src.illumos.org is lagging behind with repo updates, can we kick it a bit:D

jclulow might be the one to do that... ^

[illumos-gate] 15535 prof: remove prof -- Toomas Soome <tsoome⊙mc>

it's still early on the west coast

[illumos-gate] 15554 loader: add strnlen() -- Toomas Soome <tsoome⊙mc>

[illumos-gate] 15494 sun_fc: the address of 'HBA_wwn::wwn' will never be NULL -- Toomas Soome <tsoome⊙mc>

It should be better again, sorry about that

thanks!

[illumos-gate] 15562 xref in the cold after assembly file rename -- Richard Lowe <richlowe⊙rn>

[illumos-gate] 15544 Makefile.buildnum's time has passed -- Richard Lowe <richlowe⊙rn>

So anyone here run the ZFS test suite using 3 zvols as DISKS? If so, do I:

1.) Need anything special in zvol creation?

2.) Need anything special to set in DISKS, as it appears to assume /dev/dsk/<entry> ?

the most reasonable way to do that is to give the zvols to bhyve

I didn't succeed at doing what you appear to be trying

I'm about to find out the hard way I guess.

(I didn't fully succeed after that either, because the zfs tests seem so damn flakey)

zfstest script seems to understand if you spec /full/path/to/file for an entry in DISKS.

:crossed_fingers:

what i remember is: illumos.org/issues/821 ("nested" zfs on zvols didn't really work for me, had other "interesting" issues), not really tried again since then :D

→ BUG 821: `zfs diff' locks up diff'ing snapshot and fs on zvol backed pool (New)

quick check, based on recent negative results, I just want to make sure a card we're getting in a new file server is actually supported. Does anyone know if a "Mellanox 25-Gigabit Ethernet Adapter ConnectX-6 Dx - PCIe 4.0 x8 - 2x SFP28" will work ok?

nice. I have found core from fmd, apparently it was result of dereferencing NULL fmd.d_errlog pointer in fmd_verror(). altho, I have no idea why fmd.d_errlog is NULL there:D

kebe: Putting ZFS on top of zvols is almost guaranteed to excite some kind of pathology, especially if you're testing changes, given the ZFS under test is on top of the ZFS being used to hold the ZFS under test etc

nomad: Do you have the PCI IDs?

I couldn't even get it to start zfstest so that's probaby good.

jclulow: I have an educated guess based on the board's specs:

1000,e[4567]

FreeBSD's version of this driver *appears* to not have needed many changes from 34xx, BUT I may be missing something so pour out your multiple grains of salt.

jclulow, I don't have the card yet. Concerned about buying it based on some info kebe has provided.

(Namely that I dont think it'll attach with stock illumos.)

I'd love to be wrong about that too, however.

oh, of course, the stack also has fmd_destroy(), so it should be death trap from stopping the service:)

How much faith should I have that it'll be solid and supported sometime in the near future?

(it being the ConnectX-6)

I guess we won't be buying that new fancy NIC this time.

maybe on the next file server.

kebe: those will attach to lmrc when it's ready

Really? I thought lmrc was HW-RAID, not pure SAS/HBA...

hm, maybe you're right. it looks like it'll just be 10e[0-7]

[illumos-gate] 15508 zfstest: zpool_initialize_verify_initialized fails to get data -- Brian Behlendorf <behlendorf1⊙lg>

tsoome: Possible to share the core?