diff options
| author | Paul Buetow <paul@buetow.org> | 2026-07-22 23:51:18 +0300 |
|---|---|---|
| committer | Paul Buetow <paul@buetow.org> | 2026-07-22 23:51:18 +0300 |
| commit | 849951be1d1a7ee9f9302006ccb187bf5b4e36f3 (patch) | |
| tree | 496c924a03a9ea6212e29bb4699e268066ebad81 /internal/server/handlers/readcommand_glob_cap_test.go | |
| parent | bf78b3abffee6d49c08ca2980156afc455994969 (diff) | |
feat: DTail fork — server/client feature development
Squashed development of the snonux/dtail fork's product code (internal/, cmd/)
since diverging from mimecast/dtail. Major areas:
- Read/output path: the former "turbo" channel-less path is now the single,
default server-side read/output path for cat/grep/tail and MapReduce; the old
channel-based path and its config/env toggles were removed.
- MapReduce: single aggregate implementation (server + serverless) fed directly
by a processor pipeline, with input-exhausted finalization via the shutdown
coordinator; high-concurrency and data-race fixes.
- Journal source reads (journal:unit.service) via journalctl, Linux-gated behind
a journal-v1 capability.
- Auth-key fast reconnect: in-memory per-user public-key cache with TTL/max-keys,
registered over an authenticated session (AUTHKEY), checked before
authorized_keys.
- Interactive query reload (--interactive-query) with SESSION START/UPDATE
generation boundaries and capability negotiation.
- Client-side deadlines: --timeout / --shutdownAfter as context deadlines;
follow shutdown handling.
- Client logging: diagnostics-only daily log by default, opt-in payload tee via
--log-payload.
- Numerous correctness fixes (buffer-pool double-recycle races, EOF-sentinel
leaks, glob-expansion cap, TOCTOU in CSV parsing) with accompanying unit tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Diffstat (limited to 'internal/server/handlers/readcommand_glob_cap_test.go')
| -rw-r--r-- | internal/server/handlers/readcommand_glob_cap_test.go | 214 |
1 files changed, 214 insertions, 0 deletions
diff --git a/internal/server/handlers/readcommand_glob_cap_test.go b/internal/server/handlers/readcommand_glob_cap_test.go new file mode 100644 index 0000000..b77d8d9 --- /dev/null +++ b/internal/server/handlers/readcommand_glob_cap_test.go @@ -0,0 +1,214 @@ +package handlers + +// Tests for the glob expansion cap (MaxGlobTargets). +// +// Security context: an authenticated user with a broad read permission (e.g. +// "readfiles:^/.*") can craft a glob that matches thousands of paths. Before +// this fix, filepath.Glob would return every match, pendingFiles would be +// incremented by the full count, and one goroutine would be spawned per path +// — all before any concurrency limiter was consulted. The cap enforced here +// truncates the expansion to MaxGlobTargets paths so that the number of +// goroutines and the memory used by a single read command are bounded. + +import ( + "context" + "fmt" + "os" + "path/filepath" + "sync/atomic" + "testing" + "time" + + "github.com/mimecast/dtail/internal/io/fs" + "github.com/mimecast/dtail/internal/lcontext" + maprserver "github.com/mimecast/dtail/internal/mapr/server" + "github.com/mimecast/dtail/internal/omode" + "github.com/mimecast/dtail/internal/regex" +) + +// globCapTestServer is a minimal readCommandServer implementation whose sole +// purpose is to count PrepareReadTarget invocations (i.e. dispatched files). +// All output/aggregate/lifecycle methods are stubs. The catLimiter is +// generously sized so it never blocks the test. +type globCapTestServer struct { + catLimiter chan struct{} + tailLimiter chan struct{} + serverMessage chan string + // preparedCount counts how many file paths were dispatched for reading. + preparedCount int32 + // pendingFiles mirrors the lifecycle counter used by readFiles. + pendingFiles int32 + // maxGlobTargets is the cap returned by MaxGlobTargets(). + maxGlobTargets int +} + +func newGlobCapTestServer(maxGlobTargets int) *globCapTestServer { + // Use a large limiter so reads are never blocked by concurrency limits + // — we only want to test the glob cap, not the semaphore. + limiter := make(chan struct{}, 10000) + return &globCapTestServer{ + catLimiter: limiter, + tailLimiter: limiter, + serverMessage: make(chan string, 128), + maxGlobTargets: maxGlobTargets, + } +} + +// PrepareReadTarget records the path and returns a valid journal-free target. +// We use a plain ValidatedReadTarget with FileKind so no actual file I/O is +// attempted; the goroutine that calls read() will fail gracefully when the +// nonexistent catFile cannot be opened, but that is fine for this test. +func (s *globCapTestServer) PrepareReadTarget(path string) (fs.ValidatedReadTarget, bool) { + atomic.AddInt32(&s.preparedCount, 1) + // Return a valid file-kind target so readFileIfPermissions proceeds past + // the permission check and reaches the actual read machinery. + return fs.ValidatedReadTarget{Kind: fs.FileKind}, true +} + +func (s *globCapTestServer) CatLimiter() chan struct{} { return s.catLimiter } +func (s *globCapTestServer) TailLimiter() chan struct{} { return s.tailLimiter } + +func (s *globCapTestServer) LogContext() interface{} { return "glob-cap-test" } +func (s *globCapTestServer) SendServerMessage(msg string) { s.drainOrStore(msg) } +func (s *globCapTestServer) ServerMessagesChannel() chan string { return s.serverMessage } +func (s *globCapTestServer) Hostname() string { return "testhost" } +func (s *globCapTestServer) PlainOutput() bool { return false } +func (s *globCapTestServer) Serverless() bool { return false } + +func (s *globCapTestServer) drainOrStore(msg string) { + select { + case s.serverMessage <- msg: + default: + } +} + +func (s *globCapTestServer) Aggregate() *maprserver.Aggregate { return nil } + +// AddPendingFiles tracks the lifecycle counter that readFiles uses to +// signal EOF; it operates on a separate field from preparedCount. +func (s *globCapTestServer) AddPendingFiles(delta int32) int32 { + return atomic.AddInt32(&s.pendingFiles, delta) +} + +func (s *globCapTestServer) CompletePendingFile() (int32, int32) { + remaining := atomic.AddInt32(&s.pendingFiles, -1) + return remaining, 0 +} + +func (s *globCapTestServer) PendingAndActive() (int32, int32) { + return atomic.LoadInt32(&s.pendingFiles), 0 +} +func (s *globCapTestServer) ActiveSessionGeneration() uint64 { return 0 } +func (s *globCapTestServer) TriggerShutdown() {} + +func (s *globCapTestServer) DirectOutputActive() bool { return false } +func (s *globCapTestServer) EnableDirectOutput() bool { return false } +func (s *globCapTestServer) HasOutputEOF() bool { return false } +func (s *globCapTestServer) FlushOutput() {} +func (s *globCapTestServer) OutputEpoch() uint64 { return 0 } +func (s *globCapTestServer) SignalOutputEOF(epoch uint64) {} +func (s *globCapTestServer) GetOutputChannel() chan []byte { return nil } +func (s *globCapTestServer) OutputChannelLen() int { return 0 } +func (s *globCapTestServer) WaitForOutputEOFAck(time.Duration) bool { return true } + +func (s *globCapTestServer) ReadGlobRetryInterval() time.Duration { return time.Millisecond } +func (s *globCapTestServer) ReadRetryInterval() time.Duration { return time.Millisecond } +func (s *globCapTestServer) MaxLineLength() int { return 