netdev - Re: [PATCH 1/1] Fix qmap header retrieval in qmimux_rx

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]

Date:   Fri, 21 Dec 2018 13:33:25 +0100
From:   Bjørn Mork <bjorn@...k.no>
To:     Daniele Palmas <dnlplm@...il.com>
Cc:     netdev@...r.kernel.org
Subject: Re: [PATCH 1/1] Fix qmap header retrieval in qmimux_rx_fixup

Daniele Palmas <dnlplm@...il.com> writes:

> This patch fixes qmap header retrieval when modem is configured for
> dl data aggregation.
>
> Signed-off-by: Daniele Palmas <dnlplm@...il.com>
> ---
> Hi Bjørn and all,
>
> I'm facing an issue when using qmi_wwan with modem configured with dl data aggregation and qmap multiplexing, e.g. something like
>
> root@...22:~# qmicli -d /dev/cdc-wdm0 --wda-set-data-format=link-layer-protocol=raw-ip,ul-protocol=qmap,dl-protocol=qmap,dl-max-datagrams=20
> [/dev/cdc-wdm0] Successfully set data format
>                         QoS flow header: no
>                     Link layer protocol: 'raw-ip'
>        Uplink data aggregation protocol: 'qmap'
>      Downlink data aggregation protocol: 'qmap'
>                           NDP signature: '0'
> Downlink data aggregation max datagrams: '20'
>      Downlink data aggregation max size: '16384'
>
> The issue is related to qmap header retrieval in qmimux_rx_fixup: basically it seems to me that it is always taken the first qmap header. Maybe the patch below should fix this issue.
>
> Note also that, by default, this won't be enough, since also rx_urb_size should be changed to the downlink data aggregation max size value: currently I'm just modifying the network interface MTU that changes also the
> rx_urb_size.
>
> Not sure if this makes sense, so I thought to share anyway this with you for confirmation.

My personal opinion - take it for that and nothing else:

Aggregation adds buffer bloat, alignment issues, extra headers with
associated magic handling, and more.  It can make sense for some use
cases where each transmission adds significant overhead. The typical
example is radio networks. But I do not think this applies to USB.  We
are much better off sending each packet as a separate USB buffer, with a
queue that is just long enough for back-to-back transmission.  The
experience with aggregation in NCM/MBIM is not good.

So I've ignored aggregation in QMI, and will probably continue doing
so.  That doesn't mean that I am going to object to you or anyone else
implementing the support if you see a usecase. That's your problem ;-)

Bjørn