#define __XFS_LOG_PRIV_H__
struct xfs_buf;
-struct ktrace;
struct log;
struct xlog_ticket;
struct xfs_buf_cancel;
#define XLOG_HEADER_SIZE 512
#define XLOG_REC_SHIFT(log) \
- BTOBB(1 << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
+ BTOBB(1 << (xfs_sb_version_haslogv2(&log->l_mp->m_sb) ? \
XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
#define XLOG_TOTAL_REC_SHIFT(log) \
- BTOBB(XLOG_MAX_ICLOGS << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
+ BTOBB(XLOG_MAX_ICLOGS << (xfs_sb_version_haslogv2(&log->l_mp->m_sb) ? \
XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
-/*
- * set lsns
- */
-#define ASSIGN_ANY_LSN_HOST(lsn,cycle,block) \
- { \
- (lsn) = ((xfs_lsn_t)(cycle)<<32)|(block); \
- }
-#define ASSIGN_ANY_LSN_DISK(lsn,cycle,block) \
- { \
- INT_SET(((uint *)&(lsn))[0], ARCH_CONVERT, (cycle)); \
- INT_SET(((uint *)&(lsn))[1], ARCH_CONVERT, (block)); \
- }
-#define ASSIGN_LSN(lsn,log) \
- ASSIGN_ANY_LSN_DISK(lsn,(log)->l_curr_cycle,(log)->l_curr_block);
-
-#define XLOG_SET(f,b) (((f) & (b)) == (b))
-
-#define GET_CYCLE(ptr, arch) \
- (INT_GET(*(uint *)(ptr), arch) == XLOG_HEADER_MAGIC_NUM ? \
- INT_GET(*((uint *)(ptr)+1), arch) : \
- INT_GET(*(uint *)(ptr), arch) \
- )
+static inline xfs_lsn_t xlog_assign_lsn(uint cycle, uint block)
+{
+ return ((xfs_lsn_t)cycle << 32) | block;
+}
-#define BLK_AVG(blk1, blk2) ((blk1+blk2) >> 1)
+static inline uint xlog_get_cycle(char *ptr)
+{
+ if (be32_to_cpu(*(__be32 *)ptr) == XLOG_HEADER_MAGIC_NUM)
+ return be32_to_cpu(*((__be32 *)ptr + 1));
+ else
+ return be32_to_cpu(*(__be32 *)ptr);
+}
+#define BLK_AVG(blk1, blk2) ((blk1+blk2) >> 1)
#ifdef __KERNEL__
*
* this has endian issues, of course.
*/
-
-#ifndef XFS_NATIVE_HOST
-#define GET_CLIENT_ID(i,arch) \
- ((i) & 0xff)
-#else
-#define GET_CLIENT_ID(i,arch) \
- ((i) >> 24)
-#endif
-
-#define GRANT_LOCK(log) mutex_spinlock(&(log)->l_grant_lock)
-#define GRANT_UNLOCK(log, s) mutex_spinunlock(&(log)->l_grant_lock, s)
+static inline uint xlog_get_client_id(__be32 i)
+{
+ return be32_to_cpu(i) >> 24;
+}
#define xlog_panic(args...) cmn_err(CE_PANIC, ## args)
#define xlog_exit(args...) cmn_err(CE_PANIC, ## args)
#define XLOG_TIC_INITED 0x1 /* has been initialized */
#define XLOG_TIC_PERM_RESERV 0x2 /* permanent reservation */
#define XLOG_TIC_IN_Q 0x4
+
+#define XLOG_TIC_FLAGS \
+ { XLOG_TIC_INITED, "XLOG_TIC_INITED" }, \
+ { XLOG_TIC_PERM_RESERV, "XLOG_TIC_PERM_RESERV" }, \
+ { XLOG_TIC_IN_Q, "XLOG_TIC_IN_Q" }
+
#endif /* __KERNEL__ */
#define XLOG_UNMOUNT_TYPE 0x556e /* Un for Unmount */
#define XLOG_RECOVERY_NEEDED 0x4 /* log was recovered */
#define XLOG_IO_ERROR 0x8 /* log hit an I/O error, and being
shutdown */
-typedef __uint32_t xlog_tid_t;
-
#ifdef __KERNEL__
/*
} xlog_res_t;
typedef struct xlog_ticket {
- sv_t t_sema; /* sleep on this semaphore : 20 */
- struct xlog_ticket *t_next; /* :4|8 */
+ sv_t t_wait; /* ticket wait queue : 20 */
+ struct xlog_ticket *t_next; /* :4|8 */
struct xlog_ticket *t_prev; /* :4|8 */
xlog_tid_t t_tid; /* transaction identifier : 4 */
+ atomic_t t_ref; /* ticket reference count : 4 */
int t_curr_res; /* current reservation in bytes : 4 */
int t_unit_res; /* unit reservation in bytes : 4 */
char t_ocnt; /* original count : 1 */
typedef struct xlog_op_header {
- xlog_tid_t oh_tid; /* transaction id of operation : 4 b */
- int oh_len; /* bytes in data region : 4 b */
- __uint8_t oh_clientid; /* who sent me this : 1 b */
- __uint8_t oh_flags; /* : 1 b */
- ushort oh_res2; /* 32 bit align : 2 b */
+ __be32 oh_tid; /* transaction id of operation : 4 b */
+ __be32 oh_len; /* bytes in data region : 4 b */
+ __u8 oh_clientid; /* who sent me this : 1 b */
+ __u8 oh_flags; /* : 1 b */
+ __u16 oh_res2; /* 32 bit align : 2 b */
} xlog_op_header_t;
#endif
typedef struct xlog_rec_header {
- uint h_magicno; /* log record (LR) identifier : 4 */
- uint h_cycle; /* write cycle of log : 4 */
- int h_version; /* LR version : 4 */
- int h_len; /* len in bytes; should be 64-bit aligned: 4 */
- xfs_lsn_t h_lsn; /* lsn of this LR : 8 */
- xfs_lsn_t h_tail_lsn; /* lsn of 1st LR w/ buffers not committed: 8 */
- uint h_chksum; /* may not be used; non-zero if used : 4 */
- int h_prev_block; /* block number to previous LR : 4 */
- int h_num_logops; /* number of log operations in this LR : 4 */
- uint h_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE];
+ __be32 h_magicno; /* log record (LR) identifier : 4 */
+ __be32 h_cycle; /* write cycle of log : 4 */
+ __be32 h_version; /* LR version : 4 */
+ __be32 h_len; /* len in bytes; should be 64-bit aligned: 4 */
+ __be64 h_lsn; /* lsn of this LR : 8 */
+ __be64 h_tail_lsn; /* lsn of 1st LR w/ buffers not committed: 8 */
+ __be32 h_chksum; /* may not be used; non-zero if used : 4 */
+ __be32 h_prev_block; /* block number to previous LR : 4 */
+ __be32 h_num_logops; /* number of log operations in this LR : 4 */
+ __be32 h_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE];
/* new fields */
- int h_fmt; /* format of log record : 4 */
- uuid_t h_fs_uuid; /* uuid of FS : 16 */
- int h_size; /* iclog size : 4 */
+ __be32 h_fmt; /* format of log record : 4 */
+ uuid_t h_fs_uuid; /* uuid of FS : 16 */
+ __be32 h_size; /* iclog size : 4 */
} xlog_rec_header_t;
typedef struct xlog_rec_ext_header {
- uint xh_cycle; /* write cycle of log : 4 */
- uint xh_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; /* : 256 */
+ __be32 xh_cycle; /* write cycle of log : 4 */
+ __be32 xh_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; /* : 256 */
} xlog_rec_ext_header_t;
#ifdef __KERNEL__
+
+/*
+ * Quite misnamed, because this union lays out the actual on-disk log buffer.
+ */
+typedef union xlog_in_core2 {
+ xlog_rec_header_t hic_header;
+ xlog_rec_ext_header_t hic_xheader;
+ char hic_sector[XLOG_HEADER_SIZE];
+} xlog_in_core_2_t;
+
/*
* - A log record header is 512 bytes. There is plenty of room to grow the
* xlog_rec_header_t into the reserved space.
* - ic_data follows, so a write to disk can start at the beginning of
* the iclog.
- * - ic_forcesema is used to implement synchronous forcing of the iclog to disk.
+ * - ic_forcewait is used to implement synchronous forcing of the iclog to disk.
* - ic_next is the pointer to the next iclog in the ring.
* - ic_bp is a pointer to the buffer used to write this incore log to disk.
* - ic_log is a pointer back to the global log structure.
* - ic_offset is the current number of bytes written to in this iclog.
* - ic_refcnt is bumped when someone is writing to the log.
* - ic_state is the state of the iclog.
+ *
+ * Because of cacheline contention on large machines, we need to separate
+ * various resources onto different cachelines. To start with, make the
+ * structure cacheline aligned. The following fields can be contended on
+ * by independent processes:
+ *
+ * - ic_callback_*
+ * - ic_refcnt
+ * - fields protected by the global l_icloglock
+ *
+ * so we need to ensure that these fields are located in separate cachelines.
+ * We'll put all the read-only and l_icloglock fields in the first cacheline,
+ * and move everything else out to subsequent cachelines.
*/
-typedef struct xlog_iclog_fields {
- sv_t ic_forcesema;
- sv_t ic_writesema;
+typedef struct xlog_in_core {
+ sv_t ic_force_wait;
+ sv_t ic_write_wait;
struct xlog_in_core *ic_next;
struct xlog_in_core *ic_prev;
struct xfs_buf *ic_bp;
struct log *ic_log;
- xfs_log_callback_t *ic_callback;
- xfs_log_callback_t **ic_callback_tail;
-#ifdef XFS_LOG_TRACE
- struct ktrace *ic_trace;
-#endif
int ic_size;
int ic_offset;
- int ic_refcnt;
int ic_bwritecnt;
- ushort_t ic_state;
+ unsigned short ic_state;
char *ic_datap; /* pointer to iclog data */
-} xlog_iclog_fields_t;
-typedef union xlog_in_core2 {
- xlog_rec_header_t hic_header;
- xlog_rec_ext_header_t hic_xheader;
- char hic_sector[XLOG_HEADER_SIZE];
-} xlog_in_core_2_t;
+ /* Callback structures need their own cacheline */
+ spinlock_t ic_callback_lock ____cacheline_aligned_in_smp;
+ xfs_log_callback_t *ic_callback;
+ xfs_log_callback_t **ic_callback_tail;
-typedef struct xlog_in_core {
- xlog_iclog_fields_t hic_fields;
- xlog_in_core_2_t *hic_data;
+ /* reference counts need their own cacheline */
+ atomic_t ic_refcnt ____cacheline_aligned_in_smp;
+ xlog_in_core_2_t *ic_data;
+#define ic_header ic_data->hic_header
} xlog_in_core_t;
/*
- * Defines to save our code from this glop.
+ * The CIL context is used to aggregate per-transaction details as well be
+ * passed to the iclog for checkpoint post-commit processing. After being
+ * passed to the iclog, another context needs to be allocated for tracking the
+ * next set of transactions to be aggregated into a checkpoint.
+ */
+struct xfs_cil;
+
+struct xfs_cil_ctx {
+ struct xfs_cil *cil;
+ xfs_lsn_t sequence; /* chkpt sequence # */
+ xfs_lsn_t start_lsn; /* first LSN of chkpt commit */
+ xfs_lsn_t commit_lsn; /* chkpt commit record lsn */
+ struct xlog_ticket *ticket; /* chkpt ticket */
+ int nvecs; /* number of regions */
+ int space_used; /* aggregate size of regions */
+ struct list_head busy_extents; /* busy extents in chkpt */
+ struct xfs_log_vec *lv_chain; /* logvecs being pushed */
+ xfs_log_callback_t log_cb; /* completion callback hook. */
+ struct list_head committing; /* ctx committing list */
+};
+
+/*
+ * Committed Item List structure
+ *
+ * This structure is used to track log items that have been committed but not
+ * yet written into the log. It is used only when the delayed logging mount
+ * option is enabled.
+ *
+ * This structure tracks the list of committing checkpoint contexts so
+ * we can avoid the problem of having to hold out new transactions during a
+ * flush until we have a the commit record LSN of the checkpoint. We can
+ * traverse the list of committing contexts in xlog_cil_push_lsn() to find a
+ * sequence match and extract the commit LSN directly from there. If the
+ * checkpoint is still in the process of committing, we can block waiting for
+ * the commit LSN to be determined as well. This should make synchronous
+ * operations almost as efficient as the old logging methods.
+ */
+struct xfs_cil {
+ struct log *xc_log;
+ struct list_head xc_cil;
+ spinlock_t xc_cil_lock;
+ struct xfs_cil_ctx *xc_ctx;
+ struct rw_semaphore xc_ctx_lock;
+ struct list_head xc_committing;
+ sv_t xc_commit_wait;
+};
+
+/*
+ * The amount of log space we should the CIL to aggregate is difficult to size.
+ * Whatever we chose we have to make we can get a reservation for the log space
+ * effectively, that it is large enough to capture sufficient relogging to
+ * reduce log buffer IO significantly, but it is not too large for the log or
+ * induces too much latency when writing out through the iclogs. We track both
+ * space consumed and the number of vectors in the checkpoint context, so we
+ * need to decide which to use for limiting.
+ *
+ * Every log buffer we write out during a push needs a header reserved, which
+ * is at least one sector and more for v2 logs. Hence we need a reservation of
+ * at least 512 bytes per 32k of log space just for the LR headers. That means
+ * 16KB of reservation per megabyte of delayed logging space we will consume,
+ * plus various headers. The number of headers will vary based on the num of
+ * io vectors, so limiting on a specific number of vectors is going to result
+ * in transactions of varying size. IOWs, it is more consistent to track and
+ * limit space consumed in the log rather than by the number of objects being
+ * logged in order to prevent checkpoint ticket overruns.
+ *
+ * Further, use of static reservations through the log grant mechanism is
+ * problematic. It introduces a lot of complexity (e.g. reserve grant vs write
+ * grant) and a significant deadlock potential because regranting write space
+ * can block on log pushes. Hence if we have to regrant log space during a log
+ * push, we can deadlock.
+ *
+ * However, we can avoid this by use of a dynamic "reservation stealing"
+ * technique during transaction commit whereby unused reservation space in the
+ * transaction ticket is transferred to the CIL ctx commit ticket to cover the
+ * space needed by the checkpoint transaction. This means that we never need to
+ * specifically reserve space for the CIL checkpoint transaction, nor do we
+ * need to regrant space once the checkpoint completes. This also means the
+ * checkpoint transaction ticket is specific to the checkpoint context, rather
+ * than the CIL itself.
+ *
+ * With dynamic reservations, we can basically make up arbitrary limits for the
+ * checkpoint size so long as they don't violate any other size rules. Hence
+ * the initial maximum size for the checkpoint transaction will be set to a
+ * quarter of the log or 8MB, which ever is smaller. 8MB is an arbitrary limit
+ * right now based on the latency of writing out a large amount of data through
+ * the circular iclog buffers.
*/
-#define ic_forcesema hic_fields.ic_forcesema
-#define ic_writesema hic_fields.ic_writesema
-#define ic_next hic_fields.ic_next
-#define ic_prev hic_fields.ic_prev
-#define ic_bp hic_fields.ic_bp
-#define ic_log hic_fields.ic_log
-#define ic_callback hic_fields.ic_callback
-#define ic_callback_tail hic_fields.ic_callback_tail
-#define ic_trace hic_fields.ic_trace
-#define ic_size hic_fields.ic_size
-#define ic_offset hic_fields.ic_offset
-#define ic_refcnt hic_fields.ic_refcnt
-#define ic_bwritecnt hic_fields.ic_bwritecnt
-#define ic_state hic_fields.ic_state
-#define ic_datap hic_fields.ic_datap
-#define ic_header hic_data->hic_header
+
+#define XLOG_CIL_SPACE_LIMIT(log) \
+ (min((log->l_logsize >> 2), (8 * 1024 * 1024)))
/*
* The reservation head lsn is not made up of a cycle number and block number.
* that round off problems won't occur when releasing partial reservations.
*/
typedef struct log {
+ /* The following fields don't need locking */
+ struct xfs_mount *l_mp; /* mount point */
+ struct xfs_ail *l_ailp; /* AIL log is working with */
+ struct xfs_cil *l_cilp; /* CIL log is working with */
+ struct xfs_buf *l_xbuf; /* extra buffer for log
+ * wrapping */
+ struct xfs_buftarg *l_targ; /* buftarg of log */
+ uint l_flags;
+ uint l_quotaoffs_flag; /* XFS_DQ_*, for QUOTAOFFs */
+ struct xfs_buf_cancel **l_buf_cancel_table;
+ int l_iclog_hsize; /* size of iclog header */
+ int l_iclog_heads; /* # of iclog header sectors */
+ uint l_sectBBsize; /* sector size in BBs (2^n) */
+ int l_iclog_size; /* size of log in bytes */
+ int l_iclog_size_log; /* log power size of log */
+ int l_iclog_bufs; /* number of iclog buffers */
+ xfs_daddr_t l_logBBstart; /* start block of log */
+ int l_logsize; /* size of log in bytes */
+ int l_logBBsize; /* size of log in BB chunks */
+
/* The following block of fields are changed while holding icloglock */
- sema_t l_flushsema; /* iclog flushing semaphore */
- int l_flushcnt; /* # of procs waiting on this
- * sema */
- int l_ticket_cnt; /* free ticket count */
- int l_ticket_tcnt; /* total ticket count */
+ sv_t l_flush_wait ____cacheline_aligned_in_smp;
+ /* waiting for iclog flush */
int l_covered_state;/* state of "covering disk
* log entries" */
- xlog_ticket_t *l_freelist; /* free list of tickets */
- xlog_ticket_t *l_unmount_free;/* kmem_free these addresses */
- xlog_ticket_t *l_tail; /* free list of tickets */
xlog_in_core_t *l_iclog; /* head log queue */
spinlock_t l_icloglock; /* grab to change iclog state */
xfs_lsn_t l_tail_lsn; /* lsn of 1st LR with unflushed
* buffers */
xfs_lsn_t l_last_sync_lsn;/* lsn of last LR on disk */
- struct xfs_mount *l_mp; /* mount point */
- struct xfs_buf *l_xbuf; /* extra buffer for log
- * wrapping */
- struct xfs_buftarg *l_targ; /* buftarg of log */
- xfs_daddr_t l_logBBstart; /* start block of log */
- int l_logsize; /* size of log in bytes */
- int l_logBBsize; /* size of log in BB chunks */
int l_curr_cycle; /* Cycle number of log writes */
int l_prev_cycle; /* Cycle number before last
* block increment */
int l_curr_block; /* current logical log block */
int l_prev_block; /* previous logical log block */
- int l_iclog_size; /* size of log in bytes */
- int l_iclog_size_log; /* log power size of log */
- int l_iclog_bufs; /* number of iclog buffers */
-
- /* The following field are used for debugging; need to hold icloglock */
- char *l_iclog_bak[XLOG_MAX_ICLOGS];
/* The following block of fields are changed while holding grant_lock */
- lock_t l_grant_lock;
+ spinlock_t l_grant_lock ____cacheline_aligned_in_smp;
xlog_ticket_t *l_reserve_headq;
xlog_ticket_t *l_write_headq;
int l_grant_reserve_cycle;
int l_grant_write_cycle;
int l_grant_write_bytes;
- /* The following fields don't need locking */
-#ifdef XFS_LOG_TRACE
- struct ktrace *l_trace;
- struct ktrace *l_grant_trace;
+ /* The following field are used for debugging; need to hold icloglock */
+#ifdef DEBUG
+ char *l_iclog_bak[XLOG_MAX_ICLOGS];
#endif
- uint l_flags;
- uint l_quotaoffs_flag; /* XFS_DQ_*, for QUOTAOFFs */
- struct xfs_buf_cancel **l_buf_cancel_table;
- int l_iclog_hsize; /* size of iclog header */
- int l_iclog_heads; /* # of iclog header sectors */
- uint l_sectbb_log; /* log2 of sector size in BBs */
- uint l_sectbb_mask; /* sector size (in BBs)
- * alignment mask */
+
} xlog_t;
#define XLOG_FORCED_SHUTDOWN(log) ((log)->l_flags & XLOG_IO_ERROR)
-
/* common routines */
extern xfs_lsn_t xlog_assign_tail_lsn(struct xfs_mount *mp);
-extern int xlog_find_tail(xlog_t *log,
- xfs_daddr_t *head_blk,
- xfs_daddr_t *tail_blk);
extern int xlog_recover(xlog_t *log);
-extern int xlog_recover_finish(xlog_t *log, int mfsi_flags);
+extern int xlog_recover_finish(xlog_t *log);
extern void xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog, int);
-extern void xlog_recover_process_iunlinks(xlog_t *log);
-extern struct xfs_buf *xlog_get_bp(xlog_t *, int);
-extern void xlog_put_bp(struct xfs_buf *);
-extern int xlog_bread(xlog_t *, xfs_daddr_t, int, struct xfs_buf *);
+extern kmem_zone_t *xfs_log_ticket_zone;
+struct xlog_ticket *xlog_ticket_alloc(struct log *log, int unit_bytes,
+ int count, char client, uint xflags,
+ int alloc_flags);
+
+
+static inline void
+xlog_write_adv_cnt(void **ptr, int *len, int *off, size_t bytes)
+{
+ *ptr += bytes;
+ *len -= bytes;
+ *off += bytes;
+}
+
+void xlog_print_tic_res(struct xfs_mount *mp, struct xlog_ticket *ticket);
+int xlog_write(struct log *log, struct xfs_log_vec *log_vector,
+ struct xlog_ticket *tic, xfs_lsn_t *start_lsn,
+ xlog_in_core_t **commit_iclog, uint flags);
+
+/*
+ * Committed Item List interfaces
+ */
+int xlog_cil_init(struct log *log);
+void xlog_cil_init_post_recovery(struct log *log);
+void xlog_cil_destroy(struct log *log);
-/* iclog tracing */
-#define XLOG_TRACE_GRAB_FLUSH 1
-#define XLOG_TRACE_REL_FLUSH 2
-#define XLOG_TRACE_SLEEP_FLUSH 3
-#define XLOG_TRACE_WAKE_FLUSH 4
+int xlog_cil_push(struct log *log, int push_now);
+xfs_lsn_t xlog_cil_push_lsn(struct log *log, xfs_lsn_t push_sequence);
/*
* Unmount record type is used as a pseudo transaction type for the ticket.