On 2026-06-24 11:31:17 [+0200], Peter Zijlstra wrote:

On Tue, Jun 23, 2026 at 04:26:49PM +0200, Sebastian Andrzej Siewior wrote:

quoted

+#ifndef WARN_ON_DEFERRED
+#define WARN_ON_DEFERRED(condition) ({					\
+	int __ret_warn_on = !!(condition);				\
+	if (unlikely(__ret_warn_on)) {					\
+		guard(preempt)();					\
+		printk_deferred_enter()					\
+		__WARN();						\
+		printk_deferred_exit()					\
+	}								\
+	unlikely(__ret_warn_on);					\
+})
+#endif

This will generate atrocious shite at the WARN sites.

You mean the missing semicolon and huge size increase?
On x86 with these guard+deffered in the upper variant, before:
    text    data     bss     dec   filename
   93910   37424     832  132166   kernel/sched/core.o
   61802    4945     152   66899   kernel/sched/fair.o
  215108   24453    3768  243329   kernel/sched/build_policy.o
   86128   30092   12704  128924   kernel/sched/build_utility.o
  456948   96914   17456  571318   total
After:
   96140   37408     832  134380   kernel/sched/core.o
   64490    4937     152   69579   kernel/sched/fair.o
  222980   24157    3768  250905   kernel/sched/build_policy.o
   86544   30100   12704  129348   kernel/sched/build_utility.o
  470154   96602   17456  584212   total + 1.3%

total went up by 1.3% or 12.59KiB.
This effects:  alpha, arc, arm, csky, hexagon, m68k, microblaze, mips,
nios2, openrisc, sparc, um, xtensa
and could motivate them to implement __WARN_FLAGS which would lower size
in general and this stunt would have no effect.

Just looked at arm and it has support for invalid opcodes somehow but
not for this.

Sebastian