Attention is currently required from: laforge, pespin.

fixeria has posted comments on this change. ( https://gerrit.osmocom.org/c/libosmo-gprs/+/32644 )

Change subject: sm: Fix APN IE content not transmitted in Act PDP Ctx Req ......................................................................

Patch Set 1:

(1 comment)

File src/sm/sm_pdu.c:

https://gerrit.osmocom.org/c/libosmo-gprs/+/32644/comment/d65c78bd_87fdc48f PS1, Line 208: rc = osmo_apn_from_str(msg->tail, msgb_tailroom(msg), sme->apn); : if (rc < 0) : return -EINVAL; : *l = rc; : msgb_put(msg, *l);

IMHO this is highly unusual: passing a pointer to msg->tail and storing some data before doing the msgb_put. All the "legacy" code that I recall does the msgb_put() first. It would at that point ASSERT if we ever overrun msgb_tailroom, before anything is memcpying over it.

I would not say it's highly unusual, but rather a common practice in cases when you don't know how much to `msgb_put()` before calling a buffer filling function, e.g. `read()` or `recv()`.

Below are grep results for all my local repositories (39 matches total):

``` $ ./gits do grep "msg->tail, msgb_tailroom" | grep "msgb_tailroom[(]" | wc -l 13 $ ./gits do grep "msg->data, msgb_tailroom" | grep "msgb_tailroom[(]" | wc -l 9 $ ./gits do grep "msgb_data(msg), msgb_tailroom" | grep "msgb_tailroom[(]" | wc -l 17 ```