tcx: Fix splat during dev unregister

During unregister_netdevice_many_notify(), the ordering of our concerned
function calls is like this:

unregister_netdevice_many_notify
dev_shutdown
qdisc_put
clsact_destroy
tcx_uninstall

The syzbot reproducer triggered a case that the qdisc refcnt is not
zero during dev_shutdown().

tcx_uninstall() will then WARN_ON_ONCE(tcx_entry(entry)->miniq_active)
because the miniq is still active and the entry should not be freed.
The latter assumed that qdisc destruction happens before tcx teardown.

This fix is to avoid tcx_uninstall() doing tcx_entry_free() when the
miniq is still alive and let the clsact_destroy() do the free later, so
that we do not assume any specific ordering for either of them.

If still active, tcx_uninstall() does clear the entry when flushing out
the prog/link. clsact_destroy() will then notice the "!tcx_entry_is_active()"
and then does the tcx_entry_free() eventually.

Fixes: e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")
Reported-by: [e-mail]
Reported-by: Leon Romanovsky <[e-mail]>
Signed-off-by: Martin KaFai Lau <[e-mail]>
Co-developed-by: Daniel Borkmann <[e-mail]>
Signed-off-by: Daniel Borkmann <[e-mail]>
Tested-by: [e-mail]
Tested-by: Leon Romanovsky <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Jakub Kicinski <[e-mail]>

include/linux/bpf_mprog.h

···
 	memcpy(dst->fp_items, src->fp_items, sizeof(src->fp_items));
 }
 
+static inline void bpf_mprog_entry_clear(struct bpf_mprog_entry *dst)
+{
+	memset(dst->fp_items, 0, sizeof(dst->fp_items));
+}
+
+static inline void bpf_mprog_clear_all(struct bpf_mprog_entry *entry,
+				       struct bpf_mprog_entry **entry_new)
+{
+	struct bpf_mprog_entry *peer;
+
+	peer = bpf_mprog_peer(entry);
+	bpf_mprog_entry_clear(peer);
+	peer->parent->count = 0;
+	*entry_new = peer;
+}
+
 static inline void bpf_mprog_entry_grow(struct bpf_mprog_entry *entry, int idx)
 {
 	int total = bpf_mprog_total(entry);

kernel/bpf/tcx.c

···
 
 void tcx_uninstall(struct net_device *dev, bool ingress)
 {
+	struct bpf_mprog_entry *entry, *entry_new = NULL;
 	struct bpf_tuple tuple = {};
-	struct bpf_mprog_entry *entry;
 	struct bpf_mprog_fp *fp;
 	struct bpf_mprog_cp *cp;
+	bool active;
 
 	entry = tcx_entry_fetch(dev, ingress);
 	if (!entry)
 		return;
-	tcx_entry_update(dev, NULL, ingress);
+	active = tcx_entry(entry)->miniq_active;
+	if (active)
+		bpf_mprog_clear_all(entry, &entry_new);
+	tcx_entry_update(dev, entry_new, ingress);
 	tcx_entry_sync();
 	bpf_mprog_foreach_tuple(entry, fp, cp, tuple) {
 		if (tuple.link)
···
 			bpf_prog_put(tuple.prog);
 		tcx_skeys_dec(ingress);
 	}
-	WARN_ON_ONCE(tcx_entry(entry)->miniq_active);
-	tcx_entry_free(entry);
+	if (!active)
+		tcx_entry_free(entry);
 }
 
 int tcx_prog_query(const union bpf_attr *attr, union bpf_attr __user *uattr)