Add tracepoints for Landlock rule addition, landlock_add_rule_fs for
filesystem rules and landlock_add_rule_net for network rules, so trace
consumers can correlate filesystem objects and network ports with their
rulesets.  Both are emitted under the ruleset lock (asserted in
TP_fast_assign) so an eBPF program reads the ruleset, including the rule
just inserted, in a consistent snapshot.

Add a version field to struct landlock_ruleset, gated on
CONFIG_TRACEPOINTS like the id field and incremented under the ruleset
lock on each successful landlock_add_rule(2), including when it only
extends an existing rule's access rights.  It fills the existing 4-byte
hole after usage, so the struct does not grow.  Pairing the ruleset ID
with the version lets a later restrict_self event record the exact
ruleset revision merged into a domain.

Resolve the filesystem rule's absolute path with d_absolute_path()
rather than the d_path() audit uses: d_absolute_path() produces
namespace-independent paths that do not depend on the tracer's chroot
state, making trace output deterministic regardless of mount namespace
configuration.  Distinguish the error cases as "<too_long>"
(-ENAMETOOLONG) and "<unreachable>" (anonymous files or detached
mounts).

Also add __trace_print_untrusted_str(), a static inline helper in the
header guarded by CREATE_TRACE_POINTS: it escapes separators, quotes,
backslashes, and non-printable bytes via string_escape_mem() so an
untrusted string (the path here, process names in later denial events)
cannot inject field separators or control characters into the ftrace
text output.

Cc: Christian Brauner <[email protected]>
Cc: Günther Noack <[email protected]>
Cc: Justin Suess <[email protected]>
Cc: Masami Hiramatsu <[email protected]>
Cc: Mathieu Desnoyers <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: Tingmao Wang <[email protected]>
Signed-off-by: Mickaël Salaün <[email protected]>
---

Changes since v3:
https://patch.msgid.link/[email protected]
- Adapt to the landlock_rule_insert() -> landlock_store_rule()
  rename.
- Move DEFINE_FREE(__putname) from include/linux/fs.h to the Landlock
  filesystem header, so the series carries no VFS change and merges
  through the Landlock tree.
- Fold the untrusted-string escaping helper into this patch as a static
  inline __trace_print_untrusted_str() in the event header under
  CREATE_TRACE_POINTS, dropping the standalone patch (reviewed by
  Tingmao Wang):
  https://patch.msgid.link/[email protected]
- Give __trace_print_untrusted_str() an explicit length argument
  (__get_dynamic_array_len() - 1) instead of strlen(), so an
  abstract socket name with embedded NUL bytes is escaped in full
  rather than truncated at the first NUL.

Changes since v2:
https://patch.msgid.link/[email protected]
- Order the add_rule_net tracepoint arguments access_rights before
  port, matching add_rule_fs.
- Reformatted TP_STRUCT__entry and TP_fast_assign to kernel tracing
  convention (requested by Steven Rostedt).
- Render access_rights as symbolic names with __print_flags(), shared
  with the audit blocker names, instead of raw hex.
- Drop the tautological version static assertion (the counter tracks
  add-rule operations, not rule count) and clarify that @version is
  incremented on access-right extensions too.
- Gate the version field and its writer on CONFIG_TRACEPOINTS (only
  tracing uses it) instead of CONFIG_SECURITY_LANDLOCK_LOG, matching the
  id field.
- Adapt rule insertion to the base's quiet flag (landlock_insert_rule()
  flags argument).

Changes since v1:
https://patch.msgid.link/[email protected]
- Added landlock_add_rule_net tracepoint for network rules.
- Dropped key=inode:0x%lx from add_rule_fs printk, using dev/ino
  instead.
- Used ruleset Landlock ID instead of kernel pointer in printk.
- Differentiated d_absolute_path() error cases (suggested by
  Tingmao Wang).
- Moved DEFINE_FREE(__putname) to include/linux/fs.h (noticed by
  Tingmao Wang).
- Added version field to struct landlock_ruleset.
- Added version to add_rule trace events (format:
  ruleset=<id>.<version>).
- Added d_absolute_path() vs d_path() rationale to commit message.
---
 include/trace/events/landlock.h | 163 +++++++++++++++++++++++++++++++-
 security/landlock/fs.c          |  19 ++++
 security/landlock/fs.h          |  32 +++++++
 security/landlock/net.c         |  11 +++
 security/landlock/ruleset.c     |  13 ++-
 security/landlock/ruleset.h     |   7 ++
 6 files changed, 240 insertions(+), 5 deletions(-)

diff --git a/include/trace/events/landlock.h b/include/trace/events/landlock.h
index 02434f770455..b378446e26a5 100644
--- a/include/trace/events/landlock.h
+++ b/include/trace/events/landlock.h
@@ -11,9 +11,57 @@
 #define _TRACE_LANDLOCK_H
 
 #include <linux/landlock.h>
+#include <linux/string.h>
+#include <linux/string_helpers.h>
 #include <linux/tracepoint.h>
+#include <linux/trace_seq.h>
 
 struct landlock_ruleset;
+struct path;
+
+#ifdef CREATE_TRACE_POINTS
+
+/*
+ * Escapes @len bytes of an untrusted string into the trace sequence @p so it
+ * cannot inject field separators or control characters into the ftrace text
+ * output, and can be unambiguously recovered.  Called from the TP_printk() of
+ * the tracepoints that expose paths and process names.  @len is passed by the
+ * caller (rather than derived with strlen()) so a name that is not
+ * NUL-terminated or carries embedded NUL bytes (an abstract socket name) is
+ * escaped in full instead of being truncated at the first NUL.
+ *
+ * Return: a pointer into @p's buffer, or NULL if @src is NULL or the buffer is
+ * exhausted (normal when the trace buffer is full).
+ */
+static inline const char *
+__trace_print_untrusted_str(struct trace_seq *p, const char *src, size_t len)
+{
+       int escaped_size;
+       char *buf;
+       size_t buf_size = seq_buf_get_buf(&p->seq, &buf);
+       const char *ret = trace_seq_buffer_ptr(p);
+
+       /* Buffer exhaustion is normal when the trace buffer is full. */
+       if (!src || buf_size == 0)
+               return NULL;
+
+       escaped_size =
+               string_escape_mem(src, len, buf, buf_size,
+                                 ESCAPE_SPACE | ESCAPE_SPECIAL | ESCAPE_NAP |
+                                         ESCAPE_APPEND | ESCAPE_OCTAL,
+                                 " ='\"\\");
+       if (unlikely(escaped_size >= buf_size)) {
+               /* We need some room for the final '\0'. */
+               seq_buf_set_overflow(&p->seq);
+               p->full = 1;
+               return NULL;
+       }
+       seq_buf_commit(&p->seq, escaped_size);
+       trace_seq_putc(p, 0);
+       return ret;
+}
+
+#endif /* CREATE_TRACE_POINTS */
 
 /* clang-format off */
 
@@ -65,6 +113,14 @@ struct landlock_ruleset;
  * lockless snapshot instead: a task's comm, and the deny_access_net struct
  * sock (whose network hook holds no socket lock), matching how the sched
  * and signal trace events sample comm.
+ *
+ * Field encoding
+ * ~~~~~~~~~~~~~~
+ *
+ * Fields that mirror the Landlock UAPI use the same C types and endianness
+ * (e.g. network ports are __u64 in host endianness, like
+ * landlock_net_port_attr.port).  Per-event details, such as where a value
+ * is byte-swapped, live in the field's own kdoc.
  */
 
 /**
@@ -86,6 +142,7 @@ TRACE_EVENT(landlock_create_ruleset,
 
        TP_STRUCT__entry(
                __field(        __u64,          ruleset_id      )
+               __field(        __u32,          ruleset_version )
                __field(        access_mask_t,  handled_fs      )
                __field(        access_mask_t,  handled_net     )
                __field(        access_mask_t,  scoped          )
@@ -93,13 +150,14 @@ TRACE_EVENT(landlock_create_ruleset,
 
        TP_fast_assign(
                __entry->ruleset_id     = ruleset->id;
+               __entry->ruleset_version = ruleset->version;
                __entry->handled_fs     = ruleset->handled_masks.fs;
                __entry->handled_net    = ruleset->handled_masks.net;
                __entry->scoped         = ruleset->handled_masks.scope;
        ),
 
-       TP_printk("ruleset=%llx handled_fs=%s handled_net=%s scoped=%s",
-               __entry->ruleset_id,
+       TP_printk("ruleset=%llx.%u handled_fs=%s handled_net=%s scoped=%s",
+               __entry->ruleset_id, __entry->ruleset_version,
                __print_flags(__entry->handled_fs, "|", 
_LANDLOCK_ACCESS_FS_NAMES),
                __print_flags(__entry->handled_net, "|", 
_LANDLOCK_ACCESS_NET_NAMES),
                __print_flags(__entry->scoped, "|", _LANDLOCK_SCOPE_NAMES))
@@ -124,13 +182,112 @@ TRACE_EVENT(landlock_free_ruleset,
 
        TP_STRUCT__entry(
                __field(        __u64,          ruleset_id      )
+               __field(        __u32,          ruleset_version )
+       ),
+
+       TP_fast_assign(
+               __entry->ruleset_id     = ruleset->id;
+               __entry->ruleset_version = ruleset->version;
+       ),
+
+       TP_printk("ruleset=%llx.%u",
+               __entry->ruleset_id, __entry->ruleset_version)
+);
+
+/**
+ * landlock_add_rule_fs - Filesystem rule added to a ruleset
+ *
+ * @ruleset: Source ruleset (never NULL).
+ * @access_rights: Effective access mask stored in the rule, not the raw
+ *                 sys_landlock_add_rule() argument (unhandled rights
+ *                 added).
+ * @path: Filesystem path for the rule (never NULL).
+ * @pathname: Resolved absolute path string (never NULL; error placeholder
+ *            on resolution failure).
+ *
+ * Emitted by sys_landlock_add_rule() under the modified ruleset's lock, so
+ * the reported ruleset is a stable snapshot that no concurrent writer can
+ * change.
+ */
+TRACE_EVENT(landlock_add_rule_fs,
+
+       TP_PROTO(const struct landlock_ruleset *ruleset,
+                access_mask_t access_rights, const struct path *path,
+                const char *pathname),
+
+       TP_ARGS(ruleset, access_rights, path, pathname),
+
+       TP_STRUCT__entry(
+               __field(        __u64,          ruleset_id      )
+               __field(        __u32,          ruleset_version )
+               __field(        access_mask_t,  access_rights   )
+               __field(        dev_t,          dev             )
+               __field(        ino_t,          ino             )
+               __string(       pathname,       pathname        )
+       ),
+
+       TP_fast_assign(
+               lockdep_assert_held(&ruleset->lock);
+               __entry->ruleset_id     = ruleset->id;
+               __entry->ruleset_version = ruleset->version;
+               __entry->access_rights  = access_rights;
+               __entry->dev            = path->dentry->d_sb->s_dev;
+               /*
+                * The inode number may not be the user-visible one,
+                * but it will be the same used by audit.
+                */
+               __entry->ino            = d_backing_inode(path->dentry)->i_ino;
+               __assign_str(pathname);
+       ),
+
+       TP_printk("ruleset=%llx.%u access_rights=%s dev=%u:%u ino=%lu path=%s",
+               __entry->ruleset_id, __entry->ruleset_version,
+               __print_flags(__entry->access_rights, "|", 
_LANDLOCK_ACCESS_FS_NAMES),
+               MAJOR(__entry->dev), MINOR(__entry->dev), __entry->ino,
+               __trace_print_untrusted_str(p, __get_str(pathname),
+                                           __get_dynamic_array_len(pathname) - 
1))
+);
+
+/**
+ * landlock_add_rule_net - Network port rule added to a ruleset
+ *
+ * @ruleset: Source ruleset (never NULL).
+ * @access_rights: Effective access mask stored in the rule, not the raw
+ *                 sys_landlock_add_rule() argument (unhandled rights
+ *                 added).
+ * @port: Network port, the landlock_net_port_attr.port UAPI value
+ *        forwarded directly.
+ *
+ * Emitted by sys_landlock_add_rule() under the modified ruleset's lock, so
+ * the reported ruleset is a stable snapshot that no concurrent writer can
+ * change.
+ */
+TRACE_EVENT(landlock_add_rule_net,
+
+       TP_PROTO(const struct landlock_ruleset *ruleset,
+                access_mask_t access_rights, __u64 port),
+
+       TP_ARGS(ruleset, access_rights, port),
+
+       TP_STRUCT__entry(
+               __field(        __u64,          ruleset_id      )
+               __field(        __u32,          ruleset_version )
+               __field(        access_mask_t,  access_rights   )
+               __field(        __u64,          port            )
        ),
 
        TP_fast_assign(
+               lockdep_assert_held(&ruleset->lock);
                __entry->ruleset_id     = ruleset->id;
+               __entry->ruleset_version = ruleset->version;
+               __entry->access_rights  = access_rights;
+               __entry->port           = port;
        ),
 
-       TP_printk("ruleset=%llx", __entry->ruleset_id)
+       TP_printk("ruleset=%llx.%u access_rights=%s port=%llu",
+               __entry->ruleset_id, __entry->ruleset_version,
+               __print_flags(__entry->access_rights, "|", 
_LANDLOCK_ACCESS_NET_NAMES),
+               __entry->port)
 );
 
 #undef _LANDLOCK_NAME_ENTRY
diff --git a/security/landlock/fs.c b/security/landlock/fs.c
index 3ff5fd24378f..3a33e7393e25 100644
--- a/security/landlock/fs.c
+++ b/security/landlock/fs.c
@@ -53,6 +53,8 @@
 #include "ruleset.h"
 #include "setup.h"
 
+#include <trace/events/landlock.h>
+
 /* Underlying object management */
 
 static void release_inode(struct landlock_object *const object)
@@ -347,7 +349,24 @@ int landlock_append_fs_rule(struct landlock_ruleset *const 
ruleset,
                return PTR_ERR(id.key.object);
        mutex_lock(&ruleset->lock);
        err = landlock_insert_rule(ruleset, id, access_rights, flags);
+
+       /*
+        * Emit after the rule insertion succeeds, so every event corresponds to
+        * a rule that is actually in the ruleset.  The ruleset lock is still
+        * held for BTF consistency (enforced by lockdep_assert_held in
+        * TP_fast_assign).
+        */
+       if (!err && trace_landlock_add_rule_fs_enabled()) {
+               char *buffer __free(__putname) = __getname();
+               const char *pathname =
+                       buffer ? resolve_path_for_trace(path, buffer) :
+                                "<no_mem>";
+
+               trace_landlock_add_rule_fs(ruleset, access_rights, path,
+                                          pathname);
+       }
        mutex_unlock(&ruleset->lock);
+
        /*
         * No need to check for an error because landlock_insert_rule()
         * increments the refcount for the new object if needed.
diff --git a/security/landlock/fs.h b/security/landlock/fs.h
index c16f24e30bd5..50b3ddb8ac1c 100644
--- a/security/landlock/fs.h
+++ b/security/landlock/fs.h
@@ -11,6 +11,7 @@
 #define _SECURITY_LANDLOCK_FS_H
 
 #include <linux/build_bug.h>
+#include <linux/cleanup.h>
 #include <linux/fs.h>
 #include <linux/init.h>
 #include <linux/rcupdate.h>
@@ -20,6 +21,8 @@
 #include "ruleset.h"
 #include "setup.h"
 
+DEFINE_FREE(__putname, char *, if (_T) __putname(_T))
+
 /**
  * struct landlock_inode_security - Inode security blob
  *
@@ -153,4 +156,33 @@ int landlock_append_fs_rule(struct landlock_ruleset *const 
ruleset,
                            const struct path *const path,
                            access_mask_t access_hierarchy, const u32 flags);
 
