diff --git a/libc/async_safe/async_safe_log.cpp b/libc/async_safe/async_safe_log.cpp
index 207035a..8b2a32b 100644
--- a/libc/async_safe/async_safe_log.cpp
+++ b/libc/async_safe/async_safe_log.cpp
@@ -30,6 +30,7 @@
 #include <ctype.h>
 #include <errno.h>
 #include <fcntl.h>
+#include <linux/net.h>
 #include <pthread.h>
 #include <stdarg.h>
 #include <stddef.h>
@@ -51,12 +52,22 @@
 #include "private/ErrnoRestorer.h"
 #include "private/ScopedPthreadMutexLocker.h"
 
-// Don't call libc's close, since it might call back into us as a result of fdsan.
+// Don't call libc's close or socket, since it might call back into us as a result of fdsan/fdtrack.
 #pragma GCC poison close
 static int __close(int fd) {
   return syscall(__NR_close, fd);
 }
 
+static int __socket(int domain, int type, int protocol) {
+#if defined(__i386__)
+  unsigned long args[3] = {static_cast<unsigned long>(domain), static_cast<unsigned long>(type),
+                           static_cast<unsigned long>(protocol)};
+  return syscall(__NR_socketcall, SYS_SOCKET, &args);
+#else
+  return syscall(__NR_socket, domain, type, protocol);
+#endif
+}
+
 // Must be kept in sync with frameworks/base/core/java/android/util/EventLog.java.
 enum AndroidEventLogType {
   EVENT_TYPE_INT = 0,
@@ -460,7 +471,7 @@
   // found that all logd crashes thus far have had no problem stuffing
   // the UNIX domain socket and moving on so not critical *today*.
 
-  int log_fd = TEMP_FAILURE_RETRY(socket(PF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK, 0));
+  int log_fd = TEMP_FAILURE_RETRY(__socket(PF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK, 0));
   if (log_fd == -1) {
     return -1;
   }