spa_min_alloc should be GCD, not min

[ixhamza]

Motivation and Context

Since spa_min_alloc may not be a power of 2, unlike ashifts, in the case of DRAID, we should not select the minimal value among several vdevs. Rounding to a multiple of it is unlikely to work for other vdevs. Instead, using the greatest common divisor produces smaller yet more reasonable results.

How Has This Been Tested?

# Create disks
truncate -s 256m /var/tmp/disk1
truncate -s 256m /var/tmp/disk2
truncate -s 256m /var/tmp/disk3
truncate -s 256m /var/tmp/disk4
truncate -s 256m /var/tmp/disk5

# Create a draid pool with ashift=9
echo 9 > /sys/module/zfs/parameters/vdev_file_logical_ashift
echo 9 > /sys/module/zfs/parameters/vdev_file_physical_ashift
zpool create tank draid1 /var/tmp/disk1 /var/tmp/disk2 /var/tmp/disk3 /var/tmp/disk4 -f

# spa_min_alloc will be 1536 [(vdc->vdc_ndata << vd->vdev_ashift) =>
# (3 << 512) => 1536]

# Add another top-level vdev with ashift=10, which produces min_alloc
# 1024 [(1 << ashift) => (1 << 10) => 1024] 
echo 10 > /sys/module/zfs/parameters/vdev_file_logical_ashift
echo 10 > /sys/module/zfs/parameters/vdev_file_physical_ashift
zpool add tank /var/tmp/disk5 -f

# New spa_min_alloc will now be 512 [gcd(1536, 1024) => 512] instead
# of 1024 [min(1536, 1024)] 

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)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS 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.

[ixhamza]

cc: @amotin

[amotin]

For a note, in #14909 I am rounding ZIL block allocation size to spa_min_alloc, and previous min value could result in allocation bigger than other vdev where allocation finally ended up can efficiently handle.

[amotin]

Thinking more about it, I think we should both keep old min value and just add new gcd. If allocation size is smaller or equal than spa_min_alloc, then it should be set to spa_min_alloc and suffer whatever inefficiency happen on bigger vdev. If allocation is bigger than spa_min_alloc, then we should use roundup(s, spa_gcd_alloc). We should probably add a function/macro implementing this logic and use it in two places where spa_min_alloc is used now.

[ixhamza]

Thank you @amotin for the feedback. Updated.

[ixhamza]

Thanks, @amotin. I will address your feedback.