Re: [PATCH 2/4] mm: Send a single IPI to TLB flush multiple pages when unmapping

From: Mel Gorman
Date: Thu Apr 16 2015 - 05:19:50 EST


On Thu, Apr 16, 2015 at 05:29:55PM +0900, Minchan Kim wrote:
> On Thu, Apr 16, 2015 at 09:07:22AM +0100, Mel Gorman wrote:
> > On Thu, Apr 16, 2015 at 03:38:26PM +0900, Minchan Kim wrote:
> > > Hello Mel,
> > >
> > > On Wed, Apr 15, 2015 at 10:28:55PM +0100, Mel Gorman wrote:
> > > > On Wed, Apr 15, 2015 at 02:16:49PM -0700, Hugh Dickins wrote:
> > > > > On Wed, 15 Apr 2015, Rik van Riel wrote:
> > > > > > On 04/15/2015 06:42 AM, Mel Gorman wrote:
> > > > > > > An IPI is sent to flush remote TLBs when a page is unmapped that was
> > > > > > > recently accessed by other CPUs. There are many circumstances where this
> > > > > > > happens but the obvious one is kswapd reclaiming pages belonging to a
> > > > > > > running process as kswapd and the task are likely running on separate CPUs.
> > > > > > >
> > > > > > > On small machines, this is not a significant problem but as machine
> > > > > > > gets larger with more cores and more memory, the cost of these IPIs can
> > > > > > > be high. This patch uses a structure similar in principle to a pagevec
> > > > > > > to collect a list of PFNs and CPUs that require flushing. It then sends
> > > > > > > one IPI to flush the list of PFNs. A new TLB flush helper is required for
> > > > > > > this and one is added for x86. Other architectures will need to decide if
> > > > > > > batching like this is both safe and worth the memory overhead. Specifically
> > > > > > > the requirement is;
> > > > > > >
> > > > > > > If a clean page is unmapped and not immediately flushed, the
> > > > > > > architecture must guarantee that a write to that page from a CPU
> > > > > > > with a cached TLB entry will trap a page fault.
> > > > > > >
> > > > > > > This is essentially what the kernel already depends on but the window is
> > > > > > > much larger with this patch applied and is worth highlighting.
> > > > > >
> > > > > > This means we already have a (hard to hit?) data corruption
> > > > > > issue in the kernel. We can lose data if we unmap a writable
> > > > > > but not dirty pte from a file page, and the task writes before
> > > > > > we flush the TLB.
> > > > >
> > > > > I don't think so. IIRC, when the CPU needs to set the dirty bit,
> > > > > it doesn't just do that in its TLB entry, but has to fetch and update
> > > > > the actual pte entry - and at that point discovers it's no longer
> > > > > valid so traps, as Mel says.
> > > > >
> > > >
> > > > This is what I'm expecting i.e. clean->dirty transition is write-through
> > > > to the PTE which is now unmapped and it traps. I'm assuming there is an
> > > > architectural guarantee that it happens but could not find an explicit
> > > > statement in the docs. I'm hoping Dave or Andi can check with the relevant
> > > > people on my behalf.
> > >
> > > A dumb question. It's not related to your patch but MADV_FREE.
> > >
> > > clean->dirty transition is *atomic* as well as write-through?
> >
> > This is the TLB cache clean->dirty transition so it's not 100% clear what you
> > are asking. It both needs to be write-through and the TLB updates must happen
> > before the actual data write to cache or memory and it must be ordered.
>
> Sorry for not clear. I will try again.
>
> In try_to_unmap_one,
>
>
> pteval = ptep_clear_flush(vma, address, pte);
> {
> pte = ptep_get_and_clear(mm, address, ptep);
> <-------------- A application write on other CPU.
> flush_tlb_page(vma, address);
> }
>
> /* Move the dirty bit to the physical page now the pte is gone. */
> dirty = pte_dirty(pteval);
> if (dirty)
> set_page_dirty(page);
> ...
>
>
> In above, ptep_clear_flush just does xchg operation to make pte zero
> in ptep_get_and_clear and return old pte_val but didn't flush TLB yet.

Correct.

> Let's assume old pte_val doesn't have dirty bit(ie, it was clean).
> If application on other CPU does write the memory at the same time,
> what happens?

The comments describe the architectural guarantee I'm looking for. Dave
says he's asking the relevant people within Intel. I revised the comment
in the unreleased V2 so it reads

/*
* We clear the PTE but do not flush so potentially a remote
* CPU could still be writing to the page. If the entry was
* previously clean then the architecture must guarantee that
* a clear->dirty transition on a cached TLB entry is written
* through and traps if the PTE is unmapped. If the entry is
* already dirty then it's handled below by the
* pte_dirty check.
*/

> I mean (pte cleaning/return old) and (dirty bit setting by CPU itself)
> should be exclusive so application on another CPU should encounter
> page fault or we should see the dirty bit.
> Is it guaranteed?
>

This is the key question. I think "yes it must be" but Dave is going to
get the definite answer in the x86 case. Each architecture will need to
examine the issue separately.

--
Mel Gorman
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/