github/ocf
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
620 Commits 1 Branch 1 Tag 4.1 MiB
d6739441de
Commit Graph

315 Commits

Author SHA1 Message Date
Michal Mielewczyk
01ce586e6a Use API instead of raw variables to update block stats. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 08:01:09 -04:00
Michal Mielewczyk
51c9c516a4 Use API instead of raw variables to update req stats. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 08:01:09 -04:00
Michal Mielewczyk
b4c384eb2d Use API instead of raw variables to update error stats. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 08:01:09 -04:00
Michal Mielewczyk
0d4f4d3294 Added API for updating statistics. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 08:01:09 -04:00
Michal Mielewczyk
42d6dbbf11 Remove low-level stats getters from public API. 
Since stats builder is implemented for retrieving cache, core and ioclass stats,
adapters should use it instead.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 08:01:09 -04:00
Michal Mielewczyk
89de6038c1 Split retrieving core stats and config info into two functions. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 08:01:09 -04:00
Michal Mielewczyk
f9da89263b Stats builder for ioclass statistics. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-10 04:32:21 -04:00
Adam Rutkowski
13cf871a13 Per-execution-context freelists 
Global free cacheline list is divided into a set of freelists, one
per execution context. When attempting to map addres to cache, first
the freelist for current execution context is considered (fast path).
If current execution context freelist is empty (fast path failure),
mapping function attempts to get freelist from other execution context
list (slow path).

The purpose of this change is improve concurrency in freelist access.
It is part of fine granularity metadata lock implementation.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-09-09 16:19:52 -04:00
Michal Mielewczyk
2450d3da4b Move block stats counters to ioclass section. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-09 02:46:45 -04:00
Michal Mielewczyk
8304ed84b8 More precise percentage value in stats builder. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-09-09 02:46:45 -04:00
Robert Bałdyga
5fd239d2bb
Merge pull request #245 from imjfckm/switching-pp 
Switching mechanism for promotion policies
 2019-08-30 11:35:16 +02:00
Robert Bałdyga
1f9f79f11d
Merge pull request #250 from robertbaldyga/parts-meta-new-section 
Move parts metadata to separate section
 2019-08-29 14:15:31 +02:00
Jan Musial
ee3fdb0506 Change nhit_hash names to avoid collision with kernel 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-08-29 13:24:54 +02:00
Jan Musial
a1d3cf0c4d Switching mechanism for promotion policies 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-08-29 13:24:26 +02:00
Robert Baldyga
d10c618065 Move parts metadata to separate section 
Fix problem introduced by increasing partition name size to 1024 bytes,
which effectively made superblock bigger than one page. Due to this
flushing superblock required more than one io, which in case of dirty
shutdown between these ios resulted in CRC missmatch and made cache
recovery impossible.

Moving parts metadata to separate sections makes superblock fitting
in one page, effectively solving described problem.

Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-29 10:23:13 +02:00
Jan Musiał
0a9f197b10
Merge pull request #238 from arutk/split_cacheline_lock_trylock 
Split cacheline locking implementation into trylock and lock
 2019-08-29 09:49:53 +02:00
Adam Rutkowski
6dd1af7df7 Split cacheline locking implementation into trylock and lock 
This change refactors the code in order to prepare for removing
global concurrency lock, which won't be needed after per-bucket
metadata locking is in place.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-08-28 18:54:19 -04:00
Robert Bałdyga
358c529eff
Merge pull request #243 from arutk/simplify_freelist_striping_init 
Initialize freelist using metadata layout implementation
 2019-08-27 17:12:04 +02:00
Michal Rakowski
e3a20c0da0
Merge pull request #239 from imjfckm/pp-implementation 
nhit promotion policy implementation
 2019-08-27 12:42:03 +02:00
Adam Rutkowski
860c8c7a0f Initialize freelist using metadata layout implementation 
Metadata layout (seq / striping) is already encapsulated
in ocf_metadata_layout_iface, no need to double this logic
in freelist initialization.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-08-26 15:48:11 -04:00
Jan Musial
62a0ccdd9f nhit promotion policy implementation 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-08-22 13:34:38 +02:00
Michal Mielewczyk
2c27e9a433 Store name for newly initialized cache in array instead of passing pointer. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-08-20 01:43:11 -04:00
Robert Bałdyga
8ed525ae7f
Merge pull request #234 from mdnfiras/sync-prim-destroyers 
Adding synchronization primitives destroyers
 2019-08-13 15:46:26 +02:00
Firas Medini
33c96d226e Verifying IO address and size 
IO address and size must be multiple of sector size. If not, ocf_io_new() will return NULL.

