Switch to DuckDuckGo Search
   May 6, 2020  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | >

Toggle Join/Part | bottom
[00:20:43] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[00:25:01] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[00:26:18] *** neirac <neirac!~neirac@pc-4-149-45-190.cm.vtr.net> has quit IRC (Quit: bye)
[00:35:23] *** andy_js <andy_js!~andy@90.218.209.121> has quit IRC (Quit: andy_js)
[00:42:43] *** phyre <phyre!~phyre___@78.30.23.107> has quit IRC (Remote host closed the connection)
[01:09:52] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[01:19:24] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[01:23:02] *** clapont <clapont!~clapont@unaffiliated/clapont> has quit IRC (Ping timeout: 260 seconds)
[01:25:57] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[01:26:30] *** tsoome__ <tsoome__!~tsoome@91.209.240.229> has joined #illumos
[01:28:54] *** tsoome <tsoome!~tsoome@91.209.240.229> has quit IRC (Ping timeout: 256 seconds)
[01:28:54] *** tsoome__ is now known as tsoome
[01:29:07] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[02:04:28] *** tomww <tomww!~tom@unaffiliated/tomww> has quit IRC (Ping timeout: 258 seconds)
[02:11:09] *** alanc <alanc!~alanc@inet-hqmc01-o.oracle.com> has quit IRC (Remote host closed the connection)
[02:26:27] <ypankov> I don't understand that commit's message
[02:29:14] <LeftWing> I
[02:29:26] <LeftWing> feel we could have improved it certainly
[02:29:48] <LeftWing> sigh
[02:30:32] <jbk> that was the openzfs message
[02:31:04] <LeftWing> I think in the future we can feel free to make it at least a sentence
[02:31:53] <LeftWing> Does it mean ... that we can now specify a probability of errors that will be assessed for each I/O?
[02:32:23] <LeftWing> That's what it looks like at least
[02:32:57] <jbk> we could before, but at larger granularities
[02:33:23] <jbk> at least some openzfs tests though are using more granular values
[02:33:42] <jbk> (such as the resilver fixes up for review)
[02:33:55] <jbk> but if there are others, it should make porting simpler
[02:35:29] <jbk> you still may need to mess with tests that use 'events', but at least other bits won't require modification
[02:37:02] * LeftWing adds clarifying comment to bug
[03:10:59] *** alanc <alanc!~alanc@inet-hqmc02-o.oracle.com> has joined #illumos
[04:04:19] *** Teknix <Teknix!~pds@69.41.134.110> has quit IRC (Ping timeout: 256 seconds)
[04:05:13] *** Teknix <Teknix!~pds@69.41.134.110> has joined #illumos
[04:05:47] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has quit IRC (Quit: jcea)
[04:06:52] *** mscheker <mscheker!uid437521@gateway/web/irccloud.com/x-cqwtuqbdxuvprobg> has quit IRC (Quit: Connection closed for inactivity)
[04:11:11] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[04:21:54] *** Obscurax <Obscurax!~Obscurax@unaffiliated/obscurax> has quit IRC (Quit: So long, and thanks for all the fish.)
