Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:2 as permitted sender) client-ip=2620:137:e000::3:2;
Date:   Mon,  4 Dec 2023 17:26:45 +0000
In-Reply-To: <20231204172646.2541916-1-jthoughton@google.com>
Mime-Version: 1.0
References: <20231204172646.2541916-1-jthoughton@google.com>
Message-ID: <20231204172646.2541916-2-jthoughton@google.com>
Subject: [PATCH 1/2] arm64: hugetlb: Distinguish between hw and sw dirtiness
 in __cont_access_flags_changed
From:   James Houghton <jthoughton@google.com>
To:     Steve Capper <steve.capper@arm.com>, Will Deacon <will@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>
Cc:     Mike Kravetz <mike.kravetz@oracle.com>,
        Muchun Song <songmuchun@bytedance.com>,
        Anshuman Khandual <anshuman.khandual@arm.com>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Ryan Roberts <ryan.roberts@arm.com>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org,
        James Houghton <jthoughton@google.com>, stable@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

__cont_access_flags_changed was originally introduced to avoid making
unnecessary changes to the PTEs. Consider the following case: all the
PTEs in the contiguous group have PTE_DIRTY | PTE_RDONLY | PTE_WRITE,
and we are running on a system without HAFDBS.  When writing via these
PTEs, we will get a page fault, and hugetlb_fault will (rightly)
attempt to update the PTEs with PTE_DIRTY | PTE_WRITE, but, as both the
original PTEs and the new PTEs are pte_dirty(),
__cont_access_flags_changed prevents the pgprot update from occurring.

To avoid the page fault loop that we get ourselves into, distinguish
between hardware-dirty and software-dirty for this check. Non-contiguous
PTEs aren't broken in the same way, as we will always write a new PTE
unless the new PTE is exactly equal to the old one.

Fixes: 031e6e6b4e12 ("arm64: hugetlb: Avoid unnecessary clearing in huge_ptep_set_access_flags")
Signed-off-by: James Houghton <jthoughton@google.com>
Cc: <stable@vger.kernel.org>

diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
index f5aae342632c..87a9564976fa 100644
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -437,7 +437,10 @@ static int __cont_access_flags_changed(pte_t *ptep, pte_t pte, int ncontig)
 	for (i = 0; i < ncontig; i++) {
 		pte_t orig_pte = ptep_get(ptep + i);
 
-		if (pte_dirty(pte) != pte_dirty(orig_pte))
+		if (pte_sw_dirty(pte) != pte_sw_dirty(orig_pte))
+			return 1;
+
+		if (pte_hw_dirty(pte) != pte_hw_dirty(orig_pte))
 			return 1;
 
 		if (pte_young(pte) != pte_young(orig_pte))
-- 
2.43.0.rc2.451.g8631bc7472-goog