On 15/07/19 14:36, Yury Kotov wrote: > Sorry, perhaps I was not accurate enough. > > To fix the bug I changed the logic of cpu_throttle_thread() function. > Before this function called qemu_mutex_(un)lock_iothread which encapsulates > work with qemu_global_mutex. > > Now, this calls qemu_cond_timedwait(..., &qemu_global_mutex, ...) which also > unlocks/locks qemu_global_mutex. But, in theory, behavior of > qemu_mutex_(un)lock_iothread may differ from simple locking/unlocking of > qemu_global_mutex. > > So, I'm not sure is such change is ok or not.
Ah, I see. No, it's okay. The only difference is setting/clearing iothread_locked which doesn't matter here. Paolo
