Hi all,
This series creates XFS_IOC_START_COMMIT and XFS_IOC_COMMIT_RANGE ioctls
to perform the exchange only if the target file has not been changed
since a given sampling point.
This new functionality uses the mechanism underlying EXCHANGE_RANGE to
stage and commit file updates such that reader programs will see either
the old contents or the new contents in their entirety, with no chance
of torn writes. A successful call completion guarantees that the new
contents will be seen even if the system fails. The pair of ioctls
allows userspace to perform what amounts to a compare and exchange
operation on entire file contents.
Note that there are ongoing arguments in the community about how best to
implement some sort of file data write counter that nfsd could also use
to signal invalidations to clients. Until such a thing is implemented,
this patch will rely on ctime/mtime updates.
Here are the proposed manual pages:
IOCTL-XFS-COMMIT-RANGE(2) System Calls ManualIOCTL-XFS-COMMIT-RANGE(2)
NAME
ioctl_xfs_start_commit - prepare to exchange the contents of
two files ioctl_xfs_commit_range - conditionally exchange the
contents of parts of two files
SYNOPSIS
#include <sys/ioctl.h>
#include <xfs/xfs_fs.h>
int ioctl(int file2_fd, XFS_IOC_START_COMMIT, struct xfs_com‐
mit_range *arg);
int ioctl(int file2_fd, XFS_IOC_COMMIT_RANGE, struct xfs_com‐
mit_range *arg);
DESCRIPTION
Given a range of bytes in a first file file1_fd and a second
range of bytes in a second file file2_fd, this ioctl(2) ex‐
changes the contents of the two ranges if file2_fd passes cer‐
tain freshness criteria.
Before exchanging the contents, the program must call the
XFS_IOC_START_COMMIT ioctl to sample freshness data for
file2_fd. If the sampled metadata does not match the file
metadata at commit time, XFS_IOC_COMMIT_RANGE will return
EBUSY.
Exchanges are atomic with regards to concurrent file opera‐
tions. Implementations must guarantee that readers see either
the old contents or the new contents in their entirety, even if
the system fails.
The system call parameters are conveyed in structures of the
following form:
struct xfs_commit_range {
__s32 file1_fd;
__u32 pad;
__u64 file1_offset;
__u64 file2_offset;
__u64 length;
__u64 flags;
__u64 file2_freshness[5];
};
The field pad must be zero.
The fields file1_fd, file1_offset, and length define the first
range of bytes to be exchanged.
The fields file2_fd, file2_offset, and length define the second
range of bytes to be exchanged.
The field file2_freshness is an opaque field whose contents are
determined by the kernel. These file attributes are used to
confirm that file2_fd has not changed by another thread since
the current thread began staging its own update.
Both files must be from the same filesystem mount. If the two
file descriptors represent the same file, the byte ranges must
not overlap. Most disk-based filesystems require that the
starts of both ranges must be aligned to the file block size.
If this is the case, the ends of the ranges must also be so
aligned unless the XFS_EXCHANGE_RANGE_TO_EOF flag is set.
The field flags control the behavior of the exchange operation.
XFS_EXCHANGE_RANGE_TO_EOF
Ignore the length parameter. All bytes in file1_fd
from file1_offset to EOF are moved to file2_fd, and
file2's size is set to (file2_offset+(file1_length-
file1_offset)). Meanwhile, all bytes in file2 from
file2_offset to EOF are moved to file1 and file1's
size is set to (file1_offset+(file2_length-
file2_offset)).
XFS_EXCHANGE_RANGE_DSYNC
Ensure that all modified in-core data in both file
ranges and all metadata updates pertaining to the
exchange operation are flushed to persistent storage
before the call returns. Opening either file de‐
scriptor with O_SYNC or O_DSYNC will have the same
effect.
XFS_EXCHANGE_RANGE_FILE1_WRITTEN
Only exchange sub-ranges of file1_fd that are known
to contain data written by application software.
Each sub-range may be expanded (both upwards and
downwards) to align with the file allocation unit.
For files on the data device, this is one filesystem
block. For files on the realtime device, this is
the realtime extent size. This facility can be used
to implement fast atomic scatter-gather writes of
any complexity for software-defined storage targets
if all writes are aligned to the file allocation
unit.
XFS_EXCHANGE_RANGE_DRY_RUN
Check the parameters and the feasibility of the op‐
eration, but do not change anything.
RETURN VALUE
On error, -1 is returned, and errno is set to indicate the er‐
ror.
ERRORS
Error codes can be one of, but are not limited to, the follow‐
ing:
EBADF file1_fd is not open for reading and writing or is open
for append-only writes; or file2_fd is not open for
reading and writing or is open for append-only writes.
