Hello dear qubes developers. I have had the opportunity to test Qubes on two similar (though not strictly identical) configurations, find, reproduce and and seemingly alleviate a functionality breaking behavior and experience some weirdness with IGFX (I am using IGFX so far because I can't quite bring myself to attempt installing nvidia GPU drivers. Will likely do so soon enough though...)
First, about the issue I've managed to seemingly alleviate Box 1: Asus ROG Hero VIII motherboard 32 GB PC-3200 RAM DDR4 Skylake core i7 6700k adaptec 5805z RAID controller vt-d enabled. Nvidia GTX 1070 GPU Double PSU config (500 w each) Box 2: Asus ROG Hero VIII motherboard 64 GB PC-2800 RAM DDR4 Skylake core i7 6700k vt-d enabled. Nvidia GTX 1070 GPU 1200 w PSU Box 1 was tested in "slight overclock" and "stock" clocks Box 2 was tested only in stock mode (it's boyfriends, don't wanna do something funny with it) First finding: the board has two USB controllers, and passing any of them to a VM (to use ASUS WL-167 G v3 USB wireless dongle) results in a certain peculiar instability that replicates perfectly on Box 1 irrespective of overclock (stock clocks make it manifest a little bit faster, oddly) and also replicates perfectly on Box 2. It manifests by intermittent loss of USB wireless dongle functionality (no packets can actually be sent or received) without any messages being present in dmesg or NetworkManager logs. After the even happens, only pulling out the USB and re-plugging it again can solve the issue (until it re-manifests), restarting netvm does not help (but pullout-replug thing can be done with VM running, and it will re-detect the dongle and continue download okay-dokay) This issue, however, manifests only when large (>1GB) files are being downloaded for a long (>10 minutes) time, so launching several downloads (such as, say, dom0 update installing fedora-24 template which is how I first encountered the issue) It appears to me that the issue is very similar to one reported here: https://groups.google.com/forum/#!topic/qubes-users/WGtXebJxbM4 Through trial and error, I have found out that manually setting PCH voltage to 1.125 and DMI voltage to 1.24 reduces the probability of this issue manifesting. It does not remove it completely (downloading 8 GB still causes the dongle to "pass out" about 4 times out of 10) but it makes connection stable enough to download fedora template. I do suspect that raising the voltages some more would "bring in more stability", but I am not convinced it is "worth it". Issue so far reproduced on two similar ASUS mobos and similar CPUs bought by two different people at different shops, and it is likely that it would be possible to reproduce it on a similar configuration, or perhaps on some other config as well. Improvement from abovementioned voltage-raising trick remains consistent irrespective of whether box is overclocked (and in fact appears slightly more effective when applied to "Box 1" when box 1 is overclocked, but I'm not sure since I can force myself to tediously re-download an 8GB file only so many time.) Oh, and, issue is present irrespective of whether debian or fedora template is used for NetVM and absent completely when downloading huge stuff through windows 10) Sadly, I was not able to completely resolve the "USB wifi fainting" thing, merely alleviate it. -------------------- Second issue (the IGFX one) that I've come to experience is intermittent artifacting (fast distortions persisting for ~1 second and/or screen turning off and on again) on HDMI connected monitor that happens whenever I start a VM that has a PCIe device (USB controller or RAID controller) attached to it. No solution was found despite extensive trial and error. Not functionality breaking so far. Reproduces on Box 1 and Box 2. I suspect it has to do with the way vt-d interacts with Intel integrated graphics, and Google suggests that integrated graphics and/or GPU in general can be excluded from vt-d/IOMMU, but I have not yet figured out how to do that on Qubes (help appreciated!) P.S.: I'd like to also figure out how to get shutdown logs that happen after the encrypted LVM and most services are already down. There seems to be a curious pcie error messages happening right before it reaches shutdown, but things happen way too fast for me to read properly. Thank you very much for Qubes, other than those two glitches it's been working out pretty okay for me. I hope this report helps figure out and perhaps fix what is going on here. -- You received this message because you are subscribed to the Google Groups "qubes-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/qubes-devel/2719bbcc-3a8f-4a51-a972-50934f67a9c5%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
