[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250630111836.GE41770@horms.kernel.org>
Date: Mon, 30 Jun 2025 12:18:36 +0100
From: Simon Horman <horms@...nel.org>
To: Alok Tiwari <alok.a.tiwari@...cle.com>
Cc: sgoutham@...vell.com, andrew+netdev@...n.ch, davem@...emloft.net,
edumazet@...gle.com, kuba@...nel.org, pabeni@...hat.com,
netdev@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH net v2] net: thunderx: avoid direct MTU assignment after
WRITE_ONCE()
On Sat, Jun 28, 2025 at 10:15:37PM -0700, Alok Tiwari wrote:
...
> @@ -1589,16 +1588,18 @@ static int nicvf_change_mtu(struct net_device *netdev, int new_mtu)
> return -EINVAL;
> }
>
> - WRITE_ONCE(netdev->mtu, new_mtu);
>
> - if (!netif_running(netdev))
> + if (!netif_running(netdev)) {
> + WRITE_ONCE(netdev->mtu, new_mtu);
> return 0;
> + }
>
> if (nicvf_update_hw_max_frs(nic, new_mtu)) {
> - netdev->mtu = orig_mtu;
> return -EINVAL;
> }
nit: curly brackets should be removed here, but see further comment below.
>
> + WRITE_ONCE(netdev->mtu, new_mtu);
> +
> return 0;
> }
Could this be more succinctly expressed as follows?
(Completely untested!)
if (netif_running(netdev) && nicvf_update_hw_max_frs(nic, new_mtu))
return -ENIVAL;
WRITE_ONCE(netdev->mtu, new_mtu);
return 0;
Powered by blists - more mailing lists