IRC channel logs

<azeem>is there a way to not use rumpdisk, but still get HPET via APIC? I am still trying to run a continuous PostgreSQL testsuite, but it keeps hanging during the first run, so I would like to see whether switching away from rumpdisk might make things more stable

<youpi>I don't think the linux glue was fixed for APIC use

<azeem>youpi: ok, thanks

<azert>yeah better focus on fixing rumpdisk that needs some fixing

<azert>one problem is clearly swapping since even with the Debian installer when I tried it naively it hanged just after creating a swap partition

<azert>and by defaults it does that

<azert>if the main reason is the race that Michael found on the mailing list, then it seems like he already has a fix which is cool

<azert>azeem: are your hangs due to out of memory? Did you try increasing it?

<azeem>doesn't look like

<azeem>I'm running free every two seconds on the console, and am never out of memory

<azeem>during the last hang, rumdisk did a lot of cpu: http://paste.debian.net/1392995/

<azeem>and yeah, I was going to rebuild gnumach/hurd with Michael's patches to see whether that helps

<azert>azeem: I see you have swap

<azert>try disabling that completely

<azert>in my opinion it hangs super fast with it

<azert>I told you how it happened to me with the installer

<azert>azeem: I didn’t put the system under any stress

<azert>it’s just that if it has swap it might end up using it evem

<azert>even if it could not

<azeem>now I got (after deactivating swap): (default pager): previous write error for object, killing it

<azeem>at least directly before, free showed 184112 KB free

<azert>azeem: maybe you can get rid of the swap partition alltogether

<youpi>azeem: 184MiB is not necessarily much, it might be trying to evict some memory to swap then

<youpi>disabling swap altogether, and making sure that mach-defaultpager is not run at all (i.e. no tmpfs mount at all), avoids the issue

<azeem>ok, I'll try that, thanks

IRC channel logs

2025-08-23.log