On 12/11/17 3:23 am, Christian Mauderer wrote: > Am 11.11.2017 um 05:14 schrieb Chris Johns: >> On 10/11/2017 17:59, Christian Mauderer wrote: >>> >>> I took a quick glance to find out the behaviour on FreeBSD: It supports >>> hot-plugging. If I for example add a >>> >>> wlans_rtwn0="wlan7" >>> ifconfig_wlan7="DHCP" >>> >>> to rc.conf a RTL8188 USB WiFi dongle is created as wlan7. I'll add that >>> pice of information to the ticket. I think further investigation is >>> necessary which services are involved into that. >> >> This makes sense and thank you. >> >>> But that will have to wait till there is time for that part. >> >> Yes. >> >>> >>> I meant that we could parse the rc.conf on the first pass and then put >>> the relevant information (rtwn0 as wlan7) into for example a linked >>> list. That list can then be used in the hot-plug-events. With that, it >>> wouldn't be necessary to do a lot of string parsing on every device >>> detection. >> >> I think the overhead of parsing will be small compared to the act of the >> device >> being added to a system. This sort of code is not time critical. It would be >> nice to be able to edit rc.conf reinsert the device and see what happens. > > The problem with "edit rc.conf" is that (as far as I understood that > interface) we have two possible sources: > > 1. from a string: rtems_bsd_run_rc_conf_script > 2. from a file: rtems_bsd_run_rc_conf and rtems_bsd_run_etc_rc_conf > > So we would have to remember the last source and parse that. Can we be > sure that the string is always there or would it be necessary to > duplicate it? The "const char* text" should in theory guarantee that it > won't change, right? >
Yes, I see what you are saying. On FreeBSD the path is set. >> >>> >>> But that is something that would have to be discussed as soon as someone >>> want's to start an implementation. >>> >> >> Yes I agree. Do you think hot-plug support is a suitable GSoC project for >> next year? > > I would expect that it is quite a lot of initial investigation to find > out how to do that (for example: finding out how to get events for new > devices). But if a student is interested in it, I would say it should be > OK as a project. > > But note that I would see the two tickets quite separate in that case > and could even be separate projects. > > The first one would be to find some events from newly created devices > and do something depending on that events (rc.conf parsing and creating > wlanX interfaces). > > The second one would be to make dhcpcd hooks useable for RTEMS and then > do something with them like calling wpa_supplicant. > > I think both shouldn't be too hard in regard to the coding. But they > might need some deeper understanding of the libbsd or dhcpcd which could > be hard for someone who isn't used to read bigger amounts of source code > - especially if it has that many layers of abstraction like FreeBSD code. > This all makes sense and I agree. It is a good next stage. Again thank you for doing this now. Chris _______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel
