Home
release-team@conference.openafs.org
Friday, June 26, 2020< ^ >
kaduk@jabber.openafs.org/barnowl has set the subject to: openafs release team
Room Configuration
Room Occupants

GMT+0
[12:08:54] mbarbosa joins the room
[13:32:09] meffie joins the room
[15:03:57] cwills joins the room
[15:54:11] yadayada joins the room
[15:59:25] wiesand joins the room
[16:01:03] <meffie> good afternoon.
[16:01:15] <wiesand> Hello
[16:01:29] <yadayada> hi all
[16:01:48] <wiesand> New RT ticket: 135122
[16:01:50] <cwills> Good day
[16:02:38] <wiesand> not that urgent I think
[16:03:33] <wiesand> other recent ones still open: #135117 and #135109
[16:04:24] <wiesand> the latter is a build error on EL6, not that urgent either IMO
[16:05:22] <wiesand> I'm not sure what the impact of #135117 is
[16:07:32] <wiesand> any insights?
[16:08:04] <yadayada> 135117 need to reproduce it. But in case directory vnode becomes orphan will that vnode can be used later on? Or some component does some garbage collection. Need to check on this.
[16:09:29] <meffie> thanks yadayada
[16:09:29] <wiesand> I figure it will stay around but no longer be visible in the parent directory?
[16:13:24] <wiesand> My other topic for today is obviously the 1.8.6 release
[16:14:14] <wiesand> I lumped the version strings and the trivial NEWS update in gerrit 14253
[16:14:23] <kaduk@jabber.openafs.org/barnowl> I was mostly assuming that 135117 would be a resource leak, but am not
sure
[16:17:01] <wiesand> Any showstoppers regarding 1.8.6?
[16:17:37] <meffie> not that i've heard
[16:17:39] <kaduk@jabber.openafs.org/barnowl> I don't remember hearing about anything
[16:18:08] <yadayada> I want to put forward one issue I saw yesterday. Not sure if it falls in showstopper but want to put it forward
[16:18:23] <kaduk@jabber.openafs.org/barnowl> please do
[16:18:30] <wiesand> go ahead
[16:20:33] <yadayada> I saw if we have rx connection struct with ref count > 0 and if ref count remains > 0 for longer time. NAT KeepAlive events result in RX version packet, and there is a case where RX version pkts are sent recursively and we flood the wire with these packets. Important thing is it does not wait for 20 sec timewait, but just flood wire with these packets
[16:21:23] <yadayada> so on one occasion ~520000 RX version packets were sent in 1 min
[16:22:10] <yadayada> I think issue is with 1.8 codebase, now this recrsion will end when ref count on connection drops to 0, But if that takes time wire is flooded with version packets.
[16:22:53] <yadayada> still in middle of debugging, will send some update and probable fix on gerrit by early next week
[16:23:36] <wiesand> We had such an issue early in the 1.6 series, and yes it could be serious.
[16:24:08] <meffie> ouch.
[16:25:26] <kaduk@jabber.openafs.org/barnowl> I guess we don't know enough yet to hold the release for, though?
[16:27:24] <yadayada> so to me it looks this issue will be more visible in case some connection structure had refernce on it for longer time. That might be rare, but if that happens we will flood n.w. As you rightly said we need to debug it further, so in case not much users have seen issue on 1.8 , we can go ahead with release.
[16:28:19] <wiesand> as long as we're reasonably confident that it's not a regression in 1.8.6
[16:28:56] <yadayada> I need to verify that, at early stage in debugging at present. Will send more details ASAP
[16:36:39] <wiesand> Hmm, there are a few rx related changes since 1.8.5. Ok, let's play it safe and wait with the release until we're sure it's not a regression.
[16:37:31] <wiesand> I don't see excessive NAT pings on my 1.8.6pre3 test system, but that wasn't used a lot lately.
[16:38:20] <yadayada> sure will drop some updates by tommorow to release team, whether it is regression in 1.8.6 or was there from start of 1.8.0
[16:38:35] <wiesand> thanks!
