Merge branch 'bpf: Implement cgroup local storage available to non-cg…

…roup-attached bpf progs'

Yonghong Song says:

====================

There already exists a local storage implementation for cgroup-attached
bpf programs. See map type BPF_MAP_TYPE_CGROUP_STORAGE and helper
bpf_get_local_storage(). But there are use cases such that non-cgroup
attached bpf progs wants to access cgroup local storage data. For example,
tc egress prog has access to sk and cgroup. It is possible to use
sk local storage to emulate cgroup local storage by storing data in socket.
But this is a waste as it could be lots of sockets belonging to a particular
cgroup. Alternatively, a separate map can be created with cgroup id as the key.
But this will introduce additional overhead to manipulate the new map.
A cgroup local storage, similar to existing sk/inode/task storage,
should help for this use case.

This patch implemented new cgroup local storage available to
non-cgroup-attached bpf programs. In the patch series, Patches 1 and 2
are preparation patches. Patch 3 implemented new cgroup local storage
kernel support. Patches 4 and 5 implemented libbpf and bpftool support.
Patches 6-8 fixed one existing test and added four new tests to validate
kernel/libbpf implementations. Patch 9 added documentation for new
BPF_MAP_TYPE_CGRP_STORAGE map type and comparison of the old and new
cgroup local storage maps.

Changelogs:
  v5 -> v6:
    . fix selftest test_libbpf_str/bpf_map_type_str due to marking
      BPF_MAP_TYPE_CGROUP_STORAGE as deprecated.
    . add cgrp_local_storage test in s390x denylist since the test
      has some fentry/fexit programs.
  v4 -> v5:
    . additional refactoring in patch 2
    . fix the call site for bpf_cgrp_storage_free() in kernel/cgroup/cgroup.c.
    . add a test for progs attaching to cgroups
    . add a negative test (the helper key is a task instead of expected cgroup)
    . some spelling fixes
  v3 -> v4:
    . fix a config guarding problem in kernel/cgroup/cgroup.c when
      cgrp_storage is deleted (CONFIG_CGROUP_BPF => CONFIG_BPF_SYSCALL).
    . rename selftest from cgroup_local_storage.c to cgrp_local_storage.c
      so the name can better align with map name.
    . fix a few misspellings.
  v2 -> v3:
    . fix a config caused kernel test complaint.
    . better description/comments in uapi bpf.h and bpf_cgrp_storage.c.
    . factor code for better resue for map_alloc/map_free.
    . improved explanation in map documentation.
  v1 -> v2:
    . change map name from BPF_MAP_TYPE_CGROUP_LOCAL_STORAGE to
      BPF_MAP_TYPE_CGRP_STORAGE.
    . removed support of sleepable programs.
    . changed the place of freeing cgrp local storage from put_css_set_locked()
      to css_free_rwork_fn().
    . added map documentation.
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Documentation/bpf/map_cgrp_storage.rst

