[illumos-gate] 15401 Spelling mistakes in section 3c of the manual -- Peter Tribble <peter.tribble⊙gc>

anyone here who is very friendly with python parser module?:)

the one removed in 3.10 ....

tsoome_: is all ident cruft removal integrated?

yes

nice, thanks

except for sparc tree

[illumos-gate] 15325 bhyve upstream sync 2023 January -- Andy Fiddaman <illumos⊙fn>

hrm...

do we build the mac module in some what that causes dtrace to lose the last few stack frames in stack() ?

There are a lot of tail calls in mac IIRC.

that's a side issue though...

i need to do some more source diving, as I think I had this question before but never got it resolved

mac_group_info(9s) suggests that the mgi_{add,rem}mac members of mac_group_info_t are optional... but mac.c:4437 suggests it's not (imagine a device that supports multiple RX rings for RSS, but only a single HW mac address)

i thought that'd imply 1 RX group, but now I'm not sure how you actually correctly express that to mac

jbk: It may be that when I wrote that I was imagining we had the combo mac/vlan entry that never actually made it in. It is required right now. I'll fix the docs, sorry.

Can you clarify what exactly the hardware features are and I can describe how to do that while working on docs in parallel?

in this case (vmxnet3s), it's multiple RX rings (I think it's implied that RSS is used) for parallelism, but only a single HW mac address is supported

OK, so you'll want to implement the entry point and basically return ENOSPC for anything other than the primary mac.

the problem is creating a vnic ontop of it then fails (instead of switcing the promisc mode)

That'll tell it to put the device in promisc.

ahh ok

Probably should make that a little clearer in the docs, sorry.

it's ok.. probably the best way to flush these things out :)

Mind filing a bug noting these and any other gotchas and assign it to me?

Just so I don't lose track of this.

sure.. (might be a little bit.. about to jump on acall)

At your convenience.

bah

Has anybody a guide to the pkgdepend output when it doesn't find a library?

it tells you where it looked, and what it looked for.

a valid library isn't in the places it looked. I think it also fails if it's for the wrong platform, etc.

if you gist it so I remember what it looks like I can be more specific.

usr/lib/stonith/plugins/stonith2/wti_nps.so is looking for an ELF dependency, in either: lib/libssp.so.0 or usr/gcc/10/lib/libssp.so.0, and not finding it.

if libssp.so.0 exists in those places, check it's the same ELF machine, etc.

if it doesn't, fix the runpaths of wti_nps.so to find where libssp actually is

/usr/lib would be a good guess, since you seem to be mysteriously not looking there

hmm yet the second exists....

libssp.so is in that place on OI

and file tells you it's for the same platform as wti_nps.so?

my guess would be wti_nps.so is for amd64, and you don't have .../64 in the runpaths.

huh. I didn't even know we built 32bit gcc still

yeah it's 32bit

I'll head out and edit runpaths :) thanks

[illumos-gate] 15404 The rup.1c manual page unnecessarily duplicates rup.1 -- Peter Tribble <peter.tribble⊙gc>