diff mbox series

[v2,4/9] ft: add FTE/RSNE building to ft_prepare_handshake

Message ID 20231206150708.2080336-5-prestwoj@gmail.com (mailing list archive)
State New
Headers show
Series Reassoc/FT roaming unification | expand

Checks

Context Check Description
tedd_an/pre-ci_am success Success
prestwoj/iwd-ci-gitlint success GitLint

Commit Message

James Prestwood Dec. 6, 2023, 3:07 p.m. UTC
In preparation to remove ft_associate build the FTE/RSNE in
ft_prepare_handshake and set into the handshake object directly.
---
 src/ft.c | 84 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 82 insertions(+), 2 deletions(-)

Comments

Denis Kenzior Dec. 6, 2023, 4:36 p.m. UTC | #1
Hi James,

On 12/6/23 09:07, James Prestwood wrote:
> In preparation to remove ft_associate build the FTE/RSNE in
> ft_prepare_handshake and set into the handshake object directly.
> ---
>   src/ft.c | 84 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
>   1 file changed, 82 insertions(+), 2 deletions(-)
> 

<snip>

> @@ -931,6 +937,80 @@ static void ft_prepare_handshake(struct ft_info *info,
>   						info->ft_info.r1khid);
>   
>   	handshake_state_derive_ptk(hs);
> +
> +	/*
> +	 * Rebuild the RSNE to include the PMKR1Name and append
> +	 * MDE + FTE.
> +	 *
> +	 * 12.8.4: "If present, the RSNE shall be set as follows:
> +	 * - Version field shall be set to 1.
> +	 * - PMKID Count field shall be set to 1.
> +	 * - PMKID field shall contain the PMKR1Name.
> +	 * - All other fields shall be as specified in 8.4.2.27
> +	 *   and 11.5.3."
> +	 */
> +	if (ie_parse_rsne_from_data(hs->supplicant_ie,
> +					hs->supplicant_ie[1] + 2,
> +					&rsn_info) < 0)
> +		return false;
> +
> +	rsn_info.num_pmkids = 1;
> +	rsn_info.pmkids = hs->pmk_r1_name;
> +	/* Always set OCVC false for FT for now */
> +	rsn_info.ocvc = false;
> +	rsne = alloca(256);
> +
> +	ie_build_rsne(&rsn_info, rsne);
> +	handshake_state_set_supplicant_ie(hs, rsne);

This is probably safe since we over-write the supplicant ie in 
netdev_connect_event() -> parse_request_ies()

> +
> +	/*
> +	 * 12.8.4: "If present, the FTE shall be set as follows:
> +	 * - ANonce, SNonce, R0KH-ID, and R1KH-ID shall be set to
> +	 *   the values contained in the second message of this
> +	 *   sequence.
> +	 * - The Element Count field of the MIC Control field shall
> +	 *   be set to the number of elements protected in this
> +	 *   frame (variable).
> +	 * [...]
> +	 * - All other fields shall be set to 0."
> +	 */
> +	memset(&ft_info, 0, sizeof(ft_info));
> +	ft_info.mic_element_count = 3;
> +	memcpy(ft_info.r0khid, hs->r0khid, hs->r0khid_len);
> +	ft_info.r0khid_len = hs->r0khid_len;
> +	memcpy(ft_info.r1khid, hs->r1khid, 6);
> +	ft_info.r1khid_present = true;
> +	memcpy(ft_info.anonce, hs->anonce, 32);
> +	memcpy(ft_info.snonce, hs->snonce, 32);
> +
> +	/*
> +	 * IEEE 802.11-2020 Section 13.7.1 FT reassociation in an RSN
> +	 *
> +	 * "If dot11RSNAOperatingChannelValidationActivated is true and
> +	 *  the FTO indicates OCVC capability, the target AP shall
> +	 *  ensure that OCI subelement of the FTE matches by ensuring
> +	 *  that all of the following are true:
> +	 *      - OCI subelement is present
> +	 *      - Channel information in the OCI matches current
> +	 *        operating channel parameters (see 12.2.9)"
> +	 */
> +	if (hs->supplicant_ocvc && hs->chandef) {
> +		oci_from_chandef(hs->chandef, ft_info.oci);
> +		ft_info.oci_present = true;
> +	}
> +
> +	fte = alloca(256);
> +	ie_build_fast_bss_transition(&ft_info, kck_len, fte);
> +
> +	if (!ft_calculate_fte_mic(hs, 5, rsne, fte, NULL, ft_info.mic))
> +		return false;
> +
> +	/* Rebuild the FT IE now with the MIC included */
> +	ie_build_fast_bss_transition(&ft_info, kck_len, fte);
> +
> +	handshake_state_set_fte(hs, fte);

However, this is less clear to me.  Looking at how FILS and FT uses this API, it 
seems that set_fte is meant for the authenticator FTE element?  So I think 
rekeying after FT would be broken by this change.

> +
> +	return true;
>   }
>   
>   static bool ft_send_action(struct wiphy_radio_work_item *work)

Regards,
-Denis
James Prestwood Dec. 6, 2023, 5:08 p.m. UTC | #2
Hi Denis,

On 12/6/23 08:36, Denis Kenzior wrote:
> Hi James,
>
> On 12/6/23 09:07, James Prestwood wrote:
>> In preparation to remove ft_associate build the FTE/RSNE in
>> ft_prepare_handshake and set into the handshake object directly.
>> ---
>>   src/ft.c | 84 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
>>   1 file changed, 82 insertions(+), 2 deletions(-)
>>
>
> <snip>
>> +    fte = alloca(256);
>> +    ie_build_fast_bss_transition(&ft_info, kck_len, fte);
>> +
>> +    if (!ft_calculate_fte_mic(hs, 5, rsne, fte, NULL, ft_info.mic))
>> +        return false;
>> +
>> +    /* Rebuild the FT IE now with the MIC included */
>> +    ie_build_fast_bss_transition(&ft_info, kck_len, fte);
>> +
>> +    handshake_state_set_fte(hs, fte);
>
> However, this is less clear to me.  Looking at how FILS and FT uses 
> this API, it seems that set_fte is meant for the authenticator FTE 
> element?  So I think rekeying after FT would be broken by this change.

Good question. Rekeys do appear to work as-is but you are right, FILS/FT 
uses set_fte() for the authenticators element, but eapol seems to use 
hs->fte for building message 2/4, as well as checks that the handshakes 
FTE matches what the authenticator sends in 3/4. maybe this is actually 
a bug in eapol? I think the reason everything "works" is because the FTE 
should be the same between both peers.

We may want to refactor and do:

handshake_state_set_authenticator_fte()

handshake_state_set_supplicant_fte()

Thanks,

James

>
>> +
>> +    return true;
>>   }
>>     static bool ft_send_action(struct wiphy_radio_work_item *work)
>
> Regards,
> -Denis
Denis Kenzior Dec. 6, 2023, 5:14 p.m. UTC | #3
Hi James,

>>> +    handshake_state_set_fte(hs, fte);
>>
>> However, this is less clear to me.  Looking at how FILS and FT uses this API, 
>> it seems that set_fte is meant for the authenticator FTE element?  So I think 
>> rekeying after FT would be broken by this change.
> 
> Good question. Rekeys do appear to work as-is but you are right, FILS/FT uses 
> set_fte() for the authenticators element, but eapol seems to use hs->fte for 

I'm pretty sure the intent was for the FTE element to be from the authenticator.

> building message 2/4, as well as checks that the handshakes FTE matches what the 

I'll have to look at how ptk_2_of_4 uses it.  Memory is fuzzy now.  Need to open 
the spec.

> authenticator sends in 3/4. maybe this is actually a bug in eapol? I think the 
> reason everything "works" is because the FTE should be the same between both peers.

Yes, but also we have logic in: netdev_connect_event() that sets the FTE from 
the response IEs.  So that's probably how things end up working in the end.

> 
> We may want to refactor and do:
> 
> handshake_state_set_authenticator_fte()
> 
> handshake_state_set_supplicant_fte()

Yeah, that seems reasonable.

Regards,
-Denis
James Prestwood Dec. 6, 2023, 5:59 p.m. UTC | #4
On 12/6/23 09:14, Denis Kenzior wrote:
> Hi James,
>
>>>> +    handshake_state_set_fte(hs, fte);
>>>
>>> However, this is less clear to me.  Looking at how FILS and FT uses 
>>> this API, it seems that set_fte is meant for the authenticator FTE 
>>> element?  So I think rekeying after FT would be broken by this change.
>>
>> Good question. Rekeys do appear to work as-is but you are right, 
>> FILS/FT uses set_fte() for the authenticators element, but eapol 
>> seems to use hs->fte for 
>
> I'm pretty sure the intent was for the FTE element to be from the 
> authenticator.
>
>> building message 2/4, as well as checks that the handshakes FTE 
>> matches what the 
>
> I'll have to look at how ptk_2_of_4 uses it.  Memory is fuzzy now.  
> Need to open the spec.

All I could find quickly was:

12.7.6.3 4-way handshake message 2

"and that the FTE and MDE are the same as those provided in the AP’s 
(Re)Association Response frame"

So it does appear for EAPoL its the authenticators FTE.

