renegade-project/linux-next
2
0
Fork 0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-23 04:34:11 +08:00
Code
766b9b168f
linux-next/drivers/md/dm-zoned-metadata.c

2538 lines
59 KiB
C
Raw Normal View History

dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
/*
* Copyright (C) 2017 Western Digital Corporation or its affiliates.
*
* This file is released under the GPL.
*/
#include "dm-zoned.h"
#include <linux/module.h>
#include <linux/crc32.h>
#define DM_MSG_PREFIX "zoned metadata"
/*
* Metadata version.
*/
#define DMZ_META_VER 1
/*
* On-disk super block magic.
*/
#define DMZ_MAGIC ((((unsigned int)('D')) << 24) | \
(((unsigned int)('Z')) << 16) | \
(((unsigned int)('B')) << 8) | \
((unsigned int)('D')))
/*
* On disk super block.
* This uses only 512 B but uses on disk a full 4KB block. This block is
* followed on disk by the mapping table of chunks to zones and the bitmap
* blocks indicating zone block validity.
* The overall resulting metadata format is:
* (1) Super block (1 block)
* (2) Chunk mapping table (nr_map_blocks)
* (3) Bitmap blocks (nr_bitmap_blocks)
* All metadata blocks are stored in conventional zones, starting from the
* the first conventional zone found on disk.
*/
struct dmz_super {
/* Magic number */
__le32 magic; /* 4 */
/* Metadata version number */
__le32 version; /* 8 */
/* Generation number */
__le64 gen; /* 16 */
/* This block number */
__le64 sb_block; /* 24 */
/* The number of metadata blocks, including this super block */
__le32 nr_meta_blocks; /* 28 */
/* The number of sequential zones reserved for reclaim */
__le32 nr_reserved_seq; /* 32 */
/* The number of entries in the mapping table */
__le32 nr_chunks; /* 36 */
/* The number of blocks used for the chunk mapping table */
__le32 nr_map_blocks; /* 40 */
/* The number of blocks used for the block bitmaps */
__le32 nr_bitmap_blocks; /* 44 */
/* Checksum */
__le32 crc; /* 48 */
/* Padding to full 512B sector */
u8 reserved[464]; /* 512 */
};
/*
* Chunk mapping entry: entries are indexed by chunk number
* and give the zone ID (dzone_id) mapping the chunk on disk.
* This zone may be sequential or random. If it is a sequential
* zone, a second zone (bzone_id) used as a write buffer may
* also be specified. This second zone will always be a randomly
* writeable zone.
*/
struct dmz_map {
__le32 dzone_id;
__le32 bzone_id;
};
/*
* Chunk mapping table metadata: 512 8-bytes entries per 4KB block.
*/
#define DMZ_MAP_ENTRIES (DMZ_BLOCK_SIZE / sizeof(struct dmz_map))
#define DMZ_MAP_ENTRIES_SHIFT (ilog2(DMZ_MAP_ENTRIES))
#define DMZ_MAP_ENTRIES_MASK (DMZ_MAP_ENTRIES - 1)
#define DMZ_MAP_UNMAPPED UINT_MAX
/*
* Meta data block descriptor (for cached metadata blocks).
*/
struct dmz_mblock {
struct rb_node node;
struct list_head link;
sector_t no;
dm zoned: fix metadata block ref counting Since the ref field of struct dmz_mblock is always used with the spinlock of struct dmz_metadata locked, there is no need to use an atomic_t type. Change the type of the ref field to an unsigne integer. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:07 +08:00
unsigned int ref;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
unsigned long state;
struct page *page;
void *data;
};
/*
* Metadata block state flags.
*/
enum {
DMZ_META_DIRTY,
DMZ_META_READING,
DMZ_META_WRITING,
DMZ_META_ERROR,
};
/*
* Super block information (one per metadata set).
*/
struct dmz_sb {
sector_t block;
struct dmz_mblock *mblk;
struct dmz_super *sb;
};
/*
* In-memory metadata.
*/
struct dmz_metadata {
struct dmz_dev *dev;
sector_t zone_bitmap_size;
unsigned int zone_nr_bitmap_blocks;
unsigned int nr_bitmap_blocks;
unsigned int nr_map_blocks;
unsigned int nr_useable_zones;
unsigned int nr_meta_blocks;
unsigned int nr_meta_zones;
unsigned int nr_data_zones;
unsigned int nr_rnd_zones;
unsigned int nr_reserved_seq;
unsigned int nr_chunks;
/* Zone information array */
struct dm_zone *zones;
struct dm_zone *sb_zone;
struct dmz_sb sb[2];
unsigned int mblk_primary;
u64 sb_gen;
unsigned int min_nr_mblks;
unsigned int max_nr_mblks;
atomic_t nr_mblks;
struct rw_semaphore mblk_sem;
struct mutex mblk_flush_lock;
spinlock_t mblk_lock;
struct rb_root mblk_rbtree;
struct list_head mblk_lru_list;
struct list_head mblk_dirty_list;
struct shrinker mblk_shrinker;
/* Zone allocation management */
struct mutex map_lock;
struct dmz_mblock **map_mblk;
unsigned int nr_rnd;
atomic_t unmap_nr_rnd;
struct list_head unmap_rnd_list;
struct list_head map_rnd_list;
unsigned int nr_seq;
atomic_t unmap_nr_seq;
struct list_head unmap_seq_list;
struct list_head map_seq_list;
atomic_t nr_reserved_seq_zones;
struct list_head reserved_seq_zones_list;
wait_queue_head_t free_wq;
};
/*
* Various accessors
*/
unsigned int dmz_id(struct dmz_metadata *zmd, struct dm_zone *zone)
{
return ((unsigned int)(zone - zmd->zones));
}
sector_t dmz_start_sect(struct dmz_metadata *zmd, struct dm_zone *zone)
{
dm zoned: fix overflow when converting zone ID to sectors A zone ID is a 32 bits unsigned int which can overflow when doing the bit shifts in dmz_start_sect(). With a 256 MB zone size drive, the overflow happens for a zone ID >= 8192. Fix this by casting the zone ID to a sector_t before doing the bit shift. While at it, similarly fix dmz_start_block(). Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-03 14:44:58 +08:00
return (sector_t)dmz_id(zmd, zone) << zmd->dev->zone_nr_sectors_shift;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
}
sector_t dmz_start_block(struct dmz_metadata *zmd, struct dm_zone *zone)
{
dm zoned: fix overflow when converting zone ID to sectors A zone ID is a 32 bits unsigned int which can overflow when doing the bit shifts in dmz_start_sect(). With a 256 MB zone size drive, the overflow happens for a zone ID >= 8192. Fix this by casting the zone ID to a sector_t before doing the bit shift. While at it, similarly fix dmz_start_block(). Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-03 14:44:58 +08:00
return (sector_t)dmz_id(zmd, zone) << zmd->dev->zone_nr_blocks_shift;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
}
unsigned int dmz_nr_chunks(struct dmz_metadata *zmd)
{
return zmd->nr_chunks;
}
unsigned int dmz_nr_rnd_zones(struct dmz_metadata *zmd)
{
return zmd->nr_rnd;
}
unsigned int dmz_nr_unmap_rnd_zones(struct dmz_metadata *zmd)
{
return atomic_read(&zmd->unmap_nr_rnd);
}
/*
* Lock/unlock mapping table.
* The map lock also protects all the zone lists.
*/
void dmz_lock_map(struct dmz_metadata *zmd)
{
mutex_lock(&zmd->map_lock);
}
void dmz_unlock_map(struct dmz_metadata *zmd)
{
mutex_unlock(&zmd->map_lock);
}
/*
* Lock/unlock metadata access. This is a "read" lock on a semaphore
* that prevents metadata flush from running while metadata are being
* modified. The actual metadata write mutual exclusion is achieved with
* the map lock and zone styate management (active and reclaim state are
* mutually exclusive).
*/
void dmz_lock_metadata(struct dmz_metadata *zmd)
{
down_read(&zmd->mblk_sem);
}
void dmz_unlock_metadata(struct dmz_metadata *zmd)
{
up_read(&zmd->mblk_sem);
}
/*
* Lock/unlock flush: prevent concurrent executions
* of dmz_flush_metadata as well as metadata modification in reclaim
* while flush is being executed.
*/
void dmz_lock_flush(struct dmz_metadata *zmd)
{
mutex_lock(&zmd->mblk_flush_lock);
}
void dmz_unlock_flush(struct dmz_metadata *zmd)
{
mutex_unlock(&zmd->mblk_flush_lock);
}
/*
* Allocate a metadata block.
*/
static struct dmz_mblock *dmz_alloc_mblock(struct dmz_metadata *zmd,
sector_t mblk_no)
{
struct dmz_mblock *mblk = NULL;
/* See if we can reuse cached blocks */
if (zmd->max_nr_mblks && atomic_read(&zmd->nr_mblks) > zmd->max_nr_mblks) {
spin_lock(&zmd->mblk_lock);
mblk = list_first_entry_or_null(&zmd->mblk_lru_list,
struct dmz_mblock, link);
if (mblk) {
list_del_init(&mblk->link);
rb_erase(&mblk->node, &zmd->mblk_rbtree);
mblk->no = mblk_no;
}
spin_unlock(&zmd->mblk_lock);
if (mblk)
return mblk;
}
/* Allocate a new block */
mblk = kmalloc(sizeof(struct dmz_mblock), GFP_NOIO);
if (!mblk)
return NULL;
mblk->page = alloc_page(GFP_NOIO);
if (!mblk->page) {
kfree(mblk);
return NULL;
}
RB_CLEAR_NODE(&mblk->node);
INIT_LIST_HEAD(&mblk->link);
dm zoned: fix metadata block ref counting Since the ref field of struct dmz_mblock is always used with the spinlock of struct dmz_metadata locked, there is no need to use an atomic_t type. Change the type of the ref field to an unsigne integer. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:07 +08:00
mblk->ref = 0;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
mblk->state = 0;
mblk->no = mblk_no;
mblk->data = page_address(mblk->page);
atomic_inc(&zmd->nr_mblks);
return mblk;
}
/*
* Free a metadata block.
*/
static void dmz_free_mblock(struct dmz_metadata *zmd, struct dmz_mblock *mblk)
{
__free_pages(mblk->page, 0);
kfree(mblk);
atomic_dec(&zmd->nr_mblks);
}
/*
* Insert a metadata block in the rbtree.
*/
static void dmz_insert_mblock(struct dmz_metadata *zmd, struct dmz_mblock *mblk)
{
struct rb_root *root = &zmd->mblk_rbtree;
struct rb_node **new = &(root->rb_node), *parent = NULL;
struct dmz_mblock *b;
/* Figure out where to put the new node */
while (*new) {
b = container_of(*new, struct dmz_mblock, node);
parent = *new;
new = (b->no < mblk->no) ? &((*new)->rb_left) : &((*new)->rb_right);
}
/* Add new node and rebalance tree */
rb_link_node(&mblk->node, parent, new);
rb_insert_color(&mblk->node, root);
}
/*
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
* Lookup a metadata block in the rbtree. If the block is found, increment
* its reference count.
