security-announce: openSUSE-SU-2020:0398-1: moderate: Security update for cni, cni-plugins, conmon, fuse-overlayfs, podman
openSUSE Security Update: Security update for cni, cni-plugins, conmon, fuse-overlayfs, podman
______________________________________________________________________________
Announcement ID: openSUSE-SU-2020:0398-1
Rating: moderate
References: #1155217 #1160460 #1164390
Cross-References: CVE-2019-18466
Affected Products:
openSUSE Leap 15.1
______________________________________________________________________________
An update that solves one vulnerability and has two fixes
is now available.
Description:
This update for cni, cni-plugins, conmon, fuse-overlayfs, podman fixes the
following issues:
podman was updated to 1.8.0:
- CVE-2019-18466: Fixed a bug where podman cp would improperly copy files
on the host when copying a symlink in the container that included a glob
operator (#3829 bsc#1155217)
- The name of the cni-bridge in the default config changed from "cni0" to
"podman-cni0" with podman-1.6.0. Add a %trigger to rename the bridge in
the system to the new default if it exists. The trigger is only excuted
when updating podman-cni-config from something older than 1.6.0. This is
mainly needed for SLE where we're updating from 1.4.4 to 1.8.0
(bsc#1160460).
Update podman to v1.8.0 (bsc#1160460):
* Features
- The podman system service command has been added, providing a preview
of Podman's new Docker-compatible API. This API is still very new, and
not yet ready for production use, but is available for early testing
- Rootless Podman now uses Rootlesskit for port forwarding, which should
greatly improve performance and capabilities
- The podman untag command has been added to remove tags from images
without deleting them
- The podman inspect command on images now displays previous names they
used
- The podman generate systemd command now supports a --new
option to generate service files that create and run new containers
instead of managing existing containers
- Support for --log-opt tag= to set logging tags has been added to the
journald log driver
- Added support for using Seccomp profiles embedded in images for podman
run and podman create via the new --seccomp-policy CLI flag
- The podman play kube command now honors pull policy
* Bugfixes
- Fixed a bug where the podman cp command would not copy the contents of
directories when paths ending in /. were given
- Fixed a bug where the podman play kube command did not properly locate
Seccomp profiles specified relative to localhost
- Fixed a bug where the podman info command for remote Podman did not
show registry information
- Fixed a bug where the podman exec command did not support having input
piped into it
- Fixed a bug where the podman cp command with rootless Podman
on CGroups v2 systems did not properly determine if the container
could be paused while copying
- Fixed a bug where the podman container prune --force command could
possible remove running containers if they were started while the
command was running
- Fixed a bug where Podman, when run as root, would not properly
configure slirp4netns networking when requested
- Fixed a bug where podman run --userns=keep-id did not work when the
user had a UID over 65535
- Fixed a bug where rootless podman run and podman create with the
--userns=keep-id option could change permissions on /run/user/$UID and
break KDE
- Fixed a bug where rootless Podman could not be run in a systemd
service on systems using CGroups v2
- Fixed a bug where podman inspect would show CPUShares as 0, instead of
the default (1024), when it was not explicitly set
- Fixed a bug where podman-remote push would segfault
- Fixed a bug where image healthchecks were not shown in the
output of podman inspect
- Fixed a bug where named volumes created with containers from pre-1.6.3
releases of Podman would be autoremoved with their containers if the
--rm flag was given, even if they were given names
- Fixed a bug where podman history was not computing image sizes
correctly
- Fixed a bug where Podman would not error on invalid values to the
--sort flag to podman images
- Fixed a bug where providing a name for the image made by podman commit
was mandatory, not optional as it should be
- Fixed a bug where the remote Podman client would append an extra " to
%PATH
- Fixed a bug where the podman build command would sometimes ignore the
-f option and build the wrong Containerfile
- Fixed a bug where the podman ps --filter command would only filter
running containers, instead of all containers, if
--all was not passed
- Fixed a bug where the podman load command on compressed images would
leave an extra copy on disk
- Fixed a bug where the podman restart command would not properly clean
up the network, causing it to function differently from podman stop;
podman start
- Fixed a bug where setting the --memory-swap flag to podman create and
podman run to -1 (to indicate unlimited) was not supported
* Misc
- Initial work on version 2 of the Podman remote API has been merged,
but is still in an alpha state and not ready for use. Read more here
- Many formatting corrections have been made to the manpages
- The changes to address (#5009) may cause anonymous volumes created by
Podman versions 1.6.3 to 1.7.0 to not be removed when their container
is removed
- Updated vendored Buildah to v1.13.1
- Updated vendored containers/storage to v1.15.8
- Updated vendored containers/image to v5.2.0
- Add apparmor-abstractions as required runtime dependency to have
`tunables/global` available.
- fixed the --force flag for the "container prune" command.
( https://github.com/containers/libpod/issues/4844)
Update podman to v1.7.0
* Features
- Added support for setting a static MAC address for containers
- Added support for creating macvlan networks with podman network
create, allowing Podman containers to be attached directly to networks
the host is connected to
- The podman image prune and podman container prune commands now support
the --filter flag to filter what will be pruned, and now prompts for
confirmation when run without --force (#4410 and #4411)
- Podman now creates CGroup namespaces by default on systems using
CGroups v2 (#4363)
- Added the podman system reset command to remove all Podman files and
perform a factory reset of the Podman installation
- Added the --history flag to podman images to display previous names
used by images (#4566)
- Added the --ignore flag to podman rm and podman stop to not error when
requested containers no longer exist
- Added the --cidfile flag to podman rm and podman stop to read the IDs
of containers to be removed or stopped from a file
- The podman play kube command now honors Seccomp annotations (#3111)
- The podman play kube command now honors RunAsUser, RunAsGroup, and
selinuxOptions
- The output format of the podman version command has been changed to
better match docker version when using the
--format flag
- Rootless Podman will no longer initialize containers/storage twice,
removing a potential deadlock preventing Podman commands from running
while an image was being pulled (#4591)
- Added tmpcopyup and notmpcopyup options to the --tmpfs and
--mount type=tmpfs flags to podman create and podman run to control
whether the content of directories are copied into tmpfs filesystems
mounted over them
- Added support for disabling detaching from containers by setting empty
detach keys via --detach-keys=""
- The podman build command now supports the --pull and
--pull-never flags to control when images are pulled during a build
- The podman ps -p command now shows the name of the pod as well as its
ID (#4703)
- The podman inspect command on containers will now display the command
used to create the container
- The podman info command now displays information on registry mirrors
(#4553)
* Bugfixes
- Fixed a bug where Podman would use an incorrect runtime directory as
root, causing state to be deleted after root logged out and making
Podman in systemd services not function properly
- Fixed a bug where the --change flag to podman import and podman commit
was not being parsed properly in many cases
- Fixed a bug where detach keys specified in libpod.conf were not used
by the podman attach and podman exec commands, which always used the
global default ctrl-p,ctrl-q key combination (#4556)
- Fixed a bug where rootless Podman was not able to run podman pod stats
even on CGroups v2 enabled systems (#4634)
- Fixed a bug where rootless Podman would fail on kernels without the
renameat2 syscall (#4570)
- Fixed a bug where containers with chained network namespace
dependencies (IE, container A using --net container=B and container B
using --net container=C) would not properly mount /etc/hosts and
/etc/resolv.conf into the container (#4626)
- Fixed a bug where podman run with the --rm flag and without
-d could, when run in the background, throw a 'container does not
exist' error when attempting to remove the container after it exited
- Fixed a bug where named volume locks were not properly reacquired
after a reboot, potentially leading to deadlocks when trying to start
containers using the volume (#4605 and #4621)
- Fixed a bug where Podman could not completely remove containers if
sent SIGKILL during removal, leaving the container name unusable
without the podman rm --storage command to complete removal (#3906)
- Fixed a bug where checkpointing containers started with --rm was
allowed when --export was not specified (the container, and
checkpoint, would be removed after checkpointing was complete by --rm)
(#3774)
- Fixed a bug where the podman pod prune command would fail if
containers were present in the pods and the --force flag was not
passed (#4346)
- Fixed a bug where containers could not set a static IP or static MAC
address if they joined a non-default CNI network (#4500)
- Fixed a bug where podman system renumber would always throw an error
if a container was mounted when it was run
- Fixed a bug where podman container restore would fail with containers
using a user namespace
- Fixed a bug where rootless Podman would attempt to use the journald
events backend even on systems without systemd installed
- Fixed a bug where podman history would sometimes not properly identify
the IDs of layers in an image (#3359)
- Fixed a bug where containers could not be restarted when Conmon v2.0.3
or later was used
- Fixed a bug where Podman did not check image OS and Architecture
against the host when starting a container
- Fixed a bug where containers in pods did not function properly with
the Kata OCI runtime (#4353)
- Fixed a bug where `podman info --format '{{ json . }}' would not
produce JSON output (#4391)
- Fixed a bug where Podman would not verify if files passed to
--authfile existed (#4328)
- Fixed a bug where podman images --digest would not always print
digests when they were available
- Fixed a bug where rootless podman run could hang due to a race with
