Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Fix stalled txg with repeated noop scans
Currently, the DSL scan code figures out when it should suspend processing and allow a txg to continue by calling the function dsl_scan_check_suspend(). Unfortunately, this function only allows the scan to suspend at a level 0 block. In the event that the system is scanning a bunch of empty snapshots or a resilver is running with a high enough scn_cur_min_txg, the scan will stop processing each dataset at the root level, deciding it has nothing left to do. This means that the check_suspend function is never called and the txg remains stuck until a dataset is found that has data to scan. This patch fixes the problem by allowing scans to suspend at the root level of the objset. For backwards compatibility, we use the bookmark <objsetid, 0, 0, 0> when we suspend here so that older versions of the code will work as intended. Reviewed-by: Matt Ahrens <[email protected]> Reviewed-by: Brian Behlendorf <[email protected]> Signed-off-by: Tom Caputi <[email protected]> Closes openzfs#9300
- Loading branch information