There were two mentions of "writebehind" cache type. From earlier declaration there are three types available: "Writethrough", "Writearound", "Writeback". Probably "writebehind" means "Writearound". replace those two mentions.
….0 support. Summary: Test Plan: Reviewers: CC: Task ID: # Blame Rev:
added 3.9.0 support
Fix flashcache_load getting cachedev from argv when using flags
Depmod should be run on the right kernel
Fix use-after-free in flashcache_destroy
The sb variable points to buf. But buf gets freed and reallocated, so the old pointer points to freed memory. The variable is then reused to invalidate the cache_sb_state, but the change never gets saved because the buf that is written to disk does not point to the same memory.
The cachedev parameter is in position 3 when no flags specificed. When -v is used then cachedev will be in position 4. Signed-off-by: Roi Dayan <firstname.lastname@example.org>
…ed in correctly. Summary: Fix for a flashcache_create bug where associativity is not being passed in correctly to the flashcache module. Test Plan: Reviewers: CC: Task ID: # Blame Rev:
It might be due to an upgrade change. The location of flashcache_version changed.
Summary: Changes to make flashcache compile on 3.6.0. Thanks to Thomas Stein for quickly verifying that the patch works. Test Plan: Reviewers: CC: Task ID: # Blame Rev:
… writearound modes. Patch submitted by Arne Wiebalck. Summary: Test Plan: Reviewers: CC: Task ID: # Blame Rev:
completion for the !WRITEBACK case, issue an uncached disk IO in the case of a SSD error. Patch submitted by Mohit Saxena. Summary: Test Plan: Reviewers: CC: Task ID: # Blame Rev:
…ng passed into vmalloc(). Summary: The vmalloc() for the superblock in flashcache_writeback_load() is passed a bad size argument. Thanks to Ping Huang for spotting this and reporting it. Test Plan: Reviewers: CC: Task ID: # Blame Rev:
Summary: Bump flashcache sw revision in stable_v2 to 2.0. Test Plan: Reviewers: CC: Task ID: # Blame Rev:
Summary: Committing an old patch to handle SSD errors in writethru and writearound modes to fall back to disk instead of returning errors back to the user. This path needs to be extended for writeback as well (failed reads on a !DIRTY block and failed new writes can be similarly handled. Thanks to Arne Wiebalck for providing the impetus for this change, reworking my initial attempt and testing the patch in production. Test Plan: Reviewers: CC: Task ID: # Blame Rev:
patch change the start order, start to demons, stop after stop demons patch test variables -n to -z fix variable name $RETVAL in condition PS sorry my bad english :)
Added basic checking for OpenVZ platforms
add ability to set version into dkms Makfile
Summary: Note about cache utilization added to SA guide. Contributed by Will Smith. Test Plan: Reviewers: CC: Task ID: # Blame Rev: