github/ocf
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1,146 Commits 1 Branch 1 Tag 4.1 MiB
be1ac09c17
Commit Graph

1146 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Robert Baldyga
9dba49192f
Merge pull request #462 from arutk/no_bits_lock 
Get rid of status bits lock
 2021-02-19 10:12:46 +01:00
Adam Rutkowski
c95f6358ab Get rid of status bits lock 
All the status bits operations are now protectec by
hash bucket locks

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-18 15:05:53 -06:00
Adam Rutkowski
cd9e42f987 Properly lock hash bucket for status bits operations 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-18 15:02:50 -06:00
Robert Baldyga
75baec5aa5
Merge pull request #456 from arutk/aalru 
Relax LRU list ordering to minimize list updates
 2021-02-18 13:48:54 +01:00
Robert Baldyga
91cbeed611
Merge pull request #461 from mmichal10/evict-pinned 
Evict pinned overflown ioclass
 2021-02-16 13:55:52 +01:00
Michal Mielewczyk
83f142c987 Functional test for overflown pinned ioclass 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-02-16 07:07:01 -05:00
Slawomir Jankowski
2741acc069 Create templates for issues 
Signed-off-by: Slawomir Jankowski <slawomir.jankowski@intel.com>
 2021-02-16 11:36:53 +01:00
Michal Mielewczyk
7f3f2ad115 Evict from overflown pinned ioclass 
If an ioclass is pinned but it exceeded its occupancy limit, it should be
evicted anyway.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-02-16 04:06:07 -05:00
Robert Baldyga
fe206a86ec
Merge pull request #452 from arutk/split_gml_master 
Split global metadata lock
 2021-02-15 18:10:36 +01:00
Adam Rutkowski
0748f33a9d Align each global metadata lock to 64B 
.. in order to move primitives intended to be accessed
concurrently in separate CPU cache line.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-15 11:27:49 -06:00
Adam Rutkowski
05780c98ed Split global metadata lock 
Divide single global lock instance into 4 to reduce contention
in multiple read-locks scenario.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-15 11:27:49 -06:00
Adam Rutkowski
10c3c3de36 Renaming hash bucket locking functions 
1. new abbreviated previx: ocf_hb (HB stands for hash bucket)
2. clear distinction between functions requiring caller to
   hold metadata shared global lock ("naked") vs the ones
   which acquire global lock on its own ("prot" for protected)
3. clear distinction between hash bucket locking functions
   accepting hash bucket id ("id"), core line and lba ("cline")
   and entire request ("req").

Resulting naming scheme:
ocf_hb_(id/cline/req)_(prot/naked)_(lock/unlock/trylock)_(rd/wr)

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-12 18:08:15 -06:00
Adam Rutkowski
c822c953ed Fix return status from hash bucket trylock wr 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-11 15:02:06 -06:00
Robert Baldyga
af8177d2ba
Merge pull request #458 from mmichal10/fix-cleaning 
Fix updating hot cachelines cleaning list
 2021-02-11 11:30:07 +01:00
Robert Baldyga
d03ea719cd
Merge pull request #451 from arutk/exact_evict_count 
only request evict size equal to request unmapped count
 2021-02-11 10:47:12 +01:00
Michal Mielewczyk
fa41d4fc88 Fix updating hot cachelines cleaning list 
Update cacheline's timestamp each time it's being written.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-02-10 10:02:57 -05:00
Adam Rutkowski
9e98eec361 Only acquire read lock to verify lru elem hotness 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 16:23:01 -06:00
Adam Rutkowski
c04bfa3962 Add macros to read lock eviction list 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 16:23:01 -06:00
Adam Rutkowski
9690f13bef Change eviction spin lock to RW lock 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 16:23:01 -06:00
Adam Rutkowski
1908707a3d LRU list unit tests 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 16:23:01 -06:00
Adam Rutkowski
b4daac11c2 Track hot items on LRU list 
Top 50% least recently used cachelines are not promoted
to list head upon access. Only after cacheline drops to
bottom 50% it is considered as a candidate to promote
to list head.

The purpose of this change is to reduce overhead of
LRU list maintanance for hot cachelines.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 16:22:55 -06:00
Adam Rutkowski
4276d65e5a unit tests for new eviction order 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 12:51:39 -06:00
Adam Rutkowski
746b32c47d Evict from overflown partitions first 
Overflown partitions now have precedence over others during
eviction, regardless of IO class priorities.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 12:51:39 -06:00
Adam Rutkowski
5538a5a95d Only request evict size equal to request unmapped count 
Removing the logic for oportunistic partition overflow
reduction by evicting more cachelines than actually
required by the request being serviced.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-02-09 11:11:15 -06:00
Robert Baldyga
fd26735157
Merge pull request #450 from mmichal10/ioclass-stats-fix 
Reset per-partition counters when adding core
 2021-02-03 13:13:30 +01:00
Michal Mielewczyk
93eccc862a Reset per-partition counters when adding core 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-02-03 06:18:44 -05:00
Robert Baldyga
3543f5c5cc
Merge pull request #443 from rafalste/update_copyright 
Update copyright statements (2021)
 2021-02-03 11:59:39 +01:00
Robert Baldyga
3ddefc9b59
Merge pull request #446 from mmichal10/lock-on-hit 
Don't evict on hit
 2021-02-02 15:18:22 +01:00
Michal Mielewczyk
3a7b55c4c2 Don't evict on hit 
If request is hit, simply try to acquire cachelines instead of verifying
whether target partition's size is not exceeded.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-01-29 17:15:32 -05:00
Rafal Stefanowski
6ed4cf8a24 Update copyright statements (2021) 
Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
 2021-01-21 13:17:34 +01:00
Robert Baldyga
7f72a9431c
Merge pull request #442 from arutk/cleaner_page_lock 
Add missing collision page lock in cleaner
 2021-01-21 09:25:21 +01:00
Adam Rutkowski
012438c279 Add missing collision page lock in cleaner 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-01-20 19:28:41 -06:00
Robert Baldyga
fe56d719d7
Merge pull request #440 from robertbaldyga/flush-collision-on-core-remove 
Flush metadata collision segment on core remove
 2021-01-19 15:11:11 +01:00
Robert Baldyga
5a88ab2d61 Flush metadata collision segment on core remove 
If there is any dirty data on the cache associated with removed core,
we must flush collision metadata after removing core to make metadata
persistent in case of dirty shutdown.

This fixes the problem when recovery procedure erroneously interprets
cache lines that belonged to removed core as valid ones.

This also fixes the problem, when after removing core containing dirty
data another core is added, and then recovery procedure following dirty
shutdown assigns cache lines from removed core to the new one, effectively
leading to data corruption.

Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2021-01-19 13:34:28 +01:00
Robert Baldyga
3f7139a814
Merge pull request #439 from mmichal10/remove_core_fine_lock 
Remove core fine lock
 2021-01-15 21:10:14 +01:00
Adam Rutkowski
f206c64ff6 Fine granularity lock in cache_mngt_core_deinit_attached_meta 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-01-15 03:11:46 -05:00
Michal Mielewczyk
6d962b38e9 API for cacheline write trylock 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-01-15 02:10:42 -05:00
Adam Rutkowski
bd20d6119b External linkage for function to sparse single cline 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-01-15 02:10:42 -05:00
Adam Rutkowski
93bda499c7 Add functions to lock specific hash bucket 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2021-01-12 15:42:21 -05:00
Robert Baldyga
fd88c2c3a4
Merge pull request #436 from mmichal10/metadata-assert 
Metadata assert
 2021-01-08 10:15:08 +01:00
Robert Baldyga
eff0047d6f
Merge pull request #434 from robertbaldyga/dont-modify-list-read-lock 
seq-cutoff: Don't modify node list under read lock
 2021-01-08 10:02:26 +01:00
Michal Mielewczyk
fcef130919 Bug on metadata access error 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-01-07 18:10:44 -05:00
Michal Mielewczyk
d0225ef1cb Prevent uint32_t overflow 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2021-01-07 02:45:05 -05:00
Robert Baldyga
ea1fc7a6d4 seq-cutoff: Don't modify node list under read lock 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2021-01-05 19:46:37 +01:00
Robert Baldyga
dd508c595f
Merge pull request #430 from rafalste/fix_attach_load_paths 
Create separate pipelines and paths for cache attach/load scenarios
 2020-12-23 16:51:37 +01:00
Robert Baldyga
69e388a10f
Merge pull request #372 from arutk/wo_test_enhancements 
Extend WO engine functional tests
 2020-12-23 16:51:16 +01:00
Rafal Stefanowski
57d4aaf7c9 Return error status from ocf_freelist_init 
Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
 2020-12-23 16:43:46 +01:00
Rafal Stefanowski
d3b61e474c Remove init_mode and use metadata.is_volatile instead 
Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
 2020-12-23 16:31:55 +01:00
Rafal Stefanowski
88b97df16d Fix pipeline attach/load paths 
Create separate pipelines for cache attach and load scenarios.

Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
 2020-12-23 16:31:49 +01:00
Robert Baldyga
efc8786ed3
Merge pull request #432 from robertbaldyga/seq-cutoff-detached-core 
Initialize sequential cutoff for detached cores
 2020-12-23 14:24:24 +01:00