[PATCH net-next] net: drop_monitor: support drop reason

[PATCH net-next] net: drop_monitor: support drop reason

[menglong8.dong]

From: Menglong Dong <imagedong@tencent.com>

In the commit c504e5c2f964 ("net: skb: introduce kfree_skb_reason()")
drop reason is introduced to the tracepoint of kfree_skb. Therefore,
drop_monitor is able to report the drop reason to users by netlink.

For now, the number of drop reason is passed to users ( seems it's
a little troublesome to pass the drop reason as string ). Therefore,
users can do some customized description of the reason.

Signed-off-by: Menglong Dong <imagedong@tencent.com>
---
 include/uapi/linux/net_dropmon.h | 1 +
 net/core/drop_monitor.c          | 6 ++++++
 2 files changed, 7 insertions(+)

diff --git a/include/uapi/linux/net_dropmon.h b/include/uapi/linux/net_dropmon.h
index 66048cc5d7b3..b2815166dbc2 100644
--- a/include/uapi/linux/net_dropmon.h
+++ b/include/uapi/linux/net_dropmon.h
@@ -93,6 +93,7 @@ enum net_dm_attr {
 	NET_DM_ATTR_SW_DROPS,			/* flag */
 	NET_DM_ATTR_HW_DROPS,			/* flag */
 	NET_DM_ATTR_FLOW_ACTION_COOKIE,		/* binary */
+	NET_DM_ATTR_REASON,			/* u32 */
 
 	__NET_DM_ATTR_MAX,
 	NET_DM_ATTR_MAX = __NET_DM_ATTR_MAX - 1
diff --git a/net/core/drop_monitor.c b/net/core/drop_monitor.c
index 7b288a121a41..4aed0e583770 100644
--- a/net/core/drop_monitor.c
+++ b/net/core/drop_monitor.c
@@ -126,6 +126,7 @@ struct net_dm_skb_cb {
 		struct devlink_trap_metadata *hw_metadata;
 		void *pc;
 	};
+	enum skb_drop_reason reason;
 };
 
 #define NET_DM_SKB_CB(__skb) ((struct net_dm_skb_cb *)&((__skb)->cb[0]))
@@ -508,6 +509,7 @@ static void net_dm_packet_trace_kfree_skb_hit(void *ignore,
 	if (!nskb)
 		return;
 
+	NET_DM_SKB_CB(nskb)->reason = reason;
 	NET_DM_SKB_CB(nskb)->pc = location;
 	/* Override the timestamp because we care about the time when the
 	 * packet was dropped.
@@ -606,6 +608,7 @@ static int net_dm_packet_report_in_port_put(struct sk_buff *msg, int ifindex,
 static int net_dm_packet_report_fill(struct sk_buff *msg, struct sk_buff *skb,
 				     size_t payload_len)
 {
+	enum skb_drop_reason reason = NET_DM_SKB_CB(skb)->reason;
 	u64 pc = (u64)(uintptr_t) NET_DM_SKB_CB(skb)->pc;
 	char buf[NET_DM_MAX_SYMBOL_LEN];
 	struct nlattr *attr;
@@ -623,6 +626,9 @@ static int net_dm_packet_report_fill(struct sk_buff *msg, struct sk_buff *skb,
 	if (nla_put_u64_64bit(msg, NET_DM_ATTR_PC, pc, NET_DM_ATTR_PAD))
 		goto nla_put_failure;
 
+	if (nla_put_u32(msg, NET_DM_ATTR_REASON, reason))
+		goto nla_put_failure;
+
 	snprintf(buf, sizeof(buf), "%pS", NET_DM_SKB_CB(skb)->pc);
 	if (nla_put_string(msg, NET_DM_ATTR_SYMBOL, buf))
 		goto nla_put_failure;
-- 
2.27.0

Re: [PATCH net-next] net: drop_monitor: support drop reason

[Jakub Kicinski]

On Mon, 24 Jan 2022 15:59:55 +0800 menglong8.dong@gmail.com wrote:
> @@ -606,6 +608,7 @@ static int net_dm_packet_report_in_port_put(struct sk_buff *msg, int ifindex,
>  static int net_dm_packet_report_fill(struct sk_buff *msg, struct sk_buff *skb,
>  				     size_t payload_len)
>  {
> +	enum skb_drop_reason reason = NET_DM_SKB_CB(skb)->reason;
>  	u64 pc = (u64)(uintptr_t) NET_DM_SKB_CB(skb)->pc;

nit: maybe it's better to get a pointer to struct net_dm_skb_cb here
instead of local var for each field?

>  	char buf[NET_DM_MAX_SYMBOL_LEN];
>  	struct nlattr *attr;
> @@ -623,6 +626,9 @@ static int net_dm_packet_report_fill(struct sk_buff *msg, struct sk_buff *skb,
>  	if (nla_put_u64_64bit(msg, NET_DM_ATTR_PC, pc, NET_DM_ATTR_PAD))
>  		goto nla_put_failure;
>  
> +	if (nla_put_u32(msg, NET_DM_ATTR_REASON, reason))
> +		goto nla_put_failure;

Re: [PATCH net-next] net: drop_monitor: support drop reason

[Menglong Dong]

On Wed, Jan 26, 2022 at 11:38 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Mon, 24 Jan 2022 15:59:55 +0800 menglong8.dong@gmail.com wrote:
> > @@ -606,6 +608,7 @@ static int net_dm_packet_report_in_port_put(struct sk_buff *msg, int ifindex,
> >  static int net_dm_packet_report_fill(struct sk_buff *msg, struct sk_buff *skb,
> >                                    size_t payload_len)
> >  {
> > +     enum skb_drop_reason reason = NET_DM_SKB_CB(skb)->reason;
> >       u64 pc = (u64)(uintptr_t) NET_DM_SKB_CB(skb)->pc;
>
> nit: maybe it's better to get a pointer to struct net_dm_skb_cb here
> instead of local var for each field?

Yeah, I think it's a good idea :/

>
> >       char buf[NET_DM_MAX_SYMBOL_LEN];
> >       struct nlattr *attr;
> > @@ -623,6 +626,9 @@ static int net_dm_packet_report_fill(struct sk_buff *msg, struct sk_buff *skb,
> >       if (nla_put_u64_64bit(msg, NET_DM_ATTR_PC, pc, NET_DM_ATTR_PAD))
> >               goto nla_put_failure;
> >
> > +     if (nla_put_u32(msg, NET_DM_ATTR_REASON, reason))
> > +             goto nla_put_failure;