etc/dists/scripts/etcnet-{add,del}_ip.sh: IPv6 supportThis adds IPv6 support for etcnet-based containers (i.e. ALT Linux).
https://bugzilla.openvz.org/2482
[kir@openvz.org: rm empty lines added by original patch]
etc/dists/scripts/postcreate.sh: Fedora 18 supportFor Fedora 18+, we need to set the following caps:
/usr/sbin/arping = cap_net_raw+ep
/usr/sbin/clockdiff = cap_net_raw+ep
/usr/bin/systemd-detect-virt = cap_dac_override,cap_sys_ptrace+ep
ALSO NOTE that in Fedora 17+ ping comes with CAP_NET_ADMIN which is
needed to use ping -m being non-root. But since CAP_NET_ADMIN is not
granted to a container by default, ping with this capability set
will ...
vzlist: warn/skip invalid CTIDs in ve.conf filesAs original bug reporter puts in, using a CTID of 12 digits in ve.conf
file name results in strange-looking error:
# mv /etc/vz/conf/2013020601.conf /etc/vz/conf/201302060101.conf ; mv
# /vz/private/2013020601 /vz/private/201302060101; mv
# /vz/root/2013020601 /vz/root/201302060101
# vzlist -a
Unable to open /etc/vz/conf/-561402811.conf: No such file or directory
CTID NPROC STATUS IP_ADD...
vzlist: more strict for cmdline-supplied CTIDsFirst, move the veid check out of 'if' statement, similar to the
previous commit aa84f2f for vzctl.
Second, check for VEID_MAX as well to filter out invalid CTIDs as early
as possible.
Related to https://bugzilla.openvz.org/2514
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl: fix check for VEID_MAXCommit af42bbd introduced check for CTID < VEID_MAX.
Apparently this check was only performed if CTID supplied was a name,
i.e. after name -> CTID conversion.
Fix it by moving the check out of if statement.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzevent: do umount/mount when rebooting CTApparently in some cases (like NFS mounts invoked by vzctl mount
scripts) vzevent should do umount/mount when restarting a container.
In general, this is the right thing to do anyway, and this is what vzctl
restart does. So, let's add umount to vzevent-reboot.
Also, since there is a hack to do reboot of Fedora 17 container in
vzevent-stop script, modify it accordingly to do umount before check...
vzlist: fix segfault on ploop CT with no DISKINODES setIf a ploop-based CT have no DISKINODES set in its config file
(which is perfectly legal), vzlist -o diskspace (or vzlist -j)
segfaults:
# vzlist 1016 -o diskspace
Segmentation fault (core dumped)
# vzlist -j 1016
Segmentation fault (core dumped)
gdb copy-paste:
Program received signal SIGSEGV, Segmentation fault.
get_ve_ploop_info (ve=0x64b060) at vzlist.c:1471
1471 ve->quota->disks...
vzctl snapshot-list: fix processing first optionFirst option to snapshot-list were ignored, because getopt_long() starts
processing options from argv[1], assuming argv[0] is binary name.
To fix, do not increment argv (and decrement argc) when calling
run_action(); instead, do this inside run_action when needed.
This way if we would add something that uses getopt() we'd be fine.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
In-container devices: add support for Fedora 18Fedora 18's systemd employs a new scheme for static devices,
this patch adds support for it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Factor out create_static_dev() out of dev_create()Very similar code is used in mk_quota_dev(), so let's factor it out.
The next patch will make use of it outside of dev.c scope.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl set --ipadd/ipdel: support for Fedora 18 CTFirst, in F18 ifconfig is not included by default, so we (finally) use
ip instead.
Second, it doesn't like having 127.0.0.1 on venet0, the simplest way to
make it like it is to have ARPCHECK="no" in ifcfg-venet0 file.
Slightly tested here and there, looks like it works.
Heavily based on patch by Konstantin Volkov <wolf@parallels.com>,
with all the bugs added by me.
Signed-off-by: Kir Kolysh...
vzctl snapshot-list: do not return 11. Introduce VZCTL_E_SNAPSHOT_LIST exit code,
use it and document it in vzctl(8).
2. Use more generic exit codes where possible in
vzctl_env_snapshot_list() and friends.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl snapshot-list: check if snapshot is supportedRather than presenting empty snapshot list, bail out with an error,
or crash, check that
1 the container is ploop-based
2 ploop is supported
Crash would happen if ploop library is not avaialble, because
vzctl_env_snapshot_list() now calls ploop functions.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
scripts/.gitignore: add vps-netns_dev_{add,del}Commit c6f7199 adds these two, and they are generated from .in
files so should be mentioned in .gitignore
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl.spec: add buildroot rm to %installAs per
http://fedoraproject.org/wiki/EPEL/GuidelinesAndPolicies#Prepping_BuildRoot_For_.25install
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl snapshot-list: make it workingCommit 9a4a853 added a new implementation of snapshot-list, with
command-line options processed directly by vzctl_env_snapshot_list().
For that to work, we need to skip options processing in
parse_action_opt(), this is what this patch does.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
man/*: correct path to scriptsCommit d3c8b95 moved scripts from /usr/lib[64] to /usr/libexec.
Fix man pages accordingly.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
init.d/vz-redhat: stop vz earlierAs it nicely laid out by the bug reporter:
> I noticed after rebooting one of our hardware nodes the other day that
> all of the 'sshd' processes in the VPSes had disappeared. It turns out
> this is related to the new "suspend/restore" feature added in vzctl 4.1,
> and also to the /etc/init.d/sshd script shipped in CentOS.
>
> The sshd init script in CentOS contains the following in its "stop"...
init.d/vz-gentoo: don't call tools by absolute pathApparently in recent Gentoo the ip tool was moved from /sbin to /bin,
so this initscript stopped working.
Apparently runscript makes sure all the needed directories are in PATH
(I have tested with runscript from baselayout-1.12.11.1, pretty old one)
so we can just drop the /sbin/ prefix.
While at it, do it not only for ip, but also for modprobe and sysctl.
http://bugzilla.openvz.org/2477
htt...
vzctl snapshot-list: re-implementThis is a new implementation of snapshot-list, with configurable
fields etc.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
util.c: introduce vzctl_get_normalized_guid()The code was already there in conf_parse_guid(), so just
extract it and move to util.c.
To be used by a following patch.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl: do now warn about VE_STOP_MODEvzctl --verbose usually complains about unknown parameters in global
configuration file, which sometimes help to discover issues like
typos in parameter names. Like this:
> Warning at /etc/vz/vz.conf:6: unknown parameter VE_STOP_MODE ("stop"),
> ignored
To prevent false positives, there is a list of well-known parameters
which are not used by vzctl itself (see more details in commit de8b7f6)....
Fix extra arguments parsing by add-on modulesModule argument behavior changed with commit 772715b1, prior to this
any arguments which were not defined in set_opt[] was passed to
mod_parse(). After this commit the entire list of available options
is passed down to vps_parse_opt() which then attempts to parse them
all with parse() including additional arguments introduced by the
module.
Fix this by passing the internally available argument...
load_ploop_lib(): prevent buffer overflowvzctl can be compiled with an old version of ploop library headers,
so size of struct ploop_functions might be smaller than the one of
the currently installed library, causing a buffer overflow in
ploop_resolve_functions() leading to all sorts of funny effects.
While it's indeed an issue with ploop_resolve_functions() design,
let's workaround it here by providing a safety bag.
Signed-off-by: ...
vz_chroot(): fix off-by-one in sigaction() argumentAccording to bits/signum.h, NSIG is the biggest signal number + 1,
so fix the for() loop accordingly.
Found while using valgrind:
==775013== Warning: bad signal number 65 in sigaction()
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzubc: fix -w/-c checkArguments -w|--watch and -c|--color can't be used together, so there
is a check for that. The problem is if -c stays before -w on the command
line, this check wouldn't work.
So move the check to after argument parsing.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzubc: remove check for watch presenseFirst, it was only performed for -w|--watch and was not performed for
-wd and -wt. Second, "watch not found" error will be found and printed
later by exec anyway, so why bother?
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
etc/init.d: move VZREBOOTDIR to vz-redhatVZREBOOT declaration should be right here in vz-redhat script
not in initd-functions.
initd-functions is a compatibility layer for Debian and Red Hat
systems, it contains everything that is different between Deb and RH,
such as VARLOCK location.
This is a fix to commit af5bf43.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
etc/init.d/vz-gentoo: fix missing VZREBOOTDIRThat was pretty lame from my side. Fix for commit af5bf43.
http://bugzilla.openvz.org/2467
Reported-by: Andrei Vinogradov <spamslepnoga@inbox.ru>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
configure: fix 'arch support openvz kernel' test caseThe file vzsyscalls.h has its constant definitions enclosed within
VZ_KERNEL_SUPPORTED. This has the desired effect of not including any
of those definitions - including the #error condition on non-suported
arches - during normal compilation.
When we compile test vzsyscalls.h, however, that constant is not yet
defined and we'll skip it all. As a side effect, the test will always
succeed.
This...
etc/init.d/vz*: clean up VZREBOOTDIRFirst, it makes sense to clean up VZREBOOTDIR just before saving
the list of now-running CTs in stop_ve()
Second, since the previous commit we are not starting CTs having
ONBOOT=no, so they are not cleared up from VZREBOOTDIR. So, clean
up VZREBOOTDIR right after getting the list.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
init.d/vz*: do not auto-restart CTs with ONBOOT=noSince vzctl 4.1 we have a feature to save the list of running containers
on init.d/vz stop, and then start those containers on init.d/vz start
(in addition to ones having explicit ONBOOT=yes).
This feature, while being great per se, creates a controversial behavior:
if a container is specifically marked with ONBOOT=no but it was running
before a reboot, it is started nevertheless.
The solutio...
vzlist: show if onboot field is unsetField onboot was just yes/no (or true/false in json output), but
internally it can be in one of three states:
0 NONE (i.e. unset)
1 YES
2 NO
This patchs adds vzlist support to display the 'unset' state,
as either '-' (in conventional output) or 'null' (in json output).
This functionality is needed by the following patch.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
allow vzsyscalls.h to always be called.To avoid spreading ifdefs all over, we would like to keep
the include of vzsyscalls.h unconditional. So after we already
defined VZ_KERNEL_SUPPORTED, we can test for that.
We also tweak the message a bit to make it more descriptive.
Signed-off-by: Glauber Costa <glommer@parallels.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
don't support vz-specific cpu functions if vz kernel is not presentWe can't call any of the vz ioctls if we are not running in a
vz-compatible kernel. So if it is compile-time disabled, don't
compile them in.
We could do it in headers, but those functions will have upstream
similars, so we need them to be callable function objects, that will
just return a meaningful error in this case.
[kir@openvz.org: add comments to #else/#endif]
Signed-off-by: Glauber Co...
provide a upstream alternative for ub functionsThe ub functions will not be ever supported in upstream kernel.
Still, they are called in a quite widespread fashion throughout the
code.
One possible course of action for this, is to skip the compilation
of ub.c, and at the same time provide an empty set of inline functions
in ub.h
[kir@openvz.org: remove #error, add comments #else/#endif */
Signed-off-by: Glauber Costa <glommer@parallels.com>
vz_open(): skip vz kernel init if not availableThe vz-kernel is not available in all architectures where mainline
Linux can run. If we detect this to be the case, the code won't be
compiled.
Signed-off-by: Glauber Costa <glommer@parallels.com>
configure: test for vz-kernel support in the target architectureThe include file "vzsyscalls.h" have a comprehensive list of vz-syscalls
and which architectures support them. If that file is included in a not
supported architecture, it will trigger an #error, which is a very
reliable way to determine that vz-specific code should not be compiled
in at all.
Since this places some new infrastructure in place for presence of
OpenVZ kernel support, this comes w...
Kir Kolyshkin
