**** BEGIN LOGGING AT Fri Apr 03 02:59:56 2020
Apr 03 06:01:35 <Noltari> wigyori DonkeyHotei you should have founder permissions on #lede now
Apr 03 07:41:41 <aparcar> blogic: ping
Apr 03 07:46:41 <ldir> Mornin'
Apr 03 08:31:38 * Tapper waves at aparcarnblogic: ldir
Apr 03 08:31:57 <Tapper> How gos it people?
Apr 03 08:36:22 <Tapper> The bordem of this lockdown is sending me around the bend!
Apr 03 08:43:16 * ldir waves at Tapper 
Apr 03 08:44:56 <Tapper> Hi mate
Apr 03 08:45:17 <Tapper> I am just tacking a look at that chat client zoom
Apr 03 08:45:25 <Tapper> taking*
Apr 03 08:45:43 <Tapper> Mite get some of my mates on there later for a chin wag
Apr 03 08:49:27 <Tapper> Hows your day going so far ldir?
Apr 03 08:50:08 <ldir> Just about to get on the exercise bike... so a numb arse shortly
Apr 03 08:50:25 <rubberduck> Tapper: Have enough Toilet Paper so everything fine.
Apr 03 08:50:55 <Tapper> lol cant have a duck with a dirty bum!
Apr 03 08:51:17 <Tapper> I have never liked them bikes ldir
Apr 03 08:52:31 <Tapper> One time wen I was going to the gym 2 girls talked me in to going to a spinning class!
Apr 03 08:53:05 <Tapper> I did not go again. It was 2 days before I could work with out wincing!
Apr 03 08:53:10 <Tapper> walk*
Apr 03 08:53:10 <ldir> lol - yeah I'm not *that* enrgetic
Apr 03 08:53:25 <Tapper> Same here!
Apr 03 08:53:27 <Tapper> lol
Apr 03 08:53:46 <Tapper> I do need to go for a walk today tho.
Apr 03 08:54:14 <Tapper> There is a field not far from my house so I think I will take the wife and kids for a walk if there is no one around.
Apr 03 08:54:48 <ldir> good idea - you need to get out for a dose of fresh air
Apr 03 08:55:32 <Tapper> rubberduck I am not joking  the other day my wife drove past a block in a van and he was selling bog role out of the back for £1 a role lol
Apr 03 08:56:06 <Tapper> Bloke*
Apr 03 08:56:35 <rubberduck> in australia there is a provider for pandemy-survival-training - 750AUD or 1 1/2 roles of toilet paper fot the training (was offered two days ago)
Apr 03 08:57:27 <Tapper> ldir Yeah mate. I like to be out and about so it's hard for me. We are all in the same boat so I should not wine to much.
Apr 03 08:59:00 <Tapper> rubberduck haha It comes to something when people are shotting bog role like bags of weed!
Apr 03 08:59:31 <Tapper> "Yo man I got the good stuff dude reel soft if you get me!"
Apr 03 09:03:20 <Tapper> OMG ldir what do you think about this? The Walkers Crisps Conspiracy That Has People Convinced We're in an Alternate Reality https://www.vice.com/en_uk/article/3a8mzy/walkers-crisps-switch-colour-packet?utm_source=dlvr.it&utm_medium=twitter
Apr 03 09:03:31 <Tapper> hahaha I told you I am bord!
Apr 03 09:03:40 <Tapper> brb kids!
Apr 03 09:08:02 <ldir> Tapper: The Internet is like a herd of performing elephants with diarrhoea – massive, difficult to re-direct, awe-inspiring, entertaining, and a source of mind-boggling amounts of excrement when you least expect it.
Apr 03 09:08:02 <ldir> — Gene Spafford
Apr 03 09:11:20 <Tapper> ldir hahahahahahah I love that. thanks
Apr 03 09:11:56 <Tapper> I was made by my 6 year old to scramble some eggs.
Apr 03 09:24:37 * ldir updates qnap firmware - that's the rest of the day gone restoring to some form of functionality
Apr 03 09:26:29 <stintel> every time I reboot this hypervisor I hope something will magically fix my IO issues :P
Apr 03 09:26:50 <stintel> but I should really just migrate it to a box with SSD
Apr 03 09:35:01 <rubberduck> a ssd is a very very good idea
Apr 03 09:36:11 <stintel> yeah well it will almost double the monthly price of the server
Apr 03 09:36:23 <rubberduck> ok
Apr 03 09:36:27 <stintel> but that's how it is :)
Apr 03 09:36:42 <rubberduck> at my hoster you get ~ 1/4 of the storage when choosing ssd
Apr 03 09:36:53 <rubberduck> at same prive
Apr 03 09:36:56 <rubberduck> price
Apr 03 09:37:33 <rubberduck> i have a vServer with only 10 GByte Storage - but this felt like a big raid-6 built of seagate ST225 disks.
