pbuf_ctor(): Stop using LK_NOWAIT, use LK_NOWITNESS

[rlibby]

Summary

The LK_NOWAIT was added to suppress a witness warning, but LK_NOWITNESS
is more what we mean. This makes pbuf_ctor() more consistent with
buf_alloc(), although, unlike buf_alloc(), for pbuf there should not be
any danger of a wild locker relying on the type stability of the buf to
attempt a lock. That is, this is essentially cosmetic.

Relevant history:

• 531f8cfea06b Use dedicated lock name for pbufs
• 5875b94c7493 buf_alloc(): lock the buffer with LK_NOWAIT
• c9e023541aef pbuf_ctor(): lock the buffer with LK_NOWAIT
• 1fb00c8f1060 buf_alloc(): Stop using LK_NOWAIT, use LK_NOWITNESS

Sponsored by: Dell EMC Isilon

[rew]

I'm fine with this change as it semantically makes sense and mimics original behavior of pbuf_ctor()

...that and LK_NOWAIT, in theory, could return an unhandled error from pbuf_ctor()

[kib]

LK_NOWAIT is the common pattern for this situation. Besides buffers, another immediate example is the vm_map lock in vmspace_fork() (look for the new_map locking and comment).

[rlibby]

In D45360#1034889, @kib wrote:

LK_NOWAIT is the common pattern for this situation. Besides buffers, another immediate example is the vm_map lock in vmspace_fork() (look for the new_map locking and comment).

Thanks for the pointer. Yeah, that case makes sense. We don't really have a more expressive API there like we do with lockmgr's flags. I wonder if we might want a way to say explicitly that this is a lock that no one else can know about, e.g. LK_INITIAL, that would avoid reporting a LOR when taken second, but also can KASSERT that the lock was not contended.

[kib]

In D45360#1034898, @rlibby wrote:

In D45360#1034889, @kib wrote:

LK_NOWAIT is the common pattern for this situation. Besides buffers, another immediate example is the vm_map lock in vmspace_fork() (look for the new_map locking and comment).

Thanks for the pointer. Yeah, that case makes sense. We don't really have a more expressive API there like we do with lockmgr's flags. I wonder if we might want a way to say explicitly that this is a lock that no one else can know about, e.g. LK_INITIAL, that would avoid reporting a LOR when taken second, but also can KASSERT that the lock was not contended.

This is an interesting proposal, I like it.