if ip-id is not increased by1 for each packet in IPV4, uor-2-ts ip-id is decoded wrong

[qingbaibai]

hello,

if ip-id is not increased by1 for each packet in IPV4, uor-2-ts ip-id is decoded wrong. For example, the ip-id of first packet is 65052 and the ip-id of next packet is 65054. The first packet is compressed as IR packet and the second packet is compressed as UOR-2-TS with extension3. When decoding UOR-2-TS with extension3, it uses "d_ip_id_decode()" function in ip_id.c. However, the ip-id is decoded as 65053.

Thank you.

[qingbaibai (sunqing-cztz)]

In RFC3095 page 75:
If value(RND) = 0, hdr(IP-ID) is compressed using Offset IP-ID
encoding (see section 4.5.5) using p = 0 and default-slope(IP-ID
offset) = 0.
If value(RND) = 1, IP-ID is the uncompressed hdr(IP-ID). IP-ID is
then passed as additional octets at the end of the compressed
header, after any extensions.

and page 85:
| IP-ID | 2 octets, if I = 1

Hence, in my opinion, if I=1, it corresponds to RND=1 and if I=0, it corresponds to RND=0.

Both in rohc-1.3.1 and rohc-1.3.2, in function"code_EXT3_packet()" in c_generic.c :
if(ip_get_version(ip) == IPV4)
  {
   if((nr_ip_id_bits > 0 && g_context->ip_flags.info.v4.rnd == 0) ||
      (g_context->ip_flags.info.v4.rnd_count < MAX_FO_COUNT &&
       g_context->ip_flags.info.v4.rnd == 0))
   {
    I = 1;
   }
   f_byte |= (I & 0x01) << 2;
  }
It sets I to 1 when rnd is 0.

[Didier Barvaux (didier-barvaux)]

> if ip-id is not increased by1 for each packet in IPV4, uor-2-ts ip-id is decoded wrong.
> For example, the ip-id of first packet is 65052 and the ip-id of next packet is 65054.
> The first packet is compressed as IR packet and the second packet is compressed
> as UOR-2-TS with extension3. When decoding UOR-2-TS with extension3, it uses
> "d_ip_id_decode()" function in ip_id.c. However, the ip-id is decoded as 65053.

You're right, there is a problem with non-regular IP-IDs and the UOR-2-TS packet. They are not compatible.

The UOR-2-TS packet type cannot be used if the IP-ID is not random (RND=1) and cannot be deduced from the reference IP-ID and the transmitted SN. In such cases, other packet types are selected. Look at the algorithm defined in the decide_SO_packet() function. In case of RTP profile with one unique IP header, the relevant part of the algorithm is as follow:

            packet = PACKET_UOR_2_TS; /* default packet */

            if(nr_sn_bits <= 4 && nr_ip_id_bits == 0 &&
               (nr_ts_bits == 0 || is_deductible(rtp_context->ts_sc)) &&
               rtp_context->tmp_variables.m_set == 0)
               packet = PACKET_UO_0;
            else if(nr_sn_bits <= 4 && nr_ip_id_bits == 0 && nr_ts_bits <= 5)
               packet = PACKET_UO_1_TS;
            else if(nr_sn_bits <= 4 && nr_ip_id_bits <= 5 &&
                    (nr_ts_bits == 0 || is_deductible(rtp_context->ts_sc)))
               packet = PACKET_UO_1_ID;
            else if(nr_ip_id_bits != 0 && nr_ts_bits <= MAX_BITS_IN_4_BYTE_SDVL)
               packet = PACKET_UOR_2_ID;

[Didier Barvaux (didier-barvaux)]

> In RFC3095 page 75:
> If value(RND) = 0, hdr(IP-ID) is compressed using Offset IP-ID
> encoding (see section 4.5.5) using p = 0 and default-slope(IP-ID
> offset) = 0.
> If value(RND) = 1, IP-ID is the uncompressed hdr(IP-ID). IP-ID is
> then passed as additional octets at the end of the compressed
> header, after any extensions.
>
> and page 85:
> | IP-ID | 2 octets, if I = 1
>
> Hence, in my opinion, if I=1, it corresponds to RND=1 and if I=0, it corresponds to RND=0.

I disagree with your analysis of the RFC. The IP-ID field that is "passed as additional octets at the end of the compressed header, after any extensions" is not the optional IP-ID in extension 3 (optional field present if I=1 as you mentioned). This is clear in the general format of the ROHC header described on page 77 of RFC 3095 [1].

Regards,
Didier

[1] http://tools.ietf.org/html/rfc3095#page-77

[qingbaibai (sunqing-cztz)]

hello,

“I disagree with your analysis of the RFC. The IP-ID field that is "passed as additional octets at the end of the compressed header, after any extensions" is not the optional IP-ID in extension 3 (optional field present if I=1 as you mentioned). This is clear in the general format of the ROHC header described on page 77 of RFC 3095 [1].”

I check the RFC 3095 and I was wrong. however,

"The UOR-2-TS packet type cannot be used if the IP-ID is not random (RND=1) and cannot be deduced from the reference IP-ID and the transmitted SN."
I think RND=1 means that ip-id is random. In RFC3095 page 82, UOR-2-TS cannot be used if there is no IPv4 header in the
context or if value(RND) and value(RND2) are both 1. It seems that you mean "The UOR-2-TS packet type can be used if the IP-ID is random (RND=1)"

[Didier Barvaux (didier-barvaux)]

> "The UOR-2-TS packet type cannot be used if the IP-ID is
> not random (RND=1) and cannot be deduced from the
> reference IP-ID and the transmitted SN."
> I think RND=1 means that ip-id is random. In RFC3095 page 82,
> UOR-2-TS cannot be used if there is no IPv4 header in the
> context or if value(RND) and value(RND2) are both 1. It seems
> that you mean "The UOR-2-TS packet type can be used if the
> IP-ID is random (RND=1)"

Yes, you're right. I made a mistake in my explanation :-/

Regards,
Didier

[Didier Barvaux (didier-barvaux)]

I forget to thank you for all the tests and code analysis you're making! Given enough eyeballs, all bugs are shallow :-)

Didier

[qingbaibai (sunqing-cztz)]

Thank you. ^.^.

