On 19/06/2019 21:10, Vijay Kumar Banerjee wrote: > > > > On Wed, Jun 19, 2019 at 9:14 PM Christian Mauderer <l...@c-mauderer.de > <mailto:l...@c-mauderer.de>> wrote: > > Bit less adaptions than I thought that you have in the code. I assumed > that I would see some more locations in that patch where you already > removed code blocks with ifdef __rtems__. I wanted to have a look > especially at these parts. I expect that you will have some problems > with memories (like allocating or copying from/to I/O memory or some DMA > stuff). Please try to be really careful with that stuff. A lot of the > RTEMS implementations are quite a bit simplified or there have been > hacks during porting such parts. > > There were some other files where I needed to do many adaptations but > I have removed them for now to reduce the number of errors I'm getting, > I'll incrementally add them only on top of working patches.
That most likely would have been the more interesting files and adaptions. But the incremental approach is OK too ;-) Like I said: Be careful with memory and DMA. That is often a little hard to get right (and not only working 95% of the time). Therefore be very careful with just removing or replacing stuff because it has been removed or replaced in other drivers. Try to understand for what the memory has been used. If you are not sure, don't hesitate to ask - ideally with small code snippets or excerpts from the incremental patches you mentioned. If you ask via mailing list it can be a reference for others. > > For example sometimes you can replace some FreeBSD I/O memory operations > by just plain memory accesses. But sometimes (like for DMA) it is > necessary to think about them and add cache handling. > > [...] _______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel
