![]() Roundoff to the log ticket rather than directly modifying grant heads. added patch to pass log ticket down to xlog_sync() so that it accounts the reworked the loop in xlog_write_single() based on Christoph's suggestion. ![]() fixed up "lv chain vector and size calculation" commit message to reflect weĪre only calculating and passin gin the vector byte count. removed stray trace_printk()s that were used for debugging. typedef uint64_t xfs_csn_t and conversion. fixed incorrect/unintended changes to xfs_log_force() behaviour Guarantees metadata vs journal ordering for both internal and external logs. With an explicit flush in the unmount record write so that it works the sameĪs the new CIL checkpoint cache pre-flush mechanism. remove the implicit external log's data device cache flush code and replace it fixed whitespace in xfs_flush_bdev_async() skip cache flush in xfs_flush_bdev_async() if the underlying queue does not pass bio to xfs_flush_bdev_async() so it doesn't need allocation. remove xfs_blkdev_issue_flush() instead of moving and renaming it. simplify xlog_wait_on_iclog_lsn() back to just a call to xlog_wait_on_iclog() aggregate many small dependent patchsets in one large one. Updated documentation with all the fixes and suggestions made. split out log vector chain conversion to list_head into separate patch (order id) were reversed, leading to intents being replayed in the wrong fixed percpu CIL item list sort order where items in the same transaction simplified the CIL sort comparison functions to a single comparison operation cleaned up the pcp, cilpcp, pcptr mess in xlog_cil_pcp_alloc() and elsewhere added the ctx->ticket->t_unit_res update back into the code even though it is factored out CIL push percpu structure aggregation into added missing cpu dead accounting transfer in xlog_cil_pcp_dead(). fixed whitespace damage in xlog_cil_pcp_dead() removed unused nvecs from struct xfs_cil_ctx added a patch for removing the optype parameter from xlog_write() no need to check for XLOG_CONTINUE_TRANS in xlog_write_partial() as it will removed the unnecessary change for reserved iclog space in fixed double initialisation of lv in xlog_write_single(). Added BUILD_BUG_ON() to check the size of unmount records. updated pahole output for checkpoint header in commit message ![]() debug overhead reduction patches merged so removed from series Added comment explaining why the CIL workqueue is limited to 4 concurrent updated comment to indicate that iclog sync state is set according to theĬaller's desire for a stable checkpoint to be performed. Wants the entire CIL checkpoint and commit record to be on stable storage when changed "push_async" to "push_commit_stable" to indicate that the push caller replaced XFS_LSN_CMP() checks with direct lsn1 = lsn2 comparisons to simplify buffer log item dirty tracking patches merged so removed from series The iclog so that multiple independent writes to an iclog doesn't clear flags ensure callers only add iclog flush/fua flags appropriately before releasing ensure that setting/clearing XLOG_ICL_NEED_FLUSH is atomic (i.e. trimmed superflous comments about not requiring REQ_PREFLUSH for iclog IO fixed completion logic for async cache flush propagate changes through patches to fix conflicts due to dependent changes. fix whitespace in xlog_cil_free_logvec() (patch 35) Patch 33 as patch 35 addresses these issues. removed unnecessary commentary about CIL ordering issues from commit msg in added space_reserved to CIL pcp structure (patch 31) add comment to xlog_cil_pcp_aggregate() (patch 30) added space_used to CIL pcp structure (patch 30) removed unused space_used/curr_res from CIL pcp structure (patch 29) removed unnecessary update of num_iovecs (patch 20) Headline numbers are an increase in transaction rate fromħ00k commits/s to 1.7M commits/s, and a reduction in fua/flush operations byĢ-3 orders of magnitude on metadata heavy workloads that don't use fsync. Performance improvements are largely documented in the change logs of the Patch remaining without at least one RVB tag is patch 30. Version 4 was posted version contains the changes Darrick requested during review. This is an update of the consolidated log scalability patchset I've been working ` (39 more replies) 0 siblings, 40 replies 56+ messages in thread 5:22 ` xfs: log stripe roundoff is a property of the log Dave Chinner Xfs: CIL and log optimisations archive mirror help / color / mirror / Atom feed * xfs: CIL and log optimisations 5:22 Dave Chinner
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |