This really clarifies things. In this situation I think I will stick with Cobalt, then.
Wojciech Domski domski.pl | blog.domski.pl W dniu 18.10.2018 o 09:52, Philippe Gerum pisze: > On 10/17/2018 10:38 PM, Wojciech Domski wrote: >> Hi! >> >> Is it possible to create a RTDM driver as a kernel module under Mercury >> and use it? >> >> I have patched kernel with preempt_rt and installed Xenomai 3.0.7 >> Mercury. However, when I try to compile my driver it says that it can >> not find <rtdm/driver.h> header file. >> >> I have checked the /usr/xenomai/include and there is no rtdm/driver.h >> only rtdm/rtdm.h. Does this mean that I can not develop driver under the >> Xenomai Mercury and I have to install Xenomai Cobalt? Or there is >> something wrong with my Xenomai installation? I have referred to >> https://gitlab.denx.de/Xenomai/xenomai/wikis/Start_Here and it does not >> mention that I can not use RTDM drivers in Mercury, actually the >> opposite but this is only showed in the Figures that for Mercury a RTDM >> driver is being used. On the other hand, it is said that: >> >> "the stock kernel drivers can be reused by real-time applications >> with no change (provided they don’t create latency issues due to >> an unfortunate implementation). On the other hand, drivers need >> to be specifically implemented for running over a dual kernel >> configuration, i.e. over the in-kernel RTDM interface." >> >> If the answer to my question is yes then what steps should I take to >> compile the RTDM driver for Mercury? >> >> And, maybe a naive question, if the answer is no then what I really gain >> from using the Mercury instead of a kernel patched with preempt_rt >> besides neat IPC mechanisms etc.? >> > Short answer: although envisioned at the beginning of the Xenomai 3.x > development cycle, RTDM over Mercury never was, and I don't know of any > plan about implementing this. As Greg mentioned earlier, the RTDM-native > work was an early PoC over Xenomai 2.3 back in 2007, but nobody kept > that effort alive. > > RTDM over Mercury would allow porting existing Cobalt drivers to a > native kernel with almost no change, but as I just mentioned, nobody > cared eventually so the idea was dropped. For new developments targeting > a native kernel, using the regular kernel interface is definitely best. > _______________________________________________ Xenomai mailing list Xenomai@xenomai.org https://xenomai.org/mailman/listinfo/xenomai
