--- Log opened Wed Jul 12 00:00:10 2017 | ||
stekern | shorne: I agree, global clocksource might be nice, but if the non-hacky timer sync works, that's good enough IMO | 02:09 |
---|---|---|
shorne | stekern: right, now ... I am getting some strange issues with spinlocks (using the qspinlocks suggested by peterz) | 02:26 |
shorne | both cpus get hung trying to aquire locks | 02:26 |
shorne | Im going to switch out to your spinlock implementation | 02:26 |
shorne | to see if that fixes it. The qspinlocks is nice because its fast+generic code using only arch specific cmpxchg | 02:27 |
shorne | but its not nice if its not working | 02:27 |
shorne | stekern: fyi, I didnt have much luck with your spinlocks either it actually looks almost the same. So that maybe rules out a spinlock issue | 05:45 |
shorne | [ 3.340000] BUG: spinlock wrong owner on CPU#1, swapper/1/0 | 05:47 |
shorne | I have a lot of the spinlock bug detection code enabled | 05:48 |
stekern | hmm, but what else have you changed? or is it just on real hw you see this? | 05:49 |
shorne | stekern: other than the timers not much has been changed. And yes, this i only on the fpga where I am seeing this | 09:22 |
shorne | I have been emulating on qemu so far | 09:23 |
shorne | it seems a bit different every time | 09:25 |
shorne | this time I am getting | 09:25 |
shorne | a stack like: | 09:26 |
shorne | ipi_icache_page_inv | 09:26 |
shorne | smp_call_function_single | 09:26 |
shorne | ompic_raise_softirq+ | 09:26 |
shorne | ipi_icache_page_inv | 09:26 |
shorne | smp_call_function_single+ | 09:27 |
shorne | ipi_icache_page_inv | 09:27 |
shorne | ... | 09:27 |
shorne | some kind of loop of ipi_icache_page_inv | 09:27 |
shorne | in update_cache(), called by handle_mm_fault() | 09:28 |
shorne | anyway, it could be the stack trace is wrong | 09:28 |
-!- [X-Scale] is now known as X-Scale | 12:17 | |
--- Log closed Thu Jul 13 00:00:12 2017 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!