Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Apr 05 13:12
    wulf7 opened #21
  • Apr 02 08:25

    evadot on master

    Make it possible to select whic… (compare)

  • Mar 30 19:41

    evadot on master

    i915kms: Remove uuencoded files… amdgpufw: Remove uuencoded file… Remove uuencoded files No need… and 3 more (compare)

  • Mar 30 18:17
    juikim commented #19
  • Mar 30 18:15
    evadot closed #19
  • Mar 30 18:15

    evadot on master

    amdgpu: Update amdgpu firmwares… amdgpu: Hook new firmwares to b… Merge pull request #19 from jui… (compare)

  • Mar 30 18:08
    juikim commented #19
  • Mar 29 17:38
    juikim commented #19
  • Mar 29 17:37
    juikim synchronize #19
  • Mar 29 17:25
    evadot commented #19
  • Mar 29 17:15
    juikim synchronize #19
  • Mar 29 17:07
    juikim edited #19
  • Mar 29 16:52
    juikim opened #19
  • Mar 29 10:49

    hselasky on drm-v4.16-fbsd12.0

    Buildfix for 12-stable after 1a… (compare)

  • Mar 18 15:41

    zeising on master

    Update README.md Update README… (compare)

  • Mar 18 15:29
    iotamudelta removed as member
  • Mar 18 15:25
    iotamudelta added as member
  • Mar 18 15:25
    iotamudelta removed as member
  • Mar 17 15:21
    jbeich commented #7
  • Mar 10 08:26
    cjayho opened #18
fbsdgitter
@fbsdgitter
[irc] <dch> in case it's not clear, I'm rebooting into a new 14.x BE each time to test the new packages.
[irc] <dch> if not I'll need to roll it all into a bhyve or rent a vm for a spot
fbsdgitter
@fbsdgitter
[irc] <dch> Erandir: I'm not intending to try to run the 14.x drivers on 13.x, but build them. I can't run 14.x for longer than 15-30m without a hard hang, atm.
[irc] <Erandir> Running a FreeBSD N+1 userland on FreeBSD N might work, but there are no guarantees.
[irc] <Erandir> Also, what happens if you simply don't load the drivers when building them in poudriere, on the system with 14_
[irc] <dch> Erandir: then I can't work (no slack/webrtc etc)
[irc] <dch> thats the complication otherwise this would be fine.
[irc] <Erandir> But the drivers build in <30 min, can't you do that on your lunch break or after the work day or the like?
fbsdgitter
@fbsdgitter
[irc] <dch> I need all of x11/nvidia-xconfig x11/nvidia-settings graphics/vulkan-tools graphics/vulkan-loader graphics/vulkan-headers x11/nvidia-driver graphics/mesa-dri graphics/mesa-libs which is a fair bit longer
[irc] <Erandir> Anyway, running a FreeBSD 14 userland on a 13 kernel is not guaranteed to work.
[irc] <Erandir> if you need more time, leave it running over night?
[irc] <Erandir> That's what I do when I do big build runs
fbsdgitter
@fbsdgitter
[irc] <dch> I think I will switch my NAS to 14.x for a while, its a very small system but with ccache it should do fine with just rebuilding a package or two
Jan Beich
@jbeich
dch: git log origin/releng/13.0..origin/main sys/kern/syscalls.master doesn't show new syscalls, so poudriere for 14.0 may just work. ioctl/sysctl with unstable kernel structs are unlikely to be used by build tools.
fbsdgitter
@fbsdgitter
[irc] <dch> jbeich thanks thats very helpful idea to check the syscalls.master diff.
fbsdgitter
@fbsdgitter
[irc] <Erandir> Yes, but if it doesn't work, or if it produces strange results, you are very much on your own.
[irc] <Erandir> Hence my "not guaranteed to work".
Tommi Pernila
@pernila_twitter
Just a heads up if anyone is interested in Vulkan 1.2 with Nvidia drivers. The latest beta 465.19.01 from 30.03.2021 adds support for FreeBSD. Release notes: https://www.nvidia.com/download/driverResults.aspx/171982/en-us
Jan Beich
@jbeich
Does vulkaninfo on NVIDIA contain VK_EXT_external_memory_host? Example consumers are emulators/rpcs3 ( RPCS3/rpcs3#9613 ) and multimedia/ffmpeg ( ffmpeg/ffmpeg@4dcb50c )
On Intel it's notoriously broken by FreeBSDDesktop/kms-drm#197
Alex
@shkhln

