On Mon, Jun 06, 2022 at 02:39:30PM +0200, Mickaël Salaün wrote:
>
> On 06/06/2022 13:20, Greg KH wrote:
> > On Mon, Jun 06, 2022 at 12:59:56PM +0200, Mickaël Salaün wrote:
> > > Hi Greg,
> > >
> > > The 21 Landlock commits merged for 5.19-rc1 and tagged with Cc stable@
> > > should indeed be backported up to 5.15 . The first commits are pure cosmetic
> > > changes but they need to be backported to avoid backport conflicts (for this
> > > series and future backports). They help maintain this subsystem, including
> > > to backport future changes.
> >
> > Ick, that's not how to submit patches for backporting ideally.
> >
> > Usually you submit the bugfixes first, and then we can backport them
> > easily.
>
> I understand, but this is are a one time cosmetic changes. Applying them
> later would have been much more difficult to handle.
Somehow other subsystems handle this just fine. Coding style cleanups
shouldn't really ever need to be backported if at all possible.
> > If you decide to reformat the codebase, well, you get to deal with the
> > backport issues later on (why is it reformatted, isn't it checkpatch
> > clean already?
> Yes, the idea was to backport early on because until now all commits (two)
> have been backported.
>
> >
> > > The following changes up to commit 8ba0005ff418
> > > ("landlock: Fix same-layer rule unions") are required to fix some edge case
> > > issues (i.e. syscall argument ordering checks and same-layer rule unions).
> > > New tests are added to check that everything work as expected for these
> > > backportable changes, and to make it possible for more test environments to
> > > run. I successfully tested the backport of all these commits to 5.15 .
> > > Please backport them to all stable branches.
> >
> > This is just backporting all files here, which seems crazy.
>
> It is backporting 21/30 commits. As maintainer it makes our work easier.
> There is no new feature introduced.
>
> >
> > >
> > > Here is the full list of the commits to backport (already marked with Cc:
> > > stable@xxxxxxxxxxxxxxx):
> > >
> > > 8ba0005ff418 landlock: Fix same-layer rule unions
> > > 2cd7cd6eed88 landlock: Create find_rule() from unmask_layers()
> > > 75c542d6c6cc landlock: Reduce the maximum number of layers to 16
> > > 5f2ff33e1084 landlock: Define access_mask_t to enforce a consistent access
> > > mask size
> > > 6533d0c3a86e selftests/landlock: Test landlock_create_ruleset(2) argument
> > > check ordering
> > > eba39ca4b155 landlock: Change landlock_restrict_self(2) check ordering
> > > 589172e5636c landlock: Change landlock_add_rule(2) argument check ordering
> > > d1788ad99087 selftests/landlock: Add tests for O_PATH
> > > 6a1bdd4a0bfc selftests/landlock: Fully test file rename with "remove" access
> > > d18955d094d0 selftests/landlock: Extend access right tests to directories
> > > c56b3bf566da selftests/landlock: Add tests for unknown access rights
> > > 291865bd7e8b selftests/landlock: Extend tests for minimal valid attribute
> > > size
> > > 87129ef13603 selftests/landlock: Make tests build with old libc
> > > a13e248ff90e landlock: Fix landlock_add_rule(2) documentation
> > > 81709f3dccac samples/landlock: Format with clang-format
> > > 9805a722db07 samples/landlock: Add clang-format exceptions
> > > 371183fa578a selftests/landlock: Format with clang-format
> > > 135464f9d29c selftests/landlock: Normalize array assignment
> > > 4598d9abf421 selftests/landlock: Add clang-format exceptions
> > > 06a1c40a09a8 landlock: Format with clang-format
> > > 6cc2df8e3a39 landlock: Add clang-format exceptions
> >
> > What order is this in? And what's the overall diffstat? And again, why
> > use clang-format at all, what is it helping with here?
>
> It is the same order as in the master branch.
So they need to be applied backwards in the list above?
> I explain about clang-format in the commit message and the related
> cover letter.
Yes, but really, that is not ok for stable backports.
I'll attempt the above list backwards and see what happens...
thanks,
greg k-h