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
*/
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
static struct dmz_mblock *dmz_get_mblock_fast(struct dmz_metadata *zmd,
sector_t mblk_no)
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
{
struct rb_root *root = &zmd->mblk_rbtree;
struct rb_node *node = root->rb_node;
struct dmz_mblock *mblk;
while (node) {
mblk = container_of(node, struct dmz_mblock, node);
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
if (mblk->no == mblk_no) {
/*
* If this is the first reference to the block,
* remove it from the LRU list.
*/
mblk->ref++;
if (mblk->ref == 1 &&
!test_bit(DMZ_META_DIRTY, &mblk->state))
list_del_init(&mblk->link);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
return mblk;
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
}
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
node = (mblk->no < mblk_no) ? node->rb_left : node->rb_right;
}
return NULL;
}
/*
* Metadata block BIO end callback.
*/
static void dmz_mblock_bio_end_io(struct bio *bio)
{
struct dmz_mblock *mblk = bio->bi_private;
int flag;
if (bio->bi_status)
set_bit(DMZ_META_ERROR, &mblk->state);
if (bio_op(bio) == REQ_OP_WRITE)
flag = DMZ_META_WRITING;
else
flag = DMZ_META_READING;
clear_bit_unlock(flag, &mblk->state);
smp_mb__after_atomic();
wake_up_bit(&mblk->state, flag);
bio_put(bio);
}
/*
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
* Read an uncached metadata block from disk and add it to the cache.
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
*/
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
static struct dmz_mblock *dmz_get_mblock_slow(struct dmz_metadata *zmd,
sector_t mblk_no)
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
{
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
struct dmz_mblock *mblk, *m;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
sector_t block = zmd->sb[zmd->mblk_primary].block + mblk_no;
struct bio *bio;
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
/* Get a new block and a BIO to read it */
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
mblk = dmz_alloc_mblock(zmd, mblk_no);
if (!mblk)
return NULL;
bio = bio_alloc(GFP_NOIO, 1);
if (!bio) {
dmz_free_mblock(zmd, mblk);
return NULL;
}
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
spin_lock(&zmd->mblk_lock);
/*
* Make sure that another context did not start reading
* the block already.
*/
m = dmz_get_mblock_fast(zmd, mblk_no);
if (m) {
spin_unlock(&zmd->mblk_lock);
dmz_free_mblock(zmd, mblk);
bio_put(bio);
return m;
}
mblk->ref++;
set_bit(DMZ_META_READING, &mblk->state);
dmz_insert_mblock(zmd, mblk);
spin_unlock(&zmd->mblk_lock);
/* Submit read BIO */
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
bio->bi_iter.bi_sector = dmz_blk2sect(block);
block: replace bi_bdev with a gendisk pointer and partitions index This way we don't need a block_device structure to submit I/O. The block_device has different life time rules from the gendisk and request_queue and is usually only available when the block device node is open. Other callers need to explicitly create one (e.g. the lightnvm passthrough code, or the new nvme multipathing code). For the actual I/O path all that we need is the gendisk, which exists once per block device. But given that the block layer also does partition remapping we additionally need a partition index, which is used for said remapping in generic_make_request. Note that all the block drivers generally want request_queue or sometimes the gendisk, so this removes a layer of indirection all over the stack. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2017-08-24 01:10:32 +08:00
bio_set_dev(bio, zmd->dev->bdev);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
bio->bi_private = mblk;
bio->bi_end_io = dmz_mblock_bio_end_io;
bio_set_op_attrs(bio, REQ_OP_READ, REQ_META | REQ_PRIO);
bio_add_page(bio, mblk->page, DMZ_BLOCK_SIZE, 0);
submit_bio(bio);
return mblk;
}
/*
* Free metadata blocks.
*/
static unsigned long dmz_shrink_mblock_cache(struct dmz_metadata *zmd,
unsigned long limit)
{
struct dmz_mblock *mblk;
unsigned long count = 0;
if (!zmd->max_nr_mblks)
return 0;
while (!list_empty(&zmd->mblk_lru_list) &&
atomic_read(&zmd->nr_mblks) > zmd->min_nr_mblks &&
count < limit) {
mblk = list_first_entry(&zmd->mblk_lru_list,
struct dmz_mblock, link);
list_del_init(&mblk->link);
rb_erase(&mblk->node, &zmd->mblk_rbtree);
dmz_free_mblock(zmd, mblk);
count++;
}
return count;
}
/*
* For mblock shrinker: get the number of unused metadata blocks in the cache.
*/
static unsigned long dmz_mblock_shrinker_count(struct shrinker *shrink,
struct shrink_control *sc)
{
struct dmz_metadata *zmd = container_of(shrink, struct dmz_metadata, mblk_shrinker);
return atomic_read(&zmd->nr_mblks);
}
/*
* For mblock shrinker: scan unused metadata blocks and shrink the cache.
*/
static unsigned long dmz_mblock_shrinker_scan(struct shrinker *shrink,
struct shrink_control *sc)
{
struct dmz_metadata *zmd = container_of(shrink, struct dmz_metadata, mblk_shrinker);
unsigned long count;
spin_lock(&zmd->mblk_lock);
count = dmz_shrink_mblock_cache(zmd, sc->nr_to_scan);
spin_unlock(&zmd->mblk_lock);
return count ? count : SHRINK_STOP;
}
/*
* Release a metadata block.
*/
static void dmz_release_mblock(struct dmz_metadata *zmd,
struct dmz_mblock *mblk)
{
if (!mblk)
return;
spin_lock(&zmd->mblk_lock);
dm zoned: fix metadata block ref counting Since the ref field of struct dmz_mblock is always used with the spinlock of struct dmz_metadata locked, there is no need to use an atomic_t type. Change the type of the ref field to an unsigne integer. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:07 +08:00
mblk->ref--;
if (mblk->ref == 0) {
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
if (test_bit(DMZ_META_ERROR, &mblk->state)) {
rb_erase(&mblk->node, &zmd->mblk_rbtree);
dmz_free_mblock(zmd, mblk);
} else if (!test_bit(DMZ_META_DIRTY, &mblk->state)) {
list_add_tail(&mblk->link, &zmd->mblk_lru_list);
dmz_shrink_mblock_cache(zmd, 1);
}
}
spin_unlock(&zmd->mblk_lock);
}
/*
* Get a metadata block from the rbtree. If the block
* is not present, read it from disk.
*/
static struct dmz_mblock *dmz_get_mblock(struct dmz_metadata *zmd,
sector_t mblk_no)
{
struct dmz_mblock *mblk;
/* Check rbtree */
spin_lock(&zmd->mblk_lock);
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
mblk = dmz_get_mblock_fast(zmd, mblk_no);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
spin_unlock(&zmd->mblk_lock);
if (!mblk) {
/* Cache miss: read the block from disk */
dm zoned: fix various dmz_get_mblock() issues dmz_fetch_mblock() called from dmz_get_mblock() has a race since the allocation of the new metadata block descriptor and its insertion in the cache rbtree with the READING state is not atomic. Two different contexts requesting the same block may end up each adding two different descriptors of the same block to the cache. Another problem for this function is that the BIO for processing the block read is allocated after the metadata block descriptor is inserted in the cache rbtree. If the BIO allocation fails, the metadata block descriptor is freed without first being removed from the rbtree. Fix the first problem by checking again if the requested block is not in the cache right before inserting the newly allocated descriptor, atomically under the mblk_lock spinlock. The second problem is fixed by simply allocating the BIO before inserting the new block in the cache. Finally, since dmz_fetch_mblock() also increments a block reference counter, rename the function to dmz_get_mblock_slow(). To be symmetric and clear, also rename dmz_lookup_mblock() to dmz_get_mblock_fast() and increment the block reference counter directly in that function rather than in dmz_get_mblock(). Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:08 +08:00
mblk = dmz_get_mblock_slow(zmd, mblk_no);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
if (!mblk)
return ERR_PTR(-ENOMEM);
}
/* Wait for on-going read I/O and check for error */
wait_on_bit_io(&mblk->state, DMZ_META_READING,
TASK_UNINTERRUPTIBLE);
if (test_bit(DMZ_META_ERROR, &mblk->state)) {
dmz_release_mblock(zmd, mblk);
return ERR_PTR(-EIO);
}
return mblk;
}
/*
* Mark a metadata block dirty.
*/
static void dmz_dirty_mblock(struct dmz_metadata *zmd, struct dmz_mblock *mblk)
{
spin_lock(&zmd->mblk_lock);
if (!test_and_set_bit(DMZ_META_DIRTY, &mblk->state))
list_add_tail(&mblk->link, &zmd->mblk_dirty_list);
spin_unlock(&zmd->mblk_lock);
}
/*
* Issue a metadata block write BIO.
*/
static void dmz_write_mblock(struct dmz_metadata *zmd, struct dmz_mblock *mblk,
unsigned int set)
{
sector_t block = zmd->sb[set].block + mblk->no;
struct bio *bio;
bio = bio_alloc(GFP_NOIO, 1);
if (!bio) {
set_bit(DMZ_META_ERROR, &mblk->state);
return;
}
set_bit(DMZ_META_WRITING, &mblk->state);
bio->bi_iter.bi_sector = dmz_blk2sect(block);
block: replace bi_bdev with a gendisk pointer and partitions index This way we don't need a block_device structure to submit I/O. The block_device has different life time rules from the gendisk and request_queue and is usually only available when the block device node is open. Other callers need to explicitly create one (e.g. the lightnvm passthrough code, or the new nvme multipathing code). For the actual I/O path all that we need is the gendisk, which exists once per block device. But given that the block layer also does partition remapping we additionally need a partition index, which is used for said remapping in generic_make_request. Note that all the block drivers generally want request_queue or sometimes the gendisk, so this removes a layer of indirection all over the stack. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2017-08-24 01:10:32 +08:00
bio_set_dev(bio, zmd->dev->bdev);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
bio->bi_private = mblk;
bio->bi_end_io = dmz_mblock_bio_end_io;
bio_set_op_attrs(bio, REQ_OP_WRITE, REQ_META | REQ_PRIO);
bio_add_page(bio, mblk->page, DMZ_BLOCK_SIZE, 0);
submit_bio(bio);
}
/*
* Read/write a metadata block.
*/
static int dmz_rdwr_block(struct dmz_metadata *zmd, int op, sector_t block,
struct page *page)
{
struct bio *bio;
int ret;
bio = bio_alloc(GFP_NOIO, 1);
if (!bio)
return -ENOMEM;
bio->bi_iter.bi_sector = dmz_blk2sect(block);
block: replace bi_bdev with a gendisk pointer and partitions index This way we don't need a block_device structure to submit I/O. The block_device has different life time rules from the gendisk and request_queue and is usually only available when the block device node is open. Other callers need to explicitly create one (e.g. the lightnvm passthrough code, or the new nvme multipathing code). For the actual I/O path all that we need is the gendisk, which exists once per block device. But given that the block layer also does partition remapping we additionally need a partition index, which is used for said remapping in generic_make_request. Note that all the block drivers generally want request_queue or sometimes the gendisk, so this removes a layer of indirection all over the stack. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2017-08-24 01:10:32 +08:00
bio_set_dev(bio, zmd->dev->bdev);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
bio_set_op_attrs(bio, op, REQ_SYNC | REQ_META | REQ_PRIO);
bio_add_page(bio, page, DMZ_BLOCK_SIZE, 0);
ret = submit_bio_wait(bio);
bio_put(bio);
return ret;
}
/*
* Write super block of the specified metadata set.
*/
static int dmz_write_sb(struct dmz_metadata *zmd, unsigned int set)
{
sector_t block = zmd->sb[set].block;
struct dmz_mblock *mblk = zmd->sb[set].mblk;
struct dmz_super *sb = zmd->sb[set].sb;
u64 sb_gen = zmd->sb_gen + 1;
int ret;
sb->magic = cpu_to_le32(DMZ_MAGIC);
sb->version = cpu_to_le32(DMZ_META_VER);
sb->gen = cpu_to_le64(sb_gen);
sb->sb_block = cpu_to_le64(block);
sb->nr_meta_blocks = cpu_to_le32(zmd->nr_meta_blocks);
sb->nr_reserved_seq = cpu_to_le32(zmd->nr_reserved_seq);
sb->nr_chunks = cpu_to_le32(zmd->nr_chunks);
sb->nr_map_blocks = cpu_to_le32(zmd->nr_map_blocks);
sb->nr_bitmap_blocks = cpu_to_le32(zmd->nr_bitmap_blocks);
sb->crc = 0;
sb->crc = cpu_to_le32(crc32_le(sb_gen, (unsigned char *)sb, DMZ_BLOCK_SIZE));
ret = dmz_rdwr_block(zmd, REQ_OP_WRITE, block, mblk->page);
if (ret == 0)
dm zoned: use GFP_NOIO in I/O path Use GFP_NOIO for memory allocations in the I/O path. Other memory allocations in the initialization path can use GFP_KERNEL. Reported-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-24 15:44:37 +08:00
ret = blkdev_issue_flush(zmd->dev->bdev, GFP_NOIO, NULL);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
return ret;
}
/*
* Write dirty metadata blocks to the specified set.
*/
static int dmz_write_dirty_mblocks(struct dmz_metadata *zmd,
struct list_head *write_list,
unsigned int set)
{
struct dmz_mblock *mblk;
struct blk_plug plug;
int ret = 0;
/* Issue writes */
blk_start_plug(&plug);
list_for_each_entry(mblk, write_list, link)
dmz_write_mblock(zmd, mblk, set);
blk_finish_plug(&plug);
/* Wait for completion */
list_for_each_entry(mblk, write_list, link) {
wait_on_bit_io(&mblk->state, DMZ_META_WRITING,
TASK_UNINTERRUPTIBLE);
if (test_bit(DMZ_META_ERROR, &mblk->state)) {
clear_bit(DMZ_META_ERROR, &mblk->state);
ret = -EIO;
}
}
/* Flush drive cache (this will also sync data) */
if (ret == 0)
dm zoned: use GFP_NOIO in I/O path Use GFP_NOIO for memory allocations in the I/O path. Other memory allocations in the initialization path can use GFP_KERNEL. Reported-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-24 15:44:37 +08:00
ret = blkdev_issue_flush(zmd->dev->bdev, GFP_NOIO, NULL);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
return ret;
}
/*
* Log dirty metadata blocks.
*/
static int dmz_log_dirty_mblocks(struct dmz_metadata *zmd,
struct list_head *write_list)
{
unsigned int log_set = zmd->mblk_primary ^ 0x1;
int ret;
/* Write dirty blocks to the log */
ret = dmz_write_dirty_mblocks(zmd, write_list, log_set);
if (ret)
return ret;
/*
* No error so far: now validate the log by updating the
* log index super block generation.
*/
ret = dmz_write_sb(zmd, log_set);
if (ret)
return ret;
return 0;
}
/*
* Flush dirty metadata blocks.
*/
int dmz_flush_metadata(struct dmz_metadata *zmd)
{
struct dmz_mblock *mblk;
struct list_head write_list;
int ret;
if (WARN_ON(!zmd))
return 0;
INIT_LIST_HEAD(&write_list);
/*
* Make sure that metadata blocks are stable before logging: take
* the write lock on the metadata semaphore to prevent target BIOs
* from modifying metadata.
*/
down_write(&zmd->mblk_sem);
/*
* This is called from the target flush work and reclaim work.
* Concurrent execution is not allowed.
*/
dmz_lock_flush(zmd);
/* Get dirty blocks */
spin_lock(&zmd->mblk_lock);
list_splice_init(&zmd->mblk_dirty_list, &write_list);
spin_unlock(&zmd->mblk_lock);
/* If there are no dirty metadata blocks, just flush the device cache */
if (list_empty(&write_list)) {
dm zoned: use GFP_NOIO in I/O path Use GFP_NOIO for memory allocations in the I/O path. Other memory allocations in the initialization path can use GFP_KERNEL. Reported-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-24 15:44:37 +08:00
ret = blkdev_issue_flush(zmd->dev->bdev, GFP_NOIO, NULL);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
goto out;
}
/*
* The primary metadata set is still clean. Keep it this way until
* all updates are successful in the secondary set. That is, use
* the secondary set as a log.
*/
ret = dmz_log_dirty_mblocks(zmd, &write_list);
if (ret)
goto out;
/*
* The log is on disk. It is now safe to update in place
* in the primary metadata set.
*/
ret = dmz_write_dirty_mblocks(zmd, &write_list, zmd->mblk_primary);
if (ret)
goto out;
ret = dmz_write_sb(zmd, zmd->mblk_primary);
if (ret)
goto out;
while (!list_empty(&write_list)) {
mblk = list_first_entry(&write_list, struct dmz_mblock, link);
list_del_init(&mblk->link);
spin_lock(&zmd->mblk_lock);
clear_bit(DMZ_META_DIRTY, &mblk->state);
dm zoned: fix metadata block ref counting Since the ref field of struct dmz_mblock is always used with the spinlock of struct dmz_metadata locked, there is no need to use an atomic_t type. Change the type of the ref field to an unsigne integer. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:07 +08:00
if (mblk->ref == 0)
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
list_add_tail(&mblk->link, &zmd->mblk_lru_list);
spin_unlock(&zmd->mblk_lock);
}
zmd->sb_gen++;
out:
if (ret && !list_empty(&write_list)) {
spin_lock(&zmd->mblk_lock);
list_splice(&write_list, &zmd->mblk_dirty_list);
spin_unlock(&zmd->mblk_lock);
}
dmz_unlock_flush(zmd);
up_write(&zmd->mblk_sem);
return ret;
}
/*
* Check super block.
*/
static int dmz_check_sb(struct dmz_metadata *zmd, struct dmz_super *sb)
{
unsigned int nr_meta_zones, nr_data_zones;
struct dmz_dev *dev = zmd->dev;
u32 crc, stored_crc;
u64 gen;
gen = le64_to_cpu(sb->gen);
stored_crc = le32_to_cpu(sb->crc);
sb->crc = 0;
crc = crc32_le(gen, (unsigned char *)sb, DMZ_BLOCK_SIZE);
if (crc != stored_crc) {
dmz_dev_err(dev, "Invalid checksum (needed 0x%08x, got 0x%08x)",
crc, stored_crc);
return -ENXIO;
}
if (le32_to_cpu(sb->magic) != DMZ_MAGIC) {
dmz_dev_err(dev, "Invalid meta magic (needed 0x%08x, got 0x%08x)",
DMZ_MAGIC, le32_to_cpu(sb->magic));
return -ENXIO;
}
if (le32_to_cpu(sb->version) != DMZ_META_VER) {
dmz_dev_err(dev, "Invalid meta version (needed %d, got %d)",
DMZ_META_VER, le32_to_cpu(sb->version));
return -ENXIO;
}
nr_meta_zones = (le32_to_cpu(sb->nr_meta_blocks) + dev->zone_nr_blocks - 1)
>> dev->zone_nr_blocks_shift;
if (!nr_meta_zones ||
nr_meta_zones >= zmd->nr_rnd_zones) {
dmz_dev_err(dev, "Invalid number of metadata blocks");
return -ENXIO;
}
if (!le32_to_cpu(sb->nr_reserved_seq) ||
le32_to_cpu(sb->nr_reserved_seq) >= (zmd->nr_useable_zones - nr_meta_zones)) {
dmz_dev_err(dev, "Invalid number of reserved sequential zones");
return -ENXIO;
}
nr_data_zones = zmd->nr_useable_zones -
(nr_meta_zones * 2 + le32_to_cpu(sb->nr_reserved_seq));
if (le32_to_cpu(sb->nr_chunks) > nr_data_zones) {
dmz_dev_err(dev, "Invalid number of chunks %u / %u",
le32_to_cpu(sb->nr_chunks), nr_data_zones);
return -ENXIO;
}
/* OK */
zmd->nr_meta_blocks = le32_to_cpu(sb->nr_meta_blocks);
zmd->nr_reserved_seq = le32_to_cpu(sb->nr_reserved_seq);
zmd->nr_chunks = le32_to_cpu(sb->nr_chunks);
zmd->nr_map_blocks = le32_to_cpu(sb->nr_map_blocks);
zmd->nr_bitmap_blocks = le32_to_cpu(sb->nr_bitmap_blocks);
zmd->nr_meta_zones = nr_meta_zones;
zmd->nr_data_zones = nr_data_zones;
return 0;
}
/*
* Read the first or second super block from disk.
*/
static int dmz_read_sb(struct dmz_metadata *zmd, unsigned int set)
{
return dmz_rdwr_block(zmd, REQ_OP_READ, zmd->sb[set].block,
zmd->sb[set].mblk->page);
}
/*
* Determine the position of the secondary super blocks on disk.
* This is used only if a corruption of the primary super block
* is detected.
*/
static int dmz_lookup_secondary_sb(struct dmz_metadata *zmd)
{
unsigned int zone_nr_blocks = zmd->dev->zone_nr_blocks;
struct dmz_mblock *mblk;
int i;
/* Allocate a block */
mblk = dmz_alloc_mblock(zmd, 0);
if (!mblk)
return -ENOMEM;
zmd->sb[1].mblk = mblk;
zmd->sb[1].sb = mblk->data;
/* Bad first super block: search for the second one */
zmd->sb[1].block = zmd->sb[0].block + zone_nr_blocks;
for (i = 0; i < zmd->nr_rnd_zones - 1; i++) {
if (dmz_read_sb(zmd, 1) != 0)
break;
if (le32_to_cpu(zmd->sb[1].sb->magic) == DMZ_MAGIC)
return 0;
zmd->sb[1].block += zone_nr_blocks;
}
dmz_free_mblock(zmd, mblk);
zmd->sb[1].mblk = NULL;
return -EIO;
}
/*
* Read the first or second super block from disk.
