All of lore.kernel.org
 help / color / mirror / Atom feed
* Can we make the io_poll queue attibute read-only
@ 2021-04-29  7:32 Christoph Hellwig
  2021-04-29 14:43 ` Jens Axboe
  0 siblings, 1 reply; 2+ messages in thread
From: Christoph Hellwig @ 2021-04-29  7:32 UTC (permalink / raw)
  To: Jens Axboe; +Cc: Ming Lei, linux-block

Hi Jens,

while looking at the polling code I wonder if we have to keep supporting
the change of the io_poll attribute and thus the poll flag on a live
queue.  The support to change this goes back to before supporting the
explicit poll queues, and was used as the prime interface to enable
polling.  Now that we use explicit poll queues the driver specific
paramters to set them up work as the prime interface to enable polling.
They can still be changed at runtime, although much more invasively
(i.e. controller reset in nvme).  The upside is that we don't need
to bother with draining the bios and two queue flags, and don't have
a confusing duplicated user interface.

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Can we make the io_poll queue attibute read-only
  2021-04-29  7:32 Can we make the io_poll queue attibute read-only Christoph Hellwig
@ 2021-04-29 14:43 ` Jens Axboe
  0 siblings, 0 replies; 2+ messages in thread
From: Jens Axboe @ 2021-04-29 14:43 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: Ming Lei, linux-block

On 4/29/21 1:32 AM, Christoph Hellwig wrote:
> Hi Jens,
> 
> while looking at the polling code I wonder if we have to keep supporting
> the change of the io_poll attribute and thus the poll flag on a live
> queue.  The support to change this goes back to before supporting the
> explicit poll queues, and was used as the prime interface to enable
> polling.  Now that we use explicit poll queues the driver specific
> paramters to set them up work as the prime interface to enable polling.
> They can still be changed at runtime, although much more invasively
> (i.e. controller reset in nvme).  The upside is that we don't need
> to bother with draining the bios and two queue flags, and don't have
> a confusing duplicated user interface.

I think so, doesn't make much sense anymore. There's no point in
configuring for poll queues and then disabling it through sysfs.

Rather than make it RO and risk breaking apps, why not just ignore the
write?

-- 
Jens Axboe


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-04-29 14:43 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-04-29  7:32 Can we make the io_poll queue attibute read-only Christoph Hellwig
2021-04-29 14:43 ` Jens Axboe

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.