A task that dumps open spufs context adds a bunch of extra elf notes
describing the SPU state. It is the only reason do_coredump() unshares
the file descriptor table. We could make this conditional on spufs but
eh.

Nothing can consume those notes anymore. gdb dropped Cell Broadband
Engine debugging in 9.1 and binutils removed it in 2.34. That's about 6
years ago. So no program can actually read an SPU note out of a core
file and probably never did in recent history.

Note that the IBM Cell blades that shipped the Cell processor were
removed in commit 05bf59fbeef3 ("powerpc/cell: Remove support for IBM
Cell Blades"). The PlayStation 3 is the only platform left and nothing
there produces or reads these notes.

So remove it.

Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
---
Christian Brauner (2):
      powerpc: remove coredump support
      coredump: stop unsharing the file descriptor table

 arch/powerpc/Kconfig                         |   1 -
 arch/powerpc/include/asm/elf.h               |   6 -
 arch/powerpc/include/asm/spu.h               |   3 -
 arch/powerpc/platforms/cell/Kconfig          |   1 -
 arch/powerpc/platforms/cell/spu_syscalls.c   |  20 ---
 arch/powerpc/platforms/cell/spufs/Makefile   |   1 -
 arch/powerpc/platforms/cell/spufs/coredump.c | 183 ---------------------------
 arch/powerpc/platforms/cell/spufs/file.c     | 114 -----------------
 arch/powerpc/platforms/cell/spufs/spufs.h    |  12 --
 arch/powerpc/platforms/cell/spufs/syscalls.c |   4 -
 fs/binfmt_elf.c                              |   4 +-
 fs/coredump.c                                |   5 -
 12 files changed, 2 insertions(+), 352 deletions(-)
---
base-commit: 66498c75b4f8017f62d720d9b59675bdf3abce91
change-id: 20260826-work-spufs-coredump-4e14ce185587

On Wed, Aug 26, 2026, at 18:06, Christian Brauner wrote:
> Note that the IBM Cell blades that shipped the Cell processor were
> removed in commit 05bf59fbeef3 ("powerpc/cell: Remove support for IBM
> Cell Blades"). The PlayStation 3 is the only platform left and nothing
> there produces or reads these notes.

I don't see how you concluded the PS3 does not produce these
notes: as far as I can tell, spufs on PS3 should still be able to do
this if a process using an SPU crashes, and old gdb binaries should
still be able to read them.

Whether anyone is still developing SPU programs on new kernels and
would benefit from the information in the dump I don't know.

> So remove it.
>
> Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>

With clarifications in the changelog:

Acked-by: Arnd Bergmann <arnd@arndb.de>

I have no objection to removing the spufs coredump. If anyone still
cares, Geoff Levand would know, so I've added him to Cc.

       Arnd

On Wed, Aug 26, 2026 at 08:47:42PM +0200, Arnd Bergmann wrote:
> On Wed, Aug 26, 2026, at 18:06, Christian Brauner wrote:
> > Note that the IBM Cell blades that shipped the Cell processor were
> > removed in commit 05bf59fbeef3 ("powerpc/cell: Remove support for IBM
> > Cell Blades"). The PlayStation 3 is the only platform left and nothing
> > there produces or reads these notes.
> 
> I don't see how you concluded the PS3 does not produce these
> notes: as far as I can tell, spufs on PS3 should still be able to do
> this if a process using an SPU crashes, and old gdb binaries should
> still be able to read them.
> 
> Whether anyone is still developing SPU programs on new kernels and
> would benefit from the information in the dump I don't know.

Sure, _old_ debuggers can do that. But new debuggers can't and have
dropped any support for this 6 years as I pointed out in the cut off
part. Similarly, gcc killed the SPU target 6 years ago in 10.1. So no
modern toolchain can build this. And the last supported kernel I found
is v6.12.

I deliberately left spufs itself in but I would really like to get rid
of the coredump special-sauce we need to have for this.

> > So remove it.
> >
> > Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
> 
> With clarifications in the changelog:

Ok, I'll point out that old tools can still do this.

> Acked-by: Arnd Bergmann <arnd@arndb.de>

Thanks.