1 files changed, 47 insertions, 0 deletions
diff --git a/libbb/bb_getgroups.c b/libbb/bb_getgroups.c
new file mode 100644
index 000000000..59ae53738
--- /dev/null
+++ b/libbb/bb_getgroups.c
@@ -0,0 +1,47 @@
+/*
+ * Utility routines.
+ *
+ * Copyright (C) 2017 Denys Vlasenko
+ *
+ * Licensed under GPLv2, see file LICENSE in this source tree.
+ */
+
+//kbuild:lib-y += bb_getgroups.o
+
+#include "libbb.h"
+
+gid_t* FAST_FUNC bb_getgroups(int *ngroups, gid_t *group_array)
+{
+	int n = ngroups ? *ngroups : 0;
+
+	/* getgroups may be a bit expensive, try to use it only once */
+	if (n < 32)
+		n = 32;
+
+	for (;;) {
+// FIXME: ash tries so hard to not die on OOM (when we are called from test),
+// and we spoil it with just one xrealloc here
+		group_array = xrealloc(group_array, (n+1) * sizeof(group_array[0]));
+		n = getgroups(n, group_array);
+		/*
+		 * If buffer is too small, kernel does not return new_n > n.
+		 * It returns -1 and EINVAL:
+		 */
+		if (n >= 0) {
+			/* Terminator for bb_getgroups(NULL, NULL) usage */
+			group_array[n] = (gid_t) -1;
+			break;
+		}
+		if (errno == EINVAL) { /* too small? */
+			/* This is the way to ask kernel how big the array is */
+			n = getgroups(0, group_array);
+			continue;
+		}
+		/* Some other error (should never happen on Linux) */
+		bb_perror_msg_and_die("getgroups");
+	}
+
+	if (ngroups)
+		*ngroups = n;
+	return group_array;
+}