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 kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by smtp.lore.kernel.org (Postfix) with ESMTP id 09B0BC87FCB for ; Tue, 5 Aug 2025 23:41:33 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id 597D18E0006; Tue, 5 Aug 2025 19:41:33 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id 56EB98E0001; Tue, 5 Aug 2025 19:41:33 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id 4848C8E0006; Tue, 5 Aug 2025 19:41:33 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0016.hostedemail.com [216.40.44.16]) by kanga.kvack.org (Postfix) with ESMTP id 378F98E0001 for ; Tue, 5 Aug 2025 19:41:33 -0400 (EDT) Received: from smtpin30.hostedemail.com (a10.router.float.18 [10.200.18.1]) by unirelay02.hostedemail.com (Postfix) with ESMTP id 92AB3135686 for ; Tue, 5 Aug 2025 23:41:32 +0000 (UTC) X-FDA: 83744328024.30.6921813 Received: from mail-qt1-f175.google.com (mail-qt1-f175.google.com [209.85.160.175]) by imf04.hostedemail.com (Postfix) with ESMTP id 9E51B40008 for ; Tue, 5 Aug 2025 23:41:30 +0000 (UTC) Authentication-Results: imf04.hostedemail.com; dkim=pass header.d=google.com header.s=20230601 header.b=maLubDS0; spf=pass (imf04.hostedemail.com: domain of surenb@google.com designates 209.85.160.175 as permitted sender) smtp.mailfrom=surenb@google.com; dmarc=pass (policy=reject) header.from=google.com ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1754437290; h=from:from:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references:dkim-signature; bh=f+S3bmPXejwFO1KMmyrSrq+aYMwqbrRLmpkvUdxnslc=; b=L0xw1lLoOMQK0Udcr8r8kmNrU8C/DUFPDgTAQnayyWF73pUbY2yxoF4tNWaDisVBo+ZN3Q boREClXqB32eOlNQwFnbdDUSOY8X0tA6gcsjY4BEcvxYj6zGkq/oYuEFnzeN3nfBboukaR aOdT8vrG3yiMR3DB8C7pGHqzkQGRzuM= ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1754437290; a=rsa-sha256; cv=none; b=jvrNzgPotbcETJedesgYp60VeNNZZl+lIVhZWo12iTDU32XoafcwNp/zsOSgTLADzHbE+U csX0mapSWihqrxZOgoYlgPPIfbLWk/GShg7hbv5brtI5TIcTHnNkcoMg1IU+8IOwEm3NSC aFrDd2W98FFx7Jw5M+ySdR/vyHZZ1dE= ARC-Authentication-Results: i=1; imf04.hostedemail.com; dkim=pass header.d=google.com header.s=20230601 header.b=maLubDS0; spf=pass (imf04.hostedemail.com: domain of surenb@google.com designates 209.85.160.175 as permitted sender) smtp.mailfrom=surenb@google.com; dmarc=pass (policy=reject) header.from=google.com Received: by mail-qt1-f175.google.com with SMTP id d75a77b69052e-4b0673b0a7cso250071cf.0 for ; Tue, 05 Aug 2025 16:41:30 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20230601; t=1754437290; x=1755042090; darn=kvack.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=f+S3bmPXejwFO1KMmyrSrq+aYMwqbrRLmpkvUdxnslc=; b=maLubDS0X9Ip6j5cgRLplEYav0nvG+7LTaIr6pPWQJrJ59mBITSSlOq+/O5rS3i19w 597ftOzu84YWoFUR+IvulyLH4ZNEyPbj8udfZOvqz87VXVrb4vVJl4iMQXiD++61Rch2 2lL4u5BgTcbFw7723yZBb4QSTza1qyyIrnkoxW0A9Evcp8kmrkSocy6FmmR67cywb3rG yDIN+3SJPLdvwzuqYbqQS75bTCwhynTeCwWosjT89BB0WZCsfBPLr3izNqkMibL6eE6t Ah7FHi15G+NGUfILhxNbUVLrGB7WvkVZT6R4H0TrtlBn9y2MeTJoQnhAFK9wN+ywnxhH JgaA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1754437290; x=1755042090; 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=f+S3bmPXejwFO1KMmyrSrq+aYMwqbrRLmpkvUdxnslc=; b=pFkp72p/4+pu83CK9XqWverMrmb3om4LerPc5AOKHlLcGOfjSurdEMJaoR9Imkpb7d vrwcDLGxstcubBy0bf9AWDMO74yJ9XnXWUvy8BhMWJVkcXSJ1qAc6WuWsqpayjmgOSVL t9VheImozcYEd6Bhod2+ExqTN4mW+nQkhqz3CQ4VEXP93mLw9Fdes5K1lUqu5QB6oGlK Nnfgxwfd/RMaTSdlSNPK/Z1XsJl3ebOopNLg3eyXybig7fVkih4gL9hLDFvQJ9tJz09C DFI2kY7fCPa1gOfR0o1XT2bQCRjqpyZ6CukNwMjvAxmXpt7iA8af1OSZju4HxvTDAG8j wgxw== X-Forwarded-Encrypted: i=1; AJvYcCVvPppB2r4Fz2VpMEuu/OBJeIrD0R64IesFZ7i7wEjtTyuo+S16IaJZeTA31hXSgTc6X5+nlBh2+w==@kvack.org X-Gm-Message-State: AOJu0YxcDWGIi+gqlteygqdG+il/C3Zi76EkMk18KgGoAeXfhlYTK9Wd 8vwbRW1ueiWoWlz0Xkll8n7PVlEySsxfo5JEAAANibAXzJ3gky1X5mGzs1ktSsUPZaGXgSovpv5 Tb+n0icaW4w9Nk8LawFdjLBucufLqG8dY0UaRWNXJ X-Gm-Gg: ASbGncsxqMeI/AOHdCb2mfqHN9yM0c1NLO4Up2hHB8bcvxRpY+YBL++mxLSyNyD7qzB IVfpVZjENcE31EBNmBlggNH4dxEnypMgEnaBg0Ss3T17HSsmS9zzWnFkNgj+ht5WEOrinxjH8S3 0jvRiFqrvxKi3FbbEvXacKCSD27Hm4wIDXGuYSDSfoKdejGK2vEjROZ9eONY3lvz+GBnnCotEC3 Z8l1PAJLQXv3vlqxtteLYdDcSxLRxT6Gb8bG2IRuKGJeRSy X-Google-Smtp-Source: AGHT+IEEmBiUQbr0GQVcfi3phUF9vukRyuZfQf9xofLE27keKB1H1qrGIdFCVd6QbGW9E6OxiTrII591nTP7vFtRWWg= X-Received: by 2002:ac8:7e90:0:b0:4ae:d2cc:ad51 with SMTP id d75a77b69052e-4b09261783cmr1338171cf.1.1754437289233; Tue, 05 Aug 2025 16:41:29 -0700 (PDT) MIME-Version: 1.0 References: In-Reply-To: From: Suren Baghdasaryan Date: Tue, 5 Aug 2025 16:41:18 -0700 X-Gm-Features: Ac12FXyFn6ZM5k32WyQsCctB41kkckIUGLUX8rOqsdud2EjylewU1LQBkK1P12w Message-ID: Subject: Re: [PATCH v2 1/1] userfaultfd: fix a crash when UFFDIO_MOVE handles a THP hole To: Peter Xu Cc: David Hildenbrand , akpm@linux-foundation.org, aarcange@redhat.com, lokeshgidra@google.com, linux-mm@kvack.org, linux-kernel@vger.kernel.org, syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com, stable@vger.kernel.org Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Rspamd-Server: rspam03 X-Rspamd-Queue-Id: 9E51B40008 X-Stat-Signature: ct6yno9rpjmfabqju1mas1oj6hc8ziwp X-Rspam-User: X-HE-Tag: 1754437290-784335 X-HE-Meta: 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 sxyGLfo5 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 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: List-Subscribe: List-Unsubscribe: On Tue, Aug 5, 2025 at 1:39=E2=80=AFPM Suren Baghdasaryan wrote: > > On Tue, Aug 5, 2025 at 7:57=E2=80=AFAM Suren Baghdasaryan wrote: > > > > On Tue, Aug 5, 2025 at 7:39=E2=80=AFAM Peter Xu wro= te: > > > > > > On Mon, Aug 04, 2025 at 07:55:42AM -0700, Suren Baghdasaryan wrote: > > > > On Fri, Aug 1, 2025 at 5:32=E2=80=AFPM Peter Xu = wrote: > > > > > > > > > > On Fri, Aug 01, 2025 at 07:30:02PM +0000, Suren Baghdasaryan wrot= e: > > > > > > On Fri, Aug 1, 2025 at 6:21=E2=80=AFPM Peter Xu wrote: > > > > > > > > > > > > > > On Fri, Aug 01, 2025 at 05:45:10PM +0000, Suren Baghdasaryan = wrote: > > > > > > > > On Fri, Aug 1, 2025 at 5:13=E2=80=AFPM Peter Xu wrote: > > > > > > > > > > > > > > > > > > On Fri, Aug 01, 2025 at 09:41:31AM -0700, Suren Baghdasar= yan wrote: > > > > > > > > > > On Fri, Aug 1, 2025 at 9:23=E2=80=AFAM Peter Xu wrote: > > > > > > > > > > > > > > > > > > > > > > On Fri, Aug 01, 2025 at 08:28:38AM -0700, Suren Baghd= asaryan wrote: > > > > > > > > > > > > On Fri, Aug 1, 2025 at 7:16=E2=80=AFAM Peter Xu wrote: > > > > > > > > > > > > > > > > > > > > > > > > > > On Fri, Aug 01, 2025 at 09:21:30AM +0200, David H= ildenbrand wrote: > > > > > > > > > > > > > > On 31.07.25 17:44, Suren Baghdasaryan wrote: > > > > > > > > > > > > > > > > > > > > > > > > > > > > Hi! > > > > > > > > > > > > > > > > > > > > > > > > > > > > Did you mean in you patch description: > > > > > > > > > > > > > > > > > > > > > > > > > > > > "userfaultfd: fix a crash in UFFDIO_MOVE with s= ome non-present PMDs" > > > > > > > > > > > > > > > > > > > > > > > > > > > > Talking about THP holes is very very confusing. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > When UFFDIO_MOVE is used with UFFDIO_MOVE_MOD= E_ALLOW_SRC_HOLES and it > > > > > > > > > > > > > > > encounters a non-present THP, it fails to pro= perly recognize an unmapped > > > > > > > > > > > > > > > > > > > > > > > > > > > > You mean a "non-present PMD that is not a migra= tion entry". > > > > > > > > > > > > > > > > > > > > > > > > > > > > > hole and tries to access a non-existent folio= , resulting in > > > > > > > > > > > > > > > a crash. Add a check to skip non-present THPs= . > > > > > > > > > > > > > > > > > > > > > > > > > > > > That makes sense. The code we have after this p= atch is rather complicated > > > > > > > > > > > > > > and hard to read. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Fixes: adef440691ba ("userfaultfd: UFFDIO_MOV= E uABI") > > > > > > > > > > > > > > > Reported-by: syzbot+b446dbe27035ef6bd6c2@syzk= aller.appspotmail.com > > > > > > > > > > > > > > > Closes: https://lore.kernel.org/all/68794b5c.= a70a0220.693ce.0050.GAE@google.com/ > > > > > > > > > > > > > > > Signed-off-by: Suren Baghdasaryan > > > > > > > > > > > > > > > Cc: stable@vger.kernel.org > > > > > > > > > > > > > > > --- > > > > > > > > > > > > > > > Changes since v1 [1] > > > > > > > > > > > > > > > - Fixed step size calculation, per Lokesh Gid= ra > > > > > > > > > > > > > > > - Added missing check for UFFDIO_MOVE_MODE_AL= LOW_SRC_HOLES, per Lokesh Gidra > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > [1] https://lore.kernel.org/all/2025073017073= 3.3829267-1-surenb@google.com/ > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > mm/userfaultfd.c | 45 +++++++++++++++++++++= ++++++++---------------- > > > > > > > > > > > > > > > 1 file changed, 29 insertions(+), 16 deleti= ons(-) > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > diff --git a/mm/userfaultfd.c b/mm/userfaultf= d.c > > > > > > > > > > > > > > > index cbed91b09640..b5af31c22731 100644 > > > > > > > > > > > > > > > --- a/mm/userfaultfd.c > > > > > > > > > > > > > > > +++ b/mm/userfaultfd.c > > > > > > > > > > > > > > > @@ -1818,28 +1818,41 @@ ssize_t move_pages(st= ruct userfaultfd_ctx *ctx, unsigned long dst_start, > > > > > > > > > > > > > > > ptl =3D pmd_trans_huge_lock(src_p= md, src_vma); > > > > > > > > > > > > > > > if (ptl) { > > > > > > > > > > > > > > > - /* Check if we can move t= he pmd without splitting it. */ > > > > > > > > > > > > > > > - if (move_splits_huge_pmd(= dst_addr, src_addr, src_start + len) || > > > > > > > > > > > > > > > - !pmd_none(dst_pmdval)= ) { > > > > > > > > > > > > > > > - struct folio *fol= io =3D pmd_folio(*src_pmd); > > > > > > > > > > > > > > > + if (pmd_present(*src_pmd)= || is_pmd_migration_entry(*src_pmd)) { > > > > > > > > > > > > > > > > > > > > > > > > > > [1] > > > > > > > > > > > > > > > > > > > > > > > > > > > > + /* Check if we ca= n move the pmd without splitting it. */ > > > > > > > > > > > > > > > + if (move_splits_h= uge_pmd(dst_addr, src_addr, src_start + len) || > > > > > > > > > > > > > > > + !pmd_none(dst= _pmdval)) { > > > > > > > > > > > > > > > + if (pmd_p= resent(*src_pmd)) { > > > > > > > > > > > > > > > > > > > > > > [2] > > > > > > > > > > > > > > > > > > > > > > > > > > + s= truct folio *folio =3D pmd_folio(*src_pmd); > > > > > > > > > > > > > > > > > > > > > > [3] > > > > > > > > > > > > > > > > > > > > > > > > > > + > > > > > > > > > > > > > > > + i= f (!folio || (!is_huge_zero_folio(folio) && > > > > > > > > > > > > > > > + = !PageAnonExclusive(&folio->page))) { > > > > > > > > > > > > > > > + = spin_unlock(ptl); > > > > > > > > > > > > > > > + = err =3D -EBUSY; > > > > > > > > > > > > > > > + = break; > > > > > > > > > > > > > > > + } > > > > > > > > > > > > > > > + } > > > > > > > > > > > > > > > > > > > > > > > > > > > > ... in particular that. Is there some way to ma= ke this code simpler / easier > > > > > > > > > > > > > > to read? Like moving that whole last folio-chec= k thingy into a helper? > > > > > > > > > > > > > > > > > > > > > > > > > > One question might be relevant is, whether the ch= eck above [1] can be > > > > > > > > > > > > > dropped. > > > > > > > > > > > > > > > > > > > > > > > > > > The thing is __pmd_trans_huge_lock() does double = check the pmd to be !none > > > > > > > > > > > > > before returning the ptl. I didn't follow closel= y on the recent changes on > > > > > > > > > > > > > mm side on possible new pmd swap entries, if migr= ation is the only possible > > > > > > > > > > > > > one then it looks like [1] can be avoided. > > > > > > > > > > > > > > > > > > > > > > > > Hi Peter, > > > > > > > > > > > > is_swap_pmd() check in __pmd_trans_huge_lock() allo= ws for (!pmd_none() > > > > > > > > > > > > && !pmd_present()) PMD to pass and that's when this= crash is hit. > > > > > > > > > > > > > > > > > > > > > > First for all, thanks for looking into the issue with= Lokesh; I am still > > > > > > > > > > > catching up with emails after taking weeks off. > > > > > > > > > > > > > > > > > > > > > > I didn't yet read into the syzbot report, but I thoug= ht the bug was about > > > > > > > > > > > referencing the folio on top of a swap entry after re= ading your current > > > > > > > > > > > patch, which has: > > > > > > > > > > > > > > > > > > > > > > if (move_splits_huge_pmd(dst_addr, src_addr, = src_start + len) || > > > > > > > > > > > !pmd_none(dst_pmdval)) { > > > > > > > > > > > struct folio *folio =3D pmd_folio(*sr= c_pmd); <---- > > > > > > > > > > > > > > > > > > > > > > Here looks like *src_pmd can be a migration entry. Is= my understanding > > > > > > > > > > > correct? > > > > > > > > > > > > > > > > > > > > Correct. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > If we drop the check at [1] then the path that take= s us to > > > > > > > > > > > > > > > > > > > > > > If my above understanding is correct, IMHO it should = be [2] above that > > > > > > > > > > > makes sure the reference won't happen on a swap entry= , not necessarily [1]? > > > > > > > > > > > > > > > > > > > > Yes, in case of migration entry this is what protects u= s. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > split_huge_pmd() will bail out inside split_huge_pm= d_locked() with no > > > > > > > > > > > > indication that split did not happen. Afterwards we= will retry > > > > > > > > > > > > > > > > > > > > > > So we're talking about the case where it's a swap pmd= entry, right? > > > > > > > > > > > > > > > > > > > > Hmm, my understanding is that it's being treated as a s= wap entry but > > > > > > > > > > in reality is not. I thought THPs are always split befo= re they get > > > > > > > > > > swapped, no? > > > > > > > > > > > > > > > > > > Yes they should be split, afaiu. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Could you elaborate why the split would fail? > > > > > > > > > > > > > > > > > > > > Just looking at the code, split_huge_pmd_locked() check= s for > > > > > > > > > > (pmd_trans_huge(*pmd) || is_pmd_migration_entry(*pmd)). > > > > > > > > > > pmd_trans_huge() is false if !pmd_present() and it's no= t a migration > > > > > > > > > > entry, so __split_huge_pmd_locked() will be skipped. > > > > > > > > > > > > > > > > > > Here might be the major part of where confusion came from= : I thought it > > > > > > > > > must be a migration pmd entry to hit the issue, so it's n= ot? > > > > > > > > > > > > > > > > > > I checked the code just now: > > > > > > > > > > > > > > > > > > __handle_mm_fault: > > > > > > > > > if (unlikely(is_swap_pmd(vmf.orig_pmd))) = { > > > > > > > > > VM_BUG_ON(thp_migration_supported= () && > > > > > > > > > !is_pmd_migrati= on_entry(vmf.orig_pmd)); > > > > > > > > > > > > > > > > > > So IIUC pmd migration entry is still the only possible wa= y to have a swap > > > > > > > > > entry. It doesn't look like we have "real" swap entries = for PMD (which can > > > > > > > > > further points to some swapfiles)? > > > > > > > > > > > > > > > > Correct. AFAIU here we stumble on a pmd entry which was all= ocated but > > > > > > > > never populated. > > > > > > > > > > > > > > Do you mean a pmd_none()? > > > > > > > > > > > > Yes. > > > > > > > > > > > > > > > > > > > > If so, that goes back to my original question, on why > > > > > > > __pmd_trans_huge_lock() returns non-NULL if it's a pmd_none()= ? IMHO it > > > > > > > really should have returned NULL for pmd_none(). > > > > > > > > > > > > That was exactly the answer I gave Lokesh when he theorized abo= ut the > > > > > > cause of this crash but after reproducing it I saw that > > > > > > pmd_trans_huge_lock() happily returns the PTL as long as PMD is= not > > > > > > pmd_none(). And that's because it passes as is_swap_pmd(). But = even if > > > > > > we change that we still need to implement the code to skip the = entire > > > > > > PMD. > > > > > > > > > > The thing is I thought if pmd_trans_huge_lock() can return non-NU= LL, it > > > > > must be either a migration entry or a present THP. So are you des= cribing a > > > > > THP but with present bit cleared? Do you know what is that entry= , and why > > > > > it has present bit cleared? > > > > > > > > In this case it's because earlier we allocated that PMD here: > > > > https://elixir.bootlin.com/linux/v6.16/source/mm/userfaultfd.c#L179= 7 > > > > > > AFAIU, this line is not about allocation of any pmd entry, but the pm= d > > > pgtable page that _holds_ the PMDs: > > > > > > static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsi= gned long address) > > > { > > > return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, addr= ess))? > > > NULL: pmd_offset(pud, address); > > > } > > > > > > It