Apr 03 09:38:21 <rubberduck> after a crash of this disk array the hoster moved to a more capable host system, now a apt-update ; apt -y upgrade lasts only ~ 5 Minutes - before it could take up to 10 Hours
Apr 03 09:38:55 <rubberduck> i'm using Debian Buster and the vServer has <128 MBytes of RAM
Apr 03 09:40:59 * ldir stares at both qnaps and asks 'what are you doing?' - it can't take this long to reboot
Apr 03 09:43:15 <stintel> rubberduck: unfortunately if you're using btrfs, 128MB RAM is not enough
Apr 03 09:43:36 <f00b4r0> ldir: reader of web-heads? :)
Apr 03 09:44:01 <ldir> nope ?
Apr 03 09:44:17 <rubberduck> ext4
Apr 03 09:44:50 <rubberduck> 128mb is not enough to even boot buster with default settings - you need to modify the initramfs settings to get this less than 128MB unpacked
Apr 03 09:45:17 <f00b4r0> ldir: https://spaf.cerias.purdue.edu/web-heads/ - you quoted Spaf
Apr 03 09:46:10 <f00b4r0> stintel: what kind of workload are you running that cripples IO? (out of curiosity)
Apr 03 09:46:22 <ldir> I've always liked the 'herd of performing elephants' quote
Apr 03 09:46:34 <Tapper> Any one know what is going on here: my build crashes if I try and build luci-app-dawn because umdns will not build. https://github.com/openwrt/packages/issues/11746
Apr 03 09:46:42 <nick[m]2> I've added a package to Luci Repro. But it is not shown under downloads? I've looked for build bod erros under fail logs, but there is nothing... What could be wrong?
Apr 03 09:46:58 <ldir> oooh baby qnap has emitted a loud beep...and the disc lights flash... maybe it is alive
Apr 03 09:47:00 <nick[m]2> With "added" I mean I made a PR that was accepted on GitHub
Apr 03 09:47:01 <nick[m]2> Yesterday
Apr 03 09:47:24 <stintel> f00b4r0: it's a single box with spinning rust and lots of VMs with mysql, rrd, zimbra (which is again mysql and rrd)
Apr 03 09:47:41 <f00b4r0> ah, databases. Makes sense.
Apr 03 09:49:25 <stintel> yeah and it doesn't help that I'm doing master-master replication between 2 VMs on the same machine. but it reduces downtime. ideally I switch this setup to 2 servers with SSDs
Apr 03 09:49:40 <stintel> but I don't earn any money from it
Apr 03 09:50:25 <stintel> it's now EUR42/mo. switchting to a single server with SSD + HDD (I need more than the 480GB SSD) would be ~75/mo
Apr 03 09:50:48 <stintel> switching to 2 with just SSD, which would be enough, would be ~110/mo iirc
Apr 03 09:51:41 <stintel> but maybe it's time to bite the bullet and just do it, the slowness is getting really frustrating
Apr 03 09:53:24 <f00b4r0> stintel: is this VPS or dedicated?
Apr 03 09:53:31 <stintel> dedi
Apr 03 09:53:35 <f00b4r0> I see.
Apr 03 09:53:46 <f00b4r0> I'm using Contabo for VPS. Darn cheap.
Apr 03 09:54:45 <stintel> 15 VMs on this single box. 4c/8t 32GB RAM 2x2TB HGST Ultrastar 7K6000
Apr 03 09:57:11 <blogic> _lore_: ping
Apr 03 09:57:47 <_lore_> blogic: pong
Apr 03 09:58:00 <blogic> _lore_: (7fa)0x007fa|state=BIND|type=IPV4_HNAPT|176.9.48.73:443->192.168.178.110:49600=>176.9.48.73:443->192.168.1.197:49600|da:c8:2b:d2:cc:e5=>00:02:c9:50:d8:26|etype=0x0800|info1=0x21408059|info2=0x7ff020|vlan1=0|vlan2=0
Apr 03 09:58:04 <blogic> (9ac)0x009ac|state=BIND|type=IPV4_HNAPT|192.168.1.197:49600->176.9.48.73:443=>192.168.178.110:49600->176.9.48.73:443|b6:aa:a3:5c:44:8b=>e8:df:70:70:bd:4d|etype=0x0800|info1=0x2140005b|info2=0x7ff040|vlan1=0|vlan2=0
Apr 03 09:58:09 <blogic> :-D
Apr 03 09:58:11 <f00b4r0> stintel: I'd wager IO isn't your only bottleneck there ;)
Apr 03 09:58:24 <stintel> f00b4r0: it is
Apr 03 09:58:29 <blogic> that is the mt7622 flow offload hw table entry of me downloading a owrt sdk from the owrt server
Apr 03 09:58:42 <f00b4r0> you seem very memory starved for such a load. Esp if these are mostly database oriented VMs
Apr 03 09:58:42 <_lore_> cool :)
Apr 03 09:58:49 <_lore_> very nice
Apr 03 09:58:50 <blogic> _lore_: just plugged the unit in band of my uplink router
Apr 03 09:59:00 <blogic> yup, i am preparing patches from trunk just now
Apr 03 09:59:05 <blogic> what about the pci issue
Apr 03 09:59:11 <blogic> i could look into it today
Apr 03 09:59:25 <blogic> is there still need to patche the driver or have you handled it ?
Apr 03 09:59:50 <f00b4r0> stintel: I'd guess no caching can happen, so all io actually hits the disk?
Apr 03 10:00:42 <Tapper> stintel at that price would it not work out cheeper to just by a server? forgive me if that's a daft thing to say!
Apr 03 10:01:20 <_lore_> I looked at it a little bit but I have not found a nice solution so far
Apr 03 10:01:27 <blogic> ok
Apr 03 10:01:36 <blogic> so msi shall be per slot and not controller
Apr 03 10:01:42 <blogic> do you have a mt7615 card ?
Apr 03 10:01:44 <stintel> f00b4r0: there isn't much room for caching in the hypervisor but that's fine, you don't want that anyway
Apr 03 10:02:02 <_lore_> the issue is:
Apr 03 10:02:19 <stintel> Tapper: sure, but it's cheaper to rent this server at OVH than just hosting a server you bought in anything close to where I live
Apr 03 10:02:33 <_lore_> in the current driver the 'inner_domain' and 'msi_domain' is per-port and not per controller
Apr 03 10:03:02 <stintel> f00b4r0: the other VMs have room for fs cache, except the database VMs they use most of the RAM for their own buffers which is also fine
Apr 03 10:03:18 <Tapper> stintelbecause of the upstreem bandwidth?
Apr 03 10:03:48 <_lore_> but the pci_device when it is discovered is associated to the bus domain that is the last added domain
Apr 03 10:03:48 <stintel> Tapper: that's one of the problems, yes. hetzner or ovh just say w/e you have 20TB traffic
Apr 03 10:04:08 <_lore_> so even if the device is on slot0 it is detected on slot 1 and it crashes
Apr 03 10:04:27 <_lore_> so to trigger the issue you need to use slot0 and not slot1
Apr 03 10:05:34 <stintel> f00b4r0: probably doesn't help that I'm doing btrfs in most of the VMs ;)
Apr 03 10:05:40 <f00b4r0> probably not
Apr 03 10:05:57 <f00b4r0> but still. 15VMs with only 32GB, that's asking for low perf anyway ;)
Apr 03 10:06:10 <blogic> _lore_: ok, will debug that
Apr 03 10:06:16 <_lore_> feel free to work on it if yo want I am a bit busy with mt7663 now
Apr 03 10:06:21 <blogic> _lore_: do you have a mt7615 card and could test something ?
Apr 03 10:06:22 <stintel> f00b4r0: these are VMs with musl libc. some of them have <512MB RAM and use like 50
Apr 03 10:06:30 <_lore_> sure
Apr 03 10:06:42 <_lore_> I have a mt7663 one
Apr 03 10:06:44 <blogic> i have a elecom board with mt7622
Apr 03 10:06:51 <_lore_> this is the only one it fits in slot0
Apr 03 10:06:59 <_lore_> this is way I discovered the issue just now
Apr 03 10:07:00 <blogic> there is a mt7615 soldered onto the pcb
Apr 03 10:07:13 <_lore_> in your board?
Apr 03 10:07:17 <blogic> the calibration data is located in mtd and not OTP
Apr 03 10:07:20 <blogic> yes
Apr 03 10:07:25 <_lore_> I have a different one
Apr 03 10:07:28 <blogic> and channel 36 works fine
Apr 03 10:07:33 <blogic> channel 100 i cannot assoc
Apr 03 10:07:41 <blogic> and channel 149 i get 10mbit throughput
Apr 03 10:07:42 <f00b4r0> stintel: i see. Now I'm curious. This sounds like embedded environment, and yet database load? I'm not sure how to reconcile the two? ;)
Apr 03 10:08:03 <blogic> I am trying to figure out if that is because calibration data is not uploaded properly or what the iussue is
Apr 03 10:08:04 <_lore_> http://www.banana-pi.org/r64.html
Apr 03 10:08:10 <stintel> f00b4r0: musl doesn't mean embedded
Apr 03 10:08:15 <Tapper> who is the maintainer of umdns?
Apr 03 10:08:19 <blogic> Tapper: me
Apr 03 10:08:22 <_lore_> this is the very similar to the one I am using now
Apr 03 10:08:25 <f00b4r0> stintel: 512MB RAM does.
Apr 03 10:08:27 <f00b4r0> ;)
Apr 03 10:08:41 <stintel> f00b4r0: it's a conscious choice due to lower memory footprint
Apr 03 10:08:42 <blogic> _lore_: could you test if mt7615 pci/e cards work right now with the version in owrt trunk on channel 100 / 149
Apr 03 10:09:07 <Tapper> https://github.com/openwrt/packages/issues/11746
Apr 03 10:09:08 <blogic> i have a feeling that the calibration data is not uploaded to fw properly on this board
Apr 03 10:09:09 <_lore_> sure
Apr 03 10:09:29 <blogic> just want to know if this is specific to the board or happens on all 7615 boards
Apr 03 10:09:30 <stintel> f00b4r0: example: I have 1 VM running an MQTT broker, just that. I assigned it 384MB RAM (because with 256MB btrfs eventually caused oom)
Apr 03 10:09:32 <blogic> awesome !