Signed-off-by: Firas Medini <mdnfiras@yahoo.com>
 2019-08-13 05:37:14 -07:00
Firas Medini
1f979f630b Adding synchronization primitives destroyers 
Environment should provide calls for destroying primitives (i.e. env_mutex_destroy()) and OCF should call these functions in its cleanup paths.

Signed-off-by: Firas Medini <mdnfiras@yahoo.com>
 2019-08-13 05:13:11 -07:00
Robert Baldyga
9730b2bd1f Remove remaining core_id occurrences from public API 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-07 16:12:06 +02:00
Michal Rakowski
d499ef1c05
Merge pull request #227 from mmichal10/mark-core-added-false 
Don't try to remove invalid cores

If valid cache metadata was read, but environment has changed (i.e. number of cache lines has changed) ocf (in error handling path) was trying to close cores which were not opened. It happened due to cores were marked in cache metadata as added, but any cache inserting operation didn't take place.

In this patch 'added' flag in cache metadata was replaced with more meaningful 'valid' - it is set if given core is stored in cache metadata. Moreover, new 'added' flag was added to core run-time metadata and it is set if given core is added to cache.
 2019-08-07 13:37:59 +02:00
Michal Mielewczyk
cdc0a575a1 Added 'added' flag in core runtime data. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-08-06 07:36:39 -04:00
Michal Mielewczyk
433a25a1b2 Renamed 'added' flag in core metadata to 'valid'. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-08-06 07:08:12 -04:00
Michal Mielewczyk
5dd09e823b Initialize promotion policy before cleaning policy. 
If cleaning policy didn't have init() function,
   '_ocf_mngt_init_instance_load_complete' returned early and promotion policy
   wasn't initialized at all.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-08-06 06:37:14 -04:00
Adam Rutkowski
9649f25415
Merge pull request #229 from robertbaldyga/cleanup-metadata 
Metadata interface cleanup
 2019-08-05 14:11:04 +02:00
Michal Mielewczyk
340ca76159 Move check for valid cache size in load. 
Loaded cache size should be compared with acutal cache size after loading
superblock.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-08-05 07:58:38 -04:00
Michal Mielewczyk
80f82cab2b Deinitialize volume types in ocf internally. 
Adapter can provide ops for data type deinitialization, if any additional
operations are required on deinitialization procedure.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-08-05 06:16:46 -04:00
Robert Baldyga
fdf6d07b85 Fix typo "collisioin" -> "collision" 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
4b1a3fa4b5 Simplify metadata_raw interface 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
a31ddd44cb Remove ocf_metadata_hash_entries_hash() 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
ff7583c6da Simplify partition metadata interface 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
3108f36e42 Simplify collision matadata interface 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
0b0546302f Remove ocf_metadata_get_core_sector() 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
0ca0ea1343 Remove ocf_metadata_line_size_is_valid() 
This function is no longer used.

Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-08-03 20:04:23 +02:00
Robert Baldyga
92c7e12549 Remove core id from public API 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-30 09:26:26 +02:00
Robert Baldyga
1100cb0b4f Remove cache id from public API 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-30 09:26:26 +02:00
Robert Baldyga
259df7ace9 Store core name in metadata 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-30 09:26:26 +02:00
Robert Baldyga
eb4272afa9 Store cache name in metadata 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-30 09:26:17 +02:00
Robert Baldyga
4f0735b503 Ensure that core name is set and unique 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-29 17:37:48 +02:00
Robert Baldyga
901b39031f Ensure that cache name is set and unique 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-29 17:37:43 +02:00
Robert Baldyga
331b99397f Change ctx lock to rmutex 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-29 15:08:44 +02:00
Kamil Łepek
05be67a72b
Merge pull request #220 from arutk/metadata_offset_hash 
New hash function formula
 2019-07-29 13:03:04 +02:00
Jan Musiał
fa14d6a4b5
Merge pull request #221 from arutk/metadata_rename_concurrency 
Restrutcure concurrency code
 2019-07-29 09:42:37 +02:00
Robert Baldyga
3355a6cd56 engine: discard: Add missing put on error path 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-27 19:06:11 +02:00
Adam Rutkowski
494861c994 Rename cache_concurrency to cache_line_concurrency 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-07-24 15:32:12 -04:00
Adam Rutkowski
e5bed8825c Move metadata concurrency to a separate file 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-07-24 15:32:09 -04:00
Adam Rutkowski
3dd4263bc8 Change hash function 
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>
 2019-07-24 15:02:34 -04:00
Adam Rutkowski
7184f7787c Cleanup map_info struct 
1. Rename hash_key -> hash
2. Better comments for hash and coll_idx

