Re: [linux-next:master 12118/12398] kernel/liveupdate/luo_core.c:402 luo_ioctl() warn: unsigned 'nr' is never less than zero.

[Andrew Morton <akpm@linux-foundation.org>]

On Sat, 29 Nov 2025 14:51:28 -0500 Pasha Tatashin <pasha.tatashin@soleen.com> wrote:

> > smatch warnings:
> > kernel/liveupdate/luo_core.c:402 luo_ioctl() warn: unsigned 'nr' is never less than zero.
> >
> > vim +/nr +402 kernel/liveupdate/luo_core.c
> >
> >    392
> >    393  static long luo_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
> >    394  {
> >    395          const struct luo_ioctl_op *op;
> >    396          struct luo_ucmd ucmd = {};
> >    397          union ucmd_buffer buf;
> >    398          unsigned int nr;
> >    399          int err;
> >    400
> >    401          nr = _IOC_NR(cmd);
> >  > 402          if (nr < LIVEUPDATE_CMD_BASE ||
> >    403              (nr - LIVEUPDATE_CMD_BASE) >= ARRAY_SIZE(luo_ioctl_ops)) {
> 
> This is a false positive. The logic is designed to work generically
> for any base, similar to how it is handled in session ioctl and
> iommufd ictl. The warning only triggers because LIVEUPDATE_CMD_BASE
> happens to be 0 in this specific case.

This is a difficult situation.

In some cases smatch has found a bug, in other cases it's just being
irritating.

I think there is value in smatch occasionally finding bugs, and to
support that we should try to rearrange false positives so that smatch
doesn't report on those.

Perhaps there's some way we can do that here, without uglifying things?