* Copyright (c) 2012, Joyent, Inc. All rights reserved.
* Copyright (c) 2011, 2016 by Delphix. All rights reserved.
* Copyright (c) 2014 by Saso Kiselkov. All rights reserved.
- * Copyright 2014 Nexenta Systems, Inc. All rights reserved.
+ * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
*/
/*
* - ARC header release, as it removes from L2ARC buflists
*/
+/*
+ * ARC operation:
+ *
+ * Every block that is in the ARC is tracked by an arc_buf_hdr_t structure.
+ * This structure can point either to a block that is still in the cache or to
+ * one that is only accessible in an L2 ARC device, or it can provide
+ * information about a block that was recently evicted. If a block is
+ * only accessible in the L2ARC, then the arc_buf_hdr_t only has enough
+ * information to retrieve it from the L2ARC device. This information is
+ * stored in the l2arc_buf_hdr_t sub-structure of the arc_buf_hdr_t. A block
+ * that is in this state cannot access the data directly.
+ *
+ * Blocks that are actively being referenced or have not been evicted
+ * are cached in the L1ARC. The L1ARC (l1arc_buf_hdr_t) is a structure within
+ * the arc_buf_hdr_t that will point to the data block in memory. A block can
+ * only be read by a consumer if it has an l1arc_buf_hdr_t. The L1ARC
+ * caches data in two ways -- in a list of arc buffers (arc_buf_t) and
+ * also in the arc_buf_hdr_t's private physical data block pointer (b_pdata).
+ * Each arc buffer (arc_buf_t) is being actively accessed by a specific ARC
+ * consumer, and always contains uncompressed data. The ARC will provide
+ * references to this data and will keep it cached until it is no longer in
+ * use. Typically, the arc will try to cache only the L1ARC's physical data
+ * block and will aggressively evict any arc_buf_t that is no longer referenced.
+ * The amount of memory consumed by the arc_buf_t's can be seen via the
+ * "overhead_size" kstat.
+ *
+ *
+ * arc_buf_hdr_t
+ * +-----------+
+ * | |
+ * | |
+ * | |
+ * +-----------+
+ * l2arc_buf_hdr_t| |
+ * | |
+ * +-----------+
+ * l1arc_buf_hdr_t| |
+ * | | arc_buf_t
+ * | b_buf +------------>+---------+ arc_buf_t
+ * | | |b_next +---->+---------+
+ * | b_pdata +-+ |---------| |b_next +-->NULL
+ * +-----------+ | | | +---------+
+ * | |b_data +-+ | |
+ * | +---------+ | |b_data +-+
+ * +->+------+ | +---------+ |
+ * (potentially) | | | |
+ * compressed | | | |
+ * data +------+ | v
+ * +->+------+ +------+
+ * uncompressed | | | |
+ * data | | | |
+ * +------+ +------+
+ *
+ * The L1ARC's data pointer, however, may or may not be uncompressed. The
+ * ARC has the ability to store the physical data (b_pdata) associated with
+ * the DVA of the arc_buf_hdr_t. Since the b_pdata is a copy of the on-disk
+ * physical block, it will match its on-disk compression characteristics.
+ * If the block on-disk is compressed, then the physical data block
+ * in the cache will also be compressed and vice-versa. This behavior
+ * can be disabled by setting 'zfs_compressed_arc_enabled' to B_FALSE. When the
+ * compressed ARC functionality is disabled, the b_pdata will point to an
+ * uncompressed version of the on-disk data.
+ *
+ * When a consumer reads a block, the ARC must first look to see if the
+ * arc_buf_hdr_t is cached. If the hdr is cached and already has an arc_buf_t,
+ * then an additional arc_buf_t is allocated and the uncompressed data is
+ * bcopied from the existing arc_buf_t. If the hdr is cached but does not
+ * have an arc_buf_t, then the ARC allocates a new arc_buf_t and decompresses
+ * the b_pdata contents into the arc_buf_t's b_data. If the arc_buf_hdr_t's
+ * b_pdata is not compressed, then the block is shared with the newly
+ * allocated arc_buf_t. This block sharing only occurs with one arc_buf_t
+ * in the arc buffer chain. Sharing the block reduces the memory overhead
+ * required when the hdr is caching uncompressed blocks or the compressed
+ * arc functionality has been disabled via 'zfs_compressed_arc_enabled'.
+ *
+ * The diagram below shows an example of an uncompressed ARC hdr that is
+ * sharing its data with an arc_buf_t:
+ *
+ * arc_buf_hdr_t
+ * +-----------+
+ * | |
+ * | |
+ * | |
+ * +-----------+
+ * l2arc_buf_hdr_t| |
+ * | |
+ * +-----------+
+ * l1arc_buf_hdr_t| |
+ * | | arc_buf_t (shared)
+ * | b_buf +------------>+---------+ arc_buf_t
+ * | | |b_next +---->+---------+
+ * | b_pdata +-+ |---------| |b_next +-->NULL
+ * +-----------+ | | | +---------+
+ * | |b_data +-+ | |
+ * | +---------+ | |b_data +-+
+ * +->+------+ | +---------+ |
+ * | | | |
+ * uncompressed | | | |
+ * data +------+ | |
+ * ^ +->+------+ |
+ * | uncompressed | | |
+ * | data | | |
+ * | +------+ |
+ * +---------------------------------+
+ *
+ * Writing to the arc requires that the ARC first discard the b_pdata
+ * since the physical block is about to be rewritten. The new data contents
+ * will be contained in the arc_buf_t (uncompressed). As the I/O pipeline
+ * performs the write, it may compress the data before writing it to disk.
+ * The ARC will be called with the transformed data and will bcopy the
+ * transformed on-disk block into a newly allocated b_pdata.
+ *
+ * When the L2ARC is in use, it will also take advantage of the b_pdata. The
+ * L2ARC will always write the contents of b_pdata to the L2ARC. This means
+ * that when compressed arc is enabled that the L2ARC blocks are identical
+ * to the on-disk block in the main data pool. This provides a significant
+ * advantage since the ARC can leverage the bp's checksum when reading from the
+ * L2ARC to determine if the contents are valid. However, if the compressed
+ * arc is disabled, then the L2ARC's block must be transformed to look
+ * like the physical block in the main data pool before comparing the
+ * checksum and determining its validity.
+ */
+
#include <sys/spa.h>
#include <sys/zio.h>
+#include <sys/spa_impl.h>
#include <sys/zio_compress.h>
+#include <sys/zio_checksum.h>
#include <sys/zfs_context.h>
#include <sys/arc.h>
#include <sys/refcount.h>
static boolean_t arc_reclaim_thread_exit;
static kcondvar_t arc_reclaim_waiters_cv;
-static kmutex_t arc_user_evicts_lock;
-static kcondvar_t arc_user_evicts_cv;
-static boolean_t arc_user_evicts_thread_exit;
-
/*
* The number of headers to evict in arc_evict_state_impl() before
* dropping the sublist lock and evicting from another sublist. A lower
*/
static boolean_t arc_warm;
+/*
+ * log2 fraction of the zio arena to keep free.
+ */
+int arc_zio_arena_free_shift = 2;
+
/*
* These tunables are for performance analysis.
*/
int zfs_arc_grow_retry = 0;
int zfs_arc_shrink_shift = 0;
int zfs_arc_p_min_shift = 0;
-int zfs_disable_dup_eviction = 0;
int zfs_arc_average_blocksize = 8 * 1024; /* 8KB */
+int zfs_compressed_arc_enabled = B_TRUE;
+
/*
* ARC will evict meta buffers that exceed arc_meta_limit. This
* tunable make arc_meta_limit adjustable for different workloads.
kstat_named_t arcstat_c_min;
kstat_named_t arcstat_c_max;
kstat_named_t arcstat_size;
+ /*
+ * Number of compressed bytes stored in the arc_buf_hdr_t's b_pdata.
+ * Note that the compressed bytes may match the uncompressed bytes
+ * if the block is either not compressed or compressed arc is disabled.
+ */
+ kstat_named_t arcstat_compressed_size;
+ /*
+ * Uncompressed size of the data stored in b_pdata. If compressed
+ * arc is disabled then this value will be identical to the stat
+ * above.
+ */
+ kstat_named_t arcstat_uncompressed_size;
+ /*
+ * Number of bytes stored in all the arc_buf_t's. This is classified
+ * as "overhead" since this data is typically short-lived and will
+ * be evicted from the arc when it becomes unreferenced unless the
+ * zfs_keep_uncompressed_metadata or zfs_keep_uncompressed_level
+ * values have been set (see comment in dbuf.c for more information).
+ */
+ kstat_named_t arcstat_overhead_size;
/*
* Number of bytes consumed by internal ARC structures necessary
* for tracking purposes; these structures are not actually
kstat_named_t arcstat_l2_writes_done;
kstat_named_t arcstat_l2_writes_error;
kstat_named_t arcstat_l2_writes_lock_retry;
- kstat_named_t arcstat_l2_writes_skip_toobig;
kstat_named_t arcstat_l2_evict_lock_retry;
kstat_named_t arcstat_l2_evict_reading;
kstat_named_t arcstat_l2_evict_l1cached;
kstat_named_t arcstat_l2_free_on_write;
- kstat_named_t arcstat_l2_cdata_free_on_write;
kstat_named_t arcstat_l2_abort_lowmem;
kstat_named_t arcstat_l2_cksum_bad;
kstat_named_t arcstat_l2_io_error;
kstat_named_t arcstat_l2_size;
kstat_named_t arcstat_l2_asize;
kstat_named_t arcstat_l2_hdr_size;
- kstat_named_t arcstat_l2_compress_successes;
- kstat_named_t arcstat_l2_compress_zeros;
- kstat_named_t arcstat_l2_compress_failures;
kstat_named_t arcstat_memory_throttle_count;
- kstat_named_t arcstat_duplicate_buffers;
- kstat_named_t arcstat_duplicate_buffers_size;
- kstat_named_t arcstat_duplicate_reads;
kstat_named_t arcstat_memory_direct_count;
kstat_named_t arcstat_memory_indirect_count;
kstat_named_t arcstat_no_grow;
{ "c_min", KSTAT_DATA_UINT64 },
{ "c_max", KSTAT_DATA_UINT64 },
{ "size", KSTAT_DATA_UINT64 },
+ { "compressed_size", KSTAT_DATA_UINT64 },
+ { "uncompressed_size", KSTAT_DATA_UINT64 },
+ { "overhead_size", KSTAT_DATA_UINT64 },
{ "hdr_size", KSTAT_DATA_UINT64 },
{ "data_size", KSTAT_DATA_UINT64 },
{ "metadata_size", KSTAT_DATA_UINT64 },
{ "l2_writes_done", KSTAT_DATA_UINT64 },
{ "l2_writes_error", KSTAT_DATA_UINT64 },
{ "l2_writes_lock_retry", KSTAT_DATA_UINT64 },
- { "l2_writes_skip_toobig", KSTAT_DATA_UINT64 },
{ "l2_evict_lock_retry", KSTAT_DATA_UINT64 },
{ "l2_evict_reading", KSTAT_DATA_UINT64 },
{ "l2_evict_l1cached", KSTAT_DATA_UINT64 },
{ "l2_free_on_write", KSTAT_DATA_UINT64 },
- { "l2_cdata_free_on_write", KSTAT_DATA_UINT64 },
{ "l2_abort_lowmem", KSTAT_DATA_UINT64 },
{ "l2_cksum_bad", KSTAT_DATA_UINT64 },
{ "l2_io_error", KSTAT_DATA_UINT64 },
{ "l2_size", KSTAT_DATA_UINT64 },
{ "l2_asize", KSTAT_DATA_UINT64 },
{ "l2_hdr_size", KSTAT_DATA_UINT64 },
- { "l2_compress_successes", KSTAT_DATA_UINT64 },
- { "l2_compress_zeros", KSTAT_DATA_UINT64 },
- { "l2_compress_failures", KSTAT_DATA_UINT64 },
{ "memory_throttle_count", KSTAT_DATA_UINT64 },
- { "duplicate_buffers", KSTAT_DATA_UINT64 },
- { "duplicate_buffers_size", KSTAT_DATA_UINT64 },
- { "duplicate_reads", KSTAT_DATA_UINT64 },
{ "memory_direct_count", KSTAT_DATA_UINT64 },
{ "memory_indirect_count", KSTAT_DATA_UINT64 },
{ "arc_no_grow", KSTAT_DATA_UINT64 },
#define arc_c ARCSTAT(arcstat_c) /* target size of cache */
#define arc_c_min ARCSTAT(arcstat_c_min) /* min target cache size */
#define arc_c_max ARCSTAT(arcstat_c_max) /* max target cache size */
-#define arc_no_grow ARCSTAT(arcstat_no_grow)
+#define arc_no_grow ARCSTAT(arcstat_no_grow) /* do not grow cache size */
#define arc_tempreserve ARCSTAT(arcstat_tempreserve)
#define arc_loaned_bytes ARCSTAT(arcstat_loaned_bytes)
#define arc_meta_limit ARCSTAT(arcstat_meta_limit) /* max size for metadata */
#define arc_need_free ARCSTAT(arcstat_need_free) /* bytes to be freed */
#define arc_sys_free ARCSTAT(arcstat_sys_free) /* target system free bytes */
-#define L2ARC_IS_VALID_COMPRESS(_c_) \
- ((_c_) == ZIO_COMPRESS_LZ4 || (_c_) == ZIO_COMPRESS_EMPTY)
+/* compressed size of entire arc */
+#define arc_compressed_size ARCSTAT(arcstat_compressed_size)
+/* uncompressed size of entire arc */
+#define arc_uncompressed_size ARCSTAT(arcstat_uncompressed_size)
+/* number of bytes in the arc from arc_buf_t's */
+#define arc_overhead_size ARCSTAT(arcstat_overhead_size)
static list_t arc_prune_list;
static kmutex_t arc_prune_mtx;
static taskq_t *arc_prune_taskq;
-static arc_buf_t *arc_eviction_list;
-static arc_buf_hdr_t arc_eviction_hdr;
#define GHOST_STATE(state) \
((state) == arc_mru_ghost || (state) == arc_mfu_ghost || \
#define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS)
#define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR)
#define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH)
-#define HDR_FREED_IN_READ(hdr) ((hdr)->b_flags & ARC_FLAG_FREED_IN_READ)
-#define HDR_BUF_AVAILABLE(hdr) ((hdr)->b_flags & ARC_FLAG_BUF_AVAILABLE)
+#define HDR_COMPRESSION_ENABLED(hdr) \
+ ((hdr)->b_flags & ARC_FLAG_COMPRESSED_ARC)
#define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE)
-#define HDR_L2COMPRESS(hdr) ((hdr)->b_flags & ARC_FLAG_L2COMPRESS)
#define HDR_L2_READING(hdr) \
- (((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \
- ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR))
+ (((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \
+ ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR))
#define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING)
#define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED)
#define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD)
+#define HDR_SHARED_DATA(hdr) ((hdr)->b_flags & ARC_FLAG_SHARED_DATA)
#define HDR_ISTYPE_METADATA(hdr) \
- ((hdr)->b_flags & ARC_FLAG_BUFC_METADATA)
+ ((hdr)->b_flags & ARC_FLAG_BUFC_METADATA)
#define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr))
#define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR)
#define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)
+/* For storing compression mode in b_flags */
+#define HDR_COMPRESS_OFFSET (highbit64(ARC_FLAG_COMPRESS_0) - 1)
+
+#define HDR_GET_COMPRESS(hdr) ((enum zio_compress)BF32_GET((hdr)->b_flags, \
+ HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS))
+#define HDR_SET_COMPRESS(hdr, cmp) BF32_SET((hdr)->b_flags, \
+ HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS, (cmp));
+
+#define ARC_BUF_LAST(buf) ((buf)->b_next == NULL)
+
/*
* Other sizes
*/
#define L2ARC_WRITE_SIZE (8 * 1024 * 1024) /* initial write max */
#define L2ARC_HEADROOM 2 /* num of writes */
-#define L2ARC_MAX_BLOCK_SIZE (16 * 1024 * 1024) /* max compress size */
/*
* If we discover during ARC scan any buffers to be compressed, we boost
#define L2ARC_FEED_SECS 1 /* caching interval secs */
#define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */
-
-/*
- * Used to distinguish headers that are being process by
- * l2arc_write_buffers(), but have yet to be assigned to a l2arc disk
- * address. This can happen when the header is added to the l2arc's list
- * of buffers to write in the first stage of l2arc_write_buffers(), but
- * has not yet been written out which happens in the second stage of
- * l2arc_write_buffers().
- */
-#define L2ARC_ADDR_UNSET ((uint64_t)(-1))
-
#define l2arc_writes_sent ARCSTAT(arcstat_l2_writes_sent)
#define l2arc_writes_done ARCSTAT(arcstat_l2_writes_done)
unsigned long l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra warmup write */
unsigned long l2arc_headroom = L2ARC_HEADROOM; /* # of dev writes */
unsigned long l2arc_headroom_boost = L2ARC_HEADROOM_BOOST;
-unsigned long l2arc_max_block_size = L2ARC_MAX_BLOCK_SIZE;
unsigned long l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */
unsigned long l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval msecs */
int l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */
-int l2arc_nocompress = B_FALSE; /* don't compress bufs */
int l2arc_feed_again = B_TRUE; /* turbo warmup */
int l2arc_norw = B_FALSE; /* no reads during writes */
static uint64_t l2arc_ndev; /* number of devices */
typedef struct l2arc_read_callback {
- arc_buf_t *l2rcb_buf; /* read buffer */
- spa_t *l2rcb_spa; /* spa */
+ arc_buf_hdr_t *l2rcb_hdr; /* read buffer */
blkptr_t l2rcb_bp; /* original blkptr */
zbookmark_phys_t l2rcb_zb; /* original bookmark */
int l2rcb_flags; /* original flags */
- enum zio_compress l2rcb_compress; /* applied compress */
} l2arc_read_callback_t;
typedef struct l2arc_data_free {
/* protected by l2arc_free_on_write_mtx */
void *l2df_data;
size_t l2df_size;
- void (*l2df_func)(void *, size_t);
+ arc_buf_contents_t l2df_type;
list_node_t l2df_list_node;
} l2arc_data_free_t;
static kcondvar_t l2arc_feed_thr_cv;
static uint8_t l2arc_thread_exit;
-static void arc_get_data_buf(arc_buf_t *);
+static void *arc_get_data_buf(arc_buf_hdr_t *, uint64_t, void *);
+static void arc_free_data_buf(arc_buf_hdr_t *, void *, uint64_t, void *);
+static void arc_hdr_free_pdata(arc_buf_hdr_t *hdr);
+static void arc_hdr_alloc_pdata(arc_buf_hdr_t *);
static void arc_access(arc_buf_hdr_t *, kmutex_t *);
static boolean_t arc_is_overflowing(void);
static void arc_buf_watch(arc_buf_t *);
static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *);
static uint32_t arc_bufc_to_flags(arc_buf_contents_t);
+static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);
+static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);
static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *);
static void l2arc_read_done(zio_t *);
-static boolean_t l2arc_compress_buf(arc_buf_hdr_t *);
-static void l2arc_decompress_zio(zio_t *, arc_buf_hdr_t *, enum zio_compress);
-static void l2arc_release_cdata_buf(arc_buf_hdr_t *);
-
static uint64_t
buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth)
{
return (crc);
}
-#define BUF_EMPTY(buf) \
- ((buf)->b_dva.dva_word[0] == 0 && \
- (buf)->b_dva.dva_word[1] == 0)
+#define HDR_EMPTY(hdr) \
+ ((hdr)->b_dva.dva_word[0] == 0 && \
+ (hdr)->b_dva.dva_word[1] == 0)
-#define BUF_EQUAL(spa, dva, birth, buf) \
- ((buf)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \
- ((buf)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \
- ((buf)->b_birth == birth) && ((buf)->b_spa == spa)
+#define HDR_EQUAL(spa, dva, birth, hdr) \
+ ((hdr)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \
+ ((hdr)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \
+ ((hdr)->b_birth == birth) && ((hdr)->b_spa == spa)
static void
buf_discard_identity(arc_buf_hdr_t *hdr)
mutex_enter(hash_lock);
for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL;
hdr = hdr->b_hash_next) {
- if (BUF_EQUAL(spa, dva, birth, hdr)) {
+ if (HDR_EQUAL(spa, dva, birth, hdr)) {
*lockp = hash_lock;
return (hdr);
}
for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL;
fhdr = fhdr->b_hash_next, i++) {
- if (BUF_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr))
+ if (HDR_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr))
return (fhdr);
}
hdr->b_hash_next = buf_hash_table.ht_table[idx];
buf_hash_table.ht_table[idx] = hdr;
- hdr->b_flags |= ARC_FLAG_IN_HASH_TABLE;
+ arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
/* collect some hash table performance data */
if (i > 0) {
hdrp = &buf_hash_table.ht_table[idx];
while ((fhdr = *hdrp) != hdr) {
- ASSERT(fhdr != NULL);
+ ASSERT3P(fhdr, !=, NULL);
hdrp = &fhdr->b_hash_next;
}
*hdrp = hdr->b_hash_next;
hdr->b_hash_next = NULL;
- hdr->b_flags &= ~ARC_FLAG_IN_HASH_TABLE;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
/* collect some hash table performance data */
ARCSTAT_BUMPDOWN(arcstat_hash_elements);
{
arc_buf_hdr_t *hdr = vbuf;
- ASSERT(BUF_EMPTY(hdr));
+ ASSERT(HDR_EMPTY(hdr));
cv_destroy(&hdr->b_l1hdr.b_cv);
refcount_destroy(&hdr->b_l1hdr.b_refcnt);
mutex_destroy(&hdr->b_l1hdr.b_freeze_lock);
{
ASSERTV(arc_buf_hdr_t *hdr = vbuf);
- ASSERT(BUF_EMPTY(hdr));
+ ASSERT(HDR_EMPTY(hdr));
arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);
}
}
}
-/*
- * Transition between the two allocation states for the arc_buf_hdr struct.
- * The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without
- * (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller
- * version is used when a cache buffer is only in the L2ARC in order to reduce
- * memory usage.
- */
-static arc_buf_hdr_t *
-arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new)
-{
- arc_buf_hdr_t *nhdr;
- l2arc_dev_t *dev;
-
- ASSERT(HDR_HAS_L2HDR(hdr));
- ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) ||
- (old == hdr_l2only_cache && new == hdr_full_cache));
-
- dev = hdr->b_l2hdr.b_dev;
- nhdr = kmem_cache_alloc(new, KM_PUSHPAGE);
-
- ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
- buf_hash_remove(hdr);
-
- bcopy(hdr, nhdr, HDR_L2ONLY_SIZE);
-
- if (new == hdr_full_cache) {
- nhdr->b_flags |= ARC_FLAG_HAS_L1HDR;
- /*
- * arc_access and arc_change_state need to be aware that a
- * header has just come out of L2ARC, so we set its state to
- * l2c_only even though it's about to change.
