From acb42ec4652450f7cd73440929fd57f1487282bf Mon Sep 17 00:00:00 2001 From: Joe Thornber Date: Tue, 30 Jan 2018 10:46:08 +0000 Subject: [PATCH] [device/bcache] Initial code drop. Compiles. Not written tests yet. --- lib/device/bcache.c | 833 ++++++++++++++++++++++++++++++++++++++++++++ lib/device/bcache.h | 83 +++++ 2 files changed, 916 insertions(+) create mode 100644 lib/device/bcache.c create mode 100644 lib/device/bcache.h diff --git a/lib/device/bcache.c b/lib/device/bcache.c new file mode 100644 index 000000000..1be626c6d --- /dev/null +++ b/lib/device/bcache.c @@ -0,0 +1,833 @@ +#define _GNU_SOURCE + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "bcache.h" +#include "dm-logging.h" +#include "log.h" + +#define SECTOR_SHIFT 9L + +//---------------------------------------------------------------- + +// Assumes the list is not empty. +static inline struct dm_list *_list_pop(struct dm_list *head) +{ + struct dm_list *l; + + l = head->n; + dm_list_del(l); + return l; +} + +//---------------------------------------------------------------- + +struct control_block { + struct dm_list list; + void *context; + struct iocb cb; +}; + +struct cb_set { + struct dm_list free; + struct dm_list allocated; + struct control_block *vec; +} control_block_set; + +static struct cb_set *_cb_set_create(unsigned nr) +{ + int i; + struct cb_set *cbs = malloc(sizeof(*cbs)); + + if (!cbs) + return NULL; + + cbs->vec = malloc(nr * sizeof(*cbs->vec)); + if (!cbs->vec) { + free(cbs); + return NULL; + } + + dm_list_init(&cbs->free); + dm_list_init(&cbs->allocated); + + for (i = 0; i < nr; i++) + dm_list_add(&cbs->free, &cbs->vec[i].list); + + return cbs; +} + +static bool _cb_set_destroy(struct cb_set *cbs) +{ + if (!dm_list_empty(&cbs->allocated)) { + // FIXME: I think we should propogate this up. + log_error("async io still in flight"); + return false; + } + + free(cbs->vec); + free(cbs); + return 0; +} + +static struct control_block *_cb_alloc(struct cb_set *cbs, void *context) +{ + struct control_block *cb; + + if (dm_list_empty(&cbs->free)) + return NULL; + + cb = dm_list_item(_list_pop(&cbs->free), struct control_block); + cb->context = context; + dm_list_add(&cbs->allocated, &cb->list); + + return cb; +} + +static void _cb_free(struct cb_set *cbs, struct control_block *cb) +{ + dm_list_del(&cb->list); + dm_list_add_h(&cbs->free, &cb->list); +} + +static struct control_block *_iocb_to_cb(struct iocb *icb) +{ + return dm_list_struct_base(icb, struct control_block, cb); +} + +//---------------------------------------------------------------- + +// FIXME: write a sync engine too +enum dir { + DIR_READ, + DIR_WRITE +}; + +struct io_engine { + io_context_t aio_context; + struct cb_set *cbs; +}; + +static struct io_engine *_engine_create(unsigned max_io) +{ + int r; + struct io_engine *e = malloc(sizeof(*e)); + + if (!e) + return NULL; + + e->aio_context = 0; + r = io_setup(max_io, &e->aio_context); + if (r < 0) { + log_warn("io_setup failed"); + return NULL; + } + + e->cbs = _cb_set_create(max_io); + if (!e->cbs) { + log_warn("couldn't create control block set"); + free(e); + return NULL; + } + + return e; +} + +static void _engine_destroy(struct io_engine *e) +{ + _cb_set_destroy(e->cbs); + io_destroy(e->aio_context); + free(e); +} + +static bool _engine_issue(struct io_engine *e, int fd, enum dir d, + sector_t sb, sector_t se, void *data, void *context) +{ + int r; + struct iocb *cb_array[1]; + struct control_block *cb; + + if (((uint64_t) data) & (PAGE_SIZE - 1)) { + log_err("misaligned data buffer"); + return false; + } + + cb = _cb_alloc(e->cbs, context); + if (!cb) { + log_err("couldn't allocate control block"); + return false; + } + + memset(&cb->cb, 0, sizeof(cb->cb)); + + cb->cb.aio_fildes = (int) fd; + cb->cb.u.c.buf = data; + cb->cb.u.c.offset = sb << SECTOR_SHIFT; + cb->cb.u.c.nbytes = (se - sb) << SECTOR_SHIFT; + cb->cb.aio_lio_opcode = (d == DIR_READ) ? IO_CMD_PREAD : IO_CMD_PWRITE; + + cb_array[0] = &cb->cb; + r = io_submit(e->aio_context, 1, cb_array); + if (r < 