Hey Linus,
/* Summary */
Note: This work came late in the cycle but the series is quite nice and
worth doing. It removes roughly double the code that it adds and
eliminates a lot of convoluted cleanup logic across the kernel.
An alternative pull request (vfs-6.19-rc1.fd_prepare.fs) is available
that contains only the more simple filesystem-focused conversions in
case you'd like to pull something more conservative.
Note this branch also contains two reverts for the KVM FD_PREPARE()
conversions as the KVM maintainers have indicated they would like to
take those changes through the KVM tree in the next cycle. Also gets rid
of a merge conflict. I chose a revert to not rebase the branch
unnecessarily so close to the merge window.
This adds the FD_ADD() and FD_PREPARE() primitive. They simplify the
common pattern of get_unused_fd_flags() + create file + fd_install()
that is used extensively throughout the kernel and currently requires
cumbersome cleanup paths.
FD_ADD() - For simple cases where a file is installed immediately:
fd = FD_ADD(O_CLOEXEC, vfio_device_open_file(device));
if (fd < 0)
vfio_device_put_registration(device);
return fd;
FD_PREPARE() - For cases requiring access to the fd or file, or
additional work before publishing:
FD_PREPARE(fdf, O_CLOEXEC, sync_file->file);
if (fdf.err) {
fput(sync_file->file);
return fdf.err;
}
data.fence = fd_prepare_fd(fdf);
if (copy_to_user((void __user *)arg, &data, sizeof(data)))
return -EFAULT;
return fd_publish(fdf);
The primitives are centered around struct fd_prepare. FD_PREPARE()
encapsulates all allocation and cleanup logic and must be followed by a
call to fd_publish() which associates the fd with the file and installs
it into the caller's fdtable. If fd_publish() isn't called, both are
deallocated automatically. FD_ADD() is a shorthand that does
fd_publish() immediately and never exposes the struct to the caller.
I've implemented this in a way that it's compatible with the cleanup
infrastructure while also being usable separately. IOW, it's centered
around struct fd_prepare which is aliased to class_fd_prepare_t and so
we can make use of all the basica guard infrastructure.
/* Testing */
gcc (Debian 14.2.0-19) 14.2.0
Debian clang version 19.1.7 (3+b1)
No build failures or warnings were observed.
/* Conflicts */
Merge conflicts with mainline or other vfs branches
===================================================
diff --cc include/linux/cleanup.h
index 19c7e475d3a4,361104bcfe92..b8bd2f15f91f
--- a/include/linux/cleanup.h
+++ b/include/linux/cleanup.h
@@@ -290,16 -294,18 +294,19 @@@ static inline class_##_name##_t class_#
class_##_name##_t var __cleanup(class_##_name##_destructor) = \
class_##_name##_constructor
+ #define CLASS_INIT(_name, _var, _init_expr) \
+ class_##_name##_t _var __cleanup(class_##_name##_destructor) = (_init_expr)
+
-#define scoped_class(_name, var, args) \
- for (CLASS(_name, var)(args); \
- __guard_ptr(_name)(&var) || !__is_cond_ptr(_name); \
- ({ goto _label; })) \
- if (0) { \
-_label: \
- break; \
+#define __scoped_class(_name, var, _label, args...) \
+ for (CLASS(_name, var)(args); ; ({ goto _label; })) \
+ if (0) { \
+_label: \
+ break; \
} else
+#define scoped_class(_name, var, args...) \
+ __scoped_class(_name, var, __UNIQUE_ID(label), args)
+
/*
* DEFINE_GUARD(name, type, lock, unlock):
* trivial wrapper around DEFINE_CLASS() above specifically
diff --cc ipc/mqueue.c
index 83d9466710d6,d3a588d0dcf6..c118ca2c377a
--- a/ipc/mqueue.c
+++ b/ipc/mqueue.c
@@@ -892,15 -892,36 +892,34 @@@ static int prepare_open(struct dentry *
return inode_permission(&nop_mnt_idmap, d_inode(dentry), acc);
}
+ static struct file *mqueue_file_open(struct filename *name,
+ struct vfsmount *mnt, int oflag, bool ro,
+ umode_t mode, struct mq_attr *attr)
+ {
- struct path path __free(path_put) = {};
+ struct dentry *dentry;
++ struct file *file;
+ int ret;
+
- dentry = lookup_noperm(&QSTR(name->name), mnt->mnt_root);
++ dentry = start_creating_noperm(mnt->mnt_root, &QSTR(name->name));
+ if (IS_ERR(dentry))
+ return ERR_CAST(dentry);
+
- path.dentry = dentry;
- path.mnt = mntget(mnt);
-
- ret = prepare_open(path.dentry, oflag, ro, mode, name, attr);
++ ret = prepare_open(dentry, oflag, ro, mode, name, attr);
+ if (ret)
- return ERR_PTR(ret);
-
- return dentry_open(&path, oflag, current_cred());
++ file = ERR_PTR(ret);
++ else
++ file = dentry_open(&(const struct path){ .mnt = mnt, .dentry = dentry },
++ oflag, current_cred());
++ end_creating(dentry);
++ return file;
+ }
+
static int do_mq_open(const char __user *u_name, int oflag, umode_t mode,
struct mq_attr *attr)
{
+ struct filename *name __free(putname) = NULL;;
struct vfsmount *mnt = current->nsproxy->ipc_ns->mq_mnt;
-- struct dentry *root = mnt->mnt_root;
- struct filename *name;
- struct path path;
- int fd, error;
- int fd;
-- int ro;
++ int fd, ro;
audit_mq_open(oflag, mode, attr);
@@@ -908,35 -929,12 +927,10 @@@
if (IS_ERR(name))
return PTR_ERR(name);
- fd = get_unused_fd_flags(O_CLOEXEC);
- if (fd < 0)
- goto out_putname;
-
ro = mnt_want_write(mnt); /* we'll drop it in any case */
- path.dentry = start_creating_noperm(root, &QSTR(name->name));
- if (IS_ERR(path.dentry)) {
- error = PTR_ERR(path.dentry);
- goto out_putfd;
- }
- path.mnt = mnt;
- error = prepare_open(path.dentry, oflag, ro, mode, name, attr);
- if (!error) {
- struct file *file = dentry_open(&path, oflag, current_cred());
- if (!IS_ERR(file))
- fd_install(fd, file);
- else
- error = PTR_ERR(file);
- }
- out_putfd:
- if (error) {
- put_unused_fd(fd);
- fd = error;
- }
- end_creating(path.dentry);
- inode_lock(d_inode(root));
+ fd = FD_ADD(O_CLOEXEC, mqueue_file_open(name, mnt, oflag, ro, mode, attr));
- inode_unlock(d_inode(root));
if (!ro)
mnt_drop_write(mnt);
- out_putname:
- putname(name);
return fd;
}
Merge conflicts with other trees
================================
[1]: https://lore.kernel.org/linux-next/20251125122934.36f75838@canb.auug.org.au
[2]: https://lore.kernel.org/linux-next/20251125171130.67ba74e1@canb.auug.org.au
The following changes since commit 3a8660878839faadb4f1a6dd72c3179c1df56787:
Linux 6.18-rc1 (2025-10-12 13:42:36 -0700)
are available in the Git repository at:
git@gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs tags/vfs-6.19-rc1.fd_prepare
for you to fetch changes up to 65c2c221846eeb157ab7cecf5a26f24d42faafcc:
Revert "kvm: FD_PREPARE() conversions" (2025-11-28 11:23:08 +0100)
Please consider pulling these changes from the signed vfs-6.19-rc1.fd_prepare tag.
Thanks!
Christian
----------------------------------------------------------------
vfs-6.19-rc1.fd_prepare
----------------------------------------------------------------
Christian Brauner (57):
file: add FD_{ADD,PREPARE}()
anon_inodes: convert to FD_ADD()
eventfd: convert do_eventfd() to FD_PREPARE()
fhandle: convert do_handle_open() to FD_ADD()
namespace: convert open_tree() to FD_ADD()
namespace: convert open_tree_attr() to FD_PREPARE()
namespace: convert fsmount() to FD_PREPARE()
fanotify: convert fanotify_init() to FD_PREPARE()
nsfs: convert open_namespace() to FD_PREPARE()
nsfs: convert ns_ioctl() to FD_PREPARE()
autofs: convert autofs_dev_ioctl_open_mountpoint() to FD_ADD()
eventpoll: convert do_epoll_create() to FD_PREPARE()
open: convert do_sys_openat2() to FD_ADD()
signalfd: convert do_signalfd4() to FD_ADD()
timerfd: convert timerfd_create() to FD_ADD()
userfaultfd: convert new_userfaultfd() to FD_PREPARE()
xfs: convert xfs_open_by_handle() to FD_PREPARE()
dma: convert dma_buf_fd() to FD_ADD()
af_unix: convert unix_file_open() to FD_ADD()
dma: convert sync_file_ioctl_merge() to FD_PREPARE()
exec: convert begin_new_exec() to FD_PREPARE()
ipc: convert do_mq_open() to FD_ADD()
bpf: convert bpf_iter_new_fd() to FD_PREPARE()
bpf: convert bpf_token_create() to FD_PREPARE()
memfd: convert memfd_create() to FD_ADD()
secretmem: convert memfd_secret() to FD_ADD()
net/handshake: convert handshake_nl_accept_doit() to FD_PREPARE()
net/kcm: convert kcm_ioctl() to FD_PREPARE()
net/sctp: convert sctp_getsockopt_peeloff_common() to FD_PREPARE()
net/socket: convert sock_map_fd() to FD_ADD()
net/socket: convert __sys_accept4_file() to FD_ADD()
spufs: convert spufs_context_open() to FD_PREPARE()
papr-hvpipe: convert papr_hvpipe_dev_create_handle() to FD_PREPARE()
spufs: convert spufs_gang_open() to FD_PREPARE()
pseries: convert papr_platform_dump_create_handle() to FD_ADD()
pseries: port papr_rtas_setup_file_interface() to FD_ADD()
dma: port sw_sync_ioctl_create_fence() to FD_PREPARE()
gpio: convert linehandle_create() to FD_PREPARE()
hv: convert mshv_ioctl_create_partition() to FD_ADD()
media: convert media_request_alloc() to FD_PREPARE()
ntsync: convert ntsync_obj_get_fd() to FD_PREPARE()
tty: convert ptm_open_peer() to FD_ADD()
vfio: convert vfio_group_ioctl_get_device_fd() to FD_ADD()
file: convert replace_fd() to FD_PREPARE()
io_uring: convert io_create_mock_file() to FD_PREPARE()
kvm: convert kvm_arch_supports_gmem_init_shared() to FD_PREPARE()
kvm: convert kvm_vcpu_ioctl_get_stats_fd() to FD_PREPARE()
Merge patch series "file: FD_{ADD,PREPARE}()"
ipc: preserve original file opening pattern
devpts: preserve original file opening pattern
dma: return zero after fd_publish()
exec: switch to FD_ADD()
handshake: return zero after fd_publish()
ntsync: only install fd on success
io_uring: return zero after fd_publish()
file: make struct fd_prepare a first-class citizen
Revert "kvm: FD_PREPARE() conversions"
Deepanshu Kartikey (1):
namespace: fix mntput of ERR_PTR in fsmount error path
Kuniyuki Iwashima (1):
fanotify: Don't call fsnotify_destroy_group() when fsnotify_alloc_group() fails.
