Bug 293351 - graphics/gpu-firmware-radeon-kmod: radeonkms is not working
Summary: graphics/gpu-firmware-radeon-kmod: radeonkms is not working
Status: Open
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: 15.0-RELEASE
Hardware: amd64 Any
: --- Affects Only Me
Assignee: freebsd-x11 (Nobody)
URL:
Keywords: regression
Depends on:
Blocks:
 
Reported: 2026-02-21 23:45 UTC by Artem Kim
Modified: 2026-08-05 12:22 UTC (History)
4 users (show)

See Also:
linimon: maintainer-feedback? (x11)


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Artem Kim 2026-02-21 23:45:35 UTC
hello,
I ran into a problem after the next source update and kernel rebuild, loading radeonkms leads to a dead lock of the system (only the reset helps).  the buffer is being initialized, its old contents are visible, but the system is blocked, I tried drm-kmod-6.6, rebuilt the system completely, and then kmod-6.6 the problem remains. Bug 276729 may have a relation.

> uname -aKU
FreeBSD uh1 15.0-STABLE FreeBSD 15.0-STABLE stable/15-93043c650548 MY amd64 1500506 1500506

> pciconf -lv
. . .

vgapci0@pci0:5:0:0:     class=0x030000 rev=0x00 hdr=0x00 vendor=0x1002 device=0x6718 subvendor=0x174b subdevice=0xe182
    vendor     = 'Advanced Micro Devices, Inc. [AMD/ATI]'
    device     = 'Cayman XT [Radeon HD 6970]'
    class      = display
    subclass   = VGA
none2@pci0:5:0:1:       class=0x040300 rev=0x00 hdr=0x00 vendor=0x1002 device=0xaa80 subvendor=0x174b subdevice=0xaa80
    vendor     = 'Advanced Micro Devices, Inc. [AMD/ATI]'
    device     = 'Cayman/Antilles HDMI Audio [Radeon HD 6930/6950/6970/6990]'
    class      = multimedia
    subclass   = HDA
. . .

> pkg info | grep drm
drm-latest-kmod-6.9.1500506_1  DRM drivers modules
drm_info-2.9.0                 Dump info about DRM devices
libdrm-2.4.131,1               Direct Rendering Manager library and headers

> # pkg info | grep firm
gpu-firmware-radeon-kmod-cayman-20220511.1500506 Firmware modules for cayman Radeon GPUs
Comment 1 Artem Kim 2026-02-22 00:50:32 UTC
I forgot to write it.  I've tried

# syscons is the legacy console driver, resembling an SCO console
device          vga                     # VGA video card driver
device          splash                  # Splash screen and screen saver support
device          sc
options         SC_PIXEL_MODE           # add support for the raster text mode

# vt is the default video console driver
device          vt
device          vt_vga
device          vt_efifb
device          vt_vbefb

and

# syscons is the legacy console driver, resembling an SCO console
#device          vga                     # VGA video card driver
#device          splash                  # Splash screen and screen saver support
#device          sc
#options         SC_PIXEL_MODE           # add support for the raster text mode

# vt is the default video console driver
device          vt
device          vt_vga
device          vt_efifb
device          vt_vbefb

and

cat /boot/loader.conf
hw.syscons.disable=1
Comment 2 Mike Rothfuss 2026-07-16 22:03:51 UTC
radeonkms works with drm-515-kmod, but not drm-kmod, in FreeBSD 15.1. I'm using it on two corebooted AMD APUs: Richland (Thinkpad E545) and Kabini (Thinkpad X140e). For the Kabini APU, loading radeonkms randomly causes the system to crash and reset around 10% of the time. I've added the "hw.vga.textmode=1" parameter to /boot/loader.conf and have not seen a reset since.
Comment 3 Artem Kim 2026-07-20 22:50:46 UTC
Thank you for your attention to my problem. I've tried different versions of kmod. but not one does not work. Moreover, the DVI outputs behave strangely: only one works - the other does not even initialize the monitor in the BIOS.  the problems started after upgrading to 15.1. The output just stopped working and had to use the other one. perhaps these are hardware problems and it's time for a great upgrade - moving to UEFI and GPT.
Comment 4 Alexey Dokuchaev freebsd_committer freebsd_triage 2026-07-21 05:56:08 UTC
(In reply to Mike Rothfuss from comment #2)
> radeonkms works with drm-515-kmod, but not drm-kmod, in FreeBSD 15.1
There were reports of various regressions with drm-kmod ports in recent years.  I haven't braced myself to upgrade from FreeBSD 13.5 and drm-510-kmod-5.10.163 where my trusty old Richland works fine.