[04:22:15] *** Obscurax <Obscurax!~Obscurax@unaffiliated/obscurax> has joined #illumos
[04:23:03] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[04:25:24] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[04:31:56] *** tomww <tomww!~tom@unaffiliated/tomww> has joined #illumos
[04:40:45] *** Riastradh <Riastradh!~riastradh@netbsd/developer/riastradh> has quit IRC (Remote host closed the connection)
[04:41:00] *** Riastradh <Riastradh!~riastradh@netbsd/developer/riastradh> has joined #illumos
[04:47:50] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[05:03:23] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[05:06:24] *** mscheker <mscheker!uid437521@gateway/web/irccloud.com/x-lmpytogjgelinzhb> has joined #illumos
[05:18:46] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[05:19:04] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[05:19:34] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[05:30:27] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[06:11:06] *** alanc <alanc!~alanc@inet-hqmc02-o.oracle.com> has quit IRC (Ping timeout: 256 seconds)
[06:11:33] *** alanc <alanc!~alanc@inet-hqmc02-o.oracle.com> has joined #illumos
[06:15:54] *** BOKALDO <BOKALDO!~BOKALDO@87.110.146.209> has joined #illumos
[06:41:14] *** awordnot <awordnot!~awordnot@c-73-210-60-203.hsd1.il.comcast.net> has quit IRC (Ping timeout: 240 seconds)
[06:41:26] *** awordnot <awordnot!~awordnot@c-73-210-60-203.hsd1.il.comcast.net> has joined #illumos
[07:21:58] *** Riastradh <Riastradh!~riastradh@netbsd/developer/riastradh> has quit IRC (Remote host closed the connection)
[07:22:31] *** Riastradh <Riastradh!~riastradh@netbsd/developer/riastradh> has joined #illumos
[07:29:40] *** clapont <clapont!~clapont@unaffiliated/clapont> has joined #illumos
[07:57:22] *** phyre <phyre!~phyre___@78.30.23.107> has joined #illumos
[08:06:41] *** nbhauke <nbhauke!~hauke@55d4c1cf.access.ecotel.net> has joined #illumos
[08:09:53] *** nbhauke <nbhauke!~hauke@55d4c1cf.access.ecotel.net> has quit IRC (Client Quit)
[08:15:16] *** neuroserve <neuroserve!~toens@ip-88-152-243-25.hsi03.unitymediagroup.de> has quit IRC (Ping timeout: 246 seconds)
[08:23:04] *** bahamas10 <bahamas10!~dave@cpe-72-231-182-75.nycap.res.rr.com> has quit IRC (Ping timeout: 256 seconds)
[08:25:04] *** pwinder <pwinder!~pwinder@86.11.191.8> has joined #illumos
[08:30:04] *** bahamas10 <bahamas10!~dave@cpe-72-231-182-75.nycap.res.rr.com> has joined #illumos
[08:50:37] *** amrfrsh <amrfrsh!~Thunderbi@134.19.189.92> has quit IRC (Ping timeout: 264 seconds)
[09:04:02] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[09:05:02] *** neuroserve <neuroserve!~toens@ip-95-222-232-201.hsi15.unitymediagroup.de> has joined #illumos
[09:07:53] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[09:08:54] *** andy_js <andy_js!~andy@90.218.209.121> has joined #illumos
[09:09:50] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[09:24:56] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[09:26:19] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[09:30:29] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[09:31:46] *** ldepandis <ldepandis!~ldepandis@unaffiliated/ldepandis> has joined #illumos
[09:37:00] *** jimklimov <jimklimov!~jimklimov@78.80.224.132> has joined #illumos
[09:45:43] *** nbhauke <nbhauke!~hauke@55d4c1cf.access.ecotel.net> has joined #illumos
[09:46:30] *** wacki <wacki!~wacki@i577B8E87.versanet.de> has joined #illumos
[09:46:52] *** mscheker <mscheker!uid437521@gateway/web/irccloud.com/x-lmpytogjgelinzhb> has quit IRC (Quit: Connection closed for inactivity)
[09:52:33] *** ptribble <ptribble!~ptribble@cpc92716-cmbg20-2-0-cust138.5-4.cable.virginm.net> has joined #illumos
[09:59:54] <ypankov> https://www.illumos.org/issues/12692#note-1 :(
[10:09:24] *** vgusev <vgusev!~vgusev@188.187.60.230> has joined #illumos
[10:17:23] *** hemi770_ <hemi770_!~hemi666@unaffiliated/hemi770> has joined #illumos
[10:18:58] *** hemi770 <hemi770!~hemi666@unaffiliated/hemi770> has quit IRC (Ping timeout: 258 seconds)
[10:21:34] *** cypa <cypa!~cypam]_@5.79.187.109> has quit IRC (Ping timeout: 256 seconds)
[10:39:22] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[10:39:59] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[10:43:13] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[10:46:29] *** PiaMuehlen <PiaMuehlen!c8c3a2f2@gateway/web/cgi-irc/kiwiirc.com/ip.200.195.162.242> has joined #illumos
[10:48:50] *** PiaMuehlen <PiaMuehlen!c8c3a2f2@gateway/web/cgi-irc/kiwiirc.com/ip.200.195.162.242> has quit IRC (Client Quit)
[10:49:16] *** clapont <clapont!~clapont@unaffiliated/clapont> has quit IRC (Ping timeout: 246 seconds)
[10:51:12] *** clapont <clapont!~clapont@unaffiliated/clapont> has joined #illumos
[10:54:11] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[10:55:13] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has joined #illumos
[10:58:46] <toastersonerson> Wow SPAM wherever you look ..... sigh
[11:00:25] <sjorge> pmooney wasn't there some cpuid change upsteram for bhyve in phabricator? vaguely remember seeing a diff for something like that. Might be totally different though.
[11:01:34] <sjorge> nvm, can't find it so my memory is clearly corrupt
[11:01:40] <sjorge> Where is the human equalivantof fecc
[11:07:02] *** jimklimov <jimklimov!~jimklimov@78.80.224.132> has quit IRC (Ping timeout: 260 seconds)
[11:21:26] *** jimklimov <jimklimov!~jimklimov@78.80.224.132> has joined #illumos
[11:25:25] *** arnoldoree <arnoldoree!~arnoldore@2001:d08:1a04:a9:e1d6:e4f0:6602:bbb1> has joined #illumos
[11:26:10] <EisNerd> hm I did some experiments today with simple webservers for ISO presentation, I have serious trouble, maybe there is sth really odd. I'll check with another box
[11:27:47] <toastersonerson> Huh no success with dirlisting?
[11:28:10] <toastersonerson> although it is apache and nginx so a phd may be required to enable the webpage properly
[11:30:51] *** wacki <wacki!~wacki@i577B8E87.versanet.de> has quit IRC (Ping timeout: 256 seconds)
[11:31:20] *** wacki <wacki!~wacki@i577B8E87.versanet.de> has joined #illumos
[11:35:15] <EisNerd> hm seems to be a general issue
[11:44:03] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[11:47:07] *** jlevon <jlevon!jlevon@2a01:7e00::f03c:92ff:fefb:3ad2> has quit IRC (Quit: Coyote finally caught me)
[11:48:04] *** jlevon <jlevon!jlevon@2a01:7e00::f03c:92ff:fefb:3ad2> has joined #illumos
[11:51:22] *** cypa <cypa!~cypam]_@5.79.187.109> has joined #illumos
[11:53:04] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[12:15:22] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[12:18:35] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[12:20:52] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[12:21:10] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[12:23:29] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[12:23:49] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[12:39:56] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[12:42:41] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[12:43:45] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[12:46:31] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[12:49:21] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[12:51:30] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[12:55:47] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[12:56:00] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[12:56:24] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[12:59:34] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[13:00:11] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[13:03:14] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[13:05:25] *** Riastradh <Riastradh!~riastradh@netbsd/developer/riastradh> has quit IRC (Ping timeout: 246 seconds)
[13:10:34] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[13:13:38] *** hemi770_ is now known as hemi770
[13:13:40] *** amrfrsh <amrfrsh!~Thunderbi@134.19.189.92> has joined #illumos
[13:19:37] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[13:19:58] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[13:25:08] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[13:26:47] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[13:30:26] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[13:35:18] *** clapont <clapont!~clapont@unaffiliated/clapont> has quit IRC (Ping timeout: 256 seconds)
[13:46:12] *** clapont <clapont!~clapont@unaffiliated/clapont> has joined #illumos
[13:54:35] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[14:01:22] *** amrfrsh <amrfrsh!~Thunderbi@134.19.189.92> has quit IRC (Ping timeout: 260 seconds)
[14:03:08] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[14:04:06] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[14:09:17] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[14:10:48] *** strugee <strugee!~strugee@strugee.net> has quit IRC (Remote host closed the connection)
[14:11:37] *** phyre <phyre!~phyre___@78.30.23.107> has quit IRC (Remote host closed the connection)
[14:12:18] <pmooney> sjorge: I dunno? If so, they never bothered to push it in.
