Out of bound when creating per-algo measurement list interfaces

[robertosassu]

Current patch: https://marc.info/?l=linux-integrity&m=176952677931446&w=2

If a TPM algorithm is not supported the PCR bank info is initialized with HASH_ALGO__LAST, which passed to hash_algo_name[] causes an out of bound.

Instead of skipping unknown TPM algorithm, create a measurement interface for it with name tpm_algo_.

Pending TODOs:

• Improve the commit message
• Make the file name in lower case

[robertosassu]

Asked the author to send the new version of the patch.

[robertosassu]

V5 of the patch (https://marc.info/?l=linux-integrity&m=177186135608747&w=2) had the problem that the template digest algorithm can be not among allocated PCR banks, thus we cannot directly take the digest size from the PCR bank info.

I made a patch to add the digest_size field to the ima_algo_desc structure, taken from the appropriate source, and used it to show measurements:

https://marc.info/?l=linux-integrity&m=177202677128752&w=2

I then applied Dmitry's patch on top of mine:

6efbd2b

Tests results:

https://github.com/robertosassu/ima-evm-utils-upstream/actions/runs/22397775765

[mimizohar]

The patch creates a measurement list for TPM algorithms, but unknown or unlisted kernel algorithms. Fix the subject line ["ima_fs: Avoid creating measurement lists for unsupported hash algos"] to match the patch. Also include a link to the TCG document defining the algorithm ids used for in the patch for naming the measurement list.

[robertosassu]

Reviewed and tested v6.

[mimizohar]

Queued in next-integrity/next-integrity-testing to be upstreamed.