Backport Xen patch for zero-length records in migration v2

[maxdev123]

Qubes OS version (e.g., R3.1): R3.2-rc1

Affected TemplateVMs (e.g., fedora-23, if applicable): all

---

Expected behavior: start DispVM

Actual behavior: DispVM doesn't work with any templates

General notes: This issue was manifested since R3.1. I tried all possible available updates

cpuinfo.txt
dmesg.txt
dmesg_xen.txt
libxl-driver.log.txt
lspci.txt
xsession-errors.txt

(current-testing, unstable)

---

Related issues:

[andrewdavidwong]

This is not appropriate for qubes-issues. Please send this to the qubes-users mailing list instead.

[andrewdavidwong]

This is not appropriate for qubes-issues. Please send this to the qubes-users mailing list instead.

[andrewdavidwong]

Also: In case this happens to be a duplicate of #2125, please try starting your DVM template and shutting it down before launching a DispVM.

[andrewdavidwong]

Also: In case this happens to be a duplicate of #2125, please try starting your DVM template and shutting it down before launching a DispVM.

[maxdev123]

I already saw the issue #2125 but the problem seem different. In particular in file libxl-driver.log.txt you can notice this line:

xc: error: X86_PV_VCPU_MSRS record truncated: length 8, min 9: Internal error

[maxdev123]

I already saw the issue #2125 but the problem seem different. In particular in file libxl-driver.log.txt you can notice this line:

xc: error: X86_PV_VCPU_MSRS record truncated: length 8, min 9: Internal error

[andrewdavidwong]

Please send this to qubes-users, at least at first. This is probably a problem somewhat unique to you rather than a bug (since many other people have no problem starting DispVMs). But if it turns out to be a bug, we can reopen this or create a new issue for it. We don't have enough information yet to make a useful ticket.

[andrewdavidwong]

Please send this to qubes-users, at least at first. This is probably a problem somewhat unique to you rather than a bug (since many other people have no problem starting DispVMs). But if it turns out to be a bug, we can reopen this or create a new issue for it. We don't have enough information yet to make a useful ticket.

[marmarek]

Fix is already discussed on xen-devel in thread "Fix issues with zero-length records in migration v2"

[marmarek]

Fix is already discussed on xen-devel in thread "Fix issues with zero-length records in migration v2"

[maxdev123]

As reported to xen-devel the patches resolve my issue. I attach patch to include the zero-length patches in qubes-builder-vmm.git (for qubes-builder the patch does apply in /path/to/qubes-builder/qubes-src/vmm-xen ). Consider this patch as experimental/start point.
zero-length-stream-v2.patch.txt

[maxdev123]

As reported to xen-devel the patches resolve my issue. I attach patch to include the zero-length patches in qubes-builder-vmm.git (for qubes-builder the patch does apply in /path/to/qubes-builder/qubes-src/vmm-xen ). Consider this patch as experimental/start point.
zero-length-stream-v2.patch.txt

[marmarek]

I'd wait for patch being committed to xen-unstable, then backport it to our version.

[marmarek]

I'd wait for patch being committed to xen-unstable, then backport it to our version.

[andrewdavidwong]

(Took my best guess at a more descriptive issue title.)

[andrewdavidwong]

(Took my best guess at a more descriptive issue title.)