Merge branch 'master' into japanese

doc: Add keyring options to Japanese lxc.containers.conf(5)

Signed-off-by: KATOH Yasufumi <karma@jazz.email.ne.jp>

Merge pull request #3288 from brauner/2020-03-11/fixes

commands: simplify lxc_cmd_fd_cleanup()

commands: simplify lxc_cmd_fd_cleanup()

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3287 from brauner/2020-03-11/fixes

fixes

commands_utils: fix command socket hashing

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

af_unix: fix return value

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

start: cleanup file descriptor closing

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3286 from brauner/2020-03-10/fixes

commands: make sure to always close the client fd

commands: make sure to always close the client fd

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3285 from brauner/2020-03-10/fixes

bugfixes

commands: improve state client cleanup

Improves: ebbca8529732 ("commands_utils: fix socket leak when adding state client")
Cc: Matthias Hardt <matthias.hardt@gmail.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

commands: switch to pid_t to send around pid

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3283 from brauner/2020-03-10/fixes

bugfixes

share_ns: improve error handling

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

share_ns: improve error handling

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

file_utils: handle libcs without fmemopen()

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

cgroups: cleanup

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

cgfsng: use __do_free_string_list all over

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

file_utils: include stdio.h for fmemopen()

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

tests/share_ns: always call pthread_exit()

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3282 from brauner/2020-03-10/fixes

memory_utils: remove unneeded inclusion of mntent.h

memory_utils: remove unneeded inclusion of mntent.h

Fixes: Android
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3281 from brauner/2020-03-10/fixes

tree-wide: cleanup

cgroups: fix memory leak and simplify code

Closes #3252.
Reported-by: LiFeng <lifeng68@huawei.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

tests/share_ns: bugfixes

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

conf: cleanup

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

commands_utils: cleanup

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

commands: cleanup

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3279 from brauner/2020-03-04/improvements

bugfixes

tree-wide: more cleanup macros

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

lxccontainer: increase cleanup macro usage

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

autotools: fix lxc-init build with clang-10

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

tree-wide: improve logging

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

tree-wide: make files cloexec whenever possible

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: cleanup various helpers

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: use logging helpers when handling no new privileges

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: use cleanup macros and logging helpers when fetching seccomp

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: use LXC_INVALID_{G,U}ID macros

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: use cleanup macros in lxc_attach_getpwshell()

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: fix fd leak

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

attach: cleanup

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

cgroup2_devices: fix logic error

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

commands: remove unused variables

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3280 from Piscolero/state_client_socket_leak_fix

commands_utils: fix socket leak when adding state client

commands_utils: fix socket leak when adding state client

If lxc_add_state_client() is called with the container already being in the desired state the client fd will never be closed and is leaking. This is due to setting stay_connected in lxc_cmd for LXC_CMD_ADD_STATE_CLIENT. If the desired state isn't already achieved the client fd will later be closed by calling lxc_cmd_fd_cleanup() but in the other case the client configuration isn't added to the handlers
state clients. So the client fd has to be closed explicitely.

This is simply tested by starting container A and calling lxc-wait -n A -s RUNNING.

Signed-off-by: Matthias Hardt <matthias.hardt@gmail.com>

Revert "commands_utils: fix socket leak in when adding state client"

This reverts commit d7aa5552448680c8ff7c4af8c19ea5dbd678e946.

Signed-off-by: Matthias Hardt <matthias.hardt@gmail.com>

Revert "commands_utils: indicate taking ownership of state_client_fd in"

This reverts commit cd0dc360ce740ba302dacc3dc70c6b20b2a5f794.

Signed-off-by: Matthias Hardt <matthias.hardt@gmail.com>

commands_utils: indicate taking ownership of state_client_fd in
lxc_add_state_client()

Signed-off-by: Matthias Hardt <matthias.hardt@gmail.com>

commands_utils: fix socket leak in when adding state client

If lxc_add_state_client() is called with the container already being in
the desired state the client fd will never be closed and is leaking.
This due to setting stay_connected in lxc_cmd for
LXC_CMD_ADD_STATE_CLIENT. If the desired state isn't already achieved
the client fd will later be closed by calling lxc_cmd_fd_cleanup() but
in the other case the client configuration isn't added to the handlers
state clients. So the client fd has to be closed explicitely.

