Linux-mtd Archive on lore.kernel.org
 help / color / Atom feed
From: Mike Snitzer <snitzer@redhat.com>
To: Christoph Hellwig <hch@lst.de>
Cc: Jens Axboe <axboe@kernel.dk>,
	linux-block@vger.kernel.org, martin.petersen@oracle.com,
	Hans de Goede <hdegoede@redhat.com>, Song Liu <song@kernel.org>,
	Richard Weinberger <richard@nod.at>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-raid@vger.kernel.org, Minchan Kim <minchan@kernel.org>,
	dm-devel@redhat.com, linux-mtd@lists.infradead.org,
	drbd-dev@tron.linbit.com, cgroups@vger.kernel.org,
	linux-mm@kvack.org
Subject: Re: [PATCH 06/14] block: lift setting the readahead size into the block layer
Date: Wed, 26 Aug 2020 18:07:38 -0400
Message-ID: <20200826220737.GA25613@redhat.com> (raw)
In-Reply-To: <20200726150333.305527-7-hch@lst.de>

On Sun, Jul 26 2020 at 11:03am -0400,
Christoph Hellwig <hch@lst.de> wrote:

> Drivers shouldn't really mess with the readahead size, as that is a VM
> concept.  Instead set it based on the optimal I/O size by lifting the
> algorithm from the md driver when registering the disk.  Also set
> bdi->io_pages there as well by applying the same scheme based on
> max_sectors.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  block/blk-settings.c         |  5 ++---
>  block/blk-sysfs.c            |  1 -
>  block/genhd.c                | 13 +++++++++++--
>  drivers/block/aoe/aoeblk.c   |  2 --
>  drivers/block/drbd/drbd_nl.c | 12 +-----------
>  drivers/md/bcache/super.c    |  4 ----
>  drivers/md/dm-table.c        |  3 ---
>  drivers/md/raid0.c           | 16 ----------------
>  drivers/md/raid10.c          | 24 +-----------------------
>  drivers/md/raid5.c           | 13 +------------
>  10 files changed, 16 insertions(+), 77 deletions(-)


In general these changes need a solid audit relative to stacking
drivers.  That is, the limits stacking methods (blk_stack_limits)
vs lower level allocation methods (__device_add_disk).

You optimized for lowlevel __device_add_disk establishing the bdi's
ra_pages and io_pages.  That is at the beginning of disk allocation,
well before any build up of stacking driver's queue_io_opt() -- which
was previously done in disk_stack_limits or driver specific methods
(e.g. dm_table_set_restrictions) that are called _after_ all the limits
stacking occurs.

By inverting the setting of the bdi's ra_pages and io_pages to be done
so early in __device_add_disk it'll break properly setting these values
for at least DM afaict.

Mike


> diff --git a/block/blk-settings.c b/block/blk-settings.c
> index 76a7e03bcd6cac..01049e9b998f1d 100644
> --- a/block/blk-settings.c
> +++ b/block/blk-settings.c
> @@ -452,6 +452,8 @@ EXPORT_SYMBOL(blk_limits_io_opt);
>  void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
>  {
>  	blk_limits_io_opt(&q->limits, opt);
> +	q->backing_dev_info->ra_pages =
> +		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
>  }
>  EXPORT_SYMBOL(blk_queue_io_opt);
>  
> @@ -628,9 +630,6 @@ void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
>  		printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
>  		       top, bottom);
>  	}
> -
> -	t->backing_dev_info->io_pages =
> -		t->limits.max_sectors >> (PAGE_SHIFT - 9);
>  }
>  EXPORT_SYMBOL(disk_stack_limits);
>  
> diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
> index 7dda709f3ccb6f..ce418d9128a0b2 100644
> --- a/block/blk-sysfs.c
> +++ b/block/blk-sysfs.c
> @@ -245,7 +245,6 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count)
>  
>  	spin_lock_irq(&q->queue_lock);
>  	q->limits.max_sectors = max_sectors_kb << 1;
> -	q->backing_dev_info->io_pages = max_sectors_kb >> (PAGE_SHIFT - 10);
>  	spin_unlock_irq(&q->queue_lock);
>  
>  	return ret;
> diff --git a/block/genhd.c b/block/genhd.c
> index 8b1e9f48957cb5..097d4e4bc0b8a2 100644
> --- a/block/genhd.c
> +++ b/block/genhd.c
> @@ -775,6 +775,7 @@ static void __device_add_disk(struct device *parent, struct gendisk *disk,
>  			      const struct attribute_group **groups,
>  			      bool register_queue)
>  {
> +	struct request_queue *q = disk->queue;
>  	dev_t devt;
>  	int retval;
>  
> @@ -785,7 +786,7 @@ static void __device_add_disk(struct device *parent, struct gendisk *disk,
>  	 * registration.
>  	 */
>  	if (register_queue)
> -		elevator_init_mq(disk->queue);
> +		elevator_init_mq(q);
>  
>  	/* minors == 0 indicates to use ext devt from part0 and should
>  	 * be accompanied with EXT_DEVT flag.  Make sure all
> @@ -815,10 +816,18 @@ static void __device_add_disk(struct device *parent, struct gendisk *disk,
>  		disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
>  		disk->flags |= GENHD_FL_NO_PART_SCAN;
>  	} else {
> -		struct backing_dev_info *bdi = disk->queue->backing_dev_info;
> +		struct backing_dev_info *bdi = q->backing_dev_info;
>  		struct device *dev = disk_to_dev(disk);
>  		int ret;
>  
> +		/*
> +		 * For read-ahead of large files to be effective, we need to
> +		 * readahead at least twice the optimal I/O size.
> +		 */
> +		bdi->ra_pages = max(queue_io_opt(q) * 2 / PAGE_SIZE,
> +				    VM_READAHEAD_PAGES);
> +		bdi->io_pages = queue_max_sectors(q) >> (PAGE_SHIFT - 9);
> +
>  		/* Register BDI before referencing it from bdev */
>  		dev->devt = devt;
>  		ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt));
> diff --git a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
> index 5ca7216e9e01f3..89b33b402b4e52 100644
> --- a/drivers/block/aoe/aoeblk.c
> +++ b/drivers/block/aoe/aoeblk.c
> @@ -347,7 +347,6 @@ aoeblk_gdalloc(void *vp)
>  	mempool_t *mp;
>  	struct request_queue *q;
>  	struct blk_mq_tag_set *set;
> -	enum { KB = 1024, MB = KB * KB, READ_AHEAD = 2 * MB, };
>  	ulong flags;
>  	int late = 0;
>  	int err;
> @@ -407,7 +406,6 @@ aoeblk_gdalloc(void *vp)
>  	WARN_ON(d->gd);
>  	WARN_ON(d->flags & DEVFL_UP);
>  	blk_queue_max_hw_sectors(q, BLK_DEF_MAX_SECTORS);
> -	q->backing_dev_info->ra_pages = READ_AHEAD / PAGE_SIZE;
>  	d->bufpool = mp;
>  	d->blkq = gd->queue = q;
>  	q->queuedata = d;
> diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
> index 650372ee2c7822..212bf711fb6b41 100644
> --- a/drivers/block/drbd/drbd_nl.c
> +++ b/drivers/block/drbd/drbd_nl.c
> @@ -1360,18 +1360,8 @@ static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backi
>  	decide_on_discard_support(device, q, b, discard_zeroes_if_aligned);
>  	decide_on_write_same_support(device, q, b, o, disable_write_same);
>  
> -	if (b) {
> +	if (b)
>  		blk_stack_limits(&q->limits, &b->limits, 0);
> -
> -		if (q->backing_dev_info->ra_pages !=
> -		    b->backing_dev_info->ra_pages) {
> -			drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
> -				 q->backing_dev_info->ra_pages,
> -				 b->backing_dev_info->ra_pages);
> -			q->backing_dev_info->ra_pages =
> -						b->backing_dev_info->ra_pages;
> -		}
> -	}
>  	fixup_discard_if_not_supported(q);
>  	fixup_write_zeroes(device, q);
>  }
> diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
> index 9e45faa054b6f4..9d3f0711be030f 100644
> --- a/drivers/md/bcache/super.c
> +++ b/drivers/md/bcache/super.c
> @@ -1367,10 +1367,6 @@ static int cached_dev_init(struct cached_dev *dc, unsigned int block_size)
>  	if (ret)
>  		return ret;
>  
