Re: [PATCH 2/3] nvme-multipath: cannot disconnect controller on stuck partition scan

[Christoph Hellwig <hch@lst.de>]

On Wed, Oct 09, 2024 at 11:32:32AM -0600, Keith Busch wrote:
> On Wed, Oct 09, 2024 at 08:23:45AM +0200, Hannes Reinecke wrote:
> > With my testcase _all_ paths return NS_NOT_READY during partition scan, so
> > I/O is constantly bounced between paths, and partition scan never returns.
> > Doesn't matter where you call it, it's stuck.
> 
> Assuming you mean a differet status, like NVME_SC_INTERNAL_PATH_ERROR, I
> can recreate a stuck scan. Let me get one more shot at fixing this by
> suppressing the scan to another context. This one is successful in my
> tests:
> 
> ---
> diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
> index 48e7a8906d012..e1fd53ff2c0ca 100644
> --- a/drivers/nvme/host/multipath.c
> +++ b/drivers/nvme/host/multipath.c
> @@ -580,6 +580,20 @@ static int nvme_add_ns_head_cdev(struct nvme_ns_head *head)
>  	return ret;
>  }
>  
> +static void nvme_scan_work(struct work_struct *work)

scan_work feels a little too generic.  Maybe part_scan_work?

Otherwise not pretty, but by far the best idea so far, so I guess should
go with it.  Can you also add a big fat comment explaining the rationale
for the actual submission?