This is simply tested by starting container A and calling lxc-wait -n A
-s RUNNING.

Signed-off-by: Matthias Hardt <matthias.hardt@gmail.com>

af_unix: cleanup

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3278 from tomponline/tp-proxy-sleep

network: Adds short sleep between veth setup and neighbour proxy add

Merge pull request #3277 from tomponline/tp-router-netlink

network: Uses netlink for IP neighbour proxy management

network: Adds short sleep between veth setup and neighbour proxy add

There is an intermittent issue, experienced on at least Ubuntu 18.04 (5.3.0-40-generic) and Alpine 3.11 (5.4.12-1-virt) when using the router network interface type that causes the IP proxy neighbour entries on the host side of the veth pair to not be created.

The `ip neigh add proxy` command returns without an error, however by the time the network up hook has started the IP neighbour proxy entries are no longer there (if they ever were).

I've also tested this using netlink rather than the ip command to add and both are equally affected.

Adding a short sleep between setting up the veth pair and adding the proxy entries appears to fix it.

Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

network: Uses netlink for IP neighbour proxy management

Removes need for ip command when managing IP neighbour proxies.

Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Merge pull request #3275 from brauner/2020-03-05/api_extension

utils: only move_fd() when fdopen() has been successful

utils: only move_fd() when fdopen() has been successful

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3274 from brauner/2020-03-05/api_extension

fixes

api-extensions: document cgroup2_devices and cgroup2 api extensions

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3272 from ffontaine/master

src/lxc/raw_syscalls.c: fix sparc assembly

src/lxc/raw_syscalls.c: fix sparc assembly

Build of lxc 3.2.1 fails with ultrasparc on:

raw_syscalls.c: In function ‘lxc_raw_clone’:
raw_syscalls.c:66:3: error: invalid 'asm': invalid operand output code
   asm volatile(
   ^~~

Issue has been added with commit
b52e8e68a61866da2af86e85905ec850f8a8b7fc which added %g1 instead of %%g1

Fixes:
 - http://autobuild.buildroot.org/results/17c2319850f02f24da6fbef9656c07f86fdc5a3a

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>

Merge pull request #3271 from brauner/2020-02-28/support_cgroup_pattern_2

cgroups: honor lxc.cgroup.pattern if set explicitly II

cgroups: honor lxc.cgroup.pattern if set explicitly II

Link: https://discuss.linuxcontainers.org/t/lxc-cgroup-pattern-is-not-being-honored
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3270 from brauner/2020-02-27/support_cgroup_pattern

cgroups: honor lxc.cgroup.pattern if set explicitly

cgroups: honor lxc.cgroup.pattern if set explicitly

Link: https://discuss.linuxcontainers.org/t/lxc-cgroup-pattern-is-not-being-honored
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3268 from brauner/2020-02-25/cgroups

cgroups: remove unused method and cleanup cgroup_exit()

cgroups: remove unused method and cleanup cgroup_exit()

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3265 from brauner/2020-02-13/fix_setgroups

tree-wide: improve setgroups() dropping

tree-wide: improve setgroups() dropping

Drop groups before we change to userns root.

Reported-by: Teddy Reed <teddy.reed@gmail.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3264 from tych0/fix-leak

lxclock: fix a small memory leak

lxclock: fix a small memory leak

if (!name), we allocate an unnamed semaphore, but if we then fail to
allocate/create the lock, we don't free this semaphore, and we just leak
it.

Signed-off-by: Tycho Andersen <tycho@tycho.ws>

Merge pull request #3261 from blenk92/fix-parse-config

container.conf: Fix parsing of config options:

container.conf: Document that order is important in config_jump_table

Add a comment that documents that more specific options of a
namespace have to be added above more generic options in
config_jump_table.

