XCP-ng 8.0.0 Release Candidate

bnrstnr

@bnrstnr said in XCP-ng 8.0.0 Beta now available!:

HP DL360 v7 with (2) Intel E5649 working good so far. Just finished installation and yum updates and everything came right up, as expected.

No issues in the 2 weeks since upgrading to beta and all updates between.

stormi

I made new ISOs meant to fix the unattended installation (either remote upgrade method or installation with answer file). That's the only change so I've not made it an official RC2, just some tests ISOs for the community to confirm that the bug is fixed, so that we are sure that it will not be in the final ISOs.

https://updates.xcp-ng.org/tmp/RC1custom1/

bvitnik

@stormi So is install.img the only thing changed in ISO or is there some changes to rpm packages?

stormi

@bvitnik no other change

bvitnik

Just to confirm. After copying new install.img to my PXE environment, unattended installation went smoothly with my answer file.

achim71

Tried the new ISO here as well and so far it is working with the answerfile now.
Is there an sytnax extension to configure an md root mirror in the answerfile in version 8 ?

stormi

@achim71 I don't think so. See https://github.com/xcp-ng/host-installer/tree/master/doc

achim71

@stormi Thanks for the link
Meanwhile i tried version 8 on an older HP G7 Microserver. The addon ipmi card does no longer work here. It requires passing module parameters to the ipmi_si module (https://gist.github.com/joshenders/9065698) and worked till now.
With the options type=kcs ports=0xca2 there is an kernel segfault now on module load.
Seems to be an issue with newer kernels.
https://bugzilla.redhat.com/show_bug.cgi?id=1551285
https://bugs.archlinux.org/task/57429

stormi

@achim71 Could you add those details in https://github.com/xcp-ng/xcp/wiki/Hardware-Compatibility-List-(HCL) ?

achim71

@stormi I try to find the commit causing the error or the one fixing it. Newer kernels (testes 5.1.3) are working. If it's not too difficult i'll try to backport the modification to 4.19.

achim71

Tracked down the necessary patch to make ipmi work again.

https://lkml.org/lkml/2019/2/21/926

*When excuting a command like:
modprobe ipmi_si ports=0xffc0e3 type=bt
The system would get an oops.

The trouble here is that ipmi_si_hardcode_find_bmc() is called before ipmi_si_platform_init(), but initialization of the hard-coded device creates an IPMI platform device, which won't be initialized yet.

The real trouble is that hard-coded devices aren't created with any device, and the fixup is done later. So do it right, create the hard-coded devices as normal platform devices.

This required adding some new resource types to the IPMI platform code for passing information required by the hard-coded device
and adding some code to remove the hard-coded platform devices on module removal.

To enforce the "hard-coded devices passed by the user take priority over firmware devices" rule, some special code was added to check and see if a hard-coded device already exists.*

The patch was backported to 4.19.37. Used the backported version from here. https://github.com/raspberrypi/linux/commit/6bba17f6bce39e46fdf8c0fe190bdc3f57ef8f8f

Once applied to the 4.19.19 sources

modprobe type=kcs ports=0xca2
ipmitool sensor

works. I tried it with an debian usb system with pristen 4.19.19 sources.
Is it woth the effort to prepare the patch for the xcp-ng kernel "kernel-4.19.19-5.0.8.1.xcpng8.0.x86_64.rpm"?
I assume it has to be signed by microsoft to make it work with secure uefi.

0

cminyard committed to raspberrypi/linux

ipmi_si: Fix crash when using hard-coded device

Backport from 41b766d661bf94a364960862cfc248a78313dbd3

When excuting a command like:
  modprobe ipmi_si ports=0xffc0e3 type=bt
The system would get an oops.

The trouble here is that ipmi_si_hardcode_find_bmc() is called before
ipmi_si_platform_init(), but initialization of the hard-coded device
creates an IPMI platform device, which won't be initialized yet.

The real trouble is that hard-coded devices aren't created with
any device, and the fixup is done later.  So do it right, create the
hard-coded devices as normal platform devices.

This required adding some new resource types to the IPMI platform
code for passing information required by the hard-coded device
and adding some code to remove the hard-coded platform devices
on module removal.

To enforce the "hard-coded devices passed by the user take priority
over firmware devices" rule, some special code was added to check
and see if a hard-coded device already exists.

The backport required some minor fixups and adding the device
id table that had been added in another change and was used
in this one.