reading and writing events
- Fixed a bug where rootless Podman would print warning-level logs
despite not be instructed to do so (#4456)
- Fixed a bug where podman pull would attempt to fetch from remote
registries when pulling an unqualified image using the docker-daemon
transport (#4434)
- Fixed a bug where podman cp would not work if STDIN was a pipe
- Fixed a bug where podman exec could stop accepting input if anything
was typed between the command being run and the exec session starting
(#4397)
- Fixed a bug where podman logs --tail 0 would print all lines
of a container's logs, instead of no lines (#4396)
- Fixed a bug where the timeout for slirp4netns was incorrectly set,
resulting in an extremely long timeout (#4344)
- Fixed a bug where the podman stats command would print CPU
utilizations figures incorrectly (#4409)
- Fixed a bug where the podman inspect --size command would not print
the size of the container's read/write layer if the size was 0 (#4744)
- Fixed a bug where the podman kill command was not properly validating
signals before use (#4746)
- Fixed a bug where the --quiet and --format flags to podman ps could
not be used at the same time
- Fixed a bug where the podman stop command was not stopping exec
sessions when a container was created without a PID namespace
(--pid=host)
- Fixed a bug where the podman pod rm --force command was not removing
anonymous volumes for containers that were removed
- Fixed a bug where the podman checkpoint command would not export all
changes to the root filesystem of the container if performed more than
once on the same container (#4606)
- Fixed a bug where containers started with --rm would not be
automatically removed on being stopped if an exec session was running
inside the container (#4666)
* Misc
- The fixes to runtime directory path as root can cause strange behavior
if an upgrade is performed while containers are running
- Updated vendored Buildah to v1.12.0
- Updated vendored containers/storage library to v1.15.4
- Updated vendored containers/image library to v5.1.0
- Kata Containers runtimes (kata-runtime, kata-qemu, and kata-fc) are
now present in the default libpod.conf, but will not be available
unless Kata containers is installed on the system
- Podman previously did not allow the creation of containers with a
memory limit lower than 4MB. This restriction has been removed, as the
crun runtime can create containers with significantly less memory
Update podman to v1.6.4
- Remove winsz FIFO on container restart to allow use with Conmon 2.03 and
higher
- Ensure volumes reacquire locks on system restart, preventing deadlocks
when starting containers
- Suppress spurious log messages when running rootless Podman
- Update vendored containers/storage to v1.13.6
- Fix a deadlock related to writing events
- Do not use the journald event logger when it is not available
Update podman to v1.6.2
* Features
- Added a --runtime flag to podman system migrate to allow the OCI
runtime for all containers to be reset, to ease transition to the crun
runtime on CGroups V2 systems until runc gains full support
- The podman rm command can now remove containers in broken states which
previously could not be removed
- The podman info command, when run without root, now shows information
on UID and GID mappings in the rootless user namespace
- Added podman build --squash-all flag, which squashes all layers
(including those of the base image) into one layer
- The --systemd flag to podman run and podman create now accepts a
string argument and allows a new value, always, which forces systemd
support without checking if the the container entrypoint is systemd
* Bugfixes
- Fixed a bug where the podman top command did not work on systems using
CGroups V2 (#4192)
- Fixed a bug where rootless Podman could double-close a file, leading
to a panic
- Fixed a bug where rootless Podman could fail to retrieve some
containers while refreshing the state
- Fixed a bug where podman start --attach --sig-proxy=false would still
proxy signals into the container
- Fixed a bug where Podman would unconditionally use a non-default path
for authentication credentials (auth.json), breaking podman login
integration with skopeo and other tools using the containers/image
library
- Fixed a bug where podman ps --format=json and podman images
--format=json would display null when no results were returned,
instead of valid JSON
- Fixed a bug where podman build --squash was incorrectly squashing all
layers into one, instead of only new layers
- Fixed a bug where rootless Podman would allow volumes with
options to be mounted (mounting volumes requires root), creating an
inconsistent state where volumes reported as mounted but were not
(#4248)
- Fixed a bug where volumes which failed to unmount could not be removed
(#4247)
- Fixed a bug where Podman incorrectly handled some errors relating to
unmounted or missing containers in containers/storage
- Fixed a bug where podman stats was broken on systems running CGroups
V2 when run rootless (#4268)
- Fixed a bug where the podman start command would print the short
container ID, instead of the full ID
- Fixed a bug where containers created with an OCI runtime that is no
longer available (uninstalled or removed from the config file) would
not appear in podman ps and could not be removed via podman rm
