Re: [PATCH 0/1] Sequence number out of range fix

From: "Gix, Brian" <brian.gix@intel.com>
To: "michal.lowas-rzechonek@silvair.com" 
	<michal.lowas-rzechonek@silvair.com>
Cc: "jakub.witowski@silvair.com" <jakub.witowski@silvair.com>,
	"linux-bluetooth@vger.kernel.org"
	<linux-bluetooth@vger.kernel.org>,
	"istotlan@ingas-xps13.amr.corp.intel.com" 
	<istotlan@ingas-xps13.amr.corp.intel.com>
Subject: Re: [PATCH 0/1] Sequence number out of range fix
Date: Wed, 15 Jan 2020 19:53:09 +0000	[thread overview]
Message-ID: <4C48D122-7887-4A5E-9E3C-0A2FE15D6A1A@intel.com> (raw)
In-Reply-To: <20200115190904.qcusabsj3ti7sid4@kynes>

Hi Michał 

> On Jan 15, 2020, at 11:09 AM, "michal.lowas-rzechonek@silvair.com" <michal.lowas-rzechonek@silvair.com> wrote:
> 
> Brian, Jakub,
> 
>> On 01/15, Gix, Brian wrote:
>> This is a very dangerous course of action.  The suggested patch might
>> potentially cause a node to re-use a sequence number more than once,
>> which would cause a "dirty nonce" condition, and allow unauthorized
>> entities to derive the encrypted data without the keys.
> 
> Good point. Note that his is also possible in the current
> implementation: since seq_num is applied to nonce with a 24bit mask,
> it's going to wrap around.

The full IV Index is in the nonce, and at 192 hours per IV index, will be unique for something like 1.4 million years.

> 
>> While this is technically possible, especially with a daemon that
>> might be looping back messages internally without ever using the OTA
>> interface, it is not really possible in an actual BT driven real life
>> system.
> 
> There is another option to exhaust the 24 bit range: we have the
> overcommit mechanism in the storage. Let's say the daemon starts, sends
> a few messages (and overcommits the sequence by a certain value), then
> either the daemon, or the system crashes.
> 
> Do that a few times, and you end up with storage containing sequence
> number exceeding 24 bits.

The over commit is calculated based on the usage rate, and the daemon would need to unexpectedly abort (not just ctrl-c or exit) for us to use the over-commit value, as on deliberate exit, the actual sequence used is saved. If an unexpected abort occurs, I think the default daemon restart is 5 seconds?
> 
>> Beacuse we store sequence numbers internally with a u32 sized data
>> type, we should *let* the value go over the max legal sequence
>> nunmbver of 0xFFFFFF (perhaps capping it at 0x01000000 to prevent
>> "super overflows").  Then we *reject* all send requests with a
>> sequence number > 0xFFFFFF.
> 
> Sounds good.
> 
> -- 
> Michał Lowas-Rzechonek <michal.lowas-rzechonek@silvair.com>
> Silvair http://silvair.com
> Jasnogórska 44, 31-358 Krakow, POLAND