netdev - Re: [PATCH v5 bpf-next 2/3] bpf: implement CAP

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: <294922f0-2dab-8392-492a-ce0e04c03cee@iogearbox.net>
Date:   Tue, 12 May 2020 22:09:49 +0200
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc:     davem@...emloft.net, netdev@...r.kernel.org, bpf@...r.kernel.org,
        kernel-team@...com, linux-security-module@...r.kernel.org,
        acme@...hat.com, jamorris@...ux.microsoft.com, jannh@...gle.com,
        kpsingh@...gle.com
Subject: Re: [PATCH v5 bpf-next 2/3] bpf: implement CAP_BPF

On 5/12/20 8:29 PM, Alexei Starovoitov wrote:
> On Tue, May 12, 2020 at 05:05:12PM +0200, Daniel Borkmann wrote:
>>> -	env->allow_ptr_leaks = is_priv;
>>> +	env->allow_ptr_leaks = perfmon_capable();
>>> +	env->bpf_capable = bpf_capable();
>>
>> Probably more of a detail, but it feels weird to tie perfmon_capable() into the BPF
>> core and use it in various places there. I would rather make this a proper bpf_*
>> prefixed helper and add a more descriptive name (what does it have to do with perf
>> or monitoring directly?). For example, all the main functionality could be under
>> `bpf_base_capable()` and everything with potential to leak pointers or mem to user
>> space as `bpf_leak_capable()`. Then inside include/linux/capability.h this can still
>> resolve under the hood to something like:
>>
>> static inline bool bpf_base_capable(void)
>> {
>> 	return capable(CAP_BPF) || capable(CAP_SYS_ADMIN);
>> }
> 
> I don't like the 'base' in the name, since 'base' implies common subset,
> but it's not the case. Also 'base' implies that something else is additive,
> but it's not the case either. The real base is unpriv. cap_bpf adds to it.
> So bpf_capable() in capability.h is the most appropriate.
> It also matches perfmon_capable() and other *_capable()

That's okay with me, naming is usually hardest. :)

>> static inline bool bpf_leak_capable(void)
>> {
>> 	return perfmon_capable();
>> }
> 
> This is ok, but not in capability.h. I can put it into bpf_verifier.h

Makes sense.