[PATCH bpf] bpf: run bpf programs with preemption disabled

[Alexei Starovoitov <ast@kernel.org>]

Disabled preemption is necessary for proper access to per-cpu maps
from BPF programs.

But the sender side of socket filters didn't have preemption disabled:
unix_dgram_sendmsg->sk_filter->sk_filter_trim_cap->bpf_prog_run_save_cb->BPF_PROG_RUN

and a combination of af_packet with tun device didn't disable either:
tpacket_snd->packet_direct_xmit->packet_pick_tx_queue->ndo_select_queue->
  tun_select_queue->tun_ebpf_select_queue->bpf_prog_run_clear_cb->BPF_PROG_RUN

Disable preemption before executing BPF programs (both classic and extended).

Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
 include/linux/filter.h | 21 ++++++++++++++++++---
 kernel/bpf/cgroup.c    |  2 +-
 2 files changed, 19 insertions(+), 4 deletions(-)

diff --git a/include/linux/filter.h b/include/linux/filter.h
index ad106d845b22..e532fcc6e4b5 100644
--- a/include/linux/filter.h
+++ b/include/linux/filter.h
@@ -591,8 +591,8 @@ static inline u8 *bpf_skb_cb(struct sk_buff *skb)
 	return qdisc_skb_cb(skb)->data;
 }
 
-static inline u32 bpf_prog_run_save_cb(const struct bpf_prog *prog,
-				       struct sk_buff *skb)
+static inline u32 __bpf_prog_run_save_cb(const struct bpf_prog *prog,
+					 struct sk_buff *skb)
 {
 	u8 *cb_data = bpf_skb_cb(skb);
 	u8 cb_saved[BPF_SKB_CB_LEN];
@@ -611,15 +611,30 @@ static inline u32 bpf_prog_run_save_cb(const struct bpf_prog *prog,
 	return res;
 }
 
+static inline u32 bpf_prog_run_save_cb(const struct bpf_prog *prog,
+				       struct sk_buff *skb)
+{
+	u32 res;
+
+	preempt_disable();
+	res = __bpf_prog_run_save_cb(prog, skb);
+	preempt_enable();
+	return res;
+}
+
 static inline u32 bpf_prog_run_clear_cb(const struct bpf_prog *prog,
 					struct sk_buff *skb)
 {
 	u8 *cb_data = bpf_skb_cb(skb);
+	u32 res;
 
 	if (unlikely(prog->cb_access))
 		memset(cb_data, 0, BPF_SKB_CB_LEN);
 
-	return BPF_PROG_RUN(prog, skb);
+	preempt_disable();
+	res = BPF_PROG_RUN(prog, skb);
+	preempt_enable();
+	return res;
 }
 
 static __always_inline u32 bpf_prog_run_xdp(const struct bpf_prog *prog,
diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
index ab612fe9862f..d17d05570a3f 100644
--- a/kernel/bpf/cgroup.c
+++ b/kernel/bpf/cgroup.c
@@ -572,7 +572,7 @@ int __cgroup_bpf_run_filter_skb(struct sock *sk,
 	bpf_compute_and_save_data_end(skb, &saved_data_end);
 
 	ret = BPF_PROG_RUN_ARRAY(cgrp->bpf.effective[type], skb,
-				 bpf_prog_run_save_cb);
+				 __bpf_prog_run_save_cb);
 	bpf_restore_data_end(skb, saved_data_end);
 	__skb_pull(skb, offset);
 	skb->sk = save_sk;
-- 
2.20.0