2.7.3

[benhillis]

What's Changed

• Update Microsoft.WSL.DeviceHost package to 1.2.14-0 by @benhillis in Update Microsoft.WSL.DeviceHost package to 1.2.14-0 #40286
• Disable shared compilation when building wslsettings by @OneBlue in Disable shared compilation when building wslsettings #40289
• diagnostics: collect MSI verbose install logs in diagnostic bundle by @yeelam-gordon in diagnostics: collect MSI verbose install logs in diagnostic bundle #40218
• Fix NuGet package path validation in nuget-stage pipeline by @benhillis in Fix NuGet package path validation in nuget-stage pipeline #40293
• test: Add arm64 test distro support by @benhillis in test: Add arm64 test distro support #14500
• test: remove duplicated DNS test coverage by @benhillis in test: remove duplicated DNS test coverage #14522
• Fix: Fail and warn the user when --uninstall is given parameters by @chemwolf6922 in Fix: Fail and warn the user when --uninstall is given parameters #14524
• virito net: revert to previous DNS behavior while we debug an issue with DNS over TCP by @benhillis in virito net: revert to previous DNS behavior while we debug an issue with DNS over TCP #14532
• Update cgmanifest to match CMakeLists.txt by @OneBlue in Update cgmanifest to match CMakeLists.txt #14550
• socketshared: add maximum message size to avoid very large allocations by @benhillis in socketshared: add maximum message size to avoid very large allocations #40050
• Fix: bind interrupted by seccomp signal race by @chemwolf6922 in Fix: bind interrupted by seccomp signal race #14554
• Set Distro Env HOSTTYPE to aarch64 for ARM64 build by @wangxin12 in Set Distro Env HOSTTYPE to aarch64 for ARM64 build #40048
• devicehost: stop re-signing and fix MSI installer failing to replace wsldevicehost.dll by @benhillis in devicehost: stop re-signing and fix MSI installer failing to replace wsldevicehost.dll #40075
• Suppress MSI-initiated reboots during Store updates by @benhillis in Suppress MSI-initiated reboots during Store updates #40079
• build: add Source Link to embed GitHub source mappings in PDBs by @benhillis in build: add Source Link to embed GitHub source mappings in PDBs #40055
• Mask console-getty.service to prevent multi-distro failures (WSL fails to load certain services when multiple WSL instances are running, and one of them is running as root #13595) by @benhillis in Mask console-getty.service to prevent multi-distro failures (#13595) #14490
• Fix random "ERROR_FILE_NOT_FOUND" when unmounting with absolute path by @chemwolf6922 in Fix random "ERROR_FILE_NOT_FOUND" when unmounting with absolute path #40092
• Fixes to route mirroring by @FetoiuCatalin in Fixes to route mirroring #40099
• Fix wrong variable checking by @wangxin12 in Fix wrong variable checking #40185
• Bump Microsoft.NETCore.App.Runtime to 10.0.6 (CVE-2026-32178) by @benhillis in Bump Microsoft.NETCore.App.Runtime to 10.0.6 (CVE-2026-32178) #40207
• Fix VHD ownership after cross-volume move to prevent E_ACCESSDENIED by @benhillis in Fix VHD ownership after cross-volume move to prevent E_ACCESSDENIED #40159
• fix: use {} format specifier instead of %s in Linux LOG_ERROR path by @benhillis in fix: use {} format specifier instead of %s in Linux LOG_ERROR path #40227
• Update MoveVhdOwnership test to WSL2 only by @Copilot in Update MoveVhdOwnership test to WSL2 only #40223
• fix: use ssize_t for readlinkat return value in p9file.cpp by @benhillis in fix: use ssize_t for readlinkat return value in p9file.cpp #40226
• cleanup: extract SkipSignal helper to deduplicate signal skip lists by @benhillis in cleanup: extract SkipSignal helper to deduplicate signal skip lists #40228
• fix: use correct GetAddrInfoTestEntry handler for get_addr_info test by @benhillis in fix: use correct GetAddrInfoTestEntry handler for get_addr_info test #40225
• fix: implement IDisposable on WslConfigService to dispose FileSystemWatcher by @benhillis in fix: implement IDisposable on WslConfigService to dispose FileSystemWatcher #40249
• fix: add waitpid safety net after signalfd setup to prevent SIGCHLD race by @benhillis in fix: add waitpid safety net after signalfd setup to prevent SIGCHLD race #40229
• Settings: Add pending changes tracking and apply confirmation dialog by @dabhattimsft in Settings: Add pending changes tracking and apply confirmation dialog #14425
• Add boot check for ipv6 disabled via registry in mirrored mode by @FetoiuCatalin in Add boot check for ipv6 disabled via registry in mirrored mode #40235
• Fix broken WSLCorePort channel after receive timeout by @chemwolf6922 in Fix broken WSLCorePort channel after receive timeout #14455
• Fix batch of minor bugs by @chemwolf6922 in Fix batch of minor bugs #40197
• Impersonate caller in ResizeDistribution VHD operations by @benhillis in Impersonate caller in ResizeDistribution VHD operations #40253
• Add BUFFER_FIELD macro for safe pretty-printing of flexible array members by @benhillis in Add BUFFER_FIELD macro for safe pretty-printing of flexible array members #40273
• Fix socket race condition in port tracking by @chemwolf6922 in Fix socket race condition in port tracking #40187
• Bound process name buffer in crash dump handler by @benhillis in Bound process name buffer in crash dump handler #40274
• Fix sizeof typo in MessageWriter::GetRelativeIndex by @benhillis in Fix sizeof typo in MessageWriter::GetRelativeIndex #40276
• Add admin protection error message for shadow admin scenarios by @benhillis in Add admin protection error message for shadow admin scenarios #40170