>
>> authenticator sends in 3/4. maybe this is actually a bug in eapol? I 
>> think the reason everything "works" is because the FTE should be the 
>> same between both peers.
>
> Yes, but also we have logic in: netdev_connect_event() that sets the 
> FTE from the response IEs.  So that's probably how things end up 
> working in the end.
>
>>
>> We may want to refactor and do:
>>
>> handshake_state_set_authenticator_fte()
>>
>> handshake_state_set_supplicant_fte()
>
> Yeah, that seems reasonable.
>
> Regards,
> -Denis
diff mbox series

Patch

diff --git a/src/ft.c b/src/ft.c
index 2cc611b8..358a4594 100644
--- a/src/ft.c
+++ b/src/ft.c
@@ -904,9 +904,15 @@  static void ft_info_destroy(void *data)
 	l_free(info);
 }
 
-static void ft_prepare_handshake(struct ft_info *info,
+static bool ft_prepare_handshake(struct ft_info *info,
 					struct handshake_state *hs)
 {
+	uint32_t kck_len = handshake_state_get_kck_len(hs);
+	struct ie_rsn_info rsn_info;
+	struct ie_ft_info ft_info;
+	uint8_t *fte;
+	uint8_t *rsne;
+
 	handshake_state_set_authenticator_address(hs, info->aa);
 
 	memcpy(hs->mde + 2, info->mde, 3);
@@ -914,7 +920,7 @@  static void ft_prepare_handshake(struct ft_info *info,
 	handshake_state_set_chandef(hs, NULL);
 
 	if (!hs->supplicant_ie)
-		return;
+		return true;
 
 	if (info->authenticator_ie)
 		handshake_state_set_authenticator_ie(hs,
@@ -931,6 +937,80 @@  static void ft_prepare_handshake(struct ft_info *info,
 						info->ft_info.r1khid);
 
 	handshake_state_derive_ptk(hs);
+
+	/*
+	 * Rebuild the RSNE to include the PMKR1Name and append
+	 * MDE + FTE.
+	 *
+	 * 12.8.4: "If present, the RSNE shall be set as follows:
+	 * - Version field shall be set to 1.
+	 * - PMKID Count field shall be set to 1.
+	 * - PMKID field shall contain the PMKR1Name.
+	 * - All other fields shall be as specified in 8.4.2.27
+	 *   and 11.5.3."
+	 */
+	if (ie_parse_rsne_from_data(hs->supplicant_ie,
+					hs->supplicant_ie[1] + 2,
+					&rsn_info) < 0)
+		return false;
+
+	rsn_info.num_pmkids = 1;
+	rsn_info.pmkids = hs->pmk_r1_name;
+	/* Always set OCVC false for FT for now */
+	rsn_info.ocvc = false;
+	rsne = alloca(256);
+
+	ie_build_rsne(&rsn_info, rsne);
+	handshake_state_set_supplicant_ie(hs, rsne);
+
+	/*
+	 * 12.8.4: "If present, the FTE shall be set as follows:
+	 * - ANonce, SNonce, R0KH-ID, and R1KH-ID shall be set to
+	 *   the values contained in the second message of this
+	 *   sequence.
+	 * - The Element Count field of the MIC Control field shall
+	 *   be set to the number of elements protected in this
+	 *   frame (variable).
+	 * [...]
+	 * - All other fields shall be set to 0."
+	 */
+	memset(&ft_info, 0, sizeof(ft_info));
+	ft_info.mic_element_count = 3;
+	memcpy(ft_info.r0khid, hs->r0khid, hs->r0khid_len);
+	ft_info.r0khid_len = hs->r0khid_len;
+	memcpy(ft_info.r1khid, hs->r1khid, 6);
+	ft_info.r1khid_present = true;
+	memcpy(ft_info.anonce, hs->anonce, 32);
+	memcpy(ft_info.snonce, hs->snonce, 32);
+
+	/*
+	 * IEEE 802.11-2020 Section 13.7.1 FT reassociation in an RSN
+	 *
+	 * "If dot11RSNAOperatingChannelValidationActivated is true and
+	 *  the FTO indicates OCVC capability, the target AP shall
+	 *  ensure that OCI subelement of the FTE matches by ensuring
+	 *  that all of the following are true:
+	 *      - OCI subelement is present
+	 *      - Channel information in the OCI matches current
+	 *        operating channel parameters (see 12.2.9)"
+	 */
+	if (hs->supplicant_ocvc && hs->chandef) {
+		oci_from_chandef(hs->chandef, ft_info.oci);
+		ft_info.oci_present = true;
+	}
+
+	fte = alloca(256);
+	ie_build_fast_bss_transition(&ft_info, kck_len, fte);
+
+	if (!ft_calculate_fte_mic(hs, 5, rsne, fte, NULL, ft_info.mic))
+		return false;
+
+	/* Rebuild the FT IE now with the MIC included */
+	ie_build_fast_bss_transition(&ft_info, kck_len, fte);
+
+	handshake_state_set_fte(hs, fte);
+
+	return true;
 }
 
 static bool ft_send_action(struct wiphy_radio_work_item *work)