.@paulmckrcu Would it be possible to do "norcu" as "synchronize on every return to userspace and every cpu_idle entry"?
I don't know, This thread is inspired by but in some sense orthogonal to current stall problem.
-
-
Mainly frustration that embed/dt/server systems need huge complex feedback system "designed for...hundreds or thousands of CPUs"
-
...when they would perform at least as well, and much more predictably, with simple rwlocks or similar.
-
You could write a SMP RCU for small systems. However, it would still have CPU stall warnings. Not clear it would help you.
-
It seems like, in principle, TINY_RCU+locks should work for SMP.
-
In practice, as discussed earlier, use of lock for RCU in Linux kernel gets you deadlock, etc.
End of conversation
New conversation -
-
-
When you get the timers fixed, let's look again.
-
Well, when I get whatever the regression is fixed. Not sure if it's timer, scheduler, rcu, or elsewhere just affecting them.
-
I'm going to post a follow-up to the lkml thread with some questions.
End of conversation
New conversation -
-
-
If you have timer problems, and you didn't have RCU, you would likely instead be fighting softlockups. Shoot the messenger?
-
But they'd be debuggable, I think. :-)
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.