> This patchset has the "easy" reverts, there are 68 remaining ones that need to be manually reviewed. Some of them are not able to be reverted as they already have been reverted, or fixed up with follow-on patches as they were determined to be invalid. Proof that these submissions were almost universally wrong.

Skimming through subject lines of 190 commits being reverted here, every single one of them is along the lines of "add refcount/NULL/etc check and conditionally do (or do not) de-allocate memory before error-path return". I.e. worst case - this will reintroduce some rare memory leak or memory lifecycle bug.

Also, all of patches in question are in drivers. So depending on hardware used, any given system's user is likely to only have to worry about 2-3, maybe 5 of the patches, not all 190.