public inbox for linux-mtd@lists.infradead.org
 help / color / mirror / Atom feed
* [PATCH] jffs2: use fs_umode_to_dtype() for dirent type
@ 2026-02-06 11:51 Adarsh Das
  2026-02-07  2:36 ` Zhihao Cheng
  2026-02-07  8:00 ` Zhihao Cheng
  0 siblings, 2 replies; 5+ messages in thread
From: Adarsh Das @ 2026-02-06 11:51 UTC (permalink / raw)
  To: dwmw2, richard; +Cc: linux-mtd, linux-kernel, Adarsh Das

Use fs_umode_to_dtype() instead of direct bit shifting when
setting directory entry type.

Signed-off-by: Adarsh Das <adarshdas950@gmail.com>
---
 fs/jffs2/dir.c | 11 +++--------
 1 file changed, 3 insertions(+), 8 deletions(-)

diff --git a/fs/jffs2/dir.c b/fs/jffs2/dir.c
index dd91f725ded6..185cf76a061e 100644
--- a/fs/jffs2/dir.c
+++ b/fs/jffs2/dir.c
@@ -261,8 +261,7 @@ static int jffs2_link (struct dentry *old_dentry, struct inode *dir_i, struct de
 	if (d_is_dir(old_dentry))
 		return -EPERM;
 
-	/* XXX: This is ugly */
-	type = (d_inode(old_dentry)->i_mode & S_IFMT) >> 12;
+	type = fs_umode_to_dtype(d_inode(old_dentry)->i_mode);
 	if (!type) type = DT_REG;
 
 	now = JFFS2_NOW();
@@ -730,10 +729,7 @@ static int jffs2_mknod (struct mnt_idmap *idmap, struct inode *dir_i,
 	rd->ino = cpu_to_je32(inode->i_ino);
 	rd->mctime = cpu_to_je32(JFFS2_NOW());
 	rd->nsize = namelen;
-
-	/* XXX: This is ugly. */
-	rd->type = (mode & S_IFMT) >> 12;
-
+	rd->type = fs_umode_to_dtype(mode);
 	rd->node_crc = cpu_to_je32(crc32(0, rd, sizeof(*rd)-8));
 	rd->name_crc = cpu_to_je32(crc32(0, dentry->d_name.name, namelen));
 
@@ -811,8 +807,7 @@ static int jffs2_rename (struct mnt_idmap *idmap,
 
 	/* Make a hard link */
 
-	/* XXX: This is ugly */
-	type = (d_inode(old_dentry)->i_mode & S_IFMT) >> 12;
+	type = fs_umode_to_dtype(d_inode(old_dentry)->i_mode);
 	if (!type) type = DT_REG;
 
 	now = JFFS2_NOW();
-- 
2.53.0


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [PATCH] jffs2: use fs_umode_to_dtype() for dirent type
  2026-02-06 11:51 [PATCH] jffs2: use fs_umode_to_dtype() for dirent type Adarsh Das
@ 2026-02-07  2:36 ` Zhihao Cheng
  2026-02-07  7:39   ` Adarsh Das
  2026-02-07  8:00 ` Zhihao Cheng
  1 sibling, 1 reply; 5+ messages in thread
From: Zhihao Cheng @ 2026-02-07  2:36 UTC (permalink / raw)
  To: Adarsh Das, dwmw2, richard; +Cc: linux-mtd, linux-kernel

在 2026/2/6 19:51, Adarsh Das 写道:
> Use fs_umode_to_dtype() instead of direct bit shifting when
> setting directory entry type.
> 
> Signed-off-by: Adarsh Das <adarshdas950@gmail.com>
> ---
>   fs/jffs2/dir.c | 11 +++--------
>   1 file changed, 3 insertions(+), 8 deletions(-)
> 
> diff --git a/fs/jffs2/dir.c b/fs/jffs2/dir.c
> index dd91f725ded6..185cf76a061e 100644
> --- a/fs/jffs2/dir.c
> +++ b/fs/jffs2/dir.c
> @@ -261,8 +261,7 @@ static int jffs2_link (struct dentry *old_dentry, struct inode *dir_i, struct de
>   	if (d_is_dir(old_dentry))
>   		return -EPERM;
>   
> -	/* XXX: This is ugly */
> -	type = (d_inode(old_dentry)->i_mode & S_IFMT) >> 12;
> +	type = fs_umode_to_dtype(d_inode(old_dentry)->i_mode);
>   	if (!type) type = DT_REG;

Hi Adarsh,
The convertion result is wrong for jffs2_do_link.
The i_mode of old_dentry is 'S_IFIFO', the jffs2 gets type DT_FIFO(1) 
and pass it to 'rd->type' by jffs2_do_link. However, after the patch 
applied, jffs2 gets type FT_FIFO(5), which will confuse other 'rd->type' 
accessing processes(eg. jffs2_scan_dirent_node[rd->type], 
jffs2_build_filesystem[fd->type]).

>   
>   	now = JFFS2_NOW();
> @@ -730,10 +729,7 @@ static int jffs2_mknod (struct mnt_idmap *idmap, struct inode *dir_i,
>   	rd->ino = cpu_to_je32(inode->i_ino);
>   	rd->mctime = cpu_to_je32(JFFS2_NOW());
>   	rd->nsize = namelen;
> -
> -	/* XXX: This is ugly. */
> -	rd->type = (mode & S_IFMT) >> 12;
> -
> +	rd->type = fs_umode_to_dtype(mode);
>   	rd->node_crc = cpu_to_je32(crc32(0, rd, sizeof(*rd)-8));
>   	rd->name_crc = cpu_to_je32(crc32(0, dentry->d_name.name, namelen));
>   
> @@ -811,8 +807,7 @@ static int jffs2_rename (struct mnt_idmap *idmap,
>   
>   	/* Make a hard link */
>   
> -	/* XXX: This is ugly */
> -	type = (d_inode(old_dentry)->i_mode & S_IFMT) >> 12;
> +	type = fs_umode_to_dtype(d_inode(old_dentry)->i_mode);
>   	if (!type) type = DT_REG;
>   
>   	now = JFFS2_NOW();
> 


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] jffs2: use fs_umode_to_dtype() for dirent type
  2026-02-07  2:36 ` Zhihao Cheng
@ 2026-02-07  7:39   ` Adarsh Das
  2026-02-07  7:59     ` Zhihao Cheng
  0 siblings, 1 reply; 5+ messages in thread
From: Adarsh Das @ 2026-02-07  7:39 UTC (permalink / raw)
  To: Zhihao Cheng; +Cc: dwmw2, richard, linux-mtd, linux-kernel, Adarsh Das

Hi Zhihao,

Thank you for taking the time to review my patch!

I'm trying to understand the issue you've identified. I looked at the 
fs_umode_to_dtype() implementation in fs/fs_types.c:

    unsigned char fs_umode_to_dtype(umode_t mode) {
        return fs_ftype_to_dtype(fs_umode_to_ftype(mode));
    }

From what I can see, it seems like the function does produce FT_FIFO (5) 
as an intermediate value, but then converts it back before returning. 
Here's what I think happens for S_IFIFO (0010000):

Old bit-shift code:
    type = (mode & S_IFMT) >> 12
         = (0010000 & 00170000) >> 12
         = 0010000 >> 12
         = 1 (DT_FIFO)

New fs_umode_to_dtype() code:
    Step 1: fs_umode_to_ftype(0010000)
            = fs_ftype_by_dtype[S_DT(0010000)]
            = fs_ftype_by_dtype[1]
            = FT_FIFO = 5

    Step 2: fs_ftype_to_dtype(5)
            = fs_dtype_by_ftype[5]
            = DT_FIFO = 1

I'm not sure how FT_FIFO (5) would remain after the function returns, since 
it seems to get converted back to DT_FIFO (1) in step 2.

If there's an issue with this approach, I can edit my patch and keep changes in only jffs2_mknod.

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] jffs2: use fs_umode_to_dtype() for dirent type
  2026-02-07  7:39   ` Adarsh Das
@ 2026-02-07  7:59     ` Zhihao Cheng
  0 siblings, 0 replies; 5+ messages in thread
From: Zhihao Cheng @ 2026-02-07  7:59 UTC (permalink / raw)
  To: Adarsh Das; +Cc: dwmw2, richard, linux-mtd, linux-kernel

