Re: [PATCH net-next v10 7/7] r8169: support setting rx queue numbers via ethtool
Jakub Kicinski <[email protected]>
| Newsgroups | org.kernel.vger.netdev,org.kernel.vger.linux-kernel |
|---|---|
| Message-ID | <[email protected]> |
On Mon, 3 Aug 2026 10:13:05 +0800 javen wrote:
> +static int rtl8169_set_channels(struct net_device *dev,
> + struct ethtool_channels *ch)
> +{
> + struct rtl8169_private *tp = netdev_priv(dev);
> + bool if_running = netif_running(dev);
> + enum rx_desc_type old_rx_desc_type;
> + enum rx_desc_type new_desc_type;
> + struct rtl8169_rx_ring *new_rx;
> + int i, ret;
This patch is missing a lot of safeties, AFAICT.
We require that the driver is resilient to memory allocation failures -
if the memory allocations for new rings fails you should restore old
rings. Calling rtl8169_up(tp) and hoping for the best is not going to
fly.
Let's leave the live reconfig to a separate series, in the interest
of making progress. Please return EBUSY if the netif_running() and
you can tackle live reconfig separately?