@@ -0,0 +1,109 @@
+.. SPDX-License-Identifier: GPL-2.0-only
+.. Copyright (C) 2022 Meta Platforms, Inc. and affiliates.
+
+=========================
+BPF_MAP_TYPE_CGRP_STORAGE
+=========================
+
+The ``BPF_MAP_TYPE_CGRP_STORAGE`` map type represents a local fix-sized
+storage for cgroups. It is only available with ``CONFIG_CGROUPS``.
+The programs are made available by the same Kconfig. The
+data for a particular cgroup can be retrieved by looking up the map
+with that cgroup.
+
+This document describes the usage and semantics of the
+``BPF_MAP_TYPE_CGRP_STORAGE`` map type.
+
+Usage
+=====
+
+The map key must be ``sizeof(int)`` representing a cgroup fd.
+To access the storage in a program, use ``bpf_cgrp_storage_get``::
+
+    void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
+
+``flags`` could be 0 or ``BPF_LOCAL_STORAGE_GET_F_CREATE`` which indicates that
+a new local storage will be created if one does not exist.
+
+The local storage can be removed with ``bpf_cgrp_storage_delete``::
+
+    long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
+
+The map is available to all program types.
+
+Examples
+========
+
+A BPF program example with BPF_MAP_TYPE_CGRP_STORAGE::
+
+    #include <vmlinux.h>
+    #include <bpf/bpf_helpers.h>
+    #include <bpf/bpf_tracing.h>
+
+    struct {
+            __uint(type, BPF_MAP_TYPE_CGRP_STORAGE);
+            __uint(map_flags, BPF_F_NO_PREALLOC);
+            __type(key, int);
+            __type(value, long);
+    } cgrp_storage SEC(".maps");
+
+    SEC("tp_btf/sys_enter")
+    int BPF_PROG(on_enter, struct pt_regs *regs, long id)
+    {
+            struct task_struct *task = bpf_get_current_task_btf();
+            long *ptr;
+
+            ptr = bpf_cgrp_storage_get(&cgrp_storage, task->cgroups->dfl_cgrp, 0,
+                                       BPF_LOCAL_STORAGE_GET_F_CREATE);
+            if (ptr)
+                __sync_fetch_and_add(ptr, 1);
+
+            return 0;
+    }
+
+Userspace accessing map declared above::
+
+    #include <linux/bpf.h>
+    #include <linux/libbpf.h>
+
+    __u32 map_lookup(struct bpf_map *map, int cgrp_fd)
+    {
+            __u32 *value;
+            value = bpf_map_lookup_elem(bpf_map__fd(map), &cgrp_fd);
+            if (value)
+                return *value;
+            return 0;
+    }
+
+Difference Between BPF_MAP_TYPE_CGRP_STORAGE and BPF_MAP_TYPE_CGROUP_STORAGE
+============================================================================
+
+The old cgroup storage map ``BPF_MAP_TYPE_CGROUP_STORAGE`` has been marked as
+deprecated (renamed to ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED``). The new
+``BPF_MAP_TYPE_CGRP_STORAGE`` map should be used instead. The following
+illusates the main difference between ``BPF_MAP_TYPE_CGRP_STORAGE`` and
+``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED``.
+
+(1). ``BPF_MAP_TYPE_CGRP_STORAGE`` can be used by all program types while
+     ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED`` is available only to cgroup program types
+     like BPF_CGROUP_INET_INGRESS or BPF_CGROUP_SOCK_OPS, etc.
+
+(2). ``BPF_MAP_TYPE_CGRP_STORAGE`` supports local storage for more than one
+     cgroup while ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED`` only supports one cgroup
+     which is attached by a BPF program.
+
+(3). ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED`` allocates local storage at attach time so
+     ``bpf_get_local_storage()`` always returns non-NULL local storage.
+     ``BPF_MAP_TYPE_CGRP_STORAGE`` allocates local storage at runtime so
+     it is possible that ``bpf_cgrp_storage_get()`` may return null local storage.
+     To avoid such null local storage issue, user space can do
+     ``bpf_map_update_elem()`` to pre-allocate local storage before a BPF program
+     is attached.
+
+(4). ``BPF_MAP_TYPE_CGRP_STORAGE`` supports deleting local storage by a BPF program
+     while ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED`` only deletes storage during
+     prog detach time.
+
+So overall, ``BPF_MAP_TYPE_CGRP_STORAGE`` supports all ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED``
+functionality and beyond. It is recommended to use ``BPF_MAP_TYPE_CGRP_STORAGE``
+instead of ``BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED``.

include/linux/bpf.h

@@ -2041,6 +2041,7 @@ struct bpf_link *bpf_link_by_id(u32 id);
 
 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
 void bpf_task_storage_free(struct task_struct *task);
+void bpf_cgrp_storage_free(struct cgroup *cgroup);
 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
 const struct btf_func_model *
 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
@@ -2295,6 +2296,10 @@ static inline bool has_current_bpf_ctx(void)
 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog)
 {
 }
+
+static inline void bpf_cgrp_storage_free(struct cgroup *cgroup)
+{
+}
 #endif /* CONFIG_BPF_SYSCALL */
 
