[RFC PATCH 0/1] nvme-pci: detect I/O queue depth changes after reset

[Christoph Hellwig]

On Wed, May 27, 2026 at 09:03:02PM -0600, Keith Busch wrote:
> On Wed, May 27, 2026 at 03:19:51PM +0200, Christoph Hellwig wrote:
> > On Wed, May 27, 2026 at 03:53:19PM +0800, guzebing wrote:
> > > This RFC instead takes the smaller approach of detecting the reset-time
> > > CAP.MQES change and making it visible.  If the live I/O queue depth
> > > shrinks, reset recovery is failed before recreating I/O queues.  If it
> > > grows, the driver warns and continues with the existing queue resources.
> > 
> > Unlike the other version this at least sounds doable without creating
> > a complete mess.  So if we can live with this version that'd make me
> > much happier.
> 
> Abandoning the device here sounds pretty harsh when blk-mq already
> supports user space decreasing the q-depth on a live queue. Why can't
> the driver do the same thing?

Reducing capabilities with a firmware upgrade is a losing proposition,
there's just way too many things that can go wrong.

And reducing the queue depth is one of the haіriest operation in
blk-mq.  I wish we had never supported it, but allowing it with a
remove trigger sounds really bad.