Add missing zfs_refcount_destroy() in key_mapping_rele()

[gamanakis]

Motivation and Context

Testing #10228 with reference_tracking_enable=TRUE revealed a missing zfs_refcount_destroy() in key_mapping_rele().

Signed-off-by: George Amanakis gamanakis@gmail.com

Description

Testing #10228 with reference tracking enabled revealed a panic when running persist_l2arc_002_pos.ksh test and then unloading the zfs module. This suggested that the panic had to do with dataset encryption. Looking at the places where zfs_refcount_create is invoked, dsl_crypt.c came up. Upon looking into it the missing zfs_refcount_destroy was found.

Without the proposed fix, unloading of the zfs modules after mounting/unmounting an encrypted dataset with reference tracking enabled causes a panic:

Call Trace:
dump_stack+0x66/0x90
slab_err+0xcd/0xf2
? __kmalloc+0x174/0x260
? __kmem_cache_shutdown+0x158/0x240
__kmem_cache_shutdown.cold+0x1d/0x115
shutdown_cache+0x11/0x140
kmem_cache_destroy+0x210/0x230
spl_kmem_cache_destroy+0x122/0x3e0 [spl]
zfs_refcount_fini+0x11/0x20 [zfs]
spa_fini+0x4b/0x120 [zfs]
zfs_kmod_fini+0x6b/0xa0 [zfs]
_fini+0xa/0x68c [zfs]
__x64_sys_delete_module+0x19c/0x2b0
do_syscall_64+0x5b/0x1a0
entry_SYSCALL_64_after_hwframe+0x44/0xa9

How Has This Been Tested?

Mounting/unmounting an encrypted dataset with reference tracking enabled and the proposed fix does not lead to a panic when unloading the zfs module.

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.

[codecov-io]

Codecov Report

Merging #10246 into master will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master   #10246      +/-   ##
==========================================
+ Coverage   79.34%   79.35%   +0.01%     
==========================================
  Files         389      389              
  Lines      123393   123394       +1     
==========================================
+ Hits        97902    97923      +21     
+ Misses      25491    25471      -20     

Flag	Coverage Δ
#kernel	79.86% <100.00%> (-0.12%)	⬇️
#user	65.37% <100.00%> (+0.03%)	⬆️

Impacted Files	Coverage Δ
module/zfs/dsl_crypt.c	87.52% <100.00%> (+0.01%)	⬆️
module/os/linux/spl/spl-kmem-cache.c	75.22% <0.00%> (-9.59%)	⬇️
module/lua/lmem.c	83.33% <0.00%> (-4.17%)	⬇️
module/zfs/vdev_removal.c	94.36% <0.00%> (-2.54%)	⬇️
module/zfs/vdev_indirect_mapping.c	96.61% <0.00%> (-2.42%)	⬇️
module/zfs/dsl_scan.c	83.79% <0.00%> (-1.61%)	⬇️
module/zfs/spa_log_spacemap.c	93.40% <0.00%> (-0.86%)	⬇️
module/zfs/dmu_traverse.c	96.34% <0.00%> (-0.67%)	⬇️
module/zfs/vdev_queue.c	95.71% <0.00%> (-0.62%)	⬇️
module/zfs/space_map.c	92.82% <0.00%> (-0.50%)	⬇️
... and 41 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 a84c92f...c6d411d. Read the comment docs.

[behlendorf]

Thanks for running this down. Looks good, @tcaputi would you mind reviewing this as well.

[tcaputi]

Good catch. Thanks for finding this