Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 14 Dec 2022 17:01:13 -0800 (PST)
From:   Hugh Dickins <hughd@google.com>
To:     Michael Roth <michael.roth@amd.com>
cc:     kvm@vger.kernel.org, linux-coco@lists.linux.dev,
        linux-mm@kvack.org, linux-crypto@vger.kernel.org, x86@kernel.org,
        linux-kernel@vger.kernel.org, tglx@linutronix.de, mingo@redhat.com,
        jroedel@suse.de, thomas.lendacky@amd.com, hpa@zytor.com,
        ardb@kernel.org, pbonzini@redhat.com, seanjc@google.com,
        vkuznets@redhat.com, wanpengli@tencent.com, jmattson@google.com,
        luto@kernel.org, dave.hansen@linux.intel.com, slp@redhat.com,
        pgonda@google.com, peterz@infradead.org,
        srinivas.pandruvada@linux.intel.com, rientjes@google.com,
        dovmurik@linux.ibm.com, tobin@ibm.com, bp@alien8.de,
        vbabka@suse.cz, kirill@shutemov.name, ak@linux.intel.com,
        tony.luck@intel.com, marcorr@google.com,
        sathyanarayanan.kuppuswamy@linux.intel.com, alpergun@google.com,
        dgilbert@redhat.com, jarkko@kernel.org, ashish.kalra@amd.com,
        harald@profian.com, Hugh Dickins <hughd@google.com>
Subject: Re: [PATCH RFC v7 21/64] x86/fault: fix handle_split_page_fault()
 to work with memfd backed pages
In-Reply-To: <20221214194056.161492-22-michael.roth@amd.com>
Message-ID: <7f2228c4-1586-2934-7b92-1a9d23b6046@google.com>
References: <20221214194056.161492-1-michael.roth@amd.com> <20221214194056.161492-22-michael.roth@amd.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Precedence: bulk

On Wed, 14 Dec 2022, Michael Roth wrote:
> From: Hugh Dickins <hughd@google.com>
> 
> When the address is backed by a memfd, the code to split the page does
> nothing more than remove the PMD from the page tables. So immediately
> install a PTE to ensure that any other pages in that 2MB region are
> brought back as in 4K pages.
> 
> Signed-off-by: Hugh Dickins <hughd@google.com>
> Cc: Hugh Dickins <hughd@google.com>
> Signed-off-by: Ashish Kalra <ashish.kalra@amd.com>
> Signed-off-by: Michael Roth <michael.roth@amd.com>

Hah, it's good to see this again, but it was "Suggested-by" me, not
"Signed-off-by" me.  And was a neat pragmatic one-liner workaround
for the immediate problem we had, but came with caveats.

The problem is that we have one wind blowing in the split direction,
and another wind (khugepaged) blowing in the collapse direction, and
who wins for how long depends on factors I've not fully got to grips
with (and is liable to differ between kernel releases).

Good and bad timing to see it.  I was just yesterday reviewing a patch
to the collapsing wind, which reminded me of an improvement yet to be
made there, thinking I'd like to try it sometime; but recallng that
someone somewhere relies on the splitting wind, and doesn't want the
collapsing wind to blow any harder - now you remind me who!

Bad timing in that I don't have any quick answer on the right thing
to do instead, and can't give it the thought it needs at the moment -
perhaps others can chime in more usefully.

Hugh

p.s. I don't know where "handle_split_page_fault" comes in, but
"x86/fault" in the subject looks wrong, since this appears to be
in generic code; and "memfd" seems inappropriate too, but perhaps you
have a situation where only memfds can reach handle_split_page_fault().

> ---
>  mm/memory.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/mm/memory.c b/mm/memory.c
> index e68da7e403c6..33c9020ba1f8 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -4999,6 +4999,11 @@ static vm_fault_t handle_pte_fault(struct vm_fault *vmf)
>  static int handle_split_page_fault(struct vm_fault *vmf)
>  {
>  	__split_huge_pmd(vmf->vma, vmf->pmd, vmf->address, false, NULL);
> +	/*
> +	 * Install a PTE immediately to ensure that any other pages in
> +	 * this 2MB region are brought back in as 4K pages.
> +	 */
> +	__pte_alloc(vmf->vma->vm_mm, vmf->pmd);
>  	return 0;
>  }
>  
> -- 
> 2.25.1