linux-kernel - Re: [PATCH] locking/pvqspinlock: Add lock holder CPU argument to pv

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: <570EF973.4000402@hpe.com>
Date:	Wed, 13 Apr 2016 21:59:15 -0400
From:	Waiman Long <waiman.long@....com>
To:	Peter Zijlstra <peterz@...radead.org>
CC:	Ingo Molnar <mingo@...hat.com>, <linux-kernel@...r.kernel.org>,
	Pan Xinhui <xinhui@...ux.vnet.ibm.com>,
	Scott J Norton <scott.norton@....com>,
	Douglas Hatch <doug.hatch@....com>
Subject: Re: [PATCH] locking/pvqspinlock: Add lock holder CPU argument to
 pv_wait()

On 04/13/2016 08:21 PM, Peter Zijlstra wrote:
> On Wed, Apr 13, 2016 at 04:59:20PM -0400, Waiman Long wrote:
>> Pan Xinhui was asking for a lock holder cpu argument in pv_wait()
>> to help the porting of pvqspinlock to PPC. The new argument will can
>> potentially help hypervisor expediate the execution of the critical
>> section so that the lock holder vCPU can release the lock sooner.
>>
>> This patch does just that by storing the previous node vCPU number.
>> In pv_wait_head_or_lock(), pv_wait() will be called with that vCPU
>> number as it is likely to be the lock holder. In pv_wait_node(),
>> -1 will be passed to pv_wait() instead to indicate that it doesn't
>> know what the current lock holder is.
> Without knowing why he needs this, it is very hard to tell if this will
> suffice.
>
> Xinhui, what do you need the extra argument for?

I have some other people asking me if pvqspinlock was able to pass in 
information about which CPU was the lock holder before as he cited the 
PPC code has this capability to somehow kick up the vCPU that has the 
lock. That can be potentially useful in x86 too.

Cheers,
Longman