May be convert this to static function (optionally inline) ?

Assert that entry->object.vm_object == NULL.
Assert locks.

Assert locks. Assert that the entry is not a submap.

Use the opportunity to fix indentation of these two lines.

I prefer to either have {} in both branches for if(), or to not have {} in both.

Do we need this function standalone ? If vm_map_clip_start() do at the very beginning

if(start <= entry->start)
     return;

and then put the body of the _vm_map_clip_start(), it seems to make the code simpler. I believe the old structure was for some quite non-optimizing compiler.

The comment doesn't really make sense anymore; what does "now" mean?

I don't really understand the purpose of lifting this portion out into a separate function: it doesn't eliminate any duplication. I also find it hard to reason about what this function is supposed to do given its name.

Can't the assertion come before the check?

IMO it would be more natural to return the new object from vm_map_entry_attach_backing_object().

I have tried to make it more sensible.

The code appears in both clip_start and clip_end functions, so that duplication is removed.

I could change the name vm_map_entry_find_backing_object, but if I drop "vm_map_entry", Alan will object, and if I drop "backing_object", I'm dropping what the comment that precedes this code block says it's about. It either creates a backing object, or uses one already there. Maybe I can do better than "find"?

Okay.

In one place, I would cast away the returned result. In the other, I would get to drop one assignment line, but I would also need to shorten the function name to fit in 80 columns after doing proper line wrapping.

What about vm_map_entry_clip_object() ?

I'm willing to do it, if that will satisfy everyone, but I don't think that name makes much sense. It seems like the actual clipping depends on a 'start' or 'end' value that isn't passed to this function.

vm_map_entry_charge_object()?

Maybe call it vm_map_entry_back()? That is admittedly terse.

Otherwise I would remove "_backing". I think it's implied.

... implied by virtue of the fact that the function doesn't take an object parameter.

This can't compile.

You did not asserted that the map is locked there.

There is no map here. I could assert something in the caller, or pass a map here so that I could assert that it is locked. In the case of the call from *_fork, I'm not sure which map I would assert as locked.