AFAIK this is still a problem

Sorry to bother people, but could we revive this review? I was recently talking to some people who noticed this the hard way and it would be great to get something merged that fixes it, even if it isn't perfect. I think this looks good enough and is definitely not "the ugliest kludge to make this work"

I see that bootconfig used to have code for adding ESPs to all ZFS/UFS boot disks, but it was removed in https://reviews.freebsd.org/D28897. @imp reviewed that one, maybe you could share your opinion here if we should bring parts of that code back? I think all disks that are configured as boot disks by the installer should a working ESP after installation. If the wrong boot disks die then you are left with a system without any ESPs at all.

Bump

Updated diff to include context

So I experimented a little and remade this as "geom -c <consumer>". But unless I figure out a good way to filter by class it is useless. Right now it will just pick whatever class the provider first appears in, like DEV. So the proper way, IMO, to do that would be "geom <class> which <consumer>" and not just add another flag. But then "which" would only exist when using "geom" directly and not when using "gmultipath"/"gpart"/"glabel" binaries which I think would look weird. What are your thoughts?

In D37309#847728, @mav wrote:

Not that I have specific objections against this, but I think it would be much better to implement a generic way to print all consumers connected to specified GEOM provider. This code does not look anyhow specific to gmultipath.