On 09/25/2014 06:01 PM, Philippe Gerum wrote: > On 09/25/2014 05:47 PM, Philippe Gerum wrote: >> On 09/25/2014 05:42 PM, Philippe Gerum wrote: >>> On 09/25/2014 05:14 PM, Philippe Gerum wrote: >>>> On 09/25/2014 04:52 PM, Reitinger, Peter wrote: >>>>> Dear Xenomai people, >>>>> >>>>> I have recently tried to migrate from 2.x to 3.x, but when I tried to run >>>>> xeno-test, I get the following output: >>>>> >>>>> >>>> >>>> [snip] >>>> >>>>> [ 94.434282] [Xenomai] bad syscall <167772715> >>>>> Segmentation fault >>>>> root@dimm-am335x:/usr/xenomai/bin# >>>>> >>>>> >>>>> >>>>> Probably I just forgot one or another little thing to adapt...? >>>>> >>>> >>>> This is not a Xenomai 3 syscall signature, but a Xenomai 2.x's. Could >>>> you provide the output of /usr/xenomai/bin/version -a? >>>> >>> >>> Never mind, I can reproduce on igep, this looks like arm-specific. I'll >>> follow up on this. Thanks. >>> >> >> Ok, the watchdog fired while running the quata scheduling test, which is >> legitimate with the default 4s threshold. What went wrong is that some >> of the watchdog bits were not updated to the new syscall format. >> Raising CONFIG_XENO_OPT_WATCHDOG_TIMEOUT to a larger value will prevent >> the trigger (or disabling CONFIG_XENO_OPT_WATCHDOG as well), while I'm >> fixing this. >> > > Nope, disabling the watchdog won't be enough, as the Cobalt core uses a > mayday trap internally for other purpose. However, this patch should fix > the issue: > > diff --git a/kernel/cobalt/arch/arm/mayday.c
Fix confirmed, the complete non-regression test now passes with the current master and -next branches. -- Philippe. _______________________________________________ Xenomai mailing list [email protected] http://www.xenomai.org/mailman/listinfo/xenomai
