Xen project Mailing List

On Mon, Mar 21, 2011 at 10:56 AM, Shriram Rajagopalan <rshriram@xxxxxxxxx> wrote:

On Mon, Mar 21, 2011 at 7:51 AM, Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx> wrote:

Shriram Rajagopalan writes ("[Xen-devel] [PATCH] remus: fix check for installed qdiscs on ifb"):

> remus: fix check for installed qdiscs on ifb

Thanks.

> current check includes ingress and pfifo_fast.
> Add mq to the list of allowed qdiscs already installed
> on ifb. This patch fixes cases where remus fails to start,
> due to an mq qdisc already present on the vif.

Forgive me for being dense, but I don't understand this at all. What
is the problem caused by pre-existing qdiscs that the code is trying
to avoid, and why are these particular qdiscs OK ?

sorry. my bad. It is not the pre-existing qdiscs that cause an issue. It is
caused by the dummy "mq" qdisc that gets added by "default". The original
code checks for presence of only ingress/pfifo-fast qdisc. If anything else is
present, it punts. In this case, "mq" is present (added by default) and causes
remus to fail.

This is what I understood from the kernel netfilter code & docs.

from net/netfilter/sched/sched_generic.c
void dev_activate(struct net_device *dev)
{
        /* No queueing discipline is attached to device;
           create default one i.e. pfifo_fast for devices,
           which need queueing and noqueue_qdisc for
           virtual interfaces
         */

        if (dev->qdisc == &noop_qdisc)
                attach_default_qdiscs(dev);
....
static void attach_default_qdiscs(struct net_device *dev)
{
...
        if (!netif_is_multiqueue(dev) || dev->tx_queue_len == 0) {
                netdev_for_each_tx_queue(dev, attach_one_default_qdisc, NULL);
       ...
        } else {
                qdisc = qdisc_create_dflt(dev, txq, &mq_qdisc_ops, TC_H_ROOT);
       ...

sch_mq is a "Classful multiqueue dummy scheduler" and according to the
multiqueue semantics in Section 2: Documentation/networking/multiqueue.txt

"Currently two qdiscs are optimized for multiqueue devices. The first is the
default pfifo_fast qdisc. This qdisc supports one qdisc per hardware queue.
A new round-robin qdisc, sch_multiq also supports multiple hardware queues."

shriram
Thanks,
Ian.

Re: [Xen-devel] [PATCH] remus: fix check for installed qdiscs on ifb