Reported-by: Yang Yingliang <yangyingliang@huawei.com>
Cc: stable@vger.kernel.org # v4.15+
Signed-off-by: Corey Minyard <cminyard@mvista.com>
Tested-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

stormi

Thanks. We could include it in an experimental kernel package indeed. No microsoft signature of our kernel so we're free to change it if we trust the change. You should also create this very detailed bug report on https://github.com/xcp-ng/xcp/issues so that we can track its status there and also on https://bugs.xenserver.org to let XenServer develpers know and maybe get their feedback about what they think of the fix.

stormi

We should also check if that patch brought regressions and subsequent commits fixed that afterwards.

achim71

Did an grep for ipmi: and ipmi_si: over kernel 4.19 Changeslogs:

ChangeLog-4.19.31:    ipmi_si: fix use-after-free of resource->name
ChangeLog-4.19.31:    Fixes: 93c303d2045b ("ipmi_si: Clean up shutdown a bit")
ChangeLog-4.19.33:    ipmi_si: Fix crash when using hard-coded device
ChangeLog-4.19.37:    ipmi: fix sleep-in-atomic in free_user at cleanup SRCU user->release_barrier
ChangeLog-4.19.37:    Fixes: 77f8269606bf ("ipmi: fix use-after-free of user->release_barrier.rda")
ChangeLog-4.19.44:    ipmi: ipmi_si_hardcode.c: init si_type array to fix a crash
ChangeLog-4.19.45:    ipmi:ssif: compare block number correctly for multi-part return messages
ChangeLog-4.19.45:    Fixes: 7d6380cd40f79 ("ipmi:ssif: Fix handling of multi-part return messages").

The patch i mentioned earlier was applied in 4.19.33 afterwards the patch from 4.19.44 looks like an following fix for an related issue.

https://lkml.org/lkml/2019/5/7/250

Yesterday I also tried to copy the whole drivers/char/ipmi folder from 4.19.57 to 4.19.19. The kernel successfully compiled and the ipmi modues also worked. Was just an quick test and needs to be compared to the above list of patches.

r1

Hi @achim71 @stormi

Patching only for 1 issue would leave other issues - just waiting to be found.

Linux kernel gets weekly updates for fixes to stable and longterm. New features get added in new versions. I think at XCP we can follow longterm branch of 4.19 and keep pushing updates of kernels on regular basis.

I have at tracked till 4.19.48 at link and at 2 instances I found that using the next patch on top of existing patches needs certain modifications. ( e.g. 37-pre and 47-pre )

Following Linux longterm in XCP also means going back to CH updates will not be possible. But I assume, CH will anyways take updates from Linux longterm.

stormi

@r1 I think we will provide the latest 4.19.x kernel as an alternate kernel for those who need it and keep the Citrix one by default.

achim71

@stormi and @r1 Having the latest 4.19 kernel available would be great. It will also not require much effort from my side.

Update: @r1 build an rpm for your kernel with the citrix ddk 8.0.0 (and an suitable zfs-kmod package). Installed both on the G7 Microserver and ipmi_si works fine again. So thank you for providing the sources on github!

r1

@achim71 I suppose you meant "built"..

XCP-NG has its own variant of DDK at https://github.com/xcp-ng/xcp-ng-build-env

Edit: replied on wrong topic.

dave.opc

Is there any possibility to make installer nonGUI?
On my Ryzen 2400G can't load, it boots up, and after choosing installer the screen is just blank. Tried the same image on VM it starts the installer, but with some graphics.

Addon: ok, i changed 8192M to 2048M in grub.cfg and was able to install the system. Everything is working fine. Now, i got a question: i did yum update and system told me that there is a new kernel with newer xcp-ng:

kernel x86_64 4.19.19-5.0.9.xcpng8.0 xcp-ng-base 30 M
xcp-ng-release x86_64 8.0.0-12 xcp-ng-base 15 k
xcp-ng-release-config x86_64 8.0.0-12 xcp-ng-base 350 k
xenserver-firstboot noarch 1.0.11-1.1.xcpng8.0 xcp-ng-base 20 k

After installing and rebooting, it still shows that kernel is 4.19.0 - any idea how to fix? (not sure if xcp-ng itself updated to 8.0.0-12, in xsconsole it shows as 8.0.0)

stormi

The new kernel has the same version, except that it has been patched for a security issue. All good.

There's no such thing as XCP-ng 8.0.0-12 either. 12 is the release number of the xcp-ng-release RPM package, version is 8.0.0.