Fixes inability to use /dev/null when inside a container #3623
Conversation
| testPath, err := securejoin.SecureJoin("/sys", entry.Path) | ||
| if err != nil { | ||
| logrus.Errorf("error joining entry path: %s", err) | ||
| continue |
There was a problem hiding this comment.
If that's an error, I guess we should not continue but return it.
There was a problem hiding this comment.
Or maybe we should not use securejoin here, since this is merely a check that such device exists, and we do not use testPath in any other way. Worst case scenario, a bad guy will trick the system into adding a line for a device that does not exist into systemd unit properties.
So, I guess, using plain old path.Join is sufficient here.
@cyphar WDYT?
There was a problem hiding this comment.
The original patch didn't have secure join, but I saw @AkihiroSuda preferred having it (not sure if it's really needed here indeed); see #3620 (comment)
There was a problem hiding this comment.
SGTM, but probably the reason not to use securejoin should be recorded in the comment lines in the source
There was a problem hiding this comment.
No problem, I removed securejoin and added a comment why it's not used.
There was a problem hiding this comment.
Yeah securejoin is not necessary here (these are host paths anyway).
evanphx
force-pushed
the
b-remove-udev-dep-main
branch
from
87e308a to
806e837
Compare
This is a forward port of opencontainers#3620 The original code depended on the origin filesystem to have /dev/{block,char} populated. This is done by udev normally and while is very common non-containerized systemd installs, it's very easy to start systemd in a container created by runc itself and not have /dev/{block,char} populated. When this occurs, the following error output is observed: $ docker run hello-world docker: Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error reopening /dev/null inside container: open /dev/null: operation not permitted: unknown. /dev/null can't be opened because it was not added to the deviceAllowList, as there was no /dev/char directory. The change here utilizes the fact that when sysfs in in use, there is a /sys/dev/{block,char} that is kernel maintained that we can check. Signed-off-by: Evan Phoenix <[email protected]>
evanphx
force-pushed
the
b-remove-udev-dep-main
branch
from
806e837 to
462e719
Compare
kolyshkin
added
kind/bug
area/systemd
backport/1.1-todo
|
|
||
| // We don't bother with securejoin here because we create entry.Path | ||
| // right above here, so we know it's safe. | ||
| if _, err := os.Stat("/sys" + entry.Path); err != nil { |
There was a problem hiding this comment.
I just realised we probably want to do entry.Path = "/sys" + entry.Path here, don't we? Otherwise systemd will get a /dev/... path that doesn't exist?
There was a problem hiding this comment.
So I’m not entirely clear on what systemd does with these values but I can tell you that in my setup, those /dev/char entries don’t exist and systemd still does the right thing. Maybe it parses the path and uses the major:minor directly?
There was a problem hiding this comment.
Nah, the /sys/dev/{block,char}/<major>:<minor> is a directory, not a device file, so we can't use it for systemd.
Perhaps systemd has a set of built-in devices it can recognize. I'll take a look.
|
backport to 1.1 is done in #3620 |
kolyshkin
added
backport/1.1-done
This is a forward port of #3620
The original code depended on the origin filesystem to have /dev/{block,char} populated. This is done by udev normally and while is very common non-containerized systemd installs, it's very easy to start systemd in a container created by runc itself and not have /dev/{block,char} populated. When this occurs, the following error output is observed:
$ docker run hello-world
docker: Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error reopening /dev/null inside container: open /dev/null: operation not permitted: unknown.
/dev/null can't be opened because it was not added to the deviceAllowList, as there was no /dev/char directory. The change here utilizes the fact that when sysfs in in use, there is a /sys/dev/{block,char} that is kernel maintained that we can check.
Signed-off-by: Evan Phoenix [email protected]