-
-
Notifications
You must be signed in to change notification settings - Fork 962
perf(send): Parallelize per-device encryption for large group messages #900
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Changes from all commits
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -14,9 +14,11 @@ import ( | |
| "encoding/hex" | ||
| "errors" | ||
| "fmt" | ||
| "runtime" | ||
| "sort" | ||
| "strconv" | ||
| "strings" | ||
| "sync" | ||
| "time" | ||
|
|
||
| "github.com/rs/zerolog" | ||
|
|
@@ -1164,74 +1166,251 @@ func (cli *Client) encryptMessageForDevices( | |
| ) ([]waBinary.Node, bool) { | ||
| ownJID := cli.getOwnID() | ||
| ownLID := cli.getOwnLID() | ||
| includeIdentity := false | ||
| participantNodes := make([]waBinary.Node, 0, len(allDevices)) | ||
| var retryDevices, retryEncryptionIdentities []types.JID | ||
| for _, jid := range allDevices { | ||
| plaintext := msgPlaintext | ||
| if (jid.User == ownJID.User || jid.User == ownLID.User) && dsmPlaintext != nil { | ||
| if jid == ownJID || jid == ownLID { | ||
|
|
||
| // Scoped cache activation (batch load sessions & identity keys) – only if backend supports it | ||
| if cli.Store != nil && cli.Store.Cache != nil && len(allDevices) > 1 { // only bother if >1 device | ||
| addresses := make([]string, 0, len(allDevices)) | ||
| for _, jid := range allDevices { | ||
| if jid == ownJID || jid == ownLID { // skip own identity placeholders, own device sessions not needed | ||
| continue | ||
| } | ||
| plaintext = dsmPlaintext | ||
| addresses = append(addresses, jid.SignalAddress().String()) | ||
| } | ||
| encryptionIdentity := jid | ||
| if jid.Server == types.DefaultUserServer { | ||
| lidForPN, err := cli.Store.LIDs.GetLIDForPN(ctx, jid) | ||
| if len(addresses) > 0 { | ||
| // Batch fetch | ||
| sessMap, err := cli.Store.Cache.GetSessions(ctx, addresses) | ||
| if err != nil { | ||
| cli.Log.Warnf("Failed to get LID for %s: %v", jid, err) | ||
| } else if !lidForPN.IsEmpty() { | ||
| cli.migrateSessionStore(ctx, jid, lidForPN) | ||
| encryptionIdentity = lidForPN | ||
| cli.Log.Warnf("Scoped cache: failed to batch fetch sessions: %v", err) | ||
| sessMap = map[string][]byte{} | ||
| } | ||
| idMap, err := cli.Store.Cache.GetIdentityKeys(ctx, addresses) | ||
| if err != nil { | ||
| cli.Log.Warnf("Scoped cache: failed to batch fetch identity keys: %v", err) | ||
| idMap = map[string][32]byte{} | ||
| } | ||
| cli.Store.SessionsCacheLock.Lock() | ||
| cli.Store.SessionsCache = sessMap | ||
| cli.Store.IdentityKeysCache = idMap | ||
| cli.Store.SessionsCacheLock.Unlock() | ||
| defer func() { | ||
| cli.Store.SessionsCacheLock.Lock() | ||
| sessionsCopy := cli.Store.SessionsCache | ||
| identityCopy := cli.Store.IdentityKeysCache | ||
| cli.Store.SessionsCache = nil | ||
| cli.Store.IdentityKeysCache = nil | ||
| cli.Store.SessionsCacheLock.Unlock() | ||
| // Flush new/updated sessions & identities back (best effort) | ||
| if len(sessionsCopy) > 0 { | ||
| if err := cli.Store.Cache.PutSessions(context.Background(), sessionsCopy); err != nil { | ||
| cli.Log.Warnf("Scoped cache: failed to flush sessions: %v", err) | ||
| } | ||
