Re: [f2fs-dev] [PATCH v3] f2fs: introduce "fsync={posix, strict}" mount options

[Chao Yu]

On 2018/2/13 18:57, Junling Zheng wrote:
> Commit "0a007b97aad6"(f2fs: recover directory operations by fsync)
> fixed xfstest generic/342 case, but it also increased the written
> data and caused the performance degradation. In most cases, there's
> no need to do so heavy fsync actually.
> 
> So we introduce two new mount options "fsync={posix,strict}" to
> control the policy of fsync. "fsync=posix" is set by default, and
> means that f2fs uses a light fsync, which follows POSIX semantics.
> And "fsync=strict" means that it's a heavy fsync, which behaves in
> line with xfs, ext4 and btrfs, where generic/342 will pass, but the
> performance will regress.
> 
> Signed-off-by: Junling Zheng 
> ---
> Changes from v2:
>  - Change to "fsync={posix,strict}" format
>  - Set "fsync=posix" default
> Changes from v1:
>  - Add document modify
>  - Add reviewer
>  Documentation/filesystems/f2fs.txt |  7 +++
>  fs/f2fs/dir.c  |  3 ++-
>  fs/f2fs/f2fs.h |  1 +
>  fs/f2fs/file.c |  3 ++-
>  fs/f2fs/namei.c|  9 ++---
>  fs/f2fs/super.c| 10 ++
>  6 files changed, 28 insertions(+), 5 deletions(-)
> 
> diff --git a/Documentation/filesystems/f2fs.txt 
> b/Documentation/filesystems/f2fs.txt
> index 0caf7da0a532..c00d40655bbc 100644
> --- a/Documentation/filesystems/f2fs.txt
> +++ b/Documentation/filesystems/f2fs.txt
> @@ -180,6 +180,13 @@ whint_mode=%s  Control which write hints are 
> passed down to block
> down hints. In "user-based" mode, f2fs tries to pass
> down hints given by users. And in "fs-based" mode, 
> f2fs
> passes down hints with its policy.
> +fsync=%s   Control the policy of fsync. This supports "posix" and
> +   "strict" modes. In "posix" mode, which is default, 
> fsync
> +   will follow POSIX semantics and does a light 
> operation to
> +   improve the filesystem performance. In "strict" mode, 
> fsync
> +   will be heavy and behaves in line with xfs, ext4 and 
> btrfs,
> +   where xfstest generic/342 will pass, but the 
> performance
> +   will regress.
>  
>  
> 
>  DEBUGFS ENTRIES
> diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
> index f00b5ed8c011..37d1259f1b92 100644
> --- a/fs/f2fs/dir.c
> +++ b/fs/f2fs/dir.c
> @@ -713,7 +713,8 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, 
> struct page *page,
>  
>   f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
>  
> - add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
> + if (test_opt(F2FS_I_SB(dir), STRICT_FSYNC))
> + add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
>  
>   if (f2fs_has_inline_dentry(dir))
>   return f2fs_delete_inline_entry(dentry, page, dir, inode);
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index dbe87c7a266e..8cf914d12f17 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -97,6 +97,7 @@ extern char *fault_name[FAULT_MAX];
>  #define F2FS_MOUNT_QUOTA 0x0040
>  #define F2FS_MOUNT_INLINE_XATTR_SIZE 0x0080
>  #define F2FS_MOUNT_RESERVE_ROOT  0x0100
> +#define F2FS_MOUNT_STRICT_FSYNC  0x0200
>  
>  #define clear_opt(sbi, option)   ((sbi)->mount_opt.opt &= 
> ~F2FS_MOUNT_##option)
>  #define set_opt(sbi, option) ((sbi)->mount_opt.opt |= F2FS_MOUNT_##option)
> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> index 672a542e5464..509b3e045247 100644
> --- a/fs/f2fs/file.c
> +++ b/fs/f2fs/file.c
> @@ -165,7 +165,8 @@ static inline enum cp_reason_type 
> need_do_checkpoint(struct inode *inode)
>   cp_reason = CP_FASTBOOT_MODE;
>   else if (sbi->active_logs == 2)
>   cp_reason = CP_SPEC_LOG_NUM;
> - else if (need_dentry_mark(sbi, inode->i_ino) &&
> + else if (test_opt(sbi, STRICT_FSYNC) &&
> + need_dentry_mark(sbi, inode->i_ino) &&
>   exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
>   cp_reason = CP_RECOVER_DIR;
>  
> diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
> index c4c94c7e9f4f..5a385b8ab172 100644
> --- a/fs/f2fs/namei.c
> +++ b/fs/f2fs/namei.c
> @@ -936,7 +936,8 @@ static int f2fs_rename(struct inode *old_dir, struct 
> dentry *old_dentry,
>   }
>   f2fs_i_links_write(old_dir, false);
>   }
> - add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
> + if (test_opt(sbi, STRICT_FSYNC))
> + add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
>  
>   f2fs_unlock_op(sbi);
>  
> @@ -1091,8 +1092,10 @@ static int f2fs_cross_rename(struct inode *old_dir, 
> struct dentry *old_dentry,
>   }
>   f2fs_mark_inode_dirty_sync(new_dir, false);
>  
> - 

[f2fs-dev] [PATCH v3] f2fs: introduce "fsync={posix, strict}" mount options

[Junling Zheng]

Commit "0a007b97aad6"(f2fs: recover directory operations by fsync)
fixed xfstest generic/342 case, but it also increased the written
data and caused the performance degradation. In most cases, there's
no need to do so heavy fsync actually.

