Attention is currently required from: laforge, keith.

dexter has posted comments on this change. ( https://gerrit.osmocom.org/c/libosmo-abis/+/32374 )

Change subject: e1d: reconnect to osmo-e1d after connection loss ......................................................................

Patch Set 3:

(1 comment)

Patchset:

PS2:

I suppose because the issue being attended to here was observed and reported in libosmo-abis using a […]

Its a problem with the e1d driver in libosmo-e1d. The file descriptors used to exchange the data on the E1 timeslots are under the control of libosmo-abis (in the line data struct). Also this part of the implementation has to interface between the way libosmo-abis manages the E1 line and osmo-e1d. This is already very application specific.

It is true, each application using the osmo-e1d client has the same problem but the problem is no different from handling any other socket connection.

One could still think to re-design the API towards osmo-e1d to hide everything socket specific from the API user. Basically the driver would be outsourced to libosmo-e1d and replaced with a minimalistic version that just calls API functions and handles some context. Basically we would have some kind of middle layer between libosmo-abis and the original libosmo-e1d API. I am not sure, this might be overcomplicated. (at least the generation of the EV_CONN_LOST might become easier since we could tie that to the control socket)