]> git.ipfire.org Git - thirdparty/linux.git/commitdiff
ipv6: frags: cleanup __IP6_INC_STATS() confusion
authorEric Dumazet <edumazet@google.com>
Tue, 26 May 2026 14:55:29 +0000 (14:55 +0000)
committerJakub Kicinski <kuba@kernel.org>
Thu, 28 May 2026 00:27:26 +0000 (17:27 -0700)
After commits e1ae5c2ea478 ("vrf: Increment Icmp6InMsgs on the original
netdev") and bdb7cc643fc9 ("ipv6: Count interface receive statistics
on the ingress netdev") net/ipv6/reassembly.c uses three different
ways to reach idev in various __IP6_INC_STATS() calls.

- ip6_dst_idev(skb_dst(skb))
- __in6_dev_get_safely(skb->dev)
- __in6_dev_stats_get(skb->dev)

Lets centralize this from ipv6_frag_rcv() and use __in6_dev_stats_get().

Note that ipv6_frag_rcv() tests if skb->dev could be NULL already, so
I chose to also guard against NULL, but we probably can remove the
tests in a followup patch, because I do not think skb->dev could be NULL.

iif = skb->dev ? skb->dev->ifindex : 0;

idev can be NULL, __IP6_INC_STATS() deals with this possibility.

Small code size reduction as a bonus.

$ scripts/bloat-o-meter -t vmlinux.old vmlinux.new
add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-145 (-145)
Function                                     old     new   delta
ipv6_frag_rcv                               2399    2362     -37
ip6_frag_reasm                               705     597    -108
Total: Before=31455552, After=31455407, chg -0.00%

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260526145529.3587126-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/ipv6/reassembly.c

index 11f9144bebbe2cf0152992c526fb6728f83f462d..05c51f669754a1a10e944bb7c4433565bcaacb40 100644 (file)
@@ -69,7 +69,7 @@ static struct inet_frags ip6_frags;
 
 static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *skb,
                          struct sk_buff *prev_tail, struct net_device *dev,
-                         int *refs);
+                         struct inet6_dev *idev, int *refs);
 
 static void ip6_frag_expire(struct timer_list *t)
 {
@@ -107,7 +107,8 @@ fq_find(struct net *net, __be32 id, const struct ipv6hdr *hdr, int iif)
 static int ip6_frag_queue(struct net *net,
                          struct frag_queue *fq, struct sk_buff *skb,
                          struct frag_hdr *fhdr, int nhoff,
-                         u32 *prob_offset, int *refs)
+                         u32 *prob_offset, int *refs,
+                         struct inet6_dev *idev)
 {
        int offset, end, fragsize;
        struct sk_buff *prev_tail;
@@ -133,8 +134,7 @@ static int ip6_frag_queue(struct net *net,
                 * we do not free it here.
                 */
                inet_frag_kill(&fq->q, refs);
-               __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
-                               IPSTATS_MIB_REASMFAILS);
+               __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMFAILS);
                return -1;
        }
 
@@ -167,8 +167,7 @@ static int ip6_frag_queue(struct net *net,
                         */
                        *prob_offset = offsetof(struct ipv6hdr, payload_len);
                        inet_frag_kill(&fq->q, refs);
-                       __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
-                                       IPSTATS_MIB_REASMFAILS);
+                       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMFAILS);
                        return -1;
                }
                if (end > fq->q.len) {
@@ -227,7 +226,7 @@ static int ip6_frag_queue(struct net *net,
                unsigned long orefdst = skb->_skb_refdst;
 
                skb->_skb_refdst = 0UL;
-               err = ip6_frag_reasm(fq, skb, prev_tail, dev, refs);
+               err = ip6_frag_reasm(fq, skb, prev_tail, dev, idev, refs);
                skb->_skb_refdst = orefdst;
                return err;
        }
@@ -242,12 +241,10 @@ insert_error:
                goto err;
        }
        err = -EINVAL;
-       __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
-                       IPSTATS_MIB_REASM_OVERLAPS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASM_OVERLAPS);
 discard_fq:
        inet_frag_kill(&fq->q, refs);
-       __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
-                       IPSTATS_MIB_REASMFAILS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMFAILS);
 err:
        kfree_skb_reason(skb, reason);
        return err;
@@ -262,7 +259,7 @@ err:
  */
 static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *skb,
                          struct sk_buff *prev_tail, struct net_device *dev,
-                         int *refs)
+                         struct inet6_dev *idev, int *refs)
 {
        struct net *net = fq->q.fqdir->net;
        unsigned int nhoff;
@@ -311,7 +308,7 @@ static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *skb,
        skb_postpush_rcsum(skb, skb_network_header(skb),
                           skb_network_header_len(skb));
 