So we introduce two new mount options "fsync={posix,strict}" to
control the policy of fsync. "fsync=posix" is set by default, and
means that f2fs uses a light fsync, which follows POSIX semantics.
And "fsync=strict" means that it's a heavy fsync, which behaves in
line with xfs, ext4 and btrfs, where generic/342 will pass, but the
performance will regress.

Signed-off-by: Junling Zheng 
---
Changes from v2:
 - Change to "fsync={posix,strict}" format
 - Set "fsync=posix" default
Changes from v1:
 - Add document modify
 - Add reviewer
 Documentation/filesystems/f2fs.txt |  7 +++
 fs/f2fs/dir.c  |  3 ++-
 fs/f2fs/f2fs.h |  1 +
 fs/f2fs/file.c |  3 ++-
 fs/f2fs/namei.c|  9 ++---
 fs/f2fs/super.c| 10 ++
 6 files changed, 28 insertions(+), 5 deletions(-)

diff --git a/Documentation/filesystems/f2fs.txt 
b/Documentation/filesystems/f2fs.txt
index 0caf7da0a532..c00d40655bbc 100644
--- a/Documentation/filesystems/f2fs.txt
+++ b/Documentation/filesystems/f2fs.txt
@@ -180,6 +180,13 @@ whint_mode=%s  Control which write hints are 
passed down to block
down hints. In "user-based" mode, f2fs tries to pass
down hints given by users. And in "fs-based" mode, f2fs
passes down hints with its policy.
+fsync=%s   Control the policy of fsync. This supports "posix" and
+   "strict" modes. In "posix" mode, which is default, fsync
+   will follow POSIX semantics and does a light operation 
to
+   improve the filesystem performance. In "strict" mode, 
fsync
+   will be heavy and behaves in line with xfs, ext4 and 
btrfs,
+   where xfstest generic/342 will pass, but the performance
+   will regress.
 
 

 DEBUGFS ENTRIES
diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index f00b5ed8c011..37d1259f1b92 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -713,7 +713,8 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, 
struct page *page,
 
f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
 
-   add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
+   if (test_opt(F2FS_I_SB(dir), STRICT_FSYNC))
+   add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
 
if (f2fs_has_inline_dentry(dir))
return f2fs_delete_inline_entry(dentry, page, dir, inode);
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index dbe87c7a266e..8cf914d12f17 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -97,6 +97,7 @@ extern char *fault_name[FAULT_MAX];
 #define F2FS_MOUNT_QUOTA   0x0040
 #define F2FS_MOUNT_INLINE_XATTR_SIZE   0x0080
 #define F2FS_MOUNT_RESERVE_ROOT0x0100
+#define F2FS_MOUNT_STRICT_FSYNC0x0200
 
 #define clear_opt(sbi, option) ((sbi)->mount_opt.opt &= ~F2FS_MOUNT_##option)
 #define set_opt(sbi, option)   ((sbi)->mount_opt.opt |= F2FS_MOUNT_##option)
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 672a542e5464..509b3e045247 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -165,7 +165,8 @@ static inline enum cp_reason_type need_do_checkpoint(struct 
inode *inode)
cp_reason = CP_FASTBOOT_MODE;
else if (sbi->active_logs == 2)
cp_reason = CP_SPEC_LOG_NUM;
-   else if (need_dentry_mark(sbi, inode->i_ino) &&
+   else if (test_opt(sbi, STRICT_FSYNC) &&
+   need_dentry_mark(sbi, inode->i_ino) &&
exist_written_data(sbi, F2FS_I(inode)->i_pino, TRANS_DIR_INO))
cp_reason = CP_RECOVER_DIR;
 
diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
index c4c94c7e9f4f..5a385b8ab172 100644
--- a/fs/f2fs/namei.c
+++ b/fs/f2fs/namei.c
@@ -936,7 +936,8 @@ static int f2fs_rename(struct inode *old_dir, struct dentry 
*old_dentry,
}
f2fs_i_links_write(old_dir, false);
}
-   add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
+   if (test_opt(sbi, STRICT_FSYNC))
+   add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
 
f2fs_unlock_op(sbi);
 
@@ -1091,8 +1092,10 @@ static int f2fs_cross_rename(struct inode *old_dir, 
struct dentry *old_dentry,
}
f2fs_mark_inode_dirty_sync(new_dir, false);
 
-   add_ino_entry(sbi, old_dir->i_ino, TRANS_DIR_INO);
-   add_ino_entry(sbi, new_dir->i_ino, TRANS_DIR_INO);
+   if (test_opt(sbi, STRICT_FSYNC)) {
+   add_ino_entry(sbi, old_dir->i_ino, TRANS_DIR_INO);
+