[PATCH v6 0/2] fuse: atomic open

[Jimmy Zuber <jamz@amazon.com>]

Add FUSE_OPEN_ATOMIC, allowing the kernel to skip LOOKUP to improve
performance. A server implementing FUSE_OPEN_ATOMIC will look up
the name, create it if O_CREAT is supplied, and open it in one request.

The kernel sends FUSE_OPEN_ATOMIC from ->atomic_open() for names
not in the dcache. This replaces a LOOKUP and OPEN or a LOOKUP and
CREATE with a single FUSE_OPEN_ATOMIC. Names in the dcache are not
supported by FUSE_OPEN_ATOMIC, since that needs a VFS change.
FUSE_OPEN_ATOMIC does not carry O_TRUNC, so that the VFS can evaluate
permissions before truncating. Servers that don't support
FUSE_OPEN_ATOMIC are handled gracefully.

This will result in servers seeing opens that the kernel has not yet
permission checked, as with ->atomic_open() on other filesystems. If
the check fails, the kernel will release the opened file.

This series applies on top of "fuse: fix use-after-free in
fuse_create_open() error path" [1].

Tested with a new selftest under UML and KVM.

Changes in v6 (from the Sashiko review of v5 [5]):
- Keep FUSE_LOOKUP for O_EXCL without O_CREAT, which the VFS passes when
  it may not create the file, so that an existing name gives -EEXIST.
- With FUSE_ATOMIC_O_TRUNC, check O_TRUNC in file->f_flags, which keeps
  it on a read-only mount.
- Check that an opened reply is a regular file before fuse_iget().
- Test read-only mounts.

Changes in v5 (from the Sashiko review of v4 [4]):
- Limit the series to ->atomic_open() and its test. Opening cached
  names (v4 patch 2) moves to a later series.
- Let the VFS check the open after the reply. The request carries the
  open flags without O_TRUNC, and the VFS truncates with SETATTR.
- Handle FUSE_OPEN_ATOMIC in its own function; FUSE_CREATE is unchanged.

Changes in v4 (from the Sashiko review of v3 [3]): a LOOKUP when the
VFS drops O_CREAT or O_TRUNC, permission rechecks, and tests for them.

Changes in v3 (from the Sashiko review of v2 [2]): reply validation
(FOPEN_CREATED, name_max, non-regular and bad inodes, DAX).

Changes in v2: d_splice_alias() for all replies, documentation of the
server's side (NeilBrown), and a rebase onto fuse for-next.

v5: https://lore.kernel.org/fuse-devel/20261028200459.2922951-1-jamz@amazon.com/
v4: https://lore.kernel.org/fuse-devel/20261026221639.628325-1-jamz@amazon.com/
v3: https://lore.kernel.org/fuse-devel/20261026190747.2373775-1-jamz@amazon.com/
v2: https://lore.kernel.org/fuse-devel/20261026151828.562870-1-jamz@amazon.com/
v1: https://lore.kernel.org/fuse-devel/20261020141647.3271341-1-jamz@amazon.com/

[1] https://lore.kernel.org/fuse-devel/20261026180649.1265825-1-jamz@amazon.com/
[2] https://sashiko.dev/#/patchset/20261026151828.562870-1-jamz@amazon.com
[3] https://sashiko.dev/#/patchset/20261026190747.2373775-1-jamz@amazon.com
[4] https://sashiko.dev/#/patchset/20261026221639.628325-1-jamz@amazon.com
[5] https://sashiko.dev/#/patchset/20261028200459.2922951-1-jamz@amazon.com

Jimmy Zuber (2):
  fuse: add FUSE_OPEN_ATOMIC
  selftests/fuse: add test for FUSE_OPEN_ATOMIC

 fs/fuse/dir.c                                 | 193 ++++++
 fs/fuse/fuse_i.h                              |   6 +
 fs/fuse/fuse_trace.h                          |   1 +
 include/uapi/linux/fuse.h                     |  20 +-
 .../selftests/filesystems/fuse/.gitignore     |   1 +
 .../selftests/filesystems/fuse/Makefile       |   3 +-
 .../filesystems/fuse/atomic_open_test.c       | 572 ++++++++++++++++++
 7 files changed, 794 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/filesystems/fuse/atomic_open_test.c


base-commit: 9b1da7b4a4e197072e630c3b85d4100a399752ff
prerequisite-patch-id: 95a6559bc4490d7d82bc80ba961ac52e80429b34
-- 
2.50.1