Commit Graph

446 Commits

Author SHA1 Message Date
Michal Mielewczyk
05f3c22dad Occupancy per ioclass utilities
Functions to check space availability and to manage cachelines reservation

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:18:47 -05:00
Michal Mielewczyk
600bd1d859 Access partition's metadata counters via functions
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:25 -05:00
Michal Mielewczyk
9d80882b00 Remove re_part field from struct ocf_req_info
Since the request carries an explicit information about number of the
cacheliens to be reparted, no need of keeping the boolean information if some
of the request's cachelines are assigned to a wrong partition

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:25 -05:00
Michal Mielewczyk
e26ca30399 Track explicit number of cachelines to be reparted
Instead of redunant calculating number of cachlines to be reparted, keep this
information in request's info

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:25 -05:00
Michal Mielewczyk
4f228317a1 Update docs for space_managment_evict_do()
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:25 -05:00
Michal Mielewczyk
21e98a6dbc Evict request's target partition in regrular order
Instead of evicting target partition as the last one, respect eviction
priorities

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:25 -05:00
Michal Mielewczyk
e999de7232 Don't roundup when evicting single part
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:25 -05:00
Michal Mielewczyk
718dc743c8 Enable particular ioclass eviction
If partition's occupancy limit is reached, cachelines should be evicted from
 request's target partition.

Information whether particular partition eviction should be triggered is
carried as a flag by request which triggered eviction.

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 08:00:23 -05:00
Michal Mielewczyk
e9d7290078 Extend ioclass management logging
When setting ioclass, print info about it's max size

Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 06:04:16 -05:00
Michal Mielewczyk
c643a41977 Prevent adding ioclass with the same id twice
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 06:04:16 -05:00
Michal Rakowski
ac2effb83d Fix whitespaces
Signed-off-by: Michal Rakowski <michal.rakowski@intel.com>
Signed-off-by: Michal Mielewczyk <michal.mielewczyk@intel.com>
2020-12-21 06:04:16 -05:00
Adam Rutkowski
44efe3e49e Refactor LRU code to use part rather than part_id
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-12-17 14:35:27 +01:00
Adam Rutkowski
41a767de97 Multiple LRU lists
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-12-17 14:35:27 +01:00
Robert Baldyga
ac83c4ecd6 seq_cutoff: Allocate seq cutoff structures dynamically per core
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-12-17 14:35:27 +01:00
Robert Baldyga
0fd095046c
Merge pull request #426 from arutk/meta_no_memcpy
Remove memcpy from collision/eviction policy metadata api
2020-12-09 13:02:49 +01:00
Adam Rutkowski
fec61528e6 Remove memcpy from collision/eviction policy metadata api
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-12-07 17:58:44 +01:00
Robert Baldyga
7af386681d
Merge pull request #418 from robertbaldyga/inc-dep-env-headers
Remove dependency to full ocf_env.h from inc/ headers
2020-11-30 17:16:32 +01:00
Robert Baldyga
9bcafb5bfb seq_cutoff: Initialize each stream with different LBA
Initializing each stream with unique LBA ensures there are no initial
rbtree collisions, and thus helps to avoid clustering of all the streams
into one big linked list instead of forming performance friendly proper
tree structure.

Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-30 15:58:18 +01:00
Robert Baldyga
b8735f6517 rbtree: Fix swapping out-of-tree node with root
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-30 15:58:18 +01:00
Robert Baldyga
c8e7e0053c Remove dependency to full ocf_env.h from inc/ headers
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-19 13:07:16 +01:00
Robert Baldyga
a54d4461f0 seq_cutoff: Always continue the biggest stream
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-10 13:21:14 +01:00
Robert Baldyga
8b03271626 rbtree: Introduce list find callback
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-10 13:21:14 +01:00
Robert Baldyga
0ae4f4b5b2 rbtree: Add equal nodes to linked list
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-10 13:21:14 +01:00
Robert Baldyga
50c4de0495 rbtree: Make swap resistant to nodes outside the tree
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-10 13:20:45 +01:00
Robert Baldyga
694224971c rbtree: Replace spaces with tabs
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-11-09 17:32:03 +01:00
Robert Baldyga
0e3c9e740e
Merge pull request #396 from arutk/lru_refactor
Simplify and modularize LRU list code
2020-11-05 15:35:33 +01:00
root
ef08141252 Use -1 for LRU list terminator instead of collision_table_entries
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-11-04 17:43:41 -06:00
Adam Rutkowski
58f8a2218a Simplify and modularize LRU list code
Refactoring LRU list code to reduce code duplication and
improve testability.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-11-04 17:42:53 -06:00
Robert Baldyga
9a23787c6b
Merge pull request #406 from arutk/flush2
Propagate I/O flags (e.g. FUA) to metadata flush I/O
2020-10-06 12:49:22 +02:00
Adam Rutkowski
716edcc637 Flush cache volume after writing config metadata segments
After writing metadata configuration to disk we must
send a flush request to make sure configuration sections
are commited to non-volatile storage.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-09-30 10:40:03 +02:00
Adam Rutkowski
c945db356c Propagate I/O flags (e.g. FUA) to metadata flush I/O
Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-09-29 14:46:27 +02:00
Robert Baldyga
7c29110e47
Merge pull request #398 from Open-CAS/proper-core-status
Fix logging core state on cache load
2020-09-04 19:56:16 +02:00
Robert Baldyga
990f5160eb Cleanup request map entries in error handling path
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-09-02 14:30:28 +02:00
Robert Baldyga
0dfdcb05e9 Fix core volume lifecycle management
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-08-21 16:40:41 +02:00
Rafal Stefanowski
6542c2fa94 Fix memory requirement when loading cache
Load properties before checking memory needs and obtain cache line size
from context rather than from cache state.

Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
2020-08-19 03:15:18 +02:00
Rafal Stefanowski
072c9c1902 Pass only needed values to _ocf_mngt_calculate_ram_needed() function
Rather then passing whole structs, supply
_ocf_mngt_calculate_ram_needed() with just the values it actually uses.

Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
2020-08-19 03:09:05 +02:00
Jan Musial
2ee1e4c8dd Fix logging core state on cache load
Signed-off-by: Jan Musial <jan.musial@intel.com>
2020-07-28 14:52:15 +02:00
Robert Baldyga
d5ecdc16dd Make CRC mismatch on recovery a warning instead of error
Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-07-28 07:49:29 +02:00
Robert Baldyga
d946124a01 Calculate CRC for runtime metadata sections only on clean load
During recovery procedure there is no guarantee that checksums
of runtime sections were flushed correctly before dirty shutdown.

Signed-off-by: Robert Baldyga <robert.baldyga@intel.com>
2020-07-28 07:45:53 +02:00
Robert Baldyga
7d889fa1fc
Merge pull request #385 from arutk/pt_write_double_inv
Two pass write invalidate
2020-07-28 07:42:44 +02:00
Adam Rutkowski
b232f2b633 Service WA write misses in WI engine
WA write must follow follow the same two-pass pattern
as WI does. This change modifies WA engine to default to
WI in case of any miss (either partial or full), not only
partial miss.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-07-20 17:26:36 +02:00
Adam Rutkowski
91b6098fda Two pass write invalidate
Add second pass of write invalidate. It is necessary only
if concurrent I/O had inserted target LBAs to cache after
WI request did traversation. These LBAs might have been
written by WI request behind the concurrent I/O's back,
resulting in making these sectors effectively invalid.
In this case we must update these sectors' metadata to
reflect this. However we won't know about this after we
traverse the request again - hence calling ocf_write_wi
again with req->wi_second_pass set to indicate that this
is the second pass (core write should be skipped).

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-07-20 17:26:35 +02:00
Robert Baldyga
ec6eae6a5f
Merge pull request #377 from arutk/fix_map
Set entry->core_id in ocf_engine_lookup_map_entry
2020-07-10 21:32:09 +02:00
Adam Rutkowski
b14312dcef Set entry->core_id in ocf_engine_lookup_map_entry
core_id should be set in this function. The fact that
it is missing might lead to incorrect behaviour e.g. in
case of promotion policy.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-06-09 13:15:50 +02:00
Adam Rutkowski
7776bd6485 WO: read clean sectors from cache
In case of partial hit WO engine first reads data for the entire
request address range from core device. Then it plumbs it by fetching
dirty sectors from cache device.

For unindentified reason this leads to a data corruption in YCSB
workload A. After flushing dirty data and re-loading cache the
data is correct.

This change modifies WO read handler to read clean data from the
cache. This is not optimal, as the clean sectors are now read twice
in case of partial hit. For now it seems to be good enough work-around
for the data corruption problem.

The symptoms, combined with the fact that this change seems to make
the problem go away, indicates that at some point WB write handler
(and/or special I/O request handlers like discard) puts CAS in a
state where in-memory medata wrongly indicates that a sector is
clean while in fact it is dirty, as marked in the on-disk metadata.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
2020-05-27 12:31:53 +02:00
Robert Baldyga
1428376554
Merge pull request #371 from Ostrokrzew/load
Disable loading cache with 'force' flag
2020-05-22 13:52:16 +02:00
Slawomir Jankowski
248018b341 Change return code to valid OCF code
Signed-off-by: Slawomir Jankowski <slawomir.jankowski@intel.com>
2020-05-21 11:11:52 +02:00
Slawomir Jankowski
544e4086ca Disable load operation with 'force' flag
Fail `ocf_mngt_cache_load` function with `OCF_ERR_INVAL`
error code when force flag is in use.
Log error message.

Closes #361

Signed-off-by: Slawomir Jankowski <slawomir.jankowski@intel.com>
2020-05-21 11:11:52 +02:00
Slawomir Jankowski
455d554dc1 Reject zero-sized discard IOs to core
Signed-off-by: Slawomir Jankowski <slawomir.jankowski@intel.com>
2020-05-19 16:23:41 +02:00
Slawomir Jankowski
da34d5047b Typo fix
Signed-off-by: Slawomir Jankowski <slawomir.jankowski@intel.com>
2020-05-19 16:23:41 +02:00