From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id DCC18E82CDA for ; Wed, 27 Sep 2023 20:43:13 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229816AbjI0UnN (ORCPT ); Wed, 27 Sep 2023 16:43:13 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:36724 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229639AbjI0UnM (ORCPT ); Wed, 27 Sep 2023 16:43:12 -0400 Received: from mail-yw1-x1135.google.com (mail-yw1-x1135.google.com [IPv6:2607:f8b0:4864:20::1135]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 126D5BF for ; Wed, 27 Sep 2023 13:43:10 -0700 (PDT) Received: by mail-yw1-x1135.google.com with SMTP id 00721157ae682-59f4bc88f9fso117074557b3.2 for ; Wed, 27 Sep 2023 13:43:10 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20230601; t=1695847389; x=1696452189; darn=vger.kernel.org; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:from:to:cc:subject:date :message-id:reply-to; bh=yjpcdsZmpsqxOy2d6oqu0Zcfs5V1qi2ngVN6expqaEM=; b=xOgWxvPL+KgdHHjS+HMkzfP7yeGEaix+Zill5GmZVNY4FD6ptykQCkbuEmFudJLG9J bbwOobKGphaM3yofwdj+2tTOYXGZe2+TjY2gYeay4no66ek6u/y3QGamu0psn7JGRZNQ K8by8YJgWcMZzlO0ZRz1muBpmoXHbRe0KuWgPHzF7S2yZ3epM5pCvcnOMYzlUgnwQxKU WiAl8t01jqUJKZ4XwMa0jv+408ZbYq7OvMt3OvZdoiXIdST9KlGwBjFxcddgFz3417mw NeEG7G1r7mhkytCd9tZxHb9N9l234hxBVv6F7ylSslnjSHnLCw8P/FCUWTcsE4t0IffT wQAg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1695847389; x=1696452189; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=yjpcdsZmpsqxOy2d6oqu0Zcfs5V1qi2ngVN6expqaEM=; b=ZMTgea4UKAXiUxjfs9E5jkxzSlpiJxFEXuqVbqWOhfTpuxbp1yWUB8oDPDG4deoFqm 9codrooNrztpUDr5guHaLQge1De1utHApxiFCGVE0JqYPs58QBOcIHoQVOZcHufTG8uQ qd+8gw5RweesrdyfbR2QIhUHlx52TUy37dTjagifk/T6bjI+LS/o66Ka1BRx5YJ0TDi/ +ckTK70xp3vo2CQcBoeQkGNFJ8t73DMCzEu+ArNF1DjDsXVseJNn39rTSUI/J4lrESCL gMZyjYn1zgr5curdN4fQ7QNZMjW8wq9z66fPHdPY1MKNDx6lsTCav57ajXmXG5Mjnhn9 OcmA== X-Gm-Message-State: AOJu0YydF0vU2CKnXcawtwholE//v7l4o66JdxbGizyxrcidwBfywZQI lljrkniufAZHVPJ/LQtue7NwdDQqeFlrxtUh5JrSKA== X-Google-Smtp-Source: AGHT+IHbo4Hykuc6zuEaClz11a4Sav0sQWQMdUu53vQbEXMyiJQn85B+L29xbafKZEJC1C9iyT0q5ZDyxPKte7jrxeU= X-Received: by 2002:a0d:d60f:0:b0:56d:3b91:7e78 with SMTP id y15-20020a0dd60f000000b0056d3b917e78mr3517738ywd.20.1695847388923; Wed, 27 Sep 2023 13:43:08 -0700 (PDT) MIME-Version: 1.0 References: <20230923013148.1390521-1-surenb@google.com> <20230923013148.1390521-3-surenb@google.com> In-Reply-To: From: Suren Baghdasaryan Date: Wed, 27 Sep 2023 13:42:54 -0700 Message-ID: Subject: Re: [PATCH v2 2/3] userfaultfd: UFFDIO_REMAP uABI To: Jann Horn Cc: akpm@linux-foundation.org, viro@zeniv.linux.org.uk, brauner@kernel.org, shuah@kernel.org, aarcange@redhat.com, lokeshgidra@google.com, peterx@redhat.com, david@redhat.com, hughd@google.com, mhocko@suse.com, axelrasmussen@google.com, rppt@kernel.org, willy@infradead.org, Liam.Howlett@oracle.com, zhangpeng362@huawei.com, bgeffon@google.com, kaleshsingh@google.com, ngeoffray@google.com, jdduke@google.com, linux-mm@kvack.org, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org, kernel-team@android.com Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-fsdevel@vger.kernel.org On Wed, Sep 27, 2023 at 1:04=E2=80=AFPM Jann Horn wrote: > > On Wed, Sep 27, 2023 at 8:08=E2=80=AFPM Suren Baghdasaryan wrote: > > On Wed, Sep 27, 2023 at 5:47=E2=80=AFAM Jann