samples: bpf: Remove bpf_load loader completely
authorDaniel T. Lee <danieltimlee@gmail.com>
Tue, 24 Nov 2020 09:03:10 +0000 (09:03 +0000)
committerAndrii Nakryiko <andrii@kernel.org>
Fri, 27 Nov 2020 03:33:36 +0000 (19:33 -0800)
Numerous refactoring that rewrites BPF programs written with bpf_load
to use the libbpf loader was finally completed, resulting in BPF
programs using bpf_load within the kernel being completely no longer
present.

This commit removes bpf_load, an outdated bpf loader that is difficult
to keep up with the latest kernel BPF and causes confusion.

Also, this commit removes the unused trace_helper and bpf_load from
samples/bpf target objects from Makefile.

Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20201124090310.24374-8-danieltimlee@gmail.com
samples/bpf/Makefile
samples/bpf/bpf_load.c [deleted file]
samples/bpf/bpf_load.h [deleted file]
samples/bpf/xdp2skb_meta_kern.c

index 25380e0..05db041 100644 (file)
@@ -73,7 +73,7 @@ tracex5-objs := tracex5_user.o $(TRACE_HELPERS)
 tracex6-objs := tracex6_user.o
 tracex7-objs := tracex7_user.o
 test_probe_write_user-objs := test_probe_write_user_user.o
-trace_output-objs := trace_output_user.o $(TRACE_HELPERS)
+trace_output-objs := trace_output_user.o
 lathist-objs := lathist_user.o
 offwaketime-objs := offwaketime_user.o $(TRACE_HELPERS)
 spintest-objs := spintest_user.o $(TRACE_HELPERS)
@@ -91,8 +91,8 @@ test_current_task_under_cgroup-objs := $(CGROUP_HELPERS) \
                                       test_current_task_under_cgroup_user.o
 trace_event-objs := trace_event_user.o $(TRACE_HELPERS)
 sampleip-objs := sampleip_user.o $(TRACE_HELPERS)
-tc_l2_redirect-objs := bpf_load.o tc_l2_redirect_user.o
-lwt_len_hist-objs := bpf_load.o lwt_len_hist_user.o
+tc_l2_redirect-objs := tc_l2_redirect_user.o
+lwt_len_hist-objs := lwt_len_hist_user.o
 xdp_tx_iptunnel-objs := xdp_tx_iptunnel_user.o
 test_map_in_map-objs := test_map_in_map_user.o
 per_socket_stats_example-objs := cookie_uid_helper_example.o
@@ -108,7 +108,7 @@ xdpsock-objs := xdpsock_user.o
 xsk_fwd-objs := xsk_fwd.o
 xdp_fwd-objs := xdp_fwd_user.o
 task_fd_query-objs := task_fd_query_user.o $(TRACE_HELPERS)
-xdp_sample_pkts-objs := xdp_sample_pkts_user.o $(TRACE_HELPERS)
+xdp_sample_pkts-objs := xdp_sample_pkts_user.o
 ibumad-objs := ibumad_user.o
 hbm-objs := hbm.o $(CGROUP_HELPERS)
 
@@ -197,8 +197,6 @@ TPROGS_CFLAGS += --sysroot=$(SYSROOT)
 TPROGS_LDFLAGS := -L$(SYSROOT)/usr/lib
 endif
 
-TPROGCFLAGS_bpf_load.o += -Wno-unused-variable
-
 TPROGS_LDLIBS                  += $(LIBBPF) -lelf -lz
 TPROGLDLIBS_tracex4            += -lrt
 TPROGLDLIBS_trace_output       += -lrt
diff --git a/samples/bpf/bpf_load.c b/samples/bpf/bpf_load.c
deleted file mode 100644 (file)
index c5ad528..0000000
+++ /dev/null
@@ -1,667 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0
-#include <stdio.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <libelf.h>
-#include <gelf.h>
-#include <errno.h>
-#include <unistd.h>
-#include <string.h>
-#include <stdbool.h>
-#include <stdlib.h>
-#include <linux/bpf.h>
-#include <linux/filter.h>
-#include <linux/perf_event.h>
-#include <linux/netlink.h>
-#include <linux/rtnetlink.h>
-#include <linux/types.h>
-#include <sys/socket.h>
-#include <sys/syscall.h>
-#include <sys/ioctl.h>
-#include <sys/mman.h>
-#include <poll.h>
-#include <ctype.h>
-#include <assert.h>
-#include <bpf/bpf.h>
-#include "bpf_load.h"
-#include "perf-sys.h"
-
-#define DEBUGFS "/sys/kernel/debug/tracing/"
-
-static char license[128];
-static int kern_version;
-static bool processed_sec[128];
-char bpf_log_buf[BPF_LOG_BUF_SIZE];
-int map_fd[MAX_MAPS];
-int prog_fd[MAX_PROGS];
-int event_fd[MAX_PROGS];
-int prog_cnt;
-int prog_array_fd = -1;
-
-struct bpf_map_data map_data[MAX_MAPS];
-int map_data_count;
-
-static int populate_prog_array(const char *event, int prog_fd)
-{
-       int ind = atoi(event), err;
-
-       err = bpf_map_update_elem(prog_array_fd, &ind, &prog_fd, BPF_ANY);
-       if (err < 0) {
