darianf

just a though, will try to play with the usb controllers settings in bios to see if it matters how the hand-off to the OS occurs (enabled / disabled)

Same issue here... Unraid 6.3.2 My unraid box crashed when i tried to restart a Linux VM, following a dist-upgrade and nvidia driver update... Unraid froze... As a result, the windows 10 VM which was running in parallel at that time is not starting anymore. few days earlier, the same win 10 vm did not start (froze on blue windows logo, without reaching the login), after machine version upgraded from i440fx 2.5 to 2.7. Also, it did not start after passing through a usb pcie card... Overall, i feel it's hit and miss result... Will try to collect more information and post in the apropriate thread

sorry for stupid questions, but... Do we need to update the virtio drivers once windows installed? How does one update them? Is there a need to update the agent as well? thanks -d

hi, i'm trying to install Solus 1.2.1, iso from 2017.01.01 (Budgie). i chose linux template. live session is fine, installation is a breeze. but then i restart, post installation, i provide login details (user/pass), i login, then it appears it's loading, but screen resets (literally, there is a black screen for a brief moment) and i'm again back to login form which asks for user & pass. tried several times, but to no avail... i'm passing through a 1060 GPU, as second GPU. Anybody experienced such issue? It could be solus/wayland related, or budgie, or...? Other linux distro's (ubuntu, mint, fedora) are fine... it's true, none of them i tried with budgie, this is the first one... thanks, -d PS: i was just able to press a dialog button (software updates) before resetting the screen and i could select the updates options. I updated some packages, after that it remains at the budgie desktop, but no menu or button available. ps2: i tried solus mate and while it passes the point of login & i can install the driver for nvidia, restart the VM with the native resolution... after that, there is no menu displayed, just the desktop icons. No option to launch the menu. The VM loads, but i'm stuck because i cannot use the desktop environment.

how do you assign 4 GPU's to single VM? i was under the impression that only 1 GPU per VM is possible...

RobJ thanks a lot for this suggest - i followed it and was able to quickly solve it no fuss... After copying and creating new configuration with the 500GB disk as data drive, and parity built on the 640GB drive, I tested the VM's and all was ok. I then added the 525GB ssd as cache and formatted to btrfs. Changed a bit the shares policy in regards to the "use cache disk" so that my appdata and domains will stay on the cache disk and not on the array... Ran mover... Will consider some backup for the VM's - i saw gridrunner has some tutorials on this topic Thank you once again! -d

hi ashman70 I think i just experienced the same issue last night. Will give some context and behavior, but i believe it's similar... 1) My windows 10 VM was created in unraid 6.2.4, with i440fx-2.5. Passthrough of 1060 gpu. All nice and dandy 2) Upgraded to unraid 6.3.1. Now the default machine version in windows 10 template is i440fx-2.7 (but my win10 VM still defined as 2.5). VM still working fine for 3-4 days already... 3) I changed in the VM definition the version of machine to 2.7. Started the VM. Windows loading screen gets stuck to the "blue windows" logo. Had to force shutdown from unraid VM page. Tried several times. Sometimes the VM Page in the unraid gui loads only the top of the VM list, except the "Windows 10" VM - since the list is being ordered alphabetically, "Windows 10" VM it's at the bottom of page. Somehow the unraid is stuck because of this VM being hanged, and Unraid does not respond properly. I had to hard-reset the Unraid. Soft reset (pressing short on the power button) did not shut down the Unraid 4) I reverted the change in the VM from machine version 2.7 to 2.5. Then i got the "automatic windows repair" or similar. i let it for 10 minutes, but it got stuck again! 5) Restarted the entire box. Started the VM (still defined as 2.5) - windows loads properly! Finally... Restarted once again, just to be sure it's consistent: i was able to properly restart the Windows VM and login. Phew! 6) Tried to passthrough my stubbed PCIe USB 3.0 card (new card, so first time when i tested it). Windows again get's stuck. Error message in the VM log is similar to below, relevant for the device id of the PCI USB card: BAR 2 mmap unsupported. Performance may be slow this i understand it's just cosmetic? But my VM does not start 7) Removed the passthrough of PCIe USB 3.0 card from VM definition. VM starts again, consistently. All these trials were done with 3:2 cores assigned to the VM (and excluding cpu0 and it's HT pair, which is left for unraid). virtio drivers are still on 118, although the lastest is 126-2. Did not attempt to change that... After this excercise, I feel that windows 10 vm tries to adapt to the HW changes (machine version, PCIe USB card) and fails. No idea why, or where should i check further - windows logs ?? Reverting the H/W changes solved the problem, but not instantly, since it required extra restarts. Is machine v2.7 not stable? or it's just not advisable to change the version after installation is stable? However, in your case, you did not perform such changes... So i would try the following: 1) start in safe mode the VM - is this an option? 2) assign cpu0 only. remove any overclocking (are you?). remove non required passthrough devices (sound, usb controllers, etc...) 3) is there an option to try to boot from the CD iso (instead of booting from HD) then attempt to load the drivers, and do a repair ? (just thinking if possible, never tried it under unraid) Few other thoughts: 1) this is an issue i saw more frequently after upgrading to 6.3.0: etc/msr_feature_control But even if this message appears, it does not affect the boot of the VM. at least i could not find an effect... 2) i440fx vs q35 - as i understood, i440fx is for windows based and q35 is for linux based OS'. It may work to put q35 for widnows, but it's not as per the recommendation. So i think that you'r eok with i440fx and should not use q35. -d ps: edited a bit...

