templates: avoid endless loop

Closes #2512.

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

Merge pull request #2495 from 2xsec/bugfix

add default log priority & cleanups

Merge pull request #2511 from 2xsec/coverity

fix coverity issues

coverity: #1438232

Failure to restore non-local value

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

pam_cgfs: cleanups

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438231

Dereference after null check

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438230

Logically dead code

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438229

Resource leak

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438233

Resource leak

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438234

Resource leak

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438235

Resource leak

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

coverity: #1438236

Resource leak

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

tools: lxc-unshare: apply default log priority

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

log: add default log priority

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

log: function cleanups

Signed-off-by: 2xsec <dh48.jeong@samsung.com>

Merge pull request #2510 from brauner/2018-08-05/cap_fixes

tree-wide: pass unsigned long to prctl()

tree-wide: pass unsigned long to prctl()

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

Merge pull request #2508 from brauner/2018-08-04/cap_fixes

macro: add new macro header, caps: bugfixes, log: bugfixes

log: bugfixes

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

caps: bugfixes

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

macro: add new macro header

This allows us to use a bunch of macros in our static build for init.lxc.static
without having to link against all of utils.{c,h}.

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

travis: export CFLAGS=-O0 for coverity

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

travis: build with -O0 for coverity

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

Merge pull request #2505 from brauner/2018-07-31/bugfixes

coverity: #438136

READEM: update Serge's mail address

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Cc: Serge Hallyn <serge@hallyn.com>

coverity: #438136

String not null terminated

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

Merge pull request #2503 from tenforward/japanese

doc: Add incompatibility with network type=none into Japanese man

doc: Add incompatibility with network type=none into Japanese man

Update for commit e4b3e36

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

Merge pull request #2499 from brauner/lxc/master

bugfixes

Merge pull request #2475 from brauner/2018-07-16/monitor_signal_pdeath

conf: improve rootfs setup

Merge pull request #2502 from brauner/2018-07-27/fix_max_devpts_option

conf: mount devpts without "max" on EINVAL

conf: don't return a clobbered errno value

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

conf: mount devpts without "max" on EINVAL

The "max" option to devpts got introduced in kernel 3.4.

Closes #2490.

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

Merge pull request #2500 from akosiaris/patch-1

Unprivileged's incompatibility with type=none docs

Unprivileged's incompatibility with type=none docs

Unprivileged containers are not compatible with sharing the
host namespace due to an inability to mount sysfs. Add docs
in lxc.container.conf to document that out.

Refs #2463

Signed-off-by: Alexandros Kosiaris <akosiaris@gmail.com>

caps: handle EINTR in read()

We don't want to link caps.{c,h} against utils.{c,h} for the sake of our static
builds init.lxc.static. This means lxc_write_nointr() will not be available. So
handle it EINTR.

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

log: handle EINTR in read()

We don't want to link log.{c,h} against utils.{c,h} for the sake of our static
builds init.lxc.static. This means lxc_write_nointr() will not be available. So
handle it EINTR.

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

utils: s/read()/lxc_read_nointr()/g

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

terminal: s/read()/lxc_read_nointr()/g

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

monitor: s/write()/lxc_write_nointr()/g

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

monitor: s/read()/lxc_read_nointr()/g

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

apparmor: s/read()/lxc_read_nointr()/g

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

tools: s/write()/lxc_write_nointr()/g

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

tools: s/read()/lxc_read_nointr()/g

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

sync: s/write()/lxc_write_nointr()/g

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

sync: s/read()/lxc_read_nointr()/g

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

network: s/write()/lxc_write_nointr()/g

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

network: s/read()/lxc_read_nointr()/g

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

lxccontainer: s/read()/lxc_read_nointr()/g

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

lxccontainer: s/write()/lxc_write_nointr()/g

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

criu: s/write()/lxc_write_nointr()/g

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

criu: s/read()/lxc_read_nointr()/g

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

cmd: s/read()/lxc_read_nointr()/g

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

cmd: s/write()/lxc_write_nointr()/g

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

cmd: s/pipe()/pipe2()/g

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

lxccontainer: s/pipe()/pipe2()/g

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

lxccontainer: cleanup do_lxcapi_get_interfaces()

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

criu: s/pipe()/pipe2()/

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

conf: always close pipe in run_userns_fn()

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

conf: s/pipe()/pipe2()/g

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

Merge pull request #2497 from brauner/2018-07-29/nl_fix

nl: avoid NULL pointer dereference

nl: avoid NULL pointer dereference

It's a valid case to call nla_put() with NULL data and 0 len. It's done e.g. in
the nla_put_attr().

There has to be a check for data in nla_put() as passing NULL to the memcpy()
is not allowed. Even if length is 0, both pointers have to be valid.

For a reference see C99 standard (7.21.1/2), it says: "pointer arguments on
such a call shall still have valid values".

Reported-by: Daniel Gimpelevich <daniel@gimpelevich.san-francisco.ca.us>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
[christian.brauner@ubuntu.com: adapted commit message]
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

confile: split mount options into flags and data

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

conf: improve rootfs setup

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

Merge pull request #2496 from flx42/nvidia-hook-lgpl

Fix license of the nvidia hook

Fix license of the nvidia hook

Fixes: #2494
Signed-off-by: Felix Abecassis <fabecassis@nvidia.com>

Merge pull request #2493 from brauner/2018-07-26/bugfixes

utils: add lxc_iterate_parts(), compile with -Wvla and -std=gnu11

autotools: default to -Wvla -std=gnu11

We can't really support anything less than gcc-4.8 anyway.

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

include: remove VLAs

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

tests: remove VLAs

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

Makefile: add missing lxctest.h

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

utils: s/strtok_r()/lxc_iterate_parts()/g

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

tools: s/strtok_r()/lxc_iterate_parts()/g

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

storage: s/strtok_r()/lxc_iterate_parts()/g

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

state: s/strtok_r()/lxc_iterate_parts()/g

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

parse: s/strtok_r()/lxc_iterate_parts()/g

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

namespace: s/strtok_r()/lxc_iterate_parts()/g

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

lxccontainer: s/strtok_r()/lxc_iterate_parts()/g

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

confile: s/strtok_r()/lxc_iterate_parts()/g

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

conf: s/strtok_r()/lxc_iterate_parts()/g

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

cgroups: s/strtok_r()/lxc_iterate_parts()

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

utils: add lxc_iterate_parts()

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

Merge pull request #2479 from Blub/apparmor-profiles

RFC: Generated Apparmor profiles, namespaces, stacking

tests: add test for generated apparmor profiles

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

apparmor: allow start-container to change to lxc-**

For generated profiles with apparmor namespaces we get
profile names with slashes in them. To match those, we need
to allow changing to lxc-**, not just lxc-*.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

apparmor: profile generation

This copies lxd's apparmor profile generation. This tries to
detect features such as cgroup namespaces, apparmor
namespaces and stacking support, and has profile parts
conditionally for unprivileged containers.

This introduces the following changes to the configuration:
  lxc.apparmor.profile = generated
    The fixed value 'generated' will cause this
    functionality to be used, otherwise there should be no
    functional changes happening unless specifically
    requested with the next key:
  lxc.apparmor.allow_nesting
    This is a boolean which, if enabled, causes the
    following changes: When generated apparmor profiles are
    used, they will contain the necessary changes to allow
    creating a nested container. In addition to the usual
    mount points, /dev/.lxc/proc and /dev/.lxc/sys will
    contain procfs and sysfs mount points without the lxcfs
    overlays, which, if generated apparmor profiles are
    being used, will not be read/writable directly.
  lxc.apparmor.raw
    A list of raw apparmor profile lines to append to the
    profile. Only valid when using generated profiles.

The following apparmor profile lines have not been copied
from lxd:

  mount /var/lib/lxd/shmounts/ -> /var/lib/lxd/shmounts/,
  mount none -> /var/lib/lxd/shmounts/,
  mount options=bind /var/lib/lxd/shmounts/** -> /var/lib/lxd/**,

They should be added via lxc.apparmor.raw entries by lxd.

In order for apparmor_parser's cache to be of use, this adds
a --with-apparmor-cache-dir ./configure option.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

apparmor: update current profiles

remove cgmanager rules and add fstype=cgroup2 variants for
the existing fstype=cgroup rules

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

utils: add must_concat helper

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

apparmor: use fopen_cloexec

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

Merge pull request #2492 from brauner/2018-07-14/fix_indendation

lxccontainer: fix indendation

lxccontainer: fix indendation

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

lsm: fixup lsm_process_label_set_at return values

Always return -1 on error (some code paths returned -1, some
returned negative error codes), don't assume 'errno' is set
afterwards, as the function already prints errors and not
all code paths will have a usable errno value.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

tests: lxc-test-apparmor-mount: check environment early

don't kill all my processes when running it as user...

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

tests: lxc-test-apparmor-mount: show a log on error

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>

Merge pull request #2489 from 2xsec/bugfix

change log macro of error case from lxc_ambient_caps_up/down

Merge pull request #2300 from LizaTretyakova/mount_injection

Mount injection API

confile: add missing header

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

start: coding style fixes

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

conf: coding style fixes

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

confile: add strdup failure check

Signed-off-by: Liza Tretyakova <elizabet.tretyakova@gmail.com>
[christian.brauner@ubuntu.com: coding style]
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>