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;
IronPort-HdrOrdr: =?us-ascii?q?A9a23=3AEQLKFazEJguRdXz/ox3XKrPwEL1zdoMgy1kn?=
 =?us-ascii?q?xilNoH1uA6ilfqWV8cjzuiWbtN9vYhsdcLy7WZVoIkmskKKdg7NhXotKNTOO0A?=
 =?us-ascii?q?SVxepZnOnfKlPbexHWx6p00KdMV+xEAsTsMF4St63HyTj9P9E+4NTvysyVuds?=
 =?us-ascii?q?=3D?=
From:   Li Zhijian <lizhijian@cn.fujitsu.com>
To:     <linux-mm@kvack.org>, <linux-rdma@vger.kernel.org>,
        <akpm@linux-foundation.org>, <jglisse@redhat.com>, <jgg@ziepe.ca>,
        <hch@infradead.org>
CC:     <yishaih@nvidia.com>, <linux-kernel@vger.kernel.org>,
        Li Zhijian <lizhijian@cn.fujitsu.com>, <stable@vger.kernel.org>
Subject: [PATCH v2] mm/hmm: bypass devmap pte when all pfn requested flags are fulfilled
Date:   Sat, 28 Aug 2021 09:04:41 +0800
Message-ID: <20210828010441.3702-1-lizhijian@cn.fujitsu.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 7BIT
Content-Type:   text/plain; charset=US-ASCII
Precedence: bulk

Previously, we noticed the one rpma example was failed[1] since 36f30e486d,
where it will use ODP feature to do RDMA WRITE between fsdax files.

After digging into the code, we found hmm_vma_handle_pte() will still
return EFAULT even though all the its requesting flags has been
fulfilled. That's because a DAX page will be marked as
(_PAGE_SPECIAL | PAGE_DEVMAP) by pte_mkdevmap().

[1]: https://github.com/pmem/rpma/issues/1142

CC: stable@vger.kernel.org
Fixes: 405506274922 ("mm/hmm: add missing call to hmm_pte_need_fault in HMM_PFN_SPECIAL handling")
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
---
 mm/hmm.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/mm/hmm.c b/mm/hmm.c
index fad6be2bf072..d324fb1a5352 100644
--- a/mm/hmm.c
+++ b/mm/hmm.c
@@ -295,10 +295,13 @@ static int hmm_vma_handle_pte(struct mm_walk *walk, unsigned long addr,
 		goto fault;
 
 	/*
+	 * Bypass devmap pte such as DAX page when all pfn requested
+	 * flags(pfn_req_flags) are fulfilled.
 	 * Since each architecture defines a struct page for the zero page, just
 	 * fall through and treat it like a normal page.
 	 */
-	if (pte_special(pte) && !is_zero_pfn(pte_pfn(pte))) {
+	if (!pte_devmap(pte) && pte_special(pte) &&
+	    !is_zero_pfn(pte_pfn(pte))) {
 		if (hmm_pte_need_fault(hmm_vma_walk, pfn_req_flags, 0)) {
 			pte_unmap(ptep);
 			return -EFAULT;
-- 
2.31.1