New Contributors

• @wangxin12 made their first contribution in Set Distro Env HOSTTYPE to aarch64 for ARM64 build #40048
• @FetoiuCatalin made their first contribution in Fixes to route mirroring #40099
• @yeelam-gordon made their first contribution in diagnostics: collect MSI verbose install logs in diagnostic bundle #40218

Full Changelog: 2.7.1...2.7.3

---

This discussion was created from the release 2.7.3.

[marcorotondi]

Queston:
Why this version is not out with kernel 6.18.x?

Is this ready for relase?
Is planned a new version of WSL with this kernel?
Thanks

[gadgj]

+1

[benhillis]

6.18 is coming soon

[gadgj]

6.18 is coming soon

This feature is exactly what I’ve been needing. Thx

[eddy2by4]

I see a lot of Pre-releases, but the last actual release was in mid-December last year.
Is there a planned date for an actual release planned ?

[markeissler]

I don't understand this development workflow. Why are there multiple major.minor pre-releases in a series in which follow on candidates add more features instead of just focusing on correcting deficiencies in the previous candidate so that there can be a release? For instance, did someone forget to release 2.7.1? I expect to see a 2.7.1 release that follows a 2.7.1-pre-release or possibly follows a series of 2.7.1-pre-release candidates wherein each carries a shrinking change list and implements no further new functionality or addresses additional bug fixes that weren't addresses in the very first pre-release in the set.

Apologies for the verbosity but I just don't get the pattern being followed by this project.

[benhillis]

Good question, in general you can consider pre-releases more like release candidates. We monitor their health and feedback and if they meet the bar, they are released.

If they are not, we apply iterative fixes and release a new RC.

[markeissler]

Okay. But it's not just "iterative fixes" that are applied is it? It looks like there is often new functionality added which just adds needless complexity to the release process. Freeze the pre-release (aka "candidate"), focus on fixing remaining bugs surfaced through testing and feedback, release, and then move on with the next pre-release. Perhaps these incremental releases just need to cover a smaller scope so they can get out the door faster?

[benhillis]

What you describe is actually the process.

