Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

pkg/subscriptions: Modernize FIPS mounts #2174

Merged

Conversation

neverpanic
Copy link
Contributor

/etc/system-fips is deprecated in CentOS Stream 9 and has been removed from CentOS Stream 10. UBI8 containers still contain /etc/system-fips -> /run/secrets/system-fips, but UBI9 containers do not, so creating /run/secrets/system-fips on UBI9 (or later) does not serve a useful purpose. See [1, 2].

Instead of checking /etc/system-fips to determine whether FIPS mode is enabled on the host, read /proc/sys/crypto/fips_enabled, which works for all supported RHEL versions and likely even earlier.

In CentOS 10 Stream, the crypto-policies package does now contain /usr/share/crypto-policies/default-fips-config, which is meant to serve as a file to bind-mount over /etc/crypto-policies/config when in FIPS mode [3]. Manual creation of this file is thus no longer required in containers/common for modern containers. Using this file as a source also enables improvements in crypto-policies tooling which will now

  • unmount the two bind mounts when a user manually changes the policy using update-crypto-policies --set, something which was previously broken in containers because /etc/crypto-policies/config was a read-only bind-mount, and
  • unmount and restore the two bind-mounts when the crypto-policies package is updated. The crypto-policies package will only do these steps if the the bind mounts for crypto-policies use the
    /usr/share/crypto-policies/default-fips-config
    file as source, so it makes sense for containers/common to switch to that.

Closes: #2130
Related: https://issues.redhat.com/browse/CRYPTO-13556

Copy link
Member

@Honny1 Honny1 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!

The approach looks good overall, but I have a few suggestions that should reduce maintenance costs. Also, vendoring will need to be done to the projects that using c/common for example podman and buildah to make the changes take effect.

PTAL @mtrmac, @Luap99

pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
logrus.Errorf("Adding FIPS mode bind mounts to container: %v", err)
}
} else {
logrus.Debug("/proc/sys/crypto/fips_enabled does not contain '1', not adding FIPS mode bind mounts")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it worth logging the value here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think so. The value can't change without a reboot, and it's going to be the the same across the entire host, regardless of whether one is inside of a container or outside. It should be simple to obtain the actual value using cat if necessary for debugging.

The only case where this could differ from the global value is if somebody bind-mounted a file over it for testing purposes. I don't think we need to handle this eventuality.

file, err := os.Create(fipsFile)
fipsFileHost := filepath.Join(mountPoint, "etc/system-fips")
if fileutils.Lexists(fipsFileHost) != nil {
fipsFileTarget, err := filepath.EvalSymlinks(fipsFileHost)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This conceptually allows an attacker to set up /etc/system-fips inside the container to be a symlink with ../../… to break out of the container and refer to hosts’ files.

Yes, we actually only compare the fipsFileTarget text without referring to the possibly-unwanted file, but, still…

In other words, either we care about the specific symlink text, and we can just Readlink (probably not), or we care about where the symlink evaluates, whatever its form, and in that case .. in the / directory of the container should be evaluated correctly (restricted to the containers’ root).

(https://github.com/cyphar/filepath-securejoin is used for that purpose in some parts of the codebase.)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good catch, fixed by using securejoin. I still need a fileutils.Lexists, because the log messages should probably differ between /etc/system-fips not existing (which is a common config) and /etc/system-fips existing but not pointing to /run/secrets/system-fips (which is probably an uncommon config). That shouldn't be an issue, though, since fileutils.Lexists doesn't follow symlinks.

pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
Copy link
Contributor

@mtrmac mtrmac left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I didn’t actually review the substance of the PR, whether the code correctly handles all relevant versions of the OSes and behaviors.


Just looking at the mechanism, all the filepath.Join(mountPoint, …) in this file make me worried about symlink breakouts. There might not be much actually wrong there (at worst, I guess, triggering an an audit message about a DAC or SELinux permission violation by the container runtime when checking for presence of an out-of-contaienr file), but that’s just one small refactor away from disaster.

I do appreciate that almost all of that is pre-existing and not directly relevant to this PR.

@Luap99
Copy link
Member

Luap99 commented Oct 1, 2024

Just looking at the mechanism, all the filepath.Join(mountPoint, …) in this file make me worried about symlink breakouts. There might not be much actually wrong there (at worst, I guess, triggering an an audit message about a DAC or SELinux permission violation by the container runtime when checking for presence of an out-of-contaienr file), but that’s just one small refactor away from disaster.

I do appreciate that almost all of that is pre-existing and not directly relevant to this PR.

I noticed this last week when I looked at this PR and reported this internally so yes this is an pre existing issue, fix in #2185
@neverpanic once this is merged please rebase and make sure all joins with of the container mount point use securejoin.