- */
- nhdr->b_l1hdr.b_state = arc_l2c_only;
-
- /* Verify previous threads set to NULL before freeing */
- ASSERT3P(nhdr->b_l1hdr.b_tmp_cdata, ==, NULL);
- } else {
- ASSERT(hdr->b_l1hdr.b_buf == NULL);
- ASSERT0(hdr->b_l1hdr.b_datacnt);
-
- /*
- * If we've reached here, We must have been called from
- * arc_evict_hdr(), as such we should have already been
- * removed from any ghost list we were previously on
- * (which protects us from racing with arc_evict_state),
- * thus no locking is needed during this check.
- */
- ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
+#define ARC_MINTIME (hz>>4) /* 62 ms */
- /*
- * A buffer must not be moved into the arc_l2c_only
- * state if it's not finished being written out to the
- * l2arc device. Otherwise, the b_l1hdr.b_tmp_cdata field
- * might try to be accessed, even though it was removed.
- */
- VERIFY(!HDR_L2_WRITING(hdr));
- VERIFY3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL);
+static inline boolean_t
+arc_buf_is_shared(arc_buf_t *buf)
+{
+ boolean_t shared = (buf->b_data != NULL &&
+ buf->b_data == buf->b_hdr->b_l1hdr.b_pdata);
+ IMPLY(shared, HDR_SHARED_DATA(buf->b_hdr));
+ return (shared);
+}
- nhdr->b_flags &= ~ARC_FLAG_HAS_L1HDR;
+static inline void
+arc_cksum_free(arc_buf_hdr_t *hdr)
+{
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
+ if (hdr->b_l1hdr.b_freeze_cksum != NULL) {
+ kmem_free(hdr->b_l1hdr.b_freeze_cksum, sizeof (zio_cksum_t));
+ hdr->b_l1hdr.b_freeze_cksum = NULL;
}
- /*
- * The header has been reallocated so we need to re-insert it into any
- * lists it was on.
- */
- (void) buf_hash_insert(nhdr, NULL);
-
- ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node));
-
- mutex_enter(&dev->l2ad_mtx);
-
- /*
- * We must place the realloc'ed header back into the list at
- * the same spot. Otherwise, if it's placed earlier in the list,
- * l2arc_write_buffers() could find it during the function's
- * write phase, and try to write it out to the l2arc.
- */
- list_insert_after(&dev->l2ad_buflist, hdr, nhdr);
- list_remove(&dev->l2ad_buflist, hdr);
-
- mutex_exit(&dev->l2ad_mtx);
-
- /*
- * Since we're using the pointer address as the tag when
- * incrementing and decrementing the l2ad_alloc refcount, we
- * must remove the old pointer (that we're about to destroy) and
- * add the new pointer to the refcount. Otherwise we'd remove
- * the wrong pointer address when calling arc_hdr_destroy() later.
- */
-
- (void) refcount_remove_many(&dev->l2ad_alloc,
- hdr->b_l2hdr.b_asize, hdr);
-
- (void) refcount_add_many(&dev->l2ad_alloc,
- nhdr->b_l2hdr.b_asize, nhdr);
-
- buf_discard_identity(hdr);
- hdr->b_freeze_cksum = NULL;
- kmem_cache_free(old, hdr);
-
- return (nhdr);
+ mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
}
-
-#define ARC_MINTIME (hz>>4) /* 62 ms */
-
static void
arc_cksum_verify(arc_buf_t *buf)
{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
zio_cksum_t zc;
if (!(zfs_flags & ZFS_DEBUG_MODIFY))
return;
- mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock);
- if (buf->b_hdr->b_freeze_cksum == NULL || HDR_IO_ERROR(buf->b_hdr)) {
- mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock);
+ ASSERT(HDR_HAS_L1HDR(hdr));
+
+ mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
+ if (hdr->b_l1hdr.b_freeze_cksum == NULL || HDR_IO_ERROR(hdr)) {
+ mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
return;
}
- fletcher_2_native(buf->b_data, buf->b_hdr->b_size, &zc);
- if (!ZIO_CHECKSUM_EQUAL(*buf->b_hdr->b_freeze_cksum, zc))
+ fletcher_2_native(buf->b_data, HDR_GET_LSIZE(hdr), &zc);
+ if (!ZIO_CHECKSUM_EQUAL(*hdr->b_l1hdr.b_freeze_cksum, zc))
panic("buffer modified while frozen!");
- mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock);
+ mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
}
-static int
-arc_cksum_equal(arc_buf_t *buf)
+static boolean_t
+arc_cksum_is_equal(arc_buf_hdr_t *hdr, zio_t *zio)
{
- zio_cksum_t zc;
- int equal;
+ enum zio_compress compress = BP_GET_COMPRESS(zio->io_bp);
+ boolean_t valid_cksum;
- mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock);
- fletcher_2_native(buf->b_data, buf->b_hdr->b_size, &zc);
- equal = ZIO_CHECKSUM_EQUAL(*buf->b_hdr->b_freeze_cksum, zc);
- mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock);
+ ASSERT(!BP_IS_EMBEDDED(zio->io_bp));
+ VERIFY3U(BP_GET_PSIZE(zio->io_bp), ==, HDR_GET_PSIZE(hdr));
- return (equal);
+ /*
+ * We rely on the blkptr's checksum to determine if the block
+ * is valid or not. When compressed arc is enabled, the l2arc
+ * writes the block to the l2arc just as it appears in the pool.
+ * This allows us to use the blkptr's checksum to validate the
+ * data that we just read off of the l2arc without having to store
+ * a separate checksum in the arc_buf_hdr_t. However, if compressed
+ * arc is disabled, then the data written to the l2arc is always
+ * uncompressed and won't match the block as it exists in the main
+ * pool. When this is the case, we must first compress it if it is
+ * compressed on the main pool before we can validate the checksum.
+ */
+ if (!HDR_COMPRESSION_ENABLED(hdr) && compress != ZIO_COMPRESS_OFF) {
+ uint64_t lsize;
+ uint64_t csize;
+ void *cbuf;
+ ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
+
+ cbuf = zio_buf_alloc(HDR_GET_PSIZE(hdr));
+ lsize = HDR_GET_LSIZE(hdr);
+ csize = zio_compress_data(compress, zio->io_data, cbuf, lsize);
+ ASSERT3U(csize, <=, HDR_GET_PSIZE(hdr));
+ if (csize < HDR_GET_PSIZE(hdr)) {
+ /*
+ * Compressed blocks are always a multiple of the
+ * smallest ashift in the pool. Ideally, we would
+ * like to round up the csize to the next
+ * spa_min_ashift but that value may have changed
+ * since the block was last written. Instead,
+ * we rely on the fact that the hdr's psize
+ * was set to the psize of the block when it was
+ * last written. We set the csize to that value
+ * and zero out any part that should not contain
+ * data.
+ */
+ bzero((char *)cbuf + csize, HDR_GET_PSIZE(hdr) - csize);
+ csize = HDR_GET_PSIZE(hdr);
+ }
+ zio_push_transform(zio, cbuf, csize, HDR_GET_PSIZE(hdr), NULL);
+ }
+
+ /*
+ * Block pointers always store the checksum for the logical data.
+ * If the block pointer has the gang bit set, then the checksum
+ * it represents is for the reconstituted data and not for an
+ * individual gang member. The zio pipeline, however, must be able to
+ * determine the checksum of each of the gang constituents so it
+ * treats the checksum comparison differently than what we need
+ * for l2arc blocks. This prevents us from using the
+ * zio_checksum_error() interface directly. Instead we must call the
+ * zio_checksum_error_impl() so that we can ensure the checksum is
+ * generated using the correct checksum algorithm and accounts for the
+ * logical I/O size and not just a gang fragment.
+ */
+ valid_cksum = (zio_checksum_error_impl(zio->io_spa, zio->io_bp,
+ BP_GET_CHECKSUM(zio->io_bp), zio->io_data, zio->io_size,
+ zio->io_offset, NULL) == 0);
+ zio_pop_transforms(zio);
+ return (valid_cksum);
}
static void
-arc_cksum_compute(arc_buf_t *buf, boolean_t force)
+arc_cksum_compute(arc_buf_t *buf)
{
- if (!force && !(zfs_flags & ZFS_DEBUG_MODIFY))
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+
+ if (!(zfs_flags & ZFS_DEBUG_MODIFY))
return;
+ ASSERT(HDR_HAS_L1HDR(hdr));
mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock);
- if (buf->b_hdr->b_freeze_cksum != NULL) {
- mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock);
+ if (hdr->b_l1hdr.b_freeze_cksum != NULL) {
+ mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
return;
}
- buf->b_hdr->b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t), KM_SLEEP);
- fletcher_2_native(buf->b_data, buf->b_hdr->b_size,
- buf->b_hdr->b_freeze_cksum);
- mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock);
+ hdr->b_l1hdr.b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t),
+ KM_SLEEP);
+ fletcher_2_native(buf->b_data, HDR_GET_LSIZE(hdr),
+ hdr->b_l1hdr.b_freeze_cksum);
+ mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
arc_buf_watch(buf);
}
{
#ifndef _KERNEL
if (arc_watch) {
- ASSERT0(mprotect(buf->b_data, buf->b_hdr->b_size,
+ ASSERT0(mprotect(buf->b_data, HDR_GET_LSIZE(buf->b_hdr),
PROT_READ | PROT_WRITE));
}
#endif
{
#ifndef _KERNEL
if (arc_watch)
- ASSERT0(mprotect(buf->b_data, buf->b_hdr->b_size, PROT_READ));
+ ASSERT0(mprotect(buf->b_data, HDR_GET_LSIZE(buf->b_hdr),
+ PROT_READ));
#endif
}
static arc_buf_contents_t
arc_buf_type(arc_buf_hdr_t *hdr)
{
+ arc_buf_contents_t type;
if (HDR_ISTYPE_METADATA(hdr)) {
- return (ARC_BUFC_METADATA);
+ type = ARC_BUFC_METADATA;
} else {
- return (ARC_BUFC_DATA);
+ type = ARC_BUFC_DATA;
}
+ VERIFY3U(hdr->b_type, ==, type);
+ return (type);
}
static uint32_t
void
arc_buf_thaw(arc_buf_t *buf)
{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+
if (zfs_flags & ZFS_DEBUG_MODIFY) {
- if (buf->b_hdr->b_l1hdr.b_state != arc_anon)
+ if (hdr->b_l1hdr.b_state != arc_anon)
panic("modifying non-anon buffer!");
- if (HDR_IO_IN_PROGRESS(buf->b_hdr))
+ if (HDR_IO_IN_PROGRESS(hdr))
panic("modifying buffer while i/o in progress!");
arc_cksum_verify(buf);
}
- mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock);
- if (buf->b_hdr->b_freeze_cksum != NULL) {
- kmem_free(buf->b_hdr->b_freeze_cksum, sizeof (zio_cksum_t));
- buf->b_hdr->b_freeze_cksum = NULL;
- }
-
- mutex_exit(&buf->b_hdr->b_l1hdr.b_freeze_lock);
-
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ arc_cksum_free(hdr);
arc_buf_unwatch(buf);
}
void
arc_buf_freeze(arc_buf_t *buf)
{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
kmutex_t *hash_lock;
if (!(zfs_flags & ZFS_DEBUG_MODIFY))
return;
- hash_lock = HDR_LOCK(buf->b_hdr);
+ hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);
- ASSERT(buf->b_hdr->b_freeze_cksum != NULL ||
- buf->b_hdr->b_l1hdr.b_state == arc_anon);
- arc_cksum_compute(buf, B_FALSE);
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ ASSERT(hdr->b_l1hdr.b_freeze_cksum != NULL ||
+ hdr->b_l1hdr.b_state == arc_anon);
+ arc_cksum_compute(buf);
mutex_exit(hash_lock);
}
+/*
+ * The arc_buf_hdr_t's b_flags should never be modified directly. Instead,
+ * the following functions should be used to ensure that the flags are
+ * updated in a thread-safe way. When manipulating the flags either
+ * the hash_lock must be held or the hdr must be undiscoverable. This
+ * ensures that we're not racing with any other threads when updating
+ * the flags.
+ */
+static inline void
+arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
+{
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
+ hdr->b_flags |= flags;
+}
+
+static inline void
+arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
+{
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
+ hdr->b_flags &= ~flags;
+}
+
+/*
+ * Setting the compression bits in the arc_buf_hdr_t's b_flags is
+ * done in a special way since we have to clear and set bits
+ * at the same time. Consumers that wish to set the compression bits
+ * must use this function to ensure that the flags are updated in
+ * thread-safe manner.
+ */
+static void
+arc_hdr_set_compress(arc_buf_hdr_t *hdr, enum zio_compress cmp)
+{
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
+
+ /*
+ * Holes and embedded blocks will always have a psize = 0 so
+ * we ignore the compression of the blkptr and set the
+ * arc_buf_hdr_t's compression to ZIO_COMPRESS_OFF.
+ * Holes and embedded blocks remain anonymous so we don't
+ * want to uncompress them. Mark them as uncompressed.
+ */
+ if (!zfs_compressed_arc_enabled || HDR_GET_PSIZE(hdr) == 0) {
+ arc_hdr_clear_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
+ HDR_SET_COMPRESS(hdr, ZIO_COMPRESS_OFF);
+ ASSERT(!HDR_COMPRESSION_ENABLED(hdr));
+ ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
+ } else {
+ arc_hdr_set_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
+ HDR_SET_COMPRESS(hdr, cmp);
+ ASSERT3U(HDR_GET_COMPRESS(hdr), ==, cmp);
+ ASSERT(HDR_COMPRESSION_ENABLED(hdr));
+ }
+}
+
+static int
+arc_decompress(arc_buf_t *buf)
+{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+ dmu_object_byteswap_t bswap = hdr->b_l1hdr.b_byteswap;
+ int error;
+
+ if (arc_buf_is_shared(buf)) {
+ ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
+ } else if (HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF) {
+ /*
+ * The arc_buf_hdr_t is either not compressed or is
+ * associated with an embedded block or a hole in which
+ * case they remain anonymous.
+ */
+ IMPLY(HDR_COMPRESSION_ENABLED(hdr), HDR_GET_PSIZE(hdr) == 0 ||
+ HDR_GET_PSIZE(hdr) == HDR_GET_LSIZE(hdr));
+ ASSERT(!HDR_SHARED_DATA(hdr));
+ bcopy(hdr->b_l1hdr.b_pdata, buf->b_data, HDR_GET_LSIZE(hdr));
+ } else {
+ ASSERT(!HDR_SHARED_DATA(hdr));
+ ASSERT3U(HDR_GET_LSIZE(hdr), !=, HDR_GET_PSIZE(hdr));
+ error = zio_decompress_data(HDR_GET_COMPRESS(hdr),
+ hdr->b_l1hdr.b_pdata, buf->b_data, HDR_GET_PSIZE(hdr),
+ HDR_GET_LSIZE(hdr));
+ if (error != 0) {
+ zfs_dbgmsg("hdr %p, compress %d, psize %d, lsize %d",
+ hdr, HDR_GET_COMPRESS(hdr), HDR_GET_PSIZE(hdr),
+ HDR_GET_LSIZE(hdr));
+ return (SET_ERROR(EIO));
+ }
+ }
+ if (bswap != DMU_BSWAP_NUMFUNCS) {
+ ASSERT(!HDR_SHARED_DATA(hdr));
+ ASSERT3U(bswap, <, DMU_BSWAP_NUMFUNCS);
+ dmu_ot_byteswap[bswap].ob_func(buf->b_data, HDR_GET_LSIZE(hdr));
+ }
+ arc_cksum_compute(buf);
+ return (0);
+}
+
+/*
+ * Return the size of the block, b_pdata, that is stored in the arc_buf_hdr_t.
+ */
+static uint64_t
+arc_hdr_size(arc_buf_hdr_t *hdr)
+{
+ uint64_t size;
+
+ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
+ HDR_GET_PSIZE(hdr) > 0) {
+ size = HDR_GET_PSIZE(hdr);
+ } else {
+ ASSERT3U(HDR_GET_LSIZE(hdr), !=, 0);
+ size = HDR_GET_LSIZE(hdr);
+ }
+ return (size);
+}
+
+/*
+ * Increment the amount of evictable space in the arc_state_t's refcount.
+ * We account for the space used by the hdr and the arc buf individually
+ * so that we can add and remove them from the refcount individually.
+ */
static void
-add_reference(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, void *tag)
+arc_evictable_space_increment(arc_buf_hdr_t *hdr, arc_state_t *state)
+{
+ arc_buf_contents_t type = arc_buf_type(hdr);
+ uint64_t lsize = HDR_GET_LSIZE(hdr);
+ arc_buf_t *buf;
+
+ ASSERT(HDR_HAS_L1HDR(hdr));
+
+ if (GHOST_STATE(state)) {
+ ASSERT0(hdr->b_l1hdr.b_bufcnt);
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+ (void) refcount_add_many(&state->arcs_esize[type], lsize, hdr);
+ return;
+ }
+
+ ASSERT(!GHOST_STATE(state));
+ if (hdr->b_l1hdr.b_pdata != NULL) {
+ (void) refcount_add_many(&state->arcs_esize[type],
+ arc_hdr_size(hdr), hdr);
+ }
+ for (buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) {
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(ARC_BUF_LAST(buf));
+ continue;
+ }
+ (void) refcount_add_many(&state->arcs_esize[type], lsize, buf);
+ }
+}
+
+/*
+ * Decrement the amount of evictable space in the arc_state_t's refcount.
+ * We account for the space used by the hdr and the arc buf individually
+ * so that we can add and remove them from the refcount individually.
+ */
+static void
+arc_evitable_space_decrement(arc_buf_hdr_t *hdr, arc_state_t *state)
+{
+ arc_buf_contents_t type = arc_buf_type(hdr);
+ uint64_t lsize = HDR_GET_LSIZE(hdr);
+ arc_buf_t *buf;
+
+ ASSERT(HDR_HAS_L1HDR(hdr));
+
+ if (GHOST_STATE(state)) {
+ ASSERT0(hdr->b_l1hdr.b_bufcnt);
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+ (void) refcount_remove_many(&state->arcs_esize[type],
+ lsize, hdr);
+ return;
+ }
+
+ ASSERT(!GHOST_STATE(state));
+ if (hdr->b_l1hdr.b_pdata != NULL) {
+ (void) refcount_remove_many(&state->arcs_esize[type],
+ arc_hdr_size(hdr), hdr);
+ }
+ for (buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) {
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(ARC_BUF_LAST(buf));
+ continue;
+ }
+ (void) refcount_remove_many(&state->arcs_esize[type],
+ lsize, buf);
+ }
+}
+
+/*
+ * Add a reference to this hdr indicating that someone is actively
+ * referencing that memory. When the refcount transitions from 0 to 1,
+ * we remove it from the respective arc_state_t list to indicate that
+ * it is not evictable.
+ */
+static void
+add_reference(arc_buf_hdr_t *hdr, void *tag)
{
arc_state_t *state;
ASSERT(HDR_HAS_L1HDR(hdr));
- ASSERT(MUTEX_HELD(hash_lock));
+ if (!MUTEX_HELD(HDR_LOCK(hdr))) {
+ ASSERT(hdr->b_l1hdr.b_state == arc_anon);
+ ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ }
state = hdr->b_l1hdr.b_state;
(state != arc_anon)) {
/* We don't use the L2-only state list. */
if (state != arc_l2c_only) {
- arc_buf_contents_t type = arc_buf_type(hdr);
- uint64_t delta = hdr->b_size * hdr->b_l1hdr.b_datacnt;
- multilist_t *list = &state->arcs_list[type];
- uint64_t *size = &state->arcs_lsize[type];
-
- multilist_remove(list, hdr);
-
- if (GHOST_STATE(state)) {
- ASSERT0(hdr->b_l1hdr.b_datacnt);
- ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
- delta = hdr->b_size;
- }
- ASSERT(delta > 0);
- ASSERT3U(*size, >=, delta);
- atomic_add_64(size, -delta);
+ multilist_remove(&state->arcs_list[arc_buf_type(hdr)],
+ hdr);
+ arc_evitable_space_decrement(hdr, state);
}
/* remove the prefetch flag if we get a reference */
- hdr->b_flags &= ~ARC_FLAG_PREFETCH;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH);
}
}
+/*
+ * Remove a reference from this hdr. When the reference transitions from
+ * 1 to 0 and we're not anonymous, then we add this hdr to the arc_state_t's
+ * list making it eligible for eviction.
+ */
static int
remove_reference(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, void *tag)
{
*/
if (((cnt = refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) == 0) &&
(state != arc_anon)) {
- arc_buf_contents_t type = arc_buf_type(hdr);
- multilist_t *list = &state->arcs_list[type];
- uint64_t *size = &state->arcs_lsize[type];
-
- multilist_insert(list, hdr);
-
- ASSERT(hdr->b_l1hdr.b_datacnt > 0);
- atomic_add_64(size, hdr->b_size *
- hdr->b_l1hdr.b_datacnt);
+ multilist_insert(&state->arcs_list[arc_buf_type(hdr)], hdr);
+ ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0);
+ arc_evictable_space_increment(hdr, state);
}
return (cnt);
}
l2hdr = &hdr->b_l2hdr;
if (l1hdr) {
- abi->abi_datacnt = l1hdr->b_datacnt;
+ abi->abi_bufcnt = l1hdr->b_bufcnt;
abi->abi_access = l1hdr->b_arc_access;
abi->abi_mru_hits = l1hdr->b_mru_hits;
abi->abi_mru_ghost_hits = l1hdr->b_mru_ghost_hits;
if (l2hdr) {
abi->abi_l2arc_dattr = l2hdr->b_daddr;
- abi->abi_l2arc_asize = l2hdr->b_asize;
- abi->abi_l2arc_compress = l2hdr->b_compress;
abi->abi_l2arc_hits = l2hdr->b_hits;
}
abi->abi_state_type = state ? state->arcs_state : ARC_STATE_ANON;
abi->abi_state_contents = arc_buf_type(hdr);
- abi->abi_size = hdr->b_size;
+ abi->abi_size = arc_hdr_size(hdr);
}
/*
{
arc_state_t *old_state;
int64_t refcnt;
- uint32_t datacnt;
- uint64_t from_delta, to_delta;
+ uint32_t bufcnt;
+ boolean_t update_old, update_new;
arc_buf_contents_t buftype = arc_buf_type(hdr);
/*
if (HDR_HAS_L1HDR(hdr)) {
old_state = hdr->b_l1hdr.b_state;
refcnt = refcount_count(&hdr->b_l1hdr.b_refcnt);
- datacnt = hdr->b_l1hdr.b_datacnt;
+ bufcnt = hdr->b_l1hdr.b_bufcnt;
+ update_old = (bufcnt > 0 || hdr->b_l1hdr.b_pdata != NULL);
} else {
old_state = arc_l2c_only;
refcnt = 0;
- datacnt = 0;
+ bufcnt = 0;
+ update_old = B_FALSE;
}
+ update_new = update_old;
ASSERT(MUTEX_HELD(hash_lock));
ASSERT3P(new_state, !=, old_state);
- ASSERT(refcnt == 0 || datacnt > 0);
- ASSERT(!GHOST_STATE(new_state) || datacnt == 0);
- ASSERT(old_state != arc_anon || datacnt <= 1);
-
- from_delta = to_delta = datacnt * hdr->b_size;
+ ASSERT(!GHOST_STATE(new_state) || bufcnt == 0);
+ ASSERT(old_state != arc_anon || bufcnt <= 1);
/*
* If this buffer is evictable, transfer it from the
*/
if (refcnt == 0) {
if (old_state != arc_anon && old_state != arc_l2c_only) {
- uint64_t *size = &old_state->arcs_lsize[buftype];
-
ASSERT(HDR_HAS_L1HDR(hdr));
multilist_remove(&old_state->arcs_list[buftype], hdr);
- /*
- * If prefetching out of the ghost cache,
- * we will have a non-zero datacnt.
- */
- if (GHOST_STATE(old_state) && datacnt == 0) {
- /* ghost elements have a ghost size */
- ASSERT(hdr->b_l1hdr.b_buf == NULL);
- from_delta = hdr->b_size;
+ if (GHOST_STATE(old_state)) {
+ ASSERT0(bufcnt);
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ update_old = B_TRUE;
}
- ASSERT3U(*size, >=, from_delta);
- atomic_add_64(size, -from_delta);
+ arc_evitable_space_decrement(hdr, old_state);
}
if (new_state != arc_anon && new_state != arc_l2c_only) {
- uint64_t *size = &new_state->arcs_lsize[buftype];
-
/*
* An L1 header always exists here, since if we're
* moving to some L1-cached state (i.e. not l2c_only or
ASSERT(HDR_HAS_L1HDR(hdr));
multilist_insert(&new_state->arcs_list[buftype], hdr);
- /* ghost elements have a ghost size */
if (GHOST_STATE(new_state)) {
- ASSERT0(datacnt);
- ASSERT(hdr->b_l1hdr.b_buf == NULL);
- to_delta = hdr->b_size;
+ ASSERT0(bufcnt);
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ update_new = B_TRUE;
}
- atomic_add_64(size, to_delta);
+ arc_evictable_space_increment(hdr, new_state);
}
}
- ASSERT(!BUF_EMPTY(hdr));
+ ASSERT(!HDR_EMPTY(hdr));
if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr))
buf_hash_remove(hdr);
/* adjust state sizes (ignore arc_l2c_only) */
- if (to_delta && new_state != arc_l2c_only) {
+ if (update_new && new_state != arc_l2c_only) {
ASSERT(HDR_HAS_L1HDR(hdr));
if (GHOST_STATE(new_state)) {
- ASSERT0(datacnt);
+ ASSERT0(bufcnt);
/*
- * We moving a header to a ghost state, we first
+ * When moving a header to a ghost state, we first
* remove all arc buffers. Thus, we'll have a
- * datacnt of zero, and no arc buffer to use for
+ * bufcnt of zero, and no arc buffer to use for
* the reference. As a result, we use the arc
* header pointer for the reference.
*/
(void) refcount_add_many(&new_state->arcs_size,
- hdr->b_size, hdr);
+ HDR_GET_LSIZE(hdr), hdr);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
} else {
arc_buf_t *buf;
- ASSERT3U(datacnt, !=, 0);
+ uint32_t buffers = 0;
/*
* Each individual buffer holds a unique reference,
*/
for (buf = hdr->b_l1hdr.b_buf; buf != NULL;
buf = buf->b_next) {
+ ASSERT3U(bufcnt, !=, 0);
+ buffers++;
+
+ /*
+ * When the arc_buf_t is sharing the data
+ * block with the hdr, the owner of the
+ * reference belongs to the hdr. Only
+ * add to the refcount if the arc_buf_t is
+ * not shared.
+ */
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(ARC_BUF_LAST(buf));
+ continue;
+ }
+
(void) refcount_add_many(&new_state->arcs_size,
- hdr->b_size, buf);
+ HDR_GET_LSIZE(hdr), buf);
+ }
+ ASSERT3U(bufcnt, ==, buffers);
+
+ if (hdr->b_l1hdr.b_pdata != NULL) {
+ (void) refcount_add_many(&new_state->arcs_size,
+ arc_hdr_size(hdr), hdr);
+ } else {
+ ASSERT(GHOST_STATE(old_state));
}
}
}
- if (from_delta && old_state != arc_l2c_only) {
+ if (update_old && old_state != arc_l2c_only) {
ASSERT(HDR_HAS_L1HDR(hdr));
if (GHOST_STATE(old_state)) {
+ ASSERT0(bufcnt);
+
/*
* When moving a header off of a ghost state,
- * there's the possibility for datacnt to be
- * non-zero. This is because we first add the
- * arc buffer to the header prior to changing
- * the header's state. Since we used the header
- * for the reference when putting the header on
- * the ghost state, we must balance that and use
- * the header when removing off the ghost state
- * (even though datacnt is non zero).
+ * the header will not contain any arc buffers.
+ * We use the arc header pointer for the reference
+ * which is exactly what we did when we put the
+ * header on the ghost state.
*/
- IMPLY(datacnt == 0, new_state == arc_anon ||
- new_state == arc_l2c_only);
-
(void) refcount_remove_many(&old_state->arcs_size,
- hdr->b_size, hdr);
+ HDR_GET_LSIZE(hdr), hdr);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
} else {
arc_buf_t *buf;
- ASSERT3U(datacnt, !=, 0);
+ uint32_t buffers = 0;
/*
* Each individual buffer holds a unique reference,
*/
for (buf = hdr->b_l1hdr.b_buf; buf != NULL;
buf = buf->b_next) {
+ ASSERT3U(bufcnt, !=, 0);
+ buffers++;
+
+ /*
+ * When the arc_buf_t is sharing the data
+ * block with the hdr, the owner of the
+ * reference belongs to the hdr. Only
+ * add to the refcount if the arc_buf_t is
+ * not shared.
+ */
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(ARC_BUF_LAST(buf));
+ continue;
+ }
+
(void) refcount_remove_many(
- &old_state->arcs_size, hdr->b_size, buf);
+ &old_state->arcs_size, HDR_GET_LSIZE(hdr),
+ buf);
}
+ ASSERT3U(bufcnt, ==, buffers);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+ (void) refcount_remove_many(
+ &old_state->arcs_size, arc_hdr_size(hdr), hdr);
}
}
atomic_add_64(&arc_size, -space);
}
-arc_buf_t *
-arc_buf_alloc(spa_t *spa, uint64_t size, void *tag, arc_buf_contents_t type)
+/*
+ * Allocate an initial buffer for this hdr, subsequent buffers will
+ * use arc_buf_clone().
+ */
+static arc_buf_t *
+arc_buf_alloc_impl(arc_buf_hdr_t *hdr, void *tag)
{
- arc_buf_hdr_t *hdr;
arc_buf_t *buf;
- VERIFY3U(size, <=, spa_maxblocksize(spa));
- hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE);
- ASSERT(BUF_EMPTY(hdr));
- ASSERT3P(hdr->b_freeze_cksum, ==, NULL);
- hdr->b_size = size;
- hdr->b_spa = spa_load_guid(spa);
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
+ VERIFY(hdr->b_type == ARC_BUFC_DATA ||
+ hdr->b_type == ARC_BUFC_METADATA);
+
+ ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ ASSERT0(hdr->b_l1hdr.b_bufcnt);
hdr->b_l1hdr.b_mru_hits = 0;
hdr->b_l1hdr.b_mru_ghost_hits = 0;
hdr->b_l1hdr.b_mfu_hits = 0;
buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
buf->b_hdr = hdr;
buf->b_data = NULL;
- buf->b_efunc = NULL;
- buf->b_private = NULL;
buf->b_next = NULL;
- hdr->b_flags = arc_bufc_to_flags(type);
- hdr->b_flags |= ARC_FLAG_HAS_L1HDR;
+ add_reference(hdr, tag);
+
+ /*
+ * We're about to change the hdr's b_flags. We must either
+ * hold the hash_lock or be undiscoverable.
+ */
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
+
+ /*
+ * If the hdr's data can be shared (no byteswapping, hdr is
+ * uncompressed, hdr's data is not currently being written to the
+ * L2ARC write) then we share the data buffer and set the appropriate
+ * bit in the hdr's b_flags to indicate the hdr is sharing it's
+ * b_pdata with the arc_buf_t. Otherwise, we allocate a new buffer to
+ * store the buf's data.
+ */
+ if (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS &&
+ HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF && !HDR_L2_WRITING(hdr)) {
+ buf->b_data = hdr->b_l1hdr.b_pdata;
+ arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
+ } else {
+ buf->b_data = arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);
+ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr));
+ arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
+ }
+ VERIFY3P(buf->b_data, !=, NULL);
hdr->b_l1hdr.b_buf = buf;
- hdr->b_l1hdr.b_state = arc_anon;
- hdr->b_l1hdr.b_arc_access = 0;
- hdr->b_l1hdr.b_datacnt = 1;
- hdr->b_l1hdr.b_tmp_cdata = NULL;
+ hdr->b_l1hdr.b_bufcnt += 1;
- arc_get_data_buf(buf);
- ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
- (void) refcount_add(&hdr->b_l1hdr.b_refcnt, tag);
+ return (buf);
+}
+
+/*
+ * Used when allocating additional buffers.
+ */
+static arc_buf_t *
+arc_buf_clone(arc_buf_t *from)
+{
+ arc_buf_t *buf;
+ arc_buf_hdr_t *hdr = from->b_hdr;
+ uint64_t size = HDR_GET_LSIZE(hdr);
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ ASSERT(hdr->b_l1hdr.b_state != arc_anon);
+
+ buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
+ buf->b_hdr = hdr;
+ buf->b_data = NULL;
+ buf->b_next = hdr->b_l1hdr.b_buf;
+ hdr->b_l1hdr.b_buf = buf;
+ buf->b_data = arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);
+ bcopy(from->b_data, buf->b_data, size);
+ hdr->b_l1hdr.b_bufcnt += 1;
+
+ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr));
return (buf);
}
{
arc_buf_t *buf;
- buf = arc_buf_alloc(spa, size, arc_onloan_tag, ARC_BUFC_DATA);
+ buf = arc_alloc_buf(spa, size, arc_onloan_tag, ARC_BUFC_DATA);
atomic_add_64(&arc_loaned_bytes, size);
return (buf);
{
arc_buf_hdr_t *hdr = buf->b_hdr;
- ASSERT(buf->b_data != NULL);
+ ASSERT3P(buf->b_data, !=, NULL);
ASSERT(HDR_HAS_L1HDR(hdr));
(void) refcount_add(&hdr->b_l1hdr.b_refcnt, tag);
(void) refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);
- atomic_add_64(&arc_loaned_bytes, -hdr->b_size);
+ atomic_add_64(&arc_loaned_bytes, -HDR_GET_LSIZE(hdr));
}
/* Detach an arc_buf from a dbuf (tag) */
{
arc_buf_hdr_t *hdr = buf->b_hdr;
- ASSERT(buf->b_data != NULL);
+ ASSERT3P(buf->b_data, !=, NULL);
ASSERT(HDR_HAS_L1HDR(hdr));
(void) refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);
(void) refcount_remove(&hdr->b_l1hdr.b_refcnt, tag);
- buf->b_efunc = NULL;
- buf->b_private = NULL;
- atomic_add_64(&arc_loaned_bytes, hdr->b_size);
+ atomic_add_64(&arc_loaned_bytes, HDR_GET_LSIZE(hdr));
}
-static arc_buf_t *
-arc_buf_clone(arc_buf_t *from)
+static void
+l2arc_free_data_on_write(void *data, size_t size, arc_buf_contents_t type)
{
- arc_buf_t *buf;
- arc_buf_hdr_t *hdr = from->b_hdr;
- uint64_t size = hdr->b_size;
+ l2arc_data_free_t *df = kmem_alloc(sizeof (*df), KM_SLEEP);
- ASSERT(HDR_HAS_L1HDR(hdr));
- ASSERT(hdr->b_l1hdr.b_state != arc_anon);
+ df->l2df_data = data;
+ df->l2df_size = size;
+ df->l2df_type = type;
+ mutex_enter(&l2arc_free_on_write_mtx);
+ list_insert_head(l2arc_free_on_write, df);
+ mutex_exit(&l2arc_free_on_write_mtx);
+}
- buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
- buf->b_hdr = hdr;
- buf->b_data = NULL;
- buf->b_efunc = NULL;
- buf->b_private = NULL;
- buf->b_next = hdr->b_l1hdr.b_buf;
- hdr->b_l1hdr.b_buf = buf;
- arc_get_data_buf(buf);
- bcopy(from->b_data, buf->b_data, size);
+static void
+arc_hdr_free_on_write(arc_buf_hdr_t *hdr)
+{
+ arc_state_t *state = hdr->b_l1hdr.b_state;
+ arc_buf_contents_t type = arc_buf_type(hdr);
+ uint64_t size = arc_hdr_size(hdr);
- /*
- * This buffer already exists in the arc so create a duplicate
- * copy for the caller. If the buffer is associated with user data
- * then track the size and number of duplicates. These stats will be
- * updated as duplicate buffers are created and destroyed.
- */
- if (HDR_ISTYPE_DATA(hdr)) {
- ARCSTAT_BUMP(arcstat_duplicate_buffers);
- ARCSTAT_INCR(arcstat_duplicate_buffers_size, size);
+ /* protected by hash lock, if in the hash table */
+ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
+ ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
+ ASSERT(state != arc_anon && state != arc_l2c_only);
+
+ (void) refcount_remove_many(&state->arcs_esize[type],
+ size, hdr);
}
- hdr->b_l1hdr.b_datacnt += 1;
- return (buf);
+ (void) refcount_remove_many(&state->arcs_size, size, hdr);
+
+ l2arc_free_data_on_write(hdr->b_l1hdr.b_pdata, size, type);
}
-void
-arc_buf_add_ref(arc_buf_t *buf, void* tag)
+/*
+ * Share the arc_buf_t's data with the hdr. Whenever we are sharing the
+ * data buffer, we transfer the refcount ownership to the hdr and update
+ * the appropriate kstats.
+ */
+static void
+arc_share_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf)
{
- arc_buf_hdr_t *hdr;
- kmutex_t *hash_lock;
+ ASSERT(!HDR_SHARED_DATA(hdr));
+ ASSERT(!arc_buf_is_shared(buf));
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
/*
- * Check to see if this buffer is evicted. Callers
- * must verify b_data != NULL to know if the add_ref
- * was successful.
+ * Start sharing the data buffer. We transfer the
+ * refcount ownership to the hdr since it always owns
+ * the refcount whenever an arc_buf_t is shared.
*/
- mutex_enter(&buf->b_evict_lock);
- if (buf->b_data == NULL) {
- mutex_exit(&buf->b_evict_lock);
- return;
- }
- hash_lock = HDR_LOCK(buf->b_hdr);
- mutex_enter(hash_lock);
- hdr = buf->b_hdr;
- ASSERT(HDR_HAS_L1HDR(hdr));
- ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
- mutex_exit(&buf->b_evict_lock);
+ refcount_transfer_ownership(&hdr->b_l1hdr.b_state->arcs_size, buf, hdr);
+ hdr->b_l1hdr.b_pdata = buf->b_data;
+ arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
- ASSERT(hdr->b_l1hdr.b_state == arc_mru ||
- hdr->b_l1hdr.b_state == arc_mfu);
-
- add_reference(hdr, hash_lock, tag);
- DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
- arc_access(hdr, hash_lock);
- mutex_exit(hash_lock);
- ARCSTAT_BUMP(arcstat_hits);
- ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
- demand, prefetch, !HDR_ISTYPE_METADATA(hdr),
- data, metadata, hits);
+ /*
+ * Since we've transferred ownership to the hdr we need
+ * to increment its compressed and uncompressed kstats and
+ * decrement the overhead size.
+ */
+ ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr));
+ ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
+ ARCSTAT_INCR(arcstat_overhead_size, -HDR_GET_LSIZE(hdr));
}
static void
-arc_buf_free_on_write(void *data, size_t size,
- void (*free_func)(void *, size_t))
+arc_unshare_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf)
{
- l2arc_data_free_t *df;
+ ASSERT(HDR_SHARED_DATA(hdr));
+ ASSERT(arc_buf_is_shared(buf));
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
- df = kmem_alloc(sizeof (*df), KM_SLEEP);
- df->l2df_data = data;
- df->l2df_size = size;
- df->l2df_func = free_func;
- mutex_enter(&l2arc_free_on_write_mtx);
- list_insert_head(l2arc_free_on_write, df);
- mutex_exit(&l2arc_free_on_write_mtx);
+ /*
+ * We are no longer sharing this buffer so we need
+ * to transfer its ownership to the rightful owner.
+ */
+ refcount_transfer_ownership(&hdr->b_l1hdr.b_state->arcs_size, hdr, buf);
+ arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
+ hdr->b_l1hdr.b_pdata = NULL;
+
+ /*
+ * Since the buffer is no longer shared between
+ * the arc buf and the hdr, count it as overhead.
+ */
+ ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr));
+ ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
+ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr));
}
/*
- * Free the arc data buffer. If it is an l2arc write in progress,
- * the buffer is placed on l2arc_free_on_write to be freed later.
+ * Free up buf->b_data and if 'remove' is set, then pull the
+ * arc_buf_t off of the the arc_buf_hdr_t's list and free it.
*/
static void
-arc_buf_data_free(arc_buf_t *buf, void (*free_func)(void *, size_t))
+arc_buf_destroy_impl(arc_buf_t *buf, boolean_t remove)
{
+ arc_buf_t **bufp;
arc_buf_hdr_t *hdr = buf->b_hdr;
-
- if (HDR_L2_WRITING(hdr)) {
- arc_buf_free_on_write(buf->b_data, hdr->b_size, free_func);
- ARCSTAT_BUMP(arcstat_l2_free_on_write);
- } else {
- free_func(buf->b_data, hdr->b_size);
- }
-}
-
-static void
-arc_buf_l2_cdata_free(arc_buf_hdr_t *hdr)
-{
- ASSERT(HDR_HAS_L2HDR(hdr));
- ASSERT(MUTEX_HELD(&hdr->b_l2hdr.b_dev->l2ad_mtx));
+ arc_buf_t *lastbuf = NULL;
+ uint64_t size = HDR_GET_LSIZE(hdr);
+ boolean_t destroyed_buf_is_shared = arc_buf_is_shared(buf);
/*
- * The b_tmp_cdata field is linked off of the b_l1hdr, so if
- * that doesn't exist, the header is in the arc_l2c_only state,
- * and there isn't anything to free (it's already been freed).
+ * Free up the data associated with the buf but only
+ * if we're not sharing this with the hdr. If we are sharing
+ * it with the hdr, then hdr will have performed the allocation
+ * so allow it to do the free.
*/
- if (!HDR_HAS_L1HDR(hdr))
+ if (buf->b_data != NULL) {
+ /*
+ * We're about to change the hdr's b_flags. We must either
+ * hold the hash_lock or be undiscoverable.
+ */
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr));
+
+ arc_cksum_verify(buf);
+ arc_buf_unwatch(buf);
+
+ if (destroyed_buf_is_shared) {
+ ASSERT(ARC_BUF_LAST(buf));
+ ASSERT(HDR_SHARED_DATA(hdr));
+ arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
+ } else {
+ arc_free_data_buf(hdr, buf->b_data, size, buf);
+ ARCSTAT_INCR(arcstat_overhead_size, -size);
+ }
+ buf->b_data = NULL;
+
+ ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
+ hdr->b_l1hdr.b_bufcnt -= 1;
+ }
+
+ /* only remove the buf if requested */
+ if (!remove)
return;
+ /* remove the buf from the hdr list */
+ bufp = &hdr->b_l1hdr.b_buf;
+ while (*bufp != NULL) {
+ if (*bufp == buf)
+ *bufp = buf->b_next;
+
+ /*
+ * If we've removed a buffer in the middle of
+ * the list then update the lastbuf and update
+ * bufp.
+ */
+ if (*bufp != NULL) {
+ lastbuf = *bufp;
+ bufp = &(*bufp)->b_next;
+ }
+ }
+ buf->b_next = NULL;
+ ASSERT3P(lastbuf, !=, buf);
+
/*
- * The header isn't being written to the l2arc device, thus it
- * shouldn't have a b_tmp_cdata to free.
+ * If the current arc_buf_t is sharing its data
+ * buffer with the hdr, then reassign the hdr's
+ * b_pdata to share it with the new buffer at the end
+ * of the list. The shared buffer is always the last one
+ * on the hdr's buffer list.
*/
- if (!HDR_L2_WRITING(hdr)) {
- ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL);
- return;
+ if (destroyed_buf_is_shared && lastbuf != NULL) {
+ ASSERT(ARC_BUF_LAST(buf));
+ ASSERT(ARC_BUF_LAST(lastbuf));
+ VERIFY(!arc_buf_is_shared(lastbuf));
+
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+ arc_hdr_free_pdata(hdr);
+
+ /*
+ * We must setup a new shared block between the
+ * last buffer and the hdr. The data would have
+ * been allocated by the arc buf so we need to transfer
+ * ownership to the hdr since it's now being shared.
+ */
+ arc_share_buf(hdr, lastbuf);
+ } else if (HDR_SHARED_DATA(hdr)) {
+ ASSERT(arc_buf_is_shared(lastbuf));
}
+ if (hdr->b_l1hdr.b_bufcnt == 0)
+ arc_cksum_free(hdr);
+
+ /* clean up the buf */
+ buf->b_hdr = NULL;
+ kmem_cache_free(buf_cache, buf);
+}
+
+static void
+arc_hdr_alloc_pdata(arc_buf_hdr_t *hdr)
+{
+ ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ ASSERT(!HDR_SHARED_DATA(hdr));
+
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+ hdr->b_l1hdr.b_pdata = arc_get_data_buf(hdr, arc_hdr_size(hdr), hdr);
+ hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+
+ ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr));
+ ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
+}
+
+static void
+arc_hdr_free_pdata(arc_buf_hdr_t *hdr)
+{
+ ASSERT(HDR_HAS_L1HDR(hdr));
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+
/*
- * The header does not have compression enabled. This can be due
- * to the buffer not being compressible, or because we're
- * freeing the buffer before the second phase of
- * l2arc_write_buffer() has started (which does the compression
- * step). In either case, b_tmp_cdata does not point to a
- * separately compressed buffer, so there's nothing to free (it
- * points to the same buffer as the arc_buf_t's b_data field).
+ * If the hdr is currently being written to the l2arc then
+ * we defer freeing the data by adding it to the l2arc_free_on_write
+ * list. The l2arc will free the data once it's finished
+ * writing it to the l2arc device.
