Re: [PATCH-v7 1/2] Bluetooth: Add MITM mechanism to LE-SMP

[Marcel Holtmann <marcel@holtmann.org>]

Hi Brian,

> > On Tue, Dec 20, 2011 at 05:04:41PM -0800, Brian Gix wrote:
> >> To achive Man-In-The-Middle (MITM) level security with Low Energy,
> >> we have to enable User Passkey Comparison.  This commit modifies the
> >> hard-coded JUST-WORKS pairing mechanism to support query via the MGMT
> >> interface of Passkey comparison and User Confirmation.
> >>
> >> Signed-off-by: Brian Gix<bgix@codeaurora.org>
> >> ---
> >>   include/net/bluetooth/hci_core.h |    1 +
> >>   include/net/bluetooth/smp.h      |    5 +
> >>   net/bluetooth/smp.c              |  227 ++++++++++++++++++++++++++++++++++----
> >>   3 files changed, 211 insertions(+), 22 deletions(-)
> >
> > ...
> >
> >> +static int tk_request(struct l2cap_conn *conn, u8 remote_oob, u8 auth,
> >> +						u8 local_io, u8 remote_io)
> >> +{
> >> +	struct hci_conn *hcon = conn->hcon;
> >> +	struct smp_chan *smp = conn->smp_chan;
> >> +	u8 method;
> >> +	u32 passkey = 0;
> >> +	int ret = 0;
> >> +
> >> +	/* Initialize key to JUST WORKS */
> >> +	memset(smp->tk, 0, sizeof(smp->tk));
> >> +	clear_bit(SMP_FLAG_TK_VALID,&smp->smp_flags);
> >> +
> >> +	BT_DBG("tk_request: auth:%d lcl:%d rem:%d", auth, local_io, remote_io);
> >> +
> >> +	/* If neither side wants MITM, use JUST WORKS */
> >> +	/* If either side has unknown io_caps, use JUST_WORKS */
> >> +	if (!(auth&  SMP_AUTH_MITM) ||
> >> +			local_io>  SMP_IO_KEYBOARD_DISPLAY ||
> >> +			remote_io>  SMP_IO_KEYBOARD_DISPLAY) {
> >> +		auth&= ~SMP_AUTH_MITM;
> >> +		set_bit(SMP_FLAG_TK_VALID,&smp->smp_flags);
> >> +		return 0;
> >> +	}
> >> +
> >> +	/* MITM is now officially requested, but not required */
> >> +	/* Determine what we need (if anything) from the agent */
> >> +	method = gen_method[local_io][remote_io];
> >> +
> >> +	if (method == JUST_WORKS || method == JUST_CFM)
> >> +		auth&= ~SMP_AUTH_MITM;
> >> +
> >> +	/* Don't bother confirming unbonded JUST_WORKS */
> >> +	if (!(auth&  SMP_AUTH_BONDING)&&  method == JUST_CFM) {
> >> +		set_bit(SMP_FLAG_TK_VALID,&smp->smp_flags);
> >> +		return 0;
> >> +	} else if (method == JUST_WORKS) {
> >> +		set_bit(SMP_FLAG_TK_VALID,&smp->smp_flags);
> >> +		return 0;
> >> +	} else if (method == OVERLAP) {
> >> +		if (hcon->link_mode&  HCI_LM_MASTER)
> >> +			method = CFM_PASSKEY;
> >> +		else
> >> +			method = REQ_PASSKEY;
> >> +	}
> >> +
> >> +	if (method == CFM_PASSKEY) {
> >> +		u8 key[16];
> >> +		/* Generate a passkey for display. It is not valid until
> >> +		 * confirmed.
> >> +		 */
> >> +		memset(key, 0, sizeof(key));
> >> +		get_random_bytes(&passkey, sizeof(passkey));
> >> +		passkey %= 1000000;
> >> +		put_unaligned_le32(passkey, key);
> >> +		swap128(key, smp->tk);
> >> +		BT_DBG("PassKey: %d", passkey);
> >> +	}
> >
> > Would it make sense to use switch(method) instead of [if/else/else if] ?
> 
> I can use a simpler "method" switch at the end, to determine which mgmt 
> call to make to get the passkey or confirmation. However, the "truth 
> table" is a bit too complex for a switch, with the method being tweaked 
> based on rules coming from the Core Spec, and then later clarified in 
> erratum 4249.  I also wanted to avoid putting the random passkey 
> generation inside of the  hci_dev_lock block.

I am fine with if statements as long as you just return and do not
bother with else if.

Regards

Marcel