bpf: Do not allow deleting local storage in NMI

[ Upstream commit 350de5b ]

Currently, local storage may deadlock when deferring freeing selem or
local storage through kfree_rcu(), call_rcu() or call_rcu_tasks_trace()
in NMI or reentrant. Since deleting selem in NMI is an unlikely use
case, partially mitigate it by returning error when calling from
bpf_xxx_storage_delete() helpers in NMI. Note that, it is still possible
to deadlock through reentrant. A full mitigation requires returning
error when irqs_disabled() is true, which, however is too heavy-handed
for bpf_xxx_storage_delete().

The long-term solution requires _nolock versions of call_rcu. Another
possible solution is to defer the free through irq_work [0], but it
would grow the size of selem, which is non-ideal.

The check is only needed in bpf_selem_unlink(), which is used by helpers
and syscalls. bpf_selem_unlink_nofail() is fine as it is called during
map and owner tear down that never run in NMI or reentrant.

[0] https://lore.kernel.org/bpf/20260205190233.912-1-alexei.starovoitov@gmail.com/

Fixes: a10787e ("bpf: Enable task local storage for tracing programs")
Signed-off-by: Amery Hung <ameryhung@gmail.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://patch.msgid.link/20260319025716.2361065-1-ameryhung@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

Author: ameryhung

kernel/bpf/bpf_local_storage.c

@@ -393,6 +393,9 @@ int bpf_selem_unlink(struct bpf_local_storage_elem *selem)
 	unsigned long flags;
 	int err;
 
+	if (in_nmi())
+		return -EOPNOTSUPP;
+
 	if (unlikely(!selem_linked_to_storage_lockless(selem)))
 		/* selem has already been unlinked from sk */
 		return 0;