суббота

[Bug 2153096] Re: IPU6 webcam (OV2740) on X1 Nano fails on kernel 7.0.0-14/-15: int3472 cannot find GPIO chip INTC1096:00, VSC chain doesn't come up

Same model here (ThinkPad X1 Nano Gen 2, 21E80013US, BIOS N3IET54W 1.34, i7-1280P), Linux Mint 22.3 (noble base), tested with HWE 7.0.0-43 and 6.17.0-42. Same symptom: sensor and /dev/video* present, but no usable frames. Sharing my findings in case they help others and narrow the root cause. 1. The kernel capture path works. After letting libcamera configure the media graph once (cam -c1 -C2 -s role=raw), direct capture from the ISYS node with plain MMAP buffers is clean: v4l2-ctl -d /dev/video8 --stream-mmap=4 --stream-count=60 → BA10 1932x1092, consecutive sequence numbers, steady 30 fps, zero csi2-1 error messages, and the raw Bayer data contains a correct image. 2. Streaming through libcamera fails. Tested with libcamera 0.8.0 built from source, SoftISP in both CPU and GPU mode. Every libcamera session produces csi2-1 error messages (Transfer FIFO overflow, inter-frame packets discarded, CRC / packet header errors, and sometimes “stream stop time out”). Completed requests report bytesused: 0 and garbage sequence numbers and timestamps, so SoftISP outputs black frames. Even role=raw requests come back flagged this way. I suspect the buffer import path (libcamera uses DMABUF, while v4l2-ctl MMAP works), but I have not confirmed this. 3. Sensor control writes while streaming cause CSI-2 errors. In one test I changed each control during a clean v4l2-ctl stream: exposure: 0 errors analogue_gain: 7 errors digital_gain: 3 errors vertical_blanking: 4 errors Frames keep coming in plain V4L2. Gain changes from AGC therefore look harmful. 4. IVSC is a red herring on non-Computer-Vision SKUs. The DSDT puts SPFD (INTC1094) under \_SB.PC00.SPI1 unless CVFS == 2. On my unit the BIOS sets CVFS = 0 (CUPN = 8). When I forced CVFS = 2 with a small SSDT, spi- INTC1094:00 appeared on the LJCA SPI bus, but vsc-tp fails with -22 because the wakeupfw LJCA GPIO line (VGPO pin 1) is not valid (gpiochip_line_is_valid = 0). I believe this unit has no IVSC fitted: Computer Vision is optional on this model. Without IVSC the direct ov2740 → CSI2 graph is correct. Working workaround (no libcamera at runtime): Set fixed sensor controls before streaming (exposure=1102, analogue_gain=512). Stream with v4l2-ctl --stream-mmap ... --stream-to=-. Debayer with ffmpeg (-pixel_format bayer_grbg16le -video_size 1952x1092), subtract black level 64, scale 10→16-bit, apply per-channel gains. Output to v4l2loopback (exclusive_caps=1). For simple auto-exposure, adjust only the exposure control. This works in browsers and video-call apps at about 28-30 fps with Secure Boot enabled. I can share the script if useful. -- You received this bug notification because you are subscribed to linux in Ubuntu. Matching subscriptions: Bgg, Bmail, Nb https://bugs.launchpad.net/bugs/2153096 Title: IPU6 webcam (OV2740) on X1 Nano fails on kernel 7.0.0-14/-15: int3472 cannot find GPIO chip INTC1096:00, VSC chain doesn't come up Status in linux package in Ubuntu: New Bug description: Thanks in advance for taking a look. I've revised this description after some follow-up testing; I think my original framing pointed at the wrong thing. What I'm seeing: The webcam on my ThinkPad X1 Nano (Gen 2, OV2740 behind IPU6) currently does not produce frames on either 7.0.0-14 or 7.0.0-15. libcamera and pipewire both see an `ov2740` device, `/dev/video*` are populated, and the IPU6 driver seems happy: intel-ipu6 0000:00:05.0: Found supported sensor INT3474:01 intel-ipu6 0000:00:05.0: Connected 1 cameras But every app that opens the camera fails identically: spa.v4l2: /dev/video0: VIDIOC_S_PARM: Inappropriate ioctl for device spa.v4l2: '/dev/video0' VIDIOC_STREAMON: Link has been severed pw.node: (v4l2_input.pci-0000_00_05.0-110) suspended -> error (Start error: Link has been severed) Cheese, Snapshot, Firefox, Chromium and Zoom all behave the same way. BIOS privacy is ok and the physical toggle has been checked. The kernel-side warnings line up with what I'd expect if the VSC chain never comes up: int3472-discrete INT3472:05: cannot find GPIO chip INTC1096:00, deferring int3472-discrete INT3472:06: cannot find GPIO chip INTC1096:00, deferring ov2740 i2c-INT3474:01: supply AVDD not found, using dummy regulator ov2740 i2c-INT3474:01: supply DOVDD not found, using dummy regulator ov2740 i2c-INT3474:01: supply DVDD not found, using dummy regulator `mei_vsc_hw` is loaded but unbound. `mei_vsc`, `intel_vsc`, `ivsc_csi`, `ivsc_ace` aren't loaded at all. The `INT3472:05/06` deferrals are the chain that powers the OV2740 rails, which is presumably why ov2740 ends up on dummy regulators. What I now believe could be a more interesting (frustrating?) bug than a kernel-version regression: * The camera **was** working on my stock 26.04 install on 2026-04-23. * That install pulled in `linux-image-generic-hwe-26.04 → 7.0.0-14.14`. The same `.14` deb is still installed today. * `linux-firmware` is still at `20260319.git217ca6e4`, libcamera at 0.7.0-1, pipewire/wireplumber at 1.6.2-1 / 0.5.13-1 — none have been upgraded since install (per `/var/log/apt/history.log` and `/var/log/dpkg.log`). * BIOS is unchanged (N3IET53W 1.33, 2025-12-03), `fwupdmgr get-history` is empty, `/etc/modprobe.d/` has nothing camera-relevant, kernel cmdline is stock (`ro quiet splash crashkernel=…`). So as best I can tell, the same kernel/firmware/userspace binaries that worked on day one are now failing, without an apt-tracked change between those two states. The HWE bump from .14 to .15 in my original report was just the moment I noticed. My best guess (and I'd love to be told if it's silly) is that this is a non-deterministic probe ordering issue around the USB-LJCA bridge — the INTC1096 ACPI companion is present at `\_SB_.PC00.XHCI.RHUB.HS08.VGPO` but its GPIO chip never gets registered, so int3472 can't find it, so the rails never come up, so VSC never starts, so streamon severs the link. I don't knw why this would sometimes succeed and sometimes not on otherwise identical software. ACPI layout summary (full data in attached diagnostics): \_SB_.PC00.SPI1.SPFD HID INTC1094 (mei_vsc_hw match) \_SB_.PC00.SPI1.SPFD.CVFD HID INTC1095 \_SB_.PC00.XHCI.RHUB.HS08.VGPO HID INTC1096 (GPIO chip never registers) \_SB_.PC00.XHCI.RHUB.HS08.VIC0/1 HID INTC1097 \_SB_.PC00.XHCI.RHUB.HS08.VSPI HID INTC1098 Side notes in case relevant: * The PCH SPI controller (PCI 0000:00:1f.5, spi0) has one child for the BIOS flash; spi-nor fails JEDEC ID on it. I think this is unrelated. * The Intel USB-LJCA bridge (8086:0b63) at HS08 / INTC1098 brings up spi1 via `spi_ljca`; `spi1` has no children (`waiting_for_supplier=0`). * I have not tried a 6.x kernel and many other distros — my recollection is the camera also did not work on 6.x for me even with the Intel patches, so I don't think rolling back further is productive. Attached `diagnostics.txt` (filed initially against 7.0.0-15) and `diagnostics-7.0.0-14.txt` (captured today on 7.0.0-14) have DMI, package versions, lsmod, lspci/lsusb, the spi_master tree, the relevant ACPI nodes, and the IPU6/MEI/VSC/OV2740/LJCA kernel and libcamera/pipewire log lines. They're the same shape so they can be diffed directly. Happy to do whatever's useful: capture across cold-boot/warm-boot/suspend cycles, watch whether the INTC1096 GPIO chip ever registers across multiple boots, try unbinding/rebinding `usb_ljca`, or run with debug options on `int3472`/`mei_vsc`/`gpio_ljca`. --- ProblemType: Bug ApportVersion: 2.34.0-0ubuntu2 Architecture: amd64 AudioDevicesInUse: USER PID ACCESS COMMAND /dev/snd/controlC0: atb 7927 F.... pipewire atb 7941 F.... wireplumber /dev/snd/seq: atb 7927 F.... pipewire CasperMD5CheckResult: pass CurrentDesktop: ubuntu:GNOME DistroRelease: Ubuntu 26.04 InstallationDate: Installed on 2026-05-17 (2 days ago) InstallationMedia: Ubuntu 26.04 "Resolute Raccoon" - Release amd64 (20260423.1) MachineType: LENOVO 21E8002YSG Package: linux-image-7.0.0-14-generic 7.0.0-14.14 PackageArchitecture: amd64 ProcEnviron: LANG=en_US.UTF-8 PATH=(custom, no user) SHELL=/bin/bash TERM=xterm-256color XDG_RUNTIME_DIR=<set> ProcFB: 0 i915drmfb ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-7.0.0-14-generic root=UUID=61e8b5c4-2752-445a-93d2-a68c8bb7dd3c ro quiet splash crashkernel=2G-4G:320M,4G-32G:512M,32G-64G:1024M,64G-128G:2048M,128G-:4096M ProcVersionSignature: Ubuntu 7.0.0-14.14-generic 7.0.0 Tags: resolute wayland-session Uname: Linux 7.0.0-14-generic x86_64 UpgradeStatus: No upgrade log present (probably fresh install) UserGroups: adm cdrom dip lpadmin lxd plugdev sudo users _MarkForUpload: True dmi.bios.date: 12/03/2025 dmi.bios.release: 1.33 dmi.bios.vendor: LENOVO dmi.bios.version: N3IET53W (1.33 ) dmi.board.asset.tag: Not Available dmi.board.name: 21E8002YSG dmi.board.vendor: LENOVO dmi.board.version: SDK0T76530 WIN dmi.chassis.asset.tag: No Asset Information dmi.chassis.type: 10 dmi.chassis.vendor: LENOVO dmi.chassis.version: None dmi.ec.firmware.release: 1.21 dmi.modalias: dmi:bvnLENOVO:bvrN3IET53W(1.33):bd12/03/2025:br1.33:efr1.21:svnLENOVO:pn21E8002YSG:pvrThinkPadX1NanoGen2:rvnLENOVO:rn21E8002YSG:rvrSDK0T76530WIN:cvnLENOVO:ct10:cvrNone:skuLENOVO_MT_21E8_BU_Think_FM_ThinkPadX1NanoGen2:pfaThinkPadX1NanoGen2: dmi.product.family: ThinkPad X1 Nano Gen 2 dmi.product.name: 21E8002YSG dmi.product.sku: LENOVO_MT_21E8_BU_Think_FM_ThinkPad X1 Nano Gen 2 dmi.product.version: ThinkPad X1 Nano Gen 2 dmi.sys.vendor: LENOVO To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2153096/+subscriptions

Комментариев нет:

Отправить комментарий