Apr 03 10:09:55 <blogic> Tapper: i looked at dawn
Apr 03 10:09:56 <Tapper> I was trying to build dawn but umdns is a dep
Apr 03 10:10:01 <_lore_> blogic: but do you mean the on chip radio?
Apr 03 10:10:08 <blogic> yes
Apr 03 10:10:17 <_lore_> ah ok, so it is a mt7622 not mt7615
Apr 03 10:10:18 <blogic> so the mt7622 board has 2 radios
Apr 03 10:10:23 <_lore_> but the driver is the same
Apr 03 10:10:35 <stintel> f00b4r0: then one of them is running this irssi instance, bitlbee and matterircd, assigned it 512MB RAM because it is enough
Apr 03 10:10:37 <blogic> the 2,4g builtin and mt7615 pci/e 5G
Apr 03 10:10:44 <blogic> both essentially use the same driver
Apr 03 10:10:51 <blogic> the 2,4g works like a charm
Apr 03 10:10:57 <_lore_> ah, so IIUC your board has even a 5GHz radio onboard
Apr 03 10:11:00 <blogic> the 5g only on channel 36
Apr 03 10:11:20 <_lore_> right?
Apr 03 10:11:26 <blogic> 1 sec
Apr 03 10:11:50 <f00b4r0> stintel: ok. I assume all these VMs are using disk images and not raw devices, since you have only two disks. In which case, brtfs (or any journaled fs) is clearly a bad choice
Apr 03 10:12:31 <f00b4r0> even if it only enforces metadata consistency, that's going to swarm io
Apr 03 10:12:32 <stintel> f00b4r0: LVM backed virtual storage - the btrfs is in the VMs itself because snapshots and checksums
Apr 03 10:13:19 <blogic> _lore_: correct there is a dedicated pcie chip soldered
Apr 03 10:14:21 <blogic> Tapper: i looked at dawn
Apr 03 10:14:27 <blogic> Tapper: the idea is nice
Apr 03 10:14:36 <blogic> Tapper: the code is pretty nasty
Apr 03 10:14:54 <blogic> like drunk student in his first years level code
Apr 03 10:15:20 <_lore_> ack, I will try it soon
Apr 03 10:15:32 <blogic> _lore_: thanks
Apr 03 10:15:41 <f00b4r0> stintel: not sure why you'd want snapshot and csum inside the VM but ok :)
Apr 03 10:15:46 <blogic> i have bpi64 but i cannot find my mt7615 card :-D
Apr 03 10:16:06 <_lore_> btw regarding pcie I guess the right approach would be define the msi/inner domain per port as it is ... but the pci_dev should be detected on that domain
Apr 03 10:16:10 <_lore_> not on bus one
Apr 03 10:16:21 <blogic> _lore_: correct
Apr 03 10:16:24 <blogic> will look into that
Apr 03 10:16:28 <_lore_> thx
Apr 03 10:17:11 <stintel> f00b4r0: you want csum everywhere. silent data corruption is real
Apr 03 10:17:30 <f00b4r0> depends on the criticality of your workload I guess
Apr 03 10:17:49 <stintel> f00b4r0: snapshots ... difficult story. if you want that outside of the VM with libvirt you need to use qcow2 images afaik
Apr 03 10:17:56 <f00b4r0> the internet worked for eons without csuming, and the world didn't collapse ;)
Apr 03 10:18:09 <f00b4r0> (storage csums that is)
Apr 03 10:18:44 <f00b4r0> yes qcow2 makes that "easier", although iirc if you use lvm you can do it anyway
Apr 03 10:18:47 <stintel> you can take lvm snapshots but it would need some extra action to make sure the snapshot is going to be in a consistent state
Apr 03 10:18:55 <f00b4r0> *nod*
Apr 03 10:18:58 <stintel> or shut down the vm :P
Apr 03 10:19:07 <zorun> or just rely on the journal :p
Apr 03 10:19:11 <f00b4r0> ^
Apr 03 10:19:15 <stintel> famous last words :P
Apr 03 10:19:18 <f00b4r0> heheh
Apr 03 10:19:20 <zorun> :)
Apr 03 10:19:46 <zorun> that's the way I do backups on the hypervisor: snapshot the thing, mount the snapshot, backup it, destroy the snapshot
Apr 03 10:19:56 <zorun> it needs to recover the journal but it works
Apr 03 10:20:15 <zorun> I mean, backup containers from the hypervisor
Apr 03 10:20:18 <stintel> interesting strategy
Apr 03 10:20:29 <zorun> it's containers and not VM, that might make a difference
Apr 03 10:21:21 <zorun> but from the point of view of LVM there's no difference
Apr 03 10:23:03 <f00b4r0> stintel: at the end of the day I guess for you it's a tradeoff between how critical you deem these VMs to be and how much you're willing to pay to accommodate the overhead of your setup ;)
Apr 03 10:28:19 <stintel> f00b4r0: I'd say mail and IRC are mission critical for me :P
Apr 03 10:28:43 <f00b4r0> I doubt they'd suffer from a bit flip ;)
Apr 03 10:28:46 <stintel> so yeah maybe I should split it to 2 SSD servers and then start charging the few people that use my mail server :P
Apr 03 10:29:29 <f00b4r0> that you should. Advertise privacy ;)
Apr 03 10:31:35 <Tapper> blogic It's not dawn that braks my builds it's umdns.
Apr 03 10:31:53 <Tapper> breaks* or brakes*
Apr 03 10:32:03 <Tapper> I don't know which one lol
Apr 03 10:32:13 <Tapper> Or  both.
Apr 03 10:33:23 <Tapper> blogic can you take a quick look at my link pleas. I have put the build log in there.
Apr 03 10:33:39 <Tapper> https://github.com/openwrt/packages/issues/11746
Apr 03 10:36:28 <f00b4r0> Tapper: if your brakes break you'll break your neck :)
Apr 03 10:37:16 <Tapper> Thanks I will try and remember that one!
Apr 03 10:38:57 <Tapper> Small tricks like can help a lot if you can store them away in your head.
Apr 03 10:39:10 <Tapper> like that*
Apr 03 10:41:41 <Tapper> If i close my issue will that link still work?
Apr 03 10:46:54 <DonkeyHotei> yes the link will still work
Apr 03 10:48:26 <Tapper> DonkeyHotei thanks mate
Apr 03 12:31:29 <_lore_> blogic: I run some test on device I have here and I found the dfs channels work not fine
Apr 03 12:31:41 <_lore_> not sure if it is a calibration issue
Apr 03 12:32:12 <_lore_> probably
Apr 03 12:32:51 <_lore_> values I get in 5180-5200 band are align to band 5700-5825
Apr 03 12:33:05 <_lore_> dfs bands are lower
Apr 03 12:36:42 <ldir> Tapper: I take it you're trying to compile with gcc 9
Apr 03 12:53:19 <Tapper> ldir yeah mate.
Apr 03 12:53:47 <Tapper> ldir should I not use gcc 9?
Apr 03 12:54:22 <ldir> gcc 9 provokes a warning with that code.  all warnings are failures so it fails.  gcc 8 is less pedantic.
Apr 03 12:55:07 <ldir> I'm trying to work out how to fix it, but I'll be honest and say my brain, C and pointers are incompatible...and generate warnings of their own :-)
Apr 03 13:01:02 <blogic> _lore_: ok, so its a generic mt7615 pci issue then
Apr 03 13:01:06 <blogic> and not specific to the board
Apr 03 13:01:28 <_lore_> no, I do not think so
Apr 03 13:01:36 <_lore_> I can try with another device later
Apr 03 13:01:51 <blogic> so generic mt7615 issue it is then
Apr 03 13:02:15 <_lore_> I think it would be due to bad calibration data
Apr 03 13:02:27 <blogic> well
Apr 03 13:02:37 <blogic> using an SDK image on the board everything works like a charm
Apr 03 13:02:44 <_lore_> ah :)
Apr 03 13:02:45 <_lore_> ok
Apr 03 13:02:57 <blogic> the board and calibration is all good
Apr 03 13:03:10 <blogic> so for you the upper channels worked ?
Apr 03 13:03:20 <_lore_> is the sdk using calibration data from the mtd or from a bin file
Apr 03 13:03:21 <_lore_> ?
Apr 03 13:03:37 <_lore_> since AFAIK even this cabality is supported by sdk
Apr 03 13:04:01 <Tapper> ldir OK I will try a build with gcc 8 in a bit after food. I will let you know if it works
Apr 03 13:04:29 <blogic> bin i assume
Apr 03 13:04:52 <_lore_> so there is a separated file with calibration data
Apr 03 13:05:01 <blogic> https://pastebin.com/2bjS99kp
Apr 03 13:05:01 <_lore_> right? they are not the same used by mt7615
Apr 03 13:05:04 <_lore_> right?
Apr 03 13:05:15 <blogic> ?