makes sure the PUD entry, not the PMD entry, be populated. > > > > Hmm. Then I was reading this code completely wrong and need to rethink > > what is happening here. > > > > > > > > > but wouldn't that be the same if the PMD was mapped and then got > > > > unmapped later? My understanding is that we allocate the PMD at the > > > > line I pointed to make UFFDIO_MOVE_MODE_ALLOW_SRC_HOLES case the sa= me > > > > as this unmapped PMD case. If my assumption is incorrect then we co= uld > > > > skip the hole earlier instead of allocating the PMD for it. > > > > > > > > > > > > > > I think my attention got attracted to pmd migration entry too muc= h, so I > > > > > didn't really notice such possibility, as I believe migration pmd= is broken > > > > > already in this path. > > > > > > > > > > The original code: > > > > > > > > > > ptl =3D pmd_trans_huge_lock(src_pmd, src_vma); > > > > > if (ptl) { > > > > > /* Check if we can move the pmd without s= plitting it. */ > > > > > if (move_splits_huge_pmd(dst_addr, src_ad= dr, src_start + len) || > > > > > !pmd_none(dst_pmdval)) { > > > > > struct folio *folio =3D pmd_folio= (*src_pmd); > > > > > > > > > > if (!folio || (!is_huge_zero_foli= o(folio) && > > > > > !PageAnonExclusive= (&folio->page))) { > > > > > spin_unlock(ptl); > > > > > err =3D -EBUSY; > > > > > break; > > > > > } > > > > > > > > > > spin_unlock(ptl); > > > > > split_huge_pmd(src_vma, src_pmd, = src_addr); > > > > > /* The folio will be split by mov= e_pages_pte() */ > > > > > continue; > > > > > } > > > > > > > > > > err =3D move_pages_huge_pmd(mm, dst_pmd, = src_pmd, > > > > > dst_pmdval, dst= _vma, src_vma, > > > > > dst_addr, src_a= ddr); > > > > > step_size =3D HPAGE_PMD_SIZE; > > > > > } else { > > > > > > > > > > It'll get ptl for a migration pmd, then pmd_folio is risky withou= t checking > > > > > present bit. That's what my previous smaller patch wanted to fix= . > > > > > > > > > > But besides that, IIUC it's all fine at least for a pmd migration= entry, > > > > > because when with the smaller patch applied, either we'll try to = split the > > > > > pmd migration entry, or we'll do move_pages_huge_pmd(), which int= ernally > > > > > handles the pmd migration entry too by waiting on it: > > > > > > > > > > if (!pmd_trans_huge(src_pmdval)) { > > > > > spin_unlock(src_ptl); > > > > > if (is_pmd_migration_entry(src_pmdval)) { > > > > > pmd_migration_entry_wait(mm, &src_pmdval)= ; > > > > > return -EAGAIN; > > > > > } > > > > > return -ENOENT; > > > > > } > > > > > > > > > > Then logically after the migration entry got recovered, we'll eit= her see a > > > > > real THP or pmd none next time. > > > > > > > > Yes, for migration entries adding the "if (pmd_present(*src_pmd))" > > > > before getting the folio is enough. The problematic case is > > > > (!pmd_none(*src_pmd) && !pmd_present(*src_pmd)) and not a migration > > > > entry. > > > > > > I thought we could have any of below here on the pmd entry: > > > > > > (0) pmd_none, which should constantly have pmd_trans_huge_lock -> N= ULL > > > > > > (1) pmd pgtable entry, which must have PRESENT && !TRANS, so > > > pmd_trans_huge_lock -> NULL, > > > > > > (2) pmd migration, pmd_trans_huge_lock -> valid > > > > > > (3) pmd thp, pmd_trans_huge_lock -> valid > > > > > > I thought (2) was