Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
From:   "Eric W. Biederman" <ebiederm@xmission.com>
To:     Solar Designer <solar@openwall.com>
Cc:     Michal =?utf-8?Q?Koutn=C3=BD?= <mkoutny@suse.com>,
        Alexey Gladkov <legion@kernel.org>,
        Kees Cook <keescook@chromium.org>,
        Shuah Khan <shuah@kernel.org>,
        Christian Brauner <brauner@kernel.org>,
        Ran Xiaokai <ran.xiaokai@zte.com.cn>,
        linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org,
        Linux Containers <containers@lists.linux-foundation.org>
References: <20220207121800.5079-1-mkoutny@suse.com>
        <20220207121800.5079-2-mkoutny@suse.com>
        <20220210011405.GA17076@openwall.com>
Date:   Wed, 09 Feb 2022 19:57:10 -0600
In-Reply-To: <20220210011405.GA17076@openwall.com> (Solar Designer's message
        of "Thu, 10 Feb 2022 02:14:05 +0100")
Message-ID: <87v8xn8ow9.fsf@email.froward.int.ebiederm.org>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8BIT
Subject: Re: [RFC PATCH 1/6] set_user: Perform RLIMIT_NPROC capability check
 against new user credentials
Precedence: bulk

Solar Designer <solar@openwall.com> writes:

> Hi Michal,
>
> On Mon, Feb 07, 2022 at 01:17:55PM +0100, Michal Koutný wrote:
>> The check is currently against the current->cred but since those are
>> going to change and we want to check RLIMIT_NPROC condition after the
>> switch, supply the capability check with the new cred.
>> But since we're checking new_user being INIT_USER any new cred's
>> capability-based allowance may be redundant when the check fails and the
>> alternative solution would be revert of the commit 2863643fb8b9
>> ("set_user: add capability check when rlimit(RLIMIT_NPROC) exceeds")
>> 
>> Fixes: 2863643fb8b9 ("set_user: add capability check when rlimit(RLIMIT_NPROC) exceeds")
>> 
>> Cc: Solar Designer <solar@openwall.com>
>> Cc: Christian Brauner <christian.brauner@ubuntu.com>
>> Signed-off-by: Michal Koutný <mkoutny@suse.com>
>> ---
>>  kernel/sys.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>> 
>> diff --git a/kernel/sys.c b/kernel/sys.c
>> index 8ea20912103a..48c90dcceff3 100644
>> --- a/kernel/sys.c
>> +++ b/kernel/sys.c
>> @@ -481,7 +481,8 @@ static int set_user(struct cred *new)
>>  	 */
>>  	if (ucounts_limit_cmp(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) >= 0 &&
>>  			new_user != INIT_USER &&
>> -			!capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
>> +			!security_capable(new, &init_user_ns, CAP_SYS_RESOURCE, CAP_OPT_NONE) &&
>> +			!security_capable(new, &init_user_ns, CAP_SYS_ADMIN, CAP_OPT_NONE))
>>  		current->flags |= PF_NPROC_EXCEEDED;
>>  	else
>>  		current->flags &= ~PF_NPROC_EXCEEDED;
>
> Thank you for working on this and CC'ing me on it.  This is related to
> the discussion Christian and I had in September:
>
> https://lore.kernel.org/all/20210913100140.bxqlg47pushoqa3r@wittgenstein/
>
> Christian was going to revert 2863643fb8b9, but apparently that never
> happened.  Back then, I also suggested:
>
> "Alternatively, we could postpone the set_user() calls until we're
> running with the new user's capabilities, but that's an invasive change
> that's likely to create its own issues."

I really think we need to do something like that.  Probably just set a
flag in commit_creds and test later.

I was working on fixes that looked cleaner and I just recently realized
that the test in fork is almost as bad.  The function has_capability can
be used but the same kind of problems exist.

I thought I was very quickly going to have patches to post but I need
to redo everything now that I have noticed the issue in fork, so it will
be a day or so.

Eric


> The change you propose above is similar to that, but is more limited and
> non-invasive.  That looks good to me.
>
> However, I think you need to drop the negations of the return value from
> security_capable().  security_capable() returns 0 or -EPERM, while
> capable() returns a bool, in kernel/capability.c: ns_capable_common():
>
> 	capable = security_capable(current_cred(), ns, cap, opts);
> 	if (capable == 0) {
> 		current->flags |= PF_SUPERPRIV;
> 		return true;
> 	}
> 	return false;
>
> Also, your change would result in this no longer setting PF_SUPERPRIV.
> This may be fine, but you could want to document it.
>
> On a related note, this comment in security/commoncap.c needs an update:
>
>  * NOTE WELL: cap_has_capability() cannot be used like the kernel's capable()
>  * and has_capability() functions.  That is, it has the reverse semantics:
>  * cap_has_capability() returns 0 when a task has a capability, but the
>  * kernel's capable() and has_capability() returns 1 for this case.
>
> cap_has_capability() doesn't actually exist, and perhaps the comment
> should refer to cap_capable().
>
> Alexander