Re: [PATCH net-next v5 2/4] net: dsa: mxl862xx: move phylink stubs to mxl862xx-phylink.c

[Simon Horman]

This is an AI-generated review of your patch. The human sending this
email has considered the AI review valid, or at least plausible.
Full review at: https://sashiko.dev
---
net: dsa: mxl862xx: move phylink stubs to mxl862xx-phylink.c

Move the phylink MAC operations and get_caps callback from mxl862xx.c
into a dedicated mxl862xx-phylink.c file. This prepares for the SerDes
PCS implementation which adds substantial phylink/PCS code.

> diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.c b/drivers/net/dsa/mxl862xx/mxl862xx.c
> index 2f22adedfbf67..a193f3c07d359 100644
> --- a/drivers/net/dsa/mxl862xx/mxl862xx.c
> +++ b/drivers/net/dsa/mxl862xx/mxl862xx.c

[ ... ]

[Severity: Critical]
This is a pre-existing issue, but the crc_err_work can execute after
dsa_unregister_switch(), leading to a use-after-free of ds->dst and
dsa_port structures.

In mxl862xx_remove(), dsa_unregister_switch() is called before the
asynchronous crc_err_work is cancelled via mxl862xx_host_shutdown():

drivers/net/dsa/mxl862xx/mxl862xx.c:mxl862xx_remove() {
	...
	dsa_unregister_switch(ds);

	mxl862xx_host_shutdown(priv);
	...
}

dsa_unregister_switch() tears down the DSA tree, frees all dsa_port
objects, and drops the reference to ds->dst.

If mxl862xx_crc_err_work_fn executes concurrently or after
dsa_unregister_switch() has returned, it traverses already freed
dsa_port entries:

drivers/net/dsa/mxl862xx/mxl862xx-host.c:mxl862xx_crc_err_work_fn() {
	...
	dsa_switch_for_each_cpu_port(dp, priv->ds)
		dev_close(dp->conduit);
	...
}

Could this order in mxl862xx_remove() result in a use-after-free, and
should mxl862xx_host_shutdown() be called before dsa_unregister_switch()?