ipv6: prevent fib6_run_gc() contention

On a high-traffic router with many processors and many IPv6 dst
entries, soft lockup in fib6_run_gc() can occur when number of
entries reaches gc_thresh.

This happens because fib6_run_gc() uses fib6_gc_lock to allow
only one thread to run the garbage collector but ip6_dst_gc()
doesn't update net->ipv6.ip6_rt_last_gc until fib6_run_gc()
returns. On a system with many entries, this can take some time
so that in the meantime, other threads pass the tests in
ip6_dst_gc() (ip6_rt_last_gc is still not updated) and wait for
the lock. They then have to run the garbage collector one after
another which blocks them for quite long.

Resolve this by replacing special value ~0UL of expire parameter
to fib6_run_gc() by explicit "force" parameter to choose between
spin_lock_bh() and spin_trylock_bh() and call fib6_run_gc() with
force=false if gc_thresh is reached but not max_size.

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 12 insertions, 15 deletions

diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index 5fc9c7a68d8d..d872553ca933 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c
@@ -1632,19 +1632,16 @@ static int fib6_age(struct rt6_info *rt, void *arg)
 
 static DEFINE_SPINLOCK(fib6_gc_lock);
 
-void fib6_run_gc(unsigned long expires, struct net *net)
+void fib6_run_gc(unsigned long expires, struct net *net, bool force)
 {
-        if (expires != ~0UL) {
+        if (force) {
                 spin_lock_bh(&fib6_gc_lock);
-                gc_args.timeout = expires ? (int)expires :
-                        net->ipv6.sysctl.ip6_rt_gc_interval;
-        } else {
-                if (!spin_trylock_bh(&fib6_gc_lock)) {
-                        mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
-                        return;
-                }
-                gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval;
+        } else if (!spin_trylock_bh(&fib6_gc_lock)) {
+                mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
+                return;
         }
+        gc_args.timeout = expires ? (int)expires :
+                          net->ipv6.sysctl.ip6_rt_gc_interval;
 
         gc_args.more = icmp6_dst_gc();
 
@@ -1661,7 +1658,7 @@ void fib6_run_gc(unsigned long expires, struct net *net)
 
 static void fib6_gc_timer_cb(unsigned long arg)
 {
-        fib6_run_gc(0, (struct net *)arg);
+        fib6_run_gc(0, (struct net *)arg, true);
 }
 
 static int __net_init fib6_net_init(struct net *net)
diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
index 24c03396e008..79aa9652ed86 100644
--- a/net/ipv6/ndisc.c
+++ b/net/ipv6/ndisc.c
@@ -1576,7 +1576,7 @@ static int ndisc_netdev_event(struct notifier_block *this, unsigned long event,
         switch (event) {
         case NETDEV_CHANGEADDR:
                 neigh_changeaddr(&nd_tbl, dev);
-                fib6_run_gc(~0UL, net);
+                fib6_run_gc(0, net, false);
                 idev = in6_dev_get(dev);
                 if (!idev)
                         break;
@@ -1586,7 +1586,7 @@ static int ndisc_netdev_event(struct notifier_block *this, unsigned long event,
                 break;
         case NETDEV_DOWN:
                 neigh_ifdown(&nd_tbl, dev);
-                fib6_run_gc(~0UL, net);
+                fib6_run_gc(0, net, false);
                 break;
         case NETDEV_NOTIFY_PEERS:
                 ndisc_send_unsol_na(dev);
diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index a8c891aa2464..824c424f9648 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -1326,7 +1326,7 @@ static int ip6_dst_gc(struct dst_ops *ops)
                 goto out;
 
         net->ipv6.ip6_rt_gc_expire++;
-        fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
+        fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
         net->ipv6.ip6_rt_last_gc = now;
         entries = dst_entries_get_slow(ops);
         if (entries < ops->gc_thresh)
@@ -2827,7 +2827,7 @@ int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
         net = (struct net *)ctl->extra1;
         delay = net->ipv6.sysctl.flush_delay;
         proc_dointvec(ctl, write, buffer, lenp, ppos);
-        fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
+        fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
         return 0;
 }