Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Shiyang Ruan <ruansy.fnst@cn.fujitsu.com>
To:     <linux-xfs@vger.kernel.org>, <linux-nvdimm@lists.01.org>,
        <darrick.wong@oracle.com>, <rgoldwyn@suse.de>, <hch@infradead.org>,
        <david@fromorbit.com>
CC:     <linux-kernel@vger.kernel.org>, <gujx@cn.fujitsu.com>,
        <qi.fuli@fujitsu.com>, <caoj.fnst@cn.fujitsu.com>,
        <ruansy.fnst@cn.fujitsu.com>
Subject: [RFC PATCH v2 2/7] dax: copy data before write.
Date:   Wed, 30 Oct 2019 12:13:53 +0800
Message-ID: <20191030041358.14450-3-ruansy.fnst@cn.fujitsu.com>
In-Reply-To: <20191030041358.14450-1-ruansy.fnst@cn.fujitsu.com>
References: <20191030041358.14450-1-ruansy.fnst@cn.fujitsu.com>
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Add dax_copy_edges() into each dax actor functions to perform
copy on write.

Signed-off-by: Shiyang Ruan <ruansy.fnst@cn.fujitsu.com>
---
 fs/dax.c | 36 +++++++++++++++++++++++++++++++++---
 1 file changed, 33 insertions(+), 3 deletions(-)

diff --git a/fs/dax.c b/fs/dax.c
index e1f4493ce56a..949a40cf1fe7 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -1159,7 +1159,7 @@ dax_iomap_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 			return iov_iter_zero(min(length, end - pos), iter);
 	}
 
-	if (WARN_ON_ONCE(iomap->type != IOMAP_MAPPED))
+	if (WARN_ON_ONCE(iomap->type != IOMAP_MAPPED && iomap == srcmap))
 		return -EIO;
 
 	/*
@@ -1198,6 +1198,12 @@ dax_iomap_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 			break;
 		}
 
+		if (iomap != srcmap) {
+			ret = dax_copy_edges(pos, length, srcmap, kaddr, false);
+			if (ret)
+				break;
+		}
+
 		map_len = PFN_PHYS(map_len);
 		kaddr += offset;
 		map_len -= offset;
@@ -1306,6 +1312,7 @@ static vm_fault_t dax_iomap_pte_fault(struct vm_fault *vmf, pfn_t *pfnp,
 	vm_fault_t ret = 0;
 	void *entry;
 	pfn_t pfn;
+	void *kaddr;
 
 	trace_dax_pte_fault(inode, vmf, ret);
 	/*
@@ -1387,19 +1394,27 @@ static vm_fault_t dax_iomap_pte_fault(struct vm_fault *vmf, pfn_t *pfnp,
 
 	switch (iomap.type) {
 	case IOMAP_MAPPED:
+cow:
 		if (iomap.flags & IOMAP_F_NEW) {
 			count_vm_event(PGMAJFAULT);
 			count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
 			major = VM_FAULT_MAJOR;
 		}
 		error = dax_iomap_direct_access(&iomap, pos, PAGE_SIZE, &pfn,
-						NULL);
+						&kaddr);
 		if (error < 0)
 			goto error_finish_iomap;
 
 		entry = dax_insert_entry(&xas, mapping, vmf, entry, pfn,
 						 0, write && !sync);
 
+		if (srcmap.type != IOMAP_HOLE) {
+			error = dax_copy_edges(pos, PAGE_SIZE, &srcmap, kaddr,
+					       false);
+			if (error)
+				goto error_finish_iomap;
+		}
+
 		/*
 		 * If we are doing synchronous page fault and inode needs fsync,
 		 * we can insert PTE into page tables only after that happens.
@@ -1423,6 +1438,9 @@ static vm_fault_t dax_iomap_pte_fault(struct vm_fault *vmf, pfn_t *pfnp,
 
 		goto finish_iomap;
 	case IOMAP_UNWRITTEN:
+		if (srcmap.type != IOMAP_HOLE)
+			goto cow;
+		/*FALLTHRU*/
 	case IOMAP_HOLE:
 		if (!write) {
 			ret = dax_load_hole(&xas, mapping, &entry, vmf);
@@ -1530,6 +1548,7 @@ static vm_fault_t dax_iomap_pmd_fault(struct vm_fault *vmf, pfn_t *pfnp,
 	loff_t pos;
 	int error;
 	pfn_t pfn;
+	void *kaddr;
 
 	/*
 	 * Check whether offset isn't beyond end of file now. Caller is
@@ -1610,15 +1629,23 @@ static vm_fault_t dax_iomap_pmd_fault(struct vm_fault *vmf, pfn_t *pfnp,
 	sync = dax_fault_is_synchronous(iomap_flags, vma, &iomap);
 
 	switch (iomap.type) {
+cow:
 	case IOMAP_MAPPED:
 		error = dax_iomap_direct_access(&iomap, pos, PMD_SIZE, &pfn,
-						NULL);
+						&kaddr);
 		if (error < 0)
 			goto finish_iomap;
 
 		entry = dax_insert_entry(&xas, mapping, vmf, entry, pfn,
 						DAX_PMD, write && !sync);
 
+		if (srcmap.type != IOMAP_HOLE) {
+			error = dax_copy_edges(pos, PMD_SIZE, &srcmap, kaddr,
+					       true);
+			if (error)
+				goto unlock_entry;
+		}
+
 		/*
 		 * If we are doing synchronous page fault and inode needs fsync,
 		 * we can insert PMD into page tables only after that happens.
@@ -1637,6 +1664,9 @@ static vm_fault_t dax_iomap_pmd_fault(struct vm_fault *vmf, pfn_t *pfnp,
 		result = vmf_insert_pfn_pmd(vmf, pfn, write);
 		break;
 	case IOMAP_UNWRITTEN:
+		if (srcmap.type != IOMAP_HOLE)
+			goto cow;
+		/*FALLTHRU*/
 	case IOMAP_HOLE:
 		if (WARN_ON_ONCE(write))
 			break;
-- 
2.23.0