在 2026/2/7 15:39, Adarsh Das 写道:
> Hi Zhihao,
> 
> Thank you for taking the time to review my patch!
> 
> I'm trying to understand the issue you've identified. I looked at the
> fs_umode_to_dtype() implementation in fs/fs_types.c:
> 
>      unsigned char fs_umode_to_dtype(umode_t mode) {
>          return fs_ftype_to_dtype(fs_umode_to_ftype(mode));
>      }
> 
>>From what I can see, it seems like the function does produce FT_FIFO (5) 
> as an intermediate value, but then converts it back before returning.
> Here's what I think happens for S_IFIFO (0010000):
> 
> Old bit-shift code:
>      type = (mode & S_IFMT) >> 12
>           = (0010000 & 00170000) >> 12
>           = 0010000 >> 12
>           = 1 (DT_FIFO)
> 
> New fs_umode_to_dtype() code:
>      Step 1: fs_umode_to_ftype(0010000)
>              = fs_ftype_by_dtype[S_DT(0010000)]
>              = fs_ftype_by_dtype[1]
>              = FT_FIFO = 5
> 
>      Step 2: fs_ftype_to_dtype(5)
>              = fs_dtype_by_ftype[5]
>              = DT_FIFO = 1
> 

Oh, that's fs_umode_to_dtype, sorry for the mistake, I saw it as 
'fs_umode_to_ftype'.

> I'm not sure how FT_FIFO (5) would remain after the function returns, since
> it seems to get converted back to DT_FIFO (1) in step 2.
> 
> If there's an issue with this approach, I can edit my patch and keep changes in only jffs2_mknod.
> .
> 


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] jffs2: use fs_umode_to_dtype() for dirent type
  2026-02-06 11:51 [PATCH] jffs2: use fs_umode_to_dtype() for dirent type Adarsh Das
  2026-02-07  2:36 ` Zhihao Cheng
@ 2026-02-07  8:00 ` Zhihao Cheng
  1 sibling, 0 replies; 5+ messages in thread
From: Zhihao Cheng @ 2026-02-07  8:00 UTC (permalink / raw)
  To: Adarsh Das, dwmw2, richard; +Cc: linux-mtd, linux-kernel

在 2026/2/6 19:51, Adarsh Das 写道:
> Use fs_umode_to_dtype() instead of direct bit shifting when
> setting directory entry type.
> 
> Signed-off-by: Adarsh Das <adarshdas950@gmail.com>
> ---
>   fs/jffs2/dir.c | 11 +++--------
>   1 file changed, 3 insertions(+), 8 deletions(-)
> 

Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
> diff --git a/fs/jffs2/dir.c b/fs/jffs2/dir.c
> index dd91f725ded6..185cf76a061e 100644
> --- a/fs/jffs2/dir.c
> +++ b/fs/jffs2/dir.c
> @@ -261,8 +261,7 @@ static int jffs2_link (struct dentry *old_dentry, struct inode *dir_i, struct de
>   	if (d_is_dir(old_dentry))
>   		return -EPERM;
>   
> -	/* XXX: This is ugly */
> -	type = (d_inode(old_dentry)->i_mode & S_IFMT) >> 12;
> +	type = fs_umode_to_dtype(d_inode(old_dentry)->i_mode);
>   	if (!type) type = DT_REG;
>   
>   	now = JFFS2_NOW();
> @@ -730,10 +729,7 @@ static int jffs2_mknod (struct mnt_idmap *idmap, struct inode *dir_i,
>   	rd->ino = cpu_to_je32(inode->i_ino);
>   	rd->mctime = cpu_to_je32(JFFS2_NOW());
>   	rd->nsize = namelen;
> -
> -	/* XXX: This is ugly. */
> -	rd->type = (mode & S_IFMT) >> 12;
> -
> +	rd->type = fs_umode_to_dtype(mode);
>   	rd->node_crc = cpu_to_je32(crc32(0, rd, sizeof(*rd)-8));
>   	rd->name_crc = cpu_to_je32(crc32(0, dentry->d_name.name, namelen));
>   
> @@ -811,8 +807,7 @@ static int jffs2_rename (struct mnt_idmap *idmap,
>   
>   	/* Make a hard link */
>   
> -	/* XXX: This is ugly */
> -	type = (d_inode(old_dentry)->i_mode & S_IFMT) >> 12;
> +	type = fs_umode_to_dtype(d_inode(old_dentry)->i_mode);
>   	if (!type) type = DT_REG;
>   
>   	now = JFFS2_NOW();
> 


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2026-02-07  8:00 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-02-06 11:51 [PATCH] jffs2: use fs_umode_to_dtype() for dirent type Adarsh Das
2026-02-07  2:36 ` Zhihao Cheng
2026-02-07  7:39   ` Adarsh Das
2026-02-07  7:59     ` Zhihao Cheng
2026-02-07  8:00 ` Zhihao Cheng

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox