xref: /linux/Documentation/bpf/kfuncs.rst (revision 857071efc362d4641673c1fbf7e8a2d7a2408e75)
1.. SPDX-License-Identifier: GPL-2.0
2
3.. _kfuncs-header-label:
4
5=============================
6BPF Kernel Functions (kfuncs)
7=============================
8
91. Introduction
10===============
11
12BPF Kernel Functions or more commonly known as kfuncs are functions in the Linux
13kernel which are exposed for use by BPF programs. Unlike normal BPF helpers,
14kfuncs do not have a stable interface and can change from one kernel release to
15another. Hence, BPF programs need to be updated in response to changes in the
16kernel. See :ref:`BPF_kfunc_lifecycle_expectations` for more information.
17
182. Defining a kfunc
19===================
20
21There are two ways to expose a kernel function to BPF programs, either make an
22existing function in the kernel visible, or add a new wrapper for BPF. In both
23cases, care must be taken that BPF program can only call such function in a
24valid context. To enforce this, visibility of a kfunc can be per program type.
25
26If you are not creating a BPF wrapper for existing kernel function, skip ahead
27to :ref:`BPF_kfunc_nodef`.
28
292.1 Creating a wrapper kfunc
30----------------------------
31
32When defining a wrapper kfunc, the wrapper function should have extern linkage.
33This prevents the compiler from optimizing away dead code, as this wrapper kfunc
34is not invoked anywhere in the kernel itself. It is not necessary to provide a
35prototype in a header for the wrapper kfunc.
36
37An example is given below::
38
39        /* Disables missing prototype warnings */
40        __bpf_kfunc_start_defs();
41
42        __bpf_kfunc struct task_struct *bpf_find_get_task_by_vpid(pid_t nr)
43        {
44                return find_get_task_by_vpid(nr);
45        }
46
47        __bpf_kfunc_end_defs();
48
49A wrapper kfunc is often needed when we need to annotate parameters of the
50kfunc. Otherwise one may directly make the kfunc visible to the BPF program by
51registering it with the BPF subsystem. See :ref:`BPF_kfunc_nodef`.
52
532.2 kfunc Parameters
54--------------------
55
56All kfuncs now require trusted arguments by default. This means that all
57pointer arguments must be valid, and all pointers to BTF objects must be
58passed in their unmodified form (at a zero offset, and without having been
59obtained from walking another pointer, with exceptions described below).
60
61There are two types of pointers to kernel objects which are considered "trusted":
62
631. Pointers which are passed as tracepoint or struct_ops callback arguments.
642. Pointers which were returned from a KF_ACQUIRE kfunc.
65
66Pointers to non-BTF objects (e.g. scalar pointers) may also be passed to
67kfuncs, and may have a non-zero offset.
68
69The definition of "valid" pointers is subject to change at any time, and has
70absolutely no ABI stability guarantees.
71
72As mentioned above, a nested pointer obtained from walking a trusted pointer is
73no longer trusted, with one exception. If a struct type has a field that is
74guaranteed to be valid (trusted or rcu, as in KF_RCU description below) as long
75as its parent pointer is valid, the following macros can be used to express
76that to the verifier:
77
78* ``BTF_TYPE_SAFE_TRUSTED``
79* ``BTF_TYPE_SAFE_RCU``
80* ``BTF_TYPE_SAFE_RCU_OR_NULL``
81
82For example,
83
84.. code-block:: c
85
86	BTF_TYPE_SAFE_TRUSTED(struct socket) {
87		struct sock *sk;
88	};
89
90or
91
92.. code-block:: c
93
94	BTF_TYPE_SAFE_RCU(struct task_struct) {
95		const cpumask_t *cpus_ptr;
96		struct css_set __rcu *cgroups;
97		struct task_struct __rcu *real_parent;
98		struct task_struct *group_leader;
99	};
100
101In other words, you must:
102
1031. Wrap the valid pointer type in a ``BTF_TYPE_SAFE_*`` macro.
104
1052. Specify the type and name of the valid nested field. This field must match
106   the field in the original type definition exactly.
107
108A new type declared by a ``BTF_TYPE_SAFE_*`` macro also needs to be emitted so
109that it appears in BTF. For example, ``BTF_TYPE_SAFE_TRUSTED(struct socket)``
110is emitted in the ``type_is_trusted()`` function as follows:
111
112.. code-block:: c
113
114	BTF_TYPE_EMIT(BTF_TYPE_SAFE_TRUSTED(struct socket));
115
1162.3 Annotating kfunc parameters
117-------------------------------
118
119Similar to BPF helpers, there is sometime need for additional context required
120by the verifier to make the usage of kernel functions safer and more useful.
121Hence, we can annotate a parameter by suffixing the name of the argument of the
122kfunc with a __tag, where tag may be one of the supported annotations.
123
1242.3.1 __sz Annotation
125---------------------
126
127This annotation is used to indicate a memory and size pair in the argument list.
128An example is given below::
129
130        __bpf_kfunc void bpf_memzero(void *mem, int mem__sz)
131        {
132        ...
133        }
134
135Here, the verifier will treat first argument as a PTR_TO_MEM, and second
136argument as its size. By default, without __sz annotation, the size of the type
137of the pointer is used. Without __sz annotation, a kfunc cannot accept a void
138pointer.
139
1402.3.2 __k Annotation
141--------------------
142
143This annotation is only understood for scalar arguments, where it indicates that
144the verifier must check the scalar argument to be a known constant, which does
145not indicate a size parameter, and the value of the constant is relevant to the
146safety of the program.
147
148An example is given below::
149
150        __bpf_kfunc void *bpf_obj_new(u32 local_type_id__k, ...)
151        {
152        ...
153        }
154
155Here, bpf_obj_new uses local_type_id argument to find out the size of that type
156ID in program's BTF and return a sized pointer to it. Each type ID will have a
157distinct size, hence it is crucial to treat each such call as distinct when
158values don't match during verifier state pruning checks.
159
160Hence, whenever a constant scalar argument is accepted by a kfunc which is not a
161size parameter, and the value of the constant matters for program safety, __k
162suffix should be used.
163
1642.3.3 __uninit Annotation
165-------------------------
166
167This annotation is used to indicate that the argument will be treated as
168uninitialized.
169
170An example is given below::
171
172        __bpf_kfunc int bpf_dynptr_from_skb(..., struct bpf_dynptr_kern *ptr__uninit)
173        {
174        ...
175        }
176
177Here, the dynptr will be treated as an uninitialized dynptr. Without this
178annotation, the verifier will reject the program if the dynptr passed in is
179not initialized.
180
1812.3.4 __nullable Annotation
182---------------------------
183
184This annotation is used to indicate that the pointer argument may be NULL.
185The verifier will allow passing NULL for such arguments.
186
187An example is given below::
188
189        __bpf_kfunc void bpf_task_release(struct task_struct *task__nullable)
190        {
191        ...
192        }
193
194Here, the task pointer may be NULL. The kfunc is responsible for checking if
195the pointer is NULL before dereferencing it.
196
197The __nullable annotation can be combined with other annotations. For example,
198when used with __sz or __szk annotations for memory and size pairs, the
199verifier will skip size validation when a NULL pointer is passed, but will
200still process the size argument to extract constant size information when
201needed::
202
203        __bpf_kfunc void *bpf_dynptr_slice(..., void *buffer__nullable,
204                                           u32 buffer__szk)
205
206Here, the buffer may be NULL. If the buffer is not NULL, it must be at least
207buffer__szk bytes in size. The kfunc is responsible for checking if the buffer
208is NULL before using it.
209
2102.3.5 __nonown_allowed Annotation
211---------------------------------
212
213This annotation is used to indicate that the parameter may be a non-owning reference.
214
215An example is given below::
216
217        __bpf_kfunc int bpf_list_add(..., struct bpf_list_node
218                                     *prev__nonown_allowed, ...)
219        {
220                ...
221        }
222
223For the ``prev__nonown_allowed`` parameter (resolved as ``KF_ARG_PTR_TO_LIST_NODE``),
224suffix ``__nonown_allowed`` retains the usual owning-pointer rules and also
225permits a non-owning reference with no ref_obj_id (e.g. the return value of
226bpf_list_front() / bpf_list_back()).
227
2282.3.6 __str Annotation
229----------------------
230This annotation is used to indicate that the argument is a constant string.
231
232An example is given below::
233
234        __bpf_kfunc bpf_get_file_xattr(..., const char *name__str, ...)
235        {
236        ...
237        }
238
239In this case, ``bpf_get_file_xattr()`` can be called as::
240
241        bpf_get_file_xattr(..., "xattr_name", ...);
242
243Or::
244
245        const char name[] = "xattr_name";  /* This need to be global */
246        int BPF_PROG(...)
247        {
248                ...
249                bpf_get_file_xattr(..., name, ...);
250                ...
251        }
252
253.. _BPF_kfunc_nodef:
254
2552.4 Using an existing kernel function
256-------------------------------------
257
258When an existing function in the kernel is fit for consumption by BPF programs,
259it can be directly registered with the BPF subsystem. However, care must still
260be taken to review the context in which it will be invoked by the BPF program
261and whether it is safe to do so.
262
2632.5 Annotating kfuncs
264---------------------
265
266In addition to kfuncs' arguments, verifier may need more information about the
267type of kfunc(s) being registered with the BPF subsystem. To do so, we define
268flags on a set of kfuncs as follows::
269
270        BTF_KFUNCS_START(bpf_task_set)
271        BTF_ID_FLAGS(func, bpf_get_task_pid, KF_ACQUIRE | KF_RET_NULL)
272        BTF_ID_FLAGS(func, bpf_put_pid, KF_RELEASE)
273        BTF_KFUNCS_END(bpf_task_set)
274
275This set encodes the BTF ID of each kfunc listed above, and encodes the flags
276along with it. It is also allowed to specify no flags.
277
278kfunc definitions should also always be annotated with the ``__bpf_kfunc``
279macro. This prevents issues such as the compiler inlining the kfunc, or the
280function being elided in an LTO build as it's not used in the rest of the
281kernel. Developers should not manually add annotations to their kfunc to prevent
282these issues. If an annotation is required to prevent such an issue with your
283kfunc, it is a bug and should be added to the definition of the macro so that
284other kfuncs are similarly protected. An example is given below::
285
286        __bpf_kfunc struct task_struct *bpf_get_task_pid(s32 pid)
287        {
288        ...
289        }
290
291Note that kfuncs must not be declared ``static``. A kfunc can be called from a
292BPF program ``*.c`` file outside the compilation unit that defines it, so its
293externally visible name must remain available for BTF ID lookup. ``static``
294linkage allows the compiler to rename the function, which can break this
295BTF-based kfunc resolution. Further note that sparse may warn that an otherwise
296unreferenced kfunc should be static. Such warnings should be ignored for kfunc
297definitions.
298
2992.5.1 KF_ACQUIRE flag
300---------------------
301
302The KF_ACQUIRE flag is used to indicate that the kfunc returns a pointer to a
303refcounted object. The verifier will then ensure that the pointer to the object
304is eventually released using a release kfunc, or transferred to a map using a
305referenced kptr (by invoking bpf_kptr_xchg). If not, the verifier fails the
306loading of the BPF program until no lingering references remain in all possible
307explored states of the program.
308
3092.5.2 KF_RET_NULL flag
310----------------------
311
312The KF_RET_NULL flag is used to indicate that the pointer returned by the kfunc
313may be NULL. Hence, it forces the user to do a NULL check on the pointer
314returned from the kfunc before making use of it (dereferencing or passing to
315another helper). This flag is often used in pairing with KF_ACQUIRE flag, but
316both are orthogonal to each other.
317
3182.5.3 KF_RELEASE flag
319---------------------
320
321The KF_RELEASE flag is used to indicate that the kfunc releases the pointer
322passed in to it. There can be only one referenced pointer that can be passed
323in. All copies of the pointer being released are invalidated as a result of
324invoking kfunc with this flag.
325
3262.5.4 KF_SLEEPABLE flag
327-----------------------
328
329The KF_SLEEPABLE flag is used for kfuncs that may sleep. Such kfuncs can only
330be called by sleepable BPF programs (BPF_F_SLEEPABLE).
331
3322.5.5 KF_DESTRUCTIVE flag
333--------------------------
334
335The KF_DESTRUCTIVE flag is used to indicate functions calling which is
336destructive to the system. For example such a call can result in system
337rebooting or panicking. Due to this additional restrictions apply to these
338calls. At the moment they only require CAP_SYS_BOOT capability, but more can be
339added later.
340
3412.5.6 KF_RCU flag
342-----------------
343
344The KF_RCU flag allows kfuncs to opt out of the default trusted args
345requirement and accept RCU pointers with weaker guarantees. The kfuncs marked
346with KF_RCU expect either PTR_TRUSTED or MEM_RCU arguments. The verifier
347guarantees that the objects are valid and there is no use-after-free. The
348pointers are not NULL, but the object's refcount could have reached zero. The
349kfuncs need to consider doing refcnt != 0 check, especially when returning a
350KF_ACQUIRE pointer. Note as well that a KF_ACQUIRE kfunc that is KF_RCU should
351very likely also be KF_RET_NULL.
352
3532.5.7 KF_RCU_PROTECTED flag
354---------------------------
355
356The KF_RCU_PROTECTED flag is used to indicate that the kfunc must be invoked in
357an RCU critical section. This is assumed by default in non-sleepable programs,
358and must be explicitly ensured by calling ``bpf_rcu_read_lock`` for sleepable
359ones.
360
361If the kfunc returns a pointer value, this flag also enforces that the returned
362pointer is RCU protected, and can only be used while the RCU critical section is
363active.
364
365The flag is distinct from the ``KF_RCU`` flag, which only ensures that its
366arguments are at least RCU protected pointers. This may transitively imply that
367RCU protection is ensured, but it does not work in cases of kfuncs which require
368RCU protection but do not take RCU protected arguments.
369
370.. _KF_deprecated_flag:
371
3722.5.8 KF_DEPRECATED flag
373------------------------
374
375The KF_DEPRECATED flag is used for kfuncs which are scheduled to be
376changed or removed in a subsequent kernel release. A kfunc that is
377marked with KF_DEPRECATED should also have any relevant information
378captured in its kernel doc. Such information typically includes the
379kfunc's expected remaining lifespan, a recommendation for new
380functionality that can replace it if any is available, and possibly a
381rationale for why it is being removed.
382
383Note that while on some occasions, a KF_DEPRECATED kfunc may continue to be
384supported and have its KF_DEPRECATED flag removed, it is likely to be far more
385difficult to remove a KF_DEPRECATED flag after it's been added than it is to
386prevent it from being added in the first place. As described in
387:ref:`BPF_kfunc_lifecycle_expectations`, users that rely on specific kfuncs are
388encouraged to make their use-cases known as early as possible, and participate
389in upstream discussions regarding whether to keep, change, deprecate, or remove
390those kfuncs if and when such discussions occur.
391
3922.5.9 KF_IMPLICIT_ARGS flag
393------------------------------------
394
395The KF_IMPLICIT_ARGS flag is used to indicate that the BPF signature
396of the kfunc is different from it's kernel signature, and the values
397for implicit arguments are provided at load time by the verifier.
398
399Only arguments of specific types are implicit.
400Currently only ``struct bpf_prog_aux *`` type is supported.
401
402A kfunc with KF_IMPLICIT_ARGS flag therefore has two types in BTF: one
403function matching the kernel declaration (with _impl suffix in the
404name by convention), and another matching the intended BPF API.
405
406Verifier only allows calls to the non-_impl version of a kfunc, that
407uses a signature without the implicit arguments.
408
409Example declaration:
410
411.. code-block:: c
412
413	__bpf_kfunc int bpf_task_work_schedule_signal(struct task_struct *task, struct bpf_task_work *tw,
414						      void *map__map, bpf_task_work_callback_t callback,
415						      struct bpf_prog_aux *aux) { ... }
416
417Example usage in BPF program:
418
419.. code-block:: c
420
421	/* note that the last argument is omitted */
422        bpf_task_work_schedule_signal(task, &work->tw, &arrmap, task_work_callback);
423
4242.6 Registering the kfuncs
425--------------------------
426
427Once the kfunc is prepared for use, the final step to making it visible is
428registering it with the BPF subsystem. Registration is done per BPF program
429type. An example is shown below::
430
431        BTF_KFUNCS_START(bpf_task_set)
432        BTF_ID_FLAGS(func, bpf_get_task_pid, KF_ACQUIRE | KF_RET_NULL)
433        BTF_ID_FLAGS(func, bpf_put_pid, KF_RELEASE)
434        BTF_KFUNCS_END(bpf_task_set)
435
436        static const struct btf_kfunc_id_set bpf_task_kfunc_set = {
437                .owner = THIS_MODULE,
438                .set   = &bpf_task_set,
439        };
440
441        static int init_subsystem(void)
442        {
443                return register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &bpf_task_kfunc_set);
444        }
445        late_initcall(init_subsystem);
446
4472.7  Specifying no-cast aliases with ___init
448--------------------------------------------
449
450The verifier will always enforce that the BTF type of a pointer passed to a
451kfunc by a BPF program, matches the type of pointer specified in the kfunc
452definition. The verifier, does, however, allow types that are equivalent
453according to the C standard to be passed to the same kfunc arg, even if their
454BTF_IDs differ.
455
456For example, for the following type definition:
457
458.. code-block:: c
459
460	struct bpf_cpumask {
461		cpumask_t cpumask;
462		refcount_t usage;
463	};
464
465The verifier would allow a ``struct bpf_cpumask *`` to be passed to a kfunc
466taking a ``cpumask_t *`` (which is a typedef of ``struct cpumask *``). For
467instance, both ``struct cpumask *`` and ``struct bpf_cpmuask *`` can be passed
468to bpf_cpumask_test_cpu().
469
470In some cases, this type-aliasing behavior is not desired. ``struct
471nf_conn___init`` is one such example:
472
473.. code-block:: c
474
475	struct nf_conn___init {
476		struct nf_conn ct;
477	};
478
479The C standard would consider these types to be equivalent, but it would not
480always be safe to pass either type to a trusted kfunc. ``struct
481nf_conn___init`` represents an allocated ``struct nf_conn`` object that has
482*not yet been initialized*, so it would therefore be unsafe to pass a ``struct
483nf_conn___init *`` to a kfunc that's expecting a fully initialized ``struct
484nf_conn *`` (e.g. ``bpf_ct_change_timeout()``).
485
486In order to accommodate such requirements, the verifier will enforce strict
487PTR_TO_BTF_ID type matching if two types have the exact same name, with one
488being suffixed with ``___init``.
489
4902.8 Accessing arena memory through kfunc arguments
491--------------------------------------------------
492
493A read or write at any address inside an arena does not oops the kernel.
494Unallocated arena pages are lazily backed by a scratch page and the
495access is reported through the program's BPF stream as an error. Only
496the BPF program's correctness is affected; the kernel itself remains
497intact.
498
499The arena is followed by a ``GUARD_SZ / 2`` (32 KiB) guard region that
500is also covered by this recovery. A kfunc handed an arena pointer may
501therefore access up to ``GUARD_SZ / 2`` past it without bounds-checking
502against the arena. Larger accesses must verify the range explicitly.
503
504.. _BPF_kfunc_lifecycle_expectations:
505
5063. kfunc lifecycle expectations
507===============================
508
509kfuncs provide a kernel <-> kernel API, and thus are not bound by any of the
510strict stability restrictions associated with kernel <-> user UAPIs. This means
511they can be thought of as similar to EXPORT_SYMBOL_GPL, and can therefore be
512modified or removed by a maintainer of the subsystem they're defined in when
513it's deemed necessary.
514
515Like any other change to the kernel, maintainers will not change or remove a
516kfunc without having a reasonable justification.  Whether or not they'll choose
517to change a kfunc will ultimately depend on a variety of factors, such as how
518widely used the kfunc is, how long the kfunc has been in the kernel, whether an
519alternative kfunc exists, what the norm is in terms of stability for the
520subsystem in question, and of course what the technical cost is of continuing
521to support the kfunc.
522
523There are several implications of this:
524
525a) kfuncs that are widely used or have been in the kernel for a long time will
526   be more difficult to justify being changed or removed by a maintainer. In
527   other words, kfuncs that are known to have a lot of users and provide
528   significant value provide stronger incentives for maintainers to invest the
529   time and complexity in supporting them. It is therefore important for
530   developers that are using kfuncs in their BPF programs to communicate and
531   explain how and why those kfuncs are being used, and to participate in
532   discussions regarding those kfuncs when they occur upstream.
533
534b) Unlike regular kernel symbols marked with EXPORT_SYMBOL_GPL, BPF programs
535   that call kfuncs are generally not part of the kernel tree. This means that
536   refactoring cannot typically change callers in-place when a kfunc changes,
537   as is done for e.g. an upstreamed driver being updated in place when a
538   kernel symbol is changed.
539
540   Unlike with regular kernel symbols, this is expected behavior for BPF
541   symbols, and out-of-tree BPF programs that use kfuncs should be considered
542   relevant to discussions and decisions around modifying and removing those
543   kfuncs. The BPF community will take an active role in participating in
544   upstream discussions when necessary to ensure that the perspectives of such
545   users are taken into account.
