Re: [PATCH v3 00/12] signal: sort out si_trapno and si_perf

Geert Uytterhoeven [off-list ref] writes:

Hi Eric,

On Tue, May 4, 2021 at 11:14 PM Eric W. Biederman [off-list ref] wrote:

quoted

This set of changes sorts out the ABI issues with SIGTRAP TRAP_PERF, and
hopefully will can get merged before any userspace code starts using the
new ABI.

The big ideas are:
- Placing the asserts first to prevent unexpected ABI changes
- si_trapno becomming ordinary fault subfield.
- struct signalfd_siginfo is almost full

This set of changes starts out with Marco's static_assert changes and
additional one of my own that enforces the fact that the alignment of
siginfo_t is also part of the ABI.  Together these build time
checks verify there are no unexpected ABI changes in the changes
that follow.

The field si_trapno is changed to become an ordinary extension of the
_sigfault member of siginfo.

The code is refactored a bit and then si_perf_type is added along side
si_perf_data in the _perf subfield of _sigfault of siginfo_t.

Finally the signalfd_siginfo fields are removed as they appear to be
filling up the structure without userspace actually being able to use
them.

Thanks for your series, which is now in next-20210506.

quoted

 arch/alpha/include/uapi/asm/siginfo.h              |   2 -
 arch/alpha/kernel/osf_sys.c                        |   2 +-
 arch/alpha/kernel/signal.c                         |   4 +-
 arch/alpha/kernel/traps.c                          |  24 ++---
 arch/alpha/mm/fault.c                              |   4 +-
 arch/arm/kernel/signal.c                           |  39 +++++++
 arch/arm64/kernel/signal.c                         |  39 +++++++
 arch/arm64/kernel/signal32.c                       |  39 +++++++
 arch/mips/include/uapi/asm/siginfo.h               |   2 -
 arch/sparc/include/uapi/asm/siginfo.h              |   3 -
 arch/sparc/kernel/process_64.c                     |   2 +-
 arch/sparc/kernel/signal32.c                       |  37 +++++++
 arch/sparc/kernel/signal_64.c                      |  36 +++++++
 arch/sparc/kernel/sys_sparc_32.c                   |   2 +-
 arch/sparc/kernel/sys_sparc_64.c                   |   2 +-
 arch/sparc/kernel/traps_32.c                       |  22 ++--
 arch/sparc/kernel/traps_64.c                       |  44 ++++----
 arch/sparc/kernel/unaligned_32.c                   |   2 +-
 arch/sparc/mm/fault_32.c                           |   2 +-
 arch/sparc/mm/fault_64.c                           |   2 +-
 arch/x86/kernel/signal_compat.c                    |  15 ++-

No changes needed for other architectures?
All m68k configs are broken with

Thanks.  I hadn't realized that si_perf asserts existed on m68k.
Thankfully linux-next caught this these.

Looking a little more deeply, it is strange that this is tested on m68k.
The architecture does not implement HAVE_PERF_EVENTS so it is impossible
for this signal to be generated.

On the off chance this these new signals will appear on m68k someday I
will update the assertion.

arch/m68k/kernel/signal.c:626:35: error: 'siginfo_t' {aka 'struct
siginfo'} has no member named 'si_perf'; did you mean 'si_errno'?

See e.g. http://kisskb.ellerman.id.au/kisskb/buildresult/14537820/

There are still a few more references left to si_perf:

$ git grep -n -w si_perf
Next/merge.log:2902:Merging userns/for-next (4cf4e48fff05 signal: sort
out si_trapno and si_perf)
arch/m68k/kernel/signal.c:626:  BUILD_BUG_ON(offsetof(siginfo_t,
si_perf) != 0x10);
include/uapi/linux/perf_event.h:467:     * siginfo_t::si_perf, e.g. to
permit user to identify the event.
tools/testing/selftests/perf_events/sigtrap_threads.c:46:/* Unique
value to check si_perf is correctly set from
perf_event_attr::sig_data. */

I will sweep them up as well.

Eric