- Fixed a bug where containers restored via podman container restore
--import would retain the CGroup path of the original container, even
if their container ID changed; thus, multiple containers created from
the same checkpoint would all share the same CGroup
* Misc
- The default PID limit for containers is now set to 4096. It can be
adjusted back to the old default (unlimited) by passing
--pids-limit 0 to podman create and podman run
- The podman start --attach command now automatically attaches STDIN if
the container was created with -i
- The podman network create command now validates network names using
the same regular expression as container and pod names
- The --systemd flag to podman run and podman create will now
only enable systemd mode when the binary being run inside the
container is /sbin/init, /usr/sbin/init, or ends in systemd
(previously detected any path ending in init or systemd)
- Updated vendored Buildah to 1.11.3
- Updated vendored containers/storage to 1.13.5
- Updated vendored containers/image to 4.0.1
Update podman to v1.6.1
* Features
- The podman network create, podman network rm, podman network inspect,
and podman network ls commands have been added to manage CNI networks
used by Podman
- The podman volume create command can now create and mount volumes with
options, allowing volumes backed by NFS, tmpfs, and many other
filesystems
- Podman can now run containers without CGroups for better integration
with systemd by using the --cgroups=disabled flag with podman create
and podman run. This is presently only supported with the crun OCI
runtime
- The podman volume rm and podman volume inspect commands can now refer
to volumes by an unambiguous partial name, in addition to full name
(e.g. podman volume rm myvol to remove a volume named myvolume) (#3891)
- The podman run and podman create commands now support the
--pull flag to allow forced re-pulling of images (#3734)
- Mounting volumes into a container using --volume, --mount, and
--tmpfs now allows the suid, dev, and exec mount options (the inverse
of nosuid, nodev, noexec) (#3819)
- Mounting volumes into a container using --mount now allows the
relabel=Z and relabel=z options to relabel mounts.
- The podman push command now supports the --digestfile option to save a
file containing the pushed digest
- Pods can now have their hostname set via podman pod create
--hostname or providing Pod YAML with a hostname set to podman play
kube (#3732)
- The podman image sign command now supports the --cert-dir flag
- The podman run and podman create commands now support the
--security-opt label=filetype:$LABEL flag to set the SELinux label for
container files
- The remote Podman client now supports healthchecks
* Bugfixes
- Fixed a bug where remote podman pull would panic if a Varlink
connection was not available (#4013)
- Fixed a bug where podman exec would not properly set terminal size
when creating a new exec session (#3903)
- Fixed a bug where podman exec would not clean up socket symlinks on
the host (#3962)
- Fixed a bug where Podman could not run systemd in containers that
created a CGroup namespace
- Fixed a bug where podman prune -a would attempt to prune images used
by Buildah and CRI-O, causing errors (#3983)
- Fixed a bug where improper permissions on the ~/.config directory
could cause rootless Podman to use an incorrect directory for storing
some files
- Fixed a bug where the bash completions for podman import threw errors
- Fixed a bug where Podman volumes created with podman volume create
would not copy the contents of their mountpoint the first time they
were mounted into a container (#3945)
- Fixed a bug where rootless Podman could not run podman exec when the
container was not run inside a CGroup owned by the user (#3937)
- Fixed a bug where podman play kube would panic when given Pod YAML
without a securityContext (#3956)
- Fixed a bug where Podman would place files incorrectly when
storage.conf configuration items were set to the empty string (#3952)
- Fixed a bug where podman build did not correctly inherit Podman's
CGroup configuration, causing crashed on CGroups V2 systems (#3938)
- Fixed a bug where remote podman run --rm would exit before the
container was completely removed, allowing race conditions when
removing container resources (#3870)
- Fixed a bug where rootless Podman would not properly handle changes to
/etc/subuid and /etc/subgid after a container was launched
- Fixed a bug where rootless Podman could not include some devices in a
container using the --device flag (#3905)
- Fixed a bug where the commit Varlink API would segfault if provided
incorrect arguments (#3897)
- Fixed a bug where temporary files were not properly cleaned up after a
build using remote Podman (#3869)
- Fixed a bug where podman remote cp crashed instead of reporting it was
not yet supported (#3861)
- Fixed a bug where podman exec would run as the wrong user when execing
into a container was started from an image with Dockerfile USER (or a
user specified via podman run --user) (#3838)
- Fixed a bug where images pulled using the oci: transport would be
improperly named
- Fixed a bug where podman varlink would hang when managed by systemd
due to SD_NOTIFY support conflicting with Varlink (#3572)
- Fixed a bug where mounts to the same destination would sometimes not
trigger a conflict, causing a race as to which was actually mounted
