bpf: Implement bpf_local_storage for inodes
authorKP Singh <kpsingh@google.com>
Tue, 25 Aug 2020 18:29:17 +0000 (20:29 +0200)
committerAlexei Starovoitov <ast@kernel.org>
Tue, 25 Aug 2020 22:00:04 +0000 (15:00 -0700)
Similar to bpf_local_storage for sockets, add local storage for inodes.
The life-cycle of storage is managed with the life-cycle of the inode.
i.e. the storage is destroyed along with the owning inode.

The BPF LSM allocates an __rcu pointer to the bpf_local_storage in the
security blob which are now stackable and can co-exist with other LSMs.

Signed-off-by: KP Singh <kpsingh@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200825182919.1118197-6-kpsingh@chromium.org
13 files changed:
include/linux/bpf_lsm.h
include/linux/bpf_types.h
include/uapi/linux/bpf.h
kernel/bpf/Makefile
kernel/bpf/bpf_inode_storage.c [new file with mode: 0644]
kernel/bpf/syscall.c
kernel/bpf/verifier.c
security/bpf/hooks.c
tools/bpf/bpftool/Documentation/bpftool-map.rst
tools/bpf/bpftool/bash-completion/bpftool
tools/bpf/bpftool/map.c
tools/include/uapi/linux/bpf.h
tools/lib/bpf/libbpf_probes.c

index af74712..aaacb6a 100644 (file)
 #include <linux/lsm_hook_defs.h>
 #undef LSM_HOOK
 
+struct bpf_storage_blob {
+       struct bpf_local_storage __rcu *storage;
+};
+
+extern struct lsm_blob_sizes bpf_lsm_blob_sizes;
+
 int bpf_lsm_verify_prog(struct bpf_verifier_log *vlog,
                        const struct bpf_prog *prog);
 
+static inline struct bpf_storage_blob *bpf_inode(
+       const struct inode *inode)
+{
+       if (unlikely(!inode->i_security))
+               return NULL;
+
+       return inode->i_security + bpf_lsm_blob_sizes.lbs_inode;
+}
+
+extern const struct bpf_func_proto bpf_inode_storage_get_proto;
+extern const struct bpf_func_proto bpf_inode_storage_delete_proto;
+void bpf_inode_storage_free(struct inode *inode);
+
 #else /* !CONFIG_BPF_LSM */
 
 static inline int bpf_lsm_verify_prog(struct bpf_verifier_log *vlog,
@@ -28,6 +47,16 @@ static inline int bpf_lsm_verify_prog(struct bpf_verifier_log *vlog,
        return -EOPNOTSUPP;
 }
 
+static inline struct bpf_storage_blob *bpf_inode(
+       const struct inode *inode)
+{
+       return NULL;
+}
+
+static inline void bpf_inode_storage_free(struct inode *inode)
+{
+}
+
 #endif /* CONFIG_BPF_LSM */
 
 #endif /* _LINUX_BPF_LSM_H */
index a52a568..2e6f568 100644 (file)
@@ -107,6 +107,9 @@ BPF_MAP_TYPE(BPF_MAP_TYPE_SK_STORAGE, sk_storage_map_ops)
 BPF_MAP_TYPE(BPF_MAP_TYPE_SOCKMAP, sock_map_ops)
 BPF_MAP_TYPE(BPF_MAP_TYPE_SOCKHASH, sock_hash_ops)
 #endif
+#ifdef CONFIG_BPF_LSM
+BPF_MAP_TYPE(BPF_MAP_TYPE_INODE_STORAGE, inode_storage_map_ops)
+#endif
 BPF_MAP_TYPE(BPF_MAP_TYPE_CPUMAP, cpu_map_ops)
 #if defined(CONFIG_XDP_SOCKETS)
 BPF_MAP_TYPE(BPF_MAP_TYPE_XSKMAP, xsk_map_ops)
index 2cbd137..b6bfcd0 100644 (file)
@@ -155,6 +155,7 @@ enum bpf_map_type {
        BPF_MAP_TYPE_DEVMAP_HASH,
        BPF_MAP_TYPE_STRUCT_OPS,
        BPF_MAP_TYPE_RINGBUF,
+       BPF_MAP_TYPE_INODE_STORAGE,
 };
 
 /* Note that tracing related programs such as
@@ -3509,6 +3510,41 @@ union bpf_attr {
  *
  *             **-EPERM** This helper cannot be used under the
  *                        current sock_ops->op.
+ * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
+ *     Description
+ *             Get a bpf_local_storage from an *inode*.
+ *
+ *             Logically, it could be thought of as getting the value from
+ *             a *map* with *inode* as the **key**.  From this
+ *             perspective,  the usage is not much different from
+ *             **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
+ *             helper enforces the key must be an inode and the map must also
+ *             be a **BPF_MAP_TYPE_INODE_STORAGE**.
+ *
+ *             Underneath, the value is stored locally at *inode* instead of
+ *             the *map*.  The *map* is used as the bpf-local-storage
+ *             "type". The bpf-local-storage "type" (i.e. the *map*) is
+ *             searched against all bpf_local_storage residing at *inode*.
+ *
+ *             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.
+ *
+ * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
+ *     Description
+ *             Delete a bpf_local_storage from an *inode*.
+ *     Return
+ *             0 on success.
+ *
+ *             **-ENOENT** if the bpf_local_storage cannot be found.
  */
 #define __BPF_FUNC_MAPPER(FN)          \
        FN(unspec),                     \
@@ -3655,7 +3691,9 @@ union bpf_attr {
        FN(get_task_stack),             \
        FN(load_hdr_opt),               \
        FN(store_hdr_opt),              \
-       FN(reserve_hdr_opt),
+       FN(reserve_hdr_opt),            \
+       FN(inode_storage_get),          \
+       FN(inode_storage_delete),       \
        /* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
index 6961ff4..bdc8cd1 100644 (file)
@@ -5,6 +5,7 @@ CFLAGS_core.o += $(call cc-disable-warning, override-init)
 obj-$(CONFIG_BPF_SYSCALL) += syscall.o verifier.o inode.o helpers.o tnum.o bpf_iter.o map_iter.o task_iter.o prog_iter.o
 obj-$(CONFIG_BPF_SYSCALL) += hashtab.o arraymap.o percpu_freelist.o bpf_lru_list.o lpm_trie.o map_in_map.o
 obj-$(CONFIG_BPF_SYSCALL) += local_storage.o queue_stack_maps.o ringbuf.o
+obj-${CONFIG_BPF_LSM}    += bpf_inode_storage.o
 obj-$(CONFIG_BPF_SYSCALL) += disasm.o
 obj-$(CONFIG_BPF_JIT) += trampoline.o
 obj-$(CONFIG_BPF_SYSCALL) += btf.o
diff --git a/kernel/bpf/bpf_inode_storage.c b/kernel/bpf/bpf_inode_storage.c
new file mode 100644 (file)
index 0000000..f3a44e9
--- /dev/null
@@ -0,0 +1,273 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2019 Facebook
+ * Copyright 2020 Google LLC.
+ */
+
+#include <linux/rculist.h>
+#include <linux/list.h>
+#include <linux/hash.h>
+#include <linux/types.h>
+#include <linux/spinlock.h>
+#include <linux/bpf.h>
+#include <linux/bpf_local_storage.h>
+#include <net/sock.h>
+#include <uapi/linux/sock_diag.h>
+#include <uapi/linux/btf.h>
+#include <linux/bpf_lsm.h>
+#include <linux/btf_ids.h>
+#include <linux/fdtable.h>
+
+DEFINE_BPF_STORAGE_CACHE(inode_cache);
+
+static struct bpf_local_storage __rcu **
+inode_storage_ptr(void *owner)
+{
+       struct inode *inode = owner;
+       struct bpf_storage_blob *bsb;
+
+       bsb = bpf_inode(inode);
+       if (!bsb)
+               return NULL;
+       return &bsb->storage;
+}
+
+static struct bpf_local_storage_data *inode_storage_lookup(struct inode *inode,
+                                                          struct bpf_map *map,
+                                                          bool cacheit_lockit)
+{
+       struct bpf_local_storage *inode_storage;
+       struct bpf_local_storage_map *smap;
+       struct bpf_storage_blob *bsb;
+
+       bsb = bpf_inode(inode);
+       if (!bsb)
+               return NULL;
+
+       inode_storage = rcu_dereference(bsb->storage);
+       if (!inode_storage)
+               return NULL;
+
+       smap = (struct bpf_local_storage_map *)map;
+       return bpf_local_storage_lookup(inode_storage, smap, cacheit_lockit);
+}
+
+void bpf_inode_storage_free(struct inode *inode)
+{
+       struct bpf_local_storage_elem *selem;
+       struct bpf_local_storage *local_storage;
+       bool free_inode_storage = false;
+       struct bpf_storage_blob *bsb;
+       struct hlist_node *n;
+
+       bsb = bpf_inode(inode);
+       if (!bsb)
+               return;
+
+       rcu_read_lock();
+
+       local_storage = rcu_dereference(bsb->storage);
+       if (!local_storage) {
+               rcu_read_unlock();
+               return;
+       }
+
+       /* Netiher the bpf_prog nor the bpf-map's syscall
+        * could be modifying the local_storage->list now.
+        * Thus, no elem can be added-to or deleted-from the
+        * local_storage->list by the bpf_prog or by the bpf-map's syscall.
+        *
+        * It is racing with bpf_local_storage_map_free() alone
+        * when unlinking elem from the local_storage->list and
+        * the map's bucket->list.
+        */
+       raw_spin_lock_bh(&local_storage->lock);
+       hlist_for_each_entry_safe(selem, n, &local_storage->list, snode) {
+               /* Always unlink from map before unlinking from
+                * local_storage.
+                */
+               bpf_selem_unlink_map(selem);
+               free_inode_storage = bpf_selem_unlink_storage_nolock(
+                       local_storage, selem, false);
+       }
+       raw_spin_unlock_bh(&local_storage->lock);
+       rcu_read_unlock();
+
+       /* free_inoode_storage should always be true as long as
+        * local_storage->list was non-empty.
+        */
+       if (free_inode_storage)
+               kfree_rcu(local_storage, rcu);
+}
+
+static void *bpf_fd_inode_storage_lookup_elem(struct bpf_map *map, void *key)
+{
+       struct bpf_local_storage_data *sdata;
+       struct file *f;
+       int fd;
+
+       fd = *(int *)key;
+       f = fget_raw(fd);
+       if (!f)
+               return NULL;
+
+       sdata = inode_storage_lookup(f->f_inode, map, true);
+       fput(f);
+       return sdata ? sdata->data : NULL;
+}
+
+static int bpf_fd_inode_storage_update_elem(struct bpf_map *map, void *key,
+                                        void *value, u64 map_flags)
+{
+       struct bpf_local_storage_data *sdata;
+       struct file *f;
+       int fd;
+
+       fd = *(int *)key;
+       f = fget_raw(fd);
+       if (!f || !inode_storage_ptr(f->f_inode))
+               return -EBADF;
+
+       sdata = bpf_local_storage_update(f->f_inode,
+                                        (struct bpf_local_storage_map *)map,
+                                        value, map_flags);
+       fput(f);
+       return PTR_ERR_OR_ZERO(sdata);
+}
+
+static int inode_storage_delete(struct inode *inode, struct bpf_map *map)
+{
+       struct bpf_local_storage_data *sdata;
+
+       sdata = inode_storage_lookup(inode, map, false);
+       if (!sdata)
+               return -ENOENT;
+
+       bpf_selem_unlink(SELEM(sdata));
+
+       return 0;
+}
+
+static int bpf_fd_inode_storage_delete_elem(struct bpf_map *map, void *key)
+{
+       struct file *f;
+       int fd, err;
+
+       fd = *(int *)key;
+       f = fget_raw(fd);
+       if (!f)
+               return -EBADF;
+
+       err = inode_storage_delete(f->f_inode, map);
+       fput(f);
+       return err;
+}
+
+BPF_CALL_4(bpf_inode_storage_get, struct bpf_map *, map, struct inode *, inode,
+          void *, value, u64, flags)
+{
+       struct bpf_local_storage_data *sdata;
+
+       if (flags & ~(BPF_LOCAL_STORAGE_GET_F_CREATE))
+               return (unsigned long)NULL;
+
+       /* explicitly check that the inode_storage_ptr is not
+        * NULL as inode_storage_lookup returns NULL in this case and
+        * bpf_local_storage_update expects the owner to have a
+        * valid storage pointer.
+        */
+       if (!inode_storage_ptr(inode))
+               return (unsigned long)NULL;
+
+       sdata = inode_storage_lookup(inode, map, true);
+       if (sdata)
+               return (unsigned long)sdata->data;
+
+       /* This helper must only called from where the inode is gurranteed
+        * to have a refcount and cannot be freed.
+        */
+       if (flags & BPF_LOCAL_STORAGE_GET_F_CREATE) {
+               sdata = bpf_local_storage_update(
+                       inode, (struct bpf_local_storage_map *)map, value,
+                       BPF_NOEXIST);
+               return IS_ERR(sdata) ? (unsigned long)NULL :
+                                            (unsigned long)sdata->data;
+       }
+
+       return (unsigned long)NULL;
+}
+
+BPF_CALL_2(bpf_inode_storage_delete,
+          struct bpf_map *, map, struct inode *, inode)
+{
