Re: [PATCH] livepatch: fix merge bug in klp_cleanup_module_patches_limited()

[Josh Poimboeuf]

On Tue, Oct 31, 2017 at 09:12:07PM +0100, Jiri Kosina wrote:
> On Tue, 31 Oct 2017, Josh Poimboeuf wrote:
> 
> > When the patch for "livepatch: add (un)patch callbacks" was posted to
> > the mailing list at:
> > 
> >   https://lkml.kernel.org/r/1507921723-1996-2-git-send-email-joe.lawrence@xxxxxxxxxx
> > 
> > ... it was based on top of the following commit:
> > 
> >   ef8daf8eeb5b ("livepatch: unpatch all klp_objects if klp_module_coming fails").
> > 
> > But when it was applied to git on the livepatching
> > for-4.15/shadow-variables branch, that branch was missing the above
> > commit.  The patch was rebased accordingly:
> > 
> >   93862e385ded ("livepatch: add (un)patch callbacks")
> > 
> > Then the two above commits were merged into livepatching/for-next:
> > 
> >   6a18fdf2d8d3 ("Merge branch 'for-4.15/callbacks' into for-next"),
> > 
> > The result of the merge commit was that the klp_post_unpatch_callback()
> > call in the klp_cleanup_module_patches_limited() function disappeared.
> > The original patch on the mailing list had that call, so put it back.
> > 
> > Fixes: 6a18fdf2d8d3 ("Merge branch 'for-4.15/callbacks' into for-next")
> > Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
> > ---
> >  kernel/livepatch/core.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
> > index 30e392858281..de9e45dca70f 100644
> > --- a/kernel/livepatch/core.c
> > +++ b/kernel/livepatch/core.c
> > @@ -875,6 +875,8 @@ static void klp_cleanup_module_patches_limited(struct module *mod,
> >  				pr_notice("reverting patch '%s' on unloading module '%s'\n",
> >  					  patch->mod->name, obj->mod->name);
> >  				klp_unpatch_object(obj);
> > +
> > +				klp_post_unpatch_callback(obj);
> 
> Bah, thanks a lot for spotting and fixing my brainfart, really 
> appreciated.
> 
> Now, the only branch I could actually technically apply this patch to 
> would be for-next, but that'd be useless, as that is only an 
> integration/testing branch, that doesn't get merged anywhere else than 
> linux-next.

Makes sense, I didn't know the rebase policy for that branch.

> Would you be okay with me just redoing the merge into for-next with this 
> fixup attributed to you in the merge message (and my git-rerere picking it 
> up for upstream)?

Sure, no problem.

-- 
Josh