EBUSY The file2 inode number and timestamps supplied do not
match file2_fd.
EINVAL The parameters are not correct for these files. This
error can also appear if either file descriptor repre‐
sents a device, FIFO, or socket. Disk filesystems gen‐
erally require the offset and length arguments to be
aligned to the fundamental block sizes of both files.
EIO An I/O error occurred.
EISDIR One of the files is a directory.
ENOMEM The kernel was unable to allocate sufficient memory to
perform the operation.
ENOSPC There is not enough free space in the filesystem ex‐
change the contents safely.
EOPNOTSUPP
The filesystem does not support exchanging bytes between
the two files.
EPERM file1_fd or file2_fd are immutable.
ETXTBSY
One of the files is a swap file.
EUCLEAN
The filesystem is corrupt.
EXDEV file1_fd and file2_fd are not on the same mounted
filesystem.
CONFORMING TO
This API is XFS-specific.
USE CASES
Several use cases are imagined for this system call. Coordina‐
tion between multiple threads is performed by the kernel.
The first is a filesystem defragmenter, which copies the con‐
tents of a file into another file and wishes to exchange the
space mappings of the two files, provided that the original
file has not changed.
An example program might look like this:
int fd = open("/some/file", O_RDWR);
int temp_fd = open("/some", O_TMPFILE | O_RDWR);
struct stat sb;
struct xfs_commit_range args = {
.flags = XFS_EXCHANGE_RANGE_TO_EOF,
};
/* gather file2's freshness information */
ioctl(fd, XFS_IOC_START_COMMIT, &args);
fstat(fd, &sb);
/* make a fresh copy of the file with terrible alignment to avoid reflink */
clone_file_range(fd, NULL, temp_fd, NULL, 1, 0);
clone_file_range(fd, NULL, temp_fd, NULL, sb.st_size - 1, 0);
/* commit the entire update */
args.file1_fd = temp_fd;
ret = ioctl(fd, XFS_IOC_COMMIT_RANGE, &args);
if (ret && errno == EBUSY)
printf("file changed while defrag was underway
");
The second is a data storage program that wants to commit non-
contiguous updates to a file atomically. This program cannot
coordinate updates to the file and therefore relies on the ker‐
nel to reject the COMMIT_RANGE command if the file has been up‐
dated by someone else. This can be done by creating a tempo‐
rary file, calling FICLONE(2) to share the contents, and stag‐
ing the updates into the temporary file. The FULL_FILES flag
is recommended for this purpose. The temporary file can be
deleted or punched out afterwards.
An example program might look like this:
int fd = open("/some/file", O_RDWR);
int temp_fd = open("/some", O_TMPFILE | O_RDWR);
struct xfs_commit_range args = {
.flags = XFS_EXCHANGE_RANGE_TO_EOF,
};
/* gather file2's freshness information */
ioctl(fd, XFS_IOC_START_COMMIT, &args);
ioctl(temp_fd, FICLONE, fd);
/* append 1MB of records */
lseek(temp_fd, 0, SEEK_END);
write(temp_fd, data1, 1000000);
/* update record index */
pwrite(temp_fd, data1, 600, 98765);
pwrite(temp_fd, data2, 320, 54321);
pwrite(temp_fd, data2, 15, 0);
/* commit the entire update */
args.file1_fd = temp_fd;
ret = ioctl(fd, XFS_IOC_COMMIT_RANGE, &args);
if (ret && errno == EBUSY)
printf("file changed before commit; will roll back
");
NOTES
Some filesystems may limit the amount of data or the number of
extents that can be exchanged in a single call.
SEE ALSO
ioctl(2)
XFS 2024-02-18 IOCTL-XFS-COMMIT-RANGE(2)
If you're going to start using this code, I strongly recommend pulling
from my git trees, which are linked below.
This has been running on the djcloud for months with no problems. Enjoy!
Comments and questions are, as always, welcome.
--D
kernel git tree:
https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=atomic-file-commits
xfsprogs git tree:
https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=atomic-file-commits
fstests git tree:
https://git.kernel.org/cgit/linux/kernel/git/djwong/xfstests-dev.git/log/?h=atomic-file-commits
---
Commits in this patchset:
* xfs: introduce new file range commit ioctls
---
fs/xfs/libxfs/xfs_fs.h | 26 +++++++++
fs/xfs/xfs_exchrange.c | 143 ++++++++++++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_exchrange.h | 16 +++++
fs/xfs/xfs_ioctl.c | 4 +
fs/xfs/xfs_trace.h | 57 +++++++++++++++++++
5 files changed, 243 insertions(+), 3 deletions(-)