 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
@@ -2535,6 +2540,8 @@ extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
 extern const struct bpf_func_proto bpf_set_retval_proto;
 extern const struct bpf_func_proto bpf_get_retval_proto;
 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto;
+extern const struct bpf_func_proto bpf_cgrp_storage_get_proto;
+extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto;
 
 const struct bpf_func_proto *tracing_prog_func_proto(
   enum bpf_func_id func_id, const struct bpf_prog *prog);

include/linux/bpf_local_storage.h

@@ -116,21 +116,22 @@ static struct bpf_local_storage_cache name = {			\
 	.idx_lock = __SPIN_LOCK_UNLOCKED(name.idx_lock),	\
 }
 
-u16 bpf_local_storage_cache_idx_get(struct bpf_local_storage_cache *cache);
-void bpf_local_storage_cache_idx_free(struct bpf_local_storage_cache *cache,
-				      u16 idx);
-
 /* Helper functions for bpf_local_storage */
 int bpf_local_storage_map_alloc_check(union bpf_attr *attr);
 
-struct bpf_local_storage_map *bpf_local_storage_map_alloc(union bpf_attr *attr);
+struct bpf_map *
+bpf_local_storage_map_alloc(union bpf_attr *attr,
+			    struct bpf_local_storage_cache *cache);
 
 struct bpf_local_storage_data *
 bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
 			 struct bpf_local_storage_map *smap,
 			 bool cacheit_lockit);
 
-void bpf_local_storage_map_free(struct bpf_local_storage_map *smap,
+bool bpf_local_storage_unlink_nolock(struct bpf_local_storage *local_storage);
+
+void bpf_local_storage_map_free(struct bpf_map *map,
+				struct bpf_local_storage_cache *cache,
 				int __percpu *busy_counter);
 
 int bpf_local_storage_map_check_btf(const struct bpf_map *map,
@@ -141,10 +142,6 @@ int bpf_local_storage_map_check_btf(const struct bpf_map *map,
 void bpf_selem_link_storage_nolock(struct bpf_local_storage *local_storage,
 				   struct bpf_local_storage_elem *selem);
 
-bool bpf_selem_unlink_storage_nolock(struct bpf_local_storage *local_storage,
-				     struct bpf_local_storage_elem *selem,
-				     bool uncharge_omem, bool use_trace_rcu);
-
 void bpf_selem_unlink(struct bpf_local_storage_elem *selem, bool use_trace_rcu);
 
 void bpf_selem_link_map(struct bpf_local_storage_map *smap,

include/linux/bpf_types.h

@@ -86,6 +86,7 @@ BPF_MAP_TYPE(BPF_MAP_TYPE_PROG_ARRAY, prog_array_map_ops)
 BPF_MAP_TYPE(BPF_MAP_TYPE_PERF_EVENT_ARRAY, perf_event_array_map_ops)
 #ifdef CONFIG_CGROUPS
 BPF_MAP_TYPE(BPF_MAP_TYPE_CGROUP_ARRAY, cgroup_array_map_ops)
+BPF_MAP_TYPE(BPF_MAP_TYPE_CGRP_STORAGE, cgrp_storage_map_ops)
 #endif
 #ifdef CONFIG_CGROUP_BPF
 BPF_MAP_TYPE(BPF_MAP_TYPE_CGROUP_STORAGE, cgroup_storage_map_ops)

include/linux/btf_ids.h

@@ -265,5 +265,6 @@ MAX_BTF_TRACING_TYPE,
 };
 
 extern u32 btf_tracing_ids[];
+extern u32 bpf_cgroup_btf_id[];
 
 #endif

include/linux/cgroup-defs.h

@@ -504,6 +504,10 @@ struct cgroup {
 	/* Used to store internal freezer state */
 	struct cgroup_freezer_state freezer;
 
