netdev - Re: [RFC bpf-next 2/4] bpf: return EOPNOTSUPP when map lookup isn't supported

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]

Message-ID: <ef15f2b1-ae87-56a1-54b1-29ea27637d5b@lab.ntt.co.jp>
Date:   Thu, 20 Sep 2018 11:34:30 +0900
From:   Prashant Bhole <bhole_prashant_q7@....ntt.co.jp>
To:     Mauricio Vasquez <mauricio.vasquez@...ito.it>,
        Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc:     Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jakub Kicinski <jakub.kicinski@...ronome.com>,
        Quentin Monnet <quentin.monnet@...ronome.com>,
        "David S . Miller" <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: [RFC bpf-next 2/4] bpf: return EOPNOTSUPP when map lookup isn't
 supported



On 9/20/2018 3:40 AM, Mauricio Vasquez wrote:
> 
> 
> On 09/19/2018 10:14 AM, Alexei Starovoitov wrote:
>> On Wed, Sep 19, 2018 at 04:51:41PM +0900, Prashant Bhole wrote:
>>> Return ERR_PTR(-EOPNOTSUPP) from map_lookup_elem() methods of below
>>> map types:
>>> - BPF_MAP_TYPE_PROG_ARRAY
>>> - BPF_MAP_TYPE_STACK_TRACE
>>> - BPF_MAP_TYPE_XSKMAP
>>> - BPF_MAP_TYPE_SOCKMAP/BPF_MAP_TYPE_SOCKHASH
>>>
>>> Signed-off-by: Prashant Bhole <bhole_prashant_q7@....ntt.co.jp>
>>> ---
>>>   kernel/bpf/arraymap.c | 2 +-
>>>   kernel/bpf/sockmap.c  | 2 +-
>>>   kernel/bpf/stackmap.c | 2 +-
>>>   kernel/bpf/xskmap.c   | 2 +-
>>>   4 files changed, 4 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c
>>> index dded84cbe814..24583da9ffd1 100644
>>> --- a/kernel/bpf/arraymap.c
>>> +++ b/kernel/bpf/arraymap.c
>>> @@ -449,7 +449,7 @@ static void fd_array_map_free(struct bpf_map *map)
>>>   static void *fd_array_map_lookup_elem(struct bpf_map *map, void *key)
>>>   {
>>> -    return NULL;
>>> +    return ERR_PTR(-EOPNOTSUPP);
>>>   }
>> conceptually the set looks good to me.
>> Please add a test to test_verifier.c to make sure
>> that these lookup helpers cannot be called from BPF program.
>> Otherwise this diff may cause crashes.
>>
>>
> I think we can remove all these stub functions that return EOPNOTSUPP or 
> EINVAL and return the error in syscall.c if ops->map_[update, delete, 
> lookup, ...] is null.
> This will require to extend (and test) the verifier to guarantee that 
> those helpers are not called in wrong maps, for example 
> map_delete_element in array like maps.
> 
> Would it make sense?

Thanks for review and suggestion.

I had thought about this way too (except adding restrictions in the 
verifier). There is no strong reason for choosing current implementation.

I thought there must be some reason that those methods are implemented 
and just returning NULL. Also there are no NULL checks for 
map_lookup_elem stub. So I decided to simply change the return value.
If some more people agree with removing stub function idea, I will do it.

-Prashant