Patch Set 2:

(9 comments)

https://gerrit.osmocom.org/#/c/6324/2/include/osmocom/bsc/acc_ramp.h
File include/osmocom/bsc/acc_ramp.h:

Line 1: /* (C) 2018 Stefan Sperling <ssperl...@sysmocom.de>
> (C) sysmocom / Author: Stefan Sperling, please look at other examples, than
Fine, fixed in next patch set.


Line 41:        struct gsm_bts *bts; /* backpointer to BTS using this ACC ramp 
*/
> one could avoid the back-pointer if we used offsetof() to go back from acc_
I agree. I will investigate this idea later and provide a follow-up patch if I 
find a nicer way of doing this.


https://gerrit.osmocom.org/#/c/6324/2/src/libbsc/acc_ramp.c
File src/libbsc/acc_ramp.c:

Line 68: static void allow_all_allowed_accs(struct acc_ramp *acc_ramp)
> the naming is a bit ... odd.
:)

I will change the naming in the next patch set to make a distinction between 
"allowed ACC" (this ACC is not currently barred) and "enabled ACC" (the user 
has not permanently barred this ACC via VTY). I hope that will be clearer.


PS2, Line 93: DRLL
> this is not related to the Radio Link Layer (which is a sub-layer of RSL). 
I will switch these to DRSL in the next patch set. I could also introduce a 
separate log category in a follow-up patch if desired.


Line 98: static void send_bts_system_info(struct gsm_bts *bts)
> this should become a public function that's part of the general bts/trx/sys
There already is one in bsc_init.c, gsm_bts_set_system_infos(). I will use that 
instead in the next patch set.


Line 107: static void do_ramping_step(void *data)
> 'acc' in the name might be useful in backtraces/debugging/... to indicate w
Agreed. Will rename in next patch set.


https://gerrit.osmocom.org/#/c/6324/2/src/libbsc/bsc_init.c
File src/libbsc/bsc_init.c:

Line 335:               acc_ramp_start(&trx->bts->acc_ramp);
> your first ramp step is executed before the call to gsm_bts_trx_set_system_
Indeed, thanks for catching this!

I think we can solve this by again calling acc_ramp_init() when we enter this 
function, to ensure ACC barring overrides take effect in the initial system 
infos bootstrap_rsl() sends via gsm_bts_trx_set_system_infos().

And then start the actual ramping process at the end of bootstrap_rsl(). This 
call might update system infos again if needed, but that's what we do as part 
of each ramping step anyway.

Calling acc_ramp_init() in bootstrap_rsl() will also help in case ACC ramp VTY 
configuration has changed when a BTS reconnects, which happens without going 
through bootstrap_bts() again as far as can I tell.


https://gerrit.osmocom.org/#/c/6324/2/src/libbsc/bsc_vty.c
File src/libbsc/bsc_vty.c:

PS2, Line 3135: enabled (0|1)
> while we have some commands like this dating back to the early days, it is 
Changed to work as suggested (Switch/Router style) in the next patch set. The 
commands which modify step size and interval were also changed slightly as a 
side effect.


https://gerrit.osmocom.org/#/c/6324/2/src/libbsc/system_information.c
File src/libbsc/system_information.c:

Line 662: static void apply_acc_ramp(struct gsm48_rach_control *rach_control, 
struct acc_ramp *acc_ramp)
> I think that should be part of the acc related code, not here?  At that poi
In the next patch set I am moving this to acc_ramp.h as an inline function, 
named as suggested.


-- 
To view, visit https://gerrit.osmocom.org/6324
To unsubscribe, visit https://gerrit.osmocom.org/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I0a5ac3a08f992f326435944f17e0a9171911afb0
Gerrit-PatchSet: 2
Gerrit-Project: osmo-bsc
Gerrit-Branch: master
Gerrit-Owner: Stefan Sperling <ssperl...@sysmocom.de>
Gerrit-Reviewer: Harald Welte <lafo...@gnumonks.org>
Gerrit-Reviewer: Jenkins Builder
Gerrit-Reviewer: Stefan Sperling <ssperl...@sysmocom.de>
Gerrit-Reviewer: Vadim Yanitskiy <axilira...@gmail.com>
Gerrit-HasComments: Yes

Reply via email to