-               printf("failed to store prog_fd in prog_array\n");
-               return -1;
-       }
-       return 0;
-}
-
-static int write_kprobe_events(const char *val)
-{
-       int fd, ret, flags;
-
-       if (val == NULL)
-               return -1;
-       else if (val[0] == '\0')
-               flags = O_WRONLY | O_TRUNC;
-       else
-               flags = O_WRONLY | O_APPEND;
-
-       fd = open(DEBUGFS "kprobe_events", flags);
-
-       ret = write(fd, val, strlen(val));
-       close(fd);
-
-       return ret;
-}
-
-static int load_and_attach(const char *event, struct bpf_insn *prog, int size)
-{
-       bool is_socket = strncmp(event, "socket", 6) == 0;
-       bool is_kprobe = strncmp(event, "kprobe/", 7) == 0;
-       bool is_kretprobe = strncmp(event, "kretprobe/", 10) == 0;
-       bool is_tracepoint = strncmp(event, "tracepoint/", 11) == 0;
-       bool is_raw_tracepoint = strncmp(event, "raw_tracepoint/", 15) == 0;
-       bool is_xdp = strncmp(event, "xdp", 3) == 0;
-       bool is_perf_event = strncmp(event, "perf_event", 10) == 0;
-       bool is_cgroup_skb = strncmp(event, "cgroup/skb", 10) == 0;
-       bool is_cgroup_sk = strncmp(event, "cgroup/sock", 11) == 0;
-       bool is_sockops = strncmp(event, "sockops", 7) == 0;
-       bool is_sk_skb = strncmp(event, "sk_skb", 6) == 0;
-       bool is_sk_msg = strncmp(event, "sk_msg", 6) == 0;
-       size_t insns_cnt = size / sizeof(struct bpf_insn);
-       enum bpf_prog_type prog_type;
-       char buf[256];
-       int fd, efd, err, id;
-       struct perf_event_attr attr = {};
-
-       attr.type = PERF_TYPE_TRACEPOINT;
-       attr.sample_type = PERF_SAMPLE_RAW;
-       attr.sample_period = 1;
-       attr.wakeup_events = 1;
-
-       if (is_socket) {
-               prog_type = BPF_PROG_TYPE_SOCKET_FILTER;
-       } else if (is_kprobe || is_kretprobe) {
-               prog_type = BPF_PROG_TYPE_KPROBE;
-       } else if (is_tracepoint) {
-               prog_type = BPF_PROG_TYPE_TRACEPOINT;
-       } else if (is_raw_tracepoint) {
-               prog_type = BPF_PROG_TYPE_RAW_TRACEPOINT;
-       } else if (is_xdp) {
-               prog_type = BPF_PROG_TYPE_XDP;
-       } else if (is_perf_event) {
-               prog_type = BPF_PROG_TYPE_PERF_EVENT;
-       } else if (is_cgroup_skb) {
-               prog_type = BPF_PROG_TYPE_CGROUP_SKB;
-       } else if (is_cgroup_sk) {
-               prog_type = BPF_PROG_TYPE_CGROUP_SOCK;
-       } else if (is_sockops) {
-               prog_type = BPF_PROG_TYPE_SOCK_OPS;
-       } else if (is_sk_skb) {
-               prog_type = BPF_PROG_TYPE_SK_SKB;
-       } else if (is_sk_msg) {
-               prog_type = BPF_PROG_TYPE_SK_MSG;
-       } else {
-               printf("Unknown event '%s'\n", event);
-               return -1;
-       }
-
-       if (prog_cnt == MAX_PROGS)
-               return -1;
-
-       fd = bpf_load_program(prog_type, prog, insns_cnt, license, kern_version,
-                             bpf_log_buf, BPF_LOG_BUF_SIZE);
-       if (fd < 0) {
-               printf("bpf_load_program() err=%d\n%s", errno, bpf_log_buf);
-               return -1;
-       }
-
-       prog_fd[prog_cnt++] = fd;
-
-       if (is_xdp || is_perf_event || is_cgroup_skb || is_cgroup_sk)
-               return 0;
-
-       if (is_socket || is_sockops || is_sk_skb || is_sk_msg) {
-               if (is_socket)
-                       event += 6;
-               else
-                       event += 7;
-               if (*event != '/')
-                       return 0;
-               event++;
-               if (!isdigit(*event)) {
-                       printf("invalid prog number\n");
-                       return -1;
-               }
-               return populate_prog_array(event, fd);
-       }
-
-       if (is_raw_tracepoint) {
-               efd = bpf_raw_tracepoint_open(event + 15, fd);
-               if (efd < 0) {
-                       printf("tracepoint %s %s\n", event + 15, strerror(errno));
-                       return -1;
-               }
-               event_fd[prog_cnt - 1] = efd;
-               return 0;
-       }
-
-       if (is_kprobe || is_kretprobe) {
-               bool need_normal_check = true;
-               const char *event_prefix = "";
-
-               if (is_kprobe)
-                       event += 7;
-               else
-                       event += 10;
-
-               if (*event == 0) {
-                       printf("event name cannot be empty\n");
-                       return -1;
-               }
-
-               if (isdigit(*event))
-                       return populate_prog_array(event, fd);
-
-#ifdef __x86_64__
-               if (strncmp(event, "sys_", 4) == 0) {
-                       snprintf(buf, sizeof(buf), "%c:__x64_%s __x64_%s",
-                               is_kprobe ? 'p' : 'r', event, event);
-                       err = write_kprobe_events(buf);
-                       if (err >= 0) {
-                               need_normal_check = false;
-                               event_prefix = "__x64_";
-                       }
-               }
-#endif
-               if (need_normal_check) {
-                       snprintf(buf, sizeof(buf), "%c:%s %s",
-                               is_kprobe ? 'p' : 'r', event, event);
-                       err = write_kprobe_events(buf);
-                       if (err < 0) {
-                               printf("failed to create kprobe '%s' error '%s'\n",
-                                      event, strerror(errno));
-                               return -1;
-                       }
-               }
-
-               strcpy(buf, DEBUGFS);
-               strcat(buf, "events/kprobes/");
-               strcat(buf, event_prefix);
-               strcat(buf, event);
-               strcat(buf, "/id");
-       } else if (is_tracepoint) {
-               event += 11;
-
-               if (*event == 0) {
-                       printf("event name cannot be empty\n");
-                       return -1;
-               }
-               strcpy(buf, DEBUGFS);
-               strcat(buf, "events/");
-               strcat(buf, event);
-               strcat(buf, "/id");
-       }
-
-       efd = open(buf, O_RDONLY, 0);
-       if (efd < 0) {
-               printf("failed to open event %s\n", event);
-               return -1;
-       }
-
-       err = read(efd, buf, sizeof(buf));
-       if (err < 0 || err >= sizeof(buf)) {
-               printf("read from '%s' failed '%s'\n", event, strerror(errno));
-               return -1;
-       }
-
-       close(efd);
-
-       buf[err] = 0;
-       id = atoi(buf);
-       attr.config = id;
-
-       efd = sys_perf_event_open(&attr, -1/*pid*/, 0/*cpu*/, -1/*group_fd*/, 0);
-       if (efd < 0) {
-               printf("event %d fd %d err %s\n", id, efd, strerror(errno));
-               return -1;
-       }
-       event_fd[prog_cnt - 1] = efd;
-       err = ioctl(efd, PERF_EVENT_IOC_ENABLE, 0);
-       if (err < 0) {
-               printf("ioctl PERF_EVENT_IOC_ENABLE failed err %s\n",
-                      strerror(errno));
-               return -1;
-       }
-       err = ioctl(efd, PERF_EVENT_IOC_SET_BPF, fd);
-       if (err < 0) {
-               printf("ioctl PERF_EVENT_IOC_SET_BPF failed err %s\n",
-                      strerror(errno));
-               return -1;
-       }
-
-       return 0;
-}
-
-static int load_maps(struct bpf_map_data *maps, int nr_maps,
-                    fixup_map_cb fixup_map)
-{
-       int i, numa_node;
-
-       for (i = 0; i < nr_maps; i++) {
-               if (fixup_map) {
-                       fixup_map(&maps[i], i);
-                       /* Allow userspace to assign map FD prior to creation */
-                       if (maps[i].fd != -1) {
-                               map_fd[i] = maps[i].fd;
-                               continue;
-                       }
-               }
-
-               numa_node = maps[i].def.map_flags & BPF_F_NUMA_NODE ?
-                       maps[i].def.numa_node : -1;
-
-               if (maps[i].def.type == BPF_MAP_TYPE_ARRAY_OF_MAPS ||
-                   maps[i].def.type == BPF_MAP_TYPE_HASH_OF_MAPS) {
-                       int inner_map_fd = map_fd[maps[i].def.inner_map_idx];
-
-                       map_fd[i] = bpf_create_map_in_map_node(maps[i].def.type,
-                                                       maps[i].name,
-                                                       maps[i].def.key_size,
-                                                       inner_map_fd,
-                                                       maps[i].def.max_entries,
-                                                       maps[i].def.map_flags,
-                                                       numa_node);
-               } else {
-                       map_fd[i] = bpf_create_map_node(maps[i].def.type,
-                                                       maps[i].name,
-                                                       maps[i].def.key_size,
-                                                       maps[i].def.value_size,
-                                                       maps[i].def.max_entries,
-                                                       maps[i].def.map_flags,
-                                                       numa_node);
-               }
-               if (map_fd[i] < 0) {
-                       printf("failed to create map %d (%s): %d %s\n",
-                              i, maps[i].name, errno, strerror(errno));
-                       return 1;
-               }
-               maps[i].fd = map_fd[i];
-
-               if (maps[i].def.type == BPF_MAP_TYPE_PROG_ARRAY)
-                       prog_array_fd = map_fd[i];
-       }
-       return 0;
-}
-
-static int get_sec(Elf *elf, int i, GElf_Ehdr *ehdr, char **shname,
-                  GElf_Shdr *shdr, Elf_Data **data)
-{
-       Elf_Scn *scn;
-
-       scn = elf_getscn(elf, i);
-       if (!scn)
-               return 1;
-
-       if (gelf_getshdr(scn, shdr) != shdr)
-               return 2;
-
-       *shname = elf_strptr(elf, ehdr->e_shstrndx, shdr->sh_name);
-       if (!*shname || !shdr->sh_size)
-               return 3;
-
-       *data = elf_getdata(scn, 0);
-       if (!*data || elf_getdata(scn, *data) != NULL)
-               return 4;
-
-       return 0;
-}
-
-static int parse_relo_and_apply(Elf_Data *data, Elf_Data *symbols,
-                               GElf_Shdr *shdr, struct bpf_insn *insn,
-                               struct bpf_map_data *maps, int nr_maps)
-{
-       int i, nrels;
-
-       nrels = shdr->sh_size / shdr->sh_entsize;
-
-       for (i = 0; i < nrels; i++) {
-               GElf_Sym sym;
-               GElf_Rel rel;
-               unsigned int insn_idx;
-               bool match = false;
-               int j, map_idx;
-
-               gelf_getrel(data, i, &rel);
-
-               insn_idx = rel.r_offset / sizeof(struct bpf_insn);
-
-               gelf_getsym(symbols, GELF_R_SYM(rel.r_info), &sym);
-
-               if (insn[insn_idx].code != (BPF_LD | BPF_IMM | BPF_DW)) {
-                       printf("invalid relo for insn[%d].code 0x%x\n",
-                              insn_idx, insn[insn_idx].code);
-                       return 1;
-               }
-               insn[insn_idx].src_reg = BPF_PSEUDO_MAP_FD;
-
-               /* Match FD relocation against recorded map_data[] offset */
-               for (map_idx = 0; map_idx < nr_maps; map_idx++) {
