[PATCH] LSM: general protection fault in legacy_parse_param

[Casey Schaufler]

On 1/25/2022 2:18 PM, Paul Moore wrote:
> On Tue, Oct 12, 2021 at 10:27 AM Casey Schaufler <casey at schaufler-ca.com> wrote:
>> On 10/12/2021 3:32 AM, Christian Brauner wrote:
>>> On Mon, Oct 11, 2021 at 03:40:22PM -0700, Casey Schaufler wrote:
>>>> The usual LSM hook "bail on fail" scheme doesn't work for cases where
>>>> a security module may return an error code indicating that it does not
>>>> recognize an input.  In this particular case Smack sees a mount option
>>>> that it recognizes, and returns 0. A call to a BPF hook follows, which
>>>> returns -ENOPARAM, which confuses the caller because Smack has processed
>>>> its data.
>>>>
>>>> Reported-by: syzbot+d1e3b1d92d25abf97943 at syzkaller.appspotmail.com
>>>> Signed-off-by: Casey Schaufler <casey at schaufler-ca.com>
>>>> ---
>>> Thanks!
>>> Note, I think that we still have the SELinux issue we discussed in the
>>> other thread:
>>>
>>>        rc = selinux_add_opt(opt, param->string, &fc->security);
>>>        if (!rc) {
>>>                param->string = NULL;
>>>                rc = 1;
>>>        }
>>>
>>> SELinux returns 1 not the expected 0. Not sure if that got fixed or is
>>> queued-up for -next. In any case, this here seems correct independent of
>>> that:
>> The aforementioned SELinux change depends on this patch. As the SELinux
>> code is today it blocks the problem seen with Smack, but introduces a
>> different issue. It prevents the BPF hook from being called.
>>
>> So the question becomes whether the SELinux change should be included
>> here, or done separately. Without the security_fs_context_parse_param()
>> change the selinux_fs_context_parse_param() change results in messy
>> failures for SELinux mounts.
> FWIW, this patch looks good to me, so:
>
> Acked-by: Paul Moore <paul at paul-moore.com>
>
> ... and with respect to the SELinux hook implementation returning 1 on
> success, I don't have a good answer and looking through my inbox I see
> David Howells hasn't responded either.  I see nothing in the original
> commit explaining why, so I'm going to say let's just change it to
> zero and be done with it; the good news is that if we do it now we've
> got almost a full cycle in linux-next to see what falls apart.  As far
> as the question of one vs two patches, it might be good to put both
> changes into a single patch just so that folks who do backports don't
> accidentally skip one and create a bad kernel build.  Casey, did you
> want to respin this patch or would you prefer me to submit another
> version?

I can create a single patch. I tried the combination on Fedora
and it worked just fine. I'll rebase and resend.

>
>>> Acked-by: Christian Brauner <christian.brauner at ubuntu.com>
>>>
>>>>   security/security.c | 14 +++++++++++++-
>>>>   1 file changed, 13 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/security/security.c b/security/security.c
>>>> index 09533cbb7221..3cf0faaf1c5b 100644
>>>> --- a/security/security.c
>>>> +++ b/security/security.c
>>>> @@ -885,7 +885,19 @@ int security_fs_context_dup(struct fs_context *fc, struct fs_context *src_fc)
>>>>
>>>>   int security_fs_context_parse_param(struct fs_context *fc, struct fs_parameter *param)
>>>>   {
>>>> -    return call_int_hook(fs_context_parse_param, -ENOPARAM, fc, param);
>>>> +    struct security_hook_list *hp;
>>>> +    int trc;
>>>> +    int rc = -ENOPARAM;
>>>> +
>>>> +    hlist_for_each_entry(hp, &security_hook_heads.fs_context_parse_param,
>>>> +                         list) {
>>>> +            trc = hp->hook.fs_context_parse_param(fc, param);
>>>> +            if (trc == 0)
>>>> +                    rc = 0;
>>>> +            else if (trc != -ENOPARAM)
>>>> +                    return trc;
>>>> +    }
>>>> +    return rc;
>>>>   }