Re: [PATCH] x86 spinlock: Fix memory corruption on completing completions

[Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>]

On 02/11/2015 11:08 PM, Oleg Nesterov wrote:
> On 02/11, Raghavendra K T wrote:
>>
>> On 02/10/2015 06:56 PM, Oleg Nesterov wrote:
>>
>>> In this case __ticket_check_and_clear_slowpath() really needs to cmpxchg
>>> the whole .head_tail. Plus obviously more boring changes. This needs a
>>> separate patch even _if_ this can work.
>>
>> Correct, but apart from this, before doing xadd in unlock,
>> we would have to make sure lsb bit is cleared so that we can live with 1
>> bit overflow to tail which is unused. now either or both of head,tail
>> lsb bit may be set after unlock.
>
> Sorry, can't understand... could you spell?
>
> If TICKET_SLOWPATH_FLAG lives in .head arch_spin_unlock() could simply do
>
> 	head = xadd(&lock->tickets.head, TICKET_LOCK_INC);
>
> 	if (head & TICKET_SLOWPATH_FLAG)
> 		__ticket_unlock_kick(head);
>
> so it can't overflow to .tail?
>

You are right.
I totally forgot we can get rid of tail operations :)

>
> And we we do this, probably it makes sense to add something like
>
> 	bool tickets_equal(__ticket_t one, __ticket_t two)
> 	{
> 		return (one ^ two) & ~TICKET_SLOWPATH_FLAG;
> 	}
>

Very nice idea. I was tired of ~TICKET_SLOWPATH_FLAG usage all over in
the current (complex :)) implementation. These two suggestions helps
alot.

> and change kvm_lock_spinning() to use tickets_equal(tickets.head, want), plus
> it can have more users in asm/spinlock.h.