Horn wr= ote: > > > On Sat, Sep 23, 2023 at 3:31=E2=80=AFAM Suren Baghdasaryan wrote: > > > > From: Andrea Arcangeli > > > > > > > > This implements the uABI of UFFDIO_REMAP. > > > > > > > > Notably one mode bitflag is also forwarded (and in turn known) by t= he > > > > lowlevel remap_pages method. > > > > > > > > Signed-off-by: Andrea Arcangeli > > > > Signed-off-by: Suren Baghdasaryan > [...] > > > > + /* > > > > + * folio_referenced walks the anon_vma chai= n > > > > + * without the folio lock. Serialize agains= t it with > > > > + * the anon_vma lock, the folio lock is not= enough. > > > > + */ > > > > + src_anon_vma =3D folio_get_anon_vma(src_fol= io); > > > > + if (!src_anon_vma) { > > > > + /* page was unmapped from under us = */ > > > > + err =3D -EAGAIN; > > > > + goto out; > > > > + } > > > > + if (!anon_vma_trylock_write(src_anon_vma)) = { > > > > + pte_unmap(&orig_src_pte); > > > > + pte_unmap(&orig_dst_pte); > > > > + src_pte =3D dst_pte =3D NULL; > > > > + /* now we can block and wait */ > > > > + anon_vma_lock_write(src_anon_vma); > > > > + goto retry; > > > > + } > > > > + } > > > > > > So at this point we have: > > > > > > - the current src_pte > > > - some referenced+locked src_folio that used to be mapped exclusivel= y > > > at src_addr > > > - (the anon_vma associated with the src_folio) > > > > > > > + err =3D remap_anon_pte(dst_mm, src_mm, dst_vma, sr= c_vma, > > > > + dst_addr, src_addr, dst_pte, s= rc_pte, > > > > + orig_dst_pte, orig_src_pte, > > > > + dst_ptl, src_ptl, src_folio); > > > > > > And then this will, without touching folio mapcounts/refcounts, delet= e > > > the current PTE at src_addr, and create a PTE at dst_addr pointing to > > > the old src_folio, leading to incorrect refcounts/mapcounts? > > > > I assume this still points to the missing previous_src_pte check > > discussed in the previous comments. Is that correct or is there yet > > another issue? > > This is still referring to the missing previous_src_pte check. > > > > > > > > + } else { > > > [...] > > > > + } > > > > + > > > > +out: > > > > + if (src_anon_vma) { > > > > + anon_vma_unlock_write(src_anon_vma); > > > > + put_anon_vma(src_anon_vma); > > > > + } > > > > + if (src_folio) { > > > > + folio_unlock(src_folio); > > > > + folio_put(src_folio); > > > > + } > > > > + if (dst_pte) > > > > + pte_unmap(dst_pte); > > > > + if (src_pte) > > > > + pte_unmap(src_pte); > > > > + mmu_notifier_invalidate_range_end(&range); > > > > + > > > > + return err; > > > > +} > > > [...] > > > > +ssize_t remap_pages(struct mm_struct *dst_mm, struct mm_struct *sr= c_mm, > > > > + unsigned long dst_start, unsigned long src_star= t, > > > > + unsigned long len, __u64 mode) > > > > +{ > > > > + struct vm_area_struct *src_vma, *dst_vma; > > > > + unsigned long src_addr, dst_addr; > > > > + pmd_t *src_pmd, *dst_pmd; > > > > + long err =3D -EINVAL; > > > > + ssize_t moved =3D 0; > > > > + > > > > + /* > > > > + * Sanitize the command parameters: > > > > + */ > > > > + BUG_ON(src_start & ~PAGE_MASK); > > > > + BUG_ON(dst_start & ~PAGE_MASK); > > > > + BUG_ON(len & ~PAGE_MASK); > > > > + > > > > + /* Does the address range wrap, or is the span zero-sized? = */ > > > > + BUG_ON(src_start + len <=3D src_start); > > > > + BUG_ON(dst_start + len <=3D dst_start); > > > > + > > > > + /* > > > > + * Because these are read sempahores there's no risk of loc= k > > > > + * inversion. > > > > + */ > > > > + mmap_read_lock(dst_mm); > > > > + if (dst_mm !