+#ifdef CONFIG_BPF_SYSCALL
+	struct bpf_local_storage __rcu  *bpf_cgrp_storage;
+#endif
+
 	/* All ancestors including self */
 	struct cgroup *ancestors[];
 };

include/uapi/linux/bpf.h

@@ -922,7 +922,14 @@ enum bpf_map_type {
 	BPF_MAP_TYPE_CPUMAP,
 	BPF_MAP_TYPE_XSKMAP,
 	BPF_MAP_TYPE_SOCKHASH,
-	BPF_MAP_TYPE_CGROUP_STORAGE,
+	BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
+	/* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
+	 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
+	 * both cgroup-attached and other progs and supports all functionality
+	 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
+	 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
+	 */
+	BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
 	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
 	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
 	BPF_MAP_TYPE_QUEUE,
@@ -935,6 +942,7 @@ enum bpf_map_type {
 	BPF_MAP_TYPE_TASK_STORAGE,
 	BPF_MAP_TYPE_BLOOM_FILTER,
 	BPF_MAP_TYPE_USER_RINGBUF,
+	BPF_MAP_TYPE_CGRP_STORAGE,
 };
 
 /* Note that tracing related programs such as
@@ -5435,6 +5443,44 @@ union bpf_attr {
  *		**-E2BIG** if user-space has tried to publish a sample which is
  *		larger than the size of the ring buffer, or which cannot fit
  *		within a struct bpf_dynptr.
+ *
+ * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
+ *	Description
+ *		Get a bpf_local_storage from the *cgroup*.
+ *
+ *		Logically, it could be thought of as getting the value from
+ *		a *map* with *cgroup* as the **key**.  From this
+ *		perspective,  the usage is not much different from
+ *		**bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
+ *		helper enforces the key must be a cgroup struct and the map must also
+ *		be a **BPF_MAP_TYPE_CGRP_STORAGE**.
+ *
+ *		In reality, the local-storage value is embedded directly inside of the
+ *		*cgroup* object itself, rather than being located in the
+ *		**BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
+ *		queried for some *map* on a *cgroup* object, the kernel will perform an
+ *		O(n) iteration over all of the live local-storage values for that
+ *		*cgroup* object until the local-storage value for the *map* is found.
+ *
+ *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
+ *		used such that a new bpf_local_storage will be
+ *		created if one does not exist.  *value* can be used
+ *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
+ *		the initial value of a bpf_local_storage.  If *value* is
+ *		**NULL**, the new bpf_local_storage will be zero initialized.
+ *	Return
+ *		A bpf_local_storage pointer is returned on success.
+ *
+ *		**NULL** if not found or there was an error in adding
+ *		a new bpf_local_storage.
+ *
+ * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
+ *	Description
+ *		Delete a bpf_local_storage from a *cgroup*.
+ *	Return
+ *		0 on success.
+ *
+ *		**-ENOENT** if the bpf_local_storage cannot be found.
  */
 #define ___BPF_FUNC_MAPPER(FN, ctx...)			\
 	FN(unspec, 0, ##ctx)				\
@@ -5647,6 +5693,8 @@ union bpf_attr {
 	FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx)	\
 	FN(ktime_get_tai_ns, 208, ##ctx)		\
 	FN(user_ringbuf_drain, 209, ##ctx)		\
+	FN(cgrp_storage_get, 210, ##ctx)		\
+	FN(cgrp_storage_delete, 211, ##ctx)		\
 	/* */
 
 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't

kernel/bpf/Makefile

@@ -25,7 +25,7 @@ ifeq ($(CONFIG_PERF_EVENTS),y)
 obj-$(CONFIG_BPF_SYSCALL) += stackmap.o
 endif
 ifeq ($(CONFIG_CGROUPS),y)
-obj-$(CONFIG_BPF_SYSCALL) += cgroup_iter.o
+obj-$(CONFIG_BPF_SYSCALL) += cgroup_iter.o bpf_cgrp_storage.o
 endif
 obj-$(CONFIG_CGROUP_BPF) += cgroup.o
 ifeq ($(CONFIG_INET),y)