0) { + log_sys_error("io_submit", ""); + _cb_free(e->cbs, cb); + return false; + } + + return true; +} + +#define MAX_IO 64 +typedef void complete_fn(void *context, int io_error); + +static bool _engine_wait(struct io_engine *e, complete_fn fn) +{ + int i, r; + struct io_event event[MAX_IO]; + struct control_block *cb; + + memset(&event, 0, sizeof(event)); + r = io_getevents(e->aio_context, 1, MAX_IO, event, NULL); + if (r < 0) { + log_sys_error("io_getevents", ""); + return false; + } + + for (i = 0; i < r; i++) { + struct io_event *ev = event + i; + + cb = _iocb_to_cb((struct iocb *) ev->obj); + + if (ev->res == cb->cb.u.c.nbytes) + fn((void *) cb->context, 0); + + else if ((int) ev->res < 0) + fn(cb->context, (int) ev->res); + + else { + log_err("short io"); + fn(cb->context, -ENODATA); + } + + _cb_free(e->cbs, cb); + } + + return true; +} + +//---------------------------------------------------------------- + +#define MIN_BLOCKS 16 +#define WRITEBACK_LOW_THRESHOLD_PERCENT 33 +#define WRITEBACK_HIGH_THRESHOLD_PERCENT 66 + +//---------------------------------------------------------------- + +static void *_alloc_aligned(size_t len, size_t alignment) +{ + void *result = NULL; + int r = posix_memalign(&result, alignment, len); + if (r) + return NULL; + + return result; +} + +//---------------------------------------------------------------- + +static bool _test_flags(struct block *b, unsigned bits) +{ + return (b->flags & bits) != 0; +} + +static void _set_flags(struct block *b, unsigned bits) +{ + b->flags |= bits; +} + +static void _clear_flags(struct block *b, unsigned bits) +{ + b->flags &= ~bits; +} + +//---------------------------------------------------------------- + +enum block_flags { + BF_IO_PENDING = (1 << 0), + BF_DIRTY = (1 << 1), +}; + +struct bcache { + int fd; + sector_t block_sectors; + uint64_t nr_data_blocks; + uint64_t nr_cache_blocks; + + struct io_engine *engine; + + void *raw_data; + struct block *raw_blocks; + + /* + * Lists that categorise the blocks. + */ + unsigned nr_locked; + unsigned nr_dirty; + unsigned nr_io_pending; + + struct dm_list free; + struct dm_list errored; + struct dm_list dirty; + struct dm_list clean; + struct dm_list io_pending; + + /* + * Hash table. + */ + unsigned nr_buckets; + unsigned hash_mask; + struct dm_list *buckets; + + /* + * Statistics + */ + unsigned read_hits; + unsigned read_misses; + unsigned write_zeroes; + unsigned write_hits; + unsigned write_misses; + unsigned prefetches; +}; + +//---------------------------------------------------------------- + +/* 2^63 + 2^61 - 2^57 + 2^54 - 2^51 - 2^18 + 1 */ +#define GOLDEN_RATIO_PRIME_64 0x9e37fffffffc0001UL + +static unsigned _hash(struct bcache *cache, int fd, uint64_t index) +{ + uint64_t h = (index << 10) & fd; + h *= GOLDEN_RATIO_PRIME_64; + return h & cache->hash_mask; +} + +static struct block *_hash_lookup(struct bcache *cache, int fd, uint64_t index) +{ + struct block *b; + unsigned h = _hash(cache, fd, index); + + dm_list_iterate_items_gen (b, cache->buckets + h, hash) + if (b->index == index) + return b; + + return NULL; +} + +static void _hash_insert(struct block *b) +{ + unsigned h = _hash(b->cache, b->fd, b->index); + dm_list_add_h(b->cache->buckets + h, &b->hash); +} + +static void _hash_remove(struct block *b) +{ + dm_list_del(&b->hash); +} + +/* + * Must return a power of 2. + */ +static unsigned _calc_nr_buckets(unsigned nr_blocks) +{ + unsigned r = 8; + unsigned n = nr_blocks / 4; + + if (n < 8) + n = 8; + + while (r < n) + r <<= 1; + + return r; +} + +static int _hash_table_init(struct bcache *cache, unsigned nr_entries) +{ + unsigned i; + + cache->nr_buckets = _calc_nr_buckets(nr_entries); + cache->hash_mask = cache->nr_buckets - 1; + cache->buckets = malloc(cache->nr_buckets * sizeof(*cache->buckets)); + if (!cache->buckets) + return -ENOMEM; + + for (i = 0; i < cache->nr_buckets; i++) + dm_list_init(cache->buckets + i); + + return 0; +} + +static void _hash_table_exit(struct bcache *cache) +{ + free(cache->buckets); +} + +//---------------------------------------------------------------- + +static