linux-kernel - Re: [PATCH] locking/osq: Drop the overload of osq lock

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] [day] [month] [year] [list]

Date:	Mon, 27 Jun 2016 18:31:25 +0800
From:	Boqun Feng <boqun.feng@...il.com>
To:	Peter Zijlstra <peterz@...radead.org>
Cc:	panxinhui <xinhui@...ux.vnet.ibm.com>,
	Pan Xinhui <xinhui.pan@...ux.vnet.ibm.com>,
	linux-kernel@...r.kernel.org, mingo@...hat.com, dave@...olabs.net,
	will.deacon@....com, Waiman.Long@....com, benh@...nel.crashing.org
Subject: Re: [PATCH] locking/osq: Drop the overload of osq lock

On Mon, Jun 27, 2016 at 10:09:59AM +0200, Peter Zijlstra wrote:
[snip]
> 
> No, this is entirely insane, also broken.
> 
> No vectors, no actual function calls, nothing like that. You want the
> below to completely compile away and generate the exact 100% same code
> it does today.
> 

Point taken.

As Xinhui also posted something similar, which worked better on not
effecting the generated code if not enabled. I think I'd better to drop
this workload to him ;-)

> > +++ b/kernel/locking/osq_lock.c
> > @@ -1,6 +1,7 @@
> >  #include <linux/percpu.h>
> >  #include <linux/sched.h>
> >  #include <linux/osq_lock.h>
> > +#include <linux/vcpu_preempt.h>
> >  
> >  /*
> >   * An MCS like lock especially tailored for optimistic spinning for sleeping
> > @@ -87,6 +88,8 @@ bool osq_lock(struct optimistic_spin_queue *lock)
> >  	struct optimistic_spin_node *prev, *next;
> >  	int curr = encode_cpu(smp_processor_id());
> >  	int old;
> > +	int loops;
> > +	long vpc;
> >  
> >  	node->locked = 0;
> >  	node->next = NULL;
> > @@ -106,6 +109,9 @@ bool osq_lock(struct optimistic_spin_queue *lock)
> >  	node->prev = prev;
> >  	WRITE_ONCE(prev->next, node);
> >  
> > +	old = old - 1;
> 
> That's just nasty, and could result in an unconditional decrement being
> issues, even though its never used.
> 

Right, better to calculate this decrement at the argument of
vcpu_is_preempted() callsite, and define the primitive of host code as

#define vcpu_is_preempted(cpu) false

, which could probably be optimized out.

Regards,
Boqun

> > +	vpc = vcpu_preempt_count();
> > +
> >  	/*
> >  	 * Normally @prev is untouchable after the above store; because at that
> >  	 * moment unlock can proceed and wipe the node element from stack.
> > @@ -118,8 +124,14 @@ bool osq_lock(struct optimistic_spin_queue *lock)
> >  	while (!READ_ONCE(node->locked)) {
> >  		/*
> >  		 * If we need to reschedule bail... so we can block.
> > +		 * An over-committed guest with more vCPUs than pCPUs
> > +		 * might fall in this loop and cause a huge overload.
> > +		 * This is because vCPU A(prev) hold the osq lock and yield out,
> > +		 * vCPU B(node) wait ->locked to be set, IOW, wait till
> > +		 * vCPU A run and unlock the osq lock.
> > +		 * NOTE that vCPU A and vCPU B might run on same physical cpu.
> >  		 */
> > -		if (need_resched())
> > +		if (need_resched() || vcpu_is_preempted(old) || vcpu_has_preempted(vpc))
> >  			goto unqueue;
> >  
> >  		cpu_relax_lowlatency();
> > 

Download attachment "signature.asc" of type "application/pgp-signature" (474 bytes)