linux-kernel - Re: [PATCH 2/6] null_blk: allow write zeores on membacked

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 5 Oct 2022 13:57:39 +0900
From:   Damien Le Moal <damien.lemoal@...nsource.wdc.com>
To:     Chaitanya Kulkarni <kch@...dia.com>, linux-block@...r.kernel.org,
        linux-kernel@...r.kernel.org
Cc:     axboe@...nel.dk, johannes.thumshirn@....com, bvanassche@....org,
        ming.lei@...hat.com, shinichiro.kawasaki@....com,
        vincent.fu@...sung.com, yukuai3@...wei.com
Subject: Re: [PATCH 2/6] null_blk: allow write zeores on membacked

On 10/5/22 12:16, Chaitanya Kulkarni wrote:
> Add a helper functions to enable the REQ_OP_WRITE_ZEROES operations
> when null_blk is configured with the membacked mode.
> 
> Since write-zeroes is a non-trivial I/O operation we need this to
> add a blktest so we can test the non-trivial I/O path from the
> application to the block layer.

Why a separate patch for this ? Introducing the module argument and
configfs equivalent separately in 2 different patches is strange and does
not facilitate review.

This patch should be merged with patch 1.

> 
> Signed-off-by: Chaitanya Kulkarni <kch@...dia.com>
> ---
>  drivers/block/null_blk/main.c | 46 ++++++++++++++++++++++++++++++++++-
>  1 file changed, 45 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c
> index fc3e883f7b84..2d592b4eb815 100644
> --- a/drivers/block/null_blk/main.c
> +++ b/drivers/block/null_blk/main.c
> @@ -420,6 +420,7 @@ NULLB_DEVICE_ATTR(blocking, bool, NULL);
>  NULLB_DEVICE_ATTR(use_per_node_hctx, bool, NULL);
>  NULLB_DEVICE_ATTR(memory_backed, bool, NULL);
>  NULLB_DEVICE_ATTR(discard, bool, NULL);
> +NULLB_DEVICE_ATTR(write_zeroes, bool, NULL);
>  NULLB_DEVICE_ATTR(mbps, uint, NULL);
>  NULLB_DEVICE_ATTR(cache_size, ulong, NULL);
>  NULLB_DEVICE_ATTR(zoned, bool, NULL);
> @@ -544,6 +545,7 @@ static struct configfs_attribute *nullb_device_attrs[] = {
>  	&nullb_device_attr_power,
>  	&nullb_device_attr_memory_backed,
>  	&nullb_device_attr_discard,
> +	&nullb_device_attr_write_zeroes,
>  	&nullb_device_attr_mbps,
>  	&nullb_device_attr_cache_size,
>  	&nullb_device_attr_badblocks,
> @@ -618,7 +620,7 @@ static ssize_t memb_group_features_show(struct config_item *item, char *page)
>  			"poll_queues,power,queue_mode,shared_tag_bitmap,size,"
>  			"submit_queues,use_per_node_hctx,virt_boundary,zoned,"
>  			"zone_capacity,zone_max_active,zone_max_open,"
> -			"zone_nr_conv,zone_size\n");
> +			"zone_nr_conv,zone_size,write_zeroes\n");
>  }
>  
>  CONFIGFS_ATTR_RO(memb_group_, features);
> @@ -875,6 +877,24 @@ static void null_free_sector(struct nullb *nullb, sector_t sector,
>  	}
>  }
>  
> +static void null_zero_sector(struct nullb_device *d, sector_t sect,
> +			     sector_t nr_sects, bool cache)
> +{
> +	struct radix_tree_root *root = cache ? &d->cache : &d->data;
> +	struct nullb_page *t_page;
> +	unsigned int offset;
> +	void *dest;
> +
> +	t_page = radix_tree_lookup(root, sect >> PAGE_SECTORS_SHIFT);
> +	if (!t_page)
> +		return;
> +
> +	offset = (sect & SECTOR_MASK) << SECTOR_SHIFT;
> +	dest = kmap_atomic(t_page->page);
> +	memset(dest + offset, 0, SECTOR_SIZE * nr_sects);
> +	kunmap_atomic(dest);
> +}
> +
>  static struct nullb_page *null_radix_tree_insert(struct nullb *nullb, u64 idx,
>  	struct nullb_page *t_page, bool is_cache)
>  {
> @@ -1191,6 +1211,27 @@ blk_status_t null_handle_discard(struct nullb_device *dev,
>  	return BLK_STS_OK;
>  }
>  
> +static blk_status_t null_handle_write_zeroes(struct nullb_device *dev,
> +					     sector_t sector, sector_t nr_sectors)
> +{
> +	unsigned int bytes_left = nr_sectors << 9;
> +	struct nullb *nullb = dev->nullb;
> +	size_t curr_bytes;
> +
> +	spin_lock_irq(&nullb->lock);
> +	while (bytes_left > 0) {
> +		curr_bytes = min_t(size_t, bytes_left, nullb->dev->blocksize);
> +		nr_sectors = curr_bytes >> SECTOR_SHIFT;
> +		null_zero_sector(nullb->dev, sector, nr_sectors, false);
> +		if (null_cache_active(nullb))
> +			null_zero_sector(nullb->dev, sector, nr_sectors, true);
> +		sector += nr_sectors;
> +		bytes_left -= curr_bytes;
> +	}
> +	spin_unlock_irq(&nullb->lock);
> +	return BLK_STS_OK;
> +}
> +
>  static int null_handle_flush(struct nullb *nullb)
>  {
>  	int err;
> @@ -1357,6 +1398,9 @@ static inline blk_status_t null_handle_memory_backed(struct nullb_cmd *cmd,
>  	if (op == REQ_OP_DISCARD)
>  		return null_handle_discard(dev, sector, nr_sectors);
>  
> +	if (op == REQ_OP_WRITE_ZEROES)
> +		return null_handle_write_zeroes(dev, sector, nr_sectors);
> +
>  	if (dev->queue_mode == NULL_Q_BIO)
>  		err = null_handle_bio(cmd);
>  	else

-- 
Damien Le Moal
Western Digital Research