Fix crypto build on FreeBSD HEAD

[mattmacy]

Update API usage to reflect recent change.

Signed-off-by: Matt Macy mmacy@FreeBSD.org

Motivation and Context

Description

How Has This Been Tested?

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the ZFS on Linux code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[kithrup]

Looks good to me (just went and looked at FreeBSD-HEAD).

[mattmacy]

It looks like a subset of crypto operations are simply broken on HEAD. Operations which succeeded previously now return EBADMSG. Presumably John will fix this promptly - otherwise I'll have to shift ZoF to use ICP - which will need to be pulled in to FreeBSD anyway because ztest on FreeBSD still relies on it because FreeBSD's Open Crypto Framework doesn't run in user.

freebsd_crypt_newsession(0xfffffe00af48f238, { CKM_AES_CCM, 1, 24, aes-192-ccm }, { 1, 0xfffffe00af48f1f0, 192 })
	key = { de 77 50 39 1e 47 b4 6b da 05 5c 72 3d 77 b4 70 7d e4 ba d5 9c 20 b5 eb }
freebsd_crypt_uio_debug_log(encrypt, 0, { CKM_AES_CCM, 1, 24, aes-192-ccm }, 0xfffffe00af48ee60, { 1, 0xfffff800031507c0, 256 }, 0xfffffe00af48efa4, 88, 24)
	key = { 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a }
	iovec #0: <0xfffffe00af48ee90, 24>
	iovec #1: <0xfffffe00af48eff0, 24>
	iovec #2: <0xfffffe00af48efb0, 64>
	iovec #3: <0xfffffe00af48ef90, 16>
freebsd_crypt_newsession(0xfffff80008265f40, { CKM_AES_CCM, 1, 24, aes-192-ccm }, { 1, 0xfffff800031507c0, 256 })
	key = { 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a }
freebsd_crypt_uio_debug_log(decrypt, 0, { CKM_AES_CCM, 1, 24, aes-192-ccm }, 0xfffffe00af48f210, { 1, 0xfffff800031507c0, 256 }, 0xfffffe00af48f384, 88, 24)
	key = { 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a }
	iovec #0: <0xfffffe00af48f240, 24>
	iovec #1: <0xfffff800c20360a8, 24>
	iovec #2: <0xfffff800c20360c8, 64>
	iovec #3: <0xfffffe00af48f370, 16>
freebsd_crypt_newsession(0xfffff800c22b5d80, { CKM_AES_CCM, 1, 24, aes-192-ccm }, { 1, 0xfffff800031507c0, 256 })
	key = { 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a 7a }
freebsd_crypt_uio: returning error 89
panic: VERIFY3(0 == dsl_dataset_hold_obj_flags(pdd->dd_pool, obj, DS_HOLD_FLAG_DECRYPT, FTAG, &ds)) failed (0 == 13)

[mattmacy]

Good news, the panic is fixed in HEAD. Bad news, this week's AMI for HEAD useless for anything but builds.

[codecov-commenter]

Codecov Report

Merging #10384 into master will increase coverage by 0.03%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #10384      +/-   ##
==========================================
+ Coverage   79.56%   79.59%   +0.03%     
==========================================
  Files         391      391              
  Lines      123603   123603              
==========================================
+ Hits        98346    98385      +39     
+ Misses      25257    25218      -39     

Flag	Coverage Δ
#kernel	79.89% <ø> (-0.09%)	⬇️
#user	66.25% <ø> (+0.64%)	⬆️

Impacted Files	Coverage Δ
module/os/linux/spl/spl-zlib.c	55.35% <0.00%> (-28.58%)	⬇️
module/os/linux/spl/spl-kmem-cache.c	75.22% <0.00%> (-9.23%)	⬇️
module/os/linux/zfs/zfs_file_os.c	84.15% <0.00%> (-1.00%)	⬇️
module/zfs/vdev_queue.c	95.52% <0.00%> (-0.90%)	⬇️
lib/libzfs/libzfs_changelist.c	84.37% <0.00%> (-0.79%)	⬇️
module/zfs/vdev_raidz.c	92.48% <0.00%> (-0.77%)	⬇️
cmd/zed/agents/zfs_mod.c	77.55% <0.00%> (-0.67%)	⬇️
module/zfs/vdev_indirect.c	84.83% <0.00%> (-0.50%)	⬇️
module/os/linux/zfs/zfs_znode.c	84.93% <0.00%> (-0.46%)	⬇️
module/zfs/dsl_dataset.c	92.16% <0.00%> (-0.34%)	⬇️
... and 47 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a1ba120...ed06093. Read the comment docs.

[ghost]

I've rolled head back to the previous AMI for now. This PR will break the CI build if merged before the next AMI is available, whenever that will be.

[behlendorf]

@mattmacy @freqlabs thanks for the update. Just let me know when the next AMI is available and we can run this PR through again and verify it resolves the issue. Then merge it.

[ghost]

Unfortunately rolling back the AMI turns out to not be enough. We’re still fetching the latest src archive for the builds, and there aren’t previous versions of the src kept around anywhere afaik.

[ghost]

@behlendorf since the HEAD bots are going to be broken no matter what until the next snapshot, we might as well merge this ASAP so I can update the FreeBSD ports.