Merge pull request #323 from brauner/2020-02-20/cgroup2_support_7

bindings: port cpuview to new cgroup getters & cleanup macro expansion

bindings: use more cleanup macros

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

bindings: port cpuview to new cgroup getters

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

Merge pull request #322 from brauner/2020-02-20/cgroup2_support_5

bindings: port blkio to new cgroup getters

bindings: port blkio to new cgroup getters

cgroup2's io controller does not support most of the blkio files.

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

Merge pull request #321 from brauner/2020-02-20/cgroup2_support_4

bindings: port cpuset to new cgroup getters and implement cgroup2 sup…

bindings: introduce dot_or_empty()

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

bindings: port cpuset to new cgroup getters and implement cgroup2 support

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

Merge pull request #320 from brauner/2020-02-20/cgroup2_support_3

bindings: port memory to new cgroup getters and support cgroup2

bindings: port memory to new cgroup getters and implement cgroup2 support

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

bindings: rework proc meminfo helpers

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

Merge pull request #319 from brauner/2020-02-20/cgroup2_support_2

cgroups: add getter instead of open-coded cgfs_get_value()

cgroups: add getter instead of open-coded cgfs_get_value()

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

Merge pull request #318 from brauner/2020-02-20/cgroup2_support

bindings: add infrastructure for cgroup2 support

bindings: add infrastructure for cgroup2 support

Mostly based on code I've written for liblxc.

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

Merge pull request #316 from KellenRenshaw/lp1860813

handle swap reporting better when swap accounting is enabled

bindings: Adjusts the logic for calculating SwapFree to better handle conditions where swap usage is reported to be <0.

Signed-off-by: Kellen Renshaw <kellen.renshaw@canonical.com>

Merge pull request #312 from wlm86/dev

Improve version reporting

Enhance version option

Signed-off-by: LiMing Wu <19092205@suning.com>

Add compile method to README

Signed-off-by: LiMing Wu <19092205@suning.com>

Merge pull request #308 from yinhongbo/master

fixed a problem with server stability degradation caused by cgroup dy…

use is_shared_pidns helper

Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>

bindings: handle shared pidns

Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>

Merge pull request #309 from yinhongbo/read_file

fix read_file second reading empty bug

fix read_file second reading empty bug

Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>

Merge pull request #307 from yinhongbo/master

fix the problem of counting the number of CPUs when using cpuset and …

fix the problem of counting the number of CPUs when using cpuset and cpuset or using cpuset alone

Changelog: format code style

Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>

Merge pull request #305 from yinhongbo/master

fix /sys/devices/system/cpu/online in cpuset not correct bug

fix /sys/devices/system/cpu/online in cpuset not correct bug

Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>

Merge pull request #303 from yinhongbo/master

fix container run pid host mode /proc/stat error log

fix container run pid host mode /proc/stat error log

Signed-off-by: Hongbo Yin <yinhongbo@bytedance.com>

Merge pull request #302 from time-river/master

macro: fix lxcfs_{error,debug,v} build error when __VA_ARGS__ is empty

macro: fix lxcfs_{error,debug,v} build error when __VA_ARGS__ is empty

Originally, the compiler complained:
  macro.h:7:25: error: expected expression before ')' token
      __func__, __VA_ARGS__);

The reason is that GCC wouldn't abandon `,` when `__VA_ARGS__` is empty.
For emaple:
  #define eprintf(format, ...) fprintf (stderr, format, __VA_ARGS__)
  eprintf("success!\n", );
       → fprintf(stderr, "success!\n", );

According to GCC doc, it's okay when adding `##` before `__VA_ARGS__`:
  #define eprintf(format, ...) fprintf (stderr, format, ##__VA_ARGS__)
  eprintf ("success!\n")
       → fprintf(stderr, "success!\n");

Signed-off-by: river <river@vvl.me>

Release LXCFS 3.1.2

Signed-off-by: Stéphane Graber <stgraber@ubuntu.com>

Merge pull request #296 from brauner/2019-07-23/swapfix

bindings: ensure that opts is non NULL

Revert "Revert "cpu: improve cpu info virtualization""

This reverts commit 39b7df424120fa9c3c5dc209b37990fed57f1efd.

Looks like this is actually fine, un-revert.

