github/ocf
4
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Dynamic I/O queue management

Browse Source 
- Queue allocation is now separated from starting cache.
- Queue can be created and destroyed in runtime.
- All queue ops accept queue handle instead of queue id.
- Cache stores queues as list instead of array.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
This commit is contained in:
Michal Mielewczyk
2019-02-15 08:12:00 -05:00
committed by Robert Baldyga
parent 1771228a46
commit e53944d472
38 changed files with 379 additions and 445 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/eviction/eviction.c
View File

@@ -40,8 +40,8 @@ static uint32_t ocf_evict_calculate(struct ocf_user_part *part,
return to_evict;
}
static inline uint32_t ocf_evict_do(struct ocf_cache *cache,
uint32_t io_queue, const uint32_t evict_cline_no,
static inline uint32_t ocf_evict_do(ocf_cache_t cache,
ocf_queue_t io_queue, const uint32_t evict_cline_no,
ocf_core_id_t core_id, ocf_part_id_t target_part_id)
{
uint32_t to_evict = 0, evicted = 0;

22
src/eviction/eviction.h
View File

@@ -32,22 +32,22 @@ union eviction_policy_meta {
* set core_id to -2 to purge the whole cache partition
*/
struct eviction_policy_ops {
void (*init_cline)(struct ocf_cache *cache,
void (*init_cline)(ocf_cache_t cache,
ocf_cache_line_t cline);
void (*rm_cline)(struct ocf_cache *cache,
void (*rm_cline)(ocf_cache_t cache,
ocf_cache_line_t cline);
bool (*can_evict)(struct ocf_cache *cache);
uint32_t (*req_clines)(struct ocf_cache *cache,
uint32_t io_queue, ocf_part_id_t part_id,
bool (*can_evict)(ocf_cache_t cache);
uint32_t (*req_clines)(ocf_cache_t cache,
ocf_queue_t io_queue, ocf_part_id_t part_id,
uint32_t cline_no, ocf_core_id_t core_id);
void (*hot_cline)(struct ocf_cache *cache,
void (*hot_cline)(ocf_cache_t cache,
ocf_cache_line_t cline);
void (*init_evp)(struct ocf_cache *cache,
void (*init_evp)(ocf_cache_t cache,
ocf_part_id_t part_id);
void (*dirty_cline)(struct ocf_cache *cache,
void (*dirty_cline)(ocf_cache_t cache,
ocf_part_id_t part_id,
uint32_t cline_no);
void (*clean_cline)(struct ocf_cache *cache,
void (*clean_cline)(ocf_cache_t cache,
ocf_part_id_t part_id,
uint32_t cline_no);
const char *name;
@@ -62,9 +62,9 @@ extern struct eviction_policy_ops evict_policy_ops[ocf_eviction_max];
* or the destination partition ID for the free buffers
* (it matches label and is part of the object (#core_id) IO group)
*/
int space_managment_evict_do(struct ocf_cache *cache,
int space_managment_evict_do(ocf_cache_t cache,
struct ocf_request *req, uint32_t evict_cline_no);
int space_management_free(struct ocf_cache *cache, uint32_t count);
int space_management_free(ocf_cache_t cache, uint32_t count);
#endif

