Attention is currently required from: laforge, neels.
16 comments:
File include/osmocom/hnbgw/hnbgw.h:
Patch Set #6, Line 384: /* When rules to count traffic to and from this hNodeB are present, this reflects the state in nftables
Done
ACK.
Patch Set #6, Line 394: struct osmo_sockaddr_str addr_remote;
libnftables interfacing is string based, as the patch should show in other places
Done
Patch Set #6, Line 402: struct {
(in an earlier patch set i actually had the two combined in one struct.) […]
Oh I didn't mean joining the dl and ul. My proposal was to simply make the struct named so it can be declared one, used twice. That also makes it possible to pass a pointer to it.
File include/osmocom/hnbgw/nft_kpi.h:
Patch Set #6, Line 21: int nft_kpi_hnb_start(struct hnb_persistent *hnbp, const struct osmo_sockaddr_str *gtpu_remote);
nftables uses strings (also all logging), so it makes sense to use strings here. […]
Done
File src/osmo-hnbgw/hnbgw.c:
Patch Set #6, Line 346: hnb_persistent_disconnected(ctx->persistent);
(or "active" and "inactive", s.a. […]
Yeah, whatever, just give it a second though :)
exactly what i mean above. But this patch is not about fixing the HNBAP / HNB activity tracking. […]
Ack
File src/osmo-hnbgw/hnbgw_hnbap.c:
Patch Set #6, Line 571: hnb_persistent_connected(ctx->persistent);
yes, but how osmo-hnbgw handles HNBAP is weird. […]
Can actually an HNB be registered without being connected? I don't think so.
My understanding is that a HNBAP link disconnect makes the HNB implicitly become "unregistered".
File src/osmo-hnbgw/hnbgw_vty.c:
Patch Set #6, Line 884: DEFUN(cfg_hnbgw_nft_kpi, cfg_hnbgw_nft_kpi_cmd,
I wanted to, but it does not exist AFAICT: […]
Ah then it's probably the default. DONE.
File src/osmo-hnbgw/nft_kpi.c:
Patch Set #6, Line 86: static __thread struct nft_thread *g_nft_thread = NULL;
the nft_thread could be launched any number of times. […]
Ok, I'm not really a fan of having this sort of generic thread which may or may not do stuff, makes the architecture more open for no good reason, but fine.
I'd prefer having 2 (actually 3 counting the main thread) distinct threads doing specific stuff with specific context, but if you like it the other way we can keep it.
Patch Set #6, Line 120: LOGP(DNFT, LOGL_ERROR, "error running nft cmd %s#%u: rc=%d cmd=%s\n",
yes we have the rc that is carried back to the main thread in the queued response. […]
Done
Patch Set #6, Line 182: enum nft_thread_req_type {
can, but this is the part of the file that defines the thread request queue, so i'd like this to be […]
Done
Patch Set #6, Line 207: struct {
hence the naming yes
Done
Patch Set #6, Line 251: OSMO_ASSERT(g_nft_thread);
Why allocate it using talloc? […]
Done
Patch Set #6, Line 297: static void nft_t2m_enqueue(struct nft_thread *t, struct nft_thread_req *req)
The terms I use so far are "main()" and "(nft-)thread", because main() isn't started by pthread_crea […]
Done
Patch Set #6, Line 415: static int update_ctr(struct rate_ctr_group *cg, int cgidx, uint64_t *last_val, uint64_t new_val)
but the second thought made you realize that there is a cg counter group involved, right =)
Yes, after looking at the surrounding code. Just sharing that we use "ctrg" in lots of places already, so it's easier for osmocom users to understand right away.
Patch Set #6, Line 452: size_t want = g_nft_thread->counters_len + 64;
I used \*2 in an array list for a diff algorithm implementation in the https://gameoftrees. […]
Done
To view, visit change 36539. To unsubscribe, or for help writing mail filters, visit settings.