From: Daeho Jeong <[email protected]>
introduce a new ioctl to replace the whole content of a file atomically,
which means it induces truncate and content update at the same time.
We can start it with F2FS_IOC_START_ATOMIC_REPLACE and complete it with
F2FS_IOC_COMMIT_ATOMIC_WRITE. Or abort it with
F2FS_IOC_ABORT_ATOMIC_WRITE.
Signed-off-by: Daeho Jeong <[email protected]>
---
v3: move i_size change after setting atomic write flag
v2: add undefined ioctl number reported by <[email protected]>
---
fs/f2fs/data.c | 3 +++
fs/f2fs/f2fs.h | 1 +
fs/f2fs/file.c | 20 ++++++++++++++------
fs/f2fs/segment.c | 14 +++++++++++++-
include/uapi/linux/f2fs.h | 1 +
5 files changed, 32 insertions(+), 7 deletions(-)
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 6cd29a575105..d3d32db3a25d 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -3438,6 +3438,9 @@ static int prepare_atomic_write_begin(struct f2fs_sb_info *sbi,
else if (*blk_addr != NULL_ADDR)
return 0;
+ if (is_inode_flag_set(inode, FI_ATOMIC_REPLACE))
+ goto reserve_block;
+
/* Look for the block in the original inode */
err = __find_data_block(inode, index, &ori_blk_addr);
if (err)
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 539da7f12cfc..2c49da12d6d8 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -764,6 +764,7 @@ enum {
FI_COMPRESS_RELEASED, /* compressed blocks were released */
FI_ALIGNED_WRITE, /* enable aligned write */
FI_COW_FILE, /* indicate COW file */
+ FI_ATOMIC_REPLACE, /* indicate atomic replace */
FI_MAX, /* max flag, never be used */
};
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index ce2336d2f688..66d62ea42c73 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -1982,7 +1982,7 @@ static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
return put_user(inode->i_generation, (int __user *)arg);
}
-static int f2fs_ioc_start_atomic_write(struct file *filp)
+static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate)
{
struct inode *inode = file_inode(filp);
struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
@@ -2051,10 +2051,6 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
f2fs_write_inode(inode, NULL);
- isize = i_size_read(inode);
- fi->original_i_size = isize;
- f2fs_i_size_write(fi->cow_inode, isize);
-
spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
sbi->atomic_files++;
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
@@ -2064,6 +2060,16 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
+ isize = i_size_read(inode);
+ fi->original_i_size = isize;
+ if (truncate) {
+ set_inode_flag(inode, FI_ATOMIC_REPLACE);
+ truncate_inode_pages_final(inode->i_mapping);
+ f2fs_i_size_write(inode, 0);
+ isize = 0;
+ }
+ f2fs_i_size_write(fi->cow_inode, isize);
+
f2fs_update_time(sbi, REQ_TIME);
fi->atomic_write_task = current;
stat_update_max_atomic_write(inode);
@@ -4082,7 +4088,9 @@ static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
case FS_IOC_GETVERSION:
return f2fs_ioc_getversion(filp, arg);
case F2FS_IOC_START_ATOMIC_WRITE:
- return f2fs_ioc_start_atomic_write(filp);
+ return f2fs_ioc_start_atomic_write(filp, false);
+ case F2FS_IOC_START_ATOMIC_REPLACE:
+ return f2fs_ioc_start_atomic_write(filp, true);
case F2FS_IOC_COMMIT_ATOMIC_WRITE:
return f2fs_ioc_commit_atomic_write(filp);
case F2FS_IOC_ABORT_ATOMIC_WRITE:
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index abb55cd418c1..e4f7efbdc027 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -264,14 +264,26 @@ static void __complete_revoke_list(struct inode *inode, struct list_head *head,
bool revoke)
{
struct revoke_entry *cur, *tmp;
+ pgoff_t start_index = 0;
+ bool truncate = is_inode_flag_set(inode, FI_ATOMIC_REPLACE);
list_for_each_entry_safe(cur, tmp, head, list) {
- if (revoke)
+ if (revoke) {
__replace_atomic_write_block(inode, cur->index,
cur->old_addr, NULL, true);
+ } else if (truncate) {
+ f2fs_truncate_hole(inode, start_index, cur->index);
+ start_index = cur->index + 1;
+ }
+
list_del(&cur->list);
kmem_cache_free(revoke_entry_slab, cur);
}
+
+ if (!revoke && truncate) {
+ f2fs_do_truncate_blocks(inode, start_index * PAGE_SIZE, false);
+ clear_inode_flag(inode, FI_ATOMIC_REPLACE);
+ }
}
static int __f2fs_commit_atomic_write(struct inode *inode)
diff --git a/include/uapi/linux/f2fs.h b/include/uapi/linux/f2fs.h
index 3121d127d5aa..955d440be104 100644
--- a/include/uapi/linux/f2fs.h
+++ b/include/uapi/linux/f2fs.h
@@ -42,6 +42,7 @@
struct f2fs_comp_option)
#define F2FS_IOC_DECOMPRESS_FILE _IO(F2FS_IOCTL_MAGIC, 23)
#define F2FS_IOC_COMPRESS_FILE _IO(F2FS_IOCTL_MAGIC, 24)
+#define F2FS_IOC_START_ATOMIC_REPLACE _IO(F2FS_IOCTL_MAGIC, 25)
/*
* should be same as XFS_IOC_GOINGDOWN.
--
2.38.0.rc1.362.ged0d419d3c-goog
On Tue, Oct 04, 2022 at 10:13:51AM -0700, Daeho Jeong wrote:
> From: Daeho Jeong <[email protected]>
>
> introduce a new ioctl to replace the whole content of a file atomically,
> which means it induces truncate and content update at the same time.
> We can start it with F2FS_IOC_START_ATOMIC_REPLACE and complete it with
> F2FS_IOC_COMMIT_ATOMIC_WRITE. Or abort it with
> F2FS_IOC_ABORT_ATOMIC_WRITE.
It would be great to Cc Darrick and linux-fsdevel as there have been
attempts to do this properly at the VFS level instead of a completely
undocumented ioctl.
On Thu, Oct 06, 2022 at 01:33:34AM -0700, Christoph Hellwig wrote:
> On Tue, Oct 04, 2022 at 10:13:51AM -0700, Daeho Jeong wrote:
> > From: Daeho Jeong <[email protected]>
> >
> > introduce a new ioctl to replace the whole content of a file atomically,
> > which means it induces truncate and content update at the same time.
> > We can start it with F2FS_IOC_START_ATOMIC_REPLACE and complete it with
> > F2FS_IOC_COMMIT_ATOMIC_WRITE. Or abort it with
> > F2FS_IOC_ABORT_ATOMIC_WRITE.
>
> It would be great to Cc Darrick and linux-fsdevel as there have been
> attempts to do this properly at the VFS level instead of a completely
> undocumented ioctl.
It's been a while since I sent the last RFC, but yes, it's still in my
queue as part of the xfs online fsck patchserieses.
https://lore.kernel.org/linux-fsdevel/161723932606.3149451.12366114306150243052.stgit@magnolia/
More recent git branch:
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/log/?h=atomic-file-updates
--D
On Thu, Oct 6, 2022 at 9:32 AM Darrick J. Wong <[email protected]> wrote:
>
> On Thu, Oct 06, 2022 at 01:33:34AM -0700, Christoph Hellwig wrote:
> > On Tue, Oct 04, 2022 at 10:13:51AM -0700, Daeho Jeong wrote:
> > > From: Daeho Jeong <[email protected]>
> > >
> > > introduce a new ioctl to replace the whole content of a file atomically,
> > > which means it induces truncate and content update at the same time.
> > > We can start it with F2FS_IOC_START_ATOMIC_REPLACE and complete it with
> > > F2FS_IOC_COMMIT_ATOMIC_WRITE. Or abort it with
> > > F2FS_IOC_ABORT_ATOMIC_WRITE.
> >
> > It would be great to Cc Darrick and linux-fsdevel as there have been
> > attempts to do this properly at the VFS level instead of a completely
> > undocumented ioctl.
>
> It's been a while since I sent the last RFC, but yes, it's still in my
> queue as part of the xfs online fsck patchserieses.
>
> https://lore.kernel.org/linux-fsdevel/161723932606.3149451.12366114306150243052.stgit@magnolia/
>
> More recent git branch:
> https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/log/?h=atomic-file-updates
Hi,
It's a very interesting suggestion and we might use this in F2FS someday.
However, I think it's not exactly matched for what
F2FS_IOC_START_ATOMIC_REPLACE is doing now.
Thanks for bringing my attention to this.
>
> --D