libre #smartos

18 Sep 2023

SmartOS -- A Modern, Open Source, Datacenter Operating System | smartos.org | github.com/TritonDataCenter/smartos-live | github.com/TritonDataCenter/triton | Security -- TritonDataCenter.com/security-issues | Channel logs - log.omnios.org/smartos

• 10:39
  jperkin
  jzu: just to close the loop on the samba thing, I backported the latest 4.17.x to 2022Q4 and manually applied the memset_s fix as it's not available in that branch
• 15:10
  danmcd
  sjorge: lemme look...
• 15:11
  danmcd
  Maybe this? setsockopt(4, SOL_IPV6, IPV6_RECVHOPLIMIT, [1], 4) = 0
• 15:11
  danmcd
  I'll double-check.
• 15:12
  danmcd
  Nope.
• 15:12
  danmcd
  Lemme try again.
• 15:13
  danmcd
  IPV6_RECVERR
• 15:13
  danmcd
  Yeah... we don't support that in illumos, therefore:
• 15:13
  danmcd
  usr/src/uts/common/brand/lx/syscall/lx_socket.c: { OPTNOTSUP, 0 }, /* IPV6_RECVERR */
• 15:14
  danmcd
  We'd need it and its IPv4 equivalent before plumbing them both down to LX.
• 15:16
  danmcd
  usr/src/uts/common/brand/lx/sys/lx_socket.h:#define LX_IP_RECVERR 11
• 15:16
  danmcd
  usr/src/uts/common/brand/lx/sys/lx_socket.h:#define LX_IPV6_RECVERR 25
• 15:18
  danmcd
  Even FreeBSD's stock TCP/IP stack doesn't support this. It's very much a Linux-ism, and from my cursory reading of the man page... quite annoying to build.
• 15:22
  danmcd
  Analysis in the OmniOS ticket.
• 15:42
  sjorge
  Is it an option to lie somehow so that ipv6 name resolution works, It's kind of a pain. Then again I'm not sure what IP(V6)_RECVERR is supose to do.
• 15:43
  sjorge
  Aha it's in the IP man page on linux man7.org/linux/man-pages/man7/ip.7.html
• 15:43
  sjorge
  "When the user receives an error from a socket operation, the errors can be received by calling recvmsg(2) with the MSG_ERRQUEUE flag set."
• 15:44
  sjorge
  That doesn't mention they SHOULD be, so I'd be tempted to lie that we will and never do ^.^
• 15:44
  sjorge
  Then again I'm also a bit evil at times
• 15:45
  sjorge
  Seems a lot of work to properly plumb up though as you mentioned
• 15:45
  sjorge
  I vaguely remember us lying un LX at another place too ?
• 15:46
  sjorge
  I wonder if freebsd hits the same issue then with there linux layer
• 17:22
  danmcd
  See my analysis for stock FreeBSD.
• 17:22
  danmcd
  I will point out they ahve a global to enable lying in some cases. Paste in a sec...
• 17:23
  danmcd
  gist.github.com/danmcd/3742495564a2546829c86ad01ba85030
• 17:37
  sjorge
  Aha specifically for DNS too
• 17:37
  sjorge
  Would it be acceptable for us to do so too?
• 17:38
  sjorge
  I don't have high hopes that I can make the change, but after having done the viona one, I'm willing to give that a go if I know it will likely be accepted
• 17:38
  sjorge
  Well I guess, thats more of a danmcd + andyf specific question I guess
• 17:38
  sjorge
  As it's not upstream :D
• 17:38
  danmcd
  FreeBSD feels it should lie about it ONLY if a global is whacked. I wonder if we can be more subtle about it, perhaps as an LX zone property?
• 17:39
  danmcd
  There's design work to be done here.
• 17:39
  sjorge
  Heck I'd be happy with a tunable :D
• 17:39
  sjorge
  I know how that works now code wise *smirks*
• 17:39
  sjorge
  Yes, I will now use this hammer for all nails, but also all screws I think I can hit with it
• 17:40
  sjorge
  The freebsd one seems to be for IPv4 though? Or do they have one for IPv6 too?
• 17:40
  sjorge
  Because for me at least it seems opnly IPv6 name resolution is busted
• 17:44
  sjorge
  How much context does LX have about whatever it's emulating, can it detect, this is glibc's dns resolver code path ?
• 17:49
  sjorge
  Of interest, when using static /etc/hosts entries ping complains about IPV6_RECVERR too
• 17:49
  sjorge
  ping: IPV6_RECVERR: Protocol not available
• 17:49
  sjorge
  Why does linux do X, comes up to often for me -_-
• 17:49
  sjorge
  Followed by, why can't people just write software that argets posix
• 17:49
  sjorge
  *targets
• 17:51
  jbk
  unfortunately a lot of developers just don't pay attention
• 17:52
  jbk
  and I think illumos and solaris are the only ones that have been really strict about 'if you want POSIX and only POSIX, here you go'
• 17:53
  jbk
  (for better or for worse)
• 18:00
  sjorge
  Oh great
• 18:00
  sjorge
  sourceware.org/pipermail/glibc-cvs/2020q4/070915.html Seems there are 'variants' too
• 18:11
  danmcd
  Yep...
• 18:11
  danmcd
  "Interface stability... we've heard of it!"
• 18:17
  sjorge
  It should somehow be delt with somehow though under lx
• 18:17
  sjorge
  Be it by lying in some cases or just supporting it
• 18:19
  bahamat
  I'm not sure they have heard of it.
• 18:22
  sjorge
  Interface stability, posix, or both? :p
• 18:28
  sjorge
  I'm not imediatly finding any of the other BSD's having it either
• 20:03
  copec
  How do you debug ` Invalid JSON payload: Unexpected token ` so you can see where it stumbles?
• 20:03
  copec
  ^ vmadm create
• 20:08
  bahamat
  copec: json -f <file>
• 20:08
  bahamat
  That should show you exactly where the parsing error is.
• 20:15
  copec
  ty
• 20:21
  copec
  jperkin, You should add dos2unix to the smartos gz tools pkgin
• 20:33
  nbjoerg
  copec: what's wrong with tr?
• 21:00
  copec
  I dunno, it didn't work, probably because it was a spliced up file, so I just ran dos2unix in joyent brand zone that had it in pkgin and it converted it fine
• 21:02
  nbjoerg
  "tr -d '\r' " is essentially all that dos2unix is doing
• 21:02
  copec
  It has fixed stuff in the past for me, I think they have some logic in there for trailing spaces
• 21:06
  jperkin
  yeh tr should fix it but I can add it anyway