Hello Eli,
On Sun, 8 Nov 2026 at 17:24, Eli Britstein <[email protected]> wrote:
> >
> > - any physical level informations (status, speed, duplex etc...) for
> > the representors should be the same as the ESW manager port.
> > This removes the need for a specialised watchdog thread (only
> > monitoring the ESW manager port in netdev-dpdk.c should be enough).
> Representors must also start as they can have their own link status for
> example.
>
> Though not trivial in a standard configuration, there is another
> configuration in which it's more natural to see it:
>
> This configuration is MPESW (Multi-Port ESW). In such configuration, the
> uplinks of the card (e.g. p0,p1) become part of the same ESW. In such
> case, p0 is still the ESW port, while p1 is a representor of it.
>
> Then, as it is a physical device, it's easy to see it can have its own
> link status.
I am not familiar with such setups, but from what I understand, it is
problematic with my proposal.
Ok, so we must have a 1:1 relation.
It seems we may as well embed the eth port in the doca object directly:
struct netdev_doca {
PADDED_MEMBERS(CACHE_LINE_SIZE,
atomic_bool started;
struct netdev_doca_tx_stats *sw_tx_stats;
/* No datapath fields below for this block. */
struct doca_flow_port *port;
struct netdev_doca_esw_ctx *esw_ctx;
char *devargs;
bool is_representor;
);
struct netdev_dpdk_eth eth;
/* No datapath fields below for the rest of the structure. */
/* In doca_list. */
struct ovs_list list_node;
struct doca_flow_pipe_entry *rss_entries[NETDEV_DOCA_RSS_NUM_ENTRIES];
struct doca_flow_pipe_entry *egress_entry;
struct doca_dev_rep *dev_rep;
};
Note: op_state looks like a dead field in netdev_doca_esw_ctx, in your
current DOCA branch,
> It can also happen for "regular" representors.
?
Not sure what "regular" means here.
> I enhanced your approach with the shadow dpdk port for representors as
> well. For that, I needed to allow zero-queues dpdk ports.
>
> The doca code no longer handles the probe/close of the dpdk port. It
> generates the devargs and handed over to the shadow dpdk port.
>
> Regarding the closure of it, it cannot work as you wrote it. The issue
> is that calling "netdev_close(shadow)" is not allowed inside
> netdev_doca_destruct. The reason is that we are under netdev_mutex,
> which is not recursive. If calling it there, we assert.
Ah yes, an oversight on my side... sorry about that.
This idea was an implementation detail, I am not stuck on the use of netdev API.
The only specific ops seem to be netdev_doca_class_init,
netdev_doca_alloc/dealloc, netdev_doca_construct/destruct,
netdev_doca_get_config/set_config/reconfigure,
netdev_doca_send/rxq_recv, netdev_doca_get_stats/get_custom_stats.
All the netdev_dpdk_eth ops can be exported in netdev-dpdk-eth.h, and
reused in netdev-doca.c.
This is simpler than adding the deferred start/stop framework.
The doca_list fate is not clear to me.
It seems like it could move as a per ESW manager list... (at least, it
would be manipulated under this device mutex?), but I did not
investigate more.
> >
> > - Rx and Tx is just a matter of calling netdev-dpdk code with the ESW
> > manager DPDK netdev.
> > I tried the approach of exporting just those helpers in a new
> > netdev-dpdk-eth.h header, which seems enough.
After more digging with my AI colleague, I reached a point where I am
not sure the doca representors have to care about the DPDK mempool
reference at all.
With the "no_queue" concept, for a representor DPDK eth object, the
mempool setup and txq allocations are just wasted resources.
(I noticed common->buf_size field... which is dead code since day 0, I
could squash this in my preparation series).
netdev_doca_send then calls both common_send (where the mempool is
needed for converting non DPDK dp-packets into mbufs) and eth_tx_burst
for the esw manager DPDK common/eth object.
And all that is left to netdev_doca_send is to maintain the sw stats after this.
> > - (based on some AI investigation on this branch code) statistics are
> > suspicious.
> > I spotted a few oddities around Rx/Tx packets/bytes at the ESW
> > manager / representor ring level, drop counters, and most of them are
> > not reported through netdev API (?).
> > Some stats look simply copied from netdev-dpdk without any code touching
> > them.
> > The (ESW manager port) HW xstats are not reported.
> Ack. I reworked the stats area. All the NIC stats (also xstats) are
> relevant only for the ESW. I removed them from representors and added
> "hw_" prefix for them, shown only for the ESW port.
Yes, ok for me.
> >
> > - usage of coverage counters in the Rx/Tx code will make this hard to
> > debug, I would turn them into interface OVS stats.
> I assume you mean netdev_doca_rx_drop_xxx ones. Those coverage counters
> should never increase. It is just a way to see if that "couldn't happen"
> situation indeed doesn't happen. I prefer not to report them as regular
> netdev statistics and leave them as coverage ones.
Ack.
> > - there is still one open question for me, should we make sure
> > netdev-doca.c and netdev-dpdk.c don't collide in mid air?
> > Like a developer switching from one configuration type to the other
> > (especially on the ESW manager topic).
> I don't understand your meaning for this. How can they "collide"?
My concern is for developers, or during a live debugging session and
you want to switch from a netdev_dpdk to a netdev_doca configuration,
or vice versa.
Depending on your blood level of caffeine (or lack of), you may end up
with some netdev_doca port *and* some netdev_dpdk port accessing the
same DPDK port / hw resources.
This is where I suspect we need some guard.
--
David Marchand
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev