Re: [PATCH] ethtool: don't touch the parent device of a net device being unregistered

[Andrew Lunn]

> >> That's not sufficient detail. ethnl_parse_header_dev_get() is under RCU
> >> and unregistration does an RCU sync after delisting the device. Also
> >> not sure you're distinguishing struct net_device and struct device.
> >> 
> >> How did you hit this issue? What are the net devices involved?
> >
> >I've provided additional details about the reproducer of this vulnerability to Jakub and to security@xxxxxxxxxx.
> 
> Hello! May I ask about the decision on this patch?
> 
> At patchwork.kernel.org, it is marked as "Changes Requested":
> <https://patchwork.kernel.org/project/netdevbpf/patch/20260322075917.254874-1-alex.popov@xxxxxxxxx/>
> 
> However, I don't have any instructions on what to change in it.

You need to add to the commit message to explain how it can happen.

My question said i _think_ devices are arranged in a tree, and release
happens from the leaves towards the root. So a parent should not be
release first. If i'm correct, this crash indicates a problem
somewhere else, not here. So we should fix that. Your commit message
needs to convince me the change is fixing the real problem, not
papering over the cracks.

    Andrew