+/**
+ * resolve_path_for_trace - Resolve a path for tracepoint display
+ *
+ * @path: The path to resolve.
+ * @buf: A buffer of at least PATH_MAX bytes for the resolved path.
+ *
+ * Uses d_absolute_path() to produce a namespace-independent absolute path,
+ * unlike d_path() which resolves relative to the process's chroot.  This
+ * ensures trace output is deterministic regardless of the tracer's mount
+ * namespace.
+ *
+ * Return: A pointer into @buf with the resolved path, or an error string
+ * ("<too_long>", "<unreachable>").
+ */
+static inline const char *resolve_path_for_trace(const struct path *path,
+                                                char *buf)
+{
+       const char *p;
+
+       p = d_absolute_path(path, buf, PATH_MAX);
+       if (!IS_ERR_OR_NULL(p))
+               return p;
+
+       if (PTR_ERR(p) == -ENAMETOOLONG)
+               return "<too_long>";
+
+       return "<unreachable>";
+}
+
 #endif /* _SECURITY_LANDLOCK_FS_H */
diff --git a/security/landlock/net.c b/security/landlock/net.c
index e27b3ba15664..ead97fcfdcff 100644
--- a/security/landlock/net.c
+++ b/security/landlock/net.c
@@ -20,6 +20,8 @@
 #include "net.h"
 #include "ruleset.h"
 
+#include <trace/events/landlock.h>
+
 int landlock_append_net_rule(struct landlock_ruleset *const ruleset,
                             const u16 port, access_mask_t access_rights,
                             const u32 flags)
@@ -37,6 +39,15 @@ int landlock_append_net_rule(struct landlock_ruleset *const 
ruleset,
 
        mutex_lock(&ruleset->lock);
        err = landlock_insert_rule(ruleset, id, access_rights, flags);
+
+       /*
+        * Emit after the rule insertion succeeds, so every event corresponds to
+        * a rule that is actually in the ruleset.  The ruleset lock is still
+        * held for BTF consistency (enforced by lockdep_assert_held in
+        * TP_fast_assign).
+        */
+       if (!err)
+               trace_landlock_add_rule_net(ruleset, access_rights, port);
        mutex_unlock(&ruleset->lock);
 
        return err;
diff --git a/security/landlock/ruleset.c b/security/landlock/ruleset.c
index 30f3897a56d4..0d07707523cd 100644
--- a/security/landlock/ruleset.c
+++ b/security/landlock/ruleset.c
@@ -4,6 +4,7 @@
  *
  * Copyright © 2016-2020 Mickaël Salaün <[email protected]>
  * Copyright © 2018-2020 ANSSI
+ * Copyright © 2026 Cloudflare, Inc.
  */
 
 #include <linux/bits.h>
@@ -306,11 +307,19 @@ int landlock_insert_rule(struct landlock_ruleset *const 
ruleset,
                        .quiet = !!(flags & LANDLOCK_ADD_RULE_QUIET),
                },
        } };
+       int err;
 
        build_check_layer();
        lockdep_assert_held(&ruleset->lock);
-       return landlock_store_rule(&ruleset->rules, id, &layers,
-                                  ARRAY_SIZE(layers));
+       err = landlock_store_rule(&ruleset->rules, id, &layers,
+                                 ARRAY_SIZE(layers));
+
+#ifdef CONFIG_TRACEPOINTS
+       if (!err)
+               ruleset->version++;
+#endif /* CONFIG_TRACEPOINTS */
+
+       return err;
 }
 
 void landlock_free_rules(struct landlock_rules *const rules)
diff --git a/security/landlock/ruleset.h b/security/landlock/ruleset.h
index dcc363740b85..b536fa0425b7 100644
--- a/security/landlock/ruleset.h
+++ b/security/landlock/ruleset.h
@@ -167,6 +167,13 @@ struct landlock_ruleset {
        refcount_t usage;
 
 #ifdef CONFIG_TRACEPOINTS
+       /**
+        * @version: Counter incremented on each successful
+        * landlock_add_rule(2), including when it only extends an existing
+        * rule's access rights.  Used by tracepoints to correlate a domain with
+        * the exact ruleset state it was created from.  Protected by @lock.
+        */
+       u32 version;
        /**
         * @id: Unique identifier for this ruleset, used for tracing.
         */
-- 
2.54.0


Reply via email to