LVS
lvs-devel
Google
 
Web LinuxVirtualServer.org

Re: [PATCH] ipvs: Use flexible array for MH lookup table

To: Rosen Penev <rosenp@xxxxxxxxx>
Subject: Re: [PATCH] ipvs: Use flexible array for MH lookup table
Cc: netfilter-devel@xxxxxxxxxxxxxxx, Simon Horman <horms@xxxxxxxxxxxx>, Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx>, Florian Westphal <fw@xxxxxxxxx>, Phil Sutter <phil@xxxxxx>, "David S. Miller" <davem@xxxxxxxxxxxxx>, Eric Dumazet <edumazet@xxxxxxxxxx>, Jakub Kicinski <kuba@xxxxxxxxxx>, Paolo Abeni <pabeni@xxxxxxxxxx>, "open list:IPVS" <netdev@xxxxxxxxxxxxxxx>, "open list:IPVS" <lvs-devel@xxxxxxxxxxxxxxx>, "open list:NETFILTER" <coreteam@xxxxxxxxxxxxx>, open list <linux-kernel@xxxxxxxxxxxxxxx>
From: Julian Anastasov <ja@xxxxxx>
Date: Fri, 22 May 2026 20:12:31 +0300 (EEST)
        Hello,

On Mon, 18 May 2026, Rosen Penev wrote:

> Store the Maglev hash lookup table in the scheduler state
> allocation instead of allocating it separately.
> 
> This keeps the lookup table tied to the RCU-freed state lifetime and
> simplifies the allocation and cleanup paths.
> 
> Assisted-by: Codex:GPT-5.5
> Signed-off-by: Rosen Penev <rosenp@xxxxxxxxx>

        Thanks for the efforts but it is better to keep the table 
allocated separately. As Sashiko [1] points out, the table size is
properly selected and we risk double-sized allocations if we apply
this patch.

        As for the problem in ip_vs_edit_service(), I have to
provide separate bugfix...

[1]:
https://sashiko.dev/#/patchset/20260519015506.634185-1-rosenp%40gmail.com

Regards

--
Julian Anastasov <ja@xxxxxx>



<Prev in Thread] Current Thread [Next in Thread>