[PATCH bpf-next 0/2] bpf: Add more dentry kfuncs for BPF LSM programs

[PATCH bpf-next 0/2] bpf: Add more dentry kfuncs for BPF LSM programs

[David Windsor]

BPF LSM programs often need to perform path-based access control and
security monitoring that requires walking filesystem structures. Currently,
BPF programs can access basic file information but lack the ability to
safely navigate dentry relationships or perform reference-counted
operations on filesystem objects.

This series extends the existing collection of filesystem kfuncs by adding
dentry-specific operations that enable BPF LSM programs to:

1. Safely acquire and release dentry references (bpf_dget/bpf_dput)
2. Navigate parent-child relationships (bpf_dget_parent)
3. Find dentry aliases for inodes (bpf_d_find_alias)
4. Access trusted dentry/vfsmount pointers from files
   (bpf_file_dentry/bpf_file_vfsmount)

David Windsor (2):
  bpf: Add dentry kfuncs for BPF LSM programs
  selftests/bpf: Add tests for dentry kfuncs

 fs/bpf_fs_kfuncs.c                            | 104 ++++++++++++++++++
 .../selftests/bpf/prog_tests/dentry_lsm.c     |  48 ++++++++
 .../testing/selftests/bpf/progs/dentry_lsm.c  |  51 +++++++++
 3 files changed, 203 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/dentry_lsm.c
 create mode 100644 tools/testing/selftests/bpf/progs/dentry_lsm.c

-- 
2.43.0

[PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[David Windsor]

Add six new BPF kfuncs that enable BPF LSM programs to safely interact
with dentry objects:

- bpf_dget(): Acquire reference on dentry
- bpf_dput(): Release reference on dentry
- bpf_dget_parent(): Get referenced parent dentry
- bpf_d_find_alias(): Find referenced alias dentry for inode
- bpf_file_dentry(): Get dentry from file
- bpf_file_vfsmount(): Get vfsmount from file

All kfuncs are currently restricted to BPF_PROG_TYPE_LSM programs.

Signed-off-by: David Windsor <dwindsor@gmail.com>
---
 fs/bpf_fs_kfuncs.c | 104 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 104 insertions(+)

diff --git a/fs/bpf_fs_kfuncs.c b/fs/bpf_fs_kfuncs.c
index 1e36a12b88f7..988e408fe7b3 100644
--- a/fs/bpf_fs_kfuncs.c
+++ b/fs/bpf_fs_kfuncs.c
@@ -169,6 +169,104 @@ __bpf_kfunc int bpf_get_file_xattr(struct file *file, const char *name__str,
 	return bpf_get_dentry_xattr(dentry, name__str, value_p);
 }
 
+/**
+ * bpf_dget - get a reference on a dentry
+ * @dentry: dentry to get a reference on
+ *
+ * Get a reference on the supplied *dentry*. The referenced dentry pointer
+ * acquired by this BPF kfunc must be released using bpf_dput().
+ *
+ * This BPF kfunc may only be called from BPF LSM programs.
+ *
+ * Return: A referenced dentry pointer. On error, NULL is returned.
+ */
+__bpf_kfunc struct dentry *bpf_dget(struct dentry *dentry)
+{
+	return dget(dentry);
+}
+
+/**
+ * bpf_dput - put a reference on a dentry
+ * @dentry: dentry to put a reference on
+ *
+ * Put a reference on the supplied *dentry*.
+ *
+ * This BPF kfunc may only be called from BPF LSM programs.
+ */
+__bpf_kfunc void bpf_dput(struct dentry *dentry)
+{
+	dput(dentry);
+}
+
+/**
+ * bpf_dget_parent - get a reference on the parent dentry
+ * @dentry: dentry to get the parent of
+ *
+ * Get a reference on the parent of the supplied *dentry*. The referenced
+ * dentry pointer acquired by this BPF kfunc must be released using bpf_dput().
+ *
+ * This BPF kfunc may only be called from BPF LSM programs.
+ *
+ * Return: A referenced parent dentry pointer. On error, NULL is returned.
+ */
+__bpf_kfunc struct dentry *bpf_dget_parent(struct dentry *dentry)
+{
+	return dget_parent(dentry);
+}
+
+/**
+ * bpf_d_find_alias - find an alias dentry for an inode
+ * @inode: inode to find an alias for
+ *
+ * Find an alias dentry for the supplied *inode*. The referenced dentry pointer
+ * acquired by this BPF kfunc must be released using bpf_dput().
+ *
+ * This BPF kfunc may only be called from BPF LSM programs.
+ *
+ * Return: A referenced alias dentry pointer. On error, NULL is returned.
+ */
+__bpf_kfunc struct dentry *bpf_d_find_alias(struct inode *inode)
+{
+	return d_find_alias(inode);
+}
+
+/**
+ * bpf_file_dentry - get the dentry associated with a file
+ * @file: file to get the dentry from
+ *
+ * Get the dentry associated with the supplied *file*. This is a trusted
+ * accessor that allows BPF programs to safely obtain a dentry pointer
+ * from a file structure. The returned pointer is borrowed and does not
+ * require bpf_dput().
+ *
+ * This BPF kfunc may only be called from BPF LSM programs.
+ *
+ * Return: A dentry pointer. On error, NULL is returned.
+ */
+__bpf_kfunc struct dentry *bpf_file_dentry(struct file *file)
+{
+	return file_dentry(file);
+}
+
+/**
+ * bpf_file_vfsmount - get the vfsmount associated with a file
+ * @file: file to get the vfsmount from
+ *
+ * Get the vfsmount associated with the supplied *file*. This is a trusted
+ * accessor that allows BPF programs to safely obtain a vfsmount pointer
+ * from a file structure. The returned pointer is borrowed and does not
+ * require any release function.
+ *
+ * This BPF kfunc may only be called from BPF LSM programs.
+ *
+ * Return: A vfsmount pointer. On error, NULL is returned.
+ */
+__bpf_kfunc struct vfsmount *bpf_file_vfsmount(struct file *file)
+{
+	return file->f_path.mnt;
+}
+
+
 __bpf_kfunc_end_defs();
 
 static int bpf_xattr_write_permission(const char *name, struct inode *inode)
@@ -367,6 +465,12 @@ BTF_ID_FLAGS(func, bpf_get_dentry_xattr, KF_SLEEPABLE | KF_TRUSTED_ARGS)
 BTF_ID_FLAGS(func, bpf_get_file_xattr, KF_SLEEPABLE | KF_TRUSTED_ARGS)
 BTF_ID_FLAGS(func, bpf_set_dentry_xattr, KF_SLEEPABLE | KF_TRUSTED_ARGS)
 BTF_ID_FLAGS(func, bpf_remove_dentry_xattr, KF_SLEEPABLE | KF_TRUSTED_ARGS)
+BTF_ID_FLAGS(func, bpf_dget, KF_ACQUIRE | KF_RET_NULL)
+BTF_ID_FLAGS(func, bpf_dput, KF_RELEASE)
+BTF_ID_FLAGS(func, bpf_dget_parent, KF_ACQUIRE | KF_RET_NULL)
+BTF_ID_FLAGS(func, bpf_d_find_alias, KF_ACQUIRE | KF_RET_NULL)
+BTF_ID_FLAGS(func, bpf_file_dentry, KF_TRUSTED_ARGS | KF_RET_NULL)
+BTF_ID_FLAGS(func, bpf_file_vfsmount, KF_TRUSTED_ARGS | KF_RET_NULL)
 BTF_KFUNCS_END(bpf_fs_kfunc_set_ids)
 
 static int bpf_fs_kfuncs_filter(const struct bpf_prog *prog, u32 kfunc_id)
-- 
2.43.0

[PATCH bpf-next 2/2] selftests/bpf: Add tests for dentry kfuncs

[David Windsor]

Add BPF selftests that exercise the new dentry kfuncs via an LSM program
attached to the file_open hook.

Signed-off-by: David Windsor <dwindsor@gmail.com>
---
 .../selftests/bpf/prog_tests/dentry_lsm.c     | 48 +++++++++++++++++
 .../testing/selftests/bpf/progs/dentry_lsm.c  | 51 +++++++++++++++++++
 2 files changed, 99 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/dentry_lsm.c
 create mode 100644 tools/testing/selftests/bpf/progs/dentry_lsm.c

diff --git a/tools/testing/selftests/bpf/prog_tests/dentry_lsm.c b/tools/testing/selftests/bpf/prog_tests/dentry_lsm.c
new file mode 100644
index 000000000000..3e8c68017954
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/dentry_lsm.c
@@ -0,0 +1,48 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2025 David Windsor <dwindsor@gmail.com> */
+
+#include <test_progs.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <limits.h>
+#include <string.h>
+#include "dentry_lsm.skel.h"
+
+void test_dentry_lsm(void)
+{
+	struct dentry_lsm *skel;
+	char test_file[PATH_MAX];
+	int fd, ret;
+
+	skel = dentry_lsm__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "dentry_lsm__open_and_load"))
+		return;
+
+	ret = dentry_lsm__attach(skel);
+	if (!ASSERT_OK(ret, "dentry_lsm__attach"))
+		goto cleanup;
+
+	/* Create a temporary file to trigger file_open LSM hook */
+	ret = snprintf(test_file, sizeof(test_file), "/tmp/bpf_test_file_%d", getpid());
+	if (!ASSERT_GT(ret, 0, "snprintf"))
+		goto cleanup_link;
+	if (!ASSERT_LT(ret, sizeof(test_file), "snprintf"))
+		goto cleanup_link;
+
+	fd = open(test_file, O_CREAT | O_RDWR, 0644);
+	if (!ASSERT_GE(fd, 0, "open"))
+		goto cleanup_link;
+	close(fd);
+
+	/* Test passes if BPF program loaded and ran without error */
+
+	/* Clean up test file */
+	unlink(test_file);
+
+cleanup_link:
+	unlink(test_file);
+cleanup:
+	dentry_lsm__destroy(skel);
+}
diff --git a/tools/testing/selftests/bpf/progs/dentry_lsm.c b/tools/testing/selftests/bpf/progs/dentry_lsm.c
new file mode 100644
index 000000000000..fa6d65d2c50f
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/dentry_lsm.c
@@ -0,0 +1,51 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2025 David Windsor <dwindsor@gmail.com> */
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_core_read.h>
+
+extern struct dentry *bpf_dget(struct dentry *dentry) __ksym;
+extern void bpf_dput(struct dentry *dentry) __ksym;
+extern struct dentry *bpf_dget_parent(struct dentry *dentry) __ksym;
+extern struct dentry *bpf_d_find_alias(struct inode *inode) __ksym;
+extern struct dentry *bpf_file_dentry(struct file *file) __ksym;
+extern struct vfsmount *bpf_file_vfsmount(struct file *file) __ksym;
+
+SEC("lsm.s/file_open")
+int BPF_PROG(file_open, struct file *file)
+{
+	struct dentry *dentry, *parent, *alias, *dentry_ref;
+	struct vfsmount *vfs_mnt;
+
+	if (!file)
+		return 0;
+
+	dentry = bpf_file_dentry(file);
+	if (dentry) {
+		dentry_ref = bpf_dget(dentry);
+		if (dentry_ref)
+			bpf_dput(dentry_ref);
+
+		parent = bpf_dget_parent(dentry);
+		if (parent)
+			bpf_dput(parent);
+	}
+
+	if (file->f_inode) {
+		alias = bpf_d_find_alias(file->f_inode);
+		if (alias)
+			bpf_dput(alias);
+	}
+
+	vfs_mnt = bpf_file_vfsmount(file);
+	if (vfs_mnt) {
+		/* Test that we can access vfsmount */
+		(void)vfs_mnt;
+	}
+
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";
-- 
2.43.0

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[Al Viro]

On Wed, Sep 24, 2025 at 07:24:33PM -0400, David Windsor wrote:
> Add six new BPF kfuncs that enable BPF LSM programs to safely interact
> with dentry objects:
> 
> - bpf_dget(): Acquire reference on dentry
> - bpf_dput(): Release reference on dentry
> - bpf_dget_parent(): Get referenced parent dentry
> - bpf_d_find_alias(): Find referenced alias dentry for inode
> - bpf_file_dentry(): Get dentry from file
> - bpf_file_vfsmount(): Get vfsmount from file
> 
> All kfuncs are currently restricted to BPF_PROG_TYPE_LSM programs.

