On 18/06/2025 06:10, Christoph Hellwig wrote:
On Tue, Jun 17, 2025 at 01:02:16PM +0100, John Garry wrote:
On 17/06/2025 11:52, Christoph Hellwig wrote:
The extra bdev_ is weird, so drop it. The maximum size is based on the
bdev hardware limits, so add a hw_ component instead.
but the min is also based on hw limits, no?
Yes.
I also note that we have request queue limits atomic_write_unit_max and
atomic_write_hw_unit_max
and bt_awu_max_hw is written with request queue limit atomic_write_unit_max
But I don't think that this will cause confusion.
Should we switch to the request_queue names instead of the nvme
spec names here entirely?
I think what you suggest is fine (with bt_awu_{min,max}_hw). It's
concise, while the request_queue names are a bit wordy. Maybe even the
_hw can be dropped (from bt_awu_{min,max}_hw) and a comment added to
their definition in xfs_buftarg.
