2025.1 Series Release Notes

30.2.0-10

Deprecation Notes

• The add_image_location and fetch_image_location APIs policy rule default value role:admin or role:service is deprecated and will be changed to role:service in future release.

Security Issues

• Fixed a Server-Side Request Forgery (SSRF) vulnerability in the web-download image import method. With the default import_filtering_opts configuration, authenticated users could supply URIs targeting loopback or link-local addresses (for example cloud metadata at 169.254.169.254) and read the response through the imported image data.

  Glance now rejects import URIs that target loopback or link-local IP addresses unless the host is explicitly listed in allowed_hosts. Hostnames that resolve to any of those address types are also rejected. Private RFC1918 addresses are not blocked by this check so private clouds can continue to import from internal hosts.

• Fixed a Server-Side Request Forgery (SSRF) vulnerability when adding HTTP(S) image locations. Authenticated users could supply location URLs targeting loopback or link-local addresses (for example cloud metadata at 169.254.169.254). Glance would fetch those URLs from glance-api during location validation and hashing.

  HTTP(S) locations are now checked with the same import_filtering_opts host filtering used for web-download. Glance validates the URI before contacting the backend so restricted destinations never trigger an outbound request from the location-add path.

  This applies to POST /v2/images/{id}/locations and the legacy PATCH locations API when HTTP store is enabled.

• A bug in the web-download import plugin was fixed which would allow a remote URI to omit the Content-Length header and feed glance data forever until the staging disk filled. We now honor the image_size_cap limit for these streams. Note that the default value for this limit is 1TiB, which may be higher than is reasonable for some deployments. Operators should take this opportunity to review that setting for correctness.

Bug Fixes

• Bug 2127798: Automatic S3 credential update when EC2 credentials are rotated

  Glance now automatically updates S3 image location URLs when EC2 credentials are rotated in Keystone. Previously, images with S3 locations would become inaccessible after credential rotation, requiring manual intervention to update location metadata. This fix ensures seamless credential rotation without breaking image access.

• Bug 2158998: Block loopback and link-local destination addresses during web-download URI validation to prevent SSRF against local services and cloud metadata endpoints.

• Bug 2161330: Block restricted destination addresses when adding HTTP(S) image locations to prevent SSRF against local services and cloud metadata endpoints.

• Fix a regression introduced when fixing CVE-2026-71197. Now during image import via web-download, if the image host is resolved to several IP addresses by DNS, all of them that pass filtering critera will be tried one by one.

  This among others ensures that if DNS resolves to both IPv4 and IPv6 addresses but Glance itself is running in a single stack network, the correct IP address that Glance can connect to will eventually be tried.

30.2.0

Security Issues

• Fixed multiple Server-Side Request Forgery (SSRF) vulnerabilities in Glance’s image import functionality. These vulnerabilities could allow attackers to bypass URL validation and access internal resources.

  web-download Import Method SSRF:

  The web-download import method had two SSRF vulnerabilities:

  HTTP Redirect Bypass: The web-download import method did not validate redirect destinations when following HTTP redirects. An attacker could provide an initial URL that passed validation, but redirect to an internal or disallowed resource that would bypass the security checks. This has been fixed by implementing SafeRedirectHandler that validates redirect destinations before following them using the same validate_import_uri() checks as the initial URL.

  IP Address Encoding Bypass: The web-download import method’s URL validation could be bypassed by encoding IP addresses in alternative formats (decimal integer, hexadecimal, octal). For example, 127.0.0.1 could be encoded as 2130706433 (decimal) or 0x7f000001 (hexadecimal) to bypass blacklist checks. This has been fixed by implementing normalize_hostname() function that uses Python’s ipaddress module to validate IP addresses. The ipaddress module only accepts standard dotted-decimal notation for IPv4 and standard format for IPv6, automatically rejecting all encoded formats (decimal, hexadecimal, octal). Any attempt to use encoded IP formats is rejected, preventing SSRF bypass attacks.

  glance-download Import Method SSRF:

  The glance-download import method had redirect validation bypass vulnerabilities in two steps of the import flow:

  Image Data Download: When downloading image data from a remote Glance endpoint, redirects were not validated, allowing attackers to redirect to internal services.

  Metadata Fetch: When fetching image metadata from a remote Glance endpoint, redirects were not validated, allowing attackers to redirect to internal services.

  Both steps have been fixed by using SafeRedirectHandler to validate redirect destinations before following them.

  OVF Processing SSRF:

  The OVF processing functionality had a critical SSRF vulnerability with zero protection - no URI validation, no redirect validation, and no IP normalization. The code directly called urllib.request.urlopen(uri) without any validation checks. This has been fixed by adding URI validation using validate_import_uri() and redirect validation using SafeRedirectHandler.

  Affected Components:

  • glance.common.scripts.utils.get_image_data_iter()

  • glance.common.utils.validate_import_uri()

  • glance.async_.flows._internal_plugins.glance_download._DownloadGlanceImage.execute()

  • glance.async_.flows.api_image_import._ImportMetadata.execute()

  • glance.async_.flows.ovf_process._OVF_Process._get_ova_iter_objects()

  Impact:

  • Severity: High (web-download, glance-download), Critical (OVF processing)

  • Affected Versions: All versions prior to this fix

  • Workaround: Administrators can temporarily disable affected import methods by removing them from the enabled_import_methods configuration option

  Bugs #2138602, #2138672, #2138675

Bug Fixes

• Bug 2138602: Fixed SSRF vulnerability in web-download import method via HTTP redirect bypass and IP address encoding bypass. Added redirect validation using SafeRedirectHandler and IP address validation using Python’s ipaddress module to reject encoded IP formats and prevent bypass attacks.

  Bug 2138672: Fixed SSRF vulnerability in glance-download import method via HTTP redirect bypass. Added redirect validation for both image data download and metadata fetch operations.

  Bug 2138675: Fixed SSRF vulnerability in OVF processing functionality which lacked URI validation. Added URI validation and redirect validation to prevent SSRF attacks when processing OVA files.

30.0.0

New Features

• A new file, glance/wsgi/api.py, has been added as a place to hold a WSGI application object. This is intended to ease deployment by providing a consistent location for these objects. For example, if using uWSGI then instead of:

  [uwsgi]
  wsgi-file = /bin/glance-wsgi-api
  

  You can now use:

  [uwsgi]
  module = glance.wsgi.api:application
  

  This also simplifies deployment with other WSGI servers that expect module paths such as gunicorn.

• Glance now inspects image content on upload and import, checking format adherence with the declared disk_format and running safety checks on content it recognizes. The new config knob [image_format]/require_image_format_match controls whether or not images are rejected when the format set on image create does not match the content that is uploaded. Some images that are currently treated as raw may trigger safety check failures when examining the MBR record (or what glance thinks is an MBR). These may be legit failures (due to proliferation of GPT images with invalid Protective MBR structures) as well as other content that attempts to be PC-BIOS-Bootable and thus have a quasi-MBR structure ahead of the payload which may not be fully compliant. Thus a new config knob [image_format]/gpt_safety_checks_nonfatal is added to (by default) allow these failures to be non-fatal.

Upgrade Notes

• The stores_info_detail policy has deprecated the rule rule:context_is_admin in favor of rule:context_is_admin AND service_roles:service to allow service to service interaction.

• Glance will check that uploaded content matches disk_format by default, so operators should be on the lookout for any false positives and be ready for reports of upload failures if users are currently not properly representing their uploads.

• Support for Python 3.8 has been removed. Now the minimum python version supported is 3.9 .

• Support for running glance services in Windows operating systems has been removed.

Deprecation Notes

• The stores_info_detail policy has deprecated the rule rule:context_is_admin in favor of rule:context_is_admin AND service_roles:service to allow service to service interaction.

Bug Fixes

• Bug 2081009: oslo_config.cfg.NotInitializedError when switching default policy_file in oslo.policy

• Bug 2086675: Suspected performance regression for RBD back end linked to location sorting

Other Notes

• Updated the stores detail API response to include fsid field in RBD store properties.