2021-10-05 15:57:26

by Konstantin Komarov

[permalink] [raw]
Subject: [PATCH v2] fs/ntfs3: Keep prealloc for all types of files

Before we haven't kept prealloc for sparse files because we thought that
it will speed up create / write operations.
It lead to situation, when user reserved some space for sparse file,
filled volume, and wasn't able to write in reserved file.
With this commit we keep prealloc.
Now xfstest generic/274 pass.
Fixes: be71b5cba2e6 ("fs/ntfs3: Add attrib operations")

Signed-off-by: Konstantin Komarov <[email protected]>
---
v2:
Expanded commit message.

fs/ntfs3/attrib.c | 8 +-------
1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/fs/ntfs3/attrib.c b/fs/ntfs3/attrib.c
index 8a00fa978f5f..e8c00dda42ad 100644
--- a/fs/ntfs3/attrib.c
+++ b/fs/ntfs3/attrib.c
@@ -447,11 +447,8 @@ int attr_set_size(struct ntfs_inode *ni, enum ATTR_TYPE type,
again_1:
align = sbi->cluster_size;

- if (is_ext) {
+ if (is_ext)
align <<= attr_b->nres.c_unit;
- if (is_attr_sparsed(attr_b))
- keep_prealloc = false;
- }

old_valid = le64_to_cpu(attr_b->nres.valid_size);
old_size = le64_to_cpu(attr_b->nres.data_size);
@@ -461,9 +458,6 @@ int attr_set_size(struct ntfs_inode *ni, enum ATTR_TYPE type,
new_alloc = (new_size + align - 1) & ~(u64)(align - 1);
new_alen = new_alloc >> cluster_bits;

- if (keep_prealloc && is_ext)
- keep_prealloc = false;
-
if (keep_prealloc && new_size < old_size) {
attr_b->nres.data_size = cpu_to_le64(new_size);
mi_b->dirty = true;
--
2.33.0


2021-10-09 15:57:21

by Kari Argillander

[permalink] [raw]
Subject: Re: [PATCH v2] fs/ntfs3: Keep prealloc for all types of files

On Tue, Oct 05, 2021 at 06:55:44PM +0300, Konstantin Komarov wrote:
> Before we haven't kept prealloc for sparse files because we thought that
> it will speed up create / write operations.
> It lead to situation, when user reserved some space for sparse file,
> filled volume, and wasn't able to write in reserved file.
> With this commit we keep prealloc.
> Now xfstest generic/274 pass.

Is it possible that you make repo for xfstests which contain changes
what you have done for it. Testing example 274 will need modifications
to xfstests so testing this will take too much time for me.

Hopefully we will get ntfs3 support to upstream xfstest but I totally
understand if you do not have time for it. That is why I ask that just
make repo for github and just but your changes to there so that someone
can make upstream support.

Argillander

> Fixes: be71b5cba2e6 ("fs/ntfs3: Add attrib operations")
>
> Signed-off-by: Konstantin Komarov <[email protected]>
> ---
> v2:
> Expanded commit message.
>
> fs/ntfs3/attrib.c | 8 +-------
> 1 file changed, 1 insertion(+), 7 deletions(-)
>
> diff --git a/fs/ntfs3/attrib.c b/fs/ntfs3/attrib.c
> index 8a00fa978f5f..e8c00dda42ad 100644
> --- a/fs/ntfs3/attrib.c
> +++ b/fs/ntfs3/attrib.c
> @@ -447,11 +447,8 @@ int attr_set_size(struct ntfs_inode *ni, enum ATTR_TYPE type,
> again_1:
> align = sbi->cluster_size;
>
> - if (is_ext) {
> + if (is_ext)
> align <<= attr_b->nres.c_unit;
> - if (is_attr_sparsed(attr_b))
> - keep_prealloc = false;
> - }
>
> old_valid = le64_to_cpu(attr_b->nres.valid_size);
> old_size = le64_to_cpu(attr_b->nres.data_size);
> @@ -461,9 +458,6 @@ int attr_set_size(struct ntfs_inode *ni, enum ATTR_TYPE type,
> new_alloc = (new_size + align - 1) & ~(u64)(align - 1);
> new_alen = new_alloc >> cluster_bits;
>
> - if (keep_prealloc && is_ext)
> - keep_prealloc = false;
> -
> if (keep_prealloc && new_size < old_size) {
> attr_b->nres.data_size = cpu_to_le64(new_size);
> mi_b->dirty = true;
> --
> 2.33.0
>