hi guys, I currently have 1 drive in my array, a 525GB SSD, it's capacity is less than half utilized. I have additional 2 HDD's, 640GB and 500GB, currently not utilized. My intention is to end up having in my unraid: - Parity: 640GB HDD - Data Drive: 500GB HDD - Cache: 525GB SDD How to achieve this? I thought that i will add the parity 640GB, let it build. Then i'll remove the SSD from the array. But i cannot replace the 535GB ssd with the 500GB HDD in same slot, because it's smaller... If i put the 500GB HDD it in different slot, will the data be build based on the parity drive? i also have additional 500GB SSD that could be used temporarily + enough network storage... Or should i just move the vdisks/files from the 525GB SSD, start new configuration, then bring back the data? thanks -d

dlandon, guys... I'm a bit confused... multiple questions following: I have a 250GB sata ssd, passed through to a win 10 vm and used to host the system partition and to boot the Wind 10 from it - basically my only drive for this VM. in UD it appears as not mounted (gray button). 1) should it be? I understand that xfs and etx4 partitions are mounted by default with discard (trim) But this is ntfs. 2) it means that trim for ntfs is not enabled? Because I get in windows this output when running the check for trim - command prompt(admin): C:\Windows\system32>fsutil behavior query DisableDeleteNotify NTFS DisableDeleteNotify = 0 ReFS DisableDeleteNotify is not currently set first line for ntfs - sounds like trim enabled 3) but second line is a bit confusing - what does it mean? thanks -d

still having this problem? i wonder why you had to stub the gpu and add it to the vfio-pci.ids in your sysconfig...

at least i was lucky and made it work... i had the same problem with my 550ti (e.g. not able to retart the VM using 550ti as passthrough and had to restart entire box). Then i used the rom for the 550ti, even if 550ti placed in second slot, and was able to restart the respective VM's using the 550ti, without having to restart the entire unraid box. Without rom, i had to restart entire box... So i suspect that if amstel is able to take the dump from the 610, or to find a rom that will work, then this problem might be solved. See this thread with the limitations faced by me while working with 550ti http://lime-technology.com/forum/index.php?topic=52960.msg535568#msg535568 -d

So activate the igpu so that you can passthrough the 730... Not sure if the passthrough will work directly, without rom. If not working, dump the rom and use it. You might be able to disable the igpu after that...

hi is the 730 your only gpu in the server? if yes, then you need an extra GPU in slot 1, while you put your 730 in another slot. then use the steps provided in this guide to extract the rom of 730. https://lime-technology.com/forum/index.php?topic=52960.0 the guide also tells how to use the rom in the xml of our vm's... -d

do you have already a modern card? if yes, buy from same from team... in this way, if you need to switch them among the vm's, you won't need to worry about drivers. before switching to unraid, i happen to switch my 2 cards from time to time, and windows would accept the "new" card without complaining. if you have gpu's from both the teams (e.g. amd + nvidia) then you have more work with driver's configuration when switching the assigned cards to your vm's 1060 vs 480 - really not in position to provide an unbiased opinion - since i did't try them both...

amstel, not sure what is best advice. as i understood, amd gpu's does not need this trick with the rom, will pass through directly while being in 1st slot... also, so far only amd has a single-slot modern graphics card (his and xfx) if this matter to you. just an example below: http://xfxforce.com/en-us/products/amd-radeon-rx-400-series/rx-460-2gb-sf-rx-460p2tfg5 While browsing the forum, did you find somebody with reporting windows 10 with an old gpu like 610? (i don't remember) my 550ti also has some issues, so i have limited success with it. win 10 not working at all with 550ti, and because of this issue, i am currently using win 10 with 1060 (first slot). That's the reason why i'm getting a small 1050ti, to replace my 550ti and get rid of this annoyance. -d

thanks ekim, will try that. i think i saw something like this in my bios as well...

ekim, i'm very interested to know what you did in bios that separated the usb controllers... thanks

Your Vm has cpu0 only assigned? If not, give it a try...

Copy the rom line from the xml and any other manual edits (none?) Edit the VM and to leave only cpu0 assigned Add back the rom line and any other manual edits that you require... Give it a try... My GTX 550ti has similar issue (but unraid does not crash) - i cannot pass it through to a windows VM. linux mint same, also elementary i believe have same / similar issue. Fedora, ubuntu gnome were ok... gtx 1060 works without an issue to passthrough ...

can you share your vm xml? i think i saw others raising issue when switiching their VM's from vnc to gpu passthrough. try fresh install at elast you see if it works...

a bit confused... my Windows 10 VM configuration was having the hyperv enabled. and i could still passthroigh the nvidia gpu, load the drivers and benchmark without issues. this was on unraid 6.2.4, before i upgraded to 6.3.0. But i keep reading that hyperv + nvidia passthrough != LOVE... Or is this an issue only when trying to use the hyper v feature? (to be frank, i didn't use hyperv...) thanks -d

ok, now i got you... but not sure i can help further regarding the rom, it's required if you want to pass through the primary nvidia gpu. it could be the only GPU in the system, or maybe you have additional GPU's... but without the rom, you cannot make it passthrou. my understanding is that AMD gpu's work to passthrough in first slot without rom. but never tried.

hi Endy i'm trying to follow through your email (several times, since i saw it's without response), but either i'm tired or i cannot understand properly what are you trying to do... To summarize my understanding, you are trying to avoid fresh installation of the Windows 10 in a VM, so that you can use your previous physical installation by converting it. And you reached a dead end since your converted [vdisk] does not work with seabios, only with ovmf. But still, only with VNC, and no GPU passthrough. But then you say it works with seabios as fresh install in VM - so passthrough of GPU works without any issue? When you say "convert to a VM" - you mean... the disk that was used by the physical installation you are trying to convert it to a vdisk? How are you doing that? Are you preparing the physical windows installation in some way, for this conversion to happen? I'm asking because as per my understanding, there is a special process to follow when converting it. Not sure if it's still relevant, i assume it is: http://lime-technology.com/wiki/index.php/UnRAID_6/VM_Management#Physical_to_Virtual_Machine_Conversion some motherboards allow selection of primary GPU from BIOS. Some added this feature in newer bios firmware updates. YMMV not sure if it helps, but your GTX 660 might require to specify the rom in the VM xml (that was the case with my 550ti). For this, you need to have this rom as a file in the array so that unraid can read it and use it during boot of VM (so you need to dump it yourself, using the guides in this forum, or to obrain it from some place - techpowerup - again, ymmv in terms of compatibility). Also try to put the 660 in second slot. Mine didn't work at all to do passthrough in first slot (closest to the GPU) even after rom was used. -d

i'm also interested for a card that plays nice with linux and windows 10. i would say also OS X - but at this moment, this os is more like a curiosity, rather than necessity...

1812 i empirically test by moving the mouse across the screen, in various applications and games, when CPU load is high I also noticed in a linux VM (connected to it via VNC), that the mouse lagged terribly, when going over some web page links - something that was affecting also the VM that initiated the VNC session. But this was before passing through the entire usb controller with the unifying receiver attached to it and implementing the cpu pinning guides. After that, such mouse lag reduced a lot. -d