netdev - Re: [net-next,v22,4/7] net: mtip: Add net_device

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20260203110341.1ce21455@wsk>
Date: Tue, 3 Feb 2026 11:03:41 +0100
From: Łukasz Majewski <lukasz.majewski@...lbox.org>
To: Jakub Kicinski <kuba@...nel.org>
Cc: andrew@...n.ch, shawnguo@...nel.org, krzk+dt@...nel.org,
 linux-kernel@...r.kernel.org, edumazet@...gle.com, netdev@...r.kernel.org,
 pabeni@...hat.com, andrew+netdev@...n.ch, davem@...emloft.net,
 conor+dt@...nel.org, horms@...nel.org, richardcochran@...il.com,
 robh@...nel.org, imx@...ts.linux.dev, linux-arm-kernel@...ts.infradead.org,
 devicetree@...r.kernel.org, wahrenst@....net, s.hauer@...gutronix.de,
 kernel@...gutronix.de, festevam@...il.com
Subject: Re: [net-next,v22,4/7] net: mtip: Add net_device_ops functions to
 the L2 switch driver

Hi Jakub,

> This is an AI-generated review of your patch. The human sending this
> email has considered the AI review valid, or at least plausible.
> 
> For local reproduction steps see
> https://netdev-ai.bots.linux.dev/ai-local.html ---
> net: mtip: Add net_device_ops functions to the L2 switch driver
> 
> This patch adds net_device_ops callbacks for the MTIP L2 switch
> driver, including transmit, multicast configuration, MAC address
> handling, and timeout recovery functions.
> 
> > diff --git a/drivers/net/ethernet/freescale/mtipsw/mtipl2sw.c
> > b/drivers/net/ethernet/freescale/mtipsw/mtipl2sw.c ---
> > a/drivers/net/ethernet/freescale/mtipsw/mtipl2sw.c +++
> > b/drivers/net/ethernet/freescale/mtipsw/mtipl2sw.c  
> 
> [ ... ]
> 
> > +static netdev_tx_t mtip_start_xmit_port(struct sk_buff *skb,
> > +					struct net_device *dev,
> > int port) +{
> > +	struct mtip_ndev_priv *priv = netdev_priv(dev);
> > +	struct switch_enet_private *fep = priv->fep;
> > +	unsigned short status;
> > +	unsigned int index;
> > +	struct cbd_t *bdp;
> > +	void *bufaddr;
> > +
> > +	spin_lock_bh(&fep->hw_lock);  
> 
> [ ... ]
> 
> > +	/* Save skb pointer. */
> > +	fep->tx_skbuff[index] = skb;  
> 
> mtip_start_xmit_port() stores sk_buffs in fep->tx_skbuff[index] and
> creates DMA mappings, but mtip_switch_tx() which should complete TX
> operations is empty:
> 
> static void mtip_switch_tx(struct switch_enet_private *fep)
> {
> }
> 
> The NAPI poll handler mtip_rx_napi() calls mtip_switch_tx() to reclaim
> transmitted buffers. Since it does nothing, won't every transmitted
> packet leak the skb and DMA mapping? Also, dirty_tx is never
> advanced, so the TX ring will fill (cur_tx == dirty_tx) and stop
> accepting packets.
> 
> jakub: maybe add a TODO comment transiently here, since later patch
> implements the cleanup?

The mtip_switch_tx() is empty as we have agreed, that I will divide
this driver to several patches to ease the review.

Adding TODO seems to only make AI review happy, as:

- The patch which adds support for FEC_MTIP_L2SW in Kconfig will be
  sent after the driver is accepted to net-next

- Those commits are even now bisectable when FEC_MTIP_L2SW is
  enabled (when I test the setup).

Anyway, if you still would like to have the TODO comment, then please
give me a hint how it shall be written to make the AI happy...



And maybe a few my thoughts:

1. AI review seems to bring each time different issues - even the
"grammatic" ones were not provided with the first AI generated review.

2. I have tried to setup claudie> to run the patch set through it -
however, it requires a paid account on a cloud/AI vendor (and I guess
that different vendors' AI engines produce different output for the same
"AI prompt")?


Anyway, I do appreciate the AI review - it provides very deep insights
through the code.

> 
> > +static void mtip_timeout(struct net_device *dev, unsigned int
> > txqueue) +{
> > +	struct mtip_ndev_priv *priv = netdev_priv(dev);
> > +
> > +	dev->stats.tx_errors++;  
> 
> Jakub Kicinski noted in v16 review that tx_errors are already counted
> by the network stack, and the statistic is exposed per-queue in
> sysfs. Does incrementing tx_errors here double-count timeout events?
> 
> https://lore.kernel.org/all/20250725151618.0bc84bdb@kernel.org/
> 

Yes, this shall been removed.

> > +	DO_ONCE(mtip_print_hw_state, dev);
> > +
> > +	schedule_work(&priv->tx_timeout_work);
> > +}  



-- 
Best regards,

Łukasz Majewski