-       __IP6_INC_STATS(net, __in6_dev_stats_get(dev, skb), IPSTATS_MIB_REASMOKS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMOKS);
        fq->q.rb_fragments = RB_ROOT;
        fq->q.fragments_tail = NULL;
        fq->q.last_run_head = NULL;
@@ -323,7 +320,7 @@ out_oversize:
 out_oom:
        net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
 out_fail:
-       __IP6_INC_STATS(net, __in6_dev_stats_get(dev, skb), IPSTATS_MIB_REASMFAILS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMFAILS);
        inet_frag_kill(&fq->q, refs);
        return -1;
 }
@@ -332,15 +329,18 @@ static int ipv6_frag_rcv(struct sk_buff *skb)
 {
        const struct ipv6hdr *hdr = ipv6_hdr(skb);
        struct net *net = skb_dst_dev_net(skb);
+       struct inet6_dev *idev;
        struct frag_hdr *fhdr;
        struct frag_queue *fq;
        u8 nexthdr;
        int iif;
 
+       idev = skb->dev ? __in6_dev_stats_get(skb->dev, skb) : NULL;
+
        if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED)
                goto fail_hdr;
 
-       __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMREQDS);
 
        /* Jumbo payload inhibits frag. header */
        if (hdr->payload_len == 0)
@@ -356,8 +356,7 @@ static int ipv6_frag_rcv(struct sk_buff *skb)
        if (!(fhdr->frag_off & htons(IP6_OFFSET | IP6_MF))) {
                /* It is not a fragmented frame */
                skb->transport_header += sizeof(struct frag_hdr);
-               __IP6_INC_STATS(net,
-                               ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
+               __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMOKS);
 
                IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
                IP6CB(skb)->flags |= IP6SKB_FRAGMENTED;
@@ -374,8 +373,7 @@ static int ipv6_frag_rcv(struct sk_buff *skb)
         */
        nexthdr = hdr->nexthdr;
        if (ipv6frag_thdr_truncated(skb, skb_network_offset(skb) + sizeof(struct ipv6hdr), &nexthdr)) {
-               __IP6_INC_STATS(net, __in6_dev_get_safely(skb->dev),
-                               IPSTATS_MIB_INHDRERRORS);
+               __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
                icmpv6_param_prob(skb, ICMPV6_HDR_INCOMP, 0);
                return -1;
        }
@@ -391,14 +389,13 @@ static int ipv6_frag_rcv(struct sk_buff *skb)
 
                fq->iif = iif;
                ret = ip6_frag_queue(net, fq, skb, fhdr, IP6CB(skb)->nhoff,
-                                    &prob_offset, &refs);
+                                    &prob_offset, &refs, idev);
 
                spin_unlock(&fq->q.lock);
                rcu_read_unlock();
                inet_frag_putn(&fq->q, refs);
                if (prob_offset) {
-                       __IP6_INC_STATS(net, __in6_dev_get_safely(skb->dev),
-                                       IPSTATS_MIB_INHDRERRORS);
+                       __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
                        /* icmpv6_param_prob() calls kfree_skb(skb) */
                        icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, prob_offset);
                }
@@ -406,13 +403,12 @@ static int ipv6_frag_rcv(struct sk_buff *skb)
        }
        rcu_read_unlock();
 
-       __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_REASMFAILS);
        kfree_skb(skb);
        return -1;
 
 fail_hdr:
-       __IP6_INC_STATS(net, __in6_dev_get_safely(skb->dev),
-                       IPSTATS_MIB_INHDRERRORS);
+       __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
        icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
        return -1;
 }