int _init_free_list(struct bcache *cache, unsigned count) +{ + unsigned i; + size_t block_size = cache->block_sectors << SECTOR_SHIFT; + unsigned char *data = + (unsigned char *) _alloc_aligned(count * block_size, PAGE_SIZE); + + /* Allocate the data for each block. We page align the data. */ + if (!data) + return -ENOMEM; + + cache->raw_data = data; + cache->raw_blocks = malloc(count * sizeof(*cache->raw_blocks)); + + if (!cache->raw_blocks) + free(cache->raw_data); + + for (i = 0; i < count; i++) { + struct block *b = cache->raw_blocks + i; + b->cache = cache; + b->data = data + (block_size * i); + dm_list_add(&cache->free, &b->list); + } + + return 0; +} + +static void _exit_free_list(struct bcache *cache) +{ + free(cache->raw_data); + free(cache->raw_blocks); +} + +static struct block *_alloc_block(struct bcache *cache) +{ + struct block *b = dm_list_struct_base(_list_pop(&cache->free), struct block, list); + return b; +} + +/*---------------------------------------------------------------- + * Clean/dirty list management. + * Always use these methods to ensure nr_dirty_ is correct. + *--------------------------------------------------------------*/ + +static void _unlink_block(struct block *b) +{ + if (_test_flags(b, BF_DIRTY)) + b->cache->nr_dirty--; + + dm_list_del(&b->list); +} + +static void _link_block(struct block *b) +{ + struct bcache *cache = b->cache; + + if (_test_flags(b, BF_DIRTY)) { + dm_list_add(&cache->dirty, &b->list); + cache->nr_dirty++; + } else + dm_list_add(&cache->clean, &b->list); +} + +static void _relink(struct block *b) +{ + _unlink_block(b); + _link_block(b); +} + +/*---------------------------------------------------------------- + * Low level IO handling + * + * We cannot have two concurrent writes on the same block. + * eg, background writeback, put with dirty, flush? + * + * To avoid this we introduce some restrictions: + * + * i) A held block can never be written back. + * ii) You cannot get a block until writeback has completed. + * + *--------------------------------------------------------------*/ + +/* + * |b->list| should be valid (either pointing to itself, on one of the other + * lists. + */ +static bool _issue_low_level(struct block *b, enum dir d) +{ + struct bcache *cache = b->cache; + sector_t sb = b->index * cache->block_sectors; + sector_t se = sb + cache->block_sectors; + + if (_test_flags(b, BF_IO_PENDING)) + return false; + + _set_flags(b, BF_IO_PENDING); + return _engine_issue(cache->engine, cache->fd, d, sb, se, b->data, b); +} + +static inline bool _issue_read(struct block *b) +{ + return _issue_low_level(b, DIR_READ); +} + +static inline bool _issue_write(struct block *b) +{ + return _issue_low_level(b, DIR_WRITE); +} + +static void _complete_io(void *context, int err) +{ + struct block *b = context; + struct bcache *cache = b->cache; + + b->error = err; + _clear_flags(b, BF_IO_PENDING); + cache->nr_io_pending--; + + /* + * b is on the io_pending list, so we don't want to use unlink_block. + * Which would incorrectly adjust nr_dirty. + */ + dm_list_del(&b->list); + + if (b->error) + dm_list_add(&cache->errored, &b->list); + + else { + _clear_flags(b, BF_DIRTY); + _link_block(b); + } +} + +static int _wait_io(struct bcache *cache) +{ + return _engine_wait(cache->engine, _complete_io); +} + +/*---------------------------------------------------------------- + * High level IO handling + *--------------------------------------------------------------*/ + +static void _wait_all(struct bcache *cache) +{ + while (!dm_list_empty(&cache->io_pending)) + _wait_io(cache); +} + +static void _wait_specific(struct block *b) +{ + while (_test_flags(b, BF_IO_PENDING)) + _wait_io(b->cache); +} + +static unsigned _writeback(struct bcache *cache, unsigned count) +{ + unsigned actual = 0; + struct block *b, *tmp; + + dm_list_iterate_items_gen_safe (b, tmp, &cache->dirty, list) { + if (actual == count) + break; + + // We can't writeback anything that's still in use. + if (!b->ref_count) { + _issue_write(b); + actual++; + } + } + + return