Skip to main content
Topic: Some VGA monitors not working since the 6.2 kernel (Read 707 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Some VGA monitors not working since the 6.2 kernel

On 2 laptops (Dell M4400 and M4500) which have a VGA port, a Maxim MX11-LCD15 TV with a VGA port stopped working, also I tried another Panasonic TX-26LXD500 TV using VGA today and it's the same. (Using HDMI to the Panasonic from a Dell E7470 in a brief test with a current kernel it had an outsized picture, the edges were offscreen, but that might be something different, I haven't investigated that at all.) These are older TV's, from the 2000's, I guess the EDID might not be entirely compliant with modern standards.
The 6.1.12-artix1-1 kernel works, later ones don't - 6.2.2-artix1-1 and also Arch ones 6.2.1-arch1-1 and 6.2.0-arch1-1.
A newish AEC monitor (not TV) did work with the latest kernels via VGA though so this is not all VGA, just some, and it isn't the output from the Dell laptops being blocked. In the newer kernels there is simply no monitor detected at all, it doesn't appear.
As this problem now seems reasonably widespread in effect and not just one weird old TV / monitor issue (but two!) it seems worth a mention, although it's most likely more of a kernel problem than an Artix one. I hadn't noticed this until recently as I don't often connect monitors to my laptops.

Re: Some VGA monitors not working since the 6.2 kernel

Reply #1
I found this suggestion searching about:
https://askubuntu.com/questions/1481695/after-latest-linux-kernel-6-2-update-monitor-is-black-does-not-recognize-signal
But that doesn't seem to be it. I tried evdi from omniverse, evdi from the AUR (which is currently newer than omniverse), evdi-compat-git and evdi-git. Then I tried ddcci-driver-linux-dkms which is some other monitor driver from the AUR but none of those seemed to help.

Re: Some VGA monitors not working since the 6.2 kernel

Reply #2
I found the commit where this started, but it's a big one and might require further refinement somehow:
Code: [Select]
commit a594533df0f6ca391da003f43d53b336a2d23ffa (HEAD)
Merge: cdb9d3537711 66efff515a65
Author: Linus Torvalds <[email protected]>
Date:   Tue Dec 13 11:59:58 2022 -0800

    Merge tag 'drm-next-2022-12-13' of git://anongit.freedesktop.org/drm/drm
   
    Pull drm updates from Dave Airlie:
     "The biggest highlight is that the accel subsystem framework is merged.
      Hopefully for 6.3 we will be able to line up a driver to use it.
   
      In drivers land, i915 enables DG2 support by default now, and nouveau
      has a big stability refactoring and initial ampere support, AMD
      includes new hw IP support and should build on ARM again. There is
      also an ofdrm driver to take over offb on platforms it's used.
   
      Stuff outside my tree, the dma-buf patches hit a few places, the vc4
      firmware changes also do, and i915 has some interactions with MEI for
      discrete GPUs. I think all of those should have been acked/reviewed by
      relevant parties.
   
      New driver:
       - ofdrm - replacement for offb
   
      fbdev:
       - add support for nomodeset
   
      fourcc:
       - add Vivante tiled modifier
   
      core:
       - atomic-helpers: CRTC primary plane test fixes, fb access hooks
       - connector: TV API consistency, cmdline parser improvements
       - send connector hotplug on cleanup
       - sort makefile objects
   
      tests:
       - sort kunit tests
       - improve DP-MST tests
       - add kunit helpers to create a device
   
      sched:
       - module param for scheduling policy
       - refcounting fix
   
      buddy:
       - add back random seed log
   
      ttm:
       - convert ttm_resource to size_t
       - optimize pool allocations
   
      edid:
       - HFVSDB parsing support fixes
       - logging/debug improvements
       - DSC quirks
   
      dma-buf:
       - Add unlocked vmap and attachment mapping
       - move drivers to common locking convention
       - locking improvements
   
      firmware:
       - new API for rPI firmware and vc4
   
      xilinx:
       - zynqmp: displayport bridge support
       - dpsub fix
   
      bridge:
       - adv7533: Remove dynamic lane switching
       - it6505: Runtime PM support, sync improvements
       - ps8640: Handle AUX defer messages
       - tc358775: Drop soft-reset over I2C
   
      panel:
       - panel-edp: Add INX N116BGE-EA2 C2 and C4 support.
       - Jadard JD9365DA-H3
       - NewVision NV3051D
   
      amdgpu:
       - DCN support on ARM
       - DCN 2.1 secure display
       - Sienna Cichlid mode2 reset fixes
       - new GC 11.x firmware versions
       - drop AMD specific DSC workarounds in favour of drm code
       - clang warning fixes
       - scheduler rework
       - SR-IOV fixes
       - GPUVM locking fixes
       - fix memory leak in CS IOCTL error path
       - flexible array updates
       - enable new GC/PSP/SMU/NBIO IP
       - GFX preemption support for gfx9
   
      amdkfd:
       - cache size fixes
       - userptr fixes
       - enable cooperative launch on gfx 10.3
       - enable GC 11.0.4 KFD support
   
      radeon:
       - replace kmap with kmap_local_page
       - ACPI ref count fix
       - HDA audio notifier support
   
      i915:
       - DG2 enabled by default
       - MTL enablement work
       - hotplug refactoring
       - VBT improvements
       - Display and watermark refactoring
       - ADL-P workaround
       - temp disable runtime_pm for discrete-
       - fix for A380 as a secondary GPU
       - Wa_18017747507 for DG2
       - CS timestamp support fixes for gen5 and earlier
       - never purge busy TTM objects
       - use i915_sg_dma_sizes for all backends
       - demote GuC kernel contexts to normal priority
       - gvt: refactor for new MDEV interface
       - enable DC power states on eDP ports
       - fix gen 2/3 workarounds
   
      nouveau:
       - fix page fault handling
       - Ampere acceleration support
       - driver stability improvements
       - nva3 backlight support
   
      msm:
       - MSM_INFO_GET_FLAGS support
       - DPU: XR30 and P010 image formats
       - Qualcomm SM6115 support
       - DSI PHY support for QCM2290
       - HDMI: refactored dev init path
       - remove exclusive-fence hack
       - fix speed-bin detection
       - enable clamp to idle on 7c3
       - improved hangcheck detection
   
      vmwgfx:
       - fb and cursor refactoring
       - convert to generic hashtable
       - cursor improvements
   
      etnaviv:
       - hw workarounds
       - softpin MMU fixes
   
      ast:
       - atomic gamma LUT support
       - convert to SHMEM
   
      lcdif:
       - support YUV planes
       - Increase DMA burst size
       - FIFO threshold tuning
   
      meson:
       - fix return type of cvbs mode_valid
   
      mgag200:
       - fix PLL setup on some revisions
   
      sun4i:
       - A100 and D1 support
   
      udl:
       - modesetting improvements
       - hot unplug support
   
      vc4:
       - support PAL-M
       - fix regression preventing 4K @ 60Hz
       - fix NULL ptr deref
   
      v3d:
       - switch to drm managed resources
   
      renesas:
       - RZ/G2L DSI support
       - DU Kconfig cleanup
   
      mediatek:
       - fixup dpi and hdmi
       - MT8188 dpi support
       - MT8195 AFBC support
   
      tegra:
       - NVDEC hardware on Tegra234 SoC
   
      hdlcd:
       - switch to drm managed resources
   
      ingenic:
       - fix registration error path
   
      hisilicon:
       - convert to drm_mode_init
   
      maildp:
       - use managed resources
   
      mtk:
       - use drm_mode_init
   
      rockchip:
      udl:
       - modesetting improvements
       - hot unplug support
   
      vc4:
       - support PAL-M
       - fix regression preventing 4K @ 60Hz
       - fix NULL ptr deref
   
      v3d:
       - switch to drm managed resources
   
      renesas:
       - RZ/G2L DSI support
       - DU Kconfig cleanup
   
      mediatek:
       - fixup dpi and hdmi
       - MT8188 dpi support
       - MT8195 AFBC support
   
      tegra:
       - NVDEC hardware on Tegra234 SoC
   
      hdlcd:
       - switch to drm managed resources
   
      ingenic:
       - fix registration error path
   
      hisilicon:
       - convert to drm_mode_init
   
      maildp:
       - use managed resources
   
      mtk:
       - use drm_mode_init
   
      rockchip:
       - use drm_mode_copy"
   
    * tag 'drm-next-2022-12-13' of git://anongit.freedesktop.org/drm/drm: (1397 commits)
      drm/amdgpu: fix mmhub register base coding error
      drm/amdgpu: add tmz support for GC IP v11.0.4
      drm/amdgpu: enable GFX Clock Gating control for GC IP v11.0.4
      drm/amdgpu: enable GFX Power Gating for GC IP v11.0.4
      drm/amdgpu: enable GFX IP v11.0.4 CG support
      drm/amdgpu: Make amdgpu_ring_mux functions as static
      drm/amdgpu: generally allow over-commit during BO allocation
      drm/amd/display: fix array index out of bound error in DCN32 DML
      drm/amd/display: 3.2.215
      drm/amd/display: set optimized required for comp buf changes
      drm/amd/display: Add debug option to skip PSR CRTC disable
      drm/amd/display: correct DML calc error of UrgentLatency
      drm/amd/display: correct static_screen_event_mask
      drm/amd/display: Ensure commit_streams returns the DC return code
      drm/amd/display: read invalid ddc pin status cause engine busy
      drm/amd/display: Bypass DET swath fill check for max clocks
      drm/amd/display: Disable uclk pstate for subvp pipes
      drm/amd/display: Fix DCN2.1 default DSC clocks
      drm/amd/display: Enable dp_hdmi21_pcon support
      drm/amd/display: prevent seamless boot on displays that don't have the preferred dig
      ...




Re: Some VGA monitors not working since the 6.2 kernel

Reply #3
In the https://aur.archlinux.org/packages/linux-drm-next-git kernel tree it is looking like this commit, which is also a merge of several commits, which moves things a bit closer:
Code: [Select]
commit b2268e26869a56e40964f0d912b58ffd076fd453 (HEAD)
Merge: 077bd80083ab 7d579eff3e55
Author: Dave Airlie <[email protected]>
Date:   Fri Dec 9 10:29:56 2022 +1000

    Merge tag 'drm-intel-next-fixes-2022-12-01' of git://anongit.freedesktop.org/drm/drm-intel into drm-next
   
    - Multi-cast register fix (Matt)
    - Fix workarounds on gen2-3 (Tvrtko)
    - Bigjoiner fix (Ville)
    - Make Guc default_list a const data (Jani)
    - Acquire forcewake before uncore read (Umesh)
    - Selftest fix (Umesh)
    - HuC related fixes (Daniele)
    - Fix some incorrect return values (Janusz)
    - Fix a memory leak in bios related code (Xia)
    - Fix VBT send packet port selection (Mikko)
    - DG2's DMC fix bump for Register noclaims and few restore (Gustavo)
   
    Signed-off-by: Dave Airlie <[email protected]>
   
    From: Rodrigo Vivi <[email protected]>
    Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]


Re: Some VGA monitors not working since the 6.2 kernel

Reply #4
I modified the linux-drm-next-git PKGBUILD to build this linux-drm-intel-git source instead, interestingly unlike other kernel builds it output very few messages during the process except for warnings, it must be possible to enable a less verbose mode.
And finally in the third kernel tree the initial problem commit is revealed to be ...
Code: [Select]
commit f530bc60a30bee47ff51b7fb71511fdd058b774a (HEAD)
Author: Ben Skeggs <[email protected]>
Date:   Wed Jun 1 20:46:33 2022 +1000

    drm/nouveau/disp: move HDMI config into acquire + infoframe methods
   
    v2:
    - fix typo in sorhdmi/g84 struct initialiser (kbuild test robot)
    v3:
    - less convoluted flow control in nvkm_uoutp_mthd_acquire_tmds() (lyude)
    v4:
    - we don't support hdmi on original nv50, don't try
   
    Signed-off-by: Ben Skeggs <[email protected]>
    Reviewed-by: Lyude Paul <[email protected]>

$ git diff --compact-summary 9793083f1dd9da8dda0ef68e90934dd7d112203b f530bc60a30bee47ff51b7fb71511fdd058b774a
 drivers/gpu/drm/nouveau/dispnv50/disp.c             | 172 +++++++++++++++++++++++++++++++++++++++++++++--------------------------------------------------------------------
 drivers/gpu/drm/nouveau/include/nvif/cl5070.h       |  14 ----------
 drivers/gpu/drm/nouveau/include/nvif/if0012.h       |  24 ++++++++++++++--
 drivers/gpu/drm/nouveau/include/nvif/outp.h         |   5 +++-
 drivers/gpu/drm/nouveau/nvif/outp.c                 |  28 ++++++++++++++++---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/g84.c      |  91 +++++++++++++++++++++++++++++++++++++-----------------------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/ga102.c    |   5 +---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c    |  93 +++++++++++++++++++++++++++++++++++++------------------------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gk104.c    |  85 +++++++++++++++++++++++++++++++++++---------------------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gm107.c    |   4 +--
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gm200.c    |  13 ++++++---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gp100.c    |   5 +---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gt215.c    |  89 +++++++++++++++++++++++++++++++++++-----------------------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/gv100.c    |  95 +++++++++++++++++++++++++++++++++++++-------------------------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.h      |  23 +++++++++------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp77.c    |   4 +--
 drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp89.c    |   4 +--
 drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.h     |   3 ++
 drivers/gpu/drm/nouveau/nvkm/engine/disp/rootnv50.c |  45 ------------------------------
 drivers/gpu/drm/nouveau/nvkm/engine/disp/tu102.c    |   5 +---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/uoutp.c    |  74 +++++++++++++++++++++++++++++++++++++++++++++++++
 21 files changed, 503 insertions(+), 378 deletions(-)


I haven't tried reverting that in the latest kernel but it's quite large  and is followed by a whole bunch of apparently related commits to Nouveau so there seems little chance  that approach would work, it could break Nouveau entirely with unpredictable and possibly undesirable results, I should warn any unwary readers of this thread!  :'(