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-07-24 13:12:05 -04:00
Michal Rakowski
aa02f56b05
Merge pull request #218 from robertbaldyga/rm-includes-metadata-superblock 
Remove unneeded includes in metadata_superblock.h
 2019-07-23 13:56:39 +02:00
Jan Musiał
1fd1a6fe17
Merge pull request #209 from arutk/stop_no_queue 
Allow to stop cache without management queue
 2019-07-23 13:51:21 +02:00
Robert Baldyga
ab2f8b199a Remove unneeded includes in metadata_superblock.h 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-23 13:00:05 +02:00
Jan Musiał
232f0cd8d7
Merge pull request #216 from robertbaldyga/io-and-req-in-single-allocation 
Allocate io and req in single allocation
 2019-07-23 11:40:30 +02:00
Jan Musiał
08abf69224
Merge pull request #217 from Open-CAS/remove-redundant-field 
Remove redundant field from struct ocf_cache_device
 2019-07-23 09:55:03 +02:00
Jan Musial
917cbd859a Add promotion policy API and use it in I/O path 
Promotion policy is supposed to perform ALRU noise filtering by
eliminating one-hit wonders being added to cache and polluting it.

Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-07-19 13:52:00 +02:00
Robert Baldyga
6822ab6140 Remove redundant field from struct ocf_cache_device 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-19 12:59:14 +02:00
Robert Baldyga
e64bb50a4b Allocate io and request in single allocation 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-17 10:31:23 +02:00
Robert Baldyga
61414f889e Introduce OCF IO allocator 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-17 10:31:23 +02:00
Robert Baldyga
e254c9b587 Merge new_io and configure into one function 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-17 10:31:23 +02:00
Robert Baldyga
4d2d31ff76 Introduce ocf_io_internal 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-07-17 09:54:51 +02:00
Adam Rutkowski
4f04b2762f Simplified stop path for cache detached case 
When cache is detached we cannot assume there is a management
queue created. This change introduces simplified cache stop
path, performing all the necessary deinit without using
IO queues.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-07-08 18:26:35 -04:00
Michal Mielewczyk
636834dae3 Check for existing metadata when attaching cache. 
Instead of overriding existing metadata silently, return appropriate error.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-07-08 09:23:32 -04:00
Jan Musial
aab3c4fbaa Fix typo in metadata_hash_entries fcn 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-07-04 14:01:27 +02:00
Jan Musial
ede2bcdf48 Fix volumes deinit routine 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-06-28 13:03:14 +02:00
Jan Musiał
e7e0ad957d
Merge pull request #201 from mmichal10/fix-load-error-handling 
Fix error handling in case of failed load
 2019-06-26 12:24:02 +02:00
Jan Musial
2c629be4c2 Fix leaking front and core volumes 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-06-26 10:00:49 +02:00
Michal Mielewczyk
52f6d6280b Fix error handling in case of failed load 
Store default metadata values and restore them in case of failed load.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-06-25 09:31:05 -04:00
Jan Musial
4a20ce9c25 Fix leaking cache references 
Signed-off-by: Jan Musial <jan.musial@intel.com>
 2019-06-25 11:43:28 +02:00
Michal Rakowski
b1a6c467a0 Introduce core_is_dirty mngt method 
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
 2019-06-25 09:12:44 +02:00
Michal Rakowski
a818ffabfd Move ocf_mngt_cache_is_dirty to public mngt interface 2019-06-24 15:23:06 +02:00
Michal Rakowski
2925147395 Remove reduntant dirty check 
When flush completion is called there could be some clines marked as dirty since those could be in-use during flushing.
 2019-06-24 14:24:34 +02:00
Jan Musial
aec01f216c Revert "Disallow ocf_cache_mode_max in io_class config" 
This reverts commit 5ad5c521df.

This change broke setting IO-classes with allocation. We use max as a
special value to indicate that the partition should use cache global
caching mode.
 2019-06-14 14:08:28 +02:00
Michał Mielewczyk
f9f75e7265
Merge branch 'master' into wo_improvements 2019-06-13 11:32:06 +02:00
Adam Rutkowski
6bb1d21489 Do not repartition cachelines in WO read 
WO cache mode should not repartition cachelines nor affect cacheline
status in any way when servicing read. Reading data from the cache
is just an internal optimization. Also WO cache mode is designed to
be used with partitioning based on write life-time hints and read
requests do not carry write lifetime hint by definition.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-06-13 13:14:30 -04:00
Michal Mielewczyk
e8b24d811b Check if all required ops are provided by adapter. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-06-13 04:46:03 -04:00
Adam Rutkowski
9dc1381b77 Refactor ocf_submit_cache_reqs map indexing 
Refactoring ocf_submit_cache_reqs to make it clear that
req->map is accessed at index derived from offset argument,
not necesarily starting at 0.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-06-12 18:56:40 -04:00
Adam Rutkowski
b82ed3af26 Initialize @increment in ocf_read_wo_cache_do 
Static code analyzers fail to understand that this variable
is always assigned to before usage.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-06-12 13:44:01 -04:00
Adam Rutkowski
b97bb6f53b Add function to check for lazy write cache mode (WO or WB) 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-06-12 12:07:02 -04:00
Adam Rutkowski
82e8c55f4a Write-only cache mode 
Write-only cache mode is similar to writeback, however read
operations do not promote data to cache. Reads are mostly serviced
by the core device, only dirty sectors are fetched from the cache.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-06-12 12:07:02 -04:00
Adam Rutkowski
ae6164a49c Helper functions to get request start/end sector in cacheline 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-06-12 12:07:02 -04:00
Michal Mielewczyk
328d27846a Prevent overriding null terminator in cache name 2019-06-11 12:12:07 +02:00
Michal Mielewczyk
46c17144f6 Remove redundant check for core id 2019-06-11 12:12:07 +02:00
Michal Mielewczyk
6cdbac82bc Check for valid core_id value 2019-06-11 12:12:07 +02:00
Michal Mielewczyk
fc5334e419 Return memcpy result 2019-06-11 12:12:07 +02:00
Michal Rakowski
29199cb5d4 Added missing metadata_unlock 
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
 2019-06-11 12:12:07 +02:00
Michal Rakowski
4789fb38f0 Added asserts in sample env for synchronization functions 
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
 2019-06-11 11:01:06 +02:00
Michal Rakowski
b1cf6c4642 Changed always returning 0 to void foo 
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
 2019-06-11 09:53:40 +02:00
Michal Rakowski
30ac3abd74 Small get_ram_needed fix, proper volume is initialized now. 2019-06-11 09:53:40 +02:00
Michal Mielewczyk
be64173066 Fix order of operations in cache stop completion 
Prevent metadata updater being freed by pyocf before it's deinitialized by ocf.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-06-06 09:03:26 -04:00
Michal Rakowski
d714f6235b Use rate limited logging in case of engine error 
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
 2019-06-06 11:27:25 +02:00
Michal Mielewczyk
e6bedb692c Unified management functions prefix. 
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
 2019-06-05 09:31:59 -04:00
Michal Rakowski
9f4536c6e3 Error codes in IO path changed to OCF-specific 
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
 2019-06-05 09:10:54 +02:00
Robert Baldyga
f240f81641 Make request structure more compressed 
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
 2019-06-03 19:19:29 +02:00
Adam Rutkowski
d7b3a187e4 Fix condition for setting req->info.dirty_all 
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
 2019-05-31 19:10:11 -04:00
Michal Wysoczanski
d27f81c588 Add missing new line in error message 
Signed-off-by: Michal Wysoczanski <michal.wysoczanski@intel.com>
 2019-05-30 00:03:09 +02:00