-
Notifications
You must be signed in to change notification settings - Fork 1.8k
Collect WSLg logs in collect-wsl-logs.ps1 #40913
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: master
Are you sure you want to change the base?
Changes from all commits
51c4901
4f23851
9c66ebe
af22996
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,79 @@ | ||
| # Analyzing WSLg logs | ||
|
|
||
| WSLg is the component that runs graphical (GUI) and audio Linux applications on WSL. Its | ||
| source lives in a separate repository: https://github.com/microsoft/wslg. The canonical list | ||
| of diagnostics to request for a WSLg bug is the WSLg repo bug report template | ||
| (`.github/ISSUE_TEMPLATE/bug_report.yml` in microsoft/wslg). | ||
|
|
||
| `diagnostics/collect-wsl-logs.ps1` collects the logs into a `wslg/` subfolder of the log archive | ||
| (via `wsl.exe --system --user root`, with a timeout so a wedged service can't hang collection), so a | ||
| standard WSL log collection already contains the log files below. Crash dumps (`dumps/`, | ||
| `wsl-crashes/`) are only collected when the script is run with `-Dump`. | ||
|
|
||
| ## Architecture (just enough to read the logs) | ||
|
|
||
| WSLg runs a **system distro** (a small Azure Linux VM, separate from the user distro) that hosts: | ||
| - **weston** (Wayland compositor) with the `rdp-backend.so` / `rdprail-shell.so` modules | ||
| - **Xwayland** for X11 apps | ||
| - **pulseaudio** with an RDP sink/source for audio | ||
| - **FreeRDP** to stream the surfaces over an hvsocket to the Windows RDP client (`msrdc.exe`) | ||
|
|
||
| `weston.log` and `wlog.log` are written by weston/FreeRDP, `pulseaudio.log` by pulseaudio, and | ||
| `stderr.log` is the combined stderr of `WSLGd` and the processes it launches. | ||
|
|
||
| ## Files (in the `wslg/` folder, sourced from `/mnt/wslg`) | ||
|
|
||
| | File | Contents | | ||
| |---|---| | ||
| | `versions.txt` | WSLg version, architecture, build date, component git hashes (weston, FreeRDP, mesa, pulseaudio) | | ||
| | `weston.log` | Weston compositor + RDP backend log | | ||
| | `wlog.log` | FreeRDP (WLog) log | | ||
| | `pulseaudio.log` | PulseAudio log | | ||
| | `stderr.log` | `WSLGd` and child-process stderr | | ||
| | `dumps/` | Legacy WSLg crash dumps (older builds only; collected with `-Dump`) | | ||
| | `wsl-crashes/` | Host-side WSLg crash dumps copied from `%TEMP%\wsl-crashes` (newer builds, e.g. `core.weston`; collected with `-Dump`) | | ||
|
|
||
| ## Reading tips and common signatures | ||
|
|
||
| - **`weston.log` is truncated on every boot.** Weston is started with `--log=/mnt/wslg/weston.log` | ||
| and opens it truncating, so the file only ever contains the **latest** system-distro boot. To | ||
| tell whether weston restarted, compare the boot timestamp on the first line (`weston 9.0.0 ...`) | ||
| across snapshots, or correlate with `dmesg`. | ||
|
|
||
| - **System distro cycling / teardown.** In `dmesg` (collected via the ETL `GuestLog` events or | ||
| the debug console), this pair means the WSLg system distro's `init` exited and the distro was | ||
| torn down (idle timeout or shutdown), which also restarts weston: | ||
| ``` | ||
| Exception: Operation canceled @p9io.cpp:258 (AcceptAsync) | ||
| WSL (1 - init()) ERROR: InitEntryUtilityVm:2551: Init has exited. Terminating distribution | ||
| ``` | ||
| Repeating every ~25-60s with no GUI client attached is usually normal idle teardown, not a crash. | ||
| A genuine weston crash instead leaves a core dump in `wsl-crashes/` (or `dumps/`). | ||
|
|
||
| - **No GPU acceleration (software rendering).** This means the virtual GPU (`/dev/dxg`, d3d12 mesa) | ||
| is not usable, so rendering falls back to CPU: | ||
| ``` | ||
| Xwayland glamor: GBM Wayland interfaces not available | ||
| Failed to initialize glamor, falling back to sw | ||
| ``` | ||
| Investigate GPU driver presence, `/dev/dxg`, and the `mesa`/`d3d12` stack if the user reports | ||
| black windows, slow rendering, or missing 3D. | ||
|
|
||
| - **Audio sink not connected (usually benign at boot).** Expected until an RDP/audio client attaches: | ||
| ``` | ||
| [rdp-sink] RDP Sink - Trying to connect to /mnt/wslg/PulseAudioRDPSink | ||
| [rdp-sink] Connected failed | ||
| ``` | ||
| If it persists while audio is actively broken, look at the RDP client (`msrdc`) audio channel. | ||
|
|
||
| - **Mostly-benign startup noise** (not a root cause on its own): | ||
| - `WSLGd: Exception: No such file or directory @FontMonitor.cpp:280` - font share not mounted. | ||
| - `XDG_RUNTIME_DIR "/mnt/wslg/runtime-dir" ... mode 040777` (should be 0700) - common warning. | ||
| - `dbus: Unknown username "pulse"` / `Option "-listen" is deprecated` - harmless. | ||
|
|
||
| ## When to escalate to the WSLg repo | ||
|
|
||
| WSLg code is not in this repository. Fixes to weston, FreeRDP, pulseaudio, or the RDP backend | ||
| belong in https://github.com/microsoft/wslg. Use a commit-pinned permalink when referencing files | ||
| there so the link does not drift, e.g. | ||
| `https://github.com/microsoft/wslg/blob/<commit-sha>/<path>`. |
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -341,6 +341,49 @@ if ($LogProfile -eq "networking") | |
| Remove-Item $networkingBashScript | ||
| } | ||
|
|
||
| # Collect WSLg logs (https://github.com/microsoft/wslg) | ||
| $wslgFolder = "$folder/wslg" | ||
| mkdir -p $wslgFolder | Out-Null | ||
|
|
||
| # Run in a job with a timeout so a wedged WSL service can't hang collection. --system --user root | ||
| # reaches /mnt/wslg even when the default distro is WSL1 or isn't running, and can read root-only logs. | ||
| $wslgJob = Start-Job -ScriptBlock { | ||
| param($DestFull, $CollectDumps) | ||
|
|
||
| $destWsl = & wsl.exe --system --user root -e wslpath -u "$DestFull" 2>$null | ||
| if ([string]::IsNullOrWhiteSpace($destWsl)) { return } | ||
| $destWsl = $destWsl.Trim() | ||
|
|
||
| # Destination is passed as $1 so paths containing a single quote are handled safely. | ||
| & wsl.exe --system --user root -e sh -c 'cp /mnt/wslg/pulseaudio.log /mnt/wslg/weston.log /mnt/wslg/wlog.log /mnt/wslg/stderr.log /mnt/wslg/versions.txt "$1/" 2>/dev/null; exit 0' sh "$destWsl" | ||
|
Collaborator
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I'm a bit confused by this, what does the
Collaborator
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Is this an edit error maybe ? |
||
|
|
||
| if ($CollectDumps) | ||
| { | ||
| & wsl.exe --system --user root -e sh -c '[ -d /mnt/wslg/dumps ] && cp -r /mnt/wslg/dumps "$1/dumps"; exit 0' sh "$destWsl" | ||
| } | ||
| } -ArgumentList (Resolve-Path $wslgFolder).Path, ([bool]$Dump) | ||
|
|
||
| if (Wait-Job $wslgJob -Timeout 60) | ||
|
Collaborator
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. 60 seconds feels like a lot for this, I'd recommend something like maybe 20 seconds |
||
| { | ||
| Receive-Job $wslgJob | Out-Null | ||
| } | ||
| else | ||
| { | ||
| Write-Host -ForegroundColor Yellow "WSLg log collection timed out and was skipped." | ||
| Stop-Job $wslgJob | ||
| } | ||
| Remove-Job $wslgJob -Force | ||
|
|
||
| # Crash dumps are only collected with -Dump, since users may not expect dumps to be published by default. | ||
| if ($Dump) | ||
| { | ||
| $wslCrashes = "$env:TEMP\wsl-crashes" | ||
| if (Test-Path $wslCrashes) | ||
| { | ||
| Copy-Item $wslCrashes "$wslgFolder/wsl-crashes" -Recurse -ErrorAction Ignore | ||
| } | ||
| } | ||
|
|
||
| if ($Dump) | ||
| { | ||
| $Assembly = [PSObject].Assembly.GetType('System.Management.Automation.WindowsErrorReporting') | ||
|
|
||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: We should probably swap and do the Trim() before the null check