Public bug reported:
Package: linux-firmware-amd-graphics 20240318.git3b128b60-0ubuntu3.4
(noble-updates)
System: Linux Mint (Ubuntu 24.04 "noble" base), kernel 7.0.0-42-generic, Mesa
25.2.8-0ubuntu0.24.04.4
GPU: AMD Radeon RX 9070 (Navi 48, PCI 1002:7550 rev c3, gfx1201)
Board: MSI MPG Z890 EDGE (MS-7E19)
SUMMARY
In linux-firmware-amd-graphics 0ubuntu3.4, dcn_4_0_1_dmcub.bin.zst is a symlink
to dcn_3_2_0_dmcub.bin.zst, i.e. the RX 9070 (DCN 4.0.1) gets the DCN 3.2.0
DMCUB firmware. Upstream linux-firmware has these as two separate files. My
system started to hang after an update that took this package from 0ubuntu3.3
to 0ubuntu3.4 (per apt history); I cannot rule out other updates in the same
period.
EVIDENCE
1. Contents of the .deb itself:
$ dpkg -c
linux-firmware-amd-graphics_20240318.git3b128b60-0ubuntu3.4_all.deb | grep
dcn_4_0_1
lrwxrwxrwx root/root 0 2026-10-19 14:08
./lib/firmware/amdgpu/dcn_4_0_1_dmcub.bin.zst -> dcn_3_2_0_dmcub.bin.zst
2. Upstream (kernel.org linux-firmware, main, commit ba7155b16ca7, Nov 5 2026)
has two different files:
dcn_4_0_1_dmcub.bin sha256
74b567655ac89bfe03389b660e973dc3c067aebd464c6389a071662cfbe6237f
dcn_3_2_0_dmcub.bin sha256
82713e08816638b8135a3e1b8ce670762a7174c623e10cdcd18ea036604e85bb
3. The 0ubuntu3.4 changelog says "amdgpu: update dmcub fw for dcn32 and dcn401"
(LP: #2168847 is the bug listed there).
4. 0ubuntu3.4 is the only version available in noble-updates now (apt policy):
Installed: 20240318.git3b128b60-0ubuntu3.4
Candidate: 20240318.git3b128b60-0ubuntu3.4
Version table: 3.4 only, from archive.ubuntu.com noble-updates/main
SYMPTOMS (before replacing firmware)
- The system freezes. It also happens with an older kernel, so it does not look
kernel-specific.
- Errors in the kernel log:
amdgpu 0000:03:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB
error - collecting diagnostic data (repeated)
amdgpu 0000:03:00.0: SMU: No response msg_reg: 2b resp_reg: 0
amdgpu 0000:03:00.0: Failed to power gate VCN instance 0!
amdgpu 0000:03:00.0: [drm] *ERROR* DPM disable vcn failed, ret = -62.
amdgpu 0000:03:00.0: Failed to enable GFXCLK DS!
amdgpu 0000:03:00.0: Failed to set workload mask 0x00000003
- On reboot: WARNING at drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c:670
(amdgpu_irq_put) in gfx_v12_0_hw_fini, plus REG_WAIT timeouts in
dcn32_hubp_pg_control, dcn32_dsc_pg_control and optc401_disable_crtc.
- Once the GPU became very hot within seconds and the fans spun up for the
first time (they are normally stopped, Zero RPM); I powered the PC off. After
it cooled down the card was detected and worked normally with the replaced
firmware.
- The same card works fine under Windows on the same machine.
WORKAROUND / WHAT I DID
I replaced all /lib/firmware/amdgpu/*.bin.zst with files from upstream
linux-firmware main (commit ba7155b16ca7), recompressed with zstd, ran
update-initramfs and put the package on hold. After that:
- No errors from amdgpu/SMU/DMCUB in the log
- SMU firmware 104.80.0, DMUB version 0x0A005000, "SMU is resumed successfully"
- Sensors work (idle ~31-33 C), glxgears on the dGPU works
CAVEAT
I replaced all amdgpu firmware files at once, so I cannot say that the wrong
dcn_4_0_1 file alone caused the hangs. I have not checked whether 0ubuntu3.3
had the same symlink. The mismatch itself is verifiable in the package.
EXPECTED
dcn_4_0_1_dmcub.bin.zst should be the DCN 4.0.1 firmware, not a link to the DCN
3.2.0 one.
** Affects: linux-firmware (Ubuntu)
Importance: Undecided
Status: New
** Description changed:
Package: linux-firmware-amd-graphics 20240318.git3b128b60-0ubuntu3.4
(noble-updates)
System: Linux Mint (Ubuntu 24.04 "noble" base), kernel 7.0.0-42-generic, Mesa
25.2.8-0ubuntu0.24.04.4
GPU: AMD Radeon RX 9070 (Navi 48, PCI 1002:7550 rev c3, gfx1201)
Board: MSI MPG Z890 EDGE (MS-7E19), BIOS dated 07/16/2026
SUMMARY
After upgrading linux-firmware-amd-graphics from 0ubuntu3.3 to 0ubuntu3.4,
the system started to hang. In 0ubuntu3.4,
/lib/firmware/amdgpu/dcn_4_0_1_dmcub.bin.zst is a symlink to
dcn_3_2_0_dmcub.bin.zst, i.e. the RX 9070 (DCN 4.0.1) gets the DCN 3.2.0 DMCUB
firmware. Upstream linux-firmware has these as two separate files.
EVIDENCE
1. Package contents (from the cached .deb):
- [paste output of: dpkg -c
/var/cache/apt/archives/linux-firmware-amd-graphics_20240318.git3b128b60-0ubuntu3.4_all.deb
| grep dcn_4_0_1]
- Installed system showed: dcn_4_0_1_dmcub.bin.zst ->
dcn_3_2_0_dmcub.bin.zst (dated Oct 19 14:08)
+ [paste output of: dpkg -c
/var/cache/apt/archives/linux-firmware-amd-graphics_20240318.git3b128b60-0ubuntu3.4_all.deb
| grep dcn_4_0_1]
+ Installed system showed: dcn_4_0_1_dmcub.bin.zst ->
dcn_3_2_0_dmcub.bin.zst (dated Oct 19 14:08)
2. Upstream (kernel.org linux-firmware, main, commit ba7155b16ca7, Nov 5
2026) has two different files:
- dcn_4_0_1_dmcub.bin sha256
74b567655ac89bfe03389b660e973dc3c067aebd464c6389a071662cfbe6237f
- dcn_3_2_0_dmcub.bin sha256
82713e08816638b8135a3e1b8ce670762a7174c623e10cdcd18ea036604e85bb
+ dcn_4_0_1_dmcub.bin sha256
74b567655ac89bfe03389b660e973dc3c067aebd464c6389a071662cfbe6237f
+ dcn_3_2_0_dmcub.bin sha256
82713e08816638b8135a3e1b8ce670762a7174c623e10cdcd18ea036604e85bb
3. The 0ubuntu3.4 changelog says "amdgpu: update dmcub fw for dcn32 and
dcn401" (LP: #2168847 is the bug listed there).
SYMPTOMS (before replacing firmware)
- The system freezes. The same happens with an older kernel (7.0.0-38), so it
follows the firmware package rather than the kernel.
- Errors in the kernel log during boot/session:
- amdgpu 0000:03:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB
error - collecting diagnostic data (repeated)
- amdgpu 0000:03:00.0: SMU: No response msg_reg: 2b resp_reg: 0
- amdgpu 0000:03:00.0: Failed to power gate VCN instance 0!
- amdgpu 0000:03:00.0: [drm] *ERROR* DPM disable vcn failed, ret = -62.
- amdgpu 0000:03:00.0: Failed to enable GFXCLK DS!
- amdgpu 0000:03:00.0: Failed to set workload mask 0x00000003
+ amdgpu 0000:03:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB
error - collecting diagnostic data (repeated)
+ amdgpu 0000:03:00.0: SMU: No response msg_reg: 2b resp_reg: 0
+ amdgpu 0000:03:00.0: Failed to power gate VCN instance 0!
+ amdgpu 0000:03:00.0: [drm] *ERROR* DPM disable vcn failed, ret = -62.
+ amdgpu 0000:03:00.0: Failed to enable GFXCLK DS!
+ amdgpu 0000:03:00.0: Failed to set workload mask 0x00000003
- On reboot: WARNING at drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c:670
(amdgpu_irq_put) in gfx_v12_0_hw_fini, plus REG_WAIT timeouts in
dcn32_hubp_pg_control, dcn32_dsc_pg_control and optc401_disable_crtc.
- Once the GPU became very hot within seconds and the fans spun up for the
first time; I powered the PC off. The card was still detected and worked at
idle afterwards.
- The same card works fine under Windows on the same machine.
WORKAROUND / WHAT I DID
I replaced all /lib/firmware/amdgpu/*.bin.zst with files from upstream
linux-firmware main (commit ba7155b16ca7), recompressed with zstd, ran
update-initramfs and put the package on hold. After that:
- No errors from amdgpu/SMU/DMCUB in the log
- SMU firmware 104.80.0, DMUB version 0x0A005000, "SMU is resumed
successfully"
- Sensors work (idle ~31-33 C), glxgears on the dGPU works
- [Update here: result of testing with the monitor connected to the RX 9070 and
under load]
CAVEAT
Since I replaced all amdgpu firmware files at once, I cannot say that the
wrong dcn_4_0_1 file alone caused the hangs. I also have not checked whether
0ubuntu3.3 had the same symlink. The mismatch itself is verifiable in the
package.
EXPECTED
dcn_4_0_1_dmcub.bin.zst should be the DCN 4.0.1 firmware, not a link to the
DCN 3.2.0 one.
Attachments: [apt policy linux-firmware-amd-graphics]; [relevant
journalctl -b -N -p 3 output]
** Description changed:
Package: linux-firmware-amd-graphics 20240318.git3b128b60-0ubuntu3.4
(noble-updates)
System: Linux Mint (Ubuntu 24.04 "noble" base), kernel 7.0.0-42-generic, Mesa
25.2.8-0ubuntu0.24.04.4
GPU: AMD Radeon RX 9070 (Navi 48, PCI 1002:7550 rev c3, gfx1201)
- Board: MSI MPG Z890 EDGE (MS-7E19), BIOS dated 07/16/2026
+ Board: MSI MPG Z890 EDGE (MS-7E19)
SUMMARY
- After upgrading linux-firmware-amd-graphics from 0ubuntu3.3 to 0ubuntu3.4,
the system started to hang. In 0ubuntu3.4,
/lib/firmware/amdgpu/dcn_4_0_1_dmcub.bin.zst is a symlink to
dcn_3_2_0_dmcub.bin.zst, i.e. the RX 9070 (DCN 4.0.1) gets the DCN 3.2.0 DMCUB
firmware. Upstream linux-firmware has these as two separate files.
+ In linux-firmware-amd-graphics 0ubuntu3.4, dcn_4_0_1_dmcub.bin.zst is a
symlink to dcn_3_2_0_dmcub.bin.zst, i.e. the RX 9070 (DCN 4.0.1) gets the DCN
3.2.0 DMCUB firmware. Upstream linux-firmware has these as two separate files.
My system started to hang after an update that took this package from
0ubuntu3.3 to 0ubuntu3.4 (per apt history); I cannot rule out other updates in
the same period.
EVIDENCE
- 1. Package contents (from the cached .deb):
- [paste output of: dpkg -c
/var/cache/apt/archives/linux-firmware-amd-graphics_20240318.git3b128b60-0ubuntu3.4_all.deb
| grep dcn_4_0_1]
- Installed system showed: dcn_4_0_1_dmcub.bin.zst ->
dcn_3_2_0_dmcub.bin.zst (dated Oct 19 14:08)
+ 1. Contents of the .deb itself:
+ $ dpkg -c
linux-firmware-amd-graphics_20240318.git3b128b60-0ubuntu3.4_all.deb | grep
dcn_4_0_1
+ lrwxrwxrwx root/root 0 2026-10-19 14:08
./lib/firmware/amdgpu/dcn_4_0_1_dmcub.bin.zst -> dcn_3_2_0_dmcub.bin.zst
2. Upstream (kernel.org linux-firmware, main, commit ba7155b16ca7, Nov 5
2026) has two different files:
- dcn_4_0_1_dmcub.bin sha256
74b567655ac89bfe03389b660e973dc3c067aebd464c6389a071662cfbe6237f
- dcn_3_2_0_dmcub.bin sha256
82713e08816638b8135a3e1b8ce670762a7174c623e10cdcd18ea036604e85bb
+ dcn_4_0_1_dmcub.bin sha256
74b567655ac89bfe03389b660e973dc3c067aebd464c6389a071662cfbe6237f
+ dcn_3_2_0_dmcub.bin sha256
82713e08816638b8135a3e1b8ce670762a7174c623e10cdcd18ea036604e85bb
3. The 0ubuntu3.4 changelog says "amdgpu: update dmcub fw for dcn32 and
dcn401" (LP: #2168847 is the bug listed there).
+ 4. 0ubuntu3.4 is the only version available in noble-updates now (apt policy):
+ Installed: 20240318.git3b128b60-0ubuntu3.4
+ Candidate: 20240318.git3b128b60-0ubuntu3.4
+ Version table: 3.4 only, from archive.ubuntu.com noble-updates/main
SYMPTOMS (before replacing firmware)
- - The system freezes. The same happens with an older kernel (7.0.0-38), so it
follows the firmware package rather than the kernel.
- - Errors in the kernel log during boot/session:
- amdgpu 0000:03:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB
error - collecting diagnostic data (repeated)
- amdgpu 0000:03:00.0: SMU: No response msg_reg: 2b resp_reg: 0
- amdgpu 0000:03:00.0: Failed to power gate VCN instance 0!
- amdgpu 0000:03:00.0: [drm] *ERROR* DPM disable vcn failed, ret = -62.
- amdgpu 0000:03:00.0: Failed to enable GFXCLK DS!
- amdgpu 0000:03:00.0: Failed to set workload mask 0x00000003
+ - The system freezes. It also happens with an older kernel, so it does not
look kernel-specific.
+ - Errors in the kernel log:
+ amdgpu 0000:03:00.0: [drm] *ERROR* dc_dmub_srv_log_diagnostic_data: DMCUB
error - collecting diagnostic data (repeated)
+ amdgpu 0000:03:00.0: SMU: No response msg_reg: 2b resp_reg: 0
+ amdgpu 0000:03:00.0: Failed to power gate VCN instance 0!
+ amdgpu 0000:03:00.0: [drm] *ERROR* DPM disable vcn failed, ret = -62.
+ amdgpu 0000:03:00.0: Failed to enable GFXCLK DS!
+ amdgpu 0000:03:00.0: Failed to set workload mask 0x00000003
- On reboot: WARNING at drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c:670
(amdgpu_irq_put) in gfx_v12_0_hw_fini, plus REG_WAIT timeouts in
dcn32_hubp_pg_control, dcn32_dsc_pg_control and optc401_disable_crtc.
- - Once the GPU became very hot within seconds and the fans spun up for the
first time; I powered the PC off. The card was still detected and worked at
idle afterwards.
+ - Once the GPU became very hot within seconds and the fans spun up for the
first time (they are normally stopped, Zero RPM); I powered the PC off. After
it cooled down the card was detected and worked normally with the replaced
firmware.
- The same card works fine under Windows on the same machine.
WORKAROUND / WHAT I DID
I replaced all /lib/firmware/amdgpu/*.bin.zst with files from upstream
linux-firmware main (commit ba7155b16ca7), recompressed with zstd, ran
update-initramfs and put the package on hold. After that:
- No errors from amdgpu/SMU/DMCUB in the log
- SMU firmware 104.80.0, DMUB version 0x0A005000, "SMU is resumed
successfully"
- Sensors work (idle ~31-33 C), glxgears on the dGPU works
CAVEAT
- Since I replaced all amdgpu firmware files at once, I cannot say that the
wrong dcn_4_0_1 file alone caused the hangs. I also have not checked whether
0ubuntu3.3 had the same symlink. The mismatch itself is verifiable in the
package.
+ I replaced all amdgpu firmware files at once, so I cannot say that the wrong
dcn_4_0_1 file alone caused the hangs. I have not checked whether 0ubuntu3.3
had the same symlink. The mismatch itself is verifiable in the package.
EXPECTED
dcn_4_0_1_dmcub.bin.zst should be the DCN 4.0.1 firmware, not a link to the
DCN 3.2.0 one.
-
- Attachments: [apt policy linux-firmware-amd-graphics]; [relevant
- journalctl -b -N -p 3 output]
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2173334
Title:
linux-firmware-amd-graphics 0ubuntu3.4: dcn_4_0_1_dmcub.bin.zst is a
symlink to dcn_3_2_0_dmcub.bin.zst; RX 9070 (Navi 48) hangs with
SMU/DMCUB errors
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux-firmware/+bug/2173334/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
