From mboxrd@z Thu Jan 1 00:00:00 1970 References: <2990549.hKSmcVRoDB@indiana> <2013828.8eEOvBkU4r@soho> <5ed168a2-fa54-0d21-a6ff-05abe1d95b11@siemens.com> <7810159.2OkF4KFj7y@fura> From: Philippe Gerum Message-ID: Date: Thu, 6 Oct 2016 19:30:54 +0200 MIME-Version: 1.0 In-Reply-To: <7810159.2OkF4KFj7y@fura> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Subject: Re: [Xenomai] About SMAP (was Re: Xenomai 3.0.3 is broken in my system) List-Id: Discussions about the Xenomai project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Leopold Palomo-Avellaneda , Jan Kiszka Cc: xenomai@xenomai.org On 10/06/2016 07:24 PM, Leopold Palomo-Avellaneda wrote: > El Dijous, 6 d'octubre de 2016, a les 13:50:00, Jan Kiszka va escriure: >> On 2016-10-06 11:51, Leopold Palomo-Avellaneda wrote: >>> Hi, >>> >>> lasts weeks I have been in a trouble with Xenomai and RTnet. Finally it >>> was >>> solved disabling some parameters in kernel (nosmap) Thanks again Jan. >>> >>> That parameter is an interesting feature of the Intel processors [1,2], >>> that IMHO affects all the RTnet code and I hope that nothing more of >>> Xenomai. >>> >>> It seems that fix it is could be easy, but it must be done (a boring task >>> ;-) In the middle, I would propose that some note could be added to the >>> FAQ, or "Troubleshooting a dual kernel configuration" page. Something >>> like this in the "Common kernel configuration issues": >>> >>> CONFIG_X86_SMAP >>> >>> On modern Intel processors with SMAP-enabled systems, kernel is not >>> allowed to access to userspace memory, it must be done in a special way. >>> So, until all Xenomai code was updated, specially RTnet, if you find >>> crashes with " unable to handle kernel paging request at " disable this >>> feature neither in kernel configuration or in boot with nosmap. >> >> Send your proposal as patch for the included documentation or even the >> kernel Kconfig tests. > > I think that I don't understand you. AFAIK xenomai documentation is in a > Wordpress that Gilles installed and configured. No, I did it, and wordpress is fed by asciidoc input. But instead of documenting a shortcoming, I would rather fix it in the code. Sentences like "until some code is updated" in a documentation reads as "when hell freezes", simply because the effort was not put at the right place. So, since you seem to be requiring RTnet to work, why not writing a patch fixing the places where that code hurts with respect to smap? -- Philippe.