Switch to DuckDuckGo Search
   April 15, 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 | >


NOTICE: This channel is no longer actively logged.

Toggle Join/Part | bottom
[00:12:01] *** KungFuJesus <KungFuJesus!~adam@72.49.160.91> has joined #illumos
[00:12:16] <KungFuJesus> ok who do I send 800MB kernel crash dumps to?
[00:12:41] <KungFuJesus> this happened exactly one time and I haven't reproduced it since, but I'm bothered that it exists
[00:12:51] <KungFuJesus> backtrace shows it in some zfs code
[00:12:58] *** hww3 <hww3!~hww3@c-73-128-248-73.hsd1.md.comcast.net> has joined #illumos
[00:13:45] <andyf> The backtrace and some other info might be enough
[00:19:47] <wilbury> Warning: /kernel/drv/amd64/nvidia uses deprecated _depends_on interface.
[00:22:26] *** neirac <neirac!~cneira@pc-4-149-45-190.cm.vtr.net> has quit IRC (Ping timeout: 256 seconds)
[00:48:01] *** neirac <neirac!~cneira@pc-4-149-45-190.cm.vtr.net> has joined #illumos
[00:55:30] *** ngchk1_ <ngchk1_!~ngchk1@b2b-92-50-91-166.unitymedia.biz> has joined #illumos
[00:57:13] *** ngchk1 <ngchk1!~ngchk1@b2b-92-50-91-166.unitymedia.biz> has quit IRC (Ping timeout: 264 seconds)
[01:14:30] *** Tempt <Tempt!~avenger@unaffiliated/tempt> has quit IRC (Ping timeout: 256 seconds)
[01:15:29] *** Tempt <Tempt!~avenger@mexico.purplecow.org> has joined #illumos
[01:15:32] *** ChanServ sets mode: +o Tempt
[01:15:34] <danmcd> It's clean.
[01:16:19] <gitomat> [illumos-gate] 12383 Slow down and lock up in mlxcx receive interrupt path (fix smatch) -- Dan McDonald <danmcd at joyent dot com>
[01:17:22] <LeftWing> Thanks again Dan
[01:18:35] <danmcd> Now to figure out why awk-points-to-nawk won't build. (IPS stuff, I suspect.)
[01:19:06] <danmcd> (Well, that and try out the new Overwatch character that showed up today with P.)
[01:23:29] *** wacki_ <wacki_!~wacki@i577B8ED3.versanet.de> has joined #illumos
[01:24:39] <KungFuJesus> alrihgt, hold on, let me see if I can make fmdump give me that
[01:24:47] *** gh34 <gh34!~textual@cpe-184-58-181-106.wi.res.rr.com> has quit IRC (Quit: Textual IRC Client: www.textualapp.com)
[01:25:20] <KungFuJesus> https://pastebin.com/uw3VsiKi
[01:25:56] <KungFuJesus> andyf: does that look sufficient or do you want the whole 800MB shebang?
[01:27:00] <KungFuJesus> looks to be in some ACE checking for permissions
[01:27:29] <KungFuJesus> I'm pretty sure I'm not using anything fancy, I don't think I have any NFSv4 ACLs deployed in any of the file systems on this machine
[01:28:25] *** wacki_ <wacki_!~wacki@i577B8ED3.versanet.de> has quit IRC (Ping timeout: 264 seconds)
[01:46:49] *** ldepandis <ldepandis!~ldepandis@unaffiliated/ldepandis> has quit IRC (Quit: My MacBook has gone to sleep. ZZZzzz…)
[02:30:30] <KungFuJesus> rmustacc: For the kernel FPU you mention, would this work similarly to FreeBSD and Linux, where you have a small subset of XMM/YMM/ZMM register state that you need to save and restore?
[02:33:25] <rmustacc> KungFuJesus: It would allow the kernel to use it in a similar fashion to user land. I'm not sure that forcing users to have to specify a subset of registers versus just biting the bullet and using the optimized instructions for this makes more or less sense.
[02:33:45] <rmustacc> But the idea is that a thread can allocate and transition to using in a way that is nestable and works across context switches.
[02:34:05] <jbk> i actually started messing with that, but haven't had a chance to get back to it yet
[02:34:25] <KungFuJesus> I do wonder what the total savings might be for a partial FPU save/restore versus using the special x86 instructions xsave and xrstor
[02:34:27] <rmustacc> KungFuJesus: https://github.com/rmustacc/illumos-gate/commit/687dea0f7db5ce2a60940855d35739aa7b90e4e9 is a snapshot of what I was prototyping.
[02:34:51] <KungFuJesus> actually it looks like those instructions allow for partial save and restore too
[02:34:54] <rmustacc> KungFuJesus: Well, that's why what I was putting together leveraged all the existing mechanisms and leveraged Intel's optimiations.
[02:34:58] <jbk> since one thing max found w/ his zfs testing is that there's a not insignificant amount of time spent dealing with the current preemption bits (not to say it's the max, but it definitely shows up)
[02:35:43] <rmustacc> KungFuJesus: The processor does a lot of tracking to optimize the save and restore path there and to make sure you don't have to do things if you don't need to.
[02:35:54] <KungFuJesus> certainly the fletcher and SHA accelerations in current OpenZFS are affected by this
[02:35:57] <rmustacc> So I would start with an API that just leveraged all that rather than trying to be smart.
[02:37:21] <rmustacc> ZFS makes things a bit more complicated due to how work can be done both on threads dedicated to ZIOs and on user threads, but in principle, it shouldn't be too bad to use something like this for that or other parts of the system.
[02:37:41] <rmustacc> We're already paying the eager FPU cost on context switch anyways, so disabling pre-emption doesn't make sense.
[02:38:22] <KungFuJesus> you don't have to deal with this if one kernel thread is pre-empting another, anyway, unless of course it's going to use the FPU as well
[02:38:33] <rmustacc> You do.
[02:38:47] <rmustacc> Or rather eager FPU makes it so its simpler to just deal with it.
[02:39:18] <rmustacc> And a number of kernel threads have FPU state anyways due to how they have a klwp_t.
[02:39:31] <KungFuJesus> ah I see, it's when the kernel returns control back to user space, it doesn't know about other kernel threads which may have stomped on the FPU state
[02:39:54] <rmustacc> Well, historically you'd trap and deal and lazily restore it.
[02:40:05] <rmustacc> But turns out, side-channels, so no can do.
[02:40:21] <KungFuJesus> does the eager FPU vulnerability force everyone to not do lazy FPU handling anymore?
[02:40:25] <rmustacc> (And not clear it was much faster given the general broad usage of the pu)
[02:40:27] <rmustacc> Um, yes.
[02:40:37] <KungFuJesus> damn :(
[02:40:44] <rmustacc> It wasn't much of a win anyways.
[02:40:57] <rmustacc> Not when userland is doing bcopy with SSE + AVX.
[02:41:31] <KungFuJesus> I wonder if the micro architecture just dumps all this state to non-architectural registers during an xsave/xrstor cycle
[02:42:06] <rmustacc> Hmm? They write things out to memory.
[02:42:29] <rmustacc> If you're talking about the tracking and optimizations with things like xsaveopt, the Intel SDM is pretty transparent about what it does.
[02:42:32] <KungFuJesus> ah, well "memory" could be fill buffers in theory, couldn't it?
[02:43:53] <rmustacc> They could perhaps further optimize things, but when you call xsave, it is written out to memory regardless.
[02:44:09] <rmustacc> Similarly xrstor can and does work on a region that wasn't created with xsave.
[02:45:47] <KungFuJesus> ah yeah it looks like the instruction takes an arbitrary address as the destination operand
[02:46:45] <rmustacc> Yes. It has to.
[02:46:56] <rmustacc> At least, to be useful to an OS.
[02:46:57] <KungFuJesus> depending on how far spread out they are temporally, though, it could be entirely serviced by cache and/or fill buffers
[02:47:38] <rmustacc> It's certainly possible.
[02:48:36] <KungFuJesus> I spend a lot of time writing SIMD code in userspace for work, this is usually a detail I've been free to ignore, heh
[02:49:23] <rmustacc> Alas, I've not been so lucky there.
[02:49:51] <KungFuJesus> near 100% success rate with SIMD acceleration when I don't have to lease the FPU like a kernel would
[02:49:56] <rmustacc> Eager fpu, avx512 power / clocking issues, etc. Only the pain of it.
[02:50:13] <KungFuJesus> now the AVX512 clocks issue I do run up against from time to time
[02:50:15] <rmustacc> Well, good news, with the design here, you wouldn't have to.
[02:50:54] <rmustacc> At least, the goal of the kfpu design is to leverage the fact that we're always eagerly switching due to eager fpu and to also handle nesting. So you don't have to think about it as leasing as much.
[02:50:59] <rmustacc> Especially for long-running threads.
[02:51:01] <KungFuJesus> yeah getting the full 16/32 register set would probably be nice. From what I've seen you only get like the first 4 or 8 of them in Linux
[02:51:32] <rmustacc> It's written so you get them all right now (though probably better to avoid anything that aliases the %zmm)
[02:52:00] <rmustacc> Erm, not meaning the lower overlap, but rather things that trick the processor into downclocking.
[02:52:25] <KungFuJesus> Incidentally I also found out that the kmask instructions have largely been some weird anachronism that survived from knights landing. GCC seems super eager to move to GPRs even when it seems stupid to
[02:52:47] <KungFuJesus> as it turns out, kmov/b/w/d/q we're ridiculously high penalties in knights landing but not SKX
[02:53:33] <KungFuJesus> still GCC seems overly eager when you use intrinsics to generate the kmov instructions and back, even when it makes no sense
[02:54:17] <KungFuJesus> yeah, not just the downclocking but the fact that if you write to those registers, there's the hidden penalty if you forget to call zeroupper
[02:55:03] <rmustacc> I'm quite aware, unfortunately.
[02:55:17] <KungFuJesus> I found that bug in glibc's libm functions, hah
[02:55:36] <rmustacc> I messed up an initial save state in the OS so at some point it thought ymm/zmm were active when not.
[02:55:56] <KungFuJesus> oh yeah, that penalty is pretty painful
[02:55:59] <rmustacc> Which didn't help fpu perf and made broadwell look better than skylake.
[02:56:11] <rmustacc> Yeah. It was a bit pernicious.
[02:56:35] <KungFuJesus> It makes SSE code outperform AVX code if you don't catch it, it's one hell of a penalty
[02:58:30] <KungFuJesus> the k register instructions also have seem weirdness where for some reason simple arithmetic shifts and adds are 3x-4x more cycles than in a GPR, and they only execute on a single port (of which is one of the ones reserved for vector instructions)
[03:03:15] <KungFuJesus> rmustacc: bge code seems functional, if you can get jumbo frames working I can test on a more long term basis with a link aggregation
[03:04:52] <rmustacc> I'm probably pausing at the current stuff right now. I would like to see it work, but it's a bit challenging right now and I wanted to get a round of support done.
