Date: Mon, 10 Apr 2017 12:01:51 +0800 From: Yubin Ruan <ablacktshirt@gmail.com> To: Warner Losh <imp@bsdimp.com> Cc: Ryan Stone <rysto32@gmail.com>, "freebsd-hackers@freebsd.org" <freebsd-hackers@freebsd.org>, Ed Schouten <ed@nuxi.nl> Subject: Re: Understanding the FreeBSD locking mechanism Message-ID: <b69597cd-fab6-7ef8-7dfe-d097283c4064@gmail.com> In-Reply-To: <CANCZdfoyjcSU%2BNHEVJF=bd8xz-Q-H1EupMPX%2BJk45r3DKZ9F9Q@mail.gmail.com> References: <e99b6366-7d30-a889-b7db-4a3b3133ff5e@gmail.com> <CABh_MKkbVVi%2BgTkaBVDvVfRggS6pbHKJE_VbYBZpAaTCZ81b7Q@mail.gmail.com> <c72c0ee3-328d-3efc-e8a0-4d6c0d5c8cee@gmail.com> <CAFMmRNwWnaq-4vEDCByqdUzWfoiZeN0nM_M5rt8ST0P8xnUTsA@mail.gmail.com> <3f93930c-7f10-4d0b-35f2-2b07d64081f0@gmail.com> <CANCZdfoyjcSU%2BNHEVJF=bd8xz-Q-H1EupMPX%2BJk45r3DKZ9F9Q@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2017/4/10 9:51, Warner Losh wrote: > On Sun, Apr 9, 2017 at 7:28 PM, Yubin Ruan <ablacktshirt@gmail.com> wrote: >> On 2017/4/10 0:24, Ryan Stone wrote: >>> >>> >>> >>> On Sun, Apr 9, 2017 at 6:13 AM, Yubin Ruan <ablacktshirt@gmail.com >>> <mailto:ablacktshirt@gmail.com>> wrote: >>> >>> >>> #######1, spinlock used in an interrupt handler >>> If a thread A holding a spinlock T get interrupted and the interrupt >>> handler responsible for this interrupt try to acquire T, then we have >>> deadlock, because A would never have a chance to run before the >>> interrupt handler return, and the interrupt handler, unfortunately, >>> will continue to spin ... so in this situation, one has to disable >>> interrupt before spinning. >>> >>> As far as I know, in Linux, they provide two kinds of spinlocks: >>> >>> spin_lock(..); /* spinlock that does not disable interrupts */ >>> spin_lock_irqsave(...); /* spinlock that disable local interrupt * >>> >>> >>> In the FreeBSD locking style, a spinlock is only used in the case where >>> one needs to synchronize with an interrupt handler. This is why spinlocks >>> always disable local interrupts in FreeBSD. >>> >>> FreeBSD's lock for the first case is the MTX_DEF mutex, which is >>> adaptively-spinning blocking mutex implementation. In short, the MTX_DEF >>> mutex will spin waiting for the lock if the owner is running, but will >>> block if the owner is deschedules. This prevents expensive trips through >>> the scheduler for the common case where the mutex is only held for short >>> periods, without wasting CPU cycles spinning in cases where the owner >>> thread >>> is descheduled and therefore will not be completing soon. >> >> >> Great explanation! I read the man page at: >> >>> >>> https://www.freebsd.org/cgi/man.cgi?query=mutex&sektion=9&apropos=0&manpath=FreeBSD+11.0-RELEASE+and+Ports >> >> and now clear about MTX_DEF and MTX_SPIN mutexs. But, still a few more >> question, if you don't mind: >> >> Is it true that a thread holding a MTX_DEF mutex can be descheduled? >> (shouldn't it disable interrupt like a MTX_SPIN mutex?) It is said on >> the main page that MTX_DEF mutex is used by default in FreeBSD, so its >> usecase must be very common. If a thread holding a MTX_DEF mutex can be >> descheduled, which means that it did not disable interrupt, then we may >> have lots of deadlock here, right? > > Yes, they can be descheduled. But that's not a problem. No other > thread can acquire the MTX_DEF lock. If another thread tries, it will > sleep and wait for the thread that holds the MTX_DEF lock to release > it. Eventually, the thread will get time to run again, and then > release the lock. Threads that just hold a MTX_DEF lock may also > migrate from CPU to CPU too. > > Warner > Does that imply that MTX_DEF should not be used in something like interrupt handler? Putting an interrupt handler into sleep doesn't make so much sense. Yubin >>> #######2, priority inversion problem >>> If thread B with a higher priority get in and try to acquire the lock >>> that thread A currently holds, then thread B would spin, while at the >>> same time thread A has no chance to run because it has lower priority, >>> thus not being able to release the lock. >>> (I haven't investigate enough into the source code, so I don't know >>> how FreeBSD and Linux handle this priority inversion problem. Maybe >>> they use priority inheritance or random boosting?) >>> >>> >>> FreeBSD's spin locks prevent priority inversion by preventing the holder >>> thread from being descheduled. >>> >>> MTX_DEF locks implement priority inheritance. >> >> >> Nice hints. Thanks! >> >> regards, >> Yubin Ruan >> >> >> _______________________________________________ >> freebsd-hackers@freebsd.org mailing list >> https://lists.freebsd.org/mailman/listinfo/freebsd-hackers >> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?b69597cd-fab6-7ef8-7dfe-d097283c4064>