| } | ||
| if len(identityCopy) > 0 { | ||
| if err := cli.Store.Cache.PutIdentityKeys(context.Background(), identityCopy); err != nil { | ||
| cli.Log.Warnf("Scoped cache: failed to flush identity keys: %v", err) | ||
| } | ||
| } | ||
| }() | ||
| } | ||
| } | ||
| includeIdentity := false | ||
| participantNodes := make([]waBinary.Node, 0, len(allDevices)) | ||
|
|
||
| encrypted, isPreKey, err := cli.encryptMessageForDeviceAndWrap( | ||
| ctx, plaintext, jid, encryptionIdentity, nil, encAttrs, | ||
| ) | ||
| if errors.Is(err, ErrNoSession) { | ||
| retryDevices = append(retryDevices, jid) | ||
| retryEncryptionIdentities = append(retryEncryptionIdentities, encryptionIdentity) | ||
| continue | ||
| } else if err != nil { | ||
| // TODO return these errors if it's a fatal one (like context cancellation or database) | ||
| cli.Log.Warnf("Failed to encrypt %s for %s: %v", id, jid, err) | ||
| continue | ||
| } | ||
| const parallelThreshold = 8 | ||
|
|
||
| participantNodes = append(participantNodes, *encrypted) | ||
| if isPreKey { | ||
| includeIdentity = true | ||
| } | ||
| const minConcurrency = 1 | ||
| const maxConcurrency = 64 | ||
| var concurrency int | ||
| if cli.EncryptConcurrency > 0 { | ||
| concurrency = cli.EncryptConcurrency | ||
| } else { | ||
| procs := runtime.GOMAXPROCS(0) | ||
|
|
||
| concurrency = procs * 2 | ||
| } | ||
| if len(retryDevices) > 0 { | ||
| bundles, err := cli.fetchPreKeys(ctx, retryDevices) | ||
| if err != nil { | ||
| cli.Log.Warnf("Failed to fetch prekeys for %v to retry encryption: %v", retryDevices, err) | ||
| } else { | ||
| for i, jid := range retryDevices { | ||
| resp := bundles[jid] | ||
| if resp.err != nil { | ||
| cli.Log.Warnf("Failed to fetch prekey for %s: %v", jid, resp.err) | ||
|
|
||
| if concurrency < minConcurrency { | ||
| concurrency = minConcurrency | ||
| } | ||
| if concurrency > maxConcurrency { | ||
| concurrency = maxConcurrency | ||
| } | ||
|
|
||
| if concurrency > len(allDevices) { | ||
| concurrency = len(allDevices) | ||
| } | ||
|
|
||
| if len(allDevices) == 0 { | ||
| return participantNodes, includeIdentity | ||
| } | ||
|
|
||
| if len(allDevices) < parallelThreshold || concurrency <= 1 { | ||
| // Fall back to original sequential implementation for small batches | ||
|
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Is there a way to minimize the amount of duplicated code here? |
||
| var retryDevices, retryEncryptionIdentities []types.JID | ||
| for _, jid := range allDevices { | ||
| plaintext := msgPlaintext | ||
| if (jid.User == ownJID.User || jid.User == ownLID.User) && dsmPlaintext != nil { | ||
| if jid == ownJID || jid == ownLID { | ||
| continue | ||
| } | ||
| plaintext := msgPlaintext | ||
| if (jid.User == ownJID.User || jid.User == ownLID.User) && dsmPlaintext != nil { | ||
| plaintext = dsmPlaintext | ||
| } | ||
| encrypted, isPreKey, err := cli.encryptMessageForDeviceAndWrap( | ||
| ctx, plaintext, jid, retryEncryptionIdentities[i], resp.bundle, encAttrs, | ||
| ) | ||
| plaintext = dsmPlaintext | ||
| } | ||
| encryptionIdentity := jid | ||
| if jid.Server == types.DefaultUserServer { | ||
