Return version number from reads

blocking-uring-test
Vitaliy Filippov 2020-02-12 12:30:50 +03:00
parent 1100ad44bd
commit b7ccd63104
6 changed files with 26 additions and 7 deletions

View File

@ -40,7 +40,7 @@ void blockstore_t::enqueue_op_first(blockstore_op_t *op)
impl->enqueue_op(op, true); impl->enqueue_op(op, true);
} }
std::map<object_id, uint64_t> & blockstore_t::get_unstable_writes() std::unordered_map<object_id, uint64_t> & blockstore_t::get_unstable_writes()
{ {
return impl->unstable_writes; return impl->unstable_writes;
} }

View File

@ -55,8 +55,15 @@ struct blockstore_op_t
std::function<void (blockstore_op_t*)> callback; std::function<void (blockstore_op_t*)> callback;
// For reads, writes & deletes: oid is the requested object // For reads, writes & deletes: oid is the requested object
object_id oid; object_id oid;
// For reads: version=0 -> last stable, version=UINT64_MAX -> last unstable, version=X -> specific version // For reads:
// For writes & deletes: a new version is assigned automatically // version == 0 -> read the last stable version,
// version == UINT64_MAX -> read the last version,
// otherwise -> read the newest version that is <= the specified version
// after execution, version is equal to the version that was read from the blockstore
// For writes & deletes:
// if version == 0, a new version is assigned automatically
// if version != 0, it is assigned for the new write if possible, otherwise -EINVAL is returned
// after execution, version is equal to the version that was written to the blockstore
uint64_t version; uint64_t version;
// For reads & writes: offset & len are the requested part of the object, buf is the buffer // For reads & writes: offset & len are the requested part of the object, buf is the buffer
uint32_t offset; uint32_t offset;
@ -102,7 +109,7 @@ public:
void enqueue_op_first(blockstore_op_t *op); void enqueue_op_first(blockstore_op_t *op);
// Unstable writes are added here (map of object_id -> version) // Unstable writes are added here (map of object_id -> version)
std::map<object_id, uint64_t> & get_unstable_writes(); std::unordered_map<object_id, uint64_t> & get_unstable_writes();
// FIXME rename to object_size // FIXME rename to object_size
uint32_t get_block_size(); uint32_t get_block_size();

View File

@ -309,7 +309,7 @@ public:
void enqueue_op(blockstore_op_t *op, bool first = false); void enqueue_op(blockstore_op_t *op, bool first = false);
// Unstable writes are added here (map of object_id -> version) // Unstable writes are added here (map of object_id -> version)
std::map<object_id, uint64_t> unstable_writes; std::unordered_map<object_id, uint64_t> unstable_writes;
inline uint32_t get_block_size() { return block_size; } inline uint32_t get_block_size() { return block_size; }
inline uint64_t get_block_count() { return block_count; } inline uint64_t get_block_count() { return block_count; }

View File

@ -92,12 +92,14 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
{ {
// region is not allocated - return zeroes // region is not allocated - return zeroes
memset(read_op->buf, 0, read_op->len); memset(read_op->buf, 0, read_op->len);
read_op->version = 0;
read_op->retval = read_op->len; read_op->retval = read_op->len;
read_op->callback(read_op); read_op->callback(read_op);
return 1; return 1;
} }
uint64_t fulfilled = 0; uint64_t fulfilled = 0;
PRIV(read_op)->pending_ops = 0; PRIV(read_op)->pending_ops = 0;
uint64_t result_version = 0;
if (dirty_found) if (dirty_found)
{ {
while (dirty_it->first.oid == read_op->oid) while (dirty_it->first.oid == read_op->oid)
@ -112,6 +114,10 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
} }
if (version_ok) if (version_ok)
{ {
if (!result_version)
{
result_version = dirty_it->first.version;
}
if (!fulfill_read(read_op, fulfilled, dirty.offset, dirty.offset + dirty.len, if (!fulfill_read(read_op, fulfilled, dirty.offset, dirty.offset + dirty.len,
dirty.state, dirty_it->first.version, dirty.location + (IS_JOURNAL(dirty.state) ? 0 : dirty.offset))) dirty.state, dirty_it->first.version, dirty.location + (IS_JOURNAL(dirty.state) ? 0 : dirty.offset)))
{ {
@ -129,6 +135,10 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
} }
if (clean_it != clean_db.end() && fulfilled < read_op->len) if (clean_it != clean_db.end() && fulfilled < read_op->len)
{ {
if (!result_version)
{
result_version = clean_it->second.version;
}
if (!clean_entry_bitmap_size) if (!clean_entry_bitmap_size)
{ {
if (!fulfill_read(read_op, fulfilled, 0, block_size, ST_CURRENT, 0, clean_it->second.location)) if (!fulfill_read(read_op, fulfilled, 0, block_size, ST_CURRENT, 0, clean_it->second.location))
@ -190,6 +200,7 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
fulfill_read(read_op, fulfilled, 0, block_size, ST_DEL_STABLE, 0, 0); fulfill_read(read_op, fulfilled, 0, block_size, ST_DEL_STABLE, 0, 0);
} }
assert(fulfilled == read_op->len); assert(fulfilled == read_op->len);
read_op->version = result_version;
if (!PRIV(read_op)->pending_ops) if (!PRIV(read_op)->pending_ops)
{ {
// everything is fulfilled from memory // everything is fulfilled from memory

View File

@ -146,7 +146,8 @@ void osd_t::make_reply(osd_op_t *op)
op->reply.hdr.retval = op->bs_op.retval; op->reply.hdr.retval = op->bs_op.retval;
if (op->op.hdr.opcode == OSD_OP_SECONDARY_LIST) if (op->op.hdr.opcode == OSD_OP_SECONDARY_LIST)
op->reply.sec_list.stable_count = op->bs_op.version; op->reply.sec_list.stable_count = op->bs_op.version;
else if (op->op.hdr.opcode == OSD_OP_SECONDARY_WRITE) else if (op->op.hdr.opcode == OSD_OP_SECONDARY_READ ||
op->op.hdr.opcode == OSD_OP_SECONDARY_WRITE)
op->reply.sec_rw.version = op->bs_op.version; op->reply.sec_rw.version = op->bs_op.version;
else if (op->op.hdr.opcode == OSD_OP_SECONDARY_DELETE) else if (op->op.hdr.opcode == OSD_OP_SECONDARY_DELETE)
op->reply.sec_del.version = op->bs_op.version; op->reply.sec_del.version = op->bs_op.version;

View File

@ -69,7 +69,7 @@ struct __attribute__((__packed__)) osd_op_secondary_rw_t
struct __attribute__((__packed__)) osd_reply_secondary_rw_t struct __attribute__((__packed__)) osd_reply_secondary_rw_t
{ {
osd_reply_header_t header; osd_reply_header_t header;
// for writes: assigned version number // for reads and writes: assigned or read version number
uint64_t version; uint64_t version;
}; };