03:03:13 <sommerfeld> alanc: fortunately intel popcount instruction operates on regular integer registers, but it's not found on cpus older than nehalem (intel) or barcelona (amd).   (and I've got one that's too old that I still occasionally use as a test machine).
03:04:51 <jbk> and no linker capabilities in the kernel :P
07:16:52 <gitomat> [illumos-gate] 18311 format: enable smatch -- Toomas Soome <tsoome⊙mc>
16:37:10 <jbk> hrm.. is it jsut 'no one's done it yet' to add a topo module to enumerate NIC leds, or is there some other reason we don't (when present)?
16:47:04 <rmustacc> At least for me, the former.
18:40:35 <gitomat> [illumos-gate] 18411 format: remove __STDC__ -- Toomas Soome <tsoome⊙mc>
19:14:50 <gitomat> [illumos-gate] 18413 zfs: spa_sync_upgrades() should only take lock when needed -- Toomas Soome <tsoome⊙mc>
20:28:02 <jbk> given we don't have man pages for the kernel secpolicy_xxx() functions, would those not be considered DDI interfaces (at least at the moment)?
20:44:18 <rmustacc> Correct, only drv_priv() is considered DDI.
21:00:28 <richlowe> the DDI-ness of things is kind of a mess, and I wouldn't worry too much about straying if I was in illumos-gate
21:00:46 <richlowe> though some day I'd love to go through and look at the parts everyone cheats on and just promote them.
21:01:15 <richlowe> but we've had problems in the past where people misguided tried to treat the DDI as gospel, and don't do that, it's bad for you
21:08:01 <rmustacc> If you have a list of some of them, Rich, happy to go through and work on some of that.
21:08:39 <richlowe> I don't unfortunately, I just sometimes run into stuff on the arm side where I ask people to look for the DDI way and then there isn't one :)
21:08:59 <rmustacc> Okay, happy to work on gaps too.
21:09:27 <jbk> the ice NIC can generate log messages from the HW (or I guess the embedded proc inside the nic), and given the rather large complexity i thought it would be nice to have some ioctls to enable emitting those (like the freebsd driver)... but also been trying to keep the interfaces used tidy w/ mapfiles
21:09:46 <richlowe> rmustacc: the most recent thing I remember being generally useful (question mark?) was a "compatible" matcher
21:09:50 <jbk> so if there is anything needed outside of mac/ddi/random it's essentialyl documented in one place
21:10:31 <jbk> but the precedent for NIC ioctls seems to be having net_config
21:10:55 <jbk> (don't know if that's a good precedent or not, but it is what it is)
21:14:03 <jbk> i finished getting the flow director and flex pipelines for RSS (kinda gave up and just ported the fbsd code -- there's likely opportuntiy to simplify things a fair amount for our use case, but trying to tease apart all of that is going to take waaaay to long)
21:14:18 <jbk> (just ported those bits of functionality)