Linux Kernel 5.10 Regression Mystery

On Tue, Sep 07, 2021 at 02:09:52PM +0530, Rohit Keshri wrote: > Hello Team, > > A flaw was found in the "Routing decision" classifier in the Linux kernel's > Traffic Control networking subsystem in the way it handled changing > classification filters, leading to a use-after-free condition. This flaw > allows unprivileged local users to escalate their privileges on the system. > The highest threat from this vulnerability is confidentiality, integrity, > as well as system availability. > > This issue was fixed in the upstream Kernel 5.10 onward with ef299cc3fa1a9 Note, commit ef299cc3fa1a ("net_sched: cls_route: remove the right filter from hashtable") came out in the 5.6 kernel release, in March of 2020, and was also backported to all relevant stable kernel releases at the beginning of April, 2020: 4.4.218 4.9.218 4.14.175 4.19.114 5.4.29 5.5.14 How did 5.10 get messed up in this, it was not released until December 2020? thanks, greg k-h

A Linux kernel mailing-list message reports a serious privilege-escalation vulnerability and claims it was fixed upstream in kernel 5.10. The punchline is that the cited fix actually appeared in kernel 5.6 and was backported months earlier, so the writer sarcastically questions how version 5.10 could have been affected before it was even released. It is used as a reaction to baffling regressions, sloppy bug reports, or security discussions that do not add up. EN: A Linux kernel mailing-list message reports a serious privilege-escalation vulnerability and claims it was fixed upstream in kernel 5.10. The punchline is that the cited fix actually appeared in kernel 5.6 and was backported months earlier, so the writer sarcastically questions how version 5.10 could have been affected before it was even released. It is used as a reaction to baffling regressions, sloppy bug reports, or security discussions that do not add up.

• @infosecmemes / 645 2021-09-07