[14:12:33] <pmooney> I struck while the iron was hot
[14:14:26] *** cypa <cypa!~cypam]_@5.79.187.109> has quit IRC (Remote host closed the connection)
[14:16:59] *** strugee <strugee!~strugee@strugee.net> has joined #illumos
[14:20:39] *** gh34 <gh34!~textual@cpe-184-58-181-106.wi.res.rr.com> has joined #illumos
[14:35:15] *** phyre <phyre!~phyre___@78.30.23.107> has joined #illumos
[14:35:54] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[14:47:50] *** BOKALDO <BOKALDO!~BOKALDO@87.110.146.209> has quit IRC (Quit: Leaving)
[14:49:40] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[15:06:26] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has joined #illumos
[15:22:49] <richlowe> ypankov: deleted, thanks
[15:45:40] *** amrfrsh <amrfrsh!~Thunderbi@134.19.189.92> has joined #illumos
[15:50:23] *** neirac <neirac!~neirac@pc-4-149-45-190.cm.vtr.net> has joined #illumos
[15:55:59] *** BOKALDO <BOKALDO!~BOKALDO@87.110.146.209> has joined #illumos
[16:04:04] <tsoome> desperate times need desperate actions: https://www.youracclaim.com/badges/026da18c-b270-4be1-8a14-adc38d885ffe :D
[16:05:09] <wilbury> where are my solaris certificates... :-P
[16:07:02] <tsoome> I have those since 8 :)
[16:07:50] <toastersonerson> tsoome (IRC): desperate times?
[16:10:31] <KungFuJesus> lol, what is with the "verify" button
[16:10:38] <wilbury> tsoome: 6, 8, 10 and 11 here.
[16:11:36] <KungFuJesus> does it go back and reticulate splines?
[16:12:12] <toastersonerson> no thats firefox
[16:18:26] <tsoome> toastersonerson: yes, as oracles partner, we must complete some sort of certifications:)
[16:19:56] <toastersonerson> Ah yeah when you nned to bed oracle then times are truly desperate :)
[16:20:32] <ptribble> Maybe unexpected, but I don't have any Solaris certifications at all
[16:21:28] <wilbury> i was at oracle partner for almost 20 years, so for me they were mandatory (esp. hw ones, from V100 to E25k)
[16:22:01] <ptribble> I helped Sun develop the S10 exams, but never got anything for that - not even a discount on taking the test
[16:22:12] <wilbury> pow!
[16:23:00] <tsoome> the fun fact is, S11 OCP was actually hard to get:D
[16:23:32] <tsoome> S10 I got from beta exam
[16:24:33] <tsoome> beta means, instead of usual ~70 questions, you get something between 200-300
[16:24:34] <wilbury> i did all of them as upgrade from previous one
[16:27:17] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[16:29:09] <EisNerd> I get retries when testing with Iperf, any Idea on this?
[16:35:03] *** tsoome <tsoome!~tsoome@91.209.240.229> has quit IRC (Read error: Connection reset by peer)
[16:35:38] <EisNerd> how do I get retry counters on oi?
[16:35:46] <EisNerd> on ubuntu iperf shows retries
[16:36:20] *** tsoome <tsoome!~tsoome@148-52-235-80.sta.estpak.ee> has joined #illumos
[16:37:38] <rzezeski> EisNerd: for overall retries you can use netstat -s (you can also use this to get per/second retries as you run the benchmark), you can also get per conn/per second retries via connstat
[16:37:59] <rzezeski> err, retransmit, got me saying retry over here
[16:42:28] <EisNerd> thx, but not good tcpRetransSegs =147767 tcpRetransBytes =214344555
[16:44:14] <EisNerd> tcpRetransSegs =147767 (1/10 of total) tcpRetransBytes =214344555 (1/5 of total)
[16:44:29] <EisNerd> maybe a problem with i40e?
