[RAMEN9610-21968]ANDROID: xt_qtaguid: Remove tag_entry from process list on untag

A sock_tag_entry can only be part of one process's
pqd_entry->sock_tag_list. RetagGing the socket only updates
sock_tag_entry->tag, and does not add the tag entry to the current
process's pqd_entry list, nor update sock_tag_entry->pid.
So the sock_tag_entry is only ever present in the
pqd_entry list of the process that initially tagged the socket.

A sock_tag_entry can also get created and not be added to any process's
pqd_entry list. This happens if the process that initially tags the
socket has not opened /dev/xt_qtaguid.

ctrl_cmd_untag() supports untagGing from a context other than the
process that initially tagged the socket. Currently, the sock_tag_entry is
only removed from its containing pqd_entry->sock_tag_list if the
process that does the untagGing has opened /dev/xt_qtaguid. However, the
tag entry should always be deleted from its pqd entry list (if present).

Bug: 176919394
Signed-off-by: Kalesh Singh <kaleshsingh@google.com>
Change-Id: I5b6f0c36c0ebefd98cc6873a4057104c7d885ccc
(cherry picked from commit c2ab93b45b5cdc426868fb8793ada2cac20568ef)

diff --git a/net/netfilter/xt_qtaguid.c b/net/netfilter/xt_qtaguid.c
index d261932ee595a048af2941943a5615983df1291f..a61f4367451973ed7365bb7a1208436b4e80cfe7 100644 (file)
--- a/net/netfilter/xt_qtaguid.c
+++ b/net/netfilter/xt_qtaguid.c
@@ -2424,15 +2424,20 @@ int qtaguid_untag(struct socket *el_socket, bool kernel)
         * At first, we want to catch user-space code that is not
         * opening the /dev/xt_qtaguid.
         */
-       if (IS_ERR_OR_NULL(pqd_entry) || !sock_tag_entry->list.next) {
+       if (IS_ERR_OR_NULL(pqd_entry))
                pr_warn_once("qtaguid: %s(): "
                             "User space forgot to open /dev/xt_qtaguid? "
                             "pid=%u tgid=%u sk_pid=%u, uid=%u\n", __func__,
                             current->pid, current->tgid, sock_tag_entry->pid,
                             from_kuid(&init_user_ns, current_fsuid()));
-       } else {
+       /*
+        * This check is needed because tagging from a process that
+        * didn’t open /dev/xt_qtaguid still adds the sock_tag_entry
+        * to sock_tag_tree.
+        */
+       if (sock_tag_entry->list.next)
                list_del(&sock_tag_entry->list);
-       }
+
        spin_unlock_bh(&uid_tag_data_tree_lock);
        /*
         * We don't free tag_ref from the utd_entry here,