Man page updates.
Do not clear mp and tmp in the case of EEXISTS in kern_renameat(), they must be unbusied/unlocked as needed.

In D55539#1272294, @arrowd wrote:

The last revision worked for me on ZFS.

I patched Qt to make use of this syscall and verified that QFile::rename implementation now uses the renameat2 code path. What's even more important is that calling QFile::rename() now results in paired IN_MOVED_FROM and IN_MOVED_TO inotify events instead of disconnected IN_CREATE and IN_DELETE. This is a great improvement!

What's worrying me is how the previous revision failed for ZFS. The syscall returned EINVAL and after that the test program entered an uninterruptible busyloop. It was not possible to stop it with kill -KILL and running cat somefile afterwards also resulted in busy-looping. The system quickly degraded into unusable state. It feels like something in kernel does not properly handle the case when AT_RENAME_NOREPLACE is not implemented for a given FS.

In D55610#1272406, @jlduran wrote:

Note that we want to keep the current behavior in makenetvfslist().

Note that we want to keep the current behavior in makenetvfslist().

I'm interested to run collabora on Freebsd, so i downloaded the diff and built the port, and attempted to run collabora in a jail, on FreeBSD 14.3. Here is what I noticed :

Address suggestions.

If you want to fix your patch to the rdpmc problem we should land this and maybe wait for my other fix for the per-package change.

i don't really want this review to turn into a debate over whether people should use RIP. the point is, people who do want to (and i'm aware some such people exist) can continue to do so, even using routed from ports if they really want.

I suppose we could, but it would be a huge diff that changed nothing.

man page

I'm sorry -- it's still too early for my brain to be effective at code reviewing. I'll retract my prior acceptance until we get some clarity on some of the changes (most of the changes going from 1->2 look correct, but I'm not sure about some of the edge cases in skip/readin).

Sorry -- forgot to sign-off in my last submission. An answer shouldn't block this change landing since this change should definitely be made (regardless of how it's expressed) as it makes the code more "functionally correct".

Not a huge fan of complexity, but would it make sense to turn these into an enum in a later revision for human readability? I'm game for making the change after you're done if you're busy with other things.

Can we enable/disable the whole module when one or both of HMP and SMP aren't enabled? I don't see why we need to build this driver when HMP and SMP don't exist.

Remove debug print

Rebase

The last revision worked for me on ZFS.

Address suggestions

Ну то есть изменения в русскоязычных .po вычитывать смысла нет, верно?

Да, po файлы лучше игнорировать, чтобы не тратить на них время. Они записываются в репозиторий, потому что так написано в инструкции - чтобы точно видеть из какой версии перевода сгенерирован этот adoc файл.

Quick response:

Since @markj just announce that he's AFK, could you take a look at this please? @glebius