[16:48:10] <EisNerd> but this would explain non perfect network performance
[16:57:10] <jbk> very possible.. IIRC it's a rather complex driver (and rather complex NIC)
[16:58:39] *** tsoome__ <tsoome__!~tsoome@91.209.240.229> has joined #illumos
[17:01:14] *** tsoome <tsoome!~tsoome@148-52-235-80.sta.estpak.ee> has quit IRC (Ping timeout: 260 seconds)
[17:01:14] *** tsoome__ is now known as tsoome
[17:02:08] <gitomat> [illumos-gate] 12413 cxgbe: cast between incompatible function types -- Toomas Soome <tsoome at me dot com>
[17:02:16] <jbk> i still need to get back to fixing some issues w/ ixgbe (rzezeski: you might get a kick out what I found :P)
[17:03:11] <EisNerd> there is sth so mad with this stuff. I have booted a live ubuntu on our esxi system (currently out of order) and have a connectx4 edr <-> 40GBit HPE FlexFabric <-> i40e xl710 OI Hipster and uploading a 7GB ISO from the Linux to OI over smb server would take hours!!
[17:03:30] <rzezeski> The possible causes of retrans are many, if you want to solve it you have to paint a picture, not guess. First step is having a high level picture of everything int he path from sender to receiver. The next step might quantify retrans per second using said tools. What other stats increase with it? Do we see unordered segments? If so packets are probably dropping somewhere. There's various kstats to check...
[17:05:25] <rzezeski> jbk: I think I recall you talking about this before, refresh my memory
[17:05:29] <rzezeski> Something to do with bhyve
[17:05:32] <rzezeski> err,viona
[17:05:38] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has joined #illumos
[17:05:42] <rzezeski> and it's interaction with ixgbe
[17:06:16] <EisNerd> but I checked to other hosts attached to GBit switches uplinked to the HPE FF, from the ubuntu no retries, to the oi same picture and this is true for both oi nodes in this box, one using the xl710 the other uses the 10G BaseT variant linked to the said GBit Switches
[17:06:37] *** baojg <baojg!~baojg@2604:a880:0:1010::16:1> has quit IRC (Remote host closed the connection)
[17:07:22] *** Riastradh <Riastradh!~riastradh@netbsd/developer/riastradh> has joined #illumos
[17:13:43] <jbk> rzezeski: well in trying to track down leaked loaned viona dbuf_ts (still haven't been able to figure out that one yet), i noticed something else
[17:14:41] <rzezeski> okay, I have a guess on what you might have found, having to do with the descriptor logic
[17:14:53] <jbk> when doing LSO, it goes through all sorts of hard-to-follow gyrations moving around bytes in mblk_ts to get them just perfect, then... proceeds to copy all of those bits it juggled into a descriptor
[17:14:54] <rzezeski> as it's pretty gnarly
[17:17:28] <jbk> when it could just copy from the original mblk_ts into the descriptor
[17:17:34] <EisNerd> ok, got answer regarding esxi, hopefully we get progress (entire team is nervous as this downs our virtual NonStop, the preferred system for development and support lab work as a lot faster than all others)
[17:18:05] *** Kurlon_ <Kurlon_!~Kurlon@cpe-67-253-136-97.rochester.res.rr.com> has quit IRC (Remote host closed the connection)
[17:18:16] <jbk> (I get if it was mapping the dblk_ts, it might need those buffers to have special alignment requirements, but bcopy() has no such requirements)
[17:18:45] *** Kurlon <Kurlon!~Kurlon@167-48.gwi.net> has joined #illumos
[17:19:11] <EisNerd> anyway, I'll try to get a better picture of the tcp issues of the oi box
[17:21:34] <jbk> which also means it's doing unnecessary allocations in the tx path.. so both simplifying and eliminating unnecessary allocations in the tx path seems like a good thing (obviously going to need lots of testing)
[17:25:12] <rzezeski> I'd have to reread the code to be sure, but IIRC it's because the way ixgbe does it it doesn't know aprior if its going to exceed the hardware's descriptor limit. When I did i40e I think I improved upon this by trying to gain that knowledge before making the choice of bcopy or DMA. The same treatment could probably be given to ixgbe. But the question is, how often is that code path actually hit. I dunno.
[17:25:42] <rzezeski> But that's my fuzzy memory.
[17:25:51] <jbk> the dump i was looking at hit that path pretty frequently (judging by the counters)
[17:26:20] <jbk> but even there, when it hits that, it does a msgpullup
[17:28:04] <rzezeski> yea, the way ixgbe did things wasn't quite right, IIRC it just uses the mblk chain length to determine if it should msgpullup, which isn't really correct because an individual mblk might cross a page boundary, but to improve upon this I had to write some pretty darn complicated code in i40e
[17:28:40] <rzezeski> at first I thought I could write that code in a way that could be shared by the drivers, but then I think I went back on that because it relied on certain specifics
[17:32:44] <rzezeski> While I agree this sucks, if I was looking at this myself, I might only make the change if it's preventing me from hitting line rate. Granted, that's probably going to depend on the host CPU/memory, and what works for one may not work for another, but fixing this may also require a bunch of time that you might want to spend elsewhere, I dunno.
[17:32:58] <rzezeski> But I certainly remember looking at the ixgbe tx logic and wishing it was better.
[17:38:31] <Smithx10> I wonder how much longer ixgbe will be with us
[17:39:10] <Smithx10> Are there any newer cards being produced that use that driver?
[17:39:47] <jbk> my home server i bought two years ago has it built in
[17:41:19] <rzezeski> That's sort of my point. I think it will be useful for years to come. BUT, I also think it's worth putting its shortcomings in context. If it doesn't prevent us from hitting line rate, or negatively affect other parts of the system for _most people_, then perhaps it's okay to leave it. But that's just how I approach it. And I don't have to data to say either way here.
[17:41:54] *** gh34 <gh34!~textual@cpe-184-58-181-106.wi.res.rr.com> has quit IRC (Ping timeout: 240 seconds)
[17:42:45] <rzezeski> At the same time, fixing it may not be as bad as I think. And it would be nice to avoid allocation when we can.
[17:45:15] <rzezeski> Also, I might have been wrong about making my i40e descriptor logic generic. Perhaps rm has some thoughts on this based on his experience with ice.
[17:45:45] <jbk> yeah, i haven't messed with it yet (other things first), but also given how incredibly hard that code is to follow (it was tripping jerry up as well), between all of that, it's probably worth the effort if only to save future us time next time a question comes up
[17:46:11] <jbk> (it took me 2-3 days of note taking to try to follow everything in there)
[17:46:23] <rzezeski> Yea, that code is confusing, i had to study it for a couple of days to make sure I had it down 100%.
[17:46:27] <jbk> because of all the mblk_t gyrations it's doing
[17:48:22] <jbk> (it'd also be nice if it didn't use it's own home grown linked list struct, but that seems less critical, and just annoying)
[17:48:48] <rzezeski> The main problem it's trying to solve is "can this mblk layout be DMA'd without violating the hardware's descriptor requirements? If not, then bcopy into something that does". And in ixgbe we don't know that until later in the process, whereas in i40e I made it so it can DMA as much as possible but still copy when it absoutely needs to.
[17:49:59] <jbk> yeah.. but for LSO, it's doing that for the headers, then _always_ bcopying the headers
[17:50:28] <rzezeski> If that implementation could be parameterized in a way that we just feed it a few specific values for each piece of hardware, then maybe we could solve this in mac itself, but I remember having doubts it could easily be made generic
[17:54:17] <rzezeski> in i40e I choose to always copy the header
[17:54:22] <rzezeski> otherwise things get WAY more complicated
[17:54:25] <rzezeski> for no good reason
[17:54:45] <rzezeski> I'm rereading my comment from i40e_lso_chain() and it's reminding me just how gnarly all this is :)
[17:58:10] <rzezeski> Man this is bringing back some painful debugging memories.
[18:13:06] <Smithx10> rzezeski: like..... this....... https://www.youtube.com/watch?v=Pu0lghnPtck ?
[18:15:15] <rzezeski> something like that
[18:16:18] <Smithx10> hahahaaa
[18:16:41] <Smithx10> I'm wondering if there is a better ground hog like day joke
[18:40:48] *** pwinder <pwinder!~pwinder@86.11.191.8> has quit IRC (Quit: This computer has gone to sleep)
[18:44:06] *** nde <nde!uid414739@gateway/web/irccloud.com/x-wsfijcpqacawwoyi> has joined #illumos
[19:08:00] *** wacki <wacki!~wacki@i577B8E87.versanet.de> has quit IRC (Read error: Connection reset by peer)
[19:08:01] *** ypankov <ypankov!~ypankov@85.175.17.187> has quit IRC (Ping timeout: 264 seconds)
[19:16:56] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has quit IRC (Quit: Leaving)
[19:20:01] *** mnrmnaugh <mnrmnaugh!~mnrmnaugh@unaffiliated/mnrmnaugh> has quit IRC (Quit: Leaving)
[19:25:30] *** mnrmnaugh <mnrmnaugh!~mnrmnaugh@unaffiliated/mnrmnaugh> has joined #illumos
[19:35:01] *** pwinder <pwinder!~pwinder@86.11.191.8> has joined #illumos
[19:35:34] *** pwinder <pwinder!~pwinder@86.11.191.8> has quit IRC (Client Quit)
[19:46:49] *** ypankov <ypankov!~ypankov@85.175.17.187> has joined #illumos
[19:57:25] *** mnrmnaugh is now known as AllTheCows
[19:57:37] <gitomat> [illumos-gate] 12697 liblddbg: sparc is missing ROOTCOMPATLINKS -- Toomas Soome <tsoome at me dot com>
[19:58:14] *** AllTheCows is now known as ALazyGoat
[20:07:29] <Smithx10> neirac: do you have a branch for gopsutil ?
[20:07:58] <Smithx10> or are the latest bits just in your nomad fork vendor dir?
[20:11:29] *** ALazyGoat is now known as mnrmnaugh
[20:18:14] *** neuroserve <neuroserve!~toens@ip-95-222-232-201.hsi15.unitymediagroup.de> has quit IRC (Ping timeout: 240 seconds)
[20:24:49] <neirac> Smithx10 just hte ones on the nomad vendor
[20:25:15] <Smithx10> kk
[20:25:32] <Smithx10> ill have 2 weeks off, think I might try to get some PRs done
[20:26:24] *** ptribble <ptribble!~ptribble@cpc92716-cmbg20-2-0-cust138.5-4.cable.virginm.net> has quit IRC (Quit: Leaving)
[20:36:12] *** tsoome <tsoome!~tsoome@91.209.240.229> has quit IRC (Read error: Connection reset by peer)
[20:36:52] *** tsoome <tsoome!~tsoome@91.209.240.229> has joined #illumos
[20:40:19] *** wacki <wacki!~wacki@i577B8E87.versanet.de> has joined #illumos
[20:48:08] *** neuroserve <neuroserve!~toens@ip-88-152-243-25.hsi03.unitymediagroup.de> has joined #illumos
[20:53:01] *** jimklimov <jimklimov!~jimklimov@78.80.224.132> has quit IRC (Ping timeout: 246 seconds)
[20:53:52] *** nde <nde!uid414739@gateway/web/irccloud.com/x-wsfijcpqacawwoyi> has quit IRC (Quit: Connection closed for inactivity)
[21:20:37] *** BOKALDO <BOKALDO!~BOKALDO@87.110.146.209> has quit IRC (Quit: Leaving)
[21:21:03] <neirac> Smithx10 that would be awesome!, you could test nomad
[21:21:18] <Smithx10> i dont think its about testing nomad
[21:21:32] <Smithx10> more about getting gops to work.....
[21:21:37] <neirac> Smithx10 we need to test it anyways
[21:21:49] <neirac> Smithx10 to make sure gops work :)
[21:21:57] <Smithx10> I imagine GOPS will have tests :)
[21:22:23] <neirac> Smithx10 yes, it has but we found out issues anyways
[21:22:32] <Smithx10> ?
[21:51:36] <neirac> Smithx10 it has tests, but we need to add more
[21:51:45] <Smithx10> okie dokie
[22:17:41] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has joined #illumos
[22:24:22] *** nbhauke <nbhauke!~hauke@55d4c1cf.access.ecotel.net> has quit IRC (Quit: nbhauke)
[22:34:09] <toastersonerson> neirac (IRC): Does gops include mounttable parsing?
[22:34:43] <neirac> toastersonerson as I recall only cpu/mem stats
[22:34:59] <toastersonerson> ok then I look into mounttable stats
[22:36:09] <toastersonerson> lol podmans container/storage repo is really inconsistent in freebsd case. one end uses sys/unix and the other cgo but the same call
[22:38:40] <toastersonerson> do we have anything resembling getmntinfo on freebsd
[22:53:06] <KungFuJesus> Cool, gkrellm's been fixed upstream: https://git.srcbox.net/gkrellm/gkrellm/commit/13f1489df4616583b123b16d04aa3992dce56b0e
[22:53:59] <KungFuJesus> someone should ping the oi-userland gkrellm maintainer to grab the patches ontop of 2.3.11 and apply them until the next release
[23:05:20] *** ibenn <ibenn!~benn@HSI-KBW-095-208-237-062.hsi5.kabel-badenwuerttemberg.de> has quit IRC (Quit: Leaving)
[23:23:18] *** Kurlon_ <Kurlon_!~Kurlon@cpe-67-253-136-97.rochester.res.rr.com> has joined #illumos
[23:26:37] *** Kurlon <Kurlon!~Kurlon@167-48.gwi.net> has quit IRC (Ping timeout: 264 seconds)
[23:29:33] *** neirac <neirac!~neirac@pc-4-149-45-190.cm.vtr.net> has quit IRC (Quit: ZNC 1.8.0 - https://znc.in)
[23:41:16] *** andy_js <andy_js!~andy@90.218.209.121> has quit IRC (Quit: andy_js)
[23:44:01] <wilbury> is _fini() being called even if _init() returns an error?
[23:46:09] <rmustacc> I wouldn't expect so.
[23:49:43] <wilbury> (asking for a friend! :-))
[23:49:51] *** CarlyLauren <CarlyLauren!224d2f0d@gateway/web/cgi-irc/kiwiirc.com/ip.34.77.47.13> has joined #illumos
[23:49:57] <rmustacc> I can take a quick look.
[23:50:14] <rmustacc> At least, the way I always have written it it always has to clean up what it created.
[23:51:46] <wilbury> for example, i create a mutex _before_ mod_install. then mod_install will fail and as part of cleanup i call mutex_destroy(). and mutex_destroy() is also called in _fini() during "normal" cleanup. i'm only curious (as i actually moved mutex_init() after mod_install() where it's safe that it succeeded.)
[23:52:09] *** CarlyLauren <CarlyLauren!224d2f0d@gateway/web/cgi-irc/kiwiirc.com/ip.34.77.47.13> has quit IRC (Client Quit)
[23:52:23] <wilbury> as calling mutex_destroy() twice would result in a panic, i suppose.
[23:52:28] <rmustacc> It looks like I should improve the docs in _init(9E), but no. If _init() fails, _fini() will not be called based on a quick search.
[23:52:41] <rmustacc> Which matches my expectation.
[23:53:05] <wilbury> also mine. but you know, assumption is the mother of all fuckups.
[23:53:09] <rmustacc> Yup
[23:53:20] <rmustacc> Which is why I'm double checking the code.
[23:53:22] *** phyre <phyre!~phyre___@78.30.23.107> has quit IRC (Remote host closed the connection)
[23:53:33] <wilbury> i'm diving into the code as well
[23:54:00] <rmustacc> But based on the paths I'm seeing, I don't believe so. When it fails, we call a different unload path.
[23:54:21] <rmustacc> e.g. the things that moduninstall() will call itself.
[23:54:51] <rmustacc> Since moduninstall is the only thing that does the _fini(9E) call, since we don't get vectored there at all I think that's the case.
[23:55:12] <rmustacc> If you find something different, let me know.
[23:57:33] *** wacki <wacki!~wacki@i577B8E87.versanet.de> has left #illumos
[23:59:36] <wilbury> yes, i found it
[23:59:48] <wilbury> the same as you
top

   May 6, 2020  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | >