Whitelist getdents() for the speech recognition Linux sandbox

The Speech team recently made a change to the Speech On-Device API to
take in a path to the language pack directory rather than a path to the
dictation.ascii_proto config file. The getdents() syscall is required
for SODA to select the appropriate config file to use.

Bug: 1124367
Change-Id: I9265d7f178d85bb9cb0fea8175878d4db6f26dfe
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2390879Reviewed-by: Matthew Denton <mpdenton@chromium.org>
Commit-Queue: Evan Liu <evliu@google.com>
Cr-Commit-Position: refs/heads/master@{#803973}

sandbox/policy/linux/bpf_speech_recognition_policy_linux.cc

@@ -23,6 +23,16 @@ SpeechRecognitionProcessPolicy::~SpeechRecognitionProcessPolicy() = default;
 ResultExpr SpeechRecognitionProcessPolicy::EvaluateSyscall(
     int system_call_number) const {
   switch (system_call_number) {
+    // Required by the Speech On-Device API (SODA) binary to find the
+    // appropriate configuration file to use within a language pack directory.
+#if defined(__NR_getdents64)
+    case __NR_getdents64:
+      return Allow();
+#endif
+#if defined(__NR_getdents)
+    case __NR_getdents:
+      return Allow();
+#endif
     default:
       auto* broker_process = SandboxLinux::GetInstance()->broker_process();
       if (broker_process->IsSyscallAllowed(system_call_number))