Closed rstocktox closed 5 years ago
I've noticed this as well, although I've only seen artifacts on the screen, never the whole screen freezing. This is due using altp2m in Xen, but I haven't been able to track down the underlying reason. I guess swapping around the memory underneath the VM affects the framebuffer somehow. I would recommend using RDP if you really need to access the VM while DRAKVUF is running on it.
Thank you so much, Tamas
I have only noticed this a couple of times. Using RDP into the VM, I have not seen it since.
I am experiencing the same issue on my desktop, as soon as i start Drakvuf the VNC view freeze although the guest is still running fine (checked with RDP) and Drakvuf outputing events. I can 100% repro on 2 hosts OS (same computer, dual boot) one is Kubuntu 18.04 and the other is Mint 19. Have the problem with Xen 4.9 but also with 4.11 using the latest libvmi or the one from Drakvuf (submodule).
On another computer (laptop) i never had this issue, this one is running Ubuntu 16.04 but i am not convinced this is related to it since several people with different hosts have experienced the problem.
The other main difference is that the laptop is booting in legacy mode while the desktop showing the problem boots in UEFI. I will try to replicate the problem on the laptop using the same OS / software configuration than on the desktop.
Is not related to UEFI in any way, noticed the same behaviour booting in legacy mode
doesn't look tied to VNC as i switched the VM config , disabled VNC and enabled SDL => same issue :( SDL viewer is frozen. Running in VNC mode, I also noticed that if you disconnect / reconnect vncviewer then the screen is refresh and you can see that the mouse click & keyboard entries emitted during the frozen state were indeed registered ! so the mouse / keyboard event are transmitted to the guest OS , only the display is not refresh until you reconnect again.
This is only related to Xen and altp2m specifically. We don't yet understand exactly why the behavior is not consistent though. Relevant discussion from xen-devel: https://www.mail-archive.com/xen-devel@lists.xenproject.org/msg11760.html
@tklengyel Hello, the issue with screen not being refreshed still exists for me. Is there any patch for Xen that could be applied right now? The abovementioned branch is not there anymore.
Just look at the comment before yours.
Ah, okay, the patches are available after clicking on each of the "follow ups". I was not familiar with how to navigate the mailing list archive.
Direct links: https://lists.xenproject.org/archives/html/xen-devel/2018-10/msg01427.html https://lists.xenproject.org/archives/html/xen-devel/2018-10/msg01426.html
You can also just grab the latest Xen 4.12 release candidate, these patches are already part of it.
Dear Tamas, did you already tested drakvuf (and its current libvmi version) with Xen 4.12? I did a fast test some days ago and had some error condition at the beginning. I will further investigate on that, and want to know if you successfully tested it without changing the libvmi version.
Thank you, Alessandro
On Mon, Feb 25, 2019 at 3:54 AM Tamas K Lengyel notifications@github.com wrote:
You can also just grab the latest Xen 4.12 release candidate, these patches are already part of it.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tklengyel/drakvuf/issues/238#issuecomment-466854427, or mute the thread https://github.com/notifications/unsubscribe-auth/AAYiLwouDUntsbm5eK91EF7FfMBId_1Rks5vQ1BTgaJpZM4LkfF7 .
Yes, it works just fine AFAICT
I've updated Xen to tags/4.12.0-rc3
version from git://xenbits.xen.org/xen.git
. The whole machine (including host system) hangs just few seconds after I run DRAKVUF and VNC.
You should try to attach a serial cable and enable all debug loglevels for Xen. When you get a crashlog on the serial post it to the xen-devel
mailinglist.
I also just tested 4.12.0-rc3
on my system and it works just fine.
Update: I've set up Debian Stretch host with 4.12.0-rc4
on another server and it works fine. The previous one still doesn't work, for now let's assume that it's some misconfiguration on my side.
Hi,
Testing a fresh install of the current DRAKVUF version on an HP Z40 desktop server (16GB RAM, 8 cores, Debian Jessie) found that whatever plugin I ran the VNC screen is not refreshed until DRAKVUF stops running.
I'm currently using TightVNC as my VNC client from my Windows laptop and it really works like a charm connecting to the same DomU in the same desktop server (without using DRAKVUF) or even connecting to another server with XEN v4.7 and old DRAKVUF versions.
Seems strange to me and, sincerely, I don't see clearly where the problem lies, as said before DRAKVUF + Xen 4.7 didn't exhibit this behaviour on another server.
I tried and changed lot of things: enabling/disabling logs on the XEN kernel command line, increasing/decreasing DomU memory, using 1 or 2 vcpus for the DomU, increasing shadow memory DomU config option, enabling/disabling DRAKVUF debug, compiling and enabling just 1 DRAKVUF plugin... but unfortunately nothing changed
I think it's worth to mention that although the screen not been refreshed the DomU seems to work/run as usual (tested enabling the file delete plugin and deleting some files in a "blindly" way).
Would you be so kind to shed some light on this issue?
Generic system info:
xl info output before starting the DomU:
DomU configuration file (Windows 7 SP1 x64 guest):
xl create output:
Output of xl info after DomU started:
As I said before DRAKVUF was compiled with debug enabled and just the exmon plugin (tested with the filedelete and the syscalls plugins also and achieving the same result):
DRAKVUF debugging output:
That desktop server is built for the only purpose of testing/developing so feel free to ask for any kind of information you need or source code modifying.
Best regards
PD: Modified sentences