[03:06:35] <KungFuJesus> didn't have any obvious issues/errors with standard MTU. Most of the tests were against the TCP stack rather than raw datagram stuff, though
[03:08:34] <KungFuJesus> https://pastebin.com/M5UhGndT
[03:13:18] <rmustacc> I'm just glad that much is working.
[03:22:12] *** joltman <joltman!~znc@unaffiliated/joltman> has quit IRC (Quit: ZNC - http://znc.in)
[03:24:27] *** wacki_ <wacki_!~wacki@i577B8292.versanet.de> has joined #illumos
[03:24:49] *** joltman <joltman!znc@gateway/vpn/privateinternetaccess/joltman> has joined #illumos
[03:25:29] *** idodeclare <idodeclare!~textual@cpe-76-185-177-63.satx.res.rr.com> has quit IRC (Quit: Textual IRC Client: www.textualapp.com)
[03:29:26] *** wacki_ <wacki_!~wacki@i577B8292.versanet.de> has quit IRC (Ping timeout: 256 seconds)
[03:34:28] *** ke5c2fin <ke5c2fin!~ke5c2fin@c-67-176-119-108.hsd1.co.comcast.net> has quit IRC (Ping timeout: 256 seconds)
[03:52:02] *** domag02 <domag02!b03fd4e2@catv-176-63-212-226.catv.broadband.hu> has joined #illumos
[03:52:09] <domag02> hi!
[03:52:31] <rmustacc> Hello
[03:58:26] <domag02> rmustacc: I found the necessary manual pages on the Oracle Solaris 10 u11 install media (Bug 814 - need hotplug manual pages)
[03:58:49] <domag02> and all of them are CDDL licensed
[03:59:53] <rmustacc> This is a really awkward thing. We really don't want to be in a grey area.
[03:59:57] <rmustacc> And this is pretty grey.
[04:00:17] <rmustacc> Because, for example, you agreed to a large terms of service when you agreed to download and obtain that media.
[04:00:30] <rmustacc> As a project, we'd rather just not expose ourselves to risk and uncertainty there, tbh.
[04:00:50] <rmustacc> It's also the case that whatever's in their manual page may not match our actual implementation.
[04:38:18] *** KungFuJesus <KungFuJesus!~adam@72.49.160.91> has quit IRC (Quit: Lost terminal)
[04:42:59] *** neirac <neirac!~cneira@pc-4-149-45-190.cm.vtr.net> has quit IRC (Quit: Lost terminal)
[04:48:27] *** neirac <neirac!~cneira@pc-4-149-45-190.cm.vtr.net> has joined #illumos
[04:56:23] *** cantstanya <cantstanya!~chatting@gateway/tor-sasl/cantstanya> has quit IRC (Ping timeout: 240 seconds)
[05:00:13] *** cantstanya <cantstanya!~chatting@gateway/tor-sasl/cantstanya> has joined #illumos
[05:11:58] <domag02> rmustacc: it looks like it's a quite hard problem
[05:13:49] *** neirac <neirac!~cneira@pc-4-149-45-190.cm.vtr.net> has quit IRC (Ping timeout: 258 seconds)
[05:15:49] <richlowe> The position at the time was that stuff CDDL'd by mistake was probably not worth the risk
[05:17:47] <domag02> How can I write a manual about (almost) the same software, with avoiding plagiarism/copyright infringement?
[05:19:50] <richlowe> base it on the code, don't read their page, I suppose.
[05:19:58] <richlowe> but I'm not a lawyer or anything.
[05:25:32] *** wacki_ <wacki_!~wacki@i577A59B5.versanet.de> has joined #illumos
[05:30:23] *** wacki_ <wacki_!~wacki@i577A59B5.versanet.de> has quit IRC (Ping timeout: 250 seconds)
[05:44:56] *** jwit_ is now known as jwit
[06:07:49] <domag02> another thing: I found an 'interesting' way of looping in shell: http://src.illumos.org/source/xref/illumos-gate/usr/src/cmd/svc/shell/net_include.sh#198
[06:09:22] <domag02> is it worth to create an issue to fix this ugliness?
[06:10:00] <LeftWing> Wow.
[06:10:18] *** Tempt <Tempt!~avenger@unaffiliated/tempt> has quit IRC (Ping timeout: 256 seconds)
[06:11:34] <LeftWing> "what else can we do?"
[06:11:40] <LeftWing> I mean, not bloody *that*
[06:11:56] *** Tempt <Tempt!~avenger@unaffiliated/tempt> has joined #illumos
[06:11:56] *** ChanServ sets mode: +o Tempt
[06:17:21] <LeftWing> domag02: It feels like we could definitely do better there
[06:18:47] <domag02> LeftWing: I think yes, for example: while [ "$i" -ne 1000 ]; do .....
[06:19:58] <LeftWing> Right
[06:20:30] <LeftWing> If you want to make an improvement there, it will be important to test that it still works afterwards
[06:22:53] <domag02> LeftWing: yes, and that's the harder part
[06:43:16] *** nde <nde!uid414739@gateway/web/irccloud.com/x-skjotqnhnygdrjrk> has quit IRC (Quit: Connection closed for inactivity)
[07:05:22] *** Tempt <Tempt!~avenger@unaffiliated/tempt> has quit IRC (Ping timeout: 258 seconds)
[07:06:55] *** Tempt <Tempt!~avenger@unaffiliated/tempt> has joined #illumos
[07:06:55] *** ChanServ sets mode: +o Tempt
[07:26:27] *** wacki_ <wacki_!~wacki@i577A59B5.versanet.de> has joined #illumos
[07:31:14] *** wacki_ <wacki_!~wacki@i577A59B5.versanet.de> has quit IRC (Ping timeout: 240 seconds)
[07:48:33] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has joined #illumos
[07:51:02] *** arnoldoree <arnoldoree!~arnoldore@2001:d08:2183:afdb:c1cd:b139:c4b1:918f> has quit IRC (Remote host closed the connection)
[08:07:06] *** phyre <phyre!~phyre___@78.30.16.107> has joined #illumos
[08:07:07] <Agnar> moin
[08:17:24] <tsoome> moin
[08:25:07] *** wacki_ <wacki_!~wacki@i577A59B5.versanet.de> has joined #illumos
[08:48:21] *** arnoldoree <arnoldoree!~arnoldore@2001:d08:2183:afdb:c1cd:b139:c4b1:918f> has joined #illumos
[08:50:13] *** alanc <alanc!~alanc@inet-hqmc01-o.oracle.com> has quit IRC (Remote host closed the connection)
[08:50:40] *** alanc <alanc!~alanc@inet-hqmc01-o.oracle.com> has joined #illumos
[08:55:27] <sjorge> moin
[08:55:41] <sjorge> jbk: had a chance to peak at the update pam_list diff?
[08:58:28] <Agnar> moin sjorge, tsoome
[09:06:02] *** neuroserve <neuroserve!~toens@ip-84-118-128-205.unity-media.net> has quit IRC (Ping timeout: 256 seconds)
[09:12:59] <toastersonerson1> moin
[09:14:59] <toastersonerson1> Does anybody know the documents of the unix language formats? I have some people in the opencontainers list asking what people on posix systems would find acceptable.
[09:16:02] <Agnar> what do you means with language formats?
[09:16:17] <toastersonerson1> en_US es_BR etc.
[09:16:30] <Agnar> oh
[09:16:32] <toastersonerson1> the gnu docs https://www.gnu.org/software/libc/manual/html_node/Locale-Names.html
[09:16:43] *** BOKALDO <BOKALDO!~BOKALDO@81.198.20.220> has joined #illumos
[09:16:47] <toastersonerson1> was wondering if there are more original ones
[09:17:12] <toastersonerson1> seems to be specified in xpg according to gnu
[09:19:44] <Agnar> yes, seems so.
[09:20:03] <Agnar> and illumos referrs also to XPG,3,4,4v2
[09:21:26] <toastersonerson1> in the manpages?
[09:21:37] <Agnar> standards(5)
[09:22:03] <toastersonerson1> :) figured. thanks
[09:22:32] *** neuroserve <neuroserve!~toens@ip-84-118-130-67.unity-media.net> has joined #illumos
[09:22:40] <Agnar> I can't find ad hoc any direct pointer in it that locale names are in XPG
[09:23:59] *** jimklimov1 <jimklimov1!~jimklimov@78.80.224.132> has joined #illumos
[09:40:26] <toastersonerson1> i found locale(5) but only the gnu doc had the names. but thats good enough for the sample and as pointers.
[09:40:37] <toastersonerson1> thnaks
[09:44:42] *** ptribble <ptribble!~ptribble@cpc92716-cmbg20-2-0-cust138.5-4.cable.virginm.net> has joined #illumos
[10:11:33] *** wacki_ <wacki_!~wacki@i577A59B5.versanet.de> has quit IRC (Quit: Lingo: www.lingoirc.com)
[10:11:48] *** wacki <wacki!~wacki@i577A59B5.versanet.de> has joined #illumos
[10:23:24] *** andy_js <andy_js!~andy@90.218.209.121> has joined #illumos
[10:23:42] *** man_u <man_u!~manu@89-92-19-81.hfc.dyn.abo.bbox.fr> has joined #illumos
[10:24:32] *** man_u <man_u!~manu@89-92-19-81.hfc.dyn.abo.bbox.fr> has left #illumos
[10:27:27] <Aram> toastersonerson1: alanc: yes, I am still here.
[10:28:26] <Aram> toastersonerson1: the repo is here, and it works, but it's an old version of Go now: https://github.com/4ad/go
[10:29:04] <Aram> getting it up to date would be quite the challenge... which I might do someday.
[10:29:48] <toastersonerson1> Aram (IRC): I had a look. there are a couple of commits that are there. Did you have any contact with the freebsd guys?
[10:30:49] <Aram> not particularly relevant to sparc, no.
[10:31:11] <Aram> all development and testing was done on solaris, porting the sparc work to freebsd would be very easy though.
[10:31:29] <toastersonerson1> they attempted a port aswell but I have no idea if it is based/realted to your work.
[10:31:59] <toastersonerson1> they got stuck with a big endian problem. They also did their work on a later version
[10:34:08] <Aram> if I were to work on this some more, it will probably be on some BSD, as that's the only current thing I could run on SPARC right now.
[10:34:11] <Aram> although the OS part is completely orthogonal to the architecture part.
[10:34:21] <Aram> but I'd have to find more time and motivation.
[10:35:05] <Aram> actually upstream is in a better position nowadays, since there's less churn in the compilers.
[10:35:31] <am11> Debian also has decent SPARC support, no? well, that's what i heard. -.-
[10:35:47] <Aram> ah yeah, sure.
[10:37:12] <Aram> anyway, I will probably do this as some point, just because I don't like leaving unfinished work.
[10:38:28] <toastersonerson1> am11 (IRC): the debian people are/were working with that work aswell.
[10:44:38] *** _joes_ <_joes_!~Thunderbi@santoroj.plus.com> has joined #illumos
[10:46:17] <am11> yup, generally their wide array of ISA support is pretty impressive (at least for devs involved in porting stuff in various worlds). https://wiki.debian.org/Sparc64 is also complete'ish.
[11:10:09] *** _joes_ <_joes_!~Thunderbi@santoroj.plus.com> has quit IRC (Quit: _joes_)
[11:11:45] <Agnar> Oracle has golang 1.7 for Sol11.3 on sparc...
[11:14:43] *** Alasdair <Alasdair!~alasdair@95.131.253.61> has joined #illumos
[11:22:32] <Aram> yes, that's mine.
[11:38:17] *** domag02 <domag02!b03fd4e2@catv-176-63-212-226.catv.broadband.hu> has quit IRC (Remote host closed the connection)
[11:49:21] *** Alasdair <Alasdair!~alasdair@95.131.253.61> has quit IRC (Quit: ZNC 1.7.5 - https://znc.in)
[11:49:41] *** Alasdair <Alasdair!~alasdair@95.131.253.61> has joined #illumos
[12:24:58] <am11> i found four classes of kstat to get swap_free from system counters. {class: kstat, name: kstat_headers}, {class: misc, name: sysinfo}, {class: vm, name: vminfo} and {class: misc, name: var}. in order to get total (global) memory status, should i aggregate these values (like freebsd:
[12:24:59] <am11> https://github.com/dotnet/runtime/blob/b3e791d/src/coreclr/src/pal/src/misc/sysinfo.cpp#L390), or just use vminfo.freemem instead of vminfo.swap_free (per class/name pair)?
[12:25:11] *** am11 <am11!54f985dd@dsl-hkibng22-54f985-221.dhcp.inet.fi> has quit IRC (Remote host closed the connection)
[12:25:47] *** am11 <am11!54f985dd@dsl-hkibng22-54f985-221.dhcp.inet.fi> has joined #illumos
[12:27:11] <am11> i found four classes of kstat to get swap_free from system counters. {class: kstat, name: kstat_headers}, {class: misc, name: sysinfo}, {class: vm, name: vminfo} and {class: misc, name: var}. in order to get total (global) memory status, should i aggregate these values (like freebsd:
[12:27:11] <am11> https://github.com/dotnet/runtime/blob/b3e791d/src/coreclr/src/pal/src/misc/sysinfo.cpp#L390), or just use vminfo.freemem instead of vminfo.swap_free (per class/name pair)?
[12:29:35] <am11> (apologies if the above message got posted twice; webchat reception is flaky today)
[12:34:44] <am11> i've scribbled down a self contained program for experiments https://paste2.org/Fd5geds4.
[13:02:23] *** phyre <phyre!~phyre___@78.30.16.107> has quit IRC (Remote host closed the connection)
[13:03:22] *** TheGreekOwl <TheGreekOwl!~Arkomeda@2a02:587:4853:6a00:7837:e1ba:9069:2f46> has joined #illumos
[13:11:33] *** ldepandis <ldepandis!~ldepandis@unaffiliated/ldepandis> has joined #illumos
[13:18:36] *** am11 <am11!54f985dd@dsl-hkibng22-54f985-221.dhcp.inet.fi> has quit IRC (Remote host closed the connection)
[13:21:02] *** tozhu <tozhu!~tozhu@101.207.125.86> has joined #illumos
[13:26:28] *** Kurlon <Kurlon!~Kurlon@cpe-67-253-136-97.rochester.res.rr.com> has joined #illumos
[13:27:10] *** jimklimov2 <jimklimov2!~jimklimov@ip-86-49-254-26.net.upcbroadband.cz> has joined #illumos
[13:27:38] *** deejam <deejam!deejam@irc.nathanic.org> has quit IRC (Remote host closed the connection)
[13:27:56] *** jimklimov <jimklimov!~jimklimov@ip-86-49-254-26.net.upcbroadband.cz> has quit IRC (Ping timeout: 258 seconds)
[13:28:34] *** am11 <am11!54f985dd@dsl-hkibng22-54f985-221.dhcp.inet.fi> has joined #illumos
[13:28:42] *** Kurlon_ <Kurlon_!~Kurlon@cpe-67-253-136-97.rochester.res.rr.com> has quit IRC (Ping timeout: 258 seconds)
[13:31:43] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[13:32:21] *** deejam <deejam!deejam@irc.nathanic.org> has joined #illumos
[13:35:04] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[13:38:12] *** neirac <neirac!~cneira@pc-4-149-45-190.cm.vtr.net> has joined #illumos
[13:41:03] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[13:45:44] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has quit IRC (Quit: nbhauke)
[13:46:39] *** jcea <jcea!~Thunderbi@2001:bc8:2ecd:caed:7670:6e00:7670:6e00> has joined #illumos
[13:52:41] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has joined #illumos
[14:00:27] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[14:09:23] *** phyre <phyre!~phyre___@78.30.16.107> has joined #illumos
[14:20:39] *** man_u <man_u!~manu@89-92-19-81.hfc.dyn.abo.bbox.fr> has joined #illumos
[14:24:49] *** EisNerd <EisNerd!~manschwet@mail.cs-software-gmbh.de> has quit IRC (Read error: Connection reset by peer)
[14:25:17] *** EisNerd <EisNerd!~manschwet@mail.cs-software-gmbh.de> has joined #illumos
[14:28:06] *** phyre <phyre!~phyre___@78.30.16.107> has quit IRC (Remote host closed the connection)
[14:30:23] *** tozhu <tozhu!~tozhu@101.207.125.86> has quit IRC (Quit: tozhu)
[14:31:11] *** arnoldoree <arnoldoree!~arnoldore@2001:d08:2183:afdb:c1cd:b139:c4b1:918f> has quit IRC (Ping timeout: 252 seconds)
[14:37:04] *** varna <varna!~varna@36.24.136.167> has joined #illumos
[14:45:18] *** arnoldoree <arnoldoree!~arnoldore@2001:d08:2183:afdb:c1cd:b139:c4b1:918f> has joined #illumos
[14:46:41] *** nde <nde!uid414739@gateway/web/irccloud.com/x-wuhedspkfwknirwv> has joined #illumos
[14:54:01] *** TheGreekOwl <TheGreekOwl!~Arkomeda@2a02:587:4853:6a00:7837:e1ba:9069:2f46> has quit IRC (Quit: I must go, my planet needs me.)
[14:55:39] *** gh34 <gh34!~textual@cpe-184-58-181-106.wi.res.rr.com> has joined #illumos
[15:39:02] *** phyre <phyre!~phyre___@78.30.16.107> has joined #illumos
[15:39:13] *** tozhu <tozhu!~tozhu@218.89.244.24> has joined #illumos
[15:39:14] *** tozhu <tozhu!~tozhu@218.89.244.24> has quit IRC (Remote host closed the connection)
[15:47:06] *** sjorge <sjorge!~sjorge@unaffiliated/sjorge> has quit IRC (Remote host closed the connection)
[15:51:18] *** sjorge <sjorge!~sjorge@unaffiliated/sjorge> has joined #illumos
[16:11:19] *** idodeclare <idodeclare!~textual@cpe-76-185-177-63.satx.res.rr.com> has joined #illumos
[16:17:20] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has quit IRC (Quit: nbhauke)
[16:29:08] *** am11 <am11!54f985dd@dsl-hkibng22-54f985-221.dhcp.inet.fi> has quit IRC (Remote host closed the connection)
[16:29:49] *** am11 <am11!54f985dd@dsl-hkibng22-54f985-221.dhcp.inet.fi> has joined #illumos
[16:47:38] *** BOKALDO <BOKALDO!~BOKALDO@81.198.20.220> has quit IRC (Quit: Leaving)
[17:07:21] <jbk> am11: swapctl(2) can tell you about all the swap devices on the system
[17:08:10] <jbk> it's what the swap command uses
[17:08:16] <jbk> http://src.illumos.org/source/xref/illumos-gate/usr/src/cmd/swap/swap.c
[17:08:44] <jbk> sjorge: send me the link again and i'll look at it today
[17:08:46] <am11> jbk: thanks, would it provide the required information https://github.com/dotnet/runtime/issues/34996#issuecomment-614069749?
[17:11:30] <jbk> also, let me create an OS ticket for your other bits
[17:13:31] <jbk> am11: it would help to undertand what exactly they mean by 'committed' memory
[17:14:34] <alanc> FYI, if you haven't seen it yet: https://github.com/0xdea/advisories/blob/master/2020-07-solaris-whodo-w.txt
[17:17:24] <am11> jbk: [excerpt from google in win32 context] Committed Memory is the number of bytes that have been allocated by processes, and to which the operating system has committed a RAM page frame or a page slot in the page file (or both).
[17:32:08] <jbk> if no one else is on it, I'll work on getting those commands fixed this morning.. there's a number of things there that could be improved
[17:34:50] <jbk> sjorge: OS-8158 for your bhyve NVMe PRs
[17:36:31] *** BOKALDO <BOKALDO!~BOKALDO@81.198.20.220> has joined #illumos
[17:40:14] <andyf> jbk thanks! I was looking through the source and the advisory but I can't get to it until later
[17:41:33] <jbk> am11: ok.. so committed memory would probably most closely match the 'used' output of swap -l
[17:43:22] <am11> jbk: in terms of linux, we need values of `totalswap` and `freeswap` from `sysinfo(2)`. i couldn't find these values in Solaris' `sysinfo(2)`, hence took the low-level kstat_open route. if there is an simpler way to get those values in C using some system calls other than kstat, i will give it a try. checking `swapctl(2)` now :)
[17:45:00] <jbk> so ani_max would probably match the total, and ani_free is amount available
[17:45:40] <jbk> ... and of course that option isn't documented in the man page
[17:46:56] <jbk> look at the swap(1) source link I pasted
[17:49:37] *** jimklimov1 <jimklimov1!~jimklimov@78.80.224.132> has quit IRC (Quit: Leaving.)
[17:58:42] *** clapont <clapont!~clapont@unaffiliated/clapont> has quit IRC (Ping timeout: 256 seconds)
[18:02:21] *** TheGreekOwl <TheGreekOwl!~Arkomeda@2a02:587:4853:6a00:7837:e1ba:9069:2f46> has joined #illumos
[18:04:01] <am11> jbk: thanks, i get these values `ai_free: 447653, ai_max: 466585` from C impl. and `swap -l` reports: `/dev/zvol/dsk/zones/swap 90,1 8 3143672 3064176`
[18:04:56] *** varna <varna!~varna@36.24.136.167> has quit IRC (Ping timeout: 256 seconds)
[18:05:11] *** varna <varna!~varna@60.176.146.221> has joined #illumos
[18:05:45] *** clapont <clapont!~clapont@unaffiliated/clapont> has joined #illumos
[18:11:48] <jbk> those are in units of pages (from the function) IIRC
[18:12:01] <jbk> so i think you'll need to get the page size and multiply to get bytes
[18:19:39] *** TheGreekOwl <TheGreekOwl!~Arkomeda@2a02:587:4853:6a00:7837:e1ba:9069:2f46> has quit IRC (Quit: I must go, my planet needs me.)
[18:23:21] <sjorge> jbk: I'll update the PR commits when I am near my computer
[18:25:16] <sjorge> As for the pam_list, same link https://github.com/joyent/illumos-joyent/compare/master...sjorge:pam_list?expand=1
[18:26:26] *** tsoome_ <tsoome_!~tsoome@f046-b0d8-ca03-e9cd-2f80-4a40-07d0-2001.sta.estpak.ee> has quit IRC (Quit: This computer has gone to sleep)
[18:28:47] *** tsoome_ <tsoome_!~tsoome@148-52-235-80.sta.estpak.ee> has joined #illumos
[18:30:30] *** ChrisBF <ChrisBF!56890402@host86-137-4-2.range86-137.btcentralplus.com> has joined #illumos
[18:30:32] *** tsoome_ <tsoome_!~tsoome@148-52-235-80.sta.estpak.ee> has quit IRC (Client Quit)
[18:30:41] *** tsoome_ <tsoome_!~tsoome@148-52-235-80.sta.estpak.ee> has joined #illumos
[18:35:10] <am11> with `* pagesize`, it is: `free: 1833586688, total: 1911132160`
[18:35:29] <am11> i.e. `int pagesize = getpagesize();`
[18:35:58] <jbk> sjorge: no worries.. just wanted to get that knocked out for you
[18:41:13] <jbk> for the pam_list -- line 334 -- 'continue' should be on its own line..
[18:42:10] <jbk> also (not necessary, just a suggestion) -- if the realloc (could just be malloc) fails, you could just emit an error and continue, and not have to have the rest of it indented so far
[18:49:58] *** clapont <clapont!~clapont@unaffiliated/clapont> has quit IRC (Read error: Connection reset by peer)
[18:52:08] *** ChrisBF <ChrisBF!56890402@host86-137-4-2.range86-137.btcentralplus.com> has quit IRC (Remote host closed the connection)
[18:57:47] *** neuroserve <neuroserve!~toens@ip-84-118-130-67.unity-media.net> has quit IRC (Ping timeout: 256 seconds)
[19:01:16] *** clapont <clapont!~clapont@unaffiliated/clapont> has joined #illumos
[19:04:13] *** arnoldoree <arnoldoree!~arnoldore@2001:d08:2183:afdb:c1cd:b139:c4b1:918f> has quit IRC (Quit: Leaving)
[19:18:37] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has joined #illumos
[19:19:07] <sjorge> what do you mean with the realoc bit?
[19:20:18] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has quit IRC (Client Quit)
[19:22:00] *** neuroserve <neuroserve!~toens@ip-178-202-216-85.hsi09.unitymediagroup.de> has joined #illumos
[19:25:23] *** Alasdair <Alasdair!~alasdair@95.131.253.61> has quit IRC (Quit: ZNC 1.7.5 - https://znc.in)
[19:25:38] *** Alasdair <Alasdair!~alasdair@al.cloud.ec> has joined #illumos
[19:40:58] *** tsoome <tsoome!~tsoome@91.209.240.229> has quit IRC (Read error: Connection reset by peer)
[19:42:27] *** tsoome <tsoome!~tsoome@91.209.240.229> has joined #illumos
[19:51:37] *** tsoome <tsoome!~tsoome@91.209.240.229> has quit IRC (Read error: Connection reset by peer)
[19:51:56] *** tsoome <tsoome!~tsoome@91.209.240.229> has joined #illumos
[20:02:23] <jbk> oh also.. protential multiple free in there..
[20:03:01] <jbk> you free grbuf, but that appears to be inside the while loop processing each line
[20:03:31] <jbk> the next time around and you do realloc(grbuf, grbuflen)
[20:03:41] <jbk> maybe something like
[20:06:07] <jbk> prior to the start of the loop at line 224, I'd just try to allocate grbuf if check_group is set, log a message if it fails (and return PAM_BUF_ERR)
[20:06:47] <gitomat> [illumos-gate] 12278 nfs-zone needs man page changes -- Dan McDonald <danmcd at joyent dot com>
[20:06:50] <jbk> then just free it after the loop
[20:07:28] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has joined #illumos
[20:07:55] *** KeiraT- <KeiraT-!~k4ra@gateway/tor-sasl/k4ra> has joined #illumos
[20:11:23] *** KeiraT <KeiraT!~k4ra@gateway/tor-sasl/k4ra> has quit IRC (Ping timeout: 240 seconds)
[20:11:25] *** KeiraT- is now known as KeiraT
[20:13:48] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has quit IRC (Quit: nbhauke)
[20:25:13] *** mnrmnaugh <mnrmnaugh!~mnrmnaugh@unaffiliated/mnrmnaugh> has quit IRC (Ping timeout: 250 seconds)
[20:30:02] *** mnrmnaugh <mnrmnaugh!~mnrmnaugh@unaffiliated/mnrmnaugh> has joined #illumos
[20:31:08] *** tsoome <tsoome!~tsoome@91.209.240.229> has quit IRC (Read error: Connection reset by peer)
[20:31:20] *** tsoome__ <tsoome__!~tsoome@91.209.240.229> has joined #illumos
[20:51:34] <sjorge> Don't we need to realloc it each iteration?
[20:52:08] <sjorge> Ah I guess not even if we get aaaa then bbb as group it will still fit in the slot, and stop at the \0
[20:57:00] *** ghost1 <ghost1!~ghost1@c-73-245-52-228.hsd1.fl.comcast.net> has joined #illumos
[21:41:20] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has joined #illumos
[21:52:34] *** ghost1 <ghost1!~ghost1@c-73-245-52-228.hsd1.fl.comcast.net> has quit IRC (Quit: ghost1)
[21:52:51] *** ghost1 <ghost1!~ghost1@c-73-245-52-228.hsd1.fl.comcast.net> has joined #illumos
[22:14:35] <jbk> no.. i'd need to look, but what i suspect is happening is that it writes the line from /etc/group (or a constructed equivalent if using ldap or nis), then splits it into individual strings at the colons and such, and then sets the struct group members to those values
[22:14:49] *** BOKALDO <BOKALDO!~BOKALDO@81.198.20.220> has quit IRC (Quit: Leaving)
[22:23:22] *** KungFuJesus <KungFuJesus!~adam@72.49.160.91> has joined #illumos
[22:23:53] <KungFuJesus> jbk: good news, I sufficiently managed to prod Saso Kiselkov on his feature request / work
[22:23:58] <KungFuJesus> hey may be contacting you soon
[22:26:46] <KungFuJesus> he replied back saying that he's ok with you integrating his changes back
[22:26:50] <KungFuJesus> This is his reply: Absolutely he can integrate my work. All I did was
[22:26:50] <KungFuJesus> modify some CDDL code, so it's still CDDL licensed. But be aware there
[22:26:50] <KungFuJesus> were still some open issues left in the GCM code, so some testing &
[22:26:50] <KungFuJesus> bugfixing might be required. Unfortunately, my workload doesn't permit
[22:26:57] <KungFuJesus> me to work on this.
[22:27:03] <KungFuJesus> ahh, sorry for the multi-line paste
[22:33:11] *** jimklimov <jimklimov!~jimklimov@ip-86-49-254-26.net.upcbroadband.cz> has joined #illumos
[22:35:08] <jbk> KungFuJesus: cool.. the big concern was to make sure the copyright notices on the stuff he did was correct -- some of them looked like they might be missing some updates, and i didn't want to make any assumptions
[22:41:29] *** nbhauke <nbhauke!~hauke@55d4dbd3.access.ecotel.net> has quit IRC (Quit: nbhauke)
[23:00:59] <jbk> rmustacc: when you have a chance (since i suspect you'll want to review the end result, but before I get too far) -- I'd like to add a proc_get_sigact() to libproc(3PROC) -- https://us-east.manta.joyent.com/jbk/public/tmp/proc_get_sigact.html is what i'm proposing for the interface
[23:01:24] <jbk> i'm more just interested in if the proposed interace seems reasonable -- word massaging of the manpage can happen during review if needed
[23:01:31] <jbk> (before I get too far into this)
[23:02:04] *** wacki <wacki!~wacki@i577A59B5.versanet.de> has quit IRC (Quit: Be back later ...)
[23:02:28] <jbk> (AFAIK, while rare, it's possible the number of signals defined on the system could increase in the future, so I didn't want to bake in NSIG)
[23:02:46] *** wacki <wacki!~wacki@i577A59B5.versanet.de> has joined #illumos
[23:07:34] *** wacki <wacki!~wacki@i577A59B5.versanet.de> has quit IRC (Ping timeout: 240 seconds)
[23:27:50] *** man_u <man_u!~manu@89-92-19-81.hfc.dyn.abo.bbox.fr> has quit IRC (Quit: man_u)
[23:34:21] <jbk> pmooney: btw i have most of the code to demangle the newer rust scheme done (don't know if that's something you guys might need) -- there's still just a few corner cases I need to chase down (which I was doing before I got tied down on EDAR)
[23:34:35] <pmooney> that's exciting
[23:34:42] <jbk> we haven't needed it yet, but in case you might..
[23:34:44] <pmooney> it'd be nice to have a better experience for that
[23:35:01] <jbk> one of the rust guys has like a multi-gb file of test cases he was running my implementation against
[23:35:03] <pmooney> the main PR for rust on illumos is on its way in
[23:35:12] <jbk> yeah, I saw :)
[23:39:49] <KungFuJesus> EDAR, that sounds like another name for a side channel vulnerability?
[23:39:54] <jbk> haha no
[23:39:58] <jbk> encrypted data at rest
[23:40:15] <jbk> basically the bits in triton and smartos for managing zfs encryption keys
[23:40:58] <KungFuJesus> oh ok, was worried it was another flavor of MDS, hah
[23:45:42] *** gh34 <gh34!~textual@cpe-184-58-181-106.wi.res.rr.com> has quit IRC (Quit: Textual IRC Client: www.textualapp.com)
[23:47:01] *** hemi770_ <hemi770_!~hemi666@unaffiliated/hemi770> has joined #illumos
[23:49:58] *** hemi770 <hemi770!~hemi666@unaffiliated/hemi770> has quit IRC (Ping timeout: 256 seconds)
top

   April 15, 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 | >