actual; +} + +/*---------------------------------------------------------------- + * High level allocation + *--------------------------------------------------------------*/ + +static struct block *_find_unused_clean_block(struct bcache *cache) +{ + struct block *b; + + dm_list_iterate_items (b, &cache->clean) { + if (!b->ref_count) { + _unlink_block(b); + _hash_remove(b); + return b; + } + } + + return NULL; +} + +static struct block *_new_block(struct bcache *cache, block_address index) +{ + struct block *b; + + b = _alloc_block(cache); + while (!b && cache->nr_locked < cache->nr_cache_blocks) { + b = _find_unused_clean_block(cache); + if (!b) { + if (dm_list_empty(&cache->io_pending)) + _writeback(cache, 16); + _wait_io(cache); + } + } + + if (b) { + dm_list_init(&b->list); + dm_list_init(&b->hash); + b->flags = 0; + b->index = index; + b->ref_count = 0; + b->error = 0; + + _hash_insert(b); + } + + return b; +} + +/*---------------------------------------------------------------- + * Block reference counting + *--------------------------------------------------------------*/ +static void _zero_block(struct block *b) +{ + b->cache->write_zeroes++; + memset(b->data, 0, b->cache->block_sectors << SECTOR_SHIFT); + _set_flags(b, BF_DIRTY); +} + +static void _hit(struct block *b, unsigned flags) +{ + struct bcache *cache = b->cache; + + if (flags & (GF_ZERO | GF_DIRTY)) + cache->write_hits++; + else + cache->read_hits++; + + _relink(b); +} + +static void _miss(struct bcache *cache, unsigned flags) +{ + if (flags & (GF_ZERO | GF_DIRTY)) + cache->write_misses++; + else + cache->read_misses++; +} + +static struct block *_lookup_or_read_block(struct bcache *cache, + int fd, block_address index, + unsigned flags) +{ + struct block *b = _hash_lookup(cache, fd, index); + + if (b) { + // FIXME: this is insufficient. We need to also catch a read + // lock of a write locked block. Ref count needs to distinguish. + if (b->ref_count && (flags & (GF_DIRTY | GF_ZERO))) { + log_err("concurrent write lock attempted"); + return NULL; + } + + if (_test_flags(b, BF_IO_PENDING)) { + _miss(cache, flags); + _wait_specific(b); + + } else + _hit(b, flags); + + _unlink_block(b); + + if (flags & GF_ZERO) + _zero_block(b); + + } else { + _miss(cache, flags); + + b = _new_block(cache, index); + if (b) { + if (flags & GF_ZERO) + _zero_block(b); + + else { + _issue_read(b); + _wait_specific(b); + + // we know the block is clean and unerrored. + _unlink_block(b); + } + } + } + + if (b && !b->error) { + if (flags & (GF_DIRTY | GF_ZERO)) + _set_flags(b, BF_DIRTY); + + _link_block(b); + return b; + } + + return NULL; +} + +static void _preemptive_writeback(struct bcache *cache) +{ + // FIXME: this ignores those blocks that are in the error state. Track + // nr_clean instead? + unsigned nr_available = cache->nr_cache_blocks - (cache->nr_dirty - cache->nr_io_pending); + if (nr_available < (WRITEBACK_LOW_THRESHOLD_PERCENT * cache->nr_cache_blocks / 100)) + _writeback(cache, (WRITEBACK_HIGH_THRESHOLD_PERCENT * cache->nr_cache_blocks / 100) - nr_available); + +} + +/*---------------------------------------------------------------- + * Public interface + *--------------------------------------------------------------*/ +struct bcache *bcache_create(sector_t block_sectors, unsigned nr_cache_blocks) +{ + int r; + struct bcache *cache; + + cache = malloc(sizeof(*cache)); + if (!cache) + return NULL; + + cache->block_sectors = block_sectors; + cache->nr_cache_blocks = nr_cache_blocks; + + cache->engine = _engine_create(nr_cache_blocks < 1024u ? nr_cache_blocks : 1024u); + if (!cache->engine) { + free(cache); + return NULL; + } + + cache->nr_locked = 0; + cache->nr_dirty = 0; + cache->nr_io_pending = 0; + + dm_list_init(&cache->free); + dm_list_init(&cache->errored); + dm_list_init(&cache->dirty); + dm_list_init(&cache->clean); + dm_list_init(&cache->io_pending); + + if (_hash_table_init(cache, nr_cache_blocks)) { + _engine_destroy(cache->engine); + free(cache); + } + + cache->read_hits = 0; + cache->read_misses = 0; + cache->write_zeroes = 0; + cache->write_hits = 0; + cache->write_misses = 0; + cache->prefetches = 0; + + r = _init_free_list(cache, nr_cache_blocks); + if (r) { + _engine_destroy(cache->engine); + _hash_table_exit(cache); + free(cache); + } + + return cache; +} + +void bcache_destroy(struct bcache *cache) +{ + if (cache->nr_locked) + log_warn("some blocks are still locked\n"); + + bcache_flush(cache); + _wait_all(cache); + _exit_free_list(cache); + _hash_table_exit(cache); + _engine_destroy(cache->engine); + free(cache); +} + +void bcache_prefetch(struct bcache *cache, int fd, block_address index) +{ + struct block *b = _hash_lookup(cache, fd, index); + + if (!b) { + cache->prefetches++; + + b = _new_block(cache, index); + if (b) + _issue_read(b); + } +} + +bool bcache_get(struct bcache *cache, int fd, block_address index, + unsigned flags, struct block **result) +{ + struct block *b = _lookup_or_read_block(cache, fd, index, flags); + if (b) { + if (!b->ref_count) + cache->nr_locked++; + b->ref_count++; + + *result = b; + return true; + } + + *result = NULL; + log_err("couldn't get block"); + return false; +} + +void bcache_put(struct block *b) +{ + if (!b->ref_count) { + log_err("ref count on bcache block already zero"); + return; + } + + b->ref_count--; + if (!b->ref_count) + b->cache->nr_locked--; + + if (_test_flags(b, BF_DIRTY)) + _preemptive_writeback(b->cache); +} + +int bcache_flush(struct bcache *cache) +{ + while (!dm_list_empty(&cache->dirty)) { + struct block *b = dm_list_item(_list_pop(&cache->dirty), struct block); + if (b->ref_count || _test_flags(b, BF_IO_PENDING)) + // The superblock may well be still locked. + continue; + + _issue_write(b); + } + + _wait_all(cache); + + return dm_list_empty(&cache->errored) ? 0 : -EIO; +} + +//---------------------------------------------------------------- + diff --git a/lib/device/bcache.h b/lib/device/bcache.h new file mode 100644 index 000000000..1f4262e96 --- /dev/null +++ b/lib/device/bcache.h @@ -0,0 +1,83 @@ +#ifndef BCACHE_H +#define BCACHE_H + +#include + +#include "libdevmapper.h" + +/*----------------------------------------------------------------*/ + +typedef uint64_t block_address; +typedef uint64_t sector_t; + +struct bcache; +struct block { + /* clients may only access these three fields */ + int fd; + uint64_t index; + void *data; + + struct bcache *cache; + struct dm_list list; + struct dm_list hash; + + unsigned flags; + unsigned ref_count; + int error; +}; + +struct bcache *bcache_create(sector_t block_size, unsigned nr_cache_blocks); +void bcache_destroy(struct bcache *cache); + +enum bcache_get_flags { + /* + * The block will be zeroed before get_block returns it. This + * potentially avoids a read if the block is not already in the cache. + * GF_DIRTY is implicit. + */ + GF_ZERO = (1 << 0), + + /* + * Indicates the caller is intending to change the data in the block, a + * writeback will occur after the block is released. + */ + GF_DIRTY = (1 << 1) +}; + +typedef uint64_t block_address; + +unsigned bcache_get_max_prefetches(struct bcache *cache); + +/* + * Use the prefetch method to take advantage of asynchronous IO. For example, + * if you wanted to read a block from many devices concurrently you'd do + * something like this: + * + * dm_list_iterate_items (dev, &devices) + * bcache_prefetch(cache, dev->fd, block); + * + * dm_list_iterate_items (dev, &devices) { + * if (!bcache_get(cache, dev->fd, block, &b)) + * fail(); + * + * process_block(b); + * } + * + * It's slightly sub optimal, since you may not run the gets in the order that + * they complete. But we're talking a very small difference, and it's worth it + * to keep callbacks out of this interface. + */ +void bcache_prefetch(struct bcache *cache, int fd, block_address index); + +/* + * Returns true on success. + */ +bool bcache_get(struct bcache *cache, int fd, block_address index, + unsigned flags, struct block **result); +void bcache_put(struct block *b); + +int bcache_flush(struct bcache *cache); + +/*----------------------------------------------------------------*/ + +#endif