1024 * 1024 } +func (s *globCapTestServer) OutputTransmissionDelay() time.Duration { return time.Millisecond } +func (s *globCapTestServer) OutputEOFWaitDuration(int) time.Duration { return time.Millisecond } +func (s *globCapTestServer) ShutdownSerializeWait() time.Duration { return time.Millisecond } +func (s *globCapTestServer) ShutdownIdleRecheckWait() time.Duration { return time.Millisecond } +func (s *globCapTestServer) OutputEOFAckTimeout() time.Duration { return time.Millisecond } + +// MaxGlobTargets returns the configurable cap for this test server. +func (s *globCapTestServer) MaxGlobTargets() int { return s.maxGlobTargets } + +// verify the interface is satisfied at compile time +var _ readCommandServer = (*globCapTestServer)(nil) + +// createTempFiles creates n empty files named file0000.log … in dir and +// returns the glob pattern that matches all of them. +func createTempFiles(t *testing.T, dir string, n int) string { + t.Helper() + for i := 0; i < n; i++ { + name := filepath.Join(dir, fmt.Sprintf("file%04d.log", i)) + if err := os.WriteFile(name, nil, 0o600); err != nil { + t.Fatalf("create temp file %s: %v", name, err) + } + } + return filepath.Join(dir, "*.log") +} + +// TestGlobCapTruncatesExcessPaths verifies that when a glob matches more files +// than MaxGlobTargets, only MaxGlobTargets paths are dispatched. The check is +// done by counting PrepareReadTarget invocations on the mock server. +func TestGlobCapTruncatesExcessPaths(t *testing.T) { + resetServerLogger(t) + + const ( + totalFiles = 20 // files on disk — clearly above the cap + cap = 5 // deliberately low cap to prove truncation + ) + + dir := t.TempDir() + glob := createTempFiles(t, dir, totalFiles) + + srv := newGlobCapTestServer(cap) + cmd := newReadCommand(srv, omode.CatClient) + + // Run readGlob with retries=1 so it finds files immediately. + cmd.readGlob(context.Background(), lcontext.LContext{}, glob, regex.NewNoop(), 1) + + got := int(atomic.LoadInt32(&srv.preparedCount)) + if got > cap { + t.Fatalf("glob cap not enforced: dispatched %d paths, expected at most %d", got, cap) + } + if got == 0 { + t.Fatal("no paths were dispatched; expected exactly cap paths to be served") + } +} + +// TestGlobCapUnderLimitPassesAll verifies that when the number of glob matches +// is at or below MaxGlobTargets, all paths are dispatched without truncation. +func TestGlobCapUnderLimitPassesAll(t *testing.T) { + resetServerLogger(t) + + const ( + totalFiles = 5 + cap = 10 // cap is above totalFiles — nothing should be dropped + ) + + dir := t.TempDir() + glob := createTempFiles(t, dir, totalFiles) + + srv := newGlobCapTestServer(cap) + cmd := newReadCommand(srv, omode.CatClient) + + cmd.readGlob(context.Background(), lcontext.LContext{}, glob, regex.NewNoop(), 1) + + got := int(atomic.LoadInt32(&srv.preparedCount)) + if got != totalFiles { + t.Fatalf("expected all %d paths dispatched, got %d", totalFiles, got) + } +} + +// TestGlobCapExactlyAtLimit verifies that when the number of glob matches +// equals MaxGlobTargets exactly, all paths are dispatched (boundary condition). +func TestGlobCapExactlyAtLimit(t *testing.T) { + resetServerLogger(t) + + const count = 8 // cap == totalFiles + + dir := t.TempDir() + glob := createTempFiles(t, dir, count) + + srv := newGlobCapTestServer(count) + cmd := newReadCommand(srv, omode.CatClient) + + cmd.readGlob(context.Background(), lcontext.LContext{}, glob, regex.NewNoop(), 1) + + got := int(atomic.LoadInt32(&srv.preparedCount)) + if got != count { + t.Fatalf("expected exactly %d paths dispatched (no truncation at limit), got %d", count, got) + } +} |
