[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1477949046.7065.320.camel@edumazet-glaptop3.roam.corp.google.com>
Date: Mon, 31 Oct 2016 14:24:06 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: idosch@...sch.org
Cc: netdev@...r.kernel.org, davem@...emloft.net, jiri@...lanox.com,
mlxsw@...lanox.com, roopa@...ulusnetworks.com,
dsa@...ulusnetworks.com, nikolay@...ulusnetworks.com,
andy@...yhouse.net, vivien.didelot@...oirfairelinux.com,
andrew@...n.ch, f.fainelli@...il.com, alexander.h.duyck@...el.com,
kuznet@....inr.ac.ru, jmorris@...ei.org, yoshfuji@...ux-ipv6.org,
kaber@...sh.net, Ido Schimmel <idosch@...lanox.com>
Subject: Re: [PATCH net-next v2] ipv4: fib: Replay events when registering
FIB notifier
On Mon, 2016-10-31 at 23:13 +0200, idosch@...sch.org wrote:
> From: Ido Schimmel <idosch@...lanox.com>
>
> When registering a FIB notifier block we should notify the caller of
> already existing FIB entries and rules, as it would otherwise have an
> incomplete view of the FIB tables. This is especially important for
> switchdev drivers that support FIB offloads. Failing to notify them of
> existing entries may lead to packet loss.
>
> Upon registration, walk the leafs of all the routing tables and for each
> leaf send notification of existing FIB aliases. Similarly, when
> unregistering the notifier synthesize a deletion event, thereby
> relieving potential callers from the need to perform cleanup.
>
> The above is consistent with the netdevice notification chain, where
> "registration and up events are replayed to the new notifier" upon
> registration.
>
> Signed-off-by: Ido Schimmel <idosch@...lanox.com>
> Reviewed-by: Jiri Pirko <jiri@...lanox.com>
How well will this work for large FIB tables ?
Holding rtnl while sending thousands of skb will prevent consumers to
make progress ?
Powered by blists - more mailing lists