+       /* This helper must only called from where the inode is gurranteed
+        * to have a refcount and cannot be freed.
+        */
+       return inode_storage_delete(inode, map);
+}
+
+static int notsupp_get_next_key(struct bpf_map *map, void *key,
+                               void *next_key)
+{
+       return -ENOTSUPP;
+}
+
+static struct bpf_map *inode_storage_map_alloc(union bpf_attr *attr)
+{
+       struct bpf_local_storage_map *smap;
+
+       smap = bpf_local_storage_map_alloc(attr);
+       if (IS_ERR(smap))
+               return ERR_CAST(smap);
+
+       smap->cache_idx = bpf_local_storage_cache_idx_get(&inode_cache);
+       return &smap->map;
+}
+
+static void inode_storage_map_free(struct bpf_map *map)
+{
+       struct bpf_local_storage_map *smap;
+
+       smap = (struct bpf_local_storage_map *)map;
+       bpf_local_storage_cache_idx_free(&inode_cache, smap->cache_idx);
+       bpf_local_storage_map_free(smap);
+}
+
+static int inode_storage_map_btf_id;
+const struct bpf_map_ops inode_storage_map_ops = {
+       .map_alloc_check = bpf_local_storage_map_alloc_check,
+       .map_alloc = inode_storage_map_alloc,
+       .map_free = inode_storage_map_free,
+       .map_get_next_key = notsupp_get_next_key,
+       .map_lookup_elem = bpf_fd_inode_storage_lookup_elem,
+       .map_update_elem = bpf_fd_inode_storage_update_elem,
+       .map_delete_elem = bpf_fd_inode_storage_delete_elem,
+       .map_check_btf = bpf_local_storage_map_check_btf,
+       .map_btf_name = "bpf_local_storage_map",
+       .map_btf_id = &inode_storage_map_btf_id,
+       .map_owner_storage_ptr = inode_storage_ptr,
+};
+
+BTF_ID_LIST(bpf_inode_storage_btf_ids)
+BTF_ID_UNUSED
+BTF_ID(struct, inode)
+
+const struct bpf_func_proto bpf_inode_storage_get_proto = {
+       .func           = bpf_inode_storage_get,
+       .gpl_only       = false,
+       .ret_type       = RET_PTR_TO_MAP_VALUE_OR_NULL,
+       .arg1_type      = ARG_CONST_MAP_PTR,
+       .arg2_type      = ARG_PTR_TO_BTF_ID,
+       .arg3_type      = ARG_PTR_TO_MAP_VALUE_OR_NULL,
+       .arg4_type      = ARG_ANYTHING,
+       .btf_id         = bpf_inode_storage_btf_ids,
+};
+
+const struct bpf_func_proto bpf_inode_storage_delete_proto = {
+       .func           = bpf_inode_storage_delete,
+       .gpl_only       = false,
+       .ret_type       = RET_INTEGER,
+       .arg1_type      = ARG_CONST_MAP_PTR,
+       .arg2_type      = ARG_PTR_TO_BTF_ID,
+       .btf_id         = bpf_inode_storage_btf_ids,
+};
index b46e973..5443cea 100644 (file)
@@ -769,7 +769,8 @@ static int map_check_btf(struct bpf_map *map, const struct btf *btf,
                if (map->map_type != BPF_MAP_TYPE_HASH &&
                    map->map_type != BPF_MAP_TYPE_ARRAY &&
                    map->map_type != BPF_MAP_TYPE_CGROUP_STORAGE &&
-                   map->map_type != BPF_MAP_TYPE_SK_STORAGE)
+                   map->map_type != BPF_MAP_TYPE_SK_STORAGE &&
+                   map->map_type != BPF_MAP_TYPE_INODE_STORAGE)
                        return -ENOTSUPP;
                if (map->spin_lock_off + sizeof(struct bpf_spin_lock) >
                    map->value_size) {
index dd24503..3874879 100644 (file)
@@ -4311,6 +4311,11 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env,
                    func_id != BPF_FUNC_sk_storage_delete)
                        goto error;
                break;
+       case BPF_MAP_TYPE_INODE_STORAGE:
+               if (func_id != BPF_FUNC_inode_storage_get &&
+                   func_id != BPF_FUNC_inode_storage_delete)
+                       goto error;
+               break;
        default:
                break;
        }
@@ -4384,6 +4389,11 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env,
                if (map->map_type != BPF_MAP_TYPE_SK_STORAGE)
                        goto error;
                break;
+       case BPF_FUNC_inode_storage_get:
+       case BPF_FUNC_inode_storage_delete:
+               if (map->map_type != BPF_MAP_TYPE_INODE_STORAGE)
+                       goto error;
+               break;
        default:
                break;
        }
index 32d32d4..788667d 100644 (file)
@@ -11,6 +11,7 @@ static struct security_hook_list bpf_lsm_hooks[] __lsm_ro_after_init = {
        LSM_HOOK_INIT(NAME, bpf_lsm_##NAME),
        #include <linux/lsm_hook_defs.h>
        #undef LSM_HOOK
+       LSM_HOOK_INIT(inode_free_security, bpf_inode_storage_free),
 };
 
