On 20.02.2024 19:44:21, Vincent MAILHOL wrote: > On Tue. 20 Feb. 2024 at 17:35, Vincent Mailhol > <vincent.mailhol@xxxxxxxxx> wrote: > > On Tue. 20 Feb. 2024 at 17:10, Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> wrote: > > > > > > On 20.02.2024 08:49:26, Dr. Michael 'Mickey' Lauer wrote: > > > > Hi Marc, > > > > > > > > >> > > > > >> Sounds pretty cool. Our design is based on an ESP32-S3 though, so I > > > > >> have to duplicate that effort. Is there some kind of USB protocol > > > > >> definition for what gs-usb expects or do I need to learn that from the > > > > >> implementation? > > > > > > > > > > I've put together a very rough summary of the driver's side of view: > > > > > > > > > > https://gist.github.com/marckleinebudde/99f72d9db67accf08cc6efa367cfe54b > > > > > > > > > > Feel free to update. > > > > > > > > Wow, that’s very kind of you! > > > > > > :) > > > > > > > Later this year I’ll take this blueprint to implement the necessary > > > > changes in our firmware. I’ll enhance / update your summary > > > > as I’ll go. > > > > > > If you discover any odd points or areas for improvement in the USB > > > protocol, please let us know. > > > > > > Currently, TX-complete sends the _entire_ structure back to the host, > > > which is quite an overhead, especially for CAN FD frames. We're working > > > on a "short-tx-complete" feature. > > > > The es581.4 may be doing what you are looking for: > > > > - When sending a frame, the driver attaches a 32 bit index: > > https://elixir.bootlin.com/linux/v6.7/source/drivers/net/can/usb/etas_es58x/es581_4.h#L78 > > - When receiving the TX completion message, the device just send > > back that index and a timestamp: > > https://elixir.bootlin.com/linux/v6.7/source/drivers/net/can/usb/etas_es58x/es581_4.h#L97 > > > > The device is agnostic of what this index represents. The driver uses > > the index from the can_put_echo_skb() / can_get_echo_skb(). > > > > (in addition the driver can handle bulk send and receive but that's > > another story). > > I forgot to mention: it is also perfectly fine not to use the index as > long as you can guarantee that the TX completion messages are received > in order (which is not hard to achieve). In which case, the TX > completion message can be as simple as just a timestamp. ACK, but in the medium term I want to add a CTRL mode to switch CAN devices into true "send CAN frames by prio" mode. regards, Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung Nürnberg | Phone: +49-5121-206917-129 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-9 |
Attachment:
signature.asc
Description: PGP signature
