Re: [PATCH] block: Fix regression in sed-opal for a saved key.

[Jens Axboe]

On 10/11/23 2:30 AM, Milan Broz wrote:
> On 10/5/23 19:58, Greg Joyce wrote:
>> On Thu, 2023-10-05 at 08:58 +0200, Milan Broz wrote:
>>> On 10/4/23 22:54, Greg Joyce wrote:
>>>> On Tue, 2023-10-03 at 12:02 +0200, Milan Broz wrote:
>>>>> The commit 3bfeb61256643281ac4be5b8a57e9d9da3db4335
>>>>> introduced the use of keyring for sed-opal.
>>>>>
>>>>> Unfortunately, there is also a possibility to save
>>>>> the Opal key used in opal_lock_unlock().
>>>>>
>>>>> This patch switches the order of operation, so the cached
>>>>> key is used instead of failure for opal_get_key.
>>>>>
>>>>> The problem was found by the cryptsetup Opal test recently
>>>>> added to the cryptsetup tree.
>>>>>
>>>>> Fixes: 3bfeb6125664 ("block: sed-opal: keyring support for SED
>>>>> keys")
>>>>> Tested-by: Ondrej Kozina <okozina@xxxxxxxxxx>
>>>>> Signed-off-by: Milan Broz <gmazyland@xxxxxxxxx>
>>>>> ---
>>>>>    block/sed-opal.c | 7 +++----
>>>>>    1 file changed, 3 insertions(+), 4 deletions(-)
>>>>>
>>>>> diff --git a/block/sed-opal.c b/block/sed-opal.c
>>>>> index 6d7f25d1711b..04f38a3f5d95 100644
>>>>> --- a/block/sed-opal.c
>>>>> +++ b/block/sed-opal.c
>>>>> @@ -2888,12 +2888,11 @@ static int opal_lock_unlock(struct
>>>>> opal_dev
>>>>> *dev,
>>>>>        if (lk_unlk->session.who > OPAL_USER9)
>>>>>            return -EINVAL;
>>>>>
>>>>> -    ret = opal_get_key(dev, &lk_unlk->session.opal_key);
>>>>> -    if (ret)
>>>>> -        return ret;
>>>>>        mutex_lock(&dev->dev_lock);
>>>>>        opal_lock_check_for_saved_key(dev, lk_unlk);
>>>>> -    ret = __opal_lock_unlock(dev, lk_unlk);
>>>>> +    ret = opal_get_key(dev, &lk_unlk->session.opal_key);
>>>>> +    if (!ret)
>>>>> +        ret = __opal_lock_unlock(dev, lk_unlk);
>>>>
>>>> This is relying on opal_get_key() returning 0 to decide if
>>>> __opal_lock_unlock() is called. Is this really what you want? It
>>>> seems
>>>> that you would want to unlock if the key is a LUKS key, even if
>>>> opal_get_key() returns non-zero.
>>>
>>> I think it is ok. That was logic introduced in your keyring patch
>>> anyway.
>>>
>>> I just fixed that if key is cached (stored in OPAL struct), that key
>>> is used
>>> and subsequent opal_get_key() does nothing, returning 0.
>>>
>>> The story is here that both OPAL lock and unlock need key, while LUKS
>>> logic never required key for lock (deactivation), so we rely on the
>>> cached
>>> OPAL key here. We do not need any key stored for unlocking (that is
>>> always
>>> decrypted from a keyslot)
>>> (I think requiring key for locking range is a design mistake in OPAL
>>> but
>>> that's not relevant for now :-)
>>
>> Okay, if the key is such that opal_get_key() always returns 0, then I
>> agree there isn't an issue.
> 
> 
> Jens, what's the status of this patch?
> 
> It is clear regression in 6.6 (I forgot to add regression list, fixed now.)
> 
> For reference, the original report and patch is here
> #regzbot link: https://lore.kernel.org/linux-block/20231003100209.380037-1-gmazyland@xxxxxxxxx/
> #regzbot ^introduced 3bfeb6125664

Was waiting on Greg to ack/review it, which it looks like he kind of
has. But would've been nice with a formal ack on it. I've queued it up
now.

-- 
Jens Axboe