Re: [net v3] drivers/net/wan/lapbether: Use needed_headroom instead of hard_header_len

[Martin Schiller]

On 2020-08-04 21:20, Xie He wrote:
> On Tue, Aug 4, 2020 at 5:43 AM Martin Schiller <ms@xxxxxxxxxx> wrote:
> > I'm not an expert in the field, but after reading the commit message 
> > and
> > the previous comments, I'd say that makes sense.
>
> Thanks!
>
> > Shouldn't this kernel panic be intercepted by a skb_cow() before the
> > skb_push() in lapbeth_data_transmit()?
>
> When a skb is passing down a protocol stack for transmission, there
> might be several different skb_push calls to prepend different
> headers. It would be the best (in terms of performance) if we can
> allocate the needed header space in advance, so that we don't need to
> reallocate the skb every time a new header needs to be prepended.

Yes, I agree.

> Adding skb_cow before these skb_push calls would indeed help
> preventing kernel panics, but that might not be the essential issue
> here, and it might also prevent us from discovering the real issue. (I
> guess this is also the reason skb_cow is not included in skb_push
> itself.)

Well, you are right that the panic is "useful" to discover the real
problem. But on the other hand, if it is possible to prevent a panic, I
think we should do so. Maybe with adding a warning, when skb_cow() needs
to reallocate memory.

But this is getting a little bit off topic. For this patch I can say:

LGTM.

Reviewed-by: Martin Schiller <ms@xxxxxxxxxx>