*/
- if (hdr->b_l2hdr.b_compress == ZIO_COMPRESS_OFF) {
- hdr->b_l1hdr.b_tmp_cdata = NULL;
- return;
+ if (HDR_L2_WRITING(hdr)) {
+ arc_hdr_free_on_write(hdr);
+ ARCSTAT_BUMP(arcstat_l2_free_on_write);
+ } else {
+ arc_free_data_buf(hdr, hdr->b_l1hdr.b_pdata,
+ arc_hdr_size(hdr), hdr);
}
+ hdr->b_l1hdr.b_pdata = NULL;
+ hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
- /*
- * There's nothing to free since the buffer was all zero's and
- * compressed to a zero length buffer.
- */
- if (hdr->b_l2hdr.b_compress == ZIO_COMPRESS_EMPTY) {
- ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL);
- return;
- }
+ ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr));
+ ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
+}
+
+static arc_buf_hdr_t *
+arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize,
+ enum zio_compress compress, arc_buf_contents_t type)
+{
+ arc_buf_hdr_t *hdr;
+
+ ASSERT3U(lsize, >, 0);
+ VERIFY(type == ARC_BUFC_DATA || type == ARC_BUFC_METADATA);
+
+ hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE);
+ ASSERT(HDR_EMPTY(hdr));
+ ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
+ HDR_SET_PSIZE(hdr, psize);
+ HDR_SET_LSIZE(hdr, lsize);
+ hdr->b_spa = spa;
+ hdr->b_type = type;
+ hdr->b_flags = 0;
+ arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L1HDR);
+ arc_hdr_set_compress(hdr, compress);
- ASSERT(L2ARC_IS_VALID_COMPRESS(hdr->b_l2hdr.b_compress));
+ hdr->b_l1hdr.b_state = arc_anon;
+ hdr->b_l1hdr.b_arc_access = 0;
+ hdr->b_l1hdr.b_bufcnt = 0;
+ hdr->b_l1hdr.b_buf = NULL;
- arc_buf_free_on_write(hdr->b_l1hdr.b_tmp_cdata,
- hdr->b_size, zio_data_buf_free);
+ /*
+ * Allocate the hdr's buffer. This will contain either
+ * the compressed or uncompressed data depending on the block
+ * it references and compressed arc enablement.
+ */
+ arc_hdr_alloc_pdata(hdr);
+ ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
- ARCSTAT_BUMP(arcstat_l2_cdata_free_on_write);
- hdr->b_l1hdr.b_tmp_cdata = NULL;
+ return (hdr);
}
/*
- * Free up buf->b_data and if 'remove' is set, then pull the
- * arc_buf_t off of the the arc_buf_hdr_t's list and free it.
+ * Transition between the two allocation states for the arc_buf_hdr struct.
+ * The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without
+ * (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller
+ * version is used when a cache buffer is only in the L2ARC in order to reduce
+ * memory usage.
*/
-static void
-arc_buf_destroy(arc_buf_t *buf, boolean_t remove)
+static arc_buf_hdr_t *
+arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new)
{
- arc_buf_t **bufp;
+ arc_buf_hdr_t *nhdr;
+ l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;
- /* free up data associated with the buf */
- if (buf->b_data != NULL) {
- arc_state_t *state = buf->b_hdr->b_l1hdr.b_state;
- uint64_t size = buf->b_hdr->b_size;
- arc_buf_contents_t type = arc_buf_type(buf->b_hdr);
+ ASSERT(HDR_HAS_L2HDR(hdr));
+ ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) ||
+ (old == hdr_l2only_cache && new == hdr_full_cache));
- arc_cksum_verify(buf);
- arc_buf_unwatch(buf);
+ nhdr = kmem_cache_alloc(new, KM_PUSHPAGE);
- if (type == ARC_BUFC_METADATA) {
- arc_buf_data_free(buf, zio_buf_free);
- arc_space_return(size, ARC_SPACE_META);
- } else {
- ASSERT(type == ARC_BUFC_DATA);
- arc_buf_data_free(buf, zio_data_buf_free);
- arc_space_return(size, ARC_SPACE_DATA);
- }
+ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
+ buf_hash_remove(hdr);
- /* protected by hash lock, if in the hash table */
- if (multilist_link_active(&buf->b_hdr->b_l1hdr.b_arc_node)) {
- uint64_t *cnt = &state->arcs_lsize[type];
+ bcopy(hdr, nhdr, HDR_L2ONLY_SIZE);
- ASSERT(refcount_is_zero(
- &buf->b_hdr->b_l1hdr.b_refcnt));
- ASSERT(state != arc_anon && state != arc_l2c_only);
+ if (new == hdr_full_cache) {
+ arc_hdr_set_flags(nhdr, ARC_FLAG_HAS_L1HDR);
+ /*
+ * arc_access and arc_change_state need to be aware that a
+ * header has just come out of L2ARC, so we set its state to
+ * l2c_only even though it's about to change.
+ */
+ nhdr->b_l1hdr.b_state = arc_l2c_only;
- ASSERT3U(*cnt, >=, size);
- atomic_add_64(cnt, -size);
- }
+ /* Verify previous threads set to NULL before freeing */
+ ASSERT3P(nhdr->b_l1hdr.b_pdata, ==, NULL);
+ } else {
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
+ ASSERT0(hdr->b_l1hdr.b_bufcnt);
+ ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
- (void) refcount_remove_many(&state->arcs_size, size, buf);
- buf->b_data = NULL;
+ /*
+ * If we've reached here, We must have been called from
+ * arc_evict_hdr(), as such we should have already been
+ * removed from any ghost list we were previously on
+ * (which protects us from racing with arc_evict_state),
+ * thus no locking is needed during this check.
+ */
+ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
/*
- * If we're destroying a duplicate buffer make sure
- * that the appropriate statistics are updated.
+ * A buffer must not be moved into the arc_l2c_only
+ * state if it's not finished being written out to the
+ * l2arc device. Otherwise, the b_l1hdr.b_pdata field
+ * might try to be accessed, even though it was removed.
*/
- if (buf->b_hdr->b_l1hdr.b_datacnt > 1 &&
- HDR_ISTYPE_DATA(buf->b_hdr)) {
- ARCSTAT_BUMPDOWN(arcstat_duplicate_buffers);
- ARCSTAT_INCR(arcstat_duplicate_buffers_size, -size);
- }
- ASSERT(buf->b_hdr->b_l1hdr.b_datacnt > 0);
- buf->b_hdr->b_l1hdr.b_datacnt -= 1;
+ VERIFY(!HDR_L2_WRITING(hdr));
+ VERIFY3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+
+ arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR);
}
+ /*
+ * The header has been reallocated so we need to re-insert it into any
+ * lists it was on.
+ */
+ (void) buf_hash_insert(nhdr, NULL);
- /* only remove the buf if requested */
- if (!remove)
- return;
+ ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node));
- /* remove the buf from the hdr list */
- for (bufp = &buf->b_hdr->b_l1hdr.b_buf; *bufp != buf;
- bufp = &(*bufp)->b_next)
- continue;
- *bufp = buf->b_next;
- buf->b_next = NULL;
+ mutex_enter(&dev->l2ad_mtx);
+
+ /*
+ * We must place the realloc'ed header back into the list at
+ * the same spot. Otherwise, if it's placed earlier in the list,
+ * l2arc_write_buffers() could find it during the function's
+ * write phase, and try to write it out to the l2arc.
+ */
+ list_insert_after(&dev->l2ad_buflist, hdr, nhdr);
+ list_remove(&dev->l2ad_buflist, hdr);
- ASSERT(buf->b_efunc == NULL);
+ mutex_exit(&dev->l2ad_mtx);
- /* clean up the buf */
- buf->b_hdr = NULL;
- kmem_cache_free(buf_cache, buf);
+ /*
+ * Since we're using the pointer address as the tag when
+ * incrementing and decrementing the l2ad_alloc refcount, we
+ * must remove the old pointer (that we're about to destroy) and
+ * add the new pointer to the refcount. Otherwise we'd remove
+ * the wrong pointer address when calling arc_hdr_destroy() later.
+ */
+
+ (void) refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr);
+ (void) refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(nhdr), nhdr);
+
+ buf_discard_identity(hdr);
+ kmem_cache_free(old, hdr);
+
+ return (nhdr);
+}
+
+/*
+ * Allocate a new arc_buf_hdr_t and arc_buf_t and return the buf to the caller.
+ * The buf is returned thawed since we expect the consumer to modify it.
+ */
+arc_buf_t *
+arc_alloc_buf(spa_t *spa, int32_t size, void *tag, arc_buf_contents_t type)
+{
+ arc_buf_t *buf;
+ arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size,
+ ZIO_COMPRESS_OFF, type);
+ ASSERT(!MUTEX_HELD(HDR_LOCK(hdr)));
+ buf = arc_buf_alloc_impl(hdr, tag);
+ arc_buf_thaw(buf);
+ return (buf);
}
static void
{
l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
l2arc_dev_t *dev = l2hdr->b_dev;
+ uint64_t asize = arc_hdr_size(hdr);
ASSERT(MUTEX_HELD(&dev->l2ad_mtx));
ASSERT(HDR_HAS_L2HDR(hdr));
list_remove(&dev->l2ad_buflist, hdr);
- /*
- * We don't want to leak the b_tmp_cdata buffer that was
- * allocated in l2arc_write_buffers()
- */
- arc_buf_l2_cdata_free(hdr);
-
- /*
- * If the l2hdr's b_daddr is equal to L2ARC_ADDR_UNSET, then
- * this header is being processed by l2arc_write_buffers() (i.e.
- * it's in the first stage of l2arc_write_buffers()).
- * Re-affirming that truth here, just to serve as a reminder. If
- * b_daddr does not equal L2ARC_ADDR_UNSET, then the header may or
- * may not have its HDR_L2_WRITING flag set. (the write may have
- * completed, in which case HDR_L2_WRITING will be false and the
- * b_daddr field will point to the address of the buffer on disk).
- */
- IMPLY(l2hdr->b_daddr == L2ARC_ADDR_UNSET, HDR_L2_WRITING(hdr));
-
- /*
- * If b_daddr is equal to L2ARC_ADDR_UNSET, we're racing with
- * l2arc_write_buffers(). Since we've just removed this header
- * from the l2arc buffer list, this header will never reach the
- * second stage of l2arc_write_buffers(), which increments the
- * accounting stats for this header. Thus, we must be careful
- * not to decrement them for this header either.
- */
- if (l2hdr->b_daddr != L2ARC_ADDR_UNSET) {
- ARCSTAT_INCR(arcstat_l2_asize, -l2hdr->b_asize);
- ARCSTAT_INCR(arcstat_l2_size, -hdr->b_size);
-
- vdev_space_update(dev->l2ad_vdev,
- -l2hdr->b_asize, 0, 0);
+ ARCSTAT_INCR(arcstat_l2_asize, -asize);
+ ARCSTAT_INCR(arcstat_l2_size, -HDR_GET_LSIZE(hdr));
- (void) refcount_remove_many(&dev->l2ad_alloc,
- l2hdr->b_asize, hdr);
- }
+ vdev_space_update(dev->l2ad_vdev, -asize, 0, 0);
- hdr->b_flags &= ~ARC_FLAG_HAS_L2HDR;
+ (void) refcount_remove_many(&dev->l2ad_alloc, asize, hdr);
+ arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);
}
static void
{
if (HDR_HAS_L1HDR(hdr)) {
ASSERT(hdr->b_l1hdr.b_buf == NULL ||
- hdr->b_l1hdr.b_datacnt > 0);
+ hdr->b_l1hdr.b_bufcnt > 0);
ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
}
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT(!HDR_IN_HASH_TABLE(hdr));
+ if (!HDR_EMPTY(hdr))
+ buf_discard_identity(hdr);
+
if (HDR_HAS_L2HDR(hdr)) {
l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;
boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx);
mutex_exit(&dev->l2ad_mtx);
}
- if (!BUF_EMPTY(hdr))
- buf_discard_identity(hdr);
+ if (HDR_HAS_L1HDR(hdr)) {
+ arc_cksum_free(hdr);
- if (hdr->b_freeze_cksum != NULL) {
- kmem_free(hdr->b_freeze_cksum, sizeof (zio_cksum_t));
- hdr->b_freeze_cksum = NULL;
- }
+ while (hdr->b_l1hdr.b_buf != NULL)
+ arc_buf_destroy_impl(hdr->b_l1hdr.b_buf, B_TRUE);
- if (HDR_HAS_L1HDR(hdr)) {
- while (hdr->b_l1hdr.b_buf) {
- arc_buf_t *buf = hdr->b_l1hdr.b_buf;
-
- if (buf->b_efunc != NULL) {
- mutex_enter(&arc_user_evicts_lock);
- mutex_enter(&buf->b_evict_lock);
- ASSERT(buf->b_hdr != NULL);
- arc_buf_destroy(hdr->b_l1hdr.b_buf, FALSE);
- hdr->b_l1hdr.b_buf = buf->b_next;
- buf->b_hdr = &arc_eviction_hdr;
- buf->b_next = arc_eviction_list;
- arc_eviction_list = buf;
- mutex_exit(&buf->b_evict_lock);
- cv_signal(&arc_user_evicts_cv);
- mutex_exit(&arc_user_evicts_lock);
- } else {
- arc_buf_destroy(hdr->b_l1hdr.b_buf, TRUE);
- }
+ if (hdr->b_l1hdr.b_pdata != NULL) {
+ arc_hdr_free_pdata(hdr);
}
}
}
void
-arc_buf_free(arc_buf_t *buf, void *tag)
-{
- arc_buf_hdr_t *hdr = buf->b_hdr;
- int hashed = hdr->b_l1hdr.b_state != arc_anon;
-
- ASSERT(buf->b_efunc == NULL);
- ASSERT(buf->b_data != NULL);
-
- if (hashed) {
- kmutex_t *hash_lock = HDR_LOCK(hdr);
-
- mutex_enter(hash_lock);
- hdr = buf->b_hdr;
- ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
-
- (void) remove_reference(hdr, hash_lock, tag);
- if (hdr->b_l1hdr.b_datacnt > 1) {
- arc_buf_destroy(buf, TRUE);
- } else {
- ASSERT(buf == hdr->b_l1hdr.b_buf);
- ASSERT(buf->b_efunc == NULL);
- hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE;
- }
- mutex_exit(hash_lock);
- } else if (HDR_IO_IN_PROGRESS(hdr)) {
- int destroy_hdr;
- /*
- * We are in the middle of an async write. Don't destroy
- * this buffer unless the write completes before we finish
- * decrementing the reference count.
- */
- mutex_enter(&arc_user_evicts_lock);
- (void) remove_reference(hdr, NULL, tag);
- ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
- destroy_hdr = !HDR_IO_IN_PROGRESS(hdr);
- mutex_exit(&arc_user_evicts_lock);
- if (destroy_hdr)
- arc_hdr_destroy(hdr);
- } else {
- if (remove_reference(hdr, NULL, tag) > 0)
- arc_buf_destroy(buf, TRUE);
- else
- arc_hdr_destroy(hdr);
- }
-}
-
-boolean_t
-arc_buf_remove_ref(arc_buf_t *buf, void* tag)
+arc_buf_destroy(arc_buf_t *buf, void* tag)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
kmutex_t *hash_lock = HDR_LOCK(hdr);
- boolean_t no_callback = (buf->b_efunc == NULL);
if (hdr->b_l1hdr.b_state == arc_anon) {
- ASSERT(hdr->b_l1hdr.b_datacnt == 1);
- arc_buf_free(buf, tag);
- return (no_callback);
+ ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
+ ASSERT(!HDR_IO_IN_PROGRESS(hdr));
+ VERIFY0(remove_reference(hdr, NULL, tag));
+ arc_hdr_destroy(hdr);
+ return;
}
mutex_enter(hash_lock);
- hdr = buf->b_hdr;
- ASSERT(hdr->b_l1hdr.b_datacnt > 0);
+ ASSERT3P(hdr, ==, buf->b_hdr);
+ ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
- ASSERT(hdr->b_l1hdr.b_state != arc_anon);
- ASSERT(buf->b_data != NULL);
+ ASSERT3P(hdr->b_l1hdr.b_state, !=, arc_anon);
+ ASSERT3P(buf->b_data, !=, NULL);
(void) remove_reference(hdr, hash_lock, tag);
- if (hdr->b_l1hdr.b_datacnt > 1) {
- if (no_callback)
- arc_buf_destroy(buf, TRUE);
- } else if (no_callback) {
- ASSERT(hdr->b_l1hdr.b_buf == buf && buf->b_next == NULL);
- ASSERT(buf->b_efunc == NULL);
- hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE;
- }
- ASSERT(no_callback || hdr->b_l1hdr.b_datacnt > 1 ||
- refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
+ arc_buf_destroy_impl(buf, B_TRUE);
mutex_exit(hash_lock);
- return (no_callback);
}
uint64_t
arc_buf_size(arc_buf_t *buf)
{
- return (buf->b_hdr->b_size);
-}
-
-/*
- * Called from the DMU to determine if the current buffer should be
- * evicted. In order to ensure proper locking, the eviction must be initiated
- * from the DMU. Return true if the buffer is associated with user data and
- * duplicate buffers still exist.
- */
-boolean_t
-arc_buf_eviction_needed(arc_buf_t *buf)
-{
- arc_buf_hdr_t *hdr;
- boolean_t evict_needed = B_FALSE;
-
- if (zfs_disable_dup_eviction)
- return (B_FALSE);
-
- mutex_enter(&buf->b_evict_lock);
- hdr = buf->b_hdr;
- if (hdr == NULL) {
- /*
- * We are in arc_do_user_evicts(); let that function
- * perform the eviction.
- */
- ASSERT(buf->b_data == NULL);
- mutex_exit(&buf->b_evict_lock);
- return (B_FALSE);
- } else if (buf->b_data == NULL) {
- /*
- * We have already been added to the arc eviction list;
- * recommend eviction.
- */
- ASSERT3P(hdr, ==, &arc_eviction_hdr);
- mutex_exit(&buf->b_evict_lock);
- return (B_TRUE);
- }
-
- if (hdr->b_l1hdr.b_datacnt > 1 && HDR_ISTYPE_DATA(hdr))
- evict_needed = B_TRUE;
-
- mutex_exit(&buf->b_evict_lock);
- return (evict_needed);
+ return (HDR_GET_LSIZE(buf->b_hdr));
}
/*
state = hdr->b_l1hdr.b_state;
if (GHOST_STATE(state)) {
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
- ASSERT(hdr->b_l1hdr.b_buf == NULL);
+ ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
/*
* l2arc_write_buffers() relies on a header's L1 portion
- * (i.e. its b_tmp_cdata field) during its write phase.
+ * (i.e. its b_pdata field) during its write phase.
* Thus, we cannot push a header onto the arc_l2c_only
* state (removing its L1 piece) until the header is
* done being written to the l2arc.
}
ARCSTAT_BUMP(arcstat_deleted);
- bytes_evicted += hdr->b_size;
+ bytes_evicted += HDR_GET_LSIZE(hdr);
DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
if (HDR_HAS_L2HDR(hdr)) {
+ ASSERT(hdr->b_l1hdr.b_pdata == NULL);
/*
* This buffer is cached on the 2nd Level ARC;
* don't destroy the header.
hdr = arc_hdr_realloc(hdr, hdr_full_cache,
hdr_l2only_cache);
} else {
+ ASSERT(hdr->b_l1hdr.b_pdata == NULL);
arc_change_state(arc_anon, hdr, hash_lock);
arc_hdr_destroy(hdr);
}
}
ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt));
- ASSERT3U(hdr->b_l1hdr.b_datacnt, >, 0);
while (hdr->b_l1hdr.b_buf) {
arc_buf_t *buf = hdr->b_l1hdr.b_buf;
if (!mutex_tryenter(&buf->b_evict_lock)) {
break;
}
if (buf->b_data != NULL)
- bytes_evicted += hdr->b_size;
- if (buf->b_efunc != NULL) {
- mutex_enter(&arc_user_evicts_lock);
- arc_buf_destroy(buf, FALSE);
- hdr->b_l1hdr.b_buf = buf->b_next;
- buf->b_hdr = &arc_eviction_hdr;
- buf->b_next = arc_eviction_list;
- arc_eviction_list = buf;
- cv_signal(&arc_user_evicts_cv);
- mutex_exit(&arc_user_evicts_lock);
- mutex_exit(&buf->b_evict_lock);
- } else {
- mutex_exit(&buf->b_evict_lock);
- arc_buf_destroy(buf, TRUE);
- }
+ bytes_evicted += HDR_GET_LSIZE(hdr);
+ mutex_exit(&buf->b_evict_lock);
+ arc_buf_destroy_impl(buf, B_TRUE);
}
if (HDR_HAS_L2HDR(hdr)) {
- ARCSTAT_INCR(arcstat_evict_l2_cached, hdr->b_size);
+ ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr));
} else {
- if (l2arc_write_eligible(hdr->b_spa, hdr))
- ARCSTAT_INCR(arcstat_evict_l2_eligible, hdr->b_size);
- else
- ARCSTAT_INCR(arcstat_evict_l2_ineligible, hdr->b_size);
+ if (l2arc_write_eligible(hdr->b_spa, hdr)) {
+ ARCSTAT_INCR(arcstat_evict_l2_eligible,
+ HDR_GET_LSIZE(hdr));
+ } else {
+ ARCSTAT_INCR(arcstat_evict_l2_ineligible,
+ HDR_GET_LSIZE(hdr));
+ }
}
- if (hdr->b_l1hdr.b_datacnt == 0) {
+ if (hdr->b_l1hdr.b_bufcnt == 0) {
+ arc_cksum_free(hdr);
+
+ bytes_evicted += arc_hdr_size(hdr);
+
+ /*
+ * If this hdr is being evicted and has a compressed
+ * buffer then we discard it here before we change states.
+ * This ensures that the accounting is updated correctly
+ * in arc_free_data_buf().
+ */
+ arc_hdr_free_pdata(hdr);
+
arc_change_state(evicted_state, hdr, hash_lock);
ASSERT(HDR_IN_HASH_TABLE(hdr));
- hdr->b_flags |= ARC_FLAG_IN_HASH_TABLE;
- hdr->b_flags &= ~ARC_FLAG_BUF_AVAILABLE;
+ arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr);
}
* Flush all "evictable" data of the given type from the arc state
* specified. This will not evict any "active" buffers (i.e. referenced).
*
- * When 'retry' is set to FALSE, the function will make a single pass
+ * When 'retry' is set to B_FALSE, the function will make a single pass
* over the state and evict any buffers that it can. Since it doesn't
* continually retry the eviction, it might end up leaving some buffers
* in the ARC due to lock misses.
*
- * When 'retry' is set to TRUE, the function will continually retry the
+ * When 'retry' is set to B_TRUE, the function will continually retry the
* eviction until *all* evictable buffers have been removed from the
* state. As a result, if concurrent insertions into the state are
* allowed (e.g. if the ARC isn't shutting down), this function might
{
uint64_t evicted = 0;
- while (state->arcs_lsize[type] != 0) {
+ while (refcount_count(&state->arcs_esize[type]) != 0) {
evicted += arc_evict_state(state, spa, ARC_EVICT_ALL, type);
if (!retry)
{
int64_t delta;
- if (bytes > 0 && state->arcs_lsize[type] > 0) {
- delta = MIN(state->arcs_lsize[type], bytes);
+ if (bytes > 0 && refcount_count(&state->arcs_esize[type]) > 0) {
+ delta = MIN(refcount_count(&state->arcs_esize[type]), bytes);
return (arc_evict_state(state, spa, delta, type));
}
static uint64_t
arc_adjust_meta_balanced(void)
{
- int64_t adjustmnt, delta, prune = 0;
- uint64_t total_evicted = 0;
+ int64_t delta, prune = 0;
+ uint64_t adjustmnt, total_evicted = 0;
arc_buf_contents_t type = ARC_BUFC_DATA;
int restarts = MAX(zfs_arc_meta_adjust_restarts, 0);
*/
adjustmnt = arc_meta_used - arc_meta_limit;
- if (adjustmnt > 0 && arc_mru->arcs_lsize[type] > 0) {
- delta = MIN(arc_mru->arcs_lsize[type], adjustmnt);
+ if (adjustmnt > 0 && refcount_count(&arc_mru->arcs_esize[type]) > 0) {
+ delta = MIN(refcount_count(&arc_mru->arcs_esize[type]),
+ adjustmnt);
total_evicted += arc_adjust_impl(arc_mru, 0, delta, type);
adjustmnt -= delta;
}
* simply decrement the amount of data evicted from the MRU.
*/
- if (adjustmnt > 0 && arc_mfu->arcs_lsize[type] > 0) {
- delta = MIN(arc_mfu->arcs_lsize[type], adjustmnt);
+ if (adjustmnt > 0 && refcount_count(&arc_mfu->arcs_esize[type]) > 0) {
+ delta = MIN(refcount_count(&arc_mfu->arcs_esize[type]),
+ adjustmnt);
total_evicted += arc_adjust_impl(arc_mfu, 0, delta, type);
}
adjustmnt = arc_meta_used - arc_meta_limit;
- if (adjustmnt > 0 && arc_mru_ghost->arcs_lsize[type] > 0) {
+ if (adjustmnt > 0 &&
+ refcount_count(&arc_mru_ghost->arcs_esize[type]) > 0) {
delta = MIN(adjustmnt,
- arc_mru_ghost->arcs_lsize[type]);
+ refcount_count(&arc_mru_ghost->arcs_esize[type]));
total_evicted += arc_adjust_impl(arc_mru_ghost, 0, delta, type);
adjustmnt -= delta;
}
- if (adjustmnt > 0 && arc_mfu_ghost->arcs_lsize[type] > 0) {
+ if (adjustmnt > 0 &&
+ refcount_count(&arc_mfu_ghost->arcs_esize[type]) > 0) {
delta = MIN(adjustmnt,
- arc_mfu_ghost->arcs_lsize[type]);
+ refcount_count(&arc_mfu_ghost->arcs_esize[type]));
total_evicted += arc_adjust_impl(arc_mfu_ghost, 0, delta, type);
}
return (total_evicted);
}
-static void
-arc_do_user_evicts(void)
-{
- mutex_enter(&arc_user_evicts_lock);
- while (arc_eviction_list != NULL) {
- arc_buf_t *buf = arc_eviction_list;
- arc_eviction_list = buf->b_next;
- mutex_enter(&buf->b_evict_lock);
- buf->b_hdr = NULL;
- mutex_exit(&buf->b_evict_lock);
- mutex_exit(&arc_user_evicts_lock);
-
- if (buf->b_efunc != NULL)
- VERIFY0(buf->b_efunc(buf->b_private));
-
- buf->b_efunc = NULL;
- buf->b_private = NULL;
- kmem_cache_free(buf_cache, buf);
- mutex_enter(&arc_user_evicts_lock);
- }
- mutex_exit(&arc_user_evicts_lock);
-}
-
void
arc_flush(spa_t *spa, boolean_t retry)
{
uint64_t guid = 0;
/*
- * If retry is TRUE, a spa must not be specified since we have
+ * If retry is B_TRUE, a spa must not be specified since we have
* no good way to determine if all of a spa's buffers have been
* evicted from an arc state.
*/
(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry);
(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry);
-
- arc_do_user_evicts();
- ASSERT(spa || arc_eviction_list == NULL);
}
void
/*
* If zio data pages are being allocated out of a separate heap segment,
* then enforce that the size of available vmem for this arena remains
- * above about 1/16th free.
+ * above about 1/4th (1/(2^arc_zio_arena_free_shift)) free.
*
- * Note: The 1/16th arena free requirement was put in place
- * to aggressively evict memory from the arc in order to avoid
- * memory fragmentation issues.
+ * Note that reducing the arc_zio_arena_free_shift keeps more virtual
+ * memory (in the zio_arena) free, which can avoid memory
+ * fragmentation issues.
*/
if (zio_arena != NULL) {
- n = vmem_size(zio_arena, VMEM_FREE) -
- (vmem_size(zio_arena, VMEM_ALLOC) >> 4);
+ n = vmem_size(zio_arena, VMEM_FREE) - (vmem_size(zio_arena,
+ VMEM_ALLOC) >> arc_zio_arena_free_shift);
if (n < lowest) {
lowest = n;
r = FMR_ZIO_ARENA;
/*
* Determine if the system is under memory pressure and is asking
- * to reclaim memory. A return value of TRUE indicates that the system
+ * to reclaim memory. A return value of B_TRUE indicates that the system
* is under memory pressure and that the arc should adjust accordingly.
*/
static boolean_t
arc_tuning_update();
+ /*
+ * This is necessary in order for the mdb ::arc dcmd to
+ * show up to date information. Since the ::arc command
+ * does not call the kstat's update function, without
+ * this call, the command may show stale stats for the
+ * anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even
+ * with this change, the data might be up to 1 second
+ * out of date; but that should suffice. The arc_state_t
+ * structures can be queried directly if more accurate
+ * information is needed.
+ */
+#ifndef __linux__
+ if (arc_ksp != NULL)
+ arc_ksp->ks_update(arc_ksp, KSTAT_READ);
+#endif
mutex_exit(&arc_reclaim_lock);
if (free_memory < 0) {
}
}
- arc_reclaim_thread_exit = FALSE;
+ arc_reclaim_thread_exit = B_FALSE;
cv_broadcast(&arc_reclaim_thread_cv);
CALLB_CPR_EXIT(&cpr); /* drops arc_reclaim_lock */
spl_fstrans_unmark(cookie);
thread_exit();
}
-static void
-arc_user_evicts_thread(void)
-{
- fstrans_cookie_t cookie = spl_fstrans_mark();
- callb_cpr_t cpr;
-
- CALLB_CPR_INIT(&cpr, &arc_user_evicts_lock, callb_generic_cpr, FTAG);
-
- mutex_enter(&arc_user_evicts_lock);
- while (!arc_user_evicts_thread_exit) {
- mutex_exit(&arc_user_evicts_lock);
-
- arc_do_user_evicts();
-
- /*
- * This is necessary in order for the mdb ::arc dcmd to
- * show up to date information. Since the ::arc command
- * does not call the kstat's update function, without
- * this call, the command may show stale stats for the
- * anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even
- * with this change, the data might be up to 1 second
- * out of date; but that should suffice. The arc_state_t
- * structures can be queried directly if more accurate
- * information is needed.
- */
- if (arc_ksp != NULL)
- arc_ksp->ks_update(arc_ksp, KSTAT_READ);
-
- mutex_enter(&arc_user_evicts_lock);
-
- /*
- * Block until signaled, or after one second (we need to
- * call the arc's kstat update function regularly).
- */
- CALLB_CPR_SAFE_BEGIN(&cpr);
- (void) cv_timedwait_sig(&arc_user_evicts_cv,
- &arc_user_evicts_lock, ddi_get_lbolt() + hz);
- CALLB_CPR_SAFE_END(&cpr, &arc_user_evicts_lock);
- }
-
- arc_user_evicts_thread_exit = FALSE;
- cv_broadcast(&arc_user_evicts_cv);
- CALLB_CPR_EXIT(&cpr); /* drops arc_user_evicts_lock */
- spl_fstrans_unmark(cookie);
- thread_exit();
-}
-
#ifdef _KERNEL
/*
* Determine the amount of memory eligible for eviction contained in the
static uint64_t
arc_evictable_memory(void) {
uint64_t arc_clean =
- arc_mru->arcs_lsize[ARC_BUFC_DATA] +
- arc_mru->arcs_lsize[ARC_BUFC_METADATA] +
- arc_mfu->arcs_lsize[ARC_BUFC_DATA] +
- arc_mfu->arcs_lsize[ARC_BUFC_METADATA];
+ refcount_count(&arc_mru->arcs_esize[ARC_BUFC_DATA]) +
+ refcount_count(&arc_mru->arcs_esize[ARC_BUFC_METADATA]) +
+ refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_DATA]) +
+ refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
uint64_t ghost_clean =
- arc_mru_ghost->arcs_lsize[ARC_BUFC_DATA] +
- arc_mru_ghost->arcs_lsize[ARC_BUFC_METADATA] +
- arc_mfu_ghost->arcs_lsize[ARC_BUFC_DATA] +
- arc_mfu_ghost->arcs_lsize[ARC_BUFC_METADATA];
+ refcount_count(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]) +
+ refcount_count(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]) +
+ refcount_count(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]) +
+ refcount_count(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
uint64_t arc_dirty = MAX((int64_t)arc_size - (int64_t)arc_clean, 0);
if (arc_dirty >= arc_c_min)
}
/*
- * The buffer, supplied as the first argument, needs a data block. If we
- * are hitting the hard limit for the cache size, we must sleep, waiting
- * for the eviction thread to catch up. If we're past the target size
- * but below the hard limit, we'll only signal the reclaim thread and
- * continue on.
+ * Allocate a block and return it to the caller. If we are hitting the
+ * hard limit for the cache size, we must sleep, waiting for the eviction
+ * thread to catch up. If we're past the target size but below the hard
+ * limit, we'll only signal the reclaim thread and continue on.
*/
-static void
-arc_get_data_buf(arc_buf_t *buf)
+static void *
+arc_get_data_buf(arc_buf_hdr_t *hdr, uint64_t size, void *tag)
{
- arc_state_t *state = buf->b_hdr->b_l1hdr.b_state;
- uint64_t size = buf->b_hdr->b_size;
- arc_buf_contents_t type = arc_buf_type(buf->b_hdr);
+ void *datap = NULL;
+ arc_state_t *state = hdr->b_l1hdr.b_state;
+ arc_buf_contents_t type = arc_buf_type(hdr);
arc_adapt(size, state);
mutex_exit(&arc_reclaim_lock);
}
+ VERIFY3U(hdr->b_type, ==, type);
if (type == ARC_BUFC_METADATA) {
- buf->b_data = zio_buf_alloc(size);
+ datap = zio_buf_alloc(size);
arc_space_consume(size, ARC_SPACE_META);
} else {
ASSERT(type == ARC_BUFC_DATA);
- buf->b_data = zio_data_buf_alloc(size);
+ datap = zio_data_buf_alloc(size);
arc_space_consume(size, ARC_SPACE_DATA);
}
* Update the state size. Note that ghost states have a
* "ghost size" and so don't need to be updated.
*/
- if (!GHOST_STATE(buf->b_hdr->b_l1hdr.b_state)) {
- arc_buf_hdr_t *hdr = buf->b_hdr;
- arc_state_t *state = hdr->b_l1hdr.b_state;
+ if (!GHOST_STATE(state)) {
- (void) refcount_add_many(&state->arcs_size, size, buf);
+ (void) refcount_add_many(&state->arcs_size, size, tag);
/*
* If this is reached via arc_read, the link is
*/
if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
- atomic_add_64(&hdr->b_l1hdr.b_state->arcs_lsize[type],
- size);
+ (void) refcount_add_many(&state->arcs_esize[type],
+ size, tag);
}
+
/*
* If we are growing the cache, and we are adding anonymous
* data, and we have outgrown arc_p, update arc_p
refcount_count(&arc_mru->arcs_size) > arc_p))
arc_p = MIN(arc_c, arc_p + size);
}
+ return (datap);
+}
+
+/*
+ * Free the arc data buffer.
+ */
+static void
+arc_free_data_buf(arc_buf_hdr_t *hdr, void *data, uint64_t size, void *tag)
+{
+ arc_state_t *state = hdr->b_l1hdr.b_state;
+ arc_buf_contents_t type = arc_buf_type(hdr);
+
+ /* protected by hash lock, if in the hash table */
+ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
+ ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
+ ASSERT(state != arc_anon && state != arc_l2c_only);
+
+ (void) refcount_remove_many(&state->arcs_esize[type],
+ size, tag);
+ }
+ (void) refcount_remove_many(&state->arcs_size, size, tag);
+
+ VERIFY3U(hdr->b_type, ==, type);
+ if (type == ARC_BUFC_METADATA) {
+ zio_buf_free(data, size);
+ arc_space_return(size, ARC_SPACE_META);
+ } else {
+ ASSERT(type == ARC_BUFC_DATA);
+ zio_data_buf_free(data, size);
+ arc_space_return(size, ARC_SPACE_DATA);
+ }
}
/*
ASSERT(multilist_link_active(
&hdr->b_l1hdr.b_arc_node));
} else {
- hdr->b_flags &= ~ARC_FLAG_PREFETCH;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH);
atomic_inc_32(&hdr->b_l1hdr.b_mru_hits);
ARCSTAT_BUMP(arcstat_mru_hits);
}
if (HDR_PREFETCH(hdr)) {
new_state = arc_mru;
if (refcount_count(&hdr->b_l1hdr.b_refcnt) > 0)
- hdr->b_flags &= ~ARC_FLAG_PREFETCH;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH);
DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
} else {
new_state = arc_mfu;
arc_bcopy_func(zio_t *zio, arc_buf_t *buf, void *arg)
{
if (zio == NULL || zio->io_error == 0)
- bcopy(buf->b_data, arg, buf->b_hdr->b_size);
- VERIFY(arc_buf_remove_ref(buf, arg));
+ bcopy(buf->b_data, arg, HDR_GET_LSIZE(buf->b_hdr));
+ arc_buf_destroy(buf, arg);
}
/* a generic arc_done_func_t */
{
arc_buf_t **bufp = arg;
if (zio && zio->io_error) {
- VERIFY(arc_buf_remove_ref(buf, arg));
+ arc_buf_destroy(buf, arg);
*bufp = NULL;
} else {
*bufp = buf;
}
}
+static void
+arc_hdr_verify(arc_buf_hdr_t *hdr, blkptr_t *bp)
+{
+ if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) {
+ ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0);
+ ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
+ } else {
+ if (HDR_COMPRESSION_ENABLED(hdr)) {
+ ASSERT3U(HDR_GET_COMPRESS(hdr), ==,
+ BP_GET_COMPRESS(bp));
+ }
+ ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
+ ASSERT3U(HDR_GET_PSIZE(hdr), ==, BP_GET_PSIZE(bp));
+ }
+}
+
static void
arc_read_done(zio_t *zio)
{
- arc_buf_hdr_t *hdr;
- arc_buf_t *buf;
- arc_buf_t *abuf; /* buffer we're assigning to callback */
+ arc_buf_hdr_t *hdr = zio->io_private;
+ arc_buf_t *abuf = NULL; /* buffer we're assigning to callback */
kmutex_t *hash_lock = NULL;
arc_callback_t *callback_list, *acb;
- int freeable = FALSE;
-
- buf = zio->io_private;
- hdr = buf->b_hdr;
+ int freeable = B_FALSE;
/*
* The hdr was inserted into hash-table and removed from lists
ASSERT3U(hdr->b_dva.dva_word[1], ==,
BP_IDENTITY(zio->io_bp)->dva_word[1]);
- found = buf_hash_find(hdr->b_spa, zio->io_bp,
- &hash_lock);
+ found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock);
- ASSERT((found == NULL && HDR_FREED_IN_READ(hdr) &&
- hash_lock == NULL) ||
- (found == hdr &&
+ ASSERT((found == hdr &&
DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) ||
(found == hdr && HDR_L2_READING(hdr)));
+ ASSERT3P(hash_lock, !=, NULL);
+ }
+
+ if (zio->io_error == 0) {
+ /* byteswap if necessary */
+ if (BP_SHOULD_BYTESWAP(zio->io_bp)) {
+ if (BP_GET_LEVEL(zio->io_bp) > 0) {
+ hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
+ } else {
+ hdr->b_l1hdr.b_byteswap =
+ DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp));
+ }
+ } else {
+ hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
+ }
}
- hdr->b_flags &= ~ARC_FLAG_L2_EVICTED;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED);
if (l2arc_noprefetch && HDR_PREFETCH(hdr))
- hdr->b_flags &= ~ARC_FLAG_L2CACHE;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE);
- /* byteswap if necessary */
callback_list = hdr->b_l1hdr.b_acb;
- ASSERT(callback_list != NULL);
- if (BP_SHOULD_BYTESWAP(zio->io_bp) && zio->io_error == 0) {
- dmu_object_byteswap_t bswap =
- DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp));
- if (BP_GET_LEVEL(zio->io_bp) > 0)
- byteswap_uint64_array(buf->b_data, hdr->b_size);
- else
- dmu_ot_byteswap[bswap].ob_func(buf->b_data, hdr->b_size);
- }
-
- arc_cksum_compute(buf, B_FALSE);
- arc_buf_watch(buf);
+ ASSERT3P(callback_list, !=, NULL);
if (hash_lock && zio->io_error == 0 &&
hdr->b_l1hdr.b_state == arc_anon) {
}
/* create copies of the data buffer for the callers */
- abuf = buf;
for (acb = callback_list; acb; acb = acb->acb_next) {
- if (acb->acb_done) {
+ if (acb->acb_done != NULL) {
+ /*
+ * If we're here, then this must be a demand read
+ * since prefetch requests don't have callbacks.
+ * If a read request has a callback (i.e. acb_done is
+ * not NULL), then we decompress the data for the
+ * first request and clone the rest. This avoids
+ * having to waste cpu resources decompressing data
+ * that nobody is explicitly waiting to read.
+ */
if (abuf == NULL) {
- ARCSTAT_BUMP(arcstat_duplicate_reads);
- abuf = arc_buf_clone(buf);
+ acb->acb_buf = arc_buf_alloc_impl(hdr,
+ acb->acb_private);
+ if (zio->io_error == 0) {
+ zio->io_error =
+ arc_decompress(acb->acb_buf);
+ }
+ abuf = acb->acb_buf;
+ } else {
+ add_reference(hdr, acb->acb_private);
+ acb->acb_buf = arc_buf_clone(abuf);
}
- acb->acb_buf = abuf;
- abuf = NULL;
}
}
hdr->b_l1hdr.b_acb = NULL;
- hdr->b_flags &= ~ARC_FLAG_IO_IN_PROGRESS;
- ASSERT(!HDR_BUF_AVAILABLE(hdr));
- if (abuf == buf) {
- ASSERT(buf->b_efunc == NULL);
- ASSERT(hdr->b_l1hdr.b_datacnt == 1);
- hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
+ if (abuf == NULL) {
+ /*
+ * This buffer didn't have a callback so it must
+ * be a prefetch.
+ */
+ ASSERT(HDR_PREFETCH(hdr));
+ ASSERT0(hdr->b_l1hdr.b_bufcnt);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
}
ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt) ||
callback_list != NULL);
- if (zio->io_error != 0) {
- hdr->b_flags |= ARC_FLAG_IO_ERROR;
+ if (zio->io_error == 0) {
+ arc_hdr_verify(hdr, zio->io_bp);
+ } else {
+ arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
if (hdr->b_l1hdr.b_state != arc_anon)
arc_change_state(arc_anon, hdr, hash_lock);
if (HDR_IN_HASH_TABLE(hdr))
arc_flags_t *arc_flags, const zbookmark_phys_t *zb)
{
arc_buf_hdr_t *hdr = NULL;
- arc_buf_t *buf = NULL;
kmutex_t *hash_lock = NULL;
zio_t *rzio;
uint64_t guid = spa_load_guid(spa);
hdr = buf_hash_find(guid, bp, &hash_lock);
}
- if (hdr != NULL && HDR_HAS_L1HDR(hdr) && hdr->b_l1hdr.b_datacnt > 0) {
-
+ if (hdr != NULL && HDR_HAS_L1HDR(hdr) && hdr->b_l1hdr.b_pdata != NULL) {
+ arc_buf_t *buf = NULL;
*arc_flags |= ARC_FLAG_CACHED;
if (HDR_IO_IN_PROGRESS(hdr)) {
ARCSTAT_BUMP(arcstat_sync_wait_for_async);
}
if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) {
- hdr->b_flags &= ~ARC_FLAG_PREDICTIVE_PREFETCH;
+ arc_hdr_clear_flags(hdr,
+ ARC_FLAG_PREDICTIVE_PREFETCH);
}
if (*arc_flags & ARC_FLAG_WAIT) {
acb->acb_zio_dummy = zio_null(pio,
spa, NULL, NULL, NULL, zio_flags);
- ASSERT(acb->acb_done != NULL);
+ ASSERT3P(acb->acb_done, !=, NULL);
acb->acb_next = hdr->b_l1hdr.b_acb;
hdr->b_l1hdr.b_acb = acb;
- add_reference(hdr, hash_lock, private);
mutex_exit(hash_lock);
goto out;
}
arc_buf_hdr_t *, hdr);
ARCSTAT_BUMP(
arcstat_demand_hit_predictive_prefetch);
- hdr->b_flags &= ~ARC_FLAG_PREDICTIVE_PREFETCH;
+ arc_hdr_clear_flags(hdr,
+ ARC_FLAG_PREDICTIVE_PREFETCH);
}
- add_reference(hdr, hash_lock, private);
+ ASSERT(!BP_IS_EMBEDDED(bp) || !BP_IS_HOLE(bp));
+
/*
* If this block is already in use, create a new
* copy of the data so that we will be guaranteed
* that arc_release() will always succeed.
*/
buf = hdr->b_l1hdr.b_buf;
- ASSERT(buf);
- ASSERT(buf->b_data);
- if (HDR_BUF_AVAILABLE(hdr)) {
- ASSERT(buf->b_efunc == NULL);
- hdr->b_flags &= ~ARC_FLAG_BUF_AVAILABLE;
+ if (buf == NULL) {
+ ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt));
+ ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
+ buf = arc_buf_alloc_impl(hdr, private);
+ VERIFY0(arc_decompress(buf));
} else {
+ add_reference(hdr, private);
buf = arc_buf_clone(buf);
}
+ ASSERT3P(buf->b_data, !=, NULL);
} else if (*arc_flags & ARC_FLAG_PREFETCH &&
refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) {
- hdr->b_flags |= ARC_FLAG_PREFETCH;
+ arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
}
DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
arc_access(hdr, hash_lock);
if (*arc_flags & ARC_FLAG_L2CACHE)
- hdr->b_flags |= ARC_FLAG_L2CACHE;
- if (*arc_flags & ARC_FLAG_L2COMPRESS)
- hdr->b_flags |= ARC_FLAG_L2COMPRESS;
+ arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
mutex_exit(hash_lock);
ARCSTAT_BUMP(arcstat_hits);
ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
if (done)
done(NULL, buf, private);
} else {
- uint64_t size = BP_GET_LSIZE(bp);
+ uint64_t lsize = BP_GET_LSIZE(bp);
+ uint64_t psize = BP_GET_PSIZE(bp);
arc_callback_t *acb;
vdev_t *vd = NULL;
uint64_t addr = 0;
boolean_t devw = B_FALSE;
- enum zio_compress b_compress = ZIO_COMPRESS_OFF;
- int32_t b_asize = 0;
+ uint64_t size;
/*
* Gracefully handle a damaged logical block size as a
* checksum error.
*/
- if (size > spa_maxblocksize(spa)) {
- ASSERT3P(buf, ==, NULL);
+ if (lsize > spa_maxblocksize(spa)) {
rc = SET_ERROR(ECKSUM);
goto out;
}
/* this block is not in the cache */
arc_buf_hdr_t *exists = NULL;
arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp);
- buf = arc_buf_alloc(spa, size, private, type);
- hdr = buf->b_hdr;
+ hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
+ BP_GET_COMPRESS(bp), type);
+
if (!BP_IS_EMBEDDED(bp)) {
hdr->b_dva = *BP_IDENTITY(bp);
hdr->b_birth = BP_PHYSICAL_BIRTH(bp);
/* somebody beat us to the hash insert */
mutex_exit(hash_lock);
buf_discard_identity(hdr);
- (void) arc_buf_remove_ref(buf, private);
+ arc_hdr_destroy(hdr);
goto top; /* restart the IO request */
}
-
- /*
- * If there is a callback, we pass our reference to
- * it; otherwise we remove our reference.
- */
- if (done == NULL) {
- (void) remove_reference(hdr, hash_lock,
- private);
- }
- if (*arc_flags & ARC_FLAG_PREFETCH)
- hdr->b_flags |= ARC_FLAG_PREFETCH;
- if (*arc_flags & ARC_FLAG_L2CACHE)
- hdr->b_flags |= ARC_FLAG_L2CACHE;
- if (*arc_flags & ARC_FLAG_L2COMPRESS)
- hdr->b_flags |= ARC_FLAG_L2COMPRESS;
- if (BP_GET_LEVEL(bp) > 0)
- hdr->b_flags |= ARC_FLAG_INDIRECT;
} else {
/*
* This block is in the ghost cache. If it was L2-only
hdr_full_cache);
}
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
ASSERT(GHOST_STATE(hdr->b_l1hdr.b_state));
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
/*
- * If there is a callback, we pass a reference to it.
+ * This is a delicate dance that we play here.
+ * This hdr is in the ghost list so we access it
+ * to move it out of the ghost list before we
+ * initiate the read. If it's a prefetch then
+ * it won't have a callback so we'll remove the
+ * reference that arc_buf_alloc_impl() created. We
+ * do this after we've called arc_access() to
+ * avoid hitting an assert in remove_reference().
*/
- if (done != NULL)
- add_reference(hdr, hash_lock, private);
- if (*arc_flags & ARC_FLAG_PREFETCH)
- hdr->b_flags |= ARC_FLAG_PREFETCH;
- if (*arc_flags & ARC_FLAG_L2CACHE)
- hdr->b_flags |= ARC_FLAG_L2CACHE;
- if (*arc_flags & ARC_FLAG_L2COMPRESS)
- hdr->b_flags |= ARC_FLAG_L2COMPRESS;
- buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
- buf->b_hdr = hdr;
- buf->b_data = NULL;
- buf->b_efunc = NULL;
- buf->b_private = NULL;
- buf->b_next = NULL;
- hdr->b_l1hdr.b_buf = buf;
- ASSERT0(hdr->b_l1hdr.b_datacnt);
- hdr->b_l1hdr.b_datacnt = 1;
- arc_get_data_buf(buf);
arc_access(hdr, hash_lock);
+ arc_hdr_alloc_pdata(hdr);
+ }
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+ size = arc_hdr_size(hdr);
+
+ /*
+ * If compression is enabled on the hdr, then will do
+ * RAW I/O and will store the compressed data in the hdr's
+ * data block. Otherwise, the hdr's data block will contain
+ * the uncompressed data.
+ */
+ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF) {
+ zio_flags |= ZIO_FLAG_RAW;
}
+ if (*arc_flags & ARC_FLAG_PREFETCH)
+ arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
+ if (*arc_flags & ARC_FLAG_L2CACHE)
+ arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
+ if (BP_GET_LEVEL(bp) > 0)
+ arc_hdr_set_flags(hdr, ARC_FLAG_INDIRECT);
if (*arc_flags & ARC_FLAG_PREDICTIVE_PREFETCH)
- hdr->b_flags |= ARC_FLAG_PREDICTIVE_PREFETCH;
+ arc_hdr_set_flags(hdr, ARC_FLAG_PREDICTIVE_PREFETCH);
ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state));
acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP);
acb->acb_done = done;
acb->acb_private = private;
- ASSERT(hdr->b_l1hdr.b_acb == NULL);
+ ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
hdr->b_l1hdr.b_acb = acb;
- hdr->b_flags |= ARC_FLAG_IO_IN_PROGRESS;
+ arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
if (HDR_HAS_L2HDR(hdr) &&
(vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) {
devw = hdr->b_l2hdr.b_dev->l2ad_writing;
addr = hdr->b_l2hdr.b_daddr;
- b_compress = hdr->b_l2hdr.b_compress;
- b_asize = hdr->b_l2hdr.b_asize;
/*
* Lock out device removal.
*/
vd = NULL;
}
+ if (priority == ZIO_PRIORITY_ASYNC_READ)
+ arc_hdr_set_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);
+ else
+ arc_hdr_clear_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);
+
if (hash_lock != NULL)
mutex_exit(hash_lock);
* At this point, we have a level 1 cache miss. Try again in
* L2ARC if possible.
*/
- ASSERT3U(hdr->b_size, ==, size);
+ ASSERT3U(HDR_GET_LSIZE(hdr), ==, lsize);
+
DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr, blkptr_t *, bp,
- uint64_t, size, zbookmark_phys_t *, zb);
+ uint64_t, lsize, zbookmark_phys_t *, zb);
ARCSTAT_BUMP(arcstat_misses);
ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
demand, prefetch, !HDR_ISTYPE_METADATA(hdr),
data, metadata, misses);
- if (priority == ZIO_PRIORITY_ASYNC_READ)
- hdr->b_flags |= ARC_FLAG_PRIO_ASYNC_READ;
- else
- hdr->b_flags &= ~ARC_FLAG_PRIO_ASYNC_READ;
-
if (vd != NULL && l2arc_ndev != 0 && !(l2arc_norw && devw)) {
/*
* Read from the L2ARC if the following are true:
cb = kmem_zalloc(sizeof (l2arc_read_callback_t),
KM_SLEEP);
- cb->l2rcb_buf = buf;
- cb->l2rcb_spa = spa;
+ cb->l2rcb_hdr = hdr;
cb->l2rcb_bp = *bp;
cb->l2rcb_zb = *zb;
cb->l2rcb_flags = zio_flags;
- cb->l2rcb_compress = b_compress;
ASSERT(addr >= VDEV_LABEL_START_SIZE &&
- addr + size < vd->vdev_psize -
+ addr + lsize < vd->vdev_psize -
VDEV_LABEL_END_SIZE);
/*
* Issue a null zio if the underlying buffer
* was squashed to zero size by compression.
*/
- if (b_compress == ZIO_COMPRESS_EMPTY) {
- rzio = zio_null(pio, spa, vd,
- l2arc_read_done, cb,
- zio_flags | ZIO_FLAG_DONT_CACHE |
- ZIO_FLAG_CANFAIL |
- ZIO_FLAG_DONT_PROPAGATE |
- ZIO_FLAG_DONT_RETRY);
- } else {
- rzio = zio_read_phys(pio, vd, addr,
- b_asize, buf->b_data,
- ZIO_CHECKSUM_OFF,
- l2arc_read_done, cb, priority,
- zio_flags | ZIO_FLAG_DONT_CACHE |
- ZIO_FLAG_CANFAIL |
- ZIO_FLAG_DONT_PROPAGATE |
- ZIO_FLAG_DONT_RETRY, B_FALSE);
- }
+ ASSERT3U(HDR_GET_COMPRESS(hdr), !=,
+ ZIO_COMPRESS_EMPTY);
+ rzio = zio_read_phys(pio, vd, addr,
+ size, hdr->b_l1hdr.b_pdata,
+ ZIO_CHECKSUM_OFF,
+ l2arc_read_done, cb, priority,
+ zio_flags | ZIO_FLAG_DONT_CACHE |
+ ZIO_FLAG_CANFAIL |
+ ZIO_FLAG_DONT_PROPAGATE |
+ ZIO_FLAG_DONT_RETRY, B_FALSE);
+
DTRACE_PROBE2(l2arc__read, vdev_t *, vd,
zio_t *, rzio);
- ARCSTAT_INCR(arcstat_l2_read_bytes, b_asize);
+ ARCSTAT_INCR(arcstat_l2_read_bytes, size);
if (*arc_flags & ARC_FLAG_NOWAIT) {
zio_nowait(rzio);
}
}
- rzio = zio_read(pio, spa, bp, buf->b_data, size,
- arc_read_done, buf, priority, zio_flags, zb);
+ rzio = zio_read(pio, spa, bp, hdr->b_l1hdr.b_pdata, size,
+ arc_read_done, hdr, priority, zio_flags, zb);
if (*arc_flags & ARC_FLAG_WAIT) {
rc = zio_wait(rzio);
kmem_free(p, sizeof (*p));
}
-void
-arc_set_callback(arc_buf_t *buf, arc_evict_func_t *func, void *private)
-{
- ASSERT(buf->b_hdr != NULL);
- ASSERT(buf->b_hdr->b_l1hdr.b_state != arc_anon);
- ASSERT(!refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt) ||
- func == NULL);
- ASSERT(buf->b_efunc == NULL);
- ASSERT(!HDR_BUF_AVAILABLE(buf->b_hdr));
-
- buf->b_efunc = func;
- buf->b_private = private;
-}
-
/*
* Notify the arc that a block was freed, and thus will never be used again.
*/
hdr = buf_hash_find(guid, bp, &hash_lock);
if (hdr == NULL)
return;
- if (HDR_BUF_AVAILABLE(hdr)) {
- arc_buf_t *buf = hdr->b_l1hdr.b_buf;
- add_reference(hdr, hash_lock, FTAG);
- hdr->b_flags &= ~ARC_FLAG_BUF_AVAILABLE;
- mutex_exit(hash_lock);
- arc_release(buf, FTAG);
- (void) arc_buf_remove_ref(buf, FTAG);
- } else {
+ /*
+ * We might be trying to free a block that is still doing I/O
+ * (i.e. prefetch) or has a reference (i.e. a dedup-ed,
+ * dmu_sync-ed block). If this block is being prefetched, then it
+ * would still have the ARC_FLAG_IO_IN_PROGRESS flag set on the hdr
+ * until the I/O completes. A block may also have a reference if it is
+ * part of a dedup-ed, dmu_synced write. The dmu_sync() function would
+ * have written the new block to its final resting place on disk but
+ * without the dedup flag set. This would have left the hdr in the MRU
+ * state and discoverable. When the txg finally syncs it detects that
+ * the block was overridden in open context and issues an override I/O.
+ * Since this is a dedup block, the override I/O will determine if the
+ * block is already in the DDT. If so, then it will replace the io_bp
+ * with the bp from the DDT and allow the I/O to finish. When the I/O
+ * reaches the done callback, dbuf_write_override_done, it will
+ * check to see if the io_bp and io_bp_override are identical.
+ * If they are not, then it indicates that the bp was replaced with
+ * the bp in the DDT and the override bp is freed. This allows
+ * us to arrive here with a reference on a block that is being
+ * freed. So if we have an I/O in progress, or a reference to
+ * this hdr, then we don't destroy the hdr.
+ */
+ if (!HDR_HAS_L1HDR(hdr) || (!HDR_IO_IN_PROGRESS(hdr) &&
+ refcount_is_zero(&hdr->b_l1hdr.b_refcnt))) {
+ arc_change_state(arc_anon, hdr, hash_lock);
+ arc_hdr_destroy(hdr);
mutex_exit(hash_lock);
- }
-
-}
-
-/*
- * Clear the user eviction callback set by arc_set_callback(), first calling
- * it if it exists. Because the presence of a callback keeps an arc_buf cached
- * clearing the callback may result in the arc_buf being destroyed. However,
- * it will not result in the *last* arc_buf being destroyed, hence the data
- * will remain cached in the ARC. We make a copy of the arc buffer here so
- * that we can process the callback without holding any locks.
- *
- * It's possible that the callback is already in the process of being cleared
- * by another thread. In this case we can not clear the callback.
- *
- * Returns B_TRUE if the callback was successfully called and cleared.
- */
-boolean_t
-arc_clear_callback(arc_buf_t *buf)
-{
- arc_buf_hdr_t *hdr;
- kmutex_t *hash_lock;
- arc_evict_func_t *efunc = buf->b_efunc;
- void *private = buf->b_private;
-
- mutex_enter(&buf->b_evict_lock);
- hdr = buf->b_hdr;
- if (hdr == NULL) {
- /*
- * We are in arc_do_user_evicts().
- */
- ASSERT(buf->b_data == NULL);
- mutex_exit(&buf->b_evict_lock);
- return (B_FALSE);
- } else if (buf->b_data == NULL) {
- /*
- * We are on the eviction list; process this buffer now
- * but let arc_do_user_evicts() do the reaping.
- */
- buf->b_efunc = NULL;
- mutex_exit(&buf->b_evict_lock);
- VERIFY0(efunc(private));
- return (B_TRUE);
- }
- hash_lock = HDR_LOCK(hdr);
- mutex_enter(hash_lock);
- hdr = buf->b_hdr;
- ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
-
- ASSERT3U(refcount_count(&hdr->b_l1hdr.b_refcnt), <,
- hdr->b_l1hdr.b_datacnt);
- ASSERT(hdr->b_l1hdr.b_state == arc_mru ||
- hdr->b_l1hdr.b_state == arc_mfu);
-
- buf->b_efunc = NULL;
- buf->b_private = NULL;
-
- if (hdr->b_l1hdr.b_datacnt > 1) {
- mutex_exit(&buf->b_evict_lock);
- arc_buf_destroy(buf, TRUE);
} else {
- ASSERT(buf == hdr->b_l1hdr.b_buf);
- hdr->b_flags |= ARC_FLAG_BUF_AVAILABLE;
- mutex_exit(&buf->b_evict_lock);
+ mutex_exit(hash_lock);
}
- mutex_exit(hash_lock);
- VERIFY0(efunc(private));
- return (B_TRUE);
}
/*
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT(!HDR_IN_HASH_TABLE(hdr));
ASSERT(!HDR_HAS_L2HDR(hdr));
- ASSERT(BUF_EMPTY(hdr));
+ ASSERT(HDR_EMPTY(hdr));
- ASSERT3U(hdr->b_l1hdr.b_datacnt, ==, 1);
+ ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
ASSERT3S(refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1);
ASSERT(!list_link_active(&hdr->b_l1hdr.b_arc_node));
- ASSERT3P(buf->b_efunc, ==, NULL);
- ASSERT3P(buf->b_private, ==, NULL);
-
hdr->b_l1hdr.b_arc_access = 0;
+
+ /*
+ * If the buf is being overridden then it may already
+ * have a hdr that is not empty.
+ */
+ buf_discard_identity(hdr);
arc_buf_thaw(buf);
return;
/*
* Do we have more than one buf?
*/
- if (hdr->b_l1hdr.b_datacnt > 1) {
+ if (hdr->b_l1hdr.b_bufcnt > 1) {
arc_buf_hdr_t *nhdr;
arc_buf_t **bufp;
- uint64_t blksz = hdr->b_size;
uint64_t spa = hdr->b_spa;
+ uint64_t psize = HDR_GET_PSIZE(hdr);
+ uint64_t lsize = HDR_GET_LSIZE(hdr);
+ enum zio_compress compress = HDR_GET_COMPRESS(hdr);
arc_buf_contents_t type = arc_buf_type(hdr);
- uint32_t flags = hdr->b_flags;
+ arc_buf_t *lastbuf = NULL;
+ VERIFY3U(hdr->b_type, ==, type);
ASSERT(hdr->b_l1hdr.b_buf != buf || buf->b_next != NULL);
+ (void) remove_reference(hdr, hash_lock, tag);
+
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(HDR_SHARED_DATA(hdr));
+ ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
+ ASSERT(ARC_BUF_LAST(buf));
+ }
+
/*
* Pull the data off of this hdr and attach it to
- * a new anonymous hdr.
+ * a new anonymous hdr. Also find the last buffer
+ * in the hdr's buffer list.
*/
- (void) remove_reference(hdr, hash_lock, tag);
bufp = &hdr->b_l1hdr.b_buf;
- while (*bufp != buf)
- bufp = &(*bufp)->b_next;
- *bufp = buf->b_next;
+ while (*bufp != NULL) {
+ if (*bufp == buf) {
+ *bufp = buf->b_next;
+ }
+
+ /*
+ * If we've removed a buffer in the middle of
+ * the list then update the lastbuf and update
+ * bufp.
+ */
+ if (*bufp != NULL) {
+ lastbuf = *bufp;
+ bufp = &(*bufp)->b_next;
+ }
+ }
buf->b_next = NULL;
+ ASSERT3P(lastbuf, !=, buf);
+ ASSERT3P(lastbuf, !=, NULL);
+ /*
+ * If the current arc_buf_t and the hdr are sharing their data
+ * buffer, then we must stop sharing that block, transfer
+ * ownership and setup sharing with a new arc_buf_t at the end
+ * of the hdr's b_buf list.
+ */
+ if (arc_buf_is_shared(buf)) {
+ ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
+ ASSERT(ARC_BUF_LAST(lastbuf));
+ VERIFY(!arc_buf_is_shared(lastbuf));
+
+ /*
+ * First, sever the block sharing relationship between
+ * buf and the arc_buf_hdr_t. Then, setup a new
+ * block sharing relationship with the last buffer
+ * on the arc_buf_t list.
+ */
+ arc_unshare_buf(hdr, buf);
+ arc_share_buf(hdr, lastbuf);
+ VERIFY3P(lastbuf->b_data, !=, NULL);
+ } else if (HDR_SHARED_DATA(hdr)) {
+ ASSERT(arc_buf_is_shared(lastbuf));
+ }
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
ASSERT3P(state, !=, arc_l2c_only);
- (void) refcount_remove_many(
- &state->arcs_size, hdr->b_size, buf);
+ (void) refcount_remove_many(&state->arcs_size,
+ HDR_GET_LSIZE(hdr), buf);
if (refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
- uint64_t *size;
-
ASSERT3P(state, !=, arc_l2c_only);
- size = &state->arcs_lsize[type];
- ASSERT3U(*size, >=, hdr->b_size);
- atomic_add_64(size, -hdr->b_size);
+ (void) refcount_remove_many(&state->arcs_esize[type],
+ HDR_GET_LSIZE(hdr), buf);
}
- /*
- * We're releasing a duplicate user data buffer, update
- * our statistics accordingly.
- */
- if (HDR_ISTYPE_DATA(hdr)) {
- ARCSTAT_BUMPDOWN(arcstat_duplicate_buffers);
- ARCSTAT_INCR(arcstat_duplicate_buffers_size,
- -hdr->b_size);
- }
- hdr->b_l1hdr.b_datacnt -= 1;
+ hdr->b_l1hdr.b_bufcnt -= 1;
arc_cksum_verify(buf);
arc_buf_unwatch(buf);
mutex_exit(hash_lock);
- nhdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE);
- nhdr->b_size = blksz;
- nhdr->b_spa = spa;
+ /*
+ * Allocate a new hdr. The new hdr will contain a b_pdata
+ * buffer which will be freed in arc_write().
+ */
+ nhdr = arc_hdr_alloc(spa, psize, lsize, compress, type);
+ ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL);
+ ASSERT0(nhdr->b_l1hdr.b_bufcnt);
+ ASSERT0(refcount_count(&nhdr->b_l1hdr.b_refcnt));
+ VERIFY3U(nhdr->b_type, ==, type);
+ ASSERT(!HDR_SHARED_DATA(nhdr));
+ nhdr->b_l1hdr.b_buf = buf;
+ nhdr->b_l1hdr.b_bufcnt = 1;
nhdr->b_l1hdr.b_mru_hits = 0;
nhdr->b_l1hdr.b_mru_ghost_hits = 0;
nhdr->b_l1hdr.b_mfu_hits = 0;
nhdr->b_l1hdr.b_mfu_ghost_hits = 0;
nhdr->b_l1hdr.b_l2_hits = 0;
- nhdr->b_flags = flags & ARC_FLAG_L2_WRITING;
- nhdr->b_flags |= arc_bufc_to_flags(type);
- nhdr->b_flags |= ARC_FLAG_HAS_L1HDR;
-
- nhdr->b_l1hdr.b_buf = buf;
- nhdr->b_l1hdr.b_datacnt = 1;
- nhdr->b_l1hdr.b_state = arc_anon;
- nhdr->b_l1hdr.b_arc_access = 0;
- nhdr->b_l1hdr.b_tmp_cdata = NULL;
- nhdr->b_freeze_cksum = NULL;
-
(void) refcount_add(&nhdr->b_l1hdr.b_refcnt, tag);
buf->b_hdr = nhdr;
+
mutex_exit(&buf->b_evict_lock);
- (void) refcount_add_many(&arc_anon->arcs_size, blksz, buf);
+ (void) refcount_add_many(&arc_anon->arcs_size,
+ HDR_GET_LSIZE(nhdr), buf);
} else {
mutex_exit(&buf->b_evict_lock);
ASSERT(refcount_count(&hdr->b_l1hdr.b_refcnt) == 1);
buf_discard_identity(hdr);
arc_buf_thaw(buf);
}
- buf->b_efunc = NULL;
- buf->b_private = NULL;
}
int
arc_write_callback_t *callback = zio->io_private;
arc_buf_t *buf = callback->awcb_buf;
arc_buf_hdr_t *hdr = buf->b_hdr;
+ uint64_t psize = BP_IS_HOLE(zio->io_bp) ? 0 : BP_GET_PSIZE(zio->io_bp);
+ enum zio_compress compress;
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(!refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt));
- ASSERT(hdr->b_l1hdr.b_datacnt > 0);
- callback->awcb_ready(zio, buf, callback->awcb_private);
+ ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
/*
- * If the IO is already in progress, then this is a re-write
- * attempt, so we need to thaw and re-compute the cksum.
- * It is the responsibility of the callback to handle the
- * accounting for any re-write attempt.
+ * If we're reexecuting this zio because the pool suspended, then
+ * cleanup any state that was previously set the first time the
+ * callback as invoked.
