Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] mm: mmu_gather: remove __tlb_reset_range() for force
 flush
To:     Peter Zijlstra <peterz@infradead.org>,
        Nadav Amit <namit@vmware.com>
Cc:     Will Deacon <will.deacon@arm.com>,
        "jstancek@redhat.com" <jstancek@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        "stable@vger.kernel.org" <stable@vger.kernel.org>,
        "linux-mm@kvack.org" <linux-mm@kvack.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "aneesh.kumar@linux.vnet.ibm.com" <aneesh.kumar@linux.vnet.ibm.com>,
        "npiggin@gmail.com" <npiggin@gmail.com>,
        "minchan@kernel.org" <minchan@kernel.org>,
        Mel Gorman <mgorman@suse.de>
References: <1557264889-109594-1-git-send-email-yang.shi@linux.alibaba.com>
 <20190509083726.GA2209@brain-police>
 <20190509103813.GP2589@hirez.programming.kicks-ass.net>
 <F22533A7-016F-4506-809A-7E86BAF24D5A@vmware.com>
 <20190509182435.GA2623@hirez.programming.kicks-ass.net>
From:   Yang Shi <yang.shi@linux.alibaba.com>
Message-ID: <84720bb8-bf3d-8c10-d675-0670f13b2efc@linux.alibaba.com>
Date:   Thu, 9 May 2019 12:10:02 -0700
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.12; rv:52.0)
 Gecko/20100101 Thunderbird/52.7.0
MIME-Version: 1.0
In-Reply-To: <20190509182435.GA2623@hirez.programming.kicks-ass.net>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Content-Language: en-US
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 5/9/19 11:24 AM, Peter Zijlstra wrote:
> On Thu, May 09, 2019 at 05:36:29PM +0000, Nadav Amit wrote:
>>> On May 9, 2019, at 3:38 AM, Peter Zijlstra <peterz@infradead.org> wrote:
>>> diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c
>>> index 99740e1dd273..fe768f8d612e 100644
>>> --- a/mm/mmu_gather.c
>>> +++ b/mm/mmu_gather.c
>>> @@ -244,15 +244,20 @@ void tlb_finish_mmu(struct mmu_gather *tlb,
>>> 		unsigned long start, unsigned long end)
>>> {
>>> 	/*
>>> -	 * If there are parallel threads are doing PTE changes on same range
>>> -	 * under non-exclusive lock(e.g., mmap_sem read-side) but defer TLB
>>> -	 * flush by batching, a thread has stable TLB entry can fail to flush
>>> -	 * the TLB by observing pte_none|!pte_dirty, for example so flush TLB
>>> -	 * forcefully if we detect parallel PTE batching threads.
>>> +	 * Sensible comment goes here..
>>> 	 */
>>> -	if (mm_tlb_flush_nested(tlb->mm)) {
>>> -		__tlb_reset_range(tlb);
>>> -		__tlb_adjust_range(tlb, start, end - start);
>>> +	if (mm_tlb_flush_nested(tlb->mm) && !tlb->full_mm) {
>>> +		/*
>>> +		 * Since we're can't tell what we actually should have
>>> +		 * flushed flush everything in the given range.
>>> +		 */
>>> +		tlb->start = start;
>>> +		tlb->end = end;
>>> +		tlb->freed_tables = 1;
>>> +		tlb->cleared_ptes = 1;
>>> +		tlb->cleared_pmds = 1;
>>> +		tlb->cleared_puds = 1;
>>> +		tlb->cleared_p4ds = 1;
>>> 	}
>>>
>>> 	tlb_flush_mmu(tlb);
>> As a simple optimization, I think it is possible to hold multiple nesting
>> counters in the mm, similar to tlb_flush_pending, for freed_tables,
>> cleared_ptes, etc.
>>
>> The first time you set tlb->freed_tables, you also atomically increase
>> mm->tlb_flush_freed_tables. Then, in tlb_flush_mmu(), you just use
>> mm->tlb_flush_freed_tables instead of tlb->freed_tables.
> That sounds fraught with races and expensive; I would much prefer to not
> go there for this arguably rare case.
>
> Consider such fun cases as where CPU-0 sees and clears a PTE, CPU-1
> races and doesn't see that PTE. Therefore CPU-0 sets and counts
> cleared_ptes. Then if CPU-1 flushes while CPU-0 is still in mmu_gather,
> it will see cleared_ptes count increased and flush that granularity,
> OTOH if CPU-1 flushes after CPU-0 completes, it will not and potentiall
> miss an invalidate it should have had.
>
> This whole concurrent mmu_gather stuff is horrible.
>
>    /me ponders more....
>
> So I think the fundamental race here is this:
>
> 	CPU-0				CPU-1
>
> 	tlb_gather_mmu(.start=1,	tlb_gather_mmu(.start=2,
> 		       .end=3);			       .end=4);
>
> 	ptep_get_and_clear_full(2)
> 	tlb_remove_tlb_entry(2);
> 	__tlb_remove_page();
> 					if (pte_present(2)) // nope
>
> 					tlb_finish_mmu();
>
> 					// continue without TLBI(2)
> 					// whoopsie
>
> 	tlb_finish_mmu();
> 	  tlb_flush()		->	TLBI(2)

I'm not quite sure if this is the case Jan really met. But, according to 
his test, once correct tlb->freed_tables and tlb->cleared_* are set, his 
test works well.

>
>
> And we can fix that by having tlb_finish_mmu() sync up. Never let a
> concurrent tlb_finish_mmu() complete until all concurrenct mmu_gathers
> have completed.

Not sure if this will scale well.

>
> This should not be too hard to make happen.