On 7/14/20 8:09 PM, Peilin Ye wrote:
Prevent __btf_resolve_helper_id() from dereferencing `btf_vmlinux`
as NULL. This patch fixes the following syzbot bug:
https://syzkaller.appspot.com/bug?id=f823224ada908fa5c207902a5a62065e53ca0fcc
Reported-by: syzbot+ee09bda7017345f1fbe6@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Peilin Ye <yepeilin.cs@xxxxxxxxx>
Looks good, applied, thanks! As far as I can tell all the other occurrences are
gated behind btf_parse_vmlinux() where we also init struct_opts, etc.
So for bpf-next this would then end up looking like ...
int btf_resolve_helper_id(struct bpf_verifier_log *log,
const struct bpf_func_proto *fn, int arg)
{
int id;
if (fn->arg_type[arg] != ARG_PTR_TO_BTF_ID)
return -EINVAL;
id = fn->btf_id[arg];
if (!id || !btf_vmlinux || id > btf_vmlinux->nr_types)
return -EINVAL;
return id;
}
---
Sorry, I got the link wrong. Thank you for pointing that out.
Change in v3:
- Fix incorrect syzbot dashboard link.
Change in v2:
- Split NULL and IS_ERR cases.
kernel/bpf/btf.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 30721f2c2d10..092116a311f4 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -4088,6 +4088,11 @@ static int __btf_resolve_helper_id(struct bpf_verifier_log *log, void *fn,
const char *tname, *sym;
u32 btf_id, i;
+ if (!btf_vmlinux) {
+ bpf_log(log, "btf_vmlinux doesn't exist\n");
+ return -EINVAL;
+ }
+
if (IS_ERR(btf_vmlinux)) {
bpf_log(log, "btf_vmlinux is malformed\n");
return -EINVAL;