Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Ian Campbell]

Adding some CC's, including the devel list.

On Fri, 2015-01-09 at 19:49 -0600, Doug McMillan wrote:
> 
> 
> configuration(?)
> I compiled booted straight from bios with xen.efi during boot I received several errors.
> xl info works (see attachment). XL list locks the terminal session. Checking into the errors I 
> found the following under dmesg (full attached):
> 
> 
>   Ignoring BGRT: invalid status 0 (expected 1) 
>   ACPI: SCI (ACPI GSI 9) not registered
>   kvm: no hardware support
>   mce: Unable to init device /dev/mcelog (rc: -16)
> 
> 
>  systemd[1]: Set hostname to <archxen>.
>  systemd[1]: var-lib-xenstored.mount's Where= setting doesn't match unit name. Refusing.

I suppose this error message is accurate and the Where field is not
"/var/lib/xenstored"? What is it actually? What options did you pass
to ./configure when building Xen?

This seems like a silly misfeature of systemd to me, but I suppose the
fix is to rename the thing to match, except that would require all the
dependent units to have the field changed too. It might turn out to be
easier to instead arrange for @XEN_LIB_STORED@ to == /var/lib/xenstored.

@devs -- we obviously need to do something about this (too late for 4.5,
but for 4.6 + backport). Perhaps there is some alternative systemd
construction which disassociates the actual path from the abstract
service "xenstored dir mounted"?

Otherwise we'll have to somehow arrange for the file and everything
which depends on it to have a name which reflects the actual mount path,
which sounds pretty awful to me...

Ian.

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Wei Liu]

On Mon, Jan 12, 2015 at 02:58:53PM +0000, Ian Campbell wrote:
> Adding some CC's, including the devel list.
> 
> On Fri, 2015-01-09 at 19:49 -0600, Doug McMillan wrote:
> > 
> > 
> > configuration(?)
> > I compiled booted straight from bios with xen.efi during boot I received several errors.
> > xl info works (see attachment). XL list locks the terminal session. Checking into the errors I 
> > found the following under dmesg (full attached):
> > 
> > 
> >   Ignoring BGRT: invalid status 0 (expected 1) 
> >   ACPI: SCI (ACPI GSI 9) not registered
> >   kvm: no hardware support
> >   mce: Unable to init device /dev/mcelog (rc: -16)
> > 
> > 
> >  systemd[1]: Set hostname to <archxen>.
> >  systemd[1]: var-lib-xenstored.mount's Where= setting doesn't match unit name. Refusing.
> 
> I suppose this error message is accurate and the Where field is not
> "/var/lib/xenstored"? What is it actually? What options did you pass
> to ./configure when building Xen?
> 
> This seems like a silly misfeature of systemd to me, but I suppose the
> fix is to rename the thing to match, except that would require all the
> dependent units to have the field changed too. It might turn out to be
> easier to instead arrange for @XEN_LIB_STORED@ to == /var/lib/xenstored.
> 
> @devs -- we obviously need to do something about this (too late for 4.5,
> but for 4.6 + backport). Perhaps there is some alternative systemd
> construction which disassociates the actual path from the abstract
> service "xenstored dir mounted"?
> 

I check its manual with regard to Where=:

  Takes an absolute path of a directory of the mount point. If the mount
  point does not exist at the time of mounting, it is created. This string
  must be reflected in the unit filename. (See above.) This option is
  mandatory.

There doesn't seem to be an alternative. I'm happy if someone can
correct me though...

> Otherwise we'll have to somehow arrange for the file and everything
> which depends on it to have a name which reflects the actual mount path,
> which sounds pretty awful to me...
> 

Agreed...

Wei.

> Ian.
> 

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Olaf Hering]

On Mon, Jan 12, Ian Campbell wrote:

> Adding some CC's, including the devel list.
> 
> On Fri, 2015-01-09 at 19:49 -0600, Doug McMillan wrote:
> > configuration(?)
> > I compiled booted straight from bios with xen.efi during boot I received several errors.
> > xl info works (see attachment). XL list locks the terminal session. Checking into the errors I 
> > found the following under dmesg (full attached):
> > 
> >   Ignoring BGRT: invalid status 0 (expected 1) 
> >   ACPI: SCI (ACPI GSI 9) not registered
> >   kvm: no hardware support
> >   mce: Unable to init device /dev/mcelog (rc: -16)
> > 
> >  systemd[1]: Set hostname to <archxen>.
> >  systemd[1]: var-lib-xenstored.mount's Where= setting doesn't match unit name. Refusing.
> 
> I suppose this error message is accurate and the Where field is not
> "/var/lib/xenstored"? What is it actually? What options did you pass
> to ./configure when building Xen?

The report states "/run/run/xenstored"? Please tell how you did
run "configure".

Olaf

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Doug McMillan]

[-- Attachment #1.1: Type: text/plain, Size: 8858 bytes --]

> Depending on which behaviour you are after you should drop either> EFI_DIR or EFI_VENDOR from your cfg I think. I think the intention is
> that people only normally specify EFI_VENDOR.
> 
> Ian.
Ian,  Thank you I did not know that. I will add information as comment in PKGBUILD.I had an error in a previous xen ver if I didn't put i the EFI_VENDOR. I saw in docs that EFI_DIR let you put the efi directly were it was to go. I will comment out EFI_VENDOR.as soon as I get the PKGBUILD 100% stable then I will likely reformat so my /boot is FAT32 so that the kernels will be updates properly. My current test box is setup withsda1 as fat32, sda2 & xen-dom0 as ext4. I am booting off of /boot/efi/EFI/arch_grubso when I had things wrong and xen.efi failed it fell back to grub2 and booted xen.
# lsblksda            8:0    0   477G  0 disk   sda1         8:1    0   512M  0 part /boot/efi  sda2         8:2    0   244M  0 part /boot  sda3         8:3    0 476.2G  0 part   xen-dom0 254:0    0  28.2G  0 lvm  /  xen-win8 254:1    0   100G  0 lvm  sdb            8:16   1  29.8G  0 disk   sdb1         8:17   1  29.8G  0 part sr0           11:0    1   3.1G  0 rom  
The xen.conf had all of these in it. Is there something in my parameters(below) that isalso causing all these to be populated?
#xen-evtchn builtin kernel#xen-gntdev builtin kernel# xen-gntalloc builtin kernelxen-blkback#xen-netback builtin kernel#xen-pciback builtin kernel#evtchn no module#gntdev no module#netbk no module#blkbk no module#xen-scsibk no module#usbbk no module#pciback no modulexen-acpi-processor#blktap2 no module#blktap no module



> The report states "/run/run/xenstored"? Please tell how you did
> run "configure".
> 
> Olaf