You have an interesting definition of safety.

We are *NOT* letting random out-of-tree code play around with the
lifetime rules for core objects.

Not happening, whatever usecase you might have in mind.  This is
far too low-level to be exposed.

NAKed-by: Al Viro <viro@zeniv.linux.org.uk>

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[David Windsor]

On Wed, Sep 24, 2025 at 7:55 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
>
> On Wed, Sep 24, 2025 at 07:24:33PM -0400, David Windsor wrote:
> > Add six new BPF kfuncs that enable BPF LSM programs to safely interact
> > with dentry objects:
> >
> > - bpf_dget(): Acquire reference on dentry
> > - bpf_dput(): Release reference on dentry
> > - bpf_dget_parent(): Get referenced parent dentry
> > - bpf_d_find_alias(): Find referenced alias dentry for inode
> > - bpf_file_dentry(): Get dentry from file
> > - bpf_file_vfsmount(): Get vfsmount from file
> >
> > All kfuncs are currently restricted to BPF_PROG_TYPE_LSM programs.
>
> You have an interesting definition of safety.
>
> We are *NOT* letting random out-of-tree code play around with the
> lifetime rules for core objects.
>

File references are already exposed to bpf (bpf_get_task_exe_file,
bpf_put_file) with the same KF_ACQUIRE|KF_RELEASE semantics. These
follow the same pattern and are also LSM-only.

> Not happening, whatever usecase you might have in mind.  This is
> far too low-level to be exposed.
>
> NAKed-by: Al Viro <viro@zeniv.linux.org.uk>

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[Al Viro]

On Wed, Sep 24, 2025 at 08:08:03PM -0400, David Windsor wrote:
> On Wed, Sep 24, 2025 at 7:55 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
> >
> > On Wed, Sep 24, 2025 at 07:24:33PM -0400, David Windsor wrote:
> > > Add six new BPF kfuncs that enable BPF LSM programs to safely interact
> > > with dentry objects:
> > >
> > > - bpf_dget(): Acquire reference on dentry
> > > - bpf_dput(): Release reference on dentry
> > > - bpf_dget_parent(): Get referenced parent dentry
> > > - bpf_d_find_alias(): Find referenced alias dentry for inode
> > > - bpf_file_dentry(): Get dentry from file
> > > - bpf_file_vfsmount(): Get vfsmount from file
> > >
> > > All kfuncs are currently restricted to BPF_PROG_TYPE_LSM programs.
> >
> > You have an interesting definition of safety.
> >
> > We are *NOT* letting random out-of-tree code play around with the
> > lifetime rules for core objects.
> >
> 
> File references are already exposed to bpf (bpf_get_task_exe_file,
> bpf_put_file) with the same KF_ACQUIRE|KF_RELEASE semantics. These
> follow the same pattern and are also LSM-only.

You can safely clone and retain file references.  You can't do that
to dentries unless you are guaranteed an active reference to superblock
to stay around for as long as you are retaining those.  Note that
LSM hooks might be called with ->s_umount held by caller, so the locking
environment for superblocks depends upon the hook in question.

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[David Windsor]