Apr 03 13:05:26 <_lore_> no, ok
Apr 03 13:05:36 <blogic> that mt7622 works
Apr 03 13:05:36 <_lore_> I opened the pastebin
Apr 03 13:05:45 <blogic> so the 26 mt7622 wifi works
Apr 03 13:05:51 <blogic> the mt7615 pci wifi fails
Apr 03 13:06:34 <blogic> the mtd only has calibration data for mt7622
Apr 03 13:06:48 <blogic> so for mt7615 i just load the fw and rompatch
Apr 03 13:06:53 <blogic> mt7615_cr4.bin
Apr 03 13:06:56 <blogic> mt7615_n9.bin
Apr 03 13:06:59 <_lore_> sorry
Apr 03 13:07:00 <blogic> mt7615_rom_patch.bin
Apr 03 13:07:01 <_lore_> one moment
Apr 03 13:07:08 <blogic> sure
Apr 03 13:07:14 <_lore_> mt7622 loads data from mtd partition
Apr 03 13:07:19 <blogic> yes
Apr 03 13:07:20 <_lore_> and it works fine
Apr 03 13:07:23 <blogic> correct
Apr 03 13:07:31 <_lore_> mt7615 does not have calibration data, right?
Apr 03 13:07:37 <blogic> correct
Apr 03 13:07:48 <_lore_> this is a problem for sure
Apr 03 13:07:56 <_lore_> and it can cause the issue in your case
Apr 03 13:08:01 <blogic> how so ?
Apr 03 13:08:07 <blogic> i thought it is self calibrating
Apr 03 13:08:12 <_lore_> nope
Apr 03 13:08:25 <_lore_> there is a calibration partition like for mt7622
Apr 03 13:08:42 <_lore_> e.g. un pcie card it is in efuse partition
Apr 03 13:09:04 <_lore_> in your case it should be in an mtd partition
Apr 03 13:09:21 <_lore_> do you have a pcie card?
Apr 03 13:09:30 <blogic> nope
Apr 03 13:09:34 <_lore_> ok
Apr 03 13:09:58 <_lore_> I tried with a device like yours (mt7621 + mt7615) but I am not sure about calibration data
Apr 03 13:10:15 <_lore_> I will retry with a mt7615 pcie card and I will get back to you
Apr 03 13:10:21 <blogic> i would have assumed that the calibration data is in the eep
Apr 03 13:10:28 <blogic> or the OTP
Apr 03 13:10:48 <_lore_> yes, is your device has otp?
Apr 03 13:10:58 <blogic> how do i find out ?
Apr 03 13:11:05 <_lore_> e.g.
Apr 03 13:11:10 <_lore_> the mac address is constant?
Apr 03 13:11:19 <_lore_> or randomly generated
Apr 03 13:12:27 <_lore_> if your device has OTP it is ok
Apr 03 13:12:47 <blogic> let me check
Apr 03 13:12:47 <blogic> constant
Apr 03 13:12:52 <blogic> constant
Apr 03 13:12:55 <_lore_> ok
Apr 03 13:12:58 <blogic> stays the same over a reboot
Apr 03 13:13:46 <_lore_> https://github.com/openwrt/mt76/blob/master/mt7615/eeprom.c#L43
Apr 03 13:13:56 <_lore_> this is where the OTP is read
Apr 03 13:14:16 <blogic> let me printk it
Apr 03 13:14:19 <_lore_> ok
Apr 03 13:15:40 <_lore_> anyway the eeprom data is accessible from deebugfs
Apr 03 13:15:46 <_lore_> *debugfs
Apr 03 13:16:18 <_lore_> /sys/kernel/debug/ieee80211/phy,X>/mt76/eeprom
Apr 03 13:16:23 <_lore_> /sys/kernel/debug/ieee80211/phy<X>/mt76/eeprom
Apr 03 13:17:22 <blogic> https://pastebin.com/kdStsWvm
Apr 03 13:17:25 <blogic> yup its there
Apr 03 13:19:38 <_lore_> ok, I think it is correct
Apr 03 13:20:08 <_lore_> can you check how the system is working on 5800 band?
Apr 03 13:20:16 <_lore_> like 5825
Apr 03 13:23:16 <blogic> 5805 works
Apr 03 13:23:20 <blogic> with VHT40
Apr 03 13:23:38 <_lore_> ok
Apr 03 13:23:48 <_lore_> I got the same results more or less
Apr 03 13:24:18 <_lore_> I will look into it
Apr 03 13:24:37 <blogic> great
Apr 03 13:26:00 <gch981213> Hi all! I'm about to start a final review and merge the 5.4 kernel for mt7621.
Apr 03 13:26:00 <gch981213> But I'm still hesitating about switching to upstream DSA in that PR:
Apr 03 13:26:00 <gch981213> 1. mac table (aka forwarding database) isn't properly synced between host bridge and switch:
Apr 03 13:26:00 <gch981213>    https://github.com/openwrt/openwrt/pull/2798#issuecomment-602221971
Apr 03 13:26:00 <gch981213>    When a client moves between switch ports and host interfaces, it got ignored for about a minute.
Apr 03 13:26:01 <gch981213> 2. switching to DSA introduces config incompatibility from old version.
Apr 03 13:26:01 <gch981213> 3. I think DSA tagged frames won't be understood by MTK PPE and we lost HWNAT possibility.
Apr 03 13:26:02 <gch981213> Is it okay to apply the DSA change considering all these issues or should I take the swconfig driver from mediatek target instead?
Apr 03 13:32:35 <adrianschmutzler> what are the pros for dsa in this particular case, why has it been chosen in the first place?
Apr 03 13:32:48 <f00b4r0> losing hwnat will seriously kill performance :(
Apr 03 13:33:21 <adrianschmutzler> despite it being upstream
Apr 03 13:33:50 <gch981213> adrianschmutzler: I'm about to say "It's from upstream" :P
Apr 03 13:35:16 <adrianschmutzler> so it's more the general swconfig vs. DSA discussion than really particularly a discussion required by mt7621? (despite for the target-specific problems you named)
Apr 03 13:35:31 <f00b4r0> I'd wager not everything from upstream is necessarily good for owrt. My understanding is upstream doesn't want to hear about forwarding shortcut, yet even google found it good enough to use it on their devices
Apr 03 13:38:14 <gch981213> The other one pro I could come up with: it seems impossible to support SFP port on UBNT ER-X with swconfig.
Apr 03 13:40:23 <gch981213> adrianschmutzler: Yes.
Apr 03 13:41:29 <adrianschmutzler> how painful would it be to port the swconfig driver? I'm just asking myself whether it would be better to not do both at one time. So, 5.4 with swconfig first, and then DSA separately, so one could go back easier ...
Apr 03 13:42:28 <gch981213> adrianschmutzler: It seems that this discussion has happend a long time ago: https://openwrt.org/docs/techref/swconfig
Apr 03 13:42:54 <gch981213> Quote from that page: "It is considered legacy and new switch drivers should use the DSA"
Apr 03 13:43:12 <adrianschmutzler> the question is who wrote it there.
Apr 03 13:43:30 <adrianschmutzler> I assume you wouldn't ask here if you were sure about it ;-)
Apr 03 13:44:04 <adrianschmutzler> and particularly the loss of hwnat will be a big loss at least in perception.
Apr 03 13:44:17 <gch981213> adrianschmutzler: It's pretty easy to switch to a swconfig driver.
Apr 03 13:44:23 <jow> please go with DSA, its the only way forward
Apr 03 13:45:25 <adrianschmutzler> I'm too new to decide on the way to go here. But if it's easy, I'd prefer to have it two-step.
Apr 03 13:48:26 <jow> sticking with swconfig just delays the inevitable, eventually we do have to switch to DSA and need to address its deficiencies upstream
Apr 03 13:49:11 <jow> if any work is spent on swconfig at all, it should be spent to port its drivers to DSA where possible, or work on a swconfig DSA backend driver
Apr 03 13:49:47 <jakllsch> what doesn't DSA do that swconfig can?
Apr 03 13:50:08 <adrianschmutzler> hwnat
Apr 03 13:50:18 <gch981213> jow: Okay. Could I add your Acked-by in related commits?
Apr 03 13:50:27 <jow> gch981213: feel free to
Apr 03 13:50:36 <jow> jakllsch: multiple cpu ports iirc
Apr 03 13:50:53 <jakllsch> ah.  so, nothing I care about. :-)
Apr 03 13:50:54 <gch981213> jow: Thanks.
Apr 03 13:51:17 <jow> well the simple case (bridge 4 ports together) works in both
Apr 03 13:51:38 <jow> its the advanced features that might be problematic
Apr 03 13:51:56 <jow> and various switch specific features
Apr 03 13:52:23 <jow> not sure how mirror ports, eee and such stuff is handled in dsa
Apr 03 13:52:27 <jow> or mib counters
Apr 03 13:52:32 <jakllsch> oh.
Apr 03 13:52:38 <jow> I guess the latter must be shoehorned into ethtool somehow
Apr 03 13:55:13 <jow> and someone once explained me that DSA requires the hardware to inject port specific tags so that it is able to relate incoming ethernet frames to the switch ports they came from
Apr 03 13:55:26 <blogic> are we moving ralink to dsa ?
Apr 03 13:55:34 <jow> not every currently swconfig supported hw adds such proprietary tags
Apr 03 13:55:56 <gch981213> jow: Correct. And that means throwing away all hw offloads on ethernet side.
Apr 03 13:55:59 <jow> on such hw one would need to emulate that somehow by programming port based vlans internally
Apr 03 13:56:13 <gch981213> blogic: mt7621 only.
Apr 03 13:56:15 <blogic> ralink hw offload works with DSA
Apr 03 13:56:29 <blogic> why not esw ?
Apr 03 13:56:39 <blogic> i thought there was dsa driver now
Apr 03 13:56:50 <blogic> from stefan rose
Apr 03 13:57:10 <blogic> I am actually getting the mediatek hw offload driver ready for usptream right now
Apr 03 13:57:21 <blogic> been working on it last 3 days
Apr 03 13:58:32 <gch981213> blogic: The existing ramips pull request is only adding 5.4 for mt7621. I'll work on mt7620/mt7628/rt305x later but my plan is to port the old driver to 5.4.
Apr 03 13:58:46 <blogic> ok
Apr 03 14:00:18 <gch981213> blogic: Oh really? I thought PPE won't be able to receive and send those frames with proprietary tags.
Apr 03 14:01:40 <blogic> it can easily
Apr 03 14:01:53 <blogic> there is a register that has the tag mask
Apr 03 14:02:01 <blogic> right now it uses vlan
Apr 03 14:02:13 <blogic> it basically treats the dsa tag as the outer vlan_tag1
Apr 03 14:20:47 <dengqf6> blogic: I have sent a patch to add port mirroring support for MT7530 DSA
Apr 03 14:22:03 <dengqf6> Should I backport it here
Apr 03 14:22:06 <dengqf6> ?
Apr 03 14:22:33 <blogic> once owrt is on v5.4
Apr 03 14:23:55 <gch981213> blogic: dengqf6 is the author of that v5.4 PR :D
Apr 03 14:24:04 <blogic> ah ok, sorry
Apr 03 14:25:02 <dengqf6> I mean  should I include that port mirror patch in the PR?
Apr 03 14:25:21 <blogic> sure why not
Apr 03 15:23:49 <dengqf6> blogic: are you still working on qca8k hwnat?
Apr 03 15:24:04 <blogic> dengqf6: sort of yes
Apr 03 15:24:16 <blogic> i just passed all my patches on to someone to finalize it
Apr 03 15:28:26 <dengqf6> How to use port mirroring in DSA: https://github.com/Mellanox/mlxsw/wiki/Port-Mirroring
Apr 03 15:59:34 <dengqf6> blogic: where can I get them?
Apr 03 16:01:08 <tmn505> Hauke: Can You take a look at mvebu kernel 5.4 PR https://github.com/openwrt/openwrt/pull/2804?
Apr 03 16:42:00 <blogic> dengqf6: i will upload them to my github the next days
Apr 03 18:21:15 * f00b4r0 dumps a bit more code to the m-l
Apr 03 18:37:17 <owrt-snap-builds> build #309 of layerscape/armv8_64b is complete: Failure [failed kmods]  Build details are at http://buildbot.openwrt.org/master/images/builders/layerscape%2Farmv8_64b/builds/309  blamelist: Rafa? Mi?ecki <rafal@milecki.pl>, Tim Harvey <tharvey@gateworks.com>, Paul Spooren <mail@aparcar.org>, Tan Zien <nabsdh9@gmail.com>
Apr 03 18:54:59 <owrt-snap-builds> build #322 of ramips/rt3883 is complete: Failure [failed kmods]  Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Frt3883/builds/322  blamelist: Rafa? Mi?ecki <rafal@milecki.pl>, Tim Harvey <tharvey@gateworks.com>, Paul Spooren <mail@aparcar.org>, Tan Zien <nabsdh9@gmail.com>
Apr 03 22:16:58 <Toomoch> hi, how can I backup all my flash if I have booted an openwrt ramdisk?
Apr 03 22:19:14 <PaulFertser> Toomoch: what type of flash?
Apr 03 22:19:30 <Toomoch> nand I belive
Apr 03 22:20:50 <Toomoch> PaulFertser: yes, confirmed nand (https://zentel-europe.com/datasheets/A5U1GA341ATS(BF)_v1.4_Zentel.pdf)
Apr 03 22:21:30 <PaulFertser> Toomoch: do you want only file contents or out-of-band data too?
Apr 03 22:22:15 <Toomoch> everything, router is non supported so if I brick I want to be able to restore somehow
Apr 03 22:23:47 <PaulFertser> Toomoch: I know nanddump from mtd tools should be capable.
Apr 03 22:24:16 <PaulFertser> mtd-utils
Apr 03 23:25:38 <owrt-snap-builds> build #321 of ramips/rt288x is complete: Failure [failed kmods]  Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Frt288x/builds/321  blamelist: Rafa? Mi?ecki <rafal@milecki.pl>, Tim Harvey <tharvey@gateworks.com>, Paul Spooren <mail@aparcar.org>, Tan Zien <nabsdh9@gmail.com>
Apr 04 02:12:56 <owrt-snap-builds> build #254 of ramips/mt7620 is complete: Failure [failed kmods]  Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Fmt7620/builds/254  blamelist: Rafa? Mi?ecki <rafal@milecki.pl>, Tim Harvey <tharvey@gateworks.com>, Paul Spooren <mail@aparcar.org>, Tan Zien <nabsdh9@gmail.com>
Apr 04 02:38:23 <owrt-snap-builds> build #252 of ramips/mt7621 is complete: Failure [failed kmods]  Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Fmt7621/builds/252  blamelist: Rafa? Mi?ecki <rafal@milecki.pl>, Tim Harvey <tharvey@gateworks.com>, Paul Spooren <mail@aparcar.org>, Tan Zien <nabsdh9@gmail.com>
**** ENDING LOGGING AT Sat Apr 04 03:00:20 2020