netdev - Re: [PATCH bpf-next v2] bpf: reject kfunc calls that overflow insn->imm

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 9 Feb 2022 14:20:30 +0800
From:   Hou Tao <houtao1@...wei.com>
To:     Yonghong Song <yhs@...com>, Alexei Starovoitov <ast@...nel.org>
CC:     Martin KaFai Lau <kafai@...com>,
        Daniel Borkmann <daniel@...earbox.net>,
        Andrii Nakryiko <andrii@...nel.org>,
        Song Liu <songliubraving@...com>,
        John Fastabend <john.fastabend@...il.com>,
        <netdev@...r.kernel.org>, <bpf@...r.kernel.org>
Subject: Re: [PATCH bpf-next v2] bpf: reject kfunc calls that overflow
 insn->imm

Hi,

On 2/9/2022 12:57 AM, Yonghong Song wrote:
>
>
> On 2/8/22 4:33 AM, Hou Tao wrote:
>> Now kfunc call uses s32 to represent the offset between the address
>> of kfunc and __bpf_call_base, but it doesn't check whether or not
>> s32 will be overflowed, so add an extra checking to reject these
>> invalid kfunc calls.
>>
>> Signed-off-by: Hou Tao <houtao1@...wei.com>
>> ---
>> v2:
>>   * instead of checking the overflow in selftests, just reject
>>     these kfunc calls directly in verifier
>>
>> v1: https://lore.kernel.org/bpf/20220206043107.18549-1-houtao1@huawei.com
>> ---
>>   kernel/bpf/verifier.c | 13 +++++++++++++
>>   1 file changed, 13 insertions(+)
>>
>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index a39eedecc93a..fd836e64b701 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
>> @@ -1832,6 +1832,13 @@ static struct btf *find_kfunc_desc_btf(struct
>> bpf_verifier_env *env,
>>       return btf_vmlinux ?: ERR_PTR(-ENOENT);
>>   }
>>   +static inline bool is_kfunc_call_imm_overflowed(unsigned long addr)
>> +{
>> +    unsigned long offset = BPF_CALL_IMM(addr);
>> +
>> +    return (unsigned long)(s32)offset != offset;
>> +}
>> +
>>   static int add_kfunc_call(struct bpf_verifier_env *env, u32 func_id, s16
>> offset)
>>   {
>>       const struct btf_type *func, *func_proto;
>> @@ -1925,6 +1932,12 @@ static int add_kfunc_call(struct bpf_verifier_env
>> *env, u32 func_id, s16 offset)
>>           return -EINVAL;
>>       }
>>   +    if (is_kfunc_call_imm_overflowed(addr)) {
>> +        verbose(env, "address of kernel function %s is out of range\n",
>> +            func_name);
>> +        return -EINVAL;
>> +    }
>> +
>>       desc = &tab->descs[tab->nr_descs++];
>>       desc->func_id = func_id;
>>       desc->imm = BPF_CALL_IMM(addr);
>
> Thanks, I would like to call BPF_CALL_IMM only once and keep checking overflow
> and setting desc->imm close to each other. How about the following
> not-compile-tested code
>
>     unsigned long call_imm;
>
>     ...
>     call_imm = BPF_CALL_IMM(addr);
>     /* some comment here */
>     if ((unsigned long)(s32)call_imm != call_imm) {
>         verbose(env, ...);
>         return -EINVAL;
>     } else {
>         desc->imm = call_imm;
>     }
call BPF_CALL_IMM once is OK for me. but I don't think the else branch is
unnecessary and it make the code
ugly. Can we just return directly when found that imm is overflowed ?

        call_imm = BPF_CALL_IMM(addr);
        /* Check whether or not the relative offset overflows desc->imm */
        if ((unsigned long)(s32)call_imm != call_imm) {
                verbose(env, "address of kernel function %s is out of range\n",
                        func_name);
                return -EINVAL;
        }

        desc = &tab->descs[tab->nr_descs++];
        desc->func_id = func_id;
        desc->imm = call_imm;




> .