=3D src_mm) > > > > + mmap_read_lock(src_mm); > > > > + > > > > + /* > > > > + * Make sure the vma is not shared, that the src and dst re= map > > > > + * ranges are both valid and fully within a single existing > > > > + * vma. > > > > + */ > > > > + src_vma =3D find_vma(src_mm, src_start); > > > > + if (!src_vma || (src_vma->vm_flags & VM_SHARED)) > > > > + goto out; > > > > + if (src_start < src_vma->vm_start || > > > > + src_start + len > src_vma->vm_end) > > > > + goto out; > > > > + > > > > + dst_vma =3D find_vma(dst_mm, dst_start); > > > > + if (!dst_vma || (dst_vma->vm_flags & VM_SHARED)) > > > > + goto out; > > > > + if (dst_start < dst_vma->vm_start || > > > > + dst_start + len > dst_vma->vm_end) > > > > + goto out; > > > > + > > > > + err =3D validate_remap_areas(src_vma, dst_vma); > > > > + if (err) > > > > + goto out; > > > > + > > > > + for (src_addr =3D src_start, dst_addr =3D dst_start; > > > > + src_addr < src_start + len;) { > > > > + spinlock_t *ptl; > > > > + pmd_t dst_pmdval; > > > > + unsigned long step_size; > > > > + > > > > + BUG_ON(dst_addr >=3D dst_start + len); > > > > + /* > > > > + * Below works because anonymous area would not hav= e a > > > > + * transparent huge PUD. If file-backed support is = added, > > > > + * that case would need to be handled here. > > > > + */ > > > > + src_pmd =3D mm_find_pmd(src_mm, src_addr); > > > > + if (unlikely(!src_pmd)) { > > > > + if (!(mode & UFFDIO_REMAP_MODE_ALLOW_SRC_HO= LES)) { > > > > + err =3D -ENOENT; > > > > + break; > > > > + } > > > > + src_pmd =3D mm_alloc_pmd(src_mm, src_addr); > > > > + if (unlikely(!src_pmd)) { > > > > + err =3D -ENOMEM; > > > > + break; > > > > + } > > > > + } > > > > + dst_pmd =3D mm_alloc_pmd(dst_mm, dst_addr); > > > > + if (unlikely(!dst_pmd)) { > > > > + err =3D -ENOMEM; > > > > + break; > > > > + } > > > > + > > > > + dst_pmdval =3D pmdp_get_lockless(dst_pmd); > > > > + /* > > > > + * If the dst_pmd is mapped as THP don't override i= t and just > > > > + * be strict. If dst_pmd changes into TPH after thi= s check, the > > > > + * remap_pages_huge_pmd() will detect the change an= d retry > > > > + * while remap_pages_pte() will detect the change a= nd fail. > > > > + */ > > > > + if (unlikely(pmd_trans_huge(dst_pmdval))) { > > > > + err =3D -EEXIST; > > > > + break; > > > > + } > > > > + > > > > + ptl =3D pmd_trans_huge_lock(src_pmd, src_vma); > > > > + if (ptl && !pmd_trans_huge(*src_pmd)) { > > > > + spin_unlock(ptl); > > > > + ptl =3D NULL; > > > > + } > > > > > > This still looks wrong - we do still have to split_huge_pmd() > > > somewhere so that remap_pages_pte() works. > > > > Hmm, I guess this extra check is not even needed... > > Hm, and instead we'd bail at the pte_offset_map_nolock() in > remap_pages_pte()? I guess that's unusual but works... Yes, that's what I was thinking but I agree, that seems fragile. Maybe just bail out early if (ptl && !pmd_trans_huge())? > > (It would be a thing to look out for if anyone tried to backport this, > since the checks in pte_offset_map_nolock() were only introduced in > 6.5, but idk if anyone's doing that)