All,  Since what you all talked about I am really hoping I configured something incorrectly here as in the EFI_VENDOR case.My bad sorry I didn't think that most probably aren't familiar with Arch on top of which I didn't point that I was trying to make  an arch package so the parameters. compile configuration,  the dependencies etc were in the attached PKGBUILD.Here are the parameters   export xen_export_check="true"   export AS86=/usr/bin/as86   export BASH=/usr/bin/bash      export BCC=/usr/bin/bcc   export BISON=/usr/bin/bison   export CMAKE=/usr/bin/cmake   export COMPILER_PATH=/usr/bin   export CURL=/usr/bin/curl-config   export EFI_VENDOR=arch   export EFI_DIR=/boot/efi/EFI/arch_grub    export BOOT_DIR=/boot/   export FIG2DEV=/usr/bin/fig2dev   export FLEX=/usr/bin/flex   export IASL=/usr/bin/iasl   export LD86=/usr/bin/ld86   export LD_EFI=/usr/x86_64-w64-mingw32/bin/ld   export MARKDOWN=/usr/bin/markdown   export PANDOC=/usr/bin/pandoc   export PERL=/usr/bin/perl   export PKG_CONFIG=/usr/bin/pkg-config   export POD2HTML=/usr/bin/core_perl/pod2html   export POD2MAN=/usr/bin/core_perl/pod2man   export POD2TEXT=/usr/bin/core_perl/pod2text   export PYTHON="/usr/bin/python2"   export XML=/usr/bin/xml2-config   export XGETTEXT=/usr/bin/xgettext
   # Things that should not be exported that will break this git build   #export ZLIB_URL="http://xenbits.xen.org/xen-extfiles/zlib-1.2.3.tar.gz"   #export LIBPCI_URL="http://xenbits.xen.org/xen-extfiles/pciutils-2.2.9.tar.bz2"   #export NEWLIB_URL="http://xenbits.xen.org/xen-extfiles/newlib-1.16.0.tar.gz"   #export LWIP_URL=" http://xenbits.xen.org/xen-extfiles/lwip-1.3.0.tar.gz"   #export GRUB_URL="http://xenbits.xen.org/xen-extfiles/grub-0.97.tar.gz"   #export OCAML_URL="http://caml.inria.fr/ocaml/pub/distrib/ocaml-4.02/ocaml-4.02.1.tar.gz"   #export GMP_URL="http://xenbits.xen.org/xen-extfiles/gmp-4.3.2.tar.bz2"   #export POLARSSL_URL="http://xenbits.xen.org/xen-extfiles/polarssl-1.1.4-gpl.tgz"   #export TPMEMU_URL="http://xenbits.xen.org/xen-extfiles/tpm_emulator-0.7.4.tar.gz"

  msg2 "Starting make..."    ./autogen.sh    ./configure                                          \       --prefix=/usr                                     \       --libdir=/usr/lib/                                \       --bindir=/usr/bin/                                \       --sbindir=/usr/bin/                               \       --enable-rpath                                    \       --localstatedir=/run                              \       --with-sysconfig-leaf-dir="default"               \       --with-system-ovmf="/usr/share/ovmf"              \       --with-xenstored=oxenstored                       \       --enable-systemd                                  \       --with-systemd="/usr/lib/systemd/system"          \       --with-systemd-modules-load="/etc/modules-load.d" \       --enable-stubdom#     --with-initddir="/etc/xen/scripts" \#     --enable-blktap2 \ #     --with-linux-backend-modules="xen-evtchn xen-gntdev xen-gntalloc xen-blkback xen-netback xen-pciback xen-acpi-processor" # --enable-caml-stubdom  ### need to check this -> --with-initddir="/etc/xen/scripts"
I manually created /run/lib/xenstored when I rebooted it was goneI noticed for some reason in my email I cut the bottom off the error message so here is the complete message plus some additional systemctl info
[root@archxen ~]# systemctl status var-lib-xenstored.mount● var-lib-xenstored.mount - mount xenstore file system   Loaded: error (Reason: Invalid argument)   Active: inactive (dead)    Where: /run/lib/xenstored     What: xenstore
Jan 11 12:37:20 archxen systemd[1]: var-lib-xenstored.mount's Where= setting doesn't match unit name. Refusing.Jan 11 12:37:20 archxen systemd[1]: var-lib-xenstored.mount's Where= setting doesn't match unit name. Refusing.Jan 11 12:37:22 archxen systemd[1]: var-lib-xenstored.mount's Where= setting doesn't match unit name. Refusing.



[root@archxen ~]# systemctl status xenstored.service● xenstored.service - The Xen xenstore   Loaded: loaded (/usr/lib/systemd/system/xenstored.service; disabled; vendor preset: disabled)   Active: failed (Result: exit-code) since Sun 2015-01-11 12:21:57 CST; 1min 57s ago  Process: 325 ExecStart=/bin/sh -c exec $XENSTORED --no-fork $XENSTORED_ARGS (code=exited, status=2)  Process: 322 ExecStartPre=/bin/mkdir -p /run/run/xen (code=exited, status=0/SUCCESS)  Process: 319 ExecStartPre=/bin/rm -f /run/lib/xenstored/tdb* (code=exited, status=0/SUCCESS)  Process: 311 ExecStartPre=/bin/grep -q control_d /proc/xen/capabilities (code=exited, status=0/SUCCESS) Main PID: 325 (code=exited, status=2)   Status: "Mismatch on number (2): Invalid request descriptor"    Error: 53 (Invalid request descriptor)
Jan 11 12:21:57 archxen sh[325]: Expected 2 fds but given 1Jan 11 12:21:57 archxen sh[325]: Fatal error: exception Failure("ocaml_sd_listen_fds() mismatch")Jan 11 12:21:57 archxen systemd[1]: xenstored.service: main process exited, code=exited, status=2/INVALIDARGUMENTJan 11 12:21:57 archxen systemd[1]: Failed to start The Xen xenstore.Jan 11 12:21:57 archxen systemd[1]: Unit xenstored.service entered failed state.Jan 11 12:21:57 archxen systemd[1]: xenstored.service failed.


journalctl _PID=325-- Logs begin at Sat 2014-12-20 04:39:45 CST, end at Mon 2015-01-12 12:37:28 CST. --Dec 25 18:57:51 localhost systemd-fsck[325]: /dev/sda2: clean, 354/62496 files, 78342/249856 blocks-- Reboot --Dec 26 13:39:36 localhost sh[325]: Expected 2 fds but given 1Dec 26 13:39:36 localhost sh[325]: Fatal error: exception Failure("ocaml_sd_listen_fds() mismatch")-- Reboot --Dec 26 14:22:54 localhost dhcpcd[325]: version 6.6.7 startingDec 26 14:22:54 localhost dhcpcd[325]: enp5s0: adding address fe80::42d8:d41e:f5f5:d2e9Dec 26 14:22:54 localhost dhcpcd[325]: enp5s0: waiting for carrierDec 26 14:22:57 localhost dhcpcd[325]: enp5s0: carrier acquiredDec 26 14:22:57 localhost dhcpcd[325]: DUID 00:01:00:01:1c:28:0e:72:d0:50:99:27:72:4eDec 26 14:22:57 localhost dhcpcd[325]: enp5s0: IAID 99:27:72:45Dec 26 14:22:57 localhost dhcpcd[325]: enp5s0: rebinding lease of 192.168.1.234Dec 26 14:22:58 localhost dhcpcd[325]: enp5s0: soliciting an IPv6 routerDec 26 14:23:02 localhost dhcpcd[325]: enp5s0: leased 192.168.1.234 for 86400 secondsDec 26 14:23:02 localhost dhcpcd[325]: enp5s0: adding route to 192.168.1.0/24Dec 26 14:23:02 localhost dhcpcd[325]: enp5s0: adding default route via 192.168.1.1Dec 26 14:23:02 localhost dhcpcd[325]: forked to background, child pid 494-- Reboot --Jan 09 17:53:49 archxen sh[325]: Fatal error: exception Failure("ocaml_sd_listen_fds() failed to get any sockets"-- Reboot --Jan 11 11:35:55 archxen sh[325]: Fatal error: exception Failure("ocaml_sd_listen_fds() failed to get any sockets"-- Reboot --Jan 11 12:15:02 archxen sh[325]: Expected 2 fds but given 1Jan 11 12:15:02 archxen sh[325]: Fatal error: exception Failure("ocaml_sd_listen_fds() mismatch")-- Reboot --Jan 11 12:21:57 archxen sh[325]: Expected 2 fds but given 1Jan 11 12:21:57 archxen sh[325]: Fatal error: exception Failure("ocaml_sd_listen_fds() mismatch")


 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 14716 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Olaf Hering]

On Mon, Jan 12, Doug McMillan wrote:

>     ./configure                                          \
>        --localstatedir=/run                              \

This is wrong for every setup one can think of. It has to be /var.

Olaf

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Olaf Hering]

On Mon, Jan 12, Ian Campbell wrote:

> @devs -- we obviously need to do something about this (too late for 4.5,
> but for 4.6 + backport). Perhaps there is some alternative systemd
> construction which disassociates the actual path from the abstract
> service "xenstored dir mounted"?

I dont think we can do anything about this systemd brain damage. Either
it gets its Where= from such line within the file, or it gets its Where=
from the filename. In which case it has to stop looking at a Where=
line.

In any case, its wrong to use --localstatedir=/tmpfs-mount-point because
that means all mails in the spool subdirectory are in danger. If thats
the mindset of ArchLinux all we can do is to recommend to stop using it
for any serious task.

Olaf

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Ian Campbell]

On Tue, 2015-01-13 at 09:45 +0100, Olaf Hering wrote:
> On Mon, Jan 12, Ian Campbell wrote:
> 
> > @devs -- we obviously need to do something about this (too late for 4.5,
> > but for 4.6 + backport). Perhaps there is some alternative systemd
> > construction which disassociates the actual path from the abstract
> > service "xenstored dir mounted"?
> 
> I dont think we can do anything about this systemd brain damage. Either
> it gets its Where= from such line within the file, or it gets its Where=
> from the filename. In which case it has to stop looking at a Where=
> line.

We either need to make XEN_LIB_STORED be hardcoded to /var/lib/xenstored
(i.e. ignoring localstatedir) or we need to make the systemd units
reflect the --localstatedir the user provided or we need to remove
--localstatedir as a user selectable option.

(the first is probably easiest, the second is probably more correct and
the third seems like overkill to me)

> In any case, its wrong to use --localstatedir=/tmpfs-mount-point because
> that means all mails in the spool subdirectory are in danger. If thats
> the mindset of ArchLinux all we can do is to recommend to stop using it
> for any serious task.

Xen doesn't include any sort of MUA, so that seems like hyperbole to me.

Not that I would recommend configuring Xen with localstatedir on /run
either mind.

Ian.

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Ian Jackson]

Ian Campbell writes ("Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux"):
> On Tue, 2015-01-13 at 09:45 +0100, Olaf Hering wrote:
> > I dont think we can do anything about this systemd brain damage. Either
> > it gets its Where= from such line within the file, or it gets its Where=
> > from the filename. In which case it has to stop looking at a Where=
> > line.
> 
> We either need to make XEN_LIB_STORED be hardcoded to /var/lib/xenstored
> (i.e. ignoring localstatedir) or we need to make the systemd units
> reflect the --localstatedir the user provided or we need to remove
> --localstatedir as a user selectable option.
> 
> (the first is probably easiest, the second is probably more correct and
> the third seems like overkill to me)

The first and third are imposing systemd-originated brokenness on
non-systemd users and are therefore not really acceptable.

There is a fourth option: we arrange that if the user specifies
--localstatedir it still breaks with systemd, but perhaps that they
get an error message earlier.  That confines the systemd breakage to
systemd users.

Ian.

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Ian Campbell]

On Tue, 2015-01-13 at 15:11 +0000, Ian Jackson wrote:
> Ian Campbell writes ("Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux"):
> > On Tue, 2015-01-13 at 09:45 +0100, Olaf Hering wrote:
> > > I dont think we can do anything about this systemd brain damage. Either
> > > it gets its Where= from such line within the file, or it gets its Where=
> > > from the filename. In which case it has to stop looking at a Where=
> > > line.
> > 
> > We either need to make XEN_LIB_STORED be hardcoded to /var/lib/xenstored
> > (i.e. ignoring localstatedir) or we need to make the systemd units
> > reflect the --localstatedir the user provided or we need to remove
> > --localstatedir as a user selectable option.
> > 
> > (the first is probably easiest, the second is probably more correct and
> > the third seems like overkill to me)
> 
> The first and third are imposing systemd-originated brokenness on
> non-systemd users and are therefore not really acceptable.

Given that I think my preference would be for my second option.

> There is a fourth option: we arrange that if the user specifies
> --localstatedir it still breaks with systemd, but perhaps that they
> get an error message earlier.  That confines the systemd breakage to
> systemd users.

Or Five: hardcode /var/lib/xenstored for systemd only (i.e. ignore
XEN_LIB_STORED).

Ian.

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Doug McMillan]

[-- Attachment #1.1: Type: text/plain, Size: 2614 bytes --]

> > @devs -- we obviously need to do something about this (too late for 4.5,
> > but for 4.6 + backport). Perhaps there is some alternative systemd
> > construction which disassociates the actual path from the abstract
> > service "xenstored dir mounted"?
> 
> I dont think we can do anything about this systemd brain damage. Either
> it gets its Where= from such line within the file, or it gets its Where=
> from the filename. In which case it has to stop looking at a Where=
> line.
> 
> In any case, its wrong to use --localstatedir=/tmpfs-mount-point because
> that means all mails in the spool subdirectory are in danger. If thats
> the mindset of ArchLinux all we can do is to recommend to stop using it
> for any serious task.
> 
> Olaf

All  My error not ArchLinux's (see the clip below from the official ARCHLinux wikihttps://wiki.archlinux.org/index.php/arch_filesystem_hierarchy). I am not sure where I got the --localstatedir=/run. Not all the info I dig up by using net searches is correct.I do guarantee that I will double check my info better in the future. I put --localstatedir=/var in the PKGBUILD configure even though it defaults to that,  so the info is in front of anyone changing the PKGBUILD. 2.12 /run: Ephemeral runtime data..........................cut............
2.18 /var: Variable files2.18.1 /var/abs2.18.2 /var/cache/pacman/pkg2.18.3 /var/lib: State information
If I can get misinformed so can someone else. Could I suggest that a warning in the configure and/or install docs thatin 4.5 changing from the defaults /var breaks the systemd mounts that xenstore requires?? 
Also quick question if I am understanding my remaining issue [tmpfs: Bad mount option context] as described by a previous thread. Untilthe code that generates it changes I need to manually change var-lib-xenstored.mount from
   [Unit]   Description=mount xenstore file system
   Requires=proc-xen.mount
   After=proc-xen.mount
   ConditionPathExists=/proc/xen/capabilities
   RefuseManualStop=true
   [Mount]
   Environment=XENSTORED_MOUNT_CTX=none
   EnvironmentFile=-/etc/default/xenstored
   What=xenstore
   Where=/var/lib/xenstored
   Type=tmpfs
   Options=mode=755,context="$XENSTORED_MOUNT_CTX"
to
   [Unit]
   Description=mount xenstore file system
   Requires=proc-xen.mount
   After=proc-xen.mount
   ConditionPathExists=/proc/xen/capabilities
   RefuseManualStop=true
   [Mount]
   EnvironmentFile=-/etc/default/xencommons
   What=xenstore
   Where=/var/lib/xenstored
   Type=tmpfs
   Options=mode=755
Or am I misunderstanding that also??
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 7768 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [Xen-users] [TestDay] minor bug + possible configuration bug 4.5rc4 archlinux

[Olaf Hering]

On Tue, Jan 13, Doug McMillan wrote:

> Also quick question if I am understanding my remaining issue [tmpfs: Bad mount
> option context] as described by a previous thread. Until
> the code that generates it changes I need to manually change
> var-lib-xenstored.mount from
...
> Or am I misunderstanding that also??

No, thats fine. The contex= is already removed for 4.5.0.

Olaf