Vladimir Cotfas wrote: > Jan, > > >> Some remaining issues: >> >> - command timeout handling should follow rt_eepro100 (CONFIG option, >> low timeout by default - you can still raise it at configure time if >> you need that for your setup) >> > Please indicate the place in code. Will look into it tomorrow. >
http://www.rts.uni-hannover.de/rtnet/lxr/source/drivers/rt_eepro100.c#328 > >> - udelay can cause subtle crashes when used over RT, see RTDM docs or >> rt_eepro100 how to do it correctly >> > Wrapped using rtdm_task_busy_sleep. Please check the order of magnitude [I > am know to get that wrong some times]. > Order of magnitude is ok. You just converted more than required. But that is harmless can be sorted out later. > >> - remaining USE_RTAI_TASKS fragments need to go for the upstream >> version, build breakages due to removing it need to be resolved >> (BTW, have you tested the driver already without all the tasks?) >> > Yes, while I was porting the driver worked fine [for 2-3 hour stretches] w/o > the tasks and doing RX processing in the IRQ context. And no Linux timer. > > - same for USE_RTAI_IRQ, which has to become always on > It is on right now. It exists in the patch so we have less work when we hack > the driver for our FPGA ISR. They are irrelevant for anything else than your system. I will not merge them as they increase the diff to the kernel driver and degrade readability of the active code paths. > > >> - RX_RING_SIZE has to be a module parameter (it's way too huge for >> normal RT operation), see rt_8139too for an example >> > Done. > Great! Just the default is still too high. Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------------
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
