00:51:17 <gwr> Anyone here familiar with the rules about sigaltstack and signal handers?
00:51:46 <gwr> eg what if the signal handler does not return? (and we take another signal)
03:02:30 <sommerfeld> gwr: looks to me like it just pushes another frame on the alternate stack and keeps going.   see how "newstack" is computed in sendsig() in uts/intel/os/sendsig.c
03:03:42 <gwr> Thanks.  Looking at some complicated code in dotnet that switches stacks, but leaves SS_ONSTACK set.
03:04:01 <sommerfeld> it only sets the new fp to the top of the alternate stack if SS_ONSTACK is cleared in the signal stack state.
03:04:26 <gwr> I think maybe it should clear that if it uses setcontext to leave the alt stack?
03:05:27 <sommerfeld> depends if there's any live state on the alt stack that would be clobbered by a fresh signal.
03:05:37 <gwr> Though I'm not sure we ever take more than one signal that has SA_ONSTACK in the sigaction.
03:06:38 <gwr> There's some weird stuff going on, and I wondered if it was conforming to expectations around our signals implementation.
03:07:34 <richlowe> So, an interesting thing we've sort of run into before is that this is all a mess
03:07:53 <richlowe> a lot of the nitty gritty that turns out to matter isn't really specified
03:08:42 <richlowe> so you really do have to go through looking at what they tried to make happen, and what we're going to do (just implementation-wise)
03:10:56 <richlowe> gwr: have you looked at any nested stuff via uc_link?
03:11:02 <richlowe> is dotnet?
03:11:28 <gwr> yes
03:12:07 <gwr> Yes, dotnet.  I have not noticed any uc_link following
03:12:32 <gwr> Just stack switching beyond what signal handlers may do for you.
03:13:12 <gwr> and generally non-portable context mucking abount.
03:14:03 <richlowe> yeah, stuff poking in contexts is an exciting mess :\
03:22:56 <gwr> where is setcontext hiding? (opengroc doesn't show it?)
03:28:25 <gwr> oh, it's a syscall
03:31:04 <gwr> no in sigaction. huh.
11:08:37 <unixery> Hi, I've got a question about zoned zfs datasets: If I have a zoned dataset, that is mount in zoneA, can I somehow also access the files in zoneB? E.g. via some kind of bind mounts? But not via network shares (higher latency).
11:28:18 <ptribble> Not a delegated dataset. If you want to share a dataset between multiple zones, create it in the global zone and share it as an fs resource (type=lofs) into each zone that needs it
11:32:36 <unixery> OK. My plan was to run the SMB server in a zone with zoned datasets. But the media server in another zone should have non network access to the files.
11:33:03 <unixery> Then I'll put the SMB server in the global zone
16:25:33 <papertigers> unixery: fwiw you can still share via SMB in a zone from a lofs mount
22:04:00 <jbk> yeah, anything visible in the zone should be able to be shared via SMB
22:04:08 <jbk> (yay power is back)
22:04:44 <richlowe> glad you're above water, etc, jbk
22:10:59 <nomad> How long was it gone?
22:41:49 <jbk> about aa day and a half
22:42:27 <jbk> my internet is still out (using my phone).. though I think that's starting to get congested as a result
22:43:30 <jbk> richlowe: i'm on the 4th floor, so if i'm under water, there's far more serious problems :)
22:44:05 <jbk> the flooding though from the hurricane wasn't too bad..
22:48:56 <jbk> not much different from a heavy downpour and had mostly drained by the afternoon.. not like Ike where I-10 still looked like a river the next day