> -	dc->disk.disk->queue->backing_dev_info->ra_pages =
> -		max(dc->disk.disk->queue->backing_dev_info->ra_pages,
> -		    q->backing_dev_info->ra_pages);
> -
>  	atomic_set(&dc->io_errors, 0);
>  	dc->io_disable = false;
>  	dc->error_limit = DEFAULT_CACHED_DEV_ERROR_LIMIT;
> diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
> index aac4c31cfc8498..324a42ed2f8894 100644
> --- a/drivers/md/dm-table.c
> +++ b/drivers/md/dm-table.c
> @@ -1924,9 +1924,6 @@ void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
>  		q->nr_zones = blkdev_nr_zones(t->md->disk);
>  	}
>  #endif
> -
> -	/* Allow reads to exceed readahead limits */
> -	q->backing_dev_info->io_pages = limits->max_sectors >> (PAGE_SHIFT - 9);
>  }
>  
>  unsigned int dm_table_get_num_targets(struct dm_table *t)
> diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
> index f54a449f97aa79..aa2d7279176880 100644
> --- a/drivers/md/raid0.c
> +++ b/drivers/md/raid0.c
> @@ -410,22 +410,6 @@ static int raid0_run(struct mddev *mddev)
>  		 mdname(mddev),
>  		 (unsigned long long)mddev->array_sectors);
>  
> -	if (mddev->queue) {
> -		/* calculate the max read-ahead size.
> -		 * For read-ahead of large files to be effective, we need to
> -		 * readahead at least twice a whole stripe. i.e. number of devices
> -		 * multiplied by chunk size times 2.
> -		 * If an individual device has an ra_pages greater than the
> -		 * chunk size, then we will not drive that device as hard as it
> -		 * wants.  We consider this a configuration error: a larger
> -		 * chunksize should be used in that case.
> -		 */
> -		int stripe = mddev->raid_disks *
> -			(mddev->chunk_sectors << 9) / PAGE_SIZE;
> -		if (mddev->queue->backing_dev_info->ra_pages < 2* stripe)
> -			mddev->queue->backing_dev_info->ra_pages = 2* stripe;
> -	}
> -
>  	dump_zones(mddev);
>  
>  	ret = md_integrity_register(mddev);
> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
> index 9f88ff9bdee437..23d15acbf457d4 100644
> --- a/drivers/md/raid10.c
> +++ b/drivers/md/raid10.c
> @@ -3865,19 +3865,6 @@ static int raid10_run(struct mddev *mddev)
>  	mddev->resync_max_sectors = size;
>  	set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
>  
> -	if (mddev->queue) {
> -		int stripe = conf->geo.raid_disks *
> -			((mddev->chunk_sectors << 9) / PAGE_SIZE);
> -
> -		/* Calculate max read-ahead size.
> -		 * We need to readahead at least twice a whole stripe....
> -		 * maybe...
> -		 */
> -		stripe /= conf->geo.near_copies;
> -		if (mddev->queue->backing_dev_info->ra_pages < 2 * stripe)
> -			mddev->queue->backing_dev_info->ra_pages = 2 * stripe;
> -	}
> -
>  	if (md_integrity_register(mddev))
>  		goto out_free_conf;
>  
> @@ -4715,17 +4702,8 @@ static void end_reshape(struct r10conf *conf)
>  	conf->reshape_safe = MaxSector;
>  	spin_unlock_irq(&conf->device_lock);
>  
> -	/* read-ahead size must cover two whole stripes, which is
> -	 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
> -	 */
> -	if (conf->mddev->queue) {
> -		int stripe = conf->geo.raid_disks *
> -			((conf->mddev->chunk_sectors << 9) / PAGE_SIZE);
> -		stripe /= conf->geo.near_copies;
> -		if (conf->mddev->queue->backing_dev_info->ra_pages < 2 * stripe)
> -			conf->mddev->queue->backing_dev_info->ra_pages = 2 * stripe;
> +	if (conf->mddev->queue)
>  		raid10_set_io_opt(conf);
> -	}
>  	conf->fullsync = 0;
>  }
>  
> diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
> index 68e41ce3ca75cc..415ce3cc155698 100644
> --- a/drivers/md/raid5.c
> +++ b/drivers/md/raid5.c
> @@ -7413,8 +7413,6 @@ static int raid5_run(struct mddev *mddev)
>  		int data_disks = conf->previous_raid_disks - conf->max_degraded;
>  		int stripe = data_disks *
>  			((mddev->chunk_sectors << 9) / PAGE_SIZE);
> -		if (mddev->queue->backing_dev_info->ra_pages < 2 * stripe)
> -			mddev->queue->backing_dev_info->ra_pages = 2 * stripe;
>  
>  		chunk_size = mddev->chunk_sectors << 9;
>  		blk_queue_io_min(mddev->queue, chunk_size);
> @@ -8002,17 +8000,8 @@ static void end_reshape(struct r5conf *conf)
>  		spin_unlock_irq(&conf->device_lock);
>  		wake_up(&conf->wait_for_overlap);
>  
> -		/* read-ahead size must cover two whole stripes, which is
> -		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
> -		 */
> -		if (conf->mddev->queue) {
> -			int data_disks = conf->raid_disks - conf->max_degraded;
> -			int stripe = data_disks * ((conf->chunk_sectors << 9)
> -						   / PAGE_SIZE);
> -			if (conf->mddev->queue->backing_dev_info->ra_pages < 2 * stripe)
> -				conf->mddev->queue->backing_dev_info->ra_pages = 2 * stripe;
> +		if (conf->mddev->queue)
>  			raid5_set_io_opt(conf);
> -		}
>  	}
>  }
>  
> -- 
> 2.27.0
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

  reply index

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-26 15:03 bdi cleanups v3 Christoph Hellwig
2020-07-26 15:03 ` [PATCH 01/14] fs: remove the unused SB_I_MULTIROOT flag Christoph Hellwig
2020-07-26 15:03 ` [PATCH 02/14] drbd: remove dead code in device_to_statistics Christoph Hellwig
2020-07-26 15:03 ` [PATCH 03/14] drbd: remove RB_CONGESTED_REMOTE Christoph Hellwig
2020-07-26 15:03 ` [PATCH 04/14] bdi: initialize ->ra_pages in bdi_init Christoph Hellwig
2020-07-26 15:03 ` [PATCH 05/14] md: update the optimal I/O size on reshape Christoph Hellwig
2020-07-26 15:03 ` [PATCH 06/14] block: lift setting the readahead size into the block layer Christoph Hellwig
2020-08-26 22:07   ` Mike Snitzer [this message]
2020-09-02 15:11     ` Christoph Hellwig
2020-09-02 16:20       ` Mike Snitzer
2020-09-10  9:28         ` Christoph Hellwig
2020-09-10 17:15           ` Mike Snitzer
2020-09-15  7:05             ` Christoph Hellwig
2020-07-26 15:03 ` [PATCH 07/14] block: make QUEUE_SYSFS_BIT_FNS a little more useful Christoph Hellwig
2020-07-29  8:57   ` [block] 54529aac98: fsmark.files_per_sec -13.6% regression kernel test robot
2020-07-30  2:14   ` [block] 54529aac98: blktests.block.005.fail kernel test robot
2020-07-26 15:03 ` [PATCH 08/14] block: add helper macros for queue sysfs entries Christoph Hellwig
2020-07-26 15:03 ` [PATCH 09/14] bdi: remove BDI_CAP_CGROUP_WRITEBACK Christoph Hellwig
2020-07-26 15:03 ` [PATCH 10/14] bdi: remove BDI_CAP_SYNCHRONOUS_IO Christoph Hellwig
2020-07-26 19:06   ` Minchan Kim
2020-07-27  7:58     ` Christoph Hellwig
2020-07-26 15:03 ` [PATCH 11/14] mm: use SWP_SYNCHRONOUS_IO more intelligently Christoph Hellwig
2020-07-26 15:03 ` [PATCH 12/14] bdi: replace BDI_CAP_STABLE_WRITES with a queue and a sb flag Christoph Hellwig
2020-07-26 15:03 ` [PATCH 13/14] bdi: invert BDI_CAP_NO_ACCT_WB Christoph Hellwig
2020-07-26 15:03 ` [PATCH 14/14] bdi: replace BDI_CAP_NO_{WRITEBACK, ACCT_DIRTY} with a single flag Christoph Hellwig
2020-07-26 15:12 ` bdi cleanups v3 Christoph Hellwig
  -- strict thread matches above, loose matches on Subject: below --
2020-07-24  7:32 Christoph Hellwig
2020-07-24  7:33 ` [PATCH 06/14] block: lift setting the readahead size into the block layer Christoph Hellwig
2020-07-28 12:23   ` Johannes Thumshirn
2020-07-22  6:25 bdi cleanups v2 Christoph Hellwig
2020-07-22  6:25 ` [PATCH 06/14] block: lift setting the readahead size into the block layer Christoph Hellwig
2020-07-22  7:13   ` Johannes Thumshirn
2020-07-22  7:18     ` Christoph Hellwig
2020-07-20  7:51 bdi cleanups Christoph Hellwig
2020-07-20  7:51 ` [PATCH 06/14] block: lift setting the readahead size into the block layer Christoph Hellwig

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=20200826220737.GA25613@redhat.com \
    --to=snitzer@redhat.com \
    --cc=axboe@kernel.dk \
    --cc=cgroups@vger.kernel.org \
    --cc=dm-devel@redhat.com \
    --cc=drbd-dev@tron.linbit.com \
    --cc=hch@lst.de \
    --cc=hdegoede@redhat.com \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux-mtd@lists.infradead.org \
    --cc=linux-raid@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    --cc=minchan@kernel.org \
    --cc=richard@nod.at \
    --cc=song@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

Linux-mtd Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-mtd/0 linux-mtd/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 linux-mtd linux-mtd/ https://lore.kernel.org/linux-mtd \
		linux-mtd@lists.infradead.org
	public-inbox-index linux-mtd

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.infradead.lists.linux-mtd


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