netdev - Re: [PATCH net] qmi_wwan: Clone the skb when in pass-through mode

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: <20210615122604.1d68b37c@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Date:   Tue, 15 Jun 2021 12:26:04 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Bjørn Mork <bjorn@...k.no>
Cc:     Kristian Evensen <kristian.evensen@...il.com>,
        Network Development <netdev@...r.kernel.org>,
        subashab@...eaurora.org
Subject: Re: [PATCH net] qmi_wwan: Clone the skb when in pass-through mode

On Tue, 15 Jun 2021 15:39:14 +0200 Bjørn Mork wrote:
> >> I think this would be a really nice solution. The same (at least
> >> FLAG_MULTI_PACKET + usbnet_skb_return) could be applied to pass
> >> through as well, giving us consistent handling of aggregated packets.
> >> While we might not save a huge number of lines, I believe the
> >> resulting code will be easier to understand.  
> >
> > Apologies for the noise. When I check the code again, I see that as
> > long as FLAG_MULTI_PACKET is set, then we end up with usbnet freeing
> > the skb (we will always jump to done in rx_process()). So for the
> > pass-through case, I believe your initial suggestion of having
> > rx_fixup return 1 is the way to go.  
> 
> Yes, if we are to use FLAG_MULTI_PACKET then we must call
> usbnet_skb_return() for all the non-muxed cases.  There is no clean way
> to enable FLAG_MULTI_PACKET on-demand.

Tricky piece of code. Perhaps we could add another return code 
to the rx_fixup call? Seems that we expect 0 or 1 today, maybe we 
can make 2 mean "data was copied out", and use that for the qmimux 
case?

> I am fine with either solution.  Whatever Jakub wants :-)

Well, turns out I was looking at the wrong netif_rx() so take
what I say with a grain of salt ;)