Re: [PATCH 4/5] nvme: replace GENHD_FL_UP with GENHD_FL_DISK_ADDED

From: Christoph Hellwig <hch@infradead.org>
To: Luis Chamberlain <mcgrof@kernel.org>
Cc: axboe@kernel.dk, hare@suse.de, bvanassche@acm.org,
	ming.lei@redhat.com, hch@infradead.org, jack@suse.cz,
	osandov@fb.com, linux-block@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-nvme@lists.infradead.org,
	sagi@grimberg.me, kbusch@kernel.org
Subject: Re: [PATCH 4/5] nvme: replace GENHD_FL_UP with GENHD_FL_DISK_ADDED
Date: Wed, 21 Jul 2021 06:31:46 +0100	[thread overview]
Message-ID: <YPexQv88hGXxBoLE@infradead.org> (raw)
In-Reply-To: <20210720182048.1906526-5-mcgrof@kernel.org>

On Tue, Jul 20, 2021 at 11:20:47AM -0700, Luis Chamberlain wrote:
> The GENHD_FL_DISK_ADDED flag is what we really want, as the
> flag GENHD_FL_UP could be set on a semi-initialized device.
> 
> Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
> ---
>  drivers/nvme/host/core.c      | 4 ++--
>  drivers/nvme/host/multipath.c | 2 +-
>  2 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> index 11779be42186..7be78491c838 100644
> --- a/drivers/nvme/host/core.c
> +++ b/drivers/nvme/host/core.c
> @@ -1819,7 +1819,7 @@ static void nvme_update_disk_info(struct gendisk *disk,
>  static inline bool nvme_first_scan(struct gendisk *disk)
>  {
>  	/* nvme_alloc_ns() scans the disk prior to adding it */
> -	return !(disk->flags & GENHD_FL_UP);
> +	return !(blk_disk_added(disk));
>  }

So this on is obviously right as it wants to check for the probe
time scan.  Although we don't need the braces.

>
>  
>  static void nvme_set_chunk_sectors(struct nvme_ns *ns, struct nvme_id_ns *id)
> @@ -3823,7 +3823,7 @@ static void nvme_ns_remove(struct nvme_ns *ns)
>  	nvme_mpath_clear_current_path(ns);
>  	synchronize_srcu(&ns->head->srcu); /* wait for concurrent submissions */
>  
> -	if (ns->disk->flags & GENHD_FL_UP) {
> +	if (blk_disk_added(ns->disk)) {

This is something that goes back to before the damn of time, but I do
not think we actually need it.  All the errors paths before alloc_disk
and add_disk just directly free the ns and never end up here.

>  		if (!nvme_ns_head_multipath(ns->head))
>  			nvme_cdev_del(&ns->cdev, &ns->cdev_device);
>  		del_gendisk(ns->disk);
> diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
> index 0ea5298469c3..f77bd2d5c1a9 100644
> --- a/drivers/nvme/host/multipath.c
> +++ b/drivers/nvme/host/multipath.c
> @@ -764,7 +764,7 @@ void nvme_mpath_remove_disk(struct nvme_ns_head *head)
>  {
>  	if (!head->disk)
>  		return;
> -	if (head->disk->flags & GENHD_FL_UP) {
> +	if (blk_disk_added(head->disk)) {
>  		nvme_cdev_del(&head->cdev, &head->cdev_device);
>  		del_gendisk(head->disk);
>  	}

This one is sort of correct due to the lazy disk addition.  But we
could and probably should check for NVME_NSHEAD_DISK_LIVE instead.