Change in osmo-hnodeb[master]: First implementation of the LLSK gtp SAPI

[pespin]

pespin has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-hnodeb/+/26503 )

Change subject: First implementation of the LLSK gtp SAPI
......................................................................


Patch Set 2:

(3 comments)

https://gerrit.osmocom.org/c/osmo-hnodeb/+/26503/2/src/osmo-hnodeb/gtp.c 
File src/osmo-hnodeb/gtp.c:

https://gerrit.osmocom.org/c/osmo-hnodeb/+/26503/2/src/osmo-hnodeb/gtp.c@70 
PS2, Line 70: req
> comment says request, next line says indication.  Further comment below again says req. […]
I don't really think it's useful. Each path has a clear 1 specific primitive to rx/tx, so it's easier to write the string directly to be able to find it with grep, or simply when reading code.

These names were simply wrong because I forgot to change it when I reworked the code to use proper operations.


https://gerrit.osmocom.org/c/osmo-hnodeb/+/26503/2/src/osmo-hnodeb/gtp.c@126 
PS2, Line 126: }
             : 
             : 	osmo_fd_setup(&hnb->gtp.fd0, gsn->fd0, OSMO_FD_READ, hnb_gtp_fd_cb, hnb, 0);
             : 	if ((rc = osmo_fd_register(&hnb->gtp.fd0)) < 0)
             : 		goto free_ret;
             : 
             : 	osmo_fd_setup(&hnb->gtp.fd1c, gsn->fd1c, OSMO_FD_READ, hnb_gtp_fd_cb, hnb, 1);
             : 	if ((rc = osmo_fd_register(&hnb->gtp.fd1c)) < 0)
             : 		goto free_ret;
             : 
             : 	osmo_fd_setup(&hnb->gtp.fd1u, gsn->fd1u, OSMO_FD_READ, hnb_gtp_fd_cb, hnb, 2);
             : 	if ((rc = osmo_fd_register(&hnb->gtp.fd1u)) < 0)
> The Iuh interface only uses GTP-U v1, neither GTPv0 and no GTPv1-C. […]
After having a quick look at libgtp and TS 29 060 I decided to go for libgtp to avoid having to reimplement the wheel and also in order to avoid to take care or think of stuff like sequence numbers, possible retransmits, gtpu-1 vs 0, encoding/decoding fields, etc. which indeed proved useful into having already a working GTP test in TTCN3.

I went initializing the 3 sockets here to speed up the setup and avoid possible failures due to some data structures not being initialized, etc.
I plan to give a try now removing the GTPC socket (I was not sure about GTP0 support here btw).
I also though about adding a new mode GTP_MODE_GTPU_ONLY (see line 115) to make sure libgtp expects a GTP-U link to be set up too.


https://gerrit.osmocom.org/c/osmo-hnodeb/+/26503/2/src/osmo-hnodeb/gtp.c@180 
PS2, Line 180: 		LOG
> where deos this constraint come from? libgtp?
Yes, pdp_t below uses "struct in_addr" in several fields, it seems only IPv4 is supported in the transport GTP link (IP/UDP).



-- 
To view, visit https://gerrit.osmocom.org/c/osmo-hnodeb/+/26503
To unsubscribe, or for help writing mail filters, visit https://gerrit.osmocom.org/settings

Gerrit-Project: osmo-hnodeb
Gerrit-Branch: master
Gerrit-Change-Id: I5a6f5dfc4e508c92adb35210b4dc576d64353366
Gerrit-Change-Number: 26503
Gerrit-PatchSet: 2
Gerrit-Owner: pespin <pespin at sysmocom.de>
Gerrit-Reviewer: Jenkins Builder
Gerrit-Reviewer: laforge <laforge at osmocom.org>
Gerrit-Comment-Date: Mon, 13 Dec 2021 16:33:12 +0000
Gerrit-HasComments: Yes
Gerrit-Has-Labels: No
Comment-In-Reply-To: laforge <laforge at osmocom.org>
Gerrit-MessageType: comment
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osmocom.org/pipermail/gerrit-log/attachments/20211213/ccdd1a0d/attachment.htm>