On 4/11/23 6:14?PM, Josh Poimboeuf wrote:
> On Tue, Apr 11, 2023 at 04:39:39PM -0600, Jens Axboe wrote:
>>>>>>> lib/iov_iter.o: warning: objtool: .altinstr_replacement+0x0: redundant UACCESS d
>>>>>>> isable
>>>>>>> lib/iov_iter.o: warning: objtool: iovec_from_user.part.0+0xc7: call to copy_comp
>>>>>>> at_iovec_from_user.part.0() with UACCESS enabled
>>>>>>> lib/iov_iter.o: warning: objtool: __import_iovec+0x21d: call to copy_compat_iovec_from_user.part.0() with UACCESS enabled
>>>>>>>
>>>>>>> Presumably introduced by commit
>>>>>>>
>>>>>>> 6376ce56feb6 ("iov_iter: import single vector iovecs as ITER_UBUF")
>>
>> lib/iov_iter.o attached, gzip'ed. NOTE: if you disable either of the
>> copy_compat_iovec_from_user() as per diff below (commented out), then
>> it doesn't complain. Is there some bug where it thinks we'll hit both?
>> That should not be possible.
>
> Yeah, the problem is an inter-procedural compiler optimization which
> moves the user_access_begin() out of copy_compat_iovec_from_user() and
> into its callers.
Ah, I see.
> Which is fine, but objtool doesn't like it as it expects the uaccess
> enable to not cross function boundaries.
>
> Do the warnings go away if you make copy_compat_iovec_from_user()
> non-static?
Yep, if I kill the static, it stops complaining.
--
Jens Axboe