$ glxinfo -B | grep -A3 vendor
OpenGL vendor string: X.Org
OpenGL renderer string: AMD ARUBA (DRM 2.50.0 / 13.5-RELEASE, LLVM 11.0.1)
OpenGL core profile version string: 4.5 (Core Profile) Mesa 22.3.7
OpenGL core profile shading language version string: 4.50
Comment 5 Artem Kim 2026-08-04 13:53:50 UTC
(In reply to Alexey Dokuchaev from comment #4)

it looks like the problem with EDID
> edid-decode -n
is not responding.
Comment 6 Alexey Dokuchaev freebsd_committer freebsd_triage 2026-08-04 17:17:09 UTC
(In reply to Artem Kim from comment #5)
> it looks like the problem with EDID: `edid-decode -n' is not responding.
If you mean the program from the `sysutils/edid-decode' port, it needs an input to decode.  Try searching for "EDID (in hex)" string in /var/log/Xorg.*log* files or `sysutils/read-edid' port.
Comment 7 Artem Kim 2026-08-04 20:50:13 UTC
(In reply to Alexey Dokuchaev from comment #6)

I'm sorry, I thought it reads the response from the device automatically. I use the vesa driver for temporary bypass. x-org cannot use the native monitor resolution and uses 720x400.
> xrandr
xrandr: Failed to get size of gamma for output default
Screen 0: minimum 720 x 400, current 720 x 400, maximum 720 x 400
default connected primary 720x400+0+0 0mm x 0mm
   720x400        0.00*

the mode definition in xorg.conf is ignored.

        Modeline        "1920x1200"     154.00  1920 1968 2000 2080  1200 1203 1209 12
35 +hsync -vsync



> cat data | edid-decode -n
. . .
      Monitor ranges (GTF): 50-100 Hz V, 30-130 kHz H, max dotclock 300 MHz
    Display Product Name: 'HP 524pn'
    Display Product Serial Number: 'CN45100H4K'
  Extension blocks: 1
Checksum: 0x33

----------------

Native Video Resolution:
  1920x1200

> cat /var/log/Xorg.0.log

...
[ 23402.809] (II) VESA(0): Monitor name: HP 524pn
[ 23402.809] (II) VESA(0): Serial No: CN45100H4K
[ 23402.809] (II) VESA(0): Number of EDID sections to follow: 1
[ 23402.809] (II) VESA(0): EDID (in hex):
[ 23402.809] (II) VESA(0):      00ffffffffffff00220e513a00000000
[ 23402.809] (II) VESA(0):      0a230103803420782a8e35ad514ba125
[ 23402.809] (II) VESA(0):      0d5054a54b00d1c0a9c08100b3009500
[ 23402.809] (II) VESA(0):      d100a9408180283c80a070b023403020
[ 23402.809] (II) VESA(0):      360006442100001a000000fd0032641e
[ 23402.809] (II) VESA(0):      821e000a202020202020000000fc0048
[ 23402.809] (II) VESA(0):      5020353234706e0a20202020000000ff
[ 23402.809] (II) VESA(0):      00434e343531303048344b0a20200133
[ 23402.809] (II) VESA(0): EDID vendor "HPN", prod id 14929
[ 23402.809] (II) VESA(0): Using EDID range info for horizontal sync
[ 23402.809] (II) VESA(0): Using EDID range info for vertical refresh
[ 23402.809] (II) VESA(0): Printing DDC gathered Modelines:
[ 23402.809] (II) VESA(0): Modeline "1920x1200"x0.0  154.00  1920 1968 2000 2080  1200 1203 1209 1235 +hsync -vsync (74.0 kHz eP)
...
Comment 8 Gleb Popov freebsd_committer freebsd_triage 2026-08-05 09:21:46 UTC
(In reply to Artem Kim from comment #0)
> gpu-firmware-radeon-kmod-cayman-20220511.1500506 Firmware modules for cayman Radeon GPUs

This looks old. The recent version should be 20260519.
Comment 9 Artem Kim 2026-08-05 12:22:08 UTC
(In reply to Gleb Popov from comment #8)

pkg install gpu-firmware-radeon-kmod-cayman
. . .
The most recent versions of packages are already installed

from ports:

> pkg info | grep GPU
drm-latest-kmod-6.12.1501501   Direct Rendering Manager (DRM) GPU drivers
gpu-firmware-radeon-kmod-cayman-20260519.1501501 Firmware modules for cayman Radeon GPUs
libplacebo-7.360.1             Reusable library for GPU-accelerated video/image rendering

The behavior has been preserved: the deadlock after loading radeonkms.

It looks like it's hanging while determining the output/monitor. Is there a problem in the DDC/EDID area?