546
547c) A kfunc will never have any hard stability guarantees. BPF APIs cannot and
548   will not ever hard-block a change in the kernel purely for stability
549   reasons. That being said, kfuncs are features that are meant to solve
550   problems and provide value to users. The decision of whether to change or
551   remove a kfunc is a multivariate technical decision that is made on a
552   case-by-case basis, and which is informed by data points such as those
553   mentioned above. It is expected that a kfunc being removed or changed with
554   no warning will not be a common occurrence or take place without sound
555   justification, but it is a possibility that must be accepted if one is to
556   use kfuncs.
557
5583.1 kfunc deprecation
559---------------------
560
561As described above, while sometimes a maintainer may find that a kfunc must be
562changed or removed immediately to accommodate some changes in their subsystem,
563usually kfuncs will be able to accommodate a longer and more measured
564deprecation process. For example, if a new kfunc comes along which provides
565superior functionality to an existing kfunc, the existing kfunc may be
566deprecated for some period of time to allow users to migrate their BPF programs
567to use the new one. Or, if a kfunc has no known users, a decision may be made
568to remove the kfunc (without providing an alternative API) after some
569deprecation period so as to provide users with a window to notify the kfunc
570maintainer if it turns out that the kfunc is actually being used.
571
572It's expected that the common case will be that kfuncs will go through a
573deprecation period rather than being changed or removed without warning. As
574described in :ref:`KF_deprecated_flag`, the kfunc framework provides the
575KF_DEPRECATED flag to kfunc developers to signal to users that a kfunc has been
576deprecated. Once a kfunc has been marked with KF_DEPRECATED, the following
577procedure is followed for removal:
578
5791. Any relevant information for deprecated kfuncs is documented in the kfunc's
580   kernel docs. This documentation will typically include the kfunc's expected
581   remaining lifespan, a recommendation for new functionality that can replace
582   the usage of the deprecated function (or an explanation as to why no such
583   replacement exists), etc.
584
5852. The deprecated kfunc is kept in the kernel for some period of time after it
586   was first marked as deprecated. This time period will be chosen on a
587   case-by-case basis, and will typically depend on how widespread the use of
588   the kfunc is, how long it has been in the kernel, and how hard it is to move
589   to alternatives. This deprecation time period is "best effort", and as
590   described :ref:`above<BPF_kfunc_lifecycle_expectations>`, circumstances may
591   sometimes dictate that the kfunc be removed before the full intended
592   deprecation period has elapsed.
593
5943. After the deprecation period the kfunc will be removed. At this point, BPF
595   programs calling the kfunc will be rejected by the verifier.
596
5974. Core kfuncs
598==============
599
600The BPF subsystem provides a number of "core" kfuncs that are potentially
601applicable to a wide variety of different possible use cases and programs.
602Those kfuncs are documented here.
603
6044.1 struct task_struct * kfuncs
605-------------------------------
606
607There are a number of kfuncs that allow ``struct task_struct *`` objects to be
608used as kptrs:
609
610.. kernel-doc:: kernel/bpf/helpers.c
611   :identifiers: bpf_task_acquire bpf_task_release
612
613These kfuncs are useful when you want to acquire or release a reference to a
614``struct task_struct *`` that was passed as e.g. a tracepoint arg, or a
615struct_ops callback arg. For example:
616
617.. code-block:: c
618
619	/**
620	 * A trivial example tracepoint program that shows how to
621	 * acquire and release a struct task_struct * pointer.
622	 */
623	SEC("tp_btf/task_newtask")
624	int BPF_PROG(task_acquire_release_example, struct task_struct *task, u64 clone_flags)
625	{
626		struct task_struct *acquired;
627
628		acquired = bpf_task_acquire(task);
629		if (acquired)
630			/*
631			 * In a typical program you'd do something like store
632			 * the task in a map, and the map will automatically
633			 * release it later. Here, we release it manually.
634			 */
635			bpf_task_release(acquired);
636		return 0;
637	}
638
639
640References acquired on ``struct task_struct *`` objects are RCU protected.
641Therefore, when in an RCU read region, you can obtain a pointer to a task
642embedded in a map value without having to acquire a reference:
643
644.. code-block:: c
645
646	#define private(name) SEC(".data." #name) __hidden __attribute__((aligned(8)))
647	private(TASK) static struct task_struct *global;
648
649	/**
650	 * A trivial example showing how to access a task stored
651	 * in a map using RCU.