| lidForPN, err := cli.Store.LIDs.GetLIDForPN(ctx, jid) | ||
| if err != nil { | ||
| // TODO return these errors if it's a fatal one (like context cancellation or database) | ||
| cli.Log.Warnf("Failed to encrypt %s for %s (retry): %v", id, jid, err) | ||
| cli.Log.Warnf("Failed to get LID for %s: %v", jid, err) | ||
| } else if !lidForPN.IsEmpty() { | ||
| cli.migrateSessionStore(ctx, jid, lidForPN) | ||
| encryptionIdentity = lidForPN | ||
| } | ||
| } | ||
| encrypted, isPreKey, err := cli.encryptMessageForDeviceAndWrap( | ||
| ctx, plaintext, jid, encryptionIdentity, nil, encAttrs, | ||
| ) | ||
| if errors.Is(err, ErrNoSession) { | ||
| retryDevices = append(retryDevices, jid) | ||
| retryEncryptionIdentities = append(retryEncryptionIdentities, encryptionIdentity) | ||
| continue | ||
| } else if err != nil { | ||
| cli.Log.Warnf("Failed to encrypt %s for %s: %v", id, jid, err) | ||
| continue | ||
| } | ||
| participantNodes = append(participantNodes, *encrypted) | ||
| if isPreKey { | ||
| includeIdentity = true | ||
| } | ||
| } | ||
| participantNodes, includeIdentity = cli.retryEncryptMissing(ctx, id, msgPlaintext, dsmPlaintext, encAttrs, ownJID, ownLID, participantNodes, includeIdentity, retryDevices, retryEncryptionIdentities) | ||
| return participantNodes, includeIdentity | ||
| } | ||
|
|
||
| type workItem struct { | ||
| jid types.JID | ||
| plaintext []byte | ||
| encryptionIdentity types.JID | ||
| } | ||
| type resultItem struct { | ||
| jid types.JID | ||
| node *waBinary.Node | ||
| isPreKey bool | ||
| retry bool | ||
| err error | ||
| encIdent types.JID | ||
| } | ||
| jobs := make(chan workItem) | ||
| results := make(chan resultItem, len(allDevices)) | ||
| var wg sync.WaitGroup | ||
|
|
||
| worker := func() { | ||
| defer wg.Done() | ||
| for wi := range jobs { | ||
| encrypted, isPreKey, err := cli.encryptMessageForDeviceAndWrap(ctx, wi.plaintext, wi.jid, wi.encryptionIdentity, nil, encAttrs) | ||
| if errors.Is(err, ErrNoSession) { | ||
| results <- resultItem{jid: wi.jid, retry: true, err: err, encIdent: wi.encryptionIdentity} | ||
| continue | ||
| } else if err != nil { | ||
| results <- resultItem{jid: wi.jid, err: err} | ||
| continue | ||
| } | ||
| results <- resultItem{jid: wi.jid, node: encrypted, isPreKey: isPreKey} | ||
| } | ||
| } | ||
|
|
||
| for i := 0; i < concurrency; i++ { | ||
| wg.Add(1) | ||
| go worker() | ||
| } | ||
|
|
||
| go func() { | ||
| for _, jid := range allDevices { | ||
| plaintext := msgPlaintext | ||
| if (jid.User == ownJID.User || jid.User == ownLID.User) && dsmPlaintext != nil { | ||
| if jid == ownJID || jid == ownLID { | ||
| continue | ||
| } | ||
| participantNodes = append(participantNodes, *encrypted) | ||
| if isPreKey { | ||
| includeIdentity = true | ||
| plaintext = dsmPlaintext | ||
| } | ||
| encryptionIdentity := jid | ||
| if jid.Server == types.DefaultUserServer { | ||
| lidForPN, err := cli.Store.LIDs.GetLIDForPN(ctx, jid) | ||
| if err != nil { | ||
| cli.Log.Warnf("Failed to get LID for %s: %v", jid, err) | ||
| } else if !lidForPN.IsEmpty() { | ||
| cli.migrateSessionStore(ctx, jid, lidForPN) | ||
