/* * Copyright(c) 2012-2018 Intel Corporation * SPDX-License-Identifier: BSD-3-Clause-Clear */ #include "ocf/ocf.h" #include "../ocf_cache_priv.h" #include "cache_engine.h" #include "engine_common.h" #include "engine_discard.h" #include "../metadata/metadata.h" #include "../utils/utils_rq.h" #include "../utils/utils_io.h" #include "../utils/utils_cache_line.h" #include "../concurrency/ocf_concurrency.h" #define OCF_ENGINE_DEBUG 0 #define OCF_ENGINE_DEBUG_IO_NAME "discard" #include "engine_debug.h" static int _ocf_discard_step_do(struct ocf_request *rq); static int _ocf_discard_step(struct ocf_request *rq); static int _ocf_discard_flush_cache(struct ocf_request *rq); static int _ocf_discard_core(struct ocf_request *rq); static const struct ocf_io_if _io_if_discard_step = { .read = _ocf_discard_step, .write = _ocf_discard_step }; static const struct ocf_io_if _io_if_discard_step_resume = { .read = _ocf_discard_step_do, .write = _ocf_discard_step_do }; static const struct ocf_io_if _io_if_discard_flush_cache = { .read = _ocf_discard_flush_cache, .write = _ocf_discard_flush_cache, }; static const struct ocf_io_if _io_if_discard_core = { .read = _ocf_discard_core, .write = _ocf_discard_core }; static void _ocf_discard_complete_rq(struct ocf_request *req, int error) { req->complete(req, error); ocf_rq_put(req); } static void _ocf_discard_core_io(struct ocf_io *io, int error) { struct ocf_request *rq = io->priv1; OCF_DEBUG_RQ(rq, "Core DISCARD Completion"); _ocf_discard_complete_rq(rq, error); } static int _ocf_discard_core(struct ocf_request *req) { struct ocf_cache *cache = req->cache; struct ocf_io *io; io = ocf_dobj_new_io(&cache->core_obj[req->core_id].obj); if (!io) { _ocf_discard_complete_rq(req, -ENOMEM); return -ENOMEM; } ocf_io_configure(io, SECTORS_TO_BYTES(req->discard.sector), SECTORS_TO_BYTES(req->discard.nr_sects), OCF_WRITE, 0, 0); ocf_io_set_cmpl(io, req, NULL, _ocf_discard_core_io); ocf_io_set_data(io, req->data, 0); ocf_dobj_submit_discard(io); return 0; } static void _ocf_discard_cache_flush_io_cmpl(struct ocf_io *io, int error) { struct ocf_request *rq = io->priv1; if (error) { ocf_metadata_error(rq->cache); _ocf_discard_complete_rq(rq, error); return; } rq->io_if = &_io_if_discard_core; ocf_engine_push_rq_front(rq, true); } static int _ocf_discard_flush_cache(struct ocf_request *req) { struct ocf_io *io; io = ocf_dobj_new_io(&req->cache->device->obj); if (!io) { ocf_metadata_error(req->cache); _ocf_discard_complete_rq(req, -ENOMEM); return -ENOMEM; } ocf_io_configure(io, 0, 0, OCF_WRITE, 0, 0); ocf_io_set_cmpl(io, req, NULL, _ocf_discard_cache_flush_io_cmpl); ocf_dobj_submit_flush(io); return 0; } static void _ocf_discard_finish_step(struct ocf_request *rq) { rq->discard.handled += BYTES_TO_SECTORS(rq->byte_length); if (rq->discard.handled < rq->discard.nr_sects) rq->io_if = &_io_if_discard_step; else if (rq->cache->device->init_mode != ocf_init_mode_metadata_volatile) rq->io_if = &_io_if_discard_flush_cache; else rq->io_if = &_io_if_discard_core; ocf_engine_push_rq_front(rq, true); } static