libbpf: Fix BTF-defined map-in-map initialization on 32-bit host arches

Libbpf built in 32-bit mode should be careful about not conflating 64-bit BPF
pointers in BPF ELF file and host architecture pointers. This patch fixes
issue of incorrect initializating of map-in-map inner map slots due to such
difference.

Fixes: 646f02ffdd49 ("libbpf: Add BTF-defined map-in-map support")
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200813204945.1020225-4-andriin@fb.com

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 0d48c18..6accdda 100644 (file)
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -5195,7 +5195,8 @@ static int bpf_object__collect_st_ops_relos(struct bpf_object *obj,
 static int bpf_object__collect_map_relos(struct bpf_object *obj,
                                         GElf_Shdr *shdr, Elf_Data *data)
 {
-       int i, j, nrels, new_sz, ptr_sz = sizeof(void *);
+       const int bpf_ptr_sz = 8, host_ptr_sz = sizeof(void *);
+       int i, j, nrels, new_sz;
        const struct btf_var_secinfo *vi = NULL;
        const struct btf_type *sec, *var, *def;
        const struct btf_member *member;
@@ -5244,7 +5245,7 @@ static int bpf_object__collect_map_relos(struct bpf_object *obj,
 
                        vi = btf_var_secinfos(sec) + map->btf_var_idx;
                        if (vi->offset <= rel.r_offset &&
-                           rel.r_offset + sizeof(void *) <= vi->offset + vi->size)
+                           rel.r_offset + bpf_ptr_sz <= vi->offset + vi->size)
                                break;
                }
                if (j == obj->nr_maps) {
@@ -5280,17 +5281,20 @@ static int bpf_object__collect_map_relos(struct bpf_object *obj,
                        return -EINVAL;
 
                moff = rel.r_offset - vi->offset - moff;
-               if (moff % ptr_sz)
+               /* here we use BPF pointer size, which is always 64 bit, as we
+                * are parsing ELF that was built for BPF target
+                */
+               if (moff % bpf_ptr_sz)
                        return -EINVAL;
-               moff /= ptr_sz;
+               moff /= bpf_ptr_sz;
                if (moff >= map->init_slots_sz) {
                        new_sz = moff + 1;
-                       tmp = realloc(map->init_slots, new_sz * ptr_sz);
+                       tmp = realloc(map->init_slots, new_sz * host_ptr_sz);
                        if (!tmp)
                                return -ENOMEM;
                        map->init_slots = tmp;
                        memset(map->init_slots + map->init_slots_sz, 0,
-                              (new_sz - map->init_slots_sz) * ptr_sz);
+                              (new_sz - map->init_slots_sz) * host_ptr_sz);
                        map->init_slots_sz = new_sz;
                }
                map->init_slots[moff] = targ_map;