| encryptionIdentity = lidForPN | ||
| } | ||
| } | ||
| jobs <- workItem{jid: jid, plaintext: plaintext, encryptionIdentity: encryptionIdentity} | ||
| } | ||
| close(jobs) | ||
| }() | ||
|
|
||
| var retryDevices []types.JID | ||
| var retryEncryptionIdentities []types.JID | ||
| for completed := 0; completed < len(allDevices); completed++ { | ||
| res := <-results | ||
| if res.err != nil { | ||
| if res.retry { | ||
| retryDevices = append(retryDevices, res.jid) | ||
| retryEncryptionIdentities = append(retryEncryptionIdentities, res.encIdent) | ||
| } else { | ||
| cli.Log.Warnf("Failed to encrypt %s for %s: %v", id, res.jid, res.err) | ||
| } | ||
| continue | ||
| } | ||
| if res.node != nil { | ||
| participantNodes = append(participantNodes, *res.node) | ||
| if res.isPreKey { | ||
| includeIdentity = true | ||
| } | ||
| } | ||
| } | ||
|
|
||
| go func() { wg.Wait(); close(results) }() | ||
|
|
||
| participantNodes, includeIdentity = cli.retryEncryptMissing(ctx, id, msgPlaintext, dsmPlaintext, encAttrs, ownJID, ownLID, participantNodes, includeIdentity, retryDevices, retryEncryptionIdentities) | ||
| return participantNodes, includeIdentity | ||
| } | ||
|
|
||
| func (cli *Client) retryEncryptMissing( | ||
|
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Is it reasonable to abstract this logic outside of the function? 🤔 |
||
| ctx context.Context, | ||
| id string, | ||
| msgPlaintext, dsmPlaintext []byte, | ||
| encAttrs waBinary.Attrs, | ||
| ownJID, ownLID types.JID, | ||
| participantNodes []waBinary.Node, | ||
| includeIdentity bool, | ||
| retryDevices, retryEncryptionIdentities []types.JID, | ||
| ) ([]waBinary.Node, bool) { | ||
| if len(retryDevices) == 0 { | ||
| return participantNodes, includeIdentity | ||
| } | ||
| bundles, err := cli.fetchPreKeys(ctx, retryDevices) | ||
| if err != nil { | ||
| cli.Log.Warnf("Failed to fetch prekeys for %v to retry encryption: %v", retryDevices, err) | ||
| return participantNodes, includeIdentity | ||
| } | ||
| for i, jid := range retryDevices { | ||
| resp := bundles[jid] | ||
| if resp.err != nil { | ||
| cli.Log.Warnf("Failed to fetch prekey for %s: %v", jid, resp.err) | ||
| continue | ||
| } | ||
| plaintext := msgPlaintext | ||
| if (jid.User == ownJID.User || jid.User == ownLID.User) && dsmPlaintext != nil { | ||
| plaintext = dsmPlaintext | ||
| } | ||
| encrypted, isPreKey, err := cli.encryptMessageForDeviceAndWrap( | ||
| ctx, plaintext, jid, retryEncryptionIdentities[i], resp.bundle, encAttrs, | ||
| ) | ||
| if err != nil { | ||
| cli.Log.Warnf("Failed to encrypt %s for %s (retry): %v", id, jid, err) | ||
| continue | ||
| } | ||
| participantNodes = append(participantNodes, *encrypted) | ||
| if isPreKey { | ||
| includeIdentity = true | ||
| } | ||
| } | ||
| return participantNodes, includeIdentity | ||
|
|
||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🤔 How was this number set? So, if I had 9 devices, you'd start a new routine for however many EncryptConsistency is (lets say I have 4 cpus).. Have you consider if that's faster than just letting the same 1 routine process just that one more device??