addon fix for the old issue to /proc full paths not being recorded (kubescape#872)

* addon fix for the old issue to /proc full paths not being recorded

Signed-off-by: entlein <einentlein@gmail.com>

* that regex was too narrow, adding the obvious ones, but might need to RCA why this regressed in the first place

Signed-off-by: entlein <einentlein@gmail.com>

---------

Signed-off-by: entlein <einentlein@gmail.com>

Author: entlein

pkg/utils/normalize_path_test.go

@@ -50,6 +50,30 @@ func TestNormalizePath(t *testing.T) {
 			input:    "/46/task",
 			expected: "/proc/46/task",
 		},
+		{
+			// #721 regression: runc:[2:INIT] user-namespace-setup paths outside
+			// the old (task|fd) allowlist previously leaked /proc-less.
+			name:     "headless proc path (setgroups)",
+			input:    "/17/setgroups",
+			expected: "/proc/17/setgroups",
+		},
+		{
+			name:     "headless proc path (gid_map)",
+			input:    "/1/gid_map",
+			expected: "/proc/1/gid_map",
+		},
+		{
+			name:     "headless proc path (uid_map)",
+			input:    "/1/uid_map",
+			expected: "/proc/1/uid_map",
+		},
+		{
+			// A non-proc path whose leading segment is non-numeric must be
+			// untouched even though a later segment looks proc-like.
+			name:     "non-proc path with data dir",
+			input:    "/data/appendonlydir/x",
+			expected: "/data/appendonlydir/x",
+		},
 		{
 			name:     "relative path (not dot)",
 			input:    "usr/bin/ls",

pkg/utils/path.go

@@ -6,7 +6,17 @@ import (
 	"strings"
 )
 
-var headlessProcRegex = regexp.MustCompile(`^/\d+/(task|fd)(/|$)`)
+// headlessProcRegex matches a headless /proc/<pid>/<file> path — a /proc/<pid>/...
+// path stripped of its /proc root — which NormalizePath re-roots under /proc.
+//
+// The allowlist enumerates the /proc/<pid> entries opened by runc:[2:INIT]
+// during container/user-namespace setup. It was previously only (task|fd),
+// which let the sibling entries (setgroups, gid_map, uid_map, status, cgroup,
+// ...) leak /proc-less into learned ContainerProfiles — a regression of #721.
+// It stays an explicit allowlist rather than a bare `^/\d+` catch-all so a
+// genuine top-level numeric directory is never misread as a PID; extend it if
+// another /proc/<pid> entry is observed leaking.
+var headlessProcRegex = regexp.MustCompile(`^/\d+/(task|fd|setgroups|gid_map|uid_map|status|stat|cgroup|mountinfo|maps|environ|comm|cmdline|ns)(/|$)`)
 
 // NormalizePath normalizes a path by:
 // 1. Prepending "/proc" to "headless" proc paths (e.g. /46/task/46/fd -> /proc/46/task/46/fd)