Linux-mm Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Vernon Yang <vernon2gm@gmail.com>
To: tglx@kernel.org, mingo@redhat.com, bp@alien8.de,
	dave.hansen@linux.intel.com, akpm@linux-foundation.org,
	david@kernel.org
Cc: hpa@zytor.com, rmclure@linux.ibm.com,
	andrew+kernel@donnellan.id.au, pasha.tatashin@soleen.com,
	kas@kernel.org, tj@kernel.org, rppt@kernel.org,
	rick.p.edgecombe@intel.com, yu-cheng.yu@intel.com,
	orsonpeters@gmail.com, linux-kernel@vger.kernel.org,
	x86@kernel.org, linux-mm@kvack.org,
	Vernon Yang <yanglincheng@kylinos.cn>,
	stable@vger.kernel.org
Subject: [PATCH] x86/mm: Fix pmd_modify() dropping the dirty bit
Date: Thu,  3 Sep 2026 11:16:08 +0800	[thread overview]
Message-ID: <20260903031608.1194238-1-vernon2gm@gmail.com> (raw)

From: Vernon Yang <yanglincheng@kylinos.cn>

pmd_modify() masks the old value with (_HPAGE_CHG_MASK & ~_PAGE_DIRTY),
silently discarding the hardware dirty bit.  The subsequent
pmd_mksaveddirty() call is supposed to transfer _PAGE_DIRTY into
_PAGE_SAVED_DIRTY when write-protecting, but the dirty bit was already
stripped from the value, so there is nothing left to transfer.

Contrast with pte_modify(), which keeps _PAGE_DIRTY_BITS in its mask,
and pud_modify(), which keeps _HPAGE_CHG_MASK untouched: pmd_modify()
is the odd one out.  Any pmd_modify() on a writable, dirty PMD loses
the dirty state.

One visible consequence is data loss with MADV_FREE on PMD-mapped THP:

    memset(buf, 0x5A, size);          // PMD-mapped THP, PMD dirty
    madvise(buf, size, MADV_FREE);    // PMD cleaned but left writable,
                                      // folio marked lazyfree
    memset(buf, 0x5A, size);          // hardware sets _PAGE_DIRTY again
    mprotect(buf, size, PROT_READ);   // pmd_modify() drops the dirty bit
    mprotect(buf, size, PROT_READ|PROT_WRITE);
    // ... memory pressure ...

Reclaim (e.g. under memcg pressure) then finds the lazyfree folio with
no dirty bit set anywhere and frees it in
__discard_anon_folio_pmd_locked(), even though the data was rewritten
after MADV_FREE; subsequent reads fault in fresh zero pages.  NUMA
hinting alone can trigger the same loss, as do_huge_pmd_numa_page()
restores the PMD through pmd_modify() as well.

PMD-mapped file THPs are affected too: mprotect()/NUMA hinting dropping
the dirty bit means rewritten data is never written back.

Fix it by keeping _PAGE_DIRTY in the preserved mask, exactly like
pte_modify() and pud_modify() do.  The existing
pmd_mksaveddirty()/pmd_clear_saveddirty() pair then performs the
hardware-dirty <-> saved-dirty transition based on the write bit,
preserving the shadow-stack encoding rules.

Closes: https://lore.kernel.org/r/CAJxLxMUGu1-L+O_nAONOwOXnS=cNbNApCWqdthRjd76LThtSPg@mail.gmail.com/
Fixes: bb3aadf7d446 ("x86/mm: Start actually marking _PAGE_SAVED_DIRTY")
Cc: stable@vger.kernel.org
Signed-off-by: Vernon Yang <yanglincheng@kylinos.cn>
---
 arch/x86/include/asm/pgtable.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index d5f4917c1edc..d551120a7c88 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -806,7 +806,7 @@ static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 	pmdval_t val = pmd_val(pmd), oldval = val;
 	pmd_t pmd_result;
 
-	val &= (_HPAGE_CHG_MASK & ~_PAGE_DIRTY);
+	val &= _HPAGE_CHG_MASK;
 	val |= check_pgprot(newprot) & ~_HPAGE_CHG_MASK;
 	val = flip_protnone_guard(oldval, val, PHYSICAL_PMD_PAGE_MASK);
 
-- 
2.53.0



             reply	other threads:[~2026-09-03  3:17 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-03  3:16 Vernon Yang [this message]
2026-09-03  4:10 ` [PATCH] x86/mm: Fix pmd_modify() dropping the dirty bit Andrew Morton
2026-09-03  5:24   ` Vernon Yang
2026-09-03  7:34   ` Orson Peters
2026-09-03 15:40     ` Dave Hansen
2026-09-03 11:54 ` Kiryl Shutsemau
2026-09-03 14:18   ` Dave Hansen
2026-09-03 17:21     ` Andrew Morton
2026-09-03 17:58     ` Edgecombe, Rick P
2026-09-03 18:01       ` Dave Hansen
2026-09-03 18:33         ` Edgecombe, Rick P

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260903031608.1194238-1-vernon2gm@gmail.com \
    --to=vernon2gm@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=andrew+kernel@donnellan.id.au \
    --cc=bp@alien8.de \
    --cc=dave.hansen@linux.intel.com \
    --cc=david@kernel.org \
    --cc=hpa@zytor.com \
    --cc=kas@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mingo@redhat.com \
    --cc=orsonpeters@gmail.com \
    --cc=pasha.tatashin@soleen.com \
    --cc=rick.p.edgecombe@intel.com \
    --cc=rmclure@linux.ibm.com \
    --cc=rppt@kernel.org \
    --cc=stable@vger.kernel.org \
    --cc=tglx@kernel.org \
    --cc=tj@kernel.org \
    --cc=x86@kernel.org \
    --cc=yanglincheng@kylinos.cn \
    --cc=yu-cheng.yu@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox