Re: [PATCH] block: loop: remove & invalidate partitions when detaching

Re: [PATCH] block: loop: remove & invalidate partitions when detaching

[Christoph Hellwig]

On Thu, Jan 04, 2018 at 06:39:24PM +0800, Ming Lei wrote:
> When detaching loop disk, neither we remove loop partitions, nor
> invalidate them. This way may cause data loss, and often confuse
> people.
> 
> This patch fixes the above issue by removing & invalidating loop
> partitions in loop_clr_fd(),

Please add a comment explaining the call (e.g. the weird loop
case of keeping unbound devices around).

Also dasd seems to implement this functionality by iterating over
all partitions and then calling ioctl_by_bdev(bdev, BLKPG, ..) on
it.  Please switch it to you new helper assuming there is a good
reason to even do this to start with.  CCing the maintainers to
confirm that.

And while you'e at it split the block and loop bits into separate
patches.

> +void del_gendisk_partitions(struct gendisk *disk)

Add a comment explanining the function, please.

>  	struct disk_part_iter piter;
>  	struct hd_struct *part;
>  
> -	blk_integrity_del(disk);
> -	disk_del_events(disk);
> -
>  	/* invalidate stuff */
>  	disk_part_iter_init(&piter, disk,
>  			     DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
> @@ -714,7 +711,17 @@ void del_gendisk(struct gendisk *disk)
>  	disk_part_iter_exit(&piter);
>  
>  	invalidate_partition(disk, 0);
> +}
> +EXPORT_SYMBOL(del_gendisk_partitions);

EXPORT_SYMBOL_GPL, please.

Re: [PATCH] block: loop: remove & invalidate partitions when detaching

[Ming Lei]

On Thu, Jan 04, 2018 at 04:38:03AM -0800, Christoph Hellwig wrote:
> On Thu, Jan 04, 2018 at 06:39:24PM +0800, Ming Lei wrote:
> > When detaching loop disk, neither we remove loop partitions, nor
> > invalidate them. This way may cause data loss, and often confuse
> > people.
> > 
> > This patch fixes the above issue by removing & invalidating loop
> > partitions in loop_clr_fd(),
> 
> Please add a comment explaining the call (e.g. the weird loop
> case of keeping unbound devices around).

OK.

> 
> Also dasd seems to implement this functionality by iterating over
> all partitions and then calling ioctl_by_bdev(bdev, BLKPG, ..) on
> it.  Please switch it to you new helper assuming there is a good
> reason to even do this to start with.  CCing the maintainers to
> confirm that.

This patch doesn't consider to hold bdev lock, and loop's case can
be a bit complicated since loop_clr_fd() can be called in release
path. I will investigate a bit if it is possible to figure out a
general helper to cover both.

> 
> And while you'e at it split the block and loop bits into separate
> patches.

OK.

> 
> > +void del_gendisk_partitions(struct gendisk *disk)
> 
> Add a comment explanining the function, please.

OK.

> 
> >  	struct disk_part_iter piter;
> >  	struct hd_struct *part;
> >  
> > -	blk_integrity_del(disk);
> > -	disk_del_events(disk);
> > -
> >  	/* invalidate stuff */
> >  	disk_part_iter_init(&piter, disk,
> >  			     DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
> > @@ -714,7 +711,17 @@ void del_gendisk(struct gendisk *disk)
> >  	disk_part_iter_exit(&piter);
> >  
> >  	invalidate_partition(disk, 0);
> > +}
> > +EXPORT_SYMBOL(del_gendisk_partitions);
> 
> EXPORT_SYMBOL_GPL, please.

OK.

-- 
Ming