After seeking time on testing today upstream fsdax, I found it
actually doesn't work well as below:

[  186.492983] ------------[ cut here ]------------
[  186.493629] WARNING: CPU: 1 PID: 205 at fs/iomap/iter.c:33 iomap_iter+0x2f6/0x310

The problem is that m_dax_part_off should be applied to physical
addresses and very sorry about that I didn't catch this eariler.

Anyway, let's fix it up now. Also, I need to find a way to set up
a standalone testcase to look after this later.

Fixes: de2051147771 ("fsdax: shift partition offset handling into the file systems")
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/data.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/erofs/data.c b/fs/erofs/data.c
index fa7ddb7ad980..226a57c57ee6 100644
--- a/fs/erofs/data.c
+++ b/fs/erofs/data.c
@@ -252,12 +252,10 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 		return ret;
 
 	iomap->offset = map.m_la;
-	if (flags & IOMAP_DAX) {
+	if (flags & IOMAP_DAX)
 		iomap->dax_dev = mdev.m_daxdev;
-		iomap->offset += mdev.m_dax_part_off;
-	} else {
+	else
 		iomap->bdev = mdev.m_bdev;
-	}
 	iomap->length = map.m_llen;
 	iomap->flags = 0;
 	iomap->private = NULL;
@@ -284,6 +282,8 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	} else {
 		iomap->type = IOMAP_MAPPED;
 		iomap->addr = mdev.m_pa;
+		if (flags & IOMAP_DAX)
+			iomap->addr += mdev.m_dax_part_off;
 	}
 	return 0;
 }
-- 
2.24.4

Looks good, thanks for fixing this up:

Reviewed-by: Christoph Hellwig <hch@lst.de>

On Thu, Jan 13, 2022 at 08:21:22AM +0100, Christoph Hellwig wrote:
> Looks good, thanks for fixing this up:
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>

Thanks for the review. Will apply later..

Thanks,
Gao Xiang

On 2022/1/13 13:18, Gao Xiang wrote:
> After seeking time on testing today upstream fsdax, I found it
> actually doesn't work well as below:
> 
> [  186.492983] ------------[ cut here ]------------
> [  186.493629] WARNING: CPU: 1 PID: 205 at fs/iomap/iter.c:33 iomap_iter+0x2f6/0x310
> 
> The problem is that m_dax_part_off should be applied to physical
> addresses and very sorry about that I didn't catch this eariler.
> 
> Anyway, let's fix it up now. Also, I need to find a way to set up
> a standalone testcase to look after this later.
> 
> Fixes: de2051147771 ("fsdax: shift partition offset handling into the file systems")
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,