Illumos #1948 (expandsize property) and #2703 (zfs send progress)

[mmatuska]

#2703 seems to work very well (zfs send progress reporting)

I need information about #1948 (zpool_reopen) if this can be utilized in Linux in vdev_disk.c (TODO part)
And also the expandsize is currently reporting 16E on my system.

Here is the illumos implementation:
https://hg.openindiana.org/upstream/illumos/illumos-gate/diff/3411fd5f1589/usr/src/uts/common/fs/zfs/vdev_disk.c
illumos/illumos-gate@4263d13#usr/src/uts/common/fs/zfs/vdev_disk.c

[behlendorf]

@mmatuska Sorry, I've been busy. Anyway, with regards to max_psize it should be entirely safe to simply do *max_psize = *psize. However, your going to need to do it near the end of the function after *psize = bdev_capacity(bdev); or it's going to be undefined. That could pretty easily explain your 16E values.

With significantly more effort we could have it report the possible expansion size. Right now it simply returns the size of the partition, and when we do expansion we re-read the partition table and use the updated size. We could probably find a way to calculate this available space without actually re-reading the partition table.

However, I don't think we should worry about it for now.

[mmatuska]

I try to stay to be close to illumos as much as possible - this means the less intervention the better.

Btw. the zfs_send progress was relatively easy to port 181f94e
The only "major" change was that in Linux I had to compare the group_leader processes instead of the individual threads in zfs_ioctl.c. In FreeBSD our SPL is done quite well so I have to change almost nothing . Today I have imported the newest empty_bpobj zpool feature to FreeBSD -HEAD.

[behlendorf]

I try to stay to be close to illumos as much as possible - this means the less intervention the better.

Yes, I agree. The comment was more for the Illumos guys, I need to remember to bring it up when I see them next. If they are going to act as ZFS upstream they should keep this sort of thing in mind.

The only "major" change was that in Linux...

I haven't looked carefully but was this because there was a missing interface in the SPL. If so, probably the right thing to do is add it so we can stay as close to the Illumos code as possible.

[behlendorf]

Squashed in to the following two commits and merged

37abac6 Illumos #2703: add mechanism to report ZFS send progress
1bd201e Illumos #1948: zpool list should show more detailed pool info

[mmatuska]

The code is now ready to import the feature flags commits with minimal conflicts