[PATCH 1/3] drivers/dax/super: Add annotation for ops in struct dax_device

[PATCH 1/3] drivers/dax/super: Add annotation for ops in struct dax_device

[Huaisheng Ye]

From: Huaisheng Ye <yehs1@lenovo.com>

There is no annotation for member ops within struct dax_device.
Add it.

Signed-off-by: Huaisheng Ye <yehs1@lenovo.com>
---
 drivers/dax/super.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/dax/super.c b/drivers/dax/super.c
index 2b2332b..aa6382a 100644
--- a/drivers/dax/super.c
+++ b/drivers/dax/super.c
@@ -161,6 +161,7 @@ enum dax_device_flags {
  * @host: optional name for lookups where the device path is not available
  * @private: dax driver private data
  * @flags: state and boolean properties
+ * @ops: function pointers for direct_access and fs-dax
  */
 struct dax_device {
 	struct hlist_node list;
-- 
1.8.3.1

[PATCH 2/3] fs/ext2/inode: Fix a type cast error for fsdax

[Huaisheng Ye]

From: Huaisheng Ye <yehs1@lenovo.com>

The type of offset within struct iomap is loff_t, which represents
file offset of mapping.

In ext2_iomap_begin, iomap->offset shall be given a type cast as
loff_t instead of u64.

Signed-off-by: Huaisheng Ye <yehs1@lenovo.com>
---
 fs/ext2/inode.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
index 7163590..ca211bd 100644
--- a/fs/ext2/inode.c
+++ b/fs/ext2/inode.c
@@ -816,7 +816,7 @@ static int ext2_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 
 	iomap->flags = 0;
 	iomap->bdev = inode->i_sb->s_bdev;
-	iomap->offset = (u64)first_block << blkbits;
+	iomap->offset = (loff_t)first_block << blkbits;
 	iomap->dax_dev = sbi->s_daxdev;
 
 	if (ret == 0) {
-- 
1.8.3.1

[PATCH 3/3] fs/ext2/inode: Optimize the condition for iomap_begin

[Huaisheng Ye]

From: Huaisheng Ye <yehs1@lenovo.com>

If ext2_get_blocks returns negative result, ext2_iomap_begin
will return for error case.
Adjust the judging condition of ret value will be useful for
code simplification.

Signed-off-by: Huaisheng Ye <yehs1@lenovo.com>
---
 fs/ext2/inode.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
index ca211bd..9b1004d 100644
--- a/fs/ext2/inode.c
+++ b/fs/ext2/inode.c
@@ -819,15 +819,15 @@ static int ext2_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	iomap->offset = (loff_t)first_block << blkbits;
 	iomap->dax_dev = sbi->s_daxdev;
 
-	if (ret == 0) {
-		iomap->type = IOMAP_HOLE;
-		iomap->addr = IOMAP_NULL_ADDR;
-		iomap->length = 1 << blkbits;
-	} else {
+	if (ret) {
 		iomap->type = IOMAP_MAPPED;
 		iomap->addr = (u64)bno << blkbits;
 		iomap->length = (u64)ret << blkbits;
 		iomap->flags |= IOMAP_F_MERGED;
+	} else {
+		iomap->type = IOMAP_HOLE;
+		iomap->addr = IOMAP_NULL_ADDR;
+		iomap->length = 1 << blkbits;
 	}
 
 	if (new)
-- 
1.8.3.1

Re: [PATCH 2/3] fs/ext2/inode: Fix a type cast error for fsdax

[Al Viro]

On Sun, Jul 01, 2018 at 02:18:47PM +0800, Huaisheng Ye wrote:
> From: Huaisheng Ye <yehs1@lenovo.com>
> 
> The type of offset within struct iomap is loff_t, which represents
> file offset of mapping.
> 
> In ext2_iomap_begin, iomap->offset shall be given a type cast as
> loff_t instead of u64.

Why is it an error?  loff_t is uniformly typedefed to long long.
In which case the second variant is different from the first one
*and* does not step into nasal demon territory?

> -	iomap->offset = (u64)first_block << blkbits;
> +	iomap->offset = (loff_t)first_block << blkbits;

Re: [PATCH 2/3] fs/ext2/inode: Fix a type cast error for fsdax

[Huaisheng Ye]

 ---- On Mon, 02 Jul 2018 03:26:00 +0800 Al Viro <viro@ZenIV.linux.org.uk> wrote ---- 
 > On Sun, Jul 01, 2018 at 02:18:47PM +0800, Huaisheng Ye wrote: 
 > > From: Huaisheng Ye <yehs1@lenovo.com> 
 > >  
 > > The type of offset within struct iomap is loff_t, which represents 
 > > file offset of mapping. 
 > >  
 > > In ext2_iomap_begin, iomap->offset shall be given a type cast as 
 > > loff_t instead of u64. 
 >  
 > Why is it an error?  loff_t is uniformly typedefed to long long. 
 > In which case the second variant is different from the first one 
 > *and* does not step into nasal demon territory?

Sorry for my inaccuracy.
The type of iomap->offset is loff_t, is it better to cast first_block
to loff_t, then do the left shift operation?
 
 >  
 > > -    iomap->offset = (u64)first_block << blkbits; 
 > > +    iomap->offset = (loff_t)first_block << blkbits; 
 > 

---
Cheers,
Huaisheng

Re: [PATCH 2/3] fs/ext2/inode: Fix a type cast error for fsdax

[Jan Kara]

On Mon 02-07-18 14:23:42, Huaisheng Ye wrote:
>  ---- On Mon, 02 Jul 2018 03:26:00 +0800 Al Viro <viro@ZenIV.linux.org.uk> wrote ---- 
>  > On Sun, Jul 01, 2018 at 02:18:47PM +0800, Huaisheng Ye wrote: 
>  > > From: Huaisheng Ye <yehs1@lenovo.com> 
>  > >  
>  > > The type of offset within struct iomap is loff_t, which represents 
>  > > file offset of mapping. 
>  > >  
>  > > In ext2_iomap_begin, iomap->offset shall be given a type cast as 
>  > > loff_t instead of u64. 
>  >  
>  > Why is it an error?  loff_t is uniformly typedefed to long long. 
>  > In which case the second variant is different from the first one 
>  > *and* does not step into nasal demon territory?
> 
> Sorry for my inaccuracy.
> The type of iomap->offset is loff_t, is it better to cast first_block
> to loff_t, then do the left shift operation?

I think what Al meant is that in the kernel loff_t is 64-bit anyway and
furthermore it is signed which means most people have to look up C standard
to remember how right shift is (un)defined if it would overflow ;). So this
does not really make the code any more readable. Rather on contrary...

								Honza

>  
>  >  
>  > > -    iomap->offset = (u64)first_block << blkbits; 
>  > > +    iomap->offset = (loff_t)first_block << blkbits; 
>  > 
> 
> ---
> Cheers,
> Huaisheng
> 
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH 3/3] fs/ext2/inode: Optimize the condition for iomap_begin

[Jan Kara]

On Sun 01-07-18 14:18:48, Huaisheng Ye wrote:
> From: Huaisheng Ye <yehs1@lenovo.com>
> 
> If ext2_get_blocks returns negative result, ext2_iomap_begin
> will return for error case.
> Adjust the judging condition of ret value will be useful for
> code simplification.
> 
> Signed-off-by: Huaisheng Ye <yehs1@lenovo.com>

I'm sorry but how exactly is this simpler?

								Honza

> ---
>  fs/ext2/inode.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
> index ca211bd..9b1004d 100644
> --- a/fs/ext2/inode.c
> +++ b/fs/ext2/inode.c
> @@ -819,15 +819,15 @@ static int ext2_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
>  	iomap->offset = (loff_t)first_block << blkbits;
>  	iomap->dax_dev = sbi->s_daxdev;
>  
> -	if (ret == 0) {
> -		iomap->type = IOMAP_HOLE;
> -		iomap->addr = IOMAP_NULL_ADDR;
> -		iomap->length = 1 << blkbits;
> -	} else {
> +	if (ret) {
>  		iomap->type = IOMAP_MAPPED;
>  		iomap->addr = (u64)bno << blkbits;
>  		iomap->length = (u64)ret << blkbits;
>  		iomap->flags |= IOMAP_F_MERGED;
> +	} else {
> +		iomap->type = IOMAP_HOLE;
> +		iomap->addr = IOMAP_NULL_ADDR;
> +		iomap->length = 1 << blkbits;
>  	}
>  
>  	if (new)
> -- 
> 1.8.3.1
> 
> 
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR