amd64 kernel __storeload_barrier: quiet gcc -Warray-bounds

[rlibby]

Summary

Use a constant input operand instead of an output operand to tell the
compiler about OFFSETOF_MONITORBUF. If we tell it we are writing to
*(u_int *)OFFSETOF_MONITORBUF, it rightly complains, but we aren't. The
memory clobber already covers the necessary semantics for the compiler.

---

This quiets around 142 warnings building amd64 GENERIC.

Test Plan

clang:

% llvm-objdump -d --disassemble-symbols=membarrier_action_seqcst /usr/obj/usr/src/freebsd/amd64.amd64/sys/GENERIC/kernel 

/usr/obj/usr/src/freebsd/amd64.amd64/sys/GENERIC/kernel:	file format elf64-x86-64

Disassembly of section .text:

ffffffff80b1d020 <membarrier_action_seqcst>:
ffffffff80b1d020: 55                   	pushq	%rbp
ffffffff80b1d021: 48 89 e5             	movq	%rsp, %rbp
ffffffff80b1d024: f0                   	lock
ffffffff80b1d025: 65 83 04 25 00 01 00 00 00   	addl	$0x0, %gs:0x100
ffffffff80b1d02e: 5d                   	popq	%rbp
ffffffff80b1d02f: c3                   	retq

gcc13:

% llvm-objdump -d --disassemble-symbols=membarrier_action_seqcst /usr/obj/gcc13/usr/src/freebsd/amd64.amd64/sys/GENERIC/kernel

/usr/obj/gcc13/usr/src/freebsd/amd64.amd64/sys/GENERIC/kernel:	file format elf64-x86-64

Disassembly of section .text:

ffffffff81448550 <membarrier_action_seqcst>:
ffffffff81448550: f0                   	lock
ffffffff81448551: 65 83 04 25 00 01 00 00 00   	addl	$0x0, %gs:0x100
ffffffff8144855a: c3                   	retq
ffffffff8144855b: 0f 1f 44 00 00       	nopl	(%rax,%rax)

[kib]

Would you please show the gcc warning, to satisfy my curiosity?

[rlibby]

In D45694#1042256, @kib wrote:

Would you please show the gcc warning, to satisfy my curiosity?

Sure, here's an example from gcc13 for membarrier_action_seqcst():

In file included from /usr/src/freebsd-branches/main-push/sys/sys/systm.h:45,
                 from /usr/src/freebsd-branches/main-push/sys/kern/kern_membarrier.c:30:
In function '__storeload_barrier',
    inlined from 'atomic_thread_fence_seq_cst' at ./machine/atomic.h:352:2,
    inlined from 'membarrier_action_seqcst' at /usr/src/freebsd-branches/main-push/sys/kern/kern_membarrier.c:59:2:
./machine/atomic.h:298:9: warning: array subscript 0 is outside array bounds of 'u_int[0]' {aka 'unsigned int[]'} [-Warray-bounds=]
  298 |         __asm __volatile("lock; addl $0,%%gs:%0"
      |         ^~~~~
In function 'membarrier_action_seqcst':
cc1: note: source object is likely at address zero
In function '__storeload_barrier',
    inlined from 'atomic_thread_fence_seq_cst' at ./machine/atomic.h:352:2,
    inlined from 'membarrier_action_seqcst' at /usr/src/freebsd-branches/main-push/sys/kern/kern_membarrier.c:59:2:
./machine/atomic.h:298:9: warning: array subscript 0 is outside array bounds of 'u_int[0]' {aka 'unsigned int[]'} [-Warray-bounds=]
In function 'membarrier_action_seqcst':
cc1: note: source object is likely at address zero

I copied the meta file here, but it's more of the same:
https://people.freebsd.org/~rlibby/gcc13-kern_membarrier.o.meta

It looks like the warning gets issued twice per call so the 142 warnings might be just 71 sites.