blob: 73cef8913da1fe4d7f54f53324c41aa86de80298 [file] [log] [blame]
/*
* Copyright (C) 2021 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#pragma once
#include <stdlib.h>
#include <unistd.h>
#include <linux/bpf.h>
#include <linux/unistd.h>
#include <sys/file.h>
#ifdef BPF_FD_JUST_USE_INT
#define BPF_FD_TYPE int
#define BPF_FD_TO_U32(x) static_cast<__u32>(x)
#else
#include <android-base/unique_fd.h>
#define BPF_FD_TYPE base::unique_fd&
#define BPF_FD_TO_U32(x) static_cast<__u32>((x).get())
#endif
namespace android {
namespace bpf {
inline uint64_t ptr_to_u64(const void * const x) {
return (uint64_t)(uintptr_t)x;
}
/* Note: bpf_attr is a union which might have a much larger size then the anonymous struct portion
* of it that we are using. The kernel's bpf() system call will perform a strict check to ensure
* all unused portions are zero. It will fail with E2BIG if we don't fully zero bpf_attr.
*/
inline int bpf(enum bpf_cmd cmd, const bpf_attr& attr) {
return syscall(__NR_bpf, cmd, &attr, sizeof(attr));
}
// this version is meant for use with cmd's which mutate the argument
inline int bpf(enum bpf_cmd cmd, bpf_attr *attr) {
return syscall(__NR_bpf, cmd, attr, sizeof(*attr));
}
inline int createMap(bpf_map_type map_type, uint32_t key_size, uint32_t value_size,
uint32_t max_entries, uint32_t map_flags) {
return bpf(BPF_MAP_CREATE, {
.map_type = map_type,
.key_size = key_size,
.value_size = value_size,
.max_entries = max_entries,
.map_flags = map_flags,
});
}
// Note:
// 'map_type' must be one of BPF_MAP_TYPE_{ARRAY,HASH}_OF_MAPS
// 'value_size' must be sizeof(u32), ie. 4
// 'inner_map_fd' is basically a template specifying {map_type, key_size, value_size, max_entries, map_flags}
// of the inner map type (and possibly only key_size/value_size actually matter?).
inline int createOuterMap(bpf_map_type map_type, uint32_t key_size, uint32_t value_size,
uint32_t max_entries, uint32_t map_flags, const BPF_FD_TYPE inner_map_fd) {
return bpf(BPF_MAP_CREATE, {
.map_type = map_type,
.key_size = key_size,
.value_size = value_size,
.max_entries = max_entries,
.map_flags = map_flags,
.inner_map_fd = BPF_FD_TO_U32(inner_map_fd),
});
}
inline int writeToMapEntry(const BPF_FD_TYPE map_fd, const void* key, const void* value,
uint64_t flags) {
return bpf(BPF_MAP_UPDATE_ELEM, {
.map_fd = BPF_FD_TO_U32(map_fd),
.key = ptr_to_u64(key),
.value = ptr_to_u64(value),
.flags = flags,
});
}
inline int findMapEntry(const BPF_FD_TYPE map_fd, const void* key, void* value) {
return bpf(BPF_MAP_LOOKUP_ELEM, {
.map_fd = BPF_FD_TO_U32(map_fd),
.key = ptr_to_u64(key),
.value = ptr_to_u64(value),
});
}
inline int deleteMapEntry(const BPF_FD_TYPE map_fd, const void* key) {
return bpf(BPF_MAP_DELETE_ELEM, {
.map_fd = BPF_FD_TO_U32(map_fd),
.key = ptr_to_u64(key),
});
}
inline int getNextMapKey(const BPF_FD_TYPE map_fd, const void* key, void* next_key) {
return bpf(BPF_MAP_GET_NEXT_KEY, {
.map_fd = BPF_FD_TO_U32(map_fd),
.key = ptr_to_u64(key),
.next_key = ptr_to_u64(next_key),
});
}
inline int getFirstMapKey(const BPF_FD_TYPE map_fd, void* firstKey) {
return getNextMapKey(map_fd, NULL, firstKey);
}
inline int bpfFdPin(const BPF_FD_TYPE map_fd, const char* pathname) {
return bpf(BPF_OBJ_PIN, {
.pathname = ptr_to_u64(pathname),
.bpf_fd = BPF_FD_TO_U32(map_fd),
});
}
inline int bpfFdGet(const char* pathname, uint32_t flag) {
return bpf(BPF_OBJ_GET, {
.pathname = ptr_to_u64(pathname),
.file_flags = flag,
});
}
int bpfGetFdMapId(const BPF_FD_TYPE map_fd);
inline int bpfLock(int fd, short type) {
if (fd < 0) return fd; // pass any errors straight through
#ifdef BPF_MAP_LOCKLESS_FOR_TEST
return fd;
#endif
#ifdef BPF_FD_JUST_USE_INT
int mapId = bpfGetFdMapId(fd);
int saved_errno = errno;
#else
base::unique_fd ufd(fd);
int mapId = bpfGetFdMapId(ufd);
int saved_errno = errno;
(void)ufd.release();
#endif
// 4.14+ required to fetch map id, but we don't want to call isAtLeastKernelVersion
if (mapId == -1 && saved_errno == EINVAL) return fd;
if (mapId <= 0) abort(); // should not be possible
// on __LP64__ (aka. 64-bit userspace) 'struct flock64' is the same as 'struct flock'
struct flock64 fl = {
.l_type = type, // short: F_{RD,WR,UN}LCK
.l_whence = SEEK_SET, // short: SEEK_{SET,CUR,END}
.l_start = mapId, // off_t: start offset
.l_len = 1, // off_t: number of bytes
};
// see: bionic/libc/bionic/fcntl.cpp: iff !__LP64__ this uses fcntl64
int ret = fcntl(fd, F_OFD_SETLK, &fl);
if (!ret) return fd; // success
close(fd);
return ret; // most likely -1 with errno == EAGAIN, due to already held lock
}
inline int mapRetrieveLocklessRW(const char* pathname) {
return bpfFdGet(pathname, 0);
}
inline int mapRetrieveExclusiveRW(const char* pathname) {
return bpfLock(mapRetrieveLocklessRW(pathname), F_WRLCK);
}
inline int mapRetrieveRW(const char* pathname) {
return bpfLock(mapRetrieveLocklessRW(pathname), F_RDLCK);
}
inline int mapRetrieveRO(const char* pathname) {
return bpfFdGet(pathname, BPF_F_RDONLY);
}
// WARNING: it's impossible to grab a shared (ie. read) lock on a write-only fd,
// so we instead choose to grab an exclusive (ie. write) lock.
inline int mapRetrieveWO(const char* pathname) {
return bpfLock(bpfFdGet(pathname, BPF_F_WRONLY), F_WRLCK);
}
inline int retrieveProgram(const char* pathname) {
return bpfFdGet(pathname, BPF_F_RDONLY);
}
inline bool usableProgram(const char* pathname) {
int fd = retrieveProgram(pathname);
bool ok = (fd >= 0);
if (ok) close(fd);
return ok;
}
inline int attachProgram(bpf_attach_type type, const BPF_FD_TYPE prog_fd,
const BPF_FD_TYPE cg_fd, uint32_t flags = 0) {
return bpf(BPF_PROG_ATTACH, {
.target_fd = BPF_FD_TO_U32(cg_fd),
.attach_bpf_fd = BPF_FD_TO_U32(prog_fd),
.attach_type = type,
.attach_flags = flags,
});
}
inline int detachProgram(bpf_attach_type type, const BPF_FD_TYPE cg_fd) {
return bpf(BPF_PROG_DETACH, {
.target_fd = BPF_FD_TO_U32(cg_fd),
.attach_type = type,
});
}
inline int queryProgram(const BPF_FD_TYPE cg_fd,
enum bpf_attach_type attach_type,
__u32 query_flags = 0,
__u32 attach_flags = 0) {
int prog_id = -1; // equivalent to an array of one integer.
bpf_attr arg = {
.query = {
.target_fd = BPF_FD_TO_U32(cg_fd),
.attach_type = attach_type,
.query_flags = query_flags,
.attach_flags = attach_flags,
.prog_ids = ptr_to_u64(&prog_id), // pointer to output array
.prog_cnt = 1, // in: space - nr of ints in the array, out: used
}
};
int v = bpf(BPF_PROG_QUERY, &arg);
if (v) return v; // error case
if (!arg.query.prog_cnt) return 0; // no program, kernel never returns zero id
return prog_id; // return actual id
}
inline int detachSingleProgram(bpf_attach_type type, const BPF_FD_TYPE prog_fd,
const BPF_FD_TYPE cg_fd) {
return bpf(BPF_PROG_DETACH, {
.target_fd = BPF_FD_TO_U32(cg_fd),
.attach_bpf_fd = BPF_FD_TO_U32(prog_fd),
.attach_type = type,
});
}
// Available in 4.12 and later kernels.
inline int runProgram(const BPF_FD_TYPE prog_fd, const void* data,
const uint32_t data_size) {
return bpf(BPF_PROG_RUN, {
.test = {
.prog_fd = BPF_FD_TO_U32(prog_fd),
.data_size_in = data_size,
.data_in = ptr_to_u64(data),
},
});
}
// BPF_OBJ_GET_INFO_BY_FD requires 4.14+ kernel
//
// Note: some fields are only defined in newer kernels (ie. the map_info struct grows
// over time), so we need to check that the field we're interested in is actually
// supported/returned by the running kernel. We do this by checking it is fully
// within the bounds of the struct size as reported by the kernel.
#define DEFINE_BPF_GET_FD(TYPE, NAME, FIELD) \
inline int bpfGetFd ## NAME(const BPF_FD_TYPE fd) { \
struct bpf_ ## TYPE ## _info info = {}; \
union bpf_attr attr = { .info = { \
.bpf_fd = BPF_FD_TO_U32(fd), \
.info_len = sizeof(info), \
.info = ptr_to_u64(&info), \
}}; \
int rv = bpf(BPF_OBJ_GET_INFO_BY_FD, attr); \
if (rv) return rv; \
if (attr.info.info_len < offsetof(bpf_ ## TYPE ## _info, FIELD) + sizeof(info.FIELD)) { \
errno = EOPNOTSUPP; \
return -1; \
}; \
return info.FIELD; \
}
// All 7 of these fields are already present in Linux v4.14 (even ACK 4.14-P)
// while BPF_OBJ_GET_INFO_BY_FD is not implemented at all in v4.9 (even ACK 4.9-Q)
DEFINE_BPF_GET_FD(map, MapType, type) // int bpfGetFdMapType(const BPF_FD_TYPE map_fd)
DEFINE_BPF_GET_FD(map, MapId, id) // int bpfGetFdMapId(const BPF_FD_TYPE map_fd)
DEFINE_BPF_GET_FD(map, KeySize, key_size) // int bpfGetFdKeySize(const BPF_FD_TYPE map_fd)
DEFINE_BPF_GET_FD(map, ValueSize, value_size) // int bpfGetFdValueSize(const BPF_FD_TYPE map_fd)
DEFINE_BPF_GET_FD(map, MaxEntries, max_entries) // int bpfGetFdMaxEntries(const BPF_FD_TYPE map_fd)
DEFINE_BPF_GET_FD(map, MapFlags, map_flags) // int bpfGetFdMapFlags(const BPF_FD_TYPE map_fd)
DEFINE_BPF_GET_FD(prog, ProgId, id) // int bpfGetFdProgId(const BPF_FD_TYPE prog_fd)
#undef DEFINE_BPF_GET_FD
} // namespace bpf
} // namespace android
#undef BPF_FD_TO_U32
#undef BPF_FD_TYPE
#undef BPF_FD_JUST_USE_INT