The snapshot AMIs built this week do work. Thanks again!

Thank you. I hoped to test this but have been stumbling over how to build an AMI with make release. It's not part of my usual workflow. We'll see if the snapshots this week are fixed!

Please add a comment to explain that calling unregistration() before fetching nfs_minvers from the sysctl is intentional when you commit.

Fixing this here too for older makefs building images from newer sources might be reasonable, but I wonder if makefs should be made a bootstrap tool instead, along with its default being improved.

This appears to be the source of the discrepancy between newfs and makefs:

sbin/newfs/mkfs.c
338:            density = MAX(NFPI, minfragsperinode) * fsize;
341:            density = minfragsperinode * fsize;

I agree with the suggestion this would be better to fix in makefs itself.

@oshogbo okay to commit?

Yep this can go through upstream.

This is still needed but the code has changed so the patch must be adapted.

Yeah that looks like the dance everyone else is doing.

The other places using root_hold_wait try to avoid it first. I'd expect that strategy to work at least for dumpon, though I could imagine zpool import not failing so gracefully. Did you consider it?

I was reading the diff stack wrong. This doesn't depend on flua-libifconfig after all, it's just displayed in the stack because it has a common ancestor: flua-libjail.

Changed the format of lagg ports to a table keyed by the interface name, so rather than

"ports": [
    {
        "lacp_state": [
            "ACTIVITY",
            "AGGREGATION",
            "SYNC",
            "COLLECTING",
            "DISTRIBUTING"
        ],
        "flags": [
            "ACTIVE",
            "COLLECTING",
            "DISTRIBUTING"
        ],
        "laggport": "igb0"
    },
    {
        "lacp_state": [
            "ACTIVITY",
            "AGGREGATION",
            "SYNC",
            "COLLECTING",
            "DISTRIBUTING"
        ],
        "flags": [
            "ACTIVE",
            "COLLECTING",
            "DISTRIBUTING"
        ],
        "laggport": "igb1"
    }
],

we now have

"ports": {
    "igb0": {
        "lacp_state": [
            "ACTIVITY",
            "AGGREGATION",
            "SYNC",
            "COLLECTING",
            "DISTRIBUTING"
        ],
        "flags": [
            "ACTIVE",
            "COLLECTING",
            "DISTRIBUTING"
        ]
    },
    "igb1": {
        "lacp_state": [
            "ACTIVITY",
            "AGGREGATION",
            "SYNC",
            "COLLECTING",
            "DISTRIBUTING"
        ],
        "flags": [
            "ACTIVE",
            "COLLECTING",
            "DISTRIBUTING"
        ]
    }
},

In D25447#609400, @gbe wrote:

From the first look the man page should be fine. Did you run igor and mandoc -Tlint against it?

Avoid tracking stack index manually when we know the position relative to the top. We're usually using the index to set a field on a table immediately under a value on the top of the stack, so the index can be specified as -2 rather than the absolute index.

Apart from the comment nit this looks good. Tested it a bit with io.popen("ps ax") to verify we do get jailed.

Other than nits in the example, this seems to work just fine. I'll also note that making the example executable in the source tree won't make the installed script executable.

The indentation in the example additions is different as well. It would be nice to keep it consistent.

Incorporated changes to libifconfig.

Incorporated feedback.

Incorporated feedback.