Re: [RFC PATCH v2 11/20] objtool: Flesh out warning related to pv_ops[] calls

[Valentin Schneider <vschneid@redhat.com>]

On 28/07/23 10:33, Josh Poimboeuf wrote:
> On Thu, Jul 20, 2023 at 05:30:47PM +0100, Valentin Schneider wrote:
>> I had to look into objtool itself to understand what this warning was
>> about; make it more explicit.
>>
>> Signed-off-by: Valentin Schneider <vschneid@redhat.com>
>> ---
>>  tools/objtool/check.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tools/objtool/check.c b/tools/objtool/check.c
>> index 8936a05f0e5ac..d308330f2910e 100644
>> --- a/tools/objtool/check.c
>> +++ b/tools/objtool/check.c
>> @@ -3360,7 +3360,7 @@ static bool pv_call_dest(struct objtool_file *file, struct instruction *insn)
>>
>>      list_for_each_entry(target, &file->pv_ops[idx].targets, pv_target) {
>>              if (!target->sec->noinstr) {
>> -			WARN("pv_ops[%d]: %s", idx, target->name);
>> +			WARN("pv_ops[%d]: indirect call to %s() leaves .noinstr.text section", idx, target->name);
>>                      file->pv_ops[idx].clean = false;
>
> This is an improvement, though I think it still results in two warnings,
> with the second not-so-useful warning happening in validate_call().
>
> Ideally it would only show a single warning, I guess that would need a
> little bit of restructuring the code.

You're quite right - fabricating an artificial warning with a call to __flush_tlb_local():

  vmlinux.o: warning: objtool: pv_ops[1]: indirect call to native_flush_tlb_local() leaves .noinstr.text section
  vmlinux.o: warning: objtool: __flush_tlb_all_noinstr+0x4: call to {dynamic}() leaves .noinstr.text section

Interestingly the second one doesn't seem to have triggered the "pv_ops"
bit of call_dest_name. Seems like any call to insn_reloc(NULL, x) will
return NULL.

Trickling down the file yields:

  vmlinux.o: warning: objtool: pv_ops[1]: indirect call to native_flush_tlb_local() leaves .noinstr.text section
  vmlinux.o: warning: objtool: __flush_tlb_all_noinstr+0x4: call to pv_ops[0]() leaves .noinstr.text section

In my case (!PARAVIRT_XXL) pv_ops should look like:
  [0]: .cpu.io_delay
  [1]: .mmu.flush_tlb_user()

so pv_ops[1] looks right. Seems like pv_call_dest() gets it right because
it uses arch_dest_reloc_offset().

If I use the above to fix up validate_call(), would we still need
pv_call_dest() & co?

>
> --
> Josh