broken, which we seem to agree upon.. however if so= the > > > smaller patch should fix it, per explanation in my previous reply. O= TOH I > > > can't think of (4). > > > > The case I was hitting is (!pmd_none && !pmd_present && > > !is_pmd_migration_entry). My original thinking was that this entry was > > newly allocated at the line I pointed earlier but now I'm not so sure > > anymore. > > Hmm, now I can't reproduce this case... I'm pretty sure I've seen that > case before but now I hit an occasional migration entry and that's > all. I must have done something wrong when testing it before. Ok, I let the reproducer run for half a day and it did not hit this case, so I must have done something wrong during my initial investigation. Sorry for the confusion. I could have sworn that I saw this case but now it just does not happen. With migration entry being the only case that leads to that pmd_folio(), the only check we need to add is the "if (pmd_present(*src_pmd))" before pmd_folio(). Would you like me to check anything else or should I go ahead and post that fix? > > > > > > > > > Said that, I just noticed (3) can be broken as well - could it be a > > > prot_none entry? The very confusing part of this patch is it seems t= o > > > think it's pmd_none() here as holes: > > > > > > if (pmd_present(*src_pmd) || is_pmd_migration_entry(*src_pmd)= ) { > > > ... > > > } else { > > > spin_unlock(ptl); > > > if (!(mode & UFFDIO_MOVE_MODE_ALLOW_SRC_HOLES)) { > > > err =3D -ENOENT; > > > break; > > > } > > > /* nothing to do to move a hole */ > > > err =3D 0; > > > step_size =3D min(HPAGE_PMD_SIZE, src_start + len - s= rc_addr); > > > } > > > > > > But is it really? Again, I don't think pmd_none() could happen with > > > pmd_trans_huge_lock() returning the ptl. > > > > That is true, in the pmd_none() case pmd_trans_huge_lock() returns NULL= . > > > > > > > > Could you double check this? E.g. with this line if that makes sense= to > > > you: > > > > > > diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c > > > index 8bf8ff0be990f..d2d4f2a0ae69f 100644 > > > --- a/mm/userfaultfd.c > > > +++ b/mm/userfaultfd.c > > > @@ -1903,6 +1903,7 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx,= unsigned long dst_start, > > > dst_addr, s= rc_addr); > > > step_size =3D HPAGE_PMD_SIZE; > > > } else { > > > + BUG_ON(!pmd_none(*src_pmd)); > > > spin_unlock(ptl); > > > if (!(mode & UFFDIO_MOVE_MODE_ALLOW_S= RC_HOLES)) { > > > err =3D -ENOENT; > > > > > > I would expect it constantly BUG() here, if that explains my thoughts= . > > > > I'll add this and check. > > This check does trigger and I logged pmd_val=3D0x8b3714067 and I think > this is normal. _PAGE_BIT_PRESENT is set and _PAGE_BIT_PSE is not set, > so is_swap_pmd()=3D=3Dfalse and pmd_trans_huge()=3D=3Dfalse, therefore > pmd_trans_huge_lock() returns NULL. > > > > > > > > > > Now I doubt it's a prot_none THP.. aka, a THP that got numa hint to b= e > > > moved. If so, we may need to process it / move it / .. but we likely > > > should never skip it. We can double check the buggy pmd entry you hi= t > > > (besides migration entry) first. > > > > Let me log the flags of the entry when this issue happens. That should > > provide more insights. > > Thanks, > > Suren. > > > > > > > > Thanks, > > > > > > -- > > > Peter Xu > > >