- Fixed a bug where podman exec --preserve-fds caused Podman to hang
(#4020)
- Fixed a bug where removing an unmounted container that was unmounted
might sometimes not properly clean up the container (#4033)
- Fixed a bug where the Varlink server would freeze when run in a
systemd unit file (#4005)
- Fixed a bug where Podman would not properly set the $HOME environment
variable when the OCI runtime did not set it
- Fixed a bug where rootless Podman would incorrectly print warning
messages when an OCI runtime was not found (#4012)
- Fixed a bug where named volumes would conflict with, instead of
overriding, tmpfs filesystems added by the --read-only-tmpfs flag to
podman create and podman run
- Fixed a bug where podman cp would incorrectly make the target
directory when copying to a symlink which pointed to a nonexistent
directory (#3894)
- Fixed a bug where remote Podman would incorrectly read STDIN when the
-i flag was not set (#4095)
- Fixed a bug where podman play kube would create an empty pod when
given an unsupported YAML type (#4093)
- Fixed a bug where podman import --change improperly parsed CMD (#4000)
- Fixed a bug where rootless Podman on systems using CGroups V2 would
not function with the cgroupfs CGroups manager
- Fixed a bug where rootless Podman could not correctly identify the
DBus session address, causing containers to fail to start (#4162)
- Fixed a bug where rootless Podman with slirp4netns networking would
fail to start containers due to mount leaks
* Misc
- Significant changes were made to Podman volumes in this release. If
you have pre-existing volumes, it is strongly recommended to run
podman system renumber after upgrading.
- Version 0.8.1 or greater of the CNI Plugins is now required for Podman
- Version 2.0.1 or greater of Conmon is strongly recommended
- Updated vendored Buildah to v1.11.2
- Updated vendored containers/storage library to v1.13.4
- Improved error messages when trying to create a pod with no name via
podman play kube
- Improved error messages when trying to run podman pause or podman
stats on a rootless container on a system without CGroups V2 enabled
- TMPDIR has been set to /var/tmp by default to better handle large
temporary files
- podman wait has been optimized to detect stopped containers more
rapidly
- Podman containers now include a ContainerManager annotation indicating
they were created by libpod
- The podman info command now includes information about slirp4netns and
fuse-overlayfs if they are available
- Podman no longer sets a default size of 65kb for tmpfs filesystems
- The default Podman CNI network has been renamed in an attempt to
prevent conflicts with CRI-O when both are run on the same system.
This should only take effect on system restart
- The output of podman volume inspect has been more closely matched to
docker volume inspect
- Add katacontainers as a recommended package, and include it as an
additional OCI runtime in the configuration.
Update podman to v1.5.1
* Features
- The hostname of pods is now set to the pod's name
* Bugfixes
- Fixed a bug where podman run and podman create did not honor the
--authfile
option (#3730)
- Fixed a bug where containers restored with podman container restore
--import would incorrectly duplicate the Conmon PID file of the
original container
- Fixed a bug where podman build ignored the default OCI runtime
configured in libpod.conf
- Fixed a bug where podman run --rm (or force-removing any running
container with podman rm --force) were not retrieving the correct exit
code (#3795)
- Fixed a bug where Podman would exit with an error if any configured
hooks directory was not present
- Fixed a bug where podman inspect and podman commit would not use the
correct CMD for containers run with podman play kube
- Fixed a bug created pods when using rootless Podman and CGroups V2
(#3801)
- Fixed a bug where the podman events command with the --since or --until
options could take a very long time to complete
* Misc
- Rootless Podman will now inherit OCI runtime configuration from the
root configuration (#3781)
- Podman now properly sets a user agent while contacting registries
(#3788)
- Add zsh completion for podman commands
Update podman to v1.5.0
* Features
- Podman containers can now join the user namespaces of other containers
with --userns=container:$ID, or a user namespace at an arbitary path
with --userns=ns:$PATH
- Rootless Podman can experimentally squash all UIDs and GIDs in an
image to a single UID and GID (which does not require use of the
newuidmap and newgidmap executables) by passing
--storage-opt ignore_chown_errors
- The podman generate kube command now produces YAML for any bind mounts
the container has created (#2303)
- The podman container restore command now features a new flag,
--ignore-static-ip, that can be used with --import to import a single
container with a static IP multiple times on the same host
- Added the ability for podman events to output JSON by specifying
--format=json
- If the OCI runtime or conmon binary cannot be found at the paths
specified in libpod.conf, Podman will now also search for them in the
calling user's path
- Added the ability to use podman import with URLs (#3609)
- The podman ps command now supports filtering names using regular
expressions (#3394)
- Rootless Podman containers with --privileged set will now mount in all
host devices that the user can access
- The podman create and podman run commands now support the
--env-host flag to forward all environment variables from the host
into the container
- Rootless Podman now supports healthchecks (#3523)
- The format of the HostConfig portion of the output of podman inspect
on containers has been improved and synced with Docker
- Podman containers now support CGroup namespaces, and can create them
by passing --cgroupns=private to podman run or podman create
- The podman create and podman run commands now support the
--ulimit=host flag, which uses any ulimits currently set on the host
for the container
- The podman rm and podman rmi commands now use different exit codes to
indicate 'no such container' and 'container is running' errors
- Support for CGroups V2 through the crun OCI runtime has been greatly
improved, allowing resource limits to be set for rootless containers
when the CGroups V2 hierarchy is in use
* Bugfixes
- Fixed a bug where a race condition could cause podman restart to fail
to start containers with ports
- Fixed a bug where containers restored from a checkpoint would not
properly report the time they were started at
- Fixed a bug where podman search would return at most 25 results, even
when the maximum number of results was set higher
- Fixed a bug where podman play kube would not honor capabilities set in
imported YAML (#3689)
- Fixed a bug where podman run --env, when passed a single key (to use
the value from the host), would set the environment variable in the
container even if it was not set on the host (#3648)
- Fixed a bug where podman commit --changes would not properly set
environment variables
- Fixed a bug where Podman could segfault while working with images with
no history
- Fixed a bug where podman volume rm could remove arbitrary volumes if
given an ambiguous name (#3635)
- Fixed a bug where podman exec invocations leaked memory by not
cleaning up files in tmpfs
- Fixed a bug where the --dns and --net=container flags to podman run
and podman create were not mutually exclusive (#3553)
- Fixed a bug where rootless Podman would be unable to run containers
when less than 5 UIDs were available
- Fixed a bug where containers in pods could not be removed without
removing the entire pod (#3556)
- Fixed a bug where Podman would not properly clean up all CGroup
controllers for created cgroups when using the cgroupfs CGroup driver
- Fixed a bug where Podman containers did not properly clean up files in
tmpfs, resulting in a memory leak as containers stopped
- Fixed a bug where healthchecks from images would not use default
settings for interval, retries, timeout, and start period when they
were not provided by the image (#3525)
- Fixed a bug where healthchecks using the HEALTHCHECK CMD format where
not properly supported (#3507)
- Fixed a bug where volume mounts using relative source paths would not
be properly resolved (#3504)
- Fixed a bug where podman run did not use authorization credentials
when a custom path was specified (#3524)
- Fixed a bug where containers checkpointed with podman container
checkpoint did not properly set their finished time
- Fixed a bug where running podman inspect on any container not created
with podman run or podman create (for example, pod infra containers)
would result in a segfault (#3500)
- Fixed a bug where healthcheck flags for podman create and podman run
were incorrectly named (#3455)
- Fixed a bug where Podman commands would fail to find targets if a
partial ID was specified that was ambiguous between a container and
pod (#3487)
- Fixed a bug where restored containers would not have the correct
SELinux label
- Fixed a bug where Varlink endpoints were not working properly if more
was not correctly specified
- Fixed a bug where the Varlink PullImage endpoint would crash if an
error occurred (#3715)
- Fixed a bug where the --mount flag to podman create and podman run did
not allow boolean arguments for its ro and rw options (#2980)
- Fixed a bug where pods did not properly share the UTS namespace,
resulting in incorrect behavior from some utilities which rely on
hostname (#3547)
- Fixed a bug where Podman would unconditionally append ENTRYPOINT to
CMD during podman commit (and when reporting CMD in podman inspect)
(#3708)
- Fixed a bug where podman events with the journald events backend would
incorrectly print 6 previous events when only new events were
requested (#3616)
- Fixed a bug where podman port would exit prematurely when a port
number was specified (#3747)
- Fixed a bug where passing . as an argument to the --dns-search flag to
podman create and podman run was not properly clearing DNS search
domains in the container
* Misc
- Updated vendored Buildah to v1.10.1
- Updated vendored containers/image to v3.0.2
- Updated vendored containers/storage to v1.13.1
- Podman now requires conmon v2.0.0 or higher
- The podman info command now displays the events logger being in use
- The podman inspect command on containers now includes the ID of the
pod a container has joined and the PID of the container's conmon
process
- The -v short flag for podman --version has been re-added
- Error messages from podman pull should be significantly clearer
- The podman exec command is now available in the remote client
- The podman-v1.5.0.tar.gz file attached is podman packaged for MacOS.
It can be installed using Homebrew.
- Update libpod.conf to support latest path discovery feature for `runc`
and `conmon` binaries.
conmon was included in version 2.0.10. (bsc#1160460, bsc#1164390,
jsc#ECO-1048, jsc#SLE-11485, jsc#SLE-11331):
fuse-overlayfs was updated to v0.7.6 (bsc#1160460)
- do not look in lower layers for the ino if there is no origin xattr set
- attempt to use the file path if the operation on the fd fails with ENXIO
- do not expose internal xattrs through listxattr and getxattr
- fix fallocate for deleted files.
- ignore O_DIRECT. It causes issues with libfuse not using an aligned
buffer, causing write(2) to fail with EINVAL.
- on copyup, do not copy the opaque xattr.
- fix a wrong lookup for whiteout files, that could happen on a double
unlink.
- fix possible segmentation fault in direct_fsync()
- use the data store to create missing whiteouts
- after a rename, force a directory reload
- introduce inodes cache
- correctly read inode for unix sockets
- avoid hash map lookup when possible
- use st_dev for the ino key
- check whether writeback is supported
- set_attrs: don't require write to S_IFREG
- ioctl: do not reuse fi->fh for directories
- fix skip whiteout deletion optimization
- store the new mode after chmod
- support fuse writeback cache and enable it by default
- add option to disable fsync
- add option to disable xattrs
- add option to skip ino number check in lower layers
- fix fd validity check
- fix memory leak
- fix read after free
- fix type for flistxattr return
- fix warnings reported by lgtm.com
- enable parallel dirops
cni was updated to 0.7.1:
- Set correct CNI version for 99-loopback.conf
Update to version 0.7.1 (bsc#1160460):
* Library changes:
+ invoke : ensure custom envs of CNIArgs are prepended to process envs
+ add GetNetworkListCachedResult to CNI interface
+ delegate : allow delegation funcs override CNI_COMMAND env
automatically in heritance
* Documentation & Convention changes:
+ Update cnitool documentation for spec v0.4.0
+ Add cni-route-override to CNI plugin list
Update to version 0.7.0:
* Spec changes:
+ Use more RFC2119 style language in specification (must, should...)
+ add notes about ADD/DEL ordering
+ Make the container ID required and unique.
+ remove the version parameter from ADD and DEL commands.
+ Network interface name matters
+ be explicit about optional and required structure members
+ add CHECK method
+ Add a well-known error for "try again"
+ SPEC.md: clarify meaning of 'routes'
* Library changes:
+ pkg/types: Makes IPAM concrete type
+ libcni: return error if Type is empty
+ skel: VERSION shouldn't block on stdin
+ non-pointer instances of types.Route now correctly marshal to JSON
+ libcni: add ValidateNetwork and ValidateNetworkList functions
+ pkg/skel: return error if JSON config has no network name
+ skel: add support for plugin version string
+ libcni: make exec handling an interface for better downstream testing
+ libcni: api now takes a Context to allow operations to be timed out or
cancelled
+ types/version: add helper to parse PrevResult
+ skel: only print about message, not errors
+ skel,invoke,libcni: implementation of CHECK method
+ cnitool: Honor interface name supplied via CNI_IFNAME environment
variable.
+ cnitool: validate correct number of args
+ Don't copy gw from IP4.Gateway to Route.GW When converting from 0.2.0
+ add PrintTo method to Result interface
+ Return a better error when the plugin returns none
- Install sleep binary into CNI plugin directory
cni-plugins was updated to 0.8.4:
Update to version 0.8.4 (bsc#1160460):
* add support for mips64le
* Add missing cniVersion in README example
* bump go-iptables module to v0.4.5
* iptables: add idempotent functions
* portmap doesn't fail if chain doesn't exist
* fix portmap port forward flakiness
* Add Bruce Ma and Piotr Skarmuk as owners
Update to version 0.8.3:
* Enhancements:
* static: prioritize the input sources for IPs (#400).
* tuning: send gratuitous ARP in case of MAC address update (#403).
* bandwidth: use uint64 for Bandwidth value (#389).
* ptp: only override DNS conf if DNS settings provided (#388).
* loopback: When prevResults are not supplied to loopback plugin, create
results to return (#383).
* loopback support CNI CHECK and result cache (#374).
* Better input validation:
* vlan: add MTU validation to loadNetConf (#405).
* macvlan: add MTU validation to loadNetConf (#404).
* bridge: check vlan id when loading net conf (#394).
* Bugfixes:
* bugfix: defer after err check, or it may panic (#391).
* portmap: Fix dual-stack support (#379).
* firewall: don't return error in DEL if prevResult is not found (#390).
* bump up libcni back to v0.7.1 (#377).
* Docs:
* contributing doc: revise test script name to run (#396).
* contributing doc: describe cnitool installation (#397).
Update plugins to v0.8.2
+ New features:
* Support "args" in static and tuning
* Add Loopback DSR support, allow l2tunnel networks to be used with the
l2bridge plugin
* host-local: return error if same ADD request is seen twice
* bandwidth: fix collisions
* Support ips capability in static and mac capability in tuning
* pkg/veth: Make host-side veth name configurable
+ Bug fixes:
* Fix: failed to set bridge addr: could not add IP address to "cni0":
file exists
* host-device: revert name setting to make retries idempotent (#357).
* Vendor update go-iptables. Vendor update go-iptables to
obtain commit f1d0510cabcb710d5c5dd284096f81444b9d8d10
* Update go.mod & go.sub
* Remove link Down/Up in MAC address change to prevent route flush
(#364).
* pkg/ip unit test: be agnostic of Linux version, on Linux 4.4 the
syscall error message is "invalid argument" not "file exists"
* bump containernetworking/cni to v0.7.1
Updated plugins to v0.8.1:
+ Bugs:
* bridge: fix ipMasq setup to use correct source address
* fix compilation error on 386
* bandwidth: get bandwidth interface in host ns through container
interface
+ Improvements:
* host-device: add pciBusID property
Updated plugins to v0.8.0:
+ New plugins:
* bandwidth - limit incoming and outgoing bandwidth
* firewall - add containers to firewall rules
* sbr - convert container routes to source-based routes
* static - assign a fixed IP address
* win-bridge, win-overlay: Windows plugins
+ Plugin features / changelog:
* CHECK Support
* macvlan:
- Allow to configure empty ipam for macvlan
- Make master config optional
* bridge:
- Add vlan tag to the bridge cni plugin
- Allow the user to assign VLAN tag
- L2 bridge Implementation.
* dhcp:
- Include Subnet Mask option parameter in DHCPREQUEST
- Add systemd unit file to activate socket with systemd
- Add container ifName to the dhcp clientID, making the clientID value
* flannel:
- Pass through runtimeConfig to delegate
* host-local:
- host-local: add ifname to file tracking IP address used
* host-device:
- Support the IPAM in the host-device
- Handle empty netns in DEL for loopback and host-device
* tuning:
- adds 'ip link' command related feature into tuning
+ Bug fixes & minor changes
* Correctly DEL on ipam failure for all plugins
* Fix bug on ip revert if cmdAdd fails on macvlan and host-device
* host-device: Ensure device is down before rename
* Fix -hostprefix option
* some DHCP servers expect to request for explicit router options
* bridge: release IP in case of error
* change source of ipmasq rule from ipn to ip
from version v0.7.5:
+ This release takes a minor change to the portmap plugin:
* Portmap: append, rather than prepend, entry rules
+ This fixes a potential issue where firewall rules may be bypassed by
port mapping
This update was imported from the SUSE:SLE-15-SP1:Update update project.
Patch Instructions:
To install this openSUSE Security Update use the SUSE recommended installation methods
like YaST online_update or "zypper patch".
Alternatively you can run the command listed for your product:
- openSUSE Leap 15.1:
zypper in -t patch openSUSE-2020-398=1
Package List:
- openSUSE Leap 15.1 (x86_64):
cni-0.7.1-lp151.2.3.1
cni-plugins-0.8.4-lp151.2.3.1
conmon-2.0.10-lp151.2.1
conmon-debuginfo-2.0.10-lp151.2.1
fuse-overlayfs-0.7.6-lp151.5.1
fuse-overlayfs-debuginfo-0.7.6-lp151.5.1
fuse-overlayfs-debugsource-0.7.6-lp151.5.1
podman-1.8.0-lp151.3.9.1
- openSUSE Leap 15.1 (noarch):
podman-cni-config-1.8.0-lp151.3.9.1
References:
https://www.suse.com/security/cve/CVE-2019-18466.html
https://bugzilla.suse.com/1155217
https://bugzilla.suse.com/1160460
https://bugzilla.suse.com/1164390
A cni, cni-plugins, conmon, fuse-overlayfs, podman security update has been released for openSUSE Leap 15.1.
