Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 22 Sep 2021 15:33:29 -0400
From:   Peter Xu <peterx@redhat.com>
To:     linux-kernel@vger.kernel.org, linux-mm@kvack.org
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Axel Rasmussen <axelrasmussen@google.com>,
        Hugh Dickins <hughd@google.com>,
        Nadav Amit <nadav.amit@gmail.com>
Subject: Re: [PATCH] mm/khugepaged: Detecting uffd-wp vma more efficiently
Message-ID: <YUuFCcuHn74STTlZ@t490s>
References: <20210922175156.130228-1-peterx@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20210922175156.130228-1-peterx@redhat.com>
Precedence: bulk

On Wed, Sep 22, 2021 at 01:51:56PM -0400, Peter Xu wrote:
> Axel: as I asked in the other thread, please help check whether minor mode will
> work properly with shmem thp enabled.  If not, I feel like this patch could be
> part of that effort at last, but it's also possible that I missed something.

Hmm, this seems to be a false-alarm too.

UFFDIO_CONTINUE is fine with thp because shmem_getpage() only returns small
pages.

Khugepaged is fine too on merging small shmem pages into thps, the same reason
as uffd-wp: it zaps ptes only, so previous pte_none() ptes will keep the same,
it makes sure all old pte_none ptes will not continue until a UFFDIO_CONTINUE.

The only last problem is when khugepaged merged small ptes into a thp, then it
could zap ptes even if they existed before.  So for minor mode fault, the uffd
service thread needs to be prepared for false positive messages.

That shouldn't be a problem either, because file-backed memory pgtables are
unstable and prone to lost, so uffd minor fault handler should always be
prepared for false positives anyway..

In summary: please feel free to ignore the above note, and sorry for the noise.

-- 
Peter Xu