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

[Hugh Dickins <hughd@google.com>]

On Thu, 16 Apr 2015, Mel Gorman wrote:
>  
>  	/* Move the dirty bit to the physical page now the pte is gone. */
> -	if (pte_dirty(pteval))
> +	if (pte_dirty(pteval)) {
> +		/*
> +		 * If the PTE was dirty then the TLB must be flushed before
> +		 * the page is unlocked as IO can start in parallel. Without
> +		 * the flush, writes could still happen and data would be
> +		 * potentially lost.
> +		 */
> +		if (deferred)
> +			flush_tlb_page(vma, address);

Okay, yes, that should deal with it; and you're probably right that the
safe pte_dirty !pte_write case is too uncommon to be worth another test.

But it would be better to batch even in the pte_dirty case: noting that
it has occurred in the tlb_ubc, then if so, doing try_to_unmap_flush()
before leaving try_to_unmap().

Particularly as you have already set_tlb_ubc_flush_pending() above,
so shrink_lruvec() may then follow with an unnecessary flush; though
I guess a little rearrangement here could stop that.

> +
>  		set_page_dirty(page);
> +	}
>  
>  	/* Update high watermark before we lower rss */
>  	update_hiwater_rss(mm);

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>