[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <d05e56d5-2677-3c03-7a25-df2ca2681a75@gmail.com>
Date: Mon, 13 May 2019 16:01:04 -0600
From: David Ahern <dsahern@...il.com>
To: mcmahon@...sta.com, davem@...emloft.net, roopa@...ulusnetworks.com,
christian@...uner.io, khlebnikov@...dex-team.ru,
lzgrablic@...sta.com, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org, mowat@...sta.com, dmia@...sta.com
Subject: Re: getneigh: add nondump to retrieve single entry
On 5/13/19 10:03 AM, mcmahon@...sta.com wrote:
> diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> index 30f6fd8f68e0..981f1568710b 100644
> --- a/net/core/neighbour.c
> +++ b/net/core/neighbour.c
> +static int neigh_find_fill(struct neigh_table *tbl, const void *pkey,
> + struct net_device *dev, struct sk_buff *skb, u32 pid,
> + u32 seq)
> +{
> + struct neighbour *neigh;
> + int key_len = tbl->key_len;
> + u32 hash_val;
> + struct neigh_hash_table *nht;
> + int err;
reverse xmas tree ordering
...
> +static int neigh_get(struct sk_buff *skb, struct nlmsghdr *nlh)
> +{
> + struct net *net = sock_net(skb->sk);
> + struct ndmsg *ndm;
> + struct nlattr *dst_attr;
> + struct neigh_table *tbl;
> + struct net_device *dev = NULL;
> +
> + ASSERT_RTNL();
> + if (nlmsg_len(nlh) < sizeof(*ndm))
> + return -EINVAL;
> +
> + dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
> + if (dst_attr == NULL)
> + return -EINVAL;
> +
> + ndm = nlmsg_data(nlh);
> + if (ndm->ndm_ifindex) {
> + dev = __dev_get_by_index(net, ndm->ndm_ifindex);
> + if (dev == NULL)
> + return -ENODEV;
> + }
> +
> + read_lock(&neigh_tbl_lock);
this patch is clearly for a MUCH older kernel than 5.2 (like 3.18
maybe?) as that lock no longer exists.
> + for (tbl = neigh_tables; tbl; tbl = tbl->next) {
> + struct sk_buff *nskb;
> + int err;
> +
> + if (tbl->family != ndm->ndm_family)
> + continue;
Use neigh_find_table.
You need to update the patch to top of net-next tree and re-work the
locking. Run tests with RCU and lock debugging enabled to make sure you
have it right.
Powered by blists - more mailing lists