(t *testing.T)    { testIterativeRandomTrieSync(t, 100) }
   184  
   185  func testIterativeRandomTrieSync(t *testing.T, batch int) {
   186  	// Create a random trie to copy
   187  	srcDb, srcTrie, srcData := makeTestTrie()
   188  
   189  	// Create a destination trie and sync with the scheduler
   190  	diskdb := aquadb.NewMemDatabase()
   191  	triedb := NewDatabase(diskdb)
   192  	sched := NewTrieSync(srcTrie.Hash(), diskdb, nil)
   193  
   194  	queue := make(map[common.Hash]struct{})
   195  	for _, hash := range sched.Missing(batch) {
   196  		queue[hash] = struct{}{}
   197  	}
   198  	for len(queue) > 0 {
   199  		// Fetch all the queued nodes in a random order
   200  		results := make([]SyncResult, 0, len(queue))
   201  		for hash := range queue {
   202  			data, err := srcDb.Node(hash)
   203  			if err != nil {
   204  				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
   205  			}
   206  			results = append(results, SyncResult{hash, data})
   207  		}
   208  		// Feed the retrieved results back and queue new tasks
   209  		if _, index, err := sched.Process(results); err != nil {
   210  			t.Fatalf("failed to process result #%d: %v", index, err)
   211  		}
   212  		if index, err := sched.Commit(diskdb); err != nil {
   213  			t.Fatalf("failed to commit data #%d: %v", index, err)
   214  		}
   215  		queue = make(map[common.Hash]struct{})
   216  		for _, hash := range sched.Missing(batch) {
   217  			queue[hash] = struct{}{}
   218  		}
   219  	}
   220  	// Cross check that the two tries are in sync
   221  	checkTrieContents(t, triedb, srcTrie.Root(), srcData)
   222  }
   223  
   224  // Tests that the trie scheduler can correctly reconstruct the state even if only
   225  // partial results are returned (Even those randomly), others sent only later.
   226  func TestIterativeRandomDelayedTrieSync(t *testing.T) {
   227  	// Create a random trie to copy
   228  	srcDb, srcTrie, srcData := makeTestTrie()
   229  
   230  	// Create a destination trie and sync with the scheduler
   231  	diskdb := aquadb.NewMemDatabase()
   232  	triedb := NewDatabase(diskdb)
   233  	sched := NewTrieSync(srcTrie.Hash(), diskdb, nil)
   234  
   235  	queue := make(map[common.Hash]struct{})
   236  	for _, hash := range sched.Missing(10000) {
   237  		queue[hash] = struct{}{}
   238  	}
   239  	for len(queue) > 0 {
   240  		// Sync only half of the scheduled nodes, even those in random order
   241  		results := make([]SyncResult, 0, len(queue)/2+1)
   242  		for hash := range queue {
   243  			data, err := srcDb.Node(hash)
   244  			if err != nil {
   245  				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
   246  			}
   247  			results = append(results, SyncResult{hash, data})
   248  
   249  			if len(results) >= cap(results) {
   250  				break
   251  			}
   252  		}
   253  		// Feed the retrieved results back and queue new tasks
   254  		if _, index, err := sched.Process(results); err != nil {
   255  			t.Fatalf("failed to process result #%d: %v", index, err)
   256  		}
   257  		if index, err := sched.Commit(diskdb); err != nil {
   258  			t.Fatalf("failed to commit data #%d: %v", index, err)
   259  		}
   260  		for _, result := range results {
   261  			delete(queue, result.Hash)
   262  		}
   263  		for _, hash := range sched.Missing(10000) {
   264  			queue[hash] = struct{}{}
   265  		}
   266  	}
   267  	// Cross check that the two tries are in sync
   268  	checkTrieContents(t, triedb, srcTrie.Root(), srcData)
   269  }
   270  
   271  // Tests that a trie sync will not request nodes multiple times, even if they
   272  // have such references.
