RBD Encryption Support

[Greenpepper15]

Summary
This PR adds support for testing Ceph librbd client-side encryption (AES-XTS) by allowing fio to load LUKS keys for an existing image with librbd. After the loading operation any IO to the rbd image will be encrypted/decrypted.

See the Ceph documentation for this feature: https://docs.ceph.com/en/reef/rbd/rbd-encryption/

Motivation
librbd has supported encryption/decryption since Ceph Pacific (16.2.z). However, the fio rbd engine currently lacks the hooks to load encryption context. This prevents users from benchmarking the crypto overhead of this librbd feature.

Implementation Details

• Adds configure configuration to detect if the linked librbd supports rbd_encryption_load.
• Adds two new job options:
  • rbd_encryption_format: (luks1, luks2)
  • rbd_encryption_passphrase: The key/passphrase to unlock the image.
• Uses rbd_encryption_load() during the connection phase to load encryption context.

Notes on Design

• No Auto-Formatting: This patch does not use rbd_encryption_format. It is assumed the user has already provisioned and formatted the image using the Ceph CLI (rbd encryption format ...). This avoids accidental data destruction or re-formatting overhead during benchmark runtime.
  • I could add Auto-Formatting by trying to load the encryption context and formatting if that fails, if you want me to.
• Backward Compatibility: The feature is guarded by CONFIG_RBD_ENCRYPTION. If fio is built against an older Ceph version (pre-Pacific), the options are unavailable, and the engine behaves as before.

Example Job File

[encrypted-test]
ioengine=rbd
pool=bench_pool
rbdname=test_image
# New encryption options
rbd_encryption_format=luks2
rbd_encryption_passphrase=my_secret_passphrase
rw=randwrite
bs=4k

And please review the documentation I wrote. I am unsure if the phrasing and syntax I used is consistent with the rest of the documentation.

[sitsofe]

• I'm fine with your choice not to auto-format the image (but I'm not benchmarking Ceph so perhaps I'm not the target audience)
• Please squash these commits into one - this can be seen as being atomic and can go in all in one go
• You may want to move some of the PR info into the commit message

[Greenpepper15]

Cool all tests passed.
Do you have more suggestions?

[axboe]

@Greenpepper15 Just that minor nit, if you fix it up and amend the commit and force push it, I can pick it up. Rest looks fine to me.

[axboe]

Pulled with a fixup added on top, github doesn't really render things properly and you had various cases of trailing whitespace or incorrect usage of tabs vs spaces.

[Greenpepper15]

Thanks for the merge!
Sorry for the inconvenience. I will be more careful next time.