]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
openvswitch: Use nested-BH locking for ovs_pcpu_storage
authorSebastian Andrzej Siewior <bigeasy@linutronix.de>
Mon, 12 May 2025 09:27:29 +0000 (11:27 +0200)
committerPaolo Abeni <pabeni@redhat.com>
Thu, 15 May 2025 13:23:31 +0000 (15:23 +0200)
ovs_pcpu_storage is a per-CPU variable and relies on disabled BH for its
locking. Without per-CPU locking in local_bh_disable() on PREEMPT_RT
this data structure requires explicit locking.
The data structure can be referenced recursive and there is a recursion
counter to avoid too many recursions.

Add a local_lock_t to the data structure and use
local_lock_nested_bh() for locking. Add an owner of the struct which is
the current task and acquire the lock only if the structure is not owned
by the current task.

Cc: Aaron Conole <aconole@redhat.com>
Cc: Eelco Chaudron <echaudro@redhat.com>
Cc: Ilya Maximets <i.maximets@ovn.org>
Cc: dev@openvswitch.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Aaron Conole <aconole@redhat.com>
Link: https://patch.msgid.link/20250512092736.229935-9-bigeasy@linutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
net/openvswitch/actions.c
net/openvswitch/datapath.c
net/openvswitch/datapath.h

index 7e4a8d41b9ed6fcdaa3262a4b50634a1d68db5d5..435725c27a557998d619c899cdfde6b44cec5f58 100644 (file)
 #include "flow_netlink.h"
 #include "openvswitch_trace.h"
 
-struct deferred_action {
-       struct sk_buff *skb;
-       const struct nlattr *actions;
-       int actions_len;
-
-       /* Store pkt_key clone when creating deferred action. */
-       struct sw_flow_key pkt_key;
-};
-
 #define MAX_L2_LEN     (VLAN_ETH_HLEN + 3 * MPLS_HLEN)
 struct ovs_frag_data {
        unsigned long dst;
@@ -64,28 +55,10 @@ struct ovs_frag_data {
 
 static DEFINE_PER_CPU(struct ovs_frag_data, ovs_frag_data_storage);
 
-#define DEFERRED_ACTION_FIFO_SIZE 10
-#define OVS_RECURSION_LIMIT 5
-#define OVS_DEFERRED_ACTION_THRESHOLD (OVS_RECURSION_LIMIT - 2)
-struct action_fifo {
-       int head;
-       int tail;
-       /* Deferred action fifo queue storage. */
-       struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE];
-};
-
-struct action_flow_keys {
-       struct sw_flow_key key[OVS_DEFERRED_ACTION_THRESHOLD];
-};
-
-struct ovs_pcpu_storage {
-       struct action_fifo action_fifos;
-       struct action_flow_keys flow_keys;
-       int exec_level;
+DEFINE_PER_CPU(struct ovs_pcpu_storage, ovs_pcpu_storage) = {
+       .bh_lock = INIT_LOCAL_LOCK(bh_lock),
 };
 
-static DEFINE_PER_CPU(struct ovs_pcpu_storage, ovs_pcpu_storage);
-
 /* Make a clone of the 'key', using the pre-allocated percpu 'flow_keys'
  * space. Return NULL if out of key spaces.
  */
index aaa6277bb49c2506f848b9c83c3b34e68ec55986..6a304ae2d959cdba771bbb2a1ac3e2a1503c975b 100644 (file)
@@ -244,11 +244,13 @@ void ovs_dp_detach_port(struct vport *p)
 /* Must be called with rcu_read_lock. */
 void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
 {
+       struct ovs_pcpu_storage *ovs_pcpu = this_cpu_ptr(&ovs_pcpu_storage);
        const struct vport *p = OVS_CB(skb)->input_vport;
        struct datapath *dp = p->dp;
        struct sw_flow *flow;
        struct sw_flow_actions *sf_acts;
        struct dp_stats_percpu *stats;
+       bool ovs_pcpu_locked = false;
        u64 *stats_counter;
        u32 n_mask_hit;
        u32 n_cache_hit;
@@ -290,10 +292,26 @@ void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
 
        ovs_flow_stats_update(flow, key->tp.flags, skb);
        sf_acts = rcu_dereference(flow->sf_acts);
+       /* This path can be invoked recursively: Use the current task to
+        * identify recursive invocation - the lock must be acquired only once.
+        * Even with disabled bottom halves this can be preempted on PREEMPT_RT.
+        * Limit the locking to RT to avoid assigning `owner' if it can be
+        * avoided.
+        */
+       if (IS_ENABLED(CONFIG_PREEMPT_RT) && ovs_pcpu->owner != current) {
+               local_lock_nested_bh(&ovs_pcpu_storage.bh_lock);
+               ovs_pcpu->owner = current;
+               ovs_pcpu_locked = true;
+       }
+
        error = ovs_execute_actions(dp, skb, sf_acts, key);
        if (unlikely(error))
                net_dbg_ratelimited("ovs: action execution error on datapath %s: %d\n",
                                    ovs_dp_name(dp), error);
+       if (ovs_pcpu_locked) {
+               ovs_pcpu->owner = NULL;
+               local_unlock_nested_bh(&ovs_pcpu_storage.bh_lock);
+       }
 
        stats_counter = &stats->n_hit;
 
@@ -671,7 +689,13 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
        sf_acts = rcu_dereference(flow->sf_acts);
 
        local_bh_disable();
+       local_lock_nested_bh(&ovs_pcpu_storage.bh_lock);
+       if (IS_ENABLED(CONFIG_PREEMPT_RT))
+               this_cpu_write(ovs_pcpu_storage.owner, current);
        err = ovs_execute_actions(dp, packet, sf_acts, &flow->key);
+       if (IS_ENABLED(CONFIG_PREEMPT_RT))
+               this_cpu_write(ovs_pcpu_storage.owner, NULL);
+       local_unlock_nested_bh(&ovs_pcpu_storage.bh_lock);
        local_bh_enable();
        rcu_read_unlock();
 
index a12640792605858ab41b3a1f0aa264f21760efec..4a665c3cfa9065f2dbc080f9fc5c83b84ac471c7 100644 (file)
@@ -173,6 +173,39 @@ struct ovs_net {
        bool xt_label;
 };
 
+struct deferred_action {
+       struct sk_buff *skb;
+       const struct nlattr *actions;
+       int actions_len;
+
+       /* Store pkt_key clone when creating deferred action. */
+       struct sw_flow_key pkt_key;
+};
+
+#define DEFERRED_ACTION_FIFO_SIZE 10
+#define OVS_RECURSION_LIMIT 5
+#define OVS_DEFERRED_ACTION_THRESHOLD (OVS_RECURSION_LIMIT - 2)
+
+struct action_fifo {
+       int head;
+       int tail;
+       /* Deferred action fifo queue storage. */
+       struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE];
+};
+
+struct action_flow_keys {
+       struct sw_flow_key key[OVS_DEFERRED_ACTION_THRESHOLD];
+};
+
+struct ovs_pcpu_storage {
+       struct action_fifo action_fifos;
+       struct action_flow_keys flow_keys;
+       int exec_level;
+       struct task_struct *owner;
+       local_lock_t bh_lock;
+};
+DECLARE_PER_CPU(struct ovs_pcpu_storage, ovs_pcpu_storage);
+
 /**
  * enum ovs_pkt_hash_types - hash info to include with a packet
  * to send to userspace.