Merge branch 'inet-add-drop-monitor-support'

Eric Dumazet says:

====================
inet: add drop monitor support

I recently tried to analyse flakes in ip_defrag selftest.
This failed miserably.

IPv4 and IPv6 reassembly units are causing false kfree_skb()
notifications. It is time to deal with this issue.

First two patches are changing core networking to better
deal with eventual skb frag_list chains, in respect
of kfree_skb/consume_skb status.

Last three patches are adding three new drop reasons,
and make sure skbs that have been reassembled into
a large datagram are no longer viewed as dropped ones.

After this, understanding why ip_defrag selftest is flaky
is possible using standard drop monitoring tools.
====================

Link: https://lore.kernel.org/r/20221029154520.2747444-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

1 files changed, 17 insertions, 13 deletions

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 1d84a17eada5..42a35b59fb1e 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -94,6 +94,7 @@ EXPORT_SYMBOL(sysctl_max_skb_frags);
 #undef FN
 #define FN(reason) [SKB_DROP_REASON_##reason] = #reason,
 const char * const drop_reasons[] = {
+	[SKB_CONSUMED] = "CONSUMED",
 	DEFINE_DROP_REASON(FN, FN)
 };
 EXPORT_SYMBOL(drop_reasons);
@@ -768,7 +769,7 @@ static void skb_free_head(struct sk_buff *skb)
 	}
 }
 
-static void skb_release_data(struct sk_buff *skb)
+static void skb_release_data(struct sk_buff *skb, enum skb_drop_reason reason)
 {
 	struct skb_shared_info *shinfo = skb_shinfo(skb);
 	int i;
@@ -791,7 +792,7 @@ static void skb_release_data(struct sk_buff *skb)
 
 free_head:
 	if (shinfo->frag_list)
-		kfree_skb_list(shinfo->frag_list);
+		kfree_skb_list_reason(shinfo->frag_list, reason);
 
 	skb_free_head(skb);
 exit:
@@ -854,11 +855,11 @@ void skb_release_head_state(struct sk_buff *skb)
 }
 
 /* Free everything but the sk_buff shell. */
-static void skb_release_all(struct sk_buff *skb)
+static void skb_release_all(struct sk_buff *skb, enum skb_drop_reason reason)
 {
 	skb_release_head_state(skb);
 	if (likely(skb->head))
-		skb_release_data(skb);
+		skb_release_data(skb, reason);
 }
 
 /**
@@ -872,7 +873,7 @@ static void skb_release_all(struct sk_buff *skb)
 
 void __kfree_skb(struct sk_buff *skb)
 {
-	skb_release_all(skb);
+	skb_release_all(skb, SKB_DROP_REASON_NOT_SPECIFIED);
 	kfree_skbmem(skb);
 }
 EXPORT_SYMBOL(__kfree_skb);
@@ -894,7 +895,10 @@ kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason)
 
 	DEBUG_NET_WARN_ON_ONCE(reason <= 0 || reason >= SKB_DROP_REASON_MAX);
 
-	trace_kfree_skb(skb, __builtin_return_address(0), reason);
+	if (reason == SKB_CONSUMED)
+		trace_consume_skb(skb);
+	else
+		trace_kfree_skb(skb, __builtin_return_address(0), reason);
 	__kfree_skb(skb);
 }
 EXPORT_SYMBOL(kfree_skb_reason);
@@ -1052,7 +1056,7 @@ EXPORT_SYMBOL(consume_skb);
 void __consume_stateless_skb(struct sk_buff *skb)
 {
 	trace_consume_skb(skb);
-	skb_release_data(skb);
+	skb_release_data(skb, SKB_CONSUMED);
 	kfree_skbmem(skb);
 }
 
@@ -1077,7 +1081,7 @@ static void napi_skb_cache_put(struct sk_buff *skb)
 
 void __kfree_skb_defer(struct sk_buff *skb)
 {
-	skb_release_all(skb);
+	skb_release_all(skb, SKB_DROP_REASON_NOT_SPECIFIED);
 	napi_skb_cache_put(skb);
 }
 
@@ -1115,7 +1119,7 @@ void napi_consume_skb(struct sk_buff *skb, int budget)
 		return;
 	}
 
-	skb_release_all(skb);
+	skb_release_all(skb, SKB_CONSUMED);
 	napi_skb_cache_put(skb);
 }
 EXPORT_SYMBOL(napi_consume_skb);
@@ -1246,7 +1250,7 @@ EXPORT_SYMBOL_GPL(alloc_skb_for_msg);
  */
 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src)
 {
-	skb_release_all(dst);
+	skb_release_all(dst, SKB_CONSUMED);
 	return __skb_clone(dst, src);
 }
 EXPORT_SYMBOL_GPL(skb_morph);
@@ -1869,7 +1873,7 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
 		if (skb_has_frag_list(skb))
 			skb_clone_fraglist(skb);
 
-		skb_release_data(skb);
+		skb_release_data(skb, SKB_CONSUMED);
 	} else {
 		skb_free_head(skb);
 	}
@@ -6209,7 +6213,7 @@ static int pskb_carve_inside_header(struct sk_buff *skb, const u32 off,
 			skb_frag_ref(skb, i);
 		if (skb_has_frag_list(skb))
 			skb_clone_fraglist(skb);
-		skb_release_data(skb);
+		skb_release_data(skb, SKB_CONSUMED);
 	} else {
 		/* we can reuse existing recount- all we did was
 		 * relocate values
@@ -6352,7 +6356,7 @@ static int pskb_carve_inside_nonlinear(struct sk_buff *skb, const u32 off,
 		kfree(data);
 		return -ENOMEM;
 	}
-	skb_release_data(skb);
+	skb_release_data(skb, SKB_CONSUMED);
 
 	skb->head = data;
 	skb->head_frag = 0;