linux-kernel - Re: What if a TLB flush needed to sleep?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20080326123239.GG16721@parisc-linux.org>
Date:	Wed, 26 Mar 2008 06:32:39 -0600
From:	Matthew Wilcox <matthew@....cx>
To:	"Luck, Tony" <tony.luck@...el.com>
Cc:	linux-arch@...r.kernel.org, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org
Subject: Re: What if a TLB flush needed to sleep?

On Tue, Mar 25, 2008 at 01:49:54PM -0700, Luck, Tony wrote:
> 1) Is holding a spin lock a problem for any other arch when
> doing a TLB flush (I'm particularly thinking of those that
> need to use IPI shootdown for the purge)?

parisc certainly has that problem too.  It won't happen very often, but
it will do an on_each_cpu() and wait for the outcome once in a while.

> 2) Is it feasible to rearrange the MM code so that we don't
> hold any locks while doing a TLB flush?  Or should I implement
> some sort of spin_only_semaphore?

down_spin() is trivial to implement without knowing the details of the
semaphore code:

void down_spin(struct semaphore *sem)
{
	while (down_trylock(sem))
		cpu_relax();
}

Of course, someone who wrote it could do better ;-)

void down_spin(struct semaphore *sem)
{
	unsigned long flags;
	int count;

	spin_lock_irqsave(&sem->lock, flags);
	count = sem->count - 1;
	if (likely(count >= 0))
		sem->count = count;
	else
		__down_spin(sem);
	spin_unlock_irqrestore(&sem->lock, flags);
}

void __down_spin(struct semaphore *sem)
{
	struct semaphore_waiter waiter;

	list_add_tail(&waiter.list, &sem->wait_list);
	waiter.task = current;
	waiter.up = 0;

	spin_unlock_irq(&sem->lock);
	while (!waiter.up)
		cpu_relax();
	spin_lock_irq(&sem->lock);
}

This more complex implementation is better because:
 - It queues properly (see also down_timeout)
 - It spins on a stack-local variable, not on a global structure

Having done all that ... I bet parisc and ia64 aren't the only two
architectures which can sleep in their tlb flush handlers.

-- 
Intel are signing my paycheques ... these opinions are still mine
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/