Signed-off-by: Stéphane Graber <stgraber@ubuntu.com>

bindings: ensure that opts is non NULL

When the shared library is reloaded but the lxcfs binary itself will
likely not be replaced and even if will still be in-memory, i.e. the
fuse loop will not be restarted.
The code here introduced a new struct lxcfs_opts which is allocated and
gets stashed in the private_data member of the fuse file when the lxcfs
binary is first called. Now, in a scenario where the shared library is
reloaded the lxcfs binary itself won't be re-run. So now we have a
shared library that assumes there's always a valid struct lxcfs_opts
stashed in the private_data member. But there isnt'. This very likely
causes the segfaults we have seen.

Fix it by verifying a struct has been stashed.

Fixes: 7e60aa1b1540 ("option to disable swap in meminfo")
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #294 from brauner/master

Revert "cpu: improve cpu info virtualization"

Revert "cpu: improve cpu info virtualization"

This reverts commit db1b32f6ff056ce713c06b9d41bcc1eadd054abd.

We had a multitude of reports saying that they hit segfaults with
current lxcfs. After a bisect and various mentions in the error reports
of corrupt cpu information this is likely the culprit: revert for now.

Link: https://github.com/lxc/lxd/issues/6000
Link: https://lists.linuxcontainers.org/pipermail/lxc-users/2019-July/014932.html
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Release LXCFS 3.1.1

Signed-off-by: Stéphane Graber <stgraber@ubuntu.com>

Makefile: Fix typo in file name

Signed-off-by: Stéphane Graber <stgraber@ubuntu.com>

Release LXCFS 3.1.0

Signed-off-by: Stéphane Graber <stgraber@ubuntu.com>

Merge pull request #293 from brauner/2019-07-18/higher_precision_uptime

bindings: enable higher precision uptime

bindings: enable higher precision uptime

I don't know for what reason anyone but care but I guess ok.

Closes #292.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #290 from BurningXFlame/master

CPU views based on quotas: several improvements

cpu: improve cpu info virtualization

This commit introduces several improvements to CPU views based on
quotas:
- fall back to cpuacct.usage_percpu if cpuacct.usage_all not exists
- correct CPU usage
- correct CPU usage in partial CPU cases when quota/period doesn't yield
  an integer

Signed-off-by: Stephen Xiang <BurningXFlame@gmail.com>
[christian.brauner@ubuntu.com: squashed commits and fixed up commit message]
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Abstract must_strcat from must_strcat_pid. Replace stringbuild with must_strcat.

Signed-off-by: Stephen Xiang <BurningXFlame@gmail.com>

Merge pull request #282 from aither64/syscpus

mount hook: mount /sys/devices/system/cpu/online

mount hook: mount /sys/devices/system/cpu/online

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

Merge pull request #278 from wavezhang/master

Support /sys/devices/system/cpu/online

Merge pull request #280 from brauner/master

Revert "Merge pull request #272 from brauner/2019-02-19/compiler_base…

Revert "Merge pull request #272 from brauner/2019-02-19/compiler_based_housekeeping"

This reverts commit ac98f00cd89278e82a4cffe9075792d2070abe69, reversing
changes made to d8addb023a0e96b7d42e253703ba909bbc81dd26.

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

Merge pull request #272 from brauner/2019-02-19/compiler_based_housekeeping

lxcfs: coding style update

remove unused function

Signed-off-by: Yafei Bao <baoyafei@gmail.com>

sys dirs do not need to implement 'read' method

Signed-off-by: Yafei Bao <baoyafei@gmail.com>

support /sys/devices/system/cpu/online

Signed-off-by: Yafei Bao <baoyafei@gmail.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

bindings: partially convert to cleanup macros

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

lxcfs: use close cleanup attributes

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

lxcfs: coding style update

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

Merge pull request #271 from dmaj/master

option to disable swap in meminfo

option to disable swap in meminfo

Signed-off-by: Dieter Maier <dma@web.de>

Merge pull request #269 from tomponline/tp-skipcgroupmount

hooks: skip mounting lxcfs cgroup tree on centos 7 hosts

config: Adds RPM spec file.

On RHEL 7 distros the RPM spec file will run the lxc.mount.hook with the --skip-cgroup-mounts option.

Signed-off-by: tomponline <tomp@tomp.uk>

hooks: Adds --skip-cgroup-mounts flag to lxc.mount.hook script.

This can be used to avoid instability in LXCFS when running on RHEL 7 hosts.

Signed-off-by: tomponline <tomp@tomp.uk>

Merge pull request #270 from tomponline/tp-reload

config: Adds reload mode to sysvinit and systemd scripts.

Merge pull request #268 from tomponline/tp-lock

Revert "bindings: protect do_release_file_info() with mutex"

config: Adds reload mode to sysvinit and systemd scripts.

To to close down https://github.com/lxc/lxcfs/pull/188

Signed-off-by: tomponline <tomp@tomp.uk>

Revert "bindings: protect do_release_file_info() with mutex"

This reverts commit 0108dc1cc6eab76392d228ee4ca844729096473d.

I am still experiencing hangs in lxcfs even with this patch.

Signed-off-by: tomponline <tomp@tomp.uk>

Merge pull request #266 from tomponline/master

Adds mutex to do_release_file_info

bindings: protect do_release_file_info() with mutex

Add a mutex to do_release_file_info to avoid lxcfs crash when stopping
multiple containers concurrently

Signed-off-by: tomponline <tomp@tomp.uk>
[christian.brauner@ubuntu.com: use correct commit message style]
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #262 from brauner/2018-12-12/cg_readdir

bindings: prevent NULL pointer dereference

bindings: prevent NULL pointer dereference

Fixes: https://bugs.launchpad.net/ubuntu/+source/lxcfs/+bug/1807628
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>

Merge pull request #260 from aither64/cpu-views

CPU views based on quotas

stat: check for out of bound access to cpuacct usage data

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

CPU view: handle disabling/enabling of physical CPUs at runtime

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

CPU view: handle CPU hotplug at runtime

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

CPU view: add mutexes

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

CPU view: prune stale entries

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

stat: use cpu views

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

calc_hash(): do not apply modulo LOAD_SIZE

This will allow `calc_hash()` to be used for other purposes than
the loadavg implementation.

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

cpuinfo: use cpu view based on cpu quotas

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

Merge pull request #255 from brauner/2018-08-26/fix_makedev

tests: include missing sys/sysmacros.h header

tests: include missing sys/sysmacros.h header

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

Merge pull request #253 from brauner/2018-08-17/prevent_double_free

bindings: prevent double free

bindings: prevent double free

Closes #252.

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

Merge pull request #250 from brauner/2018-07-30/improve_error_message

bindings: better logging for write_string()

bindings: better logging for write_string()

Closes #249.

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

Merge pull request #248 from aither64/meminfo

meminfo: read shmem from memory cgroup

meminfo: set ShmemHugePages and ShmemPmdMapped to zero

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

meminfo: read shmem from cgroup parameter memory.stat

Shmem was passed as-is from the host, but other fields weren't (such as
Cached and SReclaimed), which has caused htop to show incorrect memory usage.
If `total_shmem` is found in `memory.stat`, it is used, otherwise `Shmem`
is reported as zero.

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

Merge pull request #245 from aither64/loadavg-leak

bindings: fix memory leak in calc_pid()

bindings: fix memory leak in calc_pid()

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

travis: fix .travis.yml

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

travis: add coverity support

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

Merge pull request #244 from aither64/perct-cpu-usage

Per-container CPU usage in /proc/stat

Per-container CPU usage in /proc/stat

Containers can see utilization of all available CPUs, even if the CPU
is utilized by other containers or by the host. The contents
of `/proc/stat` is shared across the system, except for hiding CPUs
excluded by cpuset. This commit attempts to fix that, but at a cost.
CPU usage is read from cpuacct cgroup, but that accounts only for
`user` and `system` fields from `/proc/stat`. Idle time can be
calculated, but other fields cannot, thus are always set to 0.
Additionally, idle time is based on values from the host, so even
a freshly started container has large initial idle time.

If the cpuacct cgroup is not present, or reading from it fails, LXCFS
will return CPU usage stats from the host as before.

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

Merge pull request #242 from aither64/loadavg-reload

loadavg: restart thread on library reload

loadavg: restart thread on library reload

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

Merge pull request #241 from aither64/loadavg-leak

bindings: fix memory leak in proc_loadavg_read()

bindings: fix memory leak in proc_loadavg_read()

Signed-off-by: Jakub Skokan <jakub.skokan@havefun.cz>

Merge pull request #237 from zhang2639/master

loadavg