On Wed, Sep 24, 2025 at 8:29 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
>
> On Wed, Sep 24, 2025 at 08:08:03PM -0400, David Windsor wrote:
> > On Wed, Sep 24, 2025 at 7:55 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
> > >
> > > On Wed, Sep 24, 2025 at 07:24:33PM -0400, David Windsor wrote:
> > > > Add six new BPF kfuncs that enable BPF LSM programs to safely interact
> > > > with dentry objects:
> > > >
> > > > - bpf_dget(): Acquire reference on dentry
> > > > - bpf_dput(): Release reference on dentry
> > > > - bpf_dget_parent(): Get referenced parent dentry
> > > > - bpf_d_find_alias(): Find referenced alias dentry for inode
> > > > - bpf_file_dentry(): Get dentry from file
> > > > - bpf_file_vfsmount(): Get vfsmount from file
> > > >
> > > > All kfuncs are currently restricted to BPF_PROG_TYPE_LSM programs.
> > >
> > > You have an interesting definition of safety.
> > >
> > > We are *NOT* letting random out-of-tree code play around with the
> > > lifetime rules for core objects.
> > >
> >
> > File references are already exposed to bpf (bpf_get_task_exe_file,
> > bpf_put_file) with the same KF_ACQUIRE|KF_RELEASE semantics. These
> > follow the same pattern and are also LSM-only.
>
> You can safely clone and retain file references.  You can't do that
> to dentries unless you are guaranteed an active reference to superblock
> to stay around for as long as you are retaining those.  Note that
> LSM hooks might be called with ->s_umount held by caller, so the locking
> environment for superblocks depends upon the hook in question.

Yeah good point about ->s_umount, why don't we just create a new "safe
dentry hooks" BTF ID set and restrict this to those and filter in
bpf_fs_kfuncs_filter, where there's existing filtering going on
anyway?

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[Al Viro]

On Wed, Sep 24, 2025 at 08:44:24PM -0400, David Windsor wrote:

> > You can safely clone and retain file references.  You can't do that
> > to dentries unless you are guaranteed an active reference to superblock
> > to stay around for as long as you are retaining those.  Note that
> > LSM hooks might be called with ->s_umount held by caller, so the locking
> > environment for superblocks depends upon the hook in question.
> 
> Yeah good point about ->s_umount, why don't we just create a new "safe
> dentry hooks" BTF ID set and restrict this to those and filter in
> bpf_fs_kfuncs_filter, where there's existing filtering going on
> anyway?

Again, you can't just call dget(), stash the reference into a map and move
on.  That's asking for UAF.

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[David Windsor]

On Wed, Sep 24, 2025 at 8:47 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
>
> On Wed, Sep 24, 2025 at 08:44:24PM -0400, David Windsor wrote:
>
> > > You can safely clone and retain file references.  You can't do that
> > > to dentries unless you are guaranteed an active reference to superblock
> > > to stay around for as long as you are retaining those.  Note that
> > > LSM hooks might be called with ->s_umount held by caller, so the locking
> > > environment for superblocks depends upon the hook in question.
> >
> > Yeah good point about ->s_umount, why don't we just create a new "safe
> > dentry hooks" BTF ID set and restrict this to those and filter in
> > bpf_fs_kfuncs_filter, where there's existing filtering going on
> > anyway?
>
> Again, you can't just call dget(), stash the reference into a map and move
> on.  That's asking for UAF.

These can't be stored in a map (guaranteed by verifier during addr leak checks)

Re: [PATCH bpf-next 1/2] bpf: Add dentry kfuncs for BPF LSM programs

[Christian Brauner]

On Thu, Sep 25, 2025 at 12:55:18AM +0100, Al Viro wrote:
> On Wed, Sep 24, 2025 at 07:24:33PM -0400, David Windsor wrote:
> > Add six new BPF kfuncs that enable BPF LSM programs to safely interact
> > with dentry objects:
> > 
> > - bpf_dget(): Acquire reference on dentry
> > - bpf_dput(): Release reference on dentry
> > - bpf_dget_parent(): Get referenced parent dentry
> > - bpf_d_find_alias(): Find referenced alias dentry for inode
> > - bpf_file_dentry(): Get dentry from file
> > - bpf_file_vfsmount(): Get vfsmount from file
> > 
> > All kfuncs are currently restricted to BPF_PROG_TYPE_LSM programs.
> 
> You have an interesting definition of safety.
> 
> We are *NOT* letting random out-of-tree code play around with the
> lifetime rules for core objects.
> 
> Not happening, whatever usecase you might have in mind.  This is
> far too low-level to be exposed.
> 
> NAKed-by: Al Viro <viro@zeniv.linux.org.uk>

I fully agree.