飞牛影视或jellyfin硬件转码时web端和app端资源管理仪表盘GPU利用率显示为0

杨明

目前已经升级到1.1.3104，飞牛影视或jellyfin硬件转码时web端和app端资源管理仪表盘GPU利用率显示为0

系统版本：X86

设备环境：esxi虚拟机、DDNS公网

BUG现象：目前已经升级到1.1.3104，飞牛影视或jellyfin硬件转码时web端和app端资源管理仪表盘GPU利用率显示为0

出现频率：必现

精简后jellyfin日志文件：

ffmpeg version 7.1.3-Jellyfin
built with gcc 12 (Debian 12.2.0-14+deb12u1)
configuration: --enable-opencl --enable-vaapi --enable-libvpl --enable-ffnvcodec --enable-cuda --enable-libfdk-aac ...
libva info: VA-API version 1.22.0
libva info: Trying to open /usr/lib/jellyfin-ffmpeg/lib/dri/iHD_drv_video.so
libva info: va_openDriver() returns 0

/usr/lib/jellyfin-ffmpeg/ffmpeg -**yzeduration 200M -probesize 1G -f matroska -init_hw_device vaapi=va:,vendor_id=0x8086,driver=iHD -init_hw_device qsv=qs@va -init_hw_device opencl=ocl@va -filter_hw_device qs -hwaccel vaapi -hwaccel_output_format vaapi -i file:"/media/movie/emule/A.Quiet.Place.2018.UHD.BluRay.2160p.TrueHD.Atmos.7.1.x265.10bit.HDR-BeiTai/A.Quiet.Place.2018.UHD.BluRay.2160p.TrueHD.Atmos.7.1.x265.10bit.HDR-BeiTai.mkv" -map 0:0 -map 0:1 -codec✌️0 h264_qsv -low_power 1 -preset veryfast -b:v 9616000 -filter_complex "[0:6]scale,scale=-1:804,format=bgra,hwupload=derive_device=qsv[sub];[0:0]setparams=color_primaries=bt2020:color_trc=smpte2084,scale_vaapi=w=1920:h=804,hwmap=derive_device=opencl,tonemap_opencl=format=nv12:p=bt709:t=bt709:m=bt709:tonemap=bt2390:peak=100,hwmap=derive_device=qsv,format=qsv[main];[main][sub]overlay_qsv" -codec🅰️0 libfdk_aac -ac 2 -y "/cache/transcodes/1578f763c13bf47bf33328635622b56d.m3u8"

Input #0, matroska:
Stream #0:0(eng): Video: hevc (Main 10), yuv420p10le(tv, bt2020nc/bt2020/smpte2084), 3840x1608, 23.98 fps
Stream #0:1(eng): Audio: truehd (Dolby TrueHD + Dolby Atmos), 48000 Hz, 7.1, s32 (24 bit)
Stream #0:6(chi): Subtitle: hdmv_pgs_subtitle (pgssub), 1920x1080 (default)

Stream mapping:
Stream #0:0 (hevc) -> setparams (tonemap_opencl)
Stream #0:6 (pgssub) -> scale (overlay_qsv)
Stream #0:1 -> libfdk_aac
Output #0, hls:
Stream #0:0: Video: h264, qsv(tv, bt709, progressive), 1920x804, 9616 kb/s
Stream #0:1: Audio: aac, 48000 Hz, stereo, 256 kb/s

飞牛运营同学

感谢反馈，从日志看当前硬件转码已正常调用 GPU。请确认一下在之前版本中，EXSI 直通环境下资源管理里的 GPU 利用率是否可以正常显示，试试重装一下i915驱动后看看。

杨明

root@dsm:~# cat /sys/kernel/debug/dri/0/i915_binner_status
cat: /sys/kernel/debug/dri/0/i915_binner_status: No such file or directory
root@dsm:~# cat /sys/kernel/debug/dri/0/i915_clients
cat: /sys/kernel/debug/dri/0/i915_clients: No such file or directory
root@dsm:~# mount -t debugfs none /sys/kernel/debug
mount: /sys/kernel/debug: none already mounted on /sys/fs/pstore.
       dmesg(1) may have more information after failed mount system call.
root@dsm:~# cat /sys/kernel/debug/dri/0/i915_clients
cat: /sys/kernel/debug/dri/0/i915_clients: No such file or directory
root@dsm:~# mkdir /tmp/kernel_debug && mount -t debugfs none /tmp/kernel_debug
root@dsm:~# ls -l /tmp/kernel_debug/dri/0/
total 0
-r--r--r-- 1 root root 0 May 16 07:38 clients
drwxr-xr-x 3 root root 0 May 16 07:38 crtc-0
drwxr-xr-x 3 root root 0 May 16 07:38 crtc-1
drwxr-xr-x 3 root root 0 May 16 07:38 crtc-2
drwxr-xr-x 2 root root 0 May 16 07:38 DP-1
drwxr-xr-x 2 root root 0 May 16 07:38 DP-2
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-0
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-1
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-2
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-3
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-4
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-5
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-6
drwxr-xr-x 2 root root 0 May 16 07:38 encoder-7
-r--r--r-- 1 root root 0 May 16 07:38 framebuffer
-r--r--r-- 1 root root 0 May 16 07:38 gem_names
drwxr-xr-x 3 root root 0 May 16 07:38 gt0
drwxr-xr-x 3 root root 0 May 16 07:38 HDMI-A-1
drwxr-xr-x 3 root root 0 May 16 07:38 HDMI-A-2
-r--r--r-- 1 root root 0 May 16 07:38 i915_capabilities
-r--r--r-- 1 root root 0 May 16 07:38 i915_cdclk_info
-rw-r--r-- 1 root root 0 May 16 07:38 i915_cur_wm_latency
-r--r--r-- 1 root root 0 May 16 07:38 i915_ddb_info
-r--r--r-- 1 root root 0 May 16 07:38 i915_display_info
-r--r--r-- 1 root root 0 May 16 07:38 i915_dmc_info
-r--r--r-- 1 root root 0 May 16 07:38 i915_dp_mst_info
-rw-r--r-- 1 root root 0 May 16 07:38 i915_dp_test_active
-rw-r--r-- 1 root root 0 May 16 07:38 i915_dp_test_data
-rw-r--r-- 1 root root 0 May 16 07:38 i915_dp_test_type
-rw-r--r-- 1 root root 0 May 16 07:38 i915_edp_psr_debug
-r--r--r-- 1 root root 0 May 16 07:38 i915_edp_psr_status
-r--r--r-- 1 root root 0 May 16 07:38 i915_engine_info
-rw-r--r-- 1 root root 0 May 16 07:38 i915_error_state
-rw-r--r-- 1 root root 0 May 16 07:38 i915_fbc_false_color
-r--r--r-- 1 root root 0 May 16 07:38 i915_fbc_status
-rw-r--r-- 1 root root 0 May 16 07:38 i915_fifo_underrun_reset
-r-------- 1 root root 0 May 16 07:38 i915_forcewake_user
-r--r--r-- 1 root root 0 May 16 07:38 i915_frequency_info
-r--r--r-- 1 root root 0 May 16 07:38 i915_frontbuffer_tracking
-rw-r--r-- 1 root root 0 May 16 07:38 i915_gem_drop_caches
-r--r--r-- 1 root root 0 May 16 07:38 i915_gem_framebuffer
-r--r--r-- 1 root root 0 May 16 07:38 i915_gem_objects
-rw-r--r-- 1 root root 0 May 16 07:38 i915_gpu_info
-rw-r--r-- 1 root root 0 May 16 07:38 i915_hpd_short_storm_ctl
-rw-r--r-- 1 root root 0 May 16 07:38 i915_hpd_storm_ctl
-rw-r--r-- 1 root root 0 May 16 07:38 i915_ignore_long_hpd
-rw-r--r-- 1 root root 0 May 16 07:38 i915_ipc_status
-r--r--r-- 1 root root 0 May 16 07:38 i915_lpsp_status
-r--r--r-- 1 root root 0 May 16 07:38 i915_opregion
drwxr-xr-x 2 root root 0 May 16 07:38 i915_params
-rw-r--r-- 1 root root 0 May 16 07:38 i915_perf_noa_delay
-r--r--r-- 1 root root 0 May 16 07:38 i915_power_domain_info
-rw-r--r-- 1 root root 0 May 16 07:38 i915_pri_wm_latency
-r--r--r-- 1 root root 0 May 16 07:38 i915_rps_boost_info
-r--r--r-- 1 root root 0 May 16 07:38 i915_runtime_pm_status
-r--r--r-- 1 root root 0 May 16 07:38 i915_shared_dplls_info
-rw-r--r-- 1 root root 0 May 16 07:38 i915_spr_wm_latency
-r--r--r-- 1 root root 0 May 16 07:38 i915_sr_status
-r--r--r-- 1 root root 0 May 16 07:38 i915_sseu_status
-r--r--r-- 1 root root 0 May 16 07:38 i915_swizzle_info
-r--r--r-- 1 root root 0 May 16 07:38 i915_vbt
-r--r--r-- 1 root root 0 May 16 07:38 i915_wa_registers
-rw-r--r-- 1 root root 0 May 16 07:38 i915_wedged
-r--r--r-- 1 root root 0 May 16 07:38 intel_display_caps
-r--r--r-- 1 root root 0 May 16 07:38 internal_clients
-r--r--r-- 1 root root 0 May 16 07:38 name
-r--r--r-- 1 root root 0 May 16 07:38 state
root@dsm:~# sysctl -w kernel.perf_event_paranoid=-1 && INTEL_GPU_TOP_DEBUGFS=/tmp/kernel_debug/dri/0 intel_gpu_top
kernel.perf_event_paranoid = -1
Failed to detect engines! (No such file or directory)
(Kernel 4.16 or newer is required for i915 PMU support.)
Segmentation fault

以上为由AI指导的修复过程，AI告诉我“即便是把环境路径死死地喂给它，它居然还是在旧地方（PMU 底层内核接口）摔了跟头，继续 Segmentation fault 闪退，新版的飞牛 OS（1.1.3104）在升级了 Linux 内核版本（应该升级到了 6.x 较高版本内核）后，彻底改变了对老一代 Intel GPU（Gemini Lake / 第九代及更早核显）性能监控单元（PMU）的代码兼容性。在新内核里，J4105 的硬件计数器在没有开启特定微代码（GuC）时，直接拒绝向新内核提供监控指标，新版内核在 ESXi / PVE 虚拟机直通 J4105、J3455 等老款低功耗 CPU 时，前端仪表盘无法抓取 GPU 数据。