From: Yajun Deng <yajun.deng@xxxxxxxxx>
Date: Thu, 14 Sep 2023 10:37:18 +0800
Although there is a kfree_skb_reason() helper function that can be used toI'd like to hear some arguments against having them static inlines + one
find the reason why this skb is dropped, but most callers didn't increase
one of rx_dropped, tx_dropped, rx_nohandler and rx_otherhost_dropped.
For the users, people are more concerned about why the dropped in ip
is increasing. So we can export dev_core_stats_rx_dropped_inc sets,
which users would trace them know why rx_dropped is increasing.
Export dev_core_stats_{rx_dropped, tx_dropped, rx_nohandler,
rx_otherhost_dropped}_inc for trace. Also, move dev_core_stats()
and netdev_core_stats_alloc() to dev.c, as they are not called
externally.
external that I proposed earlier.
Okay.Signed-off-by: Yajun Deng <yajun.deng@xxxxxxxxx>...as it's not at least mentioned here in the changelog.
---
v3: __cold should be added to the netdev_core_stats_alloc().
v2: use __cold instead of inline in dev_core_stats().
v1: https://lore.kernel.org/netdev/20230911082016.3694700-1-yajun.deng@xxxxxxxxx/
[...]
diff --git a/net/core/dev.c b/net/core/dev.cThis is way over 79/80 chars, break the line before "netdev_".
index ccff2b6ef958..98592e4c1df0 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -10475,7 +10475,7 @@ void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
}
EXPORT_SYMBOL(netdev_stats_to_stats64);
-struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev)
+static __cold struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev)
Okay, I'll unify with the "netdev_" prefix.
{Same for the line length.
struct net_device_core_stats __percpu *p;
@@ -10488,7 +10488,35 @@ struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device
/* This READ_ONCE() pairs with the cmpxchg() above */
return READ_ONCE(dev->core_stats);
}
-EXPORT_SYMBOL(netdev_core_stats_alloc);
+
+static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
Also notice that now some of the functions you touch have "dev_" prefix,
others have "netdev_", I'd probably take a look into unifying this.
(note for the maintainers that it would be probably better to leave
explicit "inline" here, but no bloat-o-meter was provided by the
author, so I can't say anything for sure)
+{[...]
+ /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
+ struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
+
+ if (likely(p))
+ return p;
+
+ return netdev_core_stats_alloc(dev);
+}
Thanks,
Olek