On 2020-10-28 15:00 +0900, Masahiro Yamada wrote: > On Tue, Oct 27, 2020 at 4:32 AM Sven Joachim <svenjoac@xxxxxx> wrote: >> >> Building 5.10-rc1 in a setgid directory failed with the following >> error: >> >> dpkg-deb: error: control directory has bad permissions 2755 (must be >> >=0755 and <=0775) >> >> When building with fakeroot, the earlier chown call would have removed >> the setgid bits, but in a rootless build they remain. >> > > > Applied to linux-kbuild. Thanks. I don't see it there, have you pushed it out yet? > I agreed with "g-s" but was not sure about "u-s" > because nothing is explained about setuid, > and the setuid bit against directories seems to have no effect. > > > > > > It was interesting to read this article: > https://superuser.com/questions/471844/why-is-setuid-ignored-on-directories > > > > Also, it is summarized in the wikipedia > https://en.wikipedia.org/wiki/Setuid#setuid_and_setgid_on_directories > > "The setuid permission set on a directory is ignored on most UNIX and > Linux systems.[citation needed] However FreeBSD can be configured to > interpret setuid in a manner similar to setgid, in which case it > forces all files and sub-directories created in a directory to be > owned by that directory's owner - a simple form of inheritance.[5] > This is generally not needed on most systems derived from BSD, since > by default directories are treated as if their setgid bit is always > set, regardless of the actual value. As is stated in open(2), "When a > new file is created it is given the group of the directory which > contains it."" > > > After all, I am convinced that it would not hurt to do "u-s" > although I have never tested kernel builds on FreeBSD. Agreed, setuid directories should not end up in the .deb files even if that bit does currently nothing. Cheers, Sven
