zio.h revision 44cd46cadd9aab751dae6a4023c1cb5bf316d274
2N/A/*
2N/A * CDDL HEADER START
2N/A *
2N/A * The contents of this file are subject to the terms of the
2N/A * Common Development and Distribution License (the "License").
2N/A * You may not use this file except in compliance with the License.
2N/A *
2N/A * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
2N/A * or http://www.opensolaris.org/os/licensing.
2N/A * See the License for the specific language governing permissions
2N/A * and limitations under the License.
2N/A *
2N/A * When distributing Covered Code, include this CDDL HEADER in each
2N/A * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
2N/A * If applicable, add the following below this CDDL HEADER, with the
2N/A * fields enclosed by brackets "[]" replaced with your own identifying
2N/A * information: Portions Copyright [yyyy] [name of copyright owner]
2N/A *
2N/A * CDDL HEADER END
2N/A */
2N/A/*
2N/A * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
2N/A * Use is subject to license terms.
2N/A */
2N/A
2N/A#ifndef _ZIO_H
2N/A#define _ZIO_H
2N/A
2N/A#pragma ident "%Z%%M% %I% %E% SMI"
2N/A
2N/A#include <sys/zfs_context.h>
2N/A#include <sys/spa.h>
2N/A#include <sys/txg.h>
2N/A#include <sys/avl.h>
2N/A#include <sys/dkio.h>
2N/A#include <sys/fs/zfs.h>
2N/A#include <sys/zio_impl.h>
2N/A
2N/A#ifdef __cplusplus
2N/Aextern "C" {
2N/A#endif
2N/A
2N/A#define ZBT_MAGIC 0x210da7ab10c7a11ULL /* zio data bloc tail */
2N/A
2N/Atypedef struct zio_block_tail {
2N/A uint64_t zbt_magic; /* for validation, endianness */
2N/A zio_cksum_t zbt_cksum; /* 256-bit checksum */
2N/A} zio_block_tail_t;
2N/A
2N/A/*
2N/A * Gang block headers are self-checksumming and contain an array
2N/A * of block pointers.
2N/A */
2N/A#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
2N/A#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
2N/A sizeof (zio_block_tail_t)) / sizeof (blkptr_t))
2N/A#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
2N/A sizeof (zio_block_tail_t) - \
2N/A (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
2N/A sizeof (uint64_t))
2N/A
2N/A#define ZIO_GET_IOSIZE(zio) \
2N/A (BP_IS_GANG((zio)->io_bp) ? \
2N/A SPA_GANGBLOCKSIZE : BP_GET_PSIZE((zio)->io_bp))
2N/A
2N/Atypedef struct zio_gbh {
2N/A blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
2N/A uint64_t zg_filler[SPA_GBH_FILLER];
2N/A zio_block_tail_t zg_tail;
2N/A} zio_gbh_phys_t;
2N/A
2N/Aenum zio_checksum {
2N/A ZIO_CHECKSUM_INHERIT = 0,
2N/A ZIO_CHECKSUM_ON,
2N/A ZIO_CHECKSUM_OFF,
2N/A ZIO_CHECKSUM_LABEL,
2N/A ZIO_CHECKSUM_GANG_HEADER,
2N/A ZIO_CHECKSUM_ZILOG,
2N/A ZIO_CHECKSUM_FLETCHER_2,
2N/A ZIO_CHECKSUM_FLETCHER_4,
2N/A ZIO_CHECKSUM_SHA256,
2N/A ZIO_CHECKSUM_FUNCTIONS
2N/A};
2N/A
2N/A#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_2
2N/A#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
2N/A
2N/Aenum zio_compress {
2N/A ZIO_COMPRESS_INHERIT = 0,
2N/A ZIO_COMPRESS_ON,
2N/A ZIO_COMPRESS_OFF,
2N/A ZIO_COMPRESS_LZJB,
2N/A ZIO_COMPRESS_FUNCTIONS
2N/A};
2N/A
2N/A#define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
2N/A#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
2N/A
2N/A#define ZIO_PRIORITY_NOW (zio_priority_table[0])
2N/A#define ZIO_PRIORITY_SYNC_READ (zio_priority_table[1])
2N/A#define ZIO_PRIORITY_SYNC_WRITE (zio_priority_table[2])
2N/A#define ZIO_PRIORITY_ASYNC_READ (zio_priority_table[3])
2N/A#define ZIO_PRIORITY_ASYNC_WRITE (zio_priority_table[4])
2N/A#define ZIO_PRIORITY_FREE (zio_priority_table[5])
2N/A#define ZIO_PRIORITY_CACHE_FILL (zio_priority_table[6])
2N/A#define ZIO_PRIORITY_LOG_WRITE (zio_priority_table[7])
2N/A#define ZIO_PRIORITY_RESILVER (zio_priority_table[8])
2N/A#define ZIO_PRIORITY_SCRUB (zio_priority_table[9])
2N/A#define ZIO_PRIORITY_TABLE_SIZE 10
2N/A
2N/A#define ZIO_FLAG_MUSTSUCCEED 0x00000
2N/A#define ZIO_FLAG_CANFAIL 0x00001
2N/A#define ZIO_FLAG_FAILFAST 0x00002
2N/A#define ZIO_FLAG_CONFIG_HELD 0x00004
2N/A
2N/A#define ZIO_FLAG_DONT_CACHE 0x00010
2N/A#define ZIO_FLAG_DONT_QUEUE 0x00020
2N/A#define ZIO_FLAG_DONT_PROPAGATE 0x00040
2N/A#define ZIO_FLAG_DONT_RETRY 0x00080
2N/A
2N/A#define ZIO_FLAG_PHYSICAL 0x00100
2N/A#define ZIO_FLAG_IO_BYPASS 0x00200
2N/A#define ZIO_FLAG_IO_REPAIR 0x00400
2N/A#define ZIO_FLAG_SPECULATIVE 0x00800
2N/A
2N/A#define ZIO_FLAG_RESILVER 0x01000
2N/A#define ZIO_FLAG_SCRUB 0x02000
2N/A#define ZIO_FLAG_SUBBLOCK 0x04000
2N/A
2N/A#define ZIO_FLAG_NOBOOKMARK 0x10000
2N/A
2N/A#define ZIO_FLAG_GANG_INHERIT \
2N/A (ZIO_FLAG_CANFAIL | \
2N/A ZIO_FLAG_FAILFAST | \
2N/A ZIO_FLAG_CONFIG_HELD | \
2N/A ZIO_FLAG_DONT_RETRY | \
2N/A ZIO_FLAG_IO_REPAIR | \
2N/A ZIO_FLAG_SPECULATIVE | \
2N/A ZIO_FLAG_RESILVER | \
2N/A ZIO_FLAG_SCRUB)
2N/A
2N/A#define ZIO_FLAG_VDEV_INHERIT \
2N/A (ZIO_FLAG_GANG_INHERIT | \
2N/A ZIO_FLAG_DONT_CACHE | \
2N/A ZIO_FLAG_PHYSICAL)
2N/A
2N/A/*
2N/A * We'll take the unused errno 'EBADE' (from the Convergent graveyard)
2N/A * to indicate checksum errors.
2N/A */
2N/A#define ECKSUM EBADE
2N/A
2N/Atypedef struct zio zio_t;
2N/Atypedef void zio_done_func_t(zio_t *zio);
2N/A
2N/Aextern uint8_t zio_priority_table[ZIO_PRIORITY_TABLE_SIZE];
2N/Aextern char *zio_type_name[ZIO_TYPES];
2N/A
2N/A/*
2N/A * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
2N/A * identifies any block in the pool. By convention, the meta-objset (MOS)
2N/A * is objset 0, the meta-dnode is object 0, the root block (osphys_t) is
2N/A * level -1 of the meta-dnode, and intent log blocks (which are chained
2N/A * off the root block) have blkid == sequence number. In summary:
2N/A *
2N/A * mos is objset 0
2N/A * meta-dnode is object 0
2N/A * root block is <objset, 0, -1, 0>
2N/A * intent log is <objset, 0, -1, ZIL sequence number>
2N/A *
2N/A * Note: this structure is called a bookmark because its first purpose was
2N/A * to remember where to resume a pool-wide traverse. The absolute ordering
2N/A * for block visitation during traversal is defined in compare_bookmark().
2N/A *
2N/A * Note: this structure is passed between userland and the kernel.
2N/A * Therefore it must not change size or alignment between 32/64 bit
2N/A * compilation options.
2N/A */
2N/Atypedef struct zbookmark {
2N/A uint64_t zb_objset;
2N/A uint64_t zb_object;
2N/A int64_t zb_level;
2N/A uint64_t zb_blkid;
2N/A} zbookmark_t;
2N/A
2N/Astruct zio {
2N/A /* Core information about this I/O */
2N/A zio_t *io_parent;
2N/A zio_t *io_root;
2N/A spa_t *io_spa;
2N/A zbookmark_t io_bookmark;
2N/A enum zio_checksum io_checksum;
2N/A enum zio_compress io_compress;
2N/A int io_ndvas;
2N/A uint64_t io_txg;
2N/A blkptr_t *io_bp;
2N/A blkptr_t io_bp_copy;
2N/A zio_t *io_child;
2N/A zio_t *io_sibling_prev;
2N/A zio_t *io_sibling_next;
2N/A zio_transform_t *io_transform_stack;
2N/A zio_t *io_logical;
2N/A
2N/A /* Callback info */
2N/A zio_done_func_t *io_done;
2N/A void *io_private;
2N/A blkptr_t io_bp_orig;
/* Data represented by this I/O */
void *io_data;
uint64_t io_size;
/* Stuff for the vdev stack */
vdev_t *io_vd;
void *io_vsd;
uint64_t io_offset;
uint64_t io_deadline;
uint64_t io_timestamp;
avl_node_t io_offset_node;
avl_node_t io_deadline_node;
avl_tree_t *io_vdev_tree;
zio_t *io_delegate_list;
zio_t *io_delegate_next;
/* Internal pipeline state */
int io_flags;
enum zio_type io_type;
enum zio_stage io_stage;
uint8_t io_stalled;
uint8_t io_priority;
struct dk_callback io_dk_callback;
int io_cmd;
int io_retries;
int io_error;
uint32_t io_numerrors;
uint32_t io_pipeline;
uint32_t io_async_stages;
uint64_t io_children_notready;
uint64_t io_children_notdone;
void *io_waiter;
kmutex_t io_lock;
kcondvar_t io_cv;
/* FMA state */
uint64_t io_ena;
};
extern zio_t *zio_null(zio_t *pio, spa_t *spa,
zio_done_func_t *done, void *private, int flags);
extern zio_t *zio_root(spa_t *spa,
zio_done_func_t *done, void *private, int flags);
extern zio_t *zio_read(zio_t *pio, spa_t *spa, blkptr_t *bp, void *data,
uint64_t size, zio_done_func_t *done, void *private,
int priority, int flags, zbookmark_t *zb);
extern zio_t *zio_write(zio_t *pio, spa_t *spa, int checksum, int compress,
int ncopies, uint64_t txg, blkptr_t *bp, void *data, uint64_t size,
zio_done_func_t *done, void *private, int priority, int flags,
zbookmark_t *zb);
extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, int checksum,
uint64_t txg, blkptr_t *bp, void *data, uint64_t size,
zio_done_func_t *done, void *private, int priority, int flags,
zbookmark_t *zb);
extern zio_t *zio_free(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
zio_done_func_t *done, void *private);
extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
zio_done_func_t *done, void *private);
extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
zio_done_func_t *done, void *private, int priority, int flags);
extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
uint64_t size, void *data, int checksum,
zio_done_func_t *done, void *private, int priority, int flags);
extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
uint64_t size, void *data, int checksum,
zio_done_func_t *done, void *private, int priority, int flags);
extern int zio_alloc_blk(spa_t *spa, int checksum, uint64_t size,
blkptr_t *bp, uint64_t txg);
extern void zio_free_blk(spa_t *spa, blkptr_t *bp, uint64_t txg);
extern int zio_wait(zio_t *zio);
extern void zio_nowait(zio_t *zio);
extern void *zio_buf_alloc(size_t size);
extern void zio_buf_free(void *buf, size_t size);
/*
* Move an I/O to the next stage of the pipeline and execute that stage.
* There's no locking on io_stage because there's no legitimate way for
* multiple threads to be attempting to process the same I/O.
*/
extern void zio_next_stage(zio_t *zio);
extern void zio_next_stage_async(zio_t *zio);
extern void zio_wait_children_done(zio_t *zio);
/*
* Delegate I/O to a child vdev.
*/
extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
uint64_t offset, void *data, uint64_t size, int type, int priority,
int flags, zio_done_func_t *done, void *private);
extern void zio_vdev_io_bypass(zio_t *zio);
extern void zio_vdev_io_reissue(zio_t *zio);
extern void zio_vdev_io_redone(zio_t *zio);
extern void zio_checksum_verified(zio_t *zio);
extern void zio_set_gang_verifier(zio_t *zio, zio_cksum_t *zcp);
extern uint8_t zio_checksum_select(uint8_t child, uint8_t parent);
extern uint8_t zio_compress_select(uint8_t child, uint8_t parent);
boolean_t zio_should_retry(zio_t *zio);
/*
* Initial setup and teardown.
*/
extern void zio_init(void);
extern void zio_fini(void);
/*
* Fault injection
*/
struct zinject_record;
extern uint32_t zio_injection_enabled;
extern int zio_inject_fault(char *name, int flags, int *id,
struct zinject_record *record);
extern int zio_inject_list_next(int *id, char *name, size_t buflen,
struct zinject_record *record);
extern int zio_clear_fault(int id);
extern int zio_handle_fault_injection(zio_t *zio, int error);
extern int zio_handle_device_injection(vdev_t *vd, int error);
#ifdef __cplusplus
}
#endif
#endif /* _ZIO_H */