 static int __init bpf_lsm_init(void)
@@ -20,7 +21,12 @@ static int __init bpf_lsm_init(void)
        return 0;
 }
 
+struct lsm_blob_sizes bpf_lsm_blob_sizes __lsm_ro_after_init = {
+       .lbs_inode = sizeof(struct bpf_storage_blob),
+};
+
 DEFINE_LSM(bpf) = {
        .name = "bpf",
        .init = bpf_lsm_init,
+       .blobs = &bpf_lsm_blob_sizes
 };
index 41e2a74..083db6c 100644 (file)
@@ -49,7 +49,7 @@ MAP COMMANDS
 |              | **lru_percpu_hash** | **lpm_trie** | **array_of_maps** | **hash_of_maps**
 |              | **devmap** | **devmap_hash** | **sockmap** | **cpumap** | **xskmap** | **sockhash**
 |              | **cgroup_storage** | **reuseport_sockarray** | **percpu_cgroup_storage**
-|              | **queue** | **stack** | **sk_storage** | **struct_ops** | **ringbuf** }
+|              | **queue** | **stack** | **sk_storage** | **struct_ops** | **ringbuf** | **inode_storage** }
 
 DESCRIPTION
 ===========
index f53ed2f..7b68e3c 100644 (file)
@@ -704,7 +704,8 @@ _bpftool()
                                 lru_percpu_hash lpm_trie array_of_maps \
                                 hash_of_maps devmap devmap_hash sockmap cpumap \
                                 xskmap sockhash cgroup_storage reuseport_sockarray \
-                                percpu_cgroup_storage queue stack' -- \
+                                percpu_cgroup_storage queue stack sk_storage \
+                                struct_ops inode_storage' -- \
                                                    "$cur" ) )
                             return 0
                             ;;
index 3a27d31..bc00712 100644 (file)
@@ -50,6 +50,7 @@ const char * const map_type_name[] = {
        [BPF_MAP_TYPE_SK_STORAGE]               = "sk_storage",
        [BPF_MAP_TYPE_STRUCT_OPS]               = "struct_ops",
        [BPF_MAP_TYPE_RINGBUF]                  = "ringbuf",
+       [BPF_MAP_TYPE_INODE_STORAGE]            = "inode_storage",
 };
 
