netdev - Re: [next-queue PATCH 2/3] net/sched: Introduce Credit Based Shaper (CBS) qdisc

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <4b0e1610-c9df-191d-496e-4be04c785d2f@intel.com>
Date:   Wed, 27 Sep 2017 15:57:24 -0700
From:   Jesus Sanchez-Palencia <jesus.sanchez-palencia@...el.com>
To:     Vinicius Costa Gomes <vinicius.gomes@...el.com>,
        Cong Wang <xiyou.wangcong@...il.com>
Cc:     Linux Kernel Network Developers <netdev@...r.kernel.org>,
        intel-wired-lan <intel-wired-lan@...ts.osuosl.org>,
        Jamal Hadi Salim <jhs@...atatu.com>,
        Jiri Pirko <jiri@...nulli.us>, andre.guedes@...el.com,
        ivan.briano@...el.com, boon.leong.ong@...el.com,
        richardcochran@...il.com, henrik@...tad.us
Subject: Re: [next-queue PATCH 2/3] net/sched: Introduce Credit Based Shaper
 (CBS) qdisc

Hi,


On 09/27/2017 02:14 PM, Vinicius Costa Gomes wrote:
> Hi,
> 
> Cong Wang <xiyou.wangcong@...il.com> writes:
> 
>> On Tue, Sep 26, 2017 at 4:39 PM, Vinicius Costa Gomes
>> <vinicius.gomes@...el.com> wrote:
>>> +static int cbs_init(struct Qdisc *sch, struct nlattr *opt)
>>> +{
>>> +       struct cbs_sched_data *q = qdisc_priv(sch);
>>> +       struct net_device *dev = qdisc_dev(sch);
>>> +
>>> +       if (!opt)
>>> +               return -EINVAL;
>>> +
>>> +       /* FIXME: this means that we can only install this qdisc
>>> +        * "under" mqprio. Do we need a more generic way to retrieve
>>> +        * the queue, or do we pass the netdev_queue to the driver?
>>> +        */
>>> +       q->queue = TC_H_MIN(sch->parent) - 1 - netdev_get_num_tc(dev);
>>> +
>>> +       return cbs_change(sch, opt);
>>> +}
>>
>> Yeah it is ugly to assume its parent is mqprio, at least you should
>> error out if it is not the case.
> 
> Will add an error for this, for now.
> 
>>
>> I am not sure how we can solve this elegantly, perhaps you should
>> extend mqprio rather than add a new one?
> 
> Is the alternative hinted in the FIXME worse? Instead of passing the
> index of the hardware queue to the driver we pass the pointer to a
> netdev_queue to the driver and it "discovers" the HW queue from that.

What if we keep passing the index, but calculate it from the netdev_queue
pointer instead?

i.e.:  q->queue = sch->dev_queue - netdev_get_tx_queue(dev, 0);

At least it wouldn't rely on the root qdisc being of any specific type.

Regards,
Jesus