652	 */
653	SEC("tp_btf/task_newtask")
654	int BPF_PROG(task_rcu_read_example, struct task_struct *task, u64 clone_flags)
655	{
656		struct task_struct *local_copy;
657
658		bpf_rcu_read_lock();
659		local_copy = global;
660		if (local_copy)
661			/*
662			 * We could also pass local_copy to kfuncs or helper functions here,
663			 * as we're guaranteed that local_copy will be valid until we exit
664			 * the RCU read region below.
665			 */
666			bpf_printk("Global task %s is valid", local_copy->comm);
667		else
668			bpf_printk("No global task found");
669		bpf_rcu_read_unlock();
670
671		/* At this point we can no longer reference local_copy. */
672
673		return 0;
674	}
675
676----
677
678A BPF program can also look up a task from a pid. This can be useful if the
679caller doesn't have a trusted pointer to a ``struct task_struct *`` object that
680it can acquire a reference on with bpf_task_acquire().
681
682.. kernel-doc:: kernel/bpf/helpers.c
683   :identifiers: bpf_task_from_pid
684
685Here is an example of it being used:
686
687.. code-block:: c
688
689	SEC("tp_btf/task_newtask")
690	int BPF_PROG(task_get_pid_example, struct task_struct *task, u64 clone_flags)
691	{
692		struct task_struct *lookup;
693
694		lookup = bpf_task_from_pid(task->pid);
695		if (!lookup)
696			/* A task should always be found, as %task is a tracepoint arg. */
697			return -ENOENT;
698
699		if (lookup->pid != task->pid) {
700			/* bpf_task_from_pid() looks up the task via its
701			 * globally-unique pid from the init_pid_ns. Thus,
702			 * the pid of the lookup task should always be the
703			 * same as the input task.
704			 */
705			bpf_task_release(lookup);
706			return -EINVAL;
707		}
708
709		/* bpf_task_from_pid() returns an acquired reference,
710		 * so it must be dropped before returning from the
711		 * tracepoint handler.
712		 */
713		bpf_task_release(lookup);
714		return 0;
715	}
716
7174.2 struct cgroup * kfuncs
718--------------------------
719
720``struct cgroup *`` objects also have acquire and release functions:
721
722.. kernel-doc:: kernel/bpf/helpers.c
723   :identifiers: bpf_cgroup_acquire bpf_cgroup_release
724
725These kfuncs are used in exactly the same manner as bpf_task_acquire() and
726bpf_task_release() respectively, so we won't provide examples for them.
727
728----
729
730Other kfuncs available for interacting with ``struct cgroup *`` objects are
731bpf_cgroup_ancestor() and bpf_cgroup_from_id(), allowing callers to access
732the ancestor of a cgroup and find a cgroup by its ID, respectively. Both
733return a cgroup kptr.
734
735.. kernel-doc:: kernel/bpf/helpers.c
736   :identifiers: bpf_cgroup_ancestor
737
738.. kernel-doc:: kernel/bpf/helpers.c
739   :identifiers: bpf_cgroup_from_id
740
741Eventually, BPF should be updated to allow this to happen with a normal memory
742load in the program itself. This is currently not possible without more work in
743the verifier. bpf_cgroup_ancestor() can be used as follows:
744
745.. code-block:: c
746
747	/**
748	 * Simple tracepoint example that illustrates how a cgroup's
749	 * ancestor can be accessed using bpf_cgroup_ancestor().
750	 */
751	SEC("tp_btf/cgroup_mkdir")
752	int BPF_PROG(cgrp_ancestor_example, struct cgroup *cgrp, const char *path)
753	{
754		struct cgroup *parent;
755
756		/* The parent cgroup resides at the level before the current cgroup's level. */
757		parent = bpf_cgroup_ancestor(cgrp, cgrp->level - 1);
758		if (!parent)
759			return -ENOENT;
760
761		bpf_printk("Parent id is %d", parent->self.id);
762
763		/* Return the parent cgroup that was acquired above. */
764		bpf_cgroup_release(parent);
765		return 0;
766	}
767
7684.3 struct cpumask * kfuncs
769---------------------------
770
771BPF provides a set of kfuncs that can be used to query, allocate, mutate, and
772destroy struct cpumask * objects. Please refer to :ref:`cpumasks-header-label`
773for more details.
774