-                       if (maps[map_idx].elf_offset == sym.st_value) {
-                               match = true;
-                               break;
-                       }
-               }
-               if (match) {
-                       insn[insn_idx].imm = maps[map_idx].fd;
-               } else {
-                       printf("invalid relo for insn[%d] no map_data match\n",
-                              insn_idx);
-                       return 1;
-               }
-       }
-
-       return 0;
-}
-
-static int cmp_symbols(const void *l, const void *r)
-{
-       const GElf_Sym *lsym = (const GElf_Sym *)l;
-       const GElf_Sym *rsym = (const GElf_Sym *)r;
-
-       if (lsym->st_value < rsym->st_value)
-               return -1;
-       else if (lsym->st_value > rsym->st_value)
-               return 1;
-       else
-               return 0;
-}
-
-static int load_elf_maps_section(struct bpf_map_data *maps, int maps_shndx,
-                                Elf *elf, Elf_Data *symbols, int strtabidx)
-{
-       int map_sz_elf, map_sz_copy;
-       bool validate_zero = false;
-       Elf_Data *data_maps;
-       int i, nr_maps;
-       GElf_Sym *sym;
-       Elf_Scn *scn;
-       int copy_sz;
-
-       if (maps_shndx < 0)
-               return -EINVAL;
-       if (!symbols)
-               return -EINVAL;
-
-       /* Get data for maps section via elf index */
-       scn = elf_getscn(elf, maps_shndx);
-       if (scn)
-               data_maps = elf_getdata(scn, NULL);
-       if (!scn || !data_maps) {
-               printf("Failed to get Elf_Data from maps section %d\n",
-                      maps_shndx);
-               return -EINVAL;
-       }
-
-       /* For each map get corrosponding symbol table entry */
-       sym = calloc(MAX_MAPS+1, sizeof(GElf_Sym));
-       for (i = 0, nr_maps = 0; i < symbols->d_size / sizeof(GElf_Sym); i++) {
-               assert(nr_maps < MAX_MAPS+1);
-               if (!gelf_getsym(symbols, i, &sym[nr_maps]))
-                       continue;
-               if (sym[nr_maps].st_shndx != maps_shndx)
-                       continue;
-               /* Only increment iif maps section */
-               nr_maps++;
-       }
-
-       /* Align to map_fd[] order, via sort on offset in sym.st_value */
-       qsort(sym, nr_maps, sizeof(GElf_Sym), cmp_symbols);
-
-       /* Keeping compatible with ELF maps section changes
-        * ------------------------------------------------
-        * The program size of struct bpf_load_map_def is known by loader
-        * code, but struct stored in ELF file can be different.
-        *
-        * Unfortunately sym[i].st_size is zero.  To calculate the
-        * struct size stored in the ELF file, assume all struct have
-        * the same size, and simply divide with number of map
-        * symbols.
-        */
-       map_sz_elf = data_maps->d_size / nr_maps;
-       map_sz_copy = sizeof(struct bpf_load_map_def);
-       if (map_sz_elf < map_sz_copy) {
-               /*
-                * Backward compat, loading older ELF file with
-                * smaller struct, keeping remaining bytes zero.
-                */
-               map_sz_copy = map_sz_elf;
-       } else if (map_sz_elf > map_sz_copy) {
-               /*
-                * Forward compat, loading newer ELF file with larger
-                * struct with unknown features. Assume zero means
-                * feature not used.  Thus, validate rest of struct
-                * data is zero.
-                */
-               validate_zero = true;
-       }
-
-       /* Memcpy relevant part of ELF maps data to loader maps */
-       for (i = 0; i < nr_maps; i++) {
-               struct bpf_load_map_def *def;
-               unsigned char *addr, *end;
-               const char *map_name;
-               size_t offset;
-
-               map_name = elf_strptr(elf, strtabidx, sym[i].st_name);
-               maps[i].name = strdup(map_name);
-               if (!maps[i].name) {
-                       printf("strdup(%s): %s(%d)\n", map_name,
-                              strerror(errno), errno);
-                       free(sym);
-                       return -errno;
-               }
-
-               /* Symbol value is offset into ELF maps section data area */
-               offset = sym[i].st_value;
-               def = (struct bpf_load_map_def *)(data_maps->d_buf + offset);
-               maps[i].elf_offset = offset;
-               memset(&maps[i].def, 0, sizeof(struct bpf_load_map_def));
-               memcpy(&maps[i].def, def, map_sz_copy);
-
-               /* Verify no newer features were requested */
-               if (validate_zero) {
-                       addr = (unsigned char *) def + map_sz_copy;
-                       end  = (unsigned char *) def + map_sz_elf;
-                       for (; addr < end; addr++) {
-                               if (*addr != 0) {
-                                       free(sym);
-                                       return -EFBIG;
-                               }
-                       }
-               }
-       }
-
-       free(sym);
-       return nr_maps;
-}
-
-static int do_load_bpf_file(const char *path, fixup_map_cb fixup_map)
-{
-       int fd, i, ret, maps_shndx = -1, strtabidx = -1;
-       Elf *elf;
-       GElf_Ehdr ehdr;
-       GElf_Shdr shdr, shdr_prog;
-       Elf_Data *data, *data_prog, *data_maps = NULL, *symbols = NULL;
-       char *shname, *shname_prog;
-       int nr_maps = 0;
-
-       /* reset global variables */
-       kern_version = 0;
-       memset(license, 0, sizeof(license));
-       memset(processed_sec, 0, sizeof(processed_sec));
-
-       if (elf_version(EV_CURRENT) == EV_NONE)
-               return 1;
-
-       fd = open(path, O_RDONLY, 0);
-       if (fd < 0)
-               return 1;
-
-       elf = elf_begin(fd, ELF_C_READ, NULL);
-
-       if (!elf)
-               return 1;
-
-       if (gelf_getehdr(elf, &ehdr) != &ehdr)
-               return 1;
-
-       /* clear all kprobes */
-       i = write_kprobe_events("");
-
-       /* scan over all elf sections to get license and map info */
-       for (i = 1; i < ehdr.e_shnum; i++) {
-
-               if (get_sec(elf, i, &ehdr, &shname, &shdr, &data))
-                       continue;
-
-               if (0) /* helpful for llvm debugging */
-                       printf("section %d:%s data %p size %zd link %d flags %d\n",
-                              i, shname, data->d_buf, data->d_size,
-                              shdr.sh_link, (int) shdr.sh_flags);
-
-               if (strcmp(shname, "license") == 0) {
-                       processed_sec[i] = true;
-                       memcpy(license, data->d_buf, data->d_size);
-               } else if (strcmp(shname, "version") == 0) {
-                       processed_sec[i] = true;
-                       if (data->d_size != sizeof(int)) {
-                               printf("invalid size of version section %zd\n",
-                                      data->d_size);
-                               return 1;
-                       }
-                       memcpy(&kern_version, data->d_buf, sizeof(int));
-               } else if (strcmp(shname, "maps") == 0) {
-                       int j;
-
-                       maps_shndx = i;
-                       data_maps = data;
-                       for (j = 0; j < MAX_MAPS; j++)
-                               map_data[j].fd = -1;
-               } else if (shdr.sh_type == SHT_SYMTAB) {
-                       strtabidx = shdr.sh_link;
-                       symbols = data;
-               }
-       }
-
-       ret = 1;
-
-       if (!symbols) {
-               printf("missing SHT_SYMTAB section\n");
-               goto done;
-       }
-
-       if (data_maps) {
-               nr_maps = load_elf_maps_section(map_data, maps_shndx,
-                                               elf, symbols, strtabidx);
-               if (nr_maps < 0) {
-                       printf("Error: Failed loading ELF maps (errno:%d):%s\n",
-                              nr_maps, strerror(-nr_maps));
-                       goto done;
-               }
-               if (load_maps(map_data, nr_maps, fixup_map))
-                       goto done;
-               map_data_count = nr_maps;
-
-               processed_sec[maps_shndx] = true;
-       }
-
-       /* process all relo sections, and rewrite bpf insns for maps */
-       for (i = 1; i < ehdr.e_shnum; i++) {
-               if (processed_sec[i])
-                       continue;
-
-               if (get_sec(elf, i, &ehdr, &shname, &shdr, &data))
-                       continue;
-
-               if (shdr.sh_type == SHT_REL) {
-                       struct bpf_insn *insns;
-
-                       /* locate prog sec that need map fixup (relocations) */
-                       if (get_sec(elf, shdr.sh_info, &ehdr, &shname_prog,
-                                   &shdr_prog, &data_prog))
-                               continue;
-
-                       if (shdr_prog.sh_type != SHT_PROGBITS ||
-                           !(shdr_prog.sh_flags & SHF_EXECINSTR))
-                               continue;
-
-                       insns = (struct bpf_insn *) data_prog->d_buf;
-                       processed_sec[i] = true; /* relo section */
-
-                       if (parse_relo_and_apply(data, symbols, &shdr, insns,
-                                                map_data, nr_maps))
-                               continue;
-               }
-       }
-
-       /* load programs */
-       for (i = 1; i < ehdr.e_shnum; i++) {
-
-               if (processed_sec[i])
-                       continue;
-
-               if (get_sec(elf, i, &ehdr, &shname, &shdr, &data))
-                       continue;
-
-               if (memcmp(shname, "kprobe/", 7) == 0 ||
-                   memcmp(shname, "kretprobe/", 10) == 0 ||
-                   memcmp(shname, "tracepoint/", 11) == 0 ||
-                   memcmp(shname, "raw_tracepoint/", 15) == 0 ||
-                   memcmp(shname, "xdp", 3) == 0 ||
-                   memcmp(shname, "perf_event", 10) == 0 ||
-                   memcmp(shname, "socket", 6) == 0 ||
-                   memcmp(shname, "cgroup/", 7) == 0 ||
-                   memcmp(shname, "sockops", 7) == 0 ||
-                   memcmp(shname, "sk_skb", 6) == 0 ||
-                   memcmp(shname, "sk_msg", 6) == 0) {
-                       ret = load_and_attach(shname, data->d_buf,
-                                             data->d_size);
-                       if (ret != 0)
-                               goto done;
-               }
-       }
-
-done:
-       close(fd);
-       return ret;
-}
-
-int load_bpf_file(char *path)
-{
-       return do_load_bpf_file(path, NULL);
-}
-
-int load_bpf_file_fixup_map(const char *path, fixup_map_cb fixup_map)
-{
-       return do_load_bpf_file(path, fixup_map);
-}
diff --git a/samples/bpf/bpf_load.h b/samples/bpf/bpf_load.h
deleted file mode 100644 (file)
index 4fcd258..0000000
+++ /dev/null
@@ -1,57 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#ifndef __BPF_LOAD_H
-#define __BPF_LOAD_H
-
-#include <bpf/bpf.h>
-
-#define MAX_MAPS 32
-#define MAX_PROGS 32
-
-struct bpf_load_map_def {
-       unsigned int type;
-       unsigned int key_size;
-       unsigned int value_size;
-       unsigned int max_entries;
-       unsigned int map_flags;
-       unsigned int inner_map_idx;
-       unsigned int numa_node;
-};
-
-struct bpf_map_data {
-       int fd;
-       char *name;
-       size_t elf_offset;
-       struct bpf_load_map_def def;
-};
-
-typedef void (*fixup_map_cb)(struct bpf_map_data *map, int idx);
-
-extern int prog_fd[MAX_PROGS];
-extern int event_fd[MAX_PROGS];
-extern char bpf_log_buf[BPF_LOG_BUF_SIZE];
-extern int prog_cnt;
-
-/* There is a one-to-one mapping between map_fd[] and map_data[].
- * The map_data[] just contains more rich info on the given map.
- */
-extern int map_fd[MAX_MAPS];
-extern struct bpf_map_data map_data[MAX_MAPS];
-extern int map_data_count;
-
-/* parses elf file compiled by llvm .c->.o
- * . parses 'maps' section and creates maps via BPF syscall
- * . parses 'license' section and passes it to syscall
- * . parses elf relocations for BPF maps and adjusts BPF_LD_IMM64 insns by
- *   storing map_fd into insn->imm and marking such insns as BPF_PSEUDO_MAP_FD
- * . loads eBPF programs via BPF syscall
- *
- * One ELF file can contain multiple BPF programs which will be loaded
- * and their FDs stored stored in prog_fd array
- *
- * returns zero on success
- */
-int load_bpf_file(char *path);
-int load_bpf_file_fixup_map(const char *path, fixup_map_cb fixup_map);
-
-int bpf_set_link_xdp_fd(int ifindex, int fd, __u32 flags);
-#endif
index 9b78331..d563101 100644 (file)
@@ -6,7 +6,7 @@
  * This uses the XDP data_meta infrastructure, and is a cooperation
  * between two bpf-programs (1) XDP and (2) clsact at TC-ingress hook.
  *
- * Notice: This example does not use the BPF C-loader (bpf_load.c),
+ * Notice: This example does not use the BPF C-loader,
  * but instead rely on the iproute2 TC tool for loading BPF-objects.
  */
 #include <uapi/linux/bpf.h>