Does vulkaninfo on NVIDIA contain VK_EXT_external_memory_host?

https://gist.github.com/shkhln/19c4d1c7591d8a15ed27c04553455e28#file-2-diff-to-shim-diff-L169

cc @amshafer regarding VK_EXT_external_memory_host.
Austin Shafer
@amshafer
That's disabled
which is unfortunate. I was hoping it wasn't used
unrelentingtech
@unrelentingtech
speaking of "notoriously broken", mmu notifiers have been around in linux for ages https://lwn.net/Articles/266320/ but we don't have any such facility of our own at all, we'll have to invent it first to be able to write a linux compat layer for it
Warner Losh
@bsdimp
btw, looking for feedback on https://reviews.freebsd.org/D29479 which has some improved wording for the graphics setup.
nomadlogic
@nomadlogic
hey there, is there a good way to force a system to enter a debugger if it runs into a panic? i remember there was a sysctl knob to force this back when the linux drm code was being worked on.
reason i ask is after updating to current today i get a panic/reboot pretty regularly with both drm-devel and drm-current. i have savecore enabled but no core is generated
hrm according to the Graphics wiki it seems like this is what i'm looking for: debug.debugger_on_panic=0
ekhramtsov
@ekhramtsov:matrix.org
[m]
Heads up: recent current with d36d68161517 will make x11-wm/sway session hang on start. Applying the fix from D29623 unbreaks sway. I’ve posted this to the freebsd-current mailing list, although I still don’t see it in the archive. I'll try to post it later again if I won’t see the message there in a day.
Jan Beich
@jbeich
seatd or direct session? sway with seatd seems to work fine on main-n245853-b55a927bc884
D29623 talks about TLS but mesa-libs doesn't use TLS in libGL compared to mesa-devel. But I'm using mesa-devel...
ekhramtsov
@ekhramtsov:matrix.org
[m]
seatd vt gl=mesa-devel
will pay more attention to this later once inane life stuff will leave me more time
Jan Beich
@jbeich
@ekhramtsov:matrix.org, sorry for the false negative report. I had old world but new kernel. After upgrading -CURRENT again d36d68161517 indeed makes every OpenGL and Vulkan app hang with [uwrlck] in waiting channel. Affects both mesa-{libs,dri} and mesa-devel.
Jan Beich
@jbeich
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1466 has finally landed, so mpv --hwdec=vaapi --gpu-api=vulkan should work again on Wayland with mesa-devel soon. See https://code.videolan.org/videolan/libplacebo/-/issues/117 for details.
ekhramtsov
@ekhramtsov
Apparently, fixing xkbcli interactive-wayland with XDG_RUNTIME_DIR on ZFS is as simple as changing a define: https://codeberg.org/ei/eiports-wip/src/branch/main/pfalloc.patch. Ouch. Does DragonFly need some define like this? Is this OK? (I am new to C).
Niclas Zeising
@zeising
But we want something else instead of posix_fallocate, won't we? Also, it is much better to handle the error, since posix_fallocate works on UFS.
Jan Beich
@jbeich
ekhramtsov
@ekhramtsov
This is a simple test, I doubt it requires/needs posix_fallocate. It will use ftruncate with this patch.
I don't think a check like the one that was used in Firefox before is needed, it is only a test.
I can change it if you still want this.
Jan Beich
@jbeich
posix_fallocate can fail on Linux/musl with ZFS
ekhramtsov
@ekhramtsov
Indeed, I'll check for some E-something, thanks
Jan Beich
@jbeich
Actually, https://gitlab.freedesktop.org/wayland/wayland/-/merge_requests/54 didn't explicitly name any filesystem
ekhramtsov
@ekhramtsov
Can't load "changes" for some reason, might be an ajax request with an ipv4 literal address (I am behind DNS64/NAT64).
In any way, I've thoroughly checked/tested new https://codeberg.org/ei/eiports-wip/src/branch/main/pfalloc.patch.
I don't see my mistakes before some time passes, so I'll take an hour or so to forget about this and reread this
Thanks for your help!