*/
static int dmz_get_sb(struct dmz_metadata *zmd, unsigned int set)
{
struct dmz_mblock *mblk;
int ret;
/* Allocate a block */
mblk = dmz_alloc_mblock(zmd, 0);
if (!mblk)
return -ENOMEM;
zmd->sb[set].mblk = mblk;
zmd->sb[set].sb = mblk->data;
/* Read super block */
ret = dmz_read_sb(zmd, set);
if (ret) {
dmz_free_mblock(zmd, mblk);
zmd->sb[set].mblk = NULL;
return ret;
}
return 0;
}
/*
* Recover a metadata set.
*/
static int dmz_recover_mblocks(struct dmz_metadata *zmd, unsigned int dst_set)
{
unsigned int src_set = dst_set ^ 0x1;
struct page *page;
int i, ret;
dmz_dev_warn(zmd->dev, "Metadata set %u invalid: recovering", dst_set);
if (dst_set == 0)
zmd->sb[0].block = dmz_start_block(zmd, zmd->sb_zone);
else {
zmd->sb[1].block = zmd->sb[0].block +
(zmd->nr_meta_zones << zmd->dev->zone_nr_blocks_shift);
}
dm zoned: use GFP_NOIO in I/O path Use GFP_NOIO for memory allocations in the I/O path. Other memory allocations in the initialization path can use GFP_KERNEL. Reported-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-24 15:44:37 +08:00
page = alloc_page(GFP_NOIO);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
if (!page)
return -ENOMEM;
/* Copy metadata blocks */
for (i = 1; i < zmd->nr_meta_blocks; i++) {
ret = dmz_rdwr_block(zmd, REQ_OP_READ,
zmd->sb[src_set].block + i, page);
if (ret)
goto out;
ret = dmz_rdwr_block(zmd, REQ_OP_WRITE,
zmd->sb[dst_set].block + i, page);
if (ret)
goto out;
}
/* Finalize with the super block */
if (!zmd->sb[dst_set].mblk) {
zmd->sb[dst_set].mblk = dmz_alloc_mblock(zmd, 0);
if (!zmd->sb[dst_set].mblk) {
ret = -ENOMEM;
goto out;
}
zmd->sb[dst_set].sb = zmd->sb[dst_set].mblk->data;
}
ret = dmz_write_sb(zmd, dst_set);
out:
__free_pages(page, 0);
return ret;
}
/*
* Get super block from disk.
*/
static int dmz_load_sb(struct dmz_metadata *zmd)
{
bool sb_good[2] = {false, false};
u64 sb_gen[2] = {0, 0};
int ret;
/* Read and check the primary super block */
zmd->sb[0].block = dmz_start_block(zmd, zmd->sb_zone);
ret = dmz_get_sb(zmd, 0);
if (ret) {
dmz_dev_err(zmd->dev, "Read primary super block failed");
return ret;
}
ret = dmz_check_sb(zmd, zmd->sb[0].sb);
/* Read and check secondary super block */
if (ret == 0) {
sb_good[0] = true;
zmd->sb[1].block = zmd->sb[0].block +
(zmd->nr_meta_zones << zmd->dev->zone_nr_blocks_shift);
ret = dmz_get_sb(zmd, 1);
} else
ret = dmz_lookup_secondary_sb(zmd);
if (ret) {
dmz_dev_err(zmd->dev, "Read secondary super block failed");
return ret;
}
ret = dmz_check_sb(zmd, zmd->sb[1].sb);
if (ret == 0)
sb_good[1] = true;
/* Use highest generation sb first */
if (!sb_good[0] && !sb_good[1]) {
dmz_dev_err(zmd->dev, "No valid super block found");
return -EIO;
}
if (sb_good[0])
sb_gen[0] = le64_to_cpu(zmd->sb[0].sb->gen);
else
ret = dmz_recover_mblocks(zmd, 0);
if (sb_good[1])
sb_gen[1] = le64_to_cpu(zmd->sb[1].sb->gen);
else
ret = dmz_recover_mblocks(zmd, 1);
if (ret) {
dmz_dev_err(zmd->dev, "Recovery failed");
return -EIO;
}
if (sb_gen[0] >= sb_gen[1]) {
zmd->sb_gen = sb_gen[0];
zmd->mblk_primary = 0;
} else {
zmd->sb_gen = sb_gen[1];
zmd->mblk_primary = 1;
}
dmz_dev_debug(zmd->dev, "Using super block %u (gen %llu)",
zmd->mblk_primary, zmd->sb_gen);
return 0;
}
/*
* Initialize a zone descriptor.
*/
static int dmz_init_zone(struct dmz_metadata *zmd, struct dm_zone *zone,
struct blk_zone *blkz)
{
struct dmz_dev *dev = zmd->dev;
/* Ignore the eventual last runt (smaller) zone */
if (blkz->len != dev->zone_nr_sectors) {
if (blkz->start + blkz->len == dev->capacity)
return 0;
return -ENXIO;
}
INIT_LIST_HEAD(&zone->link);
atomic_set(&zone->refcount, 0);
zone->chunk = DMZ_MAP_UNMAPPED;
if (blkz->type == BLK_ZONE_TYPE_CONVENTIONAL) {
set_bit(DMZ_RND, &zone->flags);
zmd->nr_rnd_zones++;
} else if (blkz->type == BLK_ZONE_TYPE_SEQWRITE_REQ ||
blkz->type == BLK_ZONE_TYPE_SEQWRITE_PREF) {
set_bit(DMZ_SEQ, &zone->flags);
} else
return -ENXIO;
if (blkz->cond == BLK_ZONE_COND_OFFLINE)
set_bit(DMZ_OFFLINE, &zone->flags);
else if (blkz->cond == BLK_ZONE_COND_READONLY)
set_bit(DMZ_READ_ONLY, &zone->flags);
if (dmz_is_rnd(zone))
zone->wp_block = 0;
else
zone->wp_block = dmz_sect2blk(blkz->wp - blkz->start);
if (!dmz_is_offline(zone) && !dmz_is_readonly(zone)) {
zmd->nr_useable_zones++;
if (dmz_is_rnd(zone)) {
zmd->nr_rnd_zones++;
if (!zmd->sb_zone) {
/* Super block zone */
zmd->sb_zone = zone;
}
}
}
return 0;
}
/*
* Free zones descriptors.
*/
static void dmz_drop_zones(struct dmz_metadata *zmd)
{
kfree(zmd->zones);
zmd->zones = NULL;
}
/*
* The size of a zone report in number of zones.
* This results in 4096*64B=256KB report zones commands.
*/
#define DMZ_REPORT_NR_ZONES 4096
/*
* Allocate and initialize zone descriptors using the zone
* information from disk.
*/
static int dmz_init_zones(struct dmz_metadata *zmd)
{
struct dmz_dev *dev = zmd->dev;
struct dm_zone *zone;
struct blk_zone *blkz;
unsigned int nr_blkz;
sector_t sector = 0;
int i, ret = 0;
/* Init */
zmd->zone_bitmap_size = dev->zone_nr_blocks >> 3;
zmd->zone_nr_bitmap_blocks = zmd->zone_bitmap_size >> DMZ_BLOCK_SHIFT;
/* Allocate zone array */
zmd->zones = kcalloc(dev->nr_zones, sizeof(struct dm_zone), GFP_KERNEL);
if (!zmd->zones)
return -ENOMEM;
dmz_dev_info(dev, "Using %zu B for zone information",
sizeof(struct dm_zone) * dev->nr_zones);
/* Get zone information */
nr_blkz = DMZ_REPORT_NR_ZONES;
blkz = kcalloc(nr_blkz, sizeof(struct blk_zone), GFP_KERNEL);
if (!blkz) {
ret = -ENOMEM;
goto out;
}
/*
* Get zone information and initialize zone descriptors.
* At the same time, determine where the super block
* should be: first block of the first randomly writable
* zone.
*/
zone = zmd->zones;
while (sector < dev->capacity) {
/* Get zone information */
nr_blkz = DMZ_REPORT_NR_ZONES;
ret = blkdev_report_zones(dev->bdev, sector, blkz,
&nr_blkz, GFP_KERNEL);
if (ret) {
dmz_dev_err(dev, "Report zones failed %d", ret);
goto out;
}
dm zoned: Fix zone report handling The function blkdev_report_zones() returns success even if no zone information is reported (empty report). Empty zone reports can only happen if the report start sector passed exceeds the device capacity. The conditions for this to happen are either a bug in the caller code, or, a change in the device that forced the low level driver to change the device capacity to a value that is lower than the report start sector. This situation includes a failed disk revalidation resulting in the disk capacity being changed to 0. If this change happens while dm-zoned is in its initialization phase executing dmz_init_zones(), this function may enter an infinite loop and hang the system. To avoid this, add a check to disallow empty zone reports and bail out early. Also fix the function dmz_update_zone() to make sure that the report for the requested zone was correctly obtained. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Shaun Tancheff <shaun@tancheff.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2019-04-18 17:03:07 +08:00
if (!nr_blkz)
break;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
/* Process report */
for (i = 0; i < nr_blkz; i++) {
ret = dmz_init_zone(zmd, zone, &blkz[i]);
if (ret)
goto out;
sector += dev->zone_nr_sectors;
zone++;
}
}
/* The entire zone configuration of the disk should now be known */
if (sector < dev->capacity) {
dmz_dev_err(dev, "Failed to get correct zone information");
ret = -ENXIO;
}
out:
kfree(blkz);
if (ret)
dmz_drop_zones(zmd);
return ret;
}
/*
* Update a zone information.
*/
static int dmz_update_zone(struct dmz_metadata *zmd, struct dm_zone *zone)
{
unsigned int nr_blkz = 1;
struct blk_zone blkz;
int ret;
/* Get zone information from disk */
ret = blkdev_report_zones(zmd->dev->bdev, dmz_start_sect(zmd, zone),
dm zoned: use GFP_NOIO in I/O path Use GFP_NOIO for memory allocations in the I/O path. Other memory allocations in the initialization path can use GFP_KERNEL. Reported-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-24 15:44:37 +08:00
&blkz, &nr_blkz, GFP_NOIO);
dm zoned: Fix zone report handling The function blkdev_report_zones() returns success even if no zone information is reported (empty report). Empty zone reports can only happen if the report start sector passed exceeds the device capacity. The conditions for this to happen are either a bug in the caller code, or, a change in the device that forced the low level driver to change the device capacity to a value that is lower than the report start sector. This situation includes a failed disk revalidation resulting in the disk capacity being changed to 0. If this change happens while dm-zoned is in its initialization phase executing dmz_init_zones(), this function may enter an infinite loop and hang the system. To avoid this, add a check to disallow empty zone reports and bail out early. Also fix the function dmz_update_zone() to make sure that the report for the requested zone was correctly obtained. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Shaun Tancheff <shaun@tancheff.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2019-04-18 17:03:07 +08:00
if (!nr_blkz)
ret = -EIO;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
if (ret) {
dmz_dev_err(zmd->dev, "Get zone %u report failed",
dmz_id(zmd, zone));
return ret;
}
clear_bit(DMZ_OFFLINE, &zone->flags);
clear_bit(DMZ_READ_ONLY, &zone->flags);
if (blkz.cond == BLK_ZONE_COND_OFFLINE)
set_bit(DMZ_OFFLINE, &zone->flags);
else if (blkz.cond == BLK_ZONE_COND_READONLY)
set_bit(DMZ_READ_ONLY, &zone->flags);
if (dmz_is_seq(zone))
zone->wp_block = dmz_sect2blk(blkz.wp - blkz.start);
else
zone->wp_block = 0;
return 0;
}
/*
* Check a zone write pointer position when the zone is marked
* with the sequential write error flag.
*/
static int dmz_handle_seq_write_err(struct dmz_metadata *zmd,
struct dm_zone *zone)
{
unsigned int wp = 0;
int ret;
wp = zone->wp_block;
ret = dmz_update_zone(zmd, zone);
if (ret)
return ret;
dmz_dev_warn(zmd->dev, "Processing zone %u write error (zone wp %u/%u)",
dmz_id(zmd, zone), zone->wp_block, wp);
if (zone->wp_block < wp) {
dmz_invalidate_blocks(zmd, zone, zone->wp_block,
wp - zone->wp_block);
}
return 0;
}
static struct dm_zone *dmz_get(struct dmz_metadata *zmd, unsigned int zone_id)
{
return &zmd->zones[zone_id];
}
/*
* Reset a zone write pointer.
*/
static int dmz_reset_zone(struct dmz_metadata *zmd, struct dm_zone *zone)
{
int ret;
/*
* Ignore offline zones, read only zones,
* and conventional zones.
*/
if (dmz_is_offline(zone) ||
dmz_is_readonly(zone) ||
dmz_is_rnd(zone))
return 0;
if (!dmz_is_empty(zone) || dmz_seq_write_err(zone)) {
struct dmz_dev *dev = zmd->dev;
ret = blkdev_reset_zones(dev->bdev,
dmz_start_sect(zmd, zone),
dm zoned: use GFP_NOIO in I/O path Use GFP_NOIO for memory allocations in the I/O path. Other memory allocations in the initialization path can use GFP_KERNEL. Reported-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-07-24 15:44:37 +08:00
dev->zone_nr_sectors, GFP_NOIO);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
if (ret) {
dmz_dev_err(dev, "Reset zone %u failed %d",
dmz_id(zmd, zone), ret);
return ret;
}
}
/* Clear write error bit and rewind write pointer position */
clear_bit(DMZ_SEQ_WRITE_ERR, &zone->flags);
zone->wp_block = 0;
return 0;
}
static void dmz_get_zone_weight(struct dmz_metadata *zmd, struct dm_zone *zone);
/*
* Initialize chunk mapping.
*/
static int dmz_load_mapping(struct dmz_metadata *zmd)
{
struct dmz_dev *dev = zmd->dev;
struct dm_zone *dzone, *bzone;
struct dmz_mblock *dmap_mblk = NULL;
struct dmz_map *dmap;
unsigned int i = 0, e = 0, chunk = 0;
unsigned int dzone_id;
unsigned int bzone_id;
/* Metadata block array for the chunk mapping table */
zmd->map_mblk = kcalloc(zmd->nr_map_blocks,
sizeof(struct dmz_mblk *), GFP_KERNEL);
if (!zmd->map_mblk)
return -ENOMEM;
/* Get chunk mapping table blocks and initialize zone mapping */
while (chunk < zmd->nr_chunks) {
if (!dmap_mblk) {
/* Get mapping block */
dmap_mblk = dmz_get_mblock(zmd, i + 1);
if (IS_ERR(dmap_mblk))
return PTR_ERR(dmap_mblk);
zmd->map_mblk[i] = dmap_mblk;
dmap = (struct dmz_map *) dmap_mblk->data;
i++;
e = 0;
}
/* Check data zone */
dzone_id = le32_to_cpu(dmap[e].dzone_id);
if (dzone_id == DMZ_MAP_UNMAPPED)
goto next;
if (dzone_id >= dev->nr_zones) {
dmz_dev_err(dev, "Chunk %u mapping: invalid data zone ID %u",
chunk, dzone_id);
return -EIO;
}
dzone = dmz_get(zmd, dzone_id);
set_bit(DMZ_DATA, &dzone->flags);
dzone->chunk = chunk;
dmz_get_zone_weight(zmd, dzone);
if (dmz_is_rnd(dzone))
list_add_tail(&dzone->link, &zmd->map_rnd_list);
else
list_add_tail(&dzone->link, &zmd->map_seq_list);
/* Check buffer zone */
bzone_id = le32_to_cpu(dmap[e].bzone_id);
if (bzone_id == DMZ_MAP_UNMAPPED)
goto next;
if (bzone_id >= dev->nr_zones) {
dmz_dev_err(dev, "Chunk %u mapping: invalid buffer zone ID %u",
chunk, bzone_id);
return -EIO;
}
bzone = dmz_get(zmd, bzone_id);
if (!dmz_is_rnd(bzone)) {
dmz_dev_err(dev, "Chunk %u mapping: invalid buffer zone %u",
chunk, bzone_id);
return -EIO;
}
set_bit(DMZ_DATA, &bzone->flags);
set_bit(DMZ_BUF, &bzone->flags);
bzone->chunk = chunk;
bzone->bzone = dzone;
dzone->bzone = bzone;
dmz_get_zone_weight(zmd, bzone);
list_add_tail(&bzone->link, &zmd->map_rnd_list);
next:
chunk++;
e++;
if (e >= DMZ_MAP_ENTRIES)
dmap_mblk = NULL;
}
/*
* At this point, only meta zones and mapped data zones were
* fully initialized. All remaining zones are unmapped data
* zones. Finish initializing those here.
*/
for (i = 0; i < dev->nr_zones; i++) {
dzone = dmz_get(zmd, i);
if (dmz_is_meta(dzone))
continue;
if (dmz_is_rnd(dzone))
zmd->nr_rnd++;
else
zmd->nr_seq++;
if (dmz_is_data(dzone)) {
/* Already initialized */
continue;
}
/* Unmapped data zone */
set_bit(DMZ_DATA, &dzone->flags);
dzone->chunk = DMZ_MAP_UNMAPPED;
if (dmz_is_rnd(dzone)) {
list_add_tail(&dzone->link, &zmd->unmap_rnd_list);
atomic_inc(&zmd->unmap_nr_rnd);
} else if (atomic_read(&zmd->nr_reserved_seq_zones) < zmd->nr_reserved_seq) {
list_add_tail(&dzone->link, &zmd->reserved_seq_zones_list);
atomic_inc(&zmd->nr_reserved_seq_zones);
zmd->nr_seq--;
} else {
list_add_tail(&dzone->link, &zmd->unmap_seq_list);
atomic_inc(&zmd->unmap_nr_seq);
}
}
return 0;
}
/*
* Set a data chunk mapping.
*/
static void dmz_set_chunk_mapping(struct dmz_metadata *zmd, unsigned int chunk,
unsigned int dzone_id, unsigned int bzone_id)
{
struct dmz_mblock *dmap_mblk = zmd->map_mblk[chunk >> DMZ_MAP_ENTRIES_SHIFT];
struct dmz_map *dmap = (struct dmz_map *) dmap_mblk->data;
int map_idx = chunk & DMZ_MAP_ENTRIES_MASK;
dmap[map_idx].dzone_id = cpu_to_le32(dzone_id);
dmap[map_idx].bzone_id = cpu_to_le32(bzone_id);
dmz_dirty_mblock(zmd, dmap_mblk);
}
/*
* The list of mapped zones is maintained in LRU order.
* This rotates a zone at the end of its map list.
*/
static void __dmz_lru_zone(struct dmz_metadata *zmd, struct dm_zone *zone)
{
if (list_empty(&zone->link))
return;
list_del_init(&zone->link);
if (dmz_is_seq(zone)) {
/* LRU rotate sequential zone */
list_add_tail(&zone->link, &zmd->map_seq_list);
} else {
/* LRU rotate random zone */
list_add_tail(&zone->link, &zmd->map_rnd_list);
}
}
/*
* The list of mapped random zones is maintained
* in LRU order. This rotates a zone at the end of the list.
*/
static void dmz_lru_zone(struct dmz_metadata *zmd, struct dm_zone *zone)
{
__dmz_lru_zone(zmd, zone);
if (zone->bzone)
__dmz_lru_zone(zmd, zone->bzone);
}
/*
* Wait for any zone to be freed.
*/
static void dmz_wait_for_free_zones(struct dmz_metadata *zmd)
{
DEFINE_WAIT(wait);
prepare_to_wait(&zmd->free_wq, &wait, TASK_UNINTERRUPTIBLE);
dmz_unlock_map(zmd);
dmz_unlock_metadata(zmd);
io_schedule_timeout(HZ);
dmz_lock_metadata(zmd);
dmz_lock_map(zmd);
finish_wait(&zmd->free_wq, &wait);
}
/*
* Lock a zone for reclaim (set the zone RECLAIM bit).
* Returns false if the zone cannot be locked or if it is already locked
* and 1 otherwise.
*/
int dmz_lock_zone_reclaim(struct dm_zone *zone)
{
/* Active zones cannot be reclaimed */
if (dmz_is_active(zone))
return 0;
return !test_and_set_bit(DMZ_RECLAIM, &zone->flags);
}
/*
* Clear a zone reclaim flag.
*/
void dmz_unlock_zone_reclaim(struct dm_zone *zone)
{
WARN_ON(dmz_is_active(zone));
WARN_ON(!dmz_in_reclaim(zone));
clear_bit_unlock(DMZ_RECLAIM, &zone->flags);
smp_mb__after_atomic();
wake_up_bit(&zone->flags, DMZ_RECLAIM);
}
/*
* Wait for a zone reclaim to complete.
*/
static void dmz_wait_for_reclaim(struct dmz_metadata *zmd, struct dm_zone *zone)
{
dmz_unlock_map(zmd);
dmz_unlock_metadata(zmd);
wait_on_bit_timeout(&zone->flags, DMZ_RECLAIM, TASK_UNINTERRUPTIBLE, HZ);
dmz_lock_metadata(zmd);
dmz_lock_map(zmd);
}
/*
* Select a random write zone for reclaim.
*/
static struct dm_zone *dmz_get_rnd_zone_for_reclaim(struct dmz_metadata *zmd)
{
struct dm_zone *dzone = NULL;
struct dm_zone *zone;
if (list_empty(&zmd->map_rnd_list))
return NULL;
list_for_each_entry(zone, &zmd->map_rnd_list, link) {
if (dmz_is_buf(zone))
dzone = zone->bzone;
else
dzone = zone;
if (dmz_lock_zone_reclaim(dzone))
return dzone;
}
return NULL;
}
/*
* Select a buffered sequential zone for reclaim.
*/
static struct dm_zone *dmz_get_seq_zone_for_reclaim(struct dmz_metadata *zmd)
{
struct dm_zone *zone;
if (list_empty(&zmd->map_seq_list))
return NULL;
list_for_each_entry(zone, &zmd->map_seq_list, link) {
if (!zone->bzone)
continue;
if (dmz_lock_zone_reclaim(zone))
return zone;
}
return NULL;
}
/*
* Select a zone for reclaim.
*/
struct dm_zone *dmz_get_zone_for_reclaim(struct dmz_metadata *zmd)
{
struct dm_zone *zone;
/*
* Search for a zone candidate to reclaim: 2 cases are possible.
* (1) There is no free sequential zones. Then a random data zone
* cannot be reclaimed. So choose a sequential zone to reclaim so
* that afterward a random zone can be reclaimed.
* (2) At least one free sequential zone is available, then choose
* the oldest random zone (data or buffer) that can be locked.
*/
dmz_lock_map(zmd);
if (list_empty(&zmd->reserved_seq_zones_list))
zone = dmz_get_seq_zone_for_reclaim(zmd);
else
zone = dmz_get_rnd_zone_for_reclaim(zmd);
dmz_unlock_map(zmd);
return zone;
}
/*
* Activate a zone (increment its reference count).
*/
void dmz_activate_zone(struct dm_zone *zone)
{
set_bit(DMZ_ACTIVE, &zone->flags);
atomic_inc(&zone->refcount);
}
/*
* Deactivate a zone. This decrement the zone reference counter
* and clears the active state of the zone once the count reaches 0,
* indicating that all BIOs to the zone have completed. Returns
* true if the zone was deactivated.
*/
void dmz_deactivate_zone(struct dm_zone *zone)
{
if (atomic_dec_and_test(&zone->refcount)) {
WARN_ON(!test_bit(DMZ_ACTIVE, &zone->flags));
clear_bit_unlock(DMZ_ACTIVE, &zone->flags);
smp_mb__after_atomic();
}
}
/*
* Get the zone mapping a chunk, if the chunk is mapped already.
* If no mapping exist and the operation is WRITE, a zone is
* allocated and used to map the chunk.
* The zone returned will be set to the active state.
*/
struct dm_zone *dmz_get_chunk_mapping(struct dmz_metadata *zmd, unsigned int chunk, int op)
{
struct dmz_mblock *dmap_mblk = zmd->map_mblk[chunk >> DMZ_MAP_ENTRIES_SHIFT];
struct dmz_map *dmap = (struct dmz_map *) dmap_mblk->data;
int dmap_idx = chunk & DMZ_MAP_ENTRIES_MASK;
unsigned int dzone_id;
struct dm_zone *dzone = NULL;
int ret = 0;
dmz_lock_map(zmd);
again:
/* Get the chunk mapping */
dzone_id = le32_to_cpu(dmap[dmap_idx].dzone_id);
if (dzone_id == DMZ_MAP_UNMAPPED) {
/*
* Read or discard in unmapped chunks are fine. But for
* writes, we need a mapping, so get one.
*/
if (op != REQ_OP_WRITE)
goto out;
/* Alloate a random zone */
dzone = dmz_alloc_zone(zmd, DMZ_ALLOC_RND);
if (!dzone) {
dmz_wait_for_free_zones(zmd);
goto again;
}
dmz_map_zone(zmd, dzone, chunk);
} else {
/* The chunk is already mapped: get the mapping zone */
dzone = dmz_get(zmd, dzone_id);
if (dzone->chunk != chunk) {
dzone = ERR_PTR(-EIO);
goto out;
}
/* Repair write pointer if the sequential dzone has error */
if (dmz_seq_write_err(dzone)) {
ret = dmz_handle_seq_write_err(zmd, dzone);
if (ret) {
dzone = ERR_PTR(-EIO);
goto out;
}
clear_bit(DMZ_SEQ_WRITE_ERR, &dzone->flags);
}
}
/*
* If the zone is being reclaimed, the chunk mapping may change
* to a different zone. So wait for reclaim and retry. Otherwise,
* activate the zone (this will prevent reclaim from touching it).
*/
if (dmz_in_reclaim(dzone)) {
dmz_wait_for_reclaim(zmd, dzone);
goto again;
}
dmz_activate_zone(dzone);
dmz_lru_zone(zmd, dzone);
out:
dmz_unlock_map(zmd);
return dzone;
}
/*
* Write and discard change the block validity of data zones and their buffer
* zones. Check here that valid blocks are still present. If all blocks are
* invalid, the zones can be unmapped on the fly without waiting for reclaim
* to do it.
*/
void dmz_put_chunk_mapping(struct dmz_metadata *zmd, struct dm_zone *dzone)
{
struct dm_zone *bzone;
dmz_lock_map(zmd);
bzone = dzone->bzone;
if (bzone) {
if (dmz_weight(bzone))
dmz_lru_zone(zmd, bzone);
else {
/* Empty buffer zone: reclaim it */
dmz_unmap_zone(zmd, bzone);
dmz_free_zone(zmd, bzone);
bzone = NULL;
}
}
/* Deactivate the data zone */
dmz_deactivate_zone(dzone);
if (dmz_is_active(dzone) || bzone || dmz_weight(dzone))
dmz_lru_zone(zmd, dzone);
else {
/* Unbuffered inactive empty data zone: reclaim it */
dmz_unmap_zone(zmd, dzone);
dmz_free_zone(zmd, dzone);
}
dmz_unlock_map(zmd);
}
/*
* Allocate and map a random zone to buffer a chunk
* already mapped to a sequential zone.
*/
struct dm_zone *dmz_get_chunk_buffer(struct dmz_metadata *zmd,
struct dm_zone *dzone)
{
struct dm_zone *bzone;
dmz_lock_map(zmd);
again:
bzone = dzone->bzone;
if (bzone)
goto out;
/* Alloate a random zone */
bzone = dmz_alloc_zone(zmd, DMZ_ALLOC_RND);
if (!bzone) {
dmz_wait_for_free_zones(zmd);
goto again;
}
/* Update the chunk mapping */
dmz_set_chunk_mapping(zmd, dzone->chunk, dmz_id(zmd, dzone),
dmz_id(zmd, bzone));
set_bit(DMZ_BUF, &bzone->flags);
bzone->chunk = dzone->chunk;
bzone->bzone = dzone;
dzone->bzone = bzone;
list_add_tail(&bzone->link, &zmd->map_rnd_list);
out:
dmz_unlock_map(zmd);
return bzone;
}
/*
* Get an unmapped (free) zone.
* This must be called with the mapping lock held.
*/
struct dm_zone *dmz_alloc_zone(struct dmz_metadata *zmd, unsigned long flags)
{
struct list_head *list;
struct dm_zone *zone;
if (flags & DMZ_ALLOC_RND)
list = &zmd->unmap_rnd_list;
else
list = &zmd->unmap_seq_list;
again:
if (list_empty(list)) {
/*
* No free zone: if this is for reclaim, allow using the
* reserved sequential zones.
*/
if (!(flags & DMZ_ALLOC_RECLAIM) ||
list_empty(&zmd->reserved_seq_zones_list))
return NULL;
zone = list_first_entry(&zmd->reserved_seq_zones_list,
struct dm_zone, link);
list_del_init(&zone->link);
atomic_dec(&zmd->nr_reserved_seq_zones);
return zone;
}
zone = list_first_entry(list, struct dm_zone, link);
list_del_init(&zone->link);
if (dmz_is_rnd(zone))
atomic_dec(&zmd->unmap_nr_rnd);
else
atomic_dec(&zmd->unmap_nr_seq);
if (dmz_is_offline(zone)) {
dmz_dev_warn(zmd->dev, "Zone %u is offline", dmz_id(zmd, zone));
zone = NULL;
goto again;
}
return zone;
}
/*
* Free a zone.
* This must be called with the mapping lock held.
*/
void dmz_free_zone(struct dmz_metadata *zmd, struct dm_zone *zone)
{
/* If this is a sequential zone, reset it */
if (dmz_is_seq(zone))
dmz_reset_zone(zmd, zone);
/* Return the zone to its type unmap list */
if (dmz_is_rnd(zone)) {
list_add_tail(&zone->link, &zmd->unmap_rnd_list);
atomic_inc(&zmd->unmap_nr_rnd);
} else if (atomic_read(&zmd->nr_reserved_seq_zones) <
zmd->nr_reserved_seq) {
list_add_tail(&zone->link, &zmd->reserved_seq_zones_list);
atomic_inc(&zmd->nr_reserved_seq_zones);
} else {
list_add_tail(&zone->link, &zmd->unmap_seq_list);
atomic_inc(&zmd->unmap_nr_seq);
}
wake_up_all(&zmd->free_wq);
}
/*
* Map a chunk to a zone.
* This must be called with the mapping lock held.
*/
void dmz_map_zone(struct dmz_metadata *zmd, struct dm_zone *dzone,
unsigned int chunk)
{
/* Set the chunk mapping */
dmz_set_chunk_mapping(zmd, chunk, dmz_id(zmd, dzone),
DMZ_MAP_UNMAPPED);
dzone->chunk = chunk;
if (dmz_is_rnd(dzone))
list_add_tail(&dzone->link, &zmd->map_rnd_list);
else
list_add_tail(&dzone->link, &zmd->map_seq_list);
}
/*
* Unmap a zone.
* This must be called with the mapping lock held.
*/
void dmz_unmap_zone(struct dmz_metadata *zmd, struct dm_zone *zone)
{
unsigned int chunk = zone->chunk;
unsigned int dzone_id;
if (chunk == DMZ_MAP_UNMAPPED) {
/* Already unmapped */
return;
}
if (test_and_clear_bit(DMZ_BUF, &zone->flags)) {
/*
* Unmapping the chunk buffer zone: clear only
* the chunk buffer mapping
*/
dzone_id = dmz_id(zmd, zone->bzone);
zone->bzone->bzone = NULL;
zone->bzone = NULL;
} else {
/*
* Unmapping the chunk data zone: the zone must
* not be buffered.
*/
if (WARN_ON(zone->bzone)) {
zone->bzone->bzone = NULL;
zone->bzone = NULL;
}
dzone_id = DMZ_MAP_UNMAPPED;
}
dmz_set_chunk_mapping(zmd, chunk, dzone_id, DMZ_MAP_UNMAPPED);
zone->chunk = DMZ_MAP_UNMAPPED;
list_del_init(&zone->link);
}
/*
* Set @nr_bits bits in @bitmap starting from @bit.
* Return the number of bits changed from 0 to 1.
*/
static unsigned int dmz_set_bits(unsigned long *bitmap,
unsigned int bit, unsigned int nr_bits)
{
unsigned long *addr;
unsigned int end = bit + nr_bits;
unsigned int n = 0;
while (bit < end) {
if (((bit & (BITS_PER_LONG - 1)) == 0) &&
((end - bit) >= BITS_PER_LONG)) {
/* Try to set the whole word at once */
addr = bitmap + BIT_WORD(bit);
if (*addr == 0) {
*addr = ULONG_MAX;
n += BITS_PER_LONG;
bit += BITS_PER_LONG;
continue;
}
}
if (!test_and_set_bit(bit, bitmap))
n++;
bit++;
}
return n;
}
/*
* Get the bitmap block storing the bit for chunk_block in zone.
*/
static struct dmz_mblock *dmz_get_bitmap(struct dmz_metadata *zmd,
struct dm_zone *zone,
sector_t chunk_block)
{
sector_t bitmap_block = 1 + zmd->nr_map_blocks +
(sector_t)(dmz_id(zmd, zone) * zmd->zone_nr_bitmap_blocks) +
(chunk_block >> DMZ_BLOCK_SHIFT_BITS);
return dmz_get_mblock(zmd, bitmap_block);
}
/*
* Copy the valid blocks bitmap of from_zone to the bitmap of to_zone.
*/
int dmz_copy_valid_blocks(struct dmz_metadata *zmd, struct dm_zone *from_zone,
struct dm_zone *to_zone)
{
struct dmz_mblock *from_mblk, *to_mblk;
sector_t chunk_block = 0;
/* Get the zones bitmap blocks */
while (chunk_block < zmd->dev->zone_nr_blocks) {
from_mblk = dmz_get_bitmap(zmd, from_zone, chunk_block);
if (IS_ERR(from_mblk))
return PTR_ERR(from_mblk);
to_mblk = dmz_get_bitmap(zmd, to_zone, chunk_block);
if (IS_ERR(to_mblk)) {
dmz_release_mblock(zmd, from_mblk);
return PTR_ERR(to_mblk);
}
memcpy(to_mblk->data, from_mblk->data, DMZ_BLOCK_SIZE);
dmz_dirty_mblock(zmd, to_mblk);
dmz_release_mblock(zmd, to_mblk);
dmz_release_mblock(zmd, from_mblk);
chunk_block += DMZ_BLOCK_SIZE_BITS;
}
to_zone->weight = from_zone->weight;
return 0;
}
/*
* Merge the valid blocks bitmap of from_zone into the bitmap of to_zone,
* starting from chunk_block.
*/
int dmz_merge_valid_blocks(struct dmz_metadata *zmd, struct dm_zone *from_zone,
struct dm_zone *to_zone, sector_t chunk_block)
{
unsigned int nr_blocks;
int ret;
/* Get the zones bitmap blocks */
while (chunk_block < zmd->dev->zone_nr_blocks) {
/* Get a valid region from the source zone */
ret = dmz_first_valid_block(zmd, from_zone, &chunk_block);
if (ret <= 0)
return ret;
nr_blocks = ret;
ret = dmz_validate_blocks(zmd, to_zone, chunk_block, nr_blocks);
if (ret)
return ret;
chunk_block += nr_blocks;
}
return 0;
}
/*
* Validate all the blocks in the range [block..block+nr_blocks-1].
*/
int dmz_validate_blocks(struct dmz_metadata *zmd, struct dm_zone *zone,
sector_t chunk_block, unsigned int nr_blocks)
{
unsigned int count, bit, nr_bits;
unsigned int zone_nr_blocks = zmd->dev->zone_nr_blocks;
struct dmz_mblock *mblk;
unsigned int n = 0;
dmz_dev_debug(zmd->dev, "=> VALIDATE zone %u, block %llu, %u blocks",
dmz_id(zmd, zone), (unsigned long long)chunk_block,
nr_blocks);
WARN_ON(chunk_block + nr_blocks > zone_nr_blocks);
while (nr_blocks) {
/* Get bitmap block */
mblk = dmz_get_bitmap(zmd, zone, chunk_block);
if (IS_ERR(mblk))
return PTR_ERR(mblk);
/* Set bits */
bit = chunk_block & DMZ_BLOCK_MASK_BITS;
nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit);
count = dmz_set_bits((unsigned long *)mblk->data, bit, nr_bits);
if (count) {
dmz_dirty_mblock(zmd, mblk);
n += count;
}
dmz_release_mblock(zmd, mblk);
nr_blocks -= nr_bits;
chunk_block += nr_bits;
}
if (likely(zone->weight + n <= zone_nr_blocks))
zone->weight += n;
else {
dmz_dev_warn(zmd->dev, "Zone %u: weight %u should be <= %u",
dmz_id(zmd, zone), zone->weight,
zone_nr_blocks - n);
zone->weight = zone_nr_blocks;
}
return 0;
}
/*
* Clear nr_bits bits in bitmap starting from bit.
* Return the number of bits cleared.
*/
static int dmz_clear_bits(unsigned long *bitmap, int bit, int nr_bits)
{
unsigned long *addr;
int end = bit + nr_bits;
int n = 0;
while (bit < end) {
if (((bit & (BITS_PER_LONG - 1)) == 0) &&
((end - bit) >= BITS_PER_LONG)) {
/* Try to clear whole word at once */
addr = bitmap + BIT_WORD(bit);
if (*addr == ULONG_MAX) {
*addr = 0;
n += BITS_PER_LONG;
bit += BITS_PER_LONG;
continue;
}
}
if (test_and_clear_bit(bit, bitmap))
n++;
bit++;
}
return n;
}
/*
* Invalidate all the blocks in the range [block..block+nr_blocks-1].
*/
int dmz_invalidate_blocks(struct dmz_metadata *zmd, struct dm_zone *zone,
sector_t chunk_block, unsigned int nr_blocks)
{
unsigned int count, bit, nr_bits;
struct dmz_mblock *mblk;
unsigned int n = 0;
dmz_dev_debug(zmd->dev, "=> INVALIDATE zone %u, block %llu, %u blocks",
dmz_id(zmd, zone), (u64)chunk_block, nr_blocks);
WARN_ON(chunk_block + nr_blocks > zmd->dev->zone_nr_blocks);
while (nr_blocks) {
/* Get bitmap block */
mblk = dmz_get_bitmap(zmd, zone, chunk_block);
if (IS_ERR(mblk))
return PTR_ERR(mblk);
/* Clear bits */
bit = chunk_block & DMZ_BLOCK_MASK_BITS;
nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit);
count = dmz_clear_bits((unsigned long *)mblk->data,
bit, nr_bits);
if (count) {
dmz_dirty_mblock(zmd, mblk);
n += count;
}
dmz_release_mblock(zmd, mblk);
nr_blocks -= nr_bits;
chunk_block += nr_bits;
}
if (zone->weight >= n)
zone->weight -= n;
else {
dmz_dev_warn(zmd->dev, "Zone %u: weight %u should be >= %u",
dmz_id(zmd, zone), zone->weight, n);
zone->weight = 0;
}
return 0;
}
/*
* Get a block bit value.
*/
static int dmz_test_block(struct dmz_metadata *zmd, struct dm_zone *zone,
sector_t chunk_block)
{
struct dmz_mblock *mblk;
int ret;
WARN_ON(chunk_block >= zmd->dev->zone_nr_blocks);
/* Get bitmap block */
mblk = dmz_get_bitmap(zmd, zone, chunk_block);
if (IS_ERR(mblk))
return PTR_ERR(mblk);
/* Get offset */
ret = test_bit(chunk_block & DMZ_BLOCK_MASK_BITS,
(unsigned long *) mblk->data) != 0;
dmz_release_mblock(zmd, mblk);
return ret;
}
/*
* Return the number of blocks from chunk_block to the first block with a bit
* value specified by set. Search at most nr_blocks blocks from chunk_block.
*/
static int dmz_to_next_set_block(struct dmz_metadata *zmd, struct dm_zone *zone,
sector_t chunk_block, unsigned int nr_blocks,
int set)
{
struct dmz_mblock *mblk;
unsigned int bit, set_bit, nr_bits;
unsigned long *bitmap;
int n = 0;
WARN_ON(chunk_block + nr_blocks > zmd->dev->zone_nr_blocks);
while (nr_blocks) {
/* Get bitmap block */
mblk = dmz_get_bitmap(zmd, zone, chunk_block);
if (IS_ERR(mblk))
return PTR_ERR(mblk);
/* Get offset */
bitmap = (unsigned long *) mblk->data;
bit = chunk_block & DMZ_BLOCK_MASK_BITS;
nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit);
if (set)
set_bit = find_next_bit(bitmap, DMZ_BLOCK_SIZE_BITS, bit);
else
set_bit = find_next_zero_bit(bitmap, DMZ_BLOCK_SIZE_BITS, bit);
dmz_release_mblock(zmd, mblk);
n += set_bit - bit;
if (set_bit < DMZ_BLOCK_SIZE_BITS)
break;
nr_blocks -= nr_bits;
chunk_block += nr_bits;
}
return n;
}
/*
* Test if chunk_block is valid. If it is, the number of consecutive
* valid blocks from chunk_block will be returned.
*/
int dmz_block_valid(struct dmz_metadata *zmd, struct dm_zone *zone,
sector_t chunk_block)
{
int valid;
valid = dmz_test_block(zmd, zone, chunk_block);
if (valid <= 0)
return valid;
/* The block is valid: get the number of valid blocks from block */
return dmz_to_next_set_block(zmd, zone, chunk_block,
zmd->dev->zone_nr_blocks - chunk_block, 0);
}
/*
* Find the first valid block from @chunk_block in @zone.
* If such a block is found, its number is returned using
* @chunk_block and the total number of valid blocks from @chunk_block
* is returned.
*/
int dmz_first_valid_block(struct dmz_metadata *zmd, struct dm_zone *zone,
sector_t *chunk_block)
{
sector_t start_block = *chunk_block;
int ret;
ret = dmz_to_next_set_block(zmd, zone, start_block,
zmd->dev->zone_nr_blocks - start_block, 1);
if (ret < 0)
return ret;
start_block += ret;
*chunk_block = start_block;
return dmz_to_next_set_block(zmd, zone, start_block,
zmd->dev->zone_nr_blocks - start_block, 0);
}
/*
* Count the number of bits set starting from bit up to bit + nr_bits - 1.
*/
static int dmz_count_bits(void *bitmap, int bit, int nr_bits)
{
unsigned long *addr;
int end = bit + nr_bits;
int n = 0;
while (bit < end) {
if (((bit & (BITS_PER_LONG - 1)) == 0) &&
((end - bit) >= BITS_PER_LONG)) {
addr = (unsigned long *)bitmap + BIT_WORD(bit);
if (*addr == ULONG_MAX) {
n += BITS_PER_LONG;
bit += BITS_PER_LONG;
continue;
}
}
if (test_bit(bit, bitmap))
n++;
bit++;
}
return n;
}
/*
* Get a zone weight.
*/
static void dmz_get_zone_weight(struct dmz_metadata *zmd, struct dm_zone *zone)
{
struct dmz_mblock *mblk;
sector_t chunk_block = 0;
unsigned int bit, nr_bits;
unsigned int nr_blocks = zmd->dev->zone_nr_blocks;
void *bitmap;
int n = 0;
while (nr_blocks) {
/* Get bitmap block */
mblk = dmz_get_bitmap(zmd, zone, chunk_block);
if (IS_ERR(mblk)) {
n = 0;
break;
}
/* Count bits in this block */
bitmap = mblk->data;
bit = chunk_block & DMZ_BLOCK_MASK_BITS;
nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit);
n += dmz_count_bits(bitmap, bit, nr_bits);
dmz_release_mblock(zmd, mblk);
nr_blocks -= nr_bits;
chunk_block += nr_bits;
}
zone->weight = n;
}
/*
* Cleanup the zoned metadata resources.
*/
static void dmz_cleanup_metadata(struct dmz_metadata *zmd)
{
struct rb_root *root;
struct dmz_mblock *mblk, *next;
int i;
/* Release zone mapping resources */
if (zmd->map_mblk) {
for (i = 0; i < zmd->nr_map_blocks; i++)
dmz_release_mblock(zmd, zmd->map_mblk[i]);
kfree(zmd->map_mblk);
zmd->map_mblk = NULL;
}
/* Release super blocks */
for (i = 0; i < 2; i++) {
if (zmd->sb[i].mblk) {
dmz_free_mblock(zmd, zmd->sb[i].mblk);
zmd->sb[i].mblk = NULL;
}
}
/* Free cached blocks */
while (!list_empty(&zmd->mblk_dirty_list)) {
mblk = list_first_entry(&zmd->mblk_dirty_list,
struct dmz_mblock, link);
dmz_dev_warn(zmd->dev, "mblock %llu still in dirty list (ref %u)",
dm zoned: fix metadata block ref counting Since the ref field of struct dmz_mblock is always used with the spinlock of struct dmz_metadata locked, there is no need to use an atomic_t type. Change the type of the ref field to an unsigne integer. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:07 +08:00
(u64)mblk->no, mblk->ref);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
list_del_init(&mblk->link);
rb_erase(&mblk->node, &zmd->mblk_rbtree);
dmz_free_mblock(zmd, mblk);
}
while (!list_empty(&zmd->mblk_lru_list)) {
mblk = list_first_entry(&zmd->mblk_lru_list,
struct dmz_mblock, link);
list_del_init(&mblk->link);
rb_erase(&mblk->node, &zmd->mblk_rbtree);
dmz_free_mblock(zmd, mblk);
}
/* Sanity checks: the mblock rbtree should now be empty */
root = &zmd->mblk_rbtree;
rbtree_postorder_for_each_entry_safe(mblk, next, root, node) {
dmz_dev_warn(zmd->dev, "mblock %llu ref %u still in rbtree",
dm zoned: fix metadata block ref counting Since the ref field of struct dmz_mblock is always used with the spinlock of struct dmz_metadata locked, there is no need to use an atomic_t type. Change the type of the ref field to an unsigne integer. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-10-17 17:05:07 +08:00
(u64)mblk->no, mblk->ref);
mblk->ref = 0;
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
dmz_free_mblock(zmd, mblk);
}
/* Free the zone descriptors */
dmz_drop_zones(zmd);
dm: backfill missing calls to mutex_destroy() Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2018-01-06 10:17:20 +08:00
mutex_destroy(&zmd->mblk_flush_lock);
mutex_destroy(&zmd->map_lock);
dm zoned: drive-managed zoned block device target The dm-zoned device mapper target provides transparent write access to zoned block devices (ZBC and ZAC compliant block devices). dm-zoned hides to the device user (a file system or an application doing raw block device accesses) any constraint imposed on write requests by the device, equivalent to a drive-managed zoned block device model. Write requests are processed using a combination of on-disk buffering using the device conventional zones and direct in-place processing for requests aligned to a zone sequential write pointer position. A background reclaim process implemented using dm_kcopyd_copy ensures that conventional zones are always available for executing unaligned write requests. The reclaim process overhead is minimized by managing buffer zones in a least-recently-written order and first targeting the oldest buffer zones. Doing so, blocks under regular write access (such as metadata blocks of a file system) remain stored in conventional zones, resulting in no apparent overhead. dm-zoned implementation focus on simplicity and on minimizing overhead (CPU, memory and storage overhead). For a 14TB host-managed disk with 256 MB zones, dm-zoned memory usage per disk instance is at most about 3 MB and as little as 5 zones will be used internally for storing metadata and performing buffer zone reclaim operations. This is achieved using zone level indirection rather than a full block indirection system for managing block movement between zones. dm-zoned primary target is host-managed zoned block devices but it can also be used with host-aware device models to mitigate potential device-side performance degradation due to excessive random writing. Zoned block devices can be formatted and checked for use with the dm-zoned target using the dmzadm utility available at: https://github.com/hgst/dm-zoned-tools Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [Mike Snitzer partly refactored Damien's original work to cleanup the code] Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-06-07 14:55:39 +08:00
}
/*
* Initialize the zoned metadata.
*/
int dmz_ctr_metadata(struct dmz_dev *dev, struct dmz_metadata **metadata)
{
struct dmz_metadata *zmd;
unsigned int i, zid;
struct dm_zone *zone;
int ret;
zmd = kzalloc(sizeof(struct dmz_metadata), GFP_KERNEL);
if (!zmd)
return -ENOMEM;
zmd->dev = dev;
zmd->mblk_rbtree = RB_ROOT;
init_rwsem(&zmd->mblk_sem);
mutex_init(&zmd->mblk_flush_lock);
spin_lock_init(&zmd->mblk_lock);
INIT_LIST_HEAD(&zmd->mblk_lru_list);
INIT_LIST_HEAD(&zmd->mblk_dirty_list);
mutex_init(&zmd->map_lock);
atomic_set(&zmd->unmap_nr_rnd, 0);
INIT_LIST_HEAD(&zmd->unmap_rnd_list);
INIT_LIST_HEAD(&zmd->map_rnd_list);
atomic_set(&zmd->unmap_nr_seq, 0);
INIT_LIST_HEAD(&zmd->unmap_seq_list);
INIT_LIST_HEAD(&zmd->map_seq_list);
atomic_set(&zmd->nr_reserved_seq_zones, 0);
INIT_LIST_HEAD(&zmd->reserved_seq_zones_list);
init_waitqueue_head(&zmd->free_wq);
/* Initialize zone descriptors */
ret = dmz_init_zones(zmd);
if (ret)
goto err;
/* Get super block */
ret = dmz_load_sb(zmd);
if (ret)
goto err;
/* Set metadata zones starting from sb_zone */
zid = dmz_id(zmd, zmd->sb_zone);
for (i = 0; i < zmd->nr_meta_zones << 1; i++) {
zone = dmz_get(zmd, zid + i);
if (!dmz_is_rnd(zone))
goto err;
set_bit(DMZ_META, &zone->flags);
}
/* Load mapping table */
ret = dmz_load_mapping(zmd);
if (ret)
goto err;
/*
* Cache size boundaries: allow at least 2 super blocks, the chunk map
* blocks and enough blocks to be able to cache the bitmap blocks of
* up to 16 zones when idle (min_nr_mblks). Otherwise, if busy, allow
* the cache to add 512 more metadata blocks.
*/
zmd->min_nr_mblks = 2 + zmd->nr_map_blocks + zmd->zone_nr_bitmap_blocks * 16;
zmd->max_nr_mblks = zmd->min_nr_mblks + 512;
zmd->mblk_shrinker.count_objects = dmz_mblock_shrinker_count;
zmd->mblk_shrinker.scan_objects = dmz_mblock_shrinker_scan;
zmd->mblk_shrinker.seeks = DEFAULT_SEEKS;
/* Metadata cache shrinker */
ret = register_shrinker(&zmd->mblk_shrinker);
if (ret) {
dmz_dev_err(dev, "Register metadata cache shrinker failed");
goto err;
}
dmz_dev_info(dev, "Host-%s zoned block device",
bdev_zoned_model(dev->bdev) == BLK_ZONED_HA ?
"aware" : "managed");
dmz_dev_info(dev, " %llu 512-byte logical sectors",
(u64)dev->capacity);
dmz_dev_info(dev, " %u zones of %llu 512-byte logical sectors",
dev->nr_zones, (u64)dev->zone_nr_sectors);
dmz_dev_info(dev, " %u metadata zones",
zmd->nr_meta_zones * 2);
dmz_dev_info(dev, " %u data zones for %u chunks",
zmd->nr_data_zones, zmd->nr_chunks);
dmz_dev_info(dev, " %u random zones (%u unmapped)",
zmd->nr_rnd, atomic_read(&zmd->unmap_nr_rnd));
dmz_dev_info(dev, " %u sequential zones (%u unmapped)",
zmd->nr_seq, atomic_read(&zmd->unmap_nr_seq));
dmz_dev_info(dev, " %u reserved sequential data zones",
zmd->nr_reserved_seq);
dmz_dev_debug(dev, "Format:");
dmz_dev_debug(dev, "%u metadata blocks per set (%u max cache)",
zmd->nr_meta_blocks, zmd->max_nr_mblks);
dmz_dev_debug(dev, " %u data zone mapping blocks",
zmd->nr_map_blocks);
dmz_dev_debug(dev, " %u bitmap blocks",
zmd->nr_bitmap_blocks);
*metadata = zmd;
return 0;
err:
dmz_cleanup_metadata(zmd);
kfree(zmd);
*metadata = NULL;
return ret;
}
/*
* Cleanup the zoned metadata resources.
*/
void dmz_dtr_metadata(struct dmz_metadata *zmd)
{
unregister_shrinker(&zmd->mblk_shrinker);
dmz_cleanup_metadata(zmd);
kfree(zmd);
}
/*
* Check zone information on resume.
*/
int dmz_resume_metadata(struct dmz_metadata *zmd)
{
struct dmz_dev *dev = zmd->dev;
struct dm_zone *zone;
sector_t wp_block;
unsigned int i;
int ret;
/* Check zones */
for (i = 0; i < dev->nr_zones; i++) {
zone = dmz_get(zmd, i);
if (!zone) {
dmz_dev_err(dev, "Unable to get zone %u", i);
return -EIO;
}
wp_block = zone->wp_block;
ret = dmz_update_zone(zmd, zone);
if (ret) {
dmz_dev_err(dev, "Broken zone %u", i);
return ret;
}
if (dmz_is_offline(zone)) {
dmz_dev_warn(dev, "Zone %u is offline", i);
continue;
}
/* Check write pointer */
if (!dmz_is_seq(zone))
zone->wp_block = 0;
else if (zone->wp_block != wp_block) {
dmz_dev_err(dev, "Zone %u: Invalid wp (%llu / %llu)",
i, (u64)zone->wp_block, (u64)wp_block);
zone->wp_block = wp_block;
dmz_invalidate_blocks(zmd, zone, zone->wp_block,
dev->zone_nr_blocks - zone->wp_block);
}
}
return 0;
}
Copy Permalink