*/
- if (HDR_IO_IN_PROGRESS(hdr)) {
- mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
- if (hdr->b_freeze_cksum != NULL) {
- kmem_free(hdr->b_freeze_cksum, sizeof (zio_cksum_t));
- hdr->b_freeze_cksum = NULL;
+ if (zio->io_flags & ZIO_FLAG_REEXECUTED) {
+ arc_cksum_free(hdr);
+ arc_buf_unwatch(buf);
+ if (hdr->b_l1hdr.b_pdata != NULL) {
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(HDR_SHARED_DATA(hdr));
+
+ arc_unshare_buf(hdr, buf);
+ } else {
+ arc_hdr_free_pdata(hdr);
+ }
}
- mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
}
- arc_cksum_compute(buf, B_FALSE);
- hdr->b_flags |= ARC_FLAG_IO_IN_PROGRESS;
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+ ASSERT(!HDR_SHARED_DATA(hdr));
+ ASSERT(!arc_buf_is_shared(buf));
+
+ callback->awcb_ready(zio, buf, callback->awcb_private);
+
+ if (HDR_IO_IN_PROGRESS(hdr))
+ ASSERT(zio->io_flags & ZIO_FLAG_REEXECUTED);
+
+ arc_cksum_compute(buf);
+ arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
+
+ if (BP_IS_HOLE(zio->io_bp) || BP_IS_EMBEDDED(zio->io_bp)) {
+ compress = ZIO_COMPRESS_OFF;
+ } else {
+ ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(zio->io_bp));
+ compress = BP_GET_COMPRESS(zio->io_bp);
+ }
+ HDR_SET_PSIZE(hdr, psize);
+ arc_hdr_set_compress(hdr, compress);
+
+ /*
+ * If the hdr is compressed, then copy the compressed
+ * zio contents into arc_buf_hdr_t. Otherwise, copy the original
+ * data buf into the hdr. Ideally, we would like to always copy the
+ * io_data into b_pdata but the user may have disabled compressed
+ * arc thus the on-disk block may or may not match what we maintain
+ * in the hdr's b_pdata field.
+ */
+ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF) {
+ ASSERT(BP_GET_COMPRESS(zio->io_bp) != ZIO_COMPRESS_OFF);
+ ASSERT3U(psize, >, 0);
+ arc_hdr_alloc_pdata(hdr);
+ bcopy(zio->io_data, hdr->b_l1hdr.b_pdata, psize);
+ } else {
+ ASSERT3P(buf->b_data, ==, zio->io_orig_data);
+ ASSERT3U(zio->io_orig_size, ==, HDR_GET_LSIZE(hdr));
+ ASSERT3U(hdr->b_l1hdr.b_byteswap, ==, DMU_BSWAP_NUMFUNCS);
+ ASSERT(!HDR_SHARED_DATA(hdr));
+ ASSERT(!arc_buf_is_shared(buf));
+ ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+
+ /*
+ * This hdr is not compressed so we're able to share
+ * the arc_buf_t data buffer with the hdr.
+ */
+ arc_share_buf(hdr, buf);
+ VERIFY0(bcmp(zio->io_orig_data, hdr->b_l1hdr.b_pdata,
+ HDR_GET_LSIZE(hdr)));
+ }
+ arc_hdr_verify(hdr, zio->io_bp);
}
static void
arc_buf_t *buf = callback->awcb_buf;
arc_buf_hdr_t *hdr = buf->b_hdr;
- ASSERT(hdr->b_l1hdr.b_acb == NULL);
+ ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
if (zio->io_error == 0) {
+ arc_hdr_verify(hdr, zio->io_bp);
+
if (BP_IS_HOLE(zio->io_bp) || BP_IS_EMBEDDED(zio->io_bp)) {
buf_discard_identity(hdr);
} else {
hdr->b_birth = BP_PHYSICAL_BIRTH(zio->io_bp);
}
} else {
- ASSERT(BUF_EMPTY(hdr));
+ ASSERT(HDR_EMPTY(hdr));
}
/*
* dva/birth/checksum. The buffer must therefore remain anonymous
* (and uncached).
*/
- if (!BUF_EMPTY(hdr)) {
+ if (!HDR_EMPTY(hdr)) {
arc_buf_hdr_t *exists;
kmutex_t *hash_lock;
(void *)hdr, (void *)exists);
} else {
/* Dedup */
- ASSERT(hdr->b_l1hdr.b_datacnt == 1);
+ ASSERT(hdr->b_l1hdr.b_bufcnt == 1);
ASSERT(hdr->b_l1hdr.b_state == arc_anon);
ASSERT(BP_GET_DEDUP(zio->io_bp));
ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
}
}
- hdr->b_flags &= ~ARC_FLAG_IO_IN_PROGRESS;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
/* if it's not anon, we are doing a scrub */
if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon)
arc_access(hdr, hash_lock);
mutex_exit(hash_lock);
} else {
- hdr->b_flags &= ~ARC_FLAG_IO_IN_PROGRESS;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
}
ASSERT(!refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
zio_t *
arc_write(zio_t *pio, spa_t *spa, uint64_t txg,
- blkptr_t *bp, arc_buf_t *buf, boolean_t l2arc, boolean_t l2arc_compress,
+ blkptr_t *bp, arc_buf_t *buf, boolean_t l2arc,
const zio_prop_t *zp, arc_done_func_t *ready,
arc_done_func_t *children_ready, arc_done_func_t *physdone,
arc_done_func_t *done, void *private, zio_priority_t priority,
arc_write_callback_t *callback;
zio_t *zio;
- ASSERT(ready != NULL);
- ASSERT(done != NULL);
+ ASSERT3P(ready, !=, NULL);
+ ASSERT3P(done, !=, NULL);
ASSERT(!HDR_IO_ERROR(hdr));
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
- ASSERT(hdr->b_l1hdr.b_acb == NULL);
- ASSERT(hdr->b_l1hdr.b_datacnt > 0);
+ ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
+ ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0);
if (l2arc)
- hdr->b_flags |= ARC_FLAG_L2CACHE;
- if (l2arc_compress)
- hdr->b_flags |= ARC_FLAG_L2COMPRESS;
+ arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP);
callback->awcb_ready = ready;
callback->awcb_children_ready = children_ready;
callback->awcb_private = private;
callback->awcb_buf = buf;
- zio = zio_write(pio, spa, txg, bp, buf->b_data, hdr->b_size, zp,
+ /*
+ * The hdr's b_pdata is now stale, free it now. A new data block
+ * will be allocated when the zio pipeline calls arc_write_ready().
+ */
+ if (hdr->b_l1hdr.b_pdata != NULL) {
+ /*
+ * If the buf is currently sharing the data block with
+ * the hdr then we need to break that relationship here.
+ * The hdr will remain with a NULL data pointer and the
+ * buf will take sole ownership of the block.
+ */
+ if (arc_buf_is_shared(buf)) {
+ ASSERT(ARC_BUF_LAST(buf));
+ arc_unshare_buf(hdr, buf);
+ } else {
+ arc_hdr_free_pdata(hdr);
+ }
+ VERIFY3P(buf->b_data, !=, NULL);
+ arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF);
+ }
+ ASSERT(!arc_buf_is_shared(buf));
+ ASSERT3P(hdr->b_l1hdr.b_pdata, ==, NULL);
+
+ zio = zio_write(pio, spa, txg, bp, buf->b_data, HDR_GET_LSIZE(hdr), zp,
arc_write_ready,
(children_ready != NULL) ? arc_write_children_ready : NULL,
arc_write_physdone, arc_write_done, callback,
last_txg = txg;
page_load = 0;
}
-
/*
* If we are in pageout, we know that memory is already tight,
* the arc is already going to be evicting, so we just want to
if (reserve + arc_tempreserve + anon_size > arc_c / 2 &&
anon_size > arc_c / 4) {
+ uint64_t meta_esize =
+ refcount_count(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
+ uint64_t data_esize =
+ refcount_count(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK "
"anon_data=%lluK tempreserve=%lluK arc_c=%lluK\n",
- arc_tempreserve>>10,
- arc_anon->arcs_lsize[ARC_BUFC_METADATA]>>10,
- arc_anon->arcs_lsize[ARC_BUFC_DATA]>>10,
- reserve>>10, arc_c>>10);
+ arc_tempreserve >> 10, meta_esize >> 10,
+ data_esize >> 10, reserve >> 10, arc_c >> 10);
DMU_TX_STAT_BUMP(dmu_tx_dirty_throttle);
return (SET_ERROR(ERESTART));
}
kstat_named_t *evict_data, kstat_named_t *evict_metadata)
{
size->value.ui64 = refcount_count(&state->arcs_size);
- evict_data->value.ui64 = state->arcs_lsize[ARC_BUFC_DATA];
- evict_metadata->value.ui64 = state->arcs_lsize[ARC_BUFC_METADATA];
+ evict_data->value.ui64 =
+ refcount_count(&state->arcs_esize[ARC_BUFC_DATA]);
+ evict_metadata->value.ui64 =
+ refcount_count(&state->arcs_esize[ARC_BUFC_METADATA]);
}
static int
* numbers using buf_hash below. So, as an added precaution,
* let's make sure we never add empty buffers to the arc lists.
*/
- ASSERT(!BUF_EMPTY(hdr));
+ ASSERT(!HDR_EMPTY(hdr));
/*
* The assumption here, is the hash value for a given
}
+static void
+arc_state_init(void)
+{
+ arc_anon = &ARC_anon;
+ arc_mru = &ARC_mru;
+ arc_mru_ghost = &ARC_mru_ghost;
+ arc_mfu = &ARC_mfu;
+ arc_mfu_ghost = &ARC_mfu_ghost;
+ arc_l2c_only = &ARC_l2c_only;
+
+ multilist_create(&arc_mru->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mru->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mfu->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mfu->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+ multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
+ zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
+
+ refcount_create(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_create(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
+ refcount_create(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_create(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
+ refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
+ refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
+ refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
+ refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);
+
+ refcount_create(&arc_anon->arcs_size);
+ refcount_create(&arc_mru->arcs_size);
+ refcount_create(&arc_mru_ghost->arcs_size);
+ refcount_create(&arc_mfu->arcs_size);
+ refcount_create(&arc_mfu_ghost->arcs_size);
+ refcount_create(&arc_l2c_only->arcs_size);
+
+ arc_anon->arcs_state = ARC_STATE_ANON;
+ arc_mru->arcs_state = ARC_STATE_MRU;
+ arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST;
+ arc_mfu->arcs_state = ARC_STATE_MFU;
+ arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST;
+ arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY;
+}
+
+static void
+arc_state_fini(void)
+{
+ refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
+ refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
+ refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
+ refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
+ refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
+ refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
+ refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);
+
+ refcount_destroy(&arc_anon->arcs_size);
+ refcount_destroy(&arc_mru->arcs_size);
+ refcount_destroy(&arc_mru_ghost->arcs_size);
+ refcount_destroy(&arc_mfu->arcs_size);
+ refcount_destroy(&arc_mfu_ghost->arcs_size);
+ refcount_destroy(&arc_l2c_only->arcs_size);
+
+ multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]);
+ multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]);
+ multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]);
+ multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]);
+ multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]);
+ multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]);
+ multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]);
+ multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]);
+}
+
+uint64_t
+arc_max_bytes(void)
+{
+ return (arc_c_max);
+}
+
void
arc_init(void)
{
cv_init(&arc_reclaim_thread_cv, NULL, CV_DEFAULT, NULL);
cv_init(&arc_reclaim_waiters_cv, NULL, CV_DEFAULT, NULL);
- mutex_init(&arc_user_evicts_lock, NULL, MUTEX_DEFAULT, NULL);
- cv_init(&arc_user_evicts_cv, NULL, CV_DEFAULT, NULL);
-
/* Convert seconds to clock ticks */
arc_min_prefetch_lifespan = 1 * hz;
arc_c = arc_c_max;
arc_p = (arc_c >> 1);
+ arc_size = 0;
/* Set min to 1/2 of arc_c_min */
arc_meta_min = 1ULL << SPA_MAXBLOCKSHIFT;
if (arc_c < arc_c_min)
arc_c = arc_c_min;
- arc_anon = &ARC_anon;
- arc_mru = &ARC_mru;
- arc_mru_ghost = &ARC_mru_ghost;
- arc_mfu = &ARC_mfu;
- arc_mfu_ghost = &ARC_mfu_ghost;
- arc_l2c_only = &ARC_l2c_only;
- arc_size = 0;
-
- multilist_create(&arc_mru->arcs_list[ARC_BUFC_METADATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mru->arcs_list[ARC_BUFC_DATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mfu->arcs_list[ARC_BUFC_METADATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mfu->arcs_list[ARC_BUFC_DATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
- multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_DATA],
- sizeof (arc_buf_hdr_t),
- offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
- zfs_arc_num_sublists_per_state, arc_state_multilist_index_func);
-
- arc_anon->arcs_state = ARC_STATE_ANON;
- arc_mru->arcs_state = ARC_STATE_MRU;
- arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST;
- arc_mfu->arcs_state = ARC_STATE_MFU;
- arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST;
- arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY;
-
- refcount_create(&arc_anon->arcs_size);
- refcount_create(&arc_mru->arcs_size);
- refcount_create(&arc_mru_ghost->arcs_size);
- refcount_create(&arc_mfu->arcs_size);
- refcount_create(&arc_mfu_ghost->arcs_size);
- refcount_create(&arc_l2c_only->arcs_size);
-
+ arc_state_init();
buf_init();
- arc_reclaim_thread_exit = FALSE;
- arc_user_evicts_thread_exit = FALSE;
list_create(&arc_prune_list, sizeof (arc_prune_t),
offsetof(arc_prune_t, p_node));
- arc_eviction_list = NULL;
mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL);
- bzero(&arc_eviction_hdr, sizeof (arc_buf_hdr_t));
arc_prune_taskq = taskq_create("arc_prune", max_ncpus, defclsyspri,
max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
+ arc_reclaim_thread_exit = B_FALSE;
+
arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED,
sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
(void) thread_create(NULL, 0, arc_reclaim_thread, NULL, 0, &p0,
TS_RUN, defclsyspri);
- (void) thread_create(NULL, 0, arc_user_evicts_thread, NULL, 0, &p0,
- TS_RUN, defclsyspri);
-
- arc_dead = FALSE;
+ arc_dead = B_FALSE;
arc_warm = B_FALSE;
/*
#endif /* _KERNEL */
mutex_enter(&arc_reclaim_lock);
- arc_reclaim_thread_exit = TRUE;
+ arc_reclaim_thread_exit = B_TRUE;
/*
* The reclaim thread will set arc_reclaim_thread_exit back to
- * FALSE when it is finished exiting; we're waiting for that.
+ * B_FALSE when it is finished exiting; we're waiting for that.
*/
while (arc_reclaim_thread_exit) {
cv_signal(&arc_reclaim_thread_cv);
}
mutex_exit(&arc_reclaim_lock);
- mutex_enter(&arc_user_evicts_lock);
- arc_user_evicts_thread_exit = TRUE;
- /*
- * The user evicts thread will set arc_user_evicts_thread_exit
- * to FALSE when it is finished exiting; we're waiting for that.
- */
- while (arc_user_evicts_thread_exit) {
- cv_signal(&arc_user_evicts_cv);
- cv_wait(&arc_user_evicts_cv, &arc_user_evicts_lock);
- }
- mutex_exit(&arc_user_evicts_lock);
-
- /* Use TRUE to ensure *all* buffers are evicted */
- arc_flush(NULL, TRUE);
+ /* Use B_TRUE to ensure *all* buffers are evicted */
+ arc_flush(NULL, B_TRUE);
- arc_dead = TRUE;
+ arc_dead = B_TRUE;
if (arc_ksp != NULL) {
kstat_delete(arc_ksp);
cv_destroy(&arc_reclaim_thread_cv);
cv_destroy(&arc_reclaim_waiters_cv);
- mutex_destroy(&arc_user_evicts_lock);
- cv_destroy(&arc_user_evicts_cv);
-
- refcount_destroy(&arc_anon->arcs_size);
- refcount_destroy(&arc_mru->arcs_size);
- refcount_destroy(&arc_mru_ghost->arcs_size);
- refcount_destroy(&arc_mfu->arcs_size);
- refcount_destroy(&arc_mfu_ghost->arcs_size);
- refcount_destroy(&arc_l2c_only->arcs_size);
-
- multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]);
- multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]);
- multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]);
- multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]);
- multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]);
- multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]);
- multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]);
- multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]);
- multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA]);
- multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_DATA]);
-
+ arc_state_fini();
buf_fini();
ASSERT0(arc_loaned_bytes);
* l2arc_write_max max write bytes per interval
* l2arc_write_boost extra write bytes during device warmup
* l2arc_noprefetch skip caching prefetched buffers
- * l2arc_nocompress skip compressing buffers
* l2arc_headroom number of max device writes to precache
* l2arc_headroom_boost when we find compressed buffers during ARC
* scanning, we multiply headroom by this
for (df = list_tail(buflist); df; df = df_prev) {
df_prev = list_prev(buflist, df);
- ASSERT(df->l2df_data != NULL);
- ASSERT(df->l2df_func != NULL);
- df->l2df_func(df->l2df_data, df->l2df_size);
+ ASSERT3P(df->l2df_data, !=, NULL);
+ if (df->l2df_type == ARC_BUFC_METADATA) {
+ zio_buf_free(df->l2df_data, df->l2df_size);
+ } else {
+ ASSERT(df->l2df_type == ARC_BUFC_DATA);
+ zio_data_buf_free(df->l2df_data, df->l2df_size);
+ }
list_remove(buflist, df);
kmem_free(df, sizeof (l2arc_data_free_t));
}
int64_t bytes_dropped = 0;
cb = zio->io_private;
- ASSERT(cb != NULL);
+ ASSERT3P(cb, !=, NULL);
dev = cb->l2wcb_dev;
- ASSERT(dev != NULL);
+ ASSERT3P(dev, !=, NULL);
head = cb->l2wcb_head;
- ASSERT(head != NULL);
+ ASSERT3P(head, !=, NULL);
buflist = &dev->l2ad_buflist;
- ASSERT(buflist != NULL);
+ ASSERT3P(buflist, !=, NULL);
DTRACE_PROBE2(l2arc__iodone, zio_t *, zio,
l2arc_write_callback_t *, cb);
*/
ASSERT(HDR_HAS_L1HDR(hdr));
- /*
- * We may have allocated a buffer for L2ARC compression,
- * we must release it to avoid leaking this data.
- */
- l2arc_release_cdata_buf(hdr);
-
/*
* Skipped - drop L2ARC entry and mark the header as no
* longer L2 eligibile.
*/
- if (hdr->b_l2hdr.b_daddr == L2ARC_ADDR_UNSET) {
- list_remove(buflist, hdr);
- hdr->b_flags &= ~ARC_FLAG_HAS_L2HDR;
- hdr->b_flags &= ~ARC_FLAG_L2CACHE;
-
- ARCSTAT_BUMP(arcstat_l2_writes_skip_toobig);
-
- (void) refcount_remove_many(&dev->l2ad_alloc,
- hdr->b_l2hdr.b_asize, hdr);
- } else if (zio->io_error != 0) {
+ if (zio->io_error != 0) {
/*
* Error - drop L2ARC entry.
*/
list_remove(buflist, hdr);
- hdr->b_flags &= ~ARC_FLAG_HAS_L2HDR;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);
- ARCSTAT_INCR(arcstat_l2_asize, -hdr->b_l2hdr.b_asize);
- ARCSTAT_INCR(arcstat_l2_size, -hdr->b_size);
+ ARCSTAT_INCR(arcstat_l2_asize, -arc_hdr_size(hdr));
+ ARCSTAT_INCR(arcstat_l2_size, -HDR_GET_LSIZE(hdr));
- bytes_dropped += hdr->b_l2hdr.b_asize;
+ bytes_dropped += arc_hdr_size(hdr);
(void) refcount_remove_many(&dev->l2ad_alloc,
- hdr->b_l2hdr.b_asize, hdr);
+ arc_hdr_size(hdr), hdr);
}
/*
* Allow ARC to begin reads and ghost list evictions to
* this L2ARC entry.
*/
- hdr->b_flags &= ~ARC_FLAG_L2_WRITING;
+ arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING);
mutex_exit(hash_lock);
}
{
l2arc_read_callback_t *cb;
arc_buf_hdr_t *hdr;
- arc_buf_t *buf;
kmutex_t *hash_lock;
- int equal;
+ boolean_t valid_cksum;
- ASSERT(zio->io_vd != NULL);
+ ASSERT3P(zio->io_vd, !=, NULL);
ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE);
spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd);
cb = zio->io_private;
- ASSERT(cb != NULL);
- buf = cb->l2rcb_buf;
- ASSERT(buf != NULL);
+ ASSERT3P(cb, !=, NULL);
+ hdr = cb->l2rcb_hdr;
+ ASSERT3P(hdr, !=, NULL);
- hash_lock = HDR_LOCK(buf->b_hdr);
+ hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);
- hdr = buf->b_hdr;
ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
- /*
- * If the buffer was compressed, decompress it first.
- */
- if (cb->l2rcb_compress != ZIO_COMPRESS_OFF)
- l2arc_decompress_zio(zio, hdr, cb->l2rcb_compress);
- ASSERT(zio->io_data != NULL);
- ASSERT3U(zio->io_size, ==, hdr->b_size);
- ASSERT3U(BP_GET_LSIZE(&cb->l2rcb_bp), ==, hdr->b_size);
+ ASSERT3P(zio->io_data, !=, NULL);
/*
* Check this survived the L2ARC journey.
*/
- equal = arc_cksum_equal(buf);
- if (equal && zio->io_error == 0 && !HDR_L2_EVICTED(hdr)) {
+ ASSERT3P(zio->io_data, ==, hdr->b_l1hdr.b_pdata);
+ zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */
+ zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */
+
+ valid_cksum = arc_cksum_is_equal(hdr, zio);
+ if (valid_cksum && zio->io_error == 0 && !HDR_L2_EVICTED(hdr)) {
mutex_exit(hash_lock);
- zio->io_private = buf;
- zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */
- zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */
+ zio->io_private = hdr;
arc_read_done(zio);
} else {
mutex_exit(hash_lock);
} else {
zio->io_error = SET_ERROR(EIO);
}
- if (!equal)
+ if (!valid_cksum)
ARCSTAT_BUMP(arcstat_l2_cksum_bad);
/*
ASSERT(!pio || pio->io_child_type == ZIO_CHILD_LOGICAL);
- zio_nowait(zio_read(pio, cb->l2rcb_spa, &cb->l2rcb_bp,
- buf->b_data, hdr->b_size, arc_read_done, buf,
- zio->io_priority, cb->l2rcb_flags, &cb->l2rcb_zb));
+ zio_nowait(zio_read(pio, zio->io_spa, zio->io_bp,
+ hdr->b_l1hdr.b_pdata, zio->io_size, arc_read_done,
+ hdr, zio->io_priority, cb->l2rcb_flags,
+ &cb->l2rcb_zb));
}
}
*/
if (HDR_L2_READING(hdr)) {
ARCSTAT_BUMP(arcstat_l2_evict_reading);
- hdr->b_flags |= ARC_FLAG_L2_EVICTED;
+ arc_hdr_set_flags(hdr, ARC_FLAG_L2_EVICTED);
}
/* Ensure this header has finished being written */
ASSERT(!HDR_L2_WRITING(hdr));
- ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL);
arc_hdr_l2hdr_destroy(hdr);
}
* the delta by which the device hand has changed due to alignment).
*/
static uint64_t
-l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz,
- boolean_t *headroom_boost)
+l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz)
{
arc_buf_hdr_t *hdr, *hdr_prev, *head;
- uint64_t write_asize, write_sz, headroom, buf_compress_minsz,
- stats_size;
- void *buf_data;
+ uint64_t write_asize, write_psize, write_sz, headroom;
boolean_t full;
l2arc_write_callback_t *cb;
zio_t *pio, *wzio;
uint64_t guid = spa_load_guid(spa);
int try;
- const boolean_t do_headroom_boost = *headroom_boost;
- ASSERT(dev->l2ad_vdev != NULL);
-
- /* Lower the flag now, we might want to raise it again later. */
- *headroom_boost = B_FALSE;
+ ASSERT3P(dev->l2ad_vdev, !=, NULL);
pio = NULL;
- write_sz = write_asize = 0;
+ write_sz = write_asize = write_psize = 0;
full = B_FALSE;
head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE);
- head->b_flags |= ARC_FLAG_L2_WRITE_HEAD;
- head->b_flags |= ARC_FLAG_HAS_L2HDR;
-
- /*
- * We will want to try to compress buffers that are at least 2x the
- * device sector size.
- */
- buf_compress_minsz = 2 << dev->l2ad_vdev->vdev_ashift;
+ arc_hdr_set_flags(head, ARC_FLAG_L2_WRITE_HEAD | ARC_FLAG_HAS_L2HDR);
/*
* Copy buffers for L2ARC writing.
hdr = multilist_sublist_tail(mls);
headroom = target_sz * l2arc_headroom;
- if (do_headroom_boost)
+ if (zfs_compressed_arc_enabled)
headroom = (headroom * l2arc_headroom_boost) / 100;
for (; hdr; hdr = hdr_prev) {
kmutex_t *hash_lock;
- uint64_t buf_sz;
- uint64_t buf_a_sz;
+ uint64_t asize, size;
+ void *to_write;
if (arc_warm == B_FALSE)
hdr_prev = multilist_sublist_next(mls, hdr);
continue;
}
- passed_sz += hdr->b_size;
+ passed_sz += HDR_GET_LSIZE(hdr);
if (passed_sz > headroom) {
/*
* Searched too far.
continue;
}
- /*
- * Assume that the buffer is not going to be compressed
- * and could take more space on disk because of a larger
- * disk block size.
- */
- buf_sz = hdr->b_size;
- buf_a_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz);
-
- if ((write_asize + buf_a_sz) > target_sz) {
+ if ((write_asize + HDR_GET_LSIZE(hdr)) > target_sz) {
full = B_TRUE;
mutex_exit(hash_lock);
break;
ZIO_FLAG_CANFAIL);
}
- /*
- * Create and add a new L2ARC header.
- */
hdr->b_l2hdr.b_dev = dev;
- hdr->b_flags |= ARC_FLAG_L2_WRITING;
- /*
- * Temporarily stash the data buffer in b_tmp_cdata.
- * The subsequent write step will pick it up from
- * there. This is because can't access b_l1hdr.b_buf
- * without holding the hash_lock, which we in turn
- * can't access without holding the ARC list locks
- * (which we want to avoid during compression/writing)
- */
- hdr->b_l2hdr.b_compress = ZIO_COMPRESS_OFF;
- hdr->b_l2hdr.b_asize = hdr->b_size;
hdr->b_l2hdr.b_hits = 0;
- hdr->b_l1hdr.b_tmp_cdata = hdr->b_l1hdr.b_buf->b_data;
- /*
- * Explicitly set the b_daddr field to a known
- * value which means "invalid address". This
- * enables us to differentiate which stage of
- * l2arc_write_buffers() the particular header
- * is in (e.g. this loop, or the one below).
- * ARC_FLAG_L2_WRITING is not enough to make
- * this distinction, and we need to know in
- * order to do proper l2arc vdev accounting in
- * arc_release() and arc_hdr_destroy().
- *
- * Note, we can't use a new flag to distinguish
- * the two stages because we don't hold the
- * header's hash_lock below, in the second stage
- * of this function. Thus, we can't simply
- * change the b_flags field to denote that the
- * IO has been sent. We can change the b_daddr
- * field of the L2 portion, though, since we'll
- * be holding the l2ad_mtx; which is why we're
- * using it to denote the header's state change.
- */
- hdr->b_l2hdr.b_daddr = L2ARC_ADDR_UNSET;
- hdr->b_flags |= ARC_FLAG_HAS_L2HDR;
+ hdr->b_l2hdr.b_daddr = dev->l2ad_hand;
+ arc_hdr_set_flags(hdr,
+ ARC_FLAG_L2_WRITING | ARC_FLAG_HAS_L2HDR);
mutex_enter(&dev->l2ad_mtx);
list_insert_head(&dev->l2ad_buflist, hdr);
mutex_exit(&dev->l2ad_mtx);
/*
- * Compute and store the buffer cksum before
- * writing. On debug the cksum is verified first.
+ * We rely on the L1 portion of the header below, so
+ * it's invalid for this header to have been evicted out
+ * of the ghost cache, prior to being written out. The
+ * ARC_FLAG_L2_WRITING bit ensures this won't happen.
*/
- arc_cksum_verify(hdr->b_l1hdr.b_buf);
- arc_cksum_compute(hdr->b_l1hdr.b_buf, B_TRUE);
-
- mutex_exit(hash_lock);
-
- write_sz += buf_sz;
- write_asize += buf_a_sz;
- }
-
- multilist_sublist_unlock(mls);
-
- if (full == B_TRUE)
- break;
- }
-
- /* No buffers selected for writing? */
- if (pio == NULL) {
- ASSERT0(write_sz);
- ASSERT(!HDR_HAS_L1HDR(head));
- kmem_cache_free(hdr_l2only_cache, head);
- return (0);
- }
-
- mutex_enter(&dev->l2ad_mtx);
-
- /*
- * Note that elsewhere in this file arcstat_l2_asize
- * and the used space on l2ad_vdev are updated using b_asize,
- * which is not necessarily rounded up to the device block size.
- * Too keep accounting consistent we do the same here as well:
- * stats_size accumulates the sum of b_asize of the written buffers,
- * while write_asize accumulates the sum of b_asize rounded up
- * to the device block size.
- * The latter sum is used only to validate the corectness of the code.
- */
- stats_size = 0;
- write_asize = 0;
-
- /*
- * Now start writing the buffers. We're starting at the write head
- * and work backwards, retracing the course of the buffer selector
- * loop above.
- */
- for (hdr = list_prev(&dev->l2ad_buflist, head); hdr;
- hdr = list_prev(&dev->l2ad_buflist, hdr)) {
- uint64_t buf_sz;
-
- /*
- * We rely on the L1 portion of the header below, so
- * it's invalid for this header to have been evicted out
- * of the ghost cache, prior to being written out. The
- * ARC_FLAG_L2_WRITING bit ensures this won't happen.
- */
- ASSERT(HDR_HAS_L1HDR(hdr));
-
- /*
- * We shouldn't need to lock the buffer here, since we flagged
- * it as ARC_FLAG_L2_WRITING in the previous step, but we must
- * take care to only access its L2 cache parameters. In
- * particular, hdr->l1hdr.b_buf may be invalid by now due to
- * ARC eviction.
- */
- hdr->b_l2hdr.b_daddr = dev->l2ad_hand;
-
- if ((!l2arc_nocompress && HDR_L2COMPRESS(hdr)) &&
- hdr->b_l2hdr.b_asize >= buf_compress_minsz) {
- if (l2arc_compress_buf(hdr)) {
- /*
- * If compression succeeded, enable headroom
- * boost on the next scan cycle.
- */
- *headroom_boost = B_TRUE;
- }
- }
-
- /*
- * Pick up the buffer data we had previously stashed away
- * (and now potentially also compressed).
- */
- buf_data = hdr->b_l1hdr.b_tmp_cdata;
- buf_sz = hdr->b_l2hdr.b_asize;
+ ASSERT(HDR_HAS_L1HDR(hdr));
- /*
- * We need to do this regardless if buf_sz is zero or
- * not, otherwise, when this l2hdr is evicted we'll
- * remove a reference that was never added.
- */
- (void) refcount_add_many(&dev->l2ad_alloc, buf_sz, hdr);
+ ASSERT3U(HDR_GET_PSIZE(hdr), >, 0);
+ ASSERT3P(hdr->b_l1hdr.b_pdata, !=, NULL);
+ ASSERT3U(arc_hdr_size(hdr), >, 0);
+ size = arc_hdr_size(hdr);
- /* Compression may have squashed the buffer to zero length. */
- if (buf_sz != 0) {
- uint64_t buf_a_sz;
+ (void) refcount_add_many(&dev->l2ad_alloc, size, hdr);
/*
- * Buffers which are larger than l2arc_max_block_size
- * after compression are skipped and removed from L2
- * eligibility.
+ * Normally the L2ARC can use the hdr's data, but if
+ * we're sharing data between the hdr and one of its
+ * bufs, L2ARC needs its own copy of the data so that
+ * the ZIO below can't race with the buf consumer. To
+ * ensure that this copy will be available for the
+ * lifetime of the ZIO and be cleaned up afterwards, we
+ * add it to the l2arc_free_on_write queue.
*/
- if (buf_sz > l2arc_max_block_size) {
- hdr->b_l2hdr.b_daddr = L2ARC_ADDR_UNSET;
- continue;
- }
+ if (!HDR_SHARED_DATA(hdr)) {
+ to_write = hdr->b_l1hdr.b_pdata;
+ } else {
+ arc_buf_contents_t type = arc_buf_type(hdr);
+ if (type == ARC_BUFC_METADATA) {
+ to_write = zio_buf_alloc(size);
+ } else {
+ ASSERT3U(type, ==, ARC_BUFC_DATA);
+ to_write = zio_data_buf_alloc(size);
+ }
+ bcopy(hdr->b_l1hdr.b_pdata, to_write, size);
+ l2arc_free_data_on_write(to_write, size, type);
+ }
wzio = zio_write_phys(pio, dev->l2ad_vdev,
- dev->l2ad_hand, buf_sz, buf_data, ZIO_CHECKSUM_OFF,
- NULL, NULL, ZIO_PRIORITY_ASYNC_WRITE,
+ hdr->b_l2hdr.b_daddr, size, to_write,
+ ZIO_CHECKSUM_OFF, NULL, hdr,
+ ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_CANFAIL, B_FALSE);
+ write_sz += HDR_GET_LSIZE(hdr);
DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev,
zio_t *, wzio);
- (void) zio_nowait(wzio);
-
- stats_size += buf_sz;
+ write_asize += size;
/*
* Keep the clock hand suitably device-aligned.
*/
- buf_a_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz);
- write_asize += buf_a_sz;
- dev->l2ad_hand += buf_a_sz;
+ asize = vdev_psize_to_asize(dev->l2ad_vdev, size);
+ write_psize += asize;
+ dev->l2ad_hand += asize;
+
+ mutex_exit(hash_lock);
+
+ (void) zio_nowait(wzio);
}
+
+ multilist_sublist_unlock(mls);
+
+ if (full == B_TRUE)
+ break;
}
- mutex_exit(&dev->l2ad_mtx);
+ /* No buffers selected for writing? */
+ if (pio == NULL) {
+ ASSERT0(write_sz);
+ ASSERT(!HDR_HAS_L1HDR(head));
+ kmem_cache_free(hdr_l2only_cache, head);
+ return (0);
+ }
ASSERT3U(write_asize, <=, target_sz);
ARCSTAT_BUMP(arcstat_l2_writes_sent);
ARCSTAT_INCR(arcstat_l2_write_bytes, write_asize);
ARCSTAT_INCR(arcstat_l2_size, write_sz);
- ARCSTAT_INCR(arcstat_l2_asize, stats_size);
- vdev_space_update(dev->l2ad_vdev, stats_size, 0, 0);
+ ARCSTAT_INCR(arcstat_l2_asize, write_asize);
+ vdev_space_update(dev->l2ad_vdev, write_asize, 0, 0);
/*
* Bump device hand to the device start if it is approaching the end.
return (write_asize);
}
-/*
- * Compresses an L2ARC buffer.
- * The data to be compressed must be prefilled in l1hdr.b_tmp_cdata and its
- * size in l2hdr->b_asize. This routine tries to compress the data and
- * depending on the compression result there are three possible outcomes:
- * *) The buffer was incompressible. The original l2hdr contents were left
- * untouched and are ready for writing to an L2 device.
- * *) The buffer was all-zeros, so there is no need to write it to an L2
- * device. To indicate this situation b_tmp_cdata is NULL'ed, b_asize is
- * set to zero and b_compress is set to ZIO_COMPRESS_EMPTY.
- * *) Compression succeeded and b_tmp_cdata was replaced with a temporary
- * data buffer which holds the compressed data to be written, and b_asize
- * tells us how much data there is. b_compress is set to the appropriate
- * compression algorithm. Once writing is done, invoke
- * l2arc_release_cdata_buf on this l2hdr to free this temporary buffer.
- *
- * Returns B_TRUE if compression succeeded, or B_FALSE if it didn't (the
- * buffer was incompressible).
- */
-static boolean_t
-l2arc_compress_buf(arc_buf_hdr_t *hdr)
-{
- void *cdata;
- size_t csize, len, rounded;
- l2arc_buf_hdr_t *l2hdr;
-
- ASSERT(HDR_HAS_L2HDR(hdr));
-
- l2hdr = &hdr->b_l2hdr;
-
- ASSERT(HDR_HAS_L1HDR(hdr));
- ASSERT3U(l2hdr->b_compress, ==, ZIO_COMPRESS_OFF);
- ASSERT(hdr->b_l1hdr.b_tmp_cdata != NULL);
-
- len = l2hdr->b_asize;
- cdata = zio_data_buf_alloc(len);
- ASSERT3P(cdata, !=, NULL);
- csize = zio_compress_data(ZIO_COMPRESS_LZ4, hdr->b_l1hdr.b_tmp_cdata,
- cdata, l2hdr->b_asize);
-
- rounded = P2ROUNDUP(csize, (size_t)SPA_MINBLOCKSIZE);
- if (rounded > csize) {
- bzero((char *)cdata + csize, rounded - csize);
- csize = rounded;
- }
-
- if (csize == 0) {
- /* zero block, indicate that there's nothing to write */
- zio_data_buf_free(cdata, len);
- l2hdr->b_compress = ZIO_COMPRESS_EMPTY;
- l2hdr->b_asize = 0;
- hdr->b_l1hdr.b_tmp_cdata = NULL;
- ARCSTAT_BUMP(arcstat_l2_compress_zeros);
- return (B_TRUE);
- } else if (csize > 0 && csize < len) {
- /*
- * Compression succeeded, we'll keep the cdata around for
- * writing and release it afterwards.
- */
- l2hdr->b_compress = ZIO_COMPRESS_LZ4;
- l2hdr->b_asize = csize;
- hdr->b_l1hdr.b_tmp_cdata = cdata;
- ARCSTAT_BUMP(arcstat_l2_compress_successes);
- return (B_TRUE);
- } else {
- /*
- * Compression failed, release the compressed buffer.
- * l2hdr will be left unmodified.
- */
- zio_data_buf_free(cdata, len);
- ARCSTAT_BUMP(arcstat_l2_compress_failures);
- return (B_FALSE);
- }
-}
-
-/*
- * Decompresses a zio read back from an l2arc device. On success, the
- * underlying zio's io_data buffer is overwritten by the uncompressed
- * version. On decompression error (corrupt compressed stream), the
- * zio->io_error value is set to signal an I/O error.
- *
- * Please note that the compressed data stream is not checksummed, so
- * if the underlying device is experiencing data corruption, we may feed
- * corrupt data to the decompressor, so the decompressor needs to be
- * able to handle this situation (LZ4 does).
- */
-static void
-l2arc_decompress_zio(zio_t *zio, arc_buf_hdr_t *hdr, enum zio_compress c)
-{
- uint64_t csize;
- void *cdata;
-
- ASSERT(L2ARC_IS_VALID_COMPRESS(c));
-
- if (zio->io_error != 0) {
- /*
- * An io error has occured, just restore the original io
- * size in preparation for a main pool read.
- */
- zio->io_orig_size = zio->io_size = hdr->b_size;
- return;
- }
-
- if (c == ZIO_COMPRESS_EMPTY) {
- /*
- * An empty buffer results in a null zio, which means we
- * need to fill its io_data after we're done restoring the
- * buffer's contents.
- */
- ASSERT(hdr->b_l1hdr.b_buf != NULL);
- bzero(hdr->b_l1hdr.b_buf->b_data, hdr->b_size);
- zio->io_data = zio->io_orig_data = hdr->b_l1hdr.b_buf->b_data;
- } else {
- ASSERT(zio->io_data != NULL);
- /*
- * We copy the compressed data from the start of the arc buffer
- * (the zio_read will have pulled in only what we need, the
- * rest is garbage which we will overwrite at decompression)
- * and then decompress back to the ARC data buffer. This way we
- * can minimize copying by simply decompressing back over the
- * original compressed data (rather than decompressing to an
- * aux buffer and then copying back the uncompressed buffer,
- * which is likely to be much larger).
- */
- csize = zio->io_size;
- cdata = zio_data_buf_alloc(csize);
- bcopy(zio->io_data, cdata, csize);
- if (zio_decompress_data(c, cdata, zio->io_data, csize,
- hdr->b_size) != 0)
- zio->io_error = EIO;
- zio_data_buf_free(cdata, csize);
- }
-
- /* Restore the expected uncompressed IO size. */
- zio->io_orig_size = zio->io_size = hdr->b_size;
-}
-
-/*
- * Releases the temporary b_tmp_cdata buffer in an l2arc header structure.
- * This buffer serves as a temporary holder of compressed data while
- * the buffer entry is being written to an l2arc device. Once that is
- * done, we can dispose of it.
- */
-static void
-l2arc_release_cdata_buf(arc_buf_hdr_t *hdr)
-{
- enum zio_compress comp;
-
- ASSERT(HDR_HAS_L1HDR(hdr));
- ASSERT(HDR_HAS_L2HDR(hdr));
- comp = hdr->b_l2hdr.b_compress;
- ASSERT(comp == ZIO_COMPRESS_OFF || L2ARC_IS_VALID_COMPRESS(comp));
-
- if (comp == ZIO_COMPRESS_OFF) {
- /*
- * In this case, b_tmp_cdata points to the same buffer
- * as the arc_buf_t's b_data field. We don't want to
- * free it, since the arc_buf_t will handle that.
- */
- hdr->b_l1hdr.b_tmp_cdata = NULL;
- } else if (comp == ZIO_COMPRESS_EMPTY) {
- /*
- * In this case, b_tmp_cdata was compressed to an empty
- * buffer, thus there's nothing to free and b_tmp_cdata
- * should have been set to NULL in l2arc_write_buffers().
- */
- ASSERT3P(hdr->b_l1hdr.b_tmp_cdata, ==, NULL);
- } else {
- /*
- * If the data was compressed, then we've allocated a
- * temporary buffer for it, so now we need to release it.
- */
- ASSERT(hdr->b_l1hdr.b_tmp_cdata != NULL);
- zio_data_buf_free(hdr->b_l1hdr.b_tmp_cdata,
- hdr->b_size);
- hdr->b_l1hdr.b_tmp_cdata = NULL;
- }
-
-}
-
/*
* This thread feeds the L2ARC at regular intervals. This is the beating
* heart of the L2ARC.
spa_t *spa;
uint64_t size, wrote;
clock_t begin, next = ddi_get_lbolt();
- boolean_t headroom_boost = B_FALSE;
fstrans_cookie_t cookie;
CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG);
continue;
spa = dev->l2ad_spa;
- ASSERT(spa != NULL);
+ ASSERT3P(spa, !=, NULL);
/*
* If the pool is read-only then force the feed thread to
/*
* Write ARC buffers.
*/
- wrote = l2arc_write_buffers(spa, dev, size, &headroom_boost);
+ wrote = l2arc_write_buffers(spa, dev, size);
/*
* Calculate interval between writes.
break;
}
}
- ASSERT(remdev != NULL);
+ ASSERT3P(remdev, !=, NULL);
/*
* Remove device from global list
EXPORT_SYMBOL(arc_buf_size);
EXPORT_SYMBOL(arc_write);
EXPORT_SYMBOL(arc_read);
-EXPORT_SYMBOL(arc_buf_remove_ref);
EXPORT_SYMBOL(arc_buf_info);
EXPORT_SYMBOL(arc_getbuf_func);
EXPORT_SYMBOL(arc_add_prune_callback);
module_param(zfs_arc_p_min_shift, int, 0644);
MODULE_PARM_DESC(zfs_arc_p_min_shift, "arc_c shift to calc min/max arc_p");
-module_param(zfs_disable_dup_eviction, int, 0644);
-MODULE_PARM_DESC(zfs_disable_dup_eviction, "disable duplicate buffer eviction");
-
module_param(zfs_arc_average_blocksize, int, 0444);
MODULE_PARM_DESC(zfs_arc_average_blocksize, "Target average block size");
+module_param(zfs_compressed_arc_enabled, int, 0644);
+MODULE_PARM_DESC(zfs_arc_average_blocksize, "Disable compressed arc buffers");
+
module_param(zfs_arc_min_prefetch_lifespan, int, 0644);
MODULE_PARM_DESC(zfs_arc_min_prefetch_lifespan, "Min life of prefetch block");
module_param(l2arc_headroom_boost, ulong, 0644);
MODULE_PARM_DESC(l2arc_headroom_boost, "Compressed l2arc_headroom multiplier");
-module_param(l2arc_max_block_size, ulong, 0644);
-MODULE_PARM_DESC(l2arc_max_block_size, "Skip L2ARC buffers larger than N");
-
module_param(l2arc_feed_secs, ulong, 0644);
MODULE_PARM_DESC(l2arc_feed_secs, "Seconds between L2ARC writing");
module_param(l2arc_noprefetch, int, 0644);
MODULE_PARM_DESC(l2arc_noprefetch, "Skip caching prefetched buffers");
-module_param(l2arc_nocompress, int, 0644);
-MODULE_PARM_DESC(l2arc_nocompress, "Skip compressing L2ARC buffers");
-
module_param(l2arc_feed_again, int, 0644);
MODULE_PARM_DESC(l2arc_feed_again, "Turbo L2ARC warmup");
projects / zfs / commitdiff