You mentioned "The UOR-2-TS packet type cannot be deduced from the reference IP-ID and the transmitted SN." However, I find that ip-id of the UOR-2-TS packet type with extension3 is deduced from the reference IP-ID and the transmitted SN.

here is compression:
                <compression>
                        <log>
[../rohc2-1.3.1/src/comp/rohc_comp.c:192 rohc_compress()] size of IP packet = 20
0 bytes
[../rohc2-1.3.1/src/comp/rohc_comp.c:218 rohc_compress()] IP protocol = 17
[../rohc2-1.3.1/src/comp/rohc_comp.c:978 c_get_profile_from_packet()] UDP port =
 0x139c (5020)
[../rohc2-1.3.1/src/comp/rohc_comp.c:227 rohc_compress()] using profile 'RTP / C
ompressor' (0x0001)
[../rohc2-1.3.1/src/comp/rohc_comp.c:1198 c_find_context()] using context CID =
0
[../rohc2-1.3.1/src/comp/rohc_comp.c:283 rohc_compress()] compress the packet #2

[../rohc2-1.3.1/src/comp/c_generic.c:6529 check_ip_identification()] 1) old_id =
 0xfe1c new_id = 0xfe1e
[../rohc2-1.3.1/src/comp/c_generic.c:6565 check_ip_identification()] NBO = 1, RN
D = 0
[../rohc2-1.3.1/src/common/ts_sc_comp.c:89 c_add_ts()] Timestamp = 3016314151
[../rohc2-1.3.1/src/common/ts_sc_comp.c:109 c_add_ts()] state INIT_STRIDE
[../rohc2-1.3.1/src/common/ts_sc_comp.c:114 c_add_ts()] ts_stride = 160
[../rohc2-1.3.1/src/common/ts_sc_comp.c:118 c_add_ts()] ts_offset = 3016314151 m
odulo 160 = 71
[../rohc2-1.3.1/src/common/ts_sc_comp.c:121 c_add_ts()] ts_scaled = (3016314151
- 71) / 160 = 18851963
[../rohc2-1.3.1/src/comp/c_generic.c:837 c_generic_encode()] SN = 197
[../rohc2-1.3.1/src/comp/c_generic.c:6325 changed_dynamic_both_hdr()] check for
changed fields in the outer IP header
[../rohc2-1.3.1/src/comp/c_generic.c:860 c_generic_encode()] send_static = 0, se
nd_dynamic = 0
[../rohc2-1.3.1/src/comp/c_generic.c:669 change_state()] change from state 1 to
state 3
[../rohc2-1.3.1/src/comp/c_generic.c:868 c_generic_encode()] ip_id = 0xfe1e, con
text_sn = 197
[../rohc2-1.3.1/src/comp/c_generic.c:2351 update_variables()] ip_id delta = 0xfd
59 / 64857
[../rohc2-1.3.1/src/comp/c_generic.c:2352 update_variables()] ip_id bits = -1
[../rohc2-1.3.1/src/comp/c_generic.c:2362 update_variables()] sn bits=-1
[../rohc2-1.3.1/src/comp/c_generic.c:2725 decide_packet()] SO state
[../rohc2-1.3.1/src/comp/c_generic.c:2564 decide_SO_packet()] nr_ip_bits=-1 nr_s
n_bits=-1 nr_of_ip_hdr=1 rnd=0
[../rohc2-1.3.1/src/comp/c_generic.c:4096 code_UO2_packet()] code UOR-2-TS packe
t (CID = 0)
[../rohc-1.3.1final/src/comp/cid.c:74 code_cid_values()] CID = 0 => no add-CID
[../rohc2-1.3.1/src/comp/c_generic.c:4701 code_UOR2_TS_bytes()] code UOR-2-TS pa
cket with extension 3
[../rohc2-1.3.1/src/comp/c_generic.c:4714 code_UOR2_TS_bytes()] TS to send = 0xb
3c94d27
[../rohc2-1.3.1/src/comp/c_generic.c:4716 code_UOR2_TS_bytes()] bits of TS = 32
(3 in header, 29 in EXT3)
[../rohc2-1.3.1/src/comp/c_generic.c:4732 code_UOR2_TS_bytes()] bits of TS in 1s
t byte = 0x5 (mask = 0x7)
[../rohc2-1.3.1/src/comp/c_generic.c:4171 code_UO2_packet()] f_byte = 0xc5
[../rohc2-1.3.1/src/comp/c_generic.c:4175 code_UO2_packet()] s_byte = 0x85
[../rohc2-1.3.1/src/comp/c_generic.c:4178 code_UO2_packet()] t_byte = 0xfa
[../rohc2-1.3.1/src/comp/c_gen...

[Didier Barvaux (didier-barvaux)]

When I said:
    The UOR-2-TS packet type cannot be used if the IP-ID is
    random* (RND=1) and cannot be deduced from the
    reference IP-ID and the transmitted SN.

one should read:
    The UOR-2-TS packet type cannot be used if:
       1/ the IP-ID is random* (RND=1),
       2/ the IP-ID cannot be deduced from the reference
            IP-ID and the transmitted SN.

To summarize, it means that UOR-2-TS cannot be used if there is at least 1 IP-ID bit to transmit in the ROHC packet.

* I fixed my previous mistake to avoid confusion again.

Regards,
Didier