Linux 5.13 Reverts + Fixes The Problematic University of Minnesota Patches

One month ago the University of Minnesota was banned from contributing to the Linux kernel when it was revealed the university researchers were trying to intentionally submit bugs into the kernel via new patches as "hypocrite commits" as part of a questionable research paper. Linux kernel developers have finally finished reviewing all UMN.edu patches to address problematic merges to the kernel and also cleaning up / fixing their questionable patches.

Sent in on Thursday by Greg Kroah-Hartman was char/misc fixes for 5.13-rc3. While char/misc fixes at this mid-stage of the kernel cycle tend to not be too exciting, this pull request has the changes for addressing the patches from University of Minnesota researchers. Greg noted in the pull:

The majority here is the fallout of the umn.edu re-review of all prior submissions. That resulted in a bunch of reverts along with the "correct" changes made, such that there is no regression of any of the potential fixes that were made by those individuals. I would like to thank the over 80 different developers who helped with the review and fixes for this mess.

Going by the umn.edu Git activity that puts 37 patches as having been reverted with this pull request. The reverts span from ALSA to the media subsystem, networking, and other areas. That is 37 reverts out of 150+ patches from umn.edu developers over the years.

Here are some of the technical explanations for some of the problems with their patches:

The original commit message for this change was incorrect as the code path can never result in a NULL dereference, alluding to the fact that whatever tool was used to "find this" is broken. It's just an optional resource reservation, so removing this check is fine

###

While the original commit does keep the immediate "NULL dereference" from happening, it does not properly propagate the error back to the callers, AND it does not fix this same identical issue in the drivers/net/ethernet/cavium/liquidio/lio_vf_main.c for some reason.

###

This commit is not correct, it should not have used unlikely() and is not propagating the error properly to the calling function, so it should be reverted at this point in time. Also, if the check failed, the work queue was still assumed to be allocated, so further accesses would have continued to fail, meaning this patch does nothing to solve the root issues at all.

###

Upon review, this commit was found to be not be needed at all as the change was useless because this function can only be called when of_match_device matched on something. So it should be reverted.

Among the other patches that ended up being reverted for having no substantive value or at worst incorrect/incomplete code...

Source