Date: Mon, 13 Apr 2009 12:53:36 -0400 From: John Baldwin <jhb@freebsd.org> To: freebsd-hackers@freebsd.org Cc: Robert Watson <rwatson@freebsd.org>, Andrew Brampton <brampton+freebsd-hackers@gmail.com> Subject: Re: FreeBSD memguard + spinlocks Message-ID: <200904131253.37019.jhb@freebsd.org> In-Reply-To: <d41814900904111403y6732cdfeme24904e8d1ed735a@mail.gmail.com> References: <d41814900904101616n73c30e05t3c56a6cd8d9b6c0@mail.gmail.com> <alpine.BSF.2.00.0904111607410.19879@fledge.watson.org> <d41814900904111403y6732cdfeme24904e8d1ed735a@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday 11 April 2009 5:03:58 pm Andrew Brampton wrote: > 2009/4/11 Robert Watson <rwatson@freebsd.org>: > > On Sat, 11 Apr 2009, Andrew Brampton wrote: > > > > Your understanding is mostly right. =C2=A0The missing bit is this: ther= e are=20 two > > kinds of interrupt contexts -- fast/filter interrupt handlers, which=20 borrow > > the stack and execution context of the kernel thread they preempt, and > > interrupt threads, which get their own complete thread context. > > > > Fast interrupt handlers are allowed unlock to acquire spinlocks so as to > > avoid deadlock because of the borrowed context. =C2=A0This means they c= an't > > perform any sort of sleep, or acquire any locks that might sleep, since= =20 the > > thread they've preempted may hold conflicting locks, or be the one that > > would have woken up the sleep that the handler performed. =C2=A0Almost = no code > > will run in fast handlers -- perhaps checking some device registers, do= ing > > work on a lockless or spinlock-protected queue, and waking up a worker > > thread. > > > > This is why, BTW, spin locks disable interrupt: they need to control > > preemption by other interrupt handlers to avoid deadlock, but they are = not > > intended for use except when either in the scheduler, in a few related = IPI > > contexts, or when synchronizing between normal kernel code and a fast > > handler. > > > > Full interrupt thread contexts are permitted to perform short lock slee= ps, > > such as those performed when contending default mutexes, rwlocks, and > > rmlocks. They are permitted to invoke kernel services such as malloc(9), > > UMA(9), the network stack, etc, as long as they use M_NOWAIT and don't > > invoke msleep(9) or similar unbounded sleeps -- again to avoid the > > possibility of deadlocks, since you don't want an interrupt thread=20 sleeping > > waiting for an event that only it can satisfy. > > > > So the first question, really, is whether you are or mean to be using > > fast/filter interrupt handler. =C2=A0Device drivers will never call mem= ory > > allocation, free, etc, from there, but will defer it to an ithread usin= g=20 the > > filter mechanism in 8.x, or to a task queue or other worker in 7.x and > > earlier. =C2=A0If you're using a regular INTR_MPSAFE ithread, you shoul= d be=20 able > > to use only default mutexes (a single atomic operation if uncontended) > > without disabling interrupts, etc. > > > > Robert N M Watson > > Computer Laboratory > > University of Cambridge > > >=20 > Anyway, that is why I also asked about a lighter weight spin lock > (perhaps similar to this one). I tempted to replace this custom > spinlock with the standard MTX_DEF, however I'm unsure of its impact. > The custom spin lock seems quick and light to acquire, and it does not > concern me that a interrupt can potentially interrupt the code. You should just use a MTX_DEF mutex. Also, if you use M_NOWAIT, you will n= eed=20 to handle malloc() returning NULL. In general I try to allocate things whi= le=20 not holding any locks when possible and only acquire the lock to initialize= =20 the memory returned from malloc(). =2D-=20 John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200904131253.37019.jhb>