v21.6.3

Changelog:

• Fixed stack overflow when trying to insert to disabled io class and the sequential cutoff is on

v21.6.2

Changelog:

• Introduced OCF_ERR_MIN/OCF_ERR_MAX defines to make it possible to verify if given error code is an OCF error code

v21.3.3

Changelog:

• Improved data isolation between cores of the same cache
• Fixed memleak in core remove operation
• Fixed potential stack overflow in PT engine
• Fixed race condition during cleaning policy switch
• Other minor fixes

v21.6

Changelog:

• Allocator improvements to avoid unnecessary memset's
• Remove legacy self-test to improve startup time
• Structure alignment and compiler optimizations
• Replaced metadata updater with dynamic metadata io concurrency mechanism
• Removed partition list
• LRU code refactoring
• Removed eviction policy abstraction
• Improved atomicity guarantees for metadata updates
• Asynchronous API for switching cleaning policy
• Fixed typos and other minor defects

Known issues:

• IO class max occupancy can sometimes by exceeded by small percentage

v21.3

Changelog:

• Locking improvements (split global metadata lock, remove status bits locks, parallel eviction)
• Multi-tier LRU list
• Maximum IO class occupancy improvements
• Memory allocation improvements
• Per-queue multi-level sequential cutoff
• CPU cache line friendly metadata layout

v20.12.2

Bugs fixed:

• ALRU cleaning starts to early (#455)

v20.12.1

Bugs fixed:

• System crash when seq-cutoff threshold is equal to request_size in io classification (Open-CAS/open-cas-linux#663)
• Incorrect per-partition counters value (Open-CAS/open-cas-linux#673)

v20.3.4

Merge pull request #441 from robertbaldyga/v20.3.4-flush-collision-on…

v20.12

Changelog:

• Introduced occupancy per ioclass
• Introduced multistream sequential cutoff
• Performance improvements (metadata rework, better lock granularity)

Bugs fixed:

• Possible metadata checksum mismatch on cache drives without power failure write buffer protection

v20.3.2

Changelog:

• Fixed use after free bug in core attach/detach path
• Fixed bug related to improper request map cleanup after unsuccessful mapping

Known issues:

• Possible metadata checksum mismatch on cache drives without power failure write buffer protection