-
Notifications
You must be signed in to change notification settings - Fork 359
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
ansible-runner writes artifacts to world rw location by default #738
Labels
Comments
cidrblock
changed the title
ansible-runner writes artifiacts to world rw location by default
ansible-runner writes artifacts to world rw location by default
Jun 24, 2021
I should have caught this beforehand, that whole first line probably just needs to be replaced with https://docs.python.org/3/library/tempfile.html#tempfile.mkdtemp |
yeah, and with a prefix for attribution |
abadger
added a commit
to abadger/ansible-runner
that referenced
this issue
Jun 28, 2021
…dtemp() The previous code allowed for an easily exploitable race where the attacker could pre-create a known directory name to gain access to private files created by ansible-runner. Using mkdtemp() closes that hole because mkdtemp() ensures that the directory name returned has been allocated with restrictive permissions by mkdtemp itself so the attacker has no opportunity to inject their own directory. Fixes ansible#738
Merged
abadger
added a commit
to abadger/ansible-runner
that referenced
this issue
Jun 28, 2021
…dtemp() The previous code allowed for an easily exploitable race where the attacker could pre-create a known directory name to gain access to private files created by ansible-runner. Using mkdtemp() closes that hole because mkdtemp() ensures that the directory name returned has been allocated with restrictive permissions by mkdtemp itself so the attacker has no opportunity to inject their own directory. Fixes ansible#738
ansible-zuul bot
added a commit
that referenced
this issue
Jun 29, 2021
Secure tempfiles Looking through the code, I found two separate race conditions with tempfile handling which can be exploited. One of those is mentioned in #738. The other one happens because we create a tempdir and then delete it, then later reuse its name. An attacker can pre-create the tempdir in between the deletion and the second creation with the reused name. This PR attempts to fix both. Fixes #738 Reviewed-by: Shane McDonald <me@shanemcd.com> Reviewed-by: Matthew Jones <bsdmatburt@gmail.com>
atbore-phx
pushed a commit
to atbore-phx/ansible-runner
that referenced
this issue
Jun 29, 2021
…dtemp() The previous code allowed for an easily exploitable race where the attacker could pre-create a known directory name to gain access to private files created by ansible-runner. Using mkdtemp() closes that hole because mkdtemp() ensures that the directory name returned has been allocated with restrictive permissions by mkdtemp itself so the attacker has no opportunity to inject their own directory. Fixes ansible#738
atbore-phx
pushed a commit
to atbore-phx/ansible-runner
that referenced
this issue
Jun 29, 2021
…dtemp() The previous code allowed for an easily exploitable race where the attacker could pre-create a known directory name to gain access to private files created by ansible-runner. Using mkdtemp() closes that hole because mkdtemp() ensures that the directory name returned has been allocated with restrictive permissions by mkdtemp itself so the attacker has no opportunity to inject their own directory. Fixes ansible#738
shanemcd
pushed a commit
to shanemcd/ansible-runner
that referenced
this issue
Jul 2, 2021
…dtemp() The previous code allowed for an easily exploitable race where the attacker could pre-create a known directory name to gain access to private files created by ansible-runner. Using mkdtemp() closes that hole because mkdtemp() ensures that the directory name returned has been allocated with restrictive permissions by mkdtemp itself so the attacker has no opportunity to inject their own directory. Fixes ansible#738
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
related to: #722
The text was updated successfully, but these errors were encountered: