David Miller wrote:

From: Peter Zijlstra <redacted>

quoted

Hmm, what does sk_buff::input_dev do? That seems to store the initial
device?

You can run grep on the tree just as easily as I can which is what I
did to answer this question.  It only takes a few seconds of your
time to grep the source tree for things like "skb->input_dev", so
would you please do that before asking more questions like this?

It does store the initial device, but as Thomas tried so hard to
explain to you guys these device pointers in the skb are transient and
you cannot refer to them outside of packet receive processing.

Thomas did a great job of explaining and without any flaming or ad
hominem attacks.

We have now formed a decent plan for doing the accounting in a stable
way without adding new fields to sk_buff, thankyou for the catch.

The reason is that there is no refcounting performed on these devices
when they are attached to the skb, for performance reasons, and thus
the device can be downed, the module for it removed, etc. long before
the skb is freed up.

The virtual block device can refcount the network device on virtual
device create and un-refcount on virtual device delete.  We need to
add that to the core patch and maybe package it nicely so the memalloc
reserve/unreserve happens at the same time, in a tidy little library
function to share with other virtual devices like iSCSI that also need
some anti-deadlock lovin.

Regards,

Daniel