Signed-off-by: Maximilian Blenk <Maximilian.Blenk@bmw.de>

container.conf: Fix option ordering in config_jump_table

Currently lxc.selinux.context.keyring is placed after
lxc.selinux.context in config_jump_table. This entails that if
lxc.selinux.context.keyring is specified, lxc.selinux.context will be
matched to that option. This commit places the
lxc.selinux.context.keyring before lxc.selinux.context

Signed-off-by: Maximilian Blenk <Maximilian.Blenk@bmw.de>

container.conf: Fix off by 2 in option parsing

This commit fixes a off by 2 in config option parsing
(due to missing parenthesis). The error occurs if for instance
lxc.net.0.type is parsed. In that case, the .0 is removed from the
string. However, due to the missing parenthesis, the null terminating
character is off by two which means the modified config option would
be lxc.net.typepe instead of lxc.net.type.

Signed-off-by: Maximilian Blenk <Maximilian.Blenk@bmw.de>

Merge pull request #3260 from blenk92/add-keyring-option

Add keyring option

doc: Add doc for keyring options

Signed-off-by: Maximilian Blenk <Maximilian.Blenk@bmw.de>

container.conf: Add option to disable session keyring creation

lxc set's up a new session keyring for every container by default.
There might be valid use-cases where this is not wanted / needed
(e.g. systemd by default creates a new session keyring anyway).

Signed-off-by: Maximilian Blenk <Maximilian.Blenk@bmw.de>

container.conf: Add option to set keyring SELinux context

lxc set's up a new session keyring for every container by default.
If executed on an SELinux enabled system, by default, the keyring
inherits the label of the creating process. If executed with the
currently available SELinux policy, this means that the keyring
is labeled with the lxc_t type. Applications inside the container,
however, might expect that the keyring is labeled with a certain
context (and will fail to access the keyring if it's not explicitly
allowed in the global policy). This patch introduces the config
option lxc.selinux.context.keyring which enables to specify the
label of the newly created keyring. That is, the keyring can be
labeled with the label expected by the started application.

Signed-off-by: Maximilian Blenk <Maximilian.Blenk@bmw.de>

Merge pull request #3259 from chobostar/fix_default_cgroup_pattern

cgroups: fix default cgroup pattern

cgroups: fix default cgroup pattern

Signed-off-by: Kirill Petrov <yakutskkirill@mail.ru>

Merge pull request #3255 from brauner/2020-01-16/fix_monitor_hang

start: fix container killing logic

start: fix container killing logic

We need to account for the case where pidfd's are not supported by the kernel
in question.

Closes: #3254
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3253 from tomponline/tp-network-mtu

network: Restore fixed MTU functionality

network: Restore fixed MTU functionality

If MTU setting is provided in network device config then always use it rather than inheriting from the link device.

Affected both bridge and router veth modes.

Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Merge pull request #3251 from brauner/master

test: increase timeout for api reboot tests

Merge pull request #3250 from lifeng68/fix_lxc

cgroup.c: fix memory leak at cgroup init failed

test: increase timeout for api reboot tests

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

cgroup.c: fix memory leak at cgroup init failed

Signed-off-by: LiFeng <lifeng68@huawei.com>

Merge pull request #3248 from brauner/2020-01-05/veth_devices

network: improve veth device creation

network: rework network device creation

This makes all the codepaths easier to follow and also moves and renames
devices in one single step.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

network: fix network device removal

We can't delete by netdev->ifindex since that's the ifindex of the device in
the container, not on the host. The correct thing is done below.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

tests: log api reboot test failures

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

network: fix typ and formatting in comment

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

network: improve veth device creation

This allows us to avoid having to move the network device. It also allows us to
work around a kernel bug that in combination with a recent change in systemd
244 causes uses of systemd-networkd to not get an ip address.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3249 from brauner/2020-01-09/bugfixes

handle kernel version <--> header incompatibility

start: handle kernel header and kernel incompatability

We might e.g. be compiled in a container with old kernel headers. In this
scenario CLONE_PIDFD will work but pidfd_send_signal() might not be detected
because __NR_pidfd_send_signal is not defined because the kernel headers don't
match the kernel version.

This explains and fixes test-suite hangs on Jenkins I've recently debugged.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

tests: timeout after 60 seconds

That should be more than enough to reboot.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

mainloop: add missing \n

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #3241 from brauner/2019-12-25/remove_procfs_pidfd_support

start: remove procfs pidfd support