sched/idle: Mark arch_cpu_idle_dead() __noreturn
authorJosh Poimboeuf <jpoimboe@kernel.org>
Tue, 14 Feb 2023 07:05:58 +0000 (23:05 -0800)
committerJosh Poimboeuf <jpoimboe@kernel.org>
Wed, 8 Mar 2023 16:44:28 +0000 (08:44 -0800)
commit071c44e4278156f18a6a56958617223b6bffa6ab
treee928c7fe2eb63f56a17edc6e896804cafc73d542
parentdfb0f170cadb03da4f408ae53cc2908120e1f90e
sched/idle: Mark arch_cpu_idle_dead() __noreturn

Before commit 076cbf5d2163 ("x86/xen: don't let xen_pv_play_dead()
return"), in Xen, when a previously offlined CPU was brought back
online, it unexpectedly resumed execution where it left off in the
middle of the idle loop.

There were some hacks to make that work, but the behavior was surprising
as do_idle() doesn't expect an offlined CPU to return from the dead (in
arch_cpu_idle_dead()).

Now that Xen has been fixed, and the arch-specific implementations of
arch_cpu_idle_dead() also don't return, give it a __noreturn attribute.

This will cause the compiler to complain if an arch-specific
implementation might return.  It also improves code generation for both
caller and callee.

Also fixes the following warning:

  vmlinux.o: warning: objtool: do_idle+0x25f: unreachable instruction

Reported-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/r/60d527353da8c99d4cf13b6473131d46719ed16d.1676358308.git.jpoimboe@kernel.org
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
18 files changed:
arch/alpha/kernel/process.c
arch/arm/kernel/smp.c
arch/arm64/kernel/process.c
arch/csky/kernel/smp.c
arch/ia64/kernel/process.c
arch/loongarch/kernel/process.c
arch/mips/kernel/process.c
arch/parisc/kernel/process.c
arch/powerpc/kernel/smp.c
arch/riscv/kernel/cpu-hotplug.c
arch/s390/kernel/idle.c
arch/sh/kernel/idle.c
arch/sparc/kernel/process_64.c
arch/x86/kernel/process.c
arch/xtensa/kernel/smp.c
include/linux/cpu.h
kernel/sched/idle.c
tools/objtool/check.c