[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20170617174244.132862-7-tracywwnj@gmail.com>
Date: Sat, 17 Jun 2017 10:42:29 -0700
From: Wei Wang <tracywwnj@...il.com>
To: David Miller <davem@...emloft.net>, netdev@...r.kernel.org
Cc: Eric Dumazet <edumazet@...gle.com>,
Martin KaFai Lau <kafai@...com>, Wei Wang <weiwan@...gle.com>
Subject: [PATCH v2 net-next 06/21] ipv4: take dst->__refcnt when caching dst in fib
From: Wei Wang <weiwan@...gle.com>
In IPv4 routing code, fib_nh and fib_nh_exception can hold pointers
to struct rtable but they never increment dst->__refcnt.
This leads to the need of the dst garbage collector because when user
is done with this dst and calls dst_release(), it can only decrement
dst->__refcnt and can not free the dst even it sees dst->__refcnt
drops from 1 to 0 (unless DST_NOCACHE flag is set) because the routing
code might still hold reference to it.
And when the routing code tries to delete a route, it has to put the
dst to the gc_list if dst->__refcnt is not yet 0 and have a gc thread
running periodically to check on dst->__refcnt and finally to free dst
when refcnt becomes 0.
This patch increments dst->__refcnt when
fib_nh/fib_nh_exception holds reference to this dst and properly release
the dst when fib_nh/fib_nh_exception has been updated with a new dst.
This patch is a preparation in order to fully get rid of dst gc later.
Signed-off-by: Wei Wang <weiwan@...gle.com>
Acked-by: Martin KaFai Lau <kafai@...com>
---
net/ipv4/fib_semantics.c | 5 ++++-
net/ipv4/route.c | 19 ++++++++++++++++---
2 files changed, 20 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index 2157dc08c407..53b3e9c2da4c 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -152,6 +152,7 @@ static void rt_fibinfo_free(struct rtable __rcu **rtp)
* free_fib_info_rcu()
*/
+ dst_release(&rt->dst);
dst_free(&rt->dst);
}
@@ -194,8 +195,10 @@ static void rt_fibinfo_free_cpus(struct rtable __rcu * __percpu *rtp)
struct rtable *rt;
rt = rcu_dereference_protected(*per_cpu_ptr(rtp, cpu), 1);
- if (rt)
+ if (rt) {
+ dst_release(&rt->dst);
dst_free(&rt->dst);
+ }
}
free_percpu(rtp);
}
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 0a843ef2b709..3dee0043117e 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -603,11 +603,13 @@ static void fnhe_flush_routes(struct fib_nh_exception *fnhe)
rt = rcu_dereference(fnhe->fnhe_rth_input);
if (rt) {
RCU_INIT_POINTER(fnhe->fnhe_rth_input, NULL);
+ dst_release(&rt->dst);
rt_free(rt);
}
rt = rcu_dereference(fnhe->fnhe_rth_output);
if (rt) {
RCU_INIT_POINTER(fnhe->fnhe_rth_output, NULL);
+ dst_release(&rt->dst);
rt_free(rt);
}
}
@@ -1332,9 +1334,12 @@ static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe,
rt->rt_gateway = daddr;
if (!(rt->dst.flags & DST_NOCACHE)) {
+ dst_hold(&rt->dst);
rcu_assign_pointer(*porig, rt);
- if (orig)
+ if (orig) {
+ dst_release(&orig->dst);
rt_free(orig);
+ }
ret = true;
}
@@ -1357,12 +1362,20 @@ static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt)
}
orig = *p;
+ /* hold dst before doing cmpxchg() to avoid race condition
+ * on this dst
+ */
+ dst_hold(&rt->dst);
prev = cmpxchg(p, orig, rt);
if (prev == orig) {
- if (orig)
+ if (orig) {
+ dst_release(&orig->dst);
rt_free(orig);
- } else
+ }
+ } else {
+ dst_release(&rt->dst);
ret = false;
+ }
return ret;
}
--
2.13.1.518.g3df882009-goog
Powered by blists - more mailing lists