 const size_t map_type_name_size = ARRAY_SIZE(map_type_name);
@@ -1442,7 +1443,7 @@ static int do_help(int argc, char **argv)
                "                 lru_percpu_hash | lpm_trie | array_of_maps | hash_of_maps |\n"
                "                 devmap | devmap_hash | sockmap | cpumap | xskmap | sockhash |\n"
                "                 cgroup_storage | reuseport_sockarray | percpu_cgroup_storage |\n"
-               "                 queue | stack | sk_storage | struct_ops | ringbuf }\n"
+               "                 queue | stack | sk_storage | struct_ops | ringbuf | inode_storage }\n"
                "       " HELP_SPEC_OPTIONS "\n"
                "",
                bin_name, argv[-2]);
index 2cbd137..b6bfcd0 100644 (file)
@@ -155,6 +155,7 @@ enum bpf_map_type {
        BPF_MAP_TYPE_DEVMAP_HASH,
        BPF_MAP_TYPE_STRUCT_OPS,
        BPF_MAP_TYPE_RINGBUF,
+       BPF_MAP_TYPE_INODE_STORAGE,
 };
 
 /* Note that tracing related programs such as
@@ -3509,6 +3510,41 @@ union bpf_attr {
  *
  *             **-EPERM** This helper cannot be used under the
  *                        current sock_ops->op.
+ * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
+ *     Description
+ *             Get a bpf_local_storage from an *inode*.
+ *
+ *             Logically, it could be thought of as getting the value from
+ *             a *map* with *inode* as the **key**.  From this
+ *             perspective,  the usage is not much different from
+ *             **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
+ *             helper enforces the key must be an inode and the map must also
+ *             be a **BPF_MAP_TYPE_INODE_STORAGE**.
+ *
+ *             Underneath, the value is stored locally at *inode* instead of
+ *             the *map*.  The *map* is used as the bpf-local-storage
+ *             "type". The bpf-local-storage "type" (i.e. the *map*) is
+ *             searched against all bpf_local_storage residing at *inode*.
+ *
+ *             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.
+ *
+ * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
+ *     Description
+ *             Delete a bpf_local_storage from an *inode*.
+ *     Return
+ *             0 on success.
+ *
+ *             **-ENOENT** if the bpf_local_storage cannot be found.
  */
 #define __BPF_FUNC_MAPPER(FN)          \
        FN(unspec),                     \
@@ -3655,7 +3691,9 @@ union bpf_attr {
        FN(get_task_stack),             \
        FN(load_hdr_opt),               \
        FN(store_hdr_opt),              \
-       FN(reserve_hdr_opt),
+       FN(reserve_hdr_opt),            \
+       FN(inode_storage_get),          \
+       FN(inode_storage_delete),       \
        /* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
index 010c9a7..5482a9b 100644 (file)
@@ -170,7 +170,7 @@ int libbpf__load_raw_btf(const char *raw_types, size_t types_len,
        return btf_fd;
 }
 
-static int load_sk_storage_btf(void)
+static int load_local_storage_btf(void)
 {
        const char strs[] = "\0bpf_spin_lock\0val\0cnt\0l";
        /* struct bpf_spin_lock {
@@ -229,12 +229,13 @@ bool bpf_probe_map_type(enum bpf_map_type map_type, __u32 ifindex)
                key_size        = 0;
                break;
        case BPF_MAP_TYPE_SK_STORAGE:
+       case BPF_MAP_TYPE_INODE_STORAGE:
                btf_key_type_id = 1;
                btf_value_type_id = 3;
                value_size = 8;
                max_entries = 0;
                map_flags = BPF_F_NO_PREALLOC;
-               btf_fd = load_sk_storage_btf();
+               btf_fd = load_local_storage_btf();
                if (btf_fd < 0)
                        return false;
                break;