Story #5440

As a user I can disable checksum types like md5

Added by daviddavis about 1 year ago. Updated 12 months ago.

Start date:
Due date:
% Done:


Estimated time:
Platform Release:
Sprint Candidate:



Some users may not be comfortable with the use of certain algorithms that have known vulnerabilities such as md5 and sha1. Also, some security standards discourage use of certain hash algorithms.


Allow users to set a `HASH_ALGORITHMS` setting before they run Pulp for the first time that includes a list of hash algorithms they wish to use. This list should act as a whitelist.

1. Create a constant of the hash algorithms that we support (eg AVAILABLE_HASH_ALGORITHMS) and validate that each HASH_ALGORITHMS item is in AVAILABLE_HASH_ALGORITHMS
2. Validate that sha256 is in HASH_ALGORITHMS. We use this to verify the uniqueness of artifacts.
3. Make all checksum fields on Artifact nullable except sha256
4. Look for places in core and pulp_file that use hashlib and make sure they don't use hash algorithms outside of HASH_ALGORITHMS
5. Test that artifact workflows still work
6. Test (or open an issue) to ensure that pulp_rpm properly supports HASH_ALGORITHMS
7. Document the HASH_ALGORITHMS setting and note that it must be set before Pulp is run for the first time


#1 Updated by bmbouter about 1 year ago

  • Sprint/Milestone changed from 71 to 3.0.0

#2 Updated by daviddavis 12 months ago

  • Description updated (diff)

#3 Updated by daviddavis 12 months ago

  • Description updated (diff)

#4 Updated by bmbouter 12 months ago

This all looks good. Is the first-time requirement coming from the idea that you'll never re-check the data in the db if it were changed later?

#5 Updated by daviddavis 12 months ago

If you change HASH_ALGORITHMS and add another checksum type later, the existing artifacts in the database won't have values for this checksum type. I think we'd have to have some sort of script/migration to fill in these missing values if we allowed users to change HASH_ALGORITHMS.

#6 Updated by bmbouter 12 months ago

I agree and we can for now say that you can't change it on an installation. Is there any other motivation behind that requirement for users for this setting? I can't think of any, just checking for my understanding.

#7 Updated by daviddavis 12 months ago

I haven't really thought through it TBH. Would there be some problems for example if you sync down content in RPM and then remove the checksum type that the content is using?

#8 Updated by bmbouter 12 months ago

  • Sprint/Milestone deleted (3.0.0)

Removing from 3.0 blockers due to no user requesting this and FIPS support not being required for 3.0.

Please register to edit this issue

Also available in: Atom PDF