[16:39:20] <kaduk@jabber.openafs.org/barnowl> The 1.8.5->1.8.6 rx changes don't seem like they would be relevant for
this
[16:39:20] <meffie> thank you yadayada
[16:40:25] <wiesand> I'll still merge 14253 on suspicion and prepare everything, but we should wait with the tag until we hear from yadayada.
[16:41:00] <kaduk@jabber.openafs.org/barnowl> okay
[16:43:19] <wiesand> On to 1.9/master?
[16:44:14] <kaduk@jabber.openafs.org/barnowl> 14239 has been getting updates/making progress
[16:45:00] <kaduk@jabber.openafs.org/barnowl> For linux 5.8 we are down to just 14249 left, I think.
[16:45:24] <cwills> linux 5.8-rc2 still compiles cleanly with 14249.  (Using gcc-9.3)
[16:45:30] <kaduk@jabber.openafs.org/barnowl> But mostly I am behind on reviews since my brain has been very slow
this week
[16:46:16] <kaduk@jabber.openafs.org/barnowl> I am not sure what I think about 14245 (coding style) yet, so it
might be worth more people weighing in.
[16:47:22] <cwills> bleeding edge clang will need some work ( compiler will not recognize /* fallthrough */ for switch/case statements )
[16:47:55] <meffie> heh, sorry ben!
[16:48:44] <kaduk@jabber.openafs.org/barnowl> I am kind of grumpy about the clang/fallthrough thing, but don't
expect it to actually change
[16:49:30] <meffie> yes, seems the purists have won that argument.
[16:49:44] <yadayada> Thanks for review  for 14239. But looks we will get more fixes for other locks. We are planning to submit more patches on same line.  
[16:49:50] <kaduk@jabber.openafs.org/barnowl> (Fallthrough comments have like a 20-year history via BSD lint(1))
[16:50:21] <cwills> yeah.. it seems like the clang folks have dug in their heels.  (btw .. ran into a similar issue with the musl library .. they refuse to define a macro that indicates you are using the musl headers)
[16:51:58] <cwills> So I will convert what I can to a new attribute.  Will have to add some ignore the warning for "built" code from flex and any externals
[16:58:55] <wiesand> Anything else to discuss today?
[16:59:06] <yadayada> nothing from my side
[16:59:08] <kaduk@jabber.openafs.org/barnowl> Not from me, sadly
[16:59:43] <meffie> just to mention the afs workshop site for 2021 was started.
[17:00:16] <meffie> i believe the openafs foundation will be sending out a 'save the date' announcement
[17:00:33] <meffie> http://workshop.openafs.org/afsbpw21/
[17:00:50] <yadayada> ok thanks mike for this information
[17:01:04] <meffie> you are welcome.
[17:01:29] <wiesand> I'll update the link text on the homepage
[17:01:53] <meffie> just link to http://workshop.openafs.org/ <http://workshop.openafs.org/afsbpw21/>
[17:02:16] <wiesand> Let's hope this one can actually happen. I'm not convinced…
[17:02:18] <meffie> and that will redirect to the current one. i'll add links to all of the old ones on a new page
[17:02:56] <meffie> i think it will happen. it might be a "virtual" workshop, worse case.
[17:04:17] <wiesand> The link target is already ok. So it's really just s/2020/2021/
[17:04:44] <meffie> Ah, ok, thank you wiesand.
[17:05:20] <wiesand> I may even smuggle that into the web change for the 1.8.6 release ;-)
[17:05:57] <meffie> sounds good to me.
[17:06:45] <wiesand> Looks like we're finished for today. Thanks a lot everyone!
[17:06:53] <yadayada> Thanks All
[17:07:10] <meffie> have a good weekend all.
[17:07:13] meffie leaves the room
[17:07:21] <kaduk@jabber.openafs.org/barnowl> Thanks everyone, have a good weekend
[17:07:26] wiesand leaves the room
[18:09:29] mbarbosa leaves the room
[18:09:40] mbarbosa joins the room
[18:11:24] yadayada leaves the room
[18:13:29] yadayada joins the room
[20:02:55] yadayada joins the room
[20:19:09] yadayada leaves the room
[20:27:29] yadayada leaves the room
[22:05:30] cwills leaves the room
[22:36:04] mbarbosa leaves the room