On Thu, Apr 25, 2019 at 05:45:03PM +0200, Marc Kleine-Budde wrote: > On 4/25/19 5:39 PM, David Jander wrote: > > On Thu, 25 Apr 2019 17:27:13 +0200 > > Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> wrote: > > > >> On 4/25/19 5:12 PM, Oleksij Rempel wrote: > >>> On Thu, Apr 25, 2019 at 02:54:16PM +0200, Kurt Van Dijck wrote: > >>>> On do, 25 apr 2019 14:31:06 +0200, Oleksij Rempel wrote: > >>>>> Hi all, > >>>>> > >>>>> please take a look at this patches. It is UAPI extension and it is good to > >>>>> know if it is sane way to track/recognize send packages. > >>>> > >>>> (1) This feedback reports success or failure for packets. > >>>> That is usefull at some point. > >>>> Is there a mechanism to track the real progress. This is something I had > >>>> in /proc somewhere, and which is usefull for larger transfers ... > >>> > >>> It can be done in the same way over error queue. The question is, what is the use case? > >>> 1. debugging? > >>> 2. provide progress bar for the GUI? > >>> 3. optimization? > >>> 4. coordination with some kind of watchdog? > >>> > >>> For example we can send notification for each transferred TP sized block of ETP > >>> transfer and make it configurable per setsocketopt. > >> > >> If this is of general interest, we could make a TODO item from this. > >> That could be implemented later. > > > > Providing a progress bar is something I can think of quite desirable. > > Some sort of notification for each TP sized block might be good enough for > > that purpose. But, I assume this will only work for TX and not for RX, right? > > The notion of a progress bar for incoming data on an ETP session is also > > something quite useful for an application. > > We haven't looked at the RX path at all. For now the kernel receives the > whole ETP and then pushes the whole block to any matching socket. Status update can be done for both directions. For the RX path we should implement session to socket binding first. How this functionality should be enabled? Per socket or per packet? Should it be enabled separately for TX and RX? RX will be possible only per socket. TX can be done both. > > I agree that we should implement that later and leave it as a TODO now. > > OK, maybe we could add some kind of public TODO list for this. -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