arch/powerpc/platforms/cell/spufs/inode.c | 42 ++-----
arch/powerpc/platforms/pseries/papr-hvpipe.c | 39 ++-----
.../powerpc/platforms/pseries/papr-platform-dump.c | 30 ++---
arch/powerpc/platforms/pseries/papr-rtas-common.c | 27 +----
drivers/dma-buf/dma-buf.c | 10 +-
drivers/dma-buf/sw_sync.c | 39 +++----
drivers/dma-buf/sync_file.c | 53 +++------
drivers/gpio/gpiolib-cdev.c | 66 ++++-------
drivers/hv/mshv_root_main.c | 30 +----
drivers/media/mc/mc-request.c | 34 ++----
drivers/misc/ntsync.c | 21 +---
drivers/tty/pty.c | 51 +++------
drivers/vfio/group.c | 28 +----
fs/anon_inodes.c | 23 +---
fs/autofs/dev-ioctl.c | 30 +----
fs/eventfd.c | 31 ++---
fs/eventpoll.c | 32 ++----
fs/exec.c | 3 +-
fs/fhandle.c | 30 +++--
fs/file.c | 19 ++--
fs/namespace.c | 103 ++++++-----------
fs/notify/fanotify/fanotify_user.c | 60 ++++------
fs/nsfs.c | 47 +++-----
fs/open.c | 17 +--
fs/signalfd.c | 29 ++---
fs/timerfd.c | 29 ++---
fs/userfaultfd.c | 30 ++---
fs/xfs/xfs_handle.c | 56 +++------
include/linux/cleanup.h | 7 ++
include/linux/file.h | 126 +++++++++++++++++++++
io_uring/mock_file.c | 43 +++----
ipc/mqueue.c | 54 ++++-----
kernel/bpf/bpf_iter.c | 29 ++---
kernel/bpf/token.c | 47 +++-----
mm/memfd.c | 29 +----
mm/secretmem.c | 20 +---
net/handshake/netlink.c | 38 +++----
net/kcm/kcmsock.c | 22 ++--
net/sctp/socket.c | 90 ++++-----------
net/socket.c | 34 +-----
net/unix/af_unix.c | 16 +--
41 files changed, 564 insertions(+), 1000 deletions(-)On Fri, Nov 28, 2025 at 05:48:27PM +0100, Christian Brauner wrote: > Hey Linus, > > /* Summary */ > Note: This work came late in the cycle but the series is quite nice and > worth doing. It removes roughly double the code that it adds and > eliminates a lot of convoluted cleanup logic across the kernel. > > An alternative pull request (vfs-6.19-rc1.fd_prepare.fs) is available > that contains only the more simple filesystem-focused conversions in > case you'd like to pull something more conservative. > > Note this branch also contains two reverts for the KVM FD_PREPARE() > conversions as the KVM maintainers have indicated they would like to > take those changes through the KVM tree in the next cycle. Also gets rid > of a merge conflict. I chose a revert to not rebase the branch > unnecessarily so close to the merge window. Frankly, that hadn't gotten anywhere near enough exposure in -next and it's far too large and invasive. The same lack of exposure goes for the alternative branch.
On Mon, Dec 01, 2025, Al Viro wrote: > On Fri, Nov 28, 2025 at 05:48:27PM +0100, Christian Brauner wrote: > > Hey Linus, > > > > /* Summary */ > > Note: This work came late in the cycle but the series is quite nice and > > worth doing. It removes roughly double the code that it adds and > > eliminates a lot of convoluted cleanup logic across the kernel. > > > > An alternative pull request (vfs-6.19-rc1.fd_prepare.fs) is available > > that contains only the more simple filesystem-focused conversions in > > case you'd like to pull something more conservative. > > > > Note this branch also contains two reverts for the KVM FD_PREPARE() > > conversions as the KVM maintainers have indicated they would like to > > take those changes through the KVM tree in the next cycle. Also gets rid > > of a merge conflict. I chose a revert to not rebase the branch > > unnecessarily so close to the merge window. > > Frankly, that hadn't gotten anywhere near enough exposure in -next and > it's far too large and invasive. +1. Saying that I want to take the KVM changes through the KVM tree is technically true, but glosses over why I objected (or even noticed) in the first place. https://lore.kernel.org/all/20251125155455.31c53cf9@canb.auug.org.au
© 2016 - 2025 Red Hat, Inc.