diff mbox series

[bpf,3/3] devmap: Add missing RCU read lock on flush

Message ID 20190614082015.23336-4-toshiaki.makita1@gmail.com
State Accepted
Delegated to: BPF Maintainers
Headers show
Series Devmap fixes around memory and RCU | expand

Commit Message

Toshiaki Makita June 14, 2019, 8:20 a.m. UTC
.ndo_xdp_xmit() assumes it is called under RCU. For example virtio_net
uses RCU to detect it has setup the resources for tx. The assumption
accidentally broke when introducing bulk queue in devmap.

Fixes: 5d053f9da431 ("bpf: devmap prepare xdp frames for bulking")
Reported-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Toshiaki Makita <toshiaki.makita1@gmail.com>
---
 kernel/bpf/devmap.c | 4 ++++
 1 file changed, 4 insertions(+)

Comments

Toke Høiland-Jørgensen June 14, 2019, 11:07 a.m. UTC | #1
Toshiaki Makita <toshiaki.makita1@gmail.com> writes:

> .ndo_xdp_xmit() assumes it is called under RCU. For example virtio_net
> uses RCU to detect it has setup the resources for tx. The assumption
> accidentally broke when introducing bulk queue in devmap.
>
> Fixes: 5d053f9da431 ("bpf: devmap prepare xdp frames for bulking")
> Reported-by: David Ahern <dsahern@gmail.com>
> Signed-off-by: Toshiaki Makita <toshiaki.makita1@gmail.com>
> ---

I think this is still needed, but the patch context is going to conflict
with the patch I linked above... I guess it's up to the maintainers to
decide which order to merge them in :)

-Toke
diff mbox series

Patch

diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
index a126d95..1defea4 100644
--- a/kernel/bpf/devmap.c
+++ b/kernel/bpf/devmap.c
@@ -282,6 +282,7 @@  void __dev_map_flush(struct bpf_map *map)
 	unsigned long *bitmap = this_cpu_ptr(dtab->flush_needed);
 	u32 bit;
 
+	rcu_read_lock();
 	for_each_set_bit(bit, bitmap, map->max_entries) {
 		struct bpf_dtab_netdev *dev = READ_ONCE(dtab->netdev_map[bit]);
 		struct xdp_bulk_queue *bq;
@@ -297,6 +298,7 @@  void __dev_map_flush(struct bpf_map *map)
 
 		__clear_bit(bit, bitmap);
 	}
+	rcu_read_unlock();
 }
 
 /* rcu_read_lock (from syscall and BPF contexts) ensures that if a delete and/or
@@ -389,6 +391,7 @@  static void dev_map_flush_old(struct bpf_dtab_netdev *dev)
 
 		int cpu;
 
+		rcu_read_lock();
 		for_each_online_cpu(cpu) {
 			bitmap = per_cpu_ptr(dev->dtab->flush_needed, cpu);
 			__clear_bit(dev->bit, bitmap);
@@ -396,6 +399,7 @@  static void dev_map_flush_old(struct bpf_dtab_netdev *dev)
 			bq = per_cpu_ptr(dev->bulkq, cpu);
 			bq_xmit_all(dev, bq, XDP_XMIT_FLUSH, false);
 		}
+		rcu_read_unlock();
 	}
 }