   273  func TestDuplicateAvoidanceTrieSync(t *testing.T) {
   274  	// Create a random trie to copy
   275  	srcDb, srcTrie, srcData := makeTestTrie()
   276  
   277  	// Create a destination trie and sync with the scheduler
   278  	diskdb := aquadb.NewMemDatabase()
   279  	triedb := NewDatabase(diskdb)
   280  	sched := NewTrieSync(srcTrie.Hash(), diskdb, nil)
   281  
   282  	queue := append([]common.Hash{}, sched.Missing(0)...)
   283  	requested := make(map[common.Hash]struct{})
   284  
   285  	for len(queue) > 0 {
   286  		results := make([]SyncResult, len(queue))
   287  		for i, hash := range queue {
   288  			data, err := srcDb.Node(hash)
   289  			if err != nil {
   290  				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
   291  			}
   292  			if _, ok := requested[hash]; ok {
   293  				t.Errorf("hash %x already requested once", hash)
   294  			}
   295  			requested[hash] = struct{}{}
   296  
   297  			results[i] = SyncResult{hash, data}
   298  		}
   299  		if _, index, err := sched.Process(results); err != nil {
   300  			t.Fatalf("failed to process result #%d: %v", index, err)
   301  		}
   302  		if index, err := sched.Commit(diskdb); err != nil {
   303  			t.Fatalf("failed to commit data #%d: %v", index, err)
   304  		}
   305  		queue = append(queue[:0], sched.Missing(0)...)
   306  	}
   307  	// Cross check that the two tries are in sync
   308  	checkTrieContents(t, triedb, srcTrie.Root(), srcData)
   309  }
   310  
   311  // Tests that at any point in time during a sync, only complete sub-tries are in
   312  // the database.
   313  func TestIncompleteTrieSync(t *testing.T) {
   314  	// Create a random trie to copy
   315  	srcDb, srcTrie, _ := makeTestTrie()
   316  
   317  	// Create a destination trie and sync with the scheduler
   318  	diskdb := aquadb.NewMemDatabase()
   319  	triedb := NewDatabase(diskdb)
   320  	sched := NewTrieSync(srcTrie.Hash(), diskdb, nil)
   321  
   322  	added := []common.Hash{}
   323  	queue := append([]common.Hash{}, sched.Missing(1)...)
   324  	for len(queue) > 0 {
   325  		// Fetch a batch of trie nodes
   326  		results := make([]SyncResult, len(queue))
   327  		for i, hash := range queue {
   328  			data, err := srcDb.Node(hash)
   329  			if err != nil {
   330  				t.Fatalf("failed to retrieve node data for %x: %v", hash, err)
   331  			}
   332  			results[i] = SyncResult{hash, data}
   333  		}
   334  		// Process each of the trie nodes
   335  		if _, index, err := sched.Process(results); err != nil {
   336  			t.Fatalf("failed to process result #%d: %v", index, err)
   337  		}
   338  		if index, err := sched.Commit(diskdb); err != nil {
   339  			t.Fatalf("failed to commit data #%d: %v", index, err)
   340  		}
   341  		for _, result := range results {
   342  			added = append(added, result.Hash)
   343  		}
   344  		// Check that all known sub-tries in the synced trie are complete
   345  		for _, root := range added {
   346  			if err := checkTrieConsistency(triedb, root); err != nil {
   347  				t.Fatalf("trie inconsistent: %v", err)
   348  			}
   349  		}
   350  		// Fetch the next batch to retrieve
   351  		queue = append(queue[:0], sched.Missing(1)...)
   352  	}
   353  	// Sanity check that removing any node from the database is detected
   354  	for _, node := range added[1:] {
   355  		key := node.Bytes()
   356  		value, _ := diskdb.Get(key)
   357  
   358  		diskdb.Delete(key)
   359  		if err := checkTrieConsistency(triedb, added[0]); err == nil {
   360  			t.Fatalf("trie inconsistency not caught, missing: %x", key)
   361  		}
   362  		diskdb.Put(key, value)
   363  	}
   364  }