36
src/eviction/lru.c
View File

@@ -29,7 +29,7 @@
#define is_lru_tail(x) (x == collision_table_entries)
/* Sets the given collision_index as the new _head_ of the LRU list. */
static inline void update_lru_head(struct ocf_cache *cache,
static inline void update_lru_head(ocf_cache_t cache,
int partition_id, unsigned int collision_index,
int cline_dirty)
{
@@ -43,7 +43,7 @@ static inline void update_lru_head(struct ocf_cache *cache,
}
/* Sets the given collision_index as the new _tail_ of the LRU list. */
static inline void update_lru_tail(struct ocf_cache *cache,
static inline void update_lru_tail(ocf_cache_t cache,
int partition_id, unsigned int collision_index,
int cline_dirty)
{
@@ -58,7 +58,7 @@ static inline void update_lru_tail(struct ocf_cache *cache,
/* Sets the given collision_index as the new _head_ and _tail_ of
* the LRU list.
*/
static inline void update_lru_head_tail(struct ocf_cache *cache,
static inline void update_lru_head_tail(ocf_cache_t cache,
int partition_id, unsigned int collision_index, int cline_dirty)
{
update_lru_head(cache, partition_id, collision_index, cline_dirty);
@@ -66,7 +66,7 @@ static inline void update_lru_head_tail(struct ocf_cache *cache,
}
/* Adds the given collision_index to the _head_ of the LRU list */
static void add_lru_head(struct ocf_cache *cache, int partition_id,
static void add_lru_head(ocf_cache_t cache, int partition_id,
unsigned int collision_index, int cline_dirty)
{
unsigned int curr_head_index;
@@ -121,7 +121,7 @@ static void add_lru_head(struct ocf_cache *cache, int partition_id,
}
/* Deletes the node with the given collision_index from the lru list */
static void remove_lru_list(struct ocf_cache *cache, int partition_id,
static void remove_lru_list(ocf_cache_t cache, int partition_id,
unsigned int collision_index, int cline_dirty)
{
int is_clean_head = 0, is_clean_tail = 0, is_dirty_head = 0, is_dirty_tail = 0;
@@ -250,7 +250,7 @@ static void remove_lru_list(struct ocf_cache *cache, int partition_id,
/*-- End of LRU functions*/
void evp_lru_init_cline(struct ocf_cache *cache, ocf_cache_line_t cline)
void evp_lru_init_cline(ocf_cache_t cache, ocf_cache_line_t cline)
{
union eviction_policy_meta eviction;
@@ -264,7 +264,7 @@ void evp_lru_init_cline(struct ocf_cache *cache, ocf_cache_line_t cline)
/* the caller must hold the metadata lock */
void evp_lru_rm_cline(struct ocf_cache *cache, ocf_cache_line_t cline)
void evp_lru_rm_cline(ocf_cache_t cache, ocf_cache_line_t cline)
{
ocf_part_id_t part_id = ocf_metadata_get_partition_id(cache, cline);
@@ -278,7 +278,7 @@ static void evp_lru_clean_end(void *private_data, int error)
env_atomic_set(cleaning_in_progress, 0);
}
static int evp_lru_clean_getter(struct ocf_cache *cache,
static int evp_lru_clean_getter(ocf_cache_t cache,
void *getter_context, uint32_t item, ocf_cache_line_t *line)
{
union eviction_policy_meta eviction;
@@ -306,7 +306,7 @@ static int evp_lru_clean_getter(struct ocf_cache *cache,
return -1;
}
static void evp_lru_clean(struct ocf_cache *cache, uint32_t io_queue,
static void evp_lru_clean(ocf_cache_t cache, ocf_queue_t io_queue,
ocf_part_id_t part_id, uint32_t count)
{
env_atomic *progress = &cache->cleaning[part_id];
@@ -342,7 +342,7 @@ static void evp_lru_zero_line_complete(struct ocf_request *ocf_req, int error)
env_atomic_dec(&ocf_req->cache->pending_eviction_clines);
}
static void evp_lru_zero_line(struct ocf_cache *cache, uint32_t io_queue,
static void evp_lru_zero_line(ocf_cache_t cache, ocf_queue_t io_queue,
ocf_cache_line_t line)
{
struct ocf_request *req;
@@ -352,11 +352,11 @@ static void evp_lru_zero_line(struct ocf_cache *cache, uint32_t io_queue,
ocf_metadata_get_core_info(cache, line, &id, &core_line);
addr = core_line * ocf_line_size(cache);
req = ocf_req_new(cache, id, addr, ocf_line_size(cache), OCF_WRITE);
req = ocf_req_new(io_queue, &cache->core[id], addr,
ocf_line_size(cache), OCF_WRITE);
if (req) {
req->info.internal = true;
req->complete = evp_lru_zero_line_complete;
req->io_queue = io_queue;
env_atomic_inc(&cache->pending_eviction_clines);
@@ -364,7 +364,7 @@ static void evp_lru_zero_line(struct ocf_cache *cache, uint32_t io_queue,
}
}
bool evp_lru_can_evict(struct ocf_cache *cache)
bool evp_lru_can_evict(ocf_cache_t cache)
{
if (env_atomic_read(&cache->pending_eviction_clines) >=
OCF_PENDING_EVICTION_LIMIT) {
@@ -375,7 +375,7 @@ bool evp_lru_can_evict(struct ocf_cache *cache)
}
/* the caller must hold the metadata lock */
uint32_t evp_lru_req_clines(struct ocf_cache *cache, uint32_t io_queue,
uint32_t evp_lru_req_clines(ocf_cache_t cache, ocf_queue_t io_queue,
ocf_part_id_t part_id, uint32_t cline_no, ocf_core_id_t core_id)
{
uint32_t i;
@@ -438,7 +438,7 @@ uint32_t evp_lru_req_clines(struct ocf_cache *cache, uint32_t io_queue,
}
/* the caller must hold the metadata lock */
void evp_lru_hot_cline(struct ocf_cache *cache, ocf_cache_line_t cline)
void evp_lru_hot_cline(ocf_cache_t cache, ocf_cache_line_t cline)
{
ocf_part_id_t part_id = ocf_metadata_get_partition_id(cache, cline);
struct ocf_user_part *part = &cache->user_parts[part_id];
@@ -469,7 +469,7 @@ void evp_lru_hot_cline(struct ocf_cache *cache, ocf_cache_line_t cline)
add_lru_head(cache, part_id, cline, cline_dirty);
}
void evp_lru_init_evp(struct ocf_cache *cache, ocf_part_id_t part_id)
void evp_lru_init_evp(ocf_cache_t cache, ocf_part_id_t part_id)
{
unsigned int collision_table_entries =
cache->device->collision_table_entries;
@@ -483,7 +483,7 @@ void evp_lru_init_evp(struct ocf_cache *cache, ocf_part_id_t part_id)
part->runtime->eviction.policy.lru.dirty_tail = collision_table_entries;
}
void evp_lru_clean_cline(struct ocf_cache *cache, ocf_part_id_t part_id,
void evp_lru_clean_cline(ocf_cache_t cache, ocf_part_id_t part_id,
uint32_t cline)
{
OCF_METADATA_EVICTION_LOCK();
@@ -492,7 +492,7 @@ void evp_lru_clean_cline(struct ocf_cache *cache, ocf_part_id_t part_id,
OCF_METADATA_EVICTION_UNLOCK();
}
void evp_lru_dirty_cline(struct ocf_cache *cache, ocf_part_id_t part_id,
void evp_lru_dirty_cline(ocf_cache_t cache, ocf_part_id_t part_id,
uint32_t cline)
{
OCF_METADATA_EVICTION_LOCK();

2
src/eviction/lru.h
View File

@@ -12,7 +12,7 @@ void evp_lru_init_cline(struct ocf_cache *cache,
ocf_cache_line_t cline);
void evp_lru_rm_cline(struct ocf_cache *cache, ocf_cache_line_t cline);
bool evp_lru_can_evict(struct ocf_cache *cache);
uint32_t evp_lru_req_clines(struct ocf_cache *cache, uint32_t io_queue,
uint32_t evp_lru_req_clines(struct ocf_cache *cache, ocf_queue_t io_queue,
ocf_part_id_t part_id, uint32_t cline_no,
ocf_core_id_t core_id);
void evp_lru_hot_cline(struct ocf_cache *cache, ocf_cache_line_t cline);

2
src/eviction/ops.h
View File

@@ -53,7 +53,7 @@ static inline bool ocf_eviction_can_evict(struct ocf_cache *cache)
}
static inline uint32_t ocf_eviction_need_space(struct ocf_cache *cache,
uint32_t io_queue, ocf_part_id_t part_id, uint32_t clines,
ocf_queue_t io_queue, ocf_part_id_t part_id, uint32_t clines,
ocf_core_id_t core_id)
{
uint8_t type;