On 4/10/23 00:58, Damien Le Moal wrote:
On 4/8/23 08:58, Bart Van Assche wrote:
Prepare for processing the requeue list from inside __blk_mq_run_hw_queue().
With such short comment, it is hard to see exactly what this patch is trying to
do. The first part seems to be adding debugfs stuff, which I think is fine, but
should be its own patch. The second part move the requeue work from per qeue to
per hctx as I understand it. Why ? Can you explain that here ?
Hi Damien,
This patch splits the request queue requeue list into one requeue list
per hardware queue. I split the requeue list because I do not want that
the next patch in this series would affect performance negatively if no
I/O scheduler has been attached. I will double check whether this patch
is really necessary.
Thanks,
Bart.