@neverpanic
Copy link
Contributor Author

I'm on PTO until Oct 14, I'll fix the comments after that.

pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
Comment on lines 165 to 167
fips_enabled, err := os.ReadFile("/proc/sys/crypto/fips_enabled")
if err != nil {
logrus.Errorf("Failed to read /proc/sys/crypto/fips_enabled to determine FIPS state: %v", err)
return false
}
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This will log an error when the file does not exists, it not clear to me does the kernel always has the file or is that set to some CONFIG_ options in the kernel. If so we should certainly ignore ENOENT?

Second, this logs an error and then assumes fips is disabled, Given this is a security option should we not outright fail with the error (e.g. return the error to the caller)? I do realize MountsWithUIDGID() does not return an error so we would need to add a new function so maybe it is not worth it. Maybe it is not important enough but I always very much find it odd to log an error but then start the container anyway. Either way the issue is pre existing so I would not block this on it.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The kernel always has that file, so there's no need to ignore ENOENT.

I can make the change to bubble up the error, but as you said, it's going to be a larger change because MountsWithUIDGID() does not currently fail. Maybe we should do that in a separate PR afterwards?

Copy link
Member

@kwilczynski kwilczynski Oct 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The kernel always has that file, so there's no need to ignore ENOENT.

@neverpanic not always. I work with systems where CONFIG_CRYPTO_FIPS is not always enabled.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should be fixed now.

pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
pkg/subscriptions/subscriptions.go Outdated Show resolved Hide resolved
@neverpanic neverpanic force-pushed the cllang-modernize-fips-mounts branch 4 times, most recently from 3f4818e to 1b8afdf Compare October 30, 2024 17:09
Copy link
Member

@Luap99 Luap99 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

one small nit otherwise LGTM

cc @TomSweeneyRedHat not sure if the fips changes here need any updates in the container RHEL docs.

func shouldAddFIPSMounts() bool {
fips_enabled, err := os.ReadFile("/proc/sys/crypto/fips_enabled")
if err != nil {
if !os.IsNotExist(err) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

new code should use errors.Is(err, fs.ErrNotExist)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've used os.ErrNotExist, since that was already imported.

/etc/system-fips is deprecated in CentOS Stream 9 and has been removed
from CentOS Stream 10. UBI8 containers still contain /etc/system-fips ->
/run/secrets/system-fips, but UBI9 containers do not, so creating
/run/secrets/system-fips on UBI9 (or later) does not serve a useful
purpose. See [1, 2].

Instead of checking /etc/system-fips to determine whether FIPS mode is
enabled on the host, read /proc/sys/crypto/fips_enabled, which works for
all supported RHEL versions and likely even earlier.

In CentOS 10 Stream, the crypto-policies package does now contain
/usr/share/crypto-policies/default-fips-config, which is meant to serve
as a file to bind-mount over /etc/crypto-policies/config when in FIPS
mode [3]. Manual creation of this file is thus no longer required in
containers/common for modern containers. Using this file as a source
also enables improvements in crypto-policies tooling which will now
 - unmount the two bind mounts when a user manually changes the policy
   using update-crypto-policies --set, something which was previously
   broken in containers because /etc/crypto-policies/config was
   a read-only bind-mount, and
 - unmount and restore the two bind-mounts when the crypto-policies
   package is updated.
The crypto-policies package will only do these steps if the the bind
mounts for crypto-policies use the
  /usr/share/crypto-policies/default-fips-config
file as source, so it makes sense for containers/common to switch to
that.

[1]: https://gitlab.com/redhat-crypto/fedora-crypto-policies/-/merge_requests/111
[2]: https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/9/html/9.0_release_notes/deprecated_functionality#deprecated-functionality_security
[3]: https://gitlab.com/redhat-crypto/fedora-crypto-policies/-/commit/04ceadccfc07e5946b08157d06ca5c0d5a229d92

Closes: containers#2130
Related: https://issues.redhat.com/browse/CRYPTO-13556
Signed-off-by: Clemens Lang <[email protected]>
Copy link
Member

@Luap99 Luap99 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@openshift-ci openshift-ci bot added the approved label Nov 1, 2024
@kwilczynski
Copy link
Member

Nice!

@kwilczynski
Copy link
Member

/approve
/lgtm

Copy link
Contributor

openshift-ci bot commented Nov 1, 2024

@kwilczynski: changing LGTM is restricted to collaborators

In response to this:

/approve
/lgtm

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

@Luap99
Copy link
Member

Luap99 commented Nov 7, 2024

@mtrmac Mind giving this a final review/merge? I think it would be good to get this into podman 5.3.

Copy link
Member

@giuseppe giuseppe left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

/lgtm

Copy link
Contributor

openshift-ci bot commented Nov 7, 2024

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: giuseppe, kwilczynski, Luap99, neverpanic

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@openshift-merge-bot openshift-merge-bot bot merged commit e522662 into containers:main Nov 7, 2024
16 checks passed
@neverpanic neverpanic deleted the cllang-modernize-fips-mounts branch November 7, 2024 19:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Automatic FIPS mode bind-mounts rely on the presence of deprecated /etc/system-fips
6 participants