github/ocf
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Change hash function

Browse Source 
Modify ocf_metadata_hash_func to return consecutive (modulo @hash_table_entries)
values for consecutive @core_line_num. This way it is trivial to sort all
core lines within a single request according to their hash value. This kind
of sorting will be required to assure that future hash bucket metadata locks
are always acquired in fixed order, eliminating the risk of dead locks.

This change is part of fine granularity metadata lock implementation.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
This commit is contained in:
Adam Rutkowski 2019-07-19 17:15:54 -04:00
parent 7184f7787c
commit 3dd4263bc8
2 changed files with 10 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/metadata/metadata_hash.c
View File

@ -83,8 +83,7 @@ static ocf_cache_line_t ocf_metadata_hash_get_entries(
return cache_lines;
case metadata_segment_hash:
return OCF_DIV_ROUND_UP(cache_lines / 4, OCF_HASH_PRIME) *
OCF_HASH_PRIME - 1;
return OCF_DIV_ROUND_UP(cache_lines, 4);
case metadata_segment_sb_config:
return OCF_DIV_ROUND_UP(sizeof(struct ocf_superblock_config),

20
src/metadata/metadata_misc.h
View File

@ -6,21 +6,19 @@
#ifndef __METADATA_MISC_H__
#define __METADATA_MISC_H__
/*
* Hash function needs number that has no common factors with both number
* of cores and number of entries in metadata hash container (hash lists).
* This can be easily achived by picking prime which is bigger than maximum
* number of cores and ensuring that count of hash table entries is not
* divisible by this number. Let's choose 4099, which is smallest prime
* greater than OCF_CORE_MAX (which is 4096).
/* Hash function intentionally returns consecutive (modulo @hash_table_entries)
* values for consecutive @core_line_num. This way it is trivial to sort all
* core lines within a single request in ascending hash value order. This kind
* of sorting is required to assure that (future) hash bucket metadata locks are
* always acquired in fixed order, eliminating the risk of dead locks.
*/
#define OCF_HASH_PRIME 4099
static inline ocf_cache_line_t ocf_metadata_hash_func(ocf_cache_t cache,
uint64_t core_line_num, ocf_core_id_t core_id)
{
return (ocf_cache_line_t) ((core_line_num * OCF_HASH_PRIME + core_id) %
cache->device->hash_table_entries);
const unsigned int entries = cache->device->hash_table_entries;
return (ocf_cache_line_t) ((core_line_num + (core_id * (entries / 32)))
% entries);
}
void ocf_metadata_sparse_cache_line(struct ocf_cache *cache,