DM-Devel Archive on lore.kernel.org
 help / color / Atom feed
From: Jan Kara <jack@suse.cz>
To: Christoph Hellwig <hch@lst.de>
Cc: Jens Axboe <axboe@kernel.dk>, Jan Kara <jack@suse.cz>,
	Mike Snitzer <snitzer@redhat.com>,
	linux-mm@kvack.org,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Jan Kara <jack@suse.com>, Josef Bacik <josef@toxicpanda.com>,
	Coly Li <colyli@suse.de>,
	linux-block@vger.kernel.org, linux-fsdevel@vger.kernel.org,
	dm-devel@redhat.com, linux-mtd@lists.infradead.org,
	Johannes Thumshirn <johannes.thumshirn@wdc.com>,
	Tejun Heo <tj@kernel.org>,
	linux-bcache@vger.kernel.org
Subject: Re: [dm-devel] [PATCH 20/44] block: refactor blkdev_get
Date: Thu, 26 Nov 2020 15:55:21 +0100
Message-ID: <20201126145521.GG422@quack2.suse.cz> (raw)
In-Reply-To: <20201126130422.92945-21-hch@lst.de>

On Thu 26-11-20 14:03:58, Christoph Hellwig wrote:
> Move more code that is only run on the outer open but not the open of
> the underlying whole device when opening a partition into blkdev_get,
> which leads to a much easier to follow structure.
> 
> This allows to simplify the disk and module refcounting so that one
> reference is held for each open, similar to what we do with normal
> file operations.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> Acked-by: Tejun Heo <tj@kernel.org>

The patch looks good to me. You can add:

Reviewed-by: Jan Kara <jack@suse.cz>

								Honza


> ---
>  fs/block_dev.c | 185 +++++++++++++++++++++++--------------------------
>  1 file changed, 86 insertions(+), 99 deletions(-)
> 
> diff --git a/fs/block_dev.c b/fs/block_dev.c
> index 41c50cfba864e2..86a61a2141f642 100644
> --- a/fs/block_dev.c
> +++ b/fs/block_dev.c
> @@ -1403,46 +1403,12 @@ EXPORT_SYMBOL_GPL(bdev_disk_changed);
>   *  mutex_lock(part->bd_mutex)
>   *    mutex_lock_nested(whole->bd_mutex, 1)
>   */
> -
> -static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
> -		int for_part)
> +static int __blkdev_get(struct block_device *bdev, struct gendisk *disk,
> +		int partno, fmode_t mode)
>  {
> -	struct block_device *whole = NULL, *claiming = NULL;
> -	struct gendisk *disk;
>  	int ret;
> -	int partno;
> -	bool first_open = false, unblock_events = true, need_restart;
> -
> - restart:
> -	need_restart = false;
> -	ret = -ENXIO;
> -	disk = bdev_get_gendisk(bdev, &partno);
> -	if (!disk)
> -		goto out;
> -
> -	if (partno) {
> -		whole = bdget_disk(disk, 0);
> -		if (!whole) {
> -			ret = -ENOMEM;
> -			goto out_put_disk;
> -		}
> -	}
>  
> -	if (!for_part && (mode & FMODE_EXCL)) {
> -		WARN_ON_ONCE(!holder);
> -		if (whole)
> -			claiming = whole;
> -		else
> -			claiming = bdev;
> -		ret = bd_prepare_to_claim(bdev, claiming, holder);
> -		if (ret)
> -			goto out_put_whole;
> -	}
> -
> -	disk_block_events(disk);
> -	mutex_lock_nested(&bdev->bd_mutex, for_part);
>  	if (!bdev->bd_openers) {
> -		first_open = true;
>  		bdev->bd_disk = disk;
>  		bdev->bd_contains = bdev;
>  		bdev->bd_partno = partno;
> @@ -1454,15 +1420,8 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
>  				goto out_clear;
>  
>  			ret = 0;
> -			if (disk->fops->open) {
> +			if (disk->fops->open)
>  				ret = disk->fops->open(bdev, mode);
> -				/*
> -				 * If we lost a race with 'disk' being deleted,
> -				 * try again.  See md.c
> -				 */
> -				if (ret == -ERESTARTSYS)
> -					need_restart = true;
> -			}
>  
>  			if (!ret) {
>  				bd_set_nr_sectors(bdev, get_capacity(disk));
> @@ -1482,14 +1441,23 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
>  			if (ret)
>  				goto out_clear;
>  		} else {
> -			BUG_ON(for_part);
> -			ret = __blkdev_get(whole, mode, NULL, 1);
> -			if (ret)
> +			struct block_device *whole = bdget_disk(disk, 0);
> +
> +			mutex_lock_nested(&whole->bd_mutex, 1);
> +			ret = __blkdev_get(whole, disk, 0, mode);
> +			if (ret) {
> +				mutex_unlock(&whole->bd_mutex);
> +				bdput(whole);
>  				goto out_clear;
> -			bdev->bd_contains = bdgrab(whole);
> +			}
> +			whole->bd_part_count++;
> +			mutex_unlock(&whole->bd_mutex);
> +
> +			bdev->bd_contains = whole;
>  			bdev->bd_part = disk_get_part(disk, partno);
>  			if (!(disk->flags & GENHD_FL_UP) ||
>  			    !bdev->bd_part || !bdev->bd_part->nr_sects) {
> +				__blkdev_put(whole, mode, 1);
>  				ret = -ENXIO;
>  				goto out_clear;
>  			}
> @@ -1509,58 +1477,17 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
>  			    (!ret || ret == -ENOMEDIUM))
>  				bdev_disk_changed(bdev, ret == -ENOMEDIUM);
>  			if (ret)
> -				goto out_unlock_bdev;
> +				return ret;
>  		}
>  	}
>  	bdev->bd_openers++;
> -	if (for_part)
> -		bdev->bd_part_count++;
> -	if (claiming)
> -		bd_finish_claiming(bdev, claiming, holder);
> -
> -	/*
> -	 * Block event polling for write claims if requested.  Any write holder
> -	 * makes the write_holder state stick until all are released.  This is
> -	 * good enough and tracking individual writeable reference is too
> -	 * fragile given the way @mode is used in blkdev_get/put().
> -	 */
> -	if (claiming && (mode & FMODE_WRITE) && !bdev->bd_write_holder &&
> -	    (disk->flags & GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE)) {
> -		bdev->bd_write_holder = true;
> -		unblock_events = false;
> -	}
> -	mutex_unlock(&bdev->bd_mutex);
> -
> -	if (unblock_events)
> -		disk_unblock_events(disk);
> -
> -	/* only one opener holds refs to the module and disk */
> -	if (!first_open)
> -		put_disk_and_module(disk);
> -	if (whole)
> -		bdput(whole);
>  	return 0;
>  
>   out_clear:
>  	disk_put_part(bdev->bd_part);
>  	bdev->bd_disk = NULL;
>  	bdev->bd_part = NULL;
> -	if (bdev != bdev->bd_contains)
> -		__blkdev_put(bdev->bd_contains, mode, 1);
>  	bdev->bd_contains = NULL;
> - out_unlock_bdev:
> -	if (claiming)
> -		bd_abort_claiming(bdev, claiming, holder);
> -	mutex_unlock(&bdev->bd_mutex);
> -	disk_unblock_events(disk);
> - out_put_whole:
> - 	if (whole)
> -		bdput(whole);
> - out_put_disk:
> -	put_disk_and_module(disk);
> -	if (need_restart)
> -		goto restart;
> - out:
>  	return ret;
>  }
>  
> @@ -1585,7 +1512,12 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
>   */
>  static int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
>  {
> -	int ret, perm = 0;
> +	struct block_device *claiming;
> +	bool unblock_events = true;
> +	struct gendisk *disk;
> +	int perm = 0;
> +	int partno;
> +	int ret;
>  
>  	if (mode & FMODE_READ)
>  		perm |= MAY_READ;
> @@ -1595,13 +1527,67 @@ static int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
>  	if (ret)
>  		goto bdput;
>  
> -	ret =__blkdev_get(bdev, mode, holder, 0);
> -	if (ret)
> +	/*
> +	 * If we lost a race with 'disk' being deleted, try again.  See md.c.
> +	 */
> +retry:
> +	ret = -ENXIO;
> +	disk = bdev_get_gendisk(bdev, &partno);
> +	if (!disk)
>  		goto bdput;
> -	return 0;
>  
> +	if (mode & FMODE_EXCL) {
> +		WARN_ON_ONCE(!holder);
> +	
> +		ret = -ENOMEM;
> +		claiming = bdget_disk(disk, 0);
> +		if (!claiming)
> +			goto put_disk;
> +		ret = bd_prepare_to_claim(bdev, claiming, holder);
> +		if (ret)
> +			goto put_claiming;
> +	}
> +
> +	disk_block_events(disk);
> +
> +	mutex_lock(&bdev->bd_mutex);
> +	ret =__blkdev_get(bdev, disk, partno, mode);
> +	if (!(mode & FMODE_EXCL)) {
> +		; /* nothing to do here */
> +	} else if (ret) {
> +		bd_abort_claiming(bdev, claiming, holder);
> +	} else {
> +		bd_finish_claiming(bdev, claiming, holder);
> +
> +		/*
> +		 * Block event polling for write claims if requested.  Any write
> +		 * holder makes the write_holder state stick until all are
> +		 * released.  This is good enough and tracking individual
> +		 * writeable reference is too fragile given the way @mode is
> +		 * used in blkdev_get/put().
> +		 */
> +		if ((mode & FMODE_WRITE) && !bdev->bd_write_holder &&
> +		    (disk->flags & GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE)) {
> +			bdev->bd_write_holder = true;
> +			unblock_events = false;
> +		}
> +	}
> +	mutex_unlock(&bdev->bd_mutex);
> +
> +	if (unblock_events)
> +		disk_unblock_events(disk);
> +
> +put_claiming:
> +	if (mode & FMODE_EXCL)
> +		bdput(claiming);
> +put_disk:
> +	if (ret)
> +		put_disk_and_module(disk);
> +	if (ret == -ERESTARTSYS)
> +		goto retry;
>  bdput:
> -	bdput(bdev);
> +	if (ret)
> +		bdput(bdev);
>  	return ret;
>  }
>  
> @@ -1749,8 +1735,6 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
>  		if (bdev_is_partition(bdev))
>  			victim = bdev->bd_contains;
>  		bdev->bd_contains = NULL;
> -
> -		put_disk_and_module(disk);
>  	} else {
>  		if (!bdev_is_partition(bdev) && disk->fops->release)
>  			disk->fops->release(disk, mode);
> @@ -1763,6 +1747,8 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
>  
>  void blkdev_put(struct block_device *bdev, fmode_t mode)
>  {
> +	struct gendisk *disk = bdev->bd_disk;
> +
>  	mutex_lock(&bdev->bd_mutex);
>  
>  	if (mode & FMODE_EXCL) {
> @@ -1791,7 +1777,7 @@ void blkdev_put(struct block_device *bdev, fmode_t mode)
>  		 * unblock evpoll if it was a write holder.
>  		 */
>  		if (bdev_free && bdev->bd_write_holder) {
> -			disk_unblock_events(bdev->bd_disk);
> +			disk_unblock_events(disk);
>  			bdev->bd_write_holder = false;
>  		}
>  	}
> @@ -1801,11 +1787,12 @@ void blkdev_put(struct block_device *bdev, fmode_t mode)
>  	 * event.  This is to ensure detection of media removal commanded
>  	 * from userland - e.g. eject(1).
>  	 */
> -	disk_flush_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE);
> +	disk_flush_events(disk, DISK_EVENT_MEDIA_CHANGE);
>  
>  	mutex_unlock(&bdev->bd_mutex);
>  
>  	__blkdev_put(bdev, mode, 0);
> +	put_disk_and_module(disk);
>  }
>  EXPORT_SYMBOL(blkdev_put);
>  
> -- 
> 2.29.2
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel


  reply index

Thread overview: 89+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-11-26 13:03 [dm-devel] merge struct block_device and struct hd_struct v3 Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 01/44] blk-cgroup: fix a hd_struct leak in blkcg_fill_root_iostats Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 02/44] filemap: consistently use ->f_mapping over ->i_mapping Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 03/44] fs: remove get_super_thawed and get_super_exclusive_thawed Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 04/44] fs: simplify freeze_bdev/thaw_bdev Christoph Hellwig
2020-11-26 14:13   ` Jan Kara
2020-11-26 13:03 ` [dm-devel] [PATCH 05/44] mtip32xx: remove the call to fsync_bdev on removal Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 06/44] zram: do not call set_blocksize Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 07/44] loop: " Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 08/44] dm: simplify flush_bio initialization in __send_empty_flush Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 09/44] dm: remove the block_device reference in struct mapped_device Christoph Hellwig
2020-11-27 10:36   ` Hannes Reinecke
2020-11-26 13:03 ` [dm-devel] [PATCH 10/44] block: remove a duplicate __disk_get_part prototype Christoph Hellwig
2020-11-27 10:36   ` Hannes Reinecke
2020-11-26 13:03 ` [dm-devel] [PATCH 11/44] block: remove a superflous check in blkpg_do_ioctl Christoph Hellwig
2020-11-27 10:37   ` Hannes Reinecke
2020-11-26 13:03 ` [dm-devel] [PATCH 12/44] block: add a bdev_kobj helper Christoph Hellwig
2020-11-26 14:47   ` David Sterba
2020-11-27 12:03   ` Hannes Reinecke
2020-11-26 13:03 ` [dm-devel] [PATCH 13/44] block: use disk_part_iter_exit in disk_part_iter_next Christoph Hellwig
2020-11-27 12:04   ` Hannes Reinecke
2020-11-26 13:03 ` [dm-devel] [PATCH 14/44] block: use put_device in put_disk Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 15/44] block: change the hash used for looking up block devices Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 16/44] init: refactor name_to_dev_t Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 17/44] init: refactor devt_from_partuuid Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 18/44] init: cleanup match_dev_by_uuid and match_dev_by_label Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 19/44] block: refactor __blkdev_put Christoph Hellwig
2020-11-26 13:03 ` [dm-devel] [PATCH 20/44] block: refactor blkdev_get Christoph Hellwig
2020-11-26 14:55   ` Jan Kara [this message]
2020-11-26 13:03 ` [dm-devel] [PATCH 21/44] block: move bdput() to the callers of __blkdev_get Christoph Hellwig
2020-11-26 15:05   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 22/44] block: opencode devcgroup_inode_permission Christoph Hellwig
2020-11-26 15:07   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 23/44] block: remove i_bdev Christoph Hellwig
2020-11-26 15:45   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 24/44] block: simplify bdev/disk lookup in blkdev_get Christoph Hellwig
2020-11-26 16:33   ` Jan Kara
2020-11-26 17:42     ` Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 25/44] block: remove ->bd_contains Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 26/44] block: simplify the block device claiming interface Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 27/44] block: simplify part_to_disk Christoph Hellwig
2020-11-26 16:36   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 28/44] block: initialize struct block_device in bdev_alloc Christoph Hellwig
2020-11-26 16:37   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 29/44] block: remove the nr_sects field in struct hd_struct Christoph Hellwig
2020-11-26 16:50   ` Jan Kara
2020-11-26 17:52     ` Christoph Hellwig
2020-11-26 18:04       ` Jan Kara
2020-11-27  9:41         ` Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 30/44] block: move disk stat accounting to struct block_device Christoph Hellwig
2020-11-26 16:53   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 31/44] block: move the start_sect field " Christoph Hellwig
2020-11-26 16:56   ` Jan Kara
2020-11-26 18:02     ` Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 32/44] block: move the partition_meta_info " Christoph Hellwig
2020-11-26 16:58   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 33/44] block: move holder_dir " Christoph Hellwig
2020-11-26 17:00   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 34/44] block: move make_it_fail " Christoph Hellwig
2020-11-26 17:01   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 35/44] block: move the policy field " Christoph Hellwig
2020-11-26 17:03   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 36/44] block: allocate struct hd_struct as part of struct bdev_inode Christoph Hellwig
2020-11-26 17:35   ` Jan Kara
2020-11-26 18:00     ` Jan Kara
2020-11-26 18:02       ` Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 37/44] block: switch partition lookup to use struct block_device Christoph Hellwig
2020-11-26 17:45   ` Jan Kara
2020-11-26 18:22   ` Jan Kara
2020-11-27  9:48     ` Christoph Hellwig
2020-11-27 12:45       ` Jan Kara
2020-11-27 15:36         ` Christoph Hellwig
2020-11-27 17:26           ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 38/44] block: remove the partno field from struct hd_struct Christoph Hellwig
2020-11-26 17:46   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 39/44] block: pass a block_device to blk_alloc_devt Christoph Hellwig
2020-11-26 17:47   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 40/44] block: pass a block_device to invalidate_partition Christoph Hellwig
2020-11-26 17:49   ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 41/44] block: switch disk_part_iter_* to use a struct block_device Christoph Hellwig
2020-11-27 12:53   ` Jan Kara
2020-11-27 15:24     ` Christoph Hellwig
2020-11-27 17:30       ` Jan Kara
2020-11-26 13:04 ` [dm-devel] [PATCH 42/44] f2fs: remove a few bd_part checks Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 43/44] block: merge struct block_device and struct hd_struct Christoph Hellwig
2020-11-27 13:19   ` Jan Kara
2020-11-27 15:47     ` Christoph Hellwig
2020-11-26 13:04 ` [dm-devel] [PATCH 44/44] block: stop using bdget_disk for partition 0 Christoph Hellwig
2020-11-27 13:22   ` Jan Kara

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201126145521.GG422@quack2.suse.cz \
    --to=jack@suse.cz \
    --cc=axboe@kernel.dk \
    --cc=colyli@suse.de \
    --cc=dm-devel@redhat.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=hch@lst.de \
    --cc=jack@suse.com \
    --cc=johannes.thumshirn@wdc.com \
    --cc=josef@toxicpanda.com \
    --cc=linux-bcache@vger.kernel.org \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux-mtd@lists.infradead.org \
    --cc=snitzer@redhat.com \
    --cc=tj@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

DM-Devel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/dm-devel/0 dm-devel/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dm-devel dm-devel/ https://lore.kernel.org/dm-devel \
		dm-devel@redhat.com
	public-inbox-index dm-devel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/com.redhat.dm-devel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git