void _ocf_discard_step_io(struct ocf_request *rq, int error) { if (error) rq->error |= error; if (env_atomic_dec_return(&rq->req_remaining)) return; OCF_DEBUG_RQ(rq, "Completion"); /* Release WRITE lock of request */ ocf_rq_unlock_wr(rq); if (rq->error) { ocf_metadata_error(rq->cache); _ocf_discard_complete_rq(rq, rq->error); return; } _ocf_discard_finish_step(rq); } int _ocf_discard_step_do(struct ocf_request *rq) { struct ocf_cache *cache = rq->cache; /* Get OCF request - increase reference counter */ ocf_rq_get(rq); env_atomic_set(&rq->req_remaining, 1); /* One core IO */ if (ocf_engine_mapped_count(rq)) { /* There are mapped cache line, need to remove them */ OCF_METADATA_LOCK_WR(); /*- Metadata WR access ---------------*/ /* Remove mapped cache lines from metadata */ ocf_purge_map_info(rq); if (rq->info.flush_metadata) { /* Request was dirty and need to flush metadata */ ocf_metadata_flush_do_asynch(cache, rq, _ocf_discard_step_io); } OCF_METADATA_UNLOCK_WR(); /*- END Metadata WR access ---------*/ } OCF_DEBUG_RQ(rq, "Discard"); _ocf_discard_step_io(rq, 0); /* Put OCF request - decrease reference counter */ ocf_rq_put(rq); return 0; } static void _ocf_discard_on_resume(struct ocf_request *rq) { OCF_DEBUG_RQ(rq, "On resume"); ocf_engine_push_rq_front(rq, true); } static int _ocf_discard_step(struct ocf_request *rq) { int lock; struct ocf_cache *cache = rq->cache; OCF_DEBUG_TRACE(rq->cache); rq->byte_position = SECTORS_TO_BYTES(rq->discard.sector + rq->discard.handled); rq->byte_length = MIN(SECTORS_TO_BYTES(rq->discard.nr_sects - rq->discard.handled), MAX_TRIM_RQ_SIZE); rq->core_line_first = ocf_bytes_2_lines(cache, rq->byte_position); rq->core_line_last = ocf_bytes_2_lines(cache, rq->byte_position + rq->byte_length - 1); rq->core_line_count = rq->core_line_last - rq->core_line_first + 1; rq->io_if = &_io_if_discard_step_resume; OCF_METADATA_LOCK_RD(); /*- Metadata READ access, No eviction --------*/ ENV_BUG_ON(env_memset(rq->map, sizeof(*rq->map) * rq->core_line_count, 0)); /* Travers to check if request is mapped fully */ ocf_engine_traverse(rq); if (ocf_engine_mapped_count(rq)) { /* Some cache line are mapped, lock request for WRITE access */ lock = ocf_rq_trylock_wr(rq); } else { lock = OCF_LOCK_ACQUIRED; } OCF_METADATA_UNLOCK_RD(); /*- END Metadata READ access----------------*/ if (lock >= 0) { if (OCF_LOCK_ACQUIRED == lock) { _ocf_discard_step_do(rq); } else { /* WR lock was not acquired, need to wait for resume */ OCF_DEBUG_RQ(rq, "NO LOCK") } } else { OCF_DEBUG_RQ(rq, "LOCK ERROR %d", lock); rq->error |= lock; _ocf_discard_finish_step(rq); } env_cond_resched(); return 0; } int ocf_discard(struct ocf_request *rq) { OCF_DEBUG_TRACE(rq->cache); ocf_io_start(rq->io); if (rq->rw == OCF_READ) { rq->complete(rq, -EINVAL); return 0; } /* Get OCF request - increase reference counter */ ocf_rq_get(rq); /* Set resume call backs */ rq->resume = _ocf_discard_on_resume; _ocf_discard_step(rq); /* Put OCF request - decrease reference counter */ ocf_rq_put(rq); return 0; }