[SkandarS0]

this version breaks custom domains in hosts file that points to docker services

[SkandarS0]

is there a way where I can downgrade to 2.6 ?

[benhillis]

Please file an issue if you haven't already, jus tinstall the 2.6 .msi.

[BBQ4ever]

Adding a 2.7.3 regression data point on Windows 11 26200.8246 (Insider Dev channel).

After wsl --update from 2.6.3.0 → 2.7.3.0, all registered distros fail to start with:

Error code: Wsl/Service/CreateInstance/CreateVm/HCS/ERROR_FILE_NOT_FOUND

What Microsoft-Windows-Hyper-V-Compute-Operational actually shows during the failing boot:

• Event 2000 — CreateComputeSystem, result 0xC0370103 (HCS_E_INVALID_DOCUMENT_PARAMETER)
• Event 2001 — StartComputeSystem, result 0xC0370103
• Event 2009 — system notifications queued: 6/0x80070002 and 3/0x80070002 (the downstream FILE_NOT_FOUND surfaced to the user)

Ruled out by direct check:

• Microsoft-Hyper-V-Hypervisor, VirtualMachinePlatform, HypervisorPlatform features — all Enabled
• bcdedit /set hypervisorlaunchtype Auto + reboot — 0xC0370103 unchanged
• All registered distro VHDs exist on disk
• wsl --install --no-distribution, restart of WslService / vmcompute / hvhost — no effect
• .wslconfig networkingMode=Mirrored — clearing it didn't help

Fix: downgrade in-place by installing wsl.2.6.3.0.x64.msi over the 2.7.3 install via msiexec /i ... /quiet. Distros and VHDs survive. Note that wsl --update --rollback is not available in 2.7.3 — it returns E_INVALIDARG.

@benhillis the consistent 0xC0370103 (invalid document parameter) at CreateComputeSystem time strongly suggests the 2.7.3 VM-definition schema isn't accepted by the vmcompute.exe shipped in 26200.8246. Worth calling out a minimum Windows build requirement in the 2.7.x release notes — there's nothing in the current changelog indicating a host-side dependency change.

[chemwolf6922]

Could you please check if the C:\Program Files\WSL\system.vhd and C:\Program Files\WSL\tools\modules.vhd files are missing after the failed installation?
It will also be helpful if you could collect the install logs following the instructions here. Thanks. #40488 (comment)

[SomeDeveloper13]

The system.vhd and modules.vhd files are missing for me.

This is probably unrelated, but I'm also having trouble downloading the 2.7.3 msi (x64) file from here. It starts to download, but dies partway through. (not sure if this is just a networking issue or not, but would have expected that to block the whole file.) I was able to download the msixbundle file, though. Verified that the download issue was an unrelated network issue. Was able to get it downloaded.

[SomeDeveloper13]

I was able to repair the install by downloading the 2.7.3 msi file, then shift-right-clicking on the msi, and selecting "repair".
This added the missing files back.

[BBQ4ever]

@chemwolf6922 — yes, both files were missing on the broken 2.7.3 install, and I think I found why: the MSI installer triggered an unprompted Windows restart mid-install.

Missing-file confirmation

From my diagnostic capture before downgrading, C:\Program Files\WSL\tools\ under 2.7.3 contained only:

bsdtar      1233912
init        2836528
initrd.img  2836768
kernel     16023552

A recursive search of C:\Program Files\WSL\ for *.vhd*, *.img, vmlinux* returned only tools\initrd.img. No system.vhd anywhere, no modules.vhd.

After downgrading to 2.6.3 over the top (no Windows reboot, no feature change), both files are now present:

C:\Program Files\WSL\system.vhd            407,491,072
C:\Program Files\WSL\tools\modules.vhd     195,080,704

The restart that interrupted the install

System event log, ID 1074, at the exact moment wsl --update was running:

[2026-05-11 22:28:14] User32 / Event 1074
Process msiexec.exe has initiated the restart of computer BBQ4EVER
on behalf of user NT AUTHORITY\SYSTEM for the following reason:
  Reason code: 0x80030002
  Shutdown type: restart
  Comment: The Windows Installer initiated a system restart to complete
  or continue the configuration of 'Windows Subsystem for Linux'.

msiexec.exe (running as SYSTEM, invoked by wsl --update) called the schedule-reboot API and Windows restarted without any user UAC / confirmation prompt. After the reboot, the install presumably tried to resume, but system.vhd and modules.vhd were never laid down on disk. This matches the symptom: HCS later fails CreateComputeSystem with 0xC0370103 because the VM definition references VHDs that don't exist.

This may relate to the 2.7.3 changelog item "suppressing MSI-initiated reboots during Store updates" — looks like that fix covers the Store install path but not the wsl --update MSI path.

Logs we have / don't have

• No verbose MSI log from the broken 2.7.3 upgrade — wsl --update runs msiexec without /l*v by default.
• I do have a full msiexec /l*v log from the 2.6.3 downgrade install (~2.8 MB) if a known-good install trace would be useful for comparison — happy to attach.

I'd like to capture a fresh 2.7.3 install log with /l*v armed, but this is my daily-driver workstation and 24.04 had 235 GB of working data on it, so I'd rather not re-break it without a spare window. If you can wait, I'll attempt a reproduction on a spare environment when I get a chance and attach the verbose MSI log + an updated event log dump.

In the meantime, if there's anything I can pull from the current (downgraded, working) state — registry, install receipts, downgrade MSI log — let me know what would help most.

[chemwolf6922]

Hi, could you please share these files? These do not require a repro:

"C:\Windows\Temp\wsl-install-log.txt"
"$env:TEMP\wsl-install-logs.txt"  (may not exist)

[BBQ4ever]

@chemwolf6922 — both logs were preserved from the original failed 2.7.3 install.

Uploaded as a secret gist: https://gist.github.com/BBQ4ever/52e437af3324aafe02bf511027f7573e

• wsl-install-log.txt — 2 KB, high-level WSL update timeline (from C:\Windows\Temp\)
• wsl-install-logs.txt — 3.4 MB, full MSI verbose log (from %TEMP%\)

Key markers in case they help triage quickly:

wsl-install-log.txt shows the upgrade ending with MSI return code 1641:

2026-05-12T05:25:50.5810393Z C:\Program Files\WSL\wsl.exe[2.6.3.0]: Upgrading via MSI package: ...wsl.2.7.3.0.x64.msi
...
2026-05-12T05:28:13.9846693Z C:\Windows\System32\MsiExec.exe[2.7.3.0]: MSI install: WslFinalizeInstallation
2026-05-12T05:28:14.1085876Z C:\Program Files\WSL\wsl.exe[2.6.3.0]: MSI upgrade result: 1641

1641 = ERROR_SUCCESS_REBOOT_INITIATED. The MSI verbose log corroborates and ends precisely there:

MSI (s) (E8:80) [22:28:14:052]: Windows Installer installed the product. ...
MSI (s) (E8:80) [22:28:14:053]: Value of RebootAction property is 
MSI (s) (E8:80) [22:28:14:053]: Windows Installer requires a system restart. ...RestartType: 1. RestartReason: 0.
MSI (s) (E8:80) [22:28:14:082]: MainEngineThread is returning 1641
=== Logging stopped: 2026/5/11  22:28:14 ===

So the MSI declared success (status 0) but with reboot required (RestartType = 1, RestartReason = 0), returned 1641, and the system rebooted immediately. The RebootAction property was empty — I'm guessing that means no explicit reboot-prompt action was scheduled, so Windows rebooted without any user-visible confirmation. Whatever was supposed to continue after the reboot apparently never laid down system.vhd / modules.vhd.

Also worth flagging: @SomeDeveloper13 found that shift-right-clicking the 2.7.3 MSI and choosing "Repair" restores the missing files — that's a less destructive workaround than my full downgrade, for anyone landing here from search.