Re: FreeBSD/X11: Howto revive DisplayPort?
- Reply: Marek Zarychta : "Re: FreeBSD/X11: Howto revive DisplayPort?"
- In reply to: Cy Schubert : "Re: FreeBSD/X11: Howto revive DisplayPort?"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Fri, 28 Nov 2025 06:08:56 UTC
On Thu, 27 Nov 2025 07:47:47 -0800 Cy Schubert <Cy.Schubert@cschubert.com> wrote: > One of HP 840 G5's docking station's DPs is connected to a VGA KVM > switch via a DP to VGA connector. Confirmed, this is an issue. > > The VGA KVM switch is also connected to my employer's HP 840 G9, also > via a DP to VGA connector. No problems there. Though that machine uses > the newer Thunderbolt docking station. Given that it's a newer machine > with a newer docking station I think the comparison only goes so far. In my specific case, the FreeBSD hosts are using DP, the KVM switch is a kind of hybrid (2x DP, 1x HDMI fpr usage of 3 monitors simultaneously with the computer) and the monitor itself is a 4k Lenovo, nothing fancy and/or expensive, just for doing conveniently my work at home. In my office, the equipment is similar, except the KVM switch, which is a DP only, more expensive type. In all cases, the host has a discrete nVidia GPU. > > Whether this is a FreeBSD or a UEFI BIOS problem isn't quite understood > by me. When booting the FreeBSD machine and if the KVM is not switched > to the FreeBSD machine, the text mode console will have the wrong > geometry. Even in single user the login prompt is far below the bottom > of the monitor. > > But once XDM has started it assumes a graphic geometry that fits the > laptop's screen, not the external monitor. I have a script that > disables the LVDS when the VGA is connected. Similar here, but worse when it comes to console only, as initially described. If the host booting and turning on the output for conosle is not switched to by the KVM - with monitor in ON state - the monitor stays dark, no chance to get it work without reboot of the FBSD box. Different story with X11/xdm running. To get an initial screen without further interaction, the KVM has to be switched to the host about to start. By the way, same issue with a LibreELEC/KODI driven host utilizing an Intel iGPU. But other than the issue with FBSD's console, with X11/xdm running and switching the KVM afterwards towards this DP output the FBSD box / X11 is running on, some could ssh onto the box and issue some xrandr command to refresh the output. I did this recently, sorry, I have the notes not at hand. But it is a kind of a labour and sometimes impossible to connect via ssh/reset DP output. As mentioned, Linux seems not to have those problems, at least with a RedHat based product I have to use for work - but that might be a commercial addition to the X11 server, I do not know, but I realize it works (at least for two of the scenarios I described). > > I suspect that this problem will require more tinkering (not by myself) > to better understand it. There are too many variables to say it's a > single issue. > > Another issue is, X segfaults on occasion on this machine. The segfault > always occurs when firefox is active. The screen goes blank as > drm-66-kmod (I think) is corrupted. (This happens maybe once or twice > a month, not enough to pull me away from other projects.) Are you > talking about this or the first issue above? The boxes at home are AM5 based newer systems utilizing recent Ryzen 9XXX CPUs with internal GPUs, one is using a recent nVidia GTX5XXX discrete GPU. I once used the drm-66-kmod framework to apply a high resoltution output from my server to the monitor, but using the console does not require microscopic small output, but I can state that the issue is quite the same as described before. I have not found any solution to revive a stuck DP-KVM-DP(Monitor) connection as described above with X11/xdm using the DRM-66 kmod output for the console (nor with the official nvidia kmod - speaking of nvidia, I always refer to the BLOB provided by most recent x11/nvidia-driver, x11/nvidia-kmod). I hoped for the lack of knowledge on my side when it comes to such issues, especially when the efi framebuffer for simple consoles is involved, but I realizie the problem might be bi-lateral one (both sides, hosts's DP and KVM/monitor's DP). Thank you very much for sharing. oh