Re: [PATCH 5.4 000/134] 5.4.231-rc1 review

[Guenter Roeck]

On 2/3/23 11:07, Eric Biggers wrote:
> On Fri, Feb 03, 2023 at 10:54:21AM -0800, Eric Biggers wrote:
> > On Fri, Feb 03, 2023 at 09:18:26AM -0800, Guenter Roeck wrote:
> > > On Fri, Feb 03, 2023 at 05:45:19PM +0100, Greg Kroah-Hartman wrote:
> > > > On Fri, Feb 03, 2023 at 07:56:19AM -0800, Guenter Roeck wrote:
> > > > > On Fri, Feb 03, 2023 at 11:11:45AM +0100, Greg Kroah-Hartman wrote:
> > > > > > This is the start of the stable review cycle for the 5.4.231 release.
> > > > > > There are 134 patches in this series, all will be posted as a response
> > > > > > to this one.  If anyone has any issues with these being applied, please
> > > > > > let me know.
> > > > > >
> > > > > > Responses should be made by Sun, 05 Feb 2023 10:09:58 +0000.
> > > > > > Anything received after that time might be too late.
> > > > >
> > > > > Building ia64:defconfig ... failed
> > > > > --------------
> > > > > Error log:
> > > > > <stdin>:1511:2: warning: #warning syscall clone3 not implemented [-Wcpp]
> > > > > arch/ia64/kernel/mca_drv.c: In function 'mca_handler_bh':
> > > > > arch/ia64/kernel/mca_drv.c:179:9: error: implicit declaration of function 'make_task_dead'
> > > > >
> > > > > Caused by "exit: Add and use make_task_dead.". Did that really have to be backported ?
> > > >
> > > > Yup, it does!
> > > >
> > > > Eric, any help with this?
> > >
> > > Adding "#include <linux/sched/task.h>" to the affected file would probably
> > > be the easy fix. I did a quick check, and it works.
> > >
> > > Note that the same problem is seen in v4.14.y and v4.19.y. Later
> > > kernels don't have the problem.
> >
> > This problem arises because <linux/mm.h> transitively includes
> > <linux/sched/task.h> in 5.10 and later, but not in 5.4 and earlier.
> >
> > Greg, any preference for how to handle this situation?
> >
> > Just add '#include <linux/sched/task.h>' to the affected .c file (and hope there
> > are no more affected .c files in the other arch directories) and call it a day?
> >
> > Or should we backport the transitive inclusion (i.e., the #include added by
> > commit 80fbaf1c3f29)?  Or move the declaration of make_task_dead() into
> > <linux/kernel.h> so that it's next to do_exit()?
>
> One question: do *all* the arches actually get built as part of the testing for
> each stable release?  If so, we can just add the #include to the .c files that
> need it.  If not, then it would be safer to take one of the other approaches.

Yes, I do build all architectures for each stable release.

FWIW, I only noticed that one build failure due to this problem.

Guenter