-
gitomat
[illumos-gate] 9633 cxgbe should not reinvent the bool -- Patrick Mooney <pmooney⊙pc>
-
tsoome_
sigh. # prctl -n zone.cpu-shares -v 10 -r -i zone omni
-
tsoome_
prctl: omni: No controllable process found in task, project, or zone.
-
tsoome_
prctl -i zone omni has quite normal and expected output, however...
-
tsoome_
and unfortunately, FSS has no effect on bhyve zone.
-
jclulow
tsoome_: As I recall, prctl tries to find a representative process in the zone, which is usually init, but the bhyve brand is a bit wonky (in that it doesn't include the usual tree of processes)
-
jclulow
We probably shouldn't be using the agent LWP to effect prctl actions
-
jclulow
I wonder if it's also a privileges thing
-
jclulow
(not sure what cred the agent operates with...)
-
tsoome_
hm, thats unfortunate, but if it is really looking for init, then it explains the prctl error.
-
tsoome_
I havent checked the prctl source (yet?)
-
gitomat
[illumos-gate] 17002 want wcslcat(3C) and wcslcpy(3C) -- Robert Mustacchi <rm⊙fo>
-
gitomat
[illumos-gate] 17003 fix wcsdup(3C) synopsis -- Robert Mustacchi <rm⊙fo>
-
richlowe
since I suspect we might be one of the few people who use it, since AT&T sucked for so long, do we happen to have a potted document for "how to use cscope and be happy"?
-
andyf
jclulow, tsoome_ - is is a privileges thing. In a bhyve zone (at least on OmniOS), bhyve is the zone's init process and it has very limited privileges. In particular it does not have PRIV_SYS_RESOURCE which prctl needs its victim/helper to have.
-
richlowe
makes sense
-
andyf
You can still set shares via zonecfg, I just don't know how effective they are.
-
tsoome_
they are not effective at all with bhyve zone
-
jclulow
Is that because most of the CPU usage is system time, and FSS doesn't apply there?
-
jclulow
(obviously it should!)
-
gitomat
[illumos-gate] 17037 PCI headers missing dependencies -- Robert Mustacchi <rm⊙fo>
-
andyf
I've just never validated it.
-
richlowe
I can see reasons it would probably not be great