When discussing the next portions of corrections proposed for inclusion in the nucleus of 6.11 -rc5, the author of bcachefs, Linus Torvalds, expressed regret about accepting the Bcachefs file system into the nucleus. Torvalds’ dissatisfaction stems from Kent Overstreet, the author of Bcachefs, sending overly voluminous corrections during the final candidates for releases. These corrections not only fix errors but also introduce new functionality, which goes against the development guidelines where only small corrections are accepted at the RC5 stage, not large functional changes.
Overstreet submitted a patch for bcachefs with 1309 lines added and 671 lines deleted across 39 files. In addition to error corrections, the patch included significant changes like support for a new data structure to control free elements in the cache and code for converting a hash table cache. These changes aim to eliminate unnecessary locks in the cache and prevent blocking conflicts during multi-flow loads.
This is not the first instance where substantial “corrections” for Bcachefs are submitted after the change window, promoting functional additions instead of simple error fixes towards the end of a new branch’s development. In response to the pull-request, Torvalds stated:
“Enough. The last Pull-request was also large. This PULL request is also too voluminous, affects things unrelated to BCACHEFS, and does not address any regressions. At some point, ‘fixing something’ simply turns into development, and this is that moment. Bcachefs patches have become a sort of ‘high development during the release preparation cycle,’ to the extent that I am starting to regret merging Bcachefs. If Bcachefs cannot adhere to the main development